STRUCTURAL ANALYSIS OF HISTORIC CONSTRUCTION

PROCEEDINGS OF THE SIXTH INTERNATIONAL CONFERENCE ON STRUCTURAL

ANALYSIS OF HISTORIC CONSTRUCTION, 2–4 JULY, BATH, UNITED KINGDOM

Structural Analysis of Historic

Construction
Preserving Safety and Significance

VOLUME 1

Editors
Dina D’Ayala & Enrico Fodde
Department of Architecture & Civil Engineering,
University of Bath, Bath, UK
CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business

© 2008 Taylor & Francis Group, London, UK

Typeset by Charon Tec Ltd (A Macmillan Company), Chennai, India

Printed and bound in Great Britain by Cromwell Press Ltd, Towbridge, Wiltshire

All rights reserved. No part of this publication or the information contained herein may be reproduced, stored
in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, by photocopying,
recording or otherwise, without written prior permission from the publishers.

Although all care is taken to ensure integrity and the quality of this publication and the information herein, no
responsibility is assumed by the publishers nor the author for any damage to the property or persons as a result
of operation or use of this publication and/or the information contained herein.

Published by: CRC Press/Balkema

P. O. Box 447, 2300 AK Leiden, The Netherlands
e-mail: Pub.NL@taylorandfrancis.com
www.crcpress.com – www.taylorandfrancis.co.uk – www.balkema.nl

ISBN Set: 978-0-415-46872-5 (Set of 2 Volumes + CD-Rom)

ISBN Vol.1: 978-0-415-48109-0
ISBN Vol.2: 978-0-415-48107-6
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Table of Contents

Preface XV

Committees XVII

Conference Sponsors XIX

VOLUME 1

Invited papers
Preservation of workmanship or workmanship for preservation 3
K. Van Balen & R. Hendrickx

Experiences from the Northern Italy 2004 earthquake: Vulnerability assessment and
strengthening of historic churches 13
E. Giuriani & A. Marini

20th century curtain walls – loss of redundancy and increase in complexity 25

S.J. Kelley

Strengthening of masonry structures with Fibre Reinforced Plastics: From modern

conception to historical building preservation 33
M.R. Valluzzi

The structural behaviour of masonry vaults: Limit state analysis with finite friction 47
D.F. D’Ayala & E. Tomasoni

Repairs to historic timber structures: Changing attitudes and knowledge 63

D.T. Yeomans

A multidisciplinary approach for the structural restoration of the Katholikon of Dafni

Monastery in Attica Greece 71
A. Miltiadou-Fezans

Structural interventions in English Cathedrals 89

G.A. Clifton

Monastery of Salzedas (Portugal): Intervention in the cloister and information management 95

P.B. Lourenço, L.F. Ramos, G. Vasconcelos & F. Peña

Cultural Heritage protection against flood – A European FP6 research project 109
Ch. Maierhofer, Ch. Köpp, S. Kruschwitz, M. Drdacky, Ch. Hennen, S. Lanza, M. Tomaževič,
D. Kolymbas, I. Herle, A. Virsta, L. Binda & P. Askew

Definition of significance and attribution of value

Structural definition and comparison of early medieval roof structures 121
C. Thelin & K. Linscott

V
Analysis of archaic fireproof floor systems 129
D. Friedman
Intervention limits in floor wood structure restoration of a historical building 137
R.T.M. Ribeiro & M. Hoirisch
The safety of historic load-bearing structures 143
B. Szabó & I. Kirizsán
Concrete: Too young for conservation? 151
H.A. Heinemann, R.P.J. van Hees, T.G. Nijland
Significance of a Mughal Mall and its role in achieving safety 161
S. Sanyal
Historical constructions – Authenticity and adaptation to the modern demands 169
I. Bucur-Horváth, I. Popa, M. Bulbuk & J. Virág
Champaner-Pavagadh – Managing conflicts – A conservation challenge 175
S.M. Modi

History and documentation

Construction and structural behavior of Vladimir Suchov’s Nigres tower 183
M. Beckh, R. Barthel & A. Kutnyi
New studies on Brunelleschi’s Dome in Florence 191
L. Giorgi & P. Matracchi
Wells Cathedral: The crossing and the central tower. A new hypothesis on the constructional
sequence and dating of the central tower 199
G. De Blasi
A study of historical test data for better informed assessment of wrought-iron structures 207
M. O’Sullivan & T. Swailes
Structural interpretation of post-earthquakes (19th century) retrofitting on the
Santa Maria degli Angeli Basilica, Assisi, Italy 217
L. Sorrentino, D. Bruccoleri & M. Antonini
Engineering survival and success: The contributions of historic structural features in the
spire and tower of Salisbury Cathedral 227
Robert S. Cotta
Structure of baroque church roofs in Bavaria 235
S.M. Holzer & B. Köck
Brick arch bridges in the High Cauca Region of Colombia: A forgotten construction tradition 243
J.A. Galindo, J.A. Paredes, A.M. Muñoz & Y.M. Ospina
From empirical engineering to science-based engineering: The assessment by three
mathematicians of the “San Pietro” Dome (1742) 251
O. Niglio
From static history to restoration issues: The Gesú Nuovo church in Naples (Italy) 259
S. Casiello, V. Russo & E. Vassallo
Building techniques of the Zeytinburnu Military Factory built in the 19th century 267
A. Çiftçi & N. Seçkin
Investigation of the 1716 Algiers (Algeria) earthquake from historical sources 275
A.A. Abdessemed-Foufa & D. Benouar
The Morro da Queimada Archaeological Park, Ouro Preto, MG – Brazil 283
B.T. Oliveira

VI
Structure as a document 289
J. Jain

Monitoring, non destructive evaluation and testing

Historic multiple-leaf masonry wall models under compression and cyclic shear loads 297
R. Capozucca
The masonry vaults: Geometry definition and possible approaches to the static analysis 303
O. Corbi
Damage identification in masonry structures with vibration measurements 311
L.F. Ramos, P.B. Lourenço, G. De Roeck & A. Campos-Costa
Stress analysis of masonry structures: Arches, walls and vaults 321
A. Baratta, I. Corbi & O. Corbi
Evaluation of historic Dieh-Dou buildings in Taiwan 331
P.-H. Tsai & D. D’Ayala
Experimental techniques for the evaluation of the long-term behaviour of masonry
strengthening materials 339
P. Bocca & A. Grazzini
Experimental and numerical studies on the belltower of Santa Justa y Rufina (Orihuela-Spain) 349
S. Ivorra, F. Pallarés & J.M. Adam
The analysis of the strains and stresses of the reinforced brick vaults 357
J. Jasieńko & Ł. Bednarz
Cracking simulation of brick-masonry elements subjected to the double flat-jack test 367
A. Carpinteri, S. Invernizzi & G. Lacidogna
Numerical simulation and monitoring of the Cathedral of Syracuse in Sicily 375
A. Carpinteri, S. Invernizzi, G. Lacidogna, A. Manuello & L. Binda
Shaking table tests on the large scale model of Mustafa Pasha Mosque without and
with FRP 383
L. Krstevska, Lj. Tashkov, K. Gramatikov, R. Landolfo, O. Mammana, F. Portioli &
F.M. Mazzolani
Integrated modeling method for dynamic behavior of ancient pagodas 393
J. Yuan, L. Yao, S. Li & D. Abruzzese
Monitoring and strengthening interventions on the stone tomb of Cansignorio della Scala,
Verona, Italy 403
G. Gaudini, C. Modena, F. Casarin, C. Bettio & F. Lucchin
Seismic behaviour of barrel vault systems 413
A. Marini, G. Giardina, P. Riva & E. Giuriani
Assessment of material degradation based on microcores testing 423
M. Skłodowski
Structural performance of riveted connections in historical metal structures 431
M. D’Aniello, L. Fiorino & R. Landolfo
Diagnostic tests and numerical simulations for the preservation of two stone stairways in the
historic centre of Genoa (Italy) 441
A. Brignola, A. Del Grosso, S. Podestà, S. Resemini & G. Riotto
Evaluation of iron and steel in historic bridges 451
S.P. Sparks

VII
Diagnosis and analysis of two king-post trusses 459
J.M. Branco, P.J.S. Cruz & M. Piazza
Non destructive and laboratory evaluation of strength of decayed wood members in a historic
construction located in Gorgan (North of Iran) 469
M. Madhoushi, J. Eimanian & M.P. Ansell
The structural behaviour of spires 473
E. Olosz & B. Szabó
In-situ and laboratory investigations on materials and structures for the static improvement:
“Torre Quadrata del Circo Romano” (Milan) 481
S. Bortolotto, E. Ciocchini, F. Zangheri & P.M. De Marchi
Dynamic multi-body behaviour of historic masonry buildings models 489
D. D’Ayala, Y. Shi & C. Stammers
Assessment and strengthening masonry arch bridges 497
C.L. Brookes
Field tests of a timber queen-post truss and numerical analysis 507
J.M. Branco, P.J.S. Cruz, M. Piazza & H. Varum

Assessment and analytical techniques

Modelling of vaults as equivalent diaphragms in 3D seismic analysis of masonry buildings 517
S. Cattari, S. Resemini & S. Lagomarsino
Modelling of damage accumulation in masonry subjected to a long-term compressive load 525
E. Verstrynge, S. Ignoul, L. Schueremans & D. Van Gemert
A method for studying multi-body dynamics and its applications to masonry structures analysis 533
G. Gariup, I. Pitacco & P.R. Fabregat
Load-carrying capacity of multi-span masonry arch bridges having limited ductility 541
G. de Felice
Assessment of irregular masonry walls 547
N. Augenti & A. Romano
Geometrical and static aspects of the Cupola of Santa Maria del Fiore, Florence (Italy) 555
A. Cecchi, I. Chiaverini & A. Passerini
Some considerations on out-of-plane collapse modes of masonry walls 565
G. Brandonisio, E. Mele & A. De Luca
Innovative techniques for structural assessment: The case of the Holy Shroud Chapel in Turin 575
A. De Stefano, D. Enrione & G. Ruocci
On the theory of the ellipse of elasticity as a natural discretisation method in the design of
Paderno d’Adda Bridge (Italy) 583
R. Ferrari & E. Rizzi
Lower-bound analysis of masonry vaults 593
P. Block & J. Ochsendorf
Assessment of seismic resistance of a basilica church: Modelling, analysis and
strengthening proposal 601
M. Betti & A. Vignoli
Estimation of the seismic history of the city of Thessaloniki through back analysis of its
Byzantine land walls 611
K.-A. Stylianidis & A. Sextos

VIII
Earthquake based condition monitoring of the Katholikon of Dafni Monastery 621
H.P. Mouzakis, A. Miltiadou, N. Delinikolas, P. Touliatos & J. Dourakopoulos
Continuous and discontinuous modeling of the structures in Bayon temple, Angkor 629
T. Maeda, T. Yamamoto & T. Hirai
Safety assessment of temple E7 in Mỹ So,n, Viet Nam 635
P. Condoleo, A. Taliercio & L. Binda
Construction process, damage and structural analysis. Two case studies 643
R. González, F. Caballé, J. Domenge, M. Vendrell, P. Giráldez, P. Roca & J.L. González
Analysis and diagnosis of timber structures of Porto historical centre 653
T. Ilharco, J.M. Guedes, A. Arêde, E. Paupério & A.G. Costa
Baroque roof structures in Transylvania – Research and analyses 663
D. Makay & B. Szabó
The Athens Parthenon: Analysis and interpretation of the structural failures of the orthostate of
the northern wall 673
E.E. Toumbakari
Structural assessment of the Winter Garden of the Royal Glasshouses of Laeken, Belgium 683
L. Lauriks, I. Wouters, S. Adriaenssens, M. de Bouw & B. Verbeeck
Towards a systematic diagnosis of structural damage 689
I.A.E. de Vent, R.P.J. van Hees & G.J. Hobbelman
From geometrical and cracks survey to static analysis method: The case study of Santa Maria
del Quartiere dome, in Parma (Italy) 697
F. Ottoni
A benchmarking study of the analysis of non-reinforced structures applied to the structural
behavior of domes 705
P. Morer & R. Goñi
Structural analysis of the Caserta Royal Palace timber roof connections 715
G. Fabbrocino, G. Marcari, C. Laorenza & E. Cosenza
Real-time Probabilistic health monitoring of the Portogruaro Civic Tower 723
D. Zonta, M. Pozzi, P. Zanon, G.A. Anese & A. Busetto

Author index 733

VOLUME 2
Experimental studies
On the strengthening of three-leaf stone masonry walls 739
R.A. Silva, D.V. Oliveira & P.B. Lourenço
Long-term behaviour of historical masonry – A quantitative acquisition of the damage evolution 747
E. Verstrynge, L. Schueremans, D. Van Gemert, S. Ignoul & M. Wevers
Experimental assessment and modelling of in-plane behaviour of timber floors 755
M.R. Valluzzi, E. Garbin, M. Dalla Benetta & C. Modena
Bond behaviour of CFRP and GFRP laminates on brick masonry 763
M. Panizza, E. Garbin, M.R. Valluzzi & C. Modena
Experimental shear behavior of stone masonry joints 771
G. Vasconcelos, P.B. Lourenço & D. Oliveira

IX
Large scale testing of drystone retaining walls 781
C. Mundell, P. McCombie, A. Heath, J. Harkness & P. Walker
Seismic isolation, strengthening of walls with CFRP strips and heritage masonry buildings 789
M. Tomaževič, I. Klemenc & P. Weiss
Experimental survey on seismic response of masonry models 799
A. Baratta, I. Corbi, O. Corbi & D. Rinaldis
Bond performance of fiber reinforced grout on brickwork specimens 809
I. Carbone & G. de Felice
Experimental analysis of rectangular shaped sleeve connectors for composite
timber-steel-concrete floors: Bending tests 817
L. Calado, J.M. Proença, A. Panão, F.M. Mazzolani, B. Faggiano & A. Marzo
Shaking table testing of a typical Mexican colonial temple 825
M. Chávez & R. Meli
Capacity of a traditional timber mortise and tenon joint 833
A.O. Feio, P.B. Lourenço & J.S. Machado
Behaviour of refurbished timber floors characterized by different in-plane stiffness 843
M. Piazza, C. Baldessari, R. Tomasi & E. Acler
Stone masonry walls: Strengthening with TRM (I) 851
J.T. San-José, D. García, R. San-Mateos & J. Díez
The effect of friction joint and Gongpo (bracket set) as an energy dissipation in Korean
traditional wooden structure 861
J.K. Hwang, S.G. Hong, N.H. Kim, Y.W. Lee, S.J. Jeong & S.J. Joo
Optimization of cutting processes in archaeological sites 867
C. Cennamo, B.M. Chiaia, E. Masoero & S. Scaini
Experimental and numerical study on the application of the flat-jack tests to masonry walls 875
M. Acito, L. Binda, G. Cardani, J. Guedes & L. Miranda
The impact of the water/lime ratio on the structural characteristics of air lime mortars 885
R.M.H. Lawrence & P. Walker
Soil mortar’s evaluation characteristics for their use in conservation practice 891
G.E. Bei
Experimental and numerical analyses for static retrofitting intervention on the “Cappella
dei Principi” in Firenze 899
A. Antonelli, G. Bartoli & M. Betti
Comparison the effect of different repair method applying on masonry arch bridges 909
M. Miri & T.G. Hughes

Use of traditional, alternative and innovative materials

Restoring of timber structures: Connections with timber pegs 915
C. Ceraldi, V. Mormone & E. Russo Ermolli
Research on composite strengthening of historical housebuilding: Retrofitting intervention
for masonry arches and vaults 921
A. Borri & G. Castori
Static and dynamic properties of a flexible joint working in cracked historical masonries 931
A. Kwiecień, B. Zaja˛c & R. Jankowski
Nanotechnologies applied to the restoration and maintenance of wooden built heritage 941
C. Bertolini Cestari, T. Marzi, S. Invernizzi & J.M. Tulliani

X
Cross laminated timber panels to strengthen wood floors 949
A. Gubana
Lime mortar with natural pozzolana: Historical issues and mechanical behavior 957
E. Sala, I. Giustina & G.A. Plizzari
Blended lime-cement mortars for conservation purposes: Microstructure and strength
development 965
O. Cizer, K. Van Balen, D. Van Gemert & J. Elsen
Assessing workability of mortar by means of rheological parameters and desorptivity 973
R. Hendrickx, K. Van Balen & D. Van Gemert
Consolidation and reinforcement of stone walls using a reinforced repointing grid 981
A. Borri, M. Corradi, E. Speranzini & A. Giannantoni
Innovative and sustainable local material in traditional African architecture – Socio cultural
dimension 991
T.O. Odeyale & T.O. Adekunle
Imported construction materials and techniques in 19th century Ottoman architecture 999
U. Yergün & A. Çiftçi
Advance composites applications on historical structures in Italy: Case studies and future
developments 1007
P. Casadei & E. Agneloni

Novel conservation engineering techniques, restoration and strengthening

FRP-strengthening of masonry structures: Effect of debonding phenomenon 1017
E. Grande, M. Imbimbo & E. Sacco
Methodology for in situ application of hydraulic grouts on historic masonry structures.
The case of the Katholikon of Dafni Monastery 1025
A. Miltiadou-Fezans, A. Kalagri, S. Kakkinou, A. Ziagou, N. Delinikolas,
E. Zarogianni & E. Chorafa
High reversibility techniques for in-plane stiffening of wooden floors 1035
N. Gattesco & L. Macorini
Prestressing of masonry as strengthening measure against earthquake loading – Experimental
and numerical investigations and influences on simulation results 1043
S.T. Sperbeck & H. Budelmann
Enhancement of geo-electrical techniques for NDT of masonry 1053
F. Van Rickstal, D. Van Gemert, R. Keersmaekers & D. Posen
Synergetic action of new and existing structures in re-using a 18th century monastery 1061
M. Mezzi, M. Marzullo & G. Valletta
Structural assessment of earthen walls and techniques of onsite consolidation and conservation 1069
S. Sikka
Preservation of historic structures using Screw-Pile foundations 1079
A.J. Lutenegger & J.H. Kemper
Strengthening of an industrial cylindrical shell damaged by a collision 1087
W. Figeys, S. Ignoul & D.V. Gemert
Experimental results on the use of mud-based grouts to repair seismic cracks on adobe walls 1095
J. Vargas, M. Blondet, C. Cancino, F. Ginocchio, C. Iwaki & K. Morales
Development of an in situ penetration test for the uptake of preservatives in applied wood 1101
D.F. Henriques, L. Nunes & J. de Brito

XI
Special structural solutions for adaptive use in grand hotel in Norwich, CT, USA 1107
J.F. Norden
The restoration of the foundation in the north side of the temple of Apollo Epikourios 1113
K. Papadopoulos
Structural consolidation methods for the Temple of Santa Maria della Consolazione in
Todi (Perugia, Italy), damaged by landslides and earthquakes 1123
M. Mariani
The diagnosis and arresting of settlement within Westminster Hall in the Houses of Parliament, London 1133
J.D. Miller
Istanbul – Fatih, Millet Library / Feyzullah Efendi Madrasah restoration 1141
R. Ozakin & A. Erdem
Application of non-destructive techniques at the Katholikon of Dafni Monastery for mapping
the mosaics substrata and grouting monitoring 1149
Ph. Côte, X. Dérobert, A. Miltiadou-Fezans, N. Delinikolas, O. Durand, J. Alexandre,
A. Kalagri, M. Savvidou, D. Chryssopoulos, L. Anamaterou & F. Georganis
The PROHITECH research project 1157
F.M. Mazzolani
The column of Marcian/Kıztaşı in Istanbul and an evaluation of its restorations 1167
A. Erdem & R. Ozakin
Restoration and strengthening strategies for 19th century iron pedestrian suspension bridges 1175
S. Adriaenssens, B. Verbeeck, I. Wouters & M. de Bouw
Reconstruction of the Sistani House at Bam Citadel after the collapse due to the earthquake 2003 1181
W. Jäger & C. Fuchs

Vulnerability to natural hazards, preparedness and retrofitting

Lower and upper bounds in closed form for out-of-plane strength of masonry structures with
frictional resistances 1191
C. Casapulla
Influence of friction and tensile resistance on the stability of masonry arches 1199
P. Smars
Unexplained blasting vulnerabilities in a historic town 1207
D.F. Laefer, B. Loughlin, S. Hickey, S. O’Farrell & G. O’Mahony
Vulnerability and seismic improvement starting from experimental investigation 1215
C. Algeri
Seismic vulnerability evaluation of the Fossanova Gothic church 1225
G. De Matteis, F. Colanzi, A. Eboli & F.M. Mazzolani
Simplified models for seismic vulnerability analysis of bell towers 1237
E. Curti, S. Parodi & S. Podestà
Guastavino dome analysis: A comparative approach for Jefferson’s Rotunda at the
University of Virginia 1245
J. Kaup & J.A. Matteo
Seismic vulnerability and preservation of timber roof structures 1253
M.A. Parisi, C. Chesi, C. Tardini & M. Piazza
Seismic safety of basilica churches: Analysis of ten case studies 1261
G. Brandonisio, E. Mele, R. Santaniello & A. De Luca

XII
The effects of temperature on historical stone masonry structures 1271
C. Blasi & E. Coïsson
A GIS platform on main natural hazards for Valparaíso city (Chile) and vulnerability studies
for some historical constructions and urban sectors 1277
M. Indirli, F. Geremei, C. Puglisi, A. Screpanti, D. Blersch, L. Lanzoni, N. Lopez Izquerdo,
E. Milani, M. Miglioli, G. Simonini, M. Munari & F. Romanelli
The construction of a GIS database for the historical center of San Giuliano di Puglia 1287
S. Mazzanti, M. Miglioli, M. Nascosi, G. Simonini, M. Indirli, L. Lanzoni, S. Teston,
C. Alessandri & D. Pini
Vulnerability assessment of churches at Colima by 3D limit analysis models 1297
A. Orduña, A. Preciado, F. Galván & J.C. Araiza
Vulnerability screening by visual inspection of churches of North-East Anatolia 1303
A. Korkmaz, S. Toker & H. Hastemoglu
Seismic resistance assessment of heritage masonry buildings in public use in Ljubljana 1311
M. Lutman

Codes, guidelines and methods for safeguarding safety and significance

Use of reliability methods for evaluating safety of historic structures 1321
L. Schueremans & E. Verstrynge
Experimental assessment of historic building safety: The case of the Isso Tower in Castelleone, Italy 1331
L. Binda, P. Condoleo, A. Saisi, C. Tiraboschi & L. Zanzi
Making-safe Sydney’s sandstone buildings within accepted conservation guidelines 1339
J.D. Swann
Long term compressive testing for masonry – Test procedure and practical experience 1345
L. Binda, L. Schueremans, E. Verstrynge, S. Ignoul, D.V. Oliveira, P.B. Lourenço & C. Modena
Safeguarding safety and significance – Saving an historic village in Bath 1357
M.S. Stacey & B.P. McCabe
Integration of structural analysis of monuments and historical constructions in engineering and
architecture studies 1363
A. Mosseri
The role of service-learning in heritage preservation and engineering education 1369
D.W. Porter, M.M. Dewoolkar & N.J. Hayden
The safety of Gothic roof structures 1375
I. Kirizsán & B. Szabó
Strengthening historic pedestrian suspension bridges: Public safety goes first!? 1383
I. Wouters, Y. Schoonjans, M. de Bouw & S. Adriaenssens
Displacement requirements in the nonlinear kinematic procedure for masonry structures 1391
A. Giordano, M. Guadagnuolo & G. Faella

Conservation practice
Relation between sub-soil and masonry structure of Angkor monument 1399
S. Yamada, M. Araya & T. Nagai
Beaufort Castle, Lebanon; conservation versus restoration project 1407
J. Yasmine
Structural faults in earthen archaeological sites in central Asia: Analysis and repair methods 1415
E. Fodde

XIII
Maria Birnbaum – Construction history, conservation history 1423
R. Barthel, H. Maus & C. Kayser
The structural strengthening of early and mid 20th century reinforced concrete diaphragms 1431
L. Sorrentino & C. Tocci
Strengthening design of Ganxi’s Former Residence 1441
C. Qing
The Trajan Markets and their Great Hall – The conservation problems and the structural
intervention for the improvement of the seismic safety 1445
G. Croci, A. Viskovic, A. Bozzetti, L.Ungaro & M. Vitti
Structural strengthening of the Dobrzyca Palace, Poland 1455
M.Y. Minch & J.P. Szołomicki
Structural analysis and restoration of the Guglia della Madonna di Bitonto 1461
T.M. Massarelli
Non conventional solutions for the consolidation of bell towers 1467
C. Mileto, F. Vegas López-Manzanares
Strengthening of a wooden covering built in the 18th century 1475
N. Augenti
Experiences of consolidation on archaeological UNESCO sites in the Sultanate of
Oman: The Fortress of Al Balid and the Citadel of Sumhuram 1481
M. Andreini, C. Cei, G. Mariani & M. Sassu
Conservation and restoration of a Ta Prohm temple 1491
T. Lakshmipriya
Restoration of a historical building for Cyprus Technical Chamber 1499
M. Pittas
The great hall of the Trajan Markets throughout the centuries 1509
L. Ungaro, M. Vitti & E. Speranza
The XIX century opera theatre of Catania: Conservation and seismic strengthening interventions 1519
C.F. Carocci & C. Tocci
House-towers in Campi Flegrei: History, conservation and re-use 1527
G. de Martino, R. de Martino, A. Pane & F. Delizia
Structural inspection and analysis of former British Consulate in Shanghai 1537
X.L. Gu, B. Peng, X. Li & D.F. Shang

Author index 1545

XIV
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Preface

The successful preservation of an historic building, complex, or city, unless assigned to museum status, depends
on its continued use and the daily care and maintenance. The possibility of continued use depends on the
adaptation of the building to modern standards and practice of living, and this might require changes of its
constructional or structural features.
Conservation engineering can be defined as the process of understanding, interpreting and managing the
architectural heritage to safely deliver it to posterity. The term ‘safely’ embodies the assumption that the bodies
and individuals responsible for the care of historic buildings work towards ensuring maximum private or public
utility vis a vis minimum loss of fabric and value.
The conflict between safety standards and conservation philosophy usually stems from the fact that not
just the standards themselves, but the practice of achieving the standards, are based on, and refer to, modern
materials, techniques and processes. An historic building is considered of value not only because of its age,
but most importantly because of its uniqueness, its deviation from the norm and hence, to a certain extent,
from what is standard. This constitutes the building’s significance or authenticity, a quality vital to conserve.
With increasing global interest in conservation worldwide it is essential to open the debate on more inclusive
definitions of significance and on more articulated concepts for safety and for acceptable and reliable techniques,
in an attempt to reconcile and integrate further the activity of all the professions involved in conservation.
It is in this framework and with these aims that the Department of Architecture and Civil Engineering of the
University of Bath set out to organize and deliver the sixth international conference on Structural Analysis of
Historic Construction to be held in Bath’s Assembly Rooms between 2 and 4 July 2008.
A grade I listed building designed in 1769 by John Wood the Younger, the Assembly Rooms are located in
the heart of the World Heritage city. Their function was to host events and gatherings and this makes it the
perfect venue for the conference, which is the last of a series that is organized every three or two years. Previous
conferences were organized in New Delhi (2006), Padova (2004), Guimaraes (2001), and Barcelona (1998 and
1995).
The event is sponsored by CINTEC International Ltd, Gifford, English Heritage, ICOMOS, IStructE, and the
University of Bath. This sponsorship has made possible the participation to the conference of authors from other
parts of the world which could have not attended otherwise. Their contribution is critical to our understanding
of practical and theoretical issues in different contexts and to the widening and inclusivity of the debate.
The joint Department of Architecture and Civil Engineering of the University of Bath has a long standing
tradition of teaching conservation and sustainability within the built environment. The MSc in Conservation
of Historic Buildings provides education on the fundamental principles of conservation and is one of the most
successful in the country. Furthermore, the Department offers the only course on Conservation Engineering in
the UK, combining structural aspects with the study and conservation of the built heritage.
The main objective of the conference is to promote debate on an international scale, by the exchange of knowl-
edge and approaches among researchers and practitioners in the different professions involved in conservation and
from diverse cultural backgrounds. The conference is structured into eleven themes: ‘Definition of Significance
and Attribution of Value’, ‘History and Documentation’, ‘Monitoring, non Destructive Evaluation and Testing’,
‘Assessment and Analytical Techniques’, ‘Experimental Studies’, ‘Use of Traditional, Alternative and Innova-
tive Materials’, ‘Novel Conservation Engineering Techniques, Restoration and Strengthening’, ‘Vulnerability to
Natural Hazards, Preparedness and Retrofitting’, ‘Codes, Guidelines and Methods for Safeguarding Safety and
Significance’, and ‘Conservation Practice’.
A total of 180 contributions are collected in the two volumes of proceedings, of which 10 are invited papers
presented by speakers of international renown to the conservation audience. The reading of these proceedings
shows that the discipline of structural analysis of historic buildings is nowadays critical to the success of many
conservation projects and the skills required are increasingly complex and more articulated. The papers collected
all show the need for multidisciplinary approach in conservation research and practice and the continually
evolving need for more sophisticated solutions where both safety and significance play a critical role. An
increasing number of papers tackle issues not covered in previous conferences, such as the structural conservation
of modern materials. It is hoped that the papers presented here can form the basis for future research and the

XV
formulation of relevant issues that need to be advanced. This we hope will constitute the basis for a lively debate
during the 3 days of the conference and will inform and stimulate the directions for future conferences.
The editors are grateful to Marion Harney for managing the organization of the conference and to the depart-
mental office for the great support. The editors are also indebted to the Advisory Committee members for
providing the opportunity and the needed guidance to organize this conference, to the Organizing Commit-
tee members for the efforts and support given to make the conference possible, and finally but crucially to all
members of the Scientific Committee which have agreed to advise in the papers’ selection and reviewing process.

March 2008
Dina D’Ayala, Enrico Fodde

XVI
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Committees

Advisory Committee
Paulo Lourenço, University of Minho, Portugal
Claudio Modena, University of Padua, Italy
Pere Roca, Politecnico de Cataluña, Spain

Organising Committee
Dina D’Ayala, University of Bath, UK
Marion Harney, University of Bath, UK
Enrico Fodde, University of Bath, UK
Alex Copping, University of Bath, UK
Michael Forsyth, University of Bath, UK
Peter Walker, University of Bath, UK
Susan Denyer, Secretary ICOMOS-UK
Geoff Clifton, Chairman, Gifford, UK
Terry Girdler, Chief Engineer, English Heritage, UK

Scientific Committee
Takayoshi Aoki, Nagoya City University, Japan
Görün Arun, Yildiz Technical University, Istanbul, Turkey
Koenraad van Balen, Katholieke Universiteit Leuven, Belgium
Luigia Binda, Politecnico di Milano, Italy
Marcial Blondet, Catholic University of Peru, Peru
Miloz Drdácký, Institute of Theoretical and Applied Mechanics, Czech Republic
Peter Elliott, Elliott & Company, UK
Xianglin Ling Gu, Tongji University, China
Mehrdad Hejazi, University of Isfahan, Iran
Peter James, Cintec International Ltd, UK
Debra Leafer, University College Dublin, Ireland
Giorgio Macchi, Università di Pavia, Italy
Christiane Maierhofer, Fed Inst Material Research and Testing (BAM), Germany
Roberto Meli, Universidad Nacional Autónoma de Mexico, Mexico
John Ochsendorf, MIT, USA
Daniel Oliveira, University of Minho, Portugal
Maurizio Piazza, Universitá Degli Studi di Trento, Italy
Jan G. Rots, Technical University of Delft, The Netherlands
Marek Sklodowski, Institute of Fundamental Technological Research, Poland
Tom Swailes, University of Manchester, UK
Miha Tomazevic, Slovenian National Building and Civil Engineering Institute, Slovenia
Maria Rosa Valluzzi, Universita’ di Padova, Italy
Elizabeth Vintzileou, National Technical University of Athens, Greece
David Yeomans, ICOMOS, UK

XVII
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Conference sponsors

XIX
Invited papers
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Preservation of workmanship or workmanship for preservation

K. Van Balen & R. Hendrickx

K.U. Leuven, Civil Engineering dept., Leuven, Belgium

ABSTRACT: The chain of preservation should be integrated into the authenticity debate of conservation. A
particular role can be given to workmanship. A study on repair of ancient masonry structures of a fortress and
its execution show that beside analysis, the exchange with craftsmen is essential in the final decision on the
materials used for repair. Ongoing laboratory investigation revealed that when masons working from various
geographic locations are requested to assess workability of mortar, they seem to agree better than standardized
tests would indicate. This finding together and the previously explained experience indicate that workmanship
is to be considered into the preservation process, as many other interdisciplinary issues. Evaluating the heritage
values of an architectural fabric, craftsmanship embedded in it should be considered, but also the continuation
of practice has heritage values and is part of the preservation process.

1 INTRODUCTION Evaluating the heritage values of an architectural

In a previous contribution (Van Balen 2003) we have
advocated to integrate the chain of preservation into
the authenticity debate of conservation. A particular
role was identified for “workmanship”.
A study on repair of ancient masonry structures of
a fortress (Fort IV), part of the great defence works
in Belgium in the middle of the nineteenth century
in Mortsel, close to Antwerp in Belgium and the
execution of the repair work, have shown that not
only analysis is required for identifying mortar to be
used for repairing historic masonry, but that also the
exchange with craftsmen is essential in the final deci-
sion on the mortar composition to be used. The fortress
corridor around Antwerp was one of the largest con-
struction sites in Europe in the middle of the nineteenth
century. Documentation on the construction technolo-
gies used those days is available and was used in the
technical analysis preceding the conservation works.
In the second part the remarkable consensus on
workmanship of masonry mortar between masons
from various origins is explained. Ongoing laboratory
study on workmanship of mortars revealed that when
masons working in different geographic locations
(three from Flanders, one from Wallonia, one from
the Netherlands and one from France) are requested
to identify workability of mortar, they seem to agree
quite well. Their agreement seems to be more coher-
ent than standardized tests do. This finding together
with the previously explained experience indicates that
workmanship is to be considered into the preservation
process, as many other interdisciplinary issues.
fabric, craftsmanship embedded in it should be consid-
ered, but also the continuation of practice has heritage
values and is part of the preservation process. Not all
available workmanship is compatible with this pro-
cess but it can be made compatible if the context is
appropriate.
2 REPOINTING WALLS IN FORT IV
2.1 Construction of Fort IV
Fort IV was part of the defence system the Belgian
Government built in the middle of the XIX◦ C to make
a stronghold (National “Reduit” or bomb-free shelter)
in Antwerp for the ultimate retreat of its government in
case of Belgian’s occupation. Antwerp was chosen as
it was situated on the Scheldt River that could provide
an escape route to the open sea and to England (Gils
1997). Antwerp was also one marching day further
from the French and German border than Brussels.
A series of eight fortresses at the south and east of
Antwerp was planned in 1859, later in the nineteenth
century extensions of the system were planned at the
west side of the river and to the North East.
Construction works started in 1860 at Fort III,
shortly followed by the other forts based on the job
description made by the Ministry of War in 1859. This
document is a valuable tool of information to under-
stand the construction technology and the materials
used at that time. It should be understood that at that
time this was one of the largest construction sites in

3

Figure 1. Fort IV main front building (hoofdfrontgebouw)
in March 2004 (K. Van Balen).
Europe. It required a lot of building material, the orga-
nization of many brick factories in the clay ground in
the valley of the Scheldt River.
Fort IV was one of the last fortresses to be vacated by
the Belgian Army at the end of the twentieth century.
It is now property of the local town of Mortsel and
managed as a recreation area with social and cultural
functions. Also conservation of nature is considered
of great importance, sometime counteracting the need
for preservation of the built fabric. One example of
the latter is the conservation of the trees on the vaulted
underground construction while their roots damage the
original water evacuation system.
Within a EU Interreg IIIb project “Crossing the
Lines” a methodological approach of repointing has
been experimented within the project activities deal-
ing with restoration techniques. The findings hereafter
are results from that project in which the Department
of Civil Engineering and the Raymond Lemaire Inter-
national Centre for Conservation at the K.U. Leuven
were involved as requested by the Town of Mortsel.
2.2 Problem statement and methodology
The research presented here is about the re-pointing
of the facades of the fortress. Beside the study of the
re-pointing investigation was carried out on the pro-
duction of replacement bricks as well as on climate
control in these constructions with particular boundary
physical conditions: massive walls covered by earth
and with limited possibilities for aeration.
For the repair of parts of the masonry repointing was
necessary. The study and the application of the mortar
that would closely match with the existing mortar and
that could replace the original mortar, involved the
following steps:
1. Historical research on the used construction mate-
rials and on lime mortar technology.
4
2. Determination of the composition of the original
mortar through analysis.
3. Determination of the properties of the original and
repair mortar.
4. Formulation of the repair mortar based on above
conclusions and interpretations.
5. Hiring skilled craftsmen and suitable technology,
in order to assist developing and
6. Applying repair mortar.
2.2.1 Historical research
The 1859 job description and some comparative mate-
rial allowed understanding the types of materials used
and the way they were applied in the construction
of Fort IV. This information was however not univo-
cal as some of the descriptions could be interpreted
in different ways. It is also known from construc-
tion practice, that what is executed does not always
correspond with what was prescribed. Therefore anal-
ysis on the materials themselves could complete the
picture.
The work prescription included the following arti-
cles that are relevant for understanding the mortar
used:
Art 2: Mode of execution.
§9 Masonry: foundation walls and underground
masonry are made from stone (Fr.: moellons), from
concrete or bricks; all other masonry will be made from
bricks. Shall be used: ordinary mortar, cinder track
(Fr.: cendrée)-mortar or hydraulic mortar according to
the indications given by the commander of the engi-
neer corps. The vault covering (Fr.: chape) are made
from hydraulic mortar.
Art. 3: Materials to be delivered by the contractor:
§11 Lime – lime for masonry will be hydraulic with
similar quality as lime from quarries from the neigh-
bourhood of Tournai; reduced (Fr.: réduite) into putty,
plunged immediately in water, and subjected to the
needle test (Vicat Needle?) it will have to set after 24
hours under water.
§12 Sand – the necessary sand for the preparation
of the mortar will be siliceous, free from every mixing,
without clay, it will feel coarse and will be sieved (Fr.:
passé à la claie); sand for foundations will be without
mixing from earth or clay.
§13 Cinder track (Fr.: cendrée) – ashes will come
from factories and mainly from forges, it may not be
mixed with other substances; it will be sieved.
§14 Hydraulic basis – the substances that will serve
as a hydraulic basis will be subjected to the following
tests: . . . mixed evenly with ordinary mortar and sub-
jected to the needle test, they will have to set after 48
hours of submersion. . . .
§16 Bricks – bricks for a certain construction will
derive from the same producer (Fr.: fabrication), they
should have similar dimensions and appearance;
 §18 Mortars – ordinary mortars will be composed
of one part of slaked lime (Fr.: chaux éteinte) and one
part of sand.
Information on mortar composition could be sum-
marised as follows:
– Cinder track mortar will be made from two (2) parts
of extinguished lime, one (1) part of sand and one
(1) part of cinder track;
– Hydraulic mortar 1 (for drowned works) will be
composed of one (1) part of trass of Andernach (or
another hydraulic basis approved by the War Depart-
ment); (this trass is a pozzolanic material exracted
in Andernach, Germany)
– Hydraulic mortar 2 (for vault capping) will be made
of six (6) parts of slaked lime, one (1) part of sand
and four (4) parts of Andernach trass (or another
hydraulic basis approved by the War department);
Ambiguity and freedom of action is also included in
the descriptions as:
– In case exceptional circumstances require, the use
of other type of mortar or the dosage will be deter-
mined by the commander of the engineer corps.
– If the mortar is prepared by labourers, it will have
to be mixed and will be beaten during two days, it
will be used the third day after it has been re-beaten
(with as less as possible water) as to gain a degree
of liquidity (workability?) for its use.
– Mortar preparation can be accelerated using
machines; in that case the commander of the engi-
neer corps will judge on the modifications that have
to be carried out on the previous stipulation.
Those historical descriptions elucidate which mate-
rials were used, how they were mixed and in which
proportions. Aspects of workability could also be
found in conjunction with the preparation of the mor-
tar. As well workability as other circumstances allowed
for adjustments on the prescribed methods. It is clear
that other information available today is necessary to
decide on the composition of the repointing mortar.
2.2.2 Chemical analysis of mortar
Samples for further analysis were identified and were
extracted from the building.
– M1 Bedding mortar from sampled core 1.
– M1p Pointing mortar from sampled core 1.
– M2 Bedding mortar from sampled core 2.
– M4 Bedding mortar from sampled core 4.
– M5 Bedding mortar from sampled core 5.
– M7 Bedding mortar from sampled core 7.
– M8 Bedding mortar from sampled core 8.
– M8p Pointing mortar from sampled core 8.
Results of the chemical analysis are given in Table 1.
Proper interpretation of those results also requires the
5
evaluation of the thin sections which were made from
some samples.
From the analysis possible theoretical composition
of the original mortar could be estimated. Today’s
reference materials were used for the calculation
taking profit of a program developed for that pur-
pose. Based on the possible mixtures that fit with
the chemical analysis the following conclusions were
drawn:
1. There is a difference in mortar composition between
the bedding and the pointing mortar;
2. The bedding mortar has approx. 50 weight% of sand
and 50 weight% of binder and most mortars have a
moderately hydraulic binder (except M7).
3. The pointing mortar is richer in binder and as will
be explained later different sand with less glau-
conite was used. The pointing in an average has 67
weight% of binder and 33% of sand and the binder
is moderately hydraulic. The difference in sand may
also be an indication that pointing was carried out
in another construction phase.
In the next table a re-composition is given based
on the combination of lime hydrate and trass. In
the last column an alternative binder composition is
given for which a higher fraction of the binder can
be recomposed. The compositions are theoretical as
depending on the circumstances of hydration versus
carbonation but also depending on the composition
of the raw materials that have been used; the final
chemical composition may alter for a same mortar
composition.
From written historical sources we should con-
clude that normal mortar (including pointing?) should
be made from 1 volume of lime and 1 volume of
sand which in terms of weight proportions mean 30
weight% of dry lime hydrate and 70 weight% of sand.
This is different from the results found in the chem-
ical analysis that would correspond to 2.5 volumes
dry lime hydrate to 1 volume of sand. However the
description makes reference to lime putty, made from
hydraulic lime. This means that the density of the lime
will be higher and that the water fraction of the putty
is considered part of the binder.
The preparation: slaking, beating of lime for 2 days
includes the risk that if the lime contains a consider-
able amount of hydraulic and reactive particles, they
might hydrate and turn into granulates (they will not
be able to contribute to the binding anymore) which
changes the apparent proportions as identified from
the analysis.
2.2.3 Thin sections of mortar including
mortar-brick interface
The thin sections have been prepared from the samples
taken from the masonry. As a general outcome of the
Table 1. Chemical and physical analysis, ratio of sand and binder, and recomposition of bedding and pointing mortar samples.

Sample M1 M2 M4 M5 M7 M8 Mean COV M1p M8p Mean

Analysis
W loss 540◦ C (%) 7,47 6,41 5,77 8,30 5,72 5,51 7,26 8,05
W loss 1050◦ C (%) 17,84 19,30 17,77 18,83 13,44 17,31 22,10 22,90
Insoluble residu (%) 47,38 49,32 52,04 45,44 54,34 53,74 33,82 31,70
Soluble SiO2 (%) 5,80 4,46 3,88 5,97 6,80 3,52 5,17 5,77
CaO (%) 22,70 23,07 20,02 21,97 20,78 21,18 29,76 31,17
Total (%) 93,72 96,15 93,71 92,21 95,36 95,75 90,85 91,54
CO2 (%) 14,13 16,37 15,46 16,42 8,82 13,59 19,98 19,89
App. Density 1,814 1,800 1,608 1,785 1,624 1,751 1,819 1,731
(kg/dm3 )
Porosity (vol %) 32,29 33,00 40,62 33,59 38,66 34,05 31,94 35,90
Ratio of sand and binder
Sand (mass %) 47,40 49,30 52,00 45,40 54,30 53,70 50.35 6,48 66 68 67
Binder (mass %) 52,60 50,70 48,00 54,60 45,70 46,30 49,65 6,57 34 32 33
Recomposition
Hydraulicity mod. mod. mod. mod. hyd. mod. mod. mod.
Lime hydrate (%) 70 71 69 66 71 72 69,77 2,76 76 72 74
Trass (%) 19 14 14 19 25 13 17,40 24,55 14 14 14
Alternative binder CEM NHL 2 NHL 2 NHL NHL 5 NHL 2
II + L 3,5

analysis of the thin sections the following conclusions
were drawn:
1. Difference in mortar composition between the bed-
ding mortar and the pointing mortar is clear:
pointing mortar has a higher binder content and
finer fraction while the sand has almost no green-
ish (glauconite) grains that are found in the bedding
mortar.
2. Nodules of other materials, probably cinder track
are found in the mortar;
3. Adhesion between mortar and bricks is sometimes
very good and sometimes poor; this could be due
to a multitude of reasons, including the preparation
(polishing) of the thin section;
4. It seems that between bedding and pointing mor-
tar and in case of M8 between 2 layers of pointing
mortar some “deposit” can be seen; this may be an
indication that the pointing mortar has been applied
later, which could be consistent with the use of
another sand.
2.3 Dialogue with the workman
Masonry conservation of historic structures poses sub-
stantial dilemmas for architects, conservators, and
craft workers seeking to incorporate new mortars
with composition, performance, and characteristic
properties compatible to those of original mortars.
In the late 20th century and early 21st century
conservation philosophy embraces the incorporation
of original masonry materials and techniques where
possible. Properties of the mortar used for repair are
critical when in contact with the original remaining
mortar parts that have to be conserved. Since matching
the composition is no guarantee that the two mor-
tars will have compatible characteristics and future
performance, it is essential to work toward a better
understanding of original and repair mortar properties.
Therefore it is crucial that the craftsman is involved in
formulating a repair mortar, as he has to do the job
with products or materials he may not be acquainted
with.
Therefore, an important issue during the restoration
of the Fort IV was the repointing experiment of the
masonry walls as technical, historical and aesthetical
arguments were in play. However it may show how
important the intensive discussions with the crafts-
man before and during pointing were. In general we
state that due to the lack of participation (bottom-
up) within conservation, the craftsman is excluded
(top-down) from the general thinking and decision
making process. He’s familiar with compatibility and
reversibility concepts, yet he isn’t acquainted with
general conservation philosophies or interests.
As argued by K. Van Balen (Van Balen 2003) it
is clear that a uni-directional communication would
not lead to an authentic conservation intervention. In
order to avoid poor craftsmanship due to un-addressed
know-how, it’s important that the different parties dis-
cuss from the beginning the final technical, but also
the philosophical context of the conservation.
Architects could act as initiators of this debate in the
conservation team, as the present-day craftsmanship

6
training often focus on technical aspects and hardly
includes more theoretical and philosophical questions.
This early debate could help to assure the involvement
of proper craftsmanship.
In that respect the experiment in Fort IV was unique
as it tried to embrace the knowledge of a craftsman
in an early stage of the conservation process. So,
a dynamic interaction between pointer, experts and
architects arose as the pointer was asked to give pro-
posals (bottom-up) of possible mortar mixtures within
the guidelines formulated by the architect and the
guidelines resulting from the historical investigation
and the laboratory tests. In order to prevent a top-
down approach that obliges the pointer to work with
a certain mortar composition, that he doesn’t know or
isn’t familiar with, the strategy of trial and error was
followed.
During the iterative process comments of the crafts-
man were always collected after each evaluation of a
set of samples. Many comments were gathered during
conversation when the site works were ongoing.
For the proper conservation of the decorated
masonry, two types of pointing would be needed, a
dark and a light mortar. Further flush and cut-to shape
joints were requested.
After some preliminary discussions with the crafts-
man, a first series of three samples was proposed with
the following compositions (in volume ratio):
– Sample1: 1 part trass lime (German ready-mix prod-
uct with ≥55% trass) + 4 parts sand + ½ part “grey
lime” (name and product proposed by the craftsman)
– Sample 2: 1 part trass lime + ½ part grey lime + 3
parts sand + 1 part yellow sand
– Sample 3: 1 part trass lime + 3 part sand + 1 part
yellow sand + ½ part lime hydrate
According to the pointer the different samples had
more or less equal workability. An initial trial of sam-
ple 1 with 1 part of grey lime in stead of a ½ part
resulted in a fatty lime, and was therefore according
to the pointer unsuitable, as the binder-to-sand ratio
could have been too high.
After the evaluation of these samples a more greyish
colour mortar was requested by the architect. There-
fore a new set of samples was placed by the craftsman.
Meanwhile the craftsman was shown a video that
explained the use of lime in pointing mortars:
– Sample 4: 3 parts sand + 1 part trass lime
– Sample 5: 3 parts sand + 3 parts of Rhine Sand +
1 part trass lime + 1 part of grey lime
– Sample 6: 6 parts sand + 1 part trass lime + 1 part
of cement (CEM II/A LL 32,5 R)
– Sample 7: 6 parts of Sand + 1 part of trass lime +
1 part of lime hydrate
The pointer indicated that he normally would have
added some cement in sample 4 to create a bastard
Figure 2. Samples of repointing, pointing sample on top
right is sample nr. 2 which eventually has been chosen. On
the left sample 1, at the bottom sample 3.
mortar, which he preferentially applied in his projects
as such mortars feel less “fatty”. As the demand was
to use mortars without cement he proposed the same
composition by omitting cement. He confirmed that
this sample was quite difficult to work with, possibly
because the overall binder to sand ratio was quite low.
In his opinion Sample 5 was too coarse to create cut-
to shape joints. Yet, the sample had good workability,
but he was afraid to use another type of mortar for the
cut-to shape joints as colour differences would have
been possible.
The evaluation of the samples demanded for a last
set that could help to determine the final choice:
– Sample 8: 3 parts of Rhine sand + 1 part of trass
lime
– Sample 9: 5 parts of sand + 1 part of Rhine sand +
1 part of trass lime + 1part of lime hydrate
– Sample 10: 6 parts of Sand + 1 part of trass lime +
1 part of cement
Sample 8 was inspired by sample 4, yet with another
type of sand in order to obtain a colour that would
match better. Nevertheless the pointer had to admit that
this sample was too coarse to apply in the often narrow
joints. In order to make sample 5 less coarse he made
some adjustments to obtain the mortar composition of
sample 9. Although the workability improved, and it
was easier to apply in the narrow joints, it was still too
coarse to obtain a fair result.
Sample 10 finally illustrated the usual compositions
with a binder to sand ratio of 1:3 and the use of cement
and lime.
Eventually sample 2 and sample 6 were chosen for
the light and dark mortar joints.
As in one of the buildings (the left “capponière”)
the masonry showed a patchwork of light and dark
coloured bricks, the choice for the appropriate mortars

7

Figure 3. Samples of repointing: lowest sample is sample
nr. 6 which eventually has been chosen.
was less straightforward. It was decided to use the
light coloured mortar (nr. 2) for the complete façade.
The craftsman would have preferred the dark coloured
mortars, but he said this was just a matter of taste.
After the works were finished, the pointer admit-
ted that he was initially quite reluctant to use the trass
lime that was stipulated in the original job descrip-
tions of 1859, as he wasn’t familiar with the product.
He even had problems to find the appropriate prod-
uct as he only had found the first bags in a “garden
shop” close to Antwerp. Therefore it was important
that the pointer could “experiment” with the trass lime
in order to obtain insight in the workability properties
of the product, and that he could try several different
mixtures.
Eventually he was quite surprised that the trass lime
was easy to use and that a considerable strength was
obtained after some days. He mentioned that he has
been using trass lime since in other works after his
experience in Mortsel. So, one can argue the impor-
tance of initiating craftsmen in ‘forgotten’ traditions,
as the use of a certain material is important and that
the reluctance is often more related to habits than to
bad will.
The pointer applied in almost all the cases the
so-called “velco” sand, a product that originates after
washing the so-called Lommel sand. The product is
8
somewhat coarser than the Lommel-sand and there-
fore it is more difficult to make cut-to shape joints.
The craftsman was well aware of the problem, yet he
was afraid that another composition would differ too
much in colour from the approved samples.
The pointer experienced several problems when
brushing the joints, because in a prior phase the facades
were cleaned with a slightly abrasive sand-water mix-
ture. In his opinion several bricks were damaged at the
surface by this operation, the brushing of the joints pol-
luted the bricks with mortar remains that were difficult
to remove.
Some argue that many craftsmen are only inter-
ested in technical and aesthetical compatibility and
have few questions regarding the final conservation
philosophy or principles. Nevertheless, our craftsman
tried to identify peculiar problems, that in his opin-
ion could need another approach but were difficult to
judge as his training never included these more philo-
sophical issues. The most illustrative example that he
showed was the cut-to shape repointing of an area with
original brickwork. In this case the outer layer of the
bricks was damaged. The original joints had fallen out
and needed replacement. The craftsman struggled with
a technical-historical dilemma: would it be sensible to
create cut-to shape pointing as the outer layer of the
bricks had disappeared completely? His final judge-
ment was more practical by following the general rule
that window arches were originally with cut-to-shape
pointing.
It shows that the input from the craftsman, a bottom-
up strategy, from the beginning of the project can be
very useful to guide certain decisions.
Eventually the craftsman was quite satisfied with
the overall pointing result and with the guidance and
discussions during the project. Nevertheless he had to
admit that he had lost a lot of time at the beginning
when searching a correct pointing mixture. Further,
he found it quite difficult at the start to understand the
global framework of the different trials. He was very
satisfied that he could work fast and effectively with
the trass lime, a product he had never applied before.
It should be said that the craftsman had a very open
mind towards this experiment and that therefore he
kept on trying to improve his samples to obtain the
desired result.
3 INVESTIGATION WORKABILITY AND THE
ROLE OF WORKMANSHIP
In a very different project experiments were carried
out to connect measurements of workability according
to scientific laboratory procedures with a subjec-
tive judgement of workability by professionals. Some
of the experimental results are presented in another
contribution to this conference (Hendrickx 2008).
 Workability of fresh mortar is the sum of the appli- Table 2. Description of binders.
cation properties which provide its suitability, whether
for masonry purposes, plastering or jointing. Sev- No Description
eral authors and normalising institutions published
lists of these properties in varying order of impor- 1 Calcic lime CL90S
2 Calcic lime CL90S (EN 459-1:2001) (75%) with
tance (ASTM 2004, Kampff 1961 and RILEM 1978).
hydraulic binder (15%) and pozzolana (10%)
Recent literature about the rheology of mortar has 3 Calcic lime CL90S (EN 459-1:2001) with air
focused mainly on concrete. A variety of techniques entraining agent
have previously been used in laboratory measure- 4 Calcic lime CL90S (EN 459-1:2001) (66.7%) and
ments: rotational rheometry, compressive rheometry ordinary Portland cement CEM I 42.5 R (33.3%)
(squeeze tests), slump tests and alternative methods 5 Natural hydraulic lime NHL 5 (EN 459-1:2001)
like the rotating ball method or a trowel-imitating 6 Ordinary Portland cement CEM I 42.5 R
method (Banfill 2006). Standardised methods are 7 Ordinary Portland cement CEM I 42.5 R with
available to measure consistence, plasticity, water admixtures: plasticizer and air entraining agent
retention, air content, etc.. Most of the consistence
tests provide a measure of spread or slump, or pen-
Table 3. Dimensions of masonry units.
etration depth of a falling object.
Their drawback when compared to the rheological Commercial name and producer L D H
tests is that they are not directly related to the funda- solid/perforated (cm) (cm) (cm)
mental physical properties of the material. The practi-
cal difficulties in designing suitable measuring meth- Mono 3009, Vandemoortel (B) 19 9 5
ods for masonry mortar are: poor reproducibility due (solid)
to variations in components and preparation, impor- Porotherm, Wienerberger (B) 29 14 19
tant sensitivity to deformation history (thixotropy), (perforated)
and slip and plug flow in prevalent test geometries. Poroton Plan-T 14-30,0; 37.3 30 24
Wienerberger (G) (perforated)
Assessing measuring methods is possible by com-
paring results to the findings of experienced practition-
ers. An experimental programme in the framework of
a RILEM committee TC13-MR on mortars involving
5 masons led to the conclusion that
1. all craftsmen were able to attain the desired consis-
tence independently of the (dry) mix composition,
2. each craftsman has an individual ideal consistency,
3. all known test methods are inferior to the crafts-
men’s judgement and some are not suited for
mortars made with specific binders (RILEM 1978). Figure 4. Overview of trowel shapes used in experiments.
The type on the left was used by the Dutch mason; the type on
It has been shown that the often used flow table test the right by the French mason; the others by Belgian masons.
yields contradictory results for cement mortars and
lime mortars (Van Balen and Van Gemert 1991).
work with his own tools, which lead to the use of a
variety of trowels (Fig. 4).
In the first section of the programme, with quarry
3.1 Materials and methods
sand, the binder to aggregate ratios (B:A) were fixed,
The selected binders include a wide variety of lime- and the water to binder ratios (W:B) were left free
based binders and an ordinary Portland cement for the masons to choose. All ratios are given in mass
(Table 2). The units are fired clay bricks of three dif- proportions. The B:A were derived from a practice-
ferent sizes, two of which are perforated. Binder 7 is based reference.
used in 2 subsets of mortars with a different dosage Batches of approximately 35 litres were mixed
of the air-entrainer: 0.4% of the binder weight in 7a in a floor-model Hobart mixer type M80. Approxi-
and 0.1% of the binder weight in 7b. The plasticizer mately 75% of the estimated needed water quantity
is added at 0.5% of the binder weight. The additives was poured in the bowl, then half of the aggregate,
used with binder 7 are Rheomix 359 and Micro-air 100 the binder, and the other half of the aggregate was
(BASF). added. Mixing was done at low speed, while water was
The masons in the test panel all have at least 5 years added by the mason, until the water content and homo-
of experience with mortars containing lime. The panel geneity were considered optimal. The amount of added
is composed of three nationalities: Dutch (1), French water is weighed and the consistence of each compo-
(1) and Belgian (4). Each workman was allowed to sition is measured immediately by plunger penetration

9

⫾ Standard deviation
2.50
2.00
1.50
W:B
1.00
0.50
0.00
1 2 3 4 5 6 7a
B type
Figure 5. Mean selected water to binder mass ratio (W:B)
for mortar compositions with different binder (B) types.
(EN 459-2:2001). Each batch is tested on all three
brick types: both types of perforated bricks are used
to add three bricks to a single-leaf wallet, and the fac-
ing brick is used to produce small columns of eight
layers of three bricks in alternating bond. The masons
are continuously interviewed and filmed during the
procedure.
Reference compositions for lab experiments with
the quarry sand are calculated by taking into account
the average of the W:B chosen by the masons (Fig. 5).
They are tested for flow (prEN 1015-3:1998), plunger
penetration (prEN 1015-4:1998), bulk density and air
content (EN 459-2:2001). The mixing procedure is not
standardised, but attempts to approximate the proce-
dure from the masons’ programme (see above), using
a table model Hobart mixer: 1.5 min mixing, 0.5 min
scraping and homogenising, 2 min mixing. All mixing
is done at low speed.
In the second section, with standardised sand, W:B
and B:A are free to choose for the mason for the com-
position of small batches (1.2 liter). Lab experiments
with average values for W:B and B:A from these tests
sometimes yield visibly unworkable mortars. Hence it
was decided to discard them from the lab experimental
programme.
For description of the scientific laboratory tests to
define workability aspects of mortar we refer to the
other contribution (Hendrickx 2008).
3.2 Experimental results and discussion
In the first section of the programme (fixed B:A), the
selected W:B values for mortars with each binder have
a coefficient of variation of 3% to 10% between the
different masons for one specific binder (Fig. 5). A
systematic difference between some of the masons
is observed. A correlation between W:B and the spe-
cific surface area of the binder is found. Air entrained
2.50
1
2.00
W:B (kg/kg)
1.50
5
6 4
1.00 Non air entrained
2
Air entrained
7b
0.50 Linear (Air entrained)
Linear (Non air entrained)
0.00
0 2 4 6 8 10 12 14
SSA (BET) (m²/g)
Figure 6. Mean selected water to binder mass ratio (W:B)
7b as a function of specific surface area (SSA) of the binder in
the mortar.
mortars have lower values than non air entrained
mortars (Fig. 6).
The same trend in W:B is found in the second section
of the programme, although with larger variation due
to the varying B:A for each composition. The average
selected B:A are between 0.15 and 0.2 for binders 1, 2
and 3; between 0.2 and 0.3 for binders 4, 5, 7a and 7b,
and 0.4 for binder 6. The latter value is very high: it
appears that the test persons added a high quantity of
the ordinary Portland cement because a normal dosage
gives an impression of being poor in binder.
The lab experiments on reference mortars revealed
important differences in flow, yield stress, water reten-
tion, etc., although all mixes were brought to opti-
mum water content. Different performance can be
attributed to different behaviour of the binder particles
on micro-scale.
What is important, is that the qualitative judge-
ment of the different mortars is similar for all the
masons, and this to a remarkable extent.The small vari-
ation in ideal W:B ratio between the different masons
for each of the mortars is remarkable. It shows how
craftsmanship, knowledge and experience seem to be
a very good discriminator for mix optimisation and for
quality assessment.
The applicability and use of mortar for different
types of bricks and masonry was also taken into
account by the masons. For example the masons
revealed that Binder 2 had an important problem with
the air entraining agent, which causes swelling and
subsequent shrinkage of the mortar during mixing and
in the first 10 minutes after mixing. The same prob-
lem seemed present in binder 3 to a lesser extent. The
comments from the masons also depend on the type of
brick used. For the small and large perforated bricks,
it is judged important that the mortar adheres well to
the (higher) vertical sides, that it does not fall into the
perforations and that it allows a large unit to be manipu-
lated in the mortar bed without too much force. For the
small facing bricks, staining and floating of bricks are
more important criteria. Floating must be understood
10
as instability due to lack of stiffening of the mortar,
which should occur upon dewatering in contact with
the porous brick.
An inquiry about the relative importance that the
masons of the test panel attribute to different prop-
erties of mortar (in fresh and hardened state), reveals
three major concerns: workable time, adhesion to brick
and workability. Mechanical resistance and yield are
also considered important. Cost, thermal insulation,
frost resistance and environmental nuisance have low
appreciation.
The results of the scientific assessment of worka-
bility are given in (Hendrickx 2008).
3.3 Synthesis and conclusions
The variance in chosen W:B between the different
masons in practical tests is small. This indicates that a
marked transition in the mortar’s behaviour takes place
around a well-defined water content: from a granular
material to a liquid material, from a frictional regime
to a viscous regime. Small variations of water con-
tent around this transition point, give a very different
mortar for the user. It seems more easy to identify
that transition point with masons than using laboratory
experiments, not to say with standardized tests.
Laboratory research (Hendrickx 2008) indicates
that a combination of tests may be able to approach
the masons judgement of workability. A combination
is proposed of yield stress measurement with a vane
apparatus, water retention tests with vacuum suction
or filter plates, and density and air content tests to
characterise the workability of a mortar in detail.
4 PRESERVATION OF WORKMANSHIP OR
WORKMANSHIP FOR PRESERVATION
The first case study illustrates that beside technical
and scientific analysis workmanship is essential in the
preservation of architectural heritage. Good workman-
ship influences the results of the appearance of the
historic masonry after preservation (“Workmanship
for preservation”).
Historical analysis and a sensible way of dealing
with conservation bring to light how workmanship is
embedded in heritage and how the preservation action
itself is “heritage” (“Preservation of workmanship”).
The historical sources referred to, were very infor-
mative on the materials and techniques used in the
original construction and they do reveal how concerns
of workmanship were integrated in the construction
process. Those historical sources show how some
options were kept open to the judgement of the work-
man. Eventually workmen had a very important impact
on the composition of the material mixture used as well
as on the final state of the masonry when built.
11
In the conservation work in Fort IV also the tangible
results of workmanship from people involved in the
construction in the nineteenth century were preserved.
The origin of the research presented in the sec-
ond part comes from the problem encountered when
comparing the so called workability of different types
of mortar using standardized tests. Those tests have
started to live their own life and were considered
measurements of workability although it is noticed
that they contradict the judgement of masons once
leaving the (not explicit!) boundaries of the usual play-
ground for which the standards were made. Compari-
son of workability judgements of cement mortar versus
lime mortar with standardized tests is hardly possible
through the eyes of practitioners. Defining standards
today for “new” types of mortars (sometimes based on
very old recipes that demonstrated durability) required
from the researchers to go back to the real appreciators
of workability: skilled workmen. The research project
referred to, showed that a group of masons with differ-
ent background could remarkably well agree on what
was considered a workable mortar. This from one hand
demonstrated that it is possible to define workabil-
ity and at the other hand it demonstrated that skilled
people “inherited” the concept and understanding of
workability. It is the demonstration of the (intangible)
heritage based on practices, habits and tradition that is
still worth to be transferred to future generations.
As the preservation works on the walls at Fort
IV are the outcome of a dialogue strongly involving
the craftsman acknowledging respectfully his skills,
knowledge and traditions the works themselves have
contributed to the preservation of craftsmanship, to the
preservation (revival?) of a living tradition.
ACKNOWLEDGEMENT
Town of Mortsel, EU Interreg IIIb project “Crossing
the Lines”, Lhoist R&D.
REFERENCES
ASTM, 2004. ASTM C 270-04a Standard specification for
mortar for unit masonry, West Conshohocken: ASTM.
Banfill, P. 2005. The rheology of fresh mortar – a review,
In Cincotto, M.A., Silva, D.A., Oliveira, J., Roman, H.R
(ed.), Proceedings of the 6th Brazilian and 1st Interna-
tional Symposium on Mortar Technology, Florianopolis,
Brazil: 73–82.
Cools, S. 2006. Crossing the lines: study and repair of
historic masonry structures : case study: restoration of
Hoofdfrontgebouw and Caponnières at Fort 4 (2004–
2005), Unpublished Master thesis Raymond Lemaire
International Centre for Conservation, K.U. Leuven.
Gils, R. 1997. Vesting Antwerpen Deel 1, Bakstenen
Schoonheid 1830–1885, (België onder de wapens, 5),
Erpe-Mere: De Krijger
Hendrickx R., Minet J., Van Balen K. & Van Gemert D. 2008.
Workability of mortars with building lime: assessment
by a panel of masons versus lab testing, 14th Interna-
tional brick and block masonry conference, Sydney, 16–20
February 2008. (accepted for publishing on CD-ROM)
Hendrickx R., Van Balen K. & Van Gemert D. (this volume)
Assessing workability of mortar by means of rheological
parameters and desorptivity, SAHC 2008, Bath, 2–4 July
2008. Rotterdam: Balkema.
Kampff, L. 1961. Workability of masonry mortars. In Mate-
rials research and standards 1: 7–8. ASTM.
RILEM 13-MR committee on mortars and renderings 1978.
Tentitative recommendation. Materials and structures 11:
207–216.
12
Van Balen, K. & Van Gemert, D. 1991. Kalk in metselmortel.
Over het meten van verwerkbaarheid en hechting en over
de invloed van luchtbelvormers en superplastifieerders.
Voorlopig rapport (unpublished report). Katholieke
Universiteit Leuven.
Van Balen, K. 2003. From conservation principles to materi-
alization (or the other way around: how is materialization
guided by principles?). In Van Gemert D. (Ed.), Consol-
idation of Masonry. Advances in Materials Science and
Restoration 1: 135–144. Freiburg: Aedificatio Publishers.
Van Balen, K. & Van Gemert, D. 2005. Crossing the lines –
restoration techniques Fort IV, Mortsel (B) (unpublished
Laboratory report R/30299/05). Laboratorium Reyntjens,
Katholieke Universiteit Leuven.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Experiences from the Northern Italy 2004 earthquake: Vulnerability

assessment and strengthening of historic churches

Ezio Giuriani & Alessandra Marini

University of Brescia, Italy

ABSTRACT: This work is about the lesson learnt from the analysis of the damages caused to some churches
by the strong earthquake, which struck Northern Italy in the year 2004. Focus is paid to some mechanisms which
are not well known in the literature: the excessive and differential rocking of neighbouring transverse arches, the
differential deflection of single leave vault rings, the tie over-tension induced by the transverse arch rocking. In
the paper, the structural solutions appositely studied to repair and to reduce the structure seismic vulnerability
of these buildings are presented. Among these solutions: internal perimeter ties, lightweight wooden roof box
structure, lightweight spandrel ribs are illustrated.

1 INTRODUCTION however, in the case of long spanned building perime-

In the year 2004, a strong earthquake struck the Benaco
region in Northern Italy. The earthquake caused severe
damage to both historical and modern constructions.
Within the historical heritage about 250 churches were
listed among the most severely impaired buildings.
This work is about the lesson learnt from the anal-
ysis of the damages caused by the 2004 earthquake to
some churches, and discusses the structural solutions
appositely studied to repair and to reduce the structure
seismic vulnerability of these buildings.
During the damage survey campaign, some recur-
rent and classical collapse mechanisms were iden-
tified: the overturning of the main façade and the
compound walls, the out of plane deflection of the
perimeter walls between the transverse arches and
along the interstorey height, the wall in-plane shear
failure. These mechanisms are well discussed in the
literature (De Benedectis et al. 1993, Giuffrè 1993,
Magenes and Calvi 1997, Lagomarsino 1998, D’Ayala
and Speranza 2002, Griffith et al. 2003, Lagomarsino
et al. 2004).
In some churches, where the lateral wall over-
turning was inhibited either by massive buttresses
alongside the nave walls, or by thick abutments,
excessive, unconstrained rocking of transverse arch
pillars was frequently observed (Doglioni et al. 1994,
Lagomarsino et al. 1999, Giuriani et al. 2007 and
2008).
Against these well known failure mechanisms
perimeter ties placed along the masonry walls are the
classical solution. Alternatively, steel truss works are
often built on top of the roof pitches. As a drawback
ter ties might be ineffective; whereas with the adoption
of steel truss work high stresses might be concen-
trated in rather small zones and the effectiveness of the
strengthening solution should be carefully evaluated.
In some churches vulnerability was shown to be
associated to less frequent mechanisms, which are not
so well known and discussed in the literature.
In San Pietro Church (Roè Volciano, Brescia), one
of the most severely damaged churches, two single
leave barrel vaults collapsed during the earthquake
because of the differential rocking of the diaphragm
arches, and because of the differential deflection of the
thin vault rings. In other churches the over-tensioning
of the existing steel ties caused either their unthreading
from the masonry walls, from exceeding the anchor-
age resistance, the tie yielding (San Lorenzo Church
in Clibbio) or failure (San Antonio Church in Roè
Volciano).
Against these mechanisms, different solutions were
proposed: internal perimeter ties hidden behind the
nave mouldings, lightweight plywood roof box struc-
tures, and lightweight clay mortar spandrel ribs.
These strengthening techniques, as well as the above
mentioned collapse mechanisms are discussed in the
paper.
2 SEISMIC VULNERABILITY OF CHURCHES
The vulnerability of the churches is generally asso-
ciated to the overturning of the main façade and the
perimeter walls, as well as to the excessive uncon-
strained rocking of the transverse diaphragm arches.

13
 In the churches surveyed following the 2004 earth- Table 1. Seismic vulnerability assessment of some churches
quake, other mechanisms, such as the excessive and hit by the 2004 earthquake with respect to some failure
differential rocking of the neighbouring diaphragm mechanisms.
arches and the differential deflection of the thin vaults,
were acknowledged as further causes of the severe C(*) = 2.5% (A)
C = 1.0% (B)
damages observed. C = 3.5% (C)
The seismic vulnerability of the churches was C = 5.0% (D)
assessed by reference to the limit analysis approach. C = 5.0% (E)
C = 5.2% (F) (A)
Table 1 shows the vulnerability coefficients esti- C = 3.3% (G)
Madonna della Rocca,
Sabbio Chiese, Brescia.
mated for some significant case studies. Façade overturning
C = 7.8% (H)
C = 10.7% (I)

C = 5.0 % (A)
2.1 Overturning of the main façade C = 4.0% (B)
C=- (C)
Overturning of the main façade was often observed C=- (D) (B)
during the survey campaign (Figure 1, Table 1). This C=- (E) San Lorenzo Church,
collapse mechanism was caused by the seismic action C = 6.0% (F) Clibbio, Brescia.
of the wall and by the possible lateral seismic thrust of C=- (G)
Lateral wall overturning C=- (H)
the roof beams. Provided that wooden roofs are usu- C = 12% (I)
ally unconstrained against lateral movements, the latter C = 3.5% (A) (C)
contribution to the toppling action was quite relevant C = 7.0% (B) San Pietro Church
in most of the analyzed churches. C = 6.3% (C) (Roè Volciano, Brescia)
The onset of the overturning mechanism is under- C = <0 (D)
C=0 (E)
lined by the development of cracks extending at the C=- (F)
wall edges and progressively reducing their width from C = <0 (G)
the top to the footing of the building. The cracks can be Diaphragm arch C = 7.4% (H)
excessive rocking C = 48% (I) (D)
vertical or angled, depending of the degree of anchor- Sant’Antonio,
age to the perimeter walls. However, it is worth noting remarkable
(Roè Volciano, Brescia)
that, the measured crack angle was always very small, crack pattern
thus the stabilizing contribution of the lateral walls was surveyed in A,
B, D, E, G, H
almost negligible in most cases. Collapse
Diaphragm arch surveyed in C
differential rocking (E)
2.2 Excessive rocking of the diaphragm and San Rocco,
(Roè Volciano, Brescia)
triumphal arches and existing tie over-tension remarkable
crack pattern
Excessive, unconstrained rocking of the diaphragm surveyed in A,
and triumphal arches was recognised as another source B, D, E, G, H
Collapse
of vulnerability of many churches in the Benaco area surveyed in C
(Table 1). Single leave vault differ- (F)
ential bending Santissima Trinità,
In rest condition, when only vertical loads are (Roè Volciano, Brescia)
applied to the structure, the arch lateral thrust can
be either largely or even entirely resisted by the but-
tress action developed by the abutments, depending on
their shape and massiveness (Figure 2a, Giuriani and
Gubana 1993 and 1995). As a result the possible ties (H) (I) (G)
are required to confine the arch thrust portion, which SS. Faustino e Giovita SS. Pietro e Paolo Santa Maria Assunta
(Montemaderno, Brescia) (Preseglie, Brescia) (Bione, Brescia)
exceeds the buttress action. The proportion of the con-
( )
fining contribution mainly depends on the structure * FS = C WTOT
where: F S is the seismic force triggering the mechanism; C is the
geometry and tie pretension. Conversely, in the case of load collapse multiplier (also regarded as vulnerability index); W TOT
rocking, the resisting mechanism significantly modi- is the structure total weight.
fies (Figure 2b, Giuriani et al. 2007 and 2008). The
crack pattern is characterized by two large horizon-
tal cracks developing at the abutment bases and two
cracks opening at the diaphragm arch springing (Fig- of full rocking, when the crack pattern is fully devel-
ures 2b, 3 and 4). The location of the cracks depends on oped, no abutment buttress action can be accounted
both the structure geometry and the local resistance. for, because the resisting ideal struts become parallel
The abutment buttress action dramatically dimin- to each other. As a result, the arch thrust has to be
ishes as the crack penetration progresses. In the case entirely confined by the existing ties.

14
Figure 1. Overturning of the façade induced by the roof
seismic thrust.
a)
b)
Figure 2. Transverse arch resisting mechanisms under a)
vertical loads (rest condition) and b) horizontal loads (case
of full rocking, Giuriani et al. 2008).
Furthermore, following the cracking triggered by
the rocking motion, the span of ideal arch significantly
increases and so does the arch lateral thrust (Figure 2b).
The span of the ideal arch in rocking condition can be
conservatively assumed to be equal to the span of the
ideal arch in rest condition (L∗ ) plus the thickness of
the abutment (d) (Giuriani et al. 2007, 2008). For typ-
ical case studies the ratio d/L∗ was observed to range
15
Figure 3. Madonna della Rocca Church, Sabbio Chiese,
Italy. Transverse arch crack pattern induced by the seismic
action.
Figure 4. Detail of the horizontal crack pattern at the
transverse arch abutment base induced by the seismic action.
between 0, 15 ÷ 0, 2, thus when shifting from rest con-
dition to rocking the arch thrust could be 30 ÷ 40%
larger than the thrust at rest condition.
In the case of full rocking, the decrease of the abut-
ment buttress action and the increase in the ideal arch
span cause the existing tie over-tension. When the tie
resistance is barely sufficient to withstand the traction
force in rest condition, no extra resources are available
in case of earthquake, and the collapse of the tie can
be expected.
Figure 5. Detail of the welded tie rod in San Antonio Church
(Manerba, Brescia, Italy). The tie rod collapsed during the
2004 earthquake.

Given this foreword, the failure of the existing ties

in San Antonio Church (Roè Volciano, Figure 5) was
interpreted as a consequence of the onset of the trans-
verse arch full rocking mechanism. In some other
churches, where weak existing ties were also surveyed,
the over-tension caused either the tie unthreading from
the masonry walls from exceeding the anchorage resis-
tance (Madonna della Rocca church, Sabbio Chiese), Figure 6. Crack pattern induced by the differential trans-
or the tie yielding (San Lorenzo Church in Clibbio). verse arch rocking. Differential rocking, which is maximum
As for the vulnerability assessment, the evalua- in the first and last bay of the nave, is the result of the different
tion of the safety factor with respect to the onset and stiffness of the transverse arches.
developement of the mechanism can be obtained by
reference to the traditional approach of the principle
differential displacements can be expected between
of virtual work (Abruzzese and Lanni 1999). The same
adjacent transverse arches (Figure 6). Noticeable dif-
approach can be also addressed to define the structure
ferential displacement can also occur in the case
capacity curve (Housner 1963, Curti et al. 2006), to
of out-of-phase rocking of neighboring diaphragm
study the abutment out-of-phase rocking (Como et al.
arches.
1991), and to evaluate the tie tension (Lagomarsino
The largest differential displacements are usually
et al. 2004).
experienced by the first and last bay transverse arches.
A simplified method was proposed in Giuriani et al.
As a matter of fact, the façade rarely undergo rocking,
(2007 and 2008) for the evaluation of the collapse mul-
unlike the nearby transverse arch; and the triumphal
tiplier and the tie tension in the case of either over
arch is typically much stiffer than the neighboring
resistant or weak ties. The seismic collapse multiplier
transverse arch. This way, the vaults covering the first
was evaluated by introducing an ideal additional hori-
and last bays are subjected to shear distortion. When
zontal constraint at the tie level, and by enforcing that
differential rocking is severe, shear distortions can be
its reaction force R∗ was nil (Figure 2b).
remarkable and diagonal cracks may form in the vault
ring (Figure 6). The damage is more severe for thinner
vaults, like the single leave vaults. In the case of very
2.3 Differential rocking of the diaphragm arches
strong earthquake differential rocking may even cause
and differential deflection of the thin groin
the collapse of the vault.
vault
When subjected to earthquake loads, the nave vaults
The different stiffness of the transverse arches along also experience differential deflection of the vault
the nave can be regarded as another source of seis- groins (Figure 7). Traditional masonry groins are con-
mic vulnerability of the churches (Table 1). The nected to the perimeter walls at the imposts, whereas
maximum top displacement experienced by the trans- the vault rings are built adjacently to the perime-
verse arches during the rocking motion depends on ter walls. This result in the structural discontinuities
the geometry and stiffness of the structure, therefore shown in Figure 7.

16
 contact
Fo D Fo
a b
A B
detachement
vault support
at the springing
structural
discontinuity
Figure 7. Differential flexural mechanism of the groins of
the single leave vaults.
When subjected to the seismic actions Fo illustrated
in Figure 7, groins a and b rotates counterclockwise
around points A and B. the vault groin a pushes against
the perimeter wall, whereas groin b detaches from the
perimeter wall and the discontinuity width increases
for increasing seismic actions. As a result, the central
portion of the vault is subjected to concentrated differ-
ential deflection. The mechanism is amplified by the
rocking of the diaphragm arches.
In the case of very strong seismic action, cracks
induced by the differential deflection may extend
through the vault ring thickness causing its collapse.
Severe shear distortions induced by differential
rocking and differential deflection of the vault were
recognized as the major causes of total and partial col-
lapse of the single leave barrel vaults of the first and
second bay of the main nave of San Pietro Church in
Roè Volciano Brescia (Figure 8).
3 STRENGTHENING TECHNIQUES
In the following, a few anti-seismic strengthening tech-
niques are illustrated. The presented techniques were
adopted for the seismic upgrading of some churches
after the 2004 earthquake.
3.1 Internal perimeter ties
The adoption of perimeter tie system is a traditional
technique, which was effectively adopted to secure the
17
a)
b)
Figure 8. S.Pietro Church in Roè Vociano, Brescia, Italy:
a) total and partial collapse of barrel vaults of the first and
second bay of the main nave, and b) crack pattern induced
by the differential rocking and deflection of the single leave
groin vaults.
structure against wall overturning. Historic construc-
tions often present perimeter ties embedded within the
wall thickness.
When the tie confining effect is insufficient the tie
system must be strengthened or replaced. This oper-
ation is usually done by placing a new external tie
system, provided that the embedment of the ties within
the wall width would require expensive and difficult
drilling works (Figure 9a).
The static scheme of the external perimeter tie
system is illustrated in Figure 9c. For the solution
to be effective, the perimeter ties must confine the
resisting ideal horizontal arch, developing within the
masonry wall width, at the springing. This solution
was applied in San Pietro and Paolo Church, Preseglie,
Italy (Figure 10).
The same confining action can be effectively
obtained by adopting an internal tie system, hidden
on top of the nave mouldings (Figure 9d). This solu-
tion has the obvious advantage of being hidden from
the sight, but it may also serve in case of irregular
or preciously decorated external walls. The confine-
ment of the resisting natural arch can still be obtained,
provided that the inclined compressed strut forms in
 Lx
t min
a) b)
L x t min 10
Lx
c) d)
Figure 9. Introduction of internal or external perimeter ties.
Figure 10. San Lorenzo Church (Clibbio, Brescia, Italy):
external perimeter ties adopted as safety emergency provi-
sional solution following the 2004 earthquake.
the masonry intersections (Figure 9d). This solution
was applied in San Pietro Church, Roè Volciano, Italy
(Figure 11).
Perimeter horizontal steel ties are inadequate in
long-span buildings lacking strong transverse arches,
as the wall span-to-thickness ratio is unfavourable and
little constraint is provided to the toppling masonry
walls (Figure 9b). In this case, regardless of the
positioning of the perimeter ties, the resisting arch
is excessively low-raised and its resistance is neg-
ligible. In case of long spanned churches, roof box
structure, which are described in the following, were
alternatively adopted.
18
Figure 11. Detail of the internal perimeter ties hidden over
the nave moldings in San Pietro Church, Roè Volciano, Italy.
Figure 12. Floor and roof box structure preventing perime-
ter walls overturning.
3.2 Roof box structures
As shown in the previous section, most of the anal-
ysed churches were found to be vulnerable with respect
to overturning of the perimeter walls, as well as to
excessive and to differential rocking of the diaphragm
arch pillars (Table 1). In-plane shear resistant roof
diaphragms, transforming the building into a box-
structure, were adopted in these churches as a viable
solution to avoid or limit these mechanisms (Giuriani
and Marini 2006 and 2008).
Roof box structures behave like floor diaphragms,
forming a sort of lit to the underlying masonries, thus
preventing the perimeter wall overturning (Figure 12).
The roof pitches are transformed into folded plates,
which gather and transfer the horizontal seismic
actions of the lateral walls, the diaphragm-arches and
the roof to the shear resisting walls.
The roof box structure introduces an elastic con-
straint along the roof ridge, whose stiffness depends
on the horizontal flexural deformability of the roof
box structure (ye , Figure 13b). The box structure can
 roof box structure
ideal constraint 2 facade panel nailed
Wc CWc Wc connections

eaves
chords
CWA CW
FVA 1
W W

plywood
roof box panels
ye structure head walls or
lateral 2
Fo 3 walls z triumphal arch
x y
ye'
ye h plywood panels
existing planks
nailed
steel straps
L

a) b)

Figure 13. Structure behavior a) prior to and b) following

the introduction of the roof box structure preventing excessive
and differential transverse arch rocking.
be proportioned in order to confine the maximum mid
span displacement (ye ) and the masonry wall drift (β),
as well as to limit the shear distortion near the supports
(ye ’, Figure 13b). This way the box structure can be
an effective solution to the excessive and differential
rocking of the diaphragm arches.
The main structural elements composing the roof
box structure are: the pitch-panel (1), the eaves chords
(c13 ), the head gables (2) (Figure 14).
In order to design the roof box structure of the ana-
lyzed churches, the same techniques usually adopted
for the flexural strengthening of wooden floors, could
have been addressed (Giuriani 2004). Among all pos-
sible techniques, however, lighter solutions were pre-
ferred in order to prevent the increase of the dead loads,
and thus of the seismic action.
The wooden lightweight box structure proposed in
Giuriani and Marini (2008) were adopted. With the
proposed technique roof diaphragms were obtained
by placing overlaying plywood panels on the exist-
ing wooden roof pitches (Figure 14). Plywood panels
were connected to each other by means of nailed steel
flanges. The whole pitch diaphragms w nailed to the
perimeter steel eaves chords, and to both roof rafters
and masonry walls by means of steel studs and vertical
anchored bars.
The adopted lightweight wooden box structure tech-
nique is mainly reversible and minimally impairing
of the building integrity, and thus respectful of the
modern restoration principles.
As for the structure analysis and proportioning,
reference to Giuriani and Marini (2008) was made.
Figure 14. Wooden roof box structure by means of overlay-
ing nailed plywood panels. Principal structural component
and detail of the plywood panel assembly.
p1
unit stripe
a
p3
p3
1m
Figure 15. Seismic action distribution.
By reference to the load distribution shown in Fig-
ure 15, in order to analyze the roof box structure
behavior, a two step approach was adopted. In the
first step vertical and horizontal idealized additional
constraints, having r1z and r1y reaction respectively,
were introduced along the roof ridge line (Figure 16).
In the second step these provisional constraints were
removed and their reaction forces (r1z and r1y ) were
backed out.
In the first step, the unit-width stripe (a) with
additional vertical and horizontal constraints, behaves
like a frame undergoing the seismic actions (case A
Figure 16). In this phase, significant uplifting verti-
cal force per unit length nA may arise along the top

19
 g1 2
Ry c 13 f1y L x
艎12 p1 h1
f 1y = -r1y Ry = V1 =
2
1 f1y L2x
3 F13 Mmax =
8
a p3 A: FRAME B: BOX c 13
h3 BEHAVIOR STRUCTURE F13 Mmax Mmax f1y L2x
F13 = =
Lx 2 Ly 8Ly
Ly
z
Ly
x y
y r1z
g1* 2 Wg* head gable
h1 q1
3 fz
r1y f 1y = -r1y dy
q1 fz
4 2 Wg*
α zg head
h(y) q0 q0
∆W *g gable
wall
q0
A: FRAME fz −1
BEHAVIOR V1  L y  V1
q1 =  cos α  =
z
B: nA 2  2 cos α  Ly
5 1 x y BOX STRUCTURE
r1z  q 
q o =  1  cos α = q1
g1* zw W w*  cos α 
3 n A = (p1l12 h1 + p 3h 3h1 )L y − g1*l12 / 2 f z = 2q o tan α − w*g
r 1y
v A = p 3h 3 / 2
4 2 Figure 17. Simplified schemes for the evaluation of the
vA vA f1y = 2p1l12 + p 3h 3
nA nA force distribution in the box structure.

Figure 16. Structural behavior: a) Frame action and

b) box-structure undergoing horizontal ridge loads.
lateral wall. These possible tensile forces must be bal-
anced by the wall self-weight, through anchored bars
suitably embedded and distributed along the crowning
masonries.
In the second step, the additional ridge constraint
reactions were backed out with forces of equal inten-
sity and opposite sign (f1y , f1z in Figure 16). Provided
that Frame A would be a free mechanism with respect
to the horizontal force f1y , for the balance to be
restored, the roof box-structure must sustain these
forces (case B, Figure 16).
According to Giuriani and Marini (2008), the roof
box structure behavior was interpreted by reference
to the classical analysis of the ribbed panels, which
is based on the assumption that the in-plane bend-
ing moment and the shear force be decoupled and
resisted by the chords and panels, respectively (Bruhn
1973). Considering a gable roof, the ribbed panels box
structure is assumed to behave like a simply supported
beam undergoing distributed forces f1y , in which the
eaves chords withstand the global horizontal bending
moment, and the pitch panels resist to the shear forces
induced by the horizontal load (Figure 17).
The structure proportioning was based on the
resistance criteria, by enforcing the elastic behavior
throughout the design seismic event. It is worth noting,
however, that ductility was guaranteed at the ultimate
limit state by the ductile behavior of the nailed con-
nections (Giuriani and Marini 2008), which are the
weakest link of the structure.
The expressions of the most significant internal
forces are shown in Figures 16 and 17. The eaves
chord cross section was proportioned to resist the eaves
chord maximum axial forces (F13 ); whereas the pitch
panel thickness, the panel mutual connection, and the
diaphragm connections to both the head gables and the
eaves chords were proportioned to resist the maximum
shear flow q1 .
Horizontal and vertical steel stud connections were
adopted to allow shear transferring to the shear resist-
ing head gable and longitudinal walls, respectively
(Figure 18). For the proportioning of the stud con-
nection, mindful attention was paid to the reduced
shear resistance of the masonry wall crown caused
by the lack of vertical confining load. To increase
the stud connection resistance, strengthening of the
masonry wall crown was sometime necessary. To this
end, injections of clay mortar along the perimeter
wall crown served the purpose in most cases. Alter-
natively, the whole masonry crown was replaced, or a
thin 20 ÷ 40 mm slab of high performance clay mor-
tar stiffened by means of plaster meshes was cast on
top of the masonry walls. Experimental results proving
the effectiveness of these techniques are available in
the literature (Gattesco and Del Piccolo 1998; Giuriani
2004; Tengattini et al. 2006).
Distributed vertical deep anchor rebars, embedded
along the head gable crowning masonries, were needed

20
 stud
connections
head gable chord
facade head gable
a)
plywood panels
eaves chord
stud
connections
top perimeter
masonry
strengthened
by means
deep
of injected
anchorages
clay masonry.
b)
Figure 18. Detail of the connection a) to the facade head
gable and b) to the longitudinal walls.
in some restoration works to confine the uplift traction
forces per unit length nA and fz induced by the hori-
zontal loadings (Figure 16). The tensile forces must be
balanced by the self-weight of the “lifted masonry”.
The distributed anchor bar embedment length was
carefully proportioned for the weight of the “lifted
masonry” to be sufficiently larger than the traction
forces.
Besides the structure bearing capacity, the appraisal
of the roof structure deformability was sometime nec-
essary. Unlike perimeter walls, which can usually
endure large out-of-plane displacements with little
damage, excessive in-plane shear deformability of
the masonry vaults and unconstrained drift of the
diaphragm arch pillars were shown to be responsible
of serious damages and even partial collapses in some
churches. To this end, restrictions were enforced that
the roof box structure lateral deflection ye was smaller
than the maximum allowable drift of the diaphragm
arch pillars; and the slope of the box structure trans-
verse deflection at the head gable ye was smaller than
the maximum shear deformation γ ∗ allowed by the
vaults.
For the evaluation of the horizontal deformability of
the plywood box structure, the analytical model based
on the principle of virtual works proposed in Giuri-
ani and Marini (2008) was adopted. In the model,
in order to account for the nailed connection shear
slip, which largely affect the global deformability,
21
Figure 19. General view of the anti-seismic plywood roof
box structure in San Pietro Church (Roè Volciano, Bresica,
Italy) and detail of the eaves chord connected to the underly-
ing masonry by means of steel studs and deep anchorages.
reduced equivalent young and shear elastic moduli are
entered.
The lightweight wooden box structure applied for
the anti-seismic retrofit of the San Pietro church is
shown in Figure 19.
3.3 Light spandrel ribs
In San Pietro Church (Roè Volciano, Brescia), covered
by a single leave masonry vault, the strengthening of
the structure against the differential deflection of the
groin vault was necessary (Table 1).
In masonry vaults, which are incapable to resist ten-
sile stresses, equilibrium is guaranteed by the ideal
arch developing within the vault thickness. The ideal
arch corresponds to the anti-funicular of the set of
applied loads. After cracking, the ideal arch must cross
the solid part of each cracked sections, both at the vault
key and at the springing. Single leave vaults are usually
very thin, thus the ideal resisting arch has little pos-
sibility to shift and modify within the vault thickness
to adapt to different unsymmetrical load distributions.
As a result, the structure is often quite vulnerable with
respect to seismic load distributions, which require
pronounced deviation of the ideal arch.
In order to enhance the structure resistance, span-
drel masonry walls are traditionally proposed. The
structure resistance is increased by increasing the
thickness of the vaults, thus allowing the ideal resisting
arch to adjust within the spandrel wall thickness.
In the case of single leave thin vaults, such as in
San Pietro church, attention was paid to avoid dead
load increase, which in turn could result in additional
seismic actions.Accordingly, lightweight spandrel ribs
were proposed as a new vault strengthening technique.
Spandrel ribs were designed to resist both compres-
sive axial forces and bending moment. This way, the
flexural stiffness of the structure was greatly enhanced.
The lightweight spandrel ribs tubular cross section
is shown in Figure 20. The resisting cross section is
made of clay mortar reinforced with plaster meshes.
 lightweight
spandrel ribs polystyrene ribs
cross section
clay plaster
plaster meshes

single leave vault

lightweight
spandrel ribs
Figure 22. San Pietro Church (Roè Volciano, Brescia): light
spandrel walls strengthening the single leave masonry dome
covering the presbitery.

Figure 20. San Pietro Church (Roè Volciano, Brescia): dis-

tribution of the lightweight spandrel ribs along the extrados
of the nave vault and detail of the spandrel rib cross section.

Figure 23. Experimental test on light spandrel walls (Pisa

Laboratory, University of Brescia, Italy).

The same technique was applied to strengthen the

Figure 21. Light spandrel ribs construction phases: a) a
layer of plaster mesh is placed overlaying the vault extra-
dos; b) the mesh is covered with a layer of clay mortar;
c) polystyrene ribs are placed along the vault ring and covered
by plaster meshes; d) a thin finishing layer of clay mortar is
spread over the structure.
single leave masonry dome covering the presbitery
(Figure 22).
The efficiency of the proposed technique was also
verified experimentally (Figure 23). The structure was
subjected to cyclic, unsymmetrical load conditions.
Good results were obtained both in terms of stiffness
and strength (Giuriani et al. 2007).

The inner lightweight core is made of polystyrene
elements.
In San Pietro Church, spandrel ribs were placed
along the vault extrados as shown in Figure 20. The
construction phases are briefly illustrated in Figure 21.
4 CONCLUDING REMARKS
The analysis of the damages caused by the 2004 earth-
quake to some churches, as well as the structural solu-
tions proposed to reduce their structure vulnerability
were discussed in the paper.

22
 The most significant results of the study can be
summarized as follows:
• In the case of full rocking, the decrease of the abut-
ment buttress action and the increase in the ideal
arch span can cause over-tension of the existing
ties. In the case of very weak ties, collapse of the
tie can be expected. In this case, the adoption of a
roof box structure, limiting the rocking of the trans-
verse arches, together with the strengthening of the
existing tie system might significantly reduce the
structure vulnerability.
• Differential rocking of transverse arch pillars
induces shear distortion within the vault ring, espe-
cially in the first and last bay of the nave. Depending
on the magnitude of the differential rocking, large
diagonal cracks my extend over the vault leading up
to its collapse. In this scenario, the structure vul-
nerability can be reduced by adopting a roof box
structure and by limiting the shear distortion near
the supports.
• In the case of a seismic event, pronounced differ-
ential deflection can be expected along the central
part of thin vaults. Depending on the severity of
the earthquake, differential deflection can cause the
collapse of single leave vaults. Against this failure
mechanism, lightweight spandrel ribs were pro-
posed. This solution allows to significantly enhance
the stiffness and strength of the vaults.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge Claudia Porteri
and Lorenzo Giuriani for their fundamental contribu-
tion to the vulnerability assessment of the churches
and to the design work.
This work was developed within the research project
DPC-ReLUIS 2005–2008, Research line n. 1: “Vul-
nerability assessment and anti-seismic strengthening
of masonry buildings”. ReLuis and MURST financial
contributions are gratefully acknowledged.
REFERENCES
Abruzzese, D., Lanni, G. (1999). Some developments on the
lateral strength of historical reinforced vaulted buildings.
In Structural Studies, Repairs and Mainteinance of his-
torical buildings, VI International Conference, STREMA,
Brebbia and Jager Ed., Dresda.
Bruhn E.F. (1973). Analysis and design of flight vehicle
structures. Jacobs Publishing Inc.
Como, M., Lanni, G., Sacco, E. (1991). Sul calcolo delle
catene di rinforzo negli edifici in muratura soggetti ad
azione sismica. VII convegno nazionale L’ingegneria
sismica in Italia, ANIDIS, Palermo.
De Benedectis, R., De Felice, G., Giuffrè, A., (1993).
Restauro antisismico di un edificio. In: Giuffrè A.,
23
Sicurezza e conservazione dei centri storici: Il caso
Ortigia, pag. 189–205. Editrice Laterza, Bari.
Curti, E., Lemme, A., Podestà, S., Risemini, S. (2006).
Criteri di verifica per la progettazione di interventi di
miglioramento sismico di edifici monumentali. Ingegne-
ria Sismica, Anno XXIII, n. 1 gennaio-aprile.
D’Ayala, D. and Speranza, E. (2002). An integrated proce-
dure for the assessment of seismic vulnerability of historic
buildings. In Proceedings of the 12th European Con-
ference on Earthquake Engineering, London, Elsevier
Science, paper n. 561 (CD-ROM).
Doglioni, F., Moretti, A., Petrini, V. (1994). Le chiese e il
terremoto, Trieste, Edizioni Lint.
Gattesco N., Del Piccolo G. (1998). Shear transfer between
concrete members and stone masonry wall through driver
dowels. European Earthquake Engineering, Vol. 1.
Giuffrè, A. (1993). Sicurezza e Conservazione dei centri
storici: il caso Ortigia, Bari, Ed. Laterza.
Giuriani, E., Cubana, A. (1993). Recupero e consolida-
mento di volte in muratura. In Atti dei Colloqui Inter-
nazionali Castelli e Città fortificate,Storia Recupero,
Valorizzazione. Palmanova, 24–25 settembre.
Giuriani, E., Gubana, A., (1995). Extrados ties for structural
restoration of Vault. In Proceedings of the Fourth Inter-
national Conference on Structural Studies of Historical
Buildings Strema 95, Architectural Studies, Materials and
Analysis, Vol. 1, Ed. C.A. Brebbia, Wessex Institute of
Technology, UK.
Giuriani E. (2004). L’organizzazione degli impalcati per
gli edifici storici. L’Edilizia. Speciale Legno strutturale,
N. 134. (In Italian).
Giuriani E., Marini A., Porteri C., Preti M. (2007) “Don-
dolio degli archi diaframma e vulnerabilità sismica delle
chiese”. Technical Report n.7, Università degli Studi di
Brescia.
Giuriani E., Marini A., (2008). Wooden roof box structure
for the anti-seismic strengthening of historic buildings.
Accepted for publication in the International Journal of
Architectural Heritage.
Giuriani E., Marini A., Porteri C., Preti M. (2008) “Vulnera-
bility of churches associated to transverse arch rocking”.
Submitted to the International Journal of Architectural
Heritage for possible publication.
Griffith, M.C., Magenes, G., Melis, G., Picchi, L. (2003).
Evaluation of out-of-plane stability of unreinforced
masonry walls subjected to seismic excitation. Jour-
nal of Earthquake Engineering, Vol. 7, Special Issue 1,
pp. 141–169.
Housner, G. W. (1963). The Behaviour Of inverted pen-
dulum structures during earthquakes. Bulletin of the
seismological Society of America, 53 (2).
Magenes, G., Calvi, G.M. (1997). In-plane seismic response
of brick masonry walls. Earthquake engineering and
Structural Dynamics, Vol. 26, pp. 1091–1112.
Marini A. and Giuriani E. 2006. Transformation of
wooden roof pitches into antiseismic shear resistance
diaphragms. V International Conference on Structural
Analysis of Historical Constructions, November 6–8.
New Delhi. MacMillan Ed. ISBN. 10: 1403-93156-9.
pp. 445–452.
Lagomarsino (1998). A new methodology for the post earth-
quake investigation of ancient churches. In Proceedings
of 11th European Conference on Earthquake Engineering,
Balkema, Rotterdam, ISBN 90 5410 982 3.
Lagomarsino, S., Brun, S., Giovinazzi, S., Idri, C., Penna, A.,
Podestà, S., Resemini, S., Rossi, B. (1999). Modelli di
calcolo per il miglioramento sismico delle chiese. In
Proceedings of 8th Italian Conference on Earthquake
Engineering.
Lagomarsino, S., Podestà, S., Risemini, S., Curti, E.,
Parodi, S. (2004). Mechanical models for the seismic vul-
nerability assessment of churches. In Structural analysis
24
of historical construction (C. Modena, P. B. Lourenço,
P. Roca Eds.), Proc. of IV Int. Seminar SAHC, Padova,
Italy, A.A. Balkema, London (UK), Vol. 2, pp. 1091–1101.
Tengattini C.G., Marini A., Giuriani E. (2006). Connessioni
a taglio nelle murature. Tecnichal Report 3a.1-UR11-
1 RELUIS – Progetto di ricerca N.1 – Vulnerability
assessment and anti-seismic strengthening of masonry
buildings.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

20th century curtain walls – loss of redundancy and increase in complexity

S.J. Kelley
Wiss, Janney, Elstner Associates, Inc. Chicago, Illinois, USA

ABSTRACT: Central to the development of the American skyscraper was the concept of the curtain wall. This
development was rapid and was driven by the distinct cultural climate within American urban centers such as
Chicago and New York. Skyscrapers would not have been technically feasible without the lightweight curtain
wall, and it was through the skyscraper that the curtain wall achieved its greatest realization. The principal and
confrontational factors of its development were economics and the need for fireproof buildings. The historical
development of the curtain wall is presented and a comparison of deterioration mechanisms between masonry
and metal and glass curtain walls is discussed. Finally the issues of loss or redundancy and increase in complexity
of contemporary curtain walls are discussed and strategies for diagnosis and conservation are presented.

1 INTRODUCTION

One of Oscar Wilde’s aphorisms regards the United

States: “America is the only country that went
from barbarism to decadence without civilization in
between.” The truth in these bon mots, at least with the
topic of curtain walls, was that curtain wall typology
would evolve as quickly as building codes would allow
to accommodate ever-taller buildings in America’s
wild skyscraper boom.
Central to the development of the American
skyscraper was the concept of the curtain wall, which
opened new avenues for architectural expression. As
defined in the Kidder-Parker Architects’and Builders’
Handbook in the first half of the 20th Century, a cur-
tain wall is “an enclosing wall built and supported
between columns or piers, and on girders or other sup-
port, and sustaining no weight other than its own.”
Spanning between support points at floor levels, its
primary functions are to provide weathertightness,
provide a fire wall, and transfer laterally-induced loads
to the structural frame. The curtain wall does not have
the limitations of a bearing wall so that more wall area
can be opened up for glazing.
Numerous factors of change were involved with Figure 1. Installation of the prefabricated steel curtain wall
the chief one being economic – buildings had to be at 860–880 Lake Shore Drive in Chicago, USA, one of the
frst residential skyscrapers to be clad entirely in glass and
built ever taller, faster, and cheaper. These factors
metal (Mies van der Rohe, 1949–1951).
were basic for a quickly growing and resource-rich
country. A reactive factor was the need for fireproof
buildings – this was not surprising in a country plagued changed dramatically during the course of the 20th
by urban fires throughout the 19th Century. From wall Century. This curtain wall evolution resulted in shift-
construction that consisted of heavy masonry to con- ing the functional burden onto thinner and different
temporary metal and glass, curtain wall components materials.

25
2 DUELING FACTORS IN AMERICAN
CONSTRUCTION: ECONOMY AND FIRE
SAFETY
Economy in construction of buildings is an axiom
of modern construction worldwide and the excep-
tions that one may cite would only prove the rule. In
America this axiom was very specific and tied to the
development of the American West.
Consider the development of the balloon frame, a
distinctly American framing technique that fulfilled
the requirements of a young and rapidly expanding
country. In a balloon frame structure, studs run con-
tinuously from the sill to the second floor top plate
which supports the roof structure. The second floor
joists are nailed to the studs and rest on ledger boards
that are fit flush into the studs. This system relies on
the external sheathing rather than triangular bracing
for lateral strength.
It was characterized by the use of plentiful wood
from America’s once ubiquitous forests, rapid con-
struction, and the limited skill required by the workers.
The advent of cheap machine-made nails, along with
water-powered sawmills in the early 19th century made
balloon framing highly attractive. Standard wood ele-
ments such as studs were readily available. Although
lumber was plentiful in 19th Century America, skilled
labor was not. Relatively unskilled workmen were able
to erect balloon frames quickly because the labor-
intensive mortise and tenon connection of the braced
frame technique was supplanted by the use of cut
nails. Without balloon framing, the western boom-
towns of America and Canada certainly could not have
blossomed overnight.
These important economic factors of construction –
namely rapid construction and limited skill required by
the worker – set a trend for American and then World
construction of large buildings in 20th century.
Another important but reactive factor was the need
for fireproof buildings as can be seen in Chicago.
Chicago’s building history recommenced after 1871,
when the Great Chicago fire devastated the City.
Despite this catastrophe – and due to a strong local
economy, the high cost of land, and rapidly evolving
building technologies – Chicago quickly rose like the
Phoenix and afforded an opportunity for the realization
of innovations in architecture, engineering, and con-
struction which established it as the birthplace of the
skyscraper. The answer of who invented the skyscraper
is debated to this day and depends entirely upon what
are considered to be the defining characteristics. Com-
parative characteristics are the separation of the wall
from the frame, the first use of the iron frame, appear-
ance of the beam-column moment connection, height
limits, use of the passenger elevator, and theories of
frame stiffness. For this paper we use the wall and
frame separation definition.
26
These structures needed to be fireproofed and
the urban fires revealed that iron work, though
inflammable would yield and fail beneath the flames.
George H. Johnson, an English-educated archi-
tect working with the Architectural Iron Works
of New York founded a business to manufacture the
fireproof terra cotta tile that he patented in 1871. On
trips Johnson made to Chicago to promote the sale of
his fireproof tile may have become a precedent for the
William leBaron Jenney’s 11-story Home Insurance
Building (Chicago, 1885). The use of masonry, and
lots of it, in skyscraper construction would prove to be
difficult to move beyond in curtain wall development
due to its fire resistant qualities.
Following these 19th Century building trends came
the development of Modern Movement technologies
characterized by experimentation and innovation in
construction materials and techniques, many of which
had neither the resilience nor the longevity of tradi-
tional construction. Materials such as cinder concrete,
plywood, and particleboard; and systems including
panelized construction, and thin curtain walls with lit-
tle redundancy came into use. This experimentation
with materials and systems remains unabated today.
3 THE DEVELOPMENT OF THE
SKYSCRAPER TYPOLOGY IN AMERICA
The curtain wall dichotomy can be traced back to
numerous 19th and early 20th Century antecedents.
In the United States curtain wall development became
intertwined with that of the skeleton frame. Skyscrap-
ers would not have been technically feasible without
the lightweight curtain wall. And it was through the
skyscraper that the curtain wall achieved its greatest
realization.
3.1 The skeleton frame
In the nineteenth century, engineers first utilized metal
frame construction in bridges, factories, and ware-
houses, and cast and wrought iron were the major
metals used in construction. The invention of the
Bessemer process in England in 1856 made it pos-
sible to produce large quantities of steel affordably.
In the United States, steel production on a large scale
was realized in the 1870s. The transition from iron
to steel was gradual, with iron still used in building
construction as late as the 1890s.
Arguably, the first metal skeleton-framed skyscraper
was the Home Insurance Building designed by Jenney
in Chicago. In this structure, column loads were trans-
ferred to stone pier footings via the metal frame
without load-bearing masonry walls. Each level of
the exterior wall was supported on a shelf angle fixed
to the spandrel girder. At that time skyscrapers were
designed without lateral bracing under the assumption

Figure 2. The 20 story Masonic Temple in Chicago with a
laterally braced steel frame clad in a masonry curtain wall
skyscraper (Burnham and Root, 1892).
that the heavy masonry cladding provided sufficient
rigidity for the whole structure.
As skeleton framing came into common use,
masonry bearing wall construction reached its prac-
tical limit. Burnham and Root’s 16-story Monadnock
Block (Chicago, 1891) utilized traditional load-
bearing masonry walls which at grade were almost 2
meters thick. This building also utilized the first rigid
frame for lateral stiffness. At the same time, Burn-
ham and Root developed a complete steel frame for
the Rand McNally Building (Chicago, 1890). They
also developed a steel frame laterally stiffened with
a diagonal bracing system in the 20-story Masonic
Temple (Figure 2). Within a three year period the
same firm had erected a bearing wall; steel frame; and
a diagonal-braced metal frame skyscraper revealing
the experimental and quickly evolving nature of the
typology.
Several years later, the structural innovations of the
Chicago school were taken further by D.H. Burnham
in the Reliance Building (Chicago, 1895). The exte-
rior bays were designed as rigid steel frames, and
two-story columns erected with staggered joints fur-
ther increased frame rigidity. The construction of
the Reliance Building also illustrated all of the
27
Figure 3. The New York Daily News building with a
masonry curtain wall that has been completely stripped of
detail as a cost measure (Hood and Howells, 1930).
familiar construction techniques which had come into
widespread use at the end of the 19th century to facil-
itate rapid construction. The construction site was lit
so work could continue into the night. Work spaces
were enclosed and heated so the project could proceed
during the winter months. The structural frame, which
began erection in mid July of 1894, was topped off on
August 1 and required only 15 days.
Steel and then concrete skeleton framing soon
became universally accepted for skyscrapers. There-
after, improvements of known design methods encour-
aged the construction of increasingly taller buildings.
4 THE MASONRY CURTAIN WALL AND ITS
DETERIORATION MECHANISMS
A good place to begin discussion of the early curtain
wall is with the Reliance Building in Chicago, the first
skyscraper to fully utilize terra cotta as a cladding. The
terra cotta units of the curtain wall are connected to a
gridwork of cast-iron mullions, lintels, and sills which
span between levels. Unlike the Home Insurance and
other similar buildings, the Reliance frame did not rely
upon the masonry curtain wall for lateral support.
In New York City the steel frame and masonry cur-
tain wall became established with the American Surety
Building in New York City (Price, 1894), and, once
adopted, skyscraper heights increased dramatically in
that city. The once impressive twenty-story buildings
of Chicago were overshadowed by buildings 100, 200,
and finally 238 meters with the Woolworth Building
(Cass Gilbert, 1913). The Woolworth building utilized
the latest developments in steel frame construction, but
its curtain wall had not abandoned the use of masonry
construction and ornament.
The First World War administered the final blow
to the arts and crafts movement in Europe, and the
machine became the basis of a new architecture. Mod-
ern European architecture required that the labor of
producing the parts be performed in the factory rather
than by craftsmen on site. German intellectuals were
in awe of the example of the American skyscraper,
a strong symbol of the new world for which they
endeavored. Ludwig Mies van der Rohe, prior to relo-
cating in Chicago, prepared a series of unrealized
projects in which the most famous came to be known
as the “Glass Skyscraper,” a highrise enveloped totally
in glass.
Contemporary with the European modernists,
America had entered a second skyscraper era. The post
World War I period brought a demand for increased
speed in design and erection. Curtain wall con-
struction, however, continued to utilize the masonry
techniques that had been developed by the turn-of-the-
century. Though mass production and standardization
had begun to impact the details of architecture, the
use of prefabrication and the module were not yet
extensively used.
H. R. Dowsell, one of the architects of the Empire
State Building (New York City, 1931) wrote of the
masonry curtain wall, “Tradition has clung to the her-
itage of thick masonry walls. We inherited masonry
walls and seem unable to outgrow our inheritance.
The idea that masonry is the only form of permanent
construction was so deeply rooted that practically all
building codes made masonry walls mandatory . . .”
(Engineering News Record, 19 February 1931).
The masonry trend continued into the Great Depres-
sion while stripping away any costly ornamenta-
tion. Examples would include the New York Daily
News (Figure 3), McGraw-Hill building (Hood and
Fouilhoux, 1932), and the PSFS Building (Howe and
Lescase, 1931). However, due to the Great Depression
and the Second World War, construction grinded to a
halt for almost 20 years.
Masonry curtain walls that we care for today con-
sist of facing material backed by several wythes of
brick or clay tile, for a total wall thickness of as much
as 30 cm or more. Metal shelf angles to support the
masonry infill are anchored to the structure at the floor
levels. Flashing was not commonly used and corrosion
of the shelf angles and other metal embedments is a
significant cause of distress.
28
Another issue is the differential movement of the
curtain wall relative to the structural frame. By 1894,
lateral movement in curtain wall construction was
actually being studied and analyzed. The masonry
cladding is also exposed to temperature-related move-
ments, while the embedded frame is protected. Struc-
tural frames shorten under dead load and material
creep. Conversely, fired-clay masonry curtain walls
expand due to the intake of moisture. Early curtain
walls were not built to accommodate these differential
movements resulting in the introduction of unantic-
ipated stresses into the curtain wall and frame. This
problem came to be understood as evidenced by dis-
placement measurements that were performed during
construction of the Empire State building in 1931. The
horizontal deflection of the top of the Empire State
Building was monitored by the American Institute of
Steel Construction. Measurements were also made to
determine exactly how much lower the various floors
were from their theoretical position. These measure-
ments showed that the 85th floor was 16 cm. below
its theoretical elevation. (H.G. Balcom, “New York’s
Tallest Skyscraper,” Civil Engineering 1 no 6, March
1931).
To address this differential movement, a pressure-
relieving joint or “cowing” was developed by the
1930s. Cowing composed of corrugated lead was typ-
ically laid horizontally at mid-span in selected levels
of the skyscraper facade. This form of pressure relief
was utilized in masonry envelopes until the 1960s.
Besides serving the function of transmitting wind
loads to the structural frame, the curtain wall also
must resist moisture infiltration. To enter the interior,
the moisture must pass through the mass of the wall;
however it may take a relatively long period of time
this infiltration to become apparent. The mass of the
wall may become saturated with slowly developing
detrimental results.
5 THE METAL AND GLASS CURTAIN WALL
AND ITS DETERIORATION MECHANISMS
New technologies resulting from World War II had a
great influence on the acceptance of the machine-made
metal and glass curtain wall. Given the abundant post-
war supply, aluminum was reasonably priced. There
was experimentation with mild steel, stainless steel,
and bronze as well. Extruded components were suit-
able for standardization and could be prefabricated for
delivery to the site. This was important because labor
had become a significant part of construction costs.
The new curtain wall technology further decreased
building weight and construction cost, and increased
usable floor area. Prefabricated construction was less
limited by cold temperatures which prohibited erec-
tion of “wet” walls of brick and mortar. The invention

Figure 4. The Lever House in New York where the prefabri-
cated metal and glass curtain wall realized its potential (SOM,
1952).
and development of float glass by Alistair Pilkington
in the 1950s would make large panes of glass afford-
ably available. American architectural philosophers of
the day lauded the fact that craftsmanship had been
transplanted from the site to the factory.
One of the first post-war buildings to be constructed
with a glass curtain wall was the Equitable Building
(Pietro Belluschi, 1948) in Portland, Oregon. Belluschi
was able to take advantage of leftover aluminum stock-
piled for World War II by smelters and to utilize
assembly techniques derived from West Coast airplane
plants. The 860–880 Lake Shore Drive buildings in
Chicago (Figure 1) were among the first residential
buildings in the United States to be sheathed entirely
in glass, and were the realization of Mies’ 1920 pro-
posal for a glass skyscraper. The steel, aluminum, and
glass skin was assembled on the buildings’ roofs in
two story high units, and then lowered into place on
the facade.
At the Lever House (Figure 4), the curtain wall has
an interior frame of mild steel clad with stainless steel.
At the United Nations Secretariat Building (Harrison
and Abramovitz, 1950), curtain walls were conceived
as an assembly of aluminum windows held in place
with a grid of reinforced mullions. At both buildings,
the lower portion of the curtain wall at each level was
backed up by a concrete masonry wall to provide the
29
fire rating that code officials felt was not provided
by the curtain wall. The masonry wall could not be
immediately abandoned though it became hidden from
view.
The approach to curtain wall design that quickly
evolved was to make the joints as weathertight as pos-
sible, then provide positive means for conducting any
water leakage out of the wall. Thus an interior drainage
system was provided to collect water that leaks through
the cladding and direct it back to the exterior.
The development of new curtain wall materials
occurred in the post-war years: thin stone veneers,
precast concrete, brick veneers, and structural silicone
glazed facades. To bring things full circle, the aesthetic
development of “Post-Modernism” led to a return to
earlier architectonic forms, but not a return to earlier
methods of construction.
Metal and glass curtain walls that we care for today
consist of factory-fabricated and preassembled metal
units that are connected to the structural frame. Glass
has gone through radical technical developments and
is typically no longer a monolithic material. In a cur-
tain wall it can appear as an insulated glass unit or
a sandwich of materials developed to strengthen it. It
may have clear, colored, or reflective coatings installed
on one or more of its surfaces or may have transpar-
ent or translucent colorants integral with the glass.
The assemblies may also include panels of aluminum,
ceramics, precast concrete, or stone.
Though not the vision of 1950s designers, metal and
glass curtain walls are wholly reliant on sealants to per-
form adequately. High performance sealants include
newly developed families of elastomeric sealants such
as polysulfides, solvent acrylics, urethanes, and sili-
cones. Numerous sealant products are now available
and have been developed to be used in either curtain
wall glazing, installation, or repair.
Modern structural frames are more flexible because
they are designed to tighter limits with less mate-
rial and are more exposed to temperature extremes
than the frames of a masonry-clad building. Prefab-
ricated curtain wall units are detailed to accommodate
these increased movements. However, lateral move-
ments of the frame and differential movement between
the frame and the cladding can lead to distress in the
glass and metal curtain wall.
6 LOSS OF REDUNDANCY AND INCREASE
IN COMPLEXITY
In a general sense, redundancy in design can prevent
failures of a building system. Structural redundancy,
allows for loads to follow an alternate path if the
primary supports fail. Redundancy of design for cur-
tain walls can apply not only to structural but to
performance issues such as weathertightness.
 The mass of the masonry curtain wall provided a
redundant system with empirically developed strength
capacities that extend far beyond that required. While
lack of maintenance could lead to water infiltration
problems or deterioration of the cladding system, these
problems do not become immediately apparent but
become manifest only over a period of time.
In the metal and glass curtain wall strength capac-
ities are engineered to meet specific standards and
redundancy is reduced. An engineered system of per-
formance redundancy has been introduced that pro-
vides a weathertight barrier while providing internal
drainage as well. However, a breakdown of these pro-
tective systems in the metal-and-glass curtain wall
would lead to immediate water leakage on the interior.
High-tech materials have been adopted for use on
these curtain walls that offer unique maintenance chal-
lenges. Tempered glass may fail due to nickel-sulfide
inclusions. Thin-stone veneers may become distressed
from material weakness, loss of strength hysteresis, or
deterioration due to anchorage. Incompatible materials
such as certain sealants and building stone or dissimi-
lar metals can cause staining which can pose long-term
maintenance problems. Little is known regarding the
long term performance of sealants that are now being
used to hold glass panels in place thus eliminating the
metal grid.
7 CONCLUSIONS
Due to the sophisticated engineered systems that have
been introduced, and the decreased redundancy, metal
and glass curtain wall systems pose unique and com-
plex problems. Diagnosis of this unique typology has
been made much easier by following the guidelines
provided in the ICOMOS Charter: Principles for the
Analysis, Conservation and Structural Restoration of
Architectural Heritage (ISCARSAH Principles) that
was ratified by the ICOMOS 14th General Assembly
in Zimbabwe in 2003. The ISCARSAH Principles pro-
vide a comprehensive and well considered philosophy
for building diagnostics which proves just as valu-
able (if not more so) for engineered systems as they
would for empirically designed medieval structures
composed of masonry and wood.
However there is a wide chasm that presently exists
between professionals in the preservation community
and environmental professionals. The obvious debate
is whether it is a sustainable strategy to maintain
an aesthetic and keep original material or whether
it is more important to achieve better water shed-
ding and thermal performance by compromising the
original material. Windows are sources of heat loss
and gain and curtain walls even more so. This was
not a great concern to designers of large structures
enveloped in glass prior to the Energy Crisis of 1973.
However energy use is of primary importance today
30
and regulated by the US Government. Consequently
there are numerous instances where curtain walls are
discarded and replaced rather than conserved.
Another issue is obsolescence. Curtain walls are the
embodiment of pure hard economics on the construc-
tion industry. Early metal and glass curtain walls, being
avant garde, were not designed with a discernible lifes-
pan.The aluminum was cheap and easy to fabricate and
not finished in a way that would preserve its luster.
Some glass products were manufactured using tech-
nically sophisticated and obsolete processes, and it
is no longer feasible to authentically recreate them.
Original sealants or sealants used for repair are not
always compatible with sealants that are presently
available.
A third issue is defining significance of the metal
and glass curtain wall. Is significance defined by the
material itself or by the transparency that it provides?
Perhaps a key to all of the issues cited can be found
in the Nara Document on Authenticity, “authenticity
judgments may be linked to the worth of a great vari-
ety of sources of information. Aspects of the sources
may include form and design, materials and substance,
use and function, traditions and techniques, location
and setting, and spirit and feeling, and other external
factors.” This statement infers that the historic fabric
does not necessarily need to be conserved to maintain
integrity.
Perhaps our approaches to energy inefficient build-
ing claddings will need to be reassessed as we enter
into a new energy paradigm. Be that as it may, as
our post-World War II buildings become historic land-
marks, the challenges that we face will be complex
as well and will require greater degrees of scientific
inquiry and the active participation of the engineer.
REFERENCES
Kelley, S., “The Glass and Metal Curtain Wall: History, Diag-
nostics, and Treatment,” Preserving Post-War Heritage,
Edited by Susan MacDonald, London: Donhead, 2001.
Kelley, S., “Conflicts and Challenges in Preserving Curtain
Walls,” APT Bulletin, Volume 32, Number 1, 2001.
Kelley, S., “Office Buildings of the Chicago School: The
Restoration of the Reliance Building,” Konservierung
Der Moderne, ICOMOS, Hefte des Deutschen National
Komitees XXIV, Munich 1998.
Kelley, S., and Dennis Johnson, “The Glass and Metal Cur-
tainwall: The History and Diagnostics,” Modern Move-
ment Heritage, London: E & FN Spon, 1998.
Kelley, S., and B. Kaskel. “Curtain Walls in the USA: Fail-
ures, Investigation, and Repair,” Curtain Wall Refurbish-
ment, A Challenge to Manage, Eindhoven, Netherlands:
DOCOMOMO International, 1997.
Kelley, S., “The History of the Glass and Metal Curtain Wall:
From the Reliance Building to the Lever House,” Window
Rehabilitation Guide for Historic Buildings, Washington
DC: Historic Preservation Education Foundation, 1997.
Kelley, S., “Aluminum: History and Conservation.”Twentieth
Century Building Materials.Thomas C. Jester, editor. New
York, New York: McGraw-Hill, 1995.
Kelley, S., “The History of the Curtain Wall: From Crafts-
manship to Machine Made.” Preserving the Recent Past.
D. Slaton and R. Shiffer, editors. Washington, D.C.:
Historic Preservation Education Foundation, 1995.
Hunderman, H., J. Koerber, and S. Kelley. “Curtain Wall
Development: The Loss of Redundancy.” D. Slaton and
31
R. Shiffer, editors. Preserving the Recent Past. Washing-
ton, D.C.: Historic Preservation Education Foundation,
1995.
Kelley, S., C. Paulson, and D. Slaton. “Assessment Tech-
niques Utilized with Historic American Highrises,” in
Structural Repair and Maintenance of Historical Build-
ings. Southampton, England: Computational Mechanics
Publications, 1989.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Strengthening of masonry structures with Fibre Reinforced Plastics: From

modern conception to historical building preservation

M.R. Valluzzi
DAUR, University of Padua, Italy

ABSTRACT: Modern techniques and innovative materials are often rapidly proposed and allowed in the current
practice, even for restoration of historical constructions, where fundamental preservation criteria have to be taken
into account.The large variability and complexity of masonry structures and typologies make particularly difficult
the preliminary choices for proper structural models and interventions, that should be based of suitable knowledge
of both existing and new materials, and of their interaction under environmental and loading conditions. Despite
the increasing number of specific studies of FRP reinforcement on masonry structures, still limited codes and
recommendations are available so far. Harmonization of test procedures and methods should be pursued, in order
to compare results and calibrate analytical and numerical models for design and assessment rules.

1 INTRODUCTION Despite the large accessibility to various prod-

In the last two decades, among modern and innovative
solutions of intervention on existing structures, com-
posite materials, as FRP (Fiber-Reinforced Plastics),
have been increasingly considered for strengthen-
ing and repair of both modern and historic masonry
constructions (buildings, bridges, towers) and struc-
tural components (walls, arches and vaults, piers and
columns).
FRP material systems are composed of fibers
embedded in a polymeric matrix, which bind and
protect the fibers themselves, in order to allow load
transferring and to activate their mechanical prop-
erties. High tensile strength and stiffness-to-weight
ratio, fatigue and corrosion resistance, easy in-situ
feasibility and adaptability, and progressive reduc-
tion in production and distribution costs, are the main
characteristics that encouraged the diffusion of these
materials at different levels: to improve the global
behaviour in seismic zone (tying, connections among
components, strengthening), to counteract specific
incipient or developed damage (high compression,
shear and/or flexural conditions), and to repair very
specific local weaknesses depending on the peculiar
construction typology. No yielding is exhibited prior
fibres failure, and sensitivities to impact, notching
and environmental agents could be present. Fibres
activate their characteristics along their prevalent dis-
tribution, whereas have negligible properties in the
other directions.
ucts (bars, strips, laminates, sheets, cords, grids),
made of several reinforcing materials (carbon, glass,
aramid, . . .) and applicable in different modalities on
masonry structures (embedded inside grooves or bed
mortar joints, externally bonded or anchored), design
rules for the interventions, feasibility recommenda-
tions and procedures aimed at checking the effective-
ness of the technique and monitoring are still under
definition.
Starting from extension of approaches proposed
for concrete structures, and after the critical evalu-
ation of the impact of generalized interventions on
historical structures (Giuffrè 1993, Tomazevic 1999,
Binda et al. 2006), researches and studies have been
more and more focused on the specific features of the
masonry material, by recognizing its lack of homo-
geneity and the large variability of typologies and
constituent basic materials and aggregations (brick,
stones, mortar, mixed arrangements, . . .).
This approach has led to a new impulse for the
upgrading of standards devoted to masonry in seis-
mic zone, which often include reference to specific
codes or recommendations on the possible applica-
tion of FRPs, unfortunately still very limited (e.g. the
OPCM 3431/2005 and CNR-DT 200/2004, in Italy;
ACI 440M/2004 in US).
The adoption of FRPs for strengthening is mainly
aimed at reinforcing masonry structures and compo-
nents by increasing their ultimate capacity (strength
and displacement), and often this is achieved by

33
modification of mechanisms at collapse, which can
involve further resisting phenomena.
What essentially emerges from the analysis of a
number of works available in literature aimed at inves-
tigating the mechanical performances of strength-
ened structures and components, and from the code
proposals, are still needs of:
Figure 1. Textile sheets and laminates (a), and various
– definition and putting into practice specific cau- features of FRP bars (c).
tious criteria for possible application of composite
materials in the historical construction preservation
field;
– clarification of critical aspects of application tech-
nologies (e.g., bond and anchorage of textiles and
bars);
– standardization of methods and experimental pro-
cedures for the characterization of the mechanical
performance of strengthened components to define
proper design and assessment criteria;
– definition and validation of investigation proce-
dures for the evaluation of the effectiveness and Figure 2. Wet lay-up system (a), structural repointing
durability of the intervention. (b) and specific anchorage devices for bars (c).

comparison with laminates but, due to the brittleness

2 MATERIALS AND COMPONENTS to folding, specific solution for anchorage should be
adopted (Figure 2). Small diameter bars or strips are
2.1 Strengthening materials used in the bed joints and a special care should be
taken in the repointing phase, in order to incorporate
Among different solutions addressed to masonry struc-
the new reinforcing system in the behaviour of the
tures, FRPs are usually proposed as application of
masonry. The aesthetics of the surface is commonly
near surface mounted bars (cylindrical or prismatic
preserved in this case.
reinforcement inserted in grooves cut on the masonry
surface) or structural repointing (bed joints reinforce-
ment) (Nanni et al. 2003), and externally bonded sheets 2.2 Masonry typologies and damage
or laminates (wet lay-up system) (Figure 1).
Existing masonry constructions could be part of
The use of Carbon and Glass fibres is predominant
modern constructive systems or historical contexts
in comparison with other types, due to their higher
(Figure 3). Their behaviour and proneness to dam-
accessibility on the market and of sufficiently high
age are connected to specific hazardous conditions
mechanical properties (usually compared with ordi-
(e.g., seismicity, subsidence, high lateral or vertical
nary steel), together with epoxy resins for external
loads), which consequently influence the choice of
gluing, and/or modified mortars for embedment in
intervention solution.
joints or grooves.
Therefore, classification of masonry and damage is
More recently, various inorganic products (cement,
fundamental, as it could be even very peculiar and
lime or clay grout) are introduced as support for
diversified among countries, depending on specific
fibres, in order to combine more suitable materials for
constructive and typologies history (Modena 1997,
the intervention on different components of existing
Binda et al. 2006).
structures.
Materials may be very different (solid or multi-
Textiles and sheets present the high advantage of
leaf sections), as well as their combination in struc-
flexibility, allowing disjointed portions or even whole
tural systems (load bearing or infill walls, vaults and
structures to be wrapped; nevertheless, a proper prepa-
columns).
ration of the surface of application is needed (sand-
In this context, FRP reinforcement could have
blasting, laying of primer and putty), and aesthetics
potential in (Figure 4):
reestablishment should taken into account. The bond
at the masonry-FRP interface is the main responsible – counteracting global or partial overturning (façades
for the mechanical performance of the intervention or corners of buildings) and improve collaboration
and of the strengthened component. among components;
On the contrary, application of bars do not require – in-plane or out-of-plane strengthening (walls under
particular preparation works of component surface in shear and bending);

34
Figure 3. Modern residential (a) and masonry historical
(b) buildings in severe conditions.
Figure 4. Typical application of FRP textiles and bars in
masonry: (a) vaults, (b) walls, (c) pillars.
– confining under vertical loads (columns and piers);
– bonding support for curved shapes (arches and
vaults).
3 TEST METHODS AND ANALYSIS
3.1 Experimental procedures
Several contributions from researches are focused
on the experimental behaviour of structural elements
and assemblages. The main aim is to characterize
the mechanical behaviour of strengthened elements,
in order to define simplified model for design and
assessment. Very first contributions about application
on masonry concerned general aspects, especially
35
Figure 5. Loading test on masonry assemblages: vaults
(a) and pillars (b), in-plane (c) and out-of plane (d) testing on
walls.
related to the improvement of the flexural capac-
ity and the behaviour in seismic area (Schwegler
1994, Saadatmanesh 1997, Triantafillou et al. 1997
and 1998, Borri et al. 2002, Corradi et al. 2002,
Brencich et al. 2005, Shrive 2006). Then, more spe-
cialized literature has been proposed, dealing with
specific experimental, analytical and numerical works
on subassemblages: arches and vaults (Briccoli Bati
et al. 2000 and 2001, Foraboschi 2001 and 2004,
Lourenço et al. 2001, Luciano et al. 2001, Valluzzi
et al. 2001, Barbieri et al. 2002, Basilio et al. 2004,
Oliveira et al. 2006, Borri et al. 2007, De Lorenzis
et al. 2005 and 2007, Baratta et al. 2007) (Figure 5.a);
columns and piers (Micelli et al. 2004, Aiello et al.
2005 and 2007, Krevaikas et al. 2005, Nurchi et al.
2005, Corradi et al. 2007) (Figure 5.b); in-plane
(Ehsani et al. 1997, Luciano et al. 1998, Valluzzi et al.
2002, Haroun 2003, Cecchi et al. 2004, De Lorenzis
et al. 2004, Ascione et al. 2005, Hamid et al. 2005,
El-Gawady et al. 2005, El-DAkhakhni et al. 2006) and
out-of-plane (Albert et al. 1998, Gilstrap et al. 1998,
De Lorenzis et al. 2000, Velasquez-Dimas et al. 2000,
Hamilton et al. 2001, Hamoush et al. 2001, Kiss et al.
2002, Kuzik et al. 2003, Galati et al. 2004, Li et al.
2004,Foster et al. 2005, Turco et al. 2006, Mosallam
2007) behaviour of wall panels (Figure 5.c and d); and
on bond (Ehsani et al. 1997, De Lorenzis et al. 2000,
Briccoli Bati et al. 2001, Casareto et al. 2003, Tan et al.
2003, Savoia et al. 2003, Aiello et al. 2005 and 2006,
Liu et al. 2005) (Figure 6).
Harmonization of experimental procedures is a
urgent need that international bodies of standardiza-
tion and related professional corporations should con-
sidered, in order to allow comparison among results.

Figure 6. Experimental tests on local mechanisms: pull-off
tests on surface (a), brick-FRP adhesion test (b), adhesion on
mortar joints (c), splitting of rods (d).
Especially for aspects non easy to clarify as the effec-
tive contribution of the reinforcement on shear and
combined actions, as well as the influence of delamina-
tion in reducing the efficiency of the bond at the inter-
face, homogeneous methods should also be set to allow
proper calibration of analytical and numerical models.
3.2 Standards and recommendation proposals
Despite the several documents available as codes or
guidelines for application of FRP in concrete struc-
tures, at present only the CNR-DT 200/2004 and the
drafts of ACI 440 are available.
In both cases, the formulations proposed for rein-
forced masonry, derived from of the theoretical
approach of reinforced concrete, is mainly adopted.
Bond and shear are definitely the main problematic
aspect to clarify, as standard experimental methods are
not provided, and analytical models are often not in
agreement among them.
As an example, the well-known truss analogy is
considered for the analytical formulation of shear
behaviour, particularly suitable for infill walls, where
the corner crushing limit is preferred to diagonal
cracking or sliding shear. The contribution of FRP
(Vm,f or VRd,f ) is added to the strength of plain masonry
(Vm or VRd,m ) as follows. For ACI 440:
36
where M is the maximum moment at the section under
consideration, V is the corresponding shear force,
d the distance from the extreme compression fibre to
the centroid of tension reinforcement, fm is the com-
pressive strength of masonry, An is the compressed
area of masonry, Af is the reinforcement area, P is
the axial load, ffe is the design strength of FRP, and
s in the spacing of horizontal reinforcement in the
vertical direction. For CNR DT/200, the analogy with
the design formulas proposed for reinforced masonry
beam, as in the Eurocode 6, is evident:
where t is the thickness of the wall, fvd is the design
shear strength of the wall by using a Coulomb fric-
tion law, Afw is the area of reinforcement parallel to
the shear action, pf and α are the spacing and the
inclination of the reinforcement, ffd is the strength of
the reinforcement, assumed as minimum between the
design tensile strength and the delamination load, γRd
(equal to 1.2 in this case) is the partial safety factors
for design, whereas partial factors applied on materials
are γM and γf , for masonry and FRP respectively.
Actually, in-plane rotation, and compression com-
bined with other conditions could affect the strength-
ened component subjected to lateral actions, and
proper limitation of the effectiveness of the interven-
tion due to lack of bonding and or environmental
conditions should be taken into account (Figure 7).
In this connection, the two documents still present
several dissimilarity in the definition of reducing fac-
tors of the nominal tensile strength and the rupture
strain of FRP for creep and environmental exposure,
as evidenced in (Garbin et al. 2006), ranging from
0.80 to 0.30 for the Italian standard and from 0.55 to
0.20 for the US one, respectively for Carbon or Glass
FRP. Moreover, proper coefficient (namely km and kv )
should be considered to take into account debonding
influence in flexural and shear capacity, depending on
several factors, as: the strengthening systems and con-
figurations, the materials at the interface, the amount
and distribution of FRP, etc. On the basis of exper-
imental results these coefficient can vary from 0.2
to 0.8, being particularly low under shear conditions
(Garbin et al. 2006).
By considering the results available in literature
obtained by diagonal compression tests, an integration
of these data is possible. Actually, diagonal compres-
sion led to different failure mode (usually splitting),
Figure 7. Debonding of CFRP strips due to peeling (a), FRP
rupture beyond bonding limit (b).

Table 1. Diagonal compression tests.

Reference unit type disposition sides n. tests

Figure 8. Ultimate load of strengthened configurations
compared to reference strength of clay and concrete walls
Valluzzi 2002 clay diagonal 2 4
subjected to diagonal compression tests.
Yu 2004 clay horizontal 2 1
Yu 2004 clay vertical 2 1
Grando 2003 clay horizontal 2 1
Valluzzi 2002 clay net 2 9
Valluzzi 2002 clay diagonal 1 4
Gabor 2006 clay diagonal 1 3
Yu et al. 2004 clay horizontal 1 1
Yu et al. 2004 clay vertical 1 1
Grando 2003 clay horizontal 1 1
Tinazzi 2003 clay vertical 1 1
Valluzzi 2002 clay net 1 7
Tinazzi 2003 clay net 1 1
Yu 2004 concrete horizontal 2 1
Yu 2004 concrete vertical 2 1
Yu 2004 concrete horizontal 1 1
Yu 2004 concrete vertical 1 1
Grando 2003 concrete horizontal 1 3 Figure 9. Increase of load versus normalized maximum
Morbin 2003 concrete horizontal 1 1 strength of reinforcement.

between clay and concrete masonry, and the highest

than modes involving friction (as for the application of
the Coulomb law, adopted in the formulation proposed
by standards), but it is still widely considered on exper-
imental basis, due to its particularly easy execution.
More reliable combined vertical and lateral load set-
up are progressively adopted for shear, but available
results on literature are still very limited. Uniformity
for diagonal compression test should at least concern
the dimensions of the walls, to be not lower than 1 m
for side, in order to avoid local effects in small size
elements.
The samples given in Table 1, available from appli-
cation of mainly CFRP and GFRP sheets by wet lay-up
system, are rather comparable among them, even
deriving from different reinforcement configurations
and type of support.
By taking into account a reference mean value for
the strength of unreinforced panel of about 100 kN
and 140 kN for clay and concrete walls, respectively,
the increase of strength grouped by configurations is
depicted in Figure 8. The best performances, due to
diagonal and symmetrical dispositions, are confirmed.
The improvement in shear strength is comparable
values are mainly related to GFRP.
By analyzing the data in comparison with the nor-
malized maximum strength of the reinforcement ω,
given by the ratio of the maximum FRP tensile strength
(Ef Af εf ) and the reference value of the shear strength
of masonry τm , computed on the net diagonal area,
multiplied by the gross area of the panel Am , results
are as in Figure 9.
Double values of the strength are obtained in a rather
large range of maximum strength of the reinforcement.
By defining an efficiency factor kv , as ratio between
the increase of load measured by the diagonal test
and the maximum normalized FRP strength, the high-
est effectiveness is achieved by minimum amount
of reinforcement by using the diagonal symmetri-
cal pattern (0.65–0.85 on clay bricks, halved in the
case of single side applications); unidirectional pat-
tern (usually in the horizontal direction) shows a good
behaviour, equivalent or even better if compared with
grid patterns, even if symmetrical (Figure 10).
Also for bonding, the reference models are adopted
from the concrete approach; many strength models
have been developed in the last decade, but, again,
the lack of standardization of harmonized procedures

37

Figure 10. Effectiveness factor for shear strength for differ-
ent amount and configuration of FRP reinforcement.
for experimental tests, led to results often difficult to
compare. CNR DT/200 adopts as bond length le and
strength ffdd the following formulations:


where γFk = c1 · f · fmtm (N mm) is the characteristic
value of the fracture energy, Ef is theYoung modulus of
FRP in the direction of the applied force, tf is the FRP
thickness, fmtm = 0.1fmk is the mean tensile strength
of masonry (considered coincident with the strength
of the blocks), γM and γf ,D are partial safety fac-
tors, varying from 1.1–1.25 and 1.2–1.5, respectively,
depending on the certification of the entire bonding
system on the support, or of only the single materials.
c1 is a coefficient to identify on experimental basis or
to adopt equal to 0.015 (0.03 is proposed for concrete
with the same equation).
The non homogeneity of masonry due to the pres-
ence of mortar joints, and the consequent influence in
the bonding phenomenon of the different mechanical
properties and of the geometrical discontinuity, are not
considered in the model; moreover, a unique signifi-
cant value for the fracture energy along the connection
is assumed.
A fundamental contribution to clarify these aspect
is done by several research groups (Briccoli Bati et al.
2001, Aiello et al. 2003 and 2006, Casareto et al.
2003, Basilio et al. 2005, Panizza et al. 2008), but
different test procedures are adopted, thus comparison
of results is often unreliable. In particular, the most
suitable arrangements could be the double-lap (DL)
(Figure 11) and the single-lap shear test (SL), the latter
being the most effective, due to the problem of repro-
ducing actual symmetry of load distribution in the DL
configuration, but not simple to realize, in comparison
with the former one.
The identification of ffdd is crucial, as it represents
the parameter strictly related to the efficiency of the
38
Figure 11. Double-lap shear test executed on CFRP sheets:
experimental set-up (a), fingerprint on brick after test (b),
peeling of brick surface on FRP (c) (Panizza et al. 2008).
Figure 12. Scheme of wall adopted for FE simulation: unre-
inforced wall (a), horizontal FRP strip application in lintels
(b), vertical strips on main vertical walls to reinforce flexural
behavior (c), addition of horizontal strips on main vertical
sects to reinforce shear behavior (d).
connection between reinforcement and masonry. The
comparison with the application of equations (5) and
the elaboration of experimental data able to define effi-
ciency laws for debonding as in (Panizza et al. 2008),
allowed to calibrate a FE model (with DIANA), in
order to simulate the performance of shear and flexural
strengthening with CFRP sheets applied in a masonry
wall including openings. The scheme of the model is
depicted in Figure 12, and the comparison among the
main parameters at the interface between reinforce-
ment and masonry, to be used for the model, are given
in Table 2. Experimental elaboration obtained by DL
shear tests have been considered, taking into account
a reduction of 30% to obtain reasonable characteristic
values, to compare to the ones computed according to
the CNR standard.
The main characteristics of materials were derived
from available experimental tests, or computed on the
basis of the national standards. For clay bricks, a char-
acteristic compressive strength of 41.2 MPa, a mean
tensile strength of 2.4 MPa and an elastic modulus of
16 GPa, were assumed; a mortar M2.5 (MPa, com-
pression) was considered, whereas computed global
Table 2. Mechanical properties of materials.

Specific fracture FRP maximum

energy fk strength (delamination)
Material (N/mm) ffdd (MPa)

CNR DT 200 0.04 335

DL shear tests 0.99 1660

properties of masonry were: 8.5 MPa for the compres-

sive strength, 8500 MPa and 3400 GPa for the elastic Figure 13. Comparison among plain masonry and three
and the shear moduli, respectively, and a Coulomb law strengthening configuration by using experimental elabora-
for shear strength calibrated with parameters equal tions and elastic-brittle law for FRP strips.
to 0.3 as cohesion and 0.4 as friction coefficient.
High-strength CFRP applied in strips 100 mm wide
were considerd, having nominal elastic modulus of
230 GPa and equivalent thickness of 0.165 mm. Loads
derived from a three storey building were assumed,
and the simulation of the effect on the base floor was
performed.
From Table 2, the maximum fracture energy and
the limit of experimental delamination are thus about
25 and 5 times higher than the valued proposed by
the code, respectively. Four configurations were con-
sidered, characterized by progressive reinforcement,
from plain masonry to global horizontal confinement,
up to application of horizontal and vertical strips on
Figure 14. Comparison among plain masonry and three
the main walls, 50 cm spaced (Figure 12). Non-linear strengthening configuration by using experimental elabora-
push-out analysis were performed, by considering pro- tions and elasto-plastic law for FRP strips.
gressive increase of horizontal acceleration, uniformly
distributed along the height of the wall. The comparison with models where fracture
The numerical results are given in Figures 13 to energy parameters are assumed as in the CNR stan-
15. To take into account a possible residual bond after dards revealed minor performances, thus confirm-
delamination occurrence, the comparison between the ing the proposed models being highly conservative
assumption of a elastic-brittle or elastic-plastic law for (Figure 15). This is surely safety oriented, but if not
the FRP behaviour was considered. controlled in its boundaries could induce overesti-
Despite the simplification adopted in the modelling, mation of FRP design, and consequent utilization of
several points can be discussed, by analyzing the dia- unnecessary reinforcement. Moreover, the sensitivity
grams which compare the imposed acceleration to of CNR assumptions to the elastic-plastic or elastic-
the displacement at the control point (Figure 12.a). brittle behaviour of FRP are less evident in comparison
First (Figure 13), by considering the experimental with experimental elaboration of basic parameters.
elaboration on fracture energy and delamination, the Finally, it is worth to remark that experimental
unreinforced model is subjected mainly to rocking, results on bond are available by tests where FRP is
confirming the high vulnerability of masonry build- glued only on brick surface, neglecting the signifi-
ings to rigid movements (a). The shear strength is not cant influence of mortar joints in the complex adhesion
activated in the walls, and the preliminary horizontal phenomenon involving the masonry.
tying is not able to increase the peak load, but only
to homogenize the plateau in the plastic branch (b).
By introducing vertical strips, the load is significantly
4 FEASIBILITY IN HISTORICAL BUILDINGS
increased, with minor differences between flexural (c)
and combined shear (d) reinforcement. The elasto-
4.1 Reference criteria
plastic behaviour of FRP is able to increase of about
25% the peak load in the complete strengthening con- Interventions to perform in historical environment
figuration, and to guarantee a constant plastic plateau cannot disregard to satisfy the specific requirements
(Figure 14). on which preservation is based.

39
Figure 15. Comparison among plain masonry and three
strengthening configuration by using CNR DT-200 for-
mulation and elastic-brittle or elasto-plastic law for FRP
strips.
It is worth to remark that, even in very hazardous
conditions, as mainly occur in seismic zone, a proper
compromise has to be taken among safety and preser-
vation, keeping priority the safeguard of human life.
This enable to reduce and control upgrading, in favour
of alternative measures of improvement, specifically
targeted to the large complexity of historic masonry
and constructions (ICOMOS/ISCARSAH). Recent
updating national codes (e.g., OPCM 3431/2005 seis-
mic code in Italy) finally provide specialized sections
for existing masonry, recognizing the differences in
typologies and materials, and by pointing out the great
importance of preliminary knowledge, supported by
the suitable application of in-situ non-destructive (ND)
and minor-destructive (MD) tests.
Nevertheless, the Charter of Venice (1964), the
Declaration of Amsterdam (1995) and the Charter of
Cracovia (2000), could be considered reference docu-
ments for the definition of criteria and actions devoted
to Cultural Heritage. Minimum interventions, having
characters of “reversibility” (intended as substitutabil-
ity or removability); use of materials and techniques
compatible with the original ones, able to guarantee
durability to the intervention itself and, consequently,
to the building; respect to the original functions (both
structural and of utilization), distinguishability of the
intervention; all these criteria should be considered
when proposing repair or strengthening of historic
structures.
FRP has a great potential to improve the brittle
behaviour of masonry components, but many aspects
related to its interaction with traditional materials and
durability are still under experimental study, especially
due to the use of resins as bonding system.
More recent studies are focused on the use of
organic matrixes, as FRCM (Fiber Reinforced Cemen-
ticious Matrix) or TRM (Textiles Reinforced Mortar)
(De Lorenzis et al. 2004, Prota et al. 2006, Papanicolau
et al. 2007), and could represent significant steps for
40
Figure 16. Repair intervention on masonry vault of Villa
Bruni (Padova): position and view of depressed structures
propped before intervention (a), design and assessment with
contribution of FRP sheets (b), scheme of global intervention
and detail of application of CFRP strips at extrados in severely
cracked zones (c).
the identification of compatible binders with original
masonry.
Cautious approach should dominate, as well as the
basic principle that intervention now has not to pretend
to be definitive, as further more appropriate measures
can be more reliable in the future.
The CNR DT-200 itself declare that interventions
with FRP on monuments and historical architecture
have to be justified as indispensable for the building,
and the respect of the above-mentioned principles of
restoration has to be guaranteed.
4.2 Some cases study and applications
The current use of FRPs in restoration work, should
be particularly cautious when dealing with historical
constructions, as many restoration principles cannot be
fully satisfied. On the contrary, their high mechanical
performances can be usefully exploited, where other
systems fails or are more invasive. Especially tex-
tiles and bars have the main advantage of not increase
dimensions of strengthened sections, thus their use,
if properly designed to solve specific problems, can
be targeted to large elements, but also to solve very
peculiar weaknesses, not only in standard bearing
masonry constructions. Some examples are given in
the following.
Even if in large assemblages, FRP can be very ver-
satile to repair urgent local cracks, as part of combined
Figure 17. Repair intervention on masonry vaults of S.ta
Corona Church in Vicenza: view of church and detail of
cracked cross vault (a), scheme of intervention at extrados
and view of FRP application combined with injections (b).

intervention. As an example, the consolidation of the

very depressed thin vault of the central hall of Villa
Bruni in Megliadino S. Vitale (Padova), has included
the repair with FRP strips of the main cracks at the Figure 18. Intervention on church of S.ta Maria in Organo
intrados and the extrados, together with other tying (Verona): view of church and position of longitudinal cracks
measures acted by the timber beams of the floor. FRP on main vault (a), application of FRPs on ribs and side
walls (b), scheme of intervention and final view after
strips were designed and assessed by using a specific
protection of strips (c).
method applied to vaults, taking into account the mod-
ified failure mechanisms of the strengthened structure
(Valluzzi et al. 2002 and 2004).
Other applications on vaults can be combined with
injections, as well as ties, to improve the global
behaviour, as in the church of S.ta Corona of Vicenza,
where cross vaults have been reinforced at their
extrados with strips 20 cm wide in their longitudinal
direction (Figure 17).
In the church of S.ta Maria in Organo of Verona,
the intervention with FRP has been limited to the in-
plane reinforcement of the walls stabilizing the vaults
from the sides and to existent transversal ribs, as well
as new ones, provided in the longitudinal direction
(Figure 18). Figure 19. Intervention with CFRP in capital of ‘Palazzo
Finally, specific applications where the high ver- della Ragione’ of Padova: original confining rings (left) and
satility of FRP laminates can hardly even the per- FRP strengthening (right).
formances of other materials, are the local repair of
weaknesses in stone elements, being part of structures a equestrian statue characterizing the marble monu-
or monuments. As an example, the confining of capi- mental tomb of Cansignorio in Verona, particularly
tals can be performed, to substitute or integrate actual deteriorated by metal oxidation and environmental
metal rings, as executed at the top of the columns of aggression. A system of superimposed layers of CFRP
the ‘Palazzo della Ragione’ in Padova. CFRP strips has allowed to rebuild the loose parts of the bearing
have been positioned after restoring of the capitals elements, re-establishing the aesthetics of the statue
and successively hidden beneath the historical original (Gaudini et al. 2008).
metal confining elements (Figure 19). Another pecu- Therefore, in monuments, where the aesthet-
liar example is the reinforcement of the basement of ics could be the most important requirement, the

41

Figure 20. Intervention on ‘Arca of Cansignorio’ (Verona):
view of deteriorated hoof and first phase of intervention (a),
protecting lime putty and final view of the statue after
intervention (b).
intervention can be focused on a precise defect to
repair, providing a proper covering of the surface (with
marble chipping or similar) to hide the structural inte-
gration. Other materials, like steel, despite similar
advantages from a structural point of view, cannot have
the same versatility and possibility in applications,
preserving aesthetics at the same time.
5 CONCLUSIONS
FRP materials have high potential for the improve-
ment of mechanical performance of structures, also
for masonry components, provided that preserva-
tion criteria are satisfied in the field of historical
constructions.
Practical applications on historical masonry struc-
tures are mainly focused on the use of sheets, aimed at:
– confining structural components (columns, towers)
or repairing specific damage of materials (statues,
capitals, etc..) which can be protected by oxidation
or other attacks;
– stabilization (as contribution, often together with
other interventions) of curved structures (arches and
vaults in buildings and bridges) and local repair of
cracks.
Moreover, FRP can allow very light interventions,
able to stabilize local critical conditions and/or to
integrate original materials.
The impulse in their rather common use in the last
decade, contributed to lower the costs for production
and distribution, thus their availability on the market is
42
more and more increasing. Nevertheless, many aspects
still are under investigation, and specific experimental
procedures and models need to be homogenized in
standards and recommendations.
Therefore, from more and more modern concep-
tions, today more than in the past, we need to resume
traditional values, in order to not forget our learning
from history and to respect constructive specificity
and functions of the original structures. Innovative
solutions can be very useful even in the Cultural
Heritage context, provided that we are able to recog-
nize their limits, and to pursue the clarification of all
aspects (both positive and negative) that are involved
in the delicate question of the preservation of historical
constructions.
ACKNOWLEDGEMENTS
The author wish to acknowledge C. Bettio, E. Garbin,
G. Guidi and M. Panizza for their contribution in the
researches and applications and for data processing
and modelling.
REFERENCES
ACI 440M Guide Draft-1. 2004. Guide for the Design
and Construction of Externally Bonded FRP System
for Strengthening Unreinforced Masonry Structures. ACI
Committee 440.
Aiello, M.A., Sciolti, M.S., Pecce M. 2003. Experimental
investigation on bond between FRP sheets and natural
masonry blocks. 10th Int. Conf on Structural Faults and
Repair SF&R-2003, London (UK).
Aiello, M.A., Micelli, F., Valente, L. 2005. Circular masonry
columns confined with FRP. Composites in Construction
CCC2005, proc. of the 3rd Int. Conf., Lyon, France.
Aiello, M.A., Micelli, F., Valente, L. 2007. Structural Upgrad-
ing of Masonry Columns by Using Composite Reinforce-
ments. ASCE J. of Composites for Construction, 11(6):
650–658.
Aiello, M.A., Sciolti, M.S. 2005. Influence of moistness and
high temperature on bond between FRP reinforcement
and calcarenite ashlars. Composites in Construction –
CCC2005, proc. of the 3rd Int. Conf., Lyon, France.
Aiello, M.A., Sciolti, S.M. 2006. Bond analysis of masonry
structures strengthened with CFRP sheets. Construction
and Building Materials, 20 (1): 90–100.
Albert, M.L., Cheng, J.J.R., Elwi,A.E. 1998. Rehabilitation of
unreinforced masonry walls with externally applied fiber-
reinforced polymers. Struct. Engrg. Rep. No. 226, Dept.
of Civ. and Envir. Engrg., University of Alberta, Canada.
Ascione, L., Feo, L., Fraternali, F. 2005. Load carrying
capacity of 2D FRP/strengthened masonry structures.
Composites Part B, 36 (8): 619–626.
Baratta, A., Corbi, O. 2007. Stress analysis of masonry vaults
and static efficacy of FRP repairs. International Journal
of Solids and Structures, 44 (24): 8028–8056.
Barbieri, A., Borri, A., Corradi, M., Di Tommaso, A. 2002.
Dynamic Behaviour of Masonry Vaults Repaired with
 FRP: Experimental Analysis. British Masonry Society,
proc. of the 6th Int. Conf.
Basilio, I., Oliveira, D., Lourenço, P. 2004. Optimal FRP
strengthening of masonry arches. Brick and Block
Masonry Conference, proc. of the 13th Int. Conf., Ams-
terdam, Netherlands.
Basilio I., Oliveira D., Lourenço P. 2005. Experimental Char-
acterization of FRP Masonry interface Behaviour. Proc.
of the 5th Int. Conf. on Analytical Models and New Con-
cepts in Concrete and Masonry Structures AMCM-2005,
Gliwice-Ustron (Poland).
Binda L., Cardani G., Saisi A., Valluzzi M.R. 2006. Vulnera-
bility analysis of the historical buildings in seismic area by
a multilevel approach. Asian Journal of Civil Engineering,
7(4): 343–357.
Borri, A., Casadei, P., Castori, G., Ebaugh, S. 2007. Research
on composite strengthening of masonry arches. Fiber
Reinforced Polymer Reinforcement for Concrete Struc-
tures – FRPRCS-8, proc. of the 8th Int. Symp., Patras,
Greece.
Borri, A., Corradi, M., Vignoli, A. 2002. New materials for
strengthening and seismic upgrading interventions. Ari-
adne 10, Arcchip, proc. of the Int. Workshop, Prague,
Czech Republic.
Brencich, A., Gambarotta, L. 2005. Mechanical response of
solid clay brickwork under eccentric loading. Part I and
II. Materials and Structures, 38.
Briccoli Bati, S., Rotunno, T. 2001. Environmental durabil-
ity of the bond between the CFRP composite materials
and masonry structures. Structural Analysis of historical
constructions – SAHC 2001, proc. of the 3rd Int. Conf.,
Guimaraes, Portugal, 1039–1046.
Briccoli Bati, S., Rovero, L. 2000. Consolidation of masonry
arches with carbon-fiber reinforced plastics. 12th Int.
BRICK/BLOCK Conference, proc., Madrid, Spain.
Briccoli Bati, S., Rovero, L. 2001. Experimental validation
of a proposed numerical model for the FRP consolida-
tion of masonry arches. Structural Analysis of historical
construction – SAHC 2001, proc. of the 3rd Int. Conf.,
Guimaraes, Portugal, 1057–1066.
Casareto M., Oliveri A., Romelli A., Lagomarsino S. 2003.
Bond behavior of FRP laminates adhered to masonry.
Advancing with Composites – Plast-2003, proc. of the Int.
Conf., Milan, Italy.
Cecchi, A., Milani, G., Tralli, A. 2004. In-plane loaded
CFRP reinforced masonry walls: mechanical character-
istics by homogenisation procedures. Composites Science
and Technology, 64: 2097–2112.
Charter of Krakow (2000). Principles for conservation and
restoration of built heritage. ICOMOS, Paris.
Charter of Venice (1964). International charter for the
conservation and restoration of monuments and sites.
Decision and resolutions. ICOMOS, Paris
Chen, J.F., Teng, J.G. 2001. Anchorage Strength Models for
FRP and Steel Plates Bonded to Concrete. ASCE Journal
of Structural Engineering, 127 (7): 784–791.
CNR-DT 200/2004. Guide for the Design and Construc-
tion of Externally Bonded FRP Systems for Strengthening
Existing Structures. National Research Council, Italy.
Corradi, M., Borri, A., Vignoli, A. 2002. Strengthening
techniques tested on masonry structures struck by the
Umbria-Marche earthquake of 1997–1998. Construction
and Building Materials, 16 (4): 229–239.
43
Corradi, M., Grazini, A., Borri, A. 2007. Confinement of
brick masonry columns with CFRP materials. Composites
Science and Technology, 67 (9): 1772–1783.
Dai, J.G., Ueda, T., Sato,Y. 2005. Development of the Nonlin-
ear Bond Stress-Slip Model of Fiber Reinforced Plastics
Sheet-Concrete Interfaces with a Simple Method. ASCE J.
of Composites for Construction, 9 (1): 52–62.
De Lorenzis, L., D., Tinazzi, A., Nanni, A., 2000. Near
Surface Mounted FRP Rods for Masonry Strengthening:
Bond and Flexural Testing. Meccanica delle Strutture in
Muratura Rinforzate con FRP Materials, proc. of the 2nd
Nat. Symp., Venice, Italy.
De Lorenzis, L., Dimitri, R., La Tegola, A. 2005. Strength-
ening of masonry edge vaults with FRP composites.
Composites in Construction – CCC2005, proc. of the 3rd
Int. Conf., Lyon, France.
De Lorenzis, L., Dimitri, R., La Tegola, A. 2007. Reduction
of the lateral thrust of masonry arches and vaults with FRP
composites. Construction and Building Materials, 21 (7):
1415–1430.
De Lorenzis, L., Galati, N., Ombres, L. 2004. In-plane shear
strengthening of natural masonry walls with NSM CFRP
strips and FRCM overlay. Structural Analysis of historical
constructions SAHC 2001, proc. of 4th Int. Sem., Padova,
Italy, 843–855.
De Lorenzis, L., Miller, B., Nanni, A. 2001. Bond of FRP
laminates to concrete. ACI Materials Journal, 98 (3):
256–264.
Ehsani M.R., Saadatmanesh H., Al-Saidy, A. 1997. Shear
behavior of URM retrofitted with FRP overlays. ASCE J.
of Composites for Construction, 1 (1): 17–25
El-Dakhakhni, W.W., Hamid, A.A., Hakam, Z.H.R.,
Elgaaly, M. 2006. Hazard mitigation and strengthening of
unreinforced masonry walls using composites. Composite
Structures, 73 (4): 458–477.
El-Gawady, M.A., Lestuzzi, P., Badoux, M. 2005. Aseis-
mic retrofitting of unreinforced masonry walls using FRP.
Composites Part B, 37 (2): 148–162.
EN 1996. Eurocode 6. Design of masonry structures.
Foraboschi, P. 2001. Strength Assessment of Masonry Arch
Retrofitted Using Composite Reinforcements. Masonry
International, 15 (1): 17–25.
Foraboschi, P. 2004. Strengthening of Masonry Arches with
Fiber-Reinforced Polymer Strips. ASCE J. of Composites
for Construction, 8 (3): 191–202.
Foster, P., Gergely, J., Young, D., McGinley, M. 2005.
Strengthening masonry buildings with FRP composites.
Structural Faults & Repair SF&R2006, proc. of Int. Conf.,
Edinburgh, UK.
Gabor, A., Ferrier, E., Jacquelin, E., Hamelin, P. 2006. Anal-
ysis and modelling of the in-plane shear behaviour of
hollow brick masonry panels, Construction and Building
Materials, 20: 308–321.
Galati, N., Micelli, F., Tumialan, J.G., La Tegola, A., Nanni, A.
2004. Comparison between FRP strengthening techniques
on the Out-of-Plane behaviour of URM masonry walls.
Innovative Materials and Technologies for Construction
and Restoration – IMTCR04, proc. of the 1st Int. Conf.,
Lecce, Italy, 440–457.
Garbin, E., Valluzzi, M.R., Modena, C., Galati, N., Nanni, A.
2006. In-plane design for masonry walls strengthened by
FRP materials, Structural Faults & Repair – 2006, June
13–15 2006, Edinburgh, UK (12 pp. on CD-ROM).
Gaudini, G., Modena, C., Casarin, F., Bettio, C., Lucchin F.
(2008), Monitoring and strengthening interventions on
the stone tomb of Cansignorio della Scala, Verona, Italy,
6th Int. Seminar on Structural Analysis of Historical
Constructions, 2–4 July 2008, Bath, UK.
Gilstrap, J.M., Dolan, C.W. 1998. Out-of-plane bending of
FRP-reinforced masonry walls. Composites Science and
Technology, 58 (8): 1277–1284.
Giuffrè A. 1993. Sicurezza e conservazione dei centri storici:
il caso di Ortigia, Laterza: Bari, Italy.
Grando, S., Valluzzi, M.R., Tumialan, J.G., Nanni A.
2003. Shear Strengthening of UMR Clay Walls whit
FRP Systems, 6th International Symposium on Fiber-
Reinforced Polymer (FRP) Reinforced for Concrete Struc-
tures (FRPRCS-6), Singapore, 8–10 July 2003, 1229–
1238.
Hamid, A.A., El-Dakhakhni, W.W., Hakam, Z.H.R.,
Elgaaly, M. 2005. Behavior of Composite Unreinforced
Masonry – Fiber-Reinforced Polymer Wall Assemblages
Under In-Plane Loading. ASCE J. of Composites for
Construction, 9 (1): 73–83.
Hamilton III, H.R., Dolan, C.W. 2001. Flexural Capacity of
Glass FRP Strengthened Concrete Masonry Walls. ASCE
J. of Composites for Construction, 5 (3): 170–178.
Hamoush, S., McGinley, M., Mlakar, P., Scott, D., Murray, K.
2001. Out-of-plane Strengthening of Masonry Walls with
Reinforced Composites. ASCE J. of Composites for Con-
struction, 5 (3): 139–145.
Haroun, M., Mosallam, A.S., Allam, K.H. 2003. Cyclic In-
Plane Shear of Concrete Masonry Walls Strengthened by
FRP Laminates. Journal of Engineering Mechanics, 128
(5): 562–573.
Kiss, R.M., Kollàr, L.P., Jai, J., Krawinkler, H. 2002. Masonry
Strengthened with FRP Subjected to Combined Bend-
ing and Compression, Part II: Test Results and Model
Predictions. Journal of Composite Materials, 36 (9):
1049–1063.
Krevaikas, T.D., Triantafillou, T.C. 2005. Masonry con-
finement with Fiber-Reinforced polymers. ASCE J. of
Composites for Construction, 3–4: 128–35.
Kuzik, M.D., Elwi, A.E., Roger Cheng, J.J. 2003. Cyclic Flex-
ure Tests of Masonry Walls Reinforced with Glass Fiber
Reinforced Polymer Sheets. ASCE J. of Composites for
Construction, 7 (1), 20–30.
Li, T., Galati, N., Nanni, A. 2004. Research on FRP strength-
ening of URM walls with opening. Mechanics of Masonry
Structures Strengthened with FRP-materials: Modeling,
Testing, Design, Control, proc. of the Int. Symp., Venice,
Italy.
Liu, Y., Dawe, J., McInerney, J. 2005. Behaviour of GFRP
sheets bonded to masonry walls. Bond Behaviour of FRP
in Structures – BBFS 2005, proc. of the Int. Symp., Hong
Kong, China, 473–480.
Lourenço, P., Poças Martins, J.P. 2001. Strengthening of the
architectural heritage with composite materials. Compos-
ites in Construction – CCC2001, proc. of the 1st Int. Conf.,
Porto, Portugal.
Luciano, R., Marfia, S., Sacco, E. 2001. Reinforcement of
masonry arches by FRP materials: experimental tests and
numerical investigations. Composites in Construction –
CCC2001, proc. of the 1st Int. Conf., Porto, Portugal.
Luciano, R., Sacco, E. 1998. Damage of masonry panels rein-
forced by FRP sheets. International Journal of Solids and
Structures, 35 (15): 1723–1741.
44
Micelli, F., De Lorenzis, L., La Tegola, A. 2004. FRP-
confined masonry columns under axial loads: experimen-
tal results and analytical model. Masonry International,
17 (3): 95–108
Modena, C. 1997. Criteria for cautious repair of his-
toric building. A valuation and strengthening of exist-
ing masonry structures, Binda L. and Modena C., Ed:
RILEM.
Morbin, A. 2003. Strengthening of Masonry Elements
with FRP Composites, Report No. CIES 02–23, Cen-
ter for Infrastructure Engineering Studies, University of
Missouri-Rolla.
Mosallam, A.S. 2007. Out-of-plane flexural behavior of
unreinforced red brick walls strengthened with FRP
composites. Composites: Part B, 38: 559–574.
Nanni, A., Tumialan, G. 2003. Fiber-Reinforced Compos-
ites for the Strengthening of Masonry Structures. Struc-
tural Engineering International, 13(4), 1 November 2003,
pp. 271–278
Nurchi, A., Valdes, M. 2005. Strengthening of stone masonry
columns by means of cement-based composite wrapping.
Composites in Construction – CCC2005, proc. of the 3rd
Int. Conf., Lyon, France.
Oliveira, D.V., Basilìo, I., Lourenço, P.B. 2006. FRP strength-
ening of masonry arches towards an enhanced behaviour.
Bridge Maintenance, Safety and Management – IAB-
MAS’06, proc., Porto, Portugal.
OPCM 3431/2005. Ordinanza 3431/2005 del 3 Maggio
2005: Norme tecniche per il progetto, la valutazione
e l’adeguamento sismico degli edifici. P.C.M., Italy.
Panizza, M., Garbin, E., Valluzzi, M.R., Modena, C. 2008.
Bond Behaviour of CFRP and GFRP Laminates on Brick
Masonry. Structural Analysis of Historical Construction –
SAHC08, proc. of the 6th Int. Seminar, Bath (UK) .
Papanicolaou, C.G., Triantafillou, T.C., Karlos, K., Pap-
athanasiou, M. 2007. Textile-reinforced mortar (TRM)
versus FRP as strengthening material of URM walls:
in-plane cyclic loading. Materials and Structures, 40:
1081–1097.
Prota, A., Marcari, G., Fabbrocino, G., Manfredi, G.,
Aldea, C. 2006. Experimental In-Plane Behavior of Tuff
Masonry Strengthened with Cementitious Matrix-Grid
Composites. ASCE Journal of Com-posites for Construc-
tion, 10 (3): 223–233.
Saadatmanesh, H. 1997. Extending service life of concrete
and masonry structures with fiber composites. Construc-
tion and Building Materials, 11 (5): 327–335.
Savoia, M., Ferracuti, B., Mazzotti, C. 2003. Non linear bond-
slip law for FRP-concrete interface. Fibre-Reinforced
Polymer Reinforcement for Concrete Structures FRPRCS-
6; proc. int. symp., Singapore.
Schwegler, G. 1994. Masonry construction strengthened
with fiber composites in seismically endangered zones.
Earthquake Engineering, proc. of the 10th Europ. Conf.,
Rotterdam, The Netherlands, 454–458.
Shrive, N.G. 2006. The use of fibre reinforced polymers to
improve seismic resistance of masonry. Construction and
Building Materials, 20 (4): 269–277.
Tan, K.H., Patoary, M.K.H., Roger, C.S.K. 2003. Anchor-
age Systems for Masonry Walls Strengthened with FRP
Composite Laminates. J. of Reinforced Plastics and Com-
posites, 22 (15): 1353–1371.
Tomaževič, M. 1999. Earth quake resistano design of
masonry buildings. Imperial College Press: London, UK.
Triantafillou, T.C. 1998. Strengthening of masonry struc-
tures using epoxy-bonded FRP laminates. ASCE J. of
Composites for Construction, 2(2): 96–104.
Triantafillou, T.C., Fardis, M.N. 1997. Strengthening of
historic masonry structures with composite materials.
Materials and Structures, 30: 486–496.
Turco, V., Secondin, S., Morbin, A., Valluzzi, M.R., Mod-
ena, C. 2006. Flexural and shear strengthening of un-
reinforced masonry with FRP bars, Composites Science
and Technology, 66(2):289–296.
Valluzzi, M.R., da Porto, F., Modena, C. 2004. Strengthen-
ing of a masonry vault with GFRP laminates: the case
of the church of S. Fermo in Verona. Innovative Mate-
rials and Technologies for Construction and Restoration
-IMTCR04, Lecce (Italy), 6–9 June 2004, pp. 430–439.
Valluzzi, M.R., Tinazzi, D., Modena, C. 2002. Shear behav-
ior of masonry panels strengthened by FRP laminates.
Construction and Building Materials, Special Issue,
16(7): 409-416.
45
Valluzzi, M.R., Tinazzi, D., Garbin, E., Modena, C. 2003.
FEM modelling of CFRP strips bond behaviour for bed
joints reinforcement techniques. Computer Methods in
Structural Masonry; proc. 6th int. conf., Rome, Italy.
Valluzzi, M.R., Valdemarca, M., Modena, C. 2001. Behavior
of brick masonry vaults strengthened by FRP laminates.
ASCE Journal of Composites for Construction, August
2001, vol. 5, n. 3, pp. 163–169.
Velazquez-Dimas, J.I., Ehsani, M.R., Saadatmanesh, H. 2000.
Out-of-Plane Behavior of Brick Masonry Walls Strength-
ened with Fiber Composites. ACI Structural Journal, 97
(3): 377–387.
Yu, P., Silva, P.F., Nanni, A. 2004. Application of Bond
Polyurea in Structural Strengthening of RC Beams and
UMR Walls, Final Report, Report No. CIES 01-49, Cen-
ter for Infrastructure Engineering Studies, University of
Missouri-Rolla.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

The structural behaviour of masonry vaults: Limit state analysis with

finite friction

D.F. D’Ayala
Department of Architecture and Civil Engineering, University of Bath, UK

E. Tomasoni
University of Brescia, Italy

ABSTRACT: The increasing interest in historic architectural heritage and the need for preservation of historical
structures has led to the continuous development in the past 20 years of a growing number of methods for the
analysis of masonry vaults. This notwithstanding some type of vaults, for instance pavilion vaults and fan vaults,
despite their broad use in past centuries, have not been thoroughly studied, mainly due to the difficulty of applying
simplified theories to their complex shapes. The major simplification that is usually carried out is to reduce the
vault to a series of adjacent arches, without transversal connection. Even though simple and accurate, the arch
model does not take into account the interaction between the arches and it is not able to properly simulate the
three-dimensional effects in the vaults. Although the results are conservative for uniform load distributions, the
model limits substantially the set of loading conditions that can be analysed and hence a thorough assessment
of vaults performance. This limitation and the need for a flexible method to study the different types of vaults
could be solved by using the concept of the surface of thrust within the framework of limit state analysis. The
main object of the present paper is the development of a computational procedure which allows to define the
3D structural behaviour of masonry vaults: using limit state analysis with finite friction, the proposed analytical
method, based on lower bound approach, allows to obtain, for a generic type of vault, the actual crack pattern,
the stress field and the horizontal thrust at the supports for both gravitational and localized loads. In the present
paper the limit state analysis with finite friction has been applied to pavilion vaults. The results obtained are
compared with non linear F.E. analysis simulations.

1 INTRODUCTION lies in its high computational burden when meaningful

Computational approaches to the integrity assess-
ment of structural masonry are presently conducted
by a variety of methodologies, ranging from highly
simplified methods to complex non-linear finite ele-
ment analyses using plasticity based material mod-
els, including joint and interface elements to model
planes of weaknesses. Research into modelling of
structural masonry has been very active mostly rely-
ing on non-linear continuum (Livesley 1978, 1992,
Harvey 1988, Melbourne & Gilbert 1994, Hughes
1997) or homogenisation based techniques (Lourenço
1996). While discontinuities can be incorporated by
use of smeared crack approach or contact elements,
the most reliable solution to date is the use of dis-
crete elements initially developed by Cundall (1971)
for rock mechanics and applied to dry block masonry
by Amadei (1995), Maunder (1993), Lemos (1997).
This approach is particularly useful to analyse cases
in which displacement are significant and concen-
trated at interfaces. The limitation of this technique
structural models are considered.
Limit analysis has the substantial advantage of dis-
regarding intermediate states of stress in order to
identify directly ultimate conditions. It tackles the
problem of main interest to structural engineers: con-
trol of safety levels. The assumption is that at failure
pre-existing self equilibrated states of stress are irrel-
evant so that direct correlation can be established
between external actions and collapse mechanisms.
However these assumptions are only valid for ‘stan-
dard’ materials for which the general theorems of plas-
ticity apply. Due to its lack of ductility and presence
of friction, masonry is not ‘standard’. The application
of classical plasticity is therefore open to criticism.
Drucker (1954) shows that the upper limit on collapse
loads holds for ‘non-standard’ materials. Radenkovic
(1961, 1962) formulates this result in general terms:
existence of a ‘standard’ instable mechanism proves
instability.
On the other hand, the uniqueness of the collapse
load disappears. Under some actions and depending

47
of the existing pre-stresses, a structure may or may not
collapse. Sometimes, it is nevertheless possible to find
non-null safe limits.
The existence of a safe domain within the friction
domain is extended further by Josselin de Jong (1964,
1973) by including the positive influence of normal
forces on stability. If through the analysis of the entire
structure the value of the minimal normal force acting
on the interfaces can be determined, the lower bound
safe domain can be extended. Palmer technique can
be used to construct it. The limit of applicability of
this procedure to masonry structures has been stud-
ied by Smars (2000). A prove of uniqueness of the
safety factor for vaulted and domed structures under
symmetric loading condition, once the magnitude of
the normal force is known is given by D’Ayala and
Casapulla (2001).
Since Heyman fisrt formulation of the plasticity
theorems for masonry (Heyman 1966), limit analysis
has been successfully applied to masonry in the ‘stan-
dard’ format under the assumption that the friction
coefficient would be high enough to prevent sliding
in the range of problems considered. This powerful
tool was specifically applied to the safety analysis of
masonry arch bridges and later extended to the analy-
sis of vaulted structures; in the search for lower bounds
different methods have been used to find admis-
sible state of stresses: funicular methods (Heyman
1966, Harvey 1988, Huerta 2001), membrane analysis
(Heyman 1966), network of forces (O’Dwyer 1999),
minimum of complementary energy (Maier 1990) or
FE with incremental failure analysis.
Livesley (1978, 1992), by adopting a static
approach, was the first to developed a formal linear
programming procedure to discuss the existence of
safe load factor of two-dimensional vaulted structures.
Within this approach research has developed substan-
tially in the past decade (D’Ayala 1993, Boothby 1994,
Baggio & Trovalusci 1998, Ferris 2001). Friction is
often taken into account but the question of applica-
bility of the general theorems of plasticity is not much
discussed. If it is usually recognised that the maxi-
mum safe load factor can be overestimated, it does not
seem to be clear that it is not unique. Only Mauldon
et al. (1997) make references to stable, instable and
potentially stable domains.
Among masonry structures, arches, vaults and but-
tresses represent the principal structural component of
historic masonry buildings. They epitomise masonry
features and problems. They are mainly subjected to
compressions, are shaped accordingly and commonly
present cracks. Techniques developed for vaults can
help to understand and define techniques for other
masonry structures.
However the interpretation of masonry vaults’
behaviour is difficult, especially when the vaulted sur-
face is not smooth and continuous, leading often to a
48
reduction to an ideal system of many arches. While
this oversimplification could be admissible for barrel
vaults and spherical domes under simple gravita-
tional loads, cannot do justice of the three-dimensional
effects developing in other types of vaults, ribbed or
with discontinuity lines. Such effects may be essential
to explain stability.
For Heyman (1966), the actual stress state cannot be
found. He abandons the behaviour’s ‘quest’ to concen-
trate on safety. In the spirit of ‘standard’ limit analysis,
all methods able to give an admissible stress state can
be used to give a lower bound of the collapse load.
Geometry and specific masses are the basic input data.
In this approach, the possibility of sliding is ruled out
a priori. However cracking is considered and resis-
tant mechanisms able to assure the equilibrium after
cracking are identified (Heyman, 1977).
The Heyman’s study presents a useful and intu-
itive approach to understand the behaviour of masonry
arches and vaults and provides the value of minimum
thickness to span ratio as a safe solution under the
assumptions of infinite compressive strength and fric-
tion resistance and zero tensile strength. Furthermore,
this work suggests that, for complex vaults, the junc-
tion between two shell surfaces leads to a large stress
concentration.
In real vaults however, sliding does occur, especially
if accompanied by a loss of shape. In that case, the
‘lower bound’ estimates can be unsafe. Stereotomy of
the vault, coefficient of friction and resistance to ten-
sion must then be considered. D’Ayala (1993, 1994,
2001) uses the knowledge of extreme meridian nor-
mal forces in a dome to demonstrate its stability, even
in the case of potential sliding. Smars (1993, 2000)
showed that locally stronger structures are not neces-
sarily safer, proposing a technique to build a lower
bound domain for structures having possible local
tension resistance.
Nowadays and to our knowledge, no general-
purpose software exist permitting 3D limit analysis
of vaults. Existing tools are not tailored to their actual
behaviour and complex geometry.
In the last years the developments of constitutive
laws for masonry structure are been made and most
finite element analysis of the masonry vaults are been
carried out (Lourenco, 2006).
Nevertheless, the analysis of historical masonry
constructions is a complex task and only few propos-
als specifically oriented to the non-linear analysis of
masonry domes and vaults exist.
Oppenheim et al. (1989) propose a limit state anal-
ysis of masonry dome, but does not take into account
the interaction between the slices.
Nart (2003) describes the mechanical behaviour of
masonry domes in relation to the applied loads and
geometry. The author’s analysis, based on the geomet-
rical parameters, consider a great variety of shapes and
loads, but assume zero hoop stresses and neglect the
sliding mechanisms.
Interesting is the work by Block (Block et al. 2006),
that propose a structural analysis tools based on the
limit state analysis for vaulted masonry buildings.
This study extends the graphical method for limit
analysis using the line of thrust. Nevertheless it does
not provide a complete analysis of three-dimensional
behaviour and mechanisms of vaulted structures.
Only few studies explain the tree-dimensional
effects. Modelling the principal stresses in a masonry
vault as a discrete network of forces, O’Dwyer (1999)
develops the limit analysis for vaults, able to take
into account redistribution effects. However this work
assume an initial value for the horizontal component
of the resultant of stresses under the condition that the
friction between the voussoirs is sufficient to prevent
failure due to sliding.
An iterative procedure for the analysis of the
masonry structure, that take into account the non-
associative frictional joints, is proposed by Gilbert
et al. (2006) that provide a tool for analysing the sta-
bility of masonry gravity structures. This method, easy
for masonry walls and for rib arches, could be very
difficult to apply for three-dimensional structure.
Both the three-dimensional effects and the slid-
ing mechanism are studied by D’Ayala e Casapulla
(2001).This study analyzes hemispherical domes by
a new analysis tools based on the limit state analysis
with finite friction and provides a simple proof of the
unique solution.
The same procedure, based on membrane theory
with limit state analysis, could be applied at different
type of vaults.
Hence, ascertained the value of finite friction in
the masonry analysis and the importance of three-
dimensional effects for the study to the complex vaults,
the paper present a limit state analysis with finite fric-
tion applied to a masonry domes, pavilion vaults and
fan vaults, in order to evaluate their actual structural
behaviour and the minimum thickness to define the
stability condition, very important element with regard
to structural rehabilitation.
2 STRUCTURAL ASSUMPTIONS FOR
MASONRY VAULTS
The structural behaviour of masonry vaults and their
collapse mechanisms depends on the material prop-
erty, so the present theory is founded on the assumption
of infinite compressive strength for the blocks, no
tension transmitted across the joints and finite friction.
The last hypothesis, often neglected by previous
studies, reflects a more realistic masonry mechanical
behaviour because, especially for historical structures,
after the deterioration of the contact surfaces or of
49
the binding materials, the original friction coefficient
could be substantially reduced. Therefore the shear
strength at blocks interfaces is not infinite, but deter-
mined by the cohesion and the internal friction angle,
having assumed the Mohr–Coulomb criterion as a
good representation of the real behaviour. Although
very little testing has been performed on historic
masonry to quantify both characteristic shear strength
and friction coefficient, this assumption is supported
by work conducted by Hendry et al. (1986) on new
masonry assemblies, providing value of friction coef-
ficient lower than 0.4.
Modelling a vault of any shape as a three-
dimensional discrete system of rigid blocks, along
meridians and parallels, it is possible to determine, by
use of the general shell theory, the meridian stresses,
the hoop stresses and the shear stresses (Fig. 1).
Careful observation of historical vaults and domes
highlights that, unless they are corbelled, and indepen-
dently of the bricks bond (for instance brick courses
parallel to the spring, perpendicular to the spring or
herringbone brickwork), the joints are always placed
in the arch’s normal plane, i.e. towards the centre of
curvature. Hence, as the weaker elements in the fabric
are the joints the analysis can be most usefully car-
ried out at the blocks’ interface, where the stress field
is related to the frictional (plastic) limits, and where
failure by shear or tension will first occur.
While it is evident that the discretization scale could
not be representative of each single brick or block
constituting the fabric, each portion identified by the
intersection of 2 parallels with two adjacent meridians
can be considered as a macroelement of homogenous
masonry material, with infinite internal compression
and tensile and shear strength at the interfaces between
portions defined by a Coulomb criterion.
In the case of double curvature vaulted structures,
for gravity load distributions, if the material has tensile
strength, then membrane theory might be assumed to
apply with resultants of meridian and parallel stresses
tangent to the middle surface of the vault at any point.
As masonry by hypothesis is generally not able to resist
tensile stresses for most shapes of vault the gravity load
distribution will cause tensile stresses near the spring
level that the material is not able to absorb and hence
cracks will form. This means that the stress field is
modified with respect to the membrane theory, the
meridian stress resultant is no more tangent to the
meridian geometric curve and generally not centred,
so that both resultant bending moment and shear stress
resultant are present at the block interface. The curva-
ture of the thrust surface is hence not known a-priori
and it is generally not constant.
As a consequence of the accruing of such stress
resultants, the equations of equilibrium for classic
membrane theory are not sufficient to solve the prob-
lem, which has ∞2 possible solutions. Among those,

Figure 1. Forces and angles for a generic element of vaults
with single curvature; (a) cross vaults, (b) pavilion vaults.
by imposing some condition on material and compat-
ibility, following the theory of plastic limit analysis it
is possible to identify the solution that maximise the
load-bearing capacity of the vault.
Using the same notation introduced in D’Ayala
Casapulla (2001), the first condition, imposed on the
material, is the frictional constraints, expressed as:
where:
As shown in Figure 1 (a,b) , for the generic element
of surface j, θj is the angle between the vertical and
the perpendicular to the elements’s surface (z’ in the
local system) and γj is the angle that the resultant of
stress Sj along the arch forms with the horizontal axis
at each interface. Hence Nj and Tj are the components
of Sj normal and parallel to the interface, respectively.
Equation (1) identifies the maximum and the min-
imum value of the shear resultant and identifies the
50
Figure 2. Oversimplification of forces Nj and Tj, compo-
nents of Sj normal and parallel to the interface J respectively,
and Coulomb’s cone projection.
points A and B on the Coulomb’s cone projection
(Fig. 2). If the tangential force T exceeds the limit,
the vault could fail by way of sliding. This means that,
in case of symmetric loading, there is a unique limit-
ing value of the shear force, and the local equilibrium
problem is at a limit state statically determined.
Imposing the equilibrium condition and applying
the lower bound static analysis, it is possible to identify
a unique solution of the problem, from which the line
of thrusts can be derived and the minimum thickness
quantified.
In the present paper the previous approach is
extended by relaxing the condition of axial symmetry
used for domes. A lower bound approach is devel-
oped for pavilion vaults, where the state of stress and
hence the shear resultant are variable from meridian
to meridian, and where singular conditions develop
along the diagonal as the intersection of two webs is
not smooth. It will be shown that by using optimisation
techniques is possible to fully define the state of stress,
the maximum capacity and the required thickness at
collapse. The results are compared with a non linear
finite element analysis of the same vault.
3 LOWER BOUND APPROACH WITH
FINITE FRICTION
Masonry pavilion vaults, despite their common use
in past centuries, in particular in XVI–XVII century
palaces, have not been studied in any greater depth yet,
principally because due to their singularity of shape
and presence of cuspide form along the diagonal, the
complex 3-dimensinal state of stress that develops in
presence of uniform gravity loading does not lend
itself easily to simplified approaches.
These vaults have continuous support along walls;
have finite curvature in the direction of the meridi-
ans, but infinite curvature along horizontal plane. This
means that there can be no contribution to recentre the
line of thrust from hoop stresses and hence membrane
theory is inherently non applicable. This means that
such type of vaults is often affected by cracks along the
diagonals, where the geometry of the vault is farthest
from the geometry of the thrust surface generated from
the gravity load distribution. Cracks can also appear
toward the centre of the web of each portion so, in
the past, they are been often modeled as independent
arches. Figure 3. Geometric characteristics of the vault’s generatrix.
However the simplified arch model, does not
account for the capacity of the vault to transfer load
associated with shear and for the arch effect that can
be develop within the horizontal strips due to their
non negligible thickness, similarly to the arch effect
accounted for in walls or slabs.
On the basis of this considerations and suitably sim-
ulating the boundary conditions and the condition of
equilibrium that arise along the diagonal due to the
singularity of the surfaces along the diagonals, limit
state analysis can be applied to this family of vaults to
determine their state of stress at failure.
The main objective of the present section is to under-
stand, through the limit state analysis approach, the
structural behaviour of pavilion vaults, in order to eval-
uate with more accuracy the actual stress field that can
cause cracks.
Moreover this method allows finding the thrust line
position and the admissible thrust surfaces, leading to
the calculation of the minimum thickness which sat-
isfies at the same time equilibrium and compatibility.
This is of great importance with regard to strength-
ening interventions because it allows determining the Figure 4. Geometric characteristic of the diagonal and of the
vaults’ safety factor, which can be expressed, as pro- vault’s generatrix, used for the centring work (Levi, 1932).
posed by Heyman as the ratio between the geometric
thickness over the minimum required thickness, but the equation of an ellipsis (Curioni, 1870; Breymann,
also the accurate position of the hinges at failure and 1885; Levi, 1932).
hence the appropriate positioning of ties or other thrust Assuming the n slices are made up of m blocks,
contrasting devices. each block is identified, in the global system, by the
The simplest case to analyse is a pavilion vault over a coordinates of its centre of mass:
plane square area and of semicircular curvature, which
is also the most frequent occurrence according to the
technical literature (Scamozzi, 1615), subject to self
weight loading.
Figure 3 shows the geometric characteristic of the
vault’s generatrix, where R is its radius, f is the vault’s
rise, l is the span. In the numerical procedure the slices
that make up half of the web between to successive for i = 1 to m and k = 1 to n
ribs are considered, these being present only along were αk is the horizontal angle between the x axis
diagonal. The ribs according to historic the technical in the global system and the horizontal projection of
literature, as shown in figure 3, can be described by x-axis in the local system as shown in Figure 5 and θi

51
 Hence, the eccentricity of the thrust surface at each
point is:

and the minimum constant thickness t required is:

This can be the objective function to minimise.

The meridian, hoop and shear stresses are calculated
considering the coordinates of the generic surface of
thrust at the interface between blocks. Subsequently,
by limiting the maximum values of the internal shear
force Tj , component of Sj parallel to the interface,
which can not be greater than the frictional strength at
each block interface, the thrust surface’s zi coordinates
that satisfy the material constraints are determined.
The coordinates of the generic thrust surface at the
interface are:

Figure 5. Generic block with the coordinates of its centre

of mass and the angles αk , θi and θi .

is the angle between the vertical and the perpendicular

to the generatrix. The relation between θi and θi is:

The state of stress for each slice’s element is completely

defined by equilibrium equations and the resultant of In equation (13) γj is the angle that the resultant of
stresses depends on the position of zi . stress Sj along the meridian forms with the horizontal
The zi coordinates are assumed as geometric axis at each interface and it is defined as:
unknowns and the position of the thrust surface is
found by means of a discrete approach through the
calculation of its coordinates, hence:

Hence, simplifying and neglecting the second order

terms, with regard to a slice element, the equilibrium
The distance between the origin of the axes and each along the local z axis yields the meridian force Sj :
point of the thrust surface is:

In (17), having defined ω the weight for unit surface

and n the number of meridian slices (see figure 5), the
geom
where li is the horizontal projection of the distance weight Wj of the portion of each block identified by
between the block’s centre of mass and the origin of angles ϑj and ϑj+1 , is:
the global system.
The distance between the origin of the axes and each
point of the middle surface is:

where γi is the angle that the tangent of the thrust line

at the centre of each element forms with the horizontal
axis.

52
 Figure 7. Cracked masonry showing the shear force Txθ and
Figure 6. (a) view from the top of the pavilion vault, with the force Sj transmitted between the blocks.
the forces for a generic element, (b) axonometric showing the
angles γi and α’k . Hence the horizontal force Hp along the diagonal is:
The shear force at the interface between two slices,
shown in Figure 6a, can be quantified using translation
equilibrium:
Starting from the diagonal slice, for horizontal
equilibrium, the parallel force Hp−1 is:

where the subscript p indicates the lateral surface of

the generic block.
Given the symmetry of the problem, for the inter-
face between the slices at the centre of the web, If Hp exceeds the tensile strength of material (by
corresponding to p = 0, the shear resultant Txθ = 0. assumption equal to zero), the hoop stresses are
Hence, using rotation equilibrium around an axis not present and the horizontal equilibrium equation
normal to the surface, the shear resultant Tθx at any reduces to:
location is:

For equilibrium to rotation around an axis normal

where the angle α’k is the angle αk projected in the to the element surface a resultant shear force Txθ is
plane tangent to the thrust surface at the centre of present due to the brick staggering (see Fig. 7) and
element (Fig. 6b): its components along the S direction at the j and j + 1
interfaces are the same.
Hence the meridian resultant S after cracking
becomes:
For j = 0 the shear force Tθx = 0.
The horizontal force Hp can be determined impos-
ing the boundary condition along the diagonal and
considering half web.
d
The shear force Txϑ , corresponding to shear force
Txθ along the diagonal, can be determined by consid- And the corresponding angle γ ∗ is equal to:
erations of global equilibrium:

Starting from i = 1, for a generic block m along the

diagonal, identified by the surfaces j − 1 and j, the Both the normal and shear resultant at this interface
d
shear force Txϑ calculated at its centre i is: can be obtained as components of S∗ as follows:

53
Imposing the frictional constraint:

The optimum programming problem is completely

defined if the maximum eccentricity of the thrust sur-
face is minimized as shown in (11). Conditions (35)
together with the target function (11) or (12) com-
pletely define the physical state of the vault and its
mechanism of collapse, as the values of the indexes i
and j for which the = sign applies in (11) and/or (35)
identifies the cross sections along the arch for which
either rotation or sliding are about to occur.
Furthermore the problem set in these terms can Figure 8. Division of an half web in slices and nomenclature
be easily extended to generic conditions of loading for the procedure.
and constraints as the symmetry is only invoked to
define an initial value of shear resultant at midspan,
but this can easily be obtained by using generic global
equilibrium considerations.

4 RESULTS FOR SQUARE PAVILION VAULTS

The limit state analysis with finite friction consti-

tutes a new tool to define the 3D structural behaviour
of masonry vaults. By means of an easy and rigor-
ous procedure, developed with commercial programs
(for example Excel), it is possible to obtain an opti-
mum solution considering the load redistribution after
cracking.
Identifying the optimal thrust surface, the proposed
analytical method, based on lower bound approach,
allows determining, for any vaults, some important
elements with regard to safety assessment and needs
for intervention, such as:
• the masonry vaults’ safety factor, expressed as the
ratio between the actual thickness over the minimum
thickness required by (12);
• the resultant tensile forces in the parallel direction
and hence the actual crack pattern (equation 28);
• the complete resultant stress field and the thrust
surface’s eccentricity;
• the resultant shear force at blocks interfaces;
• the horizontal thrust at the supports.
Furthermore, this approach is able to provide in very
modest time the same results obtained by means of
nonlinear FEA analyses, often too laborious because
of their high computational burden and the difficulty
in interpretation of the results.
To demonstrate the effectiveness of the procedure
and to clarify the role of interaction between arches
in complex vaults, the results on pavilion vaults are
discussed in this section and the comparison with a
FE model will be shown in the following one.
Considering a pavilion vault with generatrix radius
R equal to 3 m, vault’s rise f equal to 3 m and span l
Figure 9. Meridian force S before cracking, after cracking
and after the optimization for the central slice.
equal to 6 m, the results are presented for the slices that
make up half of the web between to successive ribs.
Figure 8 shows the four slices that compose arbitrarily
the portion of vaults considered.
4.1 Meridian force S
After cracking the meridian stress resultant is no more
tangent to the meridian geometric curve and generally
not centred. Figures 9 and 10 illustrate the variation of
meridian force S with respect to the horizontal coor-
dinate x, in absence of cracking, after cracking has
occurred and after cracking and running the optimiza-
tion procedure to obtain the minimum eccentricity, for
slices 1 and 4. The chart show that the resultant S,
before cracking, increase up to 2 m, corresponding to
θ = 45◦ , and reduce in the final portion of the slice.
This solution corresponds to the membrane solution
that might occur if the material has tensile capacity.
Since the masonry is generally not able to resist tensile
stresses, near the spring the cracks along the meridian
direction will form and the force S* departs from S
along the cracked surface.
Figures 9 and 10 also show the force S* obtained
minimizing the maximum thrust surface’s eccentricity

54
Figure 10. Meridian force S before cracking, after cracking
and after the optimization for the slice near to diagonal.
Figure 11. S*, normal force N and shear force T at the
interface for the central slice.
and imposing the material constrains. This force
presents the same trend of the S* before optimiza-
tion, but the optimum S* diverges earlier from the
membrane solution, indicating that cracks will extend
further in along the meridian but that the value of the
resultant is smaller. This change in value id directly
related to the distribution of eccentricity, especially for
the slice closer to the diagonal as it will be seen in 4.3.
4.2 Normal force N and shear force T at the
interface
The resultant meridian normal force N and shear force
T at the interface, components of S* normal and paral-
lel to the interface respectively, are shown in Figures 11
and 12. As it can be seen, the force N shows the same
trend of S* while the shear force T is equal to zero
up to the cracking point, where the thrust line is per-
pendicular to the block interface, and it increases in
the cracking area, where the funicular breaks away
from the middle surface because of the absence of axial
forces along the parallel. Besides, it can be observed
that, from the centre of the web to the diagonal, the
shear force T increase. Indeed the angle γj , repre-
senting the inclination of the thrust line, increasingly
differs from the geometric angle θj moving from the
centre of the web to the diagonal.
55
Figure 12. S*, normal force N and shear force T at the
interface for the slice near to diagonal.
Figure 13. Thrust surface’s eccentricity for the
non-optimized solution and for the optimum solution
(slice near to diagonal).
4.3 Thrust surface’s eccentricity
The thrust surface’s eccentricity is shown in Figure 13.
It should be noted that the non-optimized solution dif-
fers everywhere from the optimum solution. The non
optimized solution, indeed, is equal to zero along the
uncracked area and it has an exponential trend from
the haunches to the support, because the absence of
the hoop stresses requires an infinite thickness.
It is well known that, with a lower bound approach,
between the ∞ possible solutions, the actual solution
is the one that maximise the load-bearing capacity
of the vault while the thrust surface is constrained
within the vault’s thickness: imposing the condition
of minimum eccentricity and the condition that the
friction between the voussoir is sufficient to prevent
failure due to sliding, the actual solution can be iden-
tified. Hence, as shown in Figure 13, the eccentricity
in the optimum solution reduces drastically and the
presence of relative peaks indicate the possible for-
mation of plastic hinges at the extrados, at x equal
to 1.5 m (θ j = 30◦ ), and at the intrados, about at x
equal to 2.78 m (θ j = 65◦ ). The eccentricity presents
its absolute maximum at the support.
The chart in Figure 14 shows the corresponding
bending moment for the four slices, calculated as
the product of the meridian normal force N and the
Figure 14. Bending moment for all the slices belonging to
half web.
eccentricity. The maximum bending moments for all
slices are localized at the support confirming the
presence of extrados plastic hinges there.
4.4 Horizontal thrust at the supports
The possible intradox plastic hinge at x equal to 2.78 m
and usually the presence of a spandrel wall that con-
tribute to increase the thickness in the portion of the
vault near the support, has the effect of identifying only
the portion contained between an aperture of the sub-
tending angle of 2 θj as the actual portion of structure
behaving as a vault, the lower part being usually con-
sidered as solid with the supporting walls. This in turn
means that the critical cross section along the merid-
ian where the horizontal thrust should be assessed is
at θj = 65◦ , according to the previous results.
Furthermore, as it can be seen in Figure 15, showing
the horizontal component of S* along the meridian at
the centre of the web, the horizontal thrust is constant
in the cracked portion, where the hoop stresses are not
present.
The values of the horizontal thrust however is
also directly influenced by the lateral stiffness of the
perimeter walls supporting the vault: depending on the
effective slenderness of the walls and their connection
to the vault, they will have a direct effect on restraining
lateral movement of the vault and hence on the value
of the thrust developed within it. The limit conditions
of roller support and full fixity are both unlikely to be
met I practice. For the purpose of the procedure the
supports are considered as fixed, and this will provide
an upper limit of the value of the thrust. The value
of horizontal thrust is very important in a strengthen-
ing intervention because allows to determine the ties
position and strength.
Figure 16 shows the variation of horizontal thrust
along the support parallel, for half of the web length
before and after the optimization. It should be noted
that the first solution corresponds to considering each
slice independently and yields a minimum value at
the centre of the web were the ratio of rise over span
is higher. After the optimization redistribution occurs
56
Figure 15. Horizontal component of S* along the meridian
in the web centre.
Figure 16. Horizontal thrust for the non-optimum solution
and for the solution after the optimization.
via shear and arch effect at parallel level yielding the
opposite results with a trend slightly decreasing from
the centre to the diagonal.
5 COMPARISON WITH FINITE ELEMENT
MODEL
The validation of the proposed approach based on the
limit state analysis is carried out by comparing the
results in terms of stress resultants with the output
obtained by advanced non linear FE modelling.
The Finite Element Model has been constructed
using the Algor V19 FE program: shell elements are
used for all uncracked masonry portions and contact
elements are inserted at the interface between shell
elements where tensile stress develops.
The following material properties have been
assumed: specific weight γ = 1850 kg/m3 ; Poisson’s
ratio ν = 0.15; Elastic Modulus E = 5000 MPa.
For an easier comparison between the results of the
limit state analysis and the FE analysis, the FE mesh
has half the size of the limit state analysis along the
parallel but the same number of subdivisions along the
meridian, (Fig. 17).
To simulate the presence of perimeter walls high
3 m and thick 0.50 m, Shell elements are been arranged

Figure 17. Axonometric showing the Finite Elements model
of the pavilion vault.
Figure 18. Comparison between the meridian forces S
obtained by the limit state analysis and the Finite Elements
Analysis along the central slice.
along the supports and they are constrained by fixed
nodal boundary conditions at the abutments.
Figures 18 and 19 show the comparison between
the meridian forces S obtained by the optimised limit
state analysis and the Finite Elements Models along
the central slice and along the slice near to diago-
nal respectively. Both curves present the same trend,
characterized by a change of curvature in the crack-
ing area, this means that the new theory based on
limit state analysis with finite friction accurately pre-
dicts the behaviour of pavilion vaults, identifying the
actual crack pattern and the meridian resultant S field.
As it can be seen, in agreement with the safety theo-
rem of limit state analysis associated with lower bound
approaches, the values of S as calculated by the pro-
cedure are always slightly greater than the actual state
of stress identified by the FE analysis.
Also the normal force N and shear force T have the
same trend in the two models.
Furthermore, the eccentricity and hence the posi-
tion of the thrust surface is accurately calculated with
57
Figure 19. Comparison between the meridian forces S
obtained by the limit state analysis and the Finite Elements
Analysis along the slice near the diagonal.
Figure 20. Comparison between the eccentricities obtained
by the limit state analysis and the Finite Elements Analysis
along the slice near the diagonal.
the limit state analysis. As it can be seen, again in
favour of safety, the values computed by the proce-
dure are slightly greater than the values obtained with
the FE, however the location of relative minimum
and maximum and absolute values identify with great
accuracy the actual position of plastic hinges along the
slices.
The identification of the position of the extrados
hinge is particularly critical, as this also define the
position along the arch of the maximum horizontal
thrust and hence is essential for the correct positioning
of ties or for the construction of abutments.
Values of horizontal thrust have been compared for
the two analyses for and angle θ = 65 degrees and for
the condition of full lateral restraint at the support
for the F.E. model.
As it can be observed in Figure 21 where the thrust is
plotted along a half parallel between midspan (y = 0)
and the diagonal, the trend is again similar in the two
cases with the values estimated by the procedure being
clearly equal to the F.E at midspan and slightly over-
estimated as progresses toward the diagonal with a
maximum difference of 20% .
Hence it is possible to affirm that the new theory is
able to provide the actual crack pattern and the stresses
field in the masonry vaults.

Figure 21. Comparison between the horizontal thrust
obtained by the limit state analysis and the Finite Elements
Analysis.
Figure 22. Horizontal thrust along the supports obtained by
the simplified arch model (a) and the actual horizontal thrust
obtained by the limit state analysis (b).
6 DISCUSSION OF RESULTS
The limit state analysis with finite friction and the
FE analyses have allowed to clarify some important
aspects of the complex structural behavior of pavilion
vaults, in particular concerning the three-dimensional
effects and the importance of the sliding. The results on
pavilion vaults demonstrate that also vaults with rigid
boundary at the supports are affected by cracks along
the diagonals, unlike the common opinion that ascribes
the diagonal cracks only to the walls’ overturning.
In addition the structural behavior of masonry pavil-
ion vaults can not be reduced to a series of adjacent
arches, without transversal connection. In fact, in the
web, despite the cracks along the diagonals, any arch
can transfers the stresses at the arches beside it. This
interaction is clear observing the trend of horizontal
thrust applied by the vaults on the walls (Fig. 22): as
it can be seen, the actual horizontal thrust is about
constant, differently from the few indications about
the pavilion vaults reported in the scientific literature,
which indicates a triangular trend resulting from the
simplified arch model.
This means that the web develops a series of natural
arches defining an alternative load path from the centre
of the web to the diagonal.
The theory based on the limit state analysis also
allows to identify the thrust surface and to evalu-
ate the bending moments for any slice. The results
58
Figure 23. Possible vault’s collapse mechanism for any
slices.
Figure 24. Maximum thrust surface’s eccentricity obtained
by the limit state analysis with finite friction for different
friction coefficients µ.
show that two plastic hinges might form: the first one
located at the extrados, for x equal to 1.5 m (θ j = 30◦ ),
and the second one at the intrados, for x equal to
2.78 m (θ j = 65◦ ). Figure 24 shows the likely collapse
mechanism.
In addition, the vault could fail by way of sliding.
Using the limit state analysis with finite friction it
is possible to evaluate the minimum thickness which
satisfies at the same time equilibrium and compati-
bility. The application of this new theory on pavilion
vaults shows that, for a friction coefficient higher than
0.48, the shear force T are everywhere smaller than
N. µ, hence the results are unrelated to the friction
coefficient and sliding mechanism are prevented.
On the other hand, in the 0.3–0.48 friction coeffi-
cient range, the vault is able to find a new equilibrium
system, in which the stress resultants are unvaried,
but the eccentricity considerably increases (Fig. 24).
This means that the thickness necessary for the vault’s
stability would increase and consequently the safety
factor decrease. It is necessary to specify that the max-
imum eccentricity is located near the support, where
typically the spandrel ensures a big thickness.
For friction coefficient lower than 0.3 the Excel’s
solver cannot find a feasible solution, i.e. it cannot
satisfy all constraints.
Moreover, the structural behaviour of masonry
vaults could vary with the rise-span ratio f/l. As
pointed out by Palladio (1570) and Guarini (1737), the
most frequent geometry in XVI-XVII century palaces
would typically results in f/l ratios equal to 1/3, 1/4 and

Figure 25. Comparison between the meridian forces S
obtained by the limit state analysis after meridian cracking
and before meridian cracking for the rise-span ratio equal
to 1/3 (rise = 2 m), 1,5 (rise = 1,5 m) and 1,2 (rise = 1,2 m)
respectively (central slice).
Figure 27. Comparison between the eccentricities obtained
by the limit state analysis for the rise-span ratio equal to
1/3 (rise = 2 m), 1,5 (rise = 1,5 m) and 1,2 (rise = 1,2 m)
respectively (central slice).

Figure 28. Comparison between the hoop stresses Hp

obtained by the limit state analysis for pavilion vaults with
a rise-span ratio equal to 1/3 (rise = 2 m), 1,5 (rise = 1,5 m)
Figure 26. Comparison between the meridian forces S and 1,2 (rise = 1,2 m).
obtained by the limit state analysis after meridian cracking
and before meridian cracking for the rise-span ratio equal
to 1/3 (rise = 2 m), 1,5 (rise = 1,5 m) and 1,2 (rise = 1,2 m)
respectively (slice near to the diagonal).

1/5. Figures 25 and 26 show the comparison between

the meridian forces S obtained with the limit state anal-
ysis after cracking and before cracking for pavilion
vaults over a plane square area, of semicircular cur-
vature and of rise equal to 2 m (f/l = 1/3), to 1,5 m
(f/l = 1/4) and to 1,2 m (f/l = 1/5) respectively. As it
can be observed, the force S increase with decreasing
rise-span ratio, nevertheless, decreasing the rise, the
classical membrane solution, that represents the stress Figure 29. Comparison between horizontal thrust along the
field before meridian cracking, converges on the opti- supports obtained by the limit state analysis for pavilion
mum solution. In fact, reducing the f/l ratio, also the vaults with a rise-span ratio equal to 1/3 (rise = 2 m), 1,5
eccentricity’s trend is towards zero (Fig. 27) and the (rise = 1,5 m) and 1,2 (rise = 1,2 m).
tensile area near to the spring decreases (Fig. 27); this
means that, in shallow vaults, the geometry is closer to 1/5, the classical membrane theory could be used with
the thrust surface generated by the gravity load distri- good approximation.
bution, and compression hoop stresses Hp are present As it is well known, the horizontal thrust increase
for a grater portion of the surface (Fig. 28). This means if the vault’s rise reduces; furthermore it is possible to
that the spread of the cracks that often develops along observe that its trend along the perimeter walls tends
the diagonal and in the centre of the web depends on the to become constant for shallower vaults (Fig. 29). This
vault’s rise and hence, for f/l ratio smaller than about means that, reducing the rise/span ratio, the natural

59

Figure 30. Qualitative developments of natural arches for
pavilion vaults with a rise-span ratio equal to 1/3 (rise = 2 m),
1/5 (rise = 1.5 m) and 1/2 (rise = 1,2 m).
arches, that develop in the pavilion vaults and that
cause the transfer of meridian stresses from the centre
of the web to the diagonal, affect a smaller area near
to the diagonal, as it can been see from the qualitative
pictures in Figure 30.
7 CONCLUSIONS
The paper presents a new 3D limit state analysis for
vaulted structures, based on finite friction, able to
describe the actual structural behaviour of a wide range
of masonry vaults. Assuming finite friction between
block interface and using a lower bound approach, this
theory allows the identification of the optimal thrust
surface, so it is possible to define, for a generic type
of vaults, the actual crack pattern and the stress field.
The development of a computer procedure is of great
importance with regard to strengthening interventions
because it allows determining easily the masonry
vaults’ safety factor, given by the actual thickness over
the minimum thickness ratio, to understand the cause
of the cracks and to find the horizontal thrust on the
supports.
In addition, the limit state analysis with finite fric-
tion allows investigating aspect previously neglected.
In particular, while it is recognized that often
masonry vaulted structures fails by way of sliding,
most of existing analysis on masonry vaults don’t
consider the limited friction. However the use of fric-
tion limits is particularly valuable because, after the
deterioration of the contact surfaces or of the bind-
ing materials, the original friction coefficient could be
reduced and sliding mechanism can occur, especially
for the pointed vaults.
Furthermore the vaults are often schematized as
arches side by side, which do not work together.
60
The results obtained by the limit state analysis with
finite friction show that this schematisation, especially
for complex vaults, do not provide information about
the actual stress field. In fact, in the case of pavilion
vaults, it is emerged the importance of the interac-
tion between the arches for the evaluation of the crack
pattern and the horizontal thrust.
Therefore, the present paper provide a new anal-
ysis tool for a computer procedure able to value
both the rotational and sliding mechanisms, the three-
dimensional effects in the vaults, very important
especially for complex vaults and the vaults’ safety
factor, given by the minimum thickness over the actual
thickness ratio.
ACKNOWLEDGEMENTS
The authors wish to thank Prof. Ezio Giuriani from
the University of Brescia for suggesting the topic and
providing critical insight during the development of
this study.
REFERENCES
Amadei B. et al. 1995. Modelling the stability of masonry
structures with the discontinuous deformation analy-
sis (DDA) method. In 3dt International Symposyum on
Computer Methods in Structural Masonry, Lisbon.
Baggio C. & Trovalusci P. 1998. Limit analysis for no-
tension and frictional three-dimensional discrete systems.
Mechanics of Structures and Machines 26(3): 287–304.
Block P. et al. 2006. Real-time limit analysis of vaulted
masonry buildings. Computer and structure 84(29–30):
1841–1852.
Boothby T. E. & Brown C. B. 1993. General lower and upper
bound theorem of static stability, Engineering Structures
15(3): 189–196.
Boothby T. E. 1994. Stability of Masonry Piers and Arches
Including Sliding. Journal of Engineering Mechanics
120(2): 304–319.
Boothby T. E. 1994. Limit states analysis of masonry arches.
In Proceedings of the Structures Congress ’94, Atlanta,
24–28 April 1994. ASCE.
Breymann G. A. 2003. Archi, volte, cupole (1885). Roma:
Dedalo.
Casapulla C. & D’Ayala D. 2001. Lower bound approach to
the limit analysis of 3D vaulted block masonry structures.
In Computer Methods in Structural Masonry (STRUMAS
V); Proc. 5th intern. symp., Roma, 18–20 April 2001.
Cundall A. 1971. Formulation of three-dimensional distinct
element model-part I: a scheme to detect and represent
contacts in a system composed of masonry polyhedral
blocks. Int. J. Rock Mech. Min Sci. Geomech. 25(3):
107–116.
Curioni G. 1870. L’arte di Fabbricare. Costruzioni civili,
stradali, idrauliche. Torino: Negro.
D’Ayala D. 1994. In tema di comportamento strutturale delle
cupole in muratura. Ph.D. Thesis, Dept. of “Ing. Strutt. e
Geot.”, Univ. “La Sapienza”, Roma.
D’Ayala D. & Casapulla C. 2001 Limit state analysis of hemi-
spherical domes with finite friction. In Lourenço P.B.,
Roca P. (Eds.). Structural Analysis of Historical Construc-
tions, Proc. intern. seminar, Guimaraes 7–9 November.
Guimaraes: Balkema.
Drucker D. C. 1954. Coulomb friction, plasticity and limit
loads. Journal of Applied Mechanics 21(1): 71–74.
Ferris M. C. 2001. Limit analysis of frictional block assem-
blies as a mathematical program with complementarity
constraints. International Journal of Mechanical Sciences
43(1): 209–224.
Gilbert M. & Melbourne C. 1994. Rigid-block analysis of
masonry structures. The structural engineering 72(18):
356–361.
Gilbert M. et al. 2006. Limit analysis of masonry block
structures with non-associative frictional joints using lin-
ear programming. Computers and Structures 84 (13–14):
873–887.
Giuriani E. et al. 2001, Structural rehabilitation of masonry
vaults. In UNESCO-ICOMOS millennium congress;
Proc. intern. symp., Paris, 10–12 Septembre 2001.
Giuriani E. et al. 2002. Studio del comportamento strutturale
delle volte a padiglione in muratura. Technical Report of
University of Brescia 17: 1–327.
Guarini G. 1968. Architettura civile (1737). Milano: Il
polifilo.
Harvey W. J. 1988. Application of the mechanism analysis to
masonry arches. The structural Engineer 66(5), 77–84.
Heyman J. 1977. Equilibrium of shell structures, Oxford:
Clarendon Press.
Heyman J. 1996. Arches, Vaults and Buttresses, Norfolk:
Variorum.
Huerta S. 2001. Mechanics of masonry vaults: The equi-
librium approach. In Lourenço P.B., Roca P. (Eds.),
Structural Analysis of Historical Constructions, Proc.
intern. seminar, Guimaraes 7–9 November. Guimaraes:
Balkema.
61
Levi C. 1932. Trattato teorico-pratico di costruzioni civili,
rurali, stradali ed idrauliche. Milano: Hoepli.
Livesley R. K. 1978. Limit Analysis of structures formed
from rigid blocks. International Journal for Numerical
Methods in Engineering 12(11): 1853–1871.
Livesley R. K. 1992. A computational model for the limit
analysis of threedimensional masonry structures. Mecca-
nica 27(3): 161–172.
Lourenço P. B. 1996. Computational strategies for masonry
structures. Thesis Delft University of Technology. Delft
University Press.
Lourenço P. B. 2001. Analysis of historical constructions:
From thrust-lines to advanced simulations Historical Con-
structions. In Lourenço P.B., Roca P. (Eds.), Structural
Analysis of Historical Constructions, Proc. intern. semi-
nar, Guimaraes 7–9 November. Guimaraes: Balkema.
Lourenco P. B. 2006. Recommendations for restoration of
ancient buildings and the survival of a masonry chimney.
Construction and Building Materials 20(4): 239–251.
Nart M. 2003. Limit analysis of masonry domes. Masonry
international 16(1): 12–20.
O’Dowyer D. 1999. Funicular analysis of masonry vaults.
Computer and Structures 73(1–5): 187–197.
Oppenheim, I. J et al. 1989. Limit state analysis of masonry
domes. Journal of Structural Engineering 115(4):
868–882.
Palladio A. 1980. I quattro libri dell’architettura (1570).
Milano: Il polifilo.
Scamozzi V. 1964. L’idea dell’architettura universale (1615),
Ridgewood: Gregg.
Smars P. 2000. Etudes sur la stabilité des arcs et voûtes,
Ph.D. thesis, Department of Civil Engineering, Katholieke
universiteit Leuven.
Tomasoni E. 2008. Le volte in muratura negli edifici storici:
tecniche costruttive e comportamento strutturale. Ph.D.
Thesis, DICATA, Università degli Studi di Brescia.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Repairs to historic timber structures: Changing attitudes and knowledge

D.T. Yeomans
ICOMOS, UK

ABSTRACT: The conservation of timber structures has improved in recent years with a growing appreciation
of their historical significance. However there are still difficulties presented because of the limited number of
engineers with an understanding of timber structures and the failure of design codes to consider the kind of
detailing often used in historic structures. This paper will point out a number of areas where research on the
behaviour of ‘traditional’ carpentry would be valuable.

1 INTRODUCTION
Since the eighteenth century timber in Britain has been
a rather Cinderella material, essential to construction
but not highly regarded compared with masonry as
an architectural material. It did enjoy something of a
revival during the nineteenth century for roof struc-
tures of the Gothic revival but, apart from that, it
has been a poor relation to other materials for struc-
tural purposes. That is to some extent because Britain,
not having large supplies of structural species has
imported most of its building timbers, but also a legacy
of the Fire of London and the switch to brick. The sub-
sequent ‘Georgianising’ of timber-framed buildings
across the country was probably as much a fashion
statement as a concern for fire protection. One could
argue that it is not having a continuing tradition of tim-
ber structures that has led to a lack of concern for the
history and historic value of such structures, except
for exposed half-timbering. Be that as it may, the his-
torical significance of timber structures has only been
recognized relatively recently, so that when we con-
sider the approaches to conserving timber structures
we have to consider the development of ideas about the
significance of these structures as well as that of the
technical solutions; an understanding of their history
as much as of their structural behaviour. It is true that
it has for some time been a material of interest to a few
enthusiasts, people like the late Freddie Charles, and its
importance was recognized in those buildings where
it was a prominent architectural material: the roofs of
open halls and East Anglian churches for example. But
otherwise it was taken little account of.
In the nineteen sixties there was a series of studies
of the major monastic barns, some undertaken through
the collaboration of Freddie Charles and the American
scholar Walter Horn that recognized the significance
of what were the greatest timber structures of their
day (Charles & Horn, 1973: Horn & Charles, 1966,
1984). At much the same time, Stuart Rigold (1966)
did work on the many timber barns in Kent, and Cecil
Hewett (1967), having looked at the barns of Cressing
Temple, went on to draw, examples of major medieval
carpentry from all over the country, his work culmi-
nating in two large collections (Hewett, 1980 & 1985).
His work did a great deal to draw people’s attention to
these structures but, because of the inaccuracy of his
drawings, regrettably little to advance scholarship in
the area. (One could even argue that it set back scholar-
ship since many must have assumed that this work had
already been done.) Timber buildings were also a com-
mon interest of students of vernacular architecture, so
that by the end of the 1970s there was an extensive cov-
erage of medieval carpentry. In contrast little had been
done at that time on later carpentry work and for that
reason I chose to look at seventeenth and eighteenth
century carpentry to provide an outline of develop-
ment during that period that I hope others might follow
up on.
This still leaves the nineteenth century, both in terms
of the development of industrial structures, the tran-
sition to iron and architects’ use of exposed timber
structures in both Gothic revival buildings and their
later treatment of the material. Only Booth has done
work on nineteenth century laminated timber. There
are also significant 20th century timber structures that
we are now loosing. Oxford Road Station, Manchester,
a group of conoids, was recently imitated in steel.
The histories of early carpentry provided an account
of developments in the overall form of timber struc-
tures, their structural design if you will, but carpentry
is a craft skill and we also need to consider its details.
Again Hewett provided the general reader with draw-
ings of joints but that is something requiring careful
observation and measurement. In this the work that I
know best is that of Richard Harris, now curator of the

63
Weald and Downland Museum, and Peter McCurdy,
who I would describe as a scholar craftsman. These are
people who have worked closely with the timbers and
who have come to understand and bring to our atten-
tion the significance of setting out marks that one finds
within the timbers, and have a sound understanding of
the process from conversion of the log to the erection
of the building. Although they have published rela-
tively little, in their teaching they have shown us that
carpenters marks are more than just the numbering cut
into the timbers. For imported timbers there has been
a little interest in merchant marks that indicate their
sources, significant because of the changing patterns
of trade, particularly during the late eighteenth and
early nineteenth centuries, although this work is still
in its infancy.
What historical significance these clues have I leave
to others, but the significance of this work is that
aspects of the surviving timbers are now being rec-
ognized that were simply not seen a few decades ago.
Moreover, as many timbers have been reused, it has
sometimes been possible to reconstruct the design of
earlier structures that the timbers were a part of. For
example, the earlier roof of Lincoln Cathedral has been
reconstructed from timbers that now form part of the
present roof (Foot et al., 1986), but which were reused
from the previous. An example of more recent reuse
was that timbers from the original gallery structure at
Christ Church, Spitalfields were reused for the floor
of later pews.
This means that surviving timbers are now being
recognized as conveying evidence of the development
of structural design, as evidence of craft practices and
as evidence of the earlier form of the buildings in
which they are found. This is a change from the scant
value that they were given only a few decades ago.
I recall visiting the roof of Norwich Cathedral to look
at the timbers in the roof shortly before they were
taken away. By that time the transepts had already been
re-roofed using reinforced concrete A frames, involv-
ing the total loss of the original timbers. In today’s cli-
mate an alternative and rather different solution would
surely be sought in similar circumstances, perhaps
using a supplementary structure rather than a complete
replacement. We are perhaps beginning to recognize
that structural design is a cultural activity whose his-
tory is to be valued as much as carving or wall painting.
However, medieval structures still seem to be the main
focus of attention with far less understanding of the
history of early modern structures.
This is disappointing because without the develop-
ment of new carpentry structures in the seventeenth
and eighteenth centuries the forms taken by the great
public buildings, churches and country houses would
simply not have been possible. But the carpentry
structures that made this architecture possible are
largely invisible and so little attention is paid to them.
64
Much the same might be said of the structures of the
nineteenth century and even the early 20th century, and
the danger is that we will loose historically significant
structures before their significance is recognized.
2 ANALYSIS OF THE STRUCTURE
Recognizing the value of structures is the first step to
their preservation, but the second is their proper struc-
tural analysis. Here it is unfortunate to have to report
that their behaviour is not always properly understood,
with the result that unnecessarily heavy-handed repairs
are sometimes specified leading to the loss of historic
material and hence the loss of historic character. On
more than one occasion I have been contacted by those
concerned when their engineers have specified the use
of steel connectors for timber structures that had been
performing in a perfectly satisfactory manner. This can
be either because of an unwillingness to treat the struc-
ture on its own terms, or simply the application of
computer software better suited to steel or concrete
structures.
In some cases it is clear from the places where the
steel has been specified that the structural form has
been analysed as if it were a steel structure, result-
ing in assumed tension forces that the timber joints
cannot take. However, if these members are removed
from the analysis what remains might be a perfectly
satisfactory structure. In spite of what seems obvious
foolishness, there have even been a number of papers
published where finite element analysis has been inap-
propriately applied to historic timber structures. In one
case that I was sent to referee (I will not give a refer-
ence to protect the foolish.) the author came out with
the startling discovery that in a king post structure, it
was the post rather than the tie beam that would be
carrying the majority of the load. A moment’s thought
would have told him that the principal rafter/king post
assembly must be stiffer than the tie beam. I have a
feeling of unreality when reading such work. I have
to assume that such studies are carried out to impress
academic colleagues rather than as contributions to the
practice of either history or conservation.
The message here is that ‘work is no substitute for
thought’, but it remains a problem when the work is
done by a computer while it is the engineer who is
required to do the thinking. It is likely to remain a
problem because the modelling of structures using
computer programs has become a relatively simple
task. Entering the data into the computer is a fairly
mechanical operation that can be performed without
much thought. It is a sad reflection that there seems
to be a large number of people for whom thought is
a painful activity to be avoided at all costs. So much
easier to just put the data into the machine and let it
do the work. Of course, computer analysis can be used
intelligently if one realizes that large tension forces at
joints are unlikely to occur and the structure needs to
be modified accordingly. Sometimes one even finds
that the result is a statically determinate structure and
with a little thought one could have started from there.
One also needs to recognize that the nature of the
workmanship can affect the support conditions for
members that in turn will affect the overall structural
scheme. Also, the section properties that one can insert
as data for an historic structure can often be little more
than a guess. This is equally true of the behaviour of
connectors, and any analysis needs to be based on a
thorough qualitative grasp of the structure following
close observation of its construction and the way that
it has behaved. Often the uncertainties are such that
more than one structural scheme has to be considered.
Of course, timber is not unique in this regard but per-
haps the linear nature of its elements encourages a false
idea of simplicity.
3 REPAIR STRATEGY
This brings us to the methods of repair, and here we
should recognize that different approaches might be
appropriate for different kinds and periods of structure.
A simple contrast can be drawn, for example, between
the decorative roofs of the middle-ages and the purely
utilitarian structures of nineteenth century industrial
buildings. In spite of both being equally visible, the
former are clearly more valuable architecturally. As for
all historic structures, we should begin by considering
the possible historical value and hence the strategy to
be adopted rather than focusing immediately on the
tactics of repair. This issue of conservation strategy is
one in which the engineering historian has something
to contribute and about which the engineering consul-
tant might be better informed than other members of
the conservation team.
This means making some strategic assessment that
should take into account:
– The historical value of the fabric.
– The overall condition of the structure and hence
– The scale of the repairs required, but also
– The options for future use.
Consider the extremes here. We might try to retain
the timbers in a condition as close as possible to the
original structure, i.e. as a repaired structure carrying
out its original function of supporting the building.
At the other extreme is provision of new structure
to support the original timbers that simply remain
as an historic artefact reminding one of what was
there originally. The latter might well be the preferred
option where the decorative nature of the timbers is of
65
paramount importance. Between these two there may
be a choice of options that the engineer should be able
to assist in deciding between. This requires the to engi-
neer have a greater involvement in the process than
simply providing the technical fix after the strategic
decisions have already been taken.
In considering the overall strategy the choices are
between:
– repairing and restoring the as-found structure and
– providing supplementary structure
And if the first option is chosen one needs to
consider the choice between
– Repairing in-situ and
– Dismantling for repair, either completely or in part.
There is often resistance from conservation offi-
cers (and possibly clients) to the latter, even thought
there are some cases where this might be the cheapest
and safest approach. It is an option for timber struc-
tures because of their prefabricated nature, and there
is almost certainly a failure to understand that where a
structure has distorted over time, for whatever reasons,
to repair it in its as-found condition is to preserve a
cripple. It will often have locked in secondary stresses
that have arisen through the distortion that has taken
place and which now have to be accommodated within
the repairs, a complicating factor.
I have argued (Yeomans, 2007) that in some build-
ings, particularly barns, the structural design is what
we should be seeking to preserve. In contrast, where
the timber has been treated decoratively this will
take precedence and we may have to compromise the
integrity of the structure in order to preserve the sur-
face. Perhaps there are those occasions where the
choice is that simple, but this dichotomy means that
some judgement must often be made. In such circum-
stances the engineer responsible must be prepared to
engage with other conservation professionals and to
explain his point of view. It may be necessary to offer
alternative structural strategies that would have dif-
ferent effects on the historic fabric. One is balancing
structural safety, the heritage value of the structure, and
perhaps a choice of future use. The client and other
conservation professionals will be involved in such
decisions, and the engineer must be prepared to explain
the engineering issues in terms they can understand.
Although I have emphasised the need for con-
servation engineers to work with other conservation
professionals in determining the overall strategy, it is
also advantageous to work with the carpenters. I regard
the simple model of professionals producing a design
for contractors to price as inappropriate for the conser-
vation of timber structures. We are not dealing with an
empty site on which a building is to be assembled by
reasonably tried and tested methods. We are dealing
with unique situations where it is necessary to ensure
the continued stability of an existing structure that is
being changed while it is being repaired. This is a situ-
ation in which the methods used may affect the details
adopted, or in which the practicalities of a particular
detail might affect the scope of the work to be car-
ried out. In many cases an experienced carpenter may
make an invaluable contribution at the design stage
and, where possible, it is worth employing one as part
of the design team.
I raise all these issues before discussing repair meth-
ods in detail because both the repair recommendations
of ICOMOS UK’s Wood Committee and the Recom-
mendations of ISCARSAH, while informed by such
considerations, do not draw the conclusions as clearly
as they might. My view is that more consideration
needs to be given to this stage of the process in the
Figure 1. Repairs using steel carried out to the heel joint of
advice given to professionals. a Wren roof truss at Hampton Court.

4 REPAIR DETAILS

When we turn to the tactics of repair, timber has always

seemed to be a fairly contentious material the methods
of repair have attracted rather strongly held positions.
There has also been a developing of skills and the
two together seem to have resulted in some swings in
approach. (This might be true of conservation in gen-
eral, it just being my own perspective that makes me
think that it is more problematic for timber.) Repair
methods in timber can be loosely divided into three
kinds:
– Those using steel
– Those using modified traditional methods
– Those using epoxy resins with or without some
reinforcement.
The kind of repairs once commonly carried out by
engineers, cutting back the decayed timber and putting
in supplementary steelwork (Fig. 1), was later dep-
recated by those who favoured a more ‘traditional’
approach, although that usually meant carpentry meth-
ods supplemented by modern steel fasteners. It has
more recently been recognized that while such an
approach is more visually attractive, and so appropriate
for medieval timber frames, where the results are both
visible, and have to form a junction with secondary
materials, it does involve a greater loss of historic fab-
ric. In that respect the use of supplementary steelwork
is perhaps to be preferred where the repair is not seen.
Repairs using epoxy resins had a ‘bad press’ for a
while and suffer from a regrettable lack of research
into their long-term performance, but lack of research
also affects other areas of repair work.
One of the major repair projects carried out early
in the 20th century was to the roof of Westminster
Hall (Baines, 1914) where substantial decay had been
found. That used steel connectors, but in locations
Figure 2. A truss in the Banqueting House, Whitehall with
unnecessary reinforcing.
where it would not be visible. One wonders whether
there was general concern at the time for the adequacy
of timber structures simply because they were timber.
The roof of the Banqueting House, Whitehall, built by
Soane in 1824 to replace Inigo Jones’s original, was
a fine combination of timber and iron. Nevertheless,
sometime in the inter-war period it was felt necessary
to add steel to all the joints (Fig. 2) and to insert steel
ties to take over the job that the tie beams seem to
have been doing adequately till then. Clearly the rein-
forcing shown in Figure 2 is having no useful effect.
The steel structure added to the roof over the hall at
Greenwich Hospital also seems to be an unnecessary
supplement to the original structure because in neither
case is there obvious structural distress. But excessive
use of steel can still be seen in use today because it
is the material that engineers are familiar with and so
may instinctively turn to even though timber repairs
would be perfectly satisfactory.

66
 In the 1970s the use of epoxy-resins as either a con- sensible to use it – but this is an argument normally
solidant, or to replace lost timber appeared to offer a applied to non-structural elements.
solution for those who wanted the minimum loss of
historic fabric, but this ran into technical difficulties.
5 FASTENERS AND DESIGN CODES
Early research in the US on this method, where they
were interested in it for repairing softwood trusses in
Once the forces in a structure have been determined,
the roofs of aircraft hangars, showed the difficulty of
and unless there is particular concern over deflections,
controlling the run of what was a liquid and the obvi-
our engineering problems become a series of con-
ous change in the structural characteristics of joints.
nections with members in between; member stresses
However, it is questionable whether or not this work
are generally low. Unfortunately the jointing methods
was noticed by those in Britain who were enthusiastic
used in new structures are not always appropriate for
for the use of the material and there are some unfor-
historic structures. Moreover, historic structures use
tunate results of its early and inexpert use. Baguely
details that are not used in modern carpentry and there-
Hall, on the outskirts of Manchester, is an example of
fore for which there is no guidance within design codes
the kind of thing that can go wrong. Epoxy-resin was
framed for new buildings, using modern materials and
used to repair the main fames of this fine medieval
modern methods. This was never so clear as during
hall but was allowed to run over the surface of timbers
the lacuna when the newly introduced British Standard
and is now a permanent disfigurement. Many early
(BS5268) in 1984 failed to include oak as a structural
repairs using this material were carried out by a firm
material and conservation engineers had to continue
who appeared to have little technical knowledge and
to use the obsolete code (CP112). While that issue has
even less interest in the proper training of their opera-
been resolved, there are still questions over the per-
tives. The result was a number of failures, sometimes
formance of traditional joints and the values that one
because of the use of the technique in unsuitable loca-
might use for metal fasteners.
tions where exposure to the weather seems to have
The British code of practice is even inadequate in
resulted in accelerated decay of timber adjacent to the
the information given for steel fasteners. One example
epoxy-resin repair.
is that safe loads for screws in shear are only given up
The difficulty then was that with considerable anec-
to 10 mm diameter when carpenters may well wish to
dotal evidence about failures following the use of
use 12 mm diameter screws. (Even that is an improve-
epoxy-resin, and with little scientific evidence, there
ment on the earlier editions of the code where loads
was deep concern among some over the long-term per-
were only given up to 8 mm diameter.) An engineer
formance of such repairs. This was especially so where
who is unfamiliar with timber may be unwilling to go
they might be used in timbers with high or fluctuat-
beyond the limits of the table, particularly as the for-
ing moisture content. If there were to be deterioration
mulae for determining the loads from first principles
of the adjacent timber it would have an effect on the
are rather daunting. Bolts can be seen used in circum-
strength of the joint. One might also be suspicious of
stances where screws would have been better but where
its behaviour when used as an adhesive to join steel
M10s were presumably inadequate. Moreover it might
reinforcing plates to timber and the effects of temper-
not be clear to those unfamiliar with the code that the
ature changes. As the steel must move more than the
spacings and edge and end distances for bolts rather
timber what is happening at the interface between the
than screws apply to these larger sized screws.
materials? Here we have no information and can only
There is the serious question about how we should
wait and see.
view the allowable loads on bolts and screws used in
Opinions on the use of the material seem to still
green timbers that will dry in service. The problem for
be divided between those with long memories and an
conservators is complicated by the fact that one side
instinctive distrust of these techniques and those for
of the joint will be dry timber and the other side green.
whom it is a valuable tool. TRADA (2001) have pub-
I have no idea what reasoning or experimental results
lished a book on its use and at the Whitbread Brewery,
lie behind rather draconian reduction factor required
London, Hockley and Dawson used it to effect in the
by the code. If it is to allow for the possibility of split-
repair what may well be the largest surviving eigh-
ting as the timber dries one would have thought that
teenth century trusses. More recently the Weald and
such an event might well reduce the capacity of the
Downland Open Air Museum has been developing
joint to zero.
skills in the use of epoxy resins for repairing struc-
tures that have been dismantled. (Their techniques are
not suitable for in-situ work.) Richard Harris points 6 TRADITIONAL JOINTS
out that this material allows one to make repairs that
retain a lot more of the historic fabric. He has also The terminology is a little loose here because there
argued that if something is going to be lost through are traditional joints, either surviving in existing struc-
decay and epoxy-resin will prolong its life, then it is tures or used in repair work that are no different from

67
Figure 3. In loading to determine shear capacity of timber
the forces are applied at each end.
Figure 4. In atypical heel joint restrain is provided by the
tie beam behind the notch.
joints contemporary with the original construction.
At the same time there are those joints that have a
superficial resemblance to traditional joints but which
rely on modern fasteners, the most common being the
scarf joint. The difficulty is that any tests to deter-
mine working loads for such joints must be affected
by the workmanship in the carpentry. One can envisage
the problem of relating test results to the performance
of historic joints, both with very uncertain standards
of workmanship and the effects of time and chance –
or perhaps that should be wind and weather.
One wonders whether the standard test for shear
resistance is a suitable basis for the design of tradi-
tional carpentry joints relying upon shear along the
grain? The test loads the timber at either end so that
one might expect a sensibly uniform shear stress along
the plane of failure – the dotted line in Figure 3.
The joint where timber is loaded in shear parallel
to the grain is a typical heel joint where the normal
condition in practice is for restraint to be provided by
the tie beam behind a notch (Fig. 4) or a mortice and
tenon joint. In such configurations one would expect
the shear stress along the plane of failure to fall off
towards the unloaded end; but in what way? A com-
plicating factor in some buildings is that this timber
may be rather exposed and have a higher than normal
moisture content.
Pegged joints is one of those areas where there is an
overlap between new-framed construction and conser-
vation, the question being whether such joints can be
relied upon to transmit tensile forces. The difficulty is
that pegs do not behave in the same way as steel bolts or
dowels so that Johansen’s equations cannot be applied
to find the allowable loads. Experimental work shows
that the pegs fail in shear close to the tenon/mortice
68
Figure 5. A pegged mortice and tenon joint. The peg will
fail at the interface between the two timbers.
Figure 6. Forces on a lap-dovetail joint used between tie
beams and wall plates.
boundary without the kind of rotation that occurs in
bolts.
Unfortunately what work has been done so far has
produced conflicting results. Drawing on the work of
Jonathan Shanks carried out at Bath (Shanks & Walker,
2005), Ross et al (2007) have given possible work-
ing loads for pegs in oak, whereas a formula derived
from quite different tests by Schmidt (2004) at the
University of Wyoming suggests a working loads of
less than half their figures. Given such a wide dis-
crepancy, prudence suggests one should use the lower
figure for design, but science surely requires some
explanation for this large difference. The Bath tests
were carried out on complete joints while those in
Wyoming in more ‘traditional’ laboratory experiment
that allowed the effects of varying density in both pegs
and member timbers to be explored. What surprises me
is that the Bath tests that used carpenter-made joints,
and presumably subject to variations of workmanship,
nevertheless resulted in the higher figure.
The behaviour of dovetail joints is another problem-
atic area as these rely on timber loaded across the grain
(Fig. 6). When there is drying shrinkage in the timbers
there is a change in the angles of the two parts of the
joint resulting in a reduced contact area and a corre-
spondingly higher compressive stress.The stress might
be well above the elastic limit with collapse of some
of the cell walls. However there is no readily available
data from which one can estimate the movement that
might result.
A serious problem occurs when existing structures
requiring repair incorporate details that are ‘prohib-
ited’ by the present code. An example occurred where
rafters had birdsmouth cuts over the supporting plate
with the depth of the cuts exceeding half the depth of
Figure 7. Birdsmouthed rafters.
Figure 8. A simple half-lap scarf with bolts in shear.
the rafters (Fig. 7). This is larger than the depth of a
notch allowed by the code but any change would have
affected the geometry of the roof. Admittedly some
had split, but these were in the minority. In the event a
case had to be carefully made to allow their retention
of the sound rafters and the repair of those that had
split.
Scarf joints are commonly used in repairs where
bending moments have to be carried.A simple arrange-
ment is a half lap with bolts or other fasteners acting
in shear (Fig. 7) but a more traditional joint is often
preferred for visual reasons (Fig. 8). If a moment is
applied to this, one end will go into tension and the
preferred fixing is usually one or two coach screws
acting in withdrawal. (Screws are preferred to bolts
so that they are visible from one side only.) However,
there is currently no guidance on the detailing for such
a screw fixing and as carpenters will usually want to
sink the heads below the surface, the small amount of
room available is evident from the drawing.
Carpenters adopt a simple rule of making the length
of the joint about three times the depth of the section.
A few years ago TRADA looked at a number of differ-
ent methods of making such joints deriving measures
of ‘efficiency’ for each, i.e. comparing their resistance
moment with that of solid timber of the same section.
This simple, some might say simplistic, approach to the
assessment of joint behaviour does not tell one how the
69
Figure 9. A more traditional scarf joint with screws in
withdrawal at the tension end of the joint.
joint is behaving in practice, and the fastening meth-
ods were not those that would recommend themselves
to carpenters.
I have left the most difficult issue till last, which is
resistance to wind loads. One might assume that build-
ing that has stood for many years, possibly centuries,
has stood the test of time. While that is not unreason-
able for masonry structures with large masses it is more
problematic with lighter timber structures that might
have been subject to some deterioration, especially
if there have been periods of neglected maintenance.
Moreover we are aware that climate change will result
in higher wind speeds in future, again more of a prob-
lem for timber structures rather than masonry. The
result is that it might be difficult to demonstrate that
an existing structure can cope with the required design
loads. Of course, in attempting to do this one will often
consider the structure acting alone whereas the frames
may have always relied to some extent upon the infill
material. It is, of course, possible to use the resistance
of infill panels where these are of a construction recog-
nised by present design codes, but we know little or
nothing about the resistance provided by historic lath
and plaster or wattle and daub infills. Here we might
consider Japanese work that has explored the resis-
tance of traditional infilling material in the context of
earthquake loading. Whether such an approach can be
extended from large and infrequent loading to more
modest but very frequent loading is a moot point and
one that would need to be considered before embarking
on research in this direction.
7 CONCLUSIONS
This view of the present state of timber conservation
suggests that in many cases there is too little engineer-
ing involvement in determining the overall strategy
of repair. This might be true for other materials as
well but its effects are possibly more serious when
dealing with structures often requiring extensive and
highly visible repairs. This is in part compounded by
a failure by some engineers to come to terms with the
characteristics of this material in their analysis of its
behaviour.At the detailed design stage, there have been
some developments as a result of the present fashion
for green-oak structures. However, work in this area is
still patchy and rather inconclusive. If this seems a little
pessimistic, what I have not discussed is the quality of
the pool of carpentry skills that has been growing over
the last few decades that facilitate good standards of
conservation
REFERENCES
Baines, F. 1914, Westminster Hall, report to the first commis-
sioner of H.M. Works, etc., on the condition of the roof
timbers of Wstminster Hall, with suggestions for main-
taining the stability of the roof, London, HMSO. (Cd.
7436).
Charls, F.W.B. & W. Horn 1973, ‘The cruck-built barn
of Leigh Court, Worcestershire’, J. Soc. Architectural
Historians.
Foot, N.J.D., C.D. Litton & W.G. Simpson 1986, ‘The high
roofs of the east end of Lincoln Cathedral’ in Medieval
Art and Architecture at Lincoln Cathedral (B.A.A. Conf.
Trans., 8), pp. 47–74.
Hewett, C. 1967, ‘The barns at Cressing Temple, Essex’,
J. Soc. Architectural Historians, 26.
70
Hewett, C. 1980, English Historic Carpentry, Philimore,
London.
Hewett, C. 1985, English Cathedral and Monastic carpentry,
Philimore, London.
Horn, W. & F.W.B. Charles 1966, ‘The cruck-built barn
of Middle Littleton, Worcestershire, J. Soc. Architectural
Historians.
Horn, W. & F.W.B. Charles 1984, ‘The cruck-built barn
of Frocester in Gloucestershire’, J. Soc. Architectural
Historians.
Rigold, S.E. 1966, ‘Some major Kentish barns’,Archeaologia
Cantiana, 81, pp. 1–30.
Ross, P., C. Mettam & A. Holloway, 2007, Green Oak in
Construction, TRADA, High Wycombe, p. 163.
Schmidt, R.J. 2004, ‘Considerations for mortise and tenon
joint design’, Wood Design Focus, 14, No. 3, pp. 14–17.
Shanks, J.D. & P. Walker 2005, ‘Experimental performance
of mortice and tenon connections in green oak’, The
Structural Rngineer, (6 September), pp. 40–45.
TRADA 2001, Resin-bonded Repair Systems for Structural
Timber, Timber Research and Development Association,
High Wycombe.
Yeomans, D.T. 1992, The Trussed Roof: its history and
development, Scolar: Aldershot.
Yeomans, D.T. 2007 ‘Appropriate technologies for conserva-
tion’, Journal of Architectural Conservation, 13 No. 3,
pp. 9–20.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

A multidisciplinary approach for the structural restoration of the

Katholikon of Dafni Monastery in Attica Greece

A. Miltiadou-Fezans
Directorate for Technical Research on Restoration, Hellenic Ministry of Culture

ABSTRACT: The Katholikon of the Byzantine monastery of Dafni (world heritage list of UNESCO), is one
of the most important monuments of middle Byzantine period, famous worldwide for its mural mosaics. Being
constructed in a seismic area, the Katholikon was affected throughout its history by a large number of earthquakes
that caused structural problems and damages to its mosaics. This paper focuses first on the historical and recent
pathology studies of the monument and the emergency measures undertaken by the Hellenic Ministry of Culture
after the earthquake of 1999 in Athens. Then, a synthesis of the results of research programs, investigations
and structural restoration studies is presented. Furthermore, a brief presentation is made of the first phase of
interventions executed for the consolidation of the masonry and the control of their efficiency using novel
methodologies, non destructive techniques and monitoring systems.

1 INTRODUCTION

Dafni Monastery, situated at approximately 10 km

from Athens on the way to Eleusis, is one of the
major Byzantine monuments in Greece (Millet 1899,
Stikas 1962, Bouras 1998, Delinikolas et al. 2003).The
Katholikon (main Church) of the Monastery (Figures 1
and 3) has suffered severe damages during the 7th
September 1999 earthquake that affected the region
of Attica.
Immediately after the earthquake, a multidisci-
plinary working group (Delinikolas et al. 1999) was
formed by the Hellenic Ministry of Culture (HMC)
with the assignment to do the necessary inspections,
assess the nature and the significance of damages and
elaborate, together with a Scientific Committee set
to this purpose (composed by Professors Ch. Bouras,
T. P. Tassios, E. Mariolakos and N. Zias) and all com-
petent authorities of the Ministry, a strategic plan for
the protection, conservation and restoration of the
monument, its mosaics included.
Due to the severe damages of the structure and
the danger of eventual aftershocks, the decision was
taken for the application of emergency measures
(Miltiadou-Fezans et al. 2003a). The aim of those
measures was (a) to reduce the danger of fur-
ther deterioration of structural damage and (b) to
ensure accessibility and safe working conditions
for all the scientific and technical personnel, thus
enabling the execution of all the surveys and inves-
tigations, necessary for the design and implemen-
tation of the most adequate structural restoration
interventions.
Figure 1. N-W view of the Katholikon of Dafni Monastery.
In parallel with the design and implementation of
emergency measures, a first series of research pro-
grams were initiated, in order to record the state
of the monument immediately after the earthquake.
These programs were aiming at: (i) accurate survey of
geometry (Georgopoulos et al. 2003), (ii) preliminary
survey of damages, assessment of mechanical proper-
ties of construction materials, monitoring of the width
evolution of main cracks, and preparation of a first
detailed computer model of the monument used for
preliminary structural analyses (Vintzileou, 2002) and
(iii) assessment of the physicochemical characteristics
of the construction and pointing mortars (Papagianni
2002).
The above research works were carried out in
close collaboration with the responsible multidisci-
plinary team nominated by the competent authorities

71
of the HMC. On the basis of a synthesis of the
obtained results, adequately completed with detailed
in situ observations and archive researches, the mul-
tidisciplinary group of the HMC has elaborated the
structural restoration study (Delinikolas et al., 2003,
Miltiadou-Fezans et al. 2003b). Following the propos-
als of this study, approved by the Scientific Committee
and the Central Archaeological Council, it was decided
to implement the structural restoration design and
interventions in two phases.
Given the importance of the monument, a step
by step multidisciplinary approach was adopted, both
concerning the design and the implementation of the
structural restoration interventions. Such an approach
gives the possibility to start the execution of a series Figure 2. N-E view of Katholikon.
of interventions, and in the same time to perform the
in situ and laboratory investigations that are necessary
for the design of the next step. Furthermore, the assess-
ment of the effect of the first step interventions to the
structural behaviour of the monument can be carried
out and taken into account for the design of the works
of the second step.The first phase of interventions (that
is completed) comprises measures taken to repair and
strengthen masonry elements. The interventions of the
second phase aim to improve the overall behaviour of
the entire structure; their design is still in progress.
Moreover, within the HMC, it was judged as abso-
lutely necessary to support the two phases of works
with further research and investigations, comprising
the design of adequate mortars and grouts, the exper-
imental estimation of mechanical characteristics of
masonry before and after grouting, the seismic moni-
toring of the structure, the application of NDT’s for Figure 3. The central dome and its drum, curried by eight
investigating invisible parts of the monument and pendentives and eight arches.
controlling the grouting effect and efficiency.
In this paper, a brief synthesis of all the aforemen-
cross-shaped in plan, the hemispherical dome rising
tioned investigations, studies, emergency and struc-
over its square core. The dome is 8,2 m in diameter
tural restoration works will be presented.
and 16,4 m high, and rests on an almost cylindrical
drum with 16 piers and 16 vaulted windows. The
2 BRIEF DESCRIPTION OF THE dome and its drum are carried by eight pendentives
KATHOLIKON OF DAFNI MONASTERY and eight arches (four semicircular and four embodied
in the squinches of the corners), forming an octagon
The Byzantine monastery of Dafni comprises various and achieving in this way the transition from circle
buildings (laid out in a square plan), constructed over to square. Thus, twelve piers (laying out in a square
a long period of time, starting on the 11th century AD plan), provide support to the dome together with the
(Delinikolas et al. 2003). Currently, most of the build- groin vaulted arms of the cross, situated in a higher
ings are in ruins, with the exception of the Katholikon, level (Figs 3 & 4). All the other parts of the monument
part of the internal range of cells, the cistern and the are covered with byzantine groin vaults.
northern fortification walls. As one can clearly distinguish in Figures 1 and 3, the
The Katholikon belongs to the octagonal type and exterior face of the vertical perimeter walls are built
preserves large part of the original mural mosaics. It according to the enclosed brick system (stones with
comprises the main church, the sanctuary, the narthex bricks around them), following two different masonry
and four chapels, which complete its orthogonal plan. construction types. In fact, there is a lower zone, built
In the western part, only the perimeter walls of an with large dimension stones (often, reused material),
exonarthex or portico and those of a spiral stair- their length placed horizontally or vertically, in order to
way tower leading to the upper floor have survived form crosses, and un upper zone constructed with ash-
(Figs 1, 2, 5). The central part of the main church is lar masonry, using smaller and perfectly cut orthogonal

72
stones (Delinikolas et al. 2003). Bricks are placed in all
horizontal joints, and in most of the vertical ones. The
thickness of mortar joints is of 3 cm approximately.
Plain brick masonry was used in the construction of
all the windows and doors (Fig. 2). In the internal face
of the perimeter walls, as well as in internal masonry
elements in general, the enclosed brick system is not
followed systematically, whereas in various locations,
a mixture of cut, semi-cut and rubble stones are used
together with bricks. Thus, the thickness of masonry
joints is varying. Finally the entire vaulted roof was
constructed with plain brick masonry.
Both the lower and upper parts of the perimeter
walls are constructed following the three-leaf masonry
type, with varying widths of the leaves, as will be more Figure 4. Plan of the Katholikon and schematic presentation
analytically presented below (Section 9). of the arrangement of the raking shores and vertical props.

3 CONCEPT AND BRIEF PRESENTATION

OF EMERGENCY INTERVENTIONS

The emergency interventions were designed taking

into account specific demands deriving from the
importance of the monument and the necessity for
implementation of final restoration works without
removing the supports and scaffoldings. Thus, they
had to be reversible, easily assembled and allow-
ing for gradual disassembling in the interior, as well
as adjustable to the deformed geometry of damaged
elements. Moreover, any contact with the vulnerable
mural mosaics was to be avoided.
To this end various alternative solutions were exam- Figure 5. Raking shores installed in the NE corner.
ined (Miltiadou-Fezans et al. 2003a). Figure 4 shows
schematically the retained one. In the NE corner of
the building, three double-framed steel raking shores
was constructed, as in this area a pronounced tilting
of the external walls has been noticed and the telltales
installed just after the earthquake, indicated further
opening of cracks and a tendency of the corner to
detach (Fig. 5). In the interior, and in the exonarthex,
vertical steel props were built beneath the main arches,
in order to provide vertical support to their cracked
structure (Fig. 6).
Both the raking shores and the vertical props were
self supported elements, which would be activated Figure 6. Vertical steel props installed in the interior of the
only in case of further increase of the deformations church, and in the exonarthex.
of the structure. Thus, between the metal framework
and masonry walls, a 12 cm full layer of wooden taking special care to assure the in situ assemblage
beams and wedges (together with a 3 mm soft pack- of all these structures without harming the mural
ing) were inserted, to provide good contact with the mosaics. Furthermore, the following measures have
masonry without harming the surfaces, while allow- been undertaken: (a) the installation of adequate types
ing relative movement, unless the structure did start of scaffoldings in the interior and exterior of the
to move towards them. Moreover, the upper part of church, in order to offer safe working conditions for the
the exonathex walls and the NW and SW piers were personnel, and (b) the removal of the tiles of the roof
confined using steel plates and bars. and the application of temporary water isolation mem-
Regarding the drum of the dome, specially designed branes just below them, to protect the cracked extrados
steel elements were constructed to brace the windows of the vaulted structures, and hence the mural mosaics
and confine in two levels the masonry piers (Fig. 7), from leaking water.

73
 Figure 8. West and south façade: 1st construction phase,
drawings of Benouville as modified by Delinikolas (Millet
1899, Delinikolas et al. 2003).

Figure 7. Emergency measures applied to the dome’s drum.

4 HISTORICAL PATHOLOGY AND PAST

INTERVENTIONS

The Monastery is situated in a Neocene tectonic

graben between the mountains Egaleo and Korydallos
at the west side of Athens, 150 m away from the
Figure 9. West and south façade: 2nd construction phase,
E-W trending marginal fault between the alpine graphical representation of Stikas as modified by Delinikolas
Mesozoic limestone and the post-alpine deposits (Stikas 1962, Delinikolas et al. 2003).
(Mariolakos et al. 2000). Located in a tectonically
active area, many intensive earthquakes damaged
Dafni Monastery throughout the centuries (from 11th
c. to our era). Partial collapses and extended damage
were provoked to the structure and the mural mosaics,
which led to major interventions. Detailed presenta-
tion of the long history of the monument is reported in
Delinikolas et al. 2003, whereas a concise summary is
given in this paragraph.
Two main phases of construction have been identi-
fied in the Katholikon of Dafni Monastery, both dating
back to the middle Byzantine period. Although the
construction type of masonry is similar in the two
phases, the composition and final texture of point-
ing mortar used is different, thus allowing for the two
phases to be identified in situ. The first phase com- Figure 10. Graphical representation of the W façade of
prises the main church and the narthex (Fig. 8). A the portico as restored by the Cistercians (Delinikolas et al.
reddish mortar containing crashed bricks was used for 2003).
the pointing of the joints of the monument’s external
façades. At a second phase, a portico or exonarthex of the exonarthex have probably collapsed, provok-
was added to the church (in its west part); masonry ing severe damage to the rest of the structure. The
piers in the corners and marble ionic columns in the narthex, the dome and its supporting system were as
middle were used to support the arches of the added well severely damaged. Thus, extensive interventions
perimeter walls. Over the portico and the west part of were undertaken by the Cistercian monks mainly in the
the church a first floor was also added. Finally a spiral west part of the monument (13th–15th c.). The upper
staircase tower was constructed in the NW corner of part of the west wall and the majority of arches, were
the narthex (Fig. 9). A whitish mortar with a yellowish reconstructed with curved stones following the gothic
patina was used for the pointing of the joints of the style (Figure 10). Furthermore the exonarthex was cov-
external façades. ered with a timber roof and bastions were built on the
Severe damage was provoked to the church during top of the walls. The remains of the constructions of the
the 13th century, when Athens was occupied by the first floor, the narthex and the rest of the church were
Franks. The first floor and the three groined vaults also locally repaired. These extended interventions in

74
Figure 11. West façade of the exonarthex before and after
Stikas restoration interventions (photos Stikas, 1962).
the west part of the Katholikon can be considered as
the third construction phase of the building.
During Othoman occupation strong earthquakes
provoked further deterioration of the damages of the
entire church. Cracks appeared on the vaulted roof
of the crypt and the narthex, whereas severe dam-
age and partial collapse occurred to the exonarthex
and the spiral staircase. Thus, the arches of the
exonarthex perimeter walls were filled with masonry,
and extended alterations took place in the west part of
the monument. During that period some of the ionic
columns were taken by Lord Elgin to England.
The structural condition of the monument was
further deteriorated during the 19th and the 20th cen-
turies, due to numerous earthquakes occurred and
major interventions were applied for the preservation
of the monument. One should mention the 1889 and
1894 earthquakes (estimated magnitudes 6,7 and 7,0
on the Richter scale), the 1914 earthquake (M 6,0R),
the 1981 earthquake (M 6,7R) and finally the 1999
one (M 5,9R). After the 1889 earthquake, the heav-
ily damaged dome and its drum were removed and
reconstructed. Three concentric iron rings (I-beams)
were inserted at the base of the drum of the dome, as
identified during the recent restoration works. Before
removing the old dome, the mosaics were detached
and, after their conservation, partially re-attached to
the monument by Italian Conservators. During these
interventions, the lime based mortar of the mosaics
substrata was replaced with a hydraulic lime based one.
The 1894 strong earthquake caused further damages;
even the recently reconstructed dome and its drum
were once again cracked. Extended interventions took
place to the west part of the monument, including the
reconstructions of almost the entire narthex (Troump,
1896). In the period of 1897 to 1907, two stone but-
tresses both sides of the north entrance to the church
were constructed, while metallic trusses (fixed on new
foundation) were installed both sides of the south cen-
tral piers, together with iron plates and bars inserted at
three levels, in order to confine them. In 1914, the sug-
gestion is made for external confinement of the central
cupola with an iron ring. This solution, however, was
applied 40 years later.
In that period of 1955–1960 the ottoman interven-
tions were removed from the exonarthex (Fig. 11)
75
Figure 12. Interventions to the monument: Plan.
and the Cistercian construction phase was restored by
Stikas (Stikas 1962). The collapsed part was rebuilt
and the missing ionic columns were replaced by
masonry piers. A concrete beam was installed in the
interior of the reconstructed parts, without affecting
the byzantine remains of the SW pier. The upper part of
the SE chapel was also reconstructed, while extended
local re-pointing took place in various areas of the
Katholikon, using mainly cement based mortars. For-
tunately the non deteriorated old mortars were not
removed. Thus, they survived to our days in better
condition than those used by Stikas, which presented
extended decay and during the recent restoration works
were removed.
Figures 12 to 14, taken from Delinikolas et al. 2003,
present the historical pathology and past interventions
on drawings. The reconstructed parts of the monu-
ment are shown with solid colors.Yellow: 13th–15th c.,
orange: 1891, green: 1895, purple: 1897–1907, dark
blue: 1955–60.
The 1981 earthquake (Alkyonides islands) caused
numerous hair cracks to the building and severe dam-
ages to the mosaics. Thus, the pathology observed
in the monument after the 1999 earthquake was only
partly due to this severe earthquake.
5 SURVEY OF DAMAGES AFTER THE 1999
EARTHQUAKE
The systematic survey of damages after the September
7th, 1999 earthquake, reported in Miltiadou-Fezans
et al. 2003b and 2004, is summarized in this sec-
tion. In Figure 15 some typical drawings of cracks
and deformations survey are presented, showing the
severe damages observed in the monument (both to its
structural part and to the mosaics).
 (a) East-West section. View to South

Figure 13. Interventions to the monument: S and N façade.

(b) East-West section. View to North

Figure 14. Interventions to the monument: West façade of

(a) the church and (b) the exonarthex.

An extensive network of shear and bending cracks

(c) North-South section. View to West
(ranging from hair cracks to those several centime-
ters wide) has appeared on the walls and piers of
the monument, whereas numerous old cracks (due to
previous earthquakes) increased in length and width.
Severe structural dislocation and outwards movement
of the walls was recorded in the NE corner of the main
church (∼14 cm to the N and ∼10 cm to the E). Signif-
icant out-of-plane displacement of the N and S arms
of the cross (∼16 cm and 21 cm respectively), and of
the free standing west wall of the exonarthex were
also recorded, due to further deterioration of previ-
ous deformations (∼16 cm in the corners and ∼25 cm
in the middle).
The damages were more extensive in the higher
parts of the structure, especially in the sanctuary, the (d) North-South section. View to East
arms of the cross and all the arches below the dome
area. As shown in Figure 15, the NE and NW small Figure 15. Typical presentation of damages.

76
arches just below the squinches presented severe dis-
location near their crown, followed by out of plane
deformations of the squinches themselves. Cracks
appeared also in all the groin vaults of the church.
The structural condition of the dome (reconstructed
at 1891 and damaged soon after its reconstruction at
1894), was assessed as extremely critical immediately
after the earthquake (Delinikolas et al. 1999, Miltiadou
et al. 2003a, 2003b). Horizontal cracks have appeared
along the perimeter of the drum (both at its base and
top). In the piers of the drum that are situated perpen-
dicular to the East-West direction, horizontal cracks
(due to out-of-plane bending) have opened at their top
and bottom. In the piers that are situated parallel to the
E-W axis, diagonal or bi-diagonal (shear) cracks have
appeared. In the intermediate piers, mixed type of (less
severe) cracks was observed. This pathological image
does seem to confirm seismological data regarding the
predominant direction of the 1999 earthquake. Thanks
to the external upper metal tie-rod, confining the hemi-
spherical dome near the springing level, the occurrence
of severe cracking of the dome’s shell was prevented.
6 QUALITATIVE INTERPRETATION OF
DAMAGES
As reported in Miltiadou-Fezans et al. 2004, it was
observed that the number and the opening of crack in
the vertical elements of the Katholikon increase from
the base to the top of the monument. It was there-
fore obvious that the monument exhibits the tendency
to “open” from the base to the top along both main
directions. This deformed shape of the church, verified
by the photogrammetric survey (Georgopoulos et al.
2003), is confirmed also by the history of the monu-
ment. The South façade of the narthex reached in 1894
a total out-of-plane deformation larger than 200 mm
and it was reconstructed (Fig. 13,14). The photo in
Figure 16 shows the region between the original and
the reconstructed part of that façade, where one may
clearly see the permanent out-of-plane deformation of
the original part.
However, even this reconstructed part of the mon-
ument presents today a total out-of plane deformation
of 90 mm (drawing in Fig. 16). This is another element
proving that the feature we observe now in the mon-
ument is an inherent characteristic (due to its initial
construction scheme and the extended alterations/
interventions undertaken during its lifetime).
Thus, out of plane deformations were reported both
for the perimeter walls and for all the main arches
bearing the dome in the central area of the monument,
followed by a geometrical deformation (and loss of
initial shape) of the arches themselves. The lack of
wooden or metallic ties (typical structural elements
for the Byzantine architecture) or other horizontal ele-
ments connecting the vertical walls, pillars and piers
77
Figure 16. N-S transversal section in the narthex. The south
wall, although reconstructed in 1896 (see photo), presents
today a 90 mm out-of-plane deformation.
should have played an important role towards this
pathology.
Moreover, the increase of crack openings with
height was found more pronounced along the transver-
sal axis (N–S), than along the longitudinal one (E–W).
Such a difference in behaviour is usual in churches
with an orthogonal plan and could be attributed to
the larger number and sections of vertical elements
available along the E–W axis in the main church (see
Figure 12). This behaviour was also noticed in the past,
whereas previous interventions were applied with the
aim to alleviate this problem (external stone buttresses
in the north, metallic trusses and confinement of piers
in the south), as described in Section 4 (Figs 13 &14).
Although those corrective measures were in the right
direction (allowing the church to withstand the 20th
century earthquakes without local collapse), they were
proven to be insufficient to prevent extensive cracking
of the monument.
The damages observed in the drum of the cupola
may, therefore, be attributed to the (increasing with
height) tendency for out-of-plane deformations of the
church. It should be reminded here that the damages
that made imperative the demolition and reconstruc-
tion of the cupola at the end of the 19th century were of
the same nature, as those observed now; this is proved
by the missing parts of the mosaics. As described
in Section 3, most of the piers in the drum exhib-
ited out-of-plane deformations. Since the substructure
on which the system of the cupola rests is deforming
out-of-plane and the cupola itself (being very stiff) is
practically non deforming, the piers of the drum (being
rather flexible out-of-plane) are called to follow the
deformations of the substructure.
Some final comments should be made here, regard-
ing the foundation and foundation soil of the mon-
ument and their effect on its behaviour. The recent
pathology does not give signs of major differential set-
tlements that might have contributed to the damages of
the monument. This is also confirmed by geotechnical
investigations conducted immediately after the 1999
earthquake (O.T.M., 2000). On the other hand during
the application of urgent measures local investigations
revealed that rather shallow foundations (0.5 m deep
in the middle of South perimeter wall, 1.25 m deep in
the N–E chapel) are provided to the perimeter walls).
Thus, in order to collect qualitative information on
foundation elements and on eventual empty under-
ground spaces or local variations of the ground
material characteristics, geophysical investigation was
also undertaken, using mainly 3D seismic tomog-
raphy (Polymenacos et al. 2005). This investigation
suggested that, in the north-northwest part of the
church and exonarthex, inferior ground material qual-
ity has to be expected, as low velocities were recorded.
Moreover, archaeological excavation in the region of Figure 17. The model prepared for parameter analyses.
the N–W chapel revealed that Troump had probably
noticed that the foundation were relatively shallow and
has dictated their strengthening during the extended
reconstruction of the upper part of the north wall in
1896. For this purpose, a 2.0 m deep wall was con-
structed under the NW perimeter wall of the church. To
this end extensive excavations had to be undertaken in
that area, thus disturbing the foundation soil. This fact
could probably explain the lower velocities found with
the seismic tomography. Nevertheless, all the above
information and eventual further investigations con-
cerning the foundation conditions of the church and
exonatrhex have to be further examined during the
design of the interventions of the second phase.

7 NUMERICAL VERIFICATION OF THE Figure 18. Inner face of shell elements. Tensile stresses due
PATHOLOGICAL IMAGE to self-weight of the structure.

One of the major steps of the assessment of an existing
structure is the analytical reproduction of its patholog-
ical image. In the case of the Katholikon, preliminary
linear parameter analyses were performed (Miltiadou-
Fezans et al. 2004), as a means for selecting adequate
emergency interventions. Those analyses provided a
quite satisfactory numerical verification of the patho-
logical image of the monument. For the preliminary
analytical study, using the computer code ACORD,
the structure was modeled by shell elements (Fig. 17),
whereas the mechanical properties of elements belong-
ing to various parts of the structures were assumed
on the basis of the available data for the construc-
tion materials. Linear elastic analyses were performed
for various combinations of actions (self weight alone
or combined with seismic action). Both static and
dynamic analyses were performed.
Figure 18 shows the calculated stresses for the inner
face of shell elements, due to vertical loads. Irrespec-
tively of the accuracy of numerical values of stresses,
one may clearly distinguish the vulnerability of the
region of arches and domes (especially in the west
part), even for the self-weight of the monument alone.
As expected, tensile stresses are developed in the apex
of several arches, in the groin vaults of the narthex, in
the base of the cupola and that of its drum, as well as in
the four squinches. The obtained results are conform
with the qualitative interpretation of the pathology of
the monument and they prove the inherent vulnerabil-
ity of the structural system, in which (a) a stiff central
cupola is resting (through the drum) on four major
arches parallel to the two main axes of the church
(see Figures 3, 4), as well as on four arches oblique in
respect with the longitudinal and the transverse axis,
(b) the vertical, as well as the horizontal component
of the self weight of the whole system of (intersect-
ing) groin vaults arranged around the central dome,
are transferred to rather flexible stone masonry piers,
and (c) there are not currently ties in the arches and
vaults, or other elements to confine critical structural
elements or link the various parts among them. There-
fore, the tendency of the structural system to deform
laterally in its upper region is expected to be apparent
even under self-weight alone.
This working hypothesis seems to be confirmed by
the deformed shape of the structure in-plan, shown
in Figure 19. One can even observe the more pro-
nounced lateral deformation along the N–S axis of the

78

Figure 19. Deformed shape of the monument in-plan under
self-weight.
Figure 20. Deformed shape of the structure looking from
the N. Dynamic analysis; seismic action along the longitudi-
nal axis.
monument, as well as the out of plane deformation
of the perimeter walls of the narthex, which histor-
ically has been proved to be a vulnerable part of the
structure, already reconstructed in 1896 and deformed
again as measured after the 1999 earthquake. On the
same Figure, the excessive out-of-plane deformation
of the exonarhtex is also apparent.
Expectedly, this behaviour is deteriorating when a
seismic event occurs. In Figure 20, the out of phase
movement of the east and west parts of the monument
is shown. Such a movement can explain the severe
damages occurred to the drum of the dome, as well as
to the arches and vaults supporting the dome. In gen-
eral, the analyses for loading combinations including
the seismic action have shown a critical concentration
of tensile stresses in arches at various levels, as well
as in the piers of the drum. In addition, extensive dam-
ages in vertical elements (masonry in the perimeter of
the monument, as well as piers) were confirmed.
Furthermore, in the framework of this preliminary
work, the plots of principal tensile stresses were com-
pared with the respective drawings on which observed
cracks were reported (Miltiadou et al. 2004).This com-
parison proved to be quite satisfactory, as, in general,
the observed crack pattern (location and inclination of
cracks) seems to be confirmed by the analytical results
in all regions of the monument.
79
Although this preliminary analytical work allowed
for a better understanding of the structural behaviour
of this important monument, it was judged that a
more accurate assessment of its seismic behaviour was
of primordial importance for the design of optimum
interventions.
8 STRUCTURAL RESTORATION SCHEME
The pathological image of the monument (both the his-
torical and the current ones), as well as the fact that
the Katholikon is built in a tectonically active area,
suggest that in addition to interventions that are nec-
essary for the repair of observed damages, adequate
measures should be taken with the aim to alleviate
the inherent vulnerability of the structure and, thus, to
improve its future behaviour in some critical regions.
Due to the high values of the monument, its vulnerabil-
ity and the fact that interventions should not drastically
alter the initial structural system, the decision was
taken to investigate thoroughly its structural behaviour.
To this end additional data were necessary in order to
avoid conservative assumptions (both for actions and
resistances). Such assumptions would lead to exten-
sive interventions that might not be needed and that
would inevitably alter the architectural value of the
monument.
On the basis of the aforementioned reasoning, it was
decided to implement the structural restoration works
in two phases, thus giving the possibility for these addi-
tional data to be collected. The first phase of works
comprises all those considered necessary to achieve
the better possible repair and strengthening of masonry
elements (mainly stitching and deep re-pointing where
necessary, systematic grouting injections, local recon-
structions, etc). The second phase concerns the various
strengthening interventions that will be designed and
selected as optimum, in order to improve the overall
behaviour of the whole building (such as installation
of ties, diaphragmatic structures in the extrados of the
vaults and the exonarthex, etc). Due to the type and
extend of damages, it was judged as absolutely nec-
essary the first phase of interventions to be applied
in priority, in order to avoid further deterioration of
damages, which could lead to a total disruption of
continuity and even local collapses, in case of fur-
ther seismic events. Besides, during this first phase
of works a better structural survey of invisible parts
of the monument (internal face of masonry elements,
extrados of vaults) could be possible.
The implementation of the first phase of structural
restoration interventions has been now accomplished,
together with the most of the research and investi-
gations undertaken to support both phases of works,
briefly presented below. Preliminary proposals for the
second phase of interventions have also been approved,
but their final design is still under elaboration.
9 DESIGN OF GROUTS AND INVESTIGATION Table 1. Mechanical and injectability characteristics of the
OF MASONRY BEHAVIOUR BEFORE AND grouts selected to be injected in the wallettes.
AFTER GROUTING
Grout NHL5-based
9.1 Design of alternative grout compositions and properties Ternary (1) grout (2)
tests on cylinders simulating the infill material
T36 (sec) 19 22.5
The design of high injectability grouts was carried out Sand column
on the basis of the methodology proposed by Miltiadou 1.25/2.50 mm
& Tassios 2006. First, were taken into account the (Wnom ∼.0.2 mm)
Bleeding 2% 3%
performance requirements deriving from the struc-
App.viscosity 20.5 22
tural restoration study (Miltiadou et al. 2003b), and td=4.7 (sec)
then the following target values were set for the basic
mechanical properties of the grouted masonry: ten- Age (days)
sile strength approximately double than that one of the
masonry before grouting, and compressive strength Compressive and 28 90 180
approximately equal to 3.0 MPa. flexural strength
On the basis of the available literature (Vintzileou (MPa) fgc fgt fgc fgt fgc fgt
et al., 1995 and Tassios, 2004), it was estimated that
the compressive strength of the grout at the age of Ternary 4.08 2.11 8.16 2.29 10.6 3.13
six months should lie between 6 MPa and 10 MPa; a NHL5-based 2.82 2.47 4.50 2.52 6.36 3.87
grout
grout flexural strength of the order of 2 to 3 MPa was
required. In addition, the physical-chemical properties (1) Ternary: 30% white Danish cement, 25% Lime (pow-
of the raw materials should be selected in a way that der), 45% natural pozzolan Petr. (<45 µ m), superplas-
the durability of the structure and its precious mosaics ticer 1%, water 80% of the solid phase of the grout
would not be jeopardized. Finally, the grouts should (2) NHL5-based grout: 100% NHL5 (St Astier), superplas-
have high injectability capacity, so that, under low ticer 1%, water 80% of the solid phase of the grout
pressure (∼ 0.075 MPa), they enter and fill fine voids
and cracks, with a nominal minimum width (Wnom )
equal to two tenths of millimeter. Two main categories strength and durability requirements (Table 1). There-
of grouts could satisfy injectability, strength and dura- fore, they were selected to be applied to six three-leaf
bility requirements: (i) ternary grouts composed of stone wallettes, simulating the masonry of the upper
lime, pozzolan and a low cement content (30%) and parts of the monument. It has to be noted however
(ii) natural hydraulic lime – based grouts. that the hydraulic lime based grout presented a better
Thus, various grout mixtures, belonging to the sulphate resistance than the ternary one (Miltiadou –
above two categories, were designed and tested (in Fezans et al. 2007).
order to assess their physical, chemical and mechan-
ical properties) at the laboratory of the Directorate
9.2 Construction and testing of wallettes
for Technical Research on Restoration of the Hellenic
Ministry of Culture (DTRR/HMC). The main results In order to estimate in the best possible way the
of the research are reported in Kalagri et al. (2007) and mechanical characteristics of the masonry before and
Miltiadou et al. (2007). after grouting, considerable effort was devoted to the
In order to determine injectability characteristics, identification of the construction type of masonry.
the penetrability, fluidity and stability of the sus- For this purpose, radar and endoscopy were applied
pensions were fully examined in various water/solids in a systematic way. The in-depth geometry of the
ratios, with or without superplastisizer. The com- perimeter stone masonry was rather accurately identi-
positions presenting satisfactory injectability capac- fied (Vintzileou et al., 2004). As anticipated, both the
ity were further tested to evaluate their behavior to lower and upper parts of the perimeter walls belong
salt decay and estimate their mechanical characteris- to the three-leaf masonry type, with varying widths of
tics (compressive and flexural strength). Furthermore, the leaves. Although both regions were examined, the
six alternative grout formulations presenting similar investigation was concentrated mainly on the upper
injectability were injected into twenty eight cylindrical (more vulnerable) zone of masonry. In Figure 21 char-
specimens, simulating the infill material of three-leaf acteristic examples of vertical sections of the lower
stone masonry. The cylinders were then subjected to and upper zone of perimeter masonry are shown.
compression in different hardening ages (Kalagri et al, The materials characteristics (stones, bricks and
2007). After comparative evaluation of the results, two mortars) being determined in the Laboratory on sam-
grout compositions (a ternary grout and a NHL5-based ples taken from the monument and the type of masonry
grout) fulfilled simultaneously the injectability, the being determined in-situ, an attempt was made to

80
 Table 2. Basic mechanical properties of Wallettes before
and after grouting.

(MPa) (GPa)
fws / εv0 εvs Es /
W fw0 fws fw0 (‰) (‰) E0 Es E0
∗
1 1.82 3.00 1.65 −1.76 1.00 1.20 1.20
2 1.74 3.75 2.16 −1.6 −2.50 1.44 1.55 1.08
3 2.26 3.73 1.65 −2.25 −3.39 1.50 1.30 0.87
ft,0 ft,s ft,s /ft,0
4 0.1 0.21 2.10
5 0.1 0.33 3.30
6 0.1 0.22 2.20
Figure 21. Vertical sections of lower (a) and upper zone (b). W2 and W5 injected with the ternary grout; W1, W3, W4,
W6 injected with the hydraulic lime based grout.

again in compression or in diagonal compression to

Figure 22. The walletes prepared for grouting.
estimate its compressive strength using empirical
formulae available in the literature. The calculated
strength values were unacceptably scattered, thus mak-
ing unsafe or over-conservative the selection of any of
the calculated compressive strength values (Vintzileou
2002). Thus, the decision was made to construct
wallettes and to determine basic mechanical properties
by testing them before and after grouting (Vintzileou
et al. 2006, Miltiadou et al. 2006). The geometry of
the wallettes was chosen to simulate the upper and
more vulnerable part of perimeter masonry. In order
to avoid scale effects a scale 2:3 was selected. Six
three-leaf stone masonry wallettes were constructed
using materials of similar characteristics as the in-situ
ones. Half of the wallettes were tested in compression
and the other half in diagonal compression up to their
maximum resistance. After unloading, the wallettes
were grouted using one of the selected grouts of the
Table 1 (Fig. 22). The injections were carried out using
a specific methodology and the total grout consump-
tion (calculated as the ratio of grout volume consumed
for each wallette per total volume of wallette) was of
the order of ∼10%. Subsequently, they were tested
failure. The main results are summarized in Table 2.
Both grouts applied to the specimens were able to
achieve homogenisation of masonry by filling cracks
and voids of wallettes. Thus, the wallettes exhibited
substantial improvement of their behaviour, in terms
of compressive strength, tensile strength and reduc-
tion of the separation between the three leaves of
masonry, without substantial increase in their stiffness,
and proved to be efficient from the mechanical point of
view. From the two alternative compositions, the nat-
ural hydraulic lime based grout was selected for the
application to the Katholikon of Dafni Monastery, due
to the substantial (compressive and tensile) strength
enhancement of wallettes, the rather ductile behaviour
under diagonal compression (compared to that of
masonry grouted with the ternary grout), and the bet-
ter durability properties (Miltiadou et al. 2007) that
contribute to the protection of mosaics and frescoes.
9.3 Optimum grout composition
In order to further improve the hydraulic lime
based grout, the addition of fine natural pozzolan
(dmax < 45 µm) in various proportions was investi-
gated. The addition of a small percentage of pozzolan
(10%) was decided, on the basis of additional data,
deriving from porosity measurements, salt durability
tests and from in situ pilot trials.
The mix proportions and the injectability character-
istics of the optimum grout composition are presented
in Table 3, whereas more detailed information is given
in Miltiadou-Fezans et al. 2007, 2008. This optimum
grout composition was used for the repair of the mon-
ument. Moreover it was used for the repair of a large
scale model of a Byzantine groin vaulted structure,
constructed and tested on a earthquake simulator, as
briefly presented below.

81
Table 3. Optimum grout composition and injectability char-
acteristics measured in the laboratory and in situ at the first
pilot preparation.

Grout composition
NHL5 (St Astier) 90%
Pozzolan Petrotechniki 10%
Superplasticizer (1), (2) 1%
Water (1) 80%
Grout properties In lab In situ
T36 (sec) – Sand column 19–22
1.25/2.50 mm
(Wnom ∼ .0.2 mm)
Bleeding <1% 1%
Apparent viscosity − td=4.7 (sec) In lab In situ
0 min after mixing 21 22
60 min after mixing (agitated) 23 25
Apparent density (gr/cm3 ) In lab In situ Figure 24. The groin vaulted structure after grouting.
0 min after mixing 1.5050 1.4978
60 min after mixing (agitated) 1.4986 1.4870
It has to be noted that the groin vault was built with-
(1) % of the solid phase of the grout. out any formwork, following the traditional way of
(2) superplasticizer based on polycarboxylic ether. byzantine masons, as reported by Delinikolas et al.
2003. The total dimensions of the model are in plan
2.70 m × 2.60 m; the thickness and the height of the
walls are 0.45 m and 2.60 m respectively, whereas the
thickness of the vault in its centre is 0.20 m. The total
height of the model is 2.85 m. These dimensions were
selected taking into account the limitations imposed
by the capacity of the seismic simulator.
The model was tested after nine months from its
construction by imposing seismic loads gradually
increasing, until rupture. Then local application of
grouts and installation of ties to the arches was per-
formed and after a suitable period of time the tests
were repeated, until rupture. The model was then
injected with the final grout composition (Table 3) to
Figure 23. The groin vaulted structure on the seismic homogenize the whole structure, following the same
simulator. methodology with that used for the injection of the
wallettes and the injection of the monument itself
(Miltiadou-Fezans et al. 2006, 2008). Tests were again
9.4 Construction and testing of a large scale repeated until rupture. In Figure 24 the model after
structure covered with a byzantine groin vault the grouting application is shown. As expected, it was
As aforementioned in section 2, byzantine groin-vaults observed that the dynamic characteristics of the model
are used for covering the most of the parts of the were changed after the application of grouting to the
Katholikon of Dafni Monastery, including the arms of whole structure, and the model could suffer stronger
the cross, where severe damages have been occurred. base motions. The results of this experimentation are
Thus, important effort has been undertaken for achiev- still under elaboration and are going to be presented in
ing a better knowledge of the dynamic behaviour of a separate paper. An attempt will be made to correlate
such structures. An experimental research was carried these results with those of the wallettes and the in situ
out by DTRR/HMC in collaboration with the Labo- seismic monitoring.
ratory of Earthquake Engineering of NTUA. A model
of a byzantine groin-vaulted structure bearing locally
mural mosaics was constructed. The materials and 10 BRIEF PRESENTATION OF THE WORKS
the construction type used for its masonry walls were OF THE FIRST PHASE
exactly the same with those used for the construction
of the wallettes, whereas bricks and mortar were used As reported in the relevant study, (Mitiadou-Fezans
for the construction of the arches and the groin vault et al. 2003b), the main objective of this first phase of
(Fig. 23). structural restoration interventions was (i) to improve

82
Figure 25. Old mortars of first and second construction
phase and cracked mosaics that have to be preserved in situ.
Figure 26. Typical examples of the first phase structural
restoration interventions (Recent mortars removal, grouting
preperation, stiching of cracks, local reconstruction).
in the best possible way the mechanical behaviour
of the masonry elements by reinstating their continu-
ity (lost due to the numerous cracks) and increasing
their resistance (mainly to traction and shear), without
side effects due to possible durability matters. This
objective had to be achieved taking into account the
existence of mural mosaics, frescoes and old point-
ing mortars, that had to be preserved and conserved
in situ. In fact all uncolored joints of Figures 12 & 13
still preserve the old Byzantine pointing mortars.
This first phase of masonry repair interventions
comprised mainly the following works (Fig. 26): i)
very careful removal of plasters and deteriorated point-
ing mortars applied during previous interventions,
83
Figure 27. Typical examples of in situ mural mosaics con-
servation works (including installation of fine tubes for
grouting).
without harming the old ones, adjacent to or under-
neath them, ii) removal of tiles and other covering
and filling materials to reach the extrados of all
vaulted structures, iii) stitching of the most severe
cracks, using long stones, bricks, or thin titanium
plates, iv) few local reconstructions necessary either
for the repair of dislocated or collapsed parts or for
the restoration of past morphological alterations, v)
deep re-pointing where necessary and preparation of
the masonry for injection grouting, vi) implementa-
tion of injection grouting, vii) removal of all injection
tubes, (viii) in situ conservation of all deteriorated
old mortars using frescoes techniques, ix) execution
of all necessary works to ensure the protection of the
extrados from rainwater.
In parallel with the works for the masonry repair,
the competent Conservators have carried out all the
necessary works for the in situ conservation (includ-
ing grouting) of severely damaged mural mosaics,
(Chryssopoulos et al. 2003). As anticipated, in most
cases, the cracks on the masonry affected also the
mosaics beard on its internal face (Fig. 27).
From the above brief presentation of the works
undertaken, it is ensued that the application of ade-
quate mortars and grout compositions, constitute a key
parameter in order to ensure a successful intervention,
both for the masonry and the mosaics.
Thus, extended laboratory and in situ pilot tests
have been undertaken for the formulation of the ade-
quate mortars compositions taking into account the
characteristics of the existing materials, the require-
ments set by the architectural and structural study
and the worksite conditions. For the majority of the
re-pointing works adequate lime-pozzolan based mor-
tars were used. For the local reconstructions, deep
re-pointing of extremely damaged critical areas and
stitching with thin titanium plates, hydraulic lime
based mortars were applied. Detailed information con-
cerning the properties of existing mortars is given in
Papagianni 2002, while all the results concerning both
the properties of existing materials and those designed
and applied to the monument are reported in Anagnos-
topoulou & Miltiadou-Fezans 2007. In order to ensure
the protection and in situ conservation of old pointing
mortars, on the external facades, the upper central area
just below the dome, and the groin vaults of the sanc-
tuary, the deep re-pointing works have been executed
by experienced Conservators.
As anticipated special attention has been given to
the design of the adequate grouts, not only because
grouting constitutes the main of the works under-
taken that can substantially improve the mechanical
behaviour of the masonry, but also due to the fact that
it is an invisible and irreversible intervention, which
affects the masonry elements and all the mosaics, fres-
coes and old mortars beard on them. Furthermore,
as the proper design of a grout composition cannot
ensure on its own the successful execution of the
grouting intervention, a specific application method-
ology has been developed and applied. The most
important aspects of this methodology are reported in
Miltiadou-Fezans et al. 2008.This application method-
ology permitted the implementation of injections to
this important monument, bearing mosaics, frescoes
and old mortars that had to be preserved in situ, in a
more rational and fully controlled way. The monitor-
ing results and those obtained by sonic tomographies
summarized below, gave important information for the
effect and the efficiency of the whole grouting inter-
vention, and enhanced the validity of the methodology
applied.
During the whole project, all interventions have
been documented in a detailed way (adequate as built
drawings have been elaborated), together with all the
new findings concerning the materials, types of con-
struction, past interventions and pathology of various
invisible elements (i.e extrados of groin vaults). Thus
the historical and recent pathology drawings have been
updated and all necessary information has been col-
lected to be taken into account in the study of the
second phase interventions.
11 SEISMIC MONITORING SYSTEM AND
APPLICATION OF NDT’S
As aforementioned, in order to increase our under-
standing of the seismic structural response and to
decrease the uncertainty of the seismic action, the
installation of an earthquake monitoring system was
decided. The system was installed on the central core
of the monument at 2003, after the implementation of
the aforementioned emergency measures, and before
the beginning of restoration works.
Thus, when the hydraulic lime grouting applica-
tion started (Miltiadou et al. 2008), in June 2006,
the system was functioning and had already recorded
two weak earthquakes. The first took place before the
beginning of masonry repair (26-9-2004) and the sec-
ond during the preparation of masonry for grouting
84
0.45
0.40
0.35 032 039
Period, T (sec)
0.30
0.25
027 030 026 026
020
0.20
022
0.15
0.10
0.05
0.00
26/9/2004 8/1/2006 5/11/2006 12/12/2006
West - East South - North
Figure 28. Variation of the period of the first mode.
by stitching of cracks and local deep re-pointing
(8/1/2006). After three and four months from the end
of the first grouting period, during which grouting was
implemented up to the springing level of groin vaults,
two more weak earthquakes were recorded (5/11/2006
and 12/12/2006). The analyzed data, have shown that
the monitoring system installed can give important
information for the overall behaviour of the structure,
before during and after the implementation of var-
ious interventions. Detailed information is given in
Mouzakis et al. (2008).
As shown in Figure 28, during the earthquake of
08/01/2006 occurred after the urgent measures and
while preparing the masonry for grouting, the natu-
ral period of the first mode was increased, and hence
the stiffness of the structure was decreased. This is
totally expected, as only local deep re-pointing and
stitching of some cracks had been applied at that
moment, while the emergency measures have not been
designed to undertake seismic actions. Moreover an
increase of damping was recorded. These results may
indicate a slight deterioration of damages. On the con-
trary, as a result of the implementation of grouting
injections up to the springing level of groin vaults,
the period and the damping of the first mode were
decreased, along the EW and the NS directions. This
result was reached by analysing the data of both seis-
mic events, happened three and four months after the
first grouting period. Thus, it was proved that the effect
of interventions and more specifically of grouting on
the overall behaviour of the structure can be evalu-
ated by a seismic monitoring system. This promising
finding led to the decision to install in 2007 comple-
mentary instrumentation, with the aim to cover the
whole structure.
The extended seismic monitoring system has
recently recorded two more earthquakes (6/1/2008 and
30/1/2008), and the data are still under processing.
These recent seismic events took place after nine and
ten months from the end of masonry repair works
(28/4/20070), during which grouting has been carried
out to the entire structure below the level of the base
of the dome’s drum. Thus, it is expected that the seis-
mic monitoring records will give further information
for the effect and efficiency of all the interventions
undertaken until now.

Figure 29. Manual-sonic and sub-surface radar maps of
“Saint-Orestes”. Damaged zones are in dark.
12 APPLICATION OF NDT’S FOR MAPPING
MOSAICS SUBSTRATA AND GROUTING
MONITORING
In 2005, an exhaustive high frequency ground-
penetrating radar survey has been realized on the fifty
main mosaics of the Katholikon of Dafni Monastery,
with the aim to investigate the possibility of GPR
application to locate doubtful zones (delaminations,
changes of substrata mortar or other buried het-
erogeneities, etc), related to the mosaics very near
bearing-structure. As presented in Côte et al. 2004 &
2008, a methodology has been developed for the
execution of specific parallel radar profiles, the pro-
cessing of data and the construction of gray-color scale
maps from chosen trenches of the structure, related
to the amplitude of the echoes in the selected trench.
These gray-scale maps were correlated to a level of
detachment or other heterogeneities by comparison
with manual-sonic maps and pathology ones realized
by the Conservators of the mosaics (see Fig. 30). It has
been shown that the GPR maps, giving a more detailed
and less subjective qualitative evaluation of mosaics
substrata, can be a very useful tool for the Conserva-
tors in order to locate doubtful zones in a more refined
way. Furthermore, the possibility of using GPR proce-
dures to monitor the movement of the grout behind the
mosaics, in real time, during injection was also inves-
tigated; detailed information is given in Côte et al.
2008.
Investigations have been also realized to control the
grouting effect into the masonry walls, using seismic
techniques, including some sonic 2D travel time tomo-
graphies. The objective herein was to propose a simple
survey methodology which gives information for the
mechanical characterization of the materials inside the
structure, before and after grouting. The processing
of the results after the completion of grouting is still
85
Figure 30. Tomographical reconstruction in a grouted and
in a non grouted area.
in progress. In Figure 30 two indicative tomographi-
cal reconstructions of a grouted and in a non grouted
area are presented. The comparison of these two cases
permits to conclude that the grouting survey of such
masonry structures may be achieved by the observa-
tion of the velocities of their inner part (Côte et al.
2008).
13 CONCLUSIONS
The high values of the Katholikon of Dafni Monastery
and, hence, the need for accurate information to serve
the design of optimum structural interventions led
to the adoption of a step by step multidisciplinary
approach, both concerning the design and the imple-
mentation of the structural restoration interventions.
This approach proved to be very efficient, as it gives the
possibility to perform the in situ and laboratory inves-
tigations that are necessary for the design of the next
step, as well as for the evaluation of the previous ones.
In this framework, a series of quite novel investiga-
tions were undertaken; the most important results are
summarized below.
1. The holistic design of hydraulic grouts, carried out
on the basis of rational criteria, laboratory and in
situ pilot tests, led to the improved knowledge on
the hydraulic lime based grouts and, hence, to the
selection of appropriate mixes.
2. The experimental assessment of the mechani-
cal characteristics of three-leaf stone masonry
(before and after grouting) contributed to better
understanding of the behaviour of this type of
structures and proved that their repair with ade-
quately designed highly injectable hydraulic grouts
(hydraulic lime based ones or ternary grouts) can
be very efficient.
3. A hydraulic lime based grout was considered as
optimum composition to be applied in situ, on
the basis of injectability, durability, strength and
deformability characteristics.
4. Testing a scaled model of a byzantine groin-vaulted
structure on an earthquake simulator (before
and after grouting) provided valuable information
 regarding the structural behaviour of this com-
plex type of structures, as well as regarding the
strengthening capacity of injections.
5. The developed methodology for grouting appli-
cation gives the possibility for a more rational
and fully controlled implementation of injections,
not only to ordinary structures, but also to impor-
tant monuments bearing mosaics, frescoes and old
mortars, that have to be preserved in situ.
6. The entire procedure for the design and applica-
tion of adequate lime-pozolan mortars enhanced
our knowledge on this kind of mortars, traditionally
used in the past.
7. The monitoring system, installed to record the seis-
mic behaviour of the monument, provides valuable
data on the effect of interventions, as they are
applied in steps to the monument. Thus, the effects
of interventions to various characteristics of the
monument, as well as their efficiency can be con-
tinuously assessed. It should also be noted that, a
similar monitoring system having been installed in
another important monument (the Katholikon of
Osios Loucas Monastery), the comparative eval-
uation of data obtained in the two churches will
allow for a better understanding of the structural
seismic response of middle-Byzantine churches of
octagonal type.
8. The potential of NDT was further enhanced to serve
the needs of investigating important monuments.
High frequency GPR combined with endoscopy
proved to be efficient for the identification of the
masonry construction type. High frequency GPR
contributed also to locate unsafe regions of mosaics
substrata. Furthermore, sonic techniques, under-
taken before and after grouting, were proven very
efficient in checking the effect and efficiency of
grouting intervention, whereas geophysical inves-
tigation by means of 3D seismic tomography pro-
vided valuable information about the foundation
elements.
The entire plan of investigations has supported the
rational design and implementation of the first phase
of works, as well as the evaluation of their effect and
efficiency.
The collected results are now being used for the
design of the second phase interventions. A synthe-
sis of all data provided by the situ surveys during the
first phase of interventions, the results of the monitor-
ing system and those obtained by testing the wallettes
and the groin vaulted structure is under elaboration.
Analytical work will be extended, with further devel-
opment and calibration of detailed numerical mod-
els. This/these calibrated model/models will then be
used to check the efficiency of various strengthening
intervention techniques (installation of ties, metallic
confinements, diaphragmatic structures in the extra-
dos of the vaults and the exonarthex, etc). Thus, the
86
final design of the optimum interventions will be car-
ried out and the respective works will be implemented
during the second and final phase of restoration works.
Last but not least, it is strongly believed that the
experience and knowledge acquired by all levels of
involved stuff in the framework of the structural
restoration of this important monument of World Her-
itage constitute a solid base for similar investigations,
studies and structural restoration interventions to other
historical structures and monuments.
ACKNOWLEDGEMENTS
The contribution of Ass. Prof. E. Vintzileou for the
critical reading of the text is gratefully acknowledged.
The close collaboration othe Scientific Committee
with the personnel of the 1st Ephorate of Byzantine
Antiquities, the Directorate for Restoration of Byzan-
tine and Post Byzantine Monuments, the Directorate
for Technical Research on Restoration and the Direc-
torate for the Conservation of Ancient and Modern
Monuments was the base for this multidisciplinary
approach. The entire project was included in the Oper-
ational Program “CULTURE”. It was co-funded by the
European Regional Development Fund (ERDF-75%)
and by National Funds (25%).
REFERENCES
Bouras, Ch. 1998. The Dafni Monastic Complex
Reconsidered, AETOS Studies in honor of Cyril Mango,
B. G. Teubner Stuttgart und Leipzig.
Chryssopoulos D., Anamaterou L., Georganis F. 2003. Docu-
mentation study for the mosaics of the Katholikon of Dafni
Monastery after the 1999 earthquake. Hellenic Ministry
of Culture (in Greek).
Côte Ph., Dérobert X., Miltiadou-Fezans A., Delinikolas N.
2004. Mosaic-grouting monitoring by ground-penetrating
radar. Proc.IV Int. Sem. Struct. Anal. of Hist. Const.,
Padova, Vol. 1, pp. 401–406.
Côte Ph., Dérobert X., Miltiadou-Fezans A., Delinikolas N.
Durand O., Alexandre J. Kalagri A., Savvidou M,
Chryssopoylos D., Anamaterou L, Georganis, F. 2008.
Application of non-destructive techniques at the Katho-
likon of Daphni Monastery for mapping the mosaics
substrata and grouting monitoring. Sub. to the 6t intern.
conf. on Structural Analysis of Historical Constructions,
Bath.
Delinikolas, N., Miltiadou-Fezans, A., Tsofopoulou, E.,
Minos, N., Chrissopoulos, D. 1999. Initial Inspections,
Assessment and Proposals for Protecting the Monastery
of Dafni, Internal report. Hellenic Ministry of Culture (in
Greek).
Delinikolas N., Miltiadou-Fezans A., Chorafa E., Zaroyianni
E. 2003. Study on restoration of the Katholikon of Dafni
Monastery: Phase A – Architectural and historical Survey.
Hellenic Ministry of Culture (in Greek).
Georgopoulos A. 2003. Novel survey and digital pho-
togrammetry methods for geometrical and architectural
 digital survey of Byzantine monuments. The case Dafni
Monastery. Research Report, NTUA and Hellenic Min-
istry of Culture.
Kalagri A., Miltiadou–Fezans A., Vintzileou E. 2007. Design
and evaluation of hydraulic lime grouts for the strengthen-
ing of stone masonry historic structures. Proc. int. symp.
on Studies on Historical Heritage, Antalya, Turkey, 2007.
Mariolakos, I., Fountoulis, I., Andreadakis, E., 2000.
Engineering geological problems caused by human inter-
ference on monuments that influence their seismic
behavior. Compatible materials recommendations for the
preservation of European cultural heritage. Ac. Civil
Engineers of Greece.
Millet, G., 1899. Le Monastère de Daphni. Histoire, Archi-
tecture, Mosaiques, Monuments de l’Art Byzantine I,
Paris.
Miltiadou-Fezans A., Tassios T.P., Delinicolas N., Chorafa E.,
Zarogianni E., Chandrinos I. 2003a. Earthquake struc-
tural problems and urgent measures undertaken to support
the Katholikon of Dafni Monastery in Athens, Greece.
Proc.8th inter. conf. STREMAH 2003, Halkidiki, Greece,
7–9 May.
Miltiadou-Fezans A., Delinikolas N., Chorafa E.,
Zaroyianni E. 2003b. Study on restoration of the Katho-
likon of Dafni Monastery, Phase A – Structural Survey,
Analysis and remedial measures. Hellenic Ministry of
Culture (in Greek).
Miltiadou-Fezans A., Vintzileou E., Delinikolas N.,
Zaroyianni E., Chorafa E. 2004. Pathology of the Dafni
Monastery: survey, monitoring of cracks, interpretation
and numerical verification. Proc. 4th inter. Sem. on
Structural Analysis of Historical Constructions, Vol. 2,
Padova.
Miltiadou-Fezans A., Vintzileou, E., Papadopoulou, E.,
Kalagri, A. 2006. Mechanical properties of three-leaf
stone masonry after grouting. Proc. 5th inter. Sem. on
Structural Analysis of Historical Constructions, New
Delhi, India.
Miltiadou-Fezans A., Tassios T.P. 2006. New rational crite-
ria for the holistic design of hydraulic grouts. 1st Conf.
of the Society for Research and Promotion of Scientific
Restoration of Monuments (in Greek).
Miltiadou-Fezans A., Kalagri A., Delinikolas N. 2007.
Design of hydraulic grout and application methodol-
ogy for stone masonry structures bearing mosaics and
mural paintings: the case of the Katholikon of Dafni
Monastery. Proc. int. symp. on Studies on Historical
Heritage, Antalya, Turkey, 16–21 September, 2007.
87
Miltiadou-Fezans A., Kalagri A., Kakkinou S., Ziagou A.,
Delinikolas N., Zarogianni E., Chorafa E. 2008. Method-
ology for in situ application of hydraulic grouts on historic
masonry structures. The case of the Katholikon of Dafni
Monastery. Submitted to the 6th intern. conf. on Structural
Analysis of Historical Constructions, Bath, UK.
Mouzakis Ch., Miltiadou-FezansA.,Touliatios P., Delinikolas
N., Dourakopoulos J. 2008. Earthquake based condition
monitoring of the Katholikon of Dafni Monastery. Sub-
mitted to the 6th intern. conf. on Structural Analysis of
Historical Constructions, Bath, UK.
O.T.M. 2000. Geothechnical Investigation of Dafni Monastery.
Internal Report, Hellenic Ministry if Culture (in Greek).
Papagianni. I. 2002. Investigation of physicochemical charac-
teristics of the historic mortars of the Katholikon of Dafni
Monastery. Internal Report, AUTH and Hellenic Ministry
of Culture (in Greek)
Polymenakos L., Papamarinopoulos S., Miltiadou A.,
Charkiolakis N (2005). Investigation of the foundations
of a Byzantine church by three-dimensional seismic
tomography. Elsevier Journal of applied geophysics 57,
pp. 81–93.
Stikas, E. 1962. Restoration and consolidation of the
exonarthex of Dafni’s Katholikon, DChAE ser. 4, 3,
1962–1963.
Tassios, T.P. 2004, Rehabilitation of three-leaf masonry.
In Evoluzione nella sperimentazione per le costruzioni,
Seminario Internazionale 26 Settembre- 3Ottobre, CIAS.
Troump, E., Quelques vielles églises byzantines de la Grèce
moderne, Marseille, 1896, 5–11, 11–15.
Vintzileou, E., Tassios, T.P.1995. Three leaf stone masonry
strengthened by injecting cement grouts. Journal of Struc-
tural Engineering, ASCE, Vol. 121, No 5.
Vintzileou E 2002. Methods for documenation of pathology
and assessment of strengthening measures in Byzan-
tine churches. Application in Dafni Monastery. Research
Report, NTUA and Hellenic Ministry of Culture (in
Greek).
Vintzileou E., Miltiadou-Fezans A., Palieraki V. Delinikolas
N. 2004. The use of radar techniques and endoscopy in
investigating old masonry: the case of Dafni Monastery.
Proc. of 4th intern. sem. on Structural Analysis of Histor-
ical Constructions, Padova Vol. 2, pp. 351–360.
Vintzileou E., Miltiadou-Fezans A., Vrouva A,
Anagnostopoulou S. 2006. Mechanical properties of
three-leaf stone masonry. Proc.5th intern. Conf. on Struc-
tural Analysis of Historical Constructions, New Delhi,
India.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Structural interventions in English Cathedrals

G.A. Clifton
Director Historic Buildings – Gifford Ltd

ABSTRACT: The paper is an overview of the structural interventions made to English Cathedrals. It reviews
some of the most common problems and disasters and looks at how they have been overcome with reference to
particular events.

1 OVERVIEW

I aim to give an overview of structural interventions

that have been made to our Cathedrals in England.
Clearly this is a huge topic worthy of many volumes but
I shall attempt to sketch a flavour of the some of the key
historical events and the common structural problems
with reference to some particular case histories.
Many of our Cathedrals in England were built, and
in many cases rebuilt, during the period of great activ-
ity in the 1100 and 1200’s. As architectural fashions
changed, new prelates were appointed or local saints Figure 1. Chester Cathedral before restoration.
beatified, there was a tendency to refashion the Cathe-
drals in the latest style. The growing wealth of the
church made all this possible, but excited the envy of
the English monarchs. This led to Henry VIII abol-
ishing the monasteries and acquiring their wealth in
1536, at which point many of these fine buildings
became stone quarries for the locals. The depreda-
tions continued during the civil war in the 1640’s, when
most of the art work and fine interiors were stripped
and the buildings were sometimes used as stabling
for Cromwell’s forces. During the Victorian period,
in the latter part of the 19th century, there was a mas-
sive programme of repair, renewal and restoration in
our Cathedrals. This was funded by public donation,
Figure 2. Chester Cathedral after restoration.
much from very wealthy individuals. Repair schemes
included dramatic tasks such as the saving of the east
end of Winchester Cathedral, by the underpinning of 2 DISASTERS
the foundations by the diver Robert Walker. The work
often went well beyond restoration, with Gilbert Scott The Cathedrals have survived earthquakes, fire, flood
making many “Improvements” for example at Chester and man’s depredations over their 900 year history and
Cathedral. survive today as superb examples of man’s endeavours
Many of the restorations at the end of the 19th and his wish to glorify God.
and early 20th centuries made use of the “magic new The structural designs often pushed the limits of
material” reinforced concrete.A typical example being structural knowledge and collapses were not uncom-
the tieing together of the towers and nave of Lincoln mon, a list just of some of the major spire and tower
Cathedral with concrete beams and frames, together events is instructive.
with the tieing of walls with phosphor bronze rods and 1107 Winchester Walkelin’s central tower fell and the
extensive cement grouting. present one built

89
1170 Gloucester South tower to the west front
collapsed
1175 Worcester Western tower fell into the river Severn
1185 Lincoln The whole building collapsed during an
earthquake
1210 Chichester West tower collapses
1238 Lincoln Central tower fell
1248 Wells Central spire fell in an earthquake
1272 Norwich Spire burnt down during rioting
1322 Ely Central tower removed before it fell
1362 Norwich Replacement timber spire blown down
in a gale
1407 York Central tower and Spire collapsed
1463 Norwich Second replacement spire destroyed by
lightning The stone replacement from 1480 still stands!
1548 Lincoln Timber spire (524 ft high) destroyed by
lightening
1561 Old St. Paul’s Timber spire destroyed by lightning
1615 Ripon Central spire collapses
1635 Chichester West Tower collapses again
1640’s Durham West Tower demolished by Cromwell’s
army
1642 Lichfield Central spire demolished by Cromwell’s
army Figure 3. Chichester spire collapse.
1704 Canterbury Central spire removed due to storm
damage
from flying buttresses. On visual inspection one is
1786 Hereford Central spire removed after collapse of
immediately worried that the vault is in tension and
west front
that the slender columns are undergoing excessive
1807 Lincoln West front spires demolished as unsafe
bending.
1832 St. Albans Spire removed as unsafe
I have investigated these carefully and can show
1861 Chichester Central tower and spire collapse
that they have a factor of safety greater than 1.0 which
1880’s Peterborough Tower replaced to prevent a
is just as well as they have been standing unchanged
collapse.
for 900 years. On searching back through the archives
This is by no means an exhaustive list as it not I find that every Cathedral Architect, Engineer, and
does not include many of the major fires that have before that the master masons, were all expressing con-
caused untold loss to the Cathedrals, but does show cern about the retro choir and all setting up their own
the determination to recover from disaster. monitoring systems – just as I have done! Indeed the
Victorians had a series of specially fabricated props
which were kept close to the retro choir for instant
3 LONG TERM WORRIES installation if any cracks appeared. My own advice is
that if cracks appear everyone is to run, as there will
Over this time the buildings were subjected to much probably be localised and fast compression failures! It
repair, renovation and alteration and we know, as did gives one a profound sense of historical continuity to
our forefathers, of the value of constant maintenance find that we share the same worries as our forbears.
and the dire consequences of its omission. We have Another fascinating case is that of the Deans Eye
records of many of these campaigns of repair and window at Lincoln Cathedral.This is one of the top five
alteration and it is fascinating to find that many of rose windows in Europe with fabulous 1220 stained
our current concerns are merely the continuation of glass and a very ambitious form.
worries of the past generations. The companion window, the Bishop Eye collapsed
I have personal experience of two very clear exam- within 50 years and the archives show a long history
ples of such ongoing concerns, one which, so far, has of repairs to the Deans Eye, some successful and some
required no intervention, but a second which resulted a failure. A constant theme in Clerk of Works records
in significant works. over the centuries is how to keep the window stable.
The retro choir at Wells Cathedral is a vaulted This was initially by adding cramps and lead filling
area between the high altar and the Lady Chapel, the to the joints as they moved, then by bracing across
supporting columns are very slender and support an the central opening and latterly by adding major cross
unbalanced vault onto which are imposed large loads bracing behind the window. Until by the year 2000 no

90
Figure 4. Wells Cathedral retrochoir.

Figure 5. Deans Eye window before restoration – Lincoln

Cathedral.
more patching was possible and replacement was the
only option. A complete reconstruction of the stone
tracery was required and the opportunity was taken to
include some hidden strengthening so that the internal
and external cross bracing which had disfigured the
window was no longer needed.
4 THE LAST 50 YEARS
In the latter half of the 20th Century there has once
more been a significant round of repair and renewal,
with priming funding from English Heritage, but with
the majority raised from appeals. As opposed to the
Victorian repairs this time there was little “improve-
ment”, the emphasis being upon conservation and
restoration.
There were major interventions to stabilise the
structure of these great buildings and four typical
examples are:
1 Improving existing foundations by some means of
underpinning, such as at York Minster.
2 Tieing elements together with concrete ring beams
at roof, clerestory and triforium levels and adding
tie rods through towers and facades. Ely Cathedral
being a good example.
Figure 6. York Minster structural restoration.
3 Stabilising towers and spires with concrete or steel
frames, such as Salisbury Cathedral.
4 Reconstructions after fires, several examples unfor-
tunately and York Minster suffered more than most.
5 YORK MINSTER
York Minster is built on an historic site and overlays
previous constructions including a roman basilica. The
Minster was started in 1100’s but its present form was
finished in 1472 – much of it having to be rebuilt after
the central tower collapsed in 1407 whilst attempting
to remodel the tower piers.
5.1 Underpining
In 1956 a detailed survey showed a number of instabil-
ities in the building. The central tower had differential
settlement of 225 mm that was continuing and inves-
tigations showed severely damaged foundations, the
east wall was 600 mm out plumb and still moving and
the building was suffering a number of other similar
problems. A major programme of works was under-
taken of which the underpinning of the central tower
piers was probably the most significant. The intention

91
Figure 7. Fire at York Minster.
of which was to halve the bearing pressure below the
foundations to approx. 300 Kn/m2 , given that the sup-
porting soil is poor it is no wonder that it was still
settling.
The principle employed was to join the footings
below each of the central piers to their adjacent nave
and transept columns making one large footing twice
the size of that existing. The masonry footings were
surrounded by new concrete and the whole post-
tensioned into a single composite foundation. In order
to ensure that the new parts of the foundation would
carry load, additional footings were cast below the
new areas, with flat jacks between them and the new
foundation. The jacks were then stressed in order to
consolidate the soil below these areas, when this had
happened and all movement ceased then the gap and
jacks were grouted.
This was a massive and complex task which was
undertaken successfully and has ensured the future of
the minster.
5.2 Restoration after the fire
The instabilities were not the only problems at York.
In 1984 a major fire resulting from a lightning strike
destroyed the South Transept. A significant recon-
struction was then needed.
6 ELY CATHEDRAL
In the early 1970’s inspections showed that the
stonework in the West Tower at Ely Cathedral was
loose and decayed with many cracks through the tower.
The tower in common with many others is effectively
two almost separate shells, so that over time they
move independently, resulting in cracking and loss of
stiffness.
More detailed research showed that in the 1860’s
Gilbert Scott had carried out major repairs which
included the installation of massive wrought iron bars,
92
Figure 8. Ely Cathedral – West Front.
125 mm by 30 mm diagonally across the belfry and
four 60 mm diameter wrought iron ties across the
tower at two levels, plus the inclusion of many other
wrought iron ties and straps throughout the tower.
These were now exacerbating the problem as some of
them corroded.
The work was carried out under the control of
Jacques Heyman and was in several parts.
1 The external wrought iron ties and strapping that
was corroding, and damaging the stonework, was
removed, whilst all the internal ties that were still in
good condition were not disturbed.
2 The lower two thirds of the tower have been rein-
forced by stitching with stainless steel bars and
grouting. It having been found that the core to the
walls was loose in areas with some voids. The stitch-
ing effectively created complete ring beams at each
of four levels tieing together the two shells of the
tower.
3 The geometrical stability of the tower was further
assured by the provision of three sets of stainless
steel ties passing from face to face of the tower and
connecting together the corner stairs. One of these
replaced some of Scott’s work and the others were
additions.
4 The octagonal belfry, which is of a lighter con-
struction than the main portion of the tower was

Figure 9. Ely Cathedral – strengthening of the West Tower.
strengthened at its top by a new reinforced concrete
ring beam, with stainless steel cables inserted and
pre-tensioned at a lower level.
This is an excellent example of the progression of
repairs to Towers in English Cathedrals. As we have
seen there were numerous collapses of towers rela-
tively soon after their construction or alteration and
there was little that could be done with the technology
available at the time to save them. They were often
significantly under designed.
As the centuries passed by the continuing settle-
ment of overloaded ground below the central piers
caused ongoing cracking of the major towers and this
resulted in loss of stiffness which caused further redis-
tributions of load. This often increased the load on the
piers, with an ongoing racheting increase in cracking
and movements. Sometimes this was added to other
interventions such as the removal of the choir screen
at Chichester which ultimately caused the failure of
the spire. In other cases the inclusion of wrought iron
ties and strapping alongside the repair of crumbling
stonework was sufficient to maintain stability.
93
Figure 10. Ely Cathedral – strengthening of the West Tower.
However, as is so often the case, whilst the inter-
vention was successful it introduced other difficulties
that in time also needed remediation. So that whilst
in the original construction of these massive tower
walls timber ties were sometimes included to hold the
walls together, these rotted and the towers began to
lose their integrity. So the wrought iron ties, in their
turn, decayed and needed replacing.
7 SALISBURY CATHEDRAL
A different approach was undertaken at Salisbury
Cathedral where there were serious concerns about the
stability of the spire in the 1970’s. Salisbury spire has
a series of bands of open stonework at several levels in
its height and the lowest level of stonework was decay-
ing such that there were serious fears as to its ability
to continue to carry the wind loads and weight from
the spire. It was not possible to individually replace
the stones as it was highly likely that the spire would
collapse with the removal of some of the stones. Vari-
ous options were considered including the dismantling
and rebuilding of the spire – though needless to say that
was thought of as a last resort.
The successful solution was to introduce a stainless
steel frame inside the spire which effectively transmit-
ted the load from the upper part of the spire into the
frame and bypassed the open section. The frame was
then jacked against the upper spire, to ensure that it
was taking some of the load from the open section.
8 REPAIR AND MAINTENANCE
Alongside these major interventions there is an
ongoing need for constant repair and renewals. All
Figure 11. Salisbury Cathedral.

Cathedrals have a long programme of stone replace-

ment, working around the exterior over a 100 year
cycle replacing eroded stone. The work can vary
from the straightforward on nave walls, to the
complex on flying buttresses, to the difficult on
towers.
Wells Cathedral central tower requires replacement
of areas of heavily eroded stones and other repairs on
roughly a 120 year cycle; this being about the period
by which time pieces of stone are falling off to the
danger of the public below. In 1998 it was time to carry Figure 12. Wells Cathedral – Tower repairs.
out such a task. The biggest difficulty was getting the
access to all faces with a working scaffold as there is no
access to the ground below any face. A flying scaffold 9 FINALLY WHAT OF THE FUTURE?
had to be used supported from the transept, nave and
choir wall junctions, with an access bridge along the All over Europe we have repaired, strengthened and
ridge of the south transept from a lift/stair tower on the stiffened our major masonry buildings with concrete.
south transept gable. Is this a good thing? Is it durable? What happens to
The sequence in which the stone replacement is our lovely flexible masonry buildings in the next 900
carried out is important as with some areas requir- years after they have been stiffened and tied together?
ing extensive repairs it is vital to ensure adequate load That has to be the subject of a future paper.
transfer at all times to avoid local overstressing.
As well as these major works, repairs and replace-
ments there are always new constructions happening REFERENCES
at Cathedrals which impact upon the existing fabric. The Structural Engineer vol. 85 number 3 2007. The Deans
Many Cathedrals are improving their facilities such Eye Window – Lincoln Cathedral.
as by the addition of choir rehearsal spaces, storage, Institution of Civil Engineers proceedings 1971 paper 7415
education rooms, visitor centres and refectories. These S. York Minster Structural restoration – Dowrick and
all need to be close to the main building and often Beckmann.
structurally impact upon, it. Added to which are the Proceedings of the Institution of Civil Engineers 60 1976.
difficulties with the foundations, as invariably these The strengthening of the West Tower of Ely Cathedral.
new buildings will be located over areas previously Thesis M. Kirby. Chichester Spire collapse. Medieval Cathe-
drals their design, construction and structural problems
built upon, so with important archaeological remains
Int. journal of solids structures Oxford 1966. The Stone
below the surface. The foundations become a game of Skeleton.
finding any spaces that are clear for support and then J. Heyman ISBN 0-86078-597-1. Arches, Vaults and
trying to carry the building from these points. Buttresses.

94
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Monastery of Salzedas (Portugal): Intervention in the cloister and

information management

P.B. Lourenço, L.F. Ramos & G. Vasconcelos

Department of Civil Engineering, University of Minho, ISISE, Guimarães, Portugal

F. Peña
Instituto de Ingeniería, UNAM, Mexico

ABSTRACT: Anyone involved in the conservation of cultural heritage buildings is aware of the enormous
amount of information generated by the different specialists, which is generally not handled adequately by owners
and/or authorities involved in the process. The consequence is that valuable information is lost in a complex
process of reaching a decision that involves many different experts and information tends to get forgotten
or misplaced in the course of time. A case study with extensive damage, which recently suffered significant
conservation works, is described in detail, so that the adopted remedial measures are justified. Then, the case
study is also used to develop an application for the efficient management and visualization of the information
related to interventions of cultural heritage buildings.

1 INTRODUCTION 2 BRIEF HISTORICAL REVIEW

Monastery of Santa Maria de Salzedas (Portugal)
recently suffered major works in one of the cloisters.
This case study is addressed here from a compre-
hensive perspective, including aspects of the inter-
vention project and aspects related to information
management.
The first part of the paper describes in detail the
Monastery of Santa Maria de Salzedas and the con-
servation works in the cloister. This includes survey
and characterization of the damage, together with in
situ and laboratory testing, which allowed to gather the
information necessary to establish the need of an inter-
vention and to bound this intervention. The obtained
information allowed obtaining a computer simulation
of the building, which resulted in clear information on
its structural behavior. The details on the consolidation
project and the execution works are also provided.
The second part of the paper addresses a database
created for the management of the information gener-
ated during the intervention process and the historical
information collected, which has been produced along
the time and has been selected by different areas of
study. The information is accessible via intranet for
practioners and is planned to be accessible on the
monastery for visitors, in way that a non-specialist and
a team specialist can obtain information with different
contents. The paper discusses the underlying princi-
ples in which the application is based on and details
the main characteristics of the application.
The Monastery and Church of Salzedas are located
in Salzedas, Tarouca and the church was recently
classified as a National Monument. The church is
essentially set in an urban environment, whereas the
monastery is set in a more rural environment (Fig. 1a).
The plan dimensions are very large, 75 × 101 m2 . The
monastery and church possess a longitudinal irregu-
lar plan with different volumes, typical of a Cistercian
Abbey (Fig. 1b). The conservation works addressed
here focus in the cloister dated from the 17th century
(Main Cloister, in the picture).
2.1 Previous documented works in the
main cloister
The main cloister is regular and substitutes part of
the primitive cloisters (Fig. 2a). It possesses cross
vaults in the 1st level and barrel vaults in the 2nd
level. The walls, brackets and ribs are made of gran-
ite and the vaults are made of brick masonry with
clay or masonry filling. After repeated statements of
the pre-collapse status of the cloister, Leitão (1963)
and Cocheril (1978), the former General Directorate
for National Buildings and Monuments (DGEMN)
carried out remedial works in 1980/1981 and 1983
(Figs 2b-d), including: (a) Demolishing and replac-
ing the vault of the 2nd level of the West wing
by a reinforced concrete vault; (b) Dismounting and

95
 Church

Sacristy
Figure 3. Plan of Clairvaux, Cocheril (1978).
Main Chapter
Little
Cloister Cloister Room reassembling the wall separating the large and small
Inn cloisters, between the 1st and the 2nd levels.
The works do not comply with modern theories
of intervention in historical structures and would be,
today, very debatable.

(b)
3 VIRTUAL RECONSTRUCTION
Figure 1. Monastery and Church of Salzedas: (a) aerial
view; (b) plan and spatial units. The majority of the monastery disappeared after the
extinction of the religious orders in the Age of Enlight-
enment without letting any vestige. A 3D CAD model
of the virtual reconstruction of the medieval monastery
of Santa Maria de Salzedas was made in order to
allow visitors to better understand the monument and
to assist the conservation works.

3.1 The ideal plan of the Cistercian abbeys

Figure 2. Aspects of the cloister: (a) view before current
intervention works; (b) demolition of the barrel vault in the
1980’s; (c) dismounting of the wall between the large clois-
ter and the small cloister; (d) reconstruction of a reinforced
concrete vault.
The Cistercian Order did not ever define a unique
model for the monasteries, in which a formal model
was defined and repeated at each monastery. How-
ever, the similarities among abbeys, especially in the
distribution of spaces, put in evidence a common
or ideal plan for the monasteries (Fig. 3), see also
Leroux-Dhuys (2006).
Although the plans were very similar among
abbeys, they are more different in elevation. The
monastery followed a basic square structure; where
one of its sides, aligned along the East-West axis,
was formed by the church, always in the highest part
of the property. The other parts could be constructed
oriented to the North or to the South, according to
local relief. This is the largest variation observed in
the “replication” of the ideal plan among monasteries.
It is noted that the disposition of divisions was such
that the kitchen and the refectory were located in the
lowest part, near to a water-course.

96
 The church was oriented in the East-West direction,
with the apse to the East. The plan was a Latin cross,
with three naves, being the aisles lower than the cen-
tral nave. Usually the body of the naves was preceded
by a narthex and the transept was formed by a single
nave. The naves were divided by square pillars with (a)
engaged or truncated columns and were covered by
barrel vaults, oriented perpendicularly.
Sometimes, on the side of the transept, engaged to
the exterior wall, a bell tower could exist, built in stone
or wood. The construction of towers similar to tradi-
tional secular churches was forbidden. The church’s
facade was simple and divided in three parts.A detailed (b)
summary description on the arrangement of Cistercian
abbeys can be found in Amado et al. (2006).

3.2 Medieval vestiges in the actual monastery

Parts of the medieval construction could still be found
in the actual monastery. Most of the vestiges were
found in the church, while very few elements of the
medieval period could be found in the cloister. This
is probably due to the continuous changes made in
the 16th to 18th centuries in the cloister. In fact, it is
possible to find masonry ashlars from the medieval
monastery that were reused in the new structural
elements.
The church conserves the medieval walls up to the
openings of the lateral naves with the exception of the
main façade, which was rebuilt in the last campaign.
The church conserves an apse chapel from the prim-
itive construction in the northern side of the transept
(Fig. 4). It is semicircular and built with granite stone
blocks, with two engaged columns that divide the exte-
rior wall in three parts. In the central part, a rectangular
window provides light to the interior. In this same part
of the transept, in the western interior wall, there is a
spiral staircase that certainly conducted to the ancient
tower. In the outside, in the wall of the northern side
of the transept, there are two closed openings (Fig. 4).
The central and larger one should be Door of the Dead,
now leading to the cemetery. From prospecting pits,
it was possible to conclude that the outside walls of
the cloister were constructed on top of the medieval
foundations.
3.3 Definition of the modeling unit
The modeling unit is the base of the system of propor-
tions, which was an indispensable element to provide
the composition of the entire building. The Cistercian
buildings were based on the composition principle of
“ad quadratum”, Virgolino (1997), in which a simple
generating mesh gives the proportions of the elements.
In the Portuguese constructions from the beginning
of the second millennium, the units of measure com-
monly used were the Roman palm (0.223 m) and the
(c)
Figure 4. Medieval parts of the monastery that remains
in the church: (a) northern part of transept; (b) main nave;
(c) basement of the northern wall.
Roman foot (0.296 m). However, it is very likely that
the Cistercian Portuguese constructions had used the
“pied du Roi” or king’s foot, equivalent to 0.32484 m,
since it was the standard measure used in France in
that time, Virgolino (1997). It is possible to observe
that the modeling unit used in Salzedas was equal to
8 king’s feet, since the main elements of the church
present values very close to integer multiples of this
modeling unit. The reference module mesh is depicted
in Figure 5.
3.4 Model of the medieval monastery
The church, the cloister and the three wings (monks,
refectory and lay-brothers), which together constitute
the five main bodies of the monastery (the abbey’s
buildings), were built at different levels, in order to
adapt to the morphology of the location. Perpendicu-
larly to the body of the church, the bodies of the cloister
are built aligned with the direction of the slope of the
ground.
The model prepared for the medieval monastery
only presents part of the abbey’s buildings (Fig. 6). The
construction has a Romanesque solution, in which the
building compound was composed by simple paral-
lelepiped, massive and juxtaposed volumes. The walls,
foundations and pillars were built with large ashlars,
while the top and intermediate floors were usually
made with barrel or quadripartite vaults, in stone or
brick masonry. The roofs were of two slopes, with tim-
ber structure and tile covering. The cloister should be

97
Figure 5. Reference model mesh of the medieval church.
Figure 6. Three-dimensional model of the medieval
monastery.
supported on double flat and round columns. The cho-
sen solution is similar to the French cloisters: Romanic
with a heavy and massive clean style. Details of the
model are shown in Figures 7 and 8.
4 IN SITU SURVEY
4.1 Visual inspection
The condition of the cloister was quite poor, includ-
ing biological colonization sometimes associated with
98
Figure 7. External views of the medieval model.
moisture stains, deterioration of the bricks in the
vaults, cracks with variable thickness, crushing of
stones and excessive movements in walls and vaults.
Large cracks could be observed in the cloister
(Fig. 9). The largest and widest set of cracks occurs
in the barrel vaults of the South and East wings of
the 2nd level, as well as the SE and NW corners. The
cracks occur mostly in the longitudinal direction up
to a crack opening of 40 mm, even if some transversal
cracks also occur. The vaults of the 1st level exhibit
cracks in the South and West wings, up to a crack
opening of 15 mm. The West wing was supported on
temporary wooden poles.
The walls exhibit well distributed cracking at the
2nd level and almost no cracking at the 1st level. With
the exception of a few localized areas, cracking is
minor (crack openings in the range of 1 to 5 mm).
Vertical displacements up to 35 mm were measured
at the key of the crossed vaults of the 1st level. But all
the walls of the cloisters exhibit large horizontal move-
ments that lead to the separation between the vaults
and the walls, in a clear lack of verticality (Fig. 10).

Figure 8. Internal views of the medieval model.
The out-of-plumb displacement of the internal walls
reaches values of 0.18 m, 0.14 m, 0.09 m and 0.07 m
in the wings West, South, East and North, respectively.
The brackets supporting the crossed vaults of the
first level show signs of compression/shear damage,
particularly in the West wing (Fig. 11a). This can be
explained by the tilting movement of the walls. The
absence of connection between the infill of the crossed
vaults and the walls resulted in a much localized area to
transfer the load, i.e. only the brackets. Also, a signifi-
cant number of bricks show deterioration, particularly
around the cracked areas (Fig. 11b). This occurs at
bothflorr levels and can be explained by frost-thaw
cycles and water infiltration, as the amount of rainfall
per year in the region is high and the temperatures in
the winter are excellent for ice formation (daily cycles
with ± 0◦ ).
Other perturbing signs, less relevant from the struc-
tural point of view, include damage of the stone due to
99
Figure 9. Crack survey: (a) 1st level; (b) 2nd level;
(c) South wing, 2nd level; (d) South-West corner, 2nd level;
(e) West wing, 1st level; (f) South-West corner, 1st level; (g)
wall, West wing.
freeze-thaw cycles, effloresce and biological coloniza-
tion (Fig. 12). A detailed survey on stone deterioration
and salt effloresce in the cloister are given in Alves &
Pamplona (2006, 2007), see also Figure 13.
5 IN SITU TESTING AND LABORATORY
INVESTIGATION
In order to better characterize the materials, to justify
the observed damage and to define corrective mea-
sures, an experimental in-situ and laboratory testing
program was carried out.
5.1 Soil and foundation survey
This survey consisted of seven boring holes and
three pits to define the mechanical and physical
Figure 10. Aspects of the wall movements: (a) separation
between wall and vault (intrados of 1st level); (b) detail of
previous separation; (c) separation between wall and vault
(extrados of 2nd level) and (d) horizontal movement inwards
to the court.
Figure 11. Mechanical deterioration: crushing/shearing of
the stone brackets of the 1st level and decohesion of the
bricks.
characteristics of the soil and foundations. It was pos-
sible to define a layered soil consisting of an infill
of clayey nature (1.1 m), organic soil (0.30 m), allu-
vial soil with medium large stones, naturally wounded
and worn by the action of water (0.60), alluvial soil
with pebble (0.50 m). Between 2.5 and 2.7 m depth,
the soil is granular with some clay and below 2.7 m
depth large stones, with a size of 0.30 m to 0.40 m are
found (Fig. 14).
100
Figure 12. Material deterioration: degradation of stone
carving and biological colonization.
Figure 13. Examples of stone deterioration survey: (a) stone
detachment in West wing; (b) SEM images of salt crystals.
The foundation soil exhibits moderated resistance
and large heterogeneity for depths between 1.0 and
1.8 m, were supposedly all the cloister foundations are
set. The foundations for the walls seem to be medieval
and of good quality, but the foundations of the clois-
ter columns are inadequate. These irregular masonry
foundations are unable to distribute the loads over a
significant soil area and the foundations depth around
1.0 m seems to indicate that the foundations were
built on top of the original pavement level, directly
on organic soil.

Figure 14. Soil characterization: (a) identified layers;
(b) inspection pit 2; (c) inspection pit 3.
Figure 15. Aspects of the inspection with rigid endoscope.
5.2 Internal characterization with rigid
endoscope
In order to characterize the inner constitution of
vaults and walls, a few bore holes and several cracks
were inspected with a rigid endoscope (Fig. 15). The
inspection allowed several conclusions, among which:
(i) vaults are made with clay brick masonry with
0.22 m thickness. Infill material in the 1st level is soil
and infill material in the 2nd level is a sort of rubble
masonry. Separation between the two materials was
not found; (ii) walls are made with large granite stones,
with dry joints or a thin clay joint. The clay joint seems
to be washed out around the cracks and in the external
part of the wall due to weathering. An internal core
of weaker mechanical characteristics was not found;
(iii) internal longitudinal cracks that would compro-
mise the stability of the walls under vertical loading
were not found.
101
Figure 16. Flat-jack testing: (a) aspect of one test;
(b) stress-relative displacements results.
As a result of the inspection with the rigid endo-
scope, it was concluded that the granite walls of the
cloister are adequate and there is no danger of collapse
due to desegregation under vertical loading. Coring
and other techniques to estimate the strength of the
walls were considered not necessary and it was decided
to carry out two simple flat-jack tests.
5.3 Flat-jack testing
The results of one test are shown in Figure 16. The
average in-situ stress obtained is 1.2 MPa and no indi-
cation of bending in the walls was found. The value
expected is around 0.6 MPa. The difference between
these values might be due to the irregular shape of the
masonry blocks, indicating that only half of the stone
block is active. It is stressed that the value of 1.2 MPa
obtained can be considered as relatively low for the
particular type of masonry.
5.4 Coring
In order to confirm the internal constitution of the
vaults and to characterize the mechanical behavior of
the brick masonry, three φ75 mm cores were extracted
from the vault (Fig. 17). The cores confirmed the
borehole observation.
5.5 Chemical, physical and mechanical
characterization
The plaster, vault infill and mortar from the brick
masonry of the vaults were characterized with
X-ray diffraction, non-soluble residual and burn
loss tests (Fig. 18a). The bricks were characterized
with absorption tests and uniaxial compression tests
Figure 17. Samples for visual identification and mechanical
testing.

(Fig. 18b, c, d). The mortar from the brick masonry was
also characterized with uniaxial compression tests.
The representative samples were extracted from the
construction or the cored samples.
The tests indicated the composition of the plaster
and mortar (1:3 in volume) and the composition of the
vault infill. The bricks are of low quality and non-
durable, with an absorption in cold water around 20%
and a volume mass of 1560 kN/m3 .
The uniaxial compression tests were carried out
in samples of 45 × 45 × 45 mm3 , tested with greased
Teflon layers to avoid the plate confining effect. The
obtained Young’s modulus and strength for the bricks
were Eb = 7300 MPa and fb = 5.2 MPa, respectively.
These values are quite low and confirm the poor
quality of the bricks. The obtained Young’s modulus
and strength for the mortar were Em = 8600 MPa and
fm = 3.8 MPa, respectively. This strength value can be
considered normal for the mortar composition.
With these results it is possible to estimate the
strength of the brick masonry as:

according to ENV 1996-1-1. The Young’s modulus of

the composite might be obtained from homogenization
procedures as:

where tm represents the thickness of the mortar, tu rep-

resents the height of the brick and ρ represents an
efficiency factor associated with the deficient bond
between the two materials (assumed equal to 0.5).

6 SAFETY ASSESSMENT
The objective of the structural analysis carried was the
safety assessment of the cloister and the definition of
remedial measures. For this purpose, two models of
Figure 18. Examples of chemical, physical and mechani-
cal testing: (a) X-ray diffraction test on vault infill material;
(b) absorption test on brick; (c) samples tested in uniax-
ial compression; (d) compression test results on brick (two
samples).

102

Figure 19. Linear elastic analysis with a 3D model of the
periodic cell of the cloister: (a) finite element mesh; (b) defor-
mation in axonometric view and cross section showing the
inwards movement; (c) maximum (tensile) principle stresses;
(d) minimum (compressive) principle stresses.
the periodic structure of the cloister were considered,
namely a three-dimensional model and a plane stress
model.
The results shown above from the survey and testing
were adopted to define the geometry, constituents and
properties. Different materials were used for the vaults,
walls and infills. For the actions, only the self-weight
of the structure was considered.
6.1 Three-dimensional model
Figure 19 shows the results of the periodic part of the
cloister using linear elastic analysis. Appropriate peri-
odic boundary conditions have been added along the
longitudinal direction, whereas no boundaries were
added in the transverse direction, as a lower bound
representation of central part of the South wing, see
Lourenço et al. (2000) for details.
Twenty-noded elements with quadratic interpola-
tion were used in the model (Fig. 19a) and the obtained
deformation for the self-weight indicates a movements
inwards to the court of the cloister, as observed in the
actual structure (Fig. 19b). Thus, the structure seems to
presents insufficient buttressing in the internal walls.
The maximum tensile principal stresses are found
at the key of the barrel vault of the 2nd level, at the
key of the cross vault of the 1st level and at the key
of the door arch in the 1st level. The maximum value
of the principal stresses are +0.25 MPa and – 0.6 MPa,
which are relatively moderate values.
103
Figure 20. Adopted 2D model for the 2nd level: (a)
thickness in different parts; (b) deformation and boundary
conditions.
6.2 Two-dimensional model
In order to better understand the structure and its dam-
age, a plane stress model has been used for the barrel
vault, calibrated with the 3D model. Figure 20 shows
the thickness adopted and the sliding clamping sup-
ports considered for the inner wall. It is noted that
four different boundary conditions of the inner wall
have been tested and this has a major impact on the
vertical displacement of the key of the barrel vault,
on the maximum horizontal displacement and on the
maximum tensile stress.
A non-linear analysis has been made with the
2D model, incorporating interface elements for the
deformability of the soil and a maximum tensile
strength of 0.1 MPa. Figure 21 presents the results
in terms of load-displacement diagram and maxi-
mum tensile strains (as damage indicators). The model
replicates the most significant damage observed in
the structure, including separation between vault and
walls, and longitudinal cracking in the vault.
Also the model predicts no remaining capacity of
the structure in terms of additional vertical load. It
is stressed that this statement holds only in the case
of the weak foundation found. If the foundations are
assumed as rigid, the ultimate load of the structure
increases considerably.
6.3 Conclusions
The conclusions of the numerical analyses together
with the inspection allowed to conclude that: (a) the
non-symmetry between the internal and external walls
of the cloister result in inwards movements in the direc-
tion of the court, as observed in the construction; (b) a
linear elastic analysis of the construction results in
very limited displacements (in the order of one mil-
limeter) and moderate stress values (maximum tensile
stress of +0.25 MPa and maximum compressive stress
of −0.6 MPa). The large displacements observed in
the construction require a geometrical and physical
non-linear analysis; (c) in order to obtain horizontal

Figure 21. Results from non-linear analysis: (a) load-
displacement diagram; (b) separation between vault and
wall; (c) cracking of vault.
displacements of magnitude comparable to the values
observed in the structure, it is necessary to consider the
soil-structure interaction. It seems therefore that the
foundations play a key-role in the observed damage;
(d) the large movements recorded in the construction
and the deterioration of the brick vaults indicate that
the safety level of the structure is not compatible with
any use and immediate intervention was necessary.
7 REMEDIAL WORKS
The cloister required consolidations works and the
proposed solution included repositioning the walls in
plumb, elevation/re-centring vaults and arches, and
hidden tying of the walls as an additional strength-
ening. This strategy resulted from the inspection,
diagnosis and safety assessment, and from the previous
104
experience of the stone master in charge of the works
(Humberto Reis de Sousa), as a joint decision by the
authors and the technicians in charge of the monu-
ment (ArchitectsAngela Melo and Jorge da Costa from
the Cultural Property Service, Porto). The option not
to intervene in the foundations was made from the
beginning, as: (a) the intervention would need to be
very invasive; (b) it would lead to the destruction of
the buried remainings; (c) the authors believe an inter-
vention in the superstructure is sufficient to stabilise
the structure. In the modern spirit of a step-by-step
minimal intervention, the owner was alerted to the
fact that a (possible, but unlikely) intervention in the
foundations might be required in the future.
The viability of the proposed works depended on the
possibility of cracked masonry to accommodate move-
ments and the technical capability of the contractor, as
the structures would be moved but not dismounted.
The operation entails some risk due to the precari-
ous stability, significant weight and non-monolithic
behaviour of part of the structure. The operation was
made possible only by the careful execution of the
stone mason Humberto Reis de Sousa, which knew
how to straighten and move structures walls using
hydraulic jacks, cable tensioning tools and adjustable
props.
Figures 22 and 23 illustrate the proposed struc-
tural remedial measures. All metallic elements are in
stainless steel AISI 316, which provides the highest
corrosion resistance. The ties are applied only in the
wings that exhibit larger damage, possibly due to the
lack of external transverse walls: South and East, and
South and West. The ties are placed in the vault infill,
meaning that the horizontal thrust from the vault is not
aligned with the tie. This non-alignment produces a
bending moment, which is balanced by vertical stitch-
ing (or reinforcement) for the 2nd level and an uneven
vertical distribution of stresses for the 1st level. This
uneven distribution of stresses, which does not pro-
voke any tensile stresses, is possible due to the weight
of the upper structure.
It is noted that other possible solutions were ini-
tially considered for the 2nd level, and a final solution
remained to be decided during execution and clarifi-
cation of the composition of the vaults. It was found
that the masonry infill of the vaults from the 2nd level
was extremely hard and impossible to remove with-
out significant loss of fabric and possible damage to
the structure. This contributed to the conception of the
adopted solution for the 2nd level.
All ties possess a coupling element capable of
adjusting the tie. In the 2nd level, the ties are connected
to a vertical bar inserted in a fabric sock, capable
of containing the injected grout. As the masonry in
the internal walls is made of large ashlars, additional
anchorage is provided with a transverse element. For
the internal walls, sufficient bond occurs in the con-
tact with the irregular masonry. In the 1st level, vertical
 Figure 23. Consolidation works for 1st level: (a) plan with
location of ties; (b) transverse section view.

Figure 22. Consolidation works for 2nd level: (a) plan with
location of ties; (b) transverse section view; (c) details.

bars are not needed, and the ties are directly anchored
to the walls at a 30◦ angle.
Figure 24 shows images of the preparatory works
for protection and for propping and lifting the vaults.
Details on the installation of the ties are shown in Figure 24. Preparatory works: (a) protection of floor and
Figure 25 and several complementary works are shown walls; (b) adjustable propping for 2nd level; (c) adjustable
in Figure 26. propping for 1st level.

105
Figure 25. Execution of consolidation: (a, b) removal of
infill from wings in 2nd and 1st level; (c) openings to remove
debris and lift barrel vaults; (d) coring for vertical reinforce-
ment and tie anchorage in 2nd and 1st level; (e, f) installing
ties, repairing brackets and final view in 2nd and 1st level
ceiling.

Besides the works shown, protection against rain-

water infiltration and drainage of rainwater were also
carried out. A PVC membrane was installed in the 2nd
level roof and new gargoyles were designed. The final
aspect of the cloister, with the exception of the 2nd
level roof is of the previous untouched antique.
Figure 26. Complementary works: (a) mortar color adjust-
ment; (b) view of the completed external wall of South wing;
(c) buried remainings found; (d) local consolidation with new
masonry parts; (e,f) view of 2nd and 1st level floor.

106
Figure 27. The database can be browsed through: (a) menus; (b) photographs, 3D models, panoramic views; (c) hotspots.
8 INFORMATION MANAGEMENT TOOL
Anyone involved in the conservation of cultural her-
itage buildings is aware of the enormous amount
of information generated by the different specialists,
which is generally not handled adequately by owners
and/or authorities involved in the process. The conse-
quence is that valuable information is lost in a complex
process of reaching a decision that involves many dif-
ferent experts and information tends to get forgotten
or misplaced in the course of time.
Upon the recognition of this evidence, University
of Minho (UMinho) in partnership with Centre of
Computer Graphics (CCG) and the former Portuguese
Institute of Architectural Heritage (IPPAR) developed
an application for the efficient management and visu-
alization of the information related to interventions
of cultural heritage buildings. The development was
based in a case study with extensive damage and that
recently suffered major works in a cloister, namely
the Monastery of Santa Maria de Salzedas (Portugal).
A database was created for the management of all the
information generated during the intervention process
of 2005 and the historical information collected which
has been produced along the time and has been selected
by different areas of study. The information is acces-
sible via intranet for practitioners and, planned soon,
to be available on the monastery for visitors, in way
that a non-specialist and a team specialist can obtain
information with different contents.
8.1 Internet as a tool for information
management
Internet is an excellent tool that fulfils the ICOMOS
(1996) principles for the management, dissemination
and sharing of records, namely: (a) the original elec-
tronic version can be safely archived; (b) backups are
easily made and can be stored in a different and safe
107
place; (c) updated records can be easily incorporate;
(d) the format of the records can be standardized and
indexed; (e) information is accessible 24 hours a day,
year round, from any point of the world; (f) information
can be published, disseminated and be easily accessi-
ble and/or restricted depending on the classification of
the information and the user; (g) the information can
be easily presented in different languages in real time
allowing its dissemination around the world.
8.2 Structure of the database
The information about the Monastery is presented
using virtual representation (VR), i.e. through mul-
timedia tools, the users will have access to all infor-
mation stored in the database.
The application, and the database, is structured in
three main groups, which define the way in which
the information can be accessed, namely spatial units,
level of access and thematic areas. The spatial units
should be understood as the different areas in which
the Monastery has been divided. The level of access
defines the privileges of the user. Thematic areas are
the options for searching the information contained in
the database. The database can be browsed through
menus or hotspots in photographs, 360◦ panoramic
views or 3D model (Fig. 27). The thematic areas
depend on the level of access. A tree-like structure
was thought for the thematic areas (menus and sub-
menus), in which each one can be subdivided in three
different levels.
The main menus or thematic areas are shown in
a fixed top bar, while the submenus are displayed at
the top-left of the screen. The bottom-left area dis-
plays the image with hotspots. The hotspots link with
some particular information or display files stored in
the database. The information is displayed in the main
window. The information is presented as image ele-
ments (pictures, drawings, movies, panoramic views,

Figure 28. Typical view of the database.
3D models, etc.) in the top, with text elements in the
bottom.
The application has been developed in two dif-
ferent parts, including a Backoffice for inserting the
information gathered and the Front End, which is the
part visible to users as webpages automatically gener-
ated from the backoffice. Both parts can be accessed
remotely via Internet connection.
Besides the tree menu structure, there is a search
bar to find any particular file stored in the database.
Figure 28 shows a typical view of the navigation win-
dow of the database. This window shows the same
fields as the file form associated with each file in the
database. The download option will be turn on only
if the user has at least the same rights as the file has
(level of access).
9 CONCLUSIONS
A comprehensive program involving inspection, diag-
nosis, safety assessment and remedial measures of a
Cistercian cloister was presented. The cloister was in
very bad structural conditions and the consolidations
works aimed at stopping further degradation and at
preventing collapse. With the objective of keeping the
abandoned / ruined condition of the cloister, all works
have been hidden, while ample information is available
to document the intervention.
These tasks are complemented by a virtual recon-
struction of the medieval monastery and an informa-
tion management tool, with the objective of providing
simple technical information to the visitor and of man-
aging the significant amount of expert documentation
gathered in the process.
The application is based on the creation of a
database that makes possible the efficient manage-
ment of all information related to past, present and
future interventions of the Monastery, in way that it is
easily accessible by anyone.
108
ACKNOWLEDGEMENTS
The success of the works addressed in the paper is the
result of teamwork. In particular, the remedial works
could not have been carried out with the daily super-
vision of architects Ângela Melo and Jorge da Costa,
the careful execution of stone mason Humberto Reis de
Sousa, and the valuable discussions between all actors
and the joint decision process adopted, from the early
process of design.
The information management tool and the virtual
reconstruction of the monastery presented here are part
of the project POCTI-HEC-60431-2004, Integrated
approach for conservation and valuation of monu-
ments, funded by FCT (Portuguese Science and Tech-
nology Foundation). F. Peña acknowledges funding
from the FCT grant contract SFRH/BPD/17449/2004.
REFERENCES
Amado, M., Lourenço, P.B. & Peña, F. 2006. A three
dimensional model for Monastery of Santa Maria in the
Medieval Age (in Portuguese). Guimarães: University of
Minho.
Alves, C.A.S. & Pamplona, J.M.V. 2006. Survey and prelimi-
nary inspection of the cloister in Monastery of Salzedas –
Stone deterioration study (in Portuguese). Guimarães:
University of Minho.
Alves, C. & Pamplona, J. 2007. Salts in the church and
monastery of Salzedas (Viseu, Northern Portugal) (in Por-
tuguese). In Proceedings of the VI Iberian Congress of
Geochemistry; Proc. intern. symp., Vila Real, 16–21 July
2007: 194–197. Vila Real: Univ. de Trás-os-Montes e Alto
Douro.
Cocheril, M. 1978. Guide of the CistercianAbbeys in Portugal
(in French). Lisbon: Calouste Gulbenkian Foundation.
ICOMOS 1996. Principles for the recording of mon-
uments, groups of building and sites. International
Council of Monuments and Sites. Available from
http://www.international.icomos.org
Leitão, A. 1963. The monastery of Salzedas (in Portuguese).
Porto: Author’s edition.
Leroux-Dhuys, J.F. 2006. Cistercian abbeys: history and
architecture. Columbia: Konemann.
Lourenço, P.B., Vasconcelos, G., Martins, J.B., Ramos, L. &
Jalali, S. 2000. Diagnosis of the stability conditions of
the XVII century cloister of the Monastery of Salzedas, in
Tarouca (in Portuguese). Guimarães: University of Minho.
Virgolino, J.F. 1997. Architecture, measure and number in
the Cistercian church of S. Joao de Tarouca (Portugal) (in
Portuguese). Cistercium 1997: 431–456.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Cultural Heritage protection against flood – A European FP6

research project

Ch. Maierhofer, Ch. Köpp & S. Kruschwitz

Bundesanstalt für Materialforschung und -prüfung (BAM), Berlin, Germany

M. Drdacky
Institute of Theoretical and Applied Mechanics ASCR v.v.i, Prague, Czech Republic

Ch. Hennen
Institut für Diagnostik und Konservierung an Denkmalen in Sachsen und Sachsen-Anhalt e.V., Dresden, Germany

S. Lanza
University of Genova, Italy

M. Tomaževič
Slovenia National Building & Civil Engineering Institute, Ljubljana, Slovenia

D. Kolymbas
University of Innsbruck, Austria

I. Herle
Technische Universität Dresden, Germany

A. Virsta
UNESCO Chair ECOLAND, University of Bucharest, Romania

L. Binda
Politecnico di Milano, Italy

P. Askew
Industrial Microbiological Services Ltd., UK

ABSTRACT: For the protection of Cultural Heritage from environmental damage and in particular against
flood, a research project has been proposed under the 6th European Framework Program to support policy-
oriented research. The main aim is the development of preventive conservation strategies and the identification
of appropriate measures. Clear recommendations for mitigating or removing damage from Cultural Heritage
objects will be provided through the project results. CHEF proposes the integration of multidisciplinary research
as scientific support to European policies. In this frame the necessary technological basis and cost efficient and
effective tools for the development of new and innovative strategies will be provided.

1 INTRODUCTION
The European Community has suffered from disas-
trous floods in recent years, which caused enormous
damages (compare Fig. 1) and left hundreds of peo-
ples dead in several European countries. The flooding
occurred across regional borders and involved neigh-
bouring states along the rivers Elbe (2002), Odra
(1997), Rhine (1993 and 1995) and many more. Floods
can not be prevented. Europe will have to face further
flood catastrophes due to the change of climate and due
to further building activities in flood-prone regions [1]
[2] [3].
Although very comprehensive, projects, strate-
gies and further administrative measures, which
started after the recent flood events, do not

109
Figure 1. Historic power station in Prague (CZ) during flood
in 2002. Photo: V. Herle.
include the particularities of movable and immov-
able Cultural Heritage like objects in museums and
libraries, historic buildings, parks and sites (his-
toric cities), which are threatened by flood catas-
trophes. For protecting the common European Cul-
tural Heritage against flood hazards and environ-
mental conditions related to flood, a focus has to
be put on the mapping of the above mentioned
risk-areas including the objects worthy of protection.
It is required to assign a high protection level to par-
ticular monuments or to areas where valuable historic
sites are located [4].
This decision requires a lot of knowledge and expe-
riences, because a multitude of reasons can have an
influence on the characterisation and classification of
these objects. The aspects which have to be consid-
ered vary from historic significance and context of
the object, building structure or landside to techni-
cal problems like lack of documentation, unknown
structural condition and assembly, unknown material
properties and characteristics or unknown parameters
of exposure.
Thus, a research project titled Cultural Heritage
Protection against Flood (CHEF) has been proposed
under the 6th European Framework Program to sup-
port policy-oriented research. The main goal of the
CHEF project is avoiding or mitigating the damage
of movable and immovable Cultural Heritage related
to flood by considering preventive and emergency
measures, damage evaluation methodologies, as well
as restoration and repair techniques. The project will
bring together existing information and experiences
regarding flood problems and contains innovative
aspects based on experimental and theoretical research
activities in the laboratory and on-site.
The project CHEF contributes to the call area 8.1
B.3.6 The protection of Cultural Heritage and associ-
ated conservation strategies, Task 4: To help protect
110
Figure 2. Observed frequency of flood events in the historic
centre of Genova/Italy [5].
Cultural Heritage from environmental damage partic-
ularly in relation to flooding: to develop preventive
conservation strategies, to identify appropriate mea-
sures and to provide clear recommendations for mit-
igating or removing damages from Cultural Heritage
objects, particularly in relation to flooding; policy-
oriented research, Scientific Support to Policies (SSP)
by considering aspects varying from historic signifi-
cance and context of the object, building structure or
landside to technical problems like lack of documen-
tation, unknown structural condition and assembly,
unknown material properties and characteristics and
unknown parameters of exposure.
2 PROJECT OBJECTIVES AND POTENTIAL
IMPACT
As flooding is a very complex phenomenon, a
project, targeting at investigating flood mechanisms
and related issues, has to reach for ambitious goals.
Looking at the work plan of CHEF, it appears that
the project can be regarded as a series of interwoven
tasks and activities, which try to bring out the big pic-
ture and integrate many different aspects of flooding
and Cultural Heritage protection. In the following an
overview about the main objectives, each reflected also
by a designated work package, is given:
1 Classification of movable and immovable Cultural
Heritage according to their vulnerability and sensi-
tivity to flood
a. Analysis of previous floods and related environ-
mental hazards, quantification of their character-
istics in terms of suitable hydraulic parameters
b. Evaluation of results of related projects
 c. Analysis of exposure by means of adequate
survey
d. Risk assessment of Cultural Heritage based on
quantified hazards, exposure maps and GIS.
2 Analysis and classification of damage processes
in different materials (soil, brick, stone, wood,
metal, textile, paper, etc), structures (buildings,
bridges) and sites (cities, gardens, archaeological
sites) induced by flood and flood related environ-
mental impacts in close cooperation with museums
and further stakeholders. This includes:
a. Set-up of a damage catalogue
b. Investigation of the influence of water, salt and
biodegradation on material properties
c. Comparison to results of modelling.
3 Validation and evaluation of methods and sen-
sors for non-destructive testing and monitoring of
material and structural parameters, environmental
technologies will have to be provided.
4 Definition of threshold levels for exposure and dam-
age before restoration, also in relation to the use of
the objects/buildings.
5 Analysis of preventive measures before flood and
temporary (emergency) measures during and soon
after flood for movable and immovable Cultural Her-
itage. The influence of these measures on material
and structure will be adressed.
6 Assessment of restoration and repair techniques
after flood.
7 Assessment of running and new case studies on dif-
ferent objects concerning infrastructure, buildings,
parks, cities.
8 Generation of practical recommendations by defini-
tion of strategies for the protection, conservation and
repair of movable and immovable Cultural Heritage.
The potential impact of the project on the conser-
vation of Cultural Heritage is direct and transparent:
The strategies to be developed in the project can help
to make protection and repair of Cultural Heritage
against flood and related hazards far more effective
and allow carrying out these measures on a large scale.
With a reasonable management of Cultural Heritage
sites, whole neighbourhoods can be prevented from
deterioration. Another aspect is the understanding of
damage processes not only induced by flood but also
by other aggressive environmental conditions, such as
air pollution, microclimate, earthquakes, traffic vibra-
tions, etc. This knowledge will help to take precautions
against it and will thus reduce the impact of environ-
mental damage on historic structures. But also social
problems like unemployment, equal opportunities and
health and safety of citizens will be positively influ-
enced by CHEF. The project results will be dissem-
inated by publications, project reports, contacts with
end-users and mainly by publishing recommendations
111
Figure 3. Investigation of the moisture distribution at Vel-
trusy Castle (Czech Republic) after the Elbe 2002 flood with
radar, detecting damage at a balcony.
which will be practical and adapted to the needs of
a broad array of institutions, operational services and
specialists in the public sector dealing with civil pro-
tection, land development, regional planning and risk
management.
3 WORK PLAN
The scientific and technological approach aims at
identifying characteristic parameters for flood and
flood related hazards, which have an impact on mov-
able and immovable Cultural Heritage depending on
their vulnerability (compare Fig. 3 and 4). The under-
standing of damage processes in materials and struc-
tures by using environmental technologies will provide
the basis for evaluation and further developing strate-
gies for preventing, mitigating and repairing damages
at Cultural Heritage related to the objectives and poli-
cies of the European Community. It is planned to
implement a project with a clear multidisciplinary
approach as an obvious reaction upon the requirements
described in task 4 of the above-mentioned call.
The research will be carried out through eight work
packages. These work packages and their interaction
will provide a clear progress beyond the current state
of the art:
• WP1: Identification of typical environmental haz-
ards related to flood and being decisive to Cultural
Heritage. Here, flood mechanisms and hydrology
will be one important aspect. Movable and immov-
able Cultural Heritage will be classified according
to its sensitivity. This is going to be investigated in
the first of the work packages and serves mainly
as an input for the measures, strategies and meth-
ods, which are worked on in detail in the subsequent

Figure 4. Determination of the moisture distribution at Vel-
trusy Castle (Czech Republic) with radar and microwave
borehole method.
work packages. This part of the project will also
provide an overview about existing plans for protec-
tion of Cultural Heritage objects and will examine
the experiences (historic and recent) made during
floods especially in Europe. The threat posed at
Cultural Heritage by flooding is quantified in this
subtask, based on the analysis of existing litera-
ture and archived documents. The selection of the
best-suited technical investigation tools is also
addressed, with reference to the hydraulic and hydro-
logic models available and their interpretation in
view of Cultural Heritage vulnerability assessment.
The work of WP 1 will be carried out by the
institutions ITAM, CRUIE and IDK1 .
• WP2: Damage analysis of different materials and
movable Cultural Heritage, their properties and their
interaction with moisture, salt, pollution and other
phenomena related to flooding catastrophes. In this
context a survey of possible flood-related damage, a
validation of non-destructive and minor-destructive
methods for damage detection and quantification, a
classification of damage and the definition of dam-
age threshold will be given. The direct contact of
water and contaminants occurring during the flood
as well as the modified environmental conditions
after the flood leads to various damages and/or con-
taminations of soil, building materials and movable
heritage objects. These damages and modifications
strongly depend on the flood and environmental
parameters (duration of flood, height of water, flow
rate, contaminations, temperature, etc), on the type
1
The abbreviation and full names of the involved institutions
will be explained in the next section.
112
of material (brick, stone, metal, wood, paper, textile,
etc), on the material parameters (porosity, grain size,
density, etc) and the structure and conditions of the
objects. Therefore, the damages are manifold and
gained knowledge from recent floods and a damage
classification is urgently required.
The work of WP 2 will be carried out by the
institutions BAM, TUD and ISML.
• WP3: Classification of damage on historic struc-
tures and sites including buildings, infrastructure
(e.g. bridges) and cities for understanding the mech-
anisms of damage. Static and dynamic loading,
moisture and salt transport, contamination and ero-
sion problems will be considered as well as the
vulnerability of whole structures and sites. Inves-
tigations will be based on previous experience, new
and existing data recorded from on-site investiga-
tions, monitoring and laboratory tests and numerical
simulations. Existing databases and past experience
will be analysed and damages as they belong to
types of buildings or structures and to sites will be
summarized related to different materials. A damage
catalogue containing exposure data and applicabil-
ity tips for remedial works will be created. Damage
as well as protective measures will be analysed in
order to identify the most urgent problems for further
more detailed studies. Previously defined vulnera-
bility indicators will be validated (element-at-risk,
susceptibility and exposure). A cooperation with the
User Group will be mobilised and exploited.
The work of WP 3 will be carried out by the
institutions ITAM, POLIMI and IDK.
• WP4: According to the long term experience the best
preventive measure for movable heritage objects
is their replacement from endangered localities to
safe storage. Such a simple procedure failed in
many recent cases and this sub-task is suggested
to analyse the reasons and to prepare more robust
and reliable approaches. Therefore, its main focus
will be not only on administrative measures but
it will involve also technical measures as well as
critical evaluation of methods and possible needed
improvements. Preventive measures for immovable
heritage are designed according to the classifica-
tion of individual objects at risk. Conventionally,
both “structural” and “non-structural” interventions
are addressed in flood protection related issues; the
latter ones include both macro-structural interven-
tions, when involving interventions along the river
network, and micro-structural ones, when limited
to the passive protection of the resources located
on the flood prone territory. The management of
objects and sites is usually ensured by different sub-
jects and also the measures and their impacts may
differ quite substantially. Therefore, the objects are
treated separately from site to site. Temporary emer-
gency measures prepared before the flood and their
 effectiveness will be studied considering technical
measures. This will encompass temporary barriers
as well as construction measures for stabilization
of the whole building or strengthening of structures
based on knowledge of building condition and prob-
able exposure parameters. For restoration and repair,
post-flooding short term techniques (drying and
dehumidification, reducing temperature, after flood
care etc.) as well as long term techniques (pollution
removal, salt removal, desalination etc.) for repair of
damage of historic building materials, which require
a very distinct and careful treatment (paintings,
frescoes, original building materials) shall be col-
lected and critically analyzed. To correctly perform
a restoration activity after catastrophic events, it is
highly recommended to execute a careful diagnosis
on the damaged works of art.
The work of WP 4 will be carried out by the
institutions ITAM and ZAG.
• WP5: Assessment of running and new case stud-
ies on different objects concerning infrastructure,
buildings, parks, cities. Two to three cases will be
studied in detail concerning a multitude of aspects
like a small castle surrounded by a park and con-
taining a museum. Full-scale models in the shape
of small buildings will be also used as intermediate
case between laboratory and real buildings. It will be
studied how historic floods occurred and have geo-
graphically spread. Former concepts and methods
used to restore the affected case studies are going to
be evaluated. An analysis of previous and running
case studies carried out in similar projects will be
part of this work package in order to get a wider
base for the development of recommendations and
guidelines.
The work of WP 5 will be carried out by the insti-
tutions ECOLA ND, ZAG, ITAM, IDK, CRUIE and
POLIMI.
• WP6: Development of guidelines and recommen-
dations for the strategies concerning the assessment
of vulnerability of Cultural Heritage against flood,
the prevention and mitigation of damage, the emer-
gency and the medium and the establishment of long
term post-flooding action plans is the main sub-
ject of this work package. The collected damages
(damage catalogue) as well as the case studies will
be included in existing data bases. A development
of strategies for the protection and conservation of
buildings, historic areas, movable Cultural Heritage
objects which bear the risk of flooding is taking
place in close collaboration with the end-users.
Meetings during the project will serve as knowl-
edge exchange between consortium members and
end-users. Towards the end of the project, an open
workshop will be organised.
The work of WP 6 will be carried out by the
institutions ITAM, CRUIE and IDK.
113
• WP7:The work in the project has to be framed within
a strict project management. This work package
includes the administrative and the technical coor-
dination of the project. The coordinator is supported
by the WP-Leaders, who take on the responsibility
for the technical supervision in their work packages.
The work of WP 7 will be carried out by BAM.
• WP8: The planning of dissemination activities will
be carried out during the project under the leader-
ship of the Exploitation Manager with involvement
of WP-Leaders and User Group members. This
task includes the dissemination of non confidential
information about the project objectives and work
plan, coming events (seminars), published papers,
demonstration actions, training courses, lectures at
universities and especially the dissemination of the
final practical recommendations.
The work of WP 8 will be carried out by the
institutions ITAM, BAM and IDK
4 PARTNERS AND USER GROUP
The consortium includes members from 7 different
European countries:
1. Bundesanstalt für Materialforschung und – prü-
fung, Berlin, Germany (BAM)
2. Institute of Theoretical and Applied Mechan-
ics ITAM ASCR v.v.i., Prague, Czech Republic
(ITAM)
3. Institut für Diagnostik und Konservierung an
Denkmalen in Sachsen und Sachsen-Anhalt e.V.,
Dresden, Germany (IDK)
4. University of Genova, Italy (CRUIE)
5. Slovenia National Building & Civil Engineering
Institute, Ljubljana, Slovenia (ZAG)
6. University of Innsbruck, Austria (UIBK)
7. Technische Universität Dresden, Germany (TUD)
8. UNESCO Chair ECOLAND, University of
Bucharest, Romania (ECOLAND)
9. Department of Structural Engineering (DIS),
Politecnico di Milano, Italy (POLIMI)
10. Industrial Microbiological Services Ltd., United
Kingdom (IMSL)
These institutes provide a lot of experience on dif-
ferent skills related to flood protection of Cultural
Heritage:
• Experience with problems related to civil and struc-
tural engineering with a strong focus on historic
materials and structures
• Development of methods for testing and monitoring
of different parameters of materials, structures and
objects with only minor invasion
• Measuring climate and climate control of historic
buildings

Figure 5. CHEF Consortium members during the Kick-Off
Meeting on 26th of February 2007 in Brussels.
• Characterization of historic building material and
successful implementation of modern material in
historic structures and systems
• Numerical modelling, laboratory determination and
experimental verification of transient hygrothermal
transport processes in materials and building com-
ponents; physico-chemical modelling (influence
of salt transport and crystallisation) and physico-
biological modelling (conditions for mould and
algae growth) connected with deterioration and
aging effects of the building substance
• Investigation of continuous drying out processes
• Structural assessment of historic building
• Structural modelling
• Town and urban planning and engineering of the
territory
• Safeguard of the historic-architectonic heritage of
city centres against the hydraulic and hydrologi-
cal hazard, vulnerability assessment and mitigation
initiatives
• Water and soil protection engineering, hydrology
and water resources management
• Mechanical behaviour of soil, theoretical modelling
(theory of hypoplasticity) and experimental deter-
mination of material parameters for soil models.
The balance between old EU member countries and
new EU members together with associated states is
especially adapted to the experience gained with floods
from rivers and costs in neighbouring regions: Czech
Republic, Slovenia, Romania, Austria, Italy, United
Kingdom and Germany. These countries were heavily
affected by flood hazards during the last 10 years.
To involve the public and to raise awareness it is
important to plan dissemination actions, which target
at many different groups. It is necessary to present
the project and its objectives, results and approach in
an easily understandable and straightforward way. To
keep the balance between scientific approach and pub-
lic acceptance is a difficult task, but the experienced
consortium in cooperation with the User Group will
be able to manage this.
114
It is envisaged that another major opportunity to
promote the results obtained from the proposed project
will be in the form of advice to local authorities,
historic monument protection trusts and associations,
national governments and the EU in the field of spa-
tial development, specifically sustainable protection
of historic monuments.
This in turn should serve to put future EU policies
on protection of historic monuments onto an even more
widely respected scientific footing both in Europe and
world-wide.
The transfer of the research results will be in the
form of reports, scientific papers in journals, presen-
tations at international conferences, case study fact
sheets, patents on novel technologies developed within
the proposed project and an Internet Website. Other
possibilities will be hands-on demonstrations and
courses and the inclusion of the results in regulations
and guidelines.
In the universities presented by the partners, the
results of the research will be exploited through
workshops and continuing education courses for the
students.
Several dissemination activities are targeting at a
broad audience and are meant to demonstrate the
transformation of scientific research into a practical
context.
All these actions have the following objectives:
• Exploitation of project results
• Information of possible users
• Information of public
• Integration of related projects
• Involvement of other organisations
• Transnational networking
• Future education of staff.
Dissemination activities will be organised by the
Exploitation Manager in close co-operation with the
coordinator and those project partners with close ties
to end-user, managers or government or local officials
responsible for operating and maintaining historic
buildings. In that context the User Group plays an
important role for the integration of organisations and
other projects related to the maintenance and care of
Cultural Heritage buildings.
The consortium of CHEF has already contacted sev-
eral national trusts, authorities, companies, museums,
universities in order to establish a large User Group
for dissemination activities. Many institutions all over
Europe have expressed their interest in the project and
their willingness to support the activities.
5 FIRST RESULTS
Since the official project started in February 2007,
some results are already available.
5.1 Report on experiences from floods and includes short comments on their content and main
results.
As an overview about the whole complex topic, it was
The European Commission has launched several
necessary to perform a state of the art survey and to
international research projects on natural hazards
find out, which measures have been taken in the past
affecting Cultural Heritage. These projects are valu-
to prevent or reduce the impact of natural flooding
able for the CHEF project because most of them
disasters and to survey possible repair interventions
European relevance and provide highly objective infor-
that have proven to be suitable for the restoration of
mation. To take advantage of the synergies of similar
valuable buildings after the flood.
research projects it is necessary to use the contacts that
Several factors are influencing Cultural Heritage
have been established within many networks, espe-
and the losses due to flood and it is important to
cially with end-users, who are the most important
develop all-embracing strategies taking into account
addressees for the project results.
not only technical measures against flood but to con-
It is a major issue that during the next project years
sider also the “human factor” and other aspects that
the listed contacts with the leaders and partners of the
bare a threat to vulnerable buildings.
related projects will be activated in order to establish
An evaluation of countermeasures against flood
a common scheme for sharing experience and to col-
damage has to start with an analysis of former inci-
laborate in filling the gaps presently existing in our
dents. In many cases data are not available but still it
knowledge concerning protection of Cultural Heritage
is possible to draw valuable conclusions from reports
from flood effects.
and recordings of earlier catastrophies. Even the com-
Examples of very prominent international projects
parison of old pictures or photographs that show, how
related to damage of flooding events are NOAH’S
a Cultural Heritage building is located in its natural
ARK [6], RIMAX [7], and ELLA [8]. The objectives
surrounding allows to identify some of the relevant
of NOAH’S ARK (Global Climate Change Impact
characteristic parameters of flooding.
on Built Heritage and Cultural Landscapes, running
The “Report on experiences from previous and his-
from 2004–2007) were to determine the meteorologi-
toric floods, summary about the characteristic param-
cal and climate change parameters that are most critical
eters of these events” was worked out by the University
to the built cultural heritage. Moreover to research
of Genoa, Italy under the leadership of Prof. Luca
and address future expected damages and dissemi-
Lanza and with the contributions of other project part-
nate optimum adaption strategies. The RIMAX (Risk
ners. It provides a summary of the experiences gained
Management of Extreme Flood Events) project aims at
from major floods occurred in the past and their syn-
developing and implementing improved instruments
thetic descriptive parameters, which may help in future
of flood risk management by the integration of dif-
assessment of the severity of floods in relation to Cul-
ferent disciplines and several participants. It focuses
tural Heritage. The main types of floods are presented,
on extreme flood events in river basins which means
and typical problems are illustrated by means of recent
events with a return period of more than a 100 years and
and past examples related to the impact on Cultural
a highly destructive potential. ELLA (Preventive flood
Heritage.
management measures by transnational spatial plan-
The report is one of the agreed deliverables of the
ning, running 2004-2007) focusses on damage related
project and will be published after the next project
to flood events of the river Elbe in Germany. It is
meeting in April 2008.
worked towards trans-national strategies and regional
planning for flood control. This includes international
knowledge, experience and data exchange.
5.2 Interfaces to running Flood and Cultural
Heritage projects
Another deliverable of the project compiles a list 5.3 Literature compilation
of other projects, both national and international,
A large number of articles, publications, reports and
related to mitigation of damages from floods. This
websites have been collected and analyzed in order to
report was made by the Institute of Theoretical and
get a broad overview about the topic. Especially for
Applied Mechanics ITAM ASCR v.v.i., Prague, Czech
the recent floods in Germany and the Czech Repub-
Republic, led by Miloš Drdácký.
lic in 2002 a lot of valuable material is available. For
It provides basic information about other related
an effective analysis the material has been classified
running projects dealing with flooding, and on their
according to the following topics:
impact on Cultural Heritage. An analysis is made also
on other natural hazards, e.g. earthquake, landslides, • Recommendations on floods
subsidence etc. and their threat for Cultural Heritage. • Reports about floods – lessons learned
The report “Interfaces to running Flood & Cultural • Risk management of floods
Heritage projects” provides a list of identified projects • Damage of materials.

115
 This compilation allows the comparison of counter-
measures that have been taken in different European
countries to reduce the impact of flooding. The reports
about various regional flooding events show that the
visible and unvisible damages and cosequences and
may vary significantly even if flooding occurs in the
same region at different times.

5.4 Template for damage at Cultural Heritage

objects/buildings
A questionnaire has been developed in the frame of the
project to identify frequent and typical damage at Cul-
tural Heritage objects or buildings. This questionnaire
will be distributed among end-users and is meant to
catch the users demand for investigation methods or for
desired restoration activities. The template addresses
owners or stakeholders of buildings and compiles
information about damage that has affected a struc-
tural element, a building or an object of historic value.
The record of damage includes meaningful pictures or Figure 6. Historic masonry specimen “Obelix”, featuring
illustrations and asks, in a first step, for the following a variety of characteristics of Cultural Heritage buildings.
criteria: Investigations with the microwave-borehole method to deter-
mine moisture profile.
• Building/Object/Structural element
• Origin of damage constructed at BAM. This specimen, called “Obelix”,
• Flood parameters features several characteristics of historic masonry
• Material properties and was built for the validation of investigation meth-
• Applied analysis methods: / Applicable analysis ods. Traditional manufacturing techniques and historic
methods materials were used to create a specimen, which is very
• Proposed method or activity for future protection
similar to real historic buildings. One of its features
• Proposed method for restoration.
is the possibility of defined moistening. “Obelix” is
Flood-related damage, a classification of damage standing in a basin, allowing the simulation of ascend-
and the definition of damage threshold is the output of ing moisture and it contains several tubes, which
this task. Based on the results of this survey, the CHEF enable a complete moisture penetration in some parts.
consortium will give recommendations and guidelines The specimen is particularly well suited for the
for specific risk minimizing actions, which can then investigation of moisture and salt transport and for the
be implemented in the contingency planning. development or testing of sensors.

5.5 Laboratory research 5.6 CHEF website

To gain more knowledge about materials and movable
Cultural Heritage, their properties and their inter-
action with moisture, salt, pollution and other phe-
nomena related to flooding catastrophes the building
materials itself are investigated. Experimental inves-
tigations will be combined with numerical modeling
of structural behavior as well as of moisture trans-
port. Together with deterministic models, probabilistic
models to study the building and material vulnerability
will be implemented. For experimental investigations,
several non-destructive and minor destructive testing
methods as well as sensors and equipment for monitor-
ing will be evaluated and validated for selected testing
problems.
In the frame of the FP5 project ONSITEFOR-
MASONRY [9] [10], a large specimen has been
A project website has been established (www.chef.
bam.de), offering an overview about the project and
its work packages. Anyone interested is invited to get
into contact with the consortium members and to get
informed about recent project activities.
6 MEETINGS
The success of the project will definitely be based
on the distribution of the project results to relevant
interested parties, like owners of Cultural Heritage
buildings, stakeholders or other institutions responsi-
ble for the preservation of monuments. The mitigation
strategies developed in the project are only valuable,
if they can be implemented in those networks, which

116
have the possibility to establish contingency plans.
Therefore it is very important to run the project with
participation of end-users and to address the public
as often as possible. It is planned to invite end-users
to most of the project meetings and to discuss the
usefulness and impact of the project results.
The Kick-Off Meeting of the CHEF Consortium
members took place on 26th of February 2007 in Brus-
sels. The work plan to achieve the ambitious goals has
been presented to the representatives of the European
Commission.
The 2nd Periodic Meeting of the project was orga-
nized as a two days event. The first day was scheduled
as a regular working meeting to exchange first results
and to plan further dissemination activities. On the sec-
ond day invited end-users were present and together
with the group an exchange of experiences, gained
during previous floods took place.
Presentations about flooding as a global phe-
nomenon, followed by impressive reports about recent
flooding events in Germany, Slovenia, Czech Republic
and the United Kingdom highlighted the importance of
the topic. The introduction of methods for assessment
and diagnosis of affected buildings and sites as well
as the presentation of treatments for damaged historic
materials contributed to the exchange of knowledge
between consortium and end-users. This was extended
by lectures about preventive flood management mea-
sures, flood management in historic cities, urban
development, open space planning, monument preser-
vation and strategies for risk minimization.
Another meeting, the 3rd Periodic Meeting is sched-
uled for the 14th and 15th of April 2008 and will be
held in Genoa, Italy. Again, besides a regular project
meeting, one day will be dedicated for the involvement
of end-users and interested public.
7 CONCLUSIONS AND OUTLOOK
The risk of flooding or other natural hazards has
increased significantly in recent decades due to cli-
matic change and building activities in risk-prone
areas. The European Community is aware of this situ-
ation and has raised programmes to support research
initiatives, which aim at prevention of natural disas-
ters and minimizing the impact on Cultural Heritage.
For the 7th Framework Programme of the European
Commission a Tentative Roadmap has been set up for
the area Environment including Cultural Heritage. The
research need covers the following topics:
• Managing environmental changes, hazards, dam-
age, mitigation and preventive conservation
• Protection of archaeological sites and cultural land-
scapes
• Fostering the integration of Cultural Heritage in the
rural/urban setting.
117
The proposed project CHEF already deals with top-
ics, which are important for the framework described
above. It can be a very good starting point for further
initiatives in this field. The large consortium and the
involvement of the User Group offer many possibili-
ties for networking. It is very likely that the consortium
or groups within this partnership will react upon future
calls for projects in FP7. The work plan of the project
described here can be further extended by including
more case studies or by concentrating on different
aspects of Cultural Heritage protection. There are sev-
eral starting points for new initiatives or for extended
co-operations with other groups or institutions dealing
with Cultural Heritage. This project will fulfil the pro-
posed work plan and provide the agreed deliverables
but should also be regarded as some kind of “think
tank” for future research.
ACKNOWLEDGEMENTS
The project CHEF is funded by the European Com-
mission in the 6th Framework Program. The presented
results have been achieved by all Partners of the
project, which are listed in Chapter 4.
REFERENCES
[1] An EU policy on flood risk management, http://
www.europa.eu.int/comm/environment/water/
flood_risk/index.htm
[2] Becker, A. and Grünewald, U. (2003) Flood risk in
Central Europe. Science Vol. 300, 16.5.2003, p. 1099.
[3] Dombrowsky, W.R. and Ohlendiek, L. (1998) Flood
Management in Germany. In Rosenthal, U., ‘t Hart, P.
(publisher): Flood Response and Crisis Management
in Western Europe. A Comparative Analysis, Berlin,
New York a.o. Springer, pp. 153–187.
[4] Drdácký, M. and Slížková, Z. Damages and Failures
of Historic Materials and Structures from Natural Haz-
ards, in: Proceedings on 2nd H & mH, International
Conference on Vulnerability of 20th Century Cultural
eritage to Hazards and Prevention Measures, Island of
Kos, Greece, 3–5 October 2005.
[5] Lanza, S. (2003). Flood HazardThreat on Cultural Her-
itage in the Town of Genoa. J. Cultural Heritage, 4(3),
pp. 159–167.
[6] http://noahsark.isac.cnr.it
[7] http://rimax-hochwasser.de/index.php.?id=430&L=1
[8] http://ella-interreg.org/kontakt.htnl?&L=1
[9] http://www.onsiteformasonry.bam.de
[10] European Commission – Community Research,
Project Report (2006) Onsiteformasonry project: On-
site investigation techniques for the structural evalu-
ation of historic masonry buildings. EUR 21696 EN.
ISBN 92-894-9601-0.
Definition of significance and attribution of value
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Structural definition and comparison of early medieval roof structures

C. Thelin
Tyréns AB, Växjö, Sweden
Chalmers University of Technology, Göteborg, Sweden

K. Linscott
Göteborg University, Göteborg, Sweden

ABSTRACT: Few roof structures remain unscathed from the 12th and early 13th centuries. The relatively
large amount of well-preserved early roof structures in churches in Sweden is used as a point of departure to
structurally define medieval Romanesque roof structures and to make comparisons within North-Western Europe.
Investigations of actual roof structures along with study of written sources provide the basis for the study. The
structures are defined and compared according to load-carrying system, included parts, centre distances, slope of
roof, connection with wall, joints, outer roofing and decorations. We found many structural similarities and the
structures seemingly belong to the same building historical context while also showing regional differences. The
early medieval roof structures of North-Western Europe can generally and concisely be described as common
rafter roofs with tie beams, without longitudinal bracing systems and having strut beams and collar beams in
different combinations to support the rafters.

1 INTRODUCTION Swedish roof structures originating in Romanesque

In Sweden there are a large number of well-preserved
roof structures from as early as the beginning of the
12th century. They represent an advanced building
technology of the time. There are few written sources
from the early Middle Ages to rely on, making us
today highly dependent on physical artefacts for under-
standing the society and culture of that period. The
medieval roof structures are an important and valuable
cultural heritage, which can inform people of today
of the development that occurred, the craftsmanship
that existed, the materials available, the techniques of
putting together and erecting buildings, knowledge of
structural design and the creation of shelter, and the
will to create spaces. The structures that remain rep-
resent a unique contribution to the European heritage
of medieval timber structures.
There are at least two important incentives to raise
knowledge of historic roof structures:To facilitate their
interpretation and preservation. Preservation involves
maintaining these structures for coming generations.
Interpretation concerns their use as sources of infor-
mation in order to obtain knowledge of the past.
Taking appropriate restoration measures and interpret-
ing them both require a proper understanding of the
structural principles applying to such structures and
the structural behaviour they display.
The aim of the paper is to present a synthesis
and structural definition of early medieval types of
architecture and make comparisons between countries
in North-Western Europe. Since historic timber roof
structures are working structures and often have com-
plex structural properties, an engineering approach
has been used in surveying them and analysis of
structural behaviour have been carried out. Their struc-
tural action and structural components have been used
for definition and classification of structural types.
Seventy actual roof structures provide the basis for
the study.
2 EARLY MEDIEVAL ROOF STRUCTURES IN
SWEDEN
The oldest remaining roof structures in Sweden are
found in churches. Maybe as many as 270 medieval
roof structures are still in place, mostly in southern
Sweden. Out of these, maybe one hundred are of
Romanesque type. However, the figures are not
completely reliable; no comprehensive inventory of
medieval roof constructions has been compiled in
Sweden. The majority of the medieval structures
have not been satisfyingly investigated or recorded,
probably some are not even identified yet. Iden-
tified and believed to be Swedish medieval roofs
were surveyed and described by Linscott (2007)
in a recent report. Linscott’s report is based on
an earlier listing and reports of medieval roof

121
structures (Sjömar, 1990, 1992, 1995, Thelin 2005)
and newer, unpublished, recordings and surveys of
about 40 church roofs. The information was put into
a database together with dendrochronological dating
(www.timmerdraget.org/php_takkons4) and the report
also refers to searches in this database. Roof structures
of the same type were raised during several hundred
years in Sweden. The oldest known roof structure is in
Herrestad church 1112+5d (Bartholin 1990). Inves-
tigation and dating so far show that as many as 24
medieval churches have roof structures that most likely
were raised in the 12th century.
Several varying types of early medieval roof struc-
tures are found in Sweden. The different types can be
divided into two major groups, which are found in two
geographical regions. In southern Sweden the builders
raised roof trusses with tie beams, struts and later collar
beams. In northern Sweden roof trusses with passing
braces and collar beams were raised instead (Sjömar &
Storsletten 1993). Only four roofs of this type remain
in Sweden. They have not been included in the defini-
tion of structural type below since they are a different
structural type even though they are contemporary. At
the same time, the roof structures from the 12th and the
13th centuries show many similarities. The tools that
were used are the same, axe and draw-knife “skave”
and the wood is prepared in the same way. The crafts-
men used a special hewing technique, “sprättäljning”,
whereby the axe cut back and forth along the side of
the log, which makes a very characteristic herringbone
pattern. Some surfaces are also very finely planed with
the draw-knife. The different parts in a roof truss are
also put together in the same way, with straight lap
joints. The builders, in both northern and southern
Sweden, had the same basic idea of how the wood
should be treated and how a roof should be put together
and raised. They seem to have belonged to the same
building tradition.
The very oldest remaining roof structures are tech-
nically fully developed and the craftsmanship is excel-
lent. The builders, in both northern and southern
Sweden, that raised the roof of a remote parish church
in the first half of 12th century mastered the materials
and the construction completely. The highly developed
skill of craftsmanship shown in the structures, the per-
manence of the building tradition over several hundred
years, and the spread of the same building tradition in
such a large geographical area indicate the existence
of an older (wood) building tradition in Sweden that
had existed for a long time before the 12th century.
Dendrochronology was developed in the 1980s in
Sweden (Bartholin, 1994). There are references to pine
and oak. From the dendrochronological dating we can
get a picture of how the roof structures developed over
time in Sweden, even if very few of the younger roof
structures, connected to Gothic architecture, have been
investigated. In principle, the builders changed during
122
the 14th century, from raising Romanesque roof struc-
tures to building roof structures without tie beams
adapted to high Gothic vaults. The two different sys-
tems were used in parallel for at least 150 years. In the
middle of this period, the second half of 14th century,
it seems as if very little was built at all, probably as a
result of the Black Death. The use of the special hewing
technique, “sprättäljning”, that is found in all the 12th-
and 13th-century structures terminates completely in
this period.
3 DEFINITION OF STRUCTURAL TYPE
In order to find and to define the characteristics of
the structural types used in the early Middle Ages in
Sweden 70 roof structures have been used.The name of
the church, which part of the church (nave or chancel)
the roof structures are standing on, their geographi-
cal location and the type of structure are presented in
Table 1 and in Figure 1. Of the 70 roof structures 46
have been dated according to the database described
above. Unfortunately, many of these datings still have
uncertainties. The uncertainties are due to a part of the
roof structure being dated but not enough investiga-
tions of it having been carried out in order to certainly
ascribe the dating to the entire roof structure. Eigh-
teen have parts dated to before 1150, fifteen to before
1200, nine to before 1250, one to before 1300 and
three to before 1350. Of the 70 roof structures 21 have
been examined and documented by the author. Thir-
teen roof structures have not been examined by the
author in person but there is documentation in draw-
ings and descriptions. For the final 36 roof structures
the information consists solely of written documenta-
tion. This data cannot be considered as certain as in
the other cases.
To identify possible common characteristics of the
early Swedish medieval roof structures several criteria
have been used: load-carrying system, elements and
shape, centre distance, slope of roof, connection with
wall, joints, outer roofing system and decorations.
All of the 70 roof structures carry the vertical (grav-
itational) load cross-wise to the longitudinal direction
of the building from wall to wall, with trusses of the
same design and size which are generally repeated con-
secutively at a close distance, see Figure 2. This system
of carpentry is called common rafter roof or uniform
scantling.
The roof structures have tie beams that connect the
bases of the roof trusses and provide support for the
horizontal thrust. They do not have a separate struc-
ture for longitudinal stability. Longitudinal stability
is achieved by boarding (Fig. 10) below the outer
roofing, which possibly provides a stabilizing plate
together with roofing material and in many cases con-
nects the roof trusses with stone gables. In the early
roof structures the tie beam is commonly embedded
Table 1. Churches with roof structures that have been part
of the study and their type.

Type Church, part and province

BS Flistad chancel ÖG*, Forshem nave VG, Garda

nave G, Hagebyhöga chancel ÖG, Halla nave
G, Hammarlunda chancel S, Herrestad chancel
ÖG, Kaga nave and chancel ÖG, Norra Mellby
chancel S, Råda chancel VG, Torpa chancel SÖ,
Övraby nave S.
BX Flistad nave ÖG, Granhult chancel SM, Kinne-
Vedum chancel VG, Skalunda chancel VG, Söne
nave VG, Torpa nave SÖ. Visseltofta chancel S.
BSX Hagebyhöga nave ÖG, Råda nave VG, Söne
chancel VG.
B2S Herrestad nave ÖG.
B2Xa Arby nave and chancel SM, Asby nave ÖG,
Bringetofta nave SM, Eriksberg nave VG,
Granhult nave SM, Göteve nave VG, Hemmesjö
nave and chancel SM, Jät old chancel SM,
Kinne-Vedum nave VG, Mularp nave VG,
Perstorp chancel S, Skepperstad nave SM,
Suntak nave VG, Tidersrum nave ÖG.
B2Xb Marka nave VG, Norra Solberga nave SM,
Väversunda chancel ÖG.
B3Xa Jät old nave SM, Mosjö nave N, Visseltofta
nave S.
B3Xb Forsby nave VG, Forserum nave SM, Marum
nave VG, Väversunda nave ÖG.
B3Xc Gökhem nave VG, Mjäldrunga nave VG.
BS2Xa Skalunda nave VG. Figure 1. Showing the type of the roof structures presented
BSH Gräve nave N, Pelarne nave SM, Stenberga nave in Table 1 by the parts they consist of. B – bindbjälke = tie
SM, Torpa nave ÖG. beam, S – stödben = strut beam/angle brace, X – korsande
B2SH Kånna nave SM, Lyngsjö nave S. stödben = crossing strut beams, H – hanbjälke = collar beam.
BXH Kumla nave ÖG, Roma nave G.
BS2H Björka nave S, Källunge nave G, Vireda nave SM.
BSHS Drev old nave SM, Nävelsjö nave SM.
BSHX Dädesjö old nave SM, Bälaryd nave? SM.
BS2HX Edestad chancel B.
BSHSH Linde nave G.

∗
Province: ÖG Östergötland, VG Västergötland, G Gotland,
S Skåne, SÖ Södermanland, SM Småland, N Närke, B
Blekinge.

into the masonry on top of the walls which can also

provide some longitudinal stability, see Figure 3.
Table 1 and Figure 1 show the inner parts of the
roof trusses. The oldest roof structures have strut
beams (angle braces) that connect the rafters with the
tie beam, for example Herrestad church and Garda
church, see Figures 4 and 5. There is a great variety
in number of struts and how they are placed in rela-
tion to each other. There are a large number of roof
structures with crossed strut beams, for example Asby
church and old Jät church, see Figures 6 and 7. One
interesting aspect is the lack of collar beams in the
early roof structures from the 12th century. The collar
beam becomes common during the 13th century.
Figure 2. Spatial organization of a common rafter roof.
Other elements that exist in the early medieval
Swedish roof structures are ridge beams and steering
plates (see Figure 8). A ridge beam is a longitudinal
beam that connects the roof trusses at the ridge.
Fifteen roof structures still containing ridge beams
were identified, nine of them dated to the 12th century
and one to the 13th century, while the remaining five

123

Figure 3. The tie beam and the rafter foot are completely
embedded into the masonry on top of the wall at the Kaga
church, Östergötland.
Figure 5. The roof structure at Garda church, Gotland.
It has one pair of strut beams between the tie beam and the
rafters. The roof structure was once open and is decoratively
carved. Today there is a ceiling nailed to the tie beams.

Figure 4. The nave roof structure of Herrestad church,

Östergötland. It has two pairs of strut beams between the
tie beam and the rafters and a ridge beam. The tie beams and
rafter feet are embedded into the masonry of the wall. Figure 6. The nave roof structure of Asby church,
Östergötland. It has two pairs of crossed strut beams between
the rafters and the tie beam.
lack dating. A steering plate is a beam that lies between
the tie beams or collar beams of the roof truss in the
longitudinal direction of the building. It is not known
whether the steering plate has had any function other
than being decorative. It can possibly contribute to
keeping the tie beams and collar beams in place in
the structure and keeping them from turning over.
The material used is pine or oak. The historic tim-
ber material used is generally of high quality but can in
some cases be damaged by biological or other causes,
making it difficult to estimate its actual properties.
Timber pegs and iron nails were both used in joints
throughout the Middle Ages in Sweden. The joints
commonly are straight lap joints, see Figures 9 and 10.
The roof structures are generally placed rather close
with a centre distance less than a metre between the
tie beams. There are a few examples with longer cen-
tre distances, up to 1.25 metres. The centre distance
can also show great variations within the same roof
Figure 7. The nave roof structure of Old Jät church,
structure. Most of the early medieval roof structures
Småland. It has three pairs of crossed strut beams between
on stone churches in Sweden have a slope of the roof the rafter and the tie beam. The tie beams lie embedded into
between 37 and 50 degrees. There are some excep- the masonry on top of the wall. In the middle of the structure
tions, for example the old Jät church, Småland, and above the tie beams there is a steering plate.

124
Figure 8. The image to the left shows the ridge beam in the
Garda church, Gotland, from above and the image to the right
show the steering plate in the Old Jät church, Småland.
Figure 9. The connection between the rafter and the tie
beam and between the tie beam and the wall plate at For-
shem church, Västergötland. The wall plate and the tie beam
was originally embedded into the top of the wall but has been
exposed since the church was vaulted and cross-arms were
added.
Garda church, Gotland, having a slope of 59 and 53
degrees respectively, see Figures 5 and 7. The oldest
dated roof structure has the flattest roof, see Figure
4. The timber churches generally have a steeper pitch
between 55 and 60 degrees.
The tie beams of the early medieval Swedish roof
structures usually lie more or less embedded into the
masonry of the top of the walls, see Figure 3. There
is usually one wall plate situated at the outside of the
wall, to which the tie beam is connected by a notched
joint, see Figure 9. The close relation between the wall
and the tie beam leads to the assumption that the roof
structure has a vertical support along the width of the
wall which provides a shortening of the span since
the walls usually have a thickness of about one metre.
The connection with the wall plate as well as with
the masonry gives good possibilities for transferring
horizontal forces by friction between the roof trusses
and the wall. This works as proved by examples where
the tie beams have been cut off for various reasons.
The situation is different for timber churches where
the tie beam lies notched over the top timber of the
wall, and it is probable that the tie beams also act to
hold the timber walls together.
125
Figure 10. Timber joints of the straight lapped type and
cleaved boarding in the nave roof structure of Old Drev
church, Småland.
Figure 11. Decorations in the nave roof structure of
Gökhem church, Västergötland.
Different types of outer roof were used. In most
cases they have been changed or replaced several
times. Common types today are shingle, wood, tiles,
slate and tin. Below the outer roofing material there
is boarding. If the original boarding is preserved it is
usually of cleaved and cut timber, see Figure 10. Later
types are sawn, and if they are replaced in modern time
planed boarding is normally used.
Studies of roof structures from the 12th century
show that the church interior originally was open to the
roof (Sjömar 1992). An investigation of Hagebyhöga
church, Östergötland, raised in the 12th century with
walls of stone, shows that the interior both in the nave
and in the chancel originally had no ceiling; the roof
structure was visible. The wooden roof construction
probably gave the room a character like that in the
Norwegian stave churches today. Several roof struc-
tures from the 12th century are given sculptural forms
(see Figures 5 and 11). Some parts were probably
painted. Other signs of open roofs are plaster that
remains on the walls above later vaults or ceilings,
see Figure 3. This means that the wood craft was an
important means to create character in the interior,
also in the stone churches. Later a flat ceiling made
of planed boards was put up in Hagebyhöga church
before, finally, some hundred years later, stone vaults
were built. Traces of such wooden ceilings are also
found in other churches and some still exist today. In
many Swedish parish churches vaults were built in the
14th and 15th centuries.
4 COMPARISONS WITHIN A
NORTH-WESTERN EUROPEAN CONTEXT
When comparing the early medieval Swedish roofs
with other North-Western European roofs that were
raised about the same time, it is clear that they are
all part of the same context. Published surveys and
studies, in Denmark (Aaman Sørensen 1995, Madsen
2007), Norway (Storsletten, 2002), England (Smith,
1975, 1981, 2006, Hewett, 1981, 1985, Walker, 1999,
Courtenay 1993), France (Corvol-Dessert et al. 2002,
Epaud, 2003, Courtenay 1993) and Germany (Binding
1991, Lohrum 2006) show that the constructions in
all these countries and Sweden have similarities. Also
eastern parts such as Romania seem to have some sim-
ilarities (Szabó 2005). The studies show, just as in
Sweden, that the roofs were raised by skilled crafts-
men and that the crosswise load-carrying system of
the common rafter roof with tie beams on every truss
seems to have been used throughout Europe for the
early medieval Romanesque roof structures.
Most sources refer to the early medieval
(Romanesque) roof structures as having collar beams
often in combination with strut beams (Fletcher &
Haslop 1970, Corvol-Dessert et al. 2002, Courtenay
1993, Szabó 2005), but there are exceptions in France,
Belgium, Denmark and Germany of early medieval
roof structures without collar beams similar to those
in Sweden. Binding (1991) describes a similar devel-
opment to that in Sweden where the earlier roof
structures lack collar beams. The collar beams seem
to appear in Germany in the second half of the 12th
century. Courtenay (1993) and Epaud (2003) describe
the roof structure of St Gertrude at Nivelles, con-
structed in 1046, which no longer exists, and Binding
(1991) describes the monastery church in Billigheim
constructed in 1180/1190. Both St Gertrude and
Billigheim are very similar to the roof structure at
Herrestad church in Sweden dated to 1112. It has
four angled struts connecting the rafters with the tie
beam, see Figure 4. The roof structure of St Denis
in Liège is also similar but has one extra pair of
struts (Corvol-Dessert et al. 2002). In France there
are several examples of roof structures with vertical
struts, as in the roof structure of Saint-Barthélemy,
126
Liège, and Saint Pierre-de-Montmartre, Paris (Corvol-
Dessert et al. 2002). Randerup church in Denmark
(Madsen 2007) and St Martin in Sindelfingen (dated
to 1132) in Germany (Binding 1991) are of the type
with only one pair of strut beams, like for example
Garda church in Sweden, see Figure 5. There are also
examples of roof structures with crossed strut beams
in France and Denmark. The church of St Christophe
in Chabris, France, has a roof structure with six cross-
ing strut beams (Epaud 2003), similar to the old Jät
church (Fig. 7). Vester Nebel church in Denmark
has a roof structure with four crossing strut beams
(Madsen 2007), similar to Asby church (Fig. 6). Both
the Danish examples mentioned have in later times
been reinforced with collar beams.
The use of ridge beams (or ridge pieces) in some of
the Swedish roof structures is interesting, see Figure
8. Courtenay (1993) claims that ridge beams cannot be
found in England or France from 1050–1300 when dis-
cussing the late Norwegian stave churches that contain
both ridge beams and purlins. Maybe the ridge beams
on the early medieval roof structures demonstrate a
connection between the different Scandinavian roof
structure types.
Courtenay (1993) thinks that lack of longitudinal
bracing is characteristic of the early medieval roof
structures. This is also described by Corvol-Dessert
et al. (2002) and Binding (1991). Examples of diagonal
elements on the inside of the rafters to provide longi-
tudinal bracing are described by Fletcher & Spokes
(1964), Corvol-Dessert et al. (2002) and Madsen
(2007). This type of longitudinal bracing is used in
later medieval times in Swedish roof structures but
generally not in the early medieval types.
The centre distance between the roof trusses, nor-
mally being less than a metre, as well as the slope
of the roofs, normally between 40 and 50 degrees,
seem to be generally similar across Europe for the
early medieval roof structures. Joints are generally
described as being of the simple lap type (like the
Swedish ones) or being notched or having dovetailed
profiles (Courtenay 1993, Fletcher & Spokes 1964,
Fletcher & Haslop 1970, Szabó 2005, Madsen 2007)
Courtenay (1993) describes the earliest
Romanesque roofs in Northern Europe as being open
and visible from the inner space or intended to sup-
port a panelled ceiling. In contrast to the Swedish
ones, these types of roof structures are rare and only
fragmentary preserved. Evidence of decorated roof
structures from early medieval times outside Sweden
is difficult to find. Binding (1991) says that the ques-
tion of open roofs in Germany is debated and many
roofs of churches without vaults have or show evidence
of earlier ceilings. There are later examples of open
roofs from the 13th century but they are of a different
structural type. In England the tradition of open and
visible roof structures was developed in later medieval
times (Smith 2006). Smith also describes the use of a
“locking-plate” in Lincoln cathedral from around 1200
that seems to be similar to the Swedish steering plates.
The connection between the roof trusses and the
wall normally consists of wall plates in order to dis-
tribute loadings more uniformly to the top of the walls.
Examples of the use of masonry-embedded tie beams
and rafter feet exist in Germany, Denmark and France
(Binding 1991). One German example very similar
to the Kaga church (Fig 3) is St Ägidius in Mittel-
heim, which was lost in the 20th century. There is one
English example in the nave roof of Kempley, dated to
1120–1150 (Smith 2006). There are also examples of
roof structures totally devoid of wall plates (Courtenay
1993) although most seem to have double wall plates.

5 STRUCTURAL BEHAVIOUR

The simplest type of roof structure consists of two sin-

gle rafters that rest on each other with a hinged joint at Figure 12. Geometry and resulting moment diagrams from
the top. In carrying vertical load such a structure cre- computations of roof trusses with tie beams and with or
ates both vertical and horizontal forces that act upon without strut beams and collar beam.
the supports. The horizontal forces need to be sup-
ported to avoid collapse of the structure. The early
medieval roof structures have tie beams that provide Table 2. Results from computation of roof trusses in
Figure 12.
such a support, see Figure 12. If the tie beam is missing
or has been cut off the roof structure becomes depen- Normal force in tie Normal force in strut
dent on an outer support provided by the walls to avoid beam by rafter foot beam (mean value)
large deformation and stress or even collapse. Truss type kN kN
With an increase in span the problem of deflecting
rafters with high bending moments soon becomes evi- Case A 1.83 –
dent, see Figure 12. The early solution was to put strut Case B 1.70 −1.62
beams between the rafters and the tie beam to sup- Case C 3.20 0.18
port them. Figure 12 and Table 2 presents the results
from computations made on roof trusses with an outer Max moment in Max moment in tie
span of nine metres and a roof angle of 47◦ , which rafters beam
is a typical size and shape of early medieval roof kNm kNm
trusses. The support conditions are based on a one-
Case A 2.20 0.59
metre-wide wall where the vertical support is given at Case B 0.42 2.70
the outside, 150 mm from the outermost part of the roof Case C 0.61 0.65
truss, and on the inside of the wall. The tie beam has
230 × 150 mm cross-section and the other parts has
150 × 100 mm cross section. The strut beams are con-
nected to the tie beam 3.87 m from the joint between support, by compression in them, against deflection
the rafters. The tie beam and the joint between the strut of the rafters but at the expense of a greatly increased
beams and the rafters are halfway to the ridge in the deflection and moment load of the tie beam, see
vertical direction from the tie beam. The centreline of Table 2. The support provided by the strut beams thus
the collar beam is 2.70 m above the centreline of the becomes dependent on the bending stiffness of the tie
tie beam. The material is wood with an E-modulus of beam. The problem of too much deflection and bend-
10 GPa. The load is vertical and symmetrical and con- ing in the rafters is transferred to the tie beam instead.
sists of the timber (density 500 kg/m3 ) and the weight This led to the use of tie beams with large dimensions
of the outer roofing consisting of boarding and tiles in such roof structures.
which weigh 53 kg/m2 . The roof trusses stand one The collar beam is a clever innovation that makes
metre apart. the rafters support each other by compression in the
The moment diagram in Figure 12, case B, shows collar beam. If it is combined with strut beams it can
that the strut beams are successful in providing mean, depending on the placing of the collar beam, that

127
the strut beams act as hanging elements supporting the
tie beam instead of loading it, since the support of the
rafters has been taken care of by the stiffer load path
provided by the collar beam, see Table 2. The use of
collar beams also gives an increasing horizontal thrust
at the base of the roof that has to be transferred by the
joint between the rafters and the tie beam, see Table
2. The collar beam does not provide any wind bracing
more than making the rafters act together. The strut
beams, on the other hand, can act as wind braces in the
roof structures.
6 CONCLUSIONS
It seems that surprisingly many Romanesque roof
structures remain in Sweden. Of the very few North-
Western European standing roofs that were raised
in the first half of 12th century, at least 10 still
exist in Sweden. These, together with the rest of
the well-preserved medieval roof structures, are an
important and valuable cultural heritage in a European
perspective.
The remaining roof structures from early medieval
times in North-Western Europe clearly show that they
are part of the same context and have many structural
similarities but also some differences. They are com-
mon rafter roofs that carry their load crosswise to the
longitudinal direction of the building; they have tie
beams on every roof truss and in many cases con-
sist of similar inner elements. The sources indicate
that the chronology of early roof structures without
collar beams is true also outside Sweden. The collar
beam is shown to be an innovation that improves the
load-carrying system for vertical loads.
Additional studies are needed to complete the pic-
ture of early medieval roof structures, not least in
Sweden. More systematic investigation and surveys of
the medieval roof structures in combination with dat-
ing methods could provide a means to achieve a more
comprehensive overview that would shed further light
on this significant heritage of carpentry.
REFERENCES
Aaman Sørensen, M. 1995. Middelalderlige tagværker,
særlig på Sjælland. Hikuin (22): 41–46.
Bartholin, T. 1994, Träets År. Kulturmiljövård 1994(1): 38.
Binding, G. 1991. Das Dachwerk auf Kirchen im deutschen
Sprachraum: vom Mittelalter bis zum 18. Jahrhundert.
München: Dt. Kunstverl.
Corvol-Dessert, A., et al. 2002. Les charpentes du XIe au
XIXe siècle: Typologie et évolution en France du Nord
et en Belgique. Paris: Centre des monuments nationaux/
Monum, Éditions du patrimoine.
Courtenay, L. T. 1993. Timber Roofs and Spires. In Mark,
R. (ed.), Architectural Technology up to the Scientific
128
Revolution: The Art and Structure of Large-Scale Build-
ings. Massachusetts: The MIT Press.
Epaud, F. 2003. Evolution des techniques et des structures de
charpenterie du XIe au XIIIe siècle en Normandie: Une
approche des charpentes par l’archéologie du bâti. Paris.
Fletcher, J.M. and Haslop, F.W.O. 1970. The West Range at
Ely and its Romanesque Roof. Archaeological Journal
(126): 171–176.
Fletcher, J.M. and Spokes, P.S. 1964. The Origin and Devel-
opment of Crown-Post Roofs. Medieval Archaeology (8):
152–183.
Hewett, C. A. 1981. Church Carpentry, a Study Based on
Essex Examples. London: Phillimore.
Hewett, C. A. 1985. English Cathedral and Monastic Car-
pentry. Chichester: Phillimore
Linscott, K. 2007. Medeltida tak, bevarade takkonstruk-
tioner i svenska medeltidskyrkor, Del 1: Rapport om
kunskapsläget 2006. Göteborg: Göteborg University.
Lohrum, B. 2006. Early Variations of the “Sparrendach”
in the “Pfettendach”-Region of South-Western Germany.
In Drdácký M. (ed.), European Research on Cultural
Heritage: State of the Art Studies Volume 4. Prague: Insti-
tute of Theoretical and Applied Mechanics Academy of
Sciences of the Czech Republic.
Madsen, P. K. 2007. Middelalderlige kirketagværker i
Sydvest- og Sønderjylland – eksempler på alder og typer.
Aabøger for nordisk Oldkyndighed og Historie 2003.
København: Det Kgl. Nordiske Oldskriftselskab.
Sjömar, P. 1990, Medeltida takkonstruktioner i kyrkor, Stock-
holm: Sveriges Kyrkor, Riksantikvarieämbetet.
Sjömar, P. 1992. Romanska takkonstruktioner: ett värde-
fullt och outforskat källmaterial. In Ullén M. (ed.), Från
romanik till nygotik: studier i kyrklig konst och arkitek-
tur tillägnade Evald Gustafsson. Stockholm: Riksantik-
varieämbetet.
Sjömar, P. 1995. Romanskt och gotiskt – takkonstruktioner i
svenska medeltidskyrkor. Hikuin(22): 207–230.
Sjömar, P. and Storsletten O. 1993. Där Norge och Sverige
mötas. Kirkearkeologi og kirkekunst: studier tilegnet
Sigrid og Håkon Christie. Øvre Ervik: Alvheim & Eide.
Smith, J. T. 1975. Norwegian Stave-church Roofs from an
English Standpoint. Universitetets Oldsaksamling Arbok
1975/76/77: 123–136.
Smith, J. T. 1981. Mittelalterliche Dachkonstruktionen
in Nordwesteuropa. Frühe Holzkirchen im nördlichen
Europa: 379–390.
Smith, J. T. 2006. English Medieval Roofs. In Drdácký M.
(ed.), European Research on Cultural Heritage: State of
the Art Studies Volume 4. Prague: Institute of Theoretical
andApplied MechanicsAcademy of Sciences of the Czech
Republic.
Storsletten, O. 2002.Takene taler: Norske takstoler 100–1350
klassifisering og oprinnelse. Oslo: Arkitekthøgskolen.
Szabó, B. 2005. Illustrated Dictionary of Historic Load-
bearing Structures. Cluj-Napoca: Kriterion Utilitas.
Thelin, C. 2005. Några medeltida takstolar: En delrapport
i ett forskningsprojekt kring takfotens historiska utveck-
ling och konstruktiva verkningssätt. Göteborg: Chalmers
University of Technology.
Walker, J. 1999. Late 12th & Early 13th Century Aisled
Buildings: A Comparison http://www.ccurrie.me.uk/vag/
walker/aisled1.htm (2007).
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Analysis of archaic fireproof floor systems

D. Friedman
Old Structures Engineering, New York, US

ABSTRACT: As engineers and builders developed modern steel framing in the late nineteenth century, the
existing options for floors to span between steel beams were forms of masonry vaults. Many possible alternate
floors were developed in the United States, but few had rational bases for design. Testing programs put in place by
building officials in NewYork City promoted the use of certain systems in NewYork and, by providing a rationale
for those systems, nationwide. Ten systems are described and analyzed.

Conservation and analysis of early modern structure
often begins with the main structural material: we
have one set of techniques for steel-framed build-
ings, another for reinforced concrete, and another for
those buildings which are architecturally modern but
are constructed using traditional wood and masonry.
This focus is inevitable but tends to obscure secondary
structural materials and systems, which include the
framing for facade ornament, structural adaptations
for mechanical systems, and floor systems. This paper
will address the original design and current-day analy-
sis of floor systems used in the United States between
1890 and 1930 to fill in between iron and steel beams.
Since iron-framed floors were first used in the
late eighteenth century, floor systems have devel-
oped in parallel with the iron-framing development.
A wood-joist floor is most often associated with wood
plank flooring, and masonry vault structures provide
masonry floors, but iron or steel plates were not
used spanning between iron and steel beams. Instead,
combinations of other materials have been employed
between the beams. Because the designers and builders
who used metal-framed floors were usually interested
in increasing their buildings’ resistance to fire, floor
systems have typically been made of masonry and
concrete.
Floor systems can be divided into three broad
groups: arches, catenaries, and beams. These cate-
gories are, of course, the three basic methods of
supporting vertical load across a horizontal span, but
unlike complex structures such as bridges where two
(or, rarely, all three) basic methods may be combined,
floor systems tend to show only one method. That
the earliest floor systems were arches is unsurpris-
ing since masonry vaults predate structural iron beams
by thousands of years; it is far more interesting from
the modern perspective that catenary systems were
popular before beam systems.
There was no mechanism for the creation of national
standards in the United States during 1880s and 90s
because there were no national building codes or
authorities. Various floor systems were promoted by
manufacturers and contractors, and reviewed by local
building officials who often had no technical train-
ing. Empirical load and fire-resistance analysis was
used by some insurance companies, local building
departments, and engineering schools. After 1896, the
NewYork City Department of Buildings and Columbia
University’s School of Mines together conducted hun-
dreds of tests on samples representing dozens of floor
systems, beginning with systems already in use and
inviting floor-system manufacturers to submit their
products. Given the lack of national standards and
the rigorous testing regime, the results of these tests
were influential far beyond New York and long after
the tests ended. For example, standard E119 of the
American Society for Testing and Materials, currently
used in the United States to determine fire-rating of
floor assemblies, is closely related to the New York
testing standard and can be used to corroborate some
of the older results.
This paper will review the floor systems that dom-
inated “fireproof” construction in the Untied States
between 1890 and 1930 as well as selected less-
common systems. For each system, analysis using
current methods will be compared with available anal-
yses from the era of construction and historic test
data. Discussion will include implications for contin-
ued use and alteration in terms of design load capacity,
detailing, and fire resistance.

129
1 HISTORICAL CONTEXT
Structural steel technology and skeleton framing of
buildings matured in the United States between 1890
and 1910. At the beginning of this period, reinforced
concrete was just beginning to be used and was still, to
American engineers and builders, partly experimental.
There was not a general consensus on reinforced con-
crete until the late 1910s. Given the late development
of concrete relative to steel, designers and builders had
already addressed the need for fireproof floors before
concrete became common. The use of bar-reinforced
concrete slabs never became the dominant solution for
floors; after the invention of composite metal deck in
the 1950s, designers and builders rarely used any other
system. As a result, from 1890 to 1960, floor systems
were used which are now unfamiliar to most engineers
and builders.
As large steel-framed buildings were constructed
in the 1890s, designers and builders became aware
of an acute need for inexpensive, lightweight and
“fireproof ” structural floors. This need was not new:
the Chicago fire of 1871 and the Boston fire of
1872 – large-scale conflagrations that had between
them destroyed 18,000 buildings including much of
both commercial downtown areas – had emphasized
the need for buildings better able to resist fire. By the
1880s, various proprietary forms of terra-cotta tile arch
floors had been patented and were in use, although all
were functionally similar.
The word “fireproof,” as used at that time, meant
what is now called “fire resisting:” a building or a
structural element that could withstand some degree
of exposure to fire without failure. In this paper, “fire-
proof ” will be used in that sense, not in its literal
meaning of withstanding any exposure to fire.
As new ideas for fireproof floors were devel-
oped, building officials and insurance companies were
forced to consider how floors built with new, often pro-
prietary and patented technology, could be reviewed
for approval. In the end, the various parties, includ-
ing engineers, companies selling proprietary systems,
and insurance companies, cooperated with officials to
create standardized tests. (Hill 1895) Many of these
tests were performed in the late 1890s and early 1900s
by various agents under the supervision of the New
York City Bureau of Buildings, in accordance with the
requirements of the New York City Building Code at
that time. (no author 1905)
The portion of the code that defined floor systems
read: “All brick or stone arches, placed between iron
floor beams, shall be at least four inches thick and
have a rise of at least one and a quarter inches to each
foot of span between the beams. Arches over five feet
span shall be properly increased in thickness . . . or the
space may be filled in with sectional hollow brick of
burnt clay or some equally good fire-proof material,
130
having a depth of not less than one and one-quarter
inches to each foot of span, a variable distance being
allowed of not over six inches in the span between the
beams.” Some version of this was in the New York
code from 1891 to 1916. (Birkmire 1898, no author
1899, no author 1901)
The first clause allows for traditional brick and
stone vaulting, which was rarely used because of the
weight of the materials and the intensive labor required
for construction. The second clause allows some
segmental terra-cotta tile arches. The third clause,
centered on the vague phrase “some equally good
fire-proof material,” and led to a proliferation of pro-
prietary systems from manufacturers looking for a
portion of the lucrative New York construction mar-
ket. After several years of politicized review of various
systems, the city Superintendent of Buildings, Steven-
son Constable, arranged for a series of tests, beginning
in 1896 and performed by the Bureau of Buildings in
concert with Columbia University. (Freitag 1899)
1.1 Testing
The building code enacted in 1898 in New York
allowed for brick and segmental tile vaulted floors.
Other floors were addressed: “. . . or between the said
beams may be placed solid or hollow burnt-clay, stone,
brick, or concrete slabs in flat and curved shapes, con-
crete or other fireproof composition, and any of said
materials may be used in combination with wire cloth,
expanded metal, wire strands, or wrought-iron or steel
bars; but in any such construction and as a prece-
dent condition to the same being used, test shall be
made . . .” (no author 1899)The test defined in the code
consisted of three stages of loading on a completed
panel of the floor in question: first, with the 7.2 kPa
load in place, the floor was heated so that an average
temperature of 927 C was kept for four hours; second,
with the load in place, the floor was cooled by pres-
surized streams of water meant to imitate fire-fighters’
hoses; and third, the floor was reloaded to 29 kPa. The
physical integrity of the floor was observed at all times
and deflection measurements taken during the third
loading stage. (no author, 1897)
Because the New York code was interpreted lit-
erally, each variation on a system had to be tested
separately. A floor system had to be tested for each
possible load and span combination for which it was
sold, leading to over 190 tests between 1895 and 1915.
(Perrine & Strehan 1915)
Similar tests were used in other cities, by insurance
companies in setting fire rates for buildings using new
floor systems, and by builders and inventors attempt-
ing to prove the reliability of their systems to owners
and local building officials. The fireproofing aspect
of the tests is obvious, but in many cases there was no
structural design for the floors beyond the empirical
 Figure 2. End-construction Flat Tile Arch (Birkmire 1898).

Figure 1. Side-construction Flat Tile Arch. The “weak

point” refers more to fireproofing than structural load.
(Birkmire 1898).

proof of supporting a high level of live load while

exposed to fire. (Freitag 1899)

2 EXAMPLE FLOOR SYSTEMS

The floor systems described here are only a few of

the many tested between 1890 and 1910. They include
the most common floors and a few oddities that often
confuse modern engineers.
The floors are grouped by type (arches, catenaries,
and beams) and in the order that the types became pop-
ular. The earliest floor systems used with iron beams
in the Untied States were arches. During the 1880s
and 90s, various tile arch forms were the most popular
floors in use, and were still used into the 1920s. The
draped-mesh (catenary) floors developed in the 1890s
became the most popular form in the late 1910s and
1920s, and were only permanently replaced by a beam-
type floor when composite metal deck was developed
in the 1950s.
Figure 3. Uastavino section, elevation, and plan (Guas-
2.1 Arch floors tavino 1892).
Neither brick vault floors nor plain segmental terra-
cotta tile arches were tested in New York, as both were Guastavino company achieved fame in the construc-
allowed without limitation. They were typically not tion of domes and long-span floors, roofs, and ceilings.
tested elsewhere, as they were considered ordinary and Two or three layers of thin tiles were arranged with
familiar floors. staggered joints. (no author 1889, no author 1897,
Both generic and proprietary flat terra-cotta tile Freitag 1921, Collins 1982)
arches were tested in New York because neither met The Roebling Arch Floor is one of a class of early
the geometric requirements for exemption. These sys- concrete floors produced by several manufacturers.
tems consisted of precast terra-cotta blocks with thin (Fig. 4) Wire-mesh arches spanned between the beam
webs, arranged as flat-arch voissoirs. There were two bottom flanges and served as forms for stone concrete
major types: side construction, with terra-cotta voids vaults. More wire mesh was hung from the beams to
perpendicular to the vault span, and the newer end con- create a plaster ceiling for fireproofing. (no author
struction, with voids parallel to the span. (Figs 1, 2) 1897, Birkmire 1898, Hool 1913).
The tiles were typically set low, to cover the bottom
of the steel beams, and fill was placed over the top to
2.2 Catenary floors
protect the tops of the beams and provide a base for
wood flooring. (no author 1897) The Metropolitan System was an early draped-wire
The Guastavino Timbrel Vault was an interesting floor, approved in 1899. (Fig. 5) The reinforcing con-
variant on the standard terra-cotta tile-arch floor, using sisted of twisted pairs of wires individually strung
Catalan hard-burned tiles and thin-shell construction across the building, anchored at slab edges, and drap-
to meet the American testing standards. (Fig. 3) It ing over the floor beams and under a hold-down bar at
was not popular for ordinary floors, although the mid-span. Since the wires carried all loads, the slabs

131

Figure 4. Roebling Arch: wood sleepers for finish flooring,
concrete fill, wire-mesh arch, hung ceiling. (Birkmire 1898).
Figure 5. Metropolitan Floor (Freitag 1899).
Figure 6. Cinder-concrete, draped-mesh slab, (Buel & Hill
1906).
were structurally unstressed and were composed of
gypsum plaster mixed with wood chips. (no author
1897, Freitag 1898, Birkmire 1899, no author 1895)
Draped-mesh floors were first developed under
proprietary names, but the presence of the existing
catenary patents meant that the basically similar mesh
floors quickly became generic commodities. (Fig. 6)
Various proprietary forms of mesh were used early on,
but since the floor only required a specified standard
size of wire at a specified spacing, the generic floors
were ultimately more common than the named ones.
As with the Metropolitan floor, the mesh passed over
the top of the floor beams, near the top of the slab,
and at midspan draped down to the bottom of the slab.
Because the mesh wires take all of load in tension, the
slab serves only to provide a walking surface and fire-
proofing. Because the concrete is not stressed, material
of poor quality and low strength, such as cinder con-
crete, was frequently used. (Perrine & Strehan 1915,
Waite 1914)
132
Figure 7. Rapp Floor: wood sleepers to attach finish floor-
ing, fill (“concrete”), brick, and inverted Ts. (Birkmire
1898).
Figure 8. Expanded Metal Floor (Freitag 1899).
Figure 9. Columbian Floor, with wood flooring on sleepers,
and rebar supported by light-gage straps over the beam tops.
The bottom transverse section shows the double-cross bars.
(Freitag 1899).
2.3 Beam floors
The Rapp Floor consisted of common brick, support-
ing a layer of fill and supported by light-gage steel
inverted Ts, which span between the bottom flanges of
the floor beams. (Fig .7)(Birkmire 1899)
The Expanded Metal Company Floor is a thin con-
crete slab reinforced with a sheet of “expanded metal,”
created by slitting a light-gage steel sheet and pulling
the sheet to open the slits. (Fig. 8) The reinforcing sheet
sat directly on the tops of the floor beams. The system
had an overall slab thickness of only 5 inches, in part
because the beams were at a close spacing. (no author
1897, Freitag 1899)
The Columbian Floor system was an early bar-
reinforced concrete slab. Bars with a cross or double-
cross section were hung from the beam tops using
light-gage steel straps to serve as flexural reinforc-
ing in a cinder-concrete slab. (Fig. 9) The typical
beam encasement provided the only shear transfer. (no
author 1897, Birkmire 1898, Hool 1913)

Figure 10. Roebling Flat Slab Floor, showing strap rein-
forcing twisted at the beam top. (Freitag 1899).
The Roebling Flat Slab Floor was basically similar
to the Columbian floor, except that the reinforcing was
rectangular bars set vertically and twisted horizontally
to rest on the floor beams. (Fig. 10) (Freitag 1899, Hool
1913)
3 RE-USE AND ANALYSIS
Since the original use of these floors depended only
on empirical testing, analysis came after the fact. For
the arch floors, there was rarely analysis at all until the
floors fell out of use, while justifications for the beam
and catenary floors have changed over time.
3.1 Current codes
During the course of the twentieth century, numerical
analysis replaced the empirical tests remaining in the
structural design. Steel beam design had been ratio-
nalized in the nineteenth century, and concrete beam
design after 1900. At the time of the floor testing, all
American building codes were local, created by either
municipalities or states. Regional and national codes
were only established later, and have only allowed
analysis-based structural design. In the case of New
York, the 1916 building code removed the testing
requirement in favor of analytic design. This code was
revised but kept in use until 1968, when an entirely
rewritten code was put into effect. All references to
tile arches or other masonry floors were removed, but
draped-mesh slabs remained as “Short-span concrete
floor and roof construction supported on steel beams.”
The formulas for minimum allowable slab thickness
and wire stress were the same as the old code, and
values were given for stone concrete and unspecified-
material “lightweight aggregate concrete.” This code
remained in effect, with revisions, until July of 2007,
when it was superseded by a local version of the 2003
International Building Code, with a one-year overlap
until the summer of 2008.The new code retains one last
vestige of the old floors: a statement that wire mesh
reinforcing “is permitted to be curved from a point
near the top of the slab over the support to a point
near the bottom of the slab at midspan, provided such
reinforcement is either continuous over, or securely
133
anchored at support.” (no author 2004, no author
2007a)
The load analysis for the different floor types that
follows is based on current codes except as noted.
Modern code analysis is somewhat simpler with
regard to the fireproofing capacity of the old floors.
Under currentAmerican practice, fire ratings for struc-
tural assemblies are determined using tests defined
in the standard specification ASTM E119. The stan-
dard has been published and periodically revised by
the ASTM since 1917 and carries no authority of its
own, but is regularly adopted into building codes and
other statutes. (no author 2007b)
The provisions of E119 are similar in many respects
to that of the New York City tests. The floors
are to be loaded “to simulate a maximum load
condition . . . under nationally recognized structural
design criteria,” the heating is to follow an upward
curve from 538 C at five minutes to 1093 C at 4 hours
(and up to 1260 C at eight hours if a rating past four
hours is sought), the floor is to be cooled by a pres-
surized hose stream, and the test report is to include
information on damage and deflection. Since (a)
the maximum design load on most floors is less than
the 7.2 kPa used in the New York tests and (b) there
are no standard uniform loads of 29 kPa as was used in
the New York reloading test, the current loading crite-
rion is less stringent than the New York tests. The total
amount of heat to be absorbed in the early stages (as
measured by the area of the time/temperature curve)
is less in the ASTM tests than in the New York tests,
the New York tests are more conservative for systems
with low fire-ratings. At four hours, the total energy
is nearly identical between the two tests, so that a sys-
tem that passed the New York test would be assigned
a four-hour rating using the ASTM test. Only when
the tests are continued past four hours does the total
energy in the ASTM test exceed that of the New York
test. As few floor systems are now required to have rat-
ings greater than three hours, it can be generally stated
that any floor that passed the New York test has effec-
tively passed the ASTM test and can be considered to
have a four-hour fire rating if it has been maintained
in its original state.
The most significant difference between the New
York tests and the ASTM test is one of perception: the
older tests were used as proof of both fire-resistance
and structural load capacity, while the current tests
are used only to determine fire ratings for assemblies
assumes to have been already numerically designed
for structural loads. This gap is less of a problem for
existing buildings than it might appear, since the cur-
rent building codes allow existing structure to be load
tested to prove capacity, in a manner reminiscent of
the historic tests.
The use of historic test data, common among engi-
neers, has been given a government imprimatur in the
form of a guide to fire ratings on historic structures
compiled by the US National Institute of Building
Sciences and published by the federal Department of
Housing and Urban Development. (no author 2000)
This guide is a compilation of historic and modern
tests on various types of building elements, including
floor systems. Gypsum slabs, terra-cotta tile arches,
and tile and concrete rib systems are all included and
assigned fire ratings of up to four hours. While the
guide does not comprehensively address all of the
floor systems of interest, it does provide an official
opinion on some and it does show that the reuse of his-
toric test data can be an officially-accepted method of
research.
3.2 Arch analysis
Arch analysis is one of the oldest problems in structural
design. In ordinary masonry practice, symmetrical and
uniformly loaded arches are rarely dependent on the
compressive strength of the material, but the terra cotta
tile used in the floors was often weaker than stone or
ordinary brick, and was used in thin webs vulnerable
to local concentrations in stress.
Among the various formulas developed for masonry
arches exists a class of simplified formulas for minor
arches – those with short spans and low rises. In such
cases, the formulas also include the assumptions for
a flat arch of uniform loading and a specific location
for the thrust line.
A simplified formula suggested while terra-cotta
floors were still in use is H = WS / 8D, where H is
the thrust, W is the total uniform load, S is the span,
and D is the effective rise of the arch. This formula
assumes that the thrust line curve was 61 mm less than
the height of the terra cotta blocks, as D is the total
arch depth minus 61 mm. (no author 1919) Note that
the load and analysis are for a unit width of arch. In this
case, the main simplifying assumption, represented by
the use of D rather than the arch depth is that the thrust
line curve is proportional to the block depth excluding
the lower portion of the blocks which extend below the
steel beams. For example, a typical application span-
ning 1830 mm between beams and consisting of wood
flooring, 125 mm inches of cinder fill, 254 mm deep
terra cotta blocks making up the flat arch, 12 mm of
plaster on the block soffits as a ceiling, and an office
live load of 2.9 kPa, had a total weight, including steel
framing, of 6.8 kPa. This translates to an arch thrust of
15 N/mm to be carried by tie rods from beam to beam.
A typical terra cotta block had 19 mm thick webs
at 100 mm spacing. Since the horizontal webs are ori-
ented incorrectly to carry the arch thrust, the thrust
must lie entirely within the vertical webs. With an
effective web area of 254–61 or 193 mm by 19 mm, the
compression in the terra cotta is .41 MPa, well within
the allowable range for clay masonry.
134
A modern simplified formula for arch thrust used by
the Brick Industry Association is H = 3 WS/8d, where
H is the thrust, W is the total uniform load, S is the span,
and d is the arch depth. (no author 1986) Again, the
analysis is for a unit width of arch. In this case, the main
simplifying assumption, represented by the “3” in the
formula, is that the thrust line is within the middle
third of the arch height. This is more conservative than
the older formula in that it increases the thrust force
by confining its path. Using the previous example, the
thrust is 34 N/mm and the net compression on the webs
is 0.93 MPa, which is still within the capacity of clay
masonry.
The formulas for segmental arches do not require
the complicated assumptions regarding the location of
the thrust curve: it is taken within the curve of the vault.
Since the floor vaults based on segmental curves typi-
cally support fill above, a thrust line that extends past
the vault material on the top side and passes through
the fill may still be viable.
Using the Guastavino vault as an example of a
segmental-arch floor, with a span of 1830 mm and
a rise of 150 mm, the average total load is 7.9 kPa,
with the extra dead load resulting the deeper fill over
the arch ends. The basic thrust formula, H = WS/8d,
gives a thrust of 22 N/mm. Unlike terra cotta vaults, the
stress in a Guastavino vault is evenly distributed along
the solid masonry section, so with a vault thickness of
75 mm, the compressive stress is 0.29 MPa.
Finally, the Roebling arch is the simplest of the type.
The minimum arch depth is the distance from the floor
to the top of the arch center, and the material is stone
concrete, with a compressive strength in the range of
14 MPa, so an analysis similar to the Guastavino floor
shows a gross over-capacity.
3.3 Catenary analysis
Analyses for catenaries with fixed supports were avail-
able in the late nineteenth century, both as general
formulas based on statics and simplified formulas that
became part of building codes and manufacturers’ rec-
ommendations during the first half of the twentieth
century.
One of the simplified catenary formulas that was
publicized by the wire-products division of United
States Steel was also incorporated in the New York
City Building Code in 1916 (no author 1944) This for-
mula is W = 3CAs / L2 where W is the total allowable
floor load, L is the beam-to-beam centerline spac-
ing , As is the unit wire cross-sectional area, and
C is a constant representing the maximum allowable
wire stress and equal to 138 MPa. The referenced ver-
sions of this formula refer to stone or cinder-concrete
slabs; however, the results also agree closely with the
stated capacities of the Metropolitan Floor’s gypsum
slab. It should be noted that the maximum allowable
tension in the wire of 138 MPa is very low for steel
wire produced after 1900. The minimum yield stress
expected in historic wire systems has been conserva-
tively established as 345 MPa. (no author 1981) Given
that current reinforced-concrete codes have combined
safety factors in the range of 1.5 to 1.8, it is evident
that there is excess capacity in the wire in the original
designs.
For example, in a 100 mm-thick draped-mesh slab
with a 1830 mm span, the total load based on wood
flooring, the cinder-concrete slab, 12 mm of plaster on
the soffit as a finish ceiling, and an office live load
of 2.9 kPa was 5.1 kPa. With mesh reinforcing of 8-
gage (4.2 mm diameter) wire at 75 mm spacing, the
maximum allowable load is 41 KPa.
A more basic formula derived from statics by
approximating the catenary curve as a parabola (a rea-
sonable assumption since the actual wire curve is not
particularly accurate) is T = (wsL2 / 8 h) + wsh where
T is the tension in a wire, w is the combined dead and
live load, s is the wire spacing, L is the span, and h is
the wire sag. By setting T equal to the allowable ten-
sile stress (Ta ) multiplied by the wire cross-sectional
area (As ), the maximum floor load can be deter-
mined as W = 8 hTaAs / (sL2 + 8 sh2 ). In the previous
example, using the 138 MPa maximum stress in the
wire and assuming a wire sag of 75 mm for the 100 mm
slab thickness, the allowable load per square foot
is 25 kPa.
In a Metropolitan floor example using the simpli-
fied formula, with pairs of 12-gage (2.8 mm diameter)
wire at 25 mm spacing and a 1830 mm span, the total
load capacity is 36 kPa. Using the basic formula, the
total load capacity is 22 kPa.
In short, the catenary floors show the excess capac-
ity required to pass the load test.
3.4 Beam analysis
Beam theory is generally more complex than arch or
catenary theory as long as fixed supports are assumed.
In addition, the design of reinforced concrete in the
United States at the time of the testing was ham-
pered by over-conservative allowable stresses. When
the New York code was amended in 1911 to allow use
of reinforced-concrete floors proved by analysis rather
than by testing, the maximum compression in the con-
crete was limited to 4,5 MPa, the maximum tension
in the reinforcing was limited to 110 MPa, and n, the
ratio of the elastic moduli, was fixed at 15. (Waite
1914) Only elastic concrete theory was in common
use at the time, and using the notation common for
that theory, the location of the neutral axis from the
compression face of the slab is kd, where d is the depth
from the compression face of the slab to the centroid of
the reinforcing and the ratio k = (2 pn + p2 n2 )0.5 − pn,
where p is the reinforcing ratio.The arm of the resisting
135
moment (the distance from the centroid of the tri-
angular elastic compression block to the centroid of
the reinforcing) is jd, where j = 1 − k/3. The resisting
moment is the lesser of the two calculated moments
Ms = As fs jd and Mc = fc kjbd2 /2, where Ms is the max-
imum resisting moment if the slab is under-reinforced
and steel yield controls, As is the reinforcing area in
square inches per foot, fs is the allowable tensile stress
in the steel, Mc is the maximum resisting moment if
the slab is over-reinforced and concrete crushing con-
trols, and fc is the allowable compressive stress in the
concrete.
For example, the code-enforced maximum stress
values would show a 100 mm stone-concrete slab
with 6 mm diameter bars at 150 mm on center, and
spanning 2440 mm, to have a maximum allowable
moment of 4.9 kN-m/m, or a maximum total load of
6.5 kPa, compared to the 7.2 kPa allowed from testing
the same floor. This is not a hypothetical compari-
son: the Monier Floor System, which was essentially
a numerically-designed, modern reinforced concrete
slab as used in the in Europe, was tested for use in
New York by 1900. If the current American Concrete
Institute code and ultimate-stress theory is used on the
same slab, assuming the 14 Mpa concrete often used
in that era and a steel yield stress of 138 MPa based on
the weakest available reinforcing steel bars, the maxi-
mum allowable moment is 8.5 kN-m/m, and, assuming
that the dead load to live load ratio is approximately 1,
a maximum unfactored total load of 8.0 kPa.
The Columbia and Roebling Flat Slab floors can
be analyzed like any other reinforced slab as long as
certain restrictions are observed: the bars must be con-
verted to a cross-sectional area per unit width of slab,
the slab strength must be adjusted to match the mate-
rial, and the lack of deformations must be accounted
for by a check on the development length of the bars.
The Expanded Metal Floor can be analyzed using
ordinary reinforced-concrete formulas and converting
the cross-sectional area of the expanded-metal sheet
into the bar equivalent.
The analysis of the Rapp Floor, a beam-type floor
that did not rely on reinforced concrete, is simpler. The
bricks spanning 200 mm and the light-gage Ts span-
ning the beam-to-beam spacing can be checked for
maximum stress using a simple beam formula.
4 CONCLUSIONS
The simplest result of the analysis is the conclusion
that these floors can be demonstrated to pass mod-
ern requirements for fire resistance and load capacity.
In the course of renovation projects, many engineers
and contractors prefer the removal of unfamiliar and
archaic structural elements rather than their reuse, even
if no damage is visible. The explanation for this behav-
ior is often “better to be safe than sorry,” as if the reuse
of existing structure that has functioned properly for
decades is somehow unsafe. It must be emphasized
that unknown structural capacity and fireproofing are
not excuses for the wholesale demolition of historic
fabric that often takes place, leaving a historic facade
covering an essentially new building. It is incumbent
upon the engineers involved to investigate unfamil-
iar structures and find methods by which they can be
analyzed.
This is not to say that there are no difficulties in
reuse once an archaic floor is analyzed. There may be
damage to the floors, particularly with the more frag-
ile systems such as tile arches, that reduce their load
capacity or fire protective abilities. Modern seismic
analysis of a building frame may depend on diaphragm
action of the floor, which may or may not be available
depending on the details of original construction. The
need for a diaphragm implies a second investigation
into details and a second analysis, not automatic dis-
qualification. Finally, alterations accompanying reuse
may require special details, such as providing new
anchorage for catenary floors when new openings are
cut for circulation or HVAC shafts. Some of the floors
are particularly vulnerable to damage during alter-
ation: theTs in a Rapp floor can shift when an adjoining
portion of the floor is removed.
In a broader sense, the continued viability of many
of these floors emphasizes that the products of empir-
ical design can still be used today. The rise of “scien-
tific” numerical analysis in the twentieth century can
obscure the value of older forms of engineering design,
even as we still rely on some of the methods and data
produced by those older forms.
REFERENCES
no author. 1889. “A New System of Fireproof Floor Construc-
tion.” Engineering News, November 9, 1889: 434–35.
no author. 1895. “The Metropolitan Concrete and Wire
Floor.” Engineering News, November 14, 1895: 333.
no author. 1897. “Comparative Standard Fireproof Floor
Tests of the New York Building Department.” Engineer-
ing Record, September 18, 1897: 337–40; September 25,
1897: 359–63; October 2, 1897: 382–87; October 9, 1897:
402–405.
no author. 1899. Building Code of the City of New York. New
York: New York City Bureau of Buildings.
no author. 1901. Building Code of The City of New York. New
York: Department of Buildings.
no author. 1905. Report of the Bureau of Buildings of the City
of NewYork for the Borough of Manhattan for the Quarter
and Year Ending December 31, 1904. New York: Martin
B. Brown Co., 1905.
136
no author. 1919. Pocket Companion for Engineers, Architects
and Builders Containing Useful Information and Tables
Appertaining to the Use of Steel. Pittsburgh: Carnegie
Steel Company.
no author. 1944. American Welded Wire Fabric for Concrete
Reinforcement. Pittsburgh: United States Steel Corpora-
tion.
no author. 1981. Evaluation of Reinforcing Steel Systems in
Old Reinforced Concrete Structures. Chicago: Concrete
Reinforcing Steel Institute.
no author. 1986. “Technical Note 31a – Structural Design of
Brick Masonry Arches.” Reston, Virginia: Brick Industry
Association.
no author. 2000. Fire Ratings of Archaic Materials and
Assemblies: Guideline on Fire Ratings of Archaic Mate-
rials and Assemblies. Washington, D.C.: U.S. Department
of Housing and Urban Development, Office of Policy
Development and Research.
no author. 2004. Building Code of the City of New York. New
York: New York City Department of Citywide Adminis-
trative Services.
no author. 2007a. New York City Building Code. New York:
New York City Department of Citywide Administrative
Services.
no author. 2007b. “Standard Test Methods for Fire Tests of
Building Construction and Materials,” ASTM E119–07a.
West Conshohocken, Pennsylvania: American Society for
Testing and Materials.
Birkmire, W. 1898. The Planning and Construction of High
Office Buildings. New York: John Wiley & Sons.
Buel, A.W. & Hill, C.S. 1906. Reinforced Concrete, 2nd
edition. New York: The Engineering News Publishing Co.
Collins, G.R. 1982. “Guastavino y Moreno, Rafael and Guas-
tavino y Esposito, Rafael,” Macmillan Encyclopedia of
Architects. New York: Macmillan Publishing Co.
Guastavino, R. Jr. 1892. “Hollow Cohesive Arch,” United
States Patent 471,173. Washington DC: United States
Patent and Trademark Office.
Freitag, J.K. 1921. Fire Prevention and Fire Protection, 2nd
edition. New York: John Wiley & Sons.
Freitag, J.K. 1899. The Fireproofing of Steel Buildings. New
York: John Wiley & Sons.
Hill, G. 1895. “Tests of Fire-Proof Flooring Material,” Trans-
actions of the American Society of Civil Engineers 34:
542–568.
Hool, G. A. 1913. Reinforced Concrete Construction,
Volume II: Retaining Walls and Buildings. New York:
McGraw-Hill.
Perrine, H. & Strehan, G. 1915. “Cinder Concrete Floor
Construction Between Steel Beams.” Transactions of the
American Society of Civil Engineers. 79: 523–621.
Waite, G.B. 1914. “Cinder Concrete Floors.” Transactions of
the American Society of Civil Engineers. 77: 1773–1823.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Intervention limits in floor wood structure restoration of a historical

building

R.T.M. Ribeiro & M. Hoirisch

PROARQ/FAU/UFRJ – Federal University of Rio de Janeiro, Rio de Janeiro, Brazil

ABSTRACT: In 1995, a routine inspection carried out by the Restoration Department of Federal University
of Rio de Janeiro disclosed instability of wood floors in a room located in a Brazilian 19th century building.
This paper presents a concise report of this structural degradation. It focuses mostly on the project of structural
restoration of areas that brought about the practice works to be implemented in a historical building listed by
the Brazilian Historical Artistic Heritage Institute. Not limiting this study to a description of the strengthening
work stages, this research analyses the actions adopted to restore the damage caused by termites in wood beams,
trying to evaluate the adequacy in this specific work of the guidance of international charters of architectural
heritage. The main goal of this paper is to analyse the criteria that oriented the feasible restoration and the limits
of this intervention.

1 INTRODUCTION preservation: to value properly its original construc-

The conservation criteria, available in charters, agree-
ments and international declarations, are applied to
the restoration program of historical buildings. One
of the most controversial matters by the conservation
architects and heritage consultants’ standpoint is the
insertion of materials and constructive techniques in
historical heritage.
The debate involving the replacement and/or inser-
tion of degraded materials interplayed with the her-
itage authenticity aspects and its values will herein be
presented.
In order to exemplify the technical issues, the case
study of a wood floor restoration will be undertaken.
The practice work has been performed in the city of Rio
de Janeiro, in a historical building, widely recognized
for its artistic and cultural values.
This paper describes the structural restoration of
those wood floorings. However, its main target is
to demonstrate that the project here displayed has
respected the intervention limits of restoration char-
ters. In order to prove it, this study will show the reha-
bilitation phases followed by the suitability between
restoration theories and practice.
2 THE ISSUE OF REPLACEMENT AND/OR
INSERTION OF THE BUILDING NEW
ELEMENTS
By intervening in an ancient building, one must bear
in mind its primary aim, in order to guarantee a proper
tive elements. As a consequence, when dealing with
the restoration of the whole building or parts of it,
materials should only be replaced and/or added if
there is a technical requirement. The use viability of
the architectural monument would be guaranteed, as
well as its preservation, preventing any imitation or
counterfeit.
The restoration project must start by the identifica-
tion of the elements to be preserved and those suitable
for adaptation and/or removal. In order for this to
be accomplished, a precise and meticulous register of
the building history is needed. Besides, the physical
status and its current conditions regarding construc-
tive materials require special care when it comes to
pinpointing its frailties.
It must be underlined the need of a detailed survey
of new materials and technologies insertion, adding
distinct physical proprieties to the building, when com-
pared to the existing ones. One must also analyse
the utilisation of traditional materials and construction
techniques in a historical heritage.
One of the resolutions of the Athens Conference
in 1931 was the use of modern techniques and mate-
rials in consolidation work of historical buildings.
However, the strengthening “should whenever possi-
ble be concealed inasmuch as the aspect and character
of the restored monument may be preserved” (In:
http://www.icomos.org).
The Venice Charter, dating from 1964 goes in the
article 12: “Replacements of missing parts must inte-
grate harmoniously with the whole, but at the same
time must be distinguishable from the original so

137
that restoration does not falsify the artistic or his-
toric evidence”. In addition to that, in the article 10:
“Where traditional techniques prove inadequate, the
consolidation of a monument can be achieved by the
use of any modem technique for conservation and
construction, the efficacy of which has been shown
by scientific data and proved by experience” (In:
http://www.icomos.org).
The Italian Charter of Restoration of 1972
approaches a prevailing architectonic restoration stan-
dard: “the replacement and possible integration of
ornaments [. . .] ought to differ from the existing ele-
ments, implying the material and surfaces to stand out,
making them distinguishable from the originals”. It
also mentions the interventions must be implemented
for fear no doubt is left and by the most discrete manner
(In: http://www.icomos.org).
The General Conference of UNESCO at Nairobi
in 1976 recommends researches and systematic sur-
veys on the use of modern techniques in conserva-
tion works, as well as studies about craftsmanship
techniques, important to the building safeguard.
3 THE MATTER OF AUTHENTICITY IN
INTERVENTION
The legitimacy of the restorative intervention in his-
torical buildings was debated since the very beginning
of restoration theories. Authors like Boito stated as a
fundamental concept the distinction of the new parts
from the existing ones, as well as emphasizing the pru-
dence of limiting intervention actions to the minimum
required. All material to be added or replaced should
have a different feature, in the type of material to be
used, or should bear a plaque, indicating the successive
architectural alterations. The new parts to be inserted
should exhibit simple lines, being perfectly balanced
with the ancient building.
Later on, the Venice Charter has established that the
new elements ought to differ from the existing parts,
and the same criterion was sustained by the Italian
Charter of Restoration.
This differentiation theory was strengthened in later
congresses, aiming at the visibility of the interven-
tions to be carried out, refusing the idea of utilisation
material and techniques similar to the existing ones,
preventing historical fraud.
In his book Theory of Restoration written in 1963,
Brandi (2004) mentions the judgement of counterfeit.
The author refers to the essential attributes the subject
should possess, but is not able to earn: the falseness is
in the judgement and not in the purpose. One of two
identical objects may be considered a fake, according
to its intention.
Pursuant to him, the distinction between copy, imi-
tation and counterfeit is not in the specific diversity
138
of the production methods, but otherwise. Hereupon,
he points out two situations in which an intervention
could be considered a counterfeit:
(1) Production of a purpose with the specific inten-
tion of deceiving either the period, the material
coherence or the author, or
(2) Diffusion of an object, not necessarily intended
to deceive as if it were an authentic work, from
that period; by non-authentic authors and produc-
ers different from the genuine one, and made out
of distinct materials. In other words, it has been
created with no counterfeit intention, but meant to
be divulged as the original.
According to Brandi, to get the counterfeit judge-
ment, it is necessary to prove the swindle.
This issue has been developing, oftentimes being
debated in international congresses and seminars. In
Japan, November 1994, the Nara Conference, based
upon the Venice Charter, has discussed the authen-
ticity issue linked to the intrinsic heritage value and
the cultural identity of the region in which it is
placed.
Following the international stream, Brazil has
hosted the V Regional Meeting of ICOMOS in 1995.
A regional document on authenticity of the southern
cone countries was the result of this assembly. Brasilia
Charter established that authenticity of values is man-
ifest, supported and preserved in the truthfulness of
the heritage that we receive and transmit. Interven-
tion must rescue building or site character, underlining
its authenticity without transforming its essence and
balance, avoiding extraordinary actions and enhanc-
ing its values. The Charter avows the quality of the
treatments as fundamental and all new elements intro-
duced must be both reversible and harmonious with the
whole.
The issue of authenticity is rooted in the values
attributed to the heritage. The judgement of value
was broadly discussed in “The modern cult of mon-
uments” by Riegl in 1903. This matter was the main
point of Nara Conference, in 1994, related to scientific
studies on cultural heritage, conservation and restora-
tion planning and procedures for the World Heritage
Convention.
The Nara Document claims that the authenticity
emerges as the chief criterion when it comes to values
attribution. “The understanding of authenticity plays a
fundamental role in all scientific studies of the cultural
heritage, in conservation and restoration planning”.
(http://www.international.icomos.org).
The interventions carried out in the University
Palace of UFRJ – Federal University of Rio de Janeiro
will be herein analyzed based on these concepts, more
specifically the floor restoration of three rooms located
in this building.

Figure 1. University Palace. Photo: Soares, B., 2007.
4 THE EDIFICATION SUBJECT OF
INTERVENTION
The University Palace (Fig. 1), heritage building
located in the University Campus of Praia Vermelha
houses some faculties of Federal University of Rio de
Janeiro. It is a neoclassical style building, situated in
the city of Rio de Janeiro, listed in Brazilian Historic
and Artistic Heritage Institute.
The building exhibits restored rooms, ceilings and
walls smartly decorated, frequently used for concerts,
congresses, academic, technical and administrative
activities and solemnities. The rectory appointed these
areas to the academic community and to Rio de Janeiro
inhabitants’ use.
4.1 Restoration of wood floors
The loose fixation between the beam and the clapboard
of the floor in Education Faculty of Federal University
of Rio de Janeiro lead to a prospective action in order to
find out its cause. The primary visual survey pointed
to some clapboards disconnected, next to the walls.
These supporting walls were wet. By analyzing the
roof conservation status over these areas, it was pos-
sible to detect clogging of the rainwater down pipe,
leading to an overflow above the gutters, causing fre-
quent infiltrations in the internal face of the wall and
also percolating its core, reaching the supporting clap-
boards wood beams. This humidity has created ideal
conditions for the termite and fungus infestation in the
floor. Around 60% of the wood beams were infected
by termites. The micro-organisms and urban plagues,
as we know, burgeon at humid zones, promoting the
rottenness of part of this beams, reducing partially or
drastically its original section (Fig. 2).
The Restoration Department was designated to
work on a strengthening project of three wood floor
rooms: (1) Science and Culture Forum coordinator
139
Figure 2. Assessment discloses a drastically reduced section
of some beams. Photo: Hoirisch, M., 1995.
Figure 3. Location of interventions, DAMSTRUC (2005).
cabinet, (2) Coordination of Post-Graduation Program
of Faculty of Education Coordination. (3) Golden
room (Fig. 3). Under the supervision of the head archi-
tect, a multidisciplinary team was gathered, composed
by urban plagues biologists experts, structural engi-
neers, and a proficient firm specialized in wood struc-
tural restoration. The practice work evolved, therefore,
in three lines. In the first one, biologists determined
the termites infecting species, in order to guarantee
the efficiency of the plagues control program. In the
second one, structural engineers calculate the remain-
ing section of each wood beam, to make the structural
assessment. Finally, the practice restoration work was
carried out according to the engineers strengthening
project.
The wood beams were properly inspected. For this
purpose, the clapboards of the rooms to be restored
were removed after being mapped. The most vulnera-
ble zones, located in the extremities and under the roof
gutter called for extra care. The tongue and groove con-
nections of the clapboards to the wood beams were
previously done by means of long forged iron nails.
They were then replaced by steel screws. This system
allowed the access to the beams, as long as from that
time on the wood beams assessment was to be made
Figure 4. Fixation of floor beams in Science and Culture
Forum coordinator cabinet. Photo: Hoirisch, M., 1996.
Figure 5. Scheme showing replaced beams.
easier. The removal of the nails frequently destroys par-
tially or completely the clapboards mortises, whilst the
use of screws allows their removal without damage,
making possible the application of baits and products
to promote the termite and other wood degradation
vectors.
The Figure 4 shows the ongoing floor restoration
in the Science and Culture Forum coordinator cabinet,
University Palace.
When defining the material to be adopted in this
practice restoration work, the wood seemed to be
the best choice. Studies pointed out which Brazilian
species bore elasticity modulus fine enough to encom-
pass the room wide extension. Special cares were taken
so to guarantee the proper level of the new beams dry-
ness. Stoves were used to assure its efficiency, when
replacing and strengthening the pre-existing ones.
The Figure 5 shows a scheme of the replaced
strengthening beams in the Science and Culture Forum
140
Figure 6. Golden room after restoration. Photo: Soares, B.,
2007.
coordinator cabinet. This proceeding has been likewise
carried out in the other two restored chambers.
Regarding the project of floors structural strength-
ening, the Restoration Department had aimed in the
first place at human’s life safeguard. Therefore, each
pre-existing structural wood beam has been assessed,
to pinpoint those to be strengthened or replaced.
Another special care was required from the biologists
responsible for the termite control. The product used in
the disinfection process should not be toxic, to avoid
harming the docents, students, employees and other
haunters of the building, since the restoration practice
has been mostly achieved during academic year.
5 FLOOR RESTORATION SURVEY:
INTERVENTION LIMITS
When opting for wood to be used in beams and
clapboards replacing and strengthening, this was a
natural choice, as the material embodies the concept
of the original project. However, the species were
picked out according to their elasticity moduli, not
necessarily taking into account the existing species
found in the room. However, wood benefiting was
achieved by modern extraction and sawing techniques.
This proceeding abides restoration documents, like the
Venice Charter and the Italian Charter of Restora-
tion. This case study has respected the former: the
replacement and strengthening of the wood beams
has been integrated to the whole. However, concern-
ing the intervention period, the difference between
the inserted material and those pre-existing ones
can only be distinguished by the wood benefiting
modern system.
When it comes to the replacement of the iron forged
nails by steel screws fixing the wood floors next
to the walls, the distinction between the pre-existing
technique and the modern one is fully catered.
 The necessary strengthening in some of the tim-
bers was accomplished by the “sandwich” technique,
putting two new pieces parallel to the pre-existing
ones.
The remaining doubt refers to the highly perishable
feature of the wood, mainly when utilized in humid
tropical weather (Rio de Janeiro relative humidity is
above 70% all year round). If this high humidity level
foments the termite infestation why had not the tim-
bers been replaced by a more long lasting material,
like steel?
The option for metallic beams could pinpoint the
exact intervention period, and the advices of the her-
itage charters would be abided. A future survey could
allow the identification of steel utilization as dating
from our times. An extra benefit would consist of the
prolongation of the inserted beam durability, when
compared to the pre-existing one. By the way, one
of the main goals in restorative interventions is to
hinder new interventions on a short span. The replace-
ment of wood beams by metallic ones could be threat
a future, for the utilization of modern materials and
techniques would make a “patchwork” out of the
heritage building. Losing the constructive original fea-
tures dated from its erection period would mean losing
its authenticity as well.
Concerning its physical aspect, a primary mapping
of the pre-existing wood pieces had been performed,
before the removal of the floor wood clapboards,
allowing its fixation to the previous position. The new
inserted wood pieces followed the original pattern, the
usual proceeding in these cases.
6 CONCLUSION
The current theories tell the insertions from the whole,
and were applied to this restoration work. The mate-
rial had not been changed, likewise preserving the
authenticity of the building. Moreover, the durability
of the restoration practice carried out may disagree
with future conservation decisions.
141
The key point about the restoration proceedings def-
inition and the materials to be utilized depends upon
the limits of the intervention; in other words, up to
what extent is it possible to intervene in a heritage?
This intervention ought to be limited by the preser-
vation of a historical building, its values, authenticity
and identity of the local cultural heritage where the
monument belongs. One must dedicate special care
to the values to be preserved and what would be the
essence of restoration practice.
Regarding our heritage actions, this is a concept
and paradigm changing era. The target should not only
be the material safeguarding, but mainly the historical
building image and authenticity, as a cultural identity
offspring.
REFERENCES
Boito, C. 2003. Os restauradores. São Paulo: Ateliê Editorial.
Brandi, C. 2004. Teoria da restauração. Trad.: Beatriz M.
Kuhl. São Paulo: Ateliê Editorial.
Hoirisch, M. 2005. Restauração das Estruturas de madeira
em prédio Histórico da Universidade Federal do Rio de
Janeiro. In: 4th International Conference on the behavior
of damaged structures. Anais do DAMSTRUC 2005, João
Pessoa, p. 1159 a 1178.
Prudêncio, W.J. & Ribeiro, R.T.M., 1998. As bases éticas
da restauração do patrimônio cultural. In: Del Rio, V.
(org.). Arquitetura: Pesquisa & Projeto. Rio de Janeiro:
Pró-Editores.
Ribeiro, R.T.M., Pimentel, V.L., Coelho, C.M.T.,
Andrade, I.E. & Melo, C.M.S. 2005. Olhares sobre o
patrimônio edificado: o conceito de valor. In: XXIII Sim-
pósio Nacional de História: Guerra e Paz / Associação
Nacional de História – ANPUH. Anais. . . XXIII Simpósio
Nacional de História: Guerra e Paz. Londrina: Editora
Mídia, 2005.
Riegl, A. 1999. El Culto Moderno a los Monumentos.
Madri: Visor Dis, S.A.
http://www.icomos.org/
http://www.nardinirestauro.it/download/pdf/carte/Carta_
restauro%20_1972.pdf
http://www.international.icomos.org/naradoc_eng.htm
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

The safety of historic load-bearing structures

B. Szabó
Technical University of Cluj-N., Romania

I. Kirizsán
Built Heritage Research and Design Centre UTILITAS, Cluj-N., Romania

ABSTRACT: The safety of the historic load-bearing structures – being a part of the general issues regarding the
load-bearing structural safety – is a very topical problem for the sustainable management of the built environment.
Being to a great degree related to the objective (the characteristics of the historic load-bearing structures, as well
as of the historic and/ or current structural environment) and subjective premises of providing safety (social-
political conditions, the level of technical knowledge, research possibilities, quality control system, available
materials, aptitude etc.), the historic load-bearing structures need interventions that improve their safety and
preserve their intrinsic heritage values as completely as possible. If the “updating” of their safety requires a
technical-economic optimization, then this optimization is greatly influenced by the objective and subjective
premises of providing safety and by the condition of a minimal loss from the heritage values.

1 PREAMBLE 2 THE SAFETY OF THE ENGINEERED

The safety of the historic load-bearing structures is a
performance expectation, which requires an accept-
able behaviour of the historic load-bearing struc-
tures in the resistance and stability limit state. This
behaviour is anticipated through (i) investigations for
the determination of the extant safety level, (ii) inter-
ventions needed for the correlation of the extant level
with the contemporary safety requirements, (iii) the
maintenance of the acceptable behaviour during use.
The acceptable behaviour of the historic load-
bearing structures in the resistance and stability limit
state is necessary because: (i) human lives and mate-
rial goods have to be protected, (ii) heritage values are
to be preserved.
An important remark has to be made on the her-
itage values: the historic load-bearing structures have
intrinsic heritage values, beside the worn ones, such as
artistic elements (murals, painted wooden elements,
carved stone elements etc.) or architectural details
(carpentry, wainscot, floors etc.).The intrinsic heritage
values are connected both with the empirical-intuitive
concept of the load-bearing structure and the historic
materials and traditional technologies, through which
they are put in operation [5].
Thus the three phases of ensuring the acceptable
behaviour – the result of which is the safety of the
historic load-bearing structures – take the heritage val-
ues always into consideration, including the intrinsic
values of the historic load-bearing structure.
LOAD-BEARING STRUCTURES [1]
The major problems regarding the safety of the engi-
neered load-bearing structures are the ones related
to the form of idealization (determinist, statistical or
strategic). Not only the bases of the problem, but also
the decisions regarding the solutions have to be in
direct connection with the applied form of idealization.
The semi-probabilistic methods – that use an infor-
mation infrastructure brought almost to perfection –
create a more and more adequate framework for the
handling of safety issues regarding the engineered
load-bearing structures.
The major problems, which this field has to con-
front, are ethical or related to the insufficiencies in
the technical-scientific knowledge, or result from the
errors due to the inappropriate quality of the materials
and from the omitted, unpredictable loads etc.
3 PREMISES FOR THE SAFETY OF HISTORIC
LOAD-BEARING STRUCTURES [2] [6]
3.1 The objective premises of the safety of historic
load-bearing structures
The characteristics of the historic load-bearing struc-
tures, as well as those of the historic and/ or current
load-bearing structural environment are considered to
be the objective premises.

143
Figure 1. The structural sensitivity of the masonry to
moisture, freezing cycles.
3.1.1 The characteristics of the historic
load-bearing structures
These characteristics are well-known, including their
share in the safety of historic buildings; for example:
(a) the masonry structural sub-units:
– are rigid to gravitational actions, if these lead to
compression and if due to these no differential
settlements are produced;
– have minimum reserves for taking over the tension,
thus they are very sensitive to dynamic actions;
– are sensitive to climatic actions, moisture, freezing
cycles etc.;
– are sensitive to natural or artificial corrosion
factors;
– the fatigue, the seismic energy-absorption of the
masonry mortars and the consequences due to
corrosion can be tested in the mass of the mason-
ries with difficulty, the evaluation of their state
parameters being possible only in quality.
(b) the timber structural sub-units:
– are load-bearing structures, which work in an elas-
tic limit state, being suitable for both static and
dynamic actions;
– are subject to biological decay, some of the dam-
ages – like the ones caused by the moisture of
the supporting structure of the floors – causing
an immediate, unforeseeable collapse.
(c) being hundreds of years old, the durability of the
historic load-bearing structures is a certainty.
(d) the state of the historic load-bearing structures,
the safety reserves are continuously changing, the
periodic re-evaluation of safety being compulsory.
3.1.2 The historic and/or current characteristics
of the historic load-bearing structural
environment
These are of great significance, influencing to a
great degree the safety of load-bearing structures.
The current characteristics of the historic load-bearing
144
Figure 2. The structural sensitivity of the timber structure
to biological decay.
structural environment differ from those of the historic
environment:
– the artificial corrosion factors changed essentially
both in quality and quantity – the historic build-
ings being extremely vulnerable to some of these
actions;
– the level of the ground water was essentially modi-
fied in the 19th century, influencing the foundation
conditions and the historic foundations to a great
degree;
– the impermeable flooring and pavement basi-
cally changed the hydrostatic balance of the load-
bearing structural environment surrounding the
buildings, increasing the humidity (and dissolved
salts) absorbed by the load-bearing structures due
to the lack of horizontal waterproofing;
– the vibrations due to traffic are very damaging,
especially for the masonry mortars – operating
almost constantly, making also the efficiency of the
interventions questionable;
– the vibrations due to industrial or household appli-
ances have an effect similar to the one of the
vibrations due to traffic.
3.1.3 Special historic conditions relating to the
historic load-bearing structures
In the course of history the ecclesiastic constructions
(especially the churches) have been built according
to prototypes. These «typical projects» are made for
exactly formulated functional requirements, but are
elaborated also (more or less consciously) for technical
requirements, demanded by the climatic, seismic, cor-
rosive historic conditions etc. of the place, where the
prototype was created. The prototype being generally
propagated – for example by the Cistercian monks –,
both the functional and technical characteristics are
entirely transmitted, without being adapted to the local
technical climatic or soil conditions (shrinkable clays,

Figure 3. The level of the historic technical knowledge –
the empirical-intuitive concept in the Czech Republic.
slope stability etc.), as well as to the seismic require-
ments. If the technical conditions of the location are
very different from those of the prototype, the safety
of the constructions will be reduced. Due to the fact
that the ecclesiastic constructions are built by a com-
munity quite rarely (once in 3–800 years), the local
experience is much harder to gain.
Much more secular constructions are built – the
imported model is even adapted to the local con-
ditions – even if the initial “imports” are entirely
transposed. Due to the very frequent subsequent trans-
formations, as well as the large number of similar
constructions built in a relatively short period of time,
the local (regional) experience is much easier to gain,
creating (intuitively) load-bearing structures that are
resistant to seismic actions and actions occurring due
to shrinkable clays and unstable slopes etc.
3.2 The subjective premises for the safety of
historic load-bearing structures
The category of the subjective premises consists of
the social-political conditions, the level of techni-
cal knowledge, research possibilities, quality control
system, available materials, aptitude and others.
These premises are equally present in the building of
the constructions, the former subsequent interventions
145
Figure 4. Historical skill in England.
(«historic» ones) and those present in our time
(«contemporary» ones).
The subjective premises influence the safety of
the load-bearing structures both in a positive and a
negative way.
3.2.1 The historic subjective premises
(a) the social-political conditions:
– cheap and abundant manpower, often compelled
to work;
– wars that produce loads, which were not taken
into consideration at the conception of the load-
bearing structure;
– the disposition of the society to protect the her-
itage values (a more and more intense attitude
since the Renaissance, the doctrinal approach of
which was not always favourable to safety).
(b) the level of technical knowledge:
– the empirical-intuitive concept of the historic
load-bearing structures – without engineered
calculations;
– the scientific basis of the technical regulations
in the historic period of interventions.
(c) available materials:
– mainly local materials, limited transportation
possibilities (the load-bearing structures built of
these materials are not always suitable for local
loads);
– building materials with a reduced tensile stress
(the historic load-bearing masonry structures
are very sensitive to dynamic actions).
– building materials with a reduced resistance to
corrosion – the historic conditions were not
familiar with contemporary corrosion.
(d) skill:
– the traditional technologies are developed for
the processing of the local materials (there are
situations, in which the traditional processing
offers a higher quality);
– the manual operations present a development
in time, going into decline only after the great
industrial revolution from the 19th century.
3.2.2 Contemporary subjective premises
(a) the social-political conditions:
– wars – producing loads, which were not taken
into consideration at the conception of the load-
bearing structures;
– the consumer society – the contemporary expec-
tations regarding the durability are different
from the historical ones;
– globalization – can also have positive effects:
the most recent standards spread quickly all over
the world;
– the disposition of the society to protect the
heritage values – laws, which protect (to a
higher or lesser degree) the historic load-bearing
structures;
– the (higher or lower) level, at which society
acknowledges the research and testing activi-
ties on the load-bearing structures that require
a qualified manpower.
(b) the level of technical knowledge:
– the level of the general technical culture and the
knowledge related to historic buildings (the high
level of knowledge within the contemporary
construction industry also requires scientific
sensitivity towards the historic buildings);
– the level of technical knowledge related to his-
toric buildings has to refer to the load-bearing
and non-bearing structural issues, building
physics and biology etc. as well;
– the average specialist (researcher, designer, con-
tractor) is expected to have technical-scientific
knowledge regarding both the empirical-
intuitive concept, historic building materials
and the traditional technologies; an informa-
tion infrastructure has been available for 12–15
years, through which the modelling issues can
be widely mastered (especially in the case of
wooden load-bearing structures) – these have to
be known and used;
– the history of the load-bearing structural mate-
rials and subunits has to be studied: floods,
seismic events, fires (the micro-fissures in the
stone of the masonry resulting from the fire
hundreds of years ago etc.);
– the conceptual compatibilities of the interven-
tions demand the knowledge of the empirical-
intuitive concept of the foundations, the sup-
porting structures, the historic floor and roof
146
Figure 5. The use of the traditional technologies in
Romania.
structures, extending it in space and time over
the regions where the specialist wants to pursue
his activity (strong and weak points, frequent
deteriorations, characteristic historic interven-
tions etc.);
– the features of the historic building materials
have to be identified – the incompatibility of the
materials can be the cause of inefficient inter-
ventions – the historical stone-quarry have to be
identified, so as to use materials similar to the
original ones;
– the traditional technologies have to be known as
well, the interventions need to be compatible, in
the technological field included.
(c) The level of knowledge regarding the heritage
values:
– those have to have a knowledge of the her-
itage values, who wish to perform interventions
regarding the safety of the load-bearing struc-
tures – the issues regarding the safety and
heritage values have to be solved simultaneously
in the case of historic buildings;
– we have to know the historic load-bearing struc-
tures both with the intrinsic and the worn her-
itage values (artistic elements: murals, coffered
ceilings, stuccoes, stone frames etc., building
structures: windows and doors, flooring, tinning
etc.);
– the worn heritage values have to be considered,
as they can eliminate a part of the consolidation
works – the allowed limit states have to take
into consideration the expectations regarding
the heritage values;
– the intrinsic heritage values of the load-bearing
structures (related to the empirical-intuitive
concept, the historic materials and the tradi-
tional technologies) can reduce the intervention
possibilities;
– we have to know the doctrinal issues regarding
the interventions carried out on historic build-
ings, especially the ones related to the historic
Figure 6. The use of the traditional technologies, interven-
tion on a fractured masonry vault.
load-bearing structures; we have to choose
between preservation, restoration, retrofitting
or re-construction – only that specialist can
choose, who knows the terminology;
– the doctrinal attitudes have to be adopted care-
fully: for example the Mediterranean fashion
of natural surfaces of the 1950s and ‘60s
has always had negative results in continental
Europe, as the number of freezing cycles is
incomparable in these places.
(d) available materials:
– we have to take into consideration that the cur-
rent materials do not reach the quality level
of the historic ones (the solid brick with the
old format, the lime or the timber from the
planted forests are only a few examples for the
poor quality of the recently produced traditional
materials);
– the contemporary building materials are pro-
duced so as to be compatible with other recently
produced materials – their compatibility with
the historic building materials has to be stud-
ied in each case; laboratories, specialists, the
awareness of the issue, as well as the will to
take measures are thus needed.
(e) skill:
– one of the tasks is the studying of the traditional
technologies, avoiding the technological incom-
patibilities between the traditional technologies
and the intervention;
– we need to study the technological conditions
for the reinforcement to begin immediately to
work (reinstating the original geometry with the
introduction of counter-bending etc.).
(f) research possibilities:
– the basic research programs are necessary [the
slow transformation of the slopes, the mechan-
ics of the earth banks and of the poor-quality
foundation soils that have been loaded for
centuries (the soil mechanics regulations deal
147
Figure 7. The use of the traditional technologies, interven-
tion on a historic timber roof construction.
only with ”virgin lands”), the behaviour of the
historic timber – rheological aspects etc.];
– the use of the research results of the different
fields in the interventions carried out for the
safety of the load-bearing structures: art history
(the building of a construction in phases – the
source of damages in the load-bearing struc-
tures), archaeology (the former foundations –
damaging factors), building physics (moisture
containing dissolved salts – causing damages)
etc.;
– providing laboratories for those industrial
branches that carry out the interventions on
historic buildings, where the quality of the load-
bearing structural materials would be studied,
creating conditions for the routine experiments:
e.g. stone is generally used without a quality
check;
– in-situ testing of the building materials and a
qualified personnel.
(g) the quality control system:
– the historic load-bearing structures have to be
levelled with the current safety standards – this
is impossible without a quality control system,
taking into consideration the aforementioned
deficiencies;
– the personnel designated for the quality control
has to have thorough knowledge of the historic
load-bearing structures;
– the quality control laws have to deal with
the characteristics of the historic load-bearing
structures in detail;
– uniform measures are recommended regarding
the safety of the built environment – taking of
course into consideration the characteristics of
the historic buildings (it is not efficacious, if a
group of institutions takes care of the contempo-
rary buildings and another group watches over
the historic ones).
4 ENSURING THE SAFETY OF HISTORIC
LOAD-BEARING STRUCTURES [6], [7],
[8], [9]
The acceptable behaviour of the historic load-bearing
structures in the resistance and stability limit state –
safety measurement of load-bearing structures – is
anticipated through: (i) investigations for the deter-
mination of the extant safety level, (ii) interventions
needed for the correlation of the extant level with the
contemporary safety requirements, (iii) the mainte-
nance of the acceptable behaviour in the resistance
and stability limit state during use.
4.1 Investigations for the determination of the
extant safety level
The investigations for the determination of the extant
safety level of the historic load-bearing structures are
related to technical, as well as to heritage issues.
The technical investigations take place during [9]
the diagnosis phases carried out on the historic load-
bearing structure (characterized by the approach of
the structural issues from the point of view of build-
ing mechanics, physics and biology, as well as of
soil mechanics and building services engineering),
focusing their attention on the safety aspects: (i) the
identification (survey and presentation) of the historic
load-bearing structure; (ii) finding the insufficiencies
in the historic load-bearing structure; (iii) the formu-
lation and testing of the performance expectations at
the historic load-bearing structure; (iv) the determi-
nation of the causes of insufficiencies at the historic
load-bearing structure.
The heritage values have to be recognized – the
investigations are linked in this respect with the artis-
tic elements or architectural details, taken over from
archaeology or art history studies, as well as from the
expert opinion on the artistic elements or the non-
bearing structures. The intrinsic heritage values of the
load-bearing structure are determined by experts in the
field of historic load-bearing structures.
4.2 Interventions needed for the correlation of the
extant level with the contemporary safety
requirements
The interventions needed for the correlation of the
extant level with the contemporary safety require-
ments are also approached from a technical point of
view, as well as with a view to the heritage values.
The technical interventions take place during [9]
the diagnosis phases carried out on the historic load-
bearing structure (characterized by the approach of
the structural problems from the point of view of
building mechanics, physics and biology, as well as
of soil mechanics and building services engineering),
148
focusing their attention on the safety aspects: (i) the
elimination of the causes of insufficiencies in the
load-bearing structure, (ii) ensuring the performance
expectations at the load-bearing structure and (iii) pre-
serving the performance expectations in the case of the
load-bearing structure – ensuring the durability of the
intervention.
From the principles adopted for the safety interven-
tions the following are to be stressed:
– the interventions are in connection with the load-
bearing structure concept: the correct concepts
allow conservation or restoration works, in the
case of the wrong ones retrofitting or possibly
reconstruction is needed; safety is provided in the
case of both permanent and temporary, as well as
extraordinary (special) actions;
– reversible (that can be dismantled) interventions are
preferred, especially when the compatibility of the
intervention with the heritage values is uncertain;
– the efficiency of the interventions has to be
checked; in this respect it is recommended to test
the dynamic characteristics of the load-bearing
structure before and after the interventions: accel-
eration, speed etc.;
– the consequences of the interventions regarding the
improvement of safety at all performance expec-
tations: in connection with building physics and
biology, as well as soil mechanics;
– the reconditioning of the quality of historic materi-
als is also recommended, with a view to regaining
their initial safety level (for example re-freshening
the mortar from the vaulted slab masonries through
its moistening with lime milk [8]).
The interventions carried out on the heritage values
should be minimal. The problem consists in, whether
the intrinsic heritage values of the load-bearing struc-
ture are recognized or not: these values have to be
protected or not, or the load-bearing structure has only
the role to support, to carry the heritage values of the
non-bearing structures (artistic elements and architec-
tural details). If the intrinsic heritage values need to
be protected, then the diagnosis and prediction of the
historic load-bearing structures have to focus on the
identification and protection of the intrinsic heritage
values, going through each phase of the diagnosis and
prediction, mentioning what happens to the intrinsic
and worn heritage values every time, thus: (i) survey
– the heritage values are surveyed; (ii) insufficiencies:
it will be described to what degree the insufficiencies
affect the heritage values; (iii) testing the way in which
the performance expectations are provided: whether
the heritage values have got deficiencies in this respect
or not, (iv) the determination of the causes of insuffi-
ciencies: whether the causes dispose of heritage values
or not; (v) the elimination of the causes of insufficien-
cies: whether they are destroyed through elimination
in the case of heritage values as well, (vi) providing the
performance expectations, (vii) preserving the perfor-
mance expectations – whether the heritage values are
affected through the suggested interventions or not.
4.3 The maintenance of the acceptable behaviour
of load-bearing structures during use
The maintenance of the acceptable behaviour of
load-bearing structures during use requires – among
others – the following:
– the functional and technical (conceptual, mate-
rial and technological) compatibility of the inter-
ventions has to be provided (shear thresholds,
compressions, counter-bending etc.);
– the incompatibilities in connection with building
mechanics, physics, biology and soil mechanics
have to be avoided;
– specifying certain measures for the perpetuation of
the effects of the interventions carried out on the
load-bearing and/ or non-bearing structure (avoid-
ing heat bridges that bring about vapour, ventilation
conditions, anticorrosive treatments etc.);
– the maintenance of the interventions, with a view to
obtaining the response of the load-bearing structure
to extraordinary actions at any time (for example
earthquakes);
– reducing the aggressiveness of the environment.
5 CONCLUSIONS
The safety of the (historic) constructions is defined
by the relation between the qualities (possibilities) of
the (historic) load-bearing structures and the demand
of the environment. Being built on the basis of
an empirical-intuitive concept, of historic materials
and with traditional technologies, the historic load-
bearing structures display their qualities (possibilities)
arranged according to these conditions of existence.
The pretentiousness of the environment is always in
connection with the place, in space and time. The
spatial coordinates mark the natural and artificial cor-
rosive, technical and social conditions in which the
load-bearing structure exists. The time coordinates
present the evolution of these conditions.
5.1 A series of questions have to be asked,
which need to be answered
(a) Beside utility and economical nature, safety plays
an important role in the sustainable management
of the built environment. The question is stated: in
the case of historic buildings, is the preservation of
the heritage value a criterion of equal importance
with that of safety and economical nature?
149
(b) The load-bearing structural concept has to be ade-
quate and has to meet the contemporary safety
requirements. If it does not meet them, what will
happen to the historic building? Will it be demol-
ished? If an intervention is carried out for the
correction of the load-bearing structural concept,
what are the acceptable limits of the intervention?
What is the quantity of heritage values that can be
sacrificed in order to provide safety?
(c) The empirical-intuitive concept can be correct, but
the historic materials – of which the concept is
made – do not comply with the pretensions of the
concept to meet the contemporary safety require-
ments: Do we replace the materials? Do we change
the concept?
(d) When are the safety requirements satisfied through
the conservation of the load-bearing structure?
What about restoration? When is retrofitting or
reconstruction applied? The reconstruction is car-
ried out with the same material, or only the
geometry is re-built by using the materials that
provide the necessary safety (for example concrete
instead of masonry, possibly a concrete framework
embedded in the masonry)?
(e) Temporary and definitive interventions are
accepted in order to ensure the safety of the historic
load-bearing structures. Is the emergency inter-
vention to provide safety an accepted version or
not: in the first phase the emergency intervention
is carried out to provide safety that temporarily
guarantees the safety of the load-bearing struc-
ture, after which the final intervention follows –
after a certain period of time: days, months, years,
decades? Do these interventions have to respond to
all the possible actions, or only to some of them? Is
there a maximum duration, for which a temporary
intervention can efficiently function?
5.2 There is a series of questions
(a) The safety of the historic load-bearing structures
is influenced by the general factors that affect
all the constructions from the built environment
and by factors specific for historic constructions.
Both categories of factors deal with a significant
objective, as well as a subjective part.
(b) The characteristics of the historic load-bearing
structures, as well as of the historic and/ or
current load-bearing structural environment are
considered to be the objective premises.
(c) The social-political conditions, the level of tech-
nical knowledge, the research possibilities, the
quality control system, the available materials,
aptitude and others are considered to be subjective
premises. These premises operate at the formation
of the constructions or during the subsequent inter-
ventions (the «historic» ones), being present in our
time («contemporary» ones) as well.
(d) Local historic materials, technologies developed
in connection with the local materials and the
local conditions relating to the development of the
technologies, the local climatic conditions (load
resulting from snow, wind, the number of the
repeated annual freezing cycles etc.), the inten-
sity of the seismic activity, all these have led to the
local empirical-intuitive concept of historic load-
bearing structures (postulating also a huge her-
itage value). The investigations will compulsorily
identify all these.
(e) The investigations will pay special attention to the
time coordinate (for example to the depreciation
of the quality of the materials due to explosions,
floods, fires, earthquakes taking place over time).
(f) The safety of the historic load-bearing structures
is sufficient, if the foundation conditions are ade-
quate, the corrosive agents controlled, taking the
following into consideration in connection with
the superstructure: (i) the safety of the historic
masonry load-bearing structural sub-units is max-
imal, if the dynamic loads and differential settle-
ment can be avoided; (ii) the safety of the historic
wooden load-bearing structural sub-units is max-
imal, if they are protected against the biological
attacks.
(g) The state of the historic load-bearing structures,
the safety reserves are continuously changing, the
periodic re-evaluation of safety being compulsory.
(h) The reconditioning of the quality of historic mate-
rials is also recommended, with a view to regaining
their initial safety level (for example re-freshening
the mortar from the vaulted slab masonries through
with lime milk).
(i) Reversible (that can be dismantled) interventions
are preferred – for their reconsideration together
with the development of intervention techniques-,
especially when the heritage values are affected.
(j) Issues regarding safety and heritage values are
required to be treated simultaneously. Theoreti-
cally we apply measures that are compatible with
the intrinsic heritage values of the historic load-
bearing structures, wishing to destroy as few of
these values as possible.
(k) Uniform measures are recommended regarding
the safety of the built environment – taking of
course into consideration the characteristics of the
historic load-bearing structures, as it is not effica-
cious, if a group of institutions takes care of the
engineered structures and another group watches
over the historic ones.
(l) Regional (even local) instructions are needed for
the right handling of the issues regarding the safety
of the historic load-bearing structures; the delim-
itation of the regions can be made according to
seismic zones, slope stability, empirical-intuitive
150
concepts (Byzantine-eastern and western building
cultures) and others.
(m) The accidents due to incapacity and frauds
cannot be avoided through technical regulations –
however perfect these might be. The measures nec-
essary for the reduction of the causes of these
accidents do not influence the conditions regarding
the safety of the constructions.
REFERENCES
[1] Ferry Borges J. & Castanheta M. 1971. Structural Safety.
Lisbon: LNEC
[2] ∗ ∗ ∗ 2003. ICOMOS Charter – Principles for the Analy-
sis, conservation and structural restoration of architec-
tural heritage. Victoria Falls, 14th General Assembly of
ICOMOS
[3] Szabó B. Gy. & Kirizsán I. 2003. Built Heritage in Crisis
Situations – Disasters and Historic Load-bearing Struc-
tures. In:Imola Kirizsán, Bálint Gy. Szabó (ed.), Historic
structures and disasters; Proc. 7th Scientific Conference
Cluj-N., 6–8 November 2003, Cluj-N.: Utilitas
[4] Szabó B. Gy. 2004. Illustrated dictionary of historic
load-bearing structures. Cluj-N.: Kriterion & Utilitas.
ISBN 973-9377-40-8
[5] Szabó B. Gy. & Kirizsán I. 2004. Conservation of His-
toric Load-bearing Structures and the Protection of their
Intrinsic Heritage Values – Legacy of Historic Load-
bearing Structures. In: Imola Kirizsán, Bálint Gy. Szabó
(ed.), Heritage values of historic structures; Proc. 8th
Scientific Conference Cluj-N., 28–30 October 2004,
Cluj-N.: Utilitas. ISBN 973-9377-38-6
[6] Szabó B. Gy. & Kirizsán I. 2006. The Safety of the His-
toric Load-Bearing Structures In: Imola Kirizsán, Bálint
Gy. Szabó (ed.), Safety of Historic Load-Bearing Struc-
tures; Proc. 10th Scientific Conference Cluj-N., 26–28
October 2006, Cluj-N.: Utilitas. ISBN 973-9377-45-9
[7] Roca P. 2006. The Study and Restoration of Histori-
cal Structures. In: P. Lourenco, P. Roca, C. Modena,
S. Agrawal (ed) Structural Analysis of Historical Con-
structions; Proceedings of the 5th International Con-
ference, New Delhi 6–8 November 2006. New Delhi,
Macmillan India Ltd. ISBN 978-1403-93155-9
[8] Szabó B. Gy., Kirizsán I., Kirizsán Il. & Vass L. 2006.
The Improvement of the Safety of the Vaulted Slabs
with Lime Treatment In: Imola Kirizsán, Bálint Gy.
Szabó (ed.), Safety of Historic Load-Bearing Struc-
tures; Proc. 10th Scientific Conference Cluj-N., 26–28
October 2006, Cluj-N.: Utilitas. ISBN 973-9377-45-9
[9] Szabó B. Gy. 2007. Metoda complexa de investigare
a starii de degradare la structuri portante istorice.
In: Raport tehnic UTILITAS_07009, Cluj-N.: Utilitas.
ISSN: 1842–9866
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Concrete: Too young for conservation?

H.A. Heinemann
Chair of Building Conservation, Faculty of Architecture, Technical University of Delft, The Netherlands

R.P.J. van Hees

Chair of Building Conservation, Faculty of Architecture, Technical University of Delft, The Netherlands
TNO Built Environment & Geosciences, Delft, The Netherlands

T.G. Nijland
TNO Built Environment & Geosciences, Delft, The Netherlands

ABSTRACT: The 20th century built heritage is one of the new conservation challenges, due to its architectural
differences from the traditional heritage and new materials. One major new material is concrete; its quantity and
importance for the new heritage requires a tailored conservation approach. Until now, there is a dependence on a
repair approach. The dangers of such an approach for the cultural-historical values of the original concrete will
be shown by the means of a case study, Fort Bezuiden Spaarndam (1901), part of the UNESCO World Heritage
Site Defence Line of Amsterdam. An underlying problem is that concrete is too young for conservation experts,
causing a dependence on concrete repair expertises. The used repair techniques are developed for structural aims,
not integrating conservation aims. Fundamental research on the history of concrete and suitable conservation
approaches and techniques is therefore needed.

1 INTRODUCTION such as incompatible repair materials or a more funda-

With the aging of the 20th century building stock,
we are now facing the challenge to evaluate them
and make selections for objects which should be kept
as part of our built cultural heritage. It is consistent
that the values which justified the selection should
be preserved and respected during any intervention.
Interventions cannot be avoided as aging often goes
along with some degree of deterioration. The demands
for interventions according to (international) conser-
vation guidelines are applicable for the entire built
heritage, independent from age or construction type.
However, for the implementation of the demands
a tailored conservation approach for the 20th century
built heritage is needed.The new heritage does not only
differ in architecture, structure, quantity, and building
typology from the traditional heritage, but also in the
used materials such as concrete. The different material
properties make a transfer of knowledge and experi-
ence gained with traditional materials such as brick
and mortar less feasible.
For concrete, the current approach is often a con-
ventional repair approach as no tailored conservation
approach exists. As in practice a loss of values cor-
related to the original concrete is often encountered,
the question arises if this is a pure technical problem
mental problem such as the way concrete conservation
projects are approached.
During this research, studies of completed interven-
tions of historic concrete buildings indicated a relation
between the following of repair approaches and the
loss of monumental values. In this paper, this rela-
tion will be illustrated by means of the case study Fort
Bezuiden Spaarndam, the Netherlands. It shows that
there are general, underlying problems in the field of
concrete conservation which endanger values, inde-
pendent from the technical performance of a repair.
2 METHODOLOGY
The current research aims at understanding the rela-
tion between the loss of value and following a repair
approach. To evaluate the impact of a repair approach,
case studies of former concrete conservation projects
in the Netherlands have been carried out. In this paper
the results of the case study Fort Bezuiden Spaarndam,
part of the Defence Line of Amsterdam (Stelling van
Amsterdam), will be presented to explain the underly-
ing problems of the preservation of values of historic
concrete.

151
 During the research, the most important involved
parties were interviewed (architect, surveyor, and con-
crete repair company) to understand their background,
knowledge, and experience and to see where the
dependence on concrete repair experts starts. Liter-
ature, documents, and surveys related to the fort, its
construction and conservation were consulted to inves-
tigate the historical background, conservation aims,
and conservation process. During a visual inspection
in November 2006, the performance and impact of the
conservation was evaluated.
The study focussed on concrete which was used
for exterior elements, stressing the complexity to bal-
ance technical, aesthetical and conservation demands,
meaning a minimal loss of existing fabric and a maxi-
mal preservation of values. Additionally, petrographic
analyses of samples were made to determine the con-
crete composition and possible damage. To evaluate
the suitability of standard concrete surveys for con-
servation tasks, literature and guidelines related to
concrete assessment were studied.
3 BACKGROUNDS OF REPAIR AND
CONSERVATION
To understand why following a repair approach in a
conservational context can endanger values, it is nec-
essary to understand why and how the main aims
of repair and conservation diverge radically. Repair
methods are usually developed to preserve a structural
function for a limited lifetime, lately also influenced by
sustainability aspects. Surveys focus on physical and
mechanical properties and the criteria for the choice
of a repair approach are, beside the structural perfor-
mance, the service life, economical issues, and the
possibilities of maintenance (ENV 1504-09 1997).
Conservation, however, is much more than ful-
filling technical and financial requirements. Besides
considering structural performance and safety, the
preservation of the authenticity of and values attached
to the site, building and material must be considered.
The value and authenticity of the cultural heritage
is understood in ‘artistic, historic, social, and sci-
entific dimensions’ and the conservation principles
‘respect. . . the existing fabric’ and the cultural heri-
tage should be preserved for ‘present and future
generations’ (ICOMOS 1999).
Yet, due to the youth of the field of concrete con-
servation, concrete repair experts are entrusted with
concrete conservation projects.Although these experts
have considerable knowledge of the technical aspects,
they are not trained in specific conservation aspects,
as the complexity and variety of monumental values
and authenticity or the properties of historic con-
crete. Because conservation experts often do not have
sufficient knowledge on the material, an evaluation of
the suitability of such a process is difficult.
152
Figure 1. Site plan of Fort Bezuiden Spaarndam.
4 FORT BEZUIDEN SPAARNDAM
4.1 Historical background of the Defence Line of
Amsterdam
The Defence Line of Amsterdam, the Netherlands
(1880–1920), is a ring of fortifications (45 forts)
around Amsterdam, intended to defend the city by sys-
tematic inundation in case of an attack. This ingenious
defence technique was one justification for the listing
as a UNESCO World Heritage Site in 1996. In con-
nection with the listing, attention was drawn to the
need for conservation and re-use, as the condition of
many forts declined after the military function of the
forts ceased in the 1960s. Fort Bezuiden Spaarndam
(1901), a provincial monument in the province North
Holland, was constructed as part of the Defence Line
of Amsterdam.
4.2 Historical background of Fort Bezuiden
Spaarndam
Fort Bezuiden Spaarndam is a non-reinforced concrete
fort built in 1901. It was planned as a standard model
of the Defence Line of Amsterdam on a fort island with
two cupolas with disappearing turrets (Fig. 1). The fort
itself did not suffer from major damages during the
Second World War, though the cupolas were blown up
by the German occupation army. The remains of the
cupolas still lay spread around the grounds. Since the
cease of the military function in 1960, the fort was
used for agricultural storage.
4.3 Historical background of the concrete of
Fort Bezuiden Spaarndam
To recognise the historical value of the original con-
crete, it is necessary to understand the construction and
technology history of the late 19th century. Knowl-
edge of concrete and its construction was mainly
based on empiric studies; standard design codes only
appeared at the beginning of the 20th century. In
this context, the knowledge gained from the con-
struction of the Defence Line of Amsterdam was
important for the Netherlands, where early concrete
technology was dominated by the army corps of
engineers (e.g. Scharroo 1910). The first Dutch rein-
forced concrete guidelines Eerste Gewapend-Beton-
Voorschriften from 1912 were influenced by the expe-
rience gained during the construction of the Defence
Line of Amsterdam (Noord-Holland 1996).
Therefore, the forts of the Defence Line of Ams-
terdam are the tangible artefacts of this pioneer phase
of concrete technology. Initially, the forts were built
as traditional brick constructions; yet the invention
of brisance grenades demanded new and stronger
construction methods. Hence from 1897 onwards,
the forts were built as non-reinforced concrete con-
structions, later as reinforced concrete constructions
(ICOMOS 1996). This transition from brick to rein-
forced concrete and the still preserved high degree
of authenticity was part of the justification for the
Netherlands to nominate the Defence Line of Ams-
terdam as a UNESCO World Heritage Site (ICOMOS
1996). The value of the concrete as a preservable
cultural-historical artefact was stated prior in the
description of the Fort Bezuiden Spaarndam in con-
nection with the listing as a provincial monument in
the early 1990s (Provincie Noord-Holland 1998).
5 CONSERVATION 1996–1999
Fort Bezuiden Spaarndam was one of the first forts
to be restored after the listing of the Defence Line
of Amsterdam as a World Heritage Site in 1996. It
was supposed to be a role-model conservation project
with extra subsidies and attention, as further forts
needed restoration in the near future. The first aim
was to find a new locally orientated function for the
fort, which should be compatible with the surrounding
nature recreation area. In 1996, it was decided to use
the northern wing as a day nursery, the postern as an
art gallery, and the southern wing as storage.
The task of conservation was handed over to a
restoration architect, who had experience with the
conservation of a brick fort, but no experience with
the material concrete. This is not unusual for the
late 1990s, as little knowledge and experience with
the conservation of concrete existed. The heritage
care authorities of the province North Holland had no
specialist for the material concrete either, and an exter-
nal concrete expert was entrusted with the survey of
the fort.
5.1 The survey of the concrete
The survey did not differ from standard, non-
conservation concrete surveys. The assignment was to
153
assess the technical state of the concrete and to deter-
mine the causes of the encountered leaching, cracking,
and leakages. Several cores were taken, not only to
determine the composition and damage causes, but
also to gain information on the wall construction.
It was stated that the walls were massive concrete
and the concrete was composed of cement with ground
granulated blast furnace slag (GGBS) and crushed
rocks (granite, porphyry) as aggregates. The following
damages were assessed in the survey:
– shrinkage cracks
– cracks along casting segments caused by missing
expansion joints
– water bearing cracks with leaching
– map cracking caused by the formation of thaumasite
or ettringite.
The advice was to reconstruct the missing earth
shelter of the roof and to seal the roof to prevent
further infiltrations. Cracks that influenced the struc-
tural performance should be injected; the injection of
other cracks was optional. In general, the condition
of the concrete, considering the construction time and
method, was stated as good and with adequate repairs,
the building could last another 100 years.
5.2 Repair of the concrete shell
The main goal of the intervention was to repair the
outer shell of the fort in a way that would solve the
problems ‘for once and for all’, repairing the roof
and treating all cracks. As a detailed advice for repair
techniques was not commissioned, the choice of an
adequate repair technique was left to the executing
concrete repair company. In 1997, the repair campaign
of the fort started and included for the exterior concrete
walls following intervention steps:
– high water pressure cleaning of the façade
– filling of cracks with standard repair mortars
– injection of cracks with micro-cement
– creating expansion joints along the cracked vertical
segment joints.
No specific requirements concerning durability and
compatibility were prescribed. The concrete repair
works were not supervised by external experts. Main-
tenance and regular inspections were not planned,
since it was assumed that the repairs would be long
lasting.
6 EVALUATION OF THE CONSERVATION
PROCESS
Although intended to be a role-model conservation
project, the approach of Fort Bezuiden Spaarndam
is an example of a concrete repair approach. This
case shows the problems resulting from an unsuit-
able approach and reflects an underlying problem
of concrete conservation: unawareness of the values
of historic concrete. Examples from the survey and
execution illustrate the consequences and show how
valuable information could have been gained or values
preserved.
First, the crucial phase of assessment and decision-
making was not guided by adequate experts who
understood both the material, including its history,
and conservation principles. Consequently, values
attached to the concrete were neither determined nor
the impact of the repair techniques on the values
attached to the concrete evaluated.
However, it is possible to gain information on
possible values during a survey (Heinemann 2007).
A review of the conservation process show how values
could have been detected and thus incorporated in the
conservation process; it also reinforces that knowl-
edge on the material, its history, conservation aims and
repair techniques is needed, and that trained experts for
these tasks are still missing.
6.1 Present condition 2006
A visual inspection of the performance and impact
of the interventions was made during a site visit of
Fort Bezuiden Spaarndam in 2006. From far distance,
the impact of the repaired cracks was already visible,
changing the appearance from a monolithic structure
to a patched structure (Fig. 2). On closer examination,
it appeared that the repaired cracks next to the newly
created expansion joints re-cracked and the concrete
of the casemates showed map-cracking with leaching
and local disintegration. On several places on the main
façade, traces of camouflage paint were found that
were not documented in former surveys.The interior of
the fort showed no signs of infiltrations or re-cracking,
and according to the tenants, who appreciated the
Figure 2. Gorge façade after repair.
154
interior of the fort (day nursery and art gallery), the
climatic conditions were satisfactory.
6.2 Awareness of values
The visual impact of the conservation on the façade
raised the question why the aesthetical values of the
monument were not respected. A possible aesthetical
value of a building is usually widely accepted and the
visual impact of a repair technique can be evaluated
without large-scale research. However, standard repair
techniques do not consider aesthetical compatibility.
Therefore, the question arose how the conservation
of concrete was realised and if values were consid-
ered in the process. During our investigation, it became
apparent that the concrete was only perceived as the
structural material of the shell. This reflects the under-
lying problem of concrete conservation, that concrete
is often not seen as a valuable historical material.
In the case of Fort Bezuiden Spaarndam, it was offi-
cially stated that the original concrete was valuable in
the justification for the listing as a UNESCO World
Heritage Site (ICOMOS 1996) and as a provincial
monument in 1992 (Noord-Holland 1998). The origi-
nal concrete is important because of its uniqueness and
remains of this phase of the technical development of
concrete in Europe are rare. However, an implemen-
tation of the preservation of and awareness for these
values was missing in the conservation process.
This unawareness led to the commission of a stan-
dard (i.e. purely technical) survey of the concrete
which was not adapted to the needs of conservation.
Without sufficient knowledge of historic concrete and
possible values, the weaknesses of a purely technical
survey in this context were not seen. Consequently, the
correlation between values, the material, and the state
of conservation was not determined. Therefore, the
values were not further integrated in the conservation
process.
6.3 Survey of the surface coarse aggregate was crushed porphyry from Quenast,
Belgium, and the fine aggregate river sand (Fig. 3).
Aesthetics are criteria for the value assessment of
From the study of historical documents, we found
buildings and for the compatibility of conservation
out that crushed rock was assumed to be superior to
techniques. Therefore, it is important to describe the
gravel because of its rough and angular surface which
surface of a monument during a survey. For techni-
was supposed to improve the bond to the cement paste.
cal surveys, as it was commissioned at Fort Bezuiden
The use of crushed rock and in particular porphyry of
Spaarndam, a description of the surface is not included
Quenast as an aggregate was advised in contempo-
as it usually does not influence the mechanical or
rary books (Kloes 1908). In the construction plans of
physical properties of the concrete.
the fort from 1900, it can be seen that a differentia-
A description of the properties of the surface such
tion between the used aggregates and function of the
as colour, texture, deposits, weathering, or traces of
concrete element was made. For important elements,
former events can be used to evaluate the impact of an
the use of porphyry was designed, while secondary
intervention (e.g. material loss due to cleaning) or help
elements where designed with cheaper gravel or even
to adapt the structure and colour of a repair mortar to
with rubble.
the surrounding concrete.
The high amount of voids and the average to high
Other essential information which should have
capillary porosity of the concrete mentioned in the
been documented is the presence of camouflage paint
survey can be explained with knowledge of the com-
traces and the original surface finish. Further colour-
position and construction method. They are in this
historical research should have been carried out to
case typical, because the rough shape of the aggre-
evaluate the options for reconstruction of the paint and
gate complicated the manual compaction. To obtain
to broaden the historical knowledge of the forts.As this
the workability, the use of a higher water-cement ratio
case was supposed to be a role-model, opportunities
was common. Both the compaction and the high water-
to study the forts in connection with the survey should
cement ratio influence the porosity and durability of
have been taken.
concrete.
How far the cleaning of the façade with high-water
Another result from the composition is the occur-
pressure caused losses of the original surface, and
rence of alkali-silica reaction (ASR) (Fig. 4). The used
with it historical information, cannot be determined
aggregate, porphyry of Quenast, is now known to
now, since there is no detailed documentation of the
be alkali-silica reactive (Nijland et al. 2003). When
façade to refer to. Given experience with façade clean-
used in combination with Ordinary Portland Cement,
ing, however, such a loss is highly likely. It also shows
which, in contrast to the GGBS cement that was
how easily standard approaches are followed, and con-
wrongly considered to be present in the original survey,
sequences for the monument and its values are not
ASR is not inhibited. With knowledge of the construc-
considered.
tion history of concrete certain damage types can be
explained, here the assumed superior combination of
6.4 Survey of the composition and construction porphyry and Ordinary Portland Cement causingASR.
An analysis of the composition, structure, and cause
of damage of concrete, as for example by petrographic
examination, is not standard, as it might not always be
necessary to diagnose the cause of damage. However,
if samples are taken, they can be used to gain insight
into the composition and construction of historic con-
crete. Besides giving relevant technical information,
it can reveal important historical information, and
therefore indicate value.
Historic concrete differs from modern concrete. In
the case of Fort Spaarndam Bezuiden, it is known that
it was constructed in a time when concrete construc-
tions were not standardised. Also the statement that the
concrete is a rare remainder of concrete development
should have led to an investigation of the samples from
a historical point of view as well.
Our petrographic analysis of samples of the south-
ern cupola and of the main building show that the Figure 3. Microphotograph showing detail of the cement
concrete is composed of a coarse-grained Portland paste, with relics of unhydrated Portland clinker (indicated
cement and not ground granulated blast furnace slag as calcium silicates C2 S and C3 S), and no blast furnace slag
(GGBS) cement as described in the survey. The at all (view 0.7 × 0.45 mm, plane polarized light).

155

Figure 4. ASR-gel (A) in voids (V) and porphyry (P)
and sand (S) as aggregates in the cement paste (C) (view
1.8 × 1.12 mm, plane polarized light).
Figure 5. Microphotograph of the section through the
surface, showing three lime-based layers (plane polarized
light).
It has to be kept in mind that the perception of
cement types differed from our current perception. At
the time, the use of Ordinary Portland Cement was pre-
ferred to the use of blast furnace slag cement. Latter
was assumed to be of poor quality, mainly because it
was a cheap waste product. It became, due to economic
reasons, only widely used in the Netherlands from the
1920s onwards.
A petrographic analysis of the concrete can also
tell stories about the history of a building. A section
through the surface (Fig. 5) shows three layers of a
lime based layers on the surface. The core was taken
before the surface was cleaned with high-water pres-
sure. Therefore it cannot be determined if they are
remains of a finish (Noord-Holland 1996) or deposits.
If the indications from the thin section would have been
used prior to the intervention, research could have been
carried out in situ and decisions made if the surface is
156
preservable.Yet it is likely that the high pressure clean-
ing of the façade affected this layer of the surface. As
no references were found on the building anymore nor
could be found in the documentation of the project,
the amount of loss of original fabric cannot be said.
With historical and technical knowledge, a petro-
graphic analysis can help to determine the authenticity
of the original concrete, which is important when deci-
sions have to be made of how much material loss is
acceptable. The results from our petrographic analy-
sis show that the composition of the concrete of Fort
Bezuiden Spaarndam is of high quality relative to the
state of the art of 1900. The given historical informa-
tion confirms the value of the authentic concrete and
any intervention should respect that value.
6.5 Advices and decision-making
As the assignment of the survey was pure technical,
additional historical aspects of the concrete were not
considered. As a result, no special treatment respecting
the historical values (significance of original concrete
and the visual appearance) was advised. The choice of
the repair techniques was done by the restoration archi-
tect and the concrete repair company. Neither of them
was familiar with the properties of historic concrete
and the complexity and consequences of repair choices
were not seen. One example of the consequences is
the repair approach of the cracks caused by thermal
movements.
A common problem of the forts of the Defence Line
of Amsterdam is the lack of expansion joints. Expan-
sion joints were not common at that time, in the 1920s
there was still discussion if expansion joints are needed
(Kleinlogel 1927). Also here the missing insight into
the properties of concrete caused damage, cracking,
which becomes a historical evidence of the trials and
errors of the technological development.
The concrete forts of the Defence Line of Amster-
dam are approximately 150 m long, most of them not
reinforced. The stresses caused by thermal movement
caused regular linear cracks along the weakest points,
usually the casting segments. Cases of severe struc-
tural damage are not known, as each casting segment
is equivalent to one vault and the cracks appeared at
the points of support in the middle of the cross walls.
Yet, the ingress of aggressive agents can be facil-
itated by the cracks. As there was no reinforcement,
corrosion was not possible but the composition of the
concrete was ASR sensitive, a reaction which needs
water and can lead to severe damage. However, no
signs for increased deterioration were visible next to
the cracks.
During the restoration campaign of Fort Bezuiden
Spaarndam is was decided to repair the outer shell
‘properly’, which included a repair of the cracks
caused by thermal stresses. The idea was to compen-
sate the thermal stresses with expansion joints, which

Figure 6. New expansion joint (arrow) next to patched
cracks (CR) caused by thermal expansion.
would be intersected. The façade was cleaned with
high water pressure and the cracks injected. The injec-
tion was originally carried out with a polyurethane
injection resin. The high porosity, the numerous voids
along the segment joints and the thickness of the wall
(1.25 m) complicated the work. As a complete filling
of the cracks was unlikely, the cracks were injected
with non-elastic micro cement instead. As the thermal
stresses would be compensated by the new joints, the
elasticity of the mortar was not considered important.
After patching and injecting, the exterior façade was
vertically intersected (0.05 cm–0.10 cm depth) along
the casting segments with the idea to create expansion
joints which would prevent further cracking of the con-
crete. The dimensions (width, depth) of the cut were
not based on calculations but on the working experi-
ence of the contractor with non-reinforced industrial
floors. A possible effectiveness of the intervention was
not studied prior to application, neither tryouts on a
limited scale were done nor by simulations.
As the concrete is not reinforced, the structure is
sensitive to stresses; instead of releasing the struc-
ture from stresses, stresses concentrated at the end
of the intersection. The reappearance of cracks in the
patched area within less than a decade indicates that
the intervention does not work (Figs 6–7).
This intervention required an unnecessary material
loss and as it failed, a ‘repair of the repair’ may follow,
causing further material loss. A simpler approach by
157
Figure 7. Close up of an expansion joint with a reappearance
of the former cracks; the colour of the used patching mortar
was not adapted.
not filling the cracks and preventing water ingress by
fixing the roof, as it was done anyway, should have
been tried instead. Together with a long-time moni-
toring of the moisture ingress through the cracks and
of the crack movement, a major intervention could
have been prevented or delayed and insight would have
been gained on the damage mechanisms, allowing the
choice of an appropriate conservation approach.
7 CONCLUSIONS
The primary problem of the concrete conservation
of Fort Bezuiden Spaarndam is not technical such
as the performance of a repair technique. The prob-
lem lies in an approach, which does not respect the
concrete as a historical valuable material: a repair
approach instead of a conservation approach. Yet there
is another underlying problem in the field of conser-
vation: concrete seems to be simply too young for
heritage stakeholders.
7.1 Conservation Fort Bezuiden Spaarndam
Monumental values were lost during the conservation
of Fort Bezuiden Spaarndam, because the concrete was
primarily understood as a structural and not a historical
material. Together with a dependence on repair experts
which are trained to perceive concrete from a structural
point of view, little attention was given to the values
correlated to the original concrete. The conservation
experts could not evaluate if the approach was suitable
for historic concrete as they had no experience with the
material.
 Similar to a repair process, the survey focused
on physical properties of the concrete and did not
integrate historical relevant aspects as the unique
composition of the concrete or the traces of camou-
flage paint. Consequently, the amount of lost original
concrete and related values was not considered.
Support and guidance for the involved parties
by concrete conservation experts was not possible,
as the field of concrete conservation barely existed
and the complexity of concrete conservation was not
understood. Therefore, possible conservation aims, the
cultural-historical values, the significance of the state
of the building and their relationship could not be
discussed openly before starting the conservation cam-
paign. Such discussion should not be limited to pos-
sible re-use or economical issues, but should include
tangible and intangible values of the building/material
as well.
A critical review of the repair approaches was not
taken into account, as it was assumed that repair
experts have sufficient knowledge of the material.
There was no awareness for the fundamental differ-
ences between a repair and conservation, therefore the
influence of a repair approach on the values was not
considered.
However, this case is not isolated and it is likely that
the results would have been similar if the conservation
would have been carried out by other people. The fun-
damental problem of concrete conservation is that the
conservation process is often not guided by conserva-
tion experts which understand both the material and
conservation principles.
7.2 Concrete conservation
Values and their preservation are key issues of heritage
care. The value of age, that, as Riegl (Riegl 1903) says,
even ‘appeals the masses’ is often not seen since to
the ‘youth’ of concrete buildings and the more sophis-
ticated historical values need art-historical knowledge
to be recognised. Until now, this art-historical knowl-
edge hardly exists and this reflects itself in the (value)
assessment of historic concrete buildings. For the
majority of restoration stakeholders, concrete build-
ings are too young, as their working field is mainly
buildings prior to 1900.
It is necessary to overcome the knowledge gap con-
cerning 20th century buildings in the working field
and education of conservation experts. Parallel, it is
necessary to develop a tailored conservation approach
to overcome the dependence on a pure repair approach.
When applying repair techniques, one should not rely
on the fact that concrete repair has been carried out
for decades, but assessed whether they are applicable,
since they were developed for different aims.
As long as there is a dependence on a repair
approach, the background of repair experts has to be
158
kept in mind. It is the task of the conservation expert
to explain the diverging requirements for conserva-
tion towards repair experts. This requires sufficient
knowledge of the material concrete from the conser-
vation expert; otherwise the deficiency of a repair
approach cannot be seen and may bring to risk the
values which were the initial reason for listing and
conservation.
As long as concrete conservation projects are still
isolated cases and no clear guidelines or references for
concrete conservation exist, general guidelines such
as the Venice Charter (ICOMOS 1964) should be
consulted to evaluate possible approaches. Finally, dis-
cussions of possible values of concrete are needed to
see that it can be more than a greyish bearing mate-
rial. As concrete was initially mainly understood as a
structural material and only later the architectural val-
ues were discovered, we have to make now a similar
evolution: from structural repairs to conservation.
ACKNOWLEDGEMENTS
Figures 1 (based on the construction drawings from
1900), 2, 6, and 7 by H. A. Heinemann, figures 3, 4
and 5 by T.G. Nijland.
REFERENCES
Construction plans Fort Bezuiden Spaarndam. 1900. Nation-
aal Archief, The Hague, OPV Oorlog Plans van Vestingen,
17e–20e eeuw, access number 4.OPV, inventory number
A268–A26.
Heinemann, H.A. 2007. Bedreigde waarden: De invloed van
de technische benadering op de cultuurhistorische waar-
den van betonnen monumenten.Praktijkreeks Cultureel
Erfgoed. 1(2):1–24.
ICOMOS. 1964. International charter for the conservation
and restoration of monuments and sites. The Venice Char-
ter. http://www.international. icomos.org/e_venice.htm.
Accessed 08-05-2007.
ICOMOS. 1996. Advisory Body Evaluation no. 759
http://whc.unesco.org/archive/advisory_body_evaluation/
759.pdf. Accessed 19-11-2007.
ICOMOS. 1999. The Burra Charter. The Australia ICO-
MOS charter for the conservation of places of cultural
significance. http://www.icomos.org/australia/. Accessed
26-11-2007.
Kloes, J. A. van der. 1908. Onze Bouwmaterialen – Mortels
en Beton. Maassluis, the Netherlands: J. van der Endt &
Zoon.
Kleinlogel, A. 1927. Bewegungsfugen im Beton- und Eisen-
betonbau. Berlin: Wilhelm Ernst & Sohn.
Nederlands Normalisatie Instituut 2006. ENV 1504-09:1997
Products and systems for the protection and repair of
concrete structures – Definitions, requirements, quality
control and evaluation of conformity – Part 7: Reinforce-
ment corrosion protection. Delft: Nederlands Normal-
isatie Instituut.
Nijland, T.G., Larbi, J.A. & Siemes, A.J.M. 2003. Experi-
ence of ASR in the Netherlands. Proceedings European
marine sand and gravel – shaping the future, EMSAGG
Conference. Delft, the Netherlands.
Provincie Noord-Holland. 1996. Onderzoek naar de
bouwkundige staat van de forten van de stelling van
Amsterdam. Haarlem, the Netherlands: Provincie Noord-
Holland.
Provincie Noord-Holland. 1998. Stelling van Amster-
dam. Positie bij Spaarndam.(2nd ed.) Haarlem, the
Netherlands: Provincie Noord-Holland.
159
Riegl, A. 1903. The Modern Cult of Monuments: Its essence
and its development. Published in The Getty Conserva-
tion Institute. 1996. Historical and Philosophical Issues in
the Conservation of Cultural Heritage. Los Angeles: The
Getty Conservation Institute (originally published 1903
as Der moderne Denkmalstatus: Sein Wesen und seine
Entstehung Vienna: W. Braunmüller).
Scharroo, P.W. 1910. Inleiding tot de studie van het gewapend
beton. Amsterdam: van Mantgem & de Does.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Significance of a Mughal Mall and its role in achieving safety

Saptarshi Sanyal
School of Planning and Architecture, New Delhi, India

ABSTRACT: The fascinating and unique covered market, Chatta Chowk Bazaar, was designed for the fortress-
palace, Redfort, the magnum opus of 17th century Mughal Emperor Shahjahan’s new capital in Shahjahanabad,
Delhi. This building was the subject of documentation and comprehensive qualitative studies, undertaken by
architectural conservation students of School of Planning and Architecture in 2006–07. The direction and
approach given by the conservation studio’s faculty, one of knowledge systems, was significant in guiding
the studies selected and the way the structure was analysed, resulting in an appreciation of the many dimensions
of the historic building. These revealed architect Shahjahan’s immense knowledge of architecture, qualified
through primary studies of Mughal architecture in Delhi including the palace-complex and secondary studies
of his other projects. This paper attempts to illustrate, through this case, how superior structural know-how
and stability, are integral to a significant historic building, which is a knowledge system, and how an essential
consequence of such significance, is safety.

1 INTRODUCTION This paper firstly addresses the structural sig-

In Aug-Dec 2006, the Conservation Studio team of
School of Planning and Architecture in an exer-
cise “The Past, Present and Future of the Mall that
Shahjahan built”, documented and analysed the Chatta
Chowk Bazaar, an imperial marketplace at Redfort in
Shahjahanabad, the erstwhile Mughal capital at Delhi.
This building was built by the fifth Mughal Emperor,
Shahjahan, remembered globally because of his con-
ception and architectural accomplishment of the world
famous Taj Mahal at Agra.
The objective of this project was to formulate a
conservation proposal for the building addressing the
aspects of reuse and historic fabric. Guided by Prof.
Nalini Thakur and Ar. Sanjay Bharadwaj, the Mughal
architectural context in Delhi was studied, including
the role of Shahjahan as an architect-planner in his
works, which revealed that while his contribution to
architecture was quite significant, this Bazaar is typo-
logically unique in this part of the world. As a part of
property inscribed in the recent inclusion of Redfort
in World Heritage, the Bazaar is also the significant
“living” heritage in the Redfort.
In this exercise, the architectural Knowledge Sys-
tems approach applied to this case helped the appraisal
of its embedded values as being very relevant in con-
servation for future use. While present usage is in
conflict with these values, causing degradation of its
technological and structural memory, the building in
question still survives “safely”.
nificance of the building, as revealed through the
primary and secondary studies of the building and
its context. Secondly, it suggests how the aspect
of structural significance compliments and informs
the knowledge required to achieve safety in his-
toric constructions, as revealed in this case, in the
context of the observed situation and its current
issues.
2 THE KNOWLEDGE SYSTEMS APPROACH
APPLIED
2.1 Contextual studies: Mughal Delhi, Shahjahan,
Shahjahanabad and Redfort
The generation of knowledge started by looking at the
historic city itself, here Mughal Delhi and its archi-
tecture, as it existed on the ground. The secondary
studies on the cultural upbringing and achievements
of the architect of the Bazaar, Shahjahan, corrob-
orated inferences about his contribution to Mughal
architecture, and the significance of his buildings,
most of which could be appreciated as outstand-
ing knowledge systems. Moreover, it was found that
Shahjahan’s knowledge was not restricted to individual
buildings but encompassed palace and city planning,
as demonstrated by the Redfort and his capital of
Shahjahanabad, with the Chatta Chowk Bazaar as an
integral part of both (Figure 1).

161
Figure 1. The Chatta Chowk Bazaar in Shahjahanabad, its
articulation of the processional imperial axis from the city to
the palace.
Figure 2. Original plan and section of the Bazaar.
2.2 Primary studies of the building
The documentation of the Bazaar was through archi-
tectural drawings of the existing state. With the
advantage of having a modular system, the ‘original’
designed structure was deciphered by removing the
accretions and alterations from the measured drawings
(Figure 2).
These preliminary exercises helped in identifying
the structural form of the building holistically. The
term “Mall” thus addressed the fundamental charac-
teristics of the building, functionally, architecturally
and formally. This building can be described as a two-
storied linear arcaded structure with an octagonal court
halfway through its length, with eight arcade bays on
either side, contiguous with the western entry of the
Redfort, the Lahore Gate (Figure 3).
Secondary studies indicated that, though inher-
ited typologically from central-Asian counterparts
of the “souk” or Bazaar, integral to Persian and
Arab cities, the Chatta Chowk Bazaar responded to
the specificities of climate, cultural context as well
as to Shahjahan’s own design of the Redfort and
Shahjahanabad, all of which is a complex design
synthesis.
162
Figure 3. Chatta Chowk Bazaar, 1911–12, octagon and
arcade; photo: Archaeological Survey of India (ASI) Annual
Report.
2.3 Structural analyses and technical studies
From the structural form of the building that
emerged, critical areas were zoned and their behaviour
analysed. Unfortunately, owing to logistical lim-
itations, the foundation could not be examined
and the formal analyses are restricted to the
superstructure alone.
Technical studies were carried out from the reading
of exposed areas of the building fabric, and samples
were qualitatively studied to understand the construc-
tion system in terms of the constituents using sec-
ondary references from similar constructions. More-
over, a similar building by Shahjahan, a hunting-lodge
at Jaunti village near Delhi, was studied insitu to under-
stand construction. This building’s collapse due to
unknown reasons exposed a lot of the brick masonry
that is similar to the Chatta Chowk Bazaar and this
helped in building the knowledge base.
2.4 Chatta Chowk Bazaar as a knowledge system
Through studies of the context and the building
itself, information related to the Chatta Chowk Bazaar
revealed many facets of this structure. Primary and
secondary studies of the architecture and its history,
as well as its technology presented disaggregated
information. Amalgamating this information with the
specific purpose of conservation, helped to appreciate
this building as a knowledge system, from contex-
tual to component levels. This knowledge system has
subsystems of architectural design, structure, con-
struction etc. defining its values. For the purpose of
this paper, although the formal structural knowledge
is highlighted qualitatively, the values revealed by the
analyses shall be critical in guiding still more detailed,
quantitative structural analyses.
3 STRUCTURAL VALUES IN THE BUILDING

3.1 Components of the structure

Before proceeding to the formal aspects of the struc-
ture and their analyses, it is necessary to briefly look at
the types of constituent structural components or struc-
tural elements in this building, along with their basic
materials. These components and their behaviour in
the whole structural system allow us to divide them
into three logical groups.
The first of these, vertical elements, have either
resultant or primary load lines or both which are ver-
tical. These elements are subject to buckling under
stress. They are morphologically in the vertical plane
and include walls and piers. In Chatta Chowk Bazaar,
these are built in brick with mortar. Figure 4. The roof of the Mall; the walls have filled up the
Secondly, horizontal elements carry uniformly dis- balconies over the parapets today as seen in this image.
tributed loads or point loads and are morphologically
horizontal or in the horizontal plane. They are sub-
ject to bending when loaded. As spanning in Chatta
Chowk Bazaar uses predominantly arcuate systems,
the presence of these is rare. These elements are lin-
tels in most small openings and mezzanine floors in
the western gate. They are built in red sandstone.
The most important and commonly used, compos-
ite elements dissipate horizontal, vertical and lateral
loads in one or more planes. These elements are mor-
phologically more complex and the loads in them are
indeterminate due to three-dimensional geometries.
However, it was possible to qualitatively map the load
transfer lines in them. These are also built in brick and
mortar and include large arch-ribs, domes, segmen-
tal domes and pendentives that support domes or flat Figure 5. Diagram of structural system of the arcade.
ceilings.
The central space of a typical module is a cov-
ered street, roofed by a segmental dome at a height of
3.2 Formal structural systems and zones 10.5 m. Along the length of the street of width 8.3 m,
each arcade segment is bound by a combination of two
In discussing the findings of qualitative structural
arch-ribs of depth 1275 mm. and four piers of section
analyses, the major aspects covered include the spatial
900 × 1200 mm. (Figure 5).
zones, structural systems and sub-systems and their
The arch-ribs are subject to smaller lateral thrusts
interfaces and how they contribute to safety.
from the roof (Figure 5, A) and larger arch-actions
Spatially speaking, the entire building can be
resulting in cumulative horizontal and vertical com-
divided into three distinct zones, each of which incor-
ponents at the springing point that exerts an outward
porates a particular structural system. These systems
horizontal thrust at this point. This thrust is contained
incorporate combinations of the aforementioned ele-
by the stepped section of the transverse wall (Figure 5,
ments and structural subsystems, both intrinsically
B) of the building on either side and the piers accom-
and in their physical interfacing. These critical zones,
modate the vertical loads. In effect, shoring these piers
namely the arcade, the octagonal court and gateways
today couldn’t render the extent of safety to the covered
are explained below with their interfacing.
street that the original design of these walls deliver.
An important subsystem here is formed by the
arches that are along the length of the street on each
3.3 Arcade modules
floor.These are designed to span 3 m and rise to heights
These fundamental units, 16 in number, are repeated on of 1.5 m on the lower storey and 1.35 m on the upper.
either side of the octagonal court to give this building The first floor arch accommodates the load of the
its most identifiable architectural form: they articulate roof in excess of that carried by the arch-rib of the street
the mall character (Figure 4). while the lower arch, along with defining the cell’s

163

Figure 6. Arch subsystem along arcade with parapets on
first floor: 1912 photo from ASI Annual Reports.
opening, acts with the pier in supporting the inter-
mediate floor. The lower floor arch, characteristic to
Shahjahan’s design, is clearly designed for structural
safety, as suggested by the absence of a wall on the
upper floor; the first floor arch overlooks the street
over a low sandstone parapet (Figure 6).
In the arcade module, it is also necessary to refer
to the structural subsystems that spatially form two
cells on the lower storey and one on the upper. Each of
these measures 9 m. square. Roofs of these completely
enclosed cells are supported by the technically supe-
rior elements of pendentives (Figure 7), as opposed
to squinches or traebeate devices. On the first floor, a
semi-covered projection at the entrance inverts access
of the cells. They are approached from the outer sides
of the building. Here also, a half-dome is formed by
pendentives that partially enclose this space. This kind
of a design detail, for a small span of 3 m, demon-
strates technical knowledge, which today makes these
areas safe.
When symmetrically arranged, these arcade sys-
tems in the building cancel any possibility of unbal-
anced stresses or lateral thrusts. The following sections
examine the design of the critical asymmetrical junc-
tions of these with morphologically distinct spatial
zones to dissipate extra stresses.
3.4 Octagonal court and arcade module interface
The “Chowk” or court at the centre of the long Bazaar,
is both visually and spatially conspicuous by being a
source of light for the otherwise covered street. Besides
this significant environmental function, its role as a
structural system contributing to safety by design is
far from negligible. To appreciate this point, refer to
Figure 8 below, where the terminating arcade system
may be viewed in the absence of the built elements
164
Figure 7. Pendentives used in the cells, photo of corner.
Figure 8. Diagram of octagonal court interface with arcade:
vertical loading components shaded in plan and axonometric.
contiguous with this open space. Taking this perspec-
tive, it becomes clear that the springing point of the
last arch-rib is thrust outward by the lateral stresses
caused due to the arcade’s roof.
The plan (Figure 8, B) shows us how this very line
forms the shorter edge of the octagon that, in the ver-
tical plane has a solid wall mass, adding a vertical
component. This is in supplement to the wall above
half of the arch-rib span, as part of a long edge of
the octagon. The vertical component thus added, is
transferred to the arches in the shorter faces and subse-
quently to piers that form the vertices (Figure 8, A–B).
The lateral (non-orthogonal) component is then dissi-
pated along the staircase block down to the arched
opening on the outer edge of the building, beyond the
octagon at the lower level.
The massing of the building around the octago-
nal court suggests that risk of failure at all points
is eliminated by increasing the vertical loads at all
nodal points that are prone to asymmetric loading. This
is a demonstration of the building design achieving
safety through elements in vertical planes. However, it
should be appreciated that the outer built areas of this

Figure 9. “Safe” structural design of Lahore Gate and
arcade interface highlighting the critical components in the
diagram.
space, owing to planar design of the elements, are the
structurally the most fallible.
3.5 Gate and arcade module interface
The longer ends of Chatta Chowk Bazaar are defined
by two distinct gateways. These are the Lahore
Gate and the “Darwaza-i-Chatta-Chowk”, the Bazaar’s
entrance and exit points respectively as one walks into
the palace-citadel.Though different in scale and mildly
different in design, the structural behaviour of these
two spatial zones is quite similar. The more elaborate
and larger of the two, the Lahore Gate and its junction
with the arcade system, shall be discussed here.
Being a complex structural system in itself, this
needs to be described briefly prior to the examina-
tion of the intersection with the arcade. Spatially, the
internal clear area of this gate is an irregular octagon
(Figure 9, C) that is spanned by a dome (Figure 9,
B) touching its longer edges over a distance of 9.3 m.
The vertices of the octagon are defined by eight piers
that rise to form, two vaults laterally and two arch-ribs
longitudinally, along the longer edges of 4.8 m each.
Four pendentives are formed along the shorter edges
of 2.7 m each. These support the large dome overhead
while along the transverse direction, smaller stone
columns support a mezzanine spanning the length of
the vault. The significant transverse depth of the gate,
largely of solid masonry, accommodates any lateral
thrust.
The presence of the pendentives in this system,
illustrates a magnified, finer resolved version of the
sub-system described earlier for the cells. Here, all
walls in the cells are replaced by piers, and their
corners, by the shorter edges of the octagon. This
is a visible sign of structural respect warranted by
the threefold span. Again, structural reliability of
165
Figure 10. Sunken roof above “Darwazah-i-Chatta-
Chowk”.
the pendentives has been employed; this also min-
imizes threat to structural safety through equitable
distribution of stresses.
Examination of the intersection with the arcade sys-
tem reveals the use of a combination of a segmental
dome, which is a three-dimensional composite ele-
ment and dual piers at each end (Figure 9,A) of spans
of 4.8 m (gate octagon) and 8.3 m (arcade). The dimen-
sions of the pier sections designed in the octagon are
roughly 1.1 m square, qualifying the extra load taken
by them from the arcade arch. The horizontal thrust of
the arch-rib of the arcade is directed vertically over the
curved surface of the segmental dome that links the two
differently sized arches.. This intersection thus demon-
strates an elimination of unbalanced thrusts, arising in
two very different structural systems.
Apart from the Lahore Gate, it is also noteworthy
that the opposite gateway to the east the “Darwaza-
i-Chatta-Chowk”, responds to safety in massing. This
is done by sinking the floor above the dome to make
the roof light (Figure 10), a balance between smaller
span, lower rise of its dome and dead-load of roof.
3.6 Salient aspects in construction
Formal aspects in structural values in Chatta Chowk
Bazaar are reinforced by certain very significant built-
in values in construction. These are the links to the
building’s stability, durability and timelessness.
Along with secondary technical sources, the pri-
mary information on construction was extracted from
exposed parts of the building’s fabric itself. As men-
tioned earlier, the hunting lodge at Jaunti, a technolog-
ical contemporary to Chatta Chowk Bazaar, was used
as a resource to understand construction through its
exposed details.
Though the discussion on this aspect is limited
by sources of information, logical conjectures were
drawn to answer questions of longevity and dura-
bility of the building’s fabric. As mentioned earlier,
the Bazaar is, with rare exceptions, a compressive
brick masonry structure. Two observations described
Figure 11. Wall construction, Jaunti; and possibly Chatta
Chowk.
Figure 12. Construction of floor/roof as observed in Jaunti.
below, one about purely compressive elements and
the other about composite elements, indicate the his-
toric durability of this building’s fabric vis-à-vis brick
construction.
In walls and piers of the Chatta Bazaar, brick cours-
ing in the core is very different from what it appears
to be on the surface. An interlaced coursing system is
followed that avoids both vertical as well as horizontal
joints (Figure 11).
Such brick coursing to form walls or piers of the
different sizes mentioned earlier, makes these struc-
tural elements behave as conceivable monoliths when
loaded vertically, thus making them extremely stable.
With reference to floors, domes and pendentives,
a consistently alternating radial pattern is followed in
brickwork. In floors, the smallest faces extend outward
(Figure 12), causing contact of the largest faces to form
flat arches; in domes, they are radially laid, each course
at a time along the largest face – a method that elim-
inates formwork. In pendentives, they are interlaced,
similar to that in the walls.
In such coursing also, a high degree of stiffness is
achieved in each structural element, intrinsically and
presumably, in their interfaces, or in subsystems.These
coursing patterns ensure substantial structural safety
due to a distributed load transfer that takes place from
one structural element to another.
166
Figure 13. Built interventions and consequences: profile of
collapsed arch at the octagon; built cubicles on the first floor.
4 VALUES CURRENTLY THREATENED
In spite of the values as explained above, the Bazaar
is currently endangered by interventions arising from
lack of information about these. Observations of the
alterations that have occurred in this building indicate
a lack of knowledge of significance.
While the implications on the structure per se are
not visible, the building lacks severe structural cracks
or deflections as yet; they may be anticipated if these
practices continue. It is essential to see what these
issues are at this juncture, given the significance of
the building.
Though the logic for its use has changed over its
three hundred and fifty odd years of history, it has
functionally more or less served as a covered market
street in its lower storey. A marked transformation in
use of the building was the housing of soldiers on the
upper floors, both in pre-independence (1857–1911)
and post-independence times (1947–2003).
The introduction of new masonry viz. walling of
the balconies facing the street to form rooms, build-
ing of bathrooms and toilets on the first-floor terraces
(Figure 13, B) were some consequences of this.
Toilets were also introduced in the southern part of
the building at the octagon on the first floor, mentioned
earlier to be a structurally vulnerable area, conse-
quently leading to the collapse of the arches on the
ground floor (Figure 13, A).
A major periodic intervention is the annual renewal
of the tar layer covering the arcade’s central street as
well as on the exterior of the building, catering to the
national Independence Day celebration that happens
atop the Lahore Gate. Due to this, the shallow plat-
form, traditionally used as a seat for the shopkeepers,
is almost flushed with the ground today, suggesting no
less than 450 mm. of rise in level. In addition to this,
the roof of the entire structure was covered with tar-felt
in 1911–12. The use of the upper floors as soldiers’
quarters caused puncturing of windows through the

Figure 14. Synthetic and natural threats.
external historic walls (Figure 14, A). Since 1992,
foliated arches that form the shop openings were also
gradually walled for signboards.
Electrical interventions have been the most marked
where ceiling fans and other such fixtures have been
hung from tubular steel rods that puncture the penden-
tives in the cells on the first floor. In some cases, the
buttressing wall has also been punctured to link two or
more cells from the inside.
Apart from the synthetic interventions mentioned
above, the southeast corner of the building is also
threatened by roots of a large tree that has grown into
the wall (Figure 14, B).
5 CONCLUSIONS
The Chatta Chowk Bazaar today is an expression of a
unique technological memory. Historically conceived
and built with cognizance of processes, techniques and
systems evolved and perfected over time, it is also sig-
nificant in its geographical (here, Indian) context by
being a lone example of a building type. This typolog-
ical significance was achieved through the architect’s
knowledge of the structural and construction systems,
amongst other things, that could make this architecture
not just possible, but also durable and timeless.
For substantive knowledge generation as explained
above, the Chatta Chowk Bazaar and its context on
the ground were used as resources to firstly inform its
typological significance and then extend it to include
structural values. It substantiates how the building’s
values are linked to the contemporary need for its
safety and the safety of its users. The issues also indi-
cate that the practices that undermine the Bazaar’s
typological significance as a historic mall also under-
mine its safety because of this significance embodying
numerous structural values.
The above definition of significance notwithstand-
ing, the scope for further enquiry is immense and will
contribute further to the body of knowledge about
167
Chatta Chowk Bazaar. This building is an irreplaceable
and timeless resource in this regard and will have to
be treated with its due respect. The knowledge sys-
tems approach has helped acknowledge the value of
this resource comprehensively and direct an attitude
for best practices in the future.
In the debate between safety and significance, cases
like Chatta Chowk illustrate that they are actually sides
of the same coin. The most effective route to achieve
safety in such outstanding knowledge systems, is thus
through authentic structural conservation, informed by
a systematic method to generate knowledge, given the
significance of these structures.
ACKNOWLEDGEMENTS
The author immensely thanks Prof. Nalini Thakur,
School of Planning and Architecture (SPA), for crit-
ical guidance, encouragement and contribution in the
work on which this paper is based, especially on using
the Knowledge Systems approach that greatly helped
the studies. Sessions with co-faculty of studio, Ar.
Sanjay Bharadwaj, Morphogenesis, Delhi and on site
resource person, Ar. Rachna Vishwanathan were also
instrumental in guiding the documentation and analy-
ses. The Archaeological Survey of India (Delhi Circle)
and the Chatta Chowk Bazaar Traders’Welfare Associ-
ation were very helpful and cooperative during the site
work. Further, the author thanks all his colleagues who
worked on this project, batch of 2006–08, Masters in
Architectural Conservation, SPA: Allahyar Allahyari,
Bhavesh Patel, Moinak Bose, Muzakkir Bijli, Sidharth
Roy & Sonal Chitnis.
REFERENCES
Asher, Catherine B. 1992. The Architecture of Mughal India.
New Delhi: Cambridge
Croci, Giorgio. Structural Preservation of the Architectural
Heritage. Structural Aspects in Restoring Monuments,
symp. Rome 1993.70: 15–18
Koch, Ebba. 2002. Mughal Architecture ,New Delhi: Oxford
Thakur, Nalini. 1988. Shahjahan and his relevance to the
future of Shahjahanabad. Indian Institute of Architects
Conf.., Goa, 1988: 8–15
Thakur, Nalini. 2006. Technological Knowledge Systems
Approach - from little insights to a paradigm shift in struc-
tural knowledge and application. Structural Analysis of
Historic Constructions Vol. 1: 25–30
Thakur, Nalini. 2002. The “Architectural Knowledge Sys-
tems” Approach: as a Solution for the regeneration and
Conservation of Indian Heritage. Architexturez: South
Asia: 1–10
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Historical constructions – Authenticity and adaptation to the

modern demands

I. Bucur-Horváth, I. Popa, M. Bulbuk & J. Virág

Technical University of Cluj-Napoca, Romania

ABSTRACT: The study deals with the possibilities of rehabilitation of historical buildings while preserving
their authenticity, as much as possible. It starts with a comprehensive definition of the historical constructions
and emphasizes the features of their authenticity. The study shows that the issue of historical constructions is
included in the overall problem of rehabilitation, but with specific requirements regarding authenticity. It presents
the basic principles of the historical buildings’ rehabilitation. Case studies regarding the structural strengthening
and the possibilities of preserving the authenticity of historical buildings during their rehabilitation are presented.
They are ordered increasingly, according to the degree of the intervention needed.

1 INTRODUCTION

Historical buildings are specific for a former human

culture whose product they are. They include repre-
sentative values for a certain epoch, place, human
community, from the perspective of constructive art,
knowledge, materials and workmanship, techniques
and technologies etc., being influenced by environ-
mental conditions, local traditions, spirituality of the
local community, previous and neighboring cultures,
social and political background and so on. Therefore, Figure 1. Theoretical and real curve of degradation.
historical buildings exhibit many features that together
may constitute the authenticity of the construction and
the base for its significance.
Actual elements of the authenticity of a histor-
ical construction could refer to their fitting into
the surrounding environment, general composition of
the building, organizing the inner spaces, external
aspect and proportions, architectural details and orna-
ments, structural conformation, structural elements
and details, materials and technologies of putting into
work. Generally speaking, each construction and thus
also the historical ones come into being as a result
of a social command. It incorporates many perfor- Figure 2. Cases of rehabilitation need.
mance characteristics responding to the beneficiary’s
exigencies. In general, they may refer to architec- There could be some differences between the the-
tural aspects of functionality, aesthetics, details etc., to oretical degradation of one or more performance
structural demands of stability and serviceability, fire characteristics of the building and the real process of
and explosions safety, durability as well as economy. degradation, marked by repeated depreciation-repair
Each performance characteristic has a certain cycles. An operation of rehabilitation becomes nec-
design value that normally decreases to an allowed essary (Fig. 2) in the case of an accidental downfall
minimum during the service life of the construction of one or more performance characteristics under
(Fig. 1). the necessary minimum, because of an unexpected

169
action (ground sliding, severe earthquake, environ-
mental catastrophe etc.). The rehabilitation is also
needed when one wants to keep the building func-
tional, beyond its normal service life.This is the overall
case of the historical constructions.
2 REHABILITATION OF HISTORICAL
CONSTRUCTIONS PRESERVING
AUTHENTICITY
Obviously, the built environment may suffer vari-
ous changes imposed by urban planning necessities.
At the same time, due to its moral wear, each con-
struction should periodically undergo modernization,
functional improvement and even functional modifi-
cation. Moreover, different deficiencies and damages
could intervene during the life, inclusively over the
normal service life of the construction. In other words,
the needs of functional changes can overlap with the
general worn condition of the construction. All these
factors lead to a decision of rehabilitation.
All interventions on a historical building have to be
performed very carefully. Besides the general demands
concerning the functionality as well as structural relia-
bility (resistance and stability, serviceability, durabil-
ity and maintenance), the rehabilitation process has
to respond to some specific requirements regarding
authenticity.
At first, the rehabilitated building should harmonize
with the surrounding environment. Then, the rehabili-
tation solution has to respect the architectural style, the
geometrical proportions, the aspect and atmosphere of
the inside, the external and internal ornaments.
The structural interventions have to assure struc-
tural compatibility with the original structure. It is
recommended to keep the original structural form,
as much as possible while performing the necessary
repairs and strengthening. Modified or new structural
elements should not disturb the architectural appear-
ance and the aesthetics of the building. The repairing
or strengthening materials should be compatible with
the original ones both from the physical and chemi-
cal point of view. They should have, at least, the same
degree of durability as the original ones. In order to
perform a correct intervention one has to take into
account the different stages of the construction, each
with its own style representing a certain epoch, that
should be respected (Bucur-Horváth, 2006).
These basic principles of preserving authenticity
often come in conflict with the real needs of functional
development of the construction and fundamental
requirements of reliability that have to be fulfilled. In
these cases it is necessary to put in balance all objec-
tive data and scientific arguments. In order to preserve
the signs of authenticity of the historical construction,
as much as possible, assuring the structural reliability
170
at the same time, the rehabilitation decision on the
structural interventions could be a compromise.
Sometimes, the necessary functional improvement
of an old building implies adding new blocks, preserv-
ing the structural and architectural authenticity of the
original one.
As a general demand, it is to be mentioned that the
added structures and structural elements should have
an appropriate degree of reliability with those of the
original structure.
3 CASE STUDIES
The above presented principles and possibilities of
preserving authenticity of historical buildings during
their rehabilitation are exemplified in case studies of
structural strengthening. They are presented gradu-
ally, in increasing order depending on the degree of
intervention.
Table 1 presents structural interventions for
retrofitting the bearing capacity of a structure or struc-
tural element, preserving the entire original structure.
The table includes cases of structural strengthening
carried out on the masonry structure of Baroque
buildings. The most common solution for improving
the bearing capacity of the structural elements is to
increase their transverse sections with a covering or
to embed strengthening elements in a proper way. In
these cases, it is very important to take into considera-
tion the physical and chemical compatibility between
the original and the strengthening material.
Table 2 includes examples of structural strength-
ening by modifying the original structure with com-
patible structural elements. Normally, they consist of
additional elements that do not essentially alter the
static behavior of the structure.
Often, metallic tie-rods shall provide for horizontal
thrusts of arched and vaulted structures. These tie-
roads can be of several types or in several positions
(a, b, c), according to the structural form, height of
supports, accessibility, aesthetic consideration etc.
Joining girder (ring beam) of reinforced concrete
has to be added on the slabs level of the old masonry
structures, in many cases.
The strengthening solution with tie-bar system can
be used to reestablish the moment bearing capacity of
large-spanned beams.
Table 3 contains examples of indirect strengthening
of historical structures. In order to maintain the origi-
nal structure or structural element, additional bearing
elements are provided.
The basic idea is to maintain the original struc-
ture, but due to its weakness or very poor technical
condition, to discharge it from an important part of
the vertical load by an additional bearing system. The
Table 1. Strengthening, preserving the original structure.

171
Table 2. Strengthening, adding compatible structural elements.

172
Table 3. Indirect strengthening providing additional bearing structure.

173
Figure 3. New staircase – exterior view.
original structure shall be repaired or restored. From
that point on, it has to carry only its own weight.
Another idea is that of using additional helping sys-
tem in order to correct displacements or deflections
that affect valuable elements of authenticity of the
historical building.
Sometimes, the necessary functional improvement
of an old building could imply adding new blocks to the
building, preserving the structural and architectural
authenticity of the original one.
For instance, in the case of the Palace of Jus-
tice in Odorheiu Secuiesc, Romania, built in 1835,
belonging to the architectural heritage, an improve-
ment of the circulation flux was necessary, responding
to the current requirements concerning emergency
exits (Bucur-Horváth 2007). A new staircase made of
reinforced concrete and glass was added to the ini-
tial building (Figs 3–4). Placed in the inner courtyard,
this very effective modern solution emphasizes the
historical character of the old building.
4 CONCLUSIONS
Any intervention on a historical building has to be
performed with great attention. Besides the general
demands concerning functionality, as well as structural
reliability, the rehabilitation process has to respond to
some specific requirements regarding authenticity. In
this sense, the rehabilitation decision and technical
174
Figure 4. New staircase – interior view.
solution should respect some basic principles, but
every case represents a new challenge. The specialist
has to establish the priorities in preserving authentic-
ity, without neglecting the functional and structural
reliability of the building.
REFERENCES
Bucur-Horváth, I. et al. Study on old masonry structures in
brick vaults. In Dr. GörünArun & Dr. Nadide Seckin (eds),
Studies in ancient structures; Proc. 2nd intern. congress,
Istanbul, 9–13 July 2001. Istanbul: Yildiz Technical Uni-
versity, 2001.
Bucur-Horváth, I. et al. Theoretical and experimental stud-
ies for strengthening Bohemian brick vaults. In Clau-
dio Modena, Paulo B. Lourenço & Pere Roca (eds),
Structural Analysis of Historical Constructions; Proc.4th
itnern. seminar; Padova,10–13 November 2004. Leiden:
Balkema, 2004.
Bucur-Horváth, I. et al. Researches on historical construc-
tions built in several stages. In Paulo B. Lourenço,
Pere Roca, Claudio Modena & Shailesh Agrawal (eds),
Structural Analysis of Historical Constructions; Proc.5th
intern. conference; New Delhi, 6–8 November 2006.
Delhi: Macmillan India Ltd., 2006.
Bucur-Horváth, I. & Bulbuk, M. Original solutions for struc-
tural and functional rehabilitation of masonry buildings.
In Görün Arun (ed.), Studies on Historical Heritage; Proc.
intern. symposium; Antalya, 17–21 September 2007.
Istanbul: Yildiz Technical University, TA-MIR, 2007.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Champaner-Pavagadh – Managing conflicts – A conservation challenge

Sonal Mithal Modi

Conservation Architect, People for Heritage Concern, Surat

ABSTRACT: The present paper focuses on the experiences gained, methods used and attitudes assumed while
formulating a management system for the Champaner-Pavagadh Archaeological Park. It brings out the conflicts
which occur when there are several stake holders managing the site; who have good intention but narrow rather
limited purview of action and vision; and ponders whether solutions to these conflicts can be prioritised with
mutual consent as a strategy for effective conservation management of the site.

1 BACKGROUND then several consecutive settlements have come about

on its various plateaus. Each settlement had its own
Champaner-Pavagadh Archaeological Park is a highly self sufficient water system making it a unique water-
complex heritage site with a 16th century medieval intelligent city.
Sultanate capital city buried beneath a thick forest
cover and landscape characterised by plateaus, mounds
and streams. (Refer Figure 1). Although stripped off its
glorious past and since forgotten; it is widely visited
by pilgrims from all over the western region of India
as it is one of the Shakti Peeths and also an important
Jain pilgrimage centre. (Refer Figure 2).
Today, with obscure Rajput and Sultanate struc-
tures (Refer Figure 3), buried city, temples, myths
and legends passed down since generations by word
of mouth; Champaner-Pavagadh is really a most illus-
trious example of a heritage site with diverse heritage
components viz., natural, built, traditional knowledge
systems and intangible heritage. It spreads over an area
of approximately 14 sq. km Core Zone and 30 sq. km
Buffer Zone. (Refer Figure 4).
The earliest settlements occurred on the hill dur- Figure 2. Kalikamata temple on the summit.
ing the Rajput regime around 9th century and since

Figure 1. Pavagadh hill. Figure 3. Sultanate Remnants – Khajuri Masjid.

175
Figure 4. Quickbird satellite image of the site.

2 PRESENT STATUS OF SITE

MANAGEMENT

Champaner-Pavagadh was declared a World Heritage

Figure 5. Forest destroying the vulnerable heritage.
Site by UNESCO in 2004. This is the only example in
India to have been recognised as a site and not as a city
or a group of monuments; and is the only one to have
met four out of the six criteria. What is unique about
this site is its authentic pre-Mughal Sultanate evidence
lying absolutely untouched courtesy the overgrown
forest. This authenticity makes it a very significant
knowledge resource.
Today, the site is being managed by several stake
holders who are either owners of some heritage compo-
nent or are instrumental in the management of it. They
vary in their aims towards the site and although well-
intentioned they are frequently tied down by limited
purview of action and vision. Hence any step towards
an integrated conservation of the site leads to more
often than not a status-quo.

Figure 6. Gaben Shah Vav.

3 CONFLICTING AIMS OF STAKE HOLDERS

Forest Department has under its purview a large area
of the site, mainly the Pavagadh Hill and the buried
Sultanate city. The forest act is very powerful and
has successfully controlled development in the for-
est land thereby protecting the authenticity of the site.
But, at the same time, its obligation of planting fresh
saplings every year has proved to be a big threat to
the archaeological heritage. The roots of the planta-
tions are slowly harming the vulnerable heritage buried
below the earth. (Refer Figure 5). The overgrown for-
est also makes accessibility to most structures very
difficult leading to their subsequent deterioration.
ASI protects, out of the identified 114 structures,
only a mere 55 and that too in isolation. It has cre-
ated islands of protected territories within the entire
Archaeological Park. Paradoxically enough, it was the
ASI which nominated this place as a site to WH com-
mittee; but when it comes to protection it can not see
beyond its 100/300 m absurd irrational law.
To add to that the State Department of Archaeol-
ogy has just one monument under its jurisdiction i.e.,
the Gaben Shah Vav. (Refer Figure 6). It protects only
those buildings that are more than a 100 years of age
and only the ones in stone! Its law does not allow for
protection of significant buildings like Iteri Masjid,
Rani no Mahal, etc.; simply because they are brick
structures, in spite of their being important knowledge
resource!
The District Collector and the DDO once spent
enormous money and energy constructing the water
facility at a heritage site. But it happened to fall within
the 300 meter boundary of ASI and had to suffer
demolition even before it started functioning. These
uncalled-for conflicts and a constant struggle for asser-
tion of one authority over the other are is not helping
the site, rather they are wasting resources. A system
is needed which enables co-ordination among various
government departments.
Gram Panchayat promotes programmes for peo-
ple’s development like repairing of pilgrims’ path,
creation of tourism oriented business opportunity, etc.
Recently, Gram Panchayat has constructed a school in

176
Figure 7. Rampant growth of Kalikamata temple Precinct.
the Gaben Shah precinct! Now they are at logger heads
with the State Department of Archaeology. Where to
develop and how to develop is not known to them!
These people don’t know where to seek answers from
and also whether they are really responsible for their
provision in the most rightful sense.
The local Panchayat is taking emergency measures
to reinstate the traditional water system. The will
to conserve and awareness is there among the local
leaders, but they lack a professional insight. Such ini-
tiatives are a big asset for the site. The need is to create
situations for methodical interventions enabling con-
venience to whoever is concerned and empowered to
take actions.
Temple trust like that of Shahji Sawai nu Deru
has a precinct within the Royal Enclosure. It has
already developed it as a neat RCC structure having
dharamshalas and a new structure over the original
mazar. As it is a private property, the trust has the
freedom to develop the site as it wishes. In such sit-
uation it becomes questionable whether a uniformity
and sympathetic growth is valid at all or not. If it is so,
there is a lack of system that can ensure a sensitive-
responsibility and an effective monitoring mechanism
without one body being assertive over the other.
The living temples, which are of archaeological
value, are being modified beyond recognition in the
name of loving care and maintenance. Take for exam-
ple, Jai Mataji Temple Trust has created fortifications
and loud RCC platforms for the ‘convenience of the
pilgrims’. (Refer Figure 7). Even the pilgrims are
happy about it. But the cultural value of the precinct
is gone for a six. But again, conservation profession-
als have no right to ‘preserve’ the sanctity of this
place as they understand simply because they are con-
vinced about its value for their profession’s sake alone.
What is actually needed is a system which discourages
haphazard interventions, and encourages pertinent
development strategies resonant of the cultural values
of the place?
PWD has proposed widening of the State highway
that passes thru the site and it also seeks to widen the
road going up to Machi! Now, if the priority is effective
transportation of the people, road widening cannot be
177
Figure 8. State highway cutting through the Fort Wall.
the only solution. (Refer Figure 8). They are oblivious
to or are ignorant about proposals of battery operated
vehicles vying between the visitor centre and Machi
so as relieve parking pressures on the Machi plateau.
The state highway has been proposed to be re-routed
to avoid thorough traffic in the site. But the proposals
at times fail to reach the right target, hurdling effective
implementation.
Champaner-Pavagadh is a water-intelligent city,
envisaged so in the Rajput times and refined during
the Sultanate era. At one point of time, it catered to
a populace of 50,000 today it is unable to quench a
mere 5,000 owing to sheer lack of knowledge about the
traditional water system. In spite of recent researches
and local campaigns Irrigation Department wishes to
remain oblivious to such knowledge and is propos-
ing a huge subsidiary canal from Narmada canal to
be brought in to Wada Talao destroying the heritage
underneath. When there are ways of avoiding destruc-
tion authorities responsible should take sympathetic
attitudes, rising above individual agendas. Perhaps
in the light of a missing management system, such
conflicts are bound to occur.
Field owners have been carrying out agricultural
activities for over 200 years now, which have already
resulted in the loss of important archaeological evi-
dence. They need to irrigate their fields and so creation
of canals, digging, etc. is widespread. Their interest is
their right. Then the academic conservation becomes
weak for management of such a heritage site. Amity
has to be ensured between these conflicting situations.
NGOs like Heritage Trust have no ownership, but
they have in their kitty ‘concerned and sensitive pro-
fessionals’ and so were entrusted the task of repairing
Malik Sandal ni Vav at Mandvi village. The conserva-
tion intervention may have been done correctly; it may
have had flaws too? Who monitors such actions? The
funding agency is merely concerned about the immedi-
ate product-oriented targets set and is blithe about the
 Figure 11. Champaner village.
Figure 9. Malik Sandal Vav before repair work.

Figure 12. Tourists/pilgrims are a big stake holder.

Figure 10. Malik Sandal Vav during repair work.

relevance of individual interventions vis-à-vis the over

all site. (Refer Figures 9, 10). Because such projects
can only be granted to individual professionals, there
is a need for a system that safeguards a consistency
among all such individual conservation interventions
in such a complex site.
Residents of the Champaner village who unfortu-
nately stay within the Royal Enclosure cannot even
get a toilet constructed in their house, because it falls
within the ASI protection. (Refer Figure 11). These
residents have been staying here long before the enact- Figure 13. Wayside deities enroute to Kalikamata temple.
ment of ASI law, yet they have to suffer. It is difficult
to ensure hygiene and good life-style to local residents Western railways have an abandoned railway track
if ASI laws remain as stringent without any scope for cutting through the forest on the site. This can innova-
flexibility. tively be clubbed with the tourism experience, it being
Tourism Department is happy having their agen- a fantastic resource. A system has to look at potentials
das relegated to mere provision of public conveniences and constraints of such resources lying strewn across
when today the site needs educative, interactive, par- the site.
ticipatory modes of tourism especially, in the wake of A vast number of pilgrims and tourists visiting the
this being an intense pilgrimage site and a potential site make it very susceptible. They are a major source
knowledge resource. (Refer Figures 12, 13). granting stable economy to the local residents so key

178
efforts go into pleasing them often overlooking the
vulnerability of the site.
4 REASONS FOR OCCURRENCE OF
CONFLICTS
Main conflicts arise:
1. Due to non availability of complete information
regarding various issues to each stake holder.
2. Due to vast amount of cultural resources lying un-
possessed and hence unattended by any agenda of
any stake holder.
3. Due to pilgrim/tourist oriented opportunities which
make the site economically self-sustained but also
highly vulnerable.
4. Due to the presence of a living village whose aspi-
ration of a better life-style can never be overlooked
on any grounds whatsoever.
5 PROPOSED MANAGEMENT SYSTEM
Hence a four point agenda has been formulated
which enables management of Information, Cul-
tural Resource, Visitor/Pilgrim, Site or the Physical
Resource.
The Management plan proposed is an attempt at
prioritising actions and at reducing conflicts if not
doing away with them altogether. Govt of Gujarat, with
the constant persuasive efforts of Heritage Trust, has
recently instituted a legal authority with such inten-
tions. But it is still a long way before various stake
holders open up their constricted vision and co-operate
towards a peaceful co-existence. Convenient misinter-
pretations of the authority by lazy officials are leading
to several bottlenecks in the development procedures,
yet all efforts are being made to ensure a continued
collaborative effort between Panchayat, government
bodies and professionals.
6 CONCLUDING REMARK
It is indispensable that Conservation is inte-
gral to an Urban Planning Process enabling a
Progress-Oriented-Approach because… Heritage is
living… ageing… and… it wants to let fade away… to
give way to a people’s right to human living
conditions.
BIBLIOGRAPHY
Alfieri, Bianca M. 2000. Islamic Architecture of the Indian
Subcontinent, Ahmedabad: Mapin Publishing Pvt. Ltd.
179
Bayley, Sir Edward C. 1970. The local Mohammadan Dynas-
ties, Gujarat, New Delhi: S. Chand and Co. Ltd.
Brown, P. 1996. Indian Architecture – Islamic Period, Bom-
bay: D B Taraporewala Sons & Co. Pvt. Ltd.
Burgess, J. & Cousen, H. 1975. Architectural Antiquities of
Northern Gujarat, Delhi: Archaeological Survey of India.
Burgess, J. 1994. Muhammadan Architecture of Bharuch,
Cambay, Dholka, Champanir, and Mahmudabad in
Gujarat, New Delhi: Archaeological Survey of India.
Department of Landscape. 2001. Champaner-Pavagadh
Archaeological Park Plan, Urbana Champaign: University
of Illinois.
Chavan, A. R. 1966. The Flora of Pavagadh, Baroda: Depart-
ment of Botany, Faculty of Science, M. S. University.
People for Heritage Concern. 2001. Cultural Resource
Information System – Inventory of Built Heritage of
Champaner-Pavagadh, Baroda: Heritage Trust.
Faridi, Fazlullah L. 1990. Mirat-i-Sikandari – A Study of
Medieval History of Gujarat, Gurgaon: Vintage Books.
Modi, Sonal M. 2002. Water Intelligent City, Land-
scapes of Water, History, Innovation and Sustainable
Design, Politechnico di Bari – Facolta di Architet-
tura, Dipartimento di Scienze dell’Ingehneria Civile e
dell’Architettura (ICAR), Bari, Regione Puglia, Italy.
Goetz, H. 1949. Journal of the Gujarat Research Society,
Volume XI, Issue No. 2.
Joshi, G. 1999. Pavagadh Darshan, Ahmedabad: Chiragh
Printers.
Mehta, R. N. 1979. Champaner – An Experiment in Medieval
Archaeology, Baroda: Ajanta Publications.
Mehta, R. N. 1979. Champaner – Ek Adhyayan, Baroda:
Maharaja Sayajirao University.
Mendonca, Francis A. 1981. Chronica do Reyno de Gusarate,
Ed., Misra S. C. & Mathew K. S., Department of History,
Faculty of Arts, Maharaja Sayajirao University, Baroda.
Modi, S. 2004. Impressions of a Forgotten City – Archi-
tectural Documentation of Champaner-Pavagadh, New
Delhi: Archaeological Survey of India & Heritage Trust
Baroda.
Misra, S. C.1985. Tarikh-i-Mahmus Shahi, Baroda: Depart-
ment of History, Faculty of Arts, The Maharaja Sayaji Rao
University of Baroda.
Patel, G. D. 1972. Gazetteer of India – Panchmahals District,
Gujarat Ahmedabad: Government Publications Depot.
Sandesara, B. J. 1973. Translation of Gangadhar Prani-
tam, Gangdas Pratapvilasa Natkam, A Historical Sanskrit
Play, Composed 1449, Baroda: Oriental Institute.
Thakur, N. 1987. Champaner- draft Action Plan for Integrated
Conservation, Baroda: Heritage Trust.
Thakur, N. 2000. Note on the legal Status of Champaner,
Unpublished.
Thakur, N. 2000. Participatory Conservation Collaborative:
A Regeneration Programme for Champaner-Pavagadh,
Unpublished.
Thakur, N. 2000. Archaeological Park as a tool for Inte-
grated Protecting Heritage Management with Planning
Process: The case of the deserted 15th century capital site,
Champaner-Pavagadh, Gujarat, Goa.
Tripathi, K. 1971. Pravasdham Pavagadh; Baroda: Diamond
Publication.
Watson, J. W.1877. The Indian Antiquary, A Journal of
Oriental Research, Vol. VI.
History and documentation
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Construction and structural behavior of Vladimir Suchov’s Nigres tower

M. Beckh & R. Barthel

Department of structural design, Munich Technical University, Munich, Germany

A. Kutnyi
Department of architectural history, building archeology, heritage conservation, Munich Technical University,
Munich, Germany

ABSTRACT: From 1927–1929, the renown Russian engineer Vladimir Grigorevic Suchov (1853–1939)
designed and built several towers for the Nigres power transmission line. As one of the last and most refined
hyperboloid steel lattice towers, they represent the culmination of his lifelong quest in the development and
optimization of this structural system.

1 INTRODUCTION

Suchov designed and built the first hyperboloid steel

lattice tower for the All-Russian exhibition in Nizhniy
Novgorod in the year 1896. The 25.6 m tall structure
with a bottom diameter of 11 m and a top diameter of
4.3 m consisted of 80 straight steel angles and inter-
mediate horizontal ring elements. The extremely light
and filigree structure served as a water tower combined
with a viewing platform and attracted the attention of
the engineering community and public alike.
In the following years, his invention gained
widespread circulation throughout Russia due to its
cost-effectiveness and structural stability in compari-
son to other tower structures. Suchov standardized the
building type and constructed more than hundred tow-
ers with varying heights and proportions for different
purposes. The tallest structure of this kind is the 150 m
tall Sabolovka radio tower in Moscow, built between
1920 and 1922.
At the end of the same decade, from 1927–1929,
Suchov built six power transmission line towers at Figure 1. The two tall Nigres transmission towers at the Oka
river, photo taken by Igor Kazus in 1989 (Graefe 1990).
the Oka river, about 100 km southwest of Nizhny
Novgorod (fig. 1). The four smaller ones were com-
posed of three, the two taller ones of five hyperboloids,
reaching 130.2 m in height. With their structural clar- 2 THE NIGRES TOWERS
ity, extreme lightness, and simple detailing, these
towers mark the high point of Suchov’s advancement 2.1 Construction and geometry
of hyperbolic lattice structures. The structure consists of five hyperboloids sections
Five of the six towers have been dismantled since with decreasing diameters (34.0 m, 25.8 m, 19.4 m,
they became out of use more than ten years ago. Only 14.0 m, 10.0 m, 6.0 m) stacked upon each other. The
one of the two 130.2 m tall structures is still existent sections are 24.9 m tall, except for the highest one,
today. This remaining structure was recently severely which measures 24.3 m. The three transmission lines
damaged when 16 of the 40 elements in the lower part were supported at 128.0 m. With the trussed outrigger
of the first drum were cut out by looters. The missing at the top, the structure reaches an overall height of
elements will be substituted in the fall of 2007. 130.2 m.

183
Figure 2. Plan view of the 5 hyperboloid sections.

The first three drums are made of 40 members (steel

angles), the fourth and fifth one of 20 members. Thus,
the first, second and third hyperboloids are divided
in plan in 9◦ segments, whereas the two upper ones
are divided in 18◦ . In plan view, the angle ϕ between
the lower and upper end of each vertical expresses the
twisting of the hyperboloid. The angle ϕ is constant for
the three lowest sections (36◦ ) and becomes more pro-
nounced towards the top of the structure (54◦ and 72◦ ).
Due to the rather subtle twisting of the verticals only
the two upper sections display the typical “waisted”
hyperboloid form. The two sets of inversely running
verticals intersect each other 4 times in the lowest three
sections and the highest one, and 3 times in the fourth
sections. (See fig. 2).
Between the hyperboloid segments, horizontal lat-
tice girders act as stiffening elements (See fig. 3) The
two sets of inversely arranged verticals are slightly
offset and joined with two rivets at each “intersec-
tion”. The 10 mm distance between the faces of the
steel angles is hereby bridged with a shim plate.
Each drum is stiffened by ten horizontal ring ele-
ments, spaced 2.26 m on centre vertically. The con-
nections between the verticals and the horizontal ring
elements do not coincide to allow for more simple
connection details. (fig. 4)
The connection between the brackets, which are
cantilevering of the verticals, and ring elements is Figure 3. View of the steel structure from inside with the
bolted. All other connections of the structure (e.g. lattice girders, which separate the hyperboloid sections.

184
 Figure 5. Assembly of the structure using the telescope
method. (Graefe, 1990).

erection of the second drum started. Unlike in the

Figure 4. Steel structure of the first hyperboloid section, case of the neighbouring three-tiered towers, which
showing the slightly offset verticals and the horizontal ring
were built conventionally with an assembly platform
elements. The vertical to the right has been cut by looters.
on top of each hyperboloid, a new construction method
Table 1. Steel sizes of the structure according to the original
was used for the subsequent upper drums. Here, the
construction documents. telescope method was employed, which had already
proven efficient in the erection of the Sabolovka tower
Drum Verticals Ring elements Lattice girder a few years earlier. (See fig. 5)
The subsequent sections were assembled inside the
First L120 × 120 × L80 × 80 × 10 L100 × 100 × 12 shaft. Their feet were pulled towards the center with
12 (40) − 240 mm a wooden corset. Five wooden derricks located at the
Second L100 × 100 × L75 × 75 × 8 L90 × 90 × 9 upper lattice girder of the last drum then lifted the next
12 (40) − 200 mm
section to the top. Once the section had reached its
Third L100 × 100 × L75 × 75 × 8 L75 × 75 × 8
10 (40) − 200 mm intended height, the mountings were slowly released.
Fourth L100 × 100 × L60 × 60 × 6 L75 × 75 × 8 Then the feet were bent outwards to the ring and were
12 (20) − 200 mm connected at their final position.
Fifth L100 × 100 × L50 × 50 × 6 L75 × 75 × 8 The construction of one drum element needed about
10 (20) − 10 mm six to eight weeks. The towers were completed in
1930. The employed telescope method explains why
Suchov designed the lower sections without a real
between verticals and the lattice girder, of members “necking”, even though it would have been structurally
at the outrigger, at splices, et cetera) are riveted. advantageous.

2.2 History of the assembly

The construction of the two five-tiered towers started
in spring 1928. The first drum was constructed with
help of an inner wooden scaffolding, which supported
the structure at certain locations. By mid March, the
3 SUCHOV’S STRUCTURAL ANALYSIS
Suchov’s original calculations for the tower resurfaced
in the town archive of Nishni Novgorod in 2007. The
10 page typewritten structural analysis comprises load

185
assumptions, steel member design as well as the design
of the riveted connections and the foundations.
Referring to load combinations given by the
Russian building code, Suchov states that the high-
est forces in the tower are caused if “the transmission
lines aren’t ruptured, no ice build-up exists, and the
wind pressure is 250 kg/m2 ”.
The design wind pressure of 250 kg/m2 – constant
over the height of the structure – is used to calculate
the resulting forces on the different hyperboloid sec-
tions and at the base. This design pressure is applied
on all the vertical members, using their unprojected
length times the width of the steel angles as the
reference area.
To account for wind shielding at the sides of the ring
elements, Suchov uses the following formula with the
diameter D and the width of the steel angle b to cal-
culate the resulting horizontal wind force on one ring.
The resulting overturning moment of the tower is com-
puted as 7384.21 tm; the self weight of the tower
including the weight of the power lines is 144.26 t.
The forces on the verticals are calculated for each
section with the z following equation:
with the overturning moment M, 2n the number of
verticals, and the radius of the considered ring r.
Hence, the highest forces in the first section due to
dead load and wind are:
This load is used to design the verticals in the first
hyperboloid. As buckling length, the distance between
the support and the connection between the verticals
and the first horizontal ring is used. Suchov argues that
the connections provide enough torsional restraint, so
that column buckling of the L – shaped vertical will
not occur around the principal but the strong axis. The
allowable maximum compressive stress is computed
with the following formula:
with slenderness ratio λ = lc /i. In this manner, the
verticals of the first section are designed.
(E.g. L120×120×12 with iy =3.65 cm,A=27.54 cm2
186
This procedure is repeated with the respective over-
turning moments to design the L-shaped verticals of
the higher hyperboloid sections.
Finally, the riveted connections of the first verti-
cal of each section are designed. The combined shear
and bearing pressure appears to be at the maximum
800 kg/cm2 for 7/8 inch rivets. 7/8, 3/4 and 5/8 inch
rivets are used for the connections according to the
calculations.
In addition, the calculations entail the design of
anchor bolts, the steel members at the top, the stabil-
ity against overturning of the structure as well as the
design of the concrete foundation. There is no struc-
tural design of ring elements or lattice girders included
in the original calculations. (Suchov 1927)
4 NEW ANALYSIS
4.1 Structural model
The structural analysis model is based on the sizes
and dimensions given in Suchov’s original construc-
tion drawings. For the calculations, the FEM program
RStab was used. The structure was modelled in the
following way:
– In order to minimize the number of nodes, the slight
offset between the two sets of inversely running
verticals is only included in the first hyperboloid
section. This is due to the fact that the first section is
more susceptible to buckling (Compare fig. 6). In the
higher sections, the joints of all vertical hyperboloid
shell members are in one plane, thus neglecting the
slight offset.
– Verticals are joined at each intersection of the
hyperboloid shell with fixed connections.
– All the horizontal ring elements and lattice girders
are joined with moment-released connections to the
verticals.
Material testing showed that the historic steel
has similar mechanical properties to S 235 (fy =
235 N/mm2 , E = 210000 N/mm2 ), which was used in
the structural model.
4.2 Wind loads
Wind loads were based on the DIN 1055-4. The
code calls for the wind velocity measured 10 m above
ground with an exceeding probability of 1/50. Only
the 5 year wind velocity was available for the local-
ity of the tower. Based on various reference material
(Schueller 1981), the wind velocity was extrapolated
rather conservatively with the factor 1.39. The result-
ing reference velocity pressure for 50 years was thus
determined as 0.32 KN/m2 .
Table 2. Resulting wind pressures c∗f q(ze ) on the steel
members according to DIN 1055-4.

hu ho ze q(ze ) q(ze )∗ 1.84

Drum [m] [m] [m] [KN/m2 ] [KN/m2 ]

First 0 24.9 15 0.632 1.16

Second 24.9 49.8 40 0.909 1.67
Third 49.8 74.7 65 1.053 1.94
Fourth 74.7 99.6 90 1.139 2.10
Fifth 99.6 123.9 114 1.205 2.22
Outrigger 124.0 130.2 127 1.237 2.28

Taking reduction ratios for angular sections into

account, cf is ascertained as 1.84. The velocity pres-
sure z meters above grade level is in midland areas:

Wind pressures were applied to all the members,

taking their flange width times length as the reference
area. Unlike in Suchov’s original calculations, where
the wind pressure on the horizontal rings is reduced, no
reduction factor was applied to account for shielding
of the members. Even though there certainly will occur
some shielding on the sides of the ring elements, no
code regulation could be found that was applicable to
this specific geometry.

4.3 Stability
To assess the ultimate bearing load of the tower, it
was critical to study its stability behaviour. Therefore,
a geometrical non-linear stability analysis had to be
performed. For that reason, it was essential to find
suitable imperfection shapes.
Following recommendations by Graf, the imperfec-
tion shape of grid shell structures should be deter-
mined by scaling the first eigen-value mode shape
under ultimate load. The first oscillation period was
determined in the modal analysis as 1.02 seconds.
In the case of a tower-like structure, the eigen-value
mode shapes are obviously extremely from the buck-
ling shape of the perfect geometry. The structure of Figure 6. Buckling shape, based on the perfect geometry.
the first hyperboloid section proved to be the most The plan view shows the first hyperboloid section only.
critical. The buckling shape shows six evenly around
the perimeter line distributed bulges, which are con-
fined to the lower half of the section. This buckling
shape of the perfect geometry was scaled and imposed dividing the E-modulus of elasticity by the coefficient
on the structure as the imperfection shape. The max- γM = 1.1. The load increase factor of the imperfect
imum deformation was selected rather conservatively geometry was determined as 4.39, compared to 4.87 of
as 150 mm, a deformation considered to be visible to the perfect geometry. This is a rather modest decrease
the naked eye. compared to other grid shell geometries.
According to DIN 18800-2, the stiffness of struc- Hyperboloids and some other anticlastic formed
tures susceptible to buckling has to be reduced by shells are typically less sensitive to imperfections, due

187
Table 3. Comparison of forces in the first hyperboloid
section under different loadings.

Maximum Maximum Maximum

normal normal bending
forces in forces in moment in
verticals of lowest lowest lattice
Load case/ the lowest lattice girder, around
combination drum girder strong axis.

LC 1 −30 KN 20 KN –
LC 2 −42 KN 29 KN –
LC 3 −393 KN −245 KN 11 KNm
+309 KN +303 KN
LC4 −132 KN −39 KN 3 KNm
+48 KN +99 KN
LC 5 −281 KN −169 KN 7 KNm
+219 KN +212 KN
LC 6 −246 KN −131 KN 8 KNm
+167 KN +184 KN

to the stabilizing effect of the interaction of compres-

sion and tension forces of the two main parabolas.
(Gioncu 1992, Graf 2002).

4.4 Load cases

The subsequent load cases/combinations were exam-
ined, in which the wind was applied perpendicular to
the direction of the transmission lines, in the following
called x-direction (fig. 7). For the analyses, the load
factors 1.35 for dead load, and 1.5 for live loads were
applied.
The self weight of the tower was deter-
mined as 1090 KN. The resultant wind forces
in x-direction for the tower structure itself are
1.5 × 1002 KN = 1503 KN for the 50 year wind, com-
pared to 1239 KN as calculated by Suchov. Although
the base shears are quite similar, the overturning
moments are significantly higher if the modern code
is applied, due to the vertical distribution of wind
pressures. The forces in the members of the first
hyperboloid section under different load cases are
summarized in table 3.
LC 1: self weight of the tower
LC 2: self weight of the tower including the
transmission line, factored
LC 3: like LC 2 plus 50-year wind load in
x-direction, factored
LC 4: like LC 2 plus 24% of the 50-year wind Figure 7. Deformed shape of the tower under self weight
in x-direction, factored and wind load in x-direction, displaying larger deformations
LC 5: Self weight including transmission line at the transition between the different hyperboloid sections.
plus wind load in x-direction according
to the original calculations of Suchov. 4.5 Structural behaviour
LC 6: Current condition without transmission 4.5.1 Under self weight
lines. Self weight of the tower plus Under self weight, all verticals are subjected evenly
5-year wind in x-direction, factored. to compression forces, whereas the lattice girders are

188

Figure 8. Transition between first and second hyperboloid
section. From top to bottom: Structure, followed by qualita-
tive contours of deformation, normal force distribution, and
bending moments of the lattice girder.
subjected to tension forces. This is caused by the dif-
ferently inclined verticals above and below each lattice
girder. The outward thrust is thus balanced by the
lattice girder.
The ring elements are not subject to any forces.They
simply act as bracing elements for the verticals.
4.5.2 Under self weight and horizontal wind load
in x-direction
The tower displays a remarkably high bending stiff-
ness. The horizontal deformation of the structure
under unfactored wind load is 340 mm, resulting in
a deflection ratio of l/388.
Under self weight and wind load, the structure dis-
plays a tube-like load bearing behaviour. All verticals
are subjected to normal forces. Predictably, the verti-
cals at the front face are under tension and the ones at
the rear face under compression. At the “sides”, ver-
ticals are alternating under tension or compression,
189
depending on their twisting angle. This imposes a ver-
tical distortion of the lattice girder at the end of each
section.
The discontinuity of the verticals between the dif-
ferent hyperboloid sections results in a decrease of
the shear stiffness, causing large deformations in these
areas. The lattice girders are subject to linearly increas-
ing compression forces on the front face and tension
forces on the opposite side, thus balancing the normal
forces of the inclined verticals below.
The bending action of the tower causes the lattice
girder to ovalize, thereby inducing bending moments
around the strong axis of the member. (fig. 8) In addi-
tion, the girders are even more affected by the push-
pull action of the sidewise verticals, acting around their
weak axis. The ring elements adjacent to the lattice
girders get affected by the distortion in this area as
well, causing overstress due to bending.
Based on the current analysis and load assump-
tions, the structure would not be adequate to sustain
the 50-year wind, due to buckling of the verticals in
the first section. The structure including transmission
lines would only satisfy code requirements if the fac-
tored wind loads are reduced by 76%. Even the current
condition without the transmission lines would not be
sufficient to sustain the 5-year wind.
Despite of some minor local overstressing, almost
all other members of the structure are suitable (load
case 3).
5 CONCLUSION
The buckling behaviour and wind load assumptions of
the structure will have to be investigated more deeply,
as the results are not in line with the 80 year life span of
the tower. One critical point is the lack of the 50-year
wind velocity at the locality of the tower. Shielding
of members in an open circular framework is another
issue that requires further research. The disregard of
any reduction factor for shielding (due to the lack
of suitable specifications) certainly effectuates a too
conservative assessment of the structural safety of the
tower.
Despite of these unanswered questions, it is intrigu-
ing to comprehend the structural behaviour of these
multi-storeyed hyperboloid lattice towers. The time-
less elegance, ingenuity and courage that the NIGRES
tower exhibits still have a humbling effect on the
beholder.
ACKNOWLEDGEMENTS
The authors are part of a joint European-Russian
research group, which is led by Rainer Graefe, insti-
tute of architectural theory and history, Universität
Innsbruck, in cooperation with Rainer Barthel, Murat Gioncu, N.B. 1992. Instability behaviour of single layer retic-
Gappoev, Uta Hassler, Alexander Kolesov, Igor Molev, ulated shells. In: International journal of space structures.
Manfred Schuller, and Tatjana Vinogradova. Vol.7. No 4.
Graefe, R. 1990. Die Kunst der sparsamen Konstruktion.
Stuttgart: Deutsche Verlags Anstalt.
Graf, J. 2002. Entwurf und Konstruktion von Translationsnet-
REFERENCES zschalen. Stuttgart: Verlag Grauer.
Schueller, G. 1981. Einführung in die Sicherheit und Zuver-
De Vries, P. 1999. Morphology and structural behaviour of
lässigkeit von Tragwerken. Berlin: Verlag Ernst und Sohn.
the hyperbolic lattice. In: 4. International colloquium on
Suchov, G.1927. Structural calculations of a 128 m tall tower.
structural morphology. Delft: University of Technology.
city archive of Nizhniy Novgorod.
DIN 18800. 1990. Stahlbauten.
DIN 1055-4. 2005. Einwirkungen auf Tragwerke –
Windlasten.

190
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

New studies on Brunelleschi’s Dome in Florence

L. Giorgi & P. Matracchi

Department of Restoration and Conservation of Architectural Heritage, University of Florence, Italy

ABSTRACT: The recent investigations on the Brunelleschi’s Cupola of Florentine Cathedral were devoted
above all to identify masonry structure and to analyse the construction of the two domes (inner and external shells)
in connection with the real building problems found in the works. Exact information on the characteristics of brick
masonry has been obtained. Other important data have been supplied by georadar investigations, demonstrating
that the inner dome is made of two (intrados and extrados) curtains with interposed a different masonry. These
new studies have been put in relation with the surveys of the dome, reaching original conclusions, far from the
common mythicized interpretations of Brunelleschi’s Dome. In the construction a fundamental role was played
by the executive slowness.

1 INTRODUCTION

Twenty years ago a first book on Santa Maria del Fiore,

the Florentine cathedral, was published, as a result of
new investigations started from the facade, the aisles
and the nave (Rocchi 1988). In some way, the con-
structional sequence carried on by the builders was
followed, as the works in Arnolfian period started just
from the facade demolishing part of the naves of Santa
Reparata, the ancient cathedral.
Recently a fourth volume of the series has been pub-
lished and this includes the studies on Brunelleschi’s
dome (Rocchi C.d.Y. 2006). These recent investiga-
tions were related above all to define the masonry
structure and the execution of the two domes (the inner
and the outer ones) in relation with the real problems
found in the works.

2 THE DURATION OF THE DOME BUILDING

PHASES

In the initial segment, including the first corridor, the

dome was built in pietraforte (a local sandstone). In
April 1422, above this level, the dome was made only
of brick masonry and in June 1425 (Saalman 1980) the
macigno (another Florentine sandstone) tie of the sec-
ond corridor was under construction. Presuming that
the faces of these eight sided cloister vault were raised
nearly uniformly, the execution of this first brick dome
segment was completed in about thirty-nine months.
From March 1426 building proceeded above the
second corridor; from June 1429 the macigno ties of
Figure 1. Schematic cross section of the dome indicating,
from archival records, the levels achieved in the different
building periods.
the third corridor started to be supplied and in July
1433 were all be put in place. Given the fact that the
date 1430 has been found scratched on the plaster just
below the floor of this passage, 1430 can be assumed

191
Figure 2. Cast of the brick masonry taken after the removal
from dome intrados of some plaster with Zuccari’s mural
paintings. A black vertical line marks the corner of the dome.
as the year of completion of the second dome segment.
So, including the year 1430, the dome portion between
the second and third corridors was built in fifty-eight
months.
The construction of the upper section, between the
third corridor and the lantern, started in July 1433, was
concluded by March 1435, as the following year the
serraglio (the lantern ring base) was begun. So this
final segment was built in twenty-one months.
It has been possible to evaluate the average raise of
the wall per month in each of the three dome sections,
putting in relation the building period with the height
of the built masonry. For the first section of the dome,
high 5.50 m and built in 39 months, this average value
corresponds to a 14 cm masonry per month; for the
second section (12 meters in 58 months) the value is
nearly 21 cm per month, in the third section (15 meters
in 21 months) 71 cm per month. On this basis, the brick
courses laid each month would result about two, three
and ten respectively in the three sections.
Limiting the working period to ten months per year,
due to weather conditions, the values of dome ele-
vation are nearly the same: respectively 17, 27 and
94 cm masonry per month in the three sections. The
results would not be different from the previous ones.
In the years 1422–1430 the dome works proceeded
very slowly: up to the third corridor the wall was raised
less than a brick course per week; on the contrary, a
relevant increase is recorded in the upper part built
between 1433 and 1435.
It is necessary to consider that above the third cor-
ridor the relevant narrowing if the dome reduces the
built masonry volume to nearly 1/3 in comparison with
the masonry of the second section. However, even con-
sidering such a difference, which caused a consequent
192
Figure 3. The survey of part of dome intrados masonry
points out discontinuities in herring-bone bonds pattern at
the corner. Bricks different in size were used, frequently cut
to fit them to the space left between the herring-bone bonds.
reduction of the employed masons due to the less space
available on the scaffoldings, it is evident the increased
speed in the construction of the upper part of the dome.
It could seem contradictory that works proceeded
quicker just where the brick beds inclination was
greater. However at this level a centring was probably
laid resting on a system of beams inserted in open-
ings correspondent to the small passages opened to the
inside of the cupola. The openings were then partially
walled up to build the present oculi.
3 THE HERRING-BONE APPARATUS
Most of the recorded herring-bone bonds is located
between the second and third corridor, where a very
irregular laying of the bricks and relevant variation of
their inclination are visible.
After the dome intrados survey, the zenith angle
of the herring-bone traces has been measured. It has
been verified a very wide range of the inclination (from
circa 31◦ to 55◦ ) among the eight faces of the dome and
among the herring-bone bonds of the same face, with
the prevalent inclination from 36◦ to 40◦ . It depends on
the different way of laying the bricks, using more or
less quantity of mortar in the joints, as an evidence
of the unequal way of working of the different teams
of masons.
In particular, in the South-East face of the dome
much more inclined herring-bone bonds than else-
where have been recorded. It seems to prove the
Figure 4. The eight faces intrados (by stereophotogrammetric survey) with the traces of herring-bone lines to point out their
different inclination, ranging from 31◦ to 55◦ .
Figure 5. View of dome intrados with the lines of identified
herring-bone traces.
permanence of the same team in the same worksite
area for a long construction period.
The knowledge of the wall apparatus has been deep-
ened thanks to the detachment of plaster parts executed
during the restoration works of the wall paintings.
The detailed survey drawings (Dalla Negra 2004) and
the casts of the masonry prove the irregularity of the
herring-bone bricks, laid with continuous and varying
rotations on the plane of the dome face, using mortar
joints often thick as much as the bricks. The rows of
the vertical bricks are discontinuous: somewhere the
inclined herring-bone row is interrupted by a number
of horizontally disposed bricks, and then it starts again
inclined. The interruption of the herring-bone pattern
is located both in the middle of the dome faces, and
in the corners. In this point there is nearly no wall
toothing: nearly all the bricks, unbroken or fragmented,
arrive just at the corner. In the few cases where a single
brick bridges the faces, it is not a special element but a
ordinary brick cut to adapt to the corner; so there is no
evidence that special V-shaped bricks have been used
193
to build the corners, at least where the plaster has been
detached and the wall apparatus has been visible.
It has to be considered that the stone ties of the third
corridor probably determined some interruption of the
inclined lines of the herring-bone bricks. The same
interruptions should be caused by the correspondent
stone ties of the second corridor; but at this level it was
impossible to record them because of the absence of
wide visible herring-bone bond segments.
Also the extrados of the inner dome has been exam-
ined, at least in many parts with no plaster. The same
characteristics as the intrados have been found: often
there is no continuity in the herring-bone lines, adjoin-
ing bricks have different rotations, and thick mortar
joints are used.
The removal of part of the tiles, made possible to
take a stereo-photogrammetric survey of a meaning-
ful sample of the outer dome extrados. The resulting
information on the herring-bone bonds is comparable
with that previously acquired: different inclinations,
discontinuities, variable thickness of the joints, irreg-
ular brick disposition even in short segments of the
herring-bone bonds.
The information now available on the herring-bone
bonds is insufficient to make a comparison between
the wall apparatus of the intrados and the extrados
of the same dome portion. The available surveys dis-
continuous, limited to only few samples and related
to parts located on different levels. Nevertheless they
appear really meaningful to make some considerations
on the herring-bone structure in the whole of the dome.
It is interesting the wide variation in the horizontal
distance (measured along the brick courses) of two
adjacent herring-bone bonds. At the intrados of the
inner dome this measure is known in all the parts where
the plaster was detached: in the corner between the
East and the North-East faces, at the height of about
15 meters above the inner gallery, these distances are
cm 95, 100 e 110; in the North and the North-West cor-
ner, at the height of 25 meters above the gallery, are

Figure 6. The georadar diagram of the inner dome dis-
plays two symmetrical discontinuities (at the intrados and
the extrados) in the brick masonry, and a different inner core.
Figure 7. The drawing represents the wall structure of the
Cu-pola, basing on georadar investigations. The inner dome
is made by two outer layers where herring-bone bonds are
present, and by an inner core where bricks are probably laid
on simple curved beds (corde blande) (Rocchi C.d.Y. 2006).
cm 45, 60, 90; in the North and the North-East corner,
at the height of 27 meters above the gallery, are cm 45,
50, 60; in the North face, at the height of 29 meters,
the greatest variability has been recorded, as the her-
ring bones are distant cm 10, 15, 25, 30, 45, 50, 55. At
the extrados, above the third corridor, at the height of
the second horizontal arch (circa 26 meters above the
gallery) the distances among the herring-bone bonds
vary from 85 to 110 cm, so they are much greater than
the ones recorded on the intrados at about the same
level.
Also the wall apparatus recorded on the outer
shell has great variability (in the same sub-horizontal
brick course two adjacent herring-bone bonds may be
divided by half brick or many bricks) and discontinuity
in the herring-bone lines (short segments made by few
194
bricks, sudden changes in their inclination, splitting of
herring-bone bonds, . . .).
In the case of herring-bone bond crossing the dome
masonry through its total thickness following a unique
curve centre, very different dimensional data should be
expected: two adjacent herring-bone bonds distant 60
cm on the intrados of the inner dome, on its extrados
the distance should be 65 cm and on the outer dome
respectively 68 and 70 cm, distances only limitedly
greater than the ones recorded on the inner intrados.
The dimensional heterogeneity that has been recorded
in the building is totally different, even if the rele-
vant slowness of the works should have made easy to
control the geometry of the herring-bone lines.
For these reasons it seems improbable that the
herring-bone bonds cross radially the whole of the
dome following a unique alignment to the centre. It is
more reasonable that the intrados and extrados herring-
bone bonds of the inner dome are not continuous, but
belonging to two different masonry layers. The exis-
tence of two distinct layers, the inner and the outer
parts of the wall, has been proved by georadar inves-
tigations, and can also be supported by considerations
relating to the building technique used in the works.
Moreover, it seems proved that the herring-bone bonds
of outer dome are totally independent from those of the
inner dome.
4 THE BRICK MASONRY FACING THE RIBS
AND OCULI PASSAGES
In the space between the two shells, at each corridor
level, very refined brick masonry can be observed on
the walls of both the passages through the ribs and to
the oculi opening inside the cupola. Here the bricks,
homogeneous in size, are laid with regular and very
thin joints and likewise the herring-bone bonds are
very accurate.
Examining in detail the bricks of the passages, some
misleading wall patterns can be detected: some bricks
apparently part of a herring-bone row are illusory, as
a single brick, laid using its largest face outwards, has
been carved in order to simulate the thin sides of four
different bricks. Other special brick elements of the
passages are used to form the vertical corner of the
abutments in masonries laid on inclined beds.
The georadar investigations carried on in the cor-
ridors revealed a difference between the masonry of
the outer part of the wall and of its internal nucleus.
So it can be affirmed that, apart from the herring-bone
bonds, on the wide faces of the dome a nearly usual
masonry (thick joints, irregular bricks, somehow irreg-
ular laying of the elements, . . .) was executed, while
the very refined brick facing was limited to the pas-
sage walls; hence some diffused misunderstandings on
the real wall quality, supposing that the whole dome
Figure 8. The accurate brick pattern visible in the serraglio
at the top of the dome (left) and in the passages through
the ribs (right). Here the flat brick line carved to suggest a
different pattern is pointed out.
Figure 9. Side wall of a passage to an oculus open to the
inner space of the dome. Basing on georadar investigations
(diagrams on the right) the very accurate visible wall appara-
tus is just a facing, uneven and different from the remaining
dome masonries.
brick masonry was represented by that one of the pas-
sages. On the contrary, it seems even possible that the
passage facing was placed during further restoration
works.
Next to the serraglio the situation is the same: the
intrados masonry of the inner dome is very confus-
ing, with irregular and segmented herring-bone bonds,
whilst in the passage at the base of the serraglio the
visible brick masonry is as accurate and refined as the
one just examined in the lower passages. So this kind
of masonry appears once more to be only a facing,
and this is also confirmed by the outer face of the
same serraglio wall, visible at the top of the space
between the two domes: under the plaster a usual thick
joint masonry, the same found everywhere but in the
passages, emerges.
195
Figure 10. Axonometric scheme with the sub-horizontal
canti-lever-arches connecting corner and intermediate ribs.
5 THE ARCH CONNECTIONS BETWEEN
THE RIBS
It is meaningful the superimposition of the detailed
survey of the just described extrados masonry of
the outer shell on the measured drawing of the sub-
horizontal cantilever-arches located on the intrados
between the corner and intermediate ribs.
Above all it has to be underlined the peculiarity
of their masonry structure: in the lower part head-
ers are laid to form a segmental arch thick one brick,
while above this arch some courses of stretchers fol-
low the brick laying of the ribs and of the outer dome.
The bricks of these cantilever-arches, however laid,
are connected to the ribs and the outer shell; those
ones above the segmental arch are laid together with
the bricks of the shell (at least at the lower levels,
as proved by the presence of herring-bone bonds in
some of them). The segmental arches continue beyond
the extrados up to the external surface of the shell,
as documented by the superimposition of the upper
cantilever-arches and the extrados survey. The plaster
absence in the extrados of the outer dome gives the
opportunity to identify nearly the total straight-arch of
a cantilever, and partially a second straight-arch placed
at the same level on the opposite side.
 The reducing cross section of those arches, typi-
cal of cantilevers, and the consequent wider section
in correspondence of the corner ribs, transfers on the
latter ones most of the stresses. Furthermore, it can be
imagined that the thickening of the cantilever-arches
at the corner rib could allow to cross them diagonally
and to make possible the laying of a continuous iron
tie, shaped as a multi-sided polygon and so much more
effective.
6 THE CORDE BLANDE
The corde blande, the curved laying of brick courses,
progressively increasing from the base to the top of
the dome, are well documented by the survey of the
dome extrados. This special laying system was caused
by works needs, as it was necessary to give continuity
to the differently inclined brick beds in the two cor-
ner ribs and in the central part of a single face. The
consequent curved laying of the bricks was got and
made progressive not only with varied thickness of
the mortar joints, but introducing discontinuities in the
Figure 11. Zenith view to the third corridor intrados of the
space between the two domes, showing the cantilever-arches
at the intrados of outer dome.
Figure 12. Detail of a cantilever-arch. The vertical bricks at the intrados (left) form a segmental arch supporting the above
masonry, laid on sub-horizontal curved beds (a corda blanda) homogeneous with the outer dome masonry.
196
herring-bone row and in the sub-horizontal masonry
between them as well as.
A further aspect has to be kept in mind. At the intra-
dos the dome corner, measured from the gallery to the
base of the lantern oculus, is only a little more than
39 meters, while at the middle of a face the corre-
sponding length is 3 meters shorter. In the works, this
different length (nearly 10%) required a greater num-
ber of brick courses in the corners, and this necessarily
caused further irregularities.
During the construction these work adjustments
were made easier by the masonry discontinuity given
by the herring-bone bonds, that caused the included
parts of masonry to be independent the one from the
other. For this reason such adaptations are nowadays
nearly invisible when examining the wall.
The marble mouldings at the external base of the
lantern are apparently curved, and so they seem to
reply the curved line of the corde blande. In fact,
the masonry which supports them is straight and the
curved line of the mouldings is caused by the brick
masonry settling, which was not followed by a cor-
respondent subsidence of the marble covering of the
corner ribs. Here the joints between the elements
are extremely reduced and so subject to only lim-
ited dimensional reductions for plastic deformation or
shrinkage. On the contrary, these phenomena are rele-
vant in all the walls of the dome, built with thick joint
masonry (Petrini 1989).
7 IRON BARS EMPLOY
In the masonry of each aisle vault and of each nave
transverse arch of Santa Maria del Fiore iron ties were
inserted. This can be considered the start point of a new
construction technique (iron reinforced brick vaults)
carried on and increased in the construction of the
Cupola of Florentine cathedral.
Many iron elements are visible at the intrados of the
dome: eyes, different in size and located on the faces
at regular levels, or metal straps at the corners. Up to
now their presence has been put in relation with the
necessity to fasten intrados suspended scaffolding, or
to fix corner wooden centrings (Dalla Negra 1995).
The same works praxis has been proved to be previ-
ously utilized in the construction of aisles and nave of
the church, where iron elements used to fix centerings
have been found in transverse arches and diagonal ribs
of the nave vaults.
Some iron bars inserted in the dome masonry are
visible. They could be part of a more extended tying
system, which probably has played a relevant role in
building static condition.
At the dome spring, just above the inner gallery,
where should be located the first stone tie, many eyed
iron bars stick out of masonry among stone cantilevers.
The bars probably cross nearly the whole of the dome
masonry.
Detailed observation has been made at the level of
the third corridor in the space between the shells at
the North-Eastern side of the dome. Here many iron
elements are visible: a radial bar laid aside a central
rib; two transversal bars inserted at the intrados of
stone ties; a third bar, protruding from the masonry
of the inner dome in coincidence of its set back. In
this last case, the bar position, almost parallel to dome
extrados, indicates that probably this element is the
anchor of an iron tie crossing radially the main shell.
Broader investigations to find out iron presence in
the dome were carried on by metal-detectors by Italian
Army (Compagnia Genio Guastatori della Brigata
Motorizzata “Friuli”). Exterior wall stripes and floors
were examined at the level of the three corridors. Geo-
radar investigations made later on the inner dome
revealed signs, probably related to the presence of
iron bars.
Basing on these investigations, it comes out that
the builders kept in mind the different behaviour of
the two shells, placing reinforcing iron bars mainly
in the corridors. In any case it has been ascer-
tained the conspicuous use of radial iron bars in the
stone masonries at the dome spring. Here the pres-
ence of large stone elements could have required a
widespread use of iron cramps, unnecessary in brick
masonry.
The late introduction of the cantilever-arches, start-
ing from the second corridor, probably suggested also
the usage of almost continuous perimetral iron bars.
This kind of hooping, present in a more recogniz-
able form at this level, seems absent above the third
corridor.

Figure 13. Survey of the extrados summit of the outer North-Eastern shell. The projection on the outer surface of inner ribs
and canti-lever-arches points out the exact coincidence between the intrados arch of the latter and the vertical bricks of the
survey. The herring-bone bonds, though irregular, also go on intermediate ribs.

197

Figure 14. Masonry at the connection of the outer shell,
corner rib and cantilever-arch. Everywhere very thick mortar
joints are visible.
Figure 15. Survey detail of the outer shell at the base of the
lantern. At the centre, the marble moulding bends fitting to
the slight lowering of the dome. At the corners, where the
marble covering of the rib act as a strut, a hollow space (in
black) is formed.
Figure 16. Near the edge between the rib (left) and the extra-
dos of the inner dome (right) a metal bar comes out the set
back. The bar is probably the anchor of an iron tie crossing
the main dome.
198
The use of metallic connections along the rib cor-
ners is probably a consequence of the difficult toothing
of ribs and herring-bone masonry of the dome. Any-
how all that proves, even if in outstanding structures
as the dome and nave of Santa Maria del Fiore, the
widespread usage of iron reinforcements in Medieval
and Early-Renaissance buildings.
REFERENCES
Dalla Negra R. (ed.) 2004. La Cupola di Santa Maria del
Fiore. Città di Castello: Sillabe.
Dalla Negra R. 1995. La cupola del Brunelleschi: il cantiere le
indagini, i rilievi. In Acidini Luchinat C. & Dalla Negra R.
(eds.). La cupola di Santa Maria del Fiore. Il cantiere di
restauro, 1980–1995: 21–22. Roma: Istituto Poligrafico e
Zecca dello Stato.
Giorgi L. & Matracchi P. 2006. Santa Maria del Fiore, fac-
ciata, corpo basilicale, cupola. In Rocchi 2006: 277–324.
Petrini G., 1989. Su alcuni recenti rilievi architettonici per il
restauro dei monumenti. In XY. Dimensioni del disegno,
V, n. 11–12, pp. 147–149.
Rocchi Coopmans de Yoldi, G. (ed.) 2006. S.Maria del Fiore.
Teorie e storie dell’archeologia e del restauro nella città
delle fabbriche arnolfiane. Firenze: Alinea.
Rocchi, G. et al. 1988. S.Maria del Fiore: Rilievi, documenti,
indagini strumentali. Interpretazione. Il corpo basilicale.
Milano: Hoepli.
Saalman, H. 1980. Filippo Brunelleschi. The Cupola of Santa
Maria del Fiore: 263. London: A. Zwemmer.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Wells Cathedral: The crossing and the central tower. A new hypothesis on the
constructional sequence and dating of the central tower

G. De Blasi
Department of Art, Science and Technology (DASTEC), University of Reggio Calabria, Italy

ABSTRACT: The focus of the present study is the crossing and the central tower of Wells Cathedral. An
analysis of the published research show that this portion of the Cathedral has been inadequately treated, and
even the two most recent monographs by Sampson and Rodwell add relatively little about the inception, the
heightening of the central tower, and the insertion of the strengthening of the crossing. The present study seeks
to advance understanding of the monument in two ways: a detailed architectonic survey of the central tower and
in-depth analysis of the biographical sources and documentary evidence.

1 INTRODUCTION

The present research is part of a wider study on Wells

Cathedral carried out by the Department of Architec-
ture and Civil Engineering of the University of Bath
and constitutes a synthesis of the author’s MPhil given
at the University of Bath under the supervision of
Dr D. D’Ayala and Mr. M. Wilson-Jones.
The decision to investigate the crossing and the
central tower of Wells Cathedral in particular arises
from the singular, inspirational quality of the Scissor-
Arches, and an instinctive need to try to explain their
insertion from both a formal and an engineering point
of view. The analysis of the bibliographical sources
highlights the lack of information on the crossing and Figure 1. The nave, the Scissor-Arches and the crossing.
the central tower, in contrast with a vivid debate on the
remodeling of the east end in the first half of the 14th
century. However, an observation by Draper (1981) the presence of the oculi in the triangular spandrels
about the reasons for remodelling the east end, pointed allows visual contact with the east end beyond. In the
me in the right direction. Draper affirms “the setting Salisbury version, by contrast, this visual contact is
out of the eastern extension at Wells was designed to absent.At Wells, consequently, during the solemn litur-
make provision for the shrine of the de Marchia in gical procession, the laity, who had to stay in the nave,
the bay behind the high altar. His tomb was, in fact, could perceive what was going on in the final stages
placed in the south transept and is usually dated shortly of the processions in the east end.
after the bishop’s death in 1302.” The heightening of Linking the heightening of the central tower with
the central tower could have been motivated by the the remodeling campaign of the east end demanded an
desire to create a landmark consonant with a place of in-depth analysis of the debate over this section of
Pilgrimage. However, the fact that the lady Chapel was Wells Cathedral, as well as the execution of a fresh
intended partly as a shrine for the holy body of William architectonic survey in the attempt to establish a
de Marchia may also suggest a liturgical aspect to the chronological sequence. The results of my survey
scissor-arches, for they act as a screen separating the at Wells point to a revised chronology, one which
laity from the east end. For anyone entering from offers a new reconciliation between the physical and
the west end or the north porch the scissor-arches con- the documentary evidence as well as art-historians
stitute the main focus of attention, yet, their shape and considerations.

199

Figure 2. The elevation section of the central tower obtained
from the architectonic survey. It is divided in three main por-
tions, i.e. the first row of arches, the second row of arches,
and the heightening of central tower.
2 CRITICAL READING OF THE SURVEY OF
THE CENTRAL TOWER
Even if the topic of the present research is the cross-
ing and the central tower, the survey regards only the
central tower. For the sake of simplifying the analysis,
the fabric under investigation is divided into three sec-
tion: the first one is delimited inside by the first row of
arches and externally by a masonry course at 2.55 m.
below the roof top; the second one, inside is formed by
the second row of arches, and externally, is bounded
upward by the top of a deep recession of 2 feet and
2 inches (65 cm); the third one, from the end of the
recession to the top of the roof, i.e. the heightening of
the central tower in 1315–22.
Starting from the first section, the external masonry
surface presents signs, which are found also on the cor-
responding internal face, i.e. same masons’ marks and
diagonal tooling on the ashlar. These signs are visible
on both surfaces because the external surface is mainly
under the nave roof. Inside the decorative apparatus
presents vertical tooling. As far as the external west
surface under the nave roof is concerned, it should be
200
Figure 3. The structural section of the first row of arches
(left) and the structural section of the second row of arches.
Figure 4. Structural plan of a portion of the second row of
arches, (left), and structural plan of a portion of the first row
of arches, (right). For the first section is possible to speak of
a mere blind arcade; the second one is a gallery, very similar
to that of the triforium.
pointed out that the presence of an earlier roof level.
In the south-west corner there is a similar moulding
to that one, which frames the nave roof. Moreover, the
extrados of the nave vault presents traces of a lower
level. When it was changed is difficult to establish.
However, a possible date could be at the ‘Interdict
break’, more probably together with the upper part of
the second level of arches at the beginning of the 14th
century.
Both the first and the second section present
an articulation of the internal wall by piers and
arcades. However, the two articulation have different
sequence – the first one present 4 bays, the second
3 bays on each side; the decorative apparatus of the
first section is simpler than that one of the second
one. Nevertheless, the most striking difference is in
the conception of the wall. For the first section, it is
possible to speak of a mere blind arcade; the second
one is a gallery, very similar to that of the triforium.
Figure 5. The second row of arches elevation (left), and
plan (right). It is possible to see the difference in the masonry
blocks between the lower and upper part.
Figure 6. A portion of the plan at third level: on the left the
section at the grids level, on the right the section of the level
above the wooden floor.
They can be read as two different answers to the Anglo-
Norman research of the articulation of the thick wall.
All these considerations highlight the fact that these
two sections belong to two different building phases.
The second section has the same depth of the third
section; however, its concept is different. Here we are
not in presence of two leaves with an empty space
between them. As the connection starts at a height of
6 feet and 4 inches (192.5 cm) and is constructed of
solid masonry, which runs to the top of this level, it
would be better defined as a wall passage.
Moreover, the upper part of the masonry, both exter-
nal and internal, differs in size in respect to the lower
part. Regarding the moulding of the arches, a series of
waves interrupted by fillets is present. Morris (1996)
in his study on the arch mouldings of the relevant fly-
ers at Salisbury Cathedral, finds four types: the double
ogee, the sunk chamfer, the wave and the ‘undulating
moulding’. He dates the wave moulding not before the
beginning of 14th century. If it is correct, than it should
be assumed that the decorative apparatus of this part
was executed during the raising of the 14th-century
portion. Outside, this level corresponds to the reces-
sion above the stringcourse. These considerations lead
to the conclusion that the upper part of this row of
arches belongs to a different and later building phase
with respect to the lower part.
201
Figure 7. Two portion of the plan at the third level: left the
section is about the grids level, right the level above the floor.
At the third level, stair-wells are present in all four
corner piers, while, for the two previous levels, the
staircase is present only at the southeast corner, even if
passages capable of accommodating stairs are present
on all four corners at the level of the first row of arches.
This absence could be explained with the different con-
ception of the two first parts in respect to the third.
The stairs of the third level were thought to be used
for inspection and may be maintenance. Their absence
at the second level is due to a different function. This
level is a sort of internal gallery open towards the cross-
ing. Indeed, it has to be remembered that the present
crossing vault is dated at the second half of the 15th
century. Moreover, the gallery passage at the second
level is similar to the clerestory passage. Probably, its
role was not merely decorative, but it could have had
a liturgical function.
After analysing the moulding in detail on the third
section, it is possible to point out the recurrent presence
of what Morris defines an “interrupted double ogee”
consisting of “two opposed roll-and-fillet mouldings,
approximately a double ogee, separated by a fillet.” In
Morris, the reference to Wells regards the Lady Chapel.
However, the presence of the double interrupted ogee
moulding in the external decorative apparatus of
the central tower (1315–22) creates a precedent for
the Lady Chapel. Moreover, the fact that he dates this
kind of moulding in the Decorated period confirms the
building date of this part of the central tower. Regard-
ing a possible involvement of William Joy in this part
of the building, it could not be excluded, as Morris
shows in his study, due to the recurrent connections
between the Wells, Salisbury and Exeter cathedrals’
workshops.
Speaking of the outer shell of the central tower, Reid
(1973) dates the walling up to c. 1356, when the Abbot
in Glastonbury conceded forty loads of stone from the
quarries at Doulting ad reparacionem magni campanil
ecclesieWells. Harvey (1996) dates the walling up after
the fire of 1439 and relates the forty loads of stone
to the insertion of what he defines as grids. Draper
Figure 8. The third level in section and elevation. The sec-
tion drawing enhances the peculiarity of the structure: two
shells divided by an empty space and connected every 7
blocks by a transversal one. The elevation shows signs of
a precedent roof system.
(1981) judges that the forty loads of stone would have
been used for the insertion of the scissor-arches. Con-
sequently he postpones their construction from 1338
to 1356.
Going back to the walling up, by stylistic considera-
tion the quatrefoil circle superimposed on a trefoil arch
is an element more of the 15th century than the 14th
and surely not datable as early as 1356. Moreover, the
grids and the walling up are strictly connected and this
shows that they have been built contemporaneously.
These observations, derived from the architectonic sur-
vey, together with the example of Exeter Cathedral
and the proved fact of the interconnections between
Exeter and Wells Cathedral, constitute sufficient evi-
dence that the walling-up was executed after the fire
of 1439. Moreover the homogeneity of the walling-up
with the top of the tower further enhances the thesis of
dating the structure after the fire of 1439.
Many considerations suggest that the fire of 1439
involved only the top of the tower. Indeed, there is no
sign of blackening either inside or outside. This con-
firms the hypothesis of Sampson that the fire involved
only the roof. Moreover, at the upper level of the
tower just below the roof there are signs of a previ-
ous roof system different from the present that dates
from the 15th century. In fact there are evident signs
of the walling-up of the holes for the location of the
beams and signs of cutting into the wall fabric where
the present beams are connected with the masonry.
Besides, if the tower had been constructed or consid-
erably repaired after the fire, it would be reasonable
to expect some sign of this intervention (for example
toothing) where the piers meet the walling-up; yet this
is not the case.
At this point it is possible to give a critical reading
of a key fact that emerges from the survey of the four
202
Figure 9. External elevation at the top level. Just below the
arch the walling perfectly course with the piers.
external fronts. On all four external fronts, at the level
of the last arch just below the roof, the walling per-
fectly courses with the adjacent piers. As this area was
subject to a massive restoration during the beginning
of the 20th century, it was not possible to affirm that
the toothing belonged to the repair after the fire. How-
ever, the fact that this area corresponds inside to the
old level of the roofing provides sufficient proof to the
hypothesis that this area was altered after the fire.
Moreover, inside in the portion above the corbel,
the stone blocks present a chromatic variation towards
the red. The high temperature due to the fire could
have caused such variation. Furthermore, as only the
top of the tower was altered after the fire, it is perfectly
congruent that the walling-up was executed after 1439
and consequently that only the top is perfectly toothed
with the pillars. As for the lateral shaft, its insertion
clearly follows the alteration of the fabric after the fire
of 1439. It is supported by the pier but not connected
to it; instead it is partly superimposed over the adja-
cent wave moulding and, stylistically, contrasts with
the simpler mouldings of the piers. If we mentally sup-
press it the sequence become a wave and a fluting bead
moulding. This kind of moulding may be found in the
scissor-arches, (Morris 1996).
The alteration of the upper part of these piers, after
the fire of 1439 is confirmed by the fact that the orig-
inal mouldings of the piers end at the spring of the
arches. From here to the top of the roof, the masonry

Figure 10. External elevation at grids level. The drawing
enhances the richness of the decorative apparatus.
looses its regularity and the blocks of stone are smaller
and less regular than those of the piers. This part is
composed in total by 6 courses of stone blocks. The
first three are higher than the others. Between the third
and fifth course there are the putlog holes. At present
they support no beam. The last course presents a
cavetto moulding. It constitutes a frame at the connec-
tion between the masonry and the roof. Regarding this
element Harvey (1996) gives an explanation, which
perfectly matches with the fact that this part was built
after the fire of 1439. Indeed, he shows that this course
was built with stone blocks of the old parapet, the
design of which is a peculiar characteristic introduced
by Thomas of Witney, (Morris 1996).
The previous considerations relating to the walling
up and the alterations of the roof indicate that the
‘grids’ were added during or after the walling-up. In
fact if they were introduced in 1356 when the window
bays between the piers were totally open, they would
have been visible from outside and would not have any
connection with the double shell wall structure, and
so creating incongruence. Moreover, one of Harvey’s
arguments in favour of their insertion in 1356 is the
fact that they present similar mouldings to those of the
scissor-arches. In fact, the architectonic survey shows
that their mouldings are similar partly to the piers that
bear them, partly to the external piers. Consequently,
taking Harvey’s hypothesis to the extreme conclusion,
they should belong to the building phase of 1315–22,
and that is clearly impossible. Furthermore, the grids
present the ‘interrupted ogee moulding’, which was
already noted in the description of the external piers.
Finally it is necessary to take into consideration
the fact that the decorative apparatus of this part of
203
the tower is more complex outside than inside. Taken
together with the fact that during the external survey
some marks of whitewash were found, this could indi-
cate that the masonry was intended not to be exposed
to weather. The same kind of washes has been found
on the central tower of Salisbury Cathedral, indicating
a relatively common practice of the Middle Ages. It is
uncertain if they belong to the original building or if
they constituted a sort of ‘restoration’work, conducted
after the construction of the tower. However, an expla-
nation of the presence of the whitewash could be that in
this way the central tower became a sort of landmark,
visible from very far. This hypothesis could explain
the greater complexity and richness of the external
decorative apparatus in respect to the internal one. To
sum up it is possible to affirm that the piers belong to
the Decorated Period and the walling up is definitely
Perpendicular.
In conclusion the analysis of the architectonic sur-
vey on the central tower leads to the identification of
the following constructional phases: (1) first row of
arches; (2) second row of arches till the third course
above the annulet; (3) the upper part of the second row
of arches; (4) the heightening of the central tower up to
the present spring line of the lancets; (5) modification
of the crown of the tower, walling-up of the bay and
insertion of the grids.
3 CRITICAL READING OF THE
BIBLIOGRAPHICAL SOURCES
Both Sampson (1998) and Rodwell (2001) place the
construction of the central tower at the end of the third
phase (1200–05), before the ‘Interdict Break’. Accord-
ing to the data from the survey this is right as afar as
the first section is concerned – i.e. from the extrados of
the crossing vault to the top of the first level of arches.
In the preceding section it has already explained that
the first section and the lower part of the second section
present both the wall blocks tooled in diagonally, while
all the shafts moulding have a vertical tooling. As the
diagonally tooling is generally dated before the ‘break’
and the vertical one after, it is very probable that the
decorative apparatus was finished after the ‘break’.
However, these two sections differ also stylistically.
It has already been established that the upper part of
the second level of arches belongs to a different build-
ing phase. Nevertheless, even the first section differs
from the lower part of the second section. The main dif-
ference is constituted by the evident fact that the first
section has four bays per side. Moreover, it has already
pointed out that the nave roof started initially at a lower
level. It is possible that after the ‘break’the elevation of
the central tower was altered and probably not totally
finished. The upper part of the second row of arches
level could have been built at the beginning of the 14th
century, at the start of the heightening of the central
tower. The further discrepancy between the lower part
and upper part of the second row of arches level can
be explained with the damage due to the earthquake in
1248. The construction of the upper part of the second
level of arches increased the total height of the cen-
tral tower. These two facts, i.e. the earthquake and the
increasing of the tower height, can explain the change
of the level of the nave roof, because it belongs to the
same building phase of the upper part of second level
of arches.
To sum up, the first level of arches is dated 1200–
05; the lower part of the second level of arches and
the decorative apparatus of the first level of arches is
dated after the mid-nave break; the upper part of the
second level of arches has to be dated at the beginning
of the 14th century just before the heightening of the
central tower.
After the death of Bishop Jocelin in 1243, work
on the cathedral came to a stop until 1286, when the
construction of the Chapter House begins. Evidently
its construction, finished in 1306, took precedence in
respect of the completion of the tower for practical
reason, and probably funds did not permit both the
constructions to be undertaken at the same time. More-
over, stylistic analysis on the upper part of the second
level of arches dates this part at the beginning of the
14th century, in sympathy with these considerations,
even if all the scholars agree in attributing these first
two sections to its inception in the first decade of the
13th century.
The heightening of the central tower was carried
between 1315–22. Harvey (1996) suggests the name
ofThomas of Witney as a possible master mason for the
heightening of the central tower. The study of Morris
(1991) on the involvement of Thomas of Witney at
Exeter, Winchester and Wells, based on stylistic analy-
sis, gives further evidence in this direction. He isolates
some specific mouldings used by Witney at Exeter
Cathedral, such as the wave, the undulating mould-
ing and the use of the ogee moulding. All these kind of
mouldings are present on the external fronts of the cen-
tral tower. However, Thomas of Witney is more famous
as a structural and architectural specialist, (Morris
1991), and these decorative details could have been
the production of the masons of his workshop.
All the scholars agree that the central tower was
transformed between the 1315 and 1322 and this
caused the sinking of the western piers. Only recently,
Sampson (1998) has pointed out that “settlement was
occurring at the time of the first building, not just in
the 1330 . . .”. In a documentary statement of 19 May
1338 the church is described as inordinately fractured
and greatly deformed, by the Chapter, [R.i., 201 (Cal.
I, 239)].
However, Harvey and Colchester (1981) affirm that
“the defects were not due to the added load of upon
204
masonry of inadequate strength: the four piers were
well built and solid. No deflection – a common fault
at crossing – can be seen opposite the thrusts of the
lower arcades. So the problem had to do nothing to with
crushing and bending. It would seem that, in contrast
to the south-east pier founded on solid rock, the other
three had inadequate foundations and sank to different
depths through the subsoil. To this unequal settlement
the inevitable moving stresses set up by wind-pressure
must have contributed.”
Moreover Sampson (1998), speaking of the east
nave, has pointed out that “Colchester . . . .observed the
way in which the six eastern nave piers on the south
side all lean away from the crossing, whereas those
to the west of the mid-nave break are vertical. In the
triforium, from the third bay west to the break, extra
relieving arches have been built in the rere-arches of
the opening.”
Consequently the insertion of the scissor-arches
was due to the subsidence, the wind-pressure and the
fact that the crossing probably with a spire increased
the instability of this part of the structure. The added
load due to the heightening of the central tower
increased a problem already present, but it was not
the main cause of the threatening collapse. Some kind
of reinforcement was inserted, i.e. “moulded orders
inserted within the arches next to the crossing in the
easternmost arches of the nave and the innermost
arches of each transept on the west side”, (Harvey
1996).
Harvey has noted that these mouldings are similar
to those of the Lady Chapel, and dates the insertion of
these elements just after the completion of the height-
ening of the central tower, i.e. by c. 1325. Moreover,
in another paper on the state of conservation of Wells
cathedral, he and Colchester (1981) note that “between
the vaults and the roof, above each pier, is a project-
ing block of toothed masonry for abutment.” They
explain these elements as the start of temporary stone
flying buttresses, which were never built. Similar fly-
ing buttresses were introduced “at the north-west and
south-west corners of the crossing from the clerestory
to main arcade of nave and transept”, (Harvey and
Colchester 1981). In fact, both the mouldings and the
flying buttresses have to be read as a first attempt to
solve the problem of instability at the crossing, conse-
quently dating them after the statement of 1338. This
means that the scissor-arches were inserted in a sec-
ond time when these first remedies failed to prevent
the collapse of the tower.
Historians agree that the insertion of the scissor-
arches started ca. 1338. Only Draper (1981) raises
some doubts on this date, and analysis of the com-
plex structure of the scissor-arches, indeed makes
it very difficult to agree with the 1338 dating. As
Draper has pointed out hyperbole is frequent in the
medieval documents. Consequently the statement “at
the Chapter meeting of 19 May 1338 as enormiter con-
fracta and confracta et enormiter deformata” cannot
be taken as a certain date. Moreover, it cannot be for-
gotten that before the insertion of the scissor-arches
the flying buttresses named above were introduced.
All these considerations with the addition of the prob-
lem of the Black Death spread all over Europe in
1347, suggest the possible conclusion that the scissor-
arches were built later. Harvey (1996) agrees with
these observations, but the analysis of the mould-
ings convinced him that the scissor-arches could not
have been constructed after the 1340s. However, the
study of Morris (1996) on this topic regarding the
scissor-arches of Salisbury Cathedral shows that this
consideration is not valid. A possible starting date can
be 1356 when the Abbot in Glastonbury conceded
forty loads of stone from the quarries at Doulting “ad
reparacionem magni campanili ecclesieWell”, (Draper
1981). Harvey (1996) has always related this with
the construction of the ‘grids’. The architectonic sur-
vey has shown that these elements have been built
later, together with the walling up of the window bays
of tower. Consequently, Draper (1981) is right when
he relates this document to the construction of the
scissor-arches.
Regarding the scissor-arches it is necessary to
understand their architectonic and structural role, even
if this part was not included in the survey. From the
architectonic point of view, they can be regarded as
flying buttresses. In fact, each scissor-arches consists
of two great pointed arches, which are connected by
the respective vertices. However, it is sufficient to
divide them along a vertical axis instead of a hori-
zontal one. In this way we obtain a sinusoidal curve.
Moreover, if we isolate the crossing piers from their
more complex context, it can be easily understood that
the formal solution could be a sort of flying buttresses
applied from inside instead of the outside. As far as
the formal definition is concerned, there are a lot of
examples in the supports of the windows and openings
of most Gothic buildings of the period, i.e. Glouces-
ter Cathedral, Salisbury Cathedral, Exeter Cathedral,
etc. Moreover, from the 1230s, the use of the scissor
bracing in roof carpentry became common, (Fletcher
1976).
From the structural point of view, the suggestion
that the added weight of the tower was not the main
cause of the distress of the crossing and the conse-
quent insertion of the scissor-arches may none the less
remain valid, as the main factor of instability of the
crossing was due to subsidence and it occurred just
after the building of the crossing and the first east bay
of the nave, (Sampson 1998).
However, it is possible to give another explana-
tion. The architectonic survey executed on the external
and internal fronts of the central tower shows a very
interesting fact. The deviation from the plumb line is
205
minimum 5–6 cm outside and 1–2 cm inside. The dif-
ference between inside and outside could be caused
by the diffuse erosion of the external stone blocks.
This fact can constitute a verification of the theory
already exposed that the distress of the crossing was
present at its inception and not because of the height-
ening of the central tower. The added load of the
central tower increased the instability of the two piers
of the crossing. The fact that the east nave was leaning
was a consequence of the instability of the crossing.
The first attempt after the heightening of the central
tower was the introduction of the flying buttresses at
clerestory level. However, the problem was more com-
plex because the piers needed to be braced in all three
directions, to counteract the thrust toward south and
north-west. The solution was partly already experi-
mented at Salisbury. Morris (1996), in his study on
the scissor-arches of Salisbury Cathedral has proved
that they are earlier than to those of Wells, dating them
c. 1320–30. Consequently they become the prototype
for those of Wells cathedral. Moreover he explains the
apparent discrepancy of the more complex moulding
system of Salisbury strainer arches by differences in
the economic resources of the two cathedrals. How-
ever this cannot be the only explanation. In fact, the
Wells scissor-arches have a more complex architec-
tonic structure. The addition of complex moulding
would be a disturbing element in the whole balance
and harmony of the structure. Indeed the peculiarity
of Wells scissor-arches is constituted by their solution
to a complexity of formal architectonic and structural
problems in a compromised context. The solution was
necessarily different from other similar cases, and it is
its peculiarity that encouraged the necessary exper-
imentation to answer the problem with a unity of
purpose.
4 CONCLUSION
The findings emerged from the critical reading of the
architectonic survey and their comparison with anal-
ysis of the documentary and bibliographical sources
invite a new proposal for dating the constructional
sequence relating to the central tower of Wells Cathe-
dral. To summarize there are 8 building phases of the
tower: Phase 1 – 1200–05, corresponding to the present
first row of arches level, probably without the deco-
rative apparatus; Phase 2 – after the mid-nave break
c. 1219, second level, corresponding to the present
lower part of the second row of arches. In this period
the decorative apparatus was carried out to the third
block above the ring of the annulet shaft; Phase 3 –
beginning of the 14th century, upper part of the sec-
ond row of arches level, to present stringcourse, which
divides it from the grids level; Phase 4 – 1315–22,
heightening of the central tower, with probably a spire;
Phase 5 – after 1338, insertion of the flying buttresses
at the clerestory level; Phase 6 – from 1356 onwards,
insertion of the scissor-arches; Phase 7 – after the fire
of 1439, walling up, insertion of the grids and transfor-
mation of the upper part of the tower and roof; Phase
8 – c. 1475 onwards, construction of the present fan
vault above the crossing by William Smyth.
The solution at Wells has to be read in the context of
the complex transformation of the eastern end, i.e. the
crossing, the central tower, the quire, the presbytery
and the Lady Chapel. The aim behind this transforma-
tion was the attempt to canonize William de Marchia
and transform Wells Cathedral into a place of pilgrim-
age. The construction of the West Front in the 13th
century totally hid from the sight the first stage of the
central tower. Even the building of the second row of
arches did not modify the predominant presence of the
façade. The heightening of the central tower aspired to
make it a landmark, visible from faraway. From this
the possibility that the external fronts were covered
with whitewash assumes an outstanding significance.
However, the canons did not succeeded in obtaining
the canonization of William de Marchia, and the peti-
tion was abandoned after the 1329. Nevertheless, the
building project was carried on with the completion
of the eastern arm. The insertion of the scissor-arches
can be read as part of the project. Draper has pointed
out that “ at Wells in the 13th century, the transept were
shat off from the laity by the pulpitum on the west side
of the crossing because the stalls extended beneath
the crossing and the north transept was used as the
chapter house.” Taking this into account, with the addi-
tional fact that the eastern end was intended to be the
shrine of the de Marchia, the scissor-arches may also
have assumed the role of a screen, which separated the
canons from the laity. They in effect acted as a sort of
visual and physical frame between the two parts of the
cathedral. Moreover, their presence, especially that of
206
the west scissor-arch, counterbalances the horizontal
emphasis of the decorative apparatus of the nave. Thus
they are the perfect answer to a compromised situation,
in which the pre-existent part had to be altered the least,
and the new insertion had to harmonise with the rest.
Despite the contribution of no less than twelve gener-
ations of master masons, Wells Cathedral maintains a
unique and impressive coherence.
REFERENCES
Draper P., (1981) ‘The sequence and the dating of the decorate
work at Wells’, Medieval Art and Architecture at Wells
and Glastonbury, The British Archaeological Association
Conference Transactions, (1978), Leeds, 30–35.
Fletcher J., (1979), ‘Medieval Timberwork at Ely’, Medieval
Art and Architecture at Ely Cathedral, British Archaeo-
logical Association Conference Transaction for the Year
1976, Leeds, 58–70.
Harvey J.H., (1996), ‘The Building of Wells Cathedral, II:
1307–1508’ Wells Cathedral: A History, edited by L.S.
Colchester, Wells, Open Book, 76–101.
Harvey J.H. & Colchester L.S., (1981), ‘Wells Cathe-
dral: Architecture and Conservation’, Transactions of the
Ancient Monuments Society, New Series, Volume 25,
1981, 104–112.
Morris R. K., (1996), ‘The Style and Butressing of Salisbury
Cathedral’, Medieval Art and Architecture at Salisbury
Cathedral, British Archaeological Association Confer-
ence Transactions for the year 1991, Leeds, 46–58.
Morris R. K., (1991), Thomas of Witney at Exeter, Winch-
ester and Wells’, Medieval Art and Architecture at Exeter
Cathedral, British Archaeological Association Confer-
ence Transactions, Leeds, 1991, 57–84.
Reid R.D., (1973), Wells Cathedral, Wells.
Rodwell W., (2001), Wells Cathedral. Excavations and struc-
tural studies, 1978–93, London, English Heritage.
Sampson J., (1998), Wells Cathedral. West Front Con-
struction, Sculpture and Conservation, Stroud, Sutton
Publishing.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

A study of historical test data for better informed assessment of

wrought-iron structures

M. O’Sullivan
University of Manchester, Manchester, UK

T. Swailes
University of Manchester, Manchester, UK
Senior Engineer, Arup, UK

ABSTRACT: Wrought-iron was the dominant structural framing material from 1850 to 1890. With similar
properties to early mild steel, it is more variable, creating uncertainty in the assessment of existing structures
where sampling and testing opportunities are limited. This paper considers the factors that determine how
historical test data may be used to better inform the assessment of wrought-iron in structures. Strength values for
wrought-iron lie between wide limits. Sources of variability include test methods used, the ‘grade’ or ‘quality’
of wrought-iron, and the type of structural element tested or from which samples have been taken. Yield point
became of great importance when the use of elastic stress analysis for beam design became commonplace and
by 1879 Watertown Arsenal in the USA was publishing stress-strain data. Ten years later testing wrought-iron
was a routine part of the education of engineers at universities.

1 INTRODUCTION Wrought-iron is relatively malleable and ductile due

Over the course of the 19th century structural framing
was dominated first by cast iron, then wrought-iron
and finally steel. It was in the period 1850 to 1890 that
wrought-iron reigned supreme. This paper focuses on
material properties and historical test data. The struc-
tural uses of wrought-iron are described in compilation
volumes for the periods 1750–1850 (Sutherland 1997),
and 1850–1900 (Thorne 2000) and also in previous
work by the second author (Swailes 2006, Swailes &
Marsh 1998).
For most of the 19th century engineering quanti-
ties such as yield stress and modulus of elasticity were
not measurable. Furthermore, quality of material var-
ied considerably. The need for better understanding
of the properties of iron became ever more impor-
tant as engineers and architects designed structures of
greater span and complexity. This prompted the spread
of experimentation in many countries.
to low carbon content, typically less than 0.05%. Steel
has carbon content in the range 0.2–1.0%.The slag is
not well bonded to the ferrite and so does not enhance
the strength of the iron. “The amount of slag in wrought
iron can be up to 3 wt% of the total. It is a glassy
substance composed of iron silicate and iron oxide”
(Walker 2002). The thickness of the slag inclusions can
range from microscopic size to 3 mm. They appear as
narrow elongated strands or streaks and are given this
shape by rolling the iron in a particular direction while
the iron is hot. By dividing the metal into strands of fer-
rite the iron can be described as having a macroscopic
grain due to its fibrous appearance. This texture is best
seen when a nicked bar is bent backwards tearing open
the metal (Figure 1).
3 EFFECT OF SLAG INCLUSIONS ON
DUCTILITY AND STRENGTH

Unlike steel, which from the 1860’s onward could be

2 COMPOSITION AND TEXTURE OF
WROUGHT IRON
Wrought-iron is a composite material, not in the
structural sense but in the metallurgical sense. It
is composed of two phases, one being ferrite-iron
with a BCC crystal structure and the other slag.
produced in large quantities, wrought-iron was always
made in small separate batches in the puddling furnace.
In the puddling process, pig iron was converted into
wrought-iron by removing carbon. The manufacture of
wrought-iron was described by Skelton (1924) and in
various works by Gale (1963, 1964, 1969, 1977). The
most obvious difference between wrought-iron and

207

Figure 1. Fibrous texture of wrought-iron revealed by
tearing open a bar in a nick-bend test. (Thorneycroft 1850).
steel is that wrought-iron always contained slag while
steel did not. Iron had to be hot worked or wrought in
order to produce a metal of better quality. Repeated
working caused the slag inclusions to be refined. They
were made smaller, shorter and more evenly dispersed
with a consequent improvement in mechanical prop-
erties. It was shown by experiment that the benefits
of repeated workings reached a peak at about the
6th working and thereafter the strength diminished
with further reworking (Turner 1908). However, it was
not economical to repeat the process beyond the 4th
working. (Gale 1964).
The size, shape and distribution of the slag inclu-
sions in wrought-iron vary greatly. In cases where
there is a high degree of elongation, the ferrite matrix
is divided into columns. (Gordon 1988). In this con-
dition low ductility can be observed in test samples
even though the ferrite itself is quite ductile. Long fer-
rite columns make the metal more prone to internal
rupture. (Gordon 1988).
Excessively long slag inclusions, from insufficient
working of the metal, cause low ductility. Equally bad
are tiny, globular slag inclusions which result from
excessive working of wrought-iron. An iron with such
a microstructure “lacks the fibrous texture, typical
of good wrought-iron, and behaves like a dirty low-
carbon steel; it tends to be brittle and has poor fatigue
properties.” (Jeffrey 1959). That is why wrought-
iron exhibits improved mechanical properties up to
about the 6th working and thereafter the mechanical
208
properties deteriorate with further working. Exces-
sive working of iron may have occurred when rolling
complex shapes.
4 COMPOSITION AND ITS EFFECT ON
MECHANICAL PROPERTIES
The condition of red-shortness is a lack of cohesion
when the metal is red-hot. The metal cracks or crum-
bles when being hot worked. It is caused by an excess of
sulphur and results from the iron not being sufficiently
purified in the puddling furnace (Skelton 1924). Sul-
phur is present in wrought-iron as iron sulphide (FeS)
but it tends to segregate from the ferrite at the grain
boundaries. Because iron sulphide has a low melting
point it causes a lack of cohesion between the grains
when the iron is heated to red-hot (Johnson 1939).
Sections rolled from red-short iron are likely to have
rough edges (Johnson 1939). Cold short iron is the
condition of low ductility at normal temperatures due
to an excess of phosphorus or copper.
“A good wrought-iron would have a maximum sul-
phur content of 0.05% and a maximum phosphorus
content of 0.16%. The manganese content should be
less than 0.1% and silicon content less than 0.2%”
(Jeffrey 1959). Of all the impurity elements phos-
phorus has the most significant effect on mechanical
properties. Elevated phosphorus content causes higher
yield strength and ultimate strength but causes a sharp
fall in ductility and impact resistance (Figure 2).
5 THE EFFECTS OF COLD WORK
It is important that the metal is kept hot particularly
during the final stages of working into a finished
shape so as to avoid strain hardening the metal by cold
rolling. This can sometimes be difficult, particularly
with the rolling of long bars of small diameter, because
the smaller or thinner the section the faster it looses
heat.
Annealing removes the effects of cold work. In a
number of the testing programs discussed in this paper
the samples were annealed before tensile testing. This
was done because the mechanical history of some
of the specimens was unknown. By first annealing
the metal the ‘natural’ strength and ductility could be
determined as opposed to measuring the strength of a
strain hardened sample.
Shearing and punching iron strain hardens the area
around the cut or hole. “M. Barba showed that cut-
ting out a ring 1/8 inch thick round a punched hole, or
annealing the plate, entirely removed the prejudicial
effect of punching” (Unwin 1910). Drilling holes for
rivets did not strain harden the metal but was slower
and more expensive than punching. Kirkaldy found

Figure 2. Effect of phosphorus on Izod impact energy
(Jeffrey 1959).
that punched plates experienced a 50% loss in ductil-
ity compared with plates that were drilled. (Kirkaldy
1876).
6 STRUCTURAL DESIGN DURING THE 19TH
CENTURY
The accuracy of testing machines prior to the 1860’s
was generally not sufficient to determine precisely
elastic limit. At that time engineers based their designs
on the quantity that they could measure with certainty,
which was the ultimate strength or breaking strength
of the material. In other words, in much the same way
as modern day engineers apply a factor of safety to the
characteristic yield strength to obtain a working or safe
design strength, 19th century engineers applied a fac-
tor of safety to the ultimate strength in order to obtain a
safe design strength. Large factors of safety were used
to ensure elastic behaviour of the metal under working
loads.
Regarding the “question of safe working strength
much difference of opinion exists among engineers, the
permanent supporting power of iron being variously
estimated at from four-tenths down to one -tenth of its
breaking strength” (Colburn 1863).
In 1859 the Board of Trade in Britain imposed
a value of 5 ton/in2 (77 N/mm2 ) as the maximum
design stress for members in railway bridges which
209
became the standard value for all structural work.
(Colburn 1863). French engineers worked to a value of
3.8 ton/in2 (59 N/mm2 ) as the maximum design stress
(The Engineer 1863).
These figures were based on tests that did not
directly measure elastic limit. For example, when
David Kirkaldy conducted his investigation into the
properties or wrought-iron and steel in 1858–61 at
the Napier shipbuilding firm, his instruments could
not measure elastic limit. When Napier and Sons dis-
continued Kirkaldy’s testing program he resigned his
position with the firm so that he could devote his entire
time to materials testing. He designed a new form of
testing machine and set up Britain’s first commercial
testing works in 1865 (Smith 1980). Records of his
tests conducted in 1866 show that by this time Kirkaldy
was able to measure elastic limit.
7 THE UNITED STATES BOARD APPOINTED
TO TEST IRON, STEEL AND OTHER METALS
In America the desire for better understanding of the
mechanical properties of iron was probably greater
than anywhere else, because at that time many
American engineers viewed themselves as being too
reliant on foreign experimental work for knowledge
about the mechanical properties of iron and steel
(Pugsley 1944). The U.S. Government created a Board
in 1874 to provide a national facility for testing mate-
rials. (Gordon 1996). Congress allocated $75,000 to
the Board and in June 1875 a contract was made
with Albert H. Emery to design a precision testing
machine (Gibbons 1934). The machine was com-
pleted and installed at Watertown Arsenal, Watertown,
Massachusetts in 1879 (The Engineer 1888).An exten-
sive program of testing began which was reported on an
annual basis. The testing machine at Watertown Arse-
nal, which became known as the ‘United States Testing
Machine’, was a significant achievement as it was one
of the largest and most precise testing machines in the
world. America was now in a position to make signif-
icant contributions to the field of materials testing.
8 MATERIALS TESTING AT M.I.T
As in Britain and Europe the American universities
also set up their own materials testing laboratories. One
in particular was that at the Massachusetts Institute of
Technology. This particular university is of interest in
the present context because of the meticulous records
of tests on wrought-iron that were made there in the
1880’s and 90’s.
The mechanical engineering laboratory at M.I.T
was established in 1883 by Gaetano Lanza. “Lanza was
born in Boston in 1848. His father was an Italian count
Figure 3. United States Testing Machine at Watertown
Arsenal (The Engineer 1890).
who went to America as a teacher of languages. His
mother was from Vermont. Gaetano Lanza received his
education at the University of Virginia where his father
held the chair of professor of languages. Following his
graduation in 1870 Gaetano Lanza became an assistant
instructor of mathematics at the University of Virginia
but resigned in 1872 to take up a position as instructor
at M.I.T. He taught mathematics for a short time, but
soon began to teach mechanics, becoming professor of
theoretical and applied mechanics in 1875” (The Tech
1925).
In 1883 Lanza was put in charge of the department
of mechanical engineering and immediately began a
considerable expansion of the mechanical engineer-
ing laboratories. He was particularly interested in the
testing of full size structural members (Lanza 1912).
At M.I.T. Lanza had a Fairbanks testing machine of
50,000 lbs capacity which besides being used for mak-
ing tensile tests on iron and wire rope, could be adapted
in such a way as to enable full size beams to be tested
for transverse strength and deflection. The allowable
spans were up to 25 feet. (Lanza 1887)
Lanza retired from M.I.T in 1911 but for a number
of years after this he was associated with the Baldwin
Locomotive works in Philadelphia, where his expertise
in full size mechanical testing was required (The Tech
1925). M.I.T.’s test records are now being used in the
present research on wrought-iron.
9 TENSILE STRENGTH OF BAR IRON
The tensile strength of wrought-iron is greater along
the direction of rolling than across it. This is because
210
Figure 4. Stress-strain graph of 7 round bars with diameters
10 mm, 13 mm, 17 mm, 23 mm, 26 mm, 39 mm, and 50 mm
(Watertown Arsenal 1888).
when loaded across the grain the slag filaments run
perpendicular to the load path, and in this direction
they act as voids for the propagation of internal cracks
across the specimen. If loaded along the grain the fer-
rite is more continuous with reduced tendency for the
formation of internal rupture surfaces.
The tensile strength is greater in narrow bars and
thin plates than in thick bars or large forgings. This
is shown in Figure 4 where a series of bars of the
same material but with diameters ranging from 50 mm
down to 10 mm were tested in tension. The stress-strain
graphs are staggered to show more clearly the reduc-
tion in yield strength with increasing diameter. The
thinner bars have greater strength because they expe-
rienced a greater amount of hot rolling which makes
the ferrite grain sizes smaller and also causes greater
cohesion between grains (Johnson 1939).
Because different structural sections undergo dif-
ferent degrees of working it is reasonable to attempt
to identify certain ranges of strength for the differ-
ent principal structural sections, namely bars, plates,
angles and beams. Figures 5 and 6 and Table 1 give the
results of tensile tests on bar iron. In order to identify
possible regional variations the results were divided
according to country of manufacture.
From this data the Scandinavian iron has a lower
elastic limit but higher ductility than the British

Figure 5. Elastic limit and ductility ofAmerican, British and
Scandinavian bar iron. 390 test results. (O’Sullivan 2007).
Figure 6. Elastic limit of American and British bar iron. 355
test results. (O’Sullivan 2007).
or American irons. Scandinavian wrought-iron was
prized for its high quality being tough and eas-
ily worked. This high ductility was due to the
Scandinavian ores being naturally low in phosphorus
but also due to the use of charcoal rather than coke in
the smelting furnace (Fairbairn 1864).
211
Table 1. Summary of tensile test data on British and
American bar iron. Elastic limit values are represented in
Figure 6.
Bar iron. Tested parallel to grain. Number of tests: 355
Elastic Ultimate Elongation
limit strength at failure
N/mm2 N/mm2 %
Range 127–304 278–533 3.7–40.5
Mean 204 353 22
Standard deviation 35 26 8
Figure 7. Cross-piling to form plate-iron. (Hutchinson
1879).
Toughness is the property that should be used in
assessing the quality of wrought-iron. Toughness is
determined by both strength and ductility, and even
though the Scandinavian irons have the lowest strength
they generally have the highest ductility, and so,
can be considered to be of good quality. However,
if attempting to assign a characteristic strength to
British or American wrought-iron it is reasonable to
exclude Scandinavian iron from the statistical analy-
sis. Figure 6 is a histogram of the measured elastic
limits of British and American bar iron only. The data
refers to tests along the grain direction. For compari-
son the yield strength of chemically pure iron has been
included. This is iron with very low impurity content
and no slag.
10 TENSILE STRENGTH OF PLATE IRON
For plate iron, an effective means of equalising the
strength parallel and perpendicular to the direction of
the grain was cross-piling, in which the bars were piled
in alternating directions as shown in Figure 7 before
being rolled into a thin plate.
However, sometimes the plates were formed with
the outer layers in the same direction resulting in
greater strength in that direction (Figure 7). Further-
more the direction of final rolling may have given some
dominance to the grain in that direction. In the results
that follow the grain direction of the plate refers to
the dominant grain direction. It can be seen from the
collection of about 550 tensile tests results along the
Table 2. Summary of test data represented in Figures 8
and 9.

Plate iron. Tested parallel to grain. Number of tests: 550

Yield Ultimate Elongation

strength strength at failure
N/mm2 N/mm2 %

Range 160–363 232–470 1–36

Mean 240 345 15
Standard deviation 32 35 7

Table 3. Summary of test data for plate iron tested per-

pendicular to grain. Yield strength values are represented in
Figure 9.

Plate iron. Tested perpendicular to grain. No. of tests: 115

Yield Ultimate Elongation

strength strength at failure
N/mm2 N/mm2 %

Range 154–298 183–389 0.1–29.2

Mean 208 296 8
Standard deviation 36 39 7
Figure 8. Yield and Ultimate strength of plate iron tested
along grain direction. Vertical pairs of data points are from
the same tensile test. (O’Sullivan 2007).
grain (Table 2) and about 115 tests perpendicular to
the grain (Table 3) that plate iron is on average about
15% stronger in the direction of the dominant grain. It
is because of this that plate girders were constructed
with the grain of the plate along the longitudinal direc-
tion of the girder. From the data the characteristic yield
strength along the grain is 185 N/mm2 which is lower
than the value of 220 N/mm2 given in the Highway
Standard BD21.

11 STRENGTH OF ANGLE, TEE, AND BEAM

IRON

The tensile strength of angle and tee iron is greater

than that of plate iron, which is expected, as it is rolled
from piled iron layers of parallel grain. With regard to
the more complex rolled I-section too few data exist to
make any generalisation on this form. Tests conducted
at M.I.T. on samples cut from an I-beam gave a mean
yield strength of 165 N/mm2 while tests conducted at
UMIST on samples cut form a Belgian I-Beam gave a
mean yield strength of 319 N/mm2 (Kontos 1996).

12 COMPRESSIVE STRENGTH OF WROUGHT

IRON

Tests by Marshall in 1887 and Kirkaldy in 1866 show

that for practical purposes the tensile and compressive
strengths of wrought-iron can be taken as the same. Figure 9. Yield strength of plate iron tested along and across
However, Gordon has proposed that in cases where grain direction. (O’Sullivan 2007).

212
Table 4. Summary of tensile test data on angle and tee iron.
(O’Sullivan 2007).

Angle and tee iron. Tested parallel to grain. No. of tests: 94

Yield Ultimate Elongation

strength strength at failure
N/mm2 N/mm2 %

Range 193–351 301–448 4–37

Mean 245 370 22
Standard deviation 27 30 7

Table 5. Numerical summary of modulus of elasticity data

represented in Figure 10. (O’Sullivan 2007).

Samples tested parallel to grain. Number of tests: 242

Modulus of elasticity kN/mm2

Range 124–253
Mean 197 Figure 10. Modulus of elasticity of bar wrought-iron
Standard deviation 13 (O’Sullivan 2007).

the slag inclusions are excessively elongated test sam-
ples can exhibit lower compressive strengths due to the
ferrite matrix being divided into columns which can
buckle prior to yielding of the ferrite (Gordon 1988).
13 BRITTLE FAILURE OF WROUGHT IRON
Past failures of wrought-iron structural elements
have included suspension bridge chain links. William
Kirkaldy investigated the failure of a tie-bar from
Charing Cross Station roof in 1905 (Swailes 2005).
More recently a beam failure in a building in Leeds
was reported (Bland 1984).
The consequences of failure are more severe for
some structures than others and some structures are
more important. To take account of this in assessment
different factors of safety could be applied to different
types of structure. However, for reasons of simplicity
and lack of sufficient data, use of a single factor of
safety is prevalent.
Fatigue failure of wrought-iron is an area that
deserves further investigation. Cullimore conducted
useful work on fatigue and determined that the fatigue
limit of wrought iron can be taken to be about one-third
of the ultimate tensile strength (Cullimore 1967).
Lack of toughness rather than strength has been
attributed to various failures of structural elements.
Wrought iron from the S.S. Great Britain (Morgan
1996) and Walnut Street Bridge in the U.S. (Green
1999) showed a high ductile-to-brittle transition tem-
perature indicating that wrought-iron is potentially
prone to brittle fracture at normal temperatures. Impact
test data indicates that the toughness of wrought is
quite variable. Charpy values for wrought-iron from
an American truss bridge were in the range 34–144
Joules (Sparks 1998) while Charpy values for mate-
rial from another American bridge were in the range
10–80 Joules. (Green 1999). For a rolled wrought-iron
beam tested at UMIST the Charpy values were quite
low, 10 Joules for the flanges and 23 Joules for the web
(Steude 2000).
14 MODULUS OF ELASTICITY
Tensile tests on various American, British and Nor-
wegian wrought-irons were compiled to produce the
histogram of values for modulus of elasticity shown
in Figure 10. The mean value is 197 kN/mm2 which is
close to the BD21 value of 200 kN/mm2 . Outlying val-
ues are a consequence of experimental measurement
error.
15 CONCLUSIONS
The quality of wrought-iron is highly dependent on
manufacturing practices. Firstly the skill of the pud-
dler dictated composition while the care exercised
in the forge in piling and rolling iron determined
the microstructure of the metal. Wrought-iron was a
batch produced metal, which resulted in considerable
variation in mechanical properties.

213
 Of the various impurity elements carbon and phos-
phorus exhibit the greatest influences by causing
embrittlement if present in excess quantity. Tests have
shown excess phosphorus to result in significant loss
of impact resistance.
High tensile strength should not be used as a mea-
sure of quality. Ductility must also be taken in account
so that together with strength it is toughness that should
be used as the measure of quality.
The UK Highway Standard BD21 quotes 220 N/mm2
as the characteristic yield strength of wrought-iron.
The tests examined in the present research so far indi-
cate the characteristic yield strength of plate iron to be
185 N/mm2 , and characteristic elastic limit of bar iron
to be 154 N/mm2 .
ACKNOWLEDGEMENTS
The authors are grateful to Cass Hayward & Partners
and to Network Rail for sharing the findings of a study
of wrought-iron stresses.
The Watertown Arsenal and M.I.T. test reports were
located by Institution of Civil Engineers Archivist
Carol Morgan.
REFERENCES
Barlow, P. 1867. A treatise on the strength of materials.
London: Lockwood and Co.
Bland, R. 1984. Brittle failure of wrought-iron beams. The
Structural Engineer. Vol.62A. No.10. October 1984. Let-
ter to Verulam. Page 327.
Colburn, Z. 1863. On the relation between the safe load and
the ultimate strength of iron. The Engineer. March 6th
1863. Pages.136–137.
Cullimore, M.S.G. 1988. Fatigue and fracture investigation
carried out on the Clifton Suspension Bridge’, Proc. ICE.
Part 1. 84. (April 1988) pages 309–329.
Cullimore, M.S.G. 1967. The fatigue strength of wrought iron
after weathering in service. The Structural Engineer. Vol.
45. No.5. pages 193–199
Department of Transport. 2001. The assessment of highway
bridges and structures, Departmental standard BD 21/01.
Fairbairn, W. 1864. Iron its history, properties, & processes
of manufacture. Edinburgh: Adam and Charles Black.
Gale, W.K.V. 1977. Historic Industrial Scenes Iron and Steel.
Buxton: Moorland Publishing Company.
Gale, W.K.V. 1969 Iron and Steel. London: Longmans, Green
and Co. Ltd.
Gale, W.K.V. 1963. Wrought iron: A Valediction. Transac-
tions of the Newcomen Society. Vol.36. pages 1–11.
Gale, W.K.V. 1964. The rolling of iron. Transactions of the
Newcomen Society. Vol.37. Pages 35–46.
Gibbons, C.H. 1934. History of testing machines for materi-
als. Transactions of the Newcomen Society, Vol.15, Pages
169–184.
Gordon, R.B. 1988. Strength and Structure of Wrought Iron’,
Archeomaterials Vol. 2, No. 2, Pages 109–137.
Gordon, R.B. 1996. American Iron 1607–1900. Baltimore:
The Johns Hopkins University press.
214
Green, P.S., Connor, R.J., Higgins, C. 1999. Rehabilitation
of a nineteenth century cast and wrought iron bridge.
Proceedings of the 1999 Structures Congress. Pages
259–262.
Higgins, J.H. 1930. Investigation of the effect of phospho-
rus in wrought iron. Contained in Appendix II of the
report of committee A-2 on wrought iron, Proceedings
of the annual meeting of the American Society for Testing
Materials, Part 1, June 23-27, 1930, Pages 159–186.
Hutchinson, E. 1879. Girder-making and the practice of
bridge building in wrought-iron. London: E. & F.N. Spon.
Johnson, J.B. 1939. Johnson’s materials of construction’
Rewritten and revised by M.O. Withey & J. Aston, 8th
Edition, New York: John Wiley & Sons Inc.
Jeffrey, R., Seager, J. & Woodhead, J.H. 1959. The mechani-
cal properties of wrought iron. Safety in Mines Research
Establishment, Research report No. 154, Ministry of
Power.
Kirkaldy, D. 1876. Results of an experimental inquiry into the
relative properties of wrought-iron plates manufactured at
Essen, Rhenish Prussia, and Yorkshire, England. London:
Kirkaldy Testing and Experimenting Works.
Kirkaldy, W. G. 1891. Illustrations of David Kirkaldy’s system
of mechanical testing. London: Sampson Low. (located in
ICE archive)
Kontos, N. 1996. Investigation of wrought iron as a structural
material. MSc Dissertation. UMIST.
Lanza, G. 1887. Notes on the Engineering Laboratories of the
Massachusetts Institute of Technology. Minutes of the Pro-
ceedings of the Institution of Civil Engineers, Part 1. Vol.
91. Session 1887-1888. Pages 347–365. Paper no. 2245.
Lanza, G. 1912. The progress in testing full-size pieces under
practical conditions, together with locomotive testing in
the United States. Journal of the Franklin Institute. Vol
CLXXIV. No.6. December 1912. Pages 607–620.
Marshall, C.A. 1887. Compressive strength of steel and
iron. Transactions of the American Society of Civil Engi-
neers, Vol. XVII, paper no.364, August 1887, New York:
American Society of Civil Engineers. Pages 53–110.
Morgan, J.E. 1996. The wrought-iron of the S.S. Great
Britain. Royal Institute of Naval Architects, confer-
ence proceedings, Historic Ships Design, Restoration &
Maintenance.
M.I.T. 1897. Tests of engineering materials made at the
Massachusetts Institute of Technology, 1887–1897, Com-
piled from Technology Quarterly, Society of Arts, Mas-
sachusetts Institute of Technology. (Produced by M.I.T. in
book form. A copy is located in the ICE archives).
O’Sullivan, M. 2007. First year PhD report. University of
Manchester.
Pugsley, A.G. 1944. The history of structural testing. The
Structural Engineer, Dec 1944. Pages 492–505.
Rawdon, H.S. & Epstein, S. 1924. The nick-bend test for
wrought iron. Technological papers of the Bureau of Stan-
dards No. 252, part of Vol.18. Washington: Dept. of
Commerce, Government printing office.
Skelton, H.J. 1924. Economics of iron and steel. 2nd Edition.
London.
Smith, D. 1980. David Kirkaldy (1820–1997) and Engi-
neering Materials testing. Transactions of the Newcomen
Society, Vol.52. Pages 49–65.
Sparks, S.P. & Badoux, M.E. 1998. Non-destructive eval-
uation of a historic wrought-iron truss bridge in New
Braunfels, Texas. APT Bulletin. Vol. 24, No.1. Pages 5–10.
Steude, T. (2000), ‘The strength of wrought iron and early
mild steel beams’, MSc Dissertation, UMIST.
Sutherland, R.J.M. (ed). 1997. Structural Iron, 1750-1850.
Aldershot: Ashgate Publishing ltd.
Swailes, T. & Marsh, J.O. 1998. Structural appraisal of iron-
framed textile mills. Institution of Civil Engineers Design
and Practice Guide. London: Thomas Telford Ltd.
Swailes, T. & Marsh, J. 2005. Development of long-span iron
roof structures in Britain. Proceedings of the Institution
of Civil Engineers, Structures and Buildings 158. October
2005. Issue SB5. Paper No.13908. Pages 321–339.
Swailes, T. 2006.Scottish Iron Structures. Guide for practi-
tioners. Edinburgh: Historic Scotland.
The Engineer. 1863. The elastic limit of iron. The Engineer,
March 6th 1863, Pages 139–140.
The Engineer. 1888. The Watertown Arsenal Testing
Machine. The Engineer, January 20th 1888, Page 54–55.
The Tech. 1925. Professor-Emeritus Gaetano Lanza. The
Tech. Monday. September 28th. 1925. Cambridge: Mas-
sachusetts Institute of Technology.
215
The Tech. 1928. Obituary of Gaetano Lanza. The Tech. Friday.
March 23rd. 1925. Vol. XLVIII. No.19. Cambridge: Mas-
sachusetts Institute of Technology.
Thorne, R. (ed). 2000. Structural iron and steel 1850–1900.
Aldershot: Ashgate Publishing Ltd.
Thorneycroft, G.B. 1850. On the manufacture of malleable
iron; with the results of experiments on the strength
of railway axles. Proceedings of the Institution of Civil
Engineers. Vol.9. Pages 294–300, Paper No. 828.
Turner, T. 1908. The Metallurgy of Iron. 3rd Edition, Charles
Griffith & Company Ltd. Strand.
Unwin, W.C. 1910. The testing of materials of construction.
3rd Edition. London: Longmans, Green & Co.
Walker, R. 2002. The production, microstructure and proper-
ties of wrought iron. Journal of Chemical Education, Vol.
79, No. 4, April 2002.
Watertown Arsenal. 1883-1893. Report of tests on the
strength of structural material made at the Watertown
Arsenal, Massachusetts. (Reports appeared as annual
volumes).
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Structural interpretation of post-earthquake (19th century) retrofitting

on the Santa Maria degli Angeli Basilica, Assisi, Italy

L. Sorrentino
Structural and Geotechnical Engineering Department, Sapienza University, Rome, Italy

D. Bruccoleri
Architect, Guidonia, Italy

M. Antonini
Architect, Terni, Italy

ABSTRACT: A sequence of ground shakings occurred from October, 1831 to March, 1832 in Central Italy. Due
to the high intensity and the repetition of the events, the Basilica of Santa Maria degli Angeli, Assisi was damaged
severely, suffering the collapse of the vaults of central and left naves. After the first emergency interventions,
engineer- architect Luigi Poletti (1792–1869) was appointed to design the retrofitting and restoration works.
From an architectural point of view the works carried out by Poletti were rather respectful, but from a structural
point of view they introduced interesting innovations. Contrary to the collapsed construction, Poletti rebuilt the
roof over the central nave with an original raised tie beam and iron braced truss, not resting on the below vault,
while the vault had now a tapered transversal section. Poletti’s solution are considered from a structural point of
view, performing numerical simulations by means of limit analysis and finite element models, with reference to
his original drawings, to previous and following earthquake damages, and to contemporary technical literature.

1 INTRODUCTION well. At the intersection of central nave and transept is

On March 15, 1832, after at least six severe shaking
which took place in the previous months, the vaults
over the central and left naves of the Basilica of Santa
Maria degli Angeli, Assisi (Central Italy) were col-
lapsed. The event had a huge impact in Italy, due to
great importance of the church.
Such importance is related to a small chapel
called “La Porziuncola”, where Saint Francis discov-
ered his vocation and founded the Order of Friars
Minor, and with the neighbouring “Cappella del Tran-
sito”, where according to the tradition Saint Francis
passed away.
The Basilica which currently shrines such sig-
nificant testimonies is one of the biggest Catholic
churches. As a matter of fact, it is approximately 126 m
long, 55 m wide and the top of its lantern is about
75 m tall. The building has three naves separated by
four pillars on each side (Figure 1). The central nave is
covered by a barrel vault with a clear span of roughly
18.5 m. On each bay of the lateral naves are five cross
vaults which correspond to five chapels. The transept
and the long choir are covered by barrel vaults as
the “Porziuncola” chapel sheltered by a dome whose
internal diameter is approximately 20 m. A bell tower
is on the right side of the choir, while there is an
open gallery between the body of the building and the
façade.
The construction of the Basilica, without the extant
façade, required almost 120 years due to both its size
and the poverty of the Friars. Although during such
a time span no severe earthquake occurred, thereafter
many earthquakes took place.
2 CONSTRUCTION HISTORY AND SEISMIC
DAMAGES
2.1 Construction history of Santa Maria degli
Angeli
The construction history of the Basilica of Santa
Maria degli Angeli has been dealt with in several
previous works (refer e.g. to di Boveglio & d’Isola
Maggiore 1834, Perilli 1842, Vignoli 1989, and ref-
erences therein). Therefore, only the most significant

217

Figure 1. Plan and longitudinal vertical section of the
Basilica before the 1832 earthquake (di Boveglio & d’Isola
Maggiore 1834, pl. 4).
phases of its construction history are reported here and
neglecting the information about the convent.
After the death of Saint Francis (1226) only small
buildings, a refectory, some dormitories, a chapel, a
choir, an infirmary, were built in the area.
In 1568 Pope Pius V decided to construct an ade-
quate Basilica, because of the importance of the site
and because of the mass of pilgrims already attracted.
The Friars asked the Perugia (Central Italy) architect
Galeazzo Alessi to provide a model for the design of
the new building which he did during the same year.
The following year materials were supplied to the site
and the prominent architect Jacopo Barozzi da Vig-
nola visited the place, probably as design supervisor
on behalf of the Pope. On March 25th the church was
officially founded. Alessi delivered a plan of the build-
ing on March 1570, and died in 1572 without having
the chance to come back again to the building site.
After five years the foundations were completed and
it took twenty years to finish the masonry work of the
three naves and related chapels. In 1593 their wooden
roof was built (p. 111). During the first decade of the
17th century three pillars of the dome were built. It was
in 1622 that the construction of the church restarted
slowly in the choir. In 1637–40 the fourth pillar was
erected, and in between 1644 to 1646 three main arches
under the dome and the choir vault were completed.
218
In 1645–48 the right transept was built, but the left
one was raised in 1650–52. Even though the construc-
tion of the dome started in 1662, just a year later was
stopped by a harsh controversy (both monetary and
technical) between the master builder and the Friars.
The works restarted between 1668 and 1674, when they
were stopped due to an out-of-plumb of the façade.
The damage was ascribed to the too limited thickness
of its upper section, and the void created by the stairs
in the lower one. Between 1674 and 1675 the façade
was strengthened and partially rebuilt. Only after that
the works on the dome started again and were com-
pleted on September 25, 1677. Then, the bell tower
was built in 1679–85, thus completing the church with
the exception of the façade (Bartelli 1989).
2.2 Earthquake damages
The first information about earthquake damages in
the church date to the April 17, 1747 earthquake (VII
MCS felt intensity, Guidoboni et al 2007): a survey
reported cracks in the choir, damages at the naves pil-
lars, out-of-plumb of the façade (Boschi et al 1998,
81 and 86).
On July 27, 1751 another earthquake (VI MCS,
Guidoboni et al 2007) hit the Basilica again in the
choir (Boschi et al 1998, 62). It is possible that is was
after these earthquakes that the lantern was looped by
means of iron ties at three different levels (Menegotto
1993 and 2003).
Generic slight damages are reported for the July 28,
1799 earthquake (Boschi et al 1998, 122), felt with a
VI MCS intensity (Guidoboni et al 2007).
A very severe seismic sequence struck the Basil-
ica between 1831 and 1832 (di Boveglio & d’Isola
Maggiore 1834, ch. 14; Perilli 1842, p. 6–7; Vignoli
1989, 147–153; Boschi et al 1998, 150–151). On Octo-
ber 27 the main door and the small cupola (on top of
the main dome) were damaged. On November 6 such
damages were worsened. The January 13 and 27, 1832
and the March 13 shakes cracked the vaults and the pil-
lars on the naves. Finally, the March 15 event caused
the collapse of the vaults of the main and the left naves,
as well as of the pillars between the two (Figure 2). The
right nave, although not crumbled, was damaged badly,
and probably survived because two of its pillars have
been previously looped with iron and timber. On the
contrary, the dome was not severely stricken.
The first emergency interventions, under the
direction of engineer Antonio Mollari, were the
confinement of the survived pillars, through iron and
timber, and the construction of wooden pyramid for
the protection of the “Porziuncola”.
On August 11, engineer architect Giuseppe Brizi
was asked to advice for permanent interventions.
Among other secondary indications, he suggested to
rebuild the roof on trusses not resting on the vault,

Figure 2. The left and central naves of the Basilica after
the March 15, 1832 shock (di Boveglio & d’Isola Maggiore
1834, pl. 7).
although this would have reduced the drum of the
dome.
On August 22 engineer architect Luigi Poletti
(1792–1869) is first mentioned. Poletti was a pre-
eminent practitioner at that time (Dezzi Bardeschi
et al 1992) and during his professional lifetime was
frequently involved in interventions after catastrophic
events (Reale et al 2004).
Poletti recommended to shore the choir’s vault, the
right nave’s arches, and the three arches between the
naves and the transept.
Between 1832 and 1835 the master builder
appointed to execute his instructions added iron rings
to the dome, repaired two pillars in the right nave and
tear down the upper section of the façade.The interven-
tions on the two pillars were not faultless. According
to Perilli (1842, 5) the original masonry of the pil-
lars was a rubble core one, with a thin external brick
masonry skin, and the master builder did simply repair
them. The description of the construction technique is
very interesting, since it may help to explain the severe
damages observed in the building.
After many discussions in 1836 Poletti was con-
firmed in his role of designer and Mollari was
appointed as supervisor of the execution. Between
March and September of the same year the right nave
was torn down and rebuilt (with solid brick masonry,
Perilli 1842, 8) up to the arches, while the vault was
completed between April and June 1837. Between July
and October 1836 the demolition of the top of the
façade was completed.
The central nave vault was rebuilt in 1838. The
construction of the roof took place in 1839. Finally,
the façade was partially rebuilt between March and
August, 1840.
On February 12, 1854 another earthquake hit the
area (VII felt intensity in Assisi, Guidoboni et al 2007).
219
The Basilica suffered cracks in the choir apse, in the
transept, in the cupola, in the lateral naves, and in
some chapels. The dome and the main nave were only
slightly damaged (Boschi et al 1998, 164). Architect
Giovan Battista Tiberi suggested extensive use of iron
ties (Boschi et al 1998, 165), especially in the lateral
naves and in the chapels. His proposals have been par-
tially carried out, avoiding the visible ties, and locating
them in the attic (refer to pictures in Lunghi & Lunghi
1989, 194–198; and description in Boschi et al 1998,
166).
Shear and spalling cracks occurred in the lantern
above the dome by the April 19, 1984 earthquake
(Boschi et al 1998, 214), whose felt intensity was
VII MCS (Guidoboni et al 2007). Such damages
induced careful provisional and permanent interven-
tions (Menegotto 1993 and 2003), which have lived
through the following seismic event.
The 1997 Umbria-Marche seismic sequence has
also damaged the Basilica: the façade (rebuilt for
architectural reasons during the 20th century) rotated
slightly, cracks were detected on the lateral naves
vaults and at the junction between naves and dome
(refer to: Capalbini et al 1999, where also a complete
geometric survey of the basilica is available).
Although from an architectural point of view, Poletti
restoration was substantially faithful to the 16th cen-
tury design (Perilli 1842, 7; Vignoli 1989, 152), from a
structural point of view he introduced important inno-
vations which will be discussed in the next section. It
is important to stress that the portions reconstructed
under his supervision survived all the subsequent
events with only minor damages (Boschi et al 1998,
151). As will be shown, the strategy pursued by Poletti
was not to recur to special earthquake-resistant solu-
tions. Instead he tried to reduce the vulnerability of the
edifice first of all by means of rule-of-art construction
techniques, such as solid brick masonry instead of rub-
ble core masonry. Secondly he saved weight on vault.
Finally, he reduced the thrust exerted by horizontal
structures, both vault and roof.
3 STRUCTURAL EVALUATION OF THE
INTERVENTIONS
3.1 The tapered vault
Unfortunately there is no graphical evidence about the
construction shape of the pre-1832 vault. However, in
di Boveglio & d’Isola Maggiore (1834, pl. 7; Figure 2)
the mark of vault with the triumphal arch has a constant
thickness. Moreover, Perilli (1842, 6) explicitly states
that the previous vault had a constant thickness.
Poletti’s drawing dated July 19, 1836 shows the
cross section of his new vault (Figure 3). Although
this is of some interest, it has dragged little attention
so far. Poletti designed a tapered vault (i.e. a vault
with non constant thickness). The actual vault geome-
try is slightly different from the one originally drawn
(Lunghi & Lunghi 189, 194–195). The vault has a
crown thickness of 32 cm, and (through three 16 cm
steps) a haunch thickness of about 80 cm. Therefore,
the span (at springers) to crown thickness ratio is about
57.7, the span (at haunches) to crown thickness ratio
is about 45.7, while the crown to haunch thicknesses
ratio equals to 2.5.
It will be shown that such a profile has several
advantages and it is therefore legitimate that Poletti
purposely pursued it in order to reduce the vault’s
thrust, considered among the reasons of the 1832
collapses.
There is no much attention in the historical literature
to tapered arches or vaults. Probably the first refer-
ence is due to architect Andrea Palladio (1508–1580)
in a 1567 appraisal on the new Cathedral of Brescia in
Northern Italy (Puppi 1988, 123–125). Writing about
the dome, Palladio states that the thickness should
grow from the lantern to the drum in order to reduce the
load at mid-span. Such thickness reduction is obtained
by using two spheres, of diverse radii, with their cen-
tres at different heights. According to Huerta (2004,
200–201), Palladio used this method to design all the
domes in his treatise (Palladio 1570), and the same was
done previously by architect Sebastiano Serlio (1475–
1554). Palladio’s appraisal was reported by Zamboni
(1778) and later by Rondelet (1832–1835, Book 9,
Sec. 6, Ch. 4, Notes and Pl. 195), so it is possible that
Poletti was aware of it.
Figure 3. Cross section highlighting the tapered vault, dated July 19, 1836 (BCALPM 1836, detail). Courtesy of: Biblioteca
Civica d’Arte “Luigi Poletti”, Modena, Italy.
220
Architect Carlo Fontana (1638–1714), in is famous
treatise on the Vatican Temple (1694, 361–367) sug-
gested a partially different rule for pointed domes,
using four different centres, at the same height, in order
to get a thickness decreasing from the lantern to the
drum.This rule was based on the observation of several
existing domes and on the exam of previous geometric
rules (Huerta 2004, 272).
The first reference to tapered arches and barrel
vaults is probably due to the French military engineer
Amédée-François Frézier (1682–1773). In his treatise
(1737–1739, 2: 87) he describes a method to draw
arches of non constant thickness based on the first
scientific theories on arches due to La Hire, Parent
and Couplet (for such theories refer e.g. to: Benvenuto
1991, 321–326, 331–336, 338–344). Frézier’s method
is rather similar to that by Serlio and Palladio, since
the intrados and extrados curves are obtained consid-
ering two centres of different height. Once the crown
thickness has been established, their positions are fixed
either by fixing the springer thickness or by lowering
the centre of the intrados curve by a fraction of its
radius. In Frézier’s examples the springer/crown thick-
ness ratio equals 3 and extrados/intrados radius ratio
equals 7/6. The French engineer gave also rules for
vaults bearing heavy loads (e.g. bridges), vaults bear-
ing small loads (e.g. vaults loaded by some wooden
truss), and vaults subjected to self weight. In the
latter case he recommends a thickness at the crown
equal to 1/24 of the span, a thickness which should
be doubled in the 30◦ section close to the haunches
(2: 96–97).
 In 1748 in an unpublished manuscript (Huerta 2004,
358–360) the French engineer Jean Rodolphe Per-
ronet (1708–1794) suggested to design bridges with
a thickness doubling from the crown to the haunches.
In the case of semi-circular arches the crown thick-
ness should be equal to 5/144 of the span plus 1 foot.
A rather similar rule was suggested by the French
Engineer Bernard Forest de Bélidor (1697–1761) in
1750–1753 (2: 445).
Frézier’s and Perronet’s approach was followed by
the French architect Jean Baptiste Rondelet (1734–
1829) in his famous treatise (French edition 1812–
1817, Italian translation 1832–1835). Therefore, he
recommends to design arches of non constant thick-
ness by lowering the centre of the extrados curve (Pl.
27 and Book 9, Sec. 6, Ch. 1, Art. 1), and to differenti-
ate the crown thickness as a function of the span and of
the type of live load (heavy, medium and zero; Book 3,
Sec. 3, Ch. 1). However, also based on some exper-
iments he performed, the thicknesses suggested are
sometimes much smaller. For the case at hand, and
considering an average stone, he suggests a thickness
of 25 cm at the crown which should be doubled at the
haunches, by means of a linear increase. Moreover,
Rondelet clearly states that a tapered vault is much
more convenient than a constant-thickness one. As a
matter of fact, he writes that a semi-circular arch of
constant thickness subjected to its self-weight needs a
thickness equal to 1/17 of the span, while a tapered one
needs a crown thickness which is one fifth of the previ-
ous (ibidem). Furthermore, he declares that a tapered
vault will exert a much lower thrust of the abutments
than a constant thickness one of equal span (Book 9,
Sec. 6, Ch. 1, Art. 2; and ibidem Ch. 3). In this last
chapter Rondelet gives tables for the crown thickness
of semicircular vaults with horizontal extrados, with
solid infill up to half height and thickness constant
or tapered. The crown thickness of the constant one
is one third bigger than the crown thickness of the
variable vault.
Rondelet’s approach had a great fortune, as evi-
denced by other widespread manuals of 19th century
(Breymann 1885, Vol. 1, Ch. 8, Sec. 9 and Pl. 100).
The vaults drawn by Rondelet or Breymann are
meant to have a smooth thickness increase. Such a
result can be obtained if the vault is meant to be
made of cut stones. However, if bricks are used (as
in Santa Maria degli Angeli) it is easier to has discrete
steps in the cross section. As far as it is known to the
authors, such case is rarely considered in the litera-
ture. Körner (1895, 286) recommends to avoid such a
design, since the thrust line will be very close to the
extrados of central (thinnest) segment, while cracks
and bulging will appear on the two sides of the vault
close to the first thickness variation. As Rondelet and
Breymann he endorses an even growth of the cross
section.
221
In order to better understand such favour for
tapered vaults as well as Poletti’s design, his vault
is analysed here by means of the safe theorem of
limit analysis (Heyman 1966). Two lines of thrust
are drawn, minimising either maximum top and bot-
tom stresses or the thrust exerted on the haunches
(Méry 1840). In the second case maximum com-
pression stress was assumed 2.2 MPa (OPCM 2005).
Three different geometries are considered: (1) Poletti’s,
(2) constant thickness t = 0.54 m giving a vault of
same weight, (3) constant thickness 1/3 bigger than
Poletti’s crown thickness (t = 0.43 m), thus following
Rondelet’s abovementioned tables. Two load cases are
taken into account: (1) self-weight only, (2) self-weight
plus a 18.5 kN load applied at mid span over a 0.5 m
length (representing the load of the roof resting on
the vault, whose load-per-unit-area has been estimated
equal to 2.0 kN/m2 ).
The analyses are performed with a computer code,
drawing the aforementioned thrust lines and the asso-
ciated stress fields (Dr. Cesare Tocci, personal com-
munication). Sample outputs are presented in Figure 4
and Figure 5.
In Table 1 is presented the comparison between
Poletti’s vault and the two vaults of constant thickness.
In terms of optimal stress Poletti’s solution grants a
3 geometrical safety coefficient (Heyman 1982, 24)
with a 62.0 kN thrust. The 0.54 m constant thickness
vault gives the same geometrical coefficient with a
much higher, 74.6 kN, thrust.
The 0.43 m constant thickness exerts a little smaller
horizontal force, 62.0 kN, but with a much lower, 2.5
geometrical coefficient.
In terms of minimum thrust, Poletti’s vault gives
again the best value, 53.6 kN, with the 0.43 m vault
granting a similar 54.6 kN. However, such figure is
also due to the lower self-weight. Therefore, the ratio
between horizontal / vertical components is 0.64 in the
Figure 4. Poletti’s vault: optimal thrust line under self
weight.
Figure 5. Poletti’s vault: stress field associated to the
optimal thrust line.
Table 1. Analysis of Poletti’s tapered vault compared to two Table 2. Analysis of Poletti’s tapered vault compared to two
constant thickness vaults. Self weight only considered. constant thickness vaults. Self weight + roof.

Poletti t = 0.54 m t = 0.43 m Poletti t = 0.54 m t = 0.43 m

Minimum Stress Thrust Stress Thrust Stress Thrust Minimum Stress Thrust Stress Thrust Stress Thrust

Geometrical 3.0 1* 3.0 1* 2.5 1* Geometrical 2.24 1* 1.86 1* 1.36 1*

safety safety
coefficient coefficient
Maximum 0.28 2.2 0.32 2.2 0.36 2.2 Maximum 0.57 2.2 0.58 2.2 1.1 2.2
compression compression
stress (MPa) stress (MPa)
Vertical 83.3 83.3 83.2 83.2 67.0 67.0 Vertical 92.5 92.5 92.9 92.9 76.3 76.3
reaction at reaction at
haunch (kN) haunch (kN)
Thrust (kN) 62.0 53.6 74.6 66.4 60.3 54.6 Thrust (kN) 87.6 72.7 95.7 86.0 79.7 76.6

*Conventional value. *Conventional value.

case of Poletti’s vault, while 0.81 in the case of 0.43 m

thickness vault.
The overall performance of Poletti’s solution is thus
much more satisfactory than the other two, hence jus-
tifying both his design and the approval of tapered
arches in the technical literature.
3.2 The roof truss
The design of the roof, which Poletti performed in
September 1838, is a very good example for conflict-
ing architectural and structural. The former must have
inspired the original design characterised by a lower
roof resting on the vault in order to let the dome tower
over the entire edifice. Structural motivations, strongly
backed by observed earthquake damage, suggested to
recur to a truss leaving the vault clear and reducing the
thrust exerted on the walls.
About his final solution, Poletti stated proudly that
“[. . .] the roof truss [has] polygon shape of my own
invention, sound as the triangular, and with the two
achievements of not touching the vault and not raising
too much [. . .]” (Vignoli 1989, 150). Poletti por-
trayed his roof in several drawings (inventories 547,
548, 550, 552, 553 at the Biblioteca Civica d’Arte
“Luigi Poletti”, Modena), but there is only one dated
(BCALPM 1838; Figure 6). Most probably inventories
547 and 553 were previous studies.
According to a document of the following year
quoted by Vignoli (1989, 159) the original design of
the truss was due to Antonio Rutili, a Foligno (Cen-
tral Italy) engineer of the public waters prefecture who
in 1833 strongly recommended Poletti’s retrofitting
for the City Hall of Foligno (ASPSF 1833). However,
Perilli (1842, 10) attributes the design to Poletti.
Poletti’s roof structure consisted of a king post truss
with raised tie beam.As usual in the Italian tradition the
king post does not touch the tie beam. Collar and prin-
cipal rafters are met at their joint by an extended ashlar
strut. Iron braces partially triangulate struts and tie
beam. Poletti must has been aware of the need to pro-
vide a tension-resistant joint between raised tie beam
and principal rafter. Therefore he decided to guarantee
the tensile strength by mean of an iron tie parallel to
the timber one.
Lunghi & Lunghi (1989) provide a detailed cross
section of the naves, reporting the size of both timber
and iron structural elements of the roof and interesting
photos of the roof system. All timber elements (prin-
cipal rafters, raised tie beam, additional struts) have a
square cross section, with a 39 cm side.
The problem of a roof truss built on a vault which
projects itself above the wall top, thus making the use
of a tie beam impossible, has been dealt with at least
since Middle Age (refer e.g. to Courtenay 1985).
Limiting the attention to 19th century, the topic of
a can be found in Rondelet (1832–35, Book 5, Sec. 2,
Ch. 3 and Pl. 108). The French architect shows two
solutions adopted over the vaults of St. Geneviève,
Paris.
The trusses are much more complicated than
Poletti’s truss, since they make use of passing braces,
and do not recur to iron. Probably both these features
were due to the larger availability of timber and highly
qualified carpentry craftsmanship.
Moreover Poletti’s roof is different from previous
and contemporary English – scissors-braced, raised
tie beam plus braces – and American – truss with
inclined braces and without raised tie beam – roofs
(Yeomans 1991). The presence of a raised tie beam
makes Poletti’s roof more similar to the English roofs,
and similar were in both cases the support conditions:
heavy masonry walls.
Breymann (1885, Vol. 2, Ch. 6, Sec. 6b) shows a
roof with a raised tie beam connected, by means of
additional braces, to the foot of the principal rafters.
In a note by the Italian editor L. Mazzocchi such braces

222

Figure 6. Design drawing of the new roof, dated September
10, 1838 (BCALPM 1838, detail). Note the iron braces and
tie. Courtesy of: Biblioteca Civica d’Arte “Luigi Poletti”,
Modena, Italy.
are prolonged to the principal rafters, thus obtaining a
scissors-braced truss.
Poletti’s roof solution seem to have dragged more
attention than the vault one, although such interest has
not always brought to a correct interpretation of its
features.
According to Lunghi & Lunghi (1989) Poletti’s
iron braces are useless, since “assuming on both sides
hinges as restraints for the structure, this is fully com-
pressed and therefore the braces are not reacting”
(p. 193). Although not clearly stated by the authors,
they seem to have assumed fixed restraints. Under such
hypothesis it is true that the inclined iron elements
are compressed. However, if a slight lateral sagging is
assumed then this is not true anymore.
A similar phenomenon has been observed in the
analysis of cross vaults. If they are examined assuming
fixed restraints almost no tension is observed (Alexan-
der et al 1977). However, if the supports can give way,
elements previously compressed will be subjected to
tension (Barthel 1989).
The roof is analysed here with a linear elastic
finite element code (CSI 2004), under self-weight only
(a 1.35 partial safety factor has been considered, for
later code verifications), and assuming an increasing
lateral sagging on both sides. Although the considered
values are still rather small compared to the free span
of the roof (18.45 m), Poletti’s iron braces are subjected
223
to tension if the restraint gives way laterally 2 cm only.
Such a lateral spread is not unrealistic both due to
lack-of-fit problems, masonry local settling under the
truss thrust, deformation of the unrestrained principal
rafters ends. As a matter of fact, modelling such unre-
strained end as a clamped inclined beam subjected to
the whole vertical reaction of the masonry, the lateral
deflection will be approximately 6.5 cm. It is possi-
ble that Poletti expected such a lateral spreading and,
therefore, he added the braces. The presence of the
braces is able to markedly reduce the horizontal thrust
exerted by the truss (Table 3), a problem which Poletti
must have taken account of, since the roof has been
considered a reason of the collapse, and since this
was the same reason of the different French, English
and American solutions previously mentioned. Fur-
thermore, the additional struts remove roughly 80%
of the vertical reaction from the principal rafters, and
thus reduce the risk of their excessive bending as well
as they transfer the load much further down the wall.
It is interesting to note that as the sagging increases
maximum bending moments in principal rafter, raised
tie beam, and strut tend to become more and more sim-
ilar (Table 3). Although he certainly did not perform
explicit calculations, Poletti must have been aware that
the braces will have induced deflections in the raised
tie and in the struts. That is probably the reason why he
gave to the three fir elements the same cross section.
Finally, it is interesting to note that, according to the
Italian standards (DT206 2006; UNI 11035-2: 2003),
the design ultimate moment (capacity) in the case of
permanent loads is equal to 128 kN m. Nonetheless,
it is true that some of the principal rafters and of the
extended ashlar struts had to be reinforced by means
of L-shaped steel profiles (Lunghi & Lunghi 1989,
196–197).
4 CONCLUSIONS
Confronted with the need to retrofit a pre-eminent
building while restoring (not reconstructing) Santa
Maria degli Angeli Basilica (Assisi, Central Italy) after
the severe seismic sequence of 1831–1832, architect
Luigi Poletti did not to recur to special earthquake-
resistant solutions. Instead he tried to reduce the
vulnerability of the edifice first of all by means of
rule-of-art construction techniques, such as solid brick
masonry instead of rubble core masonry. Secondly
he saved weight on the vault of the central nave and
reduced its thrust assuming a tapered cross section. As
has been shown in the paper, by means of numerical
simulations of thrust lines and associated stress fields,
such solution is more effective compared to uniform
thickness ones, having the same weight or complying
with recommendations of the contemporary technical
literature.
 Table 3. Analysis of Poletti’s roof truss under self-weight. (WOB = Without Braces;
WB = With Braces)
Maximum Bending Moment (kN m)
Principal rafter
Sagging Thrust (kN)
cm WOB WB WOB
0 140 – 44.1
2 133 130 45.9
5 121 108 68.9
Poletti tackled also the problem of the design of
a roof not resting on the vault and not concealing
the dome. Therefore, he designed a king post truss
with raised beam, additional extended ashlar struts and
iron braces. As shown in the paper, the problem of a
roof truss built on a vault which projects itself above
the wall top has been dealt with both in actual build-
ings and in the technical literature. Nonetheless, Poletti
solution is original and effective, as proven in the paper
by means of a finite element model. The additional
struts reduce the vertical reaction at the foot of the
principal, thus limiting the bending deflection, while
the braces are effective in case of a lateral spreading
of the structure and in reducing the thrust exerted on
the walls.
More convincing than any numerical simulations,
Poletti’s retrofitted elements survived with minor dam-
ages all the following earthquakes.
REFERENCES
Alexander, K.D., Mark, R. & Abel, J.F. 1977. The Struc-
tural Behavior of Medieval Ribbed Vaulting, Journal of
the Society of Architectural Historians 36(4): 241–251.
ASPSF (Archivio di Stato di Perugia Sezione di Foligno)
1833. Archivio Moderno, Fabbriche, I, Titolo XII, 1001.
Lettera di Antonio Rutili-Gentili, Ingegnere della prefet-
tura delle acque di Foligno, al Gonfaloniere di Foligno:
descrizione e commento del progetto di restauro del
Palazzo Comunale proposto da Luigi Poletti, Foligno 20
Aprile 1833.
Bartelli, C. 1989. La nuova facciata. Il progetto di Dante
Viviani (1913) e gli interventi di Cesare Bazzani (1925–
30). In F.F. Mancini & A. Scotti (eds), La Basilica di Santa
Maria degli Angeli. 1. Storia e architettura: 207–231.
Perugia: Electa Editori Umbri Associati.
Barthel, R. 1989. Analysis and Structural Behaviour of
Masonry Cross Vaults. In C.A. Brebbia (ed.), Struc-
tural Repair and Maintenance of Historical Buildings,
119–127. Basel: Birkhäuser.
BCALPM (Biblioteca Civica d’Arte “Luigi Poletti” Modena)
1836, Inv. 551. Spaccato per traverso della Basilica di
S. Maria degli Angeli presso Asisi coll’indicazione delle
224
Raised tie beam Strut
WB WOB WB WOB WB
– 10.4 – 2.6 –
45.9 10.4 29.3 2.6 13.1
29.1 10.4 71.7 2.6 38.6
riforme di Archi, e Volti da farsi nella ricostruzione della
medesima.
BCALPM 1838, Inv. 549. Disegno della nuova incavallatura
del tetto della navata di mezzo della Madonna degli Angeli
presso Assisi, 10 settembre 1838.
Bélidor, B.F. 1737–1753. Architecture hydraulique, ou L’arte
de conduire, d’elever, et de menager les eaux pour les
differens beisoin de la vie. 2 vols. Paris: chez Charles-
Antoine Jombert.
Benvenuto, E. 1991. An introduction to the history of struc-
tural mechanics. New York: Springer.
Boschi, E., Guidoboni, E., Ferrari, G. & Valensise, G. 1998.
I terremoti dell’Appennino umbro-marchigiano, area sud
orientale dal 99 a.C. al 1984. Bologna: Compositori.
Breymann, G.A. 1885. Trattato generale di costruzioni civili
con cenni speciali intorno alle costruzioni grandiose.Vols.
1–2. Milano: Francesco Vallardi.
Capalbini, P., Ripa di Meana, D. & Salvatici, B. 1999. Basil-
ica patriarcale della “Porziuncola”: progetto di ripristino
statico e funzionale a seguito degli eventi sismici del
26-09-1997 e successivi. Ingegneria Sismica 16(2): 5–16.
Courtenay, L.T. 1985. Where roof meets wall: structural inter-
ventions and hammer-beam antecedents, 1150–1250. In P.
Long (ed.), Science and Technology in Medieval Society,
98–124. New York: New York Academy of Sciences.
CSI (Computers and Structures Inc.) 2004. SAP2000,
rel. 9.03. Berkeley, CA, USA: CSI.
Dezzi Bardeschi, M. (ed.) 1992. Luigi Poletti architetto
(1792–1869). Bologna: Nuova Alfa Editoriale.
di Boveglio, B. & d’Isola Maggiore, M. 1834. Compendio
storico del Perdono di Asisi e della chiesa detta Porziun-
cola: cui aggiungesi un maggiore ragguaglio della Basil-
ica che la circonda, della caduta di questa, e dell’attuale
sua riedificazione. 13. ed. Asisi: dalla tipografia Spariglia.
DT (DocumentoTecnico) 206 2006. Istruzioni per il Progetto,
l’Esecuzione e il Controllo delle Strutture di Legno. Roma:
Consiglio Nazionale delle Ricerche.
Frézier, A. F. 1737–1739. La théorie et la pratique de la coupe
de pierres et des bois pou la construction des voûtes et
autres parties des bâtiments civils et militaires, ou traité de
stéréotomie à l’usage de l’architecture. Strasbourg/Paris:
Charles-Antoine Jombert.
Guidoboni, E., Ferrari, G., Mariotti, D., Comastri, A.,
Tarabusi, G. & Valensise, G. 2007. CFTI4Med, Cata-
logue of Strong Earthquakes in Italy (461 B.C.-1997) and
Mediterranean Area (760 B.C.-1500). Roma-Bologna:
INGV-SGA. http://storing.ingv.it/cfti4med.
Heyman, J. 1966. The stone skeleton. International Journal
of solids and structures 2(2): 248–279.
Heyman, J. 1982. The masonry arch. Chichester: Ellis Hor-
wood Series in Engineering Science.
Huerta, S. 2004. Arcos, bóvedas y cupolas: Geometria y
equlibrio en el cálculo traditional e estructuras de fábrica.
Madrid: Instituto Juan de Herrera.
Körner, C. 1895. Gewölbte Decken. In J. Durm, Hand-
buch der Architektur, Part 3, Tome 2, Ch. 3. Darmstadt:
Bergsträsser.
Lunghi, A. & Lunghi, E. 1989. Il terremoto del 1832 e la
ricostruzione di Luigi Poletti. In F.F. Mancini & A. Scotti
(eds). La Basilica di Santa Maria degli Angeli. 1. Storia
e Architettura, 189–206. Perugia: Electa Editori Umbri
Associati.
Menegotto, M. 1993. Seismic Repair and Upgrading of a
Dome Lantern in Assisi. Structural Engineering Interna-
tional 3(1): 34–36.
Menegotto, M. 2003. Lanterna di Santa Maria degli Angeli,
Assisi, Perugia. In P. Rocchi (ed.), Trattato sul consolida-
mento, C 486–489. Roma: Mancosu.
Méry, E. 1840. Sur l’équilibre des voûtes en berceau. Annales
des Ponts et Chaussées 19(I sem.): 50–70.
OPCM (Ordinanza del Presidente del Consiglio dei Ministri)
n.3431 del 3 maggio 2005. Ulteriori modifiche e inte-
grazioni all’OPCM n.3274 del 20 marzo 2003. Gazzetta
Ufficiale, 10 maggio 2005, n. 107.
Perilli, S. 1842. Relazione storica sul risorgimento della
Basilica degli Angeli presso Asisi. Edizione: 2. ed.
225
emendata ed accresciuta dall’Autore, a cui fa seguito
un’appendice. Roma: P. Aurelj.
Puppi, L. (ed.) 1988.Andrea Palladio: scritti sull’architettura,
1554–1579. Vicenza: N. Pozza.
Reale, C., Scheibel, B.,Vignoli, F., Decanini, L. & Sorrentino,
L. 2004. Il Regolamento edilizio di Norcia del 1860:
fra storia sismica e storia urbanistica. In S. Lagomarsino
(ed.), L’ingegneria sismica in Italia; Proc. 11th Nat. Conf.,
Genova, 25–29 gennaio 2004. Padova: SGE.
Rondelet, G. 1832–1835. Trattato teorico e pratico dell’arte
di edificare. 5 vols. Mantova: Negretti]. (Original French
edition: 1812–1817).
UNI (Ente Nazionale di Unificazione) 11035-2: 2003. Legno
strutturale – Regole per la classificazione a vista secondo
la resistenza e i valori caratteristici per tipi di legname
strutturale italiani.
Vignoli, F. 1989. La fabbrica alessiana e il cantiere di Santa
Maria degli Angeli dal XVI al XIX secolo. In F.F. Mancini
& A. Scotti (eds), La Basilica di Santa Maria degli Angeli.
1. Storia e architettura, 97–159. Perugia: Electa Editori
Umbri Associati.
Yeomans, D.T. 1991. British and American solutions to a
roofing problem. Journal of the Society of Architectural
Historians 50(3): 266–272.
Zamboni, B. 1778. Memorie intorno alle pubbliche fabbriche
più insigni della città di Brescia. Brescia: Vescovi.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Engineering survival and success: The contributions of historic structural

features in the spire and tower of Salisbury Cathedral

Robert S. Cotta
Senior Structural Engineer, Linea 5, inc., Boston, MA, USA
MSc Conservation of Historic Buildings, University of Bath, Bath, UK

ABSTRACT: The 14th century spire of Salisbury Cathedral, when built, achieved unprecedented heights
in stone masonry construction. This accomplishment was made possible through several ingenious structural
systems that used iron and timber systems to supplement the traditional methods of stone masonry construction of
the time. Its survival over 700 years, though, has not been without challenges and today the engineering systems
of the tower and spire provide a superb example of evolved technologies that assure the structural integrity of
the 123 meter high structure. The actual theoretical basis for the contributions of the structural systems in the
Salisbury tower and spire will be explored in this paper with a goal of identifying the specific critical roles
of individual systems based on historic structural behavior and analysis based on twentieth century structural
concepts.

1 INTRODUCTION application of incrementally developed solutions is

Architects and builders have historically stretched the
limits of technical knowledge in the construction of
towers and spires on churches and cathedrals from
medieval times onward to create some of the most
impressive structures in Western Europe. This reach-
ing beyond known limits has not come without cost,
as tower and spire structures have also experienced
their share of recurring problems with a constant con-
cern that such structures may not stand up over time.
Indeed, so many spires have come down that today a
much smaller number of cathedrals still have spires
on their towers than was the case in the past. This is a
significant loss of heritage.
Structural challenges on church towers and spires
have historically been addressed by taking advan-
tage of the best known technology of the time to
attempt to correct these challenges with solutions that
have as long a life as possible. The lessons come
from the history of towers and spires across Europe,
but the history of the Salisbury Cathedral tower and
spire provides a particularly rich example of structural
innovation. The incremental additions to the Salis-
bury tower and spire of multiple buttresses, scissors
arches, strainer arches, metal framework and timber
scaffolding engaged as a capstone weighting sys-
tem are technologies that can transfer to (and have
been used at) other medieval towers and spires. The
critical to the survival of the Salisbury spire and is a
part of the historic tradition associated with the build-
ing. Even through the twentieth century, applications
of state-of-the-art technologies at Salisbury Cathedral
continued this tradition of applying current technology
to more soundly address the inherent structural chal-
lenges in a masonry tower and spire of such height and
delicacy.
In the twenty-first century, the spires and towers of
historic church buildings will require the correction
of structural problems that are minimally intrusive on
the historic fabric of the building and that correct these
problems with solutions that have as long a life as pos-
sible. The success of the tower and spire at Salisbury
Cathedral in surviving for 700 years provides an excel-
lent example of how to apply evolutionary technology
to secure the structural integrity of what is a state-
of-the-art spire structure while still maintaining the
architectural fabric and beauty. Salisbury also provides
an important case study of the need, when preserving
sophisticated structures, to approach the duties of con-
servation pragmatically to find a balance between the
advantages of modern technologies and the conser-
vation goals that strive to minimize intervention. It is
this pragmatism in Salisbury’s history that may indeed
be the key factor in its survival while so many other
medieval spires have disappeared or required complete
rebuilding.

227
2 THE HISTORICAL CONTEXT FOR THE century, there is little or no written documentation
SALISBURY CATHEDRAL SPIRE and the history must be conjectured primarily by an
assessment of the surviving physical built structure.
2.1 The historical problems Salisbury Cathedral, though, is fortunate to have a
comparatively extensive documentation of its history
The spire and tower of Salisbury Cathedral have pre-
since the 17th century that provides insight into the
sented problems for builders, clerks of the works,
conditions noted and the solutions implemented to
architects and engineers through the centuries. These
maintain the structure through the past 350 years. Two
problems appeared from the very start when the deci-
documents in particular are landmarks in the history of
sion was made to construct a stone tower and spire
English building archaeology and conservation. The
of unprecedented height on a supporting structure at
first of these is Sir Christopher Wren’s assessment
the great crossing that from all available evidence
on the state of the Cathedral prepared for Dr. Seth
was never intended to support a tower of any signif-
Ward, Bishop of Salisbury, in 1668. This document
icant height. The consequences of this decision have
is considered to be the first reasonably comprehen-
impacted almost every structural component from the
sive evaluation of a historic medieval ecclesiastical
supporting foundations to the top of the spire. The
building in England. The second document is titled
impacts include the following:
A Series of particular and useful Observations, Made
• In the original cathedral structure, minimal foun- with great Diligence and Care, upon that Admirable
dations under the great crossing piers have visibly Structure, the Cathedral Church of Salisbury, writ-
settled, especially at the south-west pier, resulting ten by Francis Price, the Clerk of the Works for the
in an overall ‘lean’ to the tower and spire to the Cathedral from 1738 to 1753. Not only did Price
southwest. describe the building in words, but he also produced
• The crossing piers have a visible bow, a situation detailed drawings based on careful measurements, pro-
that is believed to have prompted the installation of viding at the time a unique snapshot of the Cathedral
the strainer arches at the great crossing. at that specific point in time.
• Buttressing, much of which appears to have been
added as part of the tower construction, was prob-
ably installed in response to outward thrust from 2.2 The structural systems supporting the base
the added tower weight that caused shifting of tri- of the tower
forium piers, cracking of walls and vaulting, and
The concerns with the foundations and supporting
deflections of transept piers that can still be seen in
pillars at the great crossing of Salisbury Cathedral
the Cathedral today.
have been the topic of extensive discussion, conjecture
• The base of the tower up to just above the main
and concern over the years. Some of the conjecture
roof ridges is a comparatively light arcaded structure
has been based on incomplete evidence or concern
with comparatively thin two-foot thick outer walls
extrapolated from experience at other medieval cathe-
with multiple penetrations. Reinforcing of this level
drals. Until the 1980’s, when Peter Taylor was able
has been implemented on several occasions over the
to make specific measurements of the pier founda-
centuries and includes iron framework ties, infilled
tions at the main crossing, the actual foundation sizes
stair towers and wall openings, and stiffening walls
were unknown quantities that many had presumed
between the arcade columns and the outer walls.
were insufficient. Measurements taken in the 1980’s
• Extended periods of neglect have resulted in failure
by Peter Taylor have been used to evaluate the footing
of iron reinforcing at the base of the spire, result-
situation and using soil mechanics theory developed
ing in stone damage and the need for a reworking
in the twentieth century have for the first time been
or replacement of the iron reinforcing and major
able to provide a theoretical basis for explaining the
restoration of damaged masonry on at least two
marginal adequacy of the footings under these piers.
occasions.
Taylor notes that the historic concern with the delicacy
• The spire itself has suffered frequent damage from
of the situation has perhaps saved the spire and tower
storms and lightning, resulting in cracks and dis-
from disaster, as this current theoretical explanation
placed units in the masonry and, in at least one
highlights the importance of maintaining the confin-
instance, fire damage to the internal timber fram-
ing soils around the foundations, a contribution that
ing. There is also a visible bend twelve meters from
would have been disrupted by any attempt to excavate
the tip of the spire in the southwest direction that is
and enlarge the foundations. (Taylor, 1988, p. 3)
of unknown origin.
Similarly, the bow of the pillars at the great cross-
The structural behavior and conditions behind these ing has been the topic of much discussion over the
problems have been considered by caretakers of the centuries. The bow and associated cracking and dis-
cathedral fabric over the centuries and solutions have placement of arcade piers at the triforium level are
been implemented. For early work prior to the 17th generally believed to have originated when the tower

228
and spire were added to the original tower base. Much
of the engineering that can be found within the main
body of the Cathedral to counteract forces imposed
by the weight of the tower and spire, including scis-
sors arches at the east crossing and 112 buttresses at
the arcade and triforium levels (Price, 1753, p. 35),
were believed to have been added during or shortly
after construction of the tower and spire in the early
14th century. The exception to this are the east-west
strainer arches at the great crossing, which appear
to have been installed approximately 100 years after
the spire addition. The contribution of these strainer
arches, which are supposedly restraining through com-
pression bearing, has been called into question over the
years based on visible gaps between the arches and the
crossing piers. (Price, 1753, p.53) Strengthening mea-
sures have been recommended as recently as the late
twentieth century to stiffen the crossing pillars and
increase the capability to resist buckling forces, but
no specific deficiency has been identified (except the
visible bow). The current approach is to monitor the
situation to ascertain if the bow is a manifestation of
an active condition, or an adaptation of the building
structure to stresses from loads imposed in the past
that have found a steady state of equilibrium.
3 ENGINEERED SYSTEMS IN THE TOWER
AND SPIRE
3.1 The historic importance of the tower and spire
structural systems at salisbury
It is the systems employed in the tower and spire of
Salisbury that demonstrate concerted efforts to address
identifiable structural behavior and that have a his-
tory of upgrade and improvement over the centuries.
This is in part due to the fact that from the beginning
the structure of the tower and spire was more than
just a structural system of stone masonry acting in
compression bearing to transfer loads to the ground.
Salisbury Cathedral, partially due to the inadequacy
of the base of the tower for the support of such a tall
tower and spire, was unique among medieval cathe-
drals in making extensive use of an elaborate iron
framework in conjunction with stone masonry as a
permanent structural system. Most ironwork used in
medieval buildings elsewhere comprised simple short
bar cramps between stone blocks or tie rods across
arches, for which the primary purpose was believed to
be to stabilize the masonry during construction until
the lime mortar had set. The one other known cathedral
of comparable vintage with extensive permanent iron-
work structures was the Freiburg Münster in Germany,
where embedded iron rings were used to achieve the
open tracery spire that was a model for numerous later
continental spires.
229
Iron, though, does corrode and when exposed to
weathering conditions will require significant restora-
tion or replacement on more frequent intervals than
stone and such replacements and restorations pro-
vide opportunities for utilizing improved technologies.
Iron-based technologies have also evolved over the
centuries, and thus have also been favored for added
retrofits to address deficiencies that develop over the
years.
3.2 The metal tension framework and gallery
infill in the original lantern tower
The best evidence that a conscientious engineering
effort was involved in the raising of the tower and spire
at Salisbury is the elaborate medieval wrought iron
framework that is located in the original lantern tower
of the Cathedral. The medieval ironwork in the lantern
was later supplemented by diagonal ties installed under
George Gilbert Scott and his engineer, F. W. Shields in
the 19th century.
The original lantern at the base of the tower – the
portion from the eaves of the main roofs up to the
clear horizontal parapet band – has been identified by
Tatton-Brown and other sources as clearly not designed
for the superimposed load of a soaring tower and spire.
(Tatton-Brown, 1991, p. 328) Originally, the lantern
was believed to have been designed to support only a
short timber spire or pyramidal roof and was intended
to be open to the space below with large wall openings
to bring exterior light into the center of the building.
The load of the tower and spire carried by the
lantern, estimated to be 6000 tons at this level, exerts an
imposed stress of 1.724 N/mm2 on the approximated
effective bearing area (Reeves, Simpson & Spencer,
1992, p. 381), which is well within the minimum
16.5 N/mm2 compressive strength of the Chilmark
limestone of which it is built. What the iron tie system
at this level does accomplish, though, is to effec-
tively reduce the unbraced length of the wall and
corner column elements, limiting the unbraced length
for outward buckling at these corner stair towers to
the height of the arcade, and keeping the theoreti-
cal unbraced length to thickness ratios within limits
considered acceptable by twentieth century code stan-
dards. Inward buckling of the stair towers is resisted
through bearing of the masonry wall construction
on the four sides, while outward buckling forces are
resisted by tension in the iron framework (Figure 1,
Force Fm) that transfers the load into compression
bearing against the depth of the walls at the adjacent
corners. The lantern is thus reinforced, for purposes
of buckling, so that the unbraced length is limited
to the height of the arcade. The intent of this frame-
work appears to be valid and for over five hundred
years performed admirably in providing the confining
hoop stress to allow the thin (by medieval standards)

Figure 1. Iron tie configuration and force distribution at
lantern level.
lantern structure at the base of the tower to support the
considerable weight above.
The original medieval ironwork although intact on
the interior of the tower, appeared to have either suc-
cumbed to corrosion in places or was weakened in
other ways when George Gilbert Scott conducted an
assessment of the Cathedral in 1862. (Cocke & Kidson,
1993, p. 29) The diagonal cross ties installed by Scott
and Shields resist the same outward buckling action
in a more direct manner (refer to Figure 1, force Fss)
and also further reduce the unbraced length by provid-
ing two sets of ties mid-height between the medieval
framework at the top and bottom of the lantern, as well
as a set between the medieval frames at the top of the
lantern. These braces are each of square cross section
at the interior of the lantern, dividing at corner piers
of the tower to pass through the wall and connect to
anchor plates at the exterior face of the corner diagonal
buttresses. (Reeves, Simpson & Spencer, 1992, p. 404)
The segmental nature of the framework and the
relatively thin cross section of the members has possi-
bly been a positive factor in allowing the iron system
to accommodate the differential expansion between
stone and iron under thermal fluctuations. An embed-
ded design within the stone would likely result in
fracturing and spalling of the stone due to thermal
stresses caused by expansion of the iron. By bringing
230
the framework into the interior of the tower as an
exposed framework, and any thermal growth can be
taken by elastic buckling and/or the play in the hook
and eye connections between segments without impos-
ing excessive stress on the stone. The amount of iron
in contact with the stone (at the corners) is dimension-
ally small, thus limiting thermal loads that may impose
tensile stresses on the stone.
3.3 Tension rings and transfer structure at
the base of the spire
The spire at Salisbury sits on the square crossing
tower with the octagon faces in the primary direc-
tions (north-south-east-west) supported directly on
the tower walls. The diagonally oriented faces are
supported by ‘squinch arches’ that bridge diagonally
across the corners of the tower to transfer the load from
these faces to the tower walls. This arch introduces a
thrust perpendicular to the plane of the four tower walls
at approximately the third points of each wall. Price
observes that consequently there is no ‘buttment’ for
the [resulting thrust from] these arches except for that
provided by the ‘iron bandages’, which are provided
at the base of the octagonal spire. (Price, 1753, p. 38)
Unlike the retrofit nature of the iron framework in
the lantern, the two systems of iron framework at the
top of the tower and at the base of the spire are spe-
cific structural systems to contain the outward thrust
resulting from the dead load weight transfer on the
sloped sides of the spire. The system at the top of
the tower was completely removed in 1967–68 and
replaced by a Peter Taylor designed system of stainless
steel bars with adjustable turnbuckles at the squinch
arches that are encased in epoxy resin that is integrated
into a continuous concrete ring beam in the gutter at
the base of the spire. (Tatton-Brown, 1991, p. 342)
Structurally, the original system (and its replacement)
functions in the same manner as the lantern system
described previously, although fully embedded in the
masonry. The iron band at the base of the spire con-
sisted of a single ring of iron bars joined using pegged
scarf joint connections. This band is located external
to the stonework, but was protected by being set in a
recess that was filled with lead. This band effectively
overcomes the need to relay on buttresses to resist the
outward thrust (Th) from the transferred dead load of
the diagonal sides of the spire by providing the ten-
sion bar to tie the arch and cancel the net effect of
forces (Fr) from each side of the arch. (Fig. 2) With
the great height of the Salisbury spire, this alternative
approach is probably the only option, as the height of
the tower dimensionally limits the amount of thrust
that can be resisted by the dead weight of the tower
walls and pinnacles at the corners.
It is also worth noting that the proportions and
geometry of the Salisbury tower, with its straight

Figure 2. Iron tie arrangement and force diagram at squinch
arches. (Background Drawing Source: Price, 1753, Plate 8).
vertical sides, are contrary to the conventional stone
masonry theory that requires increasingly deep but-
tresses toward the base to contain the lines of structural
thrust from above – a convention that can be seen in
most continental European stone spires that results
in the tapered pyramidal configuration of Cologne,
St. Stephen’s Vienna and others. The Freiburg Münster,
with a slim upper profile and a pyramidal base struc-
ture was the prototype for the Germanic spire towers
that came later. Unlike Salisbury, the octagonal plan of
the spire at Freiburg is carried down through the middle
portion of the tower, avoiding the concentrated thrust
associated with Salisbury’s squinch arches and effect-
ing the transition of octagon to square plan at the top of
the main body of the Münster, where buttressing can be
blended into the wider profile of the church building.
There are, however, iron rings encased in the joints of
the stone masonry at the top of the tower and the base
of the spire at Freiburg. These provide a mechanism
for the redistribution of unbalanced forces around the
slightly uneven plan geometry of the built tower (Fritz,
1926, pp. 50 & 59) and also provide a restraint point to
resist outward buckling action on the corner elements
of the octagonal support tower that rise up individually
from the floor of the open Oktogonhalle.
231
It is the ‘iron bandage’ at the base of the Salis-
bury spire that has been most subject to the pitfalls
of using iron in exposed conditions. Price reported
that the band at the base of the spire, even though
protected in a recess covered with lead sheet, had
been swollen due to corrosion, lost material and splin-
tered the stone. (Price, 1753, p. 38) Price added an
iron band at the base of the spire in approximately
1740 to supplement the original medieval band, both
of which were replaced in 1968 when the stainless
steel band and concrete ring beam system. (Tatton-
Brown, 1991, p. 342) The medieval iron band, a system
that is essential to the structural integrity of the spire,
did however function adequately for approximately
400 years and Price’s additional ‘bandage’ provided
another 200 years of service – an impressive dura-
tion for the pre-industrial wrought iron encased in lead
system.
3.4 External rings on the spire
The upper iron bands around the perimeter of the spire
at Salisbury were initially a recommendation of Wren’s
as a reinforcement to address concerns with the per-
ceptible bend of the spire towards the top. This bend to
the southwest is in addition to the out-of-plumb align-
ment that Wren documented in his 1668 report. Wren
specifically recommended this ‘bandage’ at the loca-
tion of the bend (corresponding to the fourth and fifth
rings from the top on the spire today), but also rec-
ommended that additional bands be added to provide
greater ‘security’. (Wren, 1668, p. 24) Wren’s imme-
diate recommendation appears to have been promptly
implemented (c. 1670), but the additional ‘security’
rings recommendation was not forgotten and the rings
below were originally installed in the late 17th or
early 18th centuries. 230 years later, the iron straps
were replaced with copper by R. Beaumont. (Tatton-
Brown, 1991, p. 342) The longevity of these iron
bands, in a fully exposed condition, similar to the
iron band system at the base of the spire, is better
than can be expected for the material and probably
can be attributed to the low level of impurities in the
pre-industrial iron that was used.
Wren’s purposes appear to be twofold. His concern
with the bend in the spire appears to be with regard to
this bend as a point of structural weakness and thus a
point of failure in future storm events.The ‘upright bar-
res’ on each octagon panel serve to anchor the portion
of the spire above the bend to the portion below, with
the top of the lower portion well-secured by the exte-
rior hoop that compresses the stone shell at this anchor
point. The additional hoops appear to be in response
to Wren’s concern with settlement and observations
of ‘downward cracks’. (Wren, 1668, p. 24) Although
Wren’s immediate recommendation for this decay was
to replace with new stone and remortar, the pattern

Figure 3. Internal tie system to the spire, immediately above
the gabletted doors and openings. (Source: Taylor, 1988, p. 8,
Figure 3.)
of ‘downward cracks’ also raised a concern of split-
ting that was most likely caused by outward buckling
and would most effectively be arrested with regular
tension hoops down the length of the spire. The fact
that there appears not to have been any tension rings
on the Salisbury spire in the first 300 years of its
life can be considered as validation of the idealized
conical spire developed by Heyman. (Heyman, 1995,
pp. 127–29) It is likely, though that due to local stone
deterioration and/or weakness, construction irregular-
ities or lightning damage, that the integrity of the spire
was compromised and outward buckling action could
develop in thin slender elements created by vertical
cracks in the shell.
Outward buckling was no doubt also intuitively
noted in the Freiburg Münster spire in its conception,
where the ribs can more readily be visualized as slen-
der column elements. The Freiburg spire ribs do have
a visible outward bow that appears to exemplify this
behavior and thus to justify the need for the horizon-
tal iron bands that are embedded at regular intervals
along the height of the spire. The forces that the rings
are capable of resisting are quite small, but Fritz notes
that the small forces that are developed are due to the
‘swollen’ profile of the ribs and the irregular plan base
on which the spire was built. (Fritz, 1926, p. 59)
A more recent 20th century addition to the series of
rings along the height of the Salisbury spire is Peter
Taylor’s internal tie ring, located at the level immedi-
ately above the door and light openings in the spire
that is designed to consolidate the octagon structure
adjacent to the zone of weakness introduced by the
openings. The system (Fig. 3) is economical, simple
and largely protected from the elements. It consists of
twin high-tensile steel tendons that pass around steel
saddles at the inner side of each of the octagon cor-
ners. These saddles are then anchored by tensioned
232
bolts that pass through the saddles and the spire shell
and are anchored by bearing plates at the exterior of the
spire corners. (Taylor, 1988, p. 8) The system has the
advantage of minimizing the exterior exposed compo-
nents (which can be of high grade corrosion resistant
materials) and placing the primary structural compo-
nent in the protected interior, where the material can
be selected for its superior structural properties (i.e. a
very high Young’s modulus value).
3.5 The internal timber frame
Perhaps the most intriguing and ingenious aspect of
the structure at Salisbury is the timber frame scaffold
inside the spire. This timber frame was archaeo-
logically evaluated and documented during the late
twentieth century spire restoration. (Tatton-Brown,
1996) The scaffold is only occasionally in contact with
the interior face of the stone shell and thus does not
function as a supporting framework for the stone shell.
The history of this framework is unclear. Michael
Drury, current CathedralArchitect at Salisbury, reports
that recent dendrochronological testing appears to sup-
port an original date of around 1400, or the time of the
first significant repairs following a possible lightning
strike. This date, if accurate, would undermine previ-
ous theories that the framework was part of the original
scaffolding and would associate the framework with
scaffolding for the 15th century repairs.This may make
more plausible the possibility of forethought given to
the dual purpose of the scaffolding serving as dead
weight anchoring of the capstone, as the total system
may have been designed as a response to the damage
experienced on a spire that had been a ‘first ‘ of its
kind.
The need for an internal scaffold during the origi-
nal construction is not a given, since exterior scaffolds
were known to be typically used in medieval construc-
tion to build towers and spires and an exterior scaffold
would have been required in any case to complete the
top of the spire. Working from within the existing spire
to access damaged portions of the spire in the 14th cen-
tury may well have been more feasible than erecting
scaffolding from the ground up.
What appears to have been the purpose of inten-
tionally connecting the framework to a rod through the
capstone was to provide a dead weight load at the top of
the spire that would effectively pre-compress the stone
shell to counteract tensile bending stresses induced by
wind. Calculations of the stresses on the shell of the
spire using approximated wind loads developed from
current British Standards have found the difference
in stresses with and without the weight of the tim-
ber framework represents a significant increase in the
margin of safety provided for the upper portions of
the spire as a result of this superimposed load. These
approximate calculations indicate that the weight of
the timber scaffolding increases the factor of safety at
12 meters from the top from 1.57 to 2.29, values that
represent a modification of the factor of safety from
a marginally acceptable value (∼1.5) to a value that
would be considered good design practice (>2.0) by
current standards.
In 1738, Francis Price reported that the central piece
of the timber scaffolding had become disconnected
from the capstone approximately at the level of the
weather door. This defect was immediately repaired,
but Price surmised that this defect may have con-
tributed to the upper bend in the spire that was of
such concern to Wren in his 1668 report. (Price, 1753,
p. 59) Calculations show that the spire could accom-
modate design wind conditions without the benefit of
the additional ‘prestress’ provided by the dead weight
of the scaffold. However, there is no way of know-
ing precisely the nature of the wind loads experienced
by the spire at 123 meters above ground, an elevation
where wind speed, gusts and other factors can differ
radically from the conditions at ground level. Even
over a limited time period exposed to less than maxi-
mum wind load conditions, fatigue behavior can cause
deterioration of all structural systems, including stone.
A marginally acceptable structure will be more sus-
ceptible to crack propagation at minor defects under
repeated cyclical loadings. The ‘bend’ is noted to be to
the southwest by Price and Wren, and this would be the
side typically subject to tensile stress by the prevailing
wind direction. A temporary disconnection may have
begun this process of fatigue deterioration by allowing
the incremental opening and closing of fine cracks on
the southwest side, causing deterioration of the stone,
which was fortunately arrested by the reinstatement of
the weighted capstone system.
Less apparent is the damping contribution of this
framework that would work to temper resonant vibra-
tions of the spire in gusty high wind conditions –
a known phenomenon in tall thin structures. Wren
designed a comparable pendulum structure suspended
from the capstone in the spire in Chichester in 1721.
It is possible that Wren was inspired by what he
saw inside the Salisbury spire during his review of
the fabric in 1668. (Tatton-Brown, 1991, p. 345.) If
intended, Wren’s engineering basis for the damping
contribution of this suspended framework was at best
instinctual. Given what is known of structural behavior
under gusty wind load conditions from 20th century
research, the damping contribution of the pendulum
could be as considerable as the dead load prestress in
reducing the magnitude of the wind-induced bending
stresses and associated long term fatigue-like tensile
stresses experienced by the spire during high-wind
storm events.
There is one final system that was installed as part
of the major restoration of the Salisbury spire in the
late 1980’s/early 1990’s. Peter Taylor designed a series
233
of steel tube frames that fit against the interior face of
the spire at several elevations. These frames serve the
purpose of stiffening the spire and providing a second
structural system to temporarily carry loads of the spire
above to allow stone replacement during restorations.
In a sense, they are a permanent installation of ‘tempo-
rary shoring’. Taylor specifically designed the frames
with a geometry provided by the slope of each face
that prevents the structure from becoming a mecha-
nism. The frame is redundant structurally and contains
an ‘in-built limit to the amount of movement of which
it is capable, thus resisting any irresponsible attempt
to ‘lift’ the spire.’ (Taylor, 1988, pp. 8–9) In keeping
with appropriate principles of historic building conser-
vation, the frame installations are reversible and are of
a technology that reflects the date of installation.
4 CONCLUSIONS REGARDING THE
EFFECTIVENESS OF SALISBURY’S
STRUCTURAL SYSTEMS
The structural systems for the tower and spire at Salis-
bury Cathedral fall into two distinct groups. The first
group is the supporting base from the foundations up
to the original lantern level on which the 85 meters
of tower and spire were constructed. The problems
experienced through this zone have primarily been
attributable to the fact that the original construction
at and over the crossing was not intended to carry
such a great mass of stone. The structural systems in
this area are essentially retrofit solutions of buttress-
ing at the base to resist outward thrust at the crossing
pillars and iron ties in the lantern to resist outward
buckling from the high levels of compression stress
on the comparatively thin supporting sections. These
systems appear to be effective and necessary solutions
to resist the thrusts and structural actions resulting
from this large load. Less conclusively addressed,
though, have been concerns about the bowing and set-
tlement of the pillars. Numerous attempts to assess
the conditions at the foundations and at the pillars
have developed a tenuous explanation for the structural
behavior based on advanced soil mechanics theory and
that the supporting base for the spire has found a stable
equilibrium.
The structural systems at Salisbury Cathedral that
can be considered most applicable to other structures
are the various iron bands and framework systems that
tie together the tower and spire structures and resist
the outward thrust forces generated on these struc-
tures. The original medieval iron frameworks in the
lantern and at the base of the spire are quite elegant
in their minimal mass, simple connection joints and
concealed design. Successful modern replacements,
most notably Taylor’s tie ring above the openings at
the base of the spire, follow similar concepts while
incorporating appropriate updated materials improve-
ments. What these tie systems accomplish is to provide
the mechanism to internally counterbalance outward
thrust loads and avoid the need for external buttress-
ing or other visually intrusive modifications. Properly
designed to not impose additional stresses due to
differential thermal movement and with adequate pro-
tection from weather effects that cause corrosion, these
metal tie reinforcements are structural solutions that
minimally intrude on the historic fabric.
Metal bands along the height of the spire, such as
the ten copper bands of the Salisbury spire and the
iron rings embedded in the horizontal stone bands
of the Freiburg Münster address less straight-forward
but important issues. These bands are not theoretically
required due to any dead or wind load stresses on the
shell of the respective spires, but they do effectively
consolidate the spire cones to distribute stresses so
that local irregularities do not propagate and become
a major structural problem.
The other critical structural system in the Salisbury
spire is the weighting of the top portion of the spire
through connection of the internal scaffold system to
the capstone to provide adequate dead load resistance
to wind loads through the upper portion of the spire.
This system supplements the metal cramp connected
stones to provide a sufficiently solid structure at the
top to counteract high wind load stresses and thin
geometrical cross sections. Although specific histori-
cal evidence is insufficient to prove direct cause and
effect, it does appear possible that the temporary loss
of this connection could be a cause for the visible bend
at 12 meters from the top of the spire.
234
REFERENCES
Cocke, T. & Kidson, P. 1993. Salisbury Cathedral: Perspec-
tives on the Architectural History. London: HMSO.
Fritz, H. 1926. Der Turmhelm des Münsters zu Freiburg im
Breisgau. Dissertation (Doktor-Ingenieurs). Technischen
Hochschule zu Karlsruhe, Germany.
Heyman, J. 1995. The Stone Skeleton – Structural Engi-
neering of Masonry Architecture. Cambridge: Cambridge
University Press.
Price, F. 1997 Facsimile Edition of original 1753 Edition. A
Series of particular and useful Observation, Made with
great Diligence and Care, upon that Admirable Structure,
The Cathedral-Church of Salisbury. And over, Hampshire:
Salisbury & Stonehenge Edition Facsimiles.
Reeves, J., Simpson, G. & Spencer, P., 1992. Iron Rein-
forcement of the Tower and Spire of Salisbury Cathedral.
Archaeological Journal, 149, pp. 380–406.
Tatton-Brown, T. 1991. Building the tower and spre of Salis-
bury Cathedral. In: L.T. Courtenay, ed. The Engineering
of Medieval Cathedrals. Aldershot, Hampshire: Ashgate
Publishing Limited, pp. 327–349.
Tatton-Brown, T. 1996. The Archaeology of the Spire of Sal-
isbury Cathedral. In: L. Keen and T. Cocke, eds. Medieval
Art and Architecture at Salisbury Cathedral, Proceedings
of The British Archaeological Association Conference,
15–18 July 1991 Salisbury. Leeds: W. S. Maney and Sons
Limited, pp. 59–67.
Taylor, P. 1988. The Tower and Spire of Salisbury Cathedral.
Association for Studies in the Conservation of Historic
Buildings Trans., 12, pp. 3–10.
Wren, C. 1668. Sir Christopher Wren’s Report on Salisbury
Cathedral for Dr. Seth Ward, Bishop, 1668. In: The Wren
Society, ed.The EleventhVolume of theWren Society 1934.
Oxford: Printed for the Wren Society at the University
Press, pp. 21–26.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Structure of baroque church roofs in Bavaria

S.M. Holzer & B. Köck

University of the Federal Armed Forces Munich, Neubiberg, Germany

ABSTRACT: The authors report results of an extensive survey of baroque timber roof constructions in Bavaria.
It turns out that the majority of these roofs lack a tie-beam at their base, because the roofs harbour masonry
vaults which rise high into the roof space. In the church roofs analyzed, different historical techniques have
been identified for carrying the thrust of the roof and the vault in the case of a missing tie-beam. We analyze
these techniques both from a historical point of view (development and spread of technological knowledge by
carpentry treatises) and from a structural point of view, including the structural assessment in an exemplary case.

1 AN INTRODUCTORY CASE STUDY

We start from some observations made during an inten-

sive inspection of the southern Bavarian pilgrimage
church of Tuntenhausen, some 30 km south-east of
Munich. The church was erected in 1628, during the
ThirtyYears War. It replaced an older, gothic church, of
which only the tower in the west is persisting – a strange
twin tower, consisting of two towers built wall-to-wall.
The church is one of the earliest “baroque” churches in
the area. However, its layout resembles very much the
layout of a gothic church. The church has three parallel
naves with three bays. The main nave is in fact some-
what higher than the aisles, but the whole keeps the
general impression of a “hall” church with three naves Figure 1. A view of the vaults of the church at Tuntenhausen.
of roughly equal height (cf. Fig. 1). The main arches
of the church are supported by two pairs of slender pil-
lars. In the east, the church features a chancel with an
ambulatory, encircling the partly preserved old gothic
polygonal apse with the principal altar.
When the authors of the present paper inspected
the church, they were confronted with preparations for
restauration work, and their opinion concerning wide
cracks in the vaults was asked for. It was not clear
whether the damage was caused by defects in the roof
structure or rather by ongoing settlements. Investiga-
tions on the stucco surfaces indicated that the cracks
were not new, but had developed with a decreasing rate
since the 17th century (Rosenheim 2007).
It turned out that the present state of the structure
could not be explained without a thorough understand-
ing of its building process, which could be deciphered
from observations in the attic space. Figure 2. Section of the roof truss, Tuntenhausen (1628).
Quite like the church itself, the roof structure resem- Shaded beams correspond to the original structure, blank
bles gothic constructions very much (cf. Fig. 2). beams are later additions.

235
The main girders of the roof are constituted by two in 17th century Bavarian architecture) was already a
superimposed trussed frames. The lower one is char- reaction to immediate damage or not. However, it is
acterized by two rows of timber columns rising directly sure that considerable deformation set in immediately
above the main arches of the church. This part is well after the completion of the construction. This is tes-
stiffened by various braces rising parallel to the rafters. tified by the deformed shapes of the aisle vaults, the
The upper storey of the roof, by contrast, makes use of outward tilt of the circumferential walls, as well as a
a “liegender Stuhl” structure, the most characteristic mess of historical rehabilitation structures above the
element of 17th and 18th century roofs in Germany. vault (cf. Figs 2 and 3). These include scissor braces
The roof accommodates the rising barrel vault of the intended to carry the thrust of the roof, as well as
central nave by simply leaving out the tie-beam in the another pair of scissors anchored to the circumferential
central section. walls, obviously meant to carry the vault thrust. From
However, in the intervals between the lunettes of the a modern point of view, it is obvious that the slowly
main vault, projecting beam ends suggest that, origi- hardening historical mortar would have prevented the
nally, the principal transverse trusses of the roof in aisle vaults from performing their intended function
fact did have continuous tie-beams (cf. Fig. 3). There as abutments, and that movements would slowly come
is no immediately obvious explanation for this feature. to a standstill only as the mortar reached its final stiff-
However, it turns out that this tiny peculiarity – easily ness. Most of the thrust is now carried by the conical
overlooked – is intimately connected to the construc- lunettes. On the other hand, the scissor braces presum-
tion process of the vault and roof, and combines well ably did not do a great service towards stopping the
with the following observations on the vaults, as we process of deformation.
shall see: Anyhow, a large percentage of the present-day
deformation must have occured immediately after
– The longitudinal barrel vault of the central nave is completion, since the stucco work of the vault already
slightly pointed in cross-section
– The lunettes of the central barrel vault rise consid-
erably (1 m circa) above the capstone of the main
arches
– There is no transverse arch, neither in the nave nor
in the aisles, neither on the extrados or intrados of
the vault
– The vaults of the aisles are rounded transverse barrel
vaults, connected to each other by conical lunettes
in longitudinal direction
– In cross-section, the profile of the vaults over the
aisles is visibly distorted where the adjacent conical
lunettes meet each other.
All these observations can be accommodated in the
following hypothetical account of the vault erection:
First, the outer walls of the church and the two arcaded
walls separating the naves were built. At this point,
there was no intermediate transverse connection or
stiffening between the free-standing, slender arcades
and the outer walls. Next, the roof was erected. The
roof made use of continuous tie-beams as anchoring
beams in the main trusses. This ensured temporal sta-
bilization of the free-standing, very slender masonry
structures (even the outer walls are very slender struc-
tures, exhibiting huge windows, but no external or
internal buttresses). After roof erection, the vaulting
was started with the transverse barrel vaults of the
aisles, the anchoring beams of the roof still being in
place. After completion of the transverse barrel vaults,
the builders obviously trusted them to carry the thrust
of both the roof truss and the central barrel vault,
and dared to cut the anchoring beams. Then, finally, Figure 3. Roof structure, Tuntenhausen. Arrow indicates
the central barrel vault was built. One might speculate location of originally continuous tie-beam cut away prior to
whether the unusual shape of the pointed vault (unique vault erection in the nave.

236
takes account of the deformed vault shapes, and is also Technically even more demanding is the church type
fastened to the vault in such a way as to effectively developed in Bavaria during the 18th century, namely,
hide already existing cracks. Therefore, the authors felt churches which embed an octogon for the congrega-
entitled to conclude that the present-day cracks present tion in a rectangular outer shape, the vault being a
no hazard for the structural stability of the church, even cupola which is hidden beneath an ordinary ridge roof.
though some of the cracks are up to 10 mm wide. These cupolas easily reach 20 m spans (e.g., parish
As a remark concluding the case study, we would church of Murnau, see below), pushing the required
like to note that the timber columns of the roof are all roof span to the limit of traditional carpentry.
tilted. However, this effect seems to be mostly inde- The authors of the present paper have conducted,
pendent from the problems with the vaults. Rather, the during the last two years, an extensive survey of
upper storey of the roof, the “liegender Stuhl”, leans baroque roofs in southeastern Germany (Bavaria),
towards the east due to a lack of stiffening. The stiff with the goal of identifying typical solution strate-
construction of the lower storey prevented the upper gies followed by carpenters in that region to cope
part from toppling, even though the construction of with the problem of large span and the challenges
the hip roof over the ambulatory is very defective. Fur- of fairly low pitch and missing tie-beam. Around
thermore, it is worth noting that the horizontal beam at 40 church roofs have been inspected, covering the
mid-elevation of the lower storey of the roof (cf. fig- entire period from 1600 to 1800. Our study comple-
ure 2) cannot be considered as a raised tie-beam, being ments the survey conducted by Sachse in the region
connected to the columns only by tenon-and-mortise directly adjacent to our area in the west (Sachse 1975).
connections. The full results of the survey will be published in a
dedicated monograph (Holzer & Köck 2008). In the
remainder of the present paper, we present typical con-
structive strategies encountered during the survey, and
2 CHALLENGES OF THE BAROQUE ROOF
we address the question how much information the
contemporary carpenters could possibly have gained
The roof of Tuntenhausen is just one of the start-
from the comparatively rich production of carpenter’s
ing points of the development of the baroque roof
treatises published in German language. Also, we dis-
in Bavaria. Obviously, it did not work quite as well
cuss issues of the mechanical behaviour of the typical
as it should have. The reasons for the failure of
structures encountered.
the structural concept at Tuntenhausen and the chal-
As late as 1814 (!) Franz Sax, of Vienna, Austria,
lenge of similar situations elsewhere are quite obvious:
published a section of a barrel-vaulted church with a
Baroque roofs in southern Germany typically exhibit
roof truss in his treatise on Bau-Technologie und Bau-
a pitch of around 45◦ . Contemporary carpenter’s trea-
Oekonomie (Sax 1814, cf. Figure 4). The lower storey
tises call this pitch the “German” pitch. It is a pitch
of the roof structure is constituted by a high “liegender
that is considerably lower than the 60◦ pitch of typical
Stuhl”. The only precaution taken to carry the thrust of
gothic roofs. The low pitch aggravates problems with
the roof is a raised tie-beam, connected to the braces
thrust, particularly in situations where continuous tie-
of the “liegender Stuhl” by means of iron clamps.
beams are not feasible, and the traditional carpentry
techniques that had been developed over the centuries
in the area north of the Alps for the high-pitch gothic
roofs were ill adapted to the new challenge of low-
pitch roofs, as evidenced by many 17th century roofs,
including the one of the Hofkirche at Neuburg on the
Danube (1615), the lowest of the Bavarian roofs of
the time.
Furthermore, brick barrel vaults with small lunettes
became the standard vaulting solution in Bavarian
baroque architecture. Unlike the adjacent regions in
southwestern Germany, Bavarian baroque builders
used timber “lattice” vaults only with reluctance, with
only very few exceptions, mostly cupolas. In addition,
virtually all the brick vaults of south-eastern Bavarian
baroque churches rise high enough into the attic space
to prevent continuous tie-beams.
At the same time, even minor church buildings of Figure 4. Design of a church roof accommodating a
the 17th and 18th century easily reach the 16.50 m masonry barrel vault, from Franz Sax, Bau-Technologie, 1814
span which is a typical upper threshold on late gothic (detail from plate V). A raised tie-beam is secured by iron
church nave widths (Augsburg, St. Ulrich, 1499). clamps.

237
Figure 5. View below the raised tie-beam in the roof of the
church atAltenerding (1721). The long tie-beam in the middle
of the picture is not a continuous one, but also stops at the
extrados of the masonry vault (covered by thermal insulation).
From a modern perspective, it seems unlikely that
someone would rely on such a construction. However,
a roof corresponding exactly to Franz Sax’ layout was
completed in 1721 in Altenerding (Fig. 5). As opposed
to the roof depicted by Franz Sax, the carpenter at
Altenerding opted for a tenon-and-mortise connection
of the raised tie-beam. The tenon traverses the entire
rafter and is connected to its outer face by a long iron
strap nailed to the top of the tie-beam. The vault at
Altenerding is a very shallow, basket-handle shaped
barrel vault.
Quite obviously, this system could not be expected
to perform well, particularly with a view to the total
absence of outer abutments, side chapels, or any other
structure capable of carrying the outward thrust of the
vault. As a consequence, the roof has undergone very
heavy restoration work in the 20th century, modifying
the entire structure into a modern truss by members
added between the raised tie-beam and the collar beam.
Franz Sax also illustrated a different setup for roofs
with discontinuous tie-beam (cf. Fig. 6). Here, two
diagonal braces meet at a king post. The details of the
joinery at this critical point remain unclear. Roofs of a
similar construction were actually built in Bavaria as
early as the last decade of the 17th century. One exam-
ple is given by the former Augustinian abbey church of
Weyarn, within 15 km from Tuntenhausen (cf. Fig. 7).
The church is a typical example of the Bavarian “wall-
pillar church”, i.e. a church with a nave vaulted by a
longitudinal barrel vault, accompanied on either side
by chapels with transverse barrel vaults. In such a
church, the transverse barrel vaults typically spring
at the same level as the main barrel vault, but their rise
is much lower than that of the longitudinal vault due to
the larger span of the latter. The difference is accom-
modated by rising, conical lunettes. Provided the side
238
Figure 6. Design of a roof without a tie-beam, from Franz
Sax, Bau-Technologie, 1814 (detail from plate X). Diagonal
braces attached to king post.
Figure 7. Section of the roof at Weyarn (1693).
chapels are deep enough, the scheme provides opti-
mal abutment to the thrust of the main vault and thus
leaves only the problem of roof thrust unaddressed.
At Weyarn, the diagonal braces appear in combination
with a low first story of the roof truss, whereas the main
“liegender Stuhl” of the roof rises on top of some kind
of elevated tie-beam. Therefore, the diagonal ties only
have to carry the thrust of the lower portion of the roof.
Nevertheless, the joint between the king post and the
diagonal ties turns out to be the most critical part of the
whole structure. In Weyarn, these joints are lap joints
secured by a wooden pin. The pins have failed, but nev-
ertheless the structure still stands largely undamaged,
probably due to the fact that the thrust of the roof could
be transmitted to the abutments through the rudimen-
tary tie-beams by friction. In this respect, the actual
design at Weyarn is much better suited to the problem
than the one published by Sax, where the lower storey
of the roof is also constituted by a “liegender Stuhl”.
An almost identical roof can be found in the church
of Perlach (today part of Munich). It dates to 1728.
Figure 8. Joint of diagonal braces with king post.
Munich-Perlach (1728). Viewed from the back side. Front
side shows a dove-tailed lap joint and additional iron straps.
Figure 9. Roof of the church at Ramersdorf (Munich, first
third of the 15th century).
As opposed to Weyarn, the builders at Perlach one
generation later would not rely on traditional joinery
to carry the load in the diagonal braces, but introduced
iron straps fastened to the king post by a bolt with an
iron wedge (Fig. 8).
Surprisingly, none of the roofs discussed so far
made use of a roof truss element that was already well
known in the 15th century, namely, scissor braces. Dur-
ing the 15th century, scissor braces became something
like a standard solution for roofs where continuous
tie-beams were not feasible because of a wooden
barrel vault. Medieval wooden barrel vaults are not
frequently encountered today in Germany, but traces
prove that this kind of construction was once quite
popular, not only in town halls, but also in churches
(cf., e.g., Cramer & Eissing 1996). In what is today
Munich, the church of Ramersdorf (first quarter of
the 15th century, see Fig. 9), is an example of such
a structure. The roof with its 60◦ pitch once accom-
modated a trefoil-shaped wooden barrel vault, as is
evidenced by traces of stucco at the gable ends of the
roof (see Fig. 9, background; cf. also Hösch 1996).
239
Figure 10. Solution for the trefoil timber barrel-vault as
suggested by Johann Wilhelm (1649).
In Ramersdorf, the wooden vault was soon replaced
by a masonry one with a much lower rise. The roof
shows scissor ties connecting all pairs of rafters.
A roof of very similar structure is also depicted
in Johann Wilhelms carpentry treatise, Architectura
Civilis, first published in 1649 (Wilhelm 1649, cf.
Fig. 10), the earliest dedicated carpentry treatise to
appear in German language. However, it needs to be
noted that Wihelm’s roof is stiffened by a liegender
Stuhl frame, which collects the loads of the rafters in
transversal frames, and only these frames are equipped
with the scissor ties, rather than every pair of rafters
as in the medieval roof. Wilhelm’s treatise sparked
a rich production of further German carpentry trea-
tises, including the Architectura Civilis published by
Caspar Walter, of Augsburg, in 1704 (Walter 1704).
Each of these treatises until the end of the 18th cen-
tury presented at least one example of a roof truss
with missing tie-beam, in combination with a wooden
or lattice barrel vault (e.g., plates 18–21 in Walter
1704). Also, several examples prove that the technique
was well known to contemporary carpenters and actu-
ally employed in building practice. The roofs of the
17th and early 18th century protestant churches of
Augsburg (St. Ulrich, constructed by Caspar Walter
in 1710) and Regensburg (Trinity church) make use of
scissor braces.
Under these circumstances, it seems somewhat odd
that the scheme was only reluctantly introduced into
roofs which had a brick vault rather than a wooden
vault. In the case of the brick vault, the added weight
causes greatly increased thrusts, so that proper precau-
tions to carry the thrust become a must. Nevertheless,
carpenters adhered longtime to traditional structural
forms and did not readily transfer solutions from one
problem class to another.
A roof equipped with scissor braces between every
pair of rafters (not just in the principal trusses) can be

Figure 11. Scissor braces in the roof of the parish church
at Murnau (1721), employed to accommodate the cupola of
20 m diameter inside the roof.
found in the church of Murnau (1717–21). However,
this church was a completely singular structure at the
time of its building. The “nave” of the church is con-
stituted by a Bavarian variant of the Roman Pantheon,
namely, an octagonal space with a diameter of 20 m.
A perfect sphere is inscribed into the nave, the princi-
pal corniche coinciding with the equator of the sphere.
Strangely, the huge hemispherical cupola (one of the
largest baroque cupolas in Germany, surpassing the
one at Weingarten, Swabia, cf. Kutnyi & Wiesneth
2007, with a diameter of 15 m, and approaching the
cupolas at Ettal and Dresden, both with a diameter of
approximately 25 m) is not visible from the outside, but
hidden under an ordinary ridge roof. Accommodating
such a large cupola in the attic posed an unprecedented
challenge to the carpenter.
In Murnau, the carpenter resorted to scissor braces
connecting all pairs of rafters to each other in the area
where the cupola protrudes into the roof space (see
Fig. 11). Also, he added a bridge-like truss in the plane
of the rafters to carry the load from the “open” part of
the roof to the parts with continuous tie-beams. Last
not least, both the brick cupola and the “hole” cut into
the base of the roof are reinforced by closed rings of tie-
beams at various levels. The Murnau roof is one of the
most remarkable feats of Bavarian baroque carpentry
and would deserve a monographical study in which we
cannot give here.
Only by the middle of the 18th century, scissor
braces had developed into a commonplace solution for
the problem of interrupted tie-beams. Regularly, the
south-eastern Bavarian architect Johann Alois Mayr
(1723–1771) employed scissor-braces whenever there
was a need to use discontinuous tie-beams. His church
roofs at Baumburg (1756), Marienberg (1764) and
Kirchweidach (completed only after Mayr’s death, in
1774) all employ scissor-braces. The case of Baum-
burg (Fig. 12) is particularly interesting for two rea-
sons. Firstly, Baumburg is a standard “wall-pillar”
240
Figure 12. Section of the roof at Baumburg (1756).The scis-
sor-braced roof accommodates a longitudinal barrel vault, not
shown here.
church with a longitudinal barrel vault covering the
nave; as such it reflects the typical situation of Bavarian
baroque. Secondly, the joinery of the roof exhibits a
marked “engineering” character: The scissor-braces
are not connected to the other timber members by tra-
ditional joints, but rather, exclusively by iron bolts.
On the other hand, all the less important joints follow
tradition, even to such an extent that lap joints with
a curved end – a characteristic element from the ver-
nacular architecture of the region – abound in the roof
truss.
3 STRUCTURAL ASSESSMENT OF SCISSOR
TRUSSES
Stress analysis of historical roof trusses poses spe-
cific problems which are otherwise not common in
structural engineering. As opposed to the analysis of
modern constructions, the idealization of the structure
is not straightforward, in particular because the stiff-
nesses of the joints and the support conditions are not
well defined. On the other hand, these modeling deci-
sions have a decisive influence on the results, as we
shall demonstrate by a little example.
In the following, we present some preliminary
results from the analysis of a principal girder of the
Baumburg roof under symmetric loading. Very similar
analyses have also been conducted by Lewandoski &
Levin (2003) for historic American scissor trusses.
Although these 19th century American trusses are
only superficially akin to our baroque trusses, but dif-
fer conceptually in many details, our results compare
well to those already reported by Lewandoski & Levin
(2003).
In the finite element analyses presented, we have
considered the “liegender Stuhl” girders as frames
with rigid corners. The rafters are in contact with the
tilted columns of the “liegender Stuhl”. All traditional
timber joints have been considered as perfectly hinged,
Figure 13. Baumburg roof. Normal forces obtained for a
model based on supports that are allowed to slide horizontally.
Thick lines: compressive forces. Thin lines: tensile forces.
quasi pin-jointed connections. By contrast, all connec-
tions employing more than one iron bolt have been
modeled as rigid because the stiffness of such a con-
nection is much higher than that of a wooden pin and
a pair of iron bolts permits transfer of moments.
Firstly, we study the effect of the support conditions.
Allowing the supports to slide freely, we obtain what
one would probably expect, namely, tensile stresses in
the scissor braces (Fig. 13). The computation demon-
strates that the king post performs a good job as a
tie-rod between what is essentially the upper and lower
chords of a triangulated truss, constituted by the rafters
and the scissor braces. In this model, the upper portions
of the braces act as struts supporting the rafters, but
only with minor effect. As a natural consequence of
the overall structural response, the lower collar beam
is also in tension, which contradicts expectation and
will not be permitted by the joinery. However, neglect-
ing the effect of the collar beam will not change the
picture much.
By contrast, if we modify the model such that
the inner supports of the frame are considered fixed,
the stress in the cross-braces reverses sign (Fig. 14); the
scissor braces then turn into some kind of an arch
supporting the lower collar beam.
This model may be criticized on the grounds that
the support will never be perfectly fixed in reality, and
it is true that the scissor braces regain their function
as tie-beams immediately once one forces even a very
small outward displacement on the inner supports.
However, the role of the scissor braces is also
strongly dependent on the king-post. If we assume
that the king-post is imperfectly fixed to either the
ridge of the roof or to the collar beam, the scissor
braces are once again transformed into an arch and get
compressive stresses (Fig. 15).
In summary, the loads of a scissor-braced truss may
either be carried by a simple triangulated, statically
determinate truss constituted by the scissor braces
and the rafters, with the essential contribution of the
241
Figure 14. Baumburg roof. Normal forces obtained for a
model based on fixed supports. Thick lines: compressive
forces. Thin lines: tensile forces.
Figure 15. Baumburg roof. Normal forces obtained for a
model with reduced effectiveness of the king post. Thick
lines: compressive forces. Thin lines: tensile forces.
king-post, or, alternatively, by some kind of a polyg-
onal arch. The transition between these two distinct
and essentially incompatible load-carrying structures
depends strongly on local modeling assumptions and
stiffness ratios. Somewhat surprisingly, the “liegender
Stuhl” frame plays only a secondary role in the overall
stress distribution.
The true situation is much more difficult to assess
than could be demonstrated in this greatly simplified
example. Roofs which do not have a tie-beam are much
more sensitive to such modeling assumptions than are
roofs with a closed rafter–tie-beam triangle.
Based on experiences similar to the one pre-
sented, the authors are currently working on more
detailed computational modeling of typical baroque
roof trusses which lack tie-beams, including the appli-
cation of interval arithmetic and fuzzy set theory.
REFERENCES
Cramer, J. & Eissing, Th. 1996. Dächer in Thüringen. Bad
Homburg: Verlag Ausbildung und Wissen.
Hösch, K. 1996. Pfarr- und Wallfahrtskirche Maria Ramers-
dorf/München. Passau: Peda.
Holzer, S. & Köck, B. 2008. Barocke Bautechnik. Gewölbe,
Kirchendachwerke und Kuppeln in Südbayern. Regens-
burg: Schnell & Steiner, to appear in 2008.
Kutnyi, A. & Wiesneth, A. 2007. Die Kuppelkonstruktion der
Basilika in Weingarten. Preprint, Lehrstuhl Bauforschung
und Denkmalpflege, Tech. Univ. Munich.
Lewandoski, J. & Levin, E. 2003. Historic American roof
trusses. Part I: Scissor trusses. With an appendix: Historic
scissor truss analysis. Timber framing (69): 4–17.
242
Rosenheim 2007. Staatliches Hochbauamt Rosenheim,
report on the stuccos surfaces of Tuntenhausen, unpub-
lished personal communication.
Sachse, H.-J. 1975. Barocke Dachwerke, Decken und
Gewölbe. Berlin: Mann.
Sax, F. 1814. Bau–Technologie und Bau–Oekonomie. Vol. III:
Die Zimmermanns–Kunst. Wien: Doll.
Walter, C. 1704. Architectura Civilis. Augspurg: Wolff.
Wilhelm, J. 1649: Architectura Civilis. Frankfurt: Author.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Brick arch bridges in the High Cauca Region of Colombia: A forgotten

construction tradition

J.A. Galindo, J.A. Paredes, A.M. Muñoz & Y.M. Ospina

Universidad Nacional de Colombia, Manizales branch, Colombia

ABSTRACT: This document presents partial results of a much broader research project that has recognized
historical and technical aspects pertaining to a set of 34 historic brick arch bridges, 20 of which still exist; all
with diverse geometric characteristics and singular dimensions. The methodological process is exposed as was
carried out with one of them and conclusions are presented, which can be extended to almost the totality of the
cases studied for the purpose of offering analysis alternatives to professionals interested in the rehabilitation and
conservation of the South American architectural and engineering patrimony.

1 INTRODUCTION

1.1 The location

The geographic region of High Cauca lies in to the
southwest of Colombia’s current territory and has,
today, close to 230 thousand inhabitants. Its principal
city is Popayan, founded by the Spaniard Sebastian
de Belalcazar in 1537. The city was built some 1737
meters above sea level and it is located at 02◦ 26 39 N
and 76◦ 37 17W. Its average yearly temperature is
of 19.8◦ C, although there are strong fluctuations in
temperature for short periods of time. Its annual pre-
cipitation regime reaches 1941 mm. The region also
evidences intense seismic activity, which for centuries
has been affected earthquakes of considerable mag-
nitude; the most recent of which took place in 1983
with an intensity of 5.5 degrees on the Richter scale,
causing great material damage and loss of human
lives.
1.2 Brick arch bridges in the Cauca region
In spite of having a relatively hostile environment
threatening the conservation of buildings, the High
Cauca region was, since its colonization (in the early
part of the 16th century), of high strategic value – given
its being located at an intermediate point between the
port of Cartagena de Indias and the commercial cen-
ters of Quito and Lima (Fig. 1). This condition gave
added importance to network of roads in Popayan and
to its works of infrastructure in the regional system of
communication paths (Galindo, 2004) and motivated
Figure 1. Geographic location of the Colombian territory
and the Cauca region. Drawing by: E. Sotelo.
local authorities to build at least 34 brick arch bridges
between 1718 and 1920.
Each of the bridges has a set of particular geometric
and dimensional characteristics, following Mediter-
ranean tradition standards related to the preparation of
construction materials (brick and mortar), the fabrica-
tion of centerings, the laying of rings, the dimensioning
of piers, and in general, all those elements related
to the conformation of structures that have lasted for
decades.
This paper will explain the historical genesis, along
with the constructive and structural characteristics of
one of the most significant bridges of this group: the
bridge built near the city of Popayán between 1769
and 1778 over the Cauca River, as part of the roadway
between this city and the city of Cali, which is located
120 km to the north.

243
Figure 2. Lateral rendering of the brick arch bridge over the
Cauca River in the city of Popayán. Drawing by: E. Sotelo.

Table 1. Measurements and dimensional ratios on the

bridge over the Cauca River in Popayán.

Diameter of Thickness Ratio of ring

the arch of the ring thickness/Arch
Element M M diameter

Arch 1 8.47 0.76 0.08

Arch 2 9.00 0.78 0.08
Arch 3 10.35 0.77 0.07 Figure 3. Simon Schenherr’s blueprint for a wooden bridge
Arch 4 19.06 1.48 0.07 Project over the Cauca River. Source: Historical Archives at
∗
Universidad del Cauca, Cabildo Section, Mss, 1753.
Data obtained through direct measurements on the structure.

2 GENERAL DESCRIPTION OF THE BRIDGE

OVER THE CAUCA RIVER IN POPAYÁN

2.1 Geometry and dimensions

The brick arch bridge over the Cauca River in Popayán,
is 153 m long and has a half-point main arch with a
19.06 m diameter, which crosses the variable current
river. It also has 3 half-point leveling arches, leading
to the slope on the north side; while on the southern
side the main arch is directly supported upon a natural
rock support vertically fitted to the ground (Fig. 2).
The structure has a constant 5.84 m width through-
out its whole length. It is entirely made of solid clay
brick joined with lime mortar, although beneath the
main arch there are some rows of etched stone –
construction material often also found on the tympa-
nums of the bridge and on the clefs of the vaults. A
sole cutwater with a diamond-shaped base is found
under the extreme north section of the main arch and
it is also entirely built with brick. The dimensions and
dimensional ratios are given in Table 1.
2.2 History of the construction of the bridge
The genesis of the structure of the bridge over the
Cauca River is kept in several dispersed documents,
which allow its partial reconstruction. In the Histori-
cal Archives at Universidad del Cauca (Popayán) there
is a document dated in 1753, which registers that priest
Simon Schenherr, member of the Company of Jesus
and master builder in charge of the construction of var-
ious religious buildings in the city and participating
member in the design phase of the bridge construc-
tion. He authored two blueprints each corresponding
to the same bridge.
Figure 4. Simon Schenherr’s blueprint of a masonry bridge
project over the Cauca River. Source: Historical Archives at
Universidad del Cauca, Cabildo Section, Mss, 1753.
The first of the blueprints (Fig. 3) bears a drawn
of a covered wooden bridge, raised over four pillars
with triangular cutwaters on each face, two of them
with foundations in the water. The bridge roadway
surface was made up of 3 spans and reached a total
approximate length of 34 m, assembled over longitu-
dinal beams supported above the pillars through beams
at 45◦ angles. Next to the drawing, there are also indi-
cated the amounts of construction materials necessary
for its execution.
The second blueprint (Fig. 4) reveals a 3 arched
masonry bridge: the two arches of the extremes with
24 m diameters and the central arch with a 6 m diam-
eter, supported on 4 pillars with cutwaters of equal
geometry, two of them set at the river bottom. The
bridge roadway surface has a double slope and accord-
ing to a notation on the drawing, it was 4 m wide.
For unknown reasons, the bridge could not be built
following any of Schenherr’s designs. Additionally, he
did not participate in directing the work.

244
Figure 5. View of the main arch of the bridge over the Cauca
River in the city of Popayán (2007).
In another document kept in the Nation’s General
Archives (Bogotá) there is testimony relating how, dur-
ing a first stage of the work begun in 1769, the bridge
collapsed at the precise moment of closing one of the
brick arches – due to the lack of technical direction.
The final construction was left to Francisco Basilio de
Angulo and Josep Hidalgo de Aracena.
Once in service and for decades, the bridge over
the Cauca River was always considered one of the
best in the region and the nation; for example, English
colonel J. P. Hamilton briefly mentions it in his book
(Hamilton, 1827) as part of his journey between
Popayán and Cali, just prior to the earthquake of 1827
that affected the region and destroyed several temples
and caused severe damage to the bridge’s main arch,
rendering it useless.
There is evidence (Zawadsky, 1929) that on at
least two occasions local construction workers tried to
rebuild it, but with no satisfactory results given that in
both cases, upon removing the wooden trusses placed
under the vault, the materials fell and were dragged
away by the river. It was not till 1840 that Polish
engineer, Estanislao Zawadsky, directed the definite
reconstruction work of the bridge.
2.3 Current state of conservation
A visual inspection carried out in 2006 permitted ver-
ifying that the bridge has not been altered by recent
structural interventions, except for the restitution of
the roadway surface in asphalt. A masonry plasterwork
made with Portland-type cement mortar during the
first half of the 20th century is in poor state of conser-
vation (Fig. 5) and only remains on the upper portion of
the bridge’s tympanums. On the lower part of the rings
there is evidence of an accelerated degradation process
of the bricks by weathering of the clay-like material
due to prolonged exposure to environmental humidity
and to the presence of the type of vegetation found in
245
Figure 6. View of lower portion of the ring on the second
arch of the bridge over the Cauca River, in Popayán, show-
ing the differential degree of deterioration of the constitutive
bricks (2007).
tropical regions, which contributes to the concentra-
tion of distinct living organisms upon bridge surfaces
(Fig. 6).
The macro structure is still in service to vehicular
traffic up to 5 Tons, as indicated to users. Maintenance
work is necessary, as is the replacement of ceramic
pieces that have disappeared by using new materi-
als with similar physical and mechanical properties
to those used in its construction.
3 CHARACTERIZATION OF THE CERAMIC
PIECES ON THE BRIDGE
3.1 Mechanical properties
According to the results of the historical research, the
bricks on the bridge over the Cauca River were made
in artisan manner with local clays as raw material, the
same craftsmanship still used in the fabrication of all
types of pieces, with 40 × 20 × 10 cm standard dimen-
sions. The furnaces are still simple constructions that
use a combination of mineral coal and vegetable lum-
ber as fuel, reaching firing temperatures between 750
and 1100◦ C. Production periods for the bricks vary
between 8 and 12 days.
Due to analysis of specimens obtained from direct
extraction of the masonry of the bridge, it was possible
to perform physical and mechanical characterization
of the clay pieces by following recommendations con-
tained in the ASTM C133-97 (American Society for
Testing and Materials, 2003) norms. Some of the
results obtained in seven of the samples are shown
in Table 2.
In spite of its low resistance capacity, there is no
physical evidence of fissures or crushing caused by
compression efforts.
Table 2. Results of the cold compression test applied sam-
ples taken from bricks on the bridge in Popayán over the
Cauca River.

Maximum Resistance to Elasticity

load cold crushing modulus
Sample kN Mpa Mpa

1 7.96 5.07 208.30

2 7.89 4.94 96.58
3 29.58 9.19 232.88
4 19.57 6.04 144.66
5 6.92 2.78 87.48
6 21.29 7.53 171.85
7 10.88 4.14 89.67
∗
Source: Materials Resistance Laboratory, Universidad
Nacional de Colombia at Manizales.
Figure 8. SEM/EDAX analysis of a sample obtained from
an inner portion of the bridge masonry. Source: Plasma
Physics Laboratory, Universidad Nacional de Colombia at
Manizales, Colombia.

Table 3. Quantification of the SEM/EDAX analysis on

the surface of the sample referred to in Figure 8.

Sample 1 – Figure 8

Element Wt %∗ At %∗∗

CK 11.09 18.25
OK 40.31 49.81
NaK 01.87 01.61
MgK 01.03 00.84
Figure 7. SEM/EDX micrograph (162x) of the organic por- AlK 13.74 10.07
tion of the surface of the construction work. Source: Plasma SiK 23.94 16.85
Physics Laboratory, Universidad Nacional de Colombia at MoL 02.98 00.61
Manizales, Colombia. KK 00.45 00.23
CaK 00.57 00.28
Other physical analyses carried out permitted con- TiK 00.55 00.23
cluding that the bricks on the bridge have high porosity FeK 03.47 01.23
(39.4–40.4%), which could not always be a charac-
∗
teristic of the original sample, but could be related Percentage related to the total weight of the sample.
∗∗
to natural physical wear and to the environmental Percentage related to the atomic weight of the sample.
conditions.
of masonry located at the base of the main arch of
3.2 SEM analysis the bridge. This analysis permitted a first approach
Scanning Electron Microscopy (SEM) micrographs to understanding the structural characteristics of the
with 162x magnifications (Fig. 7) applied over surface material. The results obtained from one of the sam-
planes of the bricks reveal hyphae and the forma- ples are expressed in Figure 8, and also related on
tion of mycelium of the fungus that has grown due Table 3. Silica and oxygen contents are notable in the
to long-term environmental exposure. The porosity of total weight of the sample, slightly close to elemental
the bricks facilitates the presence of these organisms. content of the material currently used in the production
Visual inspection permits deducing that the most of ceramics.
affected areas are found on the borders of the struc-
ture and are probably caused by the incompatibility of
3.3 XRD analysis
the mortar joints with the bricks, along with the water
percolation effect for many years. A study of the mineralogical composition of three
Simultaneously, SEM analysis techniques were masonry samples was conducted via X-ray Diffraction
applied on samples extracted from an inner portion (XRD) tests to identify the origin of the raw materials,

246
Table 4. Quantification of crystalline phases present in the
samples via XRD.

% of % of % of
compounds compounds compounds
Compound in sample 1 in sample 2 in sample 3

Quartz, low 4.28 24.98 17.59

Trydimite 27.46 – –
Cristobalite. Low 44.2 45.57 52.98
Hematite – 3.89 –
Paligorskyte 4.07 9.6 6.64
Anorthite 15.08 6.7 0.47
Muscovite – – 22.32
Barrerite 4.91 – –
Potassium Sulfate – 9.26 –
∗
Source: Materials Resistance Laboratory, Universidad Figure 9. Model of the bridge over the Cauca River, showing
Nacional de Colombia at Manizales. stress caused by the structure’s own weight.

Initially, a complex model of the bridge was drawn

as well as the conditions these were subjected to in the
production of construction materials and the exposure
to atmospheric conditions (Moropoulou et al. 2002).
The results of these tests are shown in Table 4.
All the brick samples manifest the presence of
crystalline phases in different proportions of quartz,
cristobalite, paligorskyte, and anorthite. The high level
of quartz in sample 2, associated to its dark color and
to the combination of cristobalite can be attributed to
the presence large quartz particles (Rice, 1987).
The presence of cristobalite, associated to the
orange hue of the bricks, indicates burning temper-
atures of at least 1200◦ C. Anorthite is a calcium
feldspar, formed by the interaction of calcite decom-
position residues (CaCO3 ) and clay minerals during
the burning process of the raw material upon reach-
ing approximately 900◦ C, leading us to believe that
for the sample analyzed the burning temperature was
not above 1000◦ C (Cardiano et al. 2004). Neverthe-
less, the presence of hematite, which appears around
850◦ C, is the most valuable indicator of the burning
temperature (Moropoulou et al. 2003).
4 DESCRIPTION OF THE BRIDGE’S
MECHANICAL BEHAVIOR
A static analysis – due its own weight – and a modal
analysis, both based on the behavior elasticity hypoth-
esis of the masonry materials on the bridge, were
conducted prior to its analysis against seismic forces.
The ANSYS© v.10 software was used as the analysis
tool. The elaboration of the model was done in four
stages: (a) geometric reconstitution from AutoCAD©
models, (b) grid-work of the model, (c) assigning of
the properties of the materials, restrictions, and loads,
and (d) solution of the model with emphasis on its
static and dynamic analysis.
up in 3D (67742 nodes and 33045 solid95-type ele-
ments), analyzing its structural response per its own
weight. A series of geometric conditions for its static
and/or dynamic conditions were accounted for: (1) dis-
placements were null in the direction of the main axis,
due to longitudinal confinement, (2) displacement in
the transversal direction to the main axis of the bridge
was equal to zero (as an effect of the wing walls of
the bridge and the cutwaters located at the base of
the piers), and (3) in the lower supports, all displace-
ments were considered restricted. Numerical modeling
was formulated from three structural components: the
vaults, the tympanums, and the filler material.
4.1 Static analysis (analysis of the global behavior
of the bridge)
The model of the bridge under gravity forces due to
the weight of its own structure is shown in Figure 9.
Crushing forces predominate with approximate values
of 160.5 Kpa along the entire upper zone. In the lower
zone, the crushing force value is greater, reaching val-
ues close to 1300 Kpa. Tension efforts are only present
on the arches with values close to 422 Kpa.
Along the length of the longitudinal axis or the X
axis there is a distribution of efforts with a certain
degree of uniformity: in the zone between arches there
are tensions with approximate values of 30 Kpa, while
on those same zones in the lower part of the bridge the
crushing forces reach values of 500 Kpa.
On the transversal axis, or the Y axis there is a pre-
dominance of crushing forces that vary from 200 Kpa
to 400 Kpa, reaching a highest value of 750 Kpa below
the arches. Small crushing phenomena are present on
the lateral part of the bridge with values oscillating
between 360 Kpa and 640 Kpa. The greatest displace-
ment of the bridge, due to its own weight, is noted on

247
Figure 10. First vibration mode of the bridge at 1839 Hz.
Figure 11. Third vibration mode of the bridge at 2148 Hz.
the Z axis, with a value of 5.20 cm, located on the clef
zone on the main arch.
4.2 Modal analysis modal (preliminary dynamic
analysis)
Modal analysis of the bridge was conducted to iden-
tify its dynamic properties. For such, the first 10
vibration modes were obtained. Figures 10–12 show
the first, third, and tenth mode of vibration, with
modal frequencies of 1839 Hz, 2148 Hz, and 3138 Hz,
respectively.
It can be stated that the components with the greatest
contribution to the total displacement of the bridge for
each vibration mode are those corresponding to the
transverse Y axis and the perpendicular Z axis. On
the Y axis, the greatest displacement occurred on the
sixth vibration mode – reaching a displacement close
to 8.20 cm, and for the vertical Z axis displacement
occurred on the ninth vibration mode with a value of
26.9 cm.
248
Figure 12. Tenth vibration mode of the bridge at 3138 Hz.
The greatest total displacement is caused on the
ninth vibration mode, presented on the span of the
main arch with a value near 27.2 cm with a domain
in the direction of the vertical axis of the bridge at
26.9 cm. Hence, under dynamic conditions, the ninth
modal form represents the greatest risk to the structural
integrity.
4.3 Seismic analysis
To simulate the response of the bridge to seismic
effects, we used a seismic spectrum given by Colom-
bian norm NSR-98 (Colombian Association of Seis-
mic Engineering, 2002) given by the equation:
where Aa = the parameter defining the maximum
acceleration in the region during a 475 year time lapse;
S = the parameter defining the type ground under the
structure; and I=the parameter defining the importance
of the edification according to a possible seismic disas-
ter. The following values were adopted in this analysis:
Aa = 0.25; S = 1.5; I = 1.2.
As a main result, it can be said that upon seismic
incidence defined by the spectrum mentioned, the zone
with the greatest structural vulnerability is the main
arch of the bridge, herein experiencing tension forces
that reach maximum values near 5298 Kpa with its
main component in the positive direction of the verti-
cal axis of the bridge, as noted in Figure 13. That is, in
its dynamic condition, the bridge mainly experiences

Figure 13. Tension forces on the clef of the main arch of the
bridge over the Cauca River produced by seismic accelera-
tion, under the framework of the seismic spectrum estimated
for Colombia.
tension phenomena with values around 75% above the
maximum crushing values presented by its static con-
dition, aside from such forces being concentrated in
the zone of the main arch.
5 CONCLUSIONS
We have discovered the existence of a coincidence
relationship amongst the questions: what does it mean
and how does it behave, which can only be answered
by also knowing the: how is it? In this sense, is that
the methodological aim stems from a detailed descrip-
tion of the bridge studied. The “how is it” is answered
by defining three fundamental aspects: (a) the shape,
understood as all that is susceptible to being graphi-
cally expressed and keeps a close relationship with the
historical genesis of the building, (b) the matter, i.e.,
the description of the mechanical properties (physi-
cal/chemical), and (c) the structure, understood here
as the system of relationships among the materials and
the shape.
Herein, we consider that research conducted on his-
torical buildings (edifications) should be developed
around those three spaces: the first of these through
field work involving architectural drawings, histori-
cal documentation, and visual inspection. Work on the
study of materials is carried out with the aid from lab-
oratories that have reported results, which have been
employed in the numerical modeling process via the
method of finite elements, seeking to understand their
behavior.
Easy access to calculus computer programs has
made the Method of Finite Elements (MFE) an appli-
cable path in the study of historical edifications. There
249
is abundant bibliography on the issue, as well as case
studies where the MFE has been applied. One of
the great advantages is that, additionally, it permits
answering a fourth question: how will it behave? This
is an important nucleus of the problem: foresight of
the interrelationships between the existing object and
the different alternatives to intervene upon it and how
to obtain a guarantee in each of them.
Thus, precisely knowing the behavior modes, under-
standing their matter, and predicting the types of
responses to future actions are today the indispensable
means of intervention upon the historical patrimony.
The research project exposed here points, in the long
term, toward the concrete design of actions leading
to the recovery and adaptation of the greatest pos-
sible number of cases, involving the work of local
communities.
ACKNOWLEDGEMENTS
The authors thank the support of the Dirección de
Investigaciones (DIMA) at Universidad Nacional de
Colombia in Manizales, for the financial support to
carry out this project. The authors also recognize the
collaboration received from professors A. Devia and
O. Correa from the same institution.
REFERENCES
American Society for Testing and Materials. 2003. ASTM
C133-97 Standard Test Methods for Cold Crushing
Strength and Modulus of Rupture of Refractaries. West
Conshohocken: ASTM.
Asociación Colombiana de Ingeniería Sísmica. 2002.
NSR-98. Normas colombianas de diseño y construcción
sismorresistente. Bogotá: ACIS.
Cardiano, P., Ioppolo, S., De Stefano, C., Pettignano, A.,
Sergi, S., Piraino, P. 2004. Study and characterization of
the ancient bricks of monastery of San Filippo di Fragalá
(Sicily). Analytica Chimica Acta 519: 103–111.
Galindo, J. 2004. Cruzando el Cauca. Cali: Gobernación del
Valle del Cauca.
Hamilton, J.P. 1827. Travels through the interior provinces of
Colombia. London: John Murray.
Moropoulou, A., Cakmak, A. & Polikreti, K. 2002. Prove-
nance and technology investigation of the Agia Sophia
bricks, Istanbul, Turkey. J. Am. Ceram. Soc. 85(2):
366–372.
Moropoulou, A., Polikreti, K., Ruf, V. & Deodatis, G. 2003.
San Francisco Monastery, Quito, Equador: characteri-
zation of building materials, damage assessment and
conservation considerations. Journal of Cultural Heritage
4: 101–108.
Rice, P.M. 1987. Pottery Analysis: A Sourcebook. Chicago:
University of Chicago Press.
Zawadsky, R. 1929. Mis impresiones de Popayán. Relator
I/29: 4.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

From empirical engineering to science-based engineering: The assessment

by three mathematicians of the “San Pietro” dome (1742)

O. Niglio
University e-Campus, Civil Engineering, Italy

ABSTRACT: The studies on the mechanic behaviour of domes, based on the static analysis of arches and vaults,
started in the first half of the XVII century. At the end of 1741, when the Pope Benedetto XIV commissioned an
assessment to three mathematicians Roger Joseph Boscovich, François Jacquier e Thomas Le Seur, belonging
to the Repubblica Romana dei dotti. There where serious worries on the static conditions of the San Pietro
vault where important cracks could be appreciated. Some interesting studies were developed on this problem,
even followed by learned and heated arguments, that lead to compile treatises that systemized the knowledge on
this argument. The historical documents even report the opinion of other experts, for example that one of the
mathematician Giovanni Poleni from Padua. On spite of that, the study of the three previous mathematicians
was different from the others carried out before because it had an important innovation: the assessment was
accomplished, perhaps for the first time, using a scientific criterion in order to interpret the mechanic behaviour
of an architectural structure. Its historical importance resides on the scientific theoretical conceptions used to
analyze a structural problem that were completely different from the empirical rules, at most of geometric nature,
used before. The principle of virtual works was further applied in the assessment. It was used to define the size
of the metal rings for the drum of the vault. Some researchers intend this assessment as the moment when the
change from the engineering based on artisan traditions to the engineering based on the application of new
scientific theories has been accomplished.

1 INTRODUCTION subject, accompanied by learned and heated debates.

The present scientific theories aimed at analysing the
mechanical behaviour of buildings started to be devel-
oped in the second half of the 17th century. Only
then were the mechanics of materials taking their
first steps with Mariotte and Robert Hooke’s research
(Timoshenko 1953; Timoshenko 1956). This followed
the road which Galileo had opened indicating exper-
imental observation as the basis of scientific knowl-
edge. In his last work Galileo himself had presented
the first observations on the “new science relating to
the mechanics” of structures (Galilei, 1638).
The first applications of the new scientific methods
to structural problems started to be enunciated between
the end of the 17th century and the early 18th century.
It was then, in 1741, that Benedict XIV commis-
sioned three mathematicians Roger Joseph Boscovich,
François Jacquier and Thomas Le Seur from the
“Repubblica Romana dei Dotti” (Boscovich & Al.
1742) to carry out a historical assessment. Serious
concerns had arisen over the static conditions of Saint
Peter’s dome, where significant cracks had appeared.
Interesting studies had already been developed on the
This had led to the compilation of authentic treatises,
which also aimed to order the knowledge on the sub-
ject. Historical documents also report other experts’
opinions on the state of the dome, including that of
the well-known mathematician from Venice Giovanni
Poleni (Poleni, 1748).
The three Mathematicians’ study stood out for
its important innovation. It contained an assessment
based entirely, perhaps for the first time, on a sci-
entific criterion aimed at interpreting the mechanical
behaviour of an architectural building. Its historical
importance lies in the fact that, unlike the previous
practices, which were based on empirical rules, gen-
erally of a geometric nature (see for example Poleni’s
studies on the statics of arches), theoretical concep-
tions, this time of a scientific nature, were used and
applied to the study of a structural problem. Although
not entirely correctly, the PVW (Principle of Virtual
Work) was adopted in the assessment, and used as an
instrument for measuring the metal rings to be applied
to the drum of the dome (Capecchi, 1999; Capecchi,
2002). In an attempt to determine an important date,
a number of experts (von Halász 1969) regard this

251
Figure 1. Saint Peter’s dome (Curcio, 2003).
assessment as the historical moment when the change
took place from engineering based on artisan tradi-
tions, of an empirical nature, to engineering based on
the application of the new scientific theories; theories,
which were just starting to become established.
The three Mathematicians’ assessment was pre-
sented towards the end of 1742 and printed in
1743. The study method thus introduced could truly
represent the historic beginning of modern civil
engineering. Unlike the previous practices, which used
rules dictated by intuition and experience, a scientific
process was applied to assess a building’s characteris-
tics of resistance and state of stress. This consequently
started a process, which does not yet appear complete
(Di Pasquale 1996).
2 “SAINT PETER’S” DOME
The building of “Saint Peter’s” Dome was started on 15
July 1588 under Sisto V, however, it was interrupted on
13 May 1590, just before the Pope’s death. Following
Michelangelo’s project, the building work had reached
the placing of the drum. It was completed by Giacomo
Della Porta at the beginning of the 17th century (Ack-
erman, 1968). The first cracks were discovered back in
1603, under Clement VIII, just after the building was
finished, on completion of the mosaics on the vaults.
Subsequent damage was then recorded after 1631, as
252
Figure 2. Section of Saint Peter’s dome (Curcio, 2003).
we can see in Gianlorenzo Bernini’s biography written
by Filippo Baldinucci. It was suspected that the statics
of the dome had been compromised by the insertion
of the spiral staircase by Bernini, set within the pillars,
under the pontificate of Urban VIII.
The controversy relating to Bernini’s work soon
quietened down. Baldinucci himself observed that a
number of cracks inside the dome on the cornice and
on the drum had been found before Bernini. Some
had attributed the damage to phenomena of settling
of the great dome and to the different working tech-
niques used for its construction. However, the three
Mathematicians claimed in their study that the damage
described by Baldinucci was not the damage found in
1742. The numerous criticisms raised against Bernini
referred to evident conditions of instability which were
present already in 1742. Subsequently, the instability
had increased and was developing.
“Il Tempio Vaticano” by Carlo Fontana, published
in 1694 (Curcio, 2003) made it possible to carry out
an in-depth analysis of the phases of construction of
Saint Peter’s Basilica and its dome, until the complex
took on its present arrangement. The work covers the
events of the construction of the building from the
beginning, when the emperor Constantine wanted the
basilica built near the tomb of the apostle Peter, until
the end of the 17th century.
3 THE ASSIGNMENT OF THE ASSESSMENT

Details of the task assigned to the three Mathemati-

cians can be read in the introduction of the assessment.
An important point concerns the use of the words
“ristauratione” (Restoration) and “conservazione”
(Conservation). This shows a clear wish to orient the
proposed interventions at safeguarding the existing
building work: an admirable aim, promoted in a time
when the philosophy of restoration was not yet clearly
defined.
The innovative aspect of the assessment concerns
the application of the method chosen for defining
the interventions. Explicit reference was not made to
Galileo because the memory of the trial in 1633 was
still vivid, and his writings were still banned (they were
until 1822). However, the three Mathematicians were
nonetheless faced with a problem of static restoration
using, perhaps for the first time, a scientific criterion of
calculation. They highlighted the importance of acting
not only using their “own visual observations”, but in
particular using a “good theory based on Mechanics”
for reference.
The process follows a plan that can be divided into
the subsequent four phases following a coherent and
logical approach. The phases are:
(a) Diagnosis, consisting of a careful observation of
the present state to determine the amount and
importance of the phenomena; Figure 3. Survey of the dome (1742 Parere di Tre Matem-
(b) Aetiology, consisting of an assessment, realized atici, Bibliotheca Hertziana, Rome).
from the previous observations, regarding the
identification of the causes, which may have lead
to the phenomena; the kinematic mechanisms of the various relative
(c) Prognosis, consisting of an examination of the movements.
possible criteria and methodologies available, The survey consequently becomes an instrument of
which could be used to identify and calculate the knowledge and support for thematic close examina-
solutions to be adopted; tions including kinematical analyses of the instability.
(d) Therapy, consisting of a detailed definition of the In the continuation of this part of the assessment the
working methods to be followed for applying the three Mathematicians describe the damage observed,
identified solutions. grouping it into three main areas of observation: the
drum, the vault and the lantern.

4 DIAGNOSIS
The first part of the study is dedicated to a detailed
description of the dome and the creation of a detailed
geometrical survey. This instrument of knowledge is
subsequently perfected by superimposing the existing
pattern of cracks.
The precise representation aims to lay the basis for
analysing the loads and interpreting the relative move-
ments between the various structural parts, which,
according to current terminology, could be consid-
ered macro-elements. Even the variations in width of
the cracks along the development of each of them
are evidenced, with the clear intention of representing
5 AETIOLOGY
The second part of the assessment is dedicated to iden-
tifying the causes responsible for the instability. The
authors formulate a graphic diagram to show how the
movements may have occurred. The cracks are inter-
preted as hinges around which the parts of stonework,
which is not cracked, considered non-deformed, have
rotated.The kinematical interpretation of the pattern of
cracks identified enables the three Mathematicians to
exclude a number of hypotheses formulated by others.
In the absence of cracks, which can justify them,
they exclude that the causes of the instability can be

253

Figure 4. Description of the deformation mechanism (1742,
Parere di Tre Matematici, Bibliotheca Hertziana, Rome).
attributed to the subsiding of the foundations. In sub-
stance, they think that the weight of the small dome,
the ribs and the double cap have weighed down caus-
ing the drum to move outwards. As for the iron rings
and possible damage, the three Mathematicians, still
with an elasticist mentality, claim that it would not be
possible to know if, and to what extent the metal rings
are truly effective. This is because they cannot be seen
directly as they are inserted into the stonework, nor is it
possible to know the tightening tensions. Moreover, the
iron could have suffered thermal deformations chang-
ing its ring strengths, while some of the rings could
even have broken.
6 PROGNOSIS
The assessment continues defining the process that
the three Mathematicians intend to use to assess the
quantity of actions associated with the kinematical
mechanism and the tensional state of the rings whose
scarce efficiency appeared to be responsible for the
instability.
In terms of scientific innovation this is the topical
moment of the assessment because the Three decide
to apply a process of calculation based on the PVW. In
fact, the instability refers to the excess weight that bur-
dens, above all, the upper parts of the dome, pushing
the drum outwards, and to the scarce ringing resistance
of the lower buttresses.
At this point the three Mathematicians consider the
results obtained by Philippe de la Hire and Couplet
(Benvenuto, 1981) concerning the behaviour of arches
and vaults. They come to the conclusion that two
254
causes are responsible for pushing the drum out-
wards: the weight of the small dome and the weight
of the ribs including the gores of the dome. Similarly
two horizontal forces create resistance: the force of the
rings and that of the support made up of the base, drum
and buttresses.
To assess the weights of the structures the three
Mathematicians weigh a mass of travertine and one
of stonework. Thanks also to the geometric data of
the single parts surveyed, they succeed in proving
that the total weight of the dome is equal to about
56 000 tons.
Whereas, to asses the force corresponding to the
iron rings the three Mathematicians appeal to the
treatise Coesione de’ corpi solidi by Pietro Ban Muss-
chenbroek deducing that the traction resistance of the
first ring corresponds to 114 t and that of the second
to 95.
After observing that the variation in the length of
a circular chain increases in the proportion of 2π in
relation to the variation of the radius, the Three apply
the PVW equalling all of the positive and negative
works made by the elements in play. The positive work
is produced by the weights of the macro-elements,
which represent the damaged parts of the dome and
small dome. The negative work is determined both by
the resistance opposed by the drum in contrasting the
deformations outwards, and by the resistance available
in the rings on the various levels. This latter work, how-
ever, is assessed incorrectly: as the concept of potential
elastic energy was not yet clear, it is calculated directly
considering the resistance on breaking.
After obtaining the values of the forces (thrust
and resistance), the problem of balance, however, is
handled strictly in terms of energy.
The process adopted by the three Mathematicians
to solve the problem, despite containing a number
of imperfections, is daring and modern. The Mathe-
maticians renounce the use of processes based on the
polygons of the forces, and refer to a method, the PVW,
mentioned previously by René Descartes in his prin-
cipal work published in 1637 (Descartes, 1637) and
subsequently perfected by J. Bernoulli.
In this way the three Mathematicians succeed in
proving that the weight of the small dome and the dome
exert a force H on the impost ring of the dome dis-
tributed as internal pressure p (H = 2πR × p), whose
total value results from the following relation, obtained
equalling the virtual works considered (Fig. 5):
where G indicates the weights of the small dome and
the portions of the dome, where v indicates the lower-
ings of their centres of gravity and where h indicates
the horizontal opening of the impost.
The resistance W, which contrasts the thrust H is
made up partly of the resistance exerted by the rings

Figure 5. Analysis of the balance dealt with by applying
the PVW.
Figure 6. Kinematical diagram proposed by Giovanni
Poleni.
(hence the need to calculate the force of these exerted
as radial pressure p) and partly of the horizontal resis-
tances with which the drum and the buttresses resist
overturning. The state of balance between W and H is
thus calculated by the three Mathematicians applying
the new principles of Mechanics. Thanks to this pro-
cess they succeed in calculating a missing thrust equal
in total to about three million pounds, in other words
to approximately 1000 tons.
255
Consequently, the cause of the instability is
attributed to this imbalance. The rings, which had
been laid during the construction are therefore unable
to contrast the pushing action of the structures. In
substance, the three Mathematicians conclude observ-
ing that the upper part tends to move inwards under
the action of the loads of the small dome, while the
lower part tends to move outwards developing traction
tensions in the rings.
In the light of the knowledge relating to the mechan-
ics of structures, developed later on, the adopted
process does not appear without imperfections. First
of all, the work carried out to dilate the rings, calcu-
lated as work of extending an equivalent straight rod,
is assessed considering a constant force applied from
the beginning of the elastic deformations, unduly asso-
ciating a size growing elastically with a static one.
Moreover, no reference is made to the work of elas-
tic deformation of the “macro-elements” according to
which the kinematism of the instability was examined,
nor of the anelastic deformations of the unbalanced
areas. On the other hand, as we have already seen,
a mechanical theory for structures had not yet been
developed to the point of being able to assess these
aspects correctly. Therefore, the daring choice of a pro-
cess of theoretical calculation applied for interpreting
mechanical phenomena remains significant.
7 THERAPY
Again in the light of the PVW, the need is confirmed
to find a solution so the thrusts acting outwards are
rebalanced by thrusts acting in the opposite direction
to guarantee the equilibrium of the whole structure.
To do this, the three Mathematicians suggest placing
additional rings, considering a safety coefficient equal
to two, justifying the reason and consequently showing
a typical engineering approach.
The possible solutions examined are divided into
three groups, according to which different alterna-
tive solutions are proposed: placing iron “struts” and
chains; walling up the spaces, which are currently
open, to strengthen the buttresses; eliminating the
structural loads where they are not needed. Of the three
groups the three Mathematicians prefer the first, or
rather the placing of new rings. To do this they refer to
the data analysed and to the numerical results obtained
again applying the PVW. This enables them to quantify
the number of interventions and optimise the position
of the reinforcing rings.
In response to the criticism raised by various experts
on the spiral staircase made in the four pillars by
Bernini, which allegedly also affected the statics of
the dome, on the strength of the results obtained with
a scientific process, the three Mathematicians claim
that it is not necessary to fill them in since the pushing

Figure 7. Master Nicola Zabaglia.
action of the drum is clearly less than that of the small
dome; therefore they can be preserved in the state they
still are today. However, they suggest filling in a num-
ber of spaces, which support the vaults and indicate
other rings to be inserted level with the drum.
Lastly, the three Mathematicians claim that all of
the other solutions they had heard, and which had
been suggested to them, were superfluous for resolv-
ing the static problem of the dome. The six rings alone
and the various careful interventions planned would
have undoubtedly contributed to improving the situ-
ation. In fact, the other solutions would have given a
load of approximately 950 tons in relation to a total
weight of the dome of over 56 000 tons, so with rather
insignificant added value.
The indicated project solution applying a scientific
process made it possible to propose work that was
not invasive, respectful of the Michelangelesque build-
ing and its decorations and aesthetics. The assessment
generated great controversy from well-known scholars
including Poleni himself, who had also been com-
missioned to make an analysis of the same problem
(Poleni, 1748; Baggio & Da Gai, 2000). Boscovich’s
work was later praised by C.L. Navier (Navier, 1829)
who recognised its originality.
The reinforcement work was carried out under
the supervision of Luigi Vanvitelli (Buccaro, 1988)
who applied the three Mathematicians’ proposal, and
Master Nicola Zagaglia was responsible for the organ-
isation of the site (Cosatti, 1743; Zander 1991).
The assessment described in this memorial is illus-
trated in detail in the book Dall’ingegneria empirica
verso l’ingegneria della scienza (Niglio, 2007).
256
8 CONCLUSIONS
The results of the assessment were presented by
Boscovich on 20 January 1743 and published in the
same year. The incident involving Saint Peter’s Dome
had repercussions on other situations, for example
the interventions carried out on the spire of Milan
Cathedral.
This marked the start of a debate on the relationship
between Architecture and Mechanics, between consol-
idated humanistic knowledge and new science, which
was destined to revolutionise the future of building
practices.
It was an important and daring step for the Three
Mathematicians, taken at a historical time in which
no other information of equal importance emerges. It
involved basing a whole expert analysis and the con-
sequent project proposals on the use of a scientific
principle of a purely theoretical nature, which was
completely innovative and not yet used in other real
situations.
REFERENCES
Rare and archival manuscripts.
Boscovich, R.G. & Al. 1742. Parere di tre Matematici sopra i
danni che si sono trovati nella cupola di S. Pietro sul finire
dell’Anno MDCCXLII. Dato per ordine di nostro signore
Papa Benedetto XIV. Venice (printed in 1943).
Cosatti, L. 1743, Contignationes ac pontes Nicolai Zabaglia.
Rome: Nicolò Paglierini
Descartes, R. 1637. Discours de la Méthode de bien conduire
sa raison et chercher la vérité dans les sciences; plus la
Dioptrique, les Météores et la Géométrie, qui sont des
essais de cette méthode.
Galilei, G. 1638. Discorsi e dimostrazioni matematiche
intorno a due nuove scienze attenenti alla mecanica et
i movimenti locali.
Poleni, G. 1748. Memorie istoriche della Gran Cupola del
Tempio Vaticano e de’ danni di essa, e de’ ristoramenti
loro divise in libri cinque alla santità di nostro signore
Papa Benedetto XIV. Padua.
Ackerman, J.S. 1968. L’architettura di Michelangelo. Turin:
Einaudi.
Baggio C.& Da Gai E. 2000. Tra differenza ed inno-
vazione: la meccanica in architettura, in G. Curcio
and E. Kieven (ed.), Storia dell’architettura italiana, Il
Settecento. Milan: Electa Mondatori.
Buccaro A. 1988, Aspetti della cultura tecnico-scientifica
in epoca vanvitelliana: dall’architetto allo scienziato-
artista, in Tecnologia Scienza e Storia per la con-
servazione del costruito. Florence: Callisto Pontello
Foundation.
Barrow, J. D. 1992. Perché il mondo è matematico, Rome-
Bari: Laterza Editore.
Beltrami, L. 1930. Relazione delle indagini e dei lavori di
restauro alla Cupola Vaticana dal maggio 1928 a marzo
1930. Vatican City: Vatican Printing Works.
Benvenuto, E. 1981. La scienza delle costruzioni ed il suo
sviluppo storico, Florence: Sansoni Editore.
Capecchi, D. 1999. Il Principio dei Lavori Virtuali da
Aristotele a Bernoulli, Naples: Luda Editore.
Capecchi, D. 2002. La storia del Principio dei LavoriVirtuali,
Benevento: Hevelius Edizioni.
Conforti, C. 1997. Lo specchio del cielo : forme signifi-
cati tecniche e funzioni della cupola dal Pantheon al
Novecento, Milan: Electa Mondadori.
Curcio, G. 2003. Il Tempio Vaticano 1694. Carlo Fontana.
Milan: Electa Mondadori.
Di Pasquale, S. 1996. L’Arte del Costruire tra conoscenza e
scienza. Venice: Marsilio Editore.
Navier, C. L. 1829. L’application de la Mécanique a
l’etablissement des constructions et des machines. Paris.
257
Niglio, O. 2007, Dall’ingegneria empirica verso l’ingegneria
della scienza. La perizia di tre Matematici per la Cupola
di San Pietro (1742), Padua: Il Prato Editore.
Rocchi Coopmans de Yoldi, G. 1996. S.Pietro. Arte e Storia
nella Basilica Vaticana. Bergamo: Bolis Editore.
Timoshenko, S. P. 1953. History of Strength of Materials.
London: McGraw-Hill.
Timoshenko, S.P. 1956. Scienza delle Costruzioni. Turin:
Andrea Viglongo & C. Editori.
von Halász, R. 1969. La prefabbricazione nell’edilizia indus-
trializzata. Milan: I.T.E.C.
Zander, G. 1991. Storia della Scienza e dellaTecnica Edilizia.
Rome: Multigrafica Editrice.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

From static history to restoration issues: The Gesù Nuovo church

in Naples (Italy)

S. Casiello, V. Russo & E. Vassallo

Department of History of Architecture and Restoration, University of Naples Federico II, Naples, Italy

ABSTRACT: The Gesù Nuovo church – the seat of the Society of Jesus in the capital of the Spanish viceroyalty –
presents the opportunity for a significant case-study related to the connections existing between static history
and present problems in the structural field. Archival documentation allows the reconstruction of the various
hypotheses being made, starting from the 17th century, in relation to the stability of the building. The research
highlights a historical structural “weakness”, referring, in particular, to one of the pillars in the crossing and
to the arches weighing onto it. Making a close study of the interpretations, given from the 18th century, by
royal engineers until recent years, the study aims at correlating the hypotheses expressed in the past with the
present structural conditions of the church. The “historical” issue of the cross damages is studied carefully in the
conclusions in order to emphasize the necessity of a well-considered integration between the analysis of present
conditions and a complex system of multidisciplinary skills, aiming at a full comprehension of the causes of
architecture instability.

1 INTRODUCTION
1.1 Ancient and new structural interventions
in the Gesù Nuovo church
One of the restorer’s main tasks is to make a close
investigation of the various phases of a building’s con-
struction and transformation. The knowledge of all
the previous consolidation or “embellishment” inter-
ventions, permits the planning of a restoration project
as compatible as possible with the conservation of an
artefact.
This research has been carried out on a monument
of considerable importance, the Gesù Nuovo church
in Naples. It aims not only at illustrating its history
but also at assessing the damage, caused by opera-
tions conducted during the past centuries, to which,
in different ways and using the technological skills of
the various ages, there has been an attempt to find a
remedy.
Even in this specific case the research has con-
firmed the conviction that each building’s consolida-
tion is absolutely not separate from the more complex
operation of restoration. This operation is based on
historical knowledge and uses all possible technical
investigations to reach a correct diagnosis of the build-
ing damage and to preserve, as much as possible, the
material with which it is constructed. Moreover, con-
sidering the static model as a testimony of a civilization
and a historical document, any project must be based
on a precise understanding of it.
The local Soprintendenza [regional board of the
Ministry of cultural and environmental heritage]
recently restored the four pillars, the tambour and
the arches of the church’s crossing. It is interesting
to notice that, over the centuries, crushes and dam-
ages have been interpreted using the best technicians
of each age. The first dome’s double calotte col-
lapsed after the 1688 earthquake and afterwards some
hypotheses were made about the characteristics of the
crush. The dome was then rebuilt with the plan of
the architect Arcangelo Guglielmelli with its weight
discharging onto the four 16th century pillars under
the tambour. In particular, the first pillar on the left
showed, over the centuries, serious symptoms of insta-
bility. Also on this occasion, the specialists, who were
the most qualified in each period, were engaged in a
debate about the causes of the damage.
Luigi Vanvitelli, leading the debate, hypothesized
that the cause of the damage to the left pillar and of
the complex cracking pattern of the church’s crossing
was due to the hurried repair of building damage after
the 1688 earthquake.
In conclusion, during the 18th century, two tech-
nical fronts were formed: one believed that damage
was due to the weight of Guglielmelli’s dome, the oth-
ers that they were due to the covering pressure on the
vertical walls.
The remedy consisted in the replacement of the
masonry dome with a light lathwork vault supported
by wooden trusses. Nevertheless, considering the issue

259
with a view that ignores the historical perspective, if
such a drastic solution, implying the destruction of
a significant constructive element of the monument,
is to be censured, the subsequent substitution can be
judged much more seriously. In 1973, the Genio Civile
(Ministry of Public Works) replaced the light lathwork
vault with a reinforced concrete one, weighing on the
masonry structures below.
In the period after World War II the use of reinforced
concrete became more and more frequent, especially
in structures that were disguised; this happened in the
Gesù Nuovo church where the inside of the concrete
cap had been coated with stucco. Nevertheless, in this
case, with great probability, the remedy has been the
cause of further disruptions for which, in recent years,
there has been an attempt to find a solution. Fortu-
nately, today the instruments allow technicians to make
diagnoses, which are also supported by numerical cal-
culations, that are much more precise than in the past;
still, as in this case, the attention and sensitivity of the
restorers must ensure that the proposed solutions do
not involve any more drastic substitution of parts. The
solutions should instead imply operations for struc-
tural improvement with the addition of elements which
are compatible with the ancient structures.
2 THE BUILDING OVER TIME.
CONSTRUCTION AND RESTORATIONS
BETWEEN THE 16TH AND 18TH
CENTURIES
2.1 The building site and the pre-existing
structures
The study of events of the construction of the Gesù
Nuovo church provides significant grounds for reflex-
ion referred to multiple topics which are related to
the construction and to the transformation of architec-
ture since the 16th century. The analysis of historical
sources suggests an articulated combination of inter-
pretations of damage and, consequently, of methods of
consolidation proposed during past centuries. At the
same time, the examination of the planning choices
which have been adopted in the substitution of parts –
the dome in the crossing, particularly – highlights how
the integration of lacking parts has been developed
during each historical phase with a sincere expres-
sion of the contemporary “state of technique” and the
culture of planning.
Since its establishment, the church of the seat of the
Society of Jesus in the capital of the Spanish vicereign
was planned in correspondence with the pre-existing
architecture of the 15th century palace of the prince of
Sanseverino (De Frede 2000).
External facades were re-used in Giuseppe Vale-
riano’s project along three sides of the building for
economic reasons and, moreover, for the recognition
260
of its urban significance and role. The vertical
perimetric structures of the palace were preserved
and underfoundated (Pirri 1970, p. 116, p. 121, note
58) while the excavation of the four crossing pillars
started in 1584, probably with the center of the cross-
ing planned in correspondence to the palace courtyard
(Schinosi 1706–1711, lib. V, cap. VI, p. 464).
The reconstruction of the building phases is facil-
itated by information given by the Jesuit chronicler
Giovan Francesco Araldo at the end of the 16th cen-
tury (Divenuto 1998), together with historical data
provided by documentary sources preserved in the
Jesuit archive in Rome (Pirri 1970; Bösel 1985, vol. I,
pp. 416–419). Through Araldo’s narration it is possible
to understand that the foundations of the four pil-
lars were realised between November 1584 and March
1585, reaching the “monte, et ferma pietra”, that is the
tuff in the subsoil (Divenuto 1998, p. 238). Reading the
Jesuit chronicler, we learn that the four pillars on which
the dome had to be placed were built in only three
months, from March to June 1585 (Divenuto 1998,
p. 240 and p. 243), covering their tuff basements with
piperno blocks. An interruption of approximately ten
years in the building of the church followed the estab-
lishment of the pillars so that the construction was
started again only in 1594 with work in the tribune
and in its side chapels (Divenuto 1998, p. 372); con-
struction times slowed down again in 1596 because
a serious blaze burned in the church and because of
the death of Valeriano in July. The construction of the
crossing was started again only when the tribune was
completed. The erection of the four arches discharg-
ing on the pillars – built about ten years before – and,
finally, of the bottom ledge of the tambour concluded
this building phase.
The church of Gesù Nuovo was inaugurated in
October 1601 (Errichetti 1963, p. 183, note 2) without
its central dome: about thirty years passed from the
inauguration to the beginning of the construction of
the Jesuit dome in 1629 (Errichetti 1963, p. 177).
2.2 The domes of the Gesù Nuovo church in the
seventeenth century
Although it is scarce, useful information about the
construction of the dome can be deduced from the
seventeenth century cartography and, moreover, from
the description of the structure and of its dimensions
given by the Neapolitain guide Carlo Celano in 1692.
The first dome of the Gesù Nuovo church had a tam-
bour that was marked externally with buttresses with
volutes (Figs 1 and 2). Two calottoes weighed on the
tambour and they were separated by an interspace
about seven metres high in correspondence with their
keystone; in addition, a staircase leading to the lantern
on top was constructed between the two vaults (Celano
1692, p. 44). The lantern, consisting of a cylinder
of masonry surrounded by eight grey tuff columns,

Figure 1. Naples, the Gesù Nuovo church in the map of B.
Stopendael (1653).
Figure 2. Naples, plan of Gesù Nuovo church and Professed
House (17th century). The drawing shows the double calot-
toes structure of the first dome of the church (Bösel 1985,
tav. 271).
balustrades and vessels, produced a significant weight
on the lower calottoes; as we learn from Celano, the
lantern was the first element to crack shortly after its
completion so that it required the replacement of a
stone column with a brick one.
According to the above mentioned seventeenth cen-
tury author, this intervention caused a “weak point”
which was at the origin of the collapse of the Jesuit
double dome, which occurred during the earthquake of
1688. Therefore, such a hypothesis, empirically tend-
ing to highlight the role played by the lantern in the
collapse of the dome, had already been advanced in an
“official report of damages” produced approximately
one month after the earthquake; in fact, this document
stressed that “the dome fell beginning to crumble from
the lantern”, striking in its collapse the lateral chapels
of the transept (Cantabene 2004, p. 58).Through ref-
erence to the documentary sources, it is possible to
suppose that the seismic tremors caused the loss of the
two vaults and that of the lantern, while they spared the
seventeenth century tambour which remained largely
in place.This fact is confirmed by the description of the
dynamics of the collapse provided by the Neapolitan
architect Giuseppe Lucchese in 1708; this technician,
engaged in the analysis of crackings of the double
dome of the Treasury of St.Gennaro Chapel in the
261
cathedral of Naples in the same year (Russo 2007),
refused any hypothesis of a fall “on perpendicolo”
of the Gesù Nuovo dome as had been advanced by
contemporary “mathematicians”; on the contrary, he
emphasised the rotation suffered by the vaulted parts
and by the lantern towards the west transept side. As
the report of Lucchese makes clear, the four arches,
the pendentives and the tambour remained “firm and
intact” (ATSG, A/22, foll. 66v–67); a circumstance
that would not have arisen, according to Lucchese, in
case of a collapse only in a vertical direction.
The Jesuits attended to the repair of the tambour,
to the construction of the transept left vault again and
to closing the crackings with lime and bricks in only
six months (Errichetti 1963, p. 177). Thereafter, the
collapsed parts were rebuilt following the plan of the
Neapolitan architect Arcangelo Guglielmelli between
1692 and 1693 (Amirante 1990, p. 247 and p. 334).
As a drawing of 1769 clarifies (Sasso 1856, tav. 13;
Errichetti 1963, p. 178) (Fig. 3), a single vault con-
cluded by a smaller lantern was built in order to replace
the complex double dome of the beginning of the cen-
tury (Fig. 4); its weight loaded, of course, on the pillars
and arches dating, as said before, back to the end of
the sixteenth century.
2.3 From the interpretation of structural damage to
planning choices. Debate and construction in
the eighteenth century
Just over sixty years after the rebuilding of the vault
and of its lantern, the dome by Guglielmelli and, in
particular, the first pillar on the left required the atten-
tion of Jesuits because of a worrying cracking pattern.
A monitoring of this pillar was undertaken in 1767
using “swallow tails of marble” (Carrafiello 1995,
p. 358) in correspondence with the above-mentioned
pillar; at the same time, a first consultation with the
royal architect Luigi Vanvitelli was carried out in order
to have suggestions about appropriate remedies (Di
Stefano 1973, p. 232).
The expulsion of the Jesuits from the Kingdom of
Naples, which occurred in the same year, and the sub-
sequent settlement of Reformed Franciscans in the
church caused an interruption in the control of pro-
gression of the crackings and in Vanvitelli’s task. Only
from 1769 onward did the attention of engineers and
religious focus again on the conditions of the build-
ing as the survey of 1769 by the architect Giuseppe
Mauro demonstrates (Fig. 3). Beginning from 1769
the most accredited Parthenopean technicians of the
second half of the eighteenth century were involved in
the issues of the church. Among these, in particular, a
decisive role was played in the choices by Ferdinando
Fuga, who replaced Vanvitelli as the trusted techni-
cian of the Franciscans. Considering “azzardosissima”
the solution of the rebuilding only the cracked pillar,

Figure 3. Naples, the Gesù Nuovo church. Front view of the
second dome (1692–1693) drawn by the architect Giuseppe
Mauro in 1769 (in Sasso 1856–1858, table 13).
Fuga proposed in 1769 a thickening of the structures
of the four pillars in the crossing, of the upper arches
and, in addition, of the remaining pillars and arches
of the church resorting to “sottarchi” and “contropi-
lastri” (Errichetti 1963, pp. 178–179). This proposal
would have greatly expanded the resistant sections of
the construction but, at the same time, it would have
altered its space and the perception of the chapels and
the tribune as provided for in the design of Valeriano.
The contradiction between eurythmía and the rea-
soning behind firmitas led, therefore, the Franciscans
and the Royal Secretariat to require the formation of
a team of experts in order to solve the problem; a
committee composed of the Neapolitan professional
intelligentsia, with Fuga himself, and with Mario Giof-
fredo, Giuseppe Pollio, Giuseppe Astarita, Pasquale
Monzo, Lorenzo Iaccarino, Felice Bottiglieri and
Berardo Galiani, was formed in November 1769 in
order to understand the causes and to propose a
262
Figure 4. Naples, the Gesù Nuovo church. Hypothesis about
the longitudinal section before the demolition of the second
dome (from Guerra 1967, Fig. 5).
solution to the structural problems of the church
(Errichetti 1963, pp. 178–179). During 1770 and the
first half of next year, therefore, two “schools of
thought” were gradually emerging through the direct
examination of the conditions of the building; the first
one can be considered much more “interventionist”,
while the other one was characterised by a more con-
servative approach. As we have seen, Fuga may be
considered the exponent of the first approach, so that
he would even propose the demolition of the aisles
of the church, justified by economic reasons, in the
following years (Errichetti 1963, p 181).
With respect to the issue, the royal architect Mario
Gioffredo, planner and director of the construction of
the complex dome of the Spirito Santo in the same
years, would stand definitely on a more moderate front.
Opposing Fuga’s ideas, he supported the hypothesis of
the recourse to only a partial substitution of the dam-
aged pillar, avoiding more invasive interventions in
this way. The position of Gioffredo coincided, with the
appropriate distinctions, with the interpretations – in
opposition to Fuga, as well – made by a young member
of the technical Neapolitan class, Vincenzo Lamberti
and, equally, by Berardo Galiani, author of the com-
mented translation of Vitruvius De Architectura in
1758. The first, a young engineer who published the
Voltimetria retta in 1773, ascribed the damage of the
pillar and the dome to the actions of underground
waters; he hypothesized an underpinning with arches
near to the left pillar and a consolidation of the pillar
with iron plates (Di Stefano 1973, p. 233). Lamberti,
exponent of an emerging “theory of vaults” in the
Neapolitan context, showed himself to be significantly
abreast of studies conducted in other parts of Europe
in the report he presented on that occasion (Guerra
1967, pp. 392–393); this, in particular, was in relation
to the equilibrium of vaults and arches, as the quota-
tions from de la Hire and Belidor demonstrate. This
theoretical and cultural background was not enough,
however, to avoid the answer, articulate and pragmatic,
that Luigi Vanvitelli presented to the Secretariat of the
Royal House in 1772 (Di Stefano 1973, pp. 233–234):
sharing the idea of intervention of Ferdinando Fuga
and removing the issue from “mathematical” interpre-
tations, Vanvitelli brought the problem of the Gesù
Nuovo church back to the direct examination of the Figure 5. Naples, the Gesù Nuovo church. The longitudinal
phenomena.The architect attributed the main causes of section after 1786 (from Guerra 1967, fig. 6).
the pillar’s cracking and the causes of the damage still
underway to the seismic tremors and to the excessive
speed of past “restorations”.
The position taken on the issue by Galiani seems
equally interesting: a meditated approach to the con-
structive and dimensional characteristics of the build-
ing emerges from the report submitted by him in 1774,
especially when the technician highlighted the exis-
tence of planning “repentances” in correspondence to
the cross. He reported, in particular, the existence of
a system of brick arches that had been constructed
higher than the four tuff arches and connected with
them through masonry plugs of poor quality. Galiani,
like Lamberti and Gioffredo, excluded the possibility
of an excessive load transmitted from the top dome
on the four pillars and he rather attributed the dam-
age to the thrusts of coverage structures on vertical
walls (Carrafiello 1995, pp. 355–373). Despite the
above described opposing positions, the majority of
technicians involved in the question ended up attribut-
ing the damage underway to the excessive weight of
the dome of Guglielmelli and to the poor quality of
the construction; so, the demolition of the tambour
and the higher vault was proposed in 1774. Started
only in 1786, the intervention was carried out under
the direction of the royal engineer Ignazio Di Nardo
and it involved the replacing of the seventeenth cen-
tury structures with a lighter “incannucciata” bowl
(Fig. 5), sustained by a system of upper wooden beams
(Errichetti 1963, pp. 181–183). Figure 6. Naples, the Gesù Nuovo church. The iron scaffold
Di Nardo’s intervention concluded, so, a significant used to build the fourth dome (1972) (ANSI).
phase in the history of the church’s domes, character-
ized by the progressive reduction of covering masses
made in 1973 by the Ministry of Public Works,
and, consequently, of weights.
through its local office of Genio Civile (Fig. 6). The
intervention was carried out to remedy the progressive
deterioration, aggravated by the effects of the 1961
3 THE FOURTH DOME OF THE CHURCH
earthquake, of the 18th century fake lathing vault (Iap-
pelli F.S.I. 1993). At a closer look the present dome,
3.1 Damages and interventions of consolidation
covered with decorative stuccoes in late-baroque taste,
The Gesù Nuovo “dome” that today dominates the is a plate of reinforced concrete leaning on a circular
crossing of the church is the result of a substitution masonry tambour whose base is inscribed in a square.

263
 As an application of the widespread theories of that
moment on the use of reinforced concrete shells, the
“camouflage” solution adopted by the Genio Civile
complied with both structural and formal require-
ments. There was a foretaste of this choice in the
postwar institutional culture that looked favourably
on solutions that recompose architectural elements
of buildings highly compromised by war damage, –
with particular regard to the replacement of wooden
roofing – making use of the innovative technology of
suitably disguised reinforced concrete.
In a 1949 letter the Soprintendente of Campania
Monuments, Antonino Rusconi, informing the Jesuit
rector of the church about the conditions of the wooden Figure 7. Naples, the Gesù Nuovo church. The concrete
truss above the dome, calls for a review of the struc- truss in the project of 30th october 1987 (ASBANa 8/158,
ture of the cover and also, for a replacement of the Ministry of Public Works. Lavori di ripristino urgenti alla
chiesa del Gesù Nuovo danneggiata dagli eventi sismici del
lathing vault with a masonry or reinforced concrete
23/11/1980 e 14/02/1981).
one (ASBANa, Gesù Nuovo 8/158, draft n. 5846 of
the 9/12/1949). Therefore, concern about the condi-
tion of the dome is closely connected, as early as the
first post-war period, to the conditions of the wooden
trusses of the coverage.
At the same time, symptoms of masonry crushing
begin to cause worry when Rusconi brought to the
attention of the authorities to the fact that, in the accu-
mulated delay in the repair of the war damages, the
pillars of the crossing were showing a phenomenon of
expulsion of marble coatings (ASBANa, 8/158, letter
of 21/06/1952). The Genio Civile officer in his turn
(ing. L. Persico) highlighted the necessity to uncover
the pillars and, over all, to proceed to the restoration of
the dome to its former condition. This “ripristino” was
truly urgent, given the detachments of large plaster and
stucco pieces from the present dome’s structure, which Figure 8. Naples, the Gesù Nuovo church. Hand sketch of
was detrimental to the public safety, and given also the the covering above the choir (ASBANa, ing. Ettore Minervini
progressive degradation of the wooden truss. Report s.d.).
Although the cover replacement in 1973 has clearly
increased the heaviness on the walls, in the light of (Fig. 7), in addition to the spherical cover, would soon
recent investigations regarding the structure, which are show signs of its presence and, although in the past
discussed below, it has been observed that, during the the walls had already shown symptoms of a lack of
earthquake of 1980, the concrete cover played a joint cohesion, in the few years after the construction of the
function of the masonry structures in response to the new coverage, worrying signals alarmed the experts
pressures of the undulatory waves (ASBANa, Report involved.
of ing. Ettore Minervini, 20/09/1999). In 1993, a study group, appointed by the Soprinten-
In 1987, the condition of the rotting beams and of denza Generale agli Interventi Post Sismici in Cam-
the secondary frame of the roof above the crossing pania e Basilicata to draw up an executive draft
suggested the replacement of the wooden roof with a aiming at the recovery and restoration of Monumental
reinforced concrete one. Concrete slide walls, 25 cm Public Heritage of Naples Historical Center, indi-
thick, are projected above the cover to provide the cated a pattern of damage, due to crushing problems,
connection between all the longitudinal stringcourse bringing to light the need, impossible to postpone,
attaching at the roof ’s impost. The walls are configured to systematically investigate the static conditions of
at the intrados according to the shape of the vault that the elevated masonry walls (ASBANa, Gesù Nuovo
covers the crossing and at the extrados according to the 8/158, 01/09/1993, Min. BB.AA.CC., Progetto esec-
pitched roof. These walls lean on a re inforced concrete utivo per il restauro della chiesa del Gesù Nuovo).
stringcourse whose section was to be, according to the Five years later both the central organs of the Ministry
plan, 130 cm × 80 cm (ASBANa, Gesù Nuovo 8/158 (ASBANa, 14/10/1998, letter of Central Inspector
project). The weight reached by the new structure arch. R. Di Paola) and the local Soprintendenza

264
(ASBNa Report No. 103 of 20/5/1998) admitted the
urgency of a plan of investigations aimed at under-
standing the overall structural situation of the church
with particular reference to the elements that showed
signs of instability: the four pillars of the crossing, the
dome’s pendentives and the tambour.
These surveys, consisting in horizontal and inclined
logging, endoscopic investigations, non-destructive
investigations on the walls using flatjacks, topographi-
cal surveys with electromagnetic waves for the spatial
measurement of cracks and empty spaces, with the
return of data collected using surface georadar (GRP),
showed that the four pillars, despite a constructive
symmetry between them, were characterized by a
remarkable diversity of the composition of walls and
by a lack of continuity, with special anomalies found in
the south west pillar (ASBANa, Georadar report Tec-
noin srl). Along with these investigations the engineer
Ettore Minervini, entrusted by the local Soprinten-
denza, performing the necessary calculations on the
structures of the crossing, determined the tensions of
medium compression to which pillars are subjected:
• Total weight of the structure weighing on the four
pillars: 13,398 t;
• Media-resistant surface of the pillars: A. tot
4 × 4 × 4 mt = 64 square meters;
• Pressure-average = 20.9 kg/cm2 . (ASBANa, ing.
E. Minervini report s.d.)
Combining the diagnostic results with structural
calculations, engineer Minervini pointed out that the
pillars, though endowed with a certain wall thick-
ness, had a resistant surface reduced only to the outer
layer of masonry, an outer peel of “tufo-pipernoide”
(tuff-piperno), 50 cm thick; while the inner core of
masonry bag showed a low degree of cohesion. The
intervention, aimed at stemming an obvious crushing
phenomenon that was underway, had become, at the
end of the twentieth century, urgent.
The occasion of a project for a pedestrian cross-
ing below the church provided the opportunity to
investigate the church underground. The results of sur-
veys conducted by the Soprintendenza were reassur-
ing about the compactness of foundational structures
and allowed the exclusion, as a concurring cause, of
any possible sinking of the foundations (ASBANa,
29/09/2003 Geognostic investigations, Tecnoin srl).
The solution reached by the Soprintendenza’s tech-
nicians aimed to pursue a widespread consolidation
of the building, which is now historicized as a mixed
structure of masonry and concrete, at the same time
improving the masonry response to the weight that
the enormous coverage of the dome exercises on ele-
vated walls. A system – of which ample use was made
in other parts of the structure affected by crushing –
of armed injections with mortars injected at low
pressure was used in the four pillars of the crossing.
265
The question, also raised at the ministerial level
(arch. R. di Paola) of whether these injections could
have the effect of burdening the nucleus of the pilaster
rather than consolidating it and risking an increase in
stress compression, has conducted to the use of very
fluid mortars compatible with the masonry materials,
also able to exclude chemical reactions with the salts
present in masonry themselves and avoid harmful reac-
tions to the marble cladding (ASBANa, 20/05/1998,
Report n. 103). This first intensive work on the pil-
lars was followed by operations aimed at consolidating
the trabeation above the St. Ignatius chapel, above the
nave on the side of the Moscati chapel and the tambour
under the “dome” (ASBANa report n. 93, 26/05/1999).
During the execution of these operations, signs of
crushing have been noticed at the keys of the arches of
the transept, of the aisle, of the apse together with any
other secondary arches. In addition, the lower edge of
the tambour, once perfectly horizontal, was now like a
wavy line with the lowest points on the keys of these
arches (ASBANa Report n. 38 3/4/2000). Therefore,
in the further static intervention of consolidation, on
the basis of what has already been done on the pil-
lars, executive priority has been given to saturation of
cracks as a preparatory operation.
3.2 The conservation of existing constructive
systems between innovation and tradition
Currently, the work of the Soprintendenza is focus-
ing on the saturation of cracks that span the entire
vault on the main aisle. In this case the operation has
been preceded by a consolidation of the wooden roof
(ASBANa Report n. 109, 14/06/1995) which shows
how, in recent years, the attitude is to prioritize the
permanence of traditional structural systems through
consolidation. This aims at the preservation of the role
of structural components of the architectural building.
It is worthwhile to point out that the intervention
carried out on the coverage correspondent to the choir
also goes in the direction of the conservation of pre-
existing constructive systems. The constructive system
produced up to now, while showing clear signs of a
worrying history of damage, is material proof of a
local building tradition that is also rich in historical
significance. The planned coverage is made up of a
sequence of parallel walls shaped according to the
trend of the pithed roof. Each septum is composed of
a masonry arch that is slender compared to the larger
mass of the triangular wall itself, which, departing
from the underlying barrel vault, discharges its weight
on the raising walls and partially on the same vault’s
buttresses.
Due to a arch shaped central opening on the top of
the central triangle, laid upon the arch below, worry-
ing cracks and disconnections have showed at the top
and on the sides walls. In order to ease the burden of
the cover above these walls, which are about 20 metres
apart, and to help the relative stability avoiding trau-
matic interventions, supporting structures have been
projected. In particular, two Polonceau metal profiles
in contact with the opposite sides of each of the tri-
angular walls, properly connected, will be burdened
with the coverage weight and they will support the
cemented points of the same walls as well. These metal
structures will unload the entire weight on two string-
courses in reinforced concrete, shaped so as to assist
the masonry structures and to contain a channel for the
disposal of rainwater. The mechanical properties of the
steel profiles made this solution acceptable, because
they are relatively thin with respect to the high mass
of wall that they support, with the function they serve
in full view (ASBANa Report n. 134 del 21/10/1996).
The interventions that this review briefly describes
constitute the testimony of how, over the last 15
years, thanks to experiences that have shown their
ineffectiveness over time, the path to improving the
preexistent structural condition has been undertaken,
trying to avoid the wishful thinking of radical substi-
tutions. At a time when the architectural heritage is
subject to a close evaluation in terms of seismic vul-
nerability we expect to submit these choices to a safety
assessment, as required by the recent “Guidelines for
the evaluation and reduction of seismic risk to cultural
heritage” approved in Rome on June 15, 2007.
4 CONCLUSIONS
Through the examination of past and recent interven-
tions on a complex architecture, the authors aim at
focusing on present conservation issues of the Gesù
Nuovo church in Naples. Whereas in past centuries
the solution to structural problems tended toward the
massive substitution or elimination of parts, questions
of consolidation must take into consideration today a
“constructive requirement”; this means the necessity
of preserving a diachronic architectural system must
be taken into consideration. As described in the paper,
traditional and “modern” solutions contribute to the
mentioned requirement in a unique way.
REFERENCES
Celano, C. 1692. Notitie del bello, dell’antico, e del
curioso della citta di Napoli. . . In Napoli: G. Raillard.:
giorn. III, 44.
Schinosi, F.S.J. 1706–1711. Istoria della Compagnia di Giesù
appartenente al Regno di Napoli. Napoli: Stamperia
Michele Luigi Mutio: lib. V, cap. VI: 464.
Sasso, C.N. 1856–1858. Storia de’ monumenti di Napoli e
degli architetti che gli edificavano. . . Atlante. Napoli: Tip.
F. Vitale.
Ceci, G. 1898. Il palazzo dei Sanseverino principi di Salerno.
Napoli nobilissima VII: 81.
Ceci, G. 1921. Sul rifacimento della cupola della Trinità
Maggiore. Napoli nobilissima II (5): 92.
266
Montini, R.U. 1956. La chiesa del Gesù. Napoli: Azienda
Autonoma di Soggiorno Cura e Turismo.
Errichetti, M. 1963. La cupola del Gesù Nuovo. Napoli
nobilissima II (5): 177–184.
Pane, R. 1966. Antico e nuovo (Incredibilia sed vera:
la ricostruzione della cupola del Gesù Nuovo). Napoli
nobilissima 4: 161.
Guerra, G. 1967. La cupola del Gesu Nuovo: problemi statici
e curiosità storiche. Atti dell’Accademia Pontaniana 16:
384–398.
Pirri, P. 1970. Giuseppe Valeriano S.I. architetto e pittore,
1542–1596. Roma: Institutum Historicum S.I.
Di Stefano, R. 1973. Vanvitelli ingegnere e restauratore. In
AA.VV. Luigi Vanvitelli. Napoli: Edizioni Scientifiche
Italiane.
Errichetti, M. 1974. La chiesa del Gesù Nuovo. Campania
Sacra.
Bösel, R. 1985. Jesuitenarchitektur in Italien (1540–
1773). Wien: Verlag der Österreichischen Akademie der
Wissenschaften.
Amirante, G. 1990. Architettura napoletana tra Seicento e
Settecento. L’opera di Arcangelo Guglielmelli. Napoli:
Edizioni Scientifiche Italiane.
Iappelli, F.S.I. 1994. Committenza, danni, restauri. Societas
3: 68–74.
Carrafiello,T. 1995. Berardo Galiani intendente d’architettura
(1724–1774). Archivio Storico per le Province Napole-
tane. CXIII: 245–380.
Schiattarella, A. & Iappelli, F. S.I. 1997. Gesù Nuovo.
Castellammare di Stabia: Edizioni Eidos.
Divenuto, F. 1998. Napoli l’Europa e la Compagnia di Gesù
nella «Cronica» di Giovan Francesco Araldo. Napoli:
Edizioni Scientifiche Italiane.
De Frede, C. 2000 (ried.). Il principe di Salerno Roberto San-
severino e il suo palazzo in Napoli a punte di diamante.
Napoli.
Cantabene, G. 2004. Il terremoto del 1688 a Napoli. Nuove
acquisizioni documentarie. Ricerche sul ‘600 napole-
tano: 58.
Casiello, S. (ed.). 2005. Le cupole in Campania. Indagini
conoscitive e problemi di conservazione. Napoli: Arte
Tipografica Editrice.
La Regina, F. 2005. Il tema delle cupole nella conservazione
del patrimonio architettonico. Questioni di metodo per
la conoscenza e l’intervento. In Stella Casiello (ed.), Le
cupole in Campania. . ., : 107–114.
ABBREVIATIONS:
ANSI: Archivio Napoletano Societatis Jesu.
ASBANa: Archivio Soprintendenza ai Beni Architet-
tonici e ambientali di Napoli e Provincia;
ATSG: Archivio Cappella del Tesoro di San Gennaro,
Napoli.
∗
Although the present paper is the outcome of a col-
lective work among the authors, par. 1.1 is due to Stella
Casiello, par. 2.1, 2.2 and 2.3 to Valentina Russo and
par. 3.1 and 3.2 to Emanuela Vassallo. References and
conclusions are collective.
∗∗
A special thank goes to arch. M.R. Crescenzio
(Soprintendenza per i beni Ambientali e Architet-
tonici di Napoli e Provincia) whose contribution to the
documentarial research of Part 3 has been determinant.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Building techniques of the Zeytinburnu Military Factory built in

the 19th century

Aynur Çiftçi & Nadide Seçkin

Department of Architectural Restoration, Faculty of Architecture, Yıldız Technical University, Istanbul, Turkey

ABSTRACT: The Zeytinburnu Military Factory, a large complex, was built with the purpose of modernizing
the army as a result of the reform movements of 1826. The necessary materials and machines for the factory
whose construction is known to have continued between 1845–1846, were brought from London by Barutçubaşi
Hoca Ohannes Dadyan Efendi. British engineer William Frin was responsible of the construction and assembly
of the factory. The factory was established on a wide area between the Sea of Marmara and the railroad. A map
dated 1918 shows the factory buildings and contains information about their functions. Eight buildings/building
groups of the factory are still extant. They show variation in the roof structures and their supports. Three of
the 4 cranes in one of the buildings were removed. The factory buildings that house original examples of the
period’s construction techniques must first be documented in detail and then measures must be taken for their
conservation.

1 INTRODUCTION during the reign of Sultan Abdülmecid (r. 1839–1861).

From the 15th to the 19th centuries, a large portion of
the arms and other military supplies for the Ottoman
army were manufactured in the various military work-
shops and factories located in Istanbul, the capital
and the seat of the 1st army. The reforms made in
the Ottoman military organization within the scope
of the modernization movement as of the 18th cen-
tury increased with the abolition of the “Kapıkulu
Ocakları” (troops under the sultan’s command) in 1826
and buildings with various functions were built to sat-
isfy the requirements of the new army. First of all
3 military factories, namely the “İplikhane-i Amire”
(Imperial Yarn Factory), “Feshane” (Fez Factory) and
“Beykoz Deri Fabrikası” (Beykoz Leather Factory),
were established to manufacture cloth, clothes and
shoes for the soldiers. These were followed by the
Zeytinburnu military factory built in the mid-19th cen-
tury. All these military buildings reflect the Ottoman
architectural influences of the “Westernization” period
from the point of view of their architectural forms,
construction techniques and style characteristics. The
Zeytinburnu Factory stands out as an exclusive
example with the variety and originality of its
construction techniques.
At the onset, the factory was to be established
in Izmir but later it was decided to constructed it
in European style and in Zeytinburnu. Although the
exact start date of the construction is unknown, we
know that construction was ongoing in 1845–1846
Monsieur Hetk, a manufacturer who came from Eng-
land, Hoca Garabet, the palace master builder and
Rıza, an attendant worked in the construction of the
factory and the appliances, equipment and machines
for the construction were brought from London by
Barutçubaşı Hoca Ohannes Dadyan Efendi. William
Frin, an English engineer was responsible of the set-
ting up and assembly works of the factory (İlgürel,
1989).
Citing from the writings of Mc Farlane, an English
traveller who was in Istanbul during the construction
of the factory, Tuğlacı states that some of the numer-
ous craftsmen, workers and engineers brought from
Europe left work because they could not get money.
Tuğlacı also states that many workers were killed when
the factory’s chimney built by Garabet Balyan accord-
ing to the incorrect calculations of the British experts
collapsed because Balyan’s warnings were overlooked
(Tuğlacı, 1993).
The masonry blast furnace, an important unit of
the factory, and the foundry were completed in 1846
and smelting of iron ore brought from the mine in
Büyükada started (İlgürel, 1989). The “Eser-i Hadid”
built according to the plans of Phillips, a British marine
engineer, by the Dadyan Brothers in this factory in
1848 was the first home built armoured ship (Tuğlacı,
1993).
As a result of the political crisis in Europe in 1848,
the factory made a loss in the first years. The factory’s
activities increased after 1850 and steam engines,

267
pumps, equipment for small enterprises and parts
for guns were manufactured in addition to military
equipment (Anonim, 1995).
In the second half of the 19th century the factory
was transformed into a solely arms and ammuni-
tion manufacturing plant and at this time the factory
known as the “Zeytinburnu Silah Fabrikası” (Zeytin-
burnu Military Factory) manufactured goods for the
“Tophane-i Amire” (Imperial Canon Factory). Goods
manufactured at the factory were presented at the
Sultanahmet exhibition in 1861. A technical school
in connection with the factory was established at
this time (Anonim, 1995). Grooved canons and 140
firearms of different size and make manufactured at
this factory were exhibited at the “Istanbul Interna-
tional Exhibition” of 1863 (Önsoy, 1988). Special iron
posts for street lighting, iron pipes, ploughs, gardening
tools, drawgear were also manufactured at the factory
(İlgürel, 1989).
Production at the factory gradually decreased due to
its being far from a rich mine, its inefficient operation,
competition with Europe and the political problems of
the state and it was closed down during World War I
and the armistice period (İlgürel, 1989). The factory
went back to operation in 1936 and was turned over to
the “Ordonance” class of the army in 1948. Today it
provides service as the “1st Army Maintenance Centre
Commandership” and the maintenance and repairs of
army equipment as well as the production of some
equipment is executed here.
2 ARCHITECTURAL CHARACTERISTICS
The Zeytinburnu Military Factory is located over a
large area of 128.121 m2 in the west of the city
between the Sea of Marmara and the railroad line. The
Prime Ministry Ottoman Archives has various docu-
ments related to the construction, repairs and use of
the factory buildings for the period between 1261–
1334/1849–1916. The German Union Map (“Alman
Sendika Haritası”) dated 5 October 1918 shows that
the factory consisted of a number of detached and
adjacent buildings (Fig. 2). The captions on this map
contain information on the different units of the time
and their functions. However, no information could
be obtained about the original plans of the mentioned
buildings.
Eight buildings/building groups of the factory have
survived to the present. Some of the buildings in the
factory area as well as the quay by the sea and the boat
landing no longer exist (Fig. 3). The quay was filled
in during the 1950–1960s, and the factory’s connec-
tion with the sea was cut off with a seaside road. The
buildings that have survived have to a large extent pre-
served some of their original façade characteristics and
construction techniques.
268
2.1 Building no. 1
The building shown as the “muhafaza bölüǧü
koǧuşları” (guards’ division barracks) on the map
dated 1918 is located next to the entrance gate of the
factory area. The single storey rectangular building has
round headed windows and doors. The building with
plastered and painted façades is covered with a hipped
roof. The wall thickness of the solid block masonry
building varies between 65 cm and 87 cm.
2.2 Building no. 2
The structure shown as the “büyük havuz” (large tank)
on the map dated 1918 is located at the west of the
factory area. It consists of a superstructure with a cir-
cular plan scheme over an almost square base. The
base section contains a wide round headed opening,
which is used to enter the lower level section of the
structure. The base of the solid block masonry struc-
ture is in alternating courses of ashlar and brick. The
top section with a circular plan scheme is in brick. On
the façade, there is moulded dressing between the base
and upper section. There are 154 cm long metal ties on
the façades of the base. The structure is now used as a
water tank (Fig. 5).
2.3 Building group no. 3
According to the map dated 1918, this group of build-
ings located in the north-south direction in the centre
of the factory area consists of 5 adjacent one-storey
units. From north to south, the first unit housed the
“bakır hâdde-hânesi” (copper rolling mill) and the
“şâh-merdân ocakları ve buhâr kazganları” (pile driver
furnaces/drop forges and steam boilers), the 2nd unit
housed the “demir hâdde-hânesi” (iron rolling mill).
Part of the east facing section of this unit is demol-
ished. The 3rd unit contains the “buhâr kazganları”
(steam boilers), “pirinç kılıç-hânesi” (brass sword ate-
lier) and the “maden döküm-hânesi” (metal foundry).
The 4th unit has the “kılıç-hane” (sword atelier),
“buhâr kazganları” (steam boilers) and “kundak fab-
rikası tesviye-hânesi (gunstock works fitting shop) and
the 5th unit houses various functions defined as the
“büyük çark-hane” (major machine shop) and “serî’
atışlı kundak fabrikası” (repeat fire gun stock shop).
Today, the dimensions of units 1–3 are
27.70 × 71.20 m and they are divided into two spaces
with a wall extending in the north south direction.
The main walls of the solid block masonry units are
114–118 cm thick. The roofing with wooden struc-
ture rests on the main walls. The windows and doors
are round headed. The space in the east section con-
tained 4 cranes of which only one exists today (Fig. 1).
The 4th and 5th units in solid block masonry
are 49.60 × 52.75 m and have a wall thickness of
115–126 cm. In this group of buildings only the space
 “kılıçhane” (sword atelier) in the west direction is cov-
ered with a masonry barrel vault. Tie beams have later
been placed in the vault that has windows on its east
façade. The other units are covered with wooden roof-
ing supported by wooden posts with a cross section of
36 × 35 cm (Fig. 6). The windows and doors of these
units are round headed and the façades are plastered
and painted.

2.4 Building group no. 4

This building group located at the south of the fac-
tory area extends in the east west direction and
consists of one-storey adjacent units (Fig. 4). In the

Figure 3. The sea side buildings of the factory (IRCICA

Figure 1. The crane in the building group no. 3 (2001). Archive).

Figure 2. The German Union Map (“Alman Sendika Haritasi”) dated 5 October 1918 (Factory Archive).

269
Figure 4. The building group no. 4 (IRCICA Archive).
Figure 5. “Large tank”, Building no. 2 (2001).
map dated 1918 these units are designated as “şarjör
ve mukavvâ kutu dâiresi, kovan birinci ameliyyât
dâiresi, kovan ikinci ameliyyât dâiresi, dinamo dâiresi,
kurşun ve kovan-hâne dâiresi, imlâ dâiresi, çinko
sandık dâireleri, fişek fabrikası, kurşun ve gömlek
i’mâlât-hânesi, kurşun presleri, kapsül i’mâlât-hânesi”
(charger and cardboar box shop, cartridge case initial
operations shop, cartridge case secondary operations
270
Figure 6. Wooden posts in the building group no. 3 (2001).
shop, dynamo shop, bullet and cartridge case work-
shop, explosives shop, zinc can workshops, cartridge
factory, bullet and jacket workshop, bullet presses,
cap/primer workshop).
The first section of this building group that can be
divided into two is located on the southeast side and
is 16.50 × 133 m and the second section which is on
the north-west side is 38.25 × 95.25 m. Some units
Figure 7. Wooden truss roof structure in the building group
no. 4 (2001). Figure 8. Detail of an “I” shaped steel beam and wooden
beam in the building group no. 4 (2001).
consist of large single volume spaces adjacent and
parallel to one another. The façades of the buildings
with depressed arched doors and windows are plas-
tered and painted. Photographs in the archives show
that the tile covered units in the southeast have roof
windows placed so as to catch the northern light.
The main wall thickness of the group of buildings
in solid block masonry varies between 50 and 96 cm.
The measurable bricks of the wall are 10 × 22 × 6 cm.
In various units of the structure there are load bear-
ers with cross sections of 20 × 35 cm, 24 × 24 cm and
26 × 27 cm and cast-iron columns with a diameter of
21 cm. Wooden and metal truss roof structures are
used in the units with gable roofs (Figs 7–12). The
cracks formed on the walls of the buildings during the
Istanbul earthquake of 1999 were repaired and covered Figure 9. Metal truss roof structure in the building group
in 2003. no. 4 (2001).

2.5 Building (Hamam) no. 5

The hamam, a detached unit at the south end of the fac-
tory area, consists of the changing rooms, trepidarium,
caldarium and the water tank. The halvets (very hot
cubicle) in the trepidarium and caldarium are covered
with vaults while the caldarium is covered with domes.
The façades of the solid block masonry hamam are
plastered and painted. The vaults and domes are cov-
ered with lead. The changing rooms were built at a
later period.

271
Figure 10. Wooden truss roof structure and wooden posts
in the building group no. 4 (2004).
Figure 11. Wooden truss roof structure and cast-iron
columns in the building group no. 4 (2001).
2.6 Building no. 6
This is a rectangular 2 storey building located in the
south section of the factory area. It has round headed
windows. The wall thickness of the main walls of
the solid block masonry building is 83–88 cm at the
ground floor and 67–73 cm on the floor above. The
floors of the building were repaired in reinforced
concrete.
2.7 Building no. 7
According to the map dated 1918, this building in
the east of the factory area houses workshops and
offices such as the “marangoz-hâne fen me’mûrluğu
dâiresi, mermî-i mütenevvia fen me’mûrluğu dâiresi”
(carpenter’s shop construction registrar’s office, mis-
cellaneous cartridges engineering office). On both
floors, the original plan scheme of the 15.90 × 17.10 m
2 storey building consists of a central hall and the vari-
ous spaces opening onto this hall. There is an elevated
entrance accessed by a wide strait stairs on the north-
west of the building. The entrance gate is flat arched
272
Figure 12. Cast-iron columns with a diameter of 21 cm in
the building group no. 4 (2001).
and the windows of both floors have depressed arches.
The building that has moulded eaves dressing is cov-
ered with a tiled hipped roof. The main walls thickness
of the solid block masonry building is 50–60 cm on
both floors. Bricks of 10 × 25 × 5.5 cm have been used
for the walls. The original floors and stairs are of wood.
The interior partition walls have wooden bearing piles
and plastering wooden lathing.
2.8 Building no. 8
The building in the east of the factory area has been
designated as the “intizar mahalli” (waiting area) in
the map dated 1918. The windows of the single story
rectangular building are round headed. The hipped
roof building is plastered and painted. The main wall
thickness of the solid block masonry building is 77 cm.
3 CONCLUSIONS
The Zeytinburnu Military Factory is the only exam-
ple of an important industrial building that was built
in the 19th century to manufacture a wide range of
military equipment that has survived to the present.

Figure 13. Metal window frames in the building group no. 4
(2001).
The factory’s boat landing on the Marmara Sea shore
and the railroad that was built by the northwest of
the factory in the end of the 19th century were used
to transport the raw goods and products. The factory
buildings that have survived to the present are still used
as the maintenance workshops for military equipment.
Unfortunately, no conservation decision has been
given for the factory buildings that have survived and
whose connection with the sea was terminated with
the building of the shore road. The only measured
drawings for documentation purposes are those of
a building with an administrative function (Building
No. 7).
The observations and evaluations made on the exist-
ing factory buildings within the scope of this work
have remained limited as the said buildings are being
used for military purposes. Though information on the
functional distribution of the factory buildings can be
attained from the map dated 1918, an absolute eval-
uation has not be possible as the original plans and
documents could not be reached.
273
The factory reflects the modern and traditional
construction technology of the 19th century with its
original wood and iron structures, masonry technique
and materials and its production activities. Therefore,
it is a group of buildings that should be conserved for
their value for the archaeology of industry. Changes
and additions were made in the functions of the build-
ings with time, therefore, the buildings now contain
different materials and construction details. Metal roof
structures as well as timber structures dating of the last
quarter of the 19th century are of major importance. As
mentioned earlier, an important number of the factory
buildings where the materials and structural systems
of the period were used to their maximum potential
are still standing and functioning.
The factory built to manufacture a large portion of
the arms and military equipment for the Ottoman army
is also of a documentary nature with the equipment in
its various buildings. Therefore, the buildings as well
as the equipment they contain must be documented
in detail. The expansion of the factory, the traditional
and modern manufacturing methods, assembly and
operation schemes, as well as the machines, work-
benches, tools and equipment imported from Europe,
the foreign experts, foremen and administrators who
have worked at the factory, the relations established
with the Krupp and Karlsruhe factories in Germany
are all matters that must be explored in the archive
documents.
The existing factory buildings have preservation
problems arising from changes of function, earthquake
and environmental factors. The factory buildings that
have retained their original construction techniques to
a large extent and the original equipment in the build-
ings must be repaired before they lose their originality,
and must be reviewed within the context of sustainable
preservation.
REFERENCES
Anonim. 1995. Zeytinburnu Silah Fabrikası. Dünden Bugüne
İstanbul Ansiklopedisi, 8: 488. İstanbul: Kültür Bakanlıǧı
ve Tarih Vakfı Ortak Yayını.
Çiftçi, A. 2004. 19. Yüzyılda Osmanlı Devleti’nde Askeri
Mimari ve İstanbul’da İnşa Edilen AskeriYapılar. İstanbul:
Yıldız Teknik Üniversitesi Fen Bilimleri Enstitüsü
Doktora Tezi (yayınlanmamış).
İlgürel, M. 1989. Zeytinburnu’nda Bir Demir Fabrikası. Tarih
Boyunca İstanbul Semineri 29 Mayıs-1 Haziran 1988
Bildiriler, İstanbul: Edebiyat Fakültesi Basımevi.
Önsoy, R. 1988. Tanzimat Dönemi Osmanlı Sanayii ve Sanay-
ileşme Politikası, Ankara: Türkiye İş Bankası Yayınları.
Tuğlacı, P. 1993. Osmanlı Mimarlıǧında Balyan Ailesinin
Rolü, İstanbul: Yeni Çıǧır Kitabevi.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Investigation of the 1716 Algiers (Algeria) earthquake from

historical sources

A.A. Abdessemed-Foufa
Department of Architecture, Faculty of Engineering Sciences University of Blida, Algeria

D. Benouar
Department of civil engineering, Faculty of Civil Engineering, University of Algiers, Algeria

ABSTRACT: This research work presents the consequences of the 1716 Algiers earthquake. The damages
records are collected from the ottoman files, consular mails, scientists and travellers of the 18th century. The
1716 Algiers earthquake is one of the most significant historical event having affected the city (I = IX MSK).
Indeed several sources describe it and give significant information about the scope of the damage recorded. These
entire documentaries sources and in particular those of Ottoman regency were carefully studied and analyzed
allowing us to have supplement information on this earthquake. This research work makes it possible to have
a clear vision on the damage as well as on the vulnerability of the predominant constructions in Algiers. The
information obtained, constitute an excellent damage database which will make enable us to work out the future
seismic scenarios to protect cultural heritage of the old nuclei of Algiers and eventually other historical sites
with the same characteristics.

1 INTRODUCTION give a robust way of differentiating the way in which

buildings may respond to earthquake shaking. The
1.1 The structure of the EMS-98 scale Vulnerability Table is an attempt to categorise in a
manageable way the strength of structures, taking both
The Macroseismic Scale used for the classification
building type and other factors into account.
of the damages degrees on traditional buildings of
The MSK scale defined building classes by type
Algiers dating from the 18Ith century is the European
of construction as a simple attempt to express the
Macroseismic Scale (EMS-98) with twelve degrees.
vulnerability of buildings. Accordingly, six classes
This scale takes in charge historical earthquake. The
of decreasing vulnerability are proposed (A to F)
EM-98 intensity scale, like the MSK scale which pre-
of which the first three (A, B and C) represent the
ceded it, is one of a family of intensity scales which
strength of a “typical” adobe house, brick building
originated with the widely used simple ten degree scale
and reinforced concrete (RC) structure they should
by Rossi and Forel; this was revised by Mercalli, sub-
be compatible with building classes A to C in the
sequently expanded by Cancani to twelve degrees,
MSK-64 and MSK-81 scales. Classes D and E are
and then defined in a very full way by Sieberg as
intended to represent approximately linear decreases
the Mercalli-Cancani-Sieberg (MCS) scale. It is this
in vulnerability as a result of improved level of earth-
scale which forms the starting point not only for the
quake resistant design (ERD), and also provide for
MSK/EM-98 scale, but also for the numerous versions
well-built timber, reinforced or confined masonry and
of the “Modified Mercalli” scale. The major difference
steel structures, which are well-known to be resistant
between the EM-98 scale and other intensity scales
to earthquake shaking. Class F is intended to repre-
is in the detail with which different terms used are
sent the vulnerability of a structure with a high level
defined at the outset, in particular, building types,
of earthquake resistant design (see annexes).
damage grades, and quantities.

1.2 Building types and vulnerability classes 1.3 Damage grades

The European Macroseismic Scale incorporates a The way in which a building deforms under earthquake
compromise, in which a simple differentiation of the loading depends on the building type. In this work
resistance of buildings to earthquake generated shak- the type of structure is the masonry with bricks and
ing (vulnerability) has been employed in order to stones. This structure represents the A and B classes of

275
vulnerability. As a broad categorization we can group
buildings in five grades.
Grade 1: Negligible to slight damage (no structural
damage, slight non-structural damage).
Grade 2: Moderate damage (slight structural damage,
moderate non-structural damage)
Grade 3: Substantial to heavy damage (moderate
structural damage, heavy non-structural damage)
Grade 4: Very heavy damage (heavy structural dam-
age, very heavy non-structural damage)
Grade 5: Destruction (very heavy structural damage)
(See annexes).
1.4 Quantities
The use of quantitative terms (“few”, “many”, “most”)
provides an important statistical element in the scale.
It is necessary to confine this statistical element to
broad terms, since any attempt to present the scale as
a series of graphs showing exact percentages would be
impossible to apply in practice and would destroy the
robustness of the scale. The definition of quantity has
been presented, very deliberately, in graphical format
to emphasise the way these numerical categories are
blurred rather than sharply defined. In such a case as
a precisely determined quantity falls into an overlap-
ping area, the user should consider the implications
of classing it as one category or the other, in terms of
which would be more consistent with any other data
available for the same place.
1.5 Assessing intensity from historical records
The term “historical data” is frequently used to
mean descriptions of earthquake effects from his-
torical records, that is, written sources prior to the
instrumental period (before 1900). Historical accounts
often report in detail damage to special monumental
buildings (castles, churches, palaces, towers, pillars,
mosques, and so on). Less frequently do they report
the effects on ordinary buildings, which are the only
ones which can be used within the framework of the
scale. With regard to ordinary buildings, the vulnera-
bility classes of traditional houses range in most cases
from A to B, even to C and D (wooden structures. With
regard to ordinary buildings, the vulnerability classes
of traditional houses range in most cases from A to
B, even to C and D (wooden structures). Very little
is known from the general literature about building
types in Europe and Maghreb up to the 17th century,
except for the obvious facts that people used the mate-
rials nearest to hand, and that the richer the owner, the
better-built and better-maintained his house was likely
to be. But in the Middle Ages, certainly, most houses
in many parts of Europe were made of wood, and in
the Maghreb they were made of stone and adobe brick.
276
Figure 1. Map of Algiers (After anonymous 1830).
2 THE MEDINA OF ALGIERS AT 1716
In 1716, Algiers showed the typical characteristics
of a medieval Moslem Maghreb cities such as Fez
(Morocco) and Tunis (Tunisia) (Le Tourneau 1949).
Indeed, the blocks are of various sizes and contain a
several number of houses. The parcels are fully built;
the houses are overlapping and leaning against each
other forming a compact unit. In addition, a great
number of narrow and short streets were covered by
galleries on top of which the houses extended and thus,
created roofed passageways called “Sabat”. In Algiers,
just a main street crossed the city from the south to the
north in the lower part of the city. Algiers was sur-
rounded by walls built before 1517, was pierced by
five gates with twelve batteries (figure 1).
3 TYPOLOGY OF CONSTRUCTIONS AT THE
18TH CENTURY
According to various historical sources (Haedo 1578–
1580, Shaw 1808 & De Grammaye 1670) in Algiers,
all houses look alike, most of them have a central
patio or square courtyard surrounded by rooms. This
type of house is called “Dar”. A gallery supported
by columns with horseshoe pointed arches surrounds
this courtyard called “wast al dar”. A great number of
the palaces and houses have three levels. Houses with
courtyard develop another type of dwelling, which is
called “Dar shebak”, house with metal grid, where
dimension of the courtyard is reduced. This grid rep-
resents only a tiny vertical opening for daylight. Fur-
thermore, we can find houses called “Alwi”, because
of constraint of the site giving narrows plots. So, they
do not neither “patio” nor grid. As a consequence, this
houses open towards external sance. These are gener-
ally situated on the border of the streets (figure 2). The
number of houses was estimate at 5000 (De Paradis
18th century) and are about 800 today (Cellule Casbah
 Niveau Stah
Niveau Shin
4
6

5
10

17
17
4

17
6

17
16
6

7
6

7
17
4

5
17

17

17
17

Niveau Wast al-Dar

Sedda
Niveau S'hin

17
7

17
17

4
8

10
8

17

6
4

9
2
8

6
15
1

6
7

7
4
18

7
17

17

Sedda
17

Sedda

al q'bu (chambre à encorbellement). 5- Ghurfa

1- S'qifa (entrée). 2- was't al dar (patio). 3-Bit (cjambre). 4- Bit
(magazin). 9- Mat'bakh (cuisine). 10- Bit al ma'
(chambre). 6- S'hin (galerie). 7- Druj (escaliers). 8- Makh'zan
17- Sedda (demi niveau ou espace à mi hauteur).
(latrines). 15- Dukkana (banc en maçonnerie). 16- S'tah (terrasse).
18- Hanut (boutique).

Plan 2.3.3 - Typologie des maisons d'Alger (Missoum 2003)

Figure 2. Plan of Algiers traditional houses (after Missoum
2003) and axonometric view of a block (after Atelier Casbah
1980).
2003). Religious buildings as mosques, military and
government buildings are well built (Devoulx 1875). (Gru/enthal 1998) with reference to ordinary build-
The vertical structure of buildings is made up of ings. The vulnerability of monumental buildings, as
masonry walls with a thickness varying from 10 cm the religious one, can not be easily assessed in terms
for the poorest quality to 90 cm and more for defence of vulnerability classes of EMS-98 intensity scale, as
structures. The masonry walls are built out of brick the classes are conceived with reference to ordinary
masonry bound by a lime mortar (figures 3, 3a & 3b). buildings.
The horizontal structure as floors and terraces were In general, it is well known that mosques as impor-
out of squared or not squared wood. All houses are tant monument can be very vulnerable.
covered with terraces and they all are bleached with In the case of Algiers, the importance of such build-
lime (figures 4). ings may imply that some care should be take for
their construction, suggesting vulnerability of classes
D which represents the reinforced masonry.
4 CONSTRUCTION ELEMENTS OF THE
DAMAGE
5 DAMAGE DISTRIBUTION
Traditional vertical structure, built mainly to support
their own weight as well as the weight of floors and The earthquake of 3 February 1716, which caused the
terraces (case of Algiers) and at the most to resist to largest seismic catastrophe from the Mitidja plain to
the wind action. These structures are particularly vul- the Cheliff plain (southeast of Algiers), was studied
nerable to seismic effects when they are not bonded by Ambraseys and Vogt (Ambraseys & Vogt 1988).
horizontally. In the other hand, the shape, size, weight French and Foreign British archives and Sources for
and materials of horizontal structure have a great seismological purpose were used in that study.
influence on the level of seismic damage. The historical documentary sources on earth-
As a general rule, the vulnerability of the common quake effects in Algiers are spread among different
constructions (houses) can be assessed in term of vul- archives and libraries (French, British, Simancas and
nerability classes A and B of EMS-98 intensity scale Ottoman).

277
 3 cm
3 cm

3
3

3
3
Brick

Lime Mortar

Figure 3b.

55 40
Section
0 10 20 50 cm
40
55

Figure 4. Floor made of wood.

Figure 3. Masonry typology.

The ottoman files, refers to ruinous state of the city,
this sources describing the damage in a useful way to
quantify it are, unfortunately, dispersed in the different
registers (financials and fiscal records, domanial and
successional acts). There is not any published report
containing details on damaged buildings.
Sources available so far do not detail mapping of
damage, as performed for instance in the case of some
cities in Italy (Guidoboni & Boschi 2001, Moroni
et al. 1999). No official on historical detailed damage
survey exist in Algeria.
Actually, such surveys are usually found when
the intensity does not exceed 9 MCS, that is, when
a significant part of damaged buildings can be
repaired.
February 3, 1716, in the middle of morning, at
9 h 45 mn (local time) a destructive earthquake shook
Algiers and its adjacent regions. Houses collapsed
and many others were damaged in the city, even
the country houses suffered considerable damages
Figure 3a. and some of them collapsed (Ottoman Archive 1716,

278
Anonymous manuscript no date, French Archives
1716, Gazette de France 1716, Comelin 1720 &
Ibn Redjeb 1740). Not only the poorest houses but
also solid construction, such religious one as the
mosques (Great Mosque of Algiers, Sha’ban Khudja
mosque) suffered extensive cracking of walls and par-
tial collapse of cupola, therefore they needed to be
repaired in 1735 (Devoulx 1870 & Barges 1877) and
palaces as Dar Aziza where the first floor collapsed,
and the great palace where walls suffered extensive
cracking.
On February 26, a violent commotion, damaged
what resisted to the first shock.
For this earthquake, Roussel (1973) have assigned
an intensity I = X MM, while Ambraseys and Vogt
(1988) have assigned that this intensity could not be
reached because no solid construction was damaged
and they give I = 7 MSK. In this study, we assess inten-
sity I = 9 EMS, on the basis of the assumption that the
houses stock and the mosques are respectively equally
distributed between vulnerability classes A and B and
vulnerability classes D.
An anonymous map (Algiers 1830), in absence of
any map dating from the 18th century, can be used as
a reliable reference for the earthquake situation.
6 DAMAGE TO MINOR BUILDINGS
(HOUSES)
Damage description for the earthquake of February 3,
1716 is available for 240 minor buildings (houses)
and 4 monumental buildings (mosques and palaces).
For all these constructions, they are detailed and con-
cerned damage caused by this earthquake. All of
them have been classified in term of EMS-98 damage
grades.
According to the various historical sources (cited
above) and in particular, those of the ottoman files,
various damages due to earthquake were emphasized
and they were of three types:
6.1 The total collapse of the houses
According to the analysis of the ottoman files (doma-
nial acts 1716) [8], Comelin (1720), Delphin (1922),
Shaw (1808) and Carette (1850), 200 houses collapsed,
most of the dwellings were ruined and part of the city
was thrown to the ground. According to Tassy (1830)
and Burzet (1866–1869), the country houses or houses
of the fahs around Algiers collapsed completely at a
distance of about 3 km around the city. Carette and
Burzet reported that the aftershock of 26 February
added damaged to most of the houses, which did not
collapse totally after the main shock.
The grade of damage 5 was assessed.
279
6.2 Destruction of the walls
According to the ottoman files (1716), many houses
were damaged and repairs were conducted.
The grade of damage 3 was assessed.
6.3 Rupture of floors
According to the ottoman files (1716), related to the
cadastral acts, it was revealed that many first floors
of the houses collapsed following the earthquake.
Comelin (1720) reported on the event: “. . .The house
of the ambassador of France was one of the most beau-
tiful of Algiers. It did have three floors before the last
earthquake, now only two floors remain”.
The grade of damage 4 was assessed.
7 DAMAGE TO MONUMENTAL BUILDINGS
(MOSQUES AND PALACES)
There is no detailed report on damaged monumental
buildings. The information is consigned in the reg-
isters of the religious buildings (claim of real estate).
These registers defer all the operations of maintenance,
which took place regularly on the damaged buildings
by the catastrophes (earthquake, floods, war. . .). It
result from it that only two damaged mosques by the
1716 earthquake had been described:
The great Mosque of Algiers (Djama’al kabir) pre-
sented several cracks on the north-west walls or the
“qibla” walls (Devoulx 1946), and the reparations took
place only in 1732 (Barges 1877).The grade of damage
2/3 is assessing.
The cupola of Sha’ban Khudja mosque collapsed.
The grade of damage 3/4 is assess.
It was said that many palaces were destroyed or
severely damaged. The first one related is “Dar ‘Aziza”
palace, which was the French Embassy. The Consul
Clairembaut (French Archive 1716) and later Comelin
(1720) described the damage. All the higher floor of
this large house felt down. The damage was thus signif-
icant in this residence. The degree of damage assessed
was 4. The second one is “Dar al Sultan” called “Al
Djenina” palace, which was at that time the residence
of the regency and the government palace. Accord-
ing to Klein (1937), this palace suffered many cracks
that have constrained the “Dey Ali Shaush” to leave
precipitately the palace with its court to find a pro-
visional refuge in “Burdj Mulay Muhammad” called
later “fort de l’étoile”. The grade of damage 2/3 is
assessing.
Synthetic information: collapsed, destroyed, several
cracks, referring to cumulative effects of the 1716
earthquake, are available for the remaining 240 houses,
2 religious buildings and 2 palaces.
 Table 1. Assessment and damage grade of minor building After the 1716
Algiers earthquake.

ID. ID.
Type of Slightly Destruction Several Total Total
damages damages of floors cracks collapse damages

Grades 2 4 3/4 5
Houses 68 24 15 8 130
Houses 86∗ 13∗ 9∗ 7∗ 115
Total 162 37 24 14 245
% 67.5% 15.40% 8.33% 5.40%
∗
indicate the houses, which has been not located in the map

Table 2. Assessment and damage grade of important building after the 1716 Algiers earthquake.

Slightly Damages Important Damages Destruction of the

Important Cracks in the Several cracks first floor Total Collapse
Buildings walls Grade 2/3 Grade 4 Grade 4/5 Grade 5

Great x – – –
Mosque of
Algiers
Shaban – x – –
Khudja
Mosque
Dar Aziza – – x –
Palace
Dar al Sultan x – – –
Palace

Zone 1

- 10
- -
?
?

?
Zone 2
39

-
?

- ?
-
Zone 6
37

?
Zone 5
?

8
?
?

-
Zone 3 Zone 4
19 2

Area 1
10 houses damaged
Area 2
39

Area 3
19

Figure 5.

280
 D am age Grade EM S-98
5
4/5
4
3/2
D am ages not located or without inform ation
inform ation

Figure 6. Recorded damages of the 1716 Algiers earthquake.

Damage grade 5 is assigning when the description
clearly reported “total collapse”.
Grade 4 is assigning when the sources simply
reported “collapse”.
Grade 2/3 is assigning when the sources reported
“slightly damage”.
The summary of damage is reported in Tables 1
and 2.
The identification of the buildings is performing on
the historical map of Algiers (figure 5).
This data, though not useful for detecting possible
amplification or desamplification areas, can provide
useful informations on the vulnerability of the serv-
ing buildings and serve as an input for evaluating the
ground shaking.
Annexes:

8 CONCLUSION

The severity of Algiers’destruction caused by the 1716

earthquake probably explain the absence of historical
sources, e.g. in the form of damage surveys, that would
allow us to map the distribution in the whole city. The
interpretation of the available source (ottoman files)
describing the damage, limited to the houses which
were managed by the religious institution, must take
into account that it is only a partial interpretation of
the effects of this great earthquake.
The study of Roussel (1973), comes to the con-
clusion that the intensity assessed in Algiers con-
sidering damage is X MM. On the other hand the
study of Ambraseys and Vogt (1988), suggesting
an investigation of the seismicity of the region of
Algiers, concludes that it is improbable that the inten-
sity reaches X MM because no great building was
damaged.
Whereas, in this study the intensity assesses is
I = 9 EMS-98 because of the damages of important
buildings.
Damage to 240 minor buildings and 4 important
buildings was assessed in term of EMS-98 intensity. Damages Classification after the EMS 98.

281

Quantities after the EMS 98.
Damages classification after the EMS 98.
REFERENCES
Ambraseys, N et Vogt, J. 1988. Materials for the investiga-
tion of the seismicity of the region of Algiers. European
Earthquake Engineering; 3; 16–29.
Anonymous manuscript (without date and title), (in Ara-
bic), manuscript n◦ 1637, National Library of Algiers;
Unpublished sources.
Bargès, J.J.L (Abbé). 1877. Complément de l’histoire des
beni Zeiyan, rois de Tlemcen. Paris. Ed Ernest Leroux,
Burzet, (Abbé) Histoire des désastres de l’Algérie, 1866,
1867, 1868 ; 1869.Alger, Imprimerie CentraleAlgérienne.
Carette, A.E.H. 1850. Algérie. L’univers ou histoire et
description de tous les peuples de leurs religions, mœurs,
coutumes. Paris, Ed Firmin Didot Frères.
Comelin, P. 1720. Voyages pour la rédemption des captifs aux
royaumes d’Alger et de Tunis.
282
Consular mail 1716. French Consuls Correspondence at
Algiers 1642–1792, Foreign Affaires B1 120, 1716–1720
F◦ 3-4-5 février 1716, F◦ 6-7- 21 février 1716, F◦ 8-9-10
mars 1716, Unpublished sources.
De Gramaye, J.B. (1670). Alger XV–XVII siècles. Journal de
Jean Batiste Gramaye Evèque d’Afrique. Translated by
Ben Mansour A.H. 1998. Paris. Ed Cerf.
De Haedo, D. (1578–1581). Topographie et histoire d’Alger.
Translated by Monnereau & Berbrugger. Transcripted by
Rebahi. A. Alger. Ed G.A.L.
Delphin, G. 1922. Histoire des Pachas d’Alger de 1515 à
1745, Extrait d’une chronique indigène, Extrait du Journal
Asiatique (Avril-Juin 1922 et Janvier-Mars 1925), 216–
217 ; Paris. Imprimerie Nationale.
Devoulx, A. Alger. 1875. Etude archéologique et
topographique sur cette ville aux époques romaines (Ico-
sium), arabe (Djazâir Beni Mezghanna) et turque (El
Djazâir). Revue Africaine; 19 ; 295–332 and 385–428.
Devoulx, A . 1846. Histoire d’Alger. Manuscrit n ◦ 3213. f ◦
120. National Algerian Library.
Devoulx, A . 1870. Les édifices religieux de l’ancien Alger.
Alger. Ed Typographie Bastide.
Gazette de France 1716. Press.
Grüenthal,G. 1998. L’Echelle Macrosismique Européenne.
Cahier du Centre Européen de Géodynamique et de
séismologie. Ed Française A. Levret.
Guidoboni, E et Boschi E. 2001. Catania terremoti e lave del
mondo antico alla fine del novecento. Instituto Nazionale
di Geofisica e Vulcanologia, SGA Storia Geofisica Ambi-
ente. Bologne. Ed Compositori.
Ibn Redjeb, H. 1740. Chronicle in arabic, translated by
Delphin, G, 1922.
Klein, H. 1937. Feuillets d’El Djazair. Alger. Chaix.
Moroni A, Azzaro R, Stucci M, Barbano M.S Albini P.
1999. Damage in Catania on the occasion of the 1693
earthquakes.The Catania Project, Earthquake damage sce-
narios for a high risk area in the Mediterranean; GNDT.
Pessina. Ed E Facioli and V. 119–125.
Ottoman Archive in the National Archives of Algiers. Unpub-
lished sources.
Ottoman Archives in the CAOM of France. Unpublished
sources
Roussel, J. 1973. Les zones actives et la fréquence des séismes
en Algérie 1716–1970. Bull. Soc, Hist. Natur ; Afrique du
Nord ; 64 ; 3 ; 2.
Shaw, T. 1808. Travels or observations relating to several
parts of Barbary and the Levant. Edinburgh. Printed by
J.Ritchie.
Tassy, L. De. 1830. Histoire d’Alger et du bombardement de
cette ville en 1816. Paris.Librairie Piltan.
Tourneau (Le) R. 1949. Fès avant le protectorat. Etude
économique et sociale d’une ville de l’occident . Publié
par l’Institut des Hautes Etudes Marocaines.
Venture De Paradis, J-M. 1983. Tunis et Alger au XVIIIXème
siècle. Mémoires présentées par Joseph Croq. Dijon. Ed
Sindbad.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

The Morro da Queimada Archaeological Park, Ouro Preto, MG – Brazil

B.T. Oliveira
Instituto do Patrimônio Histórico e Artístico Nacional – IPHAN, Ouro Preto, MG, Brazil

ABSTRACT: This paper deals with the creation of the Morro da Queimada Archaeological Park in Ouro Preto,
MG – Brazil. The Morro da Queimada constitutes an invaluable archaeological site, being an excellent sample of
the first architectonical typologies in town, for it keeps registers of the gold exploitation period at the beginning
of the 18th century and the vestiges of one of the most dramatic moments in the Brazilian colonial history. This
project is coordinated by IPHAN and is supported by national and international institutions, at the municipal,
state and federal levels, as well as religious bodies, NGOs and community associations. It was formally proposed
by the Museu de Arte Sacra do Carmo, based at the Paróquia de Nossa Senhora do Pilar and was approved
by the Conselho Nacional de Incentivo à Cultura – CNIC (National Committee for the Promotion of Culture),
under the auspices of MinC – the Federal Ministry for Culture of Brazil.

1 INTRODUCTION

On April 18, 1881, during one of his many visits to

Ouro Preto, Dom Pedro II found an important archae-
ological site on Morro da Queimada (Burnt Hill).
The Emperor, erudite and widely-traveled, wrote in
his journal that these ruins reminded him of Pom-
peii, the Roman city buried by the eruption of Mount
Vesuvius in 79 AD, discovered in 1748 and today a
famous archeological site, visited by specialists and
tourists the world over.
The archaeological site of Morro da Queimada sim-
ilarly has its origin in tragedy. Also known as “Morro
do Paschoal” or the Village of Ouro Podre, it was
one of the earliest settlements in Ouro Preto. It was Figure 1. The Execution of Felipe dos Santos, July 1720
destroyed in 1720, following the revolt led by Felipe Reconstruction by Antônio Parreiras (Museu Antônio Par-
dos Santos against the increased taxes imposed by the reiras, Niterói, RJ, Brazil).
Portuguese Crown through the prohibition of the cir-
culation of gold powder and the establishment of royal
minting houses in what was then the Capitania of São Dr. Manuel Mosqueira da Rosa and his son, the Frei
Paulo and Minas Gerais. Ouro Preto, then known as Vicente Botelho, and even the Frei Francisco do Monte
Vila Rica, rebelled during the night of the 28th and Alverne. According to the historian Diogo de Vascon-
the 29th of June, 1720. The situation deteriorated, and celos, Felipe dos Santos, was “the only truly popular
on July 2nd, 1720, more than two thousand rebels leader” (Vasconcelos, D., p. 182 ) (Fig. 1). The upris-
marched on the Vila do Carmo, today Mariana, with ing lasted around eighteen days. On the 16th of July,
the intention of pressuring the Governor to repel the the governor, the Count of Assumar, entered Vila Rica
measures of the Portuguese Crown. At this point, the at approximately eleven o’clock in the morning at the
movement had already spread to other communities head of fifteen hundred men.
such as Sabará and Mato Dentro which went on to After its destruction by fire on the orders of the
support the revolt. Count Assumar, the area was renamed Morro da
The principal leaders of the revolt were Pascoal Queimada and the population transferred to the nearby
da Silva Guimarães, Sebastião da Veiga Cabral, Arraiais.

283
Figure 2. Aerial view of Morro da Queimada in the
mid-decades of the 20th century (IPHAN archives, Ouro
Preto, MG, Brazil).
According to Diogo deVasconcelos, Felipe dos San-
tos was arrested on either the 19th or 20th of July,
received a summary judgement and was executed. The
other heads of the revolt were transferred to Rio de
Janeiro and there condemned to exile in Lisbon, Por-
tugal. At the time, the Governor, the Count of Assumar,
stated that even the very stones of the Arraial of Ouro
Podre plotted against the Portuguese Crown.
As a result of the Sedição de Vila Rica in 1720, the
establishment of the proposed Royal minting houses
was postponed for four years, and Minas Gerais
became an administrative region independent of the
Capitania of São Paulo.
Morro da Queimada today constitutes a priceless
archaeological site, a concrete record of the first urban
architectonical typologies of Minas Gerais, evidence
of the gold rush at the beginning of the 18th cen-
tury, and a remnant of one of the most dramatic
moments in the history of Colonial Brazil. Besides
the ruins of old buildings, in the area may be found the
remains of the many gold mines that dot the hillside:
hollowed shelters carved out of rock, long galleries,
ventilation shafts, and entrance tunnels. Still found on
this archaeological site are mundéus, reservoirs con-
structed for the washing of ore, small dams, water
diversion channels, and hydraulic systems used for
transporting slurries of water and mud mixed with gold
(Fig. 2).
2 THE DETERIORATION OF MORRO
QUEIMADA
In recent decades, Morro da Queimada as an archae-
ological site has become corrupted due to the chaotic
284
Figure 3. Irregular occupation of Morro da Queimada
(IPHAN archives, Ouro Preto, MG, Brazil).
growth of the City of Ouro Preto. In the absence of
urban planning, the site was occupied, the ruins being
used by the population for building material, and the
surviving structures employed as foundations for new
constructions (Fig. 3). The lack of protection for the
archaeological remains on Morro da Queimada in
Ouro Preto constitutes one of the more serious cases
of neglect of cultural heritage by the various levels of
government.
In April, 2003 a technical mission from UNESCO
on a visit to the city identified the preservation of the
archaeological site at Morro da Queimada as one of
the measures necessary to halt the alarming deteriora-
tion of the cultural and environmental heritage of Ouro
Preto.
3 THE RECOVERY PROJECT
Following this warning, the Instituto do Patrimônio
Histórico Artístico Nacional-IPHAN (National Insti-
tute for Historical and Artistic Heritage) established
work groups (historical research, property issues,
Housing, physical planning, and community rela-
tions), necessary for the preparation of a pilot project
for an archeological park in the area.
This project sought to create a substantial and posi-
tive impact on the City of Ouro Preto by means of the
following:
– expansion of historical research and the creation of
programs for archaeological excavations, facilitat-
ing a wider knowledge of mining history and the
material culture of the period;
– protection and classification of the ruins dating
from the earliest settlements in Ouro Preto (Fig. 4);
Figure 4. View of part of the ruins at Morro da Queimada
(IPHAN archives, Ouro Preto, MG, Brazil).
Figure 5. Aerial view of the park: area of 124.88 hectares
(IPHAN archives, Ouro Preto, MG, Brazil).
– preservation of the memory of Felipe dos Santos
and of the Sedição de Vila Rica in 1720;
– creation of a Ecomuseum (Community Museum)
and archeological museum for the towns which
emerged during the “Gold Cycle”;
– creation of a unusual option distinct from the tradi-
tional tourist circuit, to encourage visitors to extend
their stay in the area;
– protection of a meaningful part of the landscape,
including the urban and architectural complex of
Ouro Preto;
– improvement in the quality of life and social inclu-
sion of the nearby communities, creating new
employment and business opportunities, as well as
ensuring the economic sustainability of the project.
During 2006 and 2007, various plans for public
works were developed, including the following:
– delineation of the area of the park and creation
– of the park program (Fig. 5);
– demolition of the recent constructions inside the
park; the families concerned were relocated to new
homes designed for improved quality of life;
285
Figure 6. Creation of the Ecomuseum (IPHAN archives,
Ouro Preto, MG, Brazil).
Figure 7. View of the ruins on Morro do Queimada, in
the, in the mid 20th century (Photo from Luiz Fontana IFAC
archives, Ouro Preto, MG, Brazil).
– creation of an Ecomuseum, in line with the require-
ments of the local population in the nearby areas,
so as to encourage social and cultural relationships
that would facilitate social development based on
the elements of living space, society and heritage
(Fig. 6).
At present, the Historical Research Group is work-
ing on the bibliographic classification of manuscript,
published and iconographic sources and on the prepa-
ration of interview itineraries for the preliminary
stages in collecting the oral history of the area. After
a critical reading and analysis of the sources and bib-
liography, a work methodology will be established for
drawing up a basic text on how to make a qualitative
approach to the space. Research in archives, libraries
and technical reports on the history, historiography and
information management of the Morro do Queimada
will assist in making diagnoses that will orient the pro-
cess of setting up archaeological park, as well as in the
elaboration of the central theme to be incorporated in
the urban planning directives for the City of Ouro Preto
(Fig. 7).

Figure 8. Demolition of irregular constructions (IPHAN
archives, Ouro Preto, MG, Brazil).
The Property Issues and Housing groups have
already carried out the survey, transfer and new
architectural projects created for those families that
lived in the area designated for the creation of the
park. The families were compensated and can opt
to buy dwellings or lots for the construction of new
houses. The projects were developed with the partici-
pation of the Programa de Arquitetura Pública (Public
Architecture Programme) of the School of Archi-
tecture of UFMG (Universidade Federal de Minas
Gerais), which seeks to align the education of its stu-
dents with the Brazilian reality, using both theory and
practice (Fig. 8).
The Physical Planning Group has the responsibil-
ity of making interventions in a site of exceptional
natural beauty and significant archeological, historical
and natural value. The first stage, already in progress,
consists of drawing up proposals and studies for the
creation of the park, taking into consideration its
archaeological, historical, constructive, social, visual,
environmental and functional aspects, so as to obtain
information and subsidies for carrying out the projects
that involve physical intervention. The interventions
involve the identification, protection and consolida-
tion of the archaeological structures and remains, as
well as proposing solutions for environmental recov-
ery, and guaranteeing the protection, surveillance and
security of the archaeological site.
The new structures will shelter programme activ-
ities and will be grouped in conformity with their
separate functions and by functional and environ-
mental affinities: entrance and access, administrative
block, parking and cafeteria.
To articulate and connect these spaces, roads and
interpretative trails will be created that will impart the
values existing in the park and its creation. Rest and
family areas, signage, and various types of communi-
cation, visual, mobile and lit, will be installed.
286
Figure 9. View of a mine entrance, 1990s (Photo from
Germano Neto, Ouro Preto, MG, Brazil).
The architectural design proposes buildings that
will be sparse, transparent and inserted into the natural
surroundings, taking on expressive and technologi-
cal aspects appropriate for contemporary architectural
languages, but that also make reference to the typo-
logical and traditional construction characteristics.
The proposal must create a fruitful dialogue
between the natural elements and the constructions in
such a way so as to produce a renovated and restated
landscape on the Morro da Queimada (Fig. 9).
The Community Relations Group is working with
various initiatives to involve the neighbouring commu-
nities in the project of creating the archaeological park.
For this the principles, concepts, methodologies and
actions of the Ecomuseum will be used, so as to meet
the needs and desires of the neighborhoods that sur-
round the archaeological site. The principle function
of this group is to establish a participative inventory
of activities for the purpose of diagnosing local poten-
tialities and educational activities in the art education
field; creative arts, theatrical arts, music, dance; hand-
crafts; sports and leisure. The expected result is a
community project aimed at a receptive tourism, sus-
tainable culture, by means of the experiences of an
ecomuseum integrated with the archaeological park.
4 CONCLUSIONS
The challenge – to implant in the community a modern
approach to sustainable development that takes into
consideration the preservation of the natural and cul-
tural heritage of the city – requires concerted action
from all parties concerned.
This project is supported by several public and reli-
gious institutions, including the following: The United
Nations Educational, Scientific and Cultural.
Organization – UNESCO, Instituto do Patrimônio
Histórico Artístico Nacional – IPHAN, The Federal
University of Minas Gerais – UFMG, The Federal
Figure 10. View overlooking Ouro Preto from Morro da
Queimada (IPHAN archives, Ouro Preto, MG, Brazil).
University of Ouro Preto – UFOP, State Institute for
Historical and Artistic Heritage – IEPHA, the State
Forestry Foundation – IEF, the State Public Fund-
ing Agency, Municipality of Ouro Preto – PMOP,
Municipal Chamber of Ouro Preto – CMOP, the Gor-
ceix Foundation, parish of Nossa Senhora do Pilar,
parish of Santa Efigênia and the Sociedade Soto Zen
do Brasil. In addition, there is support from various
NGO’s, such as the Associação de Proteção Ambien-
tal Ouro Preto – APAOP (Ouro Preto Environmental
Protection Association), the Amigos do Patimônio Cul-
tural e Natural de Ouro Preto – AMO Ouro Preto
(Friends of the Cultural and Natural Heritage of
Ouro Preto) and the Federação das Associações de
Moradores de Ouro Preto- FAMOP (Federation of
Residents Associations of Ouro Preto).
This project is coordinated by IPHAN, developed
in partnership with The Federal University of Ouro
Preto – UFOP, Municipality of Ouro Preto – PMOP,
Municipal Chamber of Ouro Preto – CMOP, man-
aged by the Museu de Arte Sacra do Carmo, based
at the Paróquia de Nossa Senhora do Pilar, and spon-
sored by the Programa Petrobras Cultural, the Caixa
Econômica Federal (Federal Development Bank) and
Novelis Brasil Ltda.
Morro da Queimada is a site endowed with rare nat-
ural beauty, from where it is possible to obtain a unique
view overlooking the historical centre of Ouro Preto
(Fig. 10), and the Peak of Itacolomi, which guided the
first European settlers, the bandeirantes, to the area,
and, still to this day, a symbol of the city (Fig. 11).
287
Figure 11. View of the Peak of Itacolomi from Morro da
Queimada (IPHAN archives, Ouro Preto, MG, Brazil).
The creation of the archaeological park of Morro da
Queimada is one of the most important recent under-
takings in Ouro Preto, and the realization of this project
will be as important to the city as the Acropolis is for
Athens or the Palatine Hill and the Ancient Forum are
for Rome.
The creation of the archaeological park of Morro
da Queimada will establish a setting favorable to the
recovery of the cultural and environmental heritage of
the city, a new start in the place where the city began:
the Arraial of Ouro Podre where once Paschoal was
master (Meirelles, C, p. 55).
REFERENCES
Barker, P. 1981. Tecniche dello scavo archeológico. Milano,
longanesi & C.
Bediaga, B. (org.). 1999. Diário do Imperador D. Pedro II,
viagem a Minas Gerais Vol. 24, – primeira parte, 26/03 a
19/04 de 1881. Petrópolis, Museu Imperial.
De Varine, H. O Tempo Social. 1987. Rio de Janeiro, Eça
Editora.
Meirelles, C. 1989. Romanceiro da Inconfidência. Rio de
Janeiro, Editora Nova Fronteira.
Mello, S. de. 1985. Barroco Mineiro. São Paulo: Editora
Brasiliense. Belo Horizonte: Editora Rona, 1979.
Vasconcelos, D. de. 1974. História Antiga de Minas Gerais.
Vol. II. Belo Horizonte: Editora Itatiaia.
Vasconcellos, S. de. 1977. Vila Rica. São Paulo: Editora
Perspectiva.
Vasconcellos, S. de. 1979. Arquitetura no Brasil: sistemas
construtivos. Belo Horizonte: Editora Rona.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Structure as a document

Jinisha Jain
Conservation Architect, Delhi, India

ABSTRACT: Known as the ‘Old Fort’, Purana Qila, the first Mughal capital city, is situated on the eastern
edge of Delhi, along the river Jumna. Humayun, the second Mughal emperor (1530–38 A.D.) began constructing
a walled city and fortress on this site in 1533 A.D., and named it Din-Panah, or ‘Refuge of Religion’. He was
temporarily deposed by Sher Shah Sur (1538–45 A.D.), who extended the Old Fort but the fort was finally
completed by Humayun (1555–56 A.D.), the founder king. The comeback of Humayun and physical additions
to the site which can visibly be discerned but not assigned to any one ruler with confirmation compound
the problems of authorship and the historic sources in this respect fail to give any conclusive or corroborative
evidence. The fort walls are over one mile long, and contain three gates, a triple storey octagonal structure
called Sher Mandal and a mosque called Qila-I-Kunha Masjid meaning the ‘Mosque of the Old Fort’. The paper
focuses on Qila-I-Kunha and uses documentation of this structure as a tool to unveil the mystery.

1 HISTORIC SITES: AN EXISTING 2 ON SITE OF PURANA QILA

TESTIMONY
Historic sites are not just places for aesthetic admira-
tion, for exercising nostalgic moments or for eliciting
evocative responses of awe, wonder and marvel but
they are an existing testimony of how and what things
were once upon a time.
The relevance of this work really lies in understand-
ing the objects of architecture as, preserver of history,
storehouse of great deal of information and as fos-
silizer of past identities. This is to say that when a
historic building or site is conserved, it is not just an
architectural object- a physical entity that is conserved
but also the history and any other information that is
associated with that structure, thus making that struc-
ture a built resource for retrieving various kinds of
information and in determining its identity.
Recognition of elements, which contribute to iden-
tity, and integration of history is thus only ethical
to sound conservation works while lack of histori-
cal research handicaps the conservation processes and
may even tamper or completely eliminate evidences of
the past.
There has always been a mystery related to the site
of Purana Qila, located in the heart of the capital city
Delhi and the answer whether one discovers or not is
not to be found anywhere else but on the
site itself, lest the proofs to be obtained from the site
are manipulated to the extent that the answer becomes
almost impossible to be recovered from the stones of
the age old site.
Known as the ‘Old Fort’ of Delhi, Purana Qila, the
first Mughal capital city, is situated on the eastern edge
of Delhi, along the river Jumna. Humayun, the sec-
ond Mughal emperor (1530–38 A.D. /1555–56 A.D.)
began constructing a walled city and fortress on this
site in 1533 A.D. and named it Din-Panah, or ‘Refuge
of Religion’. The chosen site was an ancient area
known as Indraprastha, associated with the Hindu
epic Mahabarata. Through the excavation from the
Purana Qila evidences of pieces of the painted grey
ware, relics and remains of later periods have been
discovered in the Archeological Surveys. However,
its occurrence here seems to support the tradition of
Purana Qila being the site of Indraparastha, which
was the capital of the Pandavas, heroes of the Mahab-
harata, originally situated on the bank of the River
Jumna (Sharma, 2001).
Din-panah, original name for Purana Qila and its
surrounding most appears to have been conceived as a
royal citadel attached to a larger township which orig-
inally covered the low- lying ground around the fort
on the south, north and west.
The project was not interrupted when Humayun
was temporarily deposed by Sher Shah Sur (1538–45
A.D.), for the Sur ruler completed the fortress walls
and built many structures within. He strengthened and
extended the ‘Old Fort’ raising on the same site a
citadel for the governor and renamed it as Shergarh,
‘The fort of Sher Shah’. The structures built by Sher
Shah Suri may possibly have been demolished or

289

Figure 1. Purana Qila has some interesting structures
within including the three imposing gateways, a mosque and
an octagonal structure. A baoli and three wickets on the side
of Jumna bank also exist.
altered by Humayun when he took back his throne
and the city. In any case, it is not dubitable that Sher
Shah Suri in the small although significant interreg-
num period also left the Purana Qila unfinished and it
was finally completed by Humayun, (1555–56 A.D.)
who was also the original founder of it. The repeated
entry of Humayun and physical additions to the site
which can visibly be discerned but not assigned to
any one ruler with confirmation, have compounded
the problems of ‘who built what?’ in Purana Qila.
The fort walls in well-bonded rubble masonry are
over one mile long, and contain three gates, which
are double or triple storied, built with red sand stone
and surmounted by canopies or chhatris. Among the
three main gates, the northern gate is called the Talaqi
Darwaza meaning ‘Forbidden Gate’. The southern
gate is known as the Humayun Darwaza. The west-
ern gate, the principal gate today and in all probability
the principal gate in the past as well is called Bara
Darwaza meaning ‘Big Gate’. The eastern side has
three wickets on the side of Jumna bank, the central one
being the ‘Water Gate’. Purana Qila has some interest-
ing structures besides the fort walls, the three imposing
gateways and the wickets. It has a mosque called
Qila-I-Kunha Masjid and an octagonal triple-storeyed
pavilion called Sher Mandal. A water reservoir, baoli
also exists.
The various historic sources in Purana Qila fail
to give any conclusive or corroborative evidence
regarding their authorship. Thus, the mystery always
resurfaces with contradicting opinions and conflicting
interpretations.
For example, according to one interpretation, the
buildings that survive from the time of Sher Shah
were built close to the western gate. They include
the Qila-i-Kuhna Masjid, and the Sher Mandal, the
three-storeyed, octagonal pavilion with red and white
290
color scheme and according to another interpretation
none of the structures in Purana Qila belong to Sher
Shah, except for minor additions and repairs. There
is no epigraphical evidence to support either claim.
In any case, Sher Mandal and Qila-i-Kuhna Masjid,
are the two surviving structures of certain complex-
ity, within the fort ramparts from the mid sixteenth
century, which command interest and consideration
as they mark an important and decisive stage in the
development of the subsequent styles and features of
Islamic architecture in India.
3 PRESENTATION OF HISTORICAL RECORDS
AND EXISTING VIEWS ON THE SUBJECT
The authorship of the complex has been attributed by
medieval historians variously to Humayun and Sher
Shah. The debate related to the authorship of Purana
Qila in itself is historical with contradictory state-
ments and records from the medieval historiographies,
which are taken as important references for either of
the viewpoints. These are presented below:
• Khwand Amir’s Qanun-I-Humayuni records that in
the year A.H. 940 (1533 A.D.) Humayun laid the
foundation of a city named Din Panah in Delhi on a
raised area of the embankment of the river Jumna.
This new city was at a distance of three kuroh from
the older urban area in the south. The progress of
construction, according to this text, was so rapid that
within a period of ten months the bastions, ramparts
and gates could be almost completed.
• Tarikh-I-Da’udi states that it was Sher Shah who
built a new city on the banks of Jumna in Delhi,
between Firozabad (Kotla Firoz Shah area) and
Kilokhari (near Okhla) in the village of Indrapat,
in A.H. 947 (1540 A.D.) and inhabited it, but could
not complete it, due to his untimely demise. He also
writes that Sher Shah named the fort as Shergarh.
• This statement is also corroborated by Tarikh-
I-Shershahi, written by Abbas Khan Sarwani in
Akbar’s reign, an earlier work, which records that
Sher Shah ordered destruction of the older city
of Delhi and construction of two forts to be built
in Delhi by the bank of Jumna, the smaller one for
the residence of local governor and the larger one
around the city. But the fortifications enclosing the
town remained incomplete by the time of his death.
(Abdullah in his Tarikh-I-Da’udi, says that what was
destroyed was the Qila-I-Alai that is Siri, which was
indeed the former capital city of Delhi, at a distance
from Jumna. It is then likely that Tarikh-I-Shershahi
also talks about destruction of Siri and not the city
built by Humayun, as is sometimes believed. Also,
Kotla of Firoz Shah was the closest old city of Delhi,
that could be used as a quarry and facilitate quick
 completion of work, but the fact that it was itself
included within the ramparts, suggest that not major
recycling of materials from this site was possible).
• According to the Tarikh-I-Khan Jahan, another
important work on Afghan history, Sher Shah’s son
Salim Shah Sur (1545–54 A.D.) is stated to have
built a wall encircling the city of Humayun, which
falls perfectly in place as both Tarikh-I-Daudi and
Tarikh-I-Shershahi say that the outer fortification
had begun by Sher Shah Sur but could not be
completed by him (thus completed by his son).s
• The features of Purana Qila, which have been iden-
tified as, ‘Mughal’, could also date from Akbar’s
period because we learn from Akbaranama that in
spring 1560 A.D., Shihabuddin Ahmad Khan, the
governor of Delhi, in anticipation of an attack by
Bairam Khan, proceeded to strengthen the fort of
Delhi and to repair its towers and walls. (maramat-
i-burj-o-bara).
• Zafar Hasan, a 19th C archeologist admitted that
the difficulty in judging how much of the citadel
(Purana Qila) is the work of Humayun and how
much of Sher Shah. However, he states that accord-
ing to general opinion the walls and gates are
considered to be the work of Humayun and mosque
(Qila-I-Kunha) and Sher Mandal, of Sher Shah.
• Recently, J. Burton-Page and R. Nath have sug-
gested that none of the Delhi monuments tradition-
ally ascribed to Sher Shah were actually built during
his reign, much less patronized by him.
• Architectural Historian Catherine Asher, assigns the
Qila-I-Kunha masjid as indeed the mosque of Sher
Shah- a visual symbol of royal aspirations of the
ruler.
• M.C Joshi, a modern historian, acknowledges
minor contributions of Sher Shah and through
re-examination of literary and existing epigraphi-
cal proofs credit most of the surviving structures to
Humayun.
4 ON STRUCTURES WITHIN: QILA-I-KUHNA
MASJID
Purana Qila is a subject that has been dealt with by
many architectural historians, researchers and other
scholars time and again. It is humble to acknowledge
that the results or conclusions can never be path-
breaking in the sense that there are eventually three
possibilities one encounters: the structures were built
by Humayun, the structures were built by Sher Shah
and the third-the structures were built in parts by both.
Nevertheless, it is the approach, a fresh perspective
and the rigor of the quest that has kept the intrigue on
the subject going.
The Qila-I-Kuhna mosque conforms to the archi-
tectural tradition of a five-bay mosque with triple
291
domes, a mosque typology that had developed in the
precedent Sayyid and Lodhi periods. It occupies an
area of approximately 51 m by 13.5 m, with a total
height close to 20 meters, inclusive of the dome.
The mosque is built in core masonry of stone rubble
bonded with lime mortar and with fascia of finished
blocks of Delhi quartzite and red sandstone. The inter-
nal space, a large rectilinear hall is divided laterally
through massive load-bearing stone masonry arches
into a distinct number of five. The external eastern
façade correspondingly consists of an arrangement of
five archways with the central one being sunk in an
arched recess contained within a pronounced rectangu-
lar frontage, the Liwan. The liwan is ornamented with
intricate stone relief and intarsia in multi-coloured
schemes-red and yellow sandstone, marble and black
slate. At the rear of liwan rises a characterristic Lodhi
dome with small minarets called guldastas, literally
the flower bouquets, flanking the drum and with a
finial, kalash on the top. The scheme of five also
translates internally as an arcade of mihrabs, one in
each bay, along the qibla or the west facing wall. The
mihrabs are designed on the same general principle
as the recessed arches in the eastern façade; however
they exhibit very fine workmanship. The system of
roof support consists of domes, cross-rib vaults and
semi-domes with sqinches, stalactites and corbelled
pendentives in the phase of transition. Certain innova-
tive roofing devices have been employed and there are
three roofing techniques which have been used. The
central bay has a single masonry dome with multiple
openings in the drum for ventilation and is decorated
with colored tile work. Squinch is employed in the
phase of transition. The penultimate bays on either
side have what appear to be flat domes or vaults. The
end bays, which are oblong unlike the central and the
penultimate bays which are somewhat squarish have an
unusual roofing mechanism. The last bay shows two
half arches which spring tangentially from the shoul-
ders of the lateral arches which define this bay. This
results in creation of three spaces to be spanned with
a cross-rib arrangement. Thus the central space of the
end bay has a shallow dome, while the spaces towards
the western and eastern arcade are covered through
semi-domes, nim-gumbads resting on the suspended
arches which in turn rest on the main lateral arches.The
phases of transition are covered through beautifully
carved corbelled pendentives. Attached to the ends of
the rear wall are two octagonal turrets, which are func-
tional as well as structural. Oriel windows, decorative
brackets, sunshades or chajjas and narrow turrets with
taper, reminiscent of Tughlaq architecture- an early
phase, on each side of the central rectangular bay both
in the front and rear are other architectural features
which are used in the edifice. Hexagrams or six pointed
star motifs along with colored stone inlay of vari-
ous geometric patterns form the chief ornamentation.
The mosque had private entrances on the north and
south side for the use of the royal family, which are
now kept closed. The mosque is generally accepted to
represent the culmination of its type and as also the
prototype for the more refined subsequent typologies
of Mughal mosques.
The mosque traditionally is considered to be Sher
Shah Sur’s royal chapel. In the absence of histori-
cal inscriptions on the Qila-I-Kuhna, the traditional
view of the mosque’s patron is based on the near
contemporary histories of the Sultan Sher Shah Sur.
Following presents abstracts from these histories and
their interpretation by some of the recent historians.
• Both the Tarikh-I-Shershahi by Abbas Khan and
Abdulla’s Tarikh-I- Da’udi state that among the
very first of Sher Shah’s accomplishments upon
assuming the Delhi throne in 1540 A.D. was the
construction of his own capital, which included two
fortified enclosures, the city walls and the royal res-
idence, the area known today as the Purana Qila.
(The destruction of Humayun’s capital in Delhi is
not implied directly in any of the texts). The erection
of a magnificent Jami mosque within the Purana
Qila is cited by both Tarikhs. Moreover, Abdullah
in the Tarikh-I-Da’udi states that this mosque was
completed quickly, implying its full construction
during Sher Shah’s reign.
• Tarikh-I-Shershahi says, ‘The former capital city of
Delhi was at a distance from the Jumna, and Sher
Shah destroyed and rebuilt it by the bank of the
Jumna, and ordered two forts to be built in that city,
with the might of a mountain, and loftier in height;
the smaller fort for the governor’s residence; the
other, the city wall to protect it; (literally, Jahan-
panah) and in the governor’s fort he built a Jama
masjid of stone, in the ornamenting of which much
gold, lapis lazuli, and other precious articles were
used. But the fortifications round the city were not
completed when Sher Shah died.’
• It seems unlikely that Abbas Khan, who was com-
missioned by Akbar to write the Tarikh-I-Shershahi,
would attribute the mosque within the Purana Qila
to Sher Shah if the actual patron was either Akbar
himself or his father, Humayun.
• According to Catherine Asher, the literary evidence
referring to Sher Shah’s patronage of the Qila-I-
Kuhna mosque is hard to dispute. The mosque, then,
must have been built soon after 1540 A.D. when
Sher Shah ascended the Delhi throne. A compar-
ison of central archways with bracketed entrances
of Ibrahim Sur’s tomb (Narnaul, Haryana), which
were built by Sher Shah according to an inscriptions
thereon, and those of Purana Qila further supports
the above point of view. In fact, like the Sur monu-
ments of Sasaram, the mausoleum of Ibrahim Sur is
an improved and a refined version of Bara Gumbad
or other square Lodhi tombs.
292
• M.C Joshi opines that it is not unlikely, that this
mosque was designed and founded by Humayun
but was completed by Sher Shah, who seems to
have been responsible for building the upper part
of the super structure including the central dome,
crenellated parapet of the roof- also encountered in
his Narnaul and Sasaram monuments- and intro-
ducing structural ornamentation including blue tile
work, colourful gilded paintings, lotusbud fringe in
arches, and ornate mihrabs. Perhaps the Sur ruler
also built its roof flanking the central dome and
the two sidedomes, the bases of which are still par-
tially extant. The unfinished marble facing of the
front, with rich inlay designs can in no case be older
than the time of Akbar (1556–1605 A.D.), because
the ornate geometric patterns as well as their tech-
niques of execution are purely Mughal in character
and absent in Sur or Sultanate monuments.
• Architecturally, this edifice expresses the logical
development of the five bay structural model repre-
sented by Masjid Moth (Lodhi) and Jamali-Kamal
(Baburi) mosques; yet the employment of pointed
arches, nim-gumbads (half domes) in it supporting
the roof of the side- bays and pietra dura certainly
suggests a strong ‘Mughal’ association.
5 CRITICAL REVIEW AND ANALYSIS OF THE
ARCHITECTURE
We have already come across the inadequacy of the
archeological, historical, epigraphical and toponymic
(study of names and places; Sher Mandal as an exam-
ple would bring to mind association with Sher Shah
but most of the scholars agree with Humayun being
the actual founder) proofs in the case of Purana Qila
as demonstrated above.
This paper focuses on the documentation, architc-
tural and structural analysis and subsequent interpre-
tation of one structure in Purana Qila, Qila-I-Kunha
Masjid to trounce the other research inadequacies.
Herein are presented the author’s views on the
authorship of this structure:
• The planning, masonry and the same proportions in
all the pylons inside the mosque clearly illustrate
it was conceptualized as a five bay mosque, thus
thwarting any possibilities of the extreme end bays
being added later, although the change in propor-
tions of the elevational arches of these bays is odd
and not seen in any of the previous prototypes or
other contemporary mosques of the period, though
the central liwan is almost always elevated.
• The internal staircases housed within the end bays
were planned to reach the terrace of the turret and
then further into the drum gallery to finally the ter-
race. This is more pronounced in the light of the
fact that the mosques of Moth-ki-Masjid and Jamali
 Figure 3. East elevation.
Figure 2. Qila-I-Kunha Masjid, Purana Qila, Plans at dif-
ferent levels showing relationships of the structure and
spaces.
Kamali both of which are considered as prototype
for this mosque have two set of staircases, internal
staircases that lead to the turret and the external ones
to reach the terrace. It is only here that there exists-
a set of two internal staircases. Consequently the
end bays, the octagonal turrets and the dome could
not have been built separately and were planned,
conceptualized and executed by the same patron.
• Also, if the dome is a characteristic Suri (influenced
by the preceding pathan dynasty of the Lodhis) sin-
gle dome with guldastas, tapering turrets, a certain
flatness in appearance and polychrome mosaic work
and painted ceiling work in the interiors, the mosque
293
which shows a unified concept in its execution is
also then by Sher Shah Sur, who on the authority
of medieval writer Abbas Sarwani, a court histo-
rian of Akbar ascribed it to Sher Shah Sur and not
Humayun, his patron’s father. (Also if one persists
that the mosque was indeed built by Humayun then it
must have been built by him in its entirety, including
the so called Suri dome, both because the concept of
the structure with literal ‘structural underpinnings’
is too unified to be executed by two patrons and also
because Humayun had enough time, 1533–38 A.D.
to undertake and finish the architectural pursuit of
a mosque in his royal citadel, had it been begun
by him).
The unified concept is reflected in not just the
planning but also in the ornamentation and the
decorative motifs which are used in the building.
The motifs on the base of the columns, the frieze,
the capitals and the brackets show consistency in the
interior of the mosque and also in interior of the tur-
rets, although this was achieved through cladding.
• Also the arches, the five-bay rectangular plan and
the other dominant features such as calligraphic
embellishment, stalactite pendentives, octagonal
corner towers, oriel windows, etc. can easily be seen
in Sultanate period buildings. The mosque is only a
logical development of a mosque typology existent
in Delhi Sultanate since the Lodhi times, reflected in
the Bara Gumbad mosque, the Moth ki masjid, the
Jamali Kamal mosque(completed in Babur’s reign)
and finally a synthesis of all these in Qila-i-Kunha
mosque within Purana Qila.
• The Mihrabs and are not structural in the same way
as the arches of the main eastern façade which take
the load down through their broad pylons. The load
of the rear wall is mostly translated through the
octagonal turrets at the ends, the narrow turrets in
the centre and the fat western wall. The Mihrabs are
largely ornamental and display a conscious sense
of aesthetics. The Mihrabs with alternate ribs of
dark and light colored stone and bud-fringes along
their silhouette are seen in the tomb of Ibrahim
Sur in Narnaul and also in mosques such as the
Atala mosque in Jaunpur-the learning centre, a place

Figure 4. Structural analysis of Purana Qila.
where Sher Shah Sur had spent much his time in
youth. (Although neither of them show the clas-
sicism achieved in the mihrabs of Qila-I-Kunha
with exquisite inlay work, which probably could be
ascribed to Mughals- Humayun).
• Humayun after recapturing Delhi Sultanate must
have contributed in introducing pietra dura, such
as in the liwan, in the mihrabs, in the brackets
supporting the pendentives, in the frieze with crenel-
lation motif at the rear, etc- much of which was left
unfinished because of his untimely death. The con-
tributions made by Akbar if any are not clear. Minus
the sophisticated inlay work done in parts the whole
building indeed begins to look more like a struc-
ture built in Suri time, with more dominant play of
quartzite and red sandstone carved cladding just like
the tombs in Sasaram and Narnaul.
It is interesting to note that many Timurid ideas
and a new aesthetic sensibility had been introduced by
Babur, whom Sher Shah Sur had personally served.
Thus the influence of Mughal ideas or assimilation of
their aesthetic sensibility by a later dynasty shall not
be a surprise.
Thus it can be disputed that the mosque in its
entirety was conceptualized and built by the Sur ruler,
and only non-structural modifications such as replac-
ing the cladding, carving and introducing fine persian
inlay work were made by the Mughal rulers.
6 CONCLUSION
There is a Suri tradition of architecture different from
both its predecessor and successor dynasties. The idea
has been to integrate such research and documentation
with conservation works to be able to determine the
right kind of significance for the structure and also to
aid in preservation of historic proof while opening up
possibilities for multiple interpretations for different
294
people (and not just scholars) instead of adopting a
linear method of dispersing information. Qila-I-Kunha
Masjid shows innovations, which became prototypes
for emulation for the later buildings. It set precedents.
In the absence of sound epigraphical, historical or any
other proof, the building itself has served to be a stand-
ing document and has revealed significant information
on the basis of which we can infer that Purana Qila is
one of the few sites where the two different architec-
tural traditions of two rival camps-the Mughal and the
Sur have managed to co-exist.
REFERENCES
Abdullah. Tarikh-I-Daudi. S.A. Rashid (ed). Aligarh.
Abu-l-Fazl. Akbarnama Vol. I. 1897–1939. (Translated from
the Persian by H. Beveridge) Calcutta: Asiatic Society of
Bengal.
Ahmad Khan Matta, Basheer. 2005. Sher Shah Suri- A Fresh
Perspective. Oxford University Press.
B.Asher, Catherine. 1977. The Mausoleum of Sher Shah Suri.
In Artibus Asiae Vol. XXXIX: 273–98.
B. Asher, Catherine. 1982. The Qil’a-i-Kuhna Mosque: A
Visual Symbol of RoyalAspirations. In Chhavi II: 212–17.
Varanasi: Banaras Hindu University.
B. Asher, Catherine. 1988. Legacy and Legitimacy: Sher
Shah’s Patronage of Imperial Mausolea. In Katherine P.
Ewing (ed), Shari’at Ambiguity in South Asian Islam:
79–97. Berkeley: University of California Press.
Brown, Percy. 1975. Indian Architecture (Islamic Period): 26.
Bombay: D.B. Taraporevala Sons and Co. Pvt. Ltd.
Fergusson, James. 1910. History of Indian and Eastern
Architecture. London.
Hasan, Maulvi Zafar. 1997. Monuments of Delhi, Vol. I-IV.
New Delhi: Aryan Books International.
Khan Sarwani, Abbas. 1952. Tarikh -I- Sher Shahi or Tufhat-
I-Akbar Shahi d. 16th C. (Translation in Posthumous
papers of H.M Elliot, John Dowson (ed). First Ed., 1871.
Second Ed. 1952).
Khwand, Amir. 1940. Qanun-I-Humayuni d. 1535. (Transla-
tion by Baini Prashad). Calcutta: The Asiatic Society.
Koch, Ebba. 1991. Mughal Architecture: An Outline of its
History and Development (1526–1858). Munich: Verlag.
Ni‘mat-Allāh ibn Habı̄b-Allāh Haravı̄. Tārı̄h i Khān-Jahānı̄,
d. 1024/1615. (Translation). Collection of Oriental Works.
Qanungo, K.R . 1965. Sher Shah and His Times. Orient
Longmans Limited.
R. Nath. 1978. History of Sultanate Architecture. New Delhi:
Abhinav Publications.
R. Nath. 1995. Studies in Medieval Indian Architecture. New
Delhi.
Sharma,Y.D. 2001. Delhi and its Neighbourhood. New Delhi:
Archeological Survey of India.
Srivastava, Ashirvadi. 1950. Sher Shah and his Successors.
Agra.
Stephen, Carr. The Archaeology and Monumental Remains of
Delhi.
Tūzuk-i Jahāngı̄rı̄ or Memoirs of Jahāngı̄r. 1909–1014.
(Translated by Alexander Rogers and Henry Beveridge)
London: Royal Asiatic Society.
Yadgar, Ahmad. Tarikh-i- Salatin-i- Afghani (Translation).
Monitoring, non destructive evaluation
and testing
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Historic multiple-leaf masonry wall models under compression

and cyclic shear loads

R. Capozucca
Università Politecnica delle Marche, Ancona, Italy

ABSTRACT: An investigation on historic masonry wall models – in scale 1/3th – characterised by two leaves of
brickwork masonry and intermediate mortar has been carried out in the laboratory. Experimental results obtained
by tests on wall models under compression and cyclic shear loads are discussed and compared to theoretical results
obtained by a non linear procedure of calculus based on a tie & truss modelling for multiple-leaf masonry wall.

1 INTRODUCTION towns and the most common cause of damage during

The preservation of the architectural heritage presents
one of the important challenges in civil engineering
due to the complexity of the geometry of the struc-
tures, the variability of the materials used and the
loading history of the buildings. Structures of his-
toric buildings because of their nature and history
present a number of typical aspects that limit the appli-
cation of modern codes and building standards. The
knowledge of behaviour of historic masonry under
seismic action is fundamental to preserve the architec-
tural heritage. During an earthquake, masonry walls
are subjected to shear loads that often carry out to
the ruin the whole building. The shear strength of
unreinforced masonry (URM) was analysed in the
last decades by experimental and theoretical works
(Benjamin & Williams 1958; Hendry & Sinha 1969,
1971; Turnesek & Cacovic 1971; Yokel & Fattal 1976;
Tomazevic 1977; Bernardini et al.1980; Hamid &
Drysdale 1980; Calvi et al. 1985). On the base of the-
oretical and experimental analysis, the shear criterion
for URM known as Coulomb-type failure criterion is
adopted (EC6-ENV 1996). The shear failure widely,
is characterised by joint failure as a function of the
bond strength linked to the frictional resistance at
brick-mortar interface and the compressive stress nor-
mal to the bed joints. Although many experimental
results are available on shear strength of masonry as
yet only few data are available for historic unrein-
forced masonry walls (HURM) (Capozucca & Sinha
2004, 2005, 2007). In fact, shear behaviour of masonry
has been investigated extensively for modern masonry.
Moreover, between the different types of HURM, a
scarcity of data is on multiple-leaf walls with two or
three leaves. Multiple-leaf masonry is usual in historic
an eartquake is actually due to a ruin of the build-
ings with this type of wall masonry (Pina-Henriques J.
et al. 2004). In this paper, cyclic shear tests on HURM
wall models with multiple-leaf masonry are shown
and experimental results are discussed to investigate
cracking patterns, failure mechanisms and the ultimate
shear strength. Experimental results have been com-
pared with those obtained from a non linear theoretical
analysis carried out modelling multiple-leaf masonry
wall with tie & truss finite elements.
2 EXPERIMENTAL MODEL
2.1 HURM wall specimens
To have the meaningful results, it was essential to
have similar materials as used in the HURM. Luckily,
a few full-scale solid bricks became available dur-
ing the renovation of a 18th century Italian building,
hence test specimens were built in 1/3rd scale utiliz-
ing these. The dimensions of the model bricks were
100 × 50 × 17 mm obtained from sawing the full-scale
bricks. The average compressive strength of the model
bricks was 34.3 N/mm2 . 1:1:5 (cement: lime: sand)
mortar was used for the construction of specimens.
The model was built with full masonry for flanges
and multiple-leaf masonry for web (Fig. 1). The sec-
tion is double T shape as shown in Figures 2(a) and
(b); bricks in plane in flanges and on the list in the
web (Fig. 2(a)) with intervals of bricks in plane as ties
between two leaves (Fig. 2(b)). 1:5 (lime: sand) mortar
has been placed between two masonry leaves of web.
A series of preliminary tests were done on small
wallet specimens of HURM to obtain the compressive

297
Figure 1. Masonry wall model.

Figure 3. Set-up of shear test and instruments.

allows to transfer both precompression – normal stress

equal to 1 N/mm2 – and horizontal cyclic load (Fig. 3).
Four load cells measured the applied loads. Before
the application of shear load, the pre-compression was
(a) applied both to the web and flange of the wall by three
vertical jacks (Fig. 3) and kept constant throughout
the test.
The cyclic shear load was applied by horizontal jack
with double phase at stages till the failure. Deflections
and strains were also measured at various steps. In
Figure 3 the instrumentation to measure deflections in
six points (1. . .6), vertical strain gauges (D, E, F, G, H)
and Rosetta (A, B, C) in the middle of panel, are shown.
In Figures 4(a) and 4(b) diagonal cracks on the sur-
(b) face of web at shear failure, due to cyclic tests, for both
specimens, respectively, S1 and S2, are shown.
Figure 2. Double T shape wall section of HURM model
with (a) multiple-leaf masonry and (b) full masonry courses.
2.3 Experimental results
Table 1. Mechanical values of HURM by compression tests The failure was sudden at a horizontal force value
equal to Fu = 30 kN and Fu = 36 kN, respectively, for
Strength Direction Average values S1 and S2 specimens. The average shear stress value
young’s modulus of test (N/mm2 ) referred to the gross area 50 × 840 mm2 was equal to
τ∼= 0.86 N/mm2 for S2 specimen. In Figures 5(a) and
fy Normal to 13.5 (b) the experimental diagrams of cyclic load, ± F, vs
Ey mortar joint 8.76 · 103
Fx Parallel to 10.4
deflection, ± δ, at the top of model wall (transducer
Ex mortar joint 6.90 · 103 no. 1) are shown for S1 and S2.
The experimental measures of vertical strains
recorded at the bottom of S2 specimen are shown only
for the increasing load F− value (Fig. 6).
strength and the modulus of elasticity in two orthogo-
nal directions (Tab. 1).
3 THEORETICAL MODELLING

2.2 Set-up and shear cyclic test 3.1 Theoretical modelling of the HURM model
Two specimens S1, S2 were tested under combined Masonry is a material which exhibits distinct direc-
vertical load and horizontal shear. A system of jacks tional properties due to mortar joints. In general, the

298
 Transducer 1
40

F (kN)
30
20
10
0
-1 -0,5 -10 0 0,5 1 1,5
-20
-30
-40
(a) x (mm)

Transducer 1
40

F (kN)
30
20

10
0
-1,5 -1 -0,5 -10 0 0,5 1 1,5

-20

-30
-40
(b) x (mm)

Figure 5. Experimental cyclic diagrams of horizontal load

vs. deflection values at the top of the (a) S1 and (b) S2 models.

Figure 4. Cracks at shear failure for (a) S1 and (b) S2.

approach towards its numerical representation can

focus on the micro-modelling of the individual com-
ponents of unit and mortar, or the macro-modelling of
masonry as a composite material. In this research, the
theoretical analysis of the HURM model was devel-
oped by a macro-modelling considering the masonry
as an heterogeneous material with different directions
of strength: normal to the bed joints of mortar; paral-
lel to the bed joints of mortar and diagonal directions
inclined to an angle of 45◦ degree respect to the bed
joints (Figs. 7(a), (b) and (c)). The theoretical analy-
sis of the HURM model subjected to compression and
shear has been developed by FEM code (Straus7) with
a non linear procedure.
In Figure 8 the complete modelling is shown: beam Figure 6. Experimental strain values – S2 model.
finite elements both for vertical flanges and horizon-
tal top steel plate have been adopted while cut-off
finite elements to simulate the behaviour of web. The of compressive stress τ0 = 0.30 N/mm2 obtained by
modelling has been replied for two leaves of model. triplet tests. The Young’s modulus was initially con-
Finally, rigid links have been used to simulate the ties stant of average value Eav between the values shown in
by full bricks in the web. In the theoretical analysis, the Table 1. After the linear phase theYoung’s modulus has
value of tensile strength for cut-off finite elements has been assumed decreasing at every step of horizontal
been assumed equal to the value of shear in absence load.

299
Figure 7. Theoretical modelling of HURM model: (a) tie
and truss modelling of web surface; (b) geometric section of
HURM model web; (c) tie and truss modelling of web.
Figure 8. Theoretical modelling of the HURM wall model.
3.2 Damage coefficient
In experimental diagrams horizontal load vs deflec-
tion like that shown in Figure 9, we can appreciate
different phases of loading: first a linear elastic phase
characterized by constant stiffness; successively, a non
linear behaviour with reduction of stiffness at each
increment of load. The initial stiffness is the maximum
value that is linked to a undamaged condition of wall.
Minor values of stiffness are due to a micro-cracking
or slip on mortar joints.
300
Figure 9. Typical experimental horizontal force vs
deflection.
Figure 10. Values of damage coefficient d vs lateral force
ratio for S2.
The actual behaviour of the experimental model
is non linear up to failure and it may be described
by different steps with decreasing stiffness values
Ki = Fi /δι . A damage coefficient was defined as:
where: K0 = initial stiffness of undamaged wall;
Ki = stiffness at ith step.
The damage coefficient d has been evaluated with
reference to the experimental diagrams horizontal
load, F, vs deflection, δ, obtained by shear tests. In
Figure 10, values of damage coefficient d for each load
ratio (Fi − F0 )/F0 − where F0 = value of lateral load
evaluated at the end of linear elastic phase; Fi = lateral
load at ith step – have been calculated by experimental
data of test on S2 specimen.
A linear relation for the damage coefficient d has
been evaluated with m = 1.37:

In the experimental tests, the decrease of the stiff-

ness in the wall model was mainly due to the cracking
of web under compression and shear (Figs. 4 (a) and
(b)). In the theoretical analysis the web was simulated
by tie & truss with cutt-off finite elements character-
ized by axial stiffness (EA)i . For these elements, the
Young’s modulus varies at each step of calculus, taking
in account the damage coefficient d evaluated for S2,
as it follows:

The theoretical analysis has been developed by a

non linear procedure increasing horizontal load to
reproduce the experimental tests.

3.3 Comparison between results

The theoretical non linear analysis of S2 specimens
allowed to obtain lateral deflection data at different
values of horizontal load. Envelope curve has been
determinate by the maximum value of lateral force,
± F, vs deflection, δ, for every cycle of shear test. In
Figures 11(a) and (b) experimental values have been
shown by white points.
4 CONCLUSIONS
In this paper, the experimental behaviour of historic
masonry wall model in 1/3rd scale with multiple-leaf
web is described. A theoretical modelling has been
shown and discussed. The theoretical procedure by
non linear analysis allowed to obtain lateral force vs
deflection diagrams. A comparison by experimental
and theoretical data allowed to confirm the valid-
ity of theoretical analysis characterized by a damage
coefficient useful to describe the reduction of the stiff-
ness of wall model under increasing lateral loads.
Experimental tests on two historic models under com-
bined compression and shear, furnished the following
results:
1. the capacity of strength of the HURM model with
multiple-leaf masonry web was maintained up to
failure by shear;
Figure 11. Comparison between theoretical and experi-
mental data by envelope diagrams for (a) F+ and (b) F−
loads.
2. the behaviour of the HURM model is linear up to
a value of lateral load equal to 65% of ultimate
value of lateral load. After this value a decrease of
stiffness was recorded;
3. multiple-leaf masonry appears adequate to sustain
lateral load during the shear test. The behaviour of
two masonry leaves was recorded similar to cause
of efficiency of ties represented by brick courses:
further, the failure was contemporary reached for
the two leaves with typical diagonal shear cracks.
ACKNOWLEDGMENTS
The experimental work has been developed by funds
for research of Università Politecnica delle Marche.
REFERENCES
Benjamin R. & Williams H.A. 1958. The behaviour of one-
story brick shear walls, Jour. of Struct. Division ASCE, 84,
1723, 1–30.

301
Bernardini A. et a1. 1980. A comparison of laboratory
test methods used to determine the shear resistance of
masonry. Proc. 7th World Conf. on Earthquake Eng.,
Istanbul, vol. 7: 181–184.
Calvi B.M. et al. 1985. Random cyclic behaviour of rein-
forced masonry under shear action. Proc. 7th Int. Brick
Masonry Conf., Melbourne.
Capozucca R. & Sinha B.P. 2004. Strength and Behaviour
of Historic Masonry under Lateral Loading. 13th IBMaC,
Amsterdam, vol. 1: 277–284.
Capozucca R. & Sinha B.P. 2005. Evaluation of shear strength
of historic masonry. 5th Int. Conf.AMCM, Gliwice, Poland
(in CD-ROM).
Capozucca R. & Sinha B.P. 2007. Shear Strength of Historic
Masonry. Int. Symposium Studies on Historical Heritage,
Antalya, Turkey, 435–440.
EC6. 1995. Design of Masonry Structures, Part 1-1: general
rules for buildings – rules for reinforced and un-reinforced
masonry ENV 1996 1-1: Bruxels: CEN.
Hamid A.A. & Drysdale R.G. 1980. Behaviour of Brick
Masonry under combined shear and compression loading.
Proc. 2nd Canadian Masonry Symp., Ottawa.
302
Pina-Henriques J. et al. 2004. Testing and modelling of
multiple-leaf masonry walls under shear and compression,
Proc. SAHC, Padua, vol. II 299–312.
Sinha B.P. & Hendry A.W. 1969. Racking tests storey height
shear wall structures with openings subjected to precom-
pression, Proc. Int. Conf. on Masonry Struct. Systems
Texas 1969, USA, 192–199.
Hendry A.W. & Sinha B.P. 1971. Shear tests on full scale
single storey brickwork structures subjected to pre-
compression, Civil Engineering and Public Works Review,
1339–1334.
Yokel F.Y. & Fattal S.G. 1976. Failure Hypothesis for
Masonry Shear walls, Jour. of the Struct. Division ASCE,
102, n. ST3: 515–532.
Tomazevic M. 1977. Seismic design of masonry structures.
Construction Research Communications Limited ISSN
1365-556, Earthquake Eng. and Struct. Dynamics, 88–95.
Turnesec V. & Cacovic F. 1971. Some experimental results on
the strength of brick masonry walls, Proc. 2nd Int. Brick
Masonry Conf., Stoke on Trent 1971, UK, 149–156
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

The masonry vaults: Geometry definition and possible approaches

to the static analysis

O. Corbi
Department Structural Engineering, University of Naples Federico II, Naples, Italy

ABSTRACT: The paper focuses on the analysis of masonry vaults: on one side it addresses some of the
geometrical features involved in the treatment of shells, which usually implies not negligible difficulties in
handling the vaults’ problem due to the differential nature of their geometry, and, on the other side, it outlines
possible approaches to the static analysis of masonry vaults. To this regard, the overall static approach should
be conceived in such a way to allow the selection of membrane stress surfaces able to equilibrate the applied
loads and to satisfy the admissibility conditions of the masonry material. The paper, then, deals with these two
features, outlining a general approach for analytically handling the problem of static analysis of vaults.

1 INTRODUCTION This is why it would be of fundamental relevance

The forum about the preservation of the architectural
heritage has always been object of a big interest from
many fields; in the last decades many disciplines have
differently contributed to the topic.
According to the commonly agreed opinion from
the scientific community, any intervention for the
rehabilitation of an historical monumental building
cannot be outlined autonomously; the safeguard objec-
tive makes impossible to separate the architectural
from the constructive-structural feature.
Any preservation strategy requires the deep under-
standing of the behaviour of the structure, usually
made of masonry material, and of its past history.
to have at one’s disposal proper analytic tools for per-
forming reliable static analyses of masonry structures.
In the following one reports some discussion about
the still open research problem of masonry vaults,
outlining possible future developments.
2 GEOMETRY OF SHELLS OF GENERAL
FORM
2.1 Shells of general form
A shell of general shape may be represented as a reg-
ular surface in the space (Fig. 1), and described as a
subset of points P whose position is identified by any

Figure 1. Correspondence between the two metrics for representing the mid-surface of a shell of general form relevant to the
reference systems (Ox1 ,x2 ,x3 ) and (Ou1 ,u2 ).

303
single-valued and continuously differentiable vector
function r(u1 ,u2 ) depending on two variables u1 and
u2 belonging to a bi-dimensional space U.
The components of r(u1 ,u2 ) are the coordinates
of the point P with respect to any reference frame
Ox1 x2 x3 .
By denoting the first derivative of r(u1 ,u2 ) with
respect to ui by ri (u1 ,u2 ), and, marking by “×” the
vector product, it is assumed that
The equation of the surface, as regards to cartesian co-
ordinates xi in the reference Ox1 x2 x3 , can be expressed
in the parametric form
Substitution of ui = ui (t), in Eq. (2) gives
r = r[u1 (t), u2 (t)], which is the equation of a curve 
on the surface in Eq.(2), as shown in Figure 2. The
vector
coincides with the tangent to this surface curve .
The tangents to all surface curves (Mitrinovic
& Ulcar 1969), which pass through a fixed point
(u1 ,u2 ) of the surface, lie in the tangent plane
(Fig. 3) to the surface at that point, which con-
tains r1 (u1 ,u2 ) and r2 (u1 ,u2 ). After denoting by R the
position vector of any point other than P on the tan-
gent plane, the equation of the tangent plane is the
following

After suppressing the parameters (u1 ,u2 ), the surface

equation can be put in the implicit form

One can observe that the vectors r1 and r2 are parallel

to the tangents (characterized by unit vectors t1 and
t2 ) of the coordinate curves u1 = const and u2 = const
respectively. The normal line to the surface at the point
(u1 ,u2 ) of the surface (with unit normal vector N),
whose direction is given by the vector product (r1 ×
r2 ), has the following equation (Fig. 2)

where λ is a variable parameter defining points on the

normal, µ the length of the normal and N the unit
surface normal vector at P. Figure 3. Tangent plane to the surface at P.

Figure 2. Surface normal vector N and tangent vector t of the surface curve  in P.

304
The tangent surface of a curve , r = r(s), is the surface On a surface with the first fundamental form
generated by lines tangent to . It has the equation gik dui duk , let consider the two curves u = u(t) and
v = v(t). If they intersect at a point, then the tangents
to these two curves at the point of intersection define
an angle α between the unit tangent vectors tu and tv ,
where λ is a parameter and t is the unit vector tangent which is defined by (Fig. 7)
to . The moving of a straight line (generator) in any
direction along a curve (directrix) generates a ruled
surface. In the case when every tangent plane to a
surface is tangent to a generator, the surface is called
a developable surface. where
The differential of arch length ds of a curve ui = ui
(t) (i = 1, 2) on the surface r = r(u1 ,u2 ) may be cal-
culated by means of the first fundamental form of the
surface, as follows
is the determinant of the metric tensor g containing the
gik components, which is a symmetric second-order
tensor of the type
where gik = ri · rk .
Using the summation convention, the expression of
the first fundamental form of the surface (Mitrinovic
& Ulcar 1969) can be written in the form
thus representing the first fundamental tensor of the
surface.
Let consider the net formed on the surface by the
For the surface in the form x3 = z(x1 ,x2 ), assuming coordinate curves u1 = const and u2 = const (Fig. 4).
u1 = x1 , u2 = x2 , one has that One can evaluate the area dA of the surface elements
d as

Thereafter the area A of the region (u1 ,u2 ) of the

surface r = r(u1 ,u2 ) can be reassembled and defined
in the form

representing the double integral on  of the elements

d  with area dA.

The arch length s of the curve ui = ui (t) (i = 1, 2)

between the points t1 and t2 is then given by definite
integration of ds in Eq. (9) between the two bounds t1
and t2 , as

Figure 4. The net formed on the surface by the coordinate

curves u1 = const and u2 = const.

305
Figure 5. Normal section curve n and surface curve :
relevant normal vectors and curvatures.
Since the normal vector N to the surface
r = r(u1 ,u2 ) has the direction of the vector r1 × r2 ,
one has
Let consider now a normal section n of the sur-
face at P (Fig. 5), i.e. the intersection of a surface
with a plane which passes through the surface nor-
mal at a point P of the surface (thus containing the
unit vector normal N to the surface at that point). The
normal curvature of the normal section at P is denoted
by kn = 1/ρn with ρn the radius of curvature.
According to the Meusnier Theorem, one can relate
the radius of curvature ρ of the generic surface curve
[ui = ui (t) (i = 1,2)] with curvature k = 1/rρ at the
point P and the angle θ between the normal vectors n
and N at P as follows
where hik = r1 × r2 · rik = − ri · Nk ,
with rik = ∂2 r/∂ui ∂uk and Nk = ∂N/∂uk .
The quadratic differential form hik dui duk in Eq. (19)
is the second fundamental form of the surface and
the symmetric second-order tensor formed on the
hik components
represents the second fundamental tensor of the sur-
face, with determinant h = det(h) = h11 h22 − (h12 )2 .
At the a certain point, i.e. the umbilical point, one
has that g = λh, and the curvature is the same for all
306
Figure 6. Principal curvatures of the surface at P.
normal sections. If this is not the case, then the normal
curvature kn at the point P has two extreme values
which determine two tangents at P.
The stationary values of kn are the principal curva-
tures of the surface at P and the corresponding tangents
are the principal directions at P.
The principal curvatures are denoted by k1 and k2
and are the roots of the equation
Denoting by φ the angle between any direction at P
and the principal direction at P corresponding to “1”
(Fig. 6), the normal curvature kn in the given direction
at P is defined by Euler’s formula, as
Actually by considering the product, K = k1 k2 = h/g
of the principal curvatures, which represents the Gaus-
sian curvature of the surface at P, one may check if the
surface is developable, since, in this case, the Gaussian
curvature is zero at ali points of the surface.
2.2 Shells of revolution
With reference to shells of revolution (Baratta & Corbi
2007b), one may refer to the generic element ABCD
identified by the cut of the surface along vertical planes
passing through two couples of adjacent parallels and
meridians, as shown in Fig. 7. The generic point on
the shell may be located by its Cartesian coordinates
x1 ,x2 in the reference system (O x1 ,x2 ,x3 ), whilst the
surface equation is x3 = z(x1 ,x2 ).
After denoting by r the vector detecting the position
of the generic point of the shell, whose projection on
the x1 x2 plane is denoted by r, one can move from
 Figure 8. Normal section of a shell of revolution and
principal curvatures.

Figure 7. Shell of revolution: Cartesian and polar

co-ordinates.

Cartesian coordinates x1 ,x2 to polar coordinates θ, r in

the horizontal plane, leaving <z> as the ordinate axis,
as shown in Figures 7 and 8, where the position of the
edge A of the element ABCD of the shell is identified.
Moreover a point on the meridian is located by the
angle ϕ.
Therefore the following relations exist between
polar and Cartesian coordinates at the generic point

Figure 9. Generic shell element of dimension dA.

specific case r1 = rr , r2 = rθ , after some algebraic

where the variable z depends on the shape of the merid- operations and further developments, one may obtain,
ian line of the shell, and it is thus dependent on r, i.e. according to the Meusnier theorem (Mitrinovic &
z = z(r). Ulcar 1969, Ugural 1999), the principal curvatures k1
After calculating the first order derivatives of r, one and k2 solving the equation |h − k g| = 0 with respect
can evaluate the first fundamental form of the surface to the curvature k, as
gik dxi dxk (with dxi , dxk the variables), which involves
the coefficients gik = ri · rk and, with reference to the
generic surface element, calculate the length of the
arch curves along the parallel dsθ and meridian dsr
directions respectively, and its area dA (Fig. 9) as
follows

where ρ1 and ρ2 are the principal radii of curvature.

Since the normal vector N of the surface at point P
has the direction of the vector rr × rθ , one has

By the second fundamental form of the surface

(Ugural 1999) hik dxi dxk (with dxi , dxk the variables),
involving the coefficients hik = r1 ×√r2g · rik , where in the

307
whence, after some algebraic operations, one gets
In the case of shell of revolution the local principal
radii of curvatures on the surface can be immediately
identified. With reference to Figure 8, at some point
on the shell the meridian has a radius of curvature
ρ1 , which is the radius of curvature of the small arch
length ds.
The patch of the shell cut out by two meridians and
two parallel circles has, however, a second radius of
curvature ρ2 ; the normals at adjacent points on two
meridians intersect on the axis of the shell, and ρ2 is
the length of the normal from the point on the shell to
the axis.
Thus the shell surface is described by the four
parameters θ, ϕ, ρ1 , and ρ2 , but these quantities are
not all independent and, for example, with reference
to Figure 8, r = ρ2 sin ϕ.
Figure 9 shows the small element of the
shell and it is seen that the dimensions of the
element are
which can be shown to be in agreement to what
previously found in Eq. (24).
By comparison with Eq. (24) one has
After some further developments, one gets
308
Figure 10. Local equilibrium of the generic shell element.
whence
which yield the same result as in Eq.(25).
3 POSSIBLE APPROACHES TO THE STATIC
ANALYSIS
3.1 Set up of the problem
When addressing problems relevant to the static anal-
ysis of masonry vaults one should handle the two
problems of equilibrium with applied loads an admis-
sibility with respect to the masonry material which,
as well known, exhibits a very reduced capacity of
resisting tensile stresses (Baratta 1984,1991, Baratta &
Corbi 2005, 2006, 2007).
So the above reported treatment can be used, on
one side, for describing the real vault geometry, i.e.
the geometry of its extrados and intrados surfaces z1
and z2 bounding its thickness and the geometry of
its mid-surface, and, on the other side, for describing
the geometry of its membrane surface z, envelop-
ing the local membrane stress resultants in equilib-
rium with the applied loads (Baratta & Corbi 2006a,
2007).
In this latter case both the equilibrated local stress
resultants and the membrane surface z are unknown.
So the approach for the static study of the vault
could consist of searching for the membrane surface
both satisfying equilibrium and material admissibility.
As regards to equilibrium, considerations about
stresses and external loads acting on the element
of Figure 10 yield, as usual, three equations of
equilibrium.
The external load which may be acting on the ele-
ment has been resolved into the three perpendicular
components pϕ , pθ , pn ., while stresses are those shown
in the figure, yielding, after some algebraic develop-
ments, the three equilibrium conditions respectively in
the tangential direction (direction of θ), in the tangen-
tial direction (direction of ϕ), and in the radial direction
(i.e. normal direction), as follows
In the case when one considers an axial-symmetric
loading the shear stress resultant Nθϕ is zero every-
where and one has
When referring to the equilibrium conditions under
not axial-symmetric Eqs (32) and axial-symmetric
load Eqs (33), one should remember that, as men-
tioned in the above, the involved geometric quantities
(angles and radii of curvature) depend on the variable
z expressing the equation of the membrane stress sur-
face. Therefore, besides the unknown membrane stress
resultants, one also should search for the membrane
surface expressions both satisfying equilibrium (in the
forms of Eqs (32) or (33) and material admissibility
(which implies that the surface must be contained in
the vault profile z1 ≤ z ≤ z2 ).
After individuating the set of admissible membrane
surfaces (in terms of admissibility and equilibrium),
a possible approach in order to find the solution in
terms of stresses is to set up a Complementary Energy
problem to be formulated as a kind of extension to
masonry vaults of the classical analogous energetic
approach for linearly elastic structures.
This kind of approach has been already successfully
applied to masonry arches modeled by the Not Resist-
ing Tension (NRT) assumption (Baratta 1984,1991,
309
Baratta & Corbi 2005a, 2006, 2007, Heyman 1977),
showing to produce in its numerical implementation
results in excellent agreement with experimental data
(Baratta & Corbi 2005b, 2006b).
In order to undertake this approach the expression
of the Complementary Energy embedded in a masonry
vault element should be evaluated. In case of NRT
assumption, one should consider that the generic ele-
ment appears to be partially resistant; this means that,
denoting by u the distance of the membrane surface
from the upper profile of the vault element, the posi-
tion of the neutral surface (i.e. characterized by null
stresses) is located at a distance of 3u from the extrados
of the element.
For a bi-linear distribution of the normal stress on
the volume element, when neglecting the shear stress
component, the elastic energy can be calculated as
follows
with Âr , Âθ the areas of the compressed part of the
volume element respectively in the two cross sections
containing dsr and dsθ , er and dGr respectively the
eccentricity of the solicitation centre with respect to
the mass centre Gr , and the distance of the mass cen-
tre from the neutral plane in the two cross sections,
and E and ν respectively the Joung modulus and the
Poisson coefficient.
The final Complementary Energy functional C
expression is given by adding to the elastic energy
term L, the energy related to the work developed by
the constraint reactions R, as C = L + R.
The solution in terms of stresses can then be
searched for by numerically implementing the min-
imization of the Complementary Energy functional
under the condition that the solution itself is respectful
of the above individuated equilibrium equations and of
admissibility.
As an alternative an inverse approach can be suc-
cessfully outlined also leading to the analytical expres-
sions of the membrane stress functions for the different
vault typologies. In this case the spatial surface equi-
librium problem can be coupled to the one projected
in a selected plane according to the classical Pucker’s
approach. One can then refer to the 2D-projected prob-
lem, and introduce a stress function which is built up in
such a way to a priori satisfy some of the equilibrium
conditions.
The key of the problem lays then in the coupling of
the stress function with the membrane function.
 So the stress function should be modeled in such a
way to recognize, after imposing the remaining equi-
librium conditions and admissibility, the membrane
stress functions that can be assumed as the membrane
surfaces relevant to different vault geometries.
Some practical applications relevant to the cases
of the barrel vaults with indefinite length, the barrel
vaults confined at their extremity cross-sections, and
the spherical domes are developed in details in papers
by the author (Baratta & Corbi 2007a, 2007b; Baratta
et al. 2008).
To this regard, one should emphasize that this is
not at all a trivial objective to be pursued. Even sim-
ple applications, which means the specialization of the
described approach to vaults of simple shapes, require
a pretty hard work for setting up and checking the
expressions of the relevant stress functions.
As a counterpart, this approach allows to obtain
solutions in analytical form, which represents its major
result.
Thereafter the solution for each case can be com-
pleted by setting up the energetic approach as men-
tioned in the above.
Definitively this approach is pretty complex since
it requires to hypothesize and test a number of analyt-
ical functions for any vault typology, which represents
a consistent effort even for simple vaults geometries;
anyway it has the big advantage and original research
result, never available in literature before, to produce
the explicit analytical expressions of the stress and
membrane functions for the single cases, i.e. to give
the solutions in analytical form.
4 CONCLUSIONS
The paper presents some of the fundamental features
involved in the study of masonry vaults.
The first part of the paper is devoted to show the
main difficulties encountered in the description of the
vaults geometries and to introduce the differential tools
for handling this problem.
310
In the second part one outlines some discussion on
possible approaches to the analysis of masonry vaults,
also as regards to the possibility of obtaining analyti-
cal expressions of the solutions in terms of membrane
surfaces for different vault typologies.
REFERENCES
Baratta, A. 1984. Il materiale non reagente a trazione come
modello per il calcolo delle tensioni nelle pareti murarie.
J. Restauro. 75/76: 53–77.
Baratta, A. 1991. Statics and reliability of masonry structures.
In F.Casciati & J.B.Roberts (eds), Reliability Problems:
General Principles and Applications in Mechanics of
Solids and Structures. Udine, CISM.
Baratta, A. & Corbi, O. 2005a. On variational approaches
in NRT continua. J. of Solids and Structures. 42:
5307–5321.
Baratta, A. & Corbi, O. 2005b. Fibre Reinforced Com-
posites in Civil Engineering. Experimental Validation of
C-Fibre Masonry Retrofit. Intern. Journal of Masonry
International, British Masonry Society. 18(3): 115–124.
Baratta, A. & Corbi, O. 2006a. Analysis of masonry vaulted
systems: the barrel vaults. In Structural Analysis of His-
torical Constructions, New Delhi.
Baratta, A. & Corbi, O. 2006b. Frp Composite Reinforce-
ments on Masonry Vaults: Effectiveness and Reliability.
In A. Mirmiran & A. Nanni (eds) Composites in Civil
Engineering., Miami: 367–370.
Baratta, A. & Corbi, O. 2007a. Towards a new theory for
masonry vaults assessment. 7ICM, the 7th International
Masonry Conference. London, U.K.
Baratta, A. & Corbi, O. 2007b. Basic Theory For No-Tension
Vaults: Statics And Strength. AIMETA’07, 18◦ Congresso
Italiano di Meccanica Teorica ed Applicata, Brescia.
Baratta, A., Corbi, I. & Corbi, O. 2008. Stress Analysis of
Masonry Structures: arches, walls, and vaults. SAHC08,
VI International Conference on Structural Analysis of
Historical Constructions, Bath, UK.
Heyman, J. 1977. Equilibrium of shell structures, Oxford
University Press 1977.
Mitrinovic, D.S. & Ulcar J. 1969.Differential geometry tuto-
rial texts. Wolters-Nordhoff Publishing.
Ugural, A.C. 1999. Stresses in plates and shells. McGraw-
Hill.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Damage identification in masonry structures with vibration

measurements

L.F. Ramos & P.B. Lourenço

ISISE, University of Minho, Guimarães, Portugal

G. De Roeck
Catholic University of Leuven, Belgium

A. Campos-Costa
LNEC, Lisbon, Portugal

ABSTRACT: The paper aims at exploring damage assessment in masonry structures at an early stage by
vibration measurements. For this purpose, one approach is proposed combining global and local ND methods.
To further evaluate the approach, one masonry tower in Portugal was studied together with one wall model in the
laboratory. The model was built as reference, undamaged, state. Afterwards, progressive damage was induced
and sequential modal identification analysis was performed at each damage stage, aiming to find adequate
correspondence between dynamic behavior and internal crack growth. The paper presents all the analyses carried
out with the aim to detect and locate the damage by means of vibrations measurements.

1 INTRODUCTION
Damage on masonry structures mainly relates to
cracks, foundation settlements, material degradation
and displacements. When cracks occur, generally they
are localized, splitting the structures in macro-blocks.
Dynamic based methods to assess the damage are
attractive to this type of structures due to the present
requirements of unobtrusiveness, minimum physical
intervention and respect of the original construc-
tion. The assumption that damage can be linked to
a decrease of stiffness also seems to be reasonable to
this type of structures.
The present paper deals with the problem of dam-
age identification by using Global and Local damage
identification techniques. It is advantageous to have
two categories for damage identification: (a) the vibra-
tion based damage identification methods, currently
defined as Global methods, because they do not give
sufficiently accurate information about the extent of
the damage, but they can identify its presence and
define its precise location (e.g. Chang et al., 2003); and
(b) the methods based on visual inspections or experi-
mental tests, such as sonic tests, radar tests, radiograph
and thermal field methods (e.g. Doherty, 1987), also
called as Local methods.
Many methods are presented in literature for
damage identification based on vibration signatures,
see Montalvão et al. (2006), but there are only a few
papers on the application to masonry like structures.
An important task before damage can be identified
from vibration characteristics is the study and subse-
quent elimination of the environmental effects Peeters
(2000).
2 PROPOSED APPROACH FOR DAMAGE
IDENTIFICATION
The current practice of structural health monitoring is
based mainly on periodic visual inspections or con-
dition surveys but, during the last decade, software
and hardware developments made continuous moni-
toring possible, Chang et al. (2003). Typically, one can
install hundreds of sensors in a structure and read the
data in real time. Therefore, the present focus of inter-
est is what type of information is important from the
structural point of view and how should the data be
processed and stored for damage analysis, Londoño
(2006)?
Given the latest advantages in the field of damage
identification analysis and concerning monitoring, it
is proposed to monitor historical masonry structures
in four phases (Ramos, 2007):
1. The first phase is the data collection of the
structure, including the historical information,

311
 geometrical and topographic survey, damage sur-
vey, the mechanical characterization of materials
with Non Destructive (ND) tests, a global dynamic
modal test and a numerical model analysis for static
and dynamic calibration. This is the first approach
to the structural behaviour in the assumed healthy
condition at time “zero”;
2. In the second phase the health monitoring plan can
resort to a limited number of sensors (e.g. a pair of
reference accelerometers, strain gauges at critical
sections, temperature and humidity sensors, etc).
Data should be stored periodically and the mon-
itoring system should be able to send an alarm.
Environmental and loading effects should be stud-
ied and the presence of damage should be detected
by the global modal parameters;
3. In the third phase, after alarm triggering, a full-scale
dynamic survey with more sensors and measur-
ing points should be performed. In this phase the
“health condition” of a structure is studied with
more detail. Damage identification methods should
be applied to the structure after filtering the envi-
ronmental effects. The aim of the dynamic methods
is to confirm and locate the (possible) damage in a
global way;
4. In the last phase, a local approach with visual and
complementary ND tests should be performed to
locally assess the damage and classify it.This can be
carried out with sonic test or radar tests, depending
on the access conditions of the structure. This local
approach can give a better definition of damage.
The global and local approach should be considered
as complementary tasks. For the case of historical con-
structions these two approaches seem to be suitable,
since they are ND procedures to evaluate the health
conditions.
The following sessions present one case study
and one laboratory simulation of the University of
Minho, where the first three phases of the preceding
methodology were already applied.
3 STRUCTURAL ASSESSMENT
For the first phase of the proposed approach it is pre-
sented the case study of Mogadouro Clock Tower. The
Mogadouro Clock Tower is located inside the cas-
tle perimeter of Mogadouro (see Figure 1), a small
town in the Northeast of Portugal. It has a rectangu-
lar cross section of 4.7 × 4.5 m2 with and interior hole
of 2.5 × 2.3 m2 , and a height equal to 20.4 m. It was
built in stone masonry with granite and shale, with dry
joints at the corners and with mortar joints in the cen-
tral part of the walls. The walls have, on average, 1 m
of thickness and are composed by an irregular bond of
stones.
312
Figure 1. Mogadouro Clock Tower: (a) and (b) general view
of the tower with severe damage; and (c) the tower after the
strengthening works.
Due to lack of maintenance, severe damage was
observed in the tower in 2004. The damage was char-
acterized by cracks, material degradation, out-of-plane
displacements, loss of material in some parts, and bio-
logical growth, see Figure 1a. The most severe damage
was a pair of cracks in the East and West façade, see
Figure 1b, that divided the box cross section of the
tower in two separated U bodies, leading to a lower
safety condition.
A geometrical survey of the structure with pho-
togrammetry software allowed registering all anoma-
lies and dimensions. The dimensions of the structural
survey have an error about ±5 cm.
To re-establish the tower safety, consolidation
works were carried out in 2005. The remedial treat-
ment included: injections for the walls consolidation,
replacement of deteriorated materials and the installa-
tion of tie-rods at two levels with very light prestress-
ing stress. Figure 1c presents the final appearance of
the tower after strengthening.
3.1 Modal identification before and after the
rehabilitation works
In the case of the Mogadouro Tower, damage was
evident and a dynamic based damage identification
analysis to detect its presence was not necessary. Nev-
ertheless, the dynamic based methods could be used
to locate and to assess the damage. Therefore, two
dynamic modal identification tests were performed
before and after the structural rehabilitation. The aim
was to compare the response of the tower with different
structural conditions in order to localize the damage,
to archive qualitatively the extent of damage present
before the rehabilitation works, and to evaluate the
efficiency of the rehabilitation works.
Table 1 and Figure 2 present frequencies and mode
shapes, as well the Modal Assurance Criteria (MAC)
values, for the two structural conditions. On average,
the frequencies increase 50%.
Table 1. Estimated natural frequencies values.

Before After

Mode ω CVω ω CVω ω

shape [Hz] [%] [Hz] [%] [%]

1st 2.15 1.85 2.56 0.21 +19.28

2nd 2.58 1.05 2.76 0.30 +6.70
3rd 4.98 0.69 7.15 0.27 +43.67
4th 5.74 1.56 8.86 0.47 +54.37
5th 6.76 1.13 9.21 0.21 +36.13
6th 7.69 2.94 15.21 2.24 +97.87
7th 8.98 1.21 16.91 1.40 +88.27

Figure 3. Selected updating parameters.

(FEMU). The nonlinear least square method imple-

mented in MatLab (2006), function lsqnonlin, was
used together with FE package DIANA (2006) to
compute the numerical modes. The objective func-
tion π to be minimized is composed by the residuals
formed with calculated and experimental frequencies
and mode shapes, given by:

where mω denotes the number of eigen frequencies

Figure 2. Estimated mode shapes before and after strength-
ening, together with the MAC values.
Analyzing the modal displacements, local protuber-
ances can be observed in the areas close to the cracks
and in the upper part before the rehabilitation. This is
due to the presence of severe damage. The structure
only behaves monolithically after the rehabilitation.
The global parameter results are consistent with
the two structural conditions, i.e. the structure after
rehabilitation has a new and higher stiffness.
3.2 Numerical modeling
Aiming at a better understanding of the two struc-
tural conditions and to better evaluate the efficiency
of the strengthening, FE analyses were also addressed.
The numerical structural assessment was carried out
by the application of the FE Model Updating method
taken into account, mϕ is the number of normalized
eigen modes taken into account and Wω and Wϕ are
weighting diagonal matrices for the frequencies and
mode shapes, respectively. Note that in Eq.(1) exper-
imental and numerical mode shapes are normalized
in a way that the maximum real value of the modal
displacement is equal to one, in order to be compa-
rable with the residuals from the frequency values
normalized by the experimental results.
A 3D model with brick (20 noded) elements was
prepared. The model is presented in Figure 3 with the
selected updating parameters (Young’s modulus). The
planar roof surfaces were modeled with 6 noded shell
elements to simplify the mesh.
Table 2 presents the final results together with
the difference and the relative values of the updating
parameters. All the updating variables suffered a sig-
nificant increase, with exception of the material at the
corners (E5 ), at the foundation (E8 ), and at the North
wall (E2 ). For the corners and for the foundation these
results are acceptable because the stone masonry at
the corners was better than in the central parts of the
walls, and the foundation material tries to represent
not only the structural element but also the boundary

313
Table 2. Results of the updating analysis.

Updating Before After Diff. Relative

parameters [GPa] [GPa] [GPa] values

E1 [GPa] 0.687 1.974 +1.287 2.87

E2 [GPa] 2.210 2.210 – 1.00
E3 [GPa] 0.302 1.075 +0.773 3.56
E4 [GPa] 0.276 0.804 +0.528 2.91
E5 [GPa] 3.870 3.875 +0.005 1.00
E6 [GPa] 0.380 1.210 +0.830 3.18
E7 [GPa] 0.083 0.195 +0.112 2.35
E8 [GPa] 5.997 5.997 – 1.00

conditions at the base. For the North wall a Young’s

modulus increase was expected, but in the strength-
ened condition the North wall had a higher Young’s
modulus, which might indicate that the wall was in
better condition than the others. The others updating
variables on average they increased three times the
initial value.
The results indicate that the damage on the tower
was severe on the central parts of the South, East
and West façades and at the upper part of the tower.
These areas in the structure are the ones which suffer
large mortar loss due to deterioration, meaning that
the results can be acceptable. Finally, one can con-
clude that the strengthening works were efficient at
re-establishing the tower safety.
4 MONITORING TASK
Figure 4. Monitoring system: (a) and (b) sensors location;
and (c) logger software.
The LabView (2006) software was used to measure
and acquire accelerations, see Figure 4c. In parallel, a
combined sensor connected to the laptop through serial
cable is recording the ambient temperature and relative
air humidity. The environmental data is acquired in the
laptop by interface software provided by the supplier
of the combined sensor.
Every hour, 10 minutes of ambient vibrations in the
three channels with a sampling frequency of 100 Hz
are acquired without any triggering. The data is then
saved in ASCII files.
Finally and for modal estimation, an automatic
procedure based on SSI/Ref method (Peeters and
De Roeck, 1999) was implemented in MatLab.

The second phase of the proposed approach is the

dynamic monitoring task, which can be performed
with a limited number of sensors. This task has
been carried out since April, 2006, in the Mogadouro
Clock Tower. The aim is to evaluate the environmen-
tal and loading effects and to detect any possible non
stabilized phenomena in the structure (damage), by
studying the global dynamic parameters.
A low-cost monitoring system was chosen for this
task. The system is composed by three piezoelectric
accelerometers, connected by coaxial cables to a USB
data acquisition card with 24 bits resolution, provided
with anti-aliasing filters, which is connected to a Pen-
tium II® laptop with an uninterruptible power supply
device.
Two points (A1 and A2) were selected in the middle
of Level 2 to acquire accelerations in three directions,
see Figure 4a and b. In this way, all the mode shape
components of the first five eigenfrequencies can be
studied, including the torsion mode. The environmen-
tal measurements are acquired in the point TH, close
to the datalogger (D).
4.1 Filtering the environmental and loading
effects for damage detection
The study of the environmental and loading effects
was initially based on the assumption that the modal
response is changed by three independent variables:
temperature, relative air humidity and the level of exci-
tation. For the later and for every event, it was decided
to use the average Root Mean Square (RMS) value of
the three acceleration channels.
In general, there is a positive relation between
temperature and frequency, and a negative relation
between humidity and frequency and between level
of excitation and frequency.
Here it should be stressed that a significant fre-
quency shift related to the walls water absorption
happens twice a year. Figure 5 presents the first
resonant frequency variation compared with the envi-
ronmental variables along short period. During this
period the temperature did not change significantly,
while the relative air humidity after 18th of October is
close to 100%, i.e. the first raining season on the site.

314
Figure 5. Environmental effects of the dynamic response.
After 22nd of October the frequency values decrease
linearly with humidity almost constant. This indicates
that the structure absorbs water and the mass changes,
reducing the frequency values because the two quan-
tities are inverse related. It can also occur that water
leads to a stiffness reduction, at least in the lime mor-
tars. The inverse drying phenomenon also happens
during the hot period.
Since the environmental and the loading variables
are changing the modal parameters of the tower, an
attempt to model the dynamic response according to
the three variables was carried out. It was decided to
use a procedure similar to the one used by Peeters
(2000) with AutoRegressive output with an eXoge-
neous input parts models (ARX models). Here, ARX
models with multiple inputs and a single output (MISO
models) were computed in MatLab (2006), function
arx, to model each frequency value. The multivari-
able ARX model with n inputs u and one output y
is presented by:
where Aq is a scalar with the delay operator q−1 , Bq is a
matrix 1 × n, and e is the unknown residuals. For dam-
age detection, confidence intervals were establish and
by analyzing the outliers is possible to detect damage.
Figure 6 shows for the first natural frequency the
fitting model through the normalized frequency and
simulated errors with the 95% confidence intervals
ci. In general, the model represents the frequency
315
Figure 6. Numerical simulation: (a) ARX model; and
(b) the residuals distribution.
variation, but it doesn’t account the water absorption
phenomenon, because the relative humidity varia-
tion does not totally represent that change. Further-
more, the damage is detected by frequency shifts that
significantly go outside the confident intervals ci.
Concerning the numerical frequency simulation, it
should be stressed that the calibration period of one
year might be not enough for having a tuned model. A
longer period of, at least, three years should be used
for calibration in order to have a reliable model for
frequency prediction.
5 DAMAGE LOCALIZATION
For damage localization (the third phase of the pro-
posed approach) a laboratory simulation of a wall
model is presented here. The replicate of histori-
cal constructions was built as reference, undamaged,
state. Afterwards, progressive damage was induced
and sequential modal identification analysis was per-
formed at each damage stage, aiming at finding ade-
quate correspondence between dynamic behavior and
internal crack growth.
The wall model was built with clay bricks with
210 × 105 × 55 mm3 , handmade in the Northern area
of Portugal. The adopted clay bricks have low com-
pression strength and the adopted Mapei®mortar for
the joints has low mechanical properties for the joints,
Figure 7. Wall model: (a) general view; and (b) dimensions.

trying to be representative of the materials used in the

historical constructions. The wall has a length equal to
1.08 m, a height equal to 1.10 m, and a thickness equal
to 0.105 m, which matches the bricks thickness, see
Figure 7. The thickness of the joints is about 1.0 cm.

5.1 Static tests

Fourteen Damage Scenarios (DS) were induced with
the application of two static forces: a constant vertical
force to replicate existing vertical compressive stresses
and a horizontal force to produce shear stresses. The
vertical load was transmitted through a group of three
steel beams with appropriate devices to distribute uni-
formly the load to the wall. The last beam was direct
glued to the wall. The in-plane stresses were applied
with the aim of producing bending and shear cracks,
i.e. to reproduce the common crack pattern present in
the masonry piers that suffer earthquake actions.
To achieve several controlled cracks, three series
of static tests were carried out. Figure 8 presents foo
the last series (Series C) the test apparatus, the static
response of the wall and the final crack pattern. From
the static response is possible to observe the stiffness
decrease, where the horizontal displacement indicated
corresponds to the maximum or top value (LVDT 1).
The final crack pattern includes three cracks. One
bending crack is at left bottom part, and another bend-
ing crack is at right top part of the wall, namely crack
c1 and c2 , respectively. The third crack (c3 ) is a shear
crack in the diagonal direction of the wall.
5.2 Dynamic tests
Between each DS, modal identification analysis using
output-only techniques was performed. The selected
sensors for the dynamic tests were the accelerometers
and strain gauges with quarter bridge configuration.
The strain gauges were selected to measure modal
strains in order to directly estimate modal curvatures,
which are quantities more sensitive to damage than the
modal displacements (Ramos, 2007).
Figure 8. Wall static tests: (a) test apparatus; (b) static
response of the last series of tests; and (c) final crack pattern
from the front and back sides of the wall.
A regular grid of five vertical lines and seven hori-
zontal lines was chosen for the accelerometers. In the
case of the strains gauges, it was decided to use an
array of three vertical lines and five horizontal lines.
Figure 9 shows the location of the measuring points
where the reference points are given in a grey box.
The accelerometers were bolted to aluminum plates
that were directly glued to the wall. The vertical
strain gauges have 12 cm of length and the horizontal
strain gauges have 6 cm of length. This way, the verti-
cal strain gauge crosses, at least, three joints and the
horizontal strain gauge only one joint.
Two different excitation types were used during
the identification tests: (a) natural and ambient noise
present in the laboratory; and (b) random impact exci-
tation in space and in time with a hammer. The second
excitation type was used because it was impossible
to estimate accurately modal curvatures with ambient

316

Figure 9. Wall dynamic tests: (a) and (b) accelerometers and
strain gauges locations, respectively; and (c) and (d) details
of the ambient dynamic tests.
Figure 10. Environmental effects inside the laboratory.
excitation. The impact forces were about 2% of the
specimen mass.
The long-term changes of the environmental effects
inside the laboratory where also studied. Figure 10
presents the average values for ambient and sur-
face temperature and relative air humidity measured
for each DS. The average temperature values were
about 17◦ C with a low increasing trend. The aver-
age humidity is about 69% with ±7% of maximum
amplitude. Compared with the values obtained in the
Mogadouro Clock Tower, it is expected for values of
that order that the dynamic response of the wall does
not change significantly and therefore it is assumed
that the environmental effects inside the laboratory can
317
be neglected in the subsequent damage identification
analysis.
5.3 Damage analysis
For damage detection, the analysis was carried out
by observing frequency sifts in the natural resonant
frequencies. In the case of the damage localization
analysis, a group of damage methods was selected aim-
ing at providing an adequate approach for analysis. The
selected methods are (see Doebling et al., 1996, and
Montalvão et al., 2006, for details):
• The COMAC values;
• The Parameter Method (PM);
• The Mode Shape Curvature Method (MSCM);
• The Damage Index Method (DIM);
• The Sum of all Curvature Errors method (SCE);
• The FE Model Updating method (FEMU).
The selection of a group of methods can be dis-
cussible. Up to now, there is no single method which
gives accurate results for damage localization and for
all types of structural systems (Farrar and Doebling,
1998; and Choi et al., 2005). Therefore, the main issue
is to obtain a wide perspective of the problem and con-
clusions on damage identification, taking into account
that different methods provide different results. If sig-
nificant damage is present in the structure, the results
provided from different methods would converge in the
identification, giving more confidence to the analyst.
All the selected methods have one common aspect:
they all use spatial modal information of the struc-
ture, through mass-scaled or non-scaled mode shapes
φ and ϕ, respectively (or/and through mass-scaled
or non-scaled curvatures mode shapes φ and ϕ ,
respectively). As damage is a local phenomenon, these
quantities are useful to locate damage, especially the
modal curvatures.
5.3.1 Global damage detection
The SSI/Principal Component method implemented
in the tool ARTeMIS (SVS, 2006) was used to esti-
mate the modal parameters for all the fourteen DS.
As an example of the frequency shifts along the DS,
Table 3 presents for the Test Series C (the last series)
the frequency results for ambient excitation. The fre-
quency values are presented together with the value
±2σω as a 95% confidence interval, and the fre-
quency differences ω to the respective Reference
Scenario (RS).
In general, the significant frequency decrease, i.e.
higher that 2σω (given in a grey box on Table 3), where
around the DS where the crack was visually localized,
with exception of mode 1 for Series C, as can be seen
in Table 3.
To conclude about the global detection results,
it seems that the modal properties of the masonry
specimen are sensitive to the damage progress.
Table 3. Frequency results for ambient excitation [Hz].

†
Mode RSC DSI-C DSII-C DSIII-C DSIV-C

1st ω 3.41 3.46 3.54 2.99 2.81

2σ 0.45 0.07 0.29 0.05 0.09
ω – 0.06 0.13 −0.42 −0.60
2nd ω 12.49 12.44 11.72 10.82 9.27
2σ 0.05 0.07 0.15 0.16 0.2
ω – −0.05 −0.77 −1.67 −3.22
3rd ω 18.29 18.24 17.56 16.76 16.03
2σ 0.12 0.07 0.18 0.19 0.3
ω – −0.05 −0.73 −1.53 −2.26
4th ω 35.63 35.38 34.41 33.11 32.52
2σ 1.12 0.26 0.37 0.33 2.78
ω – −0.25 −1.22 −2.52 −3.11
Figure 12. Updating parameters for the damage analysis.
†
– Damage scenario in which the crack c3 was identified by
visual observation.
Table 4. Relative values for the updating parameters.

Up DSII-C DSIII-C DSIV-C

E1 1.00 0.98 1.00

E2 1.00 1.00 1.00
E3 1.00 1.00 1.00
E4 1.00 0.53 0.01
E5 0.07 0.05 0.01
E6 1.00 0.08 0.01
E7 0.91 0.85 0.51
E8 0.03 0.09 0.29
E9 0.69 0.78 0.36

Figure 11. Combination of non-model based results for the

comparison between DSIII-C and DSIV-C .
5.3.2 Global damage location
The first methods to be applied here were the non-
model based methods, namely the COMAC, PM,
MSCM, DIM and SCE. In these methods the curva-
ture mode shape values close to zero were neglected
in order to avoid noise contamination in the results.
The methods were applied only for the case of random
impact excitation.
Concerning the performance of the non-model
based methods, the PM calculated for modal curva-
tures, the MSCM, the DIM and the SCE gave similar
results, with better agreement for the last three meth-
ods. They were able to locate the damage in the vicinity
of the experimental crack locations.
The results combination of the selected methods is
extensively presented elsewhere (Ramos, 2007). As an
example, Figure 11 shows the locations results for the
case of the last DS when is compared with the preced-
ing one (relative comparison). The damage location is
given by the circles.
The damage localization with FEMU was carried
out with an approach similar to the one presented in
Section 3.2. For the case of Series C, eight thin bands
of elements and one square of elements with different
Young’s modulus as updating parameter were consid-
ered. Figure 12 shows the selected bands together with
the final crack pattern.
In this series, the damage was expected to be local-
ized at position E5 and E6 (crack c3 ) and it was also
expected to observed the crack growth at positions
E8 and E9 (crack c2 ). Table 4 presents the relative val-
ues for the updating parameters, when compared to
the values obtained for the same type of analysis but
calculated to the RSC .
The damage pattern was obtained with the updating
analysis DSII-C and DSIII-C . In the last updating anal-
ysis (DSIV-C ) the results were not consistent with the
observed cracks, as damage at position E4 was also pin-
pointed. This might be related to the large differences
between the RSC and the last DS of test Series C.
6 CONCLUSIONS
In this paper a methodology was proposed based
on vibration measurements to identify damage in

318
masonry-like structures. The methodology was then
applied to the Clock Tower of Mogadouro, Portugal,
and to a laboratory wall specimen.
From the presented analyses, it seems that the
proposed methodology is useful and applicable to
historical masonry structures. The natural frequency
observation seems to be a reliable approach for
damage detection, while the mode shape curvatures
changes are reliable quantities in the case of damage
localization.
REFERENCES
Chang, P.C.; Flatau, A.; and Liu, S.C. 2003. Review Paper:
Health Monitoring of Civil Infrastructure, Structural
Health Monitoring, Vol. 2 (3), pp. 257–267
Choi, S.; Park, S.; and Stubbs, N. 2005. Non-destructive Dam-
age detection in Structures using Changes in Compliance,
International Journal of Soils and Structures, n.42, pp.
4494–4513
DIANA 2006. DIANA-9 Finite Element Analysis, User’s
Manual – Release 9, TNO, Netherlands
Doebling, S.W.; Farrar C.R.; Prime, M.B.; and Shevitz D.
1996. Damage identification and health monitoring of
structural and mechanical systems from changes in their
vibration characteristics: a literature review, Los Alamos
National Laboratory, NM, 132 p.
319
Doherty, J.E. 1987. Nondestructive Evaluation, Handbook on
Experimental Mechanics, A.S. Kobavashi Edt., Society
for Experimental Mechanics
Farrar, C.R. and Doebling, S.W. 1998. Damage Detection
and Evaluation II – Field Applications to Large Struc-
tures, Modal Analysis and Testing, Júlio Silva and Nuno
Maia (Editors), NATO Science Series E, Vol. 363, Kluwer
Academic Publishers, London, pp 345–378
Londoño, N.A. 2006. Use of Vibration Data for Struc-
tural Health Monitoring of Bridges, PhD Thesis, Carleton
University, Ottawa, Canada
MATLAB 2006. MATLAB User Manual, Release 7.2, The
MathWorks, USA
Montalvão, D.; Maia, N.M.M.; and Ribeiro, A.M.R. 2006.
A Review of Vibration-based Structural Health Moni-
toring with Special Emphasis on Composite Materials,
The Shock and Vibration Digest, Vol. 38, No. 4, pp.
295–324
Peeters, B. 2000. System Identification and Damage Detec-
tion in Civil Engineering, PhD Thesis, Catholic University
of Leuven, Belgium
Peeters, B.; and De Roeck, G. 1999. Reference-Based
Stochastic Subspace Identification for Output-Only
Modal Analysis. Mechanical Systems and Signal Process-
ing, 13(6), pp. 855–878
Ramos, L.F. 2007. Damage Identification on Masonry Struc-
tures Based on Vibrations Signatures, PhD Thesis, Univer-
sity of Minho, Portugal (www.civil.uminho.pt/masonry)
SVS 2006. ARTeMIS Extractor Pro User Manual, Release
3.5, Structural Vibration Solutions, Aalborg, Denmark
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Stress analysis of masonry structures: Arches, walls and vaults

A. Baratta, I. Corbi & O. Corbi

Department of Structural Engineering, University of Naples “Federico II”, Naples, Italy

ABSTRACT: In the paper, the authors present a synthetic overview of some results obtained by means of a
number of theoretical and experimental studies developed on some classical masonry typologies such as arches,
panels and vaults. The proper implementation of the analyzed structural problem, specialized to the specific case,
which derives from the extension of classical structural approaches to structures made of masonry material, is
shown to provide a reliable approach to the problem itself, also in the case of some reinforcement, as demonstrated
by experimental data which are in perfect agreement with numerical results.

1 INTRODUCTION In details, the solution of the NRT structural prob-

The basic assumption of no-tension masonry model
coincides with the hypothesis that the tensile resistance
is null. Under this hypothesis, no-tension stress fields
are selected by the body through the activation of an
additional strain field, the fractures (see Baratta, 1991,
Baratta et al. 1981, Baratta & Toscano 1982, Bazant
1996, Heyman 1966). The behavior in compression
can be modeled in a number of different ways (elas-
tic linear, elastic non-linear, elastic-plastic; isotropic,
anisotropic; etc.), without altering substantially nei-
ther the results nor the mathematical treatment of the
problem; some convenience exists for practical appli-
cations in assuming a isotropic linearly elastic model,
in order to keep limited the number of mechanical
parameters to be identified for masonry, since increas-
ing the number of data causes increasing uncertainty in
the results. Because of these reasons, and being clearly
understood that there is no difficulty in introducing
more sophisticated models, it is convenient to set up
the fundamental theory on the basis of the assump-
tion that the behavior in compression is indefinitely
linearly elastic.
Analysis of NRT (Not Resisting Tension) bod-
ies proves that the stress, strain and displacement
fields obey extreme principles of the basic energy
functionals.
Therefore the behaviour of NRT solids under ordi-
nary loading conditions can be investigated by means
of some extensions of basic energy approaches to NRT
bodies (Baratta 1984, Baratta & O. Corbi 2005a, 2007,
Baratta & al. 1981, Baratta & Toscano 1982).
lems can be referred to the two main variational
approaches:
– the minimum principle of the Potential Energy
functional;
– the minimum principle of the Complementary
Energy functional.
In the first case the displacements and the fractures
are assumed as independent variables; the solution dis-
placement and fracture strain fields are found as the
constrained minimum point of the Potential Energy
functional, under the constraint that the fracture field
is positively semi-definite at any point.
The approach based on the minimization of
the Complementary Energy functional assumes the
stresses as independent variable. The complementary
approach is widely adopted since the existence and
uniqueness of the NRT solution are always assured in
terms of stress, if some conditions on the compatibil-
ity of the loads are satisfied. The stress field can, then,
be found as the constrained minimum of the Comple-
mentary Energy functional, under the condition that
the stress field is in equilibrium with the applied loads
and is compressive everywhere in the body.
The solution of both problems can be numerically
pursued by means of Operational Research methods
(see i.e. Rao 1978) suitably operating a discretization
of the analyzed NRT continuum (Baratta & I. Corbi
2004, Baratta & O. Corbi 2003b, 2005a). One should
notice that discussion about existence of the solution
actually can be led back to some Limit Analysis of the
considered NRT continua (Baratta & O. Corbi 2005a).

321
 2,3 m
0,124 m
0,200 m
3 2
4
1,322 m
0,382 m
0,775 m 0,750 m
Transducer Strain-gauge
(a)
Figure 1. (a) Masonry panel geometry, (b) with the applications of a light reinforcement by means of FRP strips.
In this regard, a special formulation of Limit Anal-
ysis for No-tension structures has been performed,
allowing the set up of theorems analogous to the basic
kinetic and static theorems of classical Limit Analysis;
thus, one can establish efficient procedures to assess
structural safety versus the collapse limit state (see e.g.
Como & Grimaldi 1983) by specializing and applying
fundamental theorems of Limit Analysis to NRT con-
tinua (Baratta 1991, Baratta & O. Corbi 2003a, 2005a,
Bazant 1996, Como & Grimaldi 1983, Khludnev &
Kovtinenko 2000). In details, the individuation of the
collapse (live) load multiplier for NRT continua can
be referred to the approaches relying on the two main
limit analysis theorems:
– the static theorem;
– the kinetic theorem.
This means that, after defining the classes of
statically admissible and kinetically sufficient load
patterns, Limit Analysis allows individuating the value
of the live load multiplier limiting the loading capac-
ity of the body, i.e. evaluating the collapse live load
and/or the safety factor versus collapse. One should
note that in a NRT structure its own weight (the dead
load) is an essential factor of stability, while collapse
can be produced by not-admissible additions of the
variable component of the load pattern.
Duality tools may also be successfully applied
in order to check the relationships between the two
theorems of Limit Analysis (Baratta & O. Corbi
2004). In the study of plane mono-dimensional struc-
tures featuring a low degree of redundancy, the
force/stress approach appears the most convenient
322
1
2,23 m
architrave
FRP
(b)
to be adopted if compared with the displace-
ment/strain approach, whose number of governing
variables is higher and, moreover, increasing with
the order of discretization (Baratta & I. Corbi 2003,
2004, 2006).
The following section reports some results show-
ing how the proper implementation of the described
theoretical approach for classical masonry structural
typologies such as panels, arches and vaults, produces
results that are in a very good agreement with exper-
imental data, demonstrating the overall reliability of
the mentioned approach, for whose details one should
refer to cited references.
As shown in the following, results can also be suc-
cessfully extended to the case of reinforcements with
fiber-reinforced polymers (FRP).
2 TESTS ON PROTOTYPES OF MASONRY
PANELS
2.1 Experimental investigation
This section reports some of the results of the wide
experimental campaign developed at the Laboratory
of Materials and Structural Testing of the University
of Naples “Federico II” on masonry panels, which are
symmetrical, with a central hole covered by a steel
architrave, and having upper part characterized by a
concrete fascia lightly reinforced by steel (Baratta &
I. Corbi 2003, 2004, 2006).
The geometry of the panels is shown in Figure 1.a.
A laboratory prototype of masonry panel is referred to,
made of tuff bricks (type “yellow tuff of Naples”, Italy)
jointed to each other by a pozzolana mortar in order to
confer a light additional resistance to the masonry;
the masonry itself is characterized by unit weight
γ = 10300 Nm and Young modulus E = 5.5 GPa. As
regards to the loading condition, a varying force is
applied in the middle of the left side of the panel,
in such a way to mitigate the proneness of the panel
to sliding of bricks with respect to each other, and
some loading/unloading cycles are developed up to
the collapse condition.
Once reached the crisis, the panel is reinforced
by directly laminating on the masonry some FRP
strips according to the provision scheme shown in Fig-
ure 1.b, at the same time with the impregnation of
the fibers by means of a special bi-component epoxy
resin, and a further experimental investigation is devel-
oped on the reinforced structure by re-executing some
loading/unloading cycles.
The adopted reinforcement, produced by FTS,
is a BETONTEX system GV330 U-HT, made of
12 K carbon fiber, jointed by an ultra light net
of thermo-welded glass. The mechanical character-
istics of the employed carbon fibers are: tensile
limit stress σfrp = 4.89 GPa, elastic modulus in trac-
tion Efrp = 244 GPa, limit elongation σfrp = 2%. The
FRP strip is characterized by thickness of 0.177 mm
and depth of 200 mm. The induced displacements
at some selected points [the transducers 1, 2, 3
and 4 in Figure 1.a] of the panel both for the not
reinforced and for the lightly reinforced panel are
recorded by a monitoring equipment consisting of: 4
transducers, placed at different locations of the panel
in order to record the absolute displacements, and 15
strain-gauges, arranged in 3 blocks of 5 strain-gauges,
in such a way that each block is devoted to record
the related strain situation. In details two transduc-
ers are located horizontally at two different heights
on the panel right side (transducers 1 and 2), and two
are placed in correspondence of the opening, one in
horizontal position at the top of the left side of the
hole (transducer 3) and the other one under the archi-
trave, which is devoted exclusively to control the panel
deflection (transducer 4). The displacements s(mm)
versus the varying force F(N) monitored by the trans-
ducers during the experiment in the not-reinforced
and in the reinforced case with some horizontally
applied C-FRP strips are shown in Figures 2.a–c
and 2.d–f respectively, as regards to the first load-
ing cycle. By the diagrams in Figure 2, which report
the displacements s(mm) vs the varying force F(N)
read by the transducers 1–3, some considerations can
been made.
With reference to the panel’s reinforcement by
means of the application of some C-FRP strips, the
major effect of the C-FRP intervention is the reduction
of the stress in the masonry. In general lower displace-
ments at the locations monitored by the transducers can
323
be recorded in the consolidated case with comparison
to the unconsolidated case.
To this regard, the pretty light type of reinforcement
allows to read the influence of even a small provision
on the panel response, which, on the counterpart,
cannot be expected to be macroscopic.
One should emphasize that the first objective of this
application is, then, to show the sensitivity of the NRT
model even to small changes in the structural response,
very differently from the elastic model, which, on the
contrary, for the specific case, is unable to detect any
difference in the behaviour of the wall. A number of
more effective reinforcements have also been tested
by the authors obviously resulting in more appreciable
results and a much higher performance (Baratta & I.
Corbi 2006).
In the specific case, one can notice that, with refer-
ence to the same load intensity [e.g. in correspondence
of the load value 3000 N in Figures 2.a–c], lower dis-
placements can be recorded in case of FRP insertions.
Moreover, the increase of the overall stiffness of the
panel results in a higher loading capacity with respect
to the not-reinforced wall. In particular the trend of
each curve, shows that it is closer to the x-axis (repre-
senting the load variable), thus indicating an increase
in the stiffness which is also related to an higher
collapse value of the load.
2.2 Experimental/theoretical comparison
Actually the application of the general theory of NRT
structures to the considered case of the masonry panel,
also in the presence of FRP reinforcements, can pro-
duce numerical results which are in good agreement
with the results obtained by the above reported exper-
imental campaign (Baratta & I. Corbi 2006). The
specialization of the general problem to the case of
masonry walls requires the definition of a discrete
model coupled to the real structural model, the set up
of the energetic problem (in the case of masonry pan-
els the potential energy approach is to be preferred)
for the discrete problem, which, for masonry material,
results in a Non Linear programming problem to be
solved by means of Operational Research tools, and,
finally, the search of the numerical solution of the set
up OR problem by means of a suitably implemented
calculus code (Baratta & I. Corbi 2004, 2006).
Once followed the above described steps, the
numerical results can be compared to the ones coming
out from the experimental investigation, for the final
validation of the theoretical set up.
For the specific case one may compare the results
relevant to the first loading cycle with those related
to experiments. As shown in Figures 2, the theoretical
data (continuous lines) are in good agreement wit the
experimental ones (dotted lines) both as regards to the
 Transducer 1 Transducer 2 Transducer 3
(a) (b) (c)
Experimental data Experimental data Experimental data
Theoretical results 1 Theoretical results 1 Theoretical results
1

0,8 0,8
0,8

0,6 0,6

u3 (mm)
0,6

u2 (mm)
u1 (mm)

0,4 0,4
0,4

0,2 0,2
0,2

0 0
0 0 100 200 300 400 500 600
0 100 200 300 400 500 0 100 200 300 400 500 600
F (Kg) F (Kg) F (Kg)

Transducer 1 Transducer 2 Transducer 3

(d) (e) (f)
1
Experimental data 1 Experimental data 1 Experimental data
Theoretical results Theoretical results Theoretical results

0,8
0,8 0,8

0,6
0,6 0,6
u1 (mm)

u2 (mm)

u3 (mm)

0,4
0,4 0,4

0,2 0,2 0,2

0 0 0
0 100 200 300 400 500 0 100 200 300 400 500 0 100 200 300 400 500
F (Kg) F (Kg) F (Kg)

Figure 2. Comparison between the numerical (continuous line) and the experimental (dotted line) at the monitored positions
1, 2, 3 for the not reinforced panel (a, b, c) and for the reinforced panel (d, e, f).

not reinforced case (Figs 2.a–c) and to the consolidated In this case, the sliding between bricks are reduced
case (Figs 2.d–f). and do not influence the overall characteristics of
In the first unconsolidated case the masonry deformability and stiffness of the masonry panel.
exhibits a behaviour which appears lightly stiffer than
in the theoretical model: this effect is maybe due to
3 TESTS ON PROTOTYPES OF MASONRY
the micro-fractures present at the first stage of the
ARCHES
computational procedure, which are probably absent
in the real behaviour of the masonry. The transducers
3.1 Experimental investigation
2 and 3 show an overall pretty good agreement between
numerical and experimental data even if also other This section reports some of the results of the wide
phenomena as sliding between bricks, micro-fractures, experimental campaign developed at the Laboratory
etc., should be taken into account, which cause the not of Materials and Structural Testing of the University of
perfect agreement of the diagrams relevant to the first Naples “Federico II” on masonry arches, consolidated
transducer 1 (Figs 2.a–c). or not by means of FRP strips (Baratta & O. Corbi
It is indeed because of these reasons that the numer- 2003b, 2005b).
ical/experimental agreement is higher, almost perfect, The geometry of the portal arch (Figs 3 and 4)
in the reinforced case (Figs 2.d–f). is symmetrical and is characterized by span

324
 d 10 d9 d8
d 11 d7
d 12
d 13 d6
d 14 d5
d 21 d 22 d 23 d d4
24 d 25
d 15 d 20 d 26
I1 d 19 d3
d 27
d 16 d 18 d 28 d2 F
d 17 d 29
G1

E1
d 30 d 1
T1 T2
Dial Gauge G1
Inclinometer I1
Transducers T1,T 2
Extesemeter E1
Deformometric cells dk, k=1…30

Figure 3. The portal arch model with the monitoring equip-

ment: sketch of the monitoring equipment.

L = 1900 mm, rise f = 660 mm, arch thickness d =

240 mm, piles thickness b = 385 mm, piles height
h = 1700 mm; the arch shape is a semi-ellipse. The
arch depth is 400 mm, whilst the two abutments
Figure 4. The portal arch model with the monitoring equip-
are 480 mm deep. The masonry is characterized by
ment: picture from laboratory tests.
unit weight γ = 12300 N · m−3 and Young modulus
E = 5.5 GPa.
As mentioned, in the above, in the second stage of
the experimental campaign one also considers some
is intended to represent a pseudo-seismic action, able
FRP continuous reinforcement applied on the arch
to yield a measure of the structure attitude to sustain
length. In this case, the FRP reinforcements consist
earthquake shaking.
of continuous mono-directional FRP strips applied on
The monitoring stuff (Figs 3 and 4) consists of:
the extrados of the arcade.
The adopted reinforcement, produced by FTS, is – 1 dial gauge G1 , placed on the left side of the
a BETONTEX system GV330 U-HT, made of 12 K left abutment, finalized to the monitoring of the
carbon fibre, jointed by an ultra light net of thermo- absolute displacement of the pile;
welded glass. – 2 transducers T1 and T2 , vertically placed on the
The mechanical characteristics of the employed front side of the left abutment, finalized to the mon-
fibres are: tensile limit stress σfrp = 4.89 GPa, elastic itoring of the length variation of both edges of the
modulus in traction Efrp = 244 GPa, limit elongation pile;
εfrp = 2%. The FRP strip is characterized by thickness – 1 inclinometer I1 , placed on the top of the left abut-
of 0.177 mm and depth of 100 mm. ment, finalized to the monitoring of the pile average
After roughly preparing the masonry support in rotation;
order to render the application surface smoother, the – 1 extensometer E1 , placed between the two abut-
FRP is directly laminated on the masonry, at the same ments, finalized to the monitoring of the relative
time with the impregnation of the fibres by means of piles’ displacement;
a special bi-component epoxy resin. – 30 deformometer cells, placed on the front of the
As regards the execution the tests, the structure is arch, finalized to the monitoring of the arcade
subject to its constant own weight and to a lumped deformation.
horizontal force F, applied on the top right side of
For the un-reinforced structure (Baratta and Corbi,
the right abutment in the rightward direction in the
2003a,b) the critical condition is related to the activa-
increasing phase (Figs 3 and 4), which is transmitted
tion of a collapse mechanism composed by four hinges
by means of a loading equipment consisting of a load
distributed as follows:
cell placed on the right side of the portal arch.
This force is able to potentially produce collapse of – 1 at the keystone on the extrados,
the structure according to a mechanism that is typical – 2 at the reins on the intrados,
of earthquake failures of arch-portals (Fig. 6), and it – 1 at the bottom of the right pile on the extrados.

325
10 ∆ u (mm) – 1 under the load cell on the intrados of the right pile
(where shear occurs),
Experimental Data – 1 at the bottom of the right pile on the extrados.
8 Trend Line The collapse is reached at F∼800 N with an increase
in the loading capacity of the portal arch of approx-
imately 10 times with respect to the unconsolidated
6 case. The experimental force-displacement diagram is
Numerical reported in Figure 7.
Data
4
3.2 Experimental/theoretical comparison
Actually the application of the general theory of NRT
2
structures to the considered case of the masonry portal
arch, also in the presence of FRP reinforcements, can
produce numerical results which are in good agree-
0 F (N) ment with the results obtained by the above reported
0 20 40 60 80 100 experimental campaign (Baratta & O. Corbi 2005a,
2005b, 2007).
Figure 5. Unreinforced portal arch: pile displacement u The specialization of the general problem to the case
versus load F-numerical/experimental comparison.
of masonry arches requires the definition of a discrete
model coupled to the real structural model, the set up
of the energetic problem (in the case of masonry arches
the complementary energy approach is to be preferred)
The collapse condition is reached at F∼80 N; the
for the discrete problem, which, for masonry mate-
low failure value of the force shows that, due to the
rial, results in a Non Linear programming problem
chosen elliptical shape of the arch, the funicular line
(which in the specific case can be reduced to a Linear
compatible with the applied loads and admissible (i.e.
Programming problem) to be solved by means of Oper-
interior to the arch profile) is already very close to the
ational Research tools, and, finally, the search of the
upper and lower bounds of the arch profile at the rest
numerical solution of the set up OR problem by means
condition.
of a suitably implemented calculus code (Baratta &
The experimental force-displacement diagram is
O. Corbi 2003a, 2003b, 2005a).
reported in Figure 5.
Once followed the above described steps, the
After reaching the collapse condition, the portal
numerical results can be compared to the ones coming
arch is then unloaded in order to be prepared for the
out from the experimental investigation, for the final
subsequent experimental tests on FRP reinforcements.
validation of the theoretical set up.
After completing the unloading process, the portal arch
Numerical investigation on the portal arch model
is prepared for laboratory tests on FRP reinforcements,
experimentally tested results in the possibility of
which are finalized to the evaluation of the benefits
appreciating the skill of the NRT model to capture
induced on the model response by the application of
the major features of the structure behaviour. More-
carbon fibre strips.
over also the correct modelling of the reinforcement
The reinforcement consists of a continuous FRP
and of its coupling with the main structure can be
strip bonded on the extrados of the arch. Since the
evaluated. Figure 5 reports the numerical/experimental
collapse mechanism of the not reinforced simple por-
comparison relevant to the right pile top displacement
tal arch is characterized, as described in the above, by
u (mm) versus the varying load F (N) for the considered
the formation of two intrados hinges at the reins of the
un-reinforced arch.
arch, corresponding to the fractures d4 –d5 and d12 –d13
A very good agreement between the numerical and
at the extrados, the major effect of this intervention
experimental data can be observed. The calculus code
is supposed to be the prevention of these fractures,
is demonstrated to be able to capture the behaviour of
and, therefore, a wide increase in the model loading
the portal arch following the whole loading path up to
capacity.
collapse; Figure 6 depicts the collapse mechanism of
The funicular line is now free to exceed the lower
the structure as it appears directly from the calculus
contour of the portal arch cross section.
code, clearly due to the formation of four hinges: one
In this case the critical condition is related to the
at the keystone on the extrados, two at the reins on
activation of a collapse mechanism composed by four
the intrados, one at the bottom of the right pile on the
hinges, distributed as follows:
extrados.
– 1 at the top of the left pile on the intrados, Moreover one reports in Figure 7 the numeri-
– 1 at the keystone on the extrados, cal/experimental comparison relevant to the right pile

326
Figure 6. Unreinforced portal arch: picture of the collapse Figure 8. Portal arch with extrados reinforcement: picture
mechanism captured from the calculus code. of the collapse mechanism captured from the calculus code.

10 u (mm)

8 Experimental Data

Trend Line
6

Numerical
4 Data

Figure 9. Barrel vault with horizontal directrix.

2
4 PROTOTYPES OF MASONRY VAULTS
0 F (N)
4.1 The problem of barrel vaults with indefinite
0 200 400 600 800 1000 length
Figure 7. Portal arch with extrados reinforcement: pile dis- As regards to barrel vaults (Baratta & O. Corbi 2007),
placement u vs. load F-numerical/experimental comparison. first of all, one should consider that, since the vault
geometrically derives by the translation along a direc-
trix of a generating arch curve, in this case, the
meridian lines coincide with the generatrix in their
top displacement u (mm) versus the varying load F shapes; if one considers a rectilinear directrix, the vault
(N) for the arch reinforced with an extrados FRP parallels are horizontal and rectilinear as well (Fig. 9).
reinforcement. The surface of the shell of the mid-surface of the vault
Again a very good agreement can be observed may be defined by the equation z = f(x). Because of
between the numerical and experimental data. The cal- the vault geometry, one has
culus code is demonstrated to be able to capture the
behaviour of the portal arch ; Figure 8 depicts the col-
lapse mechanism of the structure as it appears directly
from the calculus code, clearly due to the formation of
four hinges: one at the top of the left pile on the intra-
dos, one at the keystone on the extrados, one under the
load cell on the intrados of the right pile, one at the bot-
tom of the right pile on the extrados. Both numerical
and experimental data agree in assessing at approxi-
mately ten times the original value the increment of the
loading capacity of the structure due to the extrados where dsx and dsy denote the length of the sides of
FRP reinforcement. the generic vault element ABCD of area dA dx and

327
dy the length of the corresponding sides on the ele-
ment A B C D projected in the xy-plane, and ϕ and θ
denote the angles formed by the meridian sides AB and
DC of the element with the x-axis and by the parallel
sides AD and BC with the y-axis, respectively. As con-
cerns equilibrium, hypothesizing that the vault is in a
membrane state of stress, a correspondence can be set
between forces acting on the element ABCD (stresses
Nx , Ny , Nxy = Nyx and applied load for unit area, px ,
py , pz ) and projected forces acting on the associated
element A B C D (Nx , Ny , Nxy = Nyx and px , py , pz )
in the xy-plane (Baratta & O. Corbi 2007).
In absence of horizontal loads and if the vertical
load is not dependent on “y”, as it happens when the
vault is subject to only vertical loads due to the self-
weight (i.e. pz = pz (x) ≥ 0), and assuming that the vault
has an indefinite length in the direction y, equilibrium
may be expressed in the form
which reduces the problem to the determination of
stress function ψ(y).
Assuming that the directrix curve of the vault is
a circular arch (Fig. 10) of radius R, with constant
thickness “s” and unit weight γ, and imposing suit-
able constraint conditions, one yields the final solution
(Baratta & O. Corbi 2007)
with
Figure 10. Cross section of a barrel vault with circular arch
generatrix.
328
where z0 and z1 are arbitrary ordinates, conditioned by
the fact that z(t) should be contained in the interior of
the profile of the vault.
After this result, it is possible to calculate the
internal forces Nx ≤ 0, Ny = Nxy = 0 and Nx ≤ 0,
Ny = Nxy = 0
It is also possible to realize that the equilibrium
solution allows the structure to behave as a sequence
of identical independent arches. From this result, one
may refer to the results reported in the previous section
for the portal arch model, reinforced or not with some
FRP strips, whose analytical problem implementation
has been shown to give theoretical results in perfect
agreement with the produced experimental data, also
exhibiting very effective results in the reinforced case.
5 CONCLUSIONS
The paper reports some results proving the successful
application of a correct theoretical treatment, based
on the NRT material assumption, of structural prob-
lems relevant to classical masonry typologies such as
arches, walls and vaults.
The set up of the general energetic approach for
analyzing masonry structures under live loads, its
specialization to the relevant discrete models, the
implementation of ad hoc built up calculus codes are
demonstrated to produce numerical results in very
good agreement with data produced by experimental
investigation.
One should emphasize that, differently from many
models which require a number of parameters allow-
ing a certain adaptation of the shape of the numerical
curve to the experimental one, the NRT model has the
big advantage that the only mechanical parameter to
be evaluated is the masonry elastic modulus. Since the
tuning of the theoretical model is pretty simple, there
would be no possibility to force it to produce theo-
retical results fitting with such a good agreement the
experimental data, because the tuning operation itself
cannot influence the shape of the numerical diagram
but only the displacements scale.
As a point of fact, the sensitivity of the mod-
elling to material assumptions reduces to the inverse
proportionality between the material elastic modulus
and displacements, without any influence on the load
capacity and on the evolution of displacements with
the loads.
Actually the extension to the case of some rein-
forcement directly applied on the masonry can also be
studied by properly modeling the reinforcement itself
and its connection with the masonry.
 The theoretical/numerical agreement, which is an
original result also for the case of masonry con-
structions with FRP reinforcements, demonstrates that
the overall approach is reliable for the treatment
of masonry constructions also in the presence of
consolidation interventions.
REFERENCES
Baratta, A. 1984. Il materiale Non Reagente a Trazione come
modello per il calcolo delle tensioni nelle pareti murarie.
J. Restauro, 75: 53–77.
Baratta, A. 1991. Statics and reliability of masonry structures.
In General Principles Applications in Mechanics of Solids
and Structures. CISM, Udine: 205–235.
Baratta, A. & Corbi, I. 2003. Investigation of FRP consol-
idated masonry panels. Proc. 9th Int. Conf. Civil and
Structural Engineering Computing. Egmond Aan Zee,
Netherlands.
Baratta, A. & Corbi, I. 2004. Iterative procedure in No-
Tension 2D problems: theoretical solution and exper-
imental applications”, In G.C.Sih & L.Nobile (eds),
Restoration, Recycling and Rejuvenation Technology for
Engineering and Architecture Application, Aracne Ed:
67–75, Bologna.
Baratta,A. & Corbi, I. 2006. On the reinforcement of masonry
walls by means of FRP provisions. In A. Mirmiran &
A. Nanni (eds) Composites in Civil Engineering., Miami,
2006.
Baratta, A. & Corbi, O. 2003a. Limit Analysis of No Tension
bodies and non-linear programming. Proc. 9th Int. Conf.
329
Civil and Structural Engineering Computing. Egmond
Aan Zee, Netherlands.
Baratta, A. & Corbi, O. 2003b. The No Tension model for the
analysis of masonry-like structures strengthened by Fiber
Reinforced Polymers. J. Masonry Intnl. 16 (3): 89–98.
Baratta, A. & Corbi, O. 2004. Applicability of Duality Theory
to L.A. Problems with Undefined Flow Law. Proc. 1st
Int. Conf. RRTEA Restoration, Recycling and Rejuvation
Technology for Engineering, Cesena.
Baratta, A. & Corbi, O. 2005a. On variational approaches in
NRT continua. J. Solids and Structures. 42: 5307–5321.
Baratta, A. & Corbi, O. 2005b. Fibre reinforced composites
in civil engineering. Experimental validation of C-Fibre
masonry retrofit. J. Masonry Intnl. 18(3): 115–124.
Baratta, A. & Corbi, O. 2007. Towards a new theory for
masonry vaults assessment. Proc. 7Th Int. Masonry Conf.
London.
Baratta, A., Vigo M. & Voiello G. 1981. Calcolo di archi in
materiale non resistente a trazione mediante il principio
del minimo lavoro complementare. Proc. 1st Nat. Conf.
ASS.I.R.C.CO. Verona.
Bazant, Z.P. 1996. Analysis of work-of fracture method for
measuring fracture energy of concrete. J. Engineering
Mechanics ASCE 122 (2): 138–144.
Como M. & Grimaldi A. 1983. A unilateral model for Limit
Analysis of masonry walls. In Unilateral Problems in
Structural Analysis, Ravello: 25–46.
Heyman, J. 1966. The stone skeleton. J. Solids and Structures,
2: 269–279.
Khludnev, A.M. & Kovtunenko, V.A.. 2000. Analysis of
cracks in solids. Applied Mechanics Review, 53 (10).
Rao, S.S., Optimization: Theory and applications, Wiley &
Sons, New Delhi, 1978.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Evaluation of historic Dieh-Dou buildings in Taiwan

Pin-Hui Tsai & Dina D’Ayala

Department of Architecture and Civil Engineering, University of Bath, UK

ABSTRACT: The typical failure modes of the Taiwanese historic timber frame temples (Dieh-Dou buildings)
are observed and classified from existing reports and surveys undertaken after the 1999 Chi-Chi earthquake.
The results of a series of laboratory full-scale joint tests are used to set two types of failure criteria, joint failure
and material failure. Static elastic and nonlinear step by step analyses are then performed after simulating the
Dieh-Dou structure with a FE software to assess their behaviour under lateral load. The results show that although
the static analysis can reasonably simulate the overall behaviour of the structure, is only with the step-by-step
analysis, where the correct failure sequence is identified by progressive removal of the elements damaged or
failed according to the failure criteria, that the response of the structure is determined in the correct way. The
critical elements of the frame are then identified, providing important indications for repairing or strengthening
of these structures for the future.

1 INTRODUCTION two identical timber frames connected to perimeter

The Taiwanese traditional temple compounds are clas-
sified as Chinese style architecture, which started
to develop in the 17th century when Chinese immi-
grants moved from south Mainland China to Taiwan
(Fujishima et al 1999). These compounds include
several buildings and each building consists of
masonry walls laid in the direction normal to them
and supporting a system of purlins. These timber
structures are commonly called Dieh-Dou. Figure 1
shows a schematic elevation of a Dieh-Dou frame. The
main characteristic feature of Dieh-Dou buildings is
the stack of small highly decorated timber elements
forming the Dou-Gon which sits on the main beam.

Figure 1. Schematic illustration of a Dieh-Dou timber frame modified from FHCR (2005).

331
 Table 1. Level of damage of Dieh-Dou buildings after Chi-
Chi earthquake.

Damage
level Failure mode %

Level 1 No structural damage 24

Level 2 Elements pulling out (10%) 24
Damage of the roof decorated ridge (14%)
Level 3 Eaves rupture (7%) 42
Masonry wall cracks (21%)
Columns sliding (14%)
Level 4 Dieh-Dou frames with in plane or out of 7
plane leaning
Level 5 Dieh-Dou frames with in plane or out 3
of plane collapse

Figure 2. Dou-Gon joint.

This arrangement allows creating large spans between

main columns and transfer the roof weight to the main
beams and then to the ground. The components of the
Dou-Gon are shaped in double notch joints and con-
nected to the out of plane members by way of shallow
dove tails (see Figure 2). The in plane bracket con-
nects one Dou-Gon to the next in the plane of the
frame, creating the slope of the roof, while the con-
cave top accommodates the round purlin, which also
seats on the out of plane (y direction in Figure 2) carved
bracket.
Dieh-Dou buildings differ from other Chinese and
Asian temple architecture and illustrate the uniqueness
of the Taiwanese culture (Fu 2005). This notwithstand-
ing, the characteristics of Dieh-Dou structures have not
been well researched to date. In addition, Taiwan is Figure 3. Element pulling out from joints (Huang. & Shu,
located in a high seismic area and to protect these his- 2001).
toric structures it is essential to thoroughly study and
understand this structural behaviour in order to repair
2 DIEH-DOU STRUCTURE
and preserve them without reducing their significance
and heritage value.
2.1 Seismic vulnerability
In this paper, first of all, the failure modes observed
after the 1999 Chi-Chi earthquake are discussed. Then Taiwan is an island located south of mainland China
a finite element model is created to simulate a Dieh- and between the Eurasian and Philippine Sea plates,
Dou frame by performing elastic and step by step an highly seismic area. Investigation carried out within
analyses. In order to examine and compare these two the present project identified at least 52 historic archi-
different types of analysis with the actual behaviour, tectural compounds built as Dieh-Dou structures over
failure criteria are set based on a series of labo- the Chi-Chi earthquake, 7.3 Richter Scale magnitude,
ratory tests on full-scale joints. The results of the stricken region. While for 29 of them was not possi-
FE analysis, when compared with the actual failures ble to gather data about their condition after the event,
after earthquake, show that the model can reproduce for the remaining 23 data were collected on level and
the behaviour of Dieh-Dou frames well. In particu- mode of damage as shown in table 1. Level 1 means
lar, the nonlinear step by step method can explain in no structural damage. Buildings with elements pulling
more detail which are the critical elements and give out from joints (Figure 3) and damage of the roof
suggestion for further invedigations and repair. decorated ridge are categorised as damage level 2.

332
As the bearing capacity of the structure as a whole
is not impaired, damage level 3 includes eaves rupture
(Figure 4), masonry wall cracks and columns sliding
from the original position: these are not considered an
immediate threat to collapse but reduce the structures
bearing capacity. Finally, Dieh-Dou frames with in
plane or out of plane leaning or collapse are classified
as damage level 4 or 5, respectively.

2.2 Failure mode of Dieh-Dou frame

In Dieh-Dou buildings the purlins linking two main
frames out of plane sit on the top of the Dou, connected
with a simple scarfed joint. Under lateral excitation, Figure 4. Eaves rupture after earthquake.
the purlins easily slide out of their seating once the ver-
tical loading is overcome. The field survey showed that
under earthquake, in several cases, once the purlins
failed the two main frames became virtually indepen-
dent and started to take load separately. Hence in the
following numerical analysis, conservatively, a single
Dieh-Dou frame is considered neglecting the effect of
the 3D assembly.
For a single frame, the damages from table 1 can
be classified into two types, joint failure and material
failure. Joint failures include element pull out and loss
of rotational stiffness or columns sliding from the orig-
inal position. Material failures, such as timber element
rupture, are caused by excessive stresses in the timber
pieces.

3 FINITE ELEMENT MODEL EVALUATION

3.1 Description of FE model

Although precedent work on Dieh-Dou structures is
relatively limited, research on other historic timber
structures have proven that numerical analysis can
be successfully applied to obtain a fair description
of Far East Asian ancient timber structures seismic
behaviour. Also, the complicated geometry configu-
ration of Dieh-Dou structures is not well understood
and finite element analysis methods can enable to iden-
tify the whole structural behaviour. To approach the
final goal of this project, the finite element commercial
package Algor © was chosen to analyse the structure.
Three different building typologies were defined.
Results are discussed here for typology C, charac-
terised and represented by the main hall of Guan-Shi
family Temple, these buildings started in 1849 (Hsu,
2002). The difference between this frame and other
main hall of temple compounds is the supporting ele-
ment between column and main beam (see part A in
Figure 1). The lack of part A may render the main hall
of Guan-Shi weaker than other frames.
The frame was created by beam elements in the FE
environment. Ever timber piece of the Dou-Gon joint
Figure 5. Model of one layer of joint.
set (see Figure 2) was considered. Figure 5 shows one
layer of the joint in perspective view and the corre-
sponding FE modelling (in plane). The centre of joint
is represented as Node 1 and all the parts of the joint set
were modelled as beam element located at the neutral
axis of each position. Part 1 is the out of plane bracket
which is continuous through the joint; two rigid ele-
ments, B1 and B2, were created to connect it with part
2 and 3, which are the ‘Dou’ and are simulated by
parts B3 and B4. Part 4 and 5 were defined as beam
elements B5 and B6, respectively, and linked with the
joint centre Node 1 with the appropriate stiffness by
using boundary elastic elements E1 and E2. Here, the
rotational stiffness was taken from test results (Tsai &
D’Ayala & Lewis 2006). B5 and B6 represent the ends

333
of the in plane beams connecting to the out of plane
brackets by means of dove tail joints.
Apart from the complicated Dou-Gon joint set,
there are other more common beam to column joints
in the Dieh-Dou buildings. The stiffness of these joints
of these mortise tenon type was assumed in relation to
the connection geometry as proposed by Chang (2006)
in which equation (1) is for beams continuous through
column and equation (2) is for flat cut beam face to
face inside the column.
Cw : column width
Bw : beam width
Bd : beam depth
E⊥ : modulus of elasticity of timber perpendicular to
grain.
Two type of analysis were performed with this FE
model, elastic and nonlinear step by step. In the elas-
tic analysis a total lateral load equal to 0.33 mg, FL , is
applied in correspondence of each purlin. The advan-
tages of linear elastic analysis is the possibility of
getting a rapid result and make a quick judgement of
the capacity of the structure under a load magnitude of
the order of the maximum which can be applied to the
structure by the strongest earthquakes. However, ele-
ments are likely to fail during the earthquake, induced
loads causing loss of structural stiffness and leading
the frame to a diverse distribution of load. A nonlinear
step by step analysis therefore was performed, where
the model was modified at each load step according to
the criteria explained below. Figure 6 shows the finite
element model of the frame.
3.2 Failure criteria
The two categories of failure leading to the failure
modes previously described, namely joint failure and
material failure, are detailed below.
Three types of joint failure modes can be further
grouped as pull out and bending failure (applying to
horizontal members) and column failure.
A series of pull out tests were first performed in
laboratory with full scale samples of one layer of
Dou-Gon joint. The results indicated that the maxi-
mum pull out forces at failure were similar and did
not depend on the magnitude of vertical load applied
(with a value varying between 8000 N and 9000 N)
if the dove tail connection was good, but reducing to
3000 N and 1500 N when applied loads were 6500 N
and 3250 N respectively, in the case of poor fit of the
334
Figure 6. FE model.
dove tail joint (then relying on the timber to timber
friction, as expected).
Assuming proper dove tail connections which are
normally guaranteed by accurate craftsmanship, a
value of 8000 N was set to define the failure due to
element pull out in the FE model. Once the forces are
over this value in the model, the element capacity of
transmitting axial force is considered lost and the cor-
responding degree of freedom is released on the beam
element in the FE model.
Differently from pull out tests, the rotational tests
showed that the results, in terms of stiffness, depend on
the magnitude of the vertical load. However, although
the rotational stiffness were different, the maximum
moment resisted by the joint was found to be always
around 1000 Nm (Tsai & D’Ayala & Lewis 2006),
after which the capacity dropped. For the beam-
column joints an elastic limit value of 1500 Nm was
taken from Chang (2006). If the bending moment in
the model was found to exceed that of the criteria
mentioned above at some point in the analysis, then
the joint will be considered failed in bending, which
implies its capacity to transmit moment to the neigh-
bour elements is lost. In this case, the beam element is
released in rotation. Finally, the foot of the column has
a shallow pin into a stone base and relies on friction to
resist lateral action.
The coefficient of static friction were calculated
from laboratory tests of the timber pieces, obtaining
values of 0.53 and 0.58, with an average of 0.55, in line
with codes indications (which range from 0.5 to 0.6 for
friction of timber to timber in dry specimens). Divid-
ing the resultant base shear applied at the bottom of
the column by the vertical load applied, if the value is
over 0.55, then the columns are assumed to be sliding
from their original position and so no more horizontal
force could be transmitted to the ground anymore. Due
to the particular connection with the ground described
above, it is further assumed that the columns are able
to rotate and do not transmit moment.
For material failure, ombined bending and tension,
combined bending and compression, and shear stress,
are examined. The material capacities are taken from
the material tests (Tsai & D’Ayala & Lewis 2006) and
from the Wood Handbook (1999). The failure criterion
for bending stress was set at 4.6 E + 7 (N/m2 ), for shear
at 5.5 E + 6 (N/m2 ) and for tension and compression
at 4.5 E + 7 (N/m2 ) and 2 E + 7 (N/m2 ) respectively.
For the elements subject to axial compression and
bending, the interaction formula of British Standard
(BS 5268-2; 2002), shown as equation (3), is used as
failure criterion:
and the failure criterion (Miyamoto al, 2004), and is
matched by the observation of the real situation after
the earthquake (Tsai & D’Ayala & Lewis 2006). For
the step by step analysis, a total of 30 steps were per-
formed until the collapse criterion was reached. The
lateral load at the thirtieth step was 80% of FL. In both
cases (elastic and step by step analysis), the columns
did not reach the failure criterion. At the seventeenth
step (50% of FL ), the roof apex has already met the
damage criterion. The structure drifts results are shown
in Figure 8.
When compared with the results of the elastic anal-
ysis where the full lateral load was applied, the step by
step analysis shows a considerable decrease of over-
all stiffness. This is because during these 30 steps
several joints lost their rotational stiffness and sev-
eral beams pulled out and lost bending transmitting
capacity, rendering the structure more deformable.
The result of elastic analysis compared with the
where σm,a,  is the applied bending stress; σm,adm,  is evidence after the earthquake (Tsai & D’Ayala &
the permissible bending stress; σc,a,  is the applied Lewis 2006), confirmed that finite element model
compression stress; σc,adm,  is the permissible com-
pression stress (including buckling); σe is the Euler
critical stress.
Equation (3) is for permissible stress design. How-
ever, because the present defined analysis of Dieh-Dou
building focuses on failure, the values above are
inserted instead of permissible values. The value K12
can be found in Table 22 of the British Standard (BS
5268-2; 2002).
If the resulting stresses at one step are over the
values set above, then the corresponding element is
regarded as failure and completely deactivated in the
corresponding part of the FE model.
3.3 Results
In the elastic analysis the maximum load FL is applied
all at once, while the step by step analysis consists
of a series of models in which 10% of the FL is ini-
tially applied and further increases of 10% of FL are
added if the results of the previous step shows that no
damage occurred. Once elements are found to reach
the failure criteria they are treated accordingly and the
same load level reapplied until no further element fail-
ure is attained before increasing the load again. The
step by step analysis stops when the roof apex lateral
displacement are over the collapse criteria proposed
by Miyamoto et al 2004 for Asian timber structures.
The procedure of analysis implemented in the pro-
gram is shown in the flow chart of Figure 7. The
results of both analyses are discussed in the following
section.
3.3.1 Structure drift
In the static elastic analysis, the displacement of the
roof apex is 55.2 mm, which lies between the damage
335
can catch the structural behaviour under horizontal
forces.
However, the physical tests showed that the max-
imum pull out load in Dieh-Dou joint is 8000 to
9000 N, and in the linear analysis results of FE model,
there are several elements already over 8000 N, imply-
ing that several joints could have suffered element
pull out leading the frame to a diverse distribution
of load which may also cause additional decrease
of capacity. The step by step analysis proved to be
more conservative than the elastic analysis. Further-
more, in the FE model analysis, timber material were
assumed in good condition, which may not always be
the case.
3.3.2 Vulnerability
In order to find out the critical elements that may
govern the structural behaviour, the maximum hori-
zontal displacements at each load level are reported in
Figure 9 for the step by step analysis.
It can be noted that until the horizontal load is below
20% of FL , the structure is stiffer; after this level the
displacement rate increases and is maintained constant
until failure. This implies that the elements and joints
whose capacity is lost before 20% of FL are critical to
the stiffness of the Dieh-Dou frame.
In figure 10 the elements that failed up to 20% of FL
are marked. It is seen that they represent 8 main beams
and corridor beams ends, together with 2 binding ele-
ment ends, and they failed mostly for loss of rotational
stiffness. Three main beams span the main ceremonial
space of the Dieh-Dou temples (see Figure 1). The
results show that the second and third main beams,
failed at both ends, can be the first critical elements
under earthquake.
Figure 7. Flow chart of step by step analysis.
11
10
9
x displacement (cm)

8
7
6
5
4
3 step by step result
result of static analysis
2
damage criterion
1 failure criterion

0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Step

Figure 8. Structure drift of static and step by step analysis.

Figure 10. Element failed up to 20% of FL .
110%
100%
90% Figure 11 shows the elements failed when the lateral
80%
70%
displacement reached the level of the failure criterion,
% of load

60% at 80% of FL (Figure 8). The majority of the ele-

50%
40%
ments failed after the initial loss of stiffness, most of
30%
step by step analysis
whose suffered pull-out, are located outside the cen-
20%
10%
elastic analysis tral zone of the main span. This compares well with
0% the failure of Guan-Shi family temple main hall after
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 earthquake, where the elements of corridor parts were
X displacement of roof apex cm
pull out while the elements of the main span were still
in position and kept contributing to the lateral transfer
Figure 9. Maximum displacement of each load level. of load.

336
Figure 11. Element failed up to 80% of FL .
4 CONCLUSION
From the survey of Dieh-Dou building following Chi-
Chi earthquake, three type of causes leading to frame
failures can be identified, namely elements pulling out
from joints, eaves rupture and columns sliding from
the original position. Combining a series of labora-
tory tests, literature references and survey of existing
structures after the earthquake, a system of failure
criteria was established to study the occurrence and
effect of joints failure and material failure in Dieh-Dou
buildings.
A finite element model was created to simulate a
Dieh-Dou frame under lateral load. The result of a
simple elastic analysis showed that failures correlate
well to observation after the earthquake and confirm
the possibility of using finite element modelling to
simulate Dieh-Dou structures. However, the elastic
model for its nature could not allow identifying critical
elements and vulnerability. Furthermore, observation
showed that connection would fail for loads lower than
the applied maximum, causing the building to be less
stiff. A step by step analysis was then performed to
improve the evaluation.
The results of step by step showed that this method
is more accurate and conservative than the static anal-
ysis, giving the possibility to evaluate such historic
buildings properly. The maximum roof apex displace-
ment against each load indicates that the connecting
main beams and corridor beams are the critical ele-
ment in Dieh-Dou frame, and they may fail at relatively
low load levels. The rate of lateral displacement of
the structure then increases quicker when these critical
parts provide a reduced contribution. For conservation
337
of Dieh-Dou structures, it is necessary to repair or
strengthen these elements and joints first.
From the step by step FE analysis, therefore, a pre-
liminary result has been obtained which can help to
identify the critical and vulnerable elements of the
Dieh-Dou buildings. Different Dieh-Dou buildings,
which may have different frame geometries, should
be analysed to observe if the critical and vulnerable
elements are similar and the structural behaviour fol-
lows the same trend. A parametric study can finally
help to address the directions the conservation should
take. Once the vulnerable and critical element has
been defined, a strengthening strategy can be put into
place.
REFERENCES
British Standard, BS 5268-2; 2002, Structural use of timber –
Part 2: Code of practice for permissible stress design,
materials and workman ship
Chang, W.-S., 2006, On rotational performance of traditional
Chuan-Dou timber joints in Taiwan, PhD dissertation of
National Chen Kung University, Taiwan
Fijishima, G. (survey and write), Chan, H-L (edit) Taiwanese
architecture, Taiuan Publishing Co (1999), in Traditional
Chinese
Fu, C.-C., 2005, Symbolic, architecture and artistic deco-
rations in structural elements in traditional architecture
in Taiwan, Proceeding of Structural Studies, Repair and
Maintenance Architecture IX, pp. 35–42
Foundation of Historic city Conservation and Regeneration.
2005. Investigation on 14 historic buildings of Pin-Tong
County. Pin-Tong County government, in Traditional
Chinese
Hsu, U.-C., 2002, Investigation project of historic build-
ing – Chang family temple in Nan-Tou, Nan-Tou County,
Taiwan, in Traditional Chinese
Huang, P. &. Shu, M. S., 2001, A study on seismic failure
and maintenance of historic buildings. I: Bamboo and
timber construction, Council for Cultural Affair, Taipei,
in Traditional Chinese
King W.S., Yen J.Y., Yen Y.N., 1996, Joint characteris-
tics of traditional Chinese wooden frames, Engineering
Structures, Vol. 18, No. 18, pp. 635–644.
Miyamoto S., Miyazawa K., Irie Y. Wu J., Nomata Y. &
Goto, O. 2004, A study on seismic performance and
338
seismic diagnosis, seismic retrofit of Japanese temple,
Proceeding of 13th World Conference on Earthquake
Engineering, August 1–6
Tsai, P.-H. & D’Ayala, D. & Lewis, G., 2006, Finite Element
Modelling of Dieh-Dou Buildings in Taiwan, Structural
Analysis of Historical Construction V. 2. pp. 1297–1304.
Delhi: MacMillan India
Wood Handbook, 1999, US department of Agriculture Forest,
USA
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Experimental techniques for the evaluation of the long-term behaviour of

masonry strengthening materials

P. Bocca & A. Grazzini

Department of Structural Engineering and Geotechnics, Turin Polytechnic, Italy

ABSTRACT: The restoration of historical buildings is a delicate, complex process, in that newly developed
strengthening materials, whose long-term effects have not yet been fully tested, may interact adversely with the
materials making up the masonry. In the analysis of the behaviour of masonry structures and their constituent
materials, increasing importance has been assumed by the study of the long-term evolution of deformation and
mechanical characteristics, that may be affected by both loading and environmental conditions, especially in
the case of structural restoration by means of innovative materials, which is compounded by the problem of
the durability of the support-mortar system. The purpose of this investigation is to develop a methodology to
be used in laboratory tests as a preliminary design stage for structural interventions, in order to pre-qualify the
strengthening mortars and be able to formulate a judgement as to their compatibility and long-term behaviour
when applied to historical masonry walls.

1 INTRODUCTION

Historical buildings often require drastic interven-

tions, performed with techniques that make extensive
use of the new products. The tests conducted are often
limited to determining their ultimate strength, dis-
regarding their durability and their interaction with
pre-existing materials. Unfortunately, where technical,
conceptual and cultural aspects are concerned, there is
not much clarity yet about the use of modern materials
in ancient constructions. In many instances, restora-
tion works are performed according to the criteria
of modern technologies, with newly-developed mate-
rials, modelling the buildings according to resisting
schemes that are not appropriate to structures cre-
ated in stages at different times, thereby giving rise
to hybrid forms of behaviour that cannot be read-
ily foreseen (Binda & Saisi 2002). As a result, it
proves virtually impossible to verify their efficacy and Figure 1. The Royal Palace of Venaria Reale.
durability.
For years, the Non Destructive Testing Labora-
tory of the Structural Engineering Department of the residences of the Savoy family) represent the most
Politecnico of Turin has been working on a line of significant restoration project under way in Europe
research within the framework of an important con- (www.lavenaria.it, Fig. 1). The goal of the research is
vention stipulated with the Cultural Heritage Division the experimental study of the long-term behaviour of
of the Piedmont Region, on the theme “Study of the the constituent materials of the masonry, with special
quality of the applications of innovative materials in regard to the mechanical interaction between historical
strengthening interventions on historical artefacts”. bricks and modern strengthening mortars. Their long-
The field of application is a site in the proximity term behaviour remains unknown in several respects,
of Turin, where the restoration works on the Royal especially when they are applied to deteriorated histor-
Palace of Venaria Reale (one of the most important ical masonry structures, whose mechanical behaviour

339
is often difficult to analyse and has to be assessed case Table 1. Elastic modulus and compressive strength of
by case. materials.
Through a rather fast laboratory procedure
Eaverage σaverage %σ
(Bocca & Grazzini 2004), this method supplies useful Material (N/mm2 ) νaverage (N/mm2 ) (6 months)
indication for selecting, from a range of alternatives,
the product that is best in keeping with the mechan- Mortar A 6208 0.12 8.27 −7.50
ical characteristics of the historical material, thereby Mortar B 7534 0.19 10.91 +111.55
avoiding the errors associated with materials that are Mortar C 12678 0.23 10.34 +146.39
not mechanically compatible (Valluzzi et al. 2002). By Mortar D 12274 0.32 24.95 +57.47
focusing on the evolution of the deformation param- Historical 4099 0.08 8.09 –
eters in accordance with recent damage models, it is brick
possible to compare the characteristics of the materi-
als, assess their interaction and fatigue behaviour. An
important part was devoted to the execution of cyclic
tests, using new monitoring criteria, leading to a more
comprehensive understanding of the fatigue phenom-
ena that jeopardise the brick-mortar system. Special
attention was devoted to thermo-hygrometric aspects,
which are often overlooked, but whose effects on the
masonry system are often significant enough to com-
promise the validity of strengthening interventions.

2 EXPERIMENTAL CAMPAIGN:
INSTRUMENTATION AND MATERIALS

The materials used consisted of historical bricks from

the Royal Palace of Venaria Reale and four differ-
ent types of mortar (referred to as “A”, “B”, “C”,
“D”) produced by different manufacturers, suitable Figure 2. Single specimens.
for the consolidation of monumental masonry struc-
tures through the following types of intervention:
application of plaster to historical masonry (A; C),
joint closing (A; D), jacketing of masonry walls (D),
reinforcement of vaults (D), consolidation by grout
injection (B). The mechanical characteristics of single
materials have been investigated (table 1) by means of
test pieces 40 × 40 × 160 mm (Fig. 2).
Special attention was devoted to the study of
mixed brick-mortar test pieces, produced for the pre-
qualification of a strengthening product. These test
pieces were subjected both to static and to cyclic load-
ing tests and to freezing-thawing tests, by selecting
appropriate measuring systems in order to analyse, in
particular, the variations in the deformation status of
the materials under the various loads that can affect the
masonry structure after the restoration process. The
material used to produce the mixed tests pieces was
obtained from historical bricks cutted, using specially Figure 3. Instrumentation for mixed brick-mortar test
pieces.
designed timber mould measuring 223 × 57 × 83 mm
(30 mm thick layer of mortar).
Each mixed piece was labelled with “X” L, where
“X” stands for the code of the relative mortar. Cyclic pair was anchored to the plates of the MTS and another
and static compressive tests were performed with the vertical pair was applied directly to the test piece on
aid of a 250 kN model 810 MTS. The mixed pieces the opposite faces of the two materials; finally, a third
were instrumented with three pairs of transducers, pair arranged horizontally was used for measuring the
according to the scheme shown in figure 3: a vertical displacements due to bulging. Completing the system,

340
Table 2. Mechanical characteristics during maturation: Table 3. Mechanical characteristics during maturation:
mortar B. mortar D.

σ σ % E  E% σ σ % E  E%
Test Piece Days (N/mm2 ) (28-d) (N/mm2 ) (28-d) Test Piece Days (N/mm2 ) (28-d) (N/mm2 ) (28-d)

MB 4.1 28 12.38 7262 MD 4.3 28 20.08 9886

MB 7.3 28 12.54 6960 MD 6.2 28 20.97 10957
MB 9.3 28 11.32 7944 MD 7.3 28 20.87 11231
Average 28 12.08 – 7389 – Average 28 20.64 – 10691 –
MB 3.3 90 19.74 10676 MD 9.3 90 28.33 12455
MB 1.3 90 19.59 10089 MD 5.3 90 28.32 12907
MB10.2 90 20.37 11290 MD 9.1 90 28.09 11947
Average 90 19.90 +64.71 10685 +44.62 Average 90 28.25 +36.87 12436 +16.32
MB 2.2 150 17.24 8347 MD 1.1 150 28.99 12849
MB 6.1 150 20.02 7923 MD10.2 150 28.46 13152
MB 6.2 150 20.22 9046 MD 6.2 150 28.10 10448
Average 150 19.16 +58.61 8439 +14.21 Average 150 28.52 +38.18 12150 +13.64
MB 2.3 210 17.47 12000 MD10.1 210 28.91 12406
MB 8.1 210 17.94 10223 MD 6.3 210 29.16 14821
MB 3.1 210 19.02 9680 MD 2.2 210 29.66 16177
Average 210 18.14 +50.16 10634 +43.92 Average 210 29.24 +41.70 14468 +35.32

a pair of electrical strain gauges on the opposing verti-

cal faces of the test pieces measured transverse strains.
In this manner it’s possible to measure all the axial
deformations (in vertical, horizontal and transverse
directions), whose algebraic sum yields the volumetric
deformation.

3 RESULTS OF THE EXPERIMENTAL

CAMPAIGN: SINGLE MATERIAL

3.1 Static and freezing-thawing tests

For mortar B and D, the evolution of mechanical char-
acteristics in the time due to the maturation effects has
been analysed, verifying that the changes in mechan-
ical behaviour are remarkable (tables 2–3). A consid-
erable increase in the compressive strength (+64.71%
for B, +36.87% for D) is registered between 28 and 90
days, in addition to which values are stabilized a slight
decrease (Figs 4–5). The elastic modulus increases
sensitive value: mortar B reaches the peak (+44.62%)
in 90 days, at 90 days mortar D records a smaller
increase (+16.32%) that remains constant over time.
The test piece of mortar B and D were subjected to
freezing-thawing tests, in order to analysis the behav-
ior of materials exposed to adverse environmental
conditions as seasonal temperature ranges (table 4).
None of the test piece B remained intact to reach the
minimum threshold of 25 cycles. On the contrary mor-
tar D after 25 cycles showed a significant increase in
resistance (+34.27%), while the elastic modulus val-
ues has been discordant showing individual variations
Figure 4. Compression test during maturation: mortar B.
(−1.95% on the average). After 50 cycles there is a
decrease in compressive strength against the first set
of 25 cycles (Fig. 6), bringing the average values close
to the virgin reference (−3.12%). The elastic modulus
decreses −5.77%.
3.2 Brittleness index
It’s possible to evaluate the mechanical behaviour
of mortars in terms of brittleness (Bocca et al.
1988, Krajcinovic 1996), extrapolating the brittle-
ness indexes bind to the subtended areas under the
load-displacement curves, corresponding to the work
developed by the material during compressive tests.

341
 Figure 7. Apf and Apstf compute.

evaluated as the ratio between the subtended area Apf

Figure 5. Compression test during maturation: mortar D.
under the rising branch until ultimate load and the total
area Atot (Fig. 7):

to increase the value of such index the brittleness will

increase. Such characteristic corresponds to the apti-
tude of the material to arrive to failure in a sudden
way. Ipstf represents, on the contrary, the tendency to
have a ductile behaviour, and it’s expressed by the ratio
between the area Apstf subtended under post-failure
curve and the total area Atot (Fig. 4):

to increase Ipstf the ductility will increase, that is the

Figure 6. Compression test after freezing-thawing capacity to distribute the strains, to happened failure,
cycles: m. D.
deforming plastically.
However just referring to these ratio it’s not pos-
Table 4. Static tests after freezing-thawing cycles: mortar D.
sible to consider the real value of failure load, since
σi σf σ % Ei Ef E % materials with different strength could have the same
Piece Cycles (N/mm2 ) (0-cycl) (N/mm2 ) (0-cycl) indexes. In order to consider the rapidity by the mate-
rial loses the capacity to resist to further strains once
MD 1.2 25 20.64 29.66 +43.70 13557 11119 −17.98 the collapse, it’s possible considering the slope of
MD 8.2 25 20.64 25.59 +23.98 13201 10376 −21.40 the unloading branch of load-displacement diagram:
MD 5.2 25 20.64 27.88 +35.08 11066 11257 +1.73
MD 3.2 25 20.64 – – 11841 13386 +13.05 to increase the slope the brittleness will increase, of
MD 4.2 25 20.64 – – 8826 12167 +37.85 against, a slope less accentuated will be index of
MD 7.1 25 20.64 – – 9003 7869 −12.59 ductility. It’s possible to express brittleness index Ib as:
Average 25 20.64 27.71 +34.27 11249 11029 −1.95

MD 3.2 50 20.64 24.08 +16.67 11841 12930 +9.20

MD 4.2 50 20.64 17.37 −15.84 8826 10453 +18.43
MD 7.1 50 20.64 18.53 −10.22 9003 8418 −6.50
Average 50 20.64 19.99 − 3.12 11249 10600 −5.77
where Ccs is the angular coefficient of the descending
branch which interpolate the values of the post-failure
curve when there is a precipitous loss of performance.
Two parameters are extrapolated: pre-failure index By means of index Ib has been possible evaluating
(Ipf ) and post-failure index (Ipstf ). Ipf gives infor- the brittle or ductile mortars behaviour during their
mations about the brittleness of material and can be maturation (tables 5–6).

342
Table 5. Brittleness index during maturation: mortar B. Table 7. Brittleness index after freez.-thawing cycles:
mortar D.
Apf Apstf Ib %
Piece d (KNmm) Ipf Ipstf | Ccs | Ib (28-d)  Ib %
Piece Cycles Ipf Ipstf | Ccs | Ib (0-cycles)
MB 4.3 28 24.64 7.73 0.761 0.239 14.27 0.011
MB 6.2 28 21.61 16.03 0.574 0.426 11.86 0.007 Average 28 days 0.073
MB 7.3 28 20.64 14.31 0.591 0.409 14.96 0.009
Average 28 0.009 – MB1.2 25 0.811 0.189 135.52 0.110
MB5.2 25 0.689 0.311 106.02 0.073
MB 3.3 90 15.00 3.39 0.816 0.180 151.89 0.124
MB 1.3 90 13.46 4.10 0.767 0.233 151.39 0.116 Average 25 0.092 +25.08
MB10.2 90 17.12 7.42 0.698 0.302 72.41 0.051
MB3.2 50 0.894 0.106 279.54 0.250
Average 90 0.097 +995
MB4.2 50 0.721 0.279 83.58 0.060
MB 2.2 150 12.19 2.60 0.824 0.176 177.03 0.146 MB7.1 50 0.718 0.282 83.71 0.060
MB 6.1 150 15.20 3.31 0.821 0.179 199.05 0.163 Average 50 0.123 +68.73
MB 6.2 150 18.82 3.22 0.854 0.146 206.54 0.176
Average 150 0.162 +1732
MB 2.3 210 10.71 1.81 0.855 0.145 316.93 0.271
MB 8.1 210 12.85 3.50 0.786 0.214 213.83 0.168 Table 8. Results of preliminary static tests on mixed test
MB 3.1 210 14.12 1.11 0.927 0.073 910.29 0.844 pieces.
Average 210 0.428 +4741
Pmax σmax σaverage E
Series Test piece (KN) (N/mm2 ) (N/mm2 ) (N/mm2 )

AL AL02 102.75 19.30 11988

Table 6. Brittleness index during maturation: mortar D. AL04 59.76 11.49 15.40 14157
BL BL01 108.51 22.17 16940
Apf Apstf Ib % BL02 52.30 11.60 16.89 4400
Piece d (KNmm) Ipf Ipstf | Ccs | Ib (28-d) CL CL01 40.78 9.71 6597
CL02 76.98 15.46 12.58 12478
MD 4.3 28 16.00 6.65 0.706 0.294 87.84 0.062
DL DL01 58.50 12.10 6191
MD 6.2 28 19.17 5.35 0.782 0.218 116.64 0.091
MD 7.3 28 17.72 6.17 0.742 0.258 89.20 0.066 DL02 60.45 11.98 12.04 8106
Average 28 0.073

MD 9.3 90 29.56 13.27 0.690 0.310 91.30 0.063

MD 5.3 90 28.23 5.01 0.849 0.151 198.67 0.169 Table 9. Results of static tests after freezing-thawing cycles.
MD 9.1 90 30.88 8.45 0.785 0.215 123.42 0.097
Average 90 0.110 +49.8
Pmax σmax σaverage E
MD 1.1 150 28.20 6.34 0.816 0.184 188.40 0.154 Series Piece Condition (KN) (N/mm2 ) %σ var. (N/mm2 )
MD10.2 150 32.75 14.33 0.696 0.304 83.68 0.058
MD 8.1 150 23.86 7.68 0.757 0.243 106.21 0.080 AL AL03 cracked 95.54 19.78 10050
Average 150 0.097 +33.2 AL06 detached 81.00 15.83 8151
AL08 detached 59.30 12.03 15.88 +3.15 6701
MD10.1 210 27.50 5.63 0.830 0.170 155.69 0.129 BL BL07 cracked 76.30 14.23 6250
MD 6.3 210 28.70 4.38 0.868 0.132 282.39 0.245 BL10 cracked 66.50 13.52 13.88 −17.81 6582
MD 2.2 210 27.72 2.67 0.912 0.088 313.81 0.286 CL CL06 whole 104.50 9.92 10604
Average 210 0.220 +201 CL08 whole 54.62 11.13 14.88 +18.25 7191
DL DL08 whole 107.40 21.52 35358
DL07 whole 129.30 24.22 22.87 +89.93 16249

The pieces of mortar B tested within 28 days show a
low brittleness index which initially justifies the duc-
tile behaviour. But during the maturation there is a
remarkable increase of Ib , about 4700%. Such ten-
dency finds on facts in the diagrams which show an
always steeper slope unloading branch as it evolves
the maturation of material (Fig. 4). On the contrary,
mortar D shows at 28 days a less ductile behaviour,
but during the maturation the increase of brittleness
is modest (50–200%) so it highlights a much more
constant behaviour in the time.
4 RESULTS OF THE EXPERIMENTAL
CAMPAIGN: MIXED TEST PIECES
4.1 Static tests
Compression tests were performed in order to deter-
mine the failure load (table 8). Others mixed test
pieces were subjected to 28 freezing-thawing cycles.
Of special interest is the medium and great increase
in strength observed in CL and DL series after the
freezing-thawing cycles (table 9).

343
 Figure 9. AL series cyclic tests: max vertical deformations.

Figure 8. Monoaxial cyclic compression test σ − ε curve.

4.2 Cyclic loading tests

These tests were the most significant part of the testing
campaign, on account of being of decisive impor-
tance in the evaluation of the long-term behaviour of
strengthening mortars. The high value selected for the
cyclic load (70% of the static load) was designed to
make the test severe enough despite the short duration
of the test (100000 cycles −1.3 Hz), and to high-
light the potential of several indicators monitored over Figure 10. BL series cyclic tests: max vertical deforma-
time. Among these, the evolution of the volumet- tions.
ric deformation of mixed pieces was analysed, since
its propensity to assume negative values (increase in
volume) can reflect a lesser degree of collaboration
between the two materials, or even their detachment at
the interface. Together with the vertical and horizontal
deformations, this is an indicator of the validity of the
combination between the historical material and the
strengthening mortar. The test was performed through
four steps:
– initial 70% loading-unloading test (3 cycles);
– 70% cyclic test (100000 cycles);
– final 70% loading-unloading test (1 cycle); Figure 11. CL series cyclic tests: max vertical deforma-
– post-cyclic compression test to failure. tions.
In a typical σ − ε curve of a cyclic fatigue test
(Fig. 8) it is possible to identify three distinct stages: goal is to ascertain whether the fatigue life of the
stage I, where deformations are seen to increase rapidly mixed brick-mortar system also depends on the rate
(accounting for ca 10% of the service life of the of increase of vertical deformations during stage II
test piece); stage II, of stabilisation, where the defor- (secondary creep rate). In figures 9–12, it can be seen
mations increase gradually at a virtually constant that the test pieces that reached failure displayed a
stress (10–80% of test piece life); stage III, with a steeper slant in the stage II section of the curve, fol-
rapid increase till failure (Anzani et al. 2006). Var- lowed, at ca 80–90% of test piece life, by a sudden
ious authors (Minh-tan et al. 1993, Mu et al. 2005) increase at stage III (failure). Conversely, the curves
have shown that the fatigue life of a material sub- obtained for the pieces that passed the 100000 cycles
jected to cyclic loading tests is strictly correlated to mark displayed a lesser slant, which remained virtu-
the evolution of the deformations during stage II. ally the same through the end, reflecting an effective
By analogy with the method suggested for concrete behaviour still far from the failure.
(Taliercio & Gobbi 1996), the evolution of vertical From the results of the cyclic tests described above,
deformations over time is analysed as the primary through linear interpolation between 20% and 80%
parameter for predicting and quantifying the fatigue of secondary creep values, derivatives ∂εv /∂n (i.e., the
strength of the material (Carpinteri et al. 2006). The variations in the deformation vs. time curve during

344
 Figure 13. AL series cyclic tests: volumetric deformation.
Figure 12. DL series cyclic tests: max vertical deforma-
tions. Table 10. Analysis of the data.

stage II) were worked out. Through a linear regression Test Log
on the logarithmic scale (Grazzini 2004), it is possible piece n ∂εv /∂n LogN (∂εv /∂n) Nthe
to plot the data in a diagram to obtain an analytical
relationship (10) between secondary creep variations, AL01 22380 0.0270 4.350 −1.569 25583
AL05 53465 0.0198 4.728 −1.703 32029
∂εv / ∂n, and the number of cycles (N) to fatigue failure:
BL03 100000 0.0047 5.000 −2.323 90605
BL05 100000 0.0040 5.000 −2.398 102716
BL06 100000 0.0024 5.000 −2.612 147056
CL05 461 5.1818 2.664 0.714 555
CL09 1223 2.5110 3.087 0.400 941
CL10 15835 0.0501 4.200 −1.300 16294
In a majority of cases, especially in the case of the DL03 1149 0.4704 3.060 −0.328 3187
test pieces which failed to complete the cyclic test, the DL05 100000 0.0015 5.000 −2.813 206028
agreement is very good. DL06 100000 0.0070 5.000 −2.155 68328
The goal is to focus on the deformation response BL04 40993 0.0340 4.613 −1.469 21612
of the material subjected to fatigue tests consisting BL09 360 9.4729 2.556 0.976 358
of different cyclic load combinations. This requires CL04 100000 0.0035 5.000 −2.454 112832
considering additional factors affecting deformation CL07 46622 0.0192 4.669 −1.717 32795
behaviour, as a study of these parameters is needed DL09 100000 0.0025 5.000 −2.594 142671
DL10 100000 0.0113 5.000 −1.947 48171
to complete the analysis of all the mechanical prop-
erties and be able to predict the fatigue response
of the material. A valid correlation was established
between secondary creep rate (∂εv /∂n) during stage II than the actual value, corresponded to the time when
and fatigue life (number of cycles to failure, N). By a significant variation in trend was recorded, i. e. for
performing a certain number of cycles on the material piece AL05 in figure 13 (table 10): horizontal defor-
until it reaches the stage where deformations increase mations growing to a significant extent or volumetric
at a constant rate, it is possible to predict fatigue life deformations shifting to a negative sign (test piece
with a good degree of approximation. propensity to bulge due to poor vertical collaboration
Failure occurs when a deformation limit (corre- or detachment at the interface between the two mate-
lated to the loading level) is reached, after which the rials). The methodology and the numerical analysis
volume begins to increase; if the deformation rate is proved very sensitive to the initial signs of weakening
too slow, the material does not reach the limit value of the brick-mortar system, indicating clearly the onset
during cyclic loading and the values of volumetric of a crisis due to fatigue (Grazzini 2004, 2006).
deformation remain positive.
In some samples, the theoretical value Nthe was
4.3 Brittleness index: comparison between static
found to be lower than the value obtained from lab-
compressive tests, before and after
oratory tests. From a case-by-case analysis it can be
freezing-thawing and loading cyclic
seen that relationship (10), which is a function of the
secondary creep rate, is able to indicate the onset of the The figures 15–18 illustrate the static tests performed
crisis of the brick-mortar systems immediately preced- on mixed test pieces. As for static strength after
ing the final value of the testing cycles. The analysis freezing- thawing cycles, the values obtained for a
of the horizontal and volumetric deformations of these majority of the series remained within the average
test pieces, in fact, demonstrated that the theoretical of the two preliminary tests, save for the series DL
value obtained from expression (10), that was smaller (Fig. 18), which displayed a considerable increase in

345
Figure 14. AL series static tests.
Figure 15. BL series static tests.
Figure 16. CL series static tests.
strength. In most cases, the static curves after freezing-
thawing revealed a more brittle behaviour (table 11):
for AL and DL series further 340%. BL test pieces
(Fig. 16), instead, retained a good degree of ductil-
ity. The test pieces of BL and DL series that passed
100000 cycles mark were tested on to failure. The
former revealed a strength value in keeping with the
average values of preliminary tests and for BL series a
slightly less brittle behaviour (−32.3%); the DL series
displayed too a noticeable increase in their mechanical
properties and brittleness (+138%). Some CL and DL
test pieces were subjected both to the freezing-thawing
test and to cycling loading: piece CL maintained a
static behaviour similar to the weakest preliminary test
346
Figure 17. DL series static tests.
result (Fig. 17) and a decrease of brittleness (−75.8%).
In the DL series, instead, an appreciable increase
in strength was mated to a lesser degree of brittle-
ness compared to the test pieces subjected to cyclic
loading only.
5 CONCLUSIONS
The methodology adopted in the testing campaign
made it possible to identify a number of key parameters
for interpreting the long-term behaviour of strengthen-
ing mortars, based on the mechanical interaction with
the historical brick structure to which the mortar is
applied. In particular the evolution in the time of the
mechanical characteristics, due to maturation, thermo-
hygrometric and cyclic loading condition has been
possible to investigate in order to make a judgment
about the durability of materials.
In accordance with the theories formulated by
different authors, the analytical formula based on sec-
ondary creep variations during stage II demonstrated
the same sensitivity and the same degree of accuracy as
the volumetric deformation, by supplying the number
of cycles after which the collaboration between the two
materials begins to be undermined, without necessarily
resulting in test piece failure. By performing a certain
number of cycles on the material, until it reaches the
stage when the deformations grow at a constant rate, it
therefore becomes possible to predict the fatigue life
of the brick-mortar combination with a good degree of
approximation, without having to perform long series
of fatigue cycles.
Finally, the severity of freezing-thawing tests helped
to achieve a more comprehensive assessment of the
long-term behaviour of the materials, confirming
the influence of thermo-hygrometric stresses on the
masonry system. The investigation highlighted the
importance of the evolution over time of deformability,
as affected by loading and environmental conditions,
in the study of the fatigue behaviour of masonry
structures and their constituent materials.
Table 11. Brittleness index. REFERENCES
Apf Apstf Anzani, A Garavaglia, E., Binda, L. 2006. Time dependent
Piece (KNmm) Ipf Ipstf | Ccs | Ib Ib % Behaviour of Historic Msonry: a probabilistic model,
Proc. of the 7th IMC, London, UK, 50, CD-ROM.
AL02 48.61 50.34 0.491 0.509 103.62 0.051 Binda, L., Saisi, A. 2002. Conservazione degli edifici: il
AL04 128.06 217.99 0.370 0.630 50.86 0.019 miglioramento strutturale e la compatibilità tra i materiali
Average AL (virgin) 0.035 e la struttura originaria (in italian), L’Edilizia, 2, 34–41.
AL03 46.64 18.27 0.719 0.281 252.31 0.181 Bocca, P., Grazzini, A. 2004. Long term mechanical inter-
AL06 46.16 11.58 0.800 0.200 278.34 0.223 action between strengthening material and pre-existing
AL08 33.13 21.20 0.610 0.390 64.68 0.039 structures, Proc. 1th Int. Conf. Innovative Materials and
AL07 38.10 41.96 0.476 0.524 74.53 0.035 Technologies for Construction and Restoration, Lecce,
Average AL after 25 freezing-thawing cycles 0.120 +342 Italy, 2, 176–190.
BL01 64.12 31.16 0.673 0.327 230.85 0.155 Bocca, P., Carpinteri, A., Valente, S. 1988. On the appli-
BL02 56.01 24.10 0.699 0.301 54.75 0.038 cability of fracture mechanics to masonry, Proc. of the
Average BL (virgin) 0.096 8th Int. Brick and Block Masonry Conference, Dublin,
1027–1039.
BL07 52.16 50.53 0.508 0.492 58.22 0.030 Carpinteri,A., Bocca, P., Lacidogna, G., Grazzini,A., Masera,
BL10 57.73 30.80 0.652 0.348 62.00 0.040
D. 2006. Damage evaluation by acoustic emission in brick-
Average BL after 25 freezing-thawing cycles 0.035 −63.5
work structures under variable amplitude loading, Proc. of
BL03 25.91 26.46 0.495 0.505 103.94 0.051 the 7th IMC, London, UK, 32, CD-ROM.
BL05 48.07 26.14 0.648 0.352 175.10 0.113 Grazzini, A. 2004. Experimental analysis of the durability
BL06 25.55 28.61 0.472 0.528 68.22 0.032 of masonry strengthening works, PhD thesis, Dep. of
Average BL after 100000 loading cycles 0.065 −32.3 Structural Engineering , Turin Polytechnic, Italy, pp. 156.
CL01 21.18 25.92 0.450 0.550 23.67 0.011 Grazzini, A. 2006. Experimental techniques for the evalua-
CL02 55.15 17.18 0.762 0.238 215.20 0.164 tion of the durability of strengthening works on historical
Average CL (virgin) 0.087 masonry, Masonry International, 19 (3).
Krajcinovic, D. 1996. Damage mechanics, Ed. Elsevier,
CL06 56.30 16.46 0.774 0.226 331.05 0.256
Amsterdam, The Netherlands.
CL08 13.32 26.89 0.331 0.669 35.43 0.012
Average CL after 25 freezing-thawing cycles 0.134 +154 Minh-tan, D., Chaallal, O., Aitcin, P.C. 1993. Fatigue
behaviour of high-performance concrete, Journal of Mate-
CL04 12.32 22.62 0.353 0.647 59.14 0.021 rials Civil Engineering ASCE, 5, 96–111.
Average CL after28 freez-t. + 100000 loading cycles −75.8 Mu, B., Shah, S.P. 2005. Fatigue behavior of concrete sub-
DL01 25.81 17.71 0.593 0.407 124.19 0.074 jected to biaxial loading in the compression region,
DL02 30.30 28.11 0.519 0.481 40.49 0.021 Materials and Structures, 38, 289–298.
Average DL (virgin) 0.047 Mutluturk, M., Altindag, R., Turk, G., 2004. A decay func-
tion model for the integrity loss of rock when subjected
DL07 76.17 23.96 0.761 0.239 309.12 0.235 to recurrent cycles of freezing-thawing and heating-
DL08 56.94 25.05 0.695 0.305 194.16 0.135
cooling. Int. J. of rock mechanics and mining sciences, 41,
DL04 52.37 12.13 0.812 0.188 203.18 0.165
Average DL after 25 freezing-thawing cycles 0.178 +379 237–244.
Taliercio, A., Gobbi, E. 1996. Experimental investigation on
DL05 59.80 91.16 0.396 0.604 204.22 0.081 the triaxial fatigue behaviour of plain concrete, Magazine
DL06 42.80 28.10 0.604 0.396 81.05 0.049 of Concrete Research, 48, 157–172.
Average DL after 100000 loading cycles 0.065 +138 Valluzzi, M.R., Binda, L., Modena, C. 2002. Experimental
DL09 44.92 45.33 0.498 0.502 88.00 0.044 and analytical studies for the choice of repair techniques
DL10 33.05 42.60 0.437 0.563 89.28 0.039 applied to historic buildings, Materials and Structures, 35,
Average DL after 28 freez-t. + 100000 load cycles 0.042 −10.6 285–292.
Web site: www.lavenaria.it

347
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Experimental and numerical studies on the belltower of Santa Justa y

Rufina (Orihuela-Spain)

Salvador Ivorra
Departamento de Ingeniería de la Construcción, OO.PP. e I.U., Universidad de Alicante, Alicante, Spain

Francisco Pallarés
Departamento de Física Aplicada, Polytechnic University of Valencia, Valencia, Spain

José Miguel Adam

Polytechnic University of Valencia, ICICTH, Valencia, Spain

ABSTRACT: The bell-tower of Santa Justa y Rufina is located in the historical city of Orihuela (Alicante-
Spain). Its construction dates from the XV century, having suffered several changes later. The monumental group
was declared National Monument by the Spanish government in 1971. For the study of the structural behaviour
of the tower it becomes necessary to know its dynamic response influenced by two important aspects: a) the
construction is located in a high seismic risk area, and b) the tower is endowed with six bells that rotate on their
axis according to the Spanish characteristic system at a speed between 20 cycles/min and 40 cycles/min. With
the purpose of studying its dynamic response a monitoring program has been carried out using accelerometers
with the aim of determining the bending and torsional natural frequencies of the tower as well as approaching
the value of the structural damping ratio of the construction. Once the experimental results were analyzed, a
numerical model of the tower was carried out. This model has been calibrated using the experimental results
to assess the seismic vulnerability of the structure to earthquakes that currently the Spanish Standard states for
this area, as well as to evaluate the effect of the swinging of the bells after the future restoration planned. The
overall elastic modulus and joint rigidity of the tower and the nave of the church have been considered as control
parameters. A foundation of 1.5 m height was considered; the connection between the tower and the foundation
has not been considered rigid.

1 INTRODUCTION a stairway that gives access to the body of bells. All the
structure is made in masonry, arranging brick masonry
1.1 General with ashlars, maintaining a constant thickness of 1.5 m
along the whole first body of the tower. The total height
This paper presents an initial work made on the bell-
is 25.5 m. The bell room reaches the height of 35.5 m
tower of Santa Justa y Rufina church in Orihuela
with 1.0 m wall thickness. Seven windows are inside
(Alicante-Spain).This structure is a gothic belfry made
this body (in each wall) to accommodate the bells with
in XV century that possesses five bells. Some restora-
approximately 5.5 m2 each window.
tion works are planned on these bells and it is necessary
The tower has a rigid joint with the lateral walls of
to know de dynamic characteristics of the tower to dis-
the main body of the church. This joint can be observed
card future problems originated by a possible dynamic
in the west, south and east sides of the tower. The joint
interaction between the natural frequencies of the
in the west wall of the nave reaches the height of 18 m,
tower and the horizontal forces introduced by the bells
although the joint of the east wall reaches 9 m high.
on the tower.
(Figure 3)

1.2 Geometric description

The belltower has tree main bodies typical from the
gothic belltowers built in the XV century: base, bell
room and crown – Figs 1 and 2. The main body pos-
sesses a square section with 8.8 m long; inside there is
2 DYNAMIC TEST
Several dynamic tests have been performed on the bell-
tower to know the mechanical parameters, vibration
modes (bending and torsion) and structural damping.

349

Figure 1. (a) West façade. (b) East façade.
Figure 2. (a) General diagram of the tower. Architectural
plan.
All of them are based on the registration of ambi-
ent vibrations at different heights and directions. Only
bending and torsional vibrations are registered due to
the high longitudinal stiffness.
350
Figure 3. (a) Simplified model. (b) Model to simulate with
FEM.
The works of Bachmann (1997) and Casolo (1998)
fix the main torsional and bending frequencies
between 0.9 and 2 Hz for slender towers, and anal-
ogous results are obtained by Gentile (2007). The
work made by Ivorra (2006) in a similar belltower
allows evaluating frequencies and the experimental
procedure.
From equation (1) proposed in NCSE-02 (2002),
belltower frequencies can be estimated:
where: L is the plan dimension along the vibration
direction and H is the height.
So, it is expected to register a first frequency round
1.7 Hz or higher, since the stiffness is higher due to the
contact with the church.
In order to make the dynamic experimental mea-
surements, eight piezoelectric seismic accelerometers
have been placed at the height of the bell room
as shown in Fig. 4. The working range of these
accelerometers varies between 0.5 and 2000 Hz, with a
conversion factor equal to 1000 mV/g. According to
 Figure 5. Accelerometers arrangement in the belltower.
(a) Acquisition system. (b) Accelerometer no. 3.
(c) Accelerometer no. 7.

Table 1. Belltower natural frequencies.

Frequency Mode
Accelerometer Direction (Hz) classification

2-3-6 N-S 2.24 Bending

1-4-5-7 E-W 2.15 Bending
1-3-5-7 – 5.95 Torsion
Figure 4. Accelerometers arrangement in the belltower.
(a) 27 m height section. (b) Top of the tower.

the arrangement commented, belltower vibrations in
the E-W and N-S directions could be determined. The
dynamic data obtained from the ambient vibration
have been registered by a Kyowa PCD-320 equip-
ment with a sample rate of 200 Hz. Fig. 5 shows the
acquisition system and some accelerometers. Tem-
porary acceleration movements have been analysed
with DAS-100a Kyowa software to obtain frequency
results.
Power spectra responses in E-W and N-S direc-
tions have been obtained from ambient vibrations, so
the modal parameters can be concluded as shown in
Table 1 and Figure 6.
The structural damping ratio is obtained from the
results in Ivorra (2006), since the structures involved
have very similar characteristics and analogous period
of construction. The average damping ratio used in the
present work for the masonry belltower is 0.0159.
3 BELL FORCES
On the windows of the belfry five bells are located (see
Figure 8). These bells swing according to the Spanish
system: A counterweight provides a high level of bal-
ance (see Figure 7) and the bells, directly anchored
on the tower windows, rotate continuously in the same
direction.
The characteristics of these bells have been deter-
mined according to the works of Heyman & Therefall
(1976), Ivorra & Llop (2002) and Ivorra et al. (2005);
they are presented in Table 2.

351
 Figure 7. Bell Santas Justa y Rufina. Big rotating bell.

Figure 8. Bell disposition on the tower.

the real model. The results of this stage are shown in

Fig. 10 (vibration modes) and Table 3.
Figure 6. Power spectrum of obtained signal. (a)
Accelerometer 6: 2.15 Hz. (b) Accelerometer 4: 2.24 Hz. (c)
Accelerometer 5: 2.24 Hz, 5.95 Hz.
4 3-D FEM
As a first approximation, a simplified 5 degrees of
freedom model has been used to know the belltower
response to bell forces. A more refined numeri-
cal model was done using the commercial software
SAP2000TM . 8-node-hexaedral finite elements have
been used to mesh the model with three degrees of free-
dom per node. An iterative process has been performed
to fit, through a modal analysis, the fundamental fre-
quencies of the initial model and those registered from
The main assumptions for the numerical model are:
• Constant average material density 18 kN/m3 , as
stated in the Spanish standard NBE AE –88 applied
to masonry structures with solid bricks. In the beam
models it is supposed to be uniformly distributed.
• The Poisson’s ratio of the masonry was held constant
and equal to 0.15.
• Linear and elastic mechanical behaviour during the
calibration stage and modal analysis.
• Displacements restrained in the W-E direction on
the East wall up to the height of 18.2 m and on the
West wall up to the height of 10.3 m because of the
contact with the lateral wall of the nave. Equally, dis-
placements in the N-S direction are restrained in the
South wall for the same reason. These constraints
can be observed in Figure 9.
• The interaction between soil and structure is con-
sidered modelling the soil by a 17.7 × 17.7 × 8.25
meters 3D model. The elastic modulus and Pois-
son’s ratio are adjusted through an iterative process
to get appropriate values for the whole model in
accordance with the experimental values.

352
 Table 2. Characteristics of the bells of the tower.

Bronze Total Swing

Unbalance1 weight weight velocity
Bell (m) (N) (N) (Hz) (2) (3)

(1) Maria Antonia 0.08 1310 2096 0.67 0.65 1.81

(2) San Jose 0.07 2350 3760 0.57 0.54 1.41
(3) N.S. Guadalupe 0.07 4360 6976 0.58 0.58 1.71
(4) N.Sra. Rosario 0.09 7500 12000 0.42 0.7 2.18
(5) Stas Justa y Rufina 0.08 12720 20352 0.30 0.64 1.92

1
Distance between centre of rotation and centre of gravity.
2
Maximum horizontal force divided by total weight.
3
Maximum vertical force divided by total weight.

Figure 9. 3D FEM of the tower. Connections with chapel

and church. (a) North and west façades (b)East and south
façades.

Figure 9 shows the model of the soil with 10460

solid elements, 14312 nodes and 10.640 degrees of
freedom. To simulate the tower 4312 solid elements
and 6932 nodes are used.
After an iterative process an average elastic modulus
for the structure of 1.500 N/mm2 was obtained. For the
basement an average elastic modulus of 200 N/mm2
was obtained.
Table 3 shows the modal participating mass ratios
of the structure. These mass modal ratios permit to
identify the different modes.
Once the numerical model has been calibrated Figure 10. 3D FEM of the tower, (a) First mode (N–S). (b)
using the experimental results, the horizontal forces Second mode (W–E) (c) Third mode. Torsion.
caused by the swinging bells are introduced in the
model. From this analysis it has been obtained that the
dynamic amplification factor is lower than 1 for all
the bells, so no dynamic interaction is found between are separated more than 20% (DIN 4078) from the
belltower and bells. first three natural frequencies of the structure, so
The first three modal frequencies of the horizon- the dynamic amplification factor due to this dynamic
tal force introduced by the swinging bells (Fig. 11) interaction is negligible in this case. (Fig. 12)

353
 Table 3. 3D model. Modal participating mass ratios.

Axis
Direction Rotation-axis
Natural
Mode Freq. N–S E–W R(N−S) R(E−W) RZ
classification (Hz) % % Axial % % %

Bending 2.17 30.1 6.5 0.0 12.7 61.1 1.6

(N–S)
Bending 2.27 5.5 28.6 0.0 59.0 11.8 20.5
(E–W)
Torsion 5.31 1.9 0.5 0.0 0.2 2.9 14.7
Bending 6.70 7.2 5.3 4.2 5.1 6.5 1.8
(N–S)
Axial 6.95 0.3 0.2 75.4 2.2 1.8 0.1

2. With this non destructive and low-cost method five

vibration modes were clearly identified within the
frequency range 0–10 Hz.
3. This gothic belltower has its first natural frequency
higher than others presented in the scientific liter-
ature. This can be a singular aspect for this type of
tower, because the ratio between wall thickness and
slenderness can suggest a more rigid belltower than
other architectural styles.
4. On this belfry, the swinging of bells swinging in
the Spanish system causes no special or singular
structural problems.
Figure 11. Horizontal force of bell number 5.
ACKNOWLEDGEMENTS

Authors want to thank the Generalitat Valenciana

for funding this research through the granted project
GV06/135.

Figure 12. Power spectrum of horizontal force of bell
number 5.
5 CONCLUSIONS
A theoretical and experimental dynamic investigation
of a historic masonry bell-tower is described in the
paper. The following conclusions can be drawn from
the study:
1. A simplified and low-cost method is described to
evaluate the dynamic effect generated by the swing
of bells on the bell tower.
REFERENCES
Bachmann, H., Ammann, W., Deischl, F., 1997. Vibration
Problems in Structures: Practical Guidelines. Springer
Verlag, Berlin, 50–55.
Casolo, S., 1998.A three-dimensional model for vulnerability
analysis of slender medieval masonry tower. Journal of
Earthquake Engineering, Vol. 2, No. 4, 487–512.
Gentile, C., Saisi, A. (2007). Ambient vibration testing of
historic masonry towers for structural identification and
damage assessment, Construction and Building Materials
Vol. 21, 1311–1321.
Ivorra, S., Pallarés, F. (2006). Dynamic investigations on a
masonry bell tower. Engineering Structures, Volume 28,
Issue 5, April 2006, Pages 660–667.
NCSE-02. Norma de Construcción Sismorresistente Española.
Parte General y Edificación (Spanish Standard). Ministe-
rio de Fomento. 2002.
Heyman, J.,Therefall, B.D. Inertia Forces due to Bell Ringing,
International Journal of Mechanical Sciences, 18, 161–
164, 1976.

354
Ivorra, S., Llop F. Determinación de algunas características
físicas de una campana. In Proc. of the XIV Congreso
de Conservación y Restauración de Bienes Culturales.
Valladolid, p. 891–900 2002[In Spainsh].
Ivorra, S. et alter (2005) Dynamic forces produced by swing-
ing bells. Meccanica. International Journal of the the
Italian Association of theoretical and applied mechanics.
Ed. Kluwer Academic Press pub. Vol. 41- 1, 47–62, 2006
DIN 4178: “Glockentürme: Berechnung und Ausfhrung”;
1978.
355
Schutz, K. G., 1994, Dynamische Beanspruchung von
Glockentürmen. Bauingenieur 69, Springer-Verlag,
211–217.
Selby, A. R., Wilson J. M., 1997. Dynamic behaviour of
masonry church bell towers, Worldwide Advances in
Structural Concrete and Masonry. In: Proceedings of the
CCMS Symposium, Chicago, ASCE, NewYork, 188–199.
Wilson, J.M., Selby A., 1993. Engineering a Cathedral,
London, Thomas Telford Ltd., London, 77–100.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

The analysis of the strains and stresses of the reinforced brick vaults

J. Jasieńko & Ł. Bednarz

Wroclaw University of Technology, Wroclaw, Poland

ABSTRACT: The article specifies the aspects of the analysis of strains and stresses of the reinforcement of
the brick vaults. It introduces the problems of restoring and reconstruction of the vaults and presents various
ways of reinforcing these types of brick constructions applied to the dorsal side, minding at the same time that
frescos and other decorations are normally applied to the intrados side. What is more, the analytical methods and
numerical methods (FEM) used in the analysis of the construction of arches, vaults and domes, are introduced
together with the findings of the research conducted within the range of technical scale in the laboratory of the
Engineering Institute of the Technical University of Wroclaw. In addition, the measurement of the effectiveness
of various types of the adopted reinforcements was also made (inserted bar steels, grids and FRP tapes).

1 INTRODUCTION 2 ANALYSIS OF STRUCTURE

Keeping and maintaining the monumental construc-
tions, particularly the arches and vaults, which were
either partly or totally devastated, is the problem which
is encountered by more and more people dealing with
construction reinforcement.
The effective accomplishment of this task requires
an extensive knowledge about the origins of the effects
which contribute to deteriorating the construction’s
properties, and both the knowledge about the old con-
struction methods and modern methods which enable
one to prevent those effects and to restore the objects
within the scope of their former carrying capacity and
stiffness.
Implementing the modern methods of restoring and
reinforcing the brick constructions or their specific
elements (e.g. arches, vaults or domes), leads to the
considerable decline in costs, allowing at the same time
saving the culture-precious objects.
Reinforcing and restoring the brick constructions
of the arches and vaults are the difficult tasks which
usually require the individual approach to each case.
The analysis of the masonry constructions (espe-
cially the monumental arches, vaults and domes) is
a very complicated topic because of many problems
resulting from the description and from the simulation
of this type of constructions. Moreover, the numerical
approach is also a complex problem (simulation of the
material which in fact behaves in a nonlinear way), as
well as the limited information about the mechanical
properties of the materials and the complexity of the
geometrical construction.
Making the simulation of the masonry construction,
there are three major groups of analyses: linear anal-
ysis, nonlinear analysis and limit analysis. It is also
worth mentioning that despite the differences resulting
from the fact that various types of methods gener-
ate different safety coefficients and different final
results, there are no ample reasons for choosing the
one particular method over the others – during the
researches, it was decided that the simplest type of
analysis will be applied, namely the linear analysis.
Such a decision was made because of the applicabil-
ity of this type of analysis in the simple constructing
solutions.
2.1 The analytical models of calculation
A mathematical analysis of the curved brick construc-
tions placed an important position in the deliberations
of mathematicians and construction theoreticians for
a long time. What is more, many instructions sub-
stantially based on the engineering practice were
established.
Associating the practical knowledge with the theo-
retical deliberations, enabled one to the better under-
standing of the problem of the static work of arches,
vaults and domes. Making the analysis of the current
state of knowledge about estimating the capacity of
arches, vaults and domes, one can easily notice that
one of the major problems appearing on the level of
the estimation of carrying capacity is accepting the
static scheme of the curved brick construction with

357
application of the proper method of calculating the
internal forces in constructing elements.
On the way to constructive deliberations about
the statistics of the curved masonry constructions,
the systematization of those deliberations seems to be
intentional. However, this systematization should be
different from that which dominates over the histori-
cal aspects of the subject. Here, the static scheme of the
work of the groups, justifies the division into groups
which can be attributed to a certain group. The knowl-
edge of the static scheme enables the calculation of
internal forces which are present in elements of the
arches, vaults and domes as well it facilitates carrying
out the analysis of their work in respect to construction Figure 1. Research model of an arch.
security and in case of a failure in accepting the proper
method of strengthening.
However, the static calculations are based on vari-
ous simplified assumptions which are not truly real.
Therefore, such calculations cannot be made with
full precision. Uncertainty of the static calculations
becomes problematic when weight is taken into con-
sideration. The construction weight (especially in rela-
tion to monumental buildings) and snow and wind
burdens, are rather unknown and depend on other
factors which cannot be precisely calculated. Even
model-scale experiment cannot ensure certainty.
The problem which ought to be precisely analysed is
the issue of static calculations of the damaged arches,
vaults and domes. It must be specified how the various
types of damage (i.e. loosening of squinch from arch Figure 2. Real view of one of the examined arches –
rib, change of construction geometry) influence the arch A2.
way the internal forces flow through the curved brick
constructions and safety of the entire construction.
material. After a profound study of the subject’s
Although not all the mathematical considerations can
literary sources, it was decided that the masonry
be easily applied, there are a few considerations which
material will be treated as homogenous and isotropic
highly influence the analysis of the curved masonry
material as it has got the same physical properties in
constructions’ behaviour. Moreover, further solutions
all directions.
relate to various types of the curved brick construc-
tions, not only to the one described type of construction
(i.e. arches).
The group of major analytical methods com- 3 DESCRIPTION OF TEST MODELS
prises of:
Laboratory tests were conducted on 7 brick arches
– graphical method of establishing the line of pressure (0.12 m thick, 0.77 m wide, 4 m span and 2 m radius
in the arches, each – as in Fig. 1 and Fig. 2). Barrel arches built of
– three-jointed arches calculation method, materials of characteristics similar to those of mate-
– Heyman’s non-moment method applied to vaults rials in historical objects (full ceramic bricks of class
and domes. 100 and pure lime mortar). Static pattern of support
All of the numerical analyses of the monumental and loading realized in the same way in all the arches
buildings related to the static analysis, are encumbered tested.
with a basic error, namely it is very difficult or almost All of the arches models underwent monotonically
impossible to describe the resistance and deformabil- increasing static loading which was realized on the
ity of materials in present masonry constructions. If entire width of the arch, 1/3 by an actuator until the
this was possible, it would enable taking the correct destruction of the arch.
calculation model. Each arch was reinforced in a different way:
The main problem of such approach is the accep- – A1 model – the arch without any reinforcement –
tance of the most precise and efficient model of witness arch,

358
Figure 3. Model and arch A1.
Figure 4. Model and arch A2.
– A2 model – the arch reinforced with FRCM –
Ruredil X Mesh C10 carbon fibre screen set in the
Ruredil X Mesh M 25 mortar on the side of the
ridge,
– A3 model – the arch reinforced with 2 CFRP S&P
Lamelle CFK 150/2000 carbon tapes (100/1.4 inter-
section) placed in a 40 cm axial spacing and with
FRCM material – Ruredil X Mesh C10 carbon fibre
screen set in the Ruredil X Mesh M 25 mortar on
the side of the ridge,
– A4 model – the arch reinforced with FRCM –
Ruredil X Mesh C10 carbon fibre screen set in the
Ruredil X Mesh M 25 mortar on the side of the ridge
and on the intrados side,
– A5 model – the arch reinforced with ø 8 steel bars,
set in furrows made in the ridge surface of brick
arch, put ins with glue made on the basis of Epidian
5 epoxide resin,
– A6 model – the arch reinforced with 2 CFRP S&P
Lamelle CFK 150/2000 carbon tapes (100/1.4 inter-
section) glued in a 40 cm axial spacing on the side
of the ridge,
– A7 model – the arch reinforced with a girder made
of GL24 glued wood, joined with the brick arch by
the use of ø 8 steel anchors glued on the side of the
ridge.
359
Figures 3–5 present a diagrammatic way of enforc-
ing compared with a real view of another arch.
4 LABORATORY FINDINGS
In the course of the researches, the tracks of the static
balance for all the arches were marked out by mea-
suring the intensity of the force loading the laboratory
models. Figure 7 presents those tracks.
The reinforcing materials were supposed to be
set by the dorsal side of the arches or inside the
arches’ section. This type of reinforcement is domi-
nant in case of the historical arches and vaults which
are vividly decorated on the intrados side. In the
subject’s literary source, one cannot find any simi-
lar cross-sectional researches with so many methods
of reinforcing the curved brick constructions which
are acceptable with respect to the doctrine in the
constructional conservation of the historical brick
arches, especially with respect to FRCM reinforce-
ments (with the use of screens e.g. carbon screens
placed in a mineral die-block). The laboratory find-
ings showed that the reinforcement favorably affects
the carrying capacity and stiffness of the examined
arches.
Figure 5. Model and arch A5.

Figure 6. Model and arch A6.

Figure 7. Track of the static balance for arches A1, A2, A5 and A6.

Figure 8 presents the laboratory findings of all the
7 models of arches with specification of the ana-
lyzed arches: A1, A2, A5, and A6. One can also
notice the considerable growth of the destructive
force in all the reinforced arches comparing to the
model arch. Table 1 shows the proportional growth of
force.
The displacements of the arches at the application
force points with the different levels of loading were
also marked (see Figure 9).

360
Figure 8. Frontier amts of destructive force Fmax in the specific laboratory models.
Table 1. Growth of destructive force Fmax.
Failure force Fmax Growth of failure force
Arch [kN] [%]
A1 2.882 – the good quality of their workship.
A2 10.866 377.03
A5 19.314 670.16
A6 16.588 575.57
Figures 10–13 show the real schemes and models
of destruction of the examined arches. One should
also notice the transformation of those models of
destruction. Starting from the typical destruction tak-
ing place in the arch A1 and described in literature
(the model with 4 articulated joints), the next arches
the forming articulated joints change their place (they
are formed within bearings and under the place of
application force. One can also observe the shearing
of the arches on the bearings’ areas. In the practical
use of constructions, there are also very massive walls
at the places of bearings, which are much higher than
arches and which have the additional loadings (e.g.
loadings of pinnacles). Those walls prevent the arches
from a thrust so as the case of shearing in bearing
cannot occur.
The reinforced models gain much bigger deflections
than the witness arch, even with an additional loading.
It relates to especially those models of reinforcement
which involve using the carbon fibre screens where
the deformation and destruction proceeds slowly and
smoothly.
There is no any rapid movement related to separa-
tion of the reinforced material as it is with the models
361
reinforced by the CFRP tapes of steel bars alone. The
change of the schemes of destruction and growth of
the carrying capacity of the particular models of arches
prove the usefulness of the applied reinforcements and
In the course of the researches, also the defor-
mations were measured (arches A5 and A6) and the
normal stresses as well as tangential stresses in the
adhesive-bonded joints, carbon tapes (A6) and steel
bars (A5) were assigned. The results of those measures
are shown on the Figures 14 and 15 (normal stresses)
and on Figures 16 and 17 (tangential stresses).
5 NUMERICAL FINDINGS
It is generally acknowledged that for the detailed
description of the functioning and destruction of
masonry constructions, it is necessary to adopt the
calculative models with respect to nonlinearity and
brittleness of walls. However, in the case the analy-
sis of the static work of the examined arches, it was
decided to adopt much less complicated linear-elastic
models. This was mainly the result of the fact that
one of the assumptions was to enable the analysis of
the effectiveness of the reinforcements in the common
engineering papers.
On the bases of the created numerical models of the
arches, the series of numerical calculations in the area
of flexibility were made with respect to resistance and
strength. Basing on those calculations, the contour line
of deformations and level of stresses for the particular
calculative models was received. Figure 18 shows the
examples of the deformations in the arch A5 which
arose as a result of unloading on 1 kN level. Figure 19
Figure 9. Relation: force “F” – displacement “u” at the loading application point.

Figure 10. Model and real scheme of the destruction of the arch A1(force: 2,882 kN).

Figure 11. Model and Real scheme of the destruction of the arch A2 (force: 10,866 kN).

362
Figure 12. Model and Real scheme of the destruction of the arch A5 (force: 19,314 kN).

Figure 13. Model and Real scheme of the destruction of the arch A6 (force: 16,588 kN).

Figure 14. Normal stresses “σ” in the steel bar at points 1–10 in the sequent amts of the loading with “F” force in the arch A5.

363
Figure 15. Normal stresses “σ” in the carbon tape at points 10–13 in the sequent amts of the loading with “F” force in the
arch A6.

Figure 16. Tangential stresses “τ” in the adhesive-bonded joint between points 1–10 in the sequent amts of the loading with
“F” force in the arch A5.

Figure 17. Tangential stresses “σ” in the carbon tape between points 10–13 in the sequent amts of the loading with “F” force
in the arch A6.

364
Figure 18. Deformations in the numerical model of the arch A5 – the loading level – 1 kN.

Figure 19. Comparison of the deformations at 5 rodman points in the arch A2 (for the real model – a full line and for the
numerical model – dashed line.

shows the comparison of the curves of deformations
at rodman points for the real model (a full line) and
the numerical model (dashed line).
6 CONCLUSIONS
It is easy to notice that the schemes of destruc-
tion in all of the reinforced arch models changed in
comparison with the A1 arch which was not reinforced
(Figure 10–13). It shows that the static work of the rein-
forced arches changed and it also proves the need for
the applied reinforcements. One should also consider
the schemes of destruction of the joints between an
enforcing material and an arch:
– within the A2 arch it was noticed that the carbon
fibre screen loosened from masonry construction
with brick elements – it shows that destruction
(discerption) took place inside the brick,

365
– within the A5 arch it was noticed that steel bars
loosened from masonry construction with brick ele-
ments – it shows that destruction (discerption) took
place inside the brick and glue composition fulfilled
its role,
– within the A6 arch it was noticed that the FRP tape
loosened from the masonry construction it shows
that destruction (discerption) took place inside the
brick.
All reinforced models of arches considerably
increased their carrying capacity. As it is presented
in the picture 8, the increase of carrying capacity is
very huge, ranging from 3,77 (A2 arch) through 5,76
(A6 arch) up to 6,70 (A5 arch).
The conclusion of the performed research on the
models in technical scale shows, it may be stated that:
– all research models were performed with refer-
ence to the practical use of analysed techniques in
conservation of the masonry constructions,
– the research proved the possibility and usability
of FRCM enforcement application, that is the car-
bon fibre screens submerged in mineral mortar
in order to enforce the historical curved brick
constructions,
366
– the serving facility of FRCM systems in arches and
vaults constructions was confirmed,
– the research proved possibility and usability of joint
application of CFRP tapes and FRCM system glu-
ing techniques and, in result use of the positive
properties of both aforementioned systems,
– the research proved the possibility and usability
of the application of steel bars glued inside the
construction (however, it seems to be very time-
consuming and labour-consuming),
– the research proved the possibility and usability
of the application of glued wood elements (joined
with the construction by the glued steel anchors)
in order to stiffen the arch layers (it should be
noticed that there are some difficulties with match-
ing of the curvature of enforcing elements made of
glued wood to the curvatures of arches and vaults
constructions),
– the intersections of CFRP tapes and steel bars are
very rarely used; the best solution of this problem
seems to be their preliminary conjunction which
enables the substantial use of their carrying capacity
what further results in increasing the effectiveness
of reinforcement. However, such solution is barely
feasible in this type of constructions.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Cracking simulation of brick-masonry elements subjected to

the double flat-jack test

A. Carpinteri, S. Invernizzi & G. Lacidogna

Dipartimento di Ingegneria Strutturale e Geotecnica, Politecnico di Torino, Torino, Italy

ABSTRACT: In the present paper, we describe the results obtained from double flat-jack tests performed
varying the size of the masonry prism involved in the test. In addition, not only the deformations have been
acquired, but also the acoustic emissions (AE), in order to get information about local cracking during the test.
We present a meso-scale numerical model of the test, where every brick of the masonry is modeled in the details.
Discrete cracks can arise both in the mortar joints and in the brick units. A good correlation is found between
the amount of cracking simulated numerically and the experimental acoustic emissions for different prism sizes.
The model is also able to catch the decrease in the compressive strength with increasing size. It is not possible
to obtain an easy direct relation between the acoustic emission and the amount of cracking; nevertheless, it is
possible to state that the two quantities are proportional to each other when increasing sizes are considered.

1 INTRODUCTION the cracking processes taking place in some portions

Nondestructive and instrumental investigation meth-
ods are currently employed to measure and check
the evolution of adverse structural phenomena, such
as damage and cracking, and to predict their subse-
quent developments. The choice of a technique for
controlling and monitoring reinforced concrete and
masonry structures is strictly correlated with the kind
of structure to be analyzed and the data to be extracted
(Carpinteri & Bocca 1991; Anzani et al. 2000). For
historical buildings, nondestructive evaluation (NDE)
techniques are used for several purposes: (1) detecting
hidden structural elements, such as floor structures,
arches, piers, etc.; (2) determining masonry character-
istics, mapping the nonhomogeneity of the materials
used in the walls (e.g., use of different bricks during
the life of a building); (3) evaluating the extent of the
mechanical damage in cracked structures; (4) detect-
ing voids and flaws; (5) determining moisture content
and rising by capillary action; (6) detecting surface
decay phenomena; and (7) evaluating the mechanical
and physical properties of mortar and brick, or stone.
This study addresses some of the aforementioned
problems deemed of special significance. The struc-
tural geometry was defined through the customary
survey methods. Damage, cracking, and the evolution
of these phenomena over time were assessed through
a number of nondestructive techniques: tests with flat-
jacks were conducted in order to evaluate the range of
stresses affecting the structures; and at the same time,
of the masonry structures were monitored using the
acoustic emission (AE) technique.
The AE technique has proved particularly effec-
tive (Carpinteri & Lacidogna 2002, 2003, 2006), in
that it makes it possible to estimate the amount of
energy released during the fracture process and to
obtain information on the criticality of the process
underway. Strictly connected to the energy detected
by AE is the energy dissipated by the structure being
monitored. The energy dissipated during crack forma-
tion in structures made of quasibrittle materials plays a
fundamental role in the behavior throughout their life.
Strong size effects are clearly observed in the energy
density dissipated during fragmentation. Recently, a
multiscale energy dissipation process has been shown
to take place in fragmentation, from a theoretical and
fractal viewpoint (Carpinteri & Pugno 2002a,b, 2003).
Based on Griffith’s assumption of local energy dissi-
pation being proportional to the newly created crack
surface area, fractal theory shows that the energy will
be globally dissipated in a fractal domain comprised
between a surface and a volume in the Euclidean
space. According to fractal concepts, an ad hoc the-
ory is employed to monitor masonry structures by
means of the AE technique. The fractal theory takes
into account the multiscale character of energy dis-
sipation and the strong size effects associated with
it. With this energetic approach it becomes possible
to introduce a useful damage parameter for structural
assessment based on a correlation between AE activity

367
in a structure and the corresponding activity recorded
on masonry elements of different sizes, tested to failure
by means of double flat-jacks.

2 NONDESTRUCTIVE EVALUATION TESTS

2.1 Flat-jack tests

The single flat-jack test concerns the measurements of
in-situ compressive stress in existing masonry struc-
tures by use of a thin flat-jack device that is installed
in a saw cut mortar joint of the masonry wall (ASTM
1991a). The method is relatively non-destructive.After
the slot is formed in the masonry, compressive stress
at that point causes the masonry above and below the
slot to get closer. Inserting the flat-jack into the slot
and increasing its internal pressure until the original
distance between points above and below the slot is
restored, can thus measure the compressive stress in
the masonry. The slots in the masonry are prepared by
removing the mortar from masonry bed joints, avoid- Figure 1. Typical set-up for in situ flat-jack test. The dimen-
sions given are those of the specimen referred to as Vol. 1.
ing disturbing the masonry. Care must be taken in order (Reprinted from Gregorczyk and Lourenço 2000).
to remove all mortar in the bed joint, so that pressure
exerted by the flat-jack can be directly applied against other, in a solid-unit masonry wall (Fig. 1). By
the cleaned surface of the masonry units. The state of gradually increasing the flat-jack pressure, a compres-
compressive stress in the masonry is approximately sive stress is induced on the masonry comprised in
equal to the flat-jack pressure multiplied by factors between.The stress-strain relation can thus be obtained
which account for the ratio Ka of the bearing area of measuring the deformation of the masonry. In addi-
the jack in contact with the masonry to the bearing area tion, the compressive strength can be obtained, if the
of the slot, and for the physical characteristic of the test is continued to local failure. However, this may
jack Km . In fact, the flat-jack has an inherent stiffness also cause damage to the masonry in the area adjacent
which opposes expansion when the jack is pressurized. to the flat-jacks. The tangent stiffness modulus at any
Therefore, the fluid pressure in the flat-jack is greater stress interval can be obtained as follows:
than the stress that the flat-jack applies to masonry,
and a conversion factor Km is necessary to relate the
internal fluid pressure to the stress really applied.
The average compressive stress in the masonry, fm ,
can be calculated as: where, δσm is the increment of stress, and δεm is the
increment of strain. On the other hand, the secant
modulus is given by:

where, p is the flat-jack pressure required to restore

the gage points to the distance initially measured
between them. We performed the tests using rectangu-
lar flat-jack 240 mm × 120 mm wide and 7 mm thick
(by BOVIAR s.r.l., Italy). Their calibration factor was
Km = 0.90–0.92. The loading procedure was synchro-
nized and the pressure was applied with a manual
equipment (pressure range between zero and 60 bar).
The usual coefficient of variation of this test method
can be estimated equal to 20%; therefore, at least three
tests have been carried out on each area of interest.
The double flat-jack test provides a relatively
non-destructive method for determining the defor-
mation properties of existing unreinforced solid-unit
masonry(ASTM 1991b). The test is carried out insert-
ing two flat-jacks into parallel slots, one above the
where, σm and εm are the actual stress and strain in the
masonry.
3 ACOUSTIC EMISSION MONITORING
Monitoring a structure by means of the AE technique
makes it possible to detect the onset and evolution
of stress-induced cracks. Crack opening, in fact, is
accompanied by the emission of elastic waves that
propagate within the bulk of the material. These waves
can be captured and recorded by transducers applied
to the surface of the structural elements (Fig. 2).

368
Figure 2. Acoustic emission measurement system.
Figure 3. AE signal identified by the transducer.
Figure 4. Counting methods in AE technique.
The signal identified by the transducer (Fig. 3) is
preamplified and transformed into electric voltage;
it is then filtered to eliminate unwanted frequencies,
such as the vibrations caused by the mechanical instru-
mentation, which are generally lower than 100 kHz.
The signal is then analyzed by a threshold measuring
unit which counts the oscillations exceeding a cer-
tain voltage value. This method of analysis is called
ring-down counting (Pollock 1973; Brindley et al.
1973).
As a first approximation, the counting number, N ,
can be correlated to the quantity of energy released
during the loading process. This technique also consid-
ers other procedures. For instance, by keeping track of
the characteristics of the transducer and, in particular,
of its damping, it is possible to consider all the oscilla-
tions produced by a single AE signal as unique events
and to replace ring-down counting with the counting
of events (Fig. 4).
369
Figure 5. Acoustic emission relationship between signal
detection distance and signal frequency.
3.1 AE data acquisition system
The AE monitoring equipment adopted by the writ-
ers consists of piezoelectric transducers fitted with
a preamplifier and calibrated on inclusive frequen-
cies between 100 and 400 kHz. The threshold level
of the signal recorded by the system, fixed at 100 µV,
is amplified up to 100 mV. The oscillation counting
capacity is limited to 255 every 120 s of signal record-
ing. In this way a single event is the result of two
recorded minutes.
As specified in the literature (Ohtsu 1996), the max-
imum amplitude of direct non amplified signals is
about 100 µV, hence, neglecting the attenuation by
reducing to a few cm the distance of the transducer
from the signal generation point, it can be assumed
that the measuring system is able to detect the most
meaningful AE events reflecting cracking phenomena
in the masonry. Attenuation properties, in fact, depend
on the frequency range: higher frequency components
propagate in masonry with greater attenuation (Fig. 5).
Based on experimental results, for a measuring area at
a distance of 10 m, onlyAE waves with frequency com-
ponents lower than 100 kHz are detectable (Carpinteri
et al. 2005). With this system, the intensity of a single
event is, by definition, proportional to the number N
recorded in the time interval (event counting). Clearly,
this hypothesis is fully justified only in the case of
slow-crack growth (Holroyd 2000).
4 FLAT-JACK AND AE TESTS
Flat-jack testing is a versatile and powerful tech-
nique that provides significant information on the
mechanical properties of historical constructions. The
first applications of this technique on some historical
Figure 6. Combined flat-jack test and AE monitoring.

monuments (Rossi 1982) clearly showed its great

potential. The test is only slightly destructive, and this
is why it is now widely accepted and used by monu-
ment monitoring and rehabilitation experts (Binda &
Tiraboschi, 1999; Gregorczyk & Lourenço, 2000).
When double jacks are used, this test works accord-
ing to the same principle as a standard compressive
test. The difference is that it is performed in situ
and the load is applied by means of two flat-jacks
instead of the loading platens. The test method is based
on the following assumptions: the masonry surround-
ing the slot notches is homogenous; the stress applied
to the masonry by the flat-jacks is uniform and the
state of stress in the test prism is uniaxial.
In order to assess the extent of damage in the
zone monitored using the AE technique, a compres-
sive test was conducted on the masonry through the
combined use of double jacks and AE sensors (Fig. 6). Figure 7. Schemes of the double flat-jack tests performed
The tests were carried out with flat-jacks measuring on different wall sizes.
24 × 12 cm2 . The cuts made into the masonry wall to
obtain a smaller-sized specimen were made into two
horizontal mortar joints spaced about 30 cm apart.
The minimum slenderness ratio of the specimens
was h/t = 2.5, where h is the height of the prism com-
prised between the two flat-jacks and t = 120 mm the
deepness of each flat-jack. This made it possible to
reduce the friction effects on masonry behavior arising
from the action of the flat-jacks.
During the tests, the stress-strain relationship of
the masonry was determined by gradually increasing
the pressure applied by the flat-jacks in the course
of three loading-unloading cycles. Peak compressive
strength was obtained from the load–displacement
diagram, when the latter became highly nonlinear,
denoting imminent failure. Compressive tests were Figure 8. Equivalent masonry prisms tested in compression
performed on three different masonry portions. The by means of double flat-jacks.
prismatic masonry volumes tested in compression
were delimited crosswise by vertical cuts (Fig. 7). Con- with the procedures specified in ASTM (1991b), other
sequently, the in-situ test is equivalent to a compression than for the vertical cuts produced in order to elim-
test performed on specimens with different sizes, as inate, in the cracked element, the influence of the
shown in Figure 8.The tests were performed in keeping adjacent masonry portions.

370
Figure 9. Double flat-jack test on Volume 2: cumulative
number of AE events (2) versus cyclic loading (1).

Table 1. Experimental values obtained from flat-jack tests

and AE measurements.

Specimen Volume (cm3 ) Peak stress (MPa) Nmax at σu

Vol. 1 8 640 2.07 ∼6 500

Vol. 2 16 992 1.61 ∼12 000
vol. 3 33 984 1.59 ∼18 000

Figure 9 shows the results obtained from these

tests for the intermediate element (Volume 2). Similar
results were obtained for the other two elements. The
figure also shows the three loading cycles performed
as a function of time and the diagram of the cumu-
lative number of AE counts. From the AE diagram it
can be clearly seen that the material releases energy
when the stress level reached previously is exceeded Figure 10. Experimental results obtained from the double
flat jack tests.
(Kaiser effect, Kaiser 1950). Moreover, from the dia-
gram, we find that the cumulative number ofAE counts
Table 2. Mechanical properties adopted in the analysis.
at failure stress (i.e. immediately before the critical
condition is reached) is Nmax ∼
= 12000. The experimen- Unit Joint
tal results obtained on the three masonry elements are
summarized in Table 1. Young’s modulus E 6.0 109 Pa 1.0 109 Pa
The stress-strain diagrams obtained from exper- Poisson ratio ν 0.15 0.15
iments are shown in Figure 10. The first cracking Tensile strength ft 3.0 106 N 3.0 105 N
load, which reasonably corresponds to the compres- Fracture energy Gf 50 N/m 10 N/m
sive strength of the masonry, is deduced not only from Shear ret. Factor β 0.01 0.01
a visual inspection during the test, but also monitoring Compress. strength fc 3 107 Pa 1 107 Pa
when the horizontal strain suddenly increases.

5 NUMERICAL SIMULATION 2005). The mechanical properties of the materials are

The numerical model of the double flat-jack test
was built exploiting the symmetry of the problem.
Quadratic elements were used to represent both the
brick units and the mortar joints. The failure of both
components was assumed as ideal plasticity in com-
pression and linear softening in tension. A fixed
smeared crack model based on total deformation was
used. All the analyses were performed with the Finite
Element Software DIANA 9.1 (de Witte & Schreppers,
summarized in Table 2.
Figure 11a shows the mesh used to model the
smallest specimen. Taking advantage of the problem
symmetry, only one quarter of the geometry has been
discretized. Figure 11b shows details about the load-
ing and boundary conditions. The following procedure
has been applied. First, a displacement is imposed to
the top of the specimen.
The amount of such displacement can be calculated
from another model of the masonry wall (“uncut”),

371
Figure 11. Finite element mesh adopted for Volume 1
exploiting symmetry (crf. Shaded area in Figure 7a). Mesh
and materials (a); loads and boundary conditions (b).

without the cut where the flat-jack is placed afterward.

This corresponds to the in situ configuration before
the test. The imposed displacement is determined
such then the vertical stress equals the in-situ value.
Afterwards, the pressure load in both sides of the cut
is applied incrementally. When the pressure reaches
the in-situ value of the vertical stress, the defor-
mation of the model approaches the configuration
obtained from the “uncut” model, exactly like in the
experimental procedure.
If the load is increased further, the material com-
prised in between the two flat-jacks starts to damage.
This behavior is caught correctly by the numerical
model. Figure 12 shows the stress-strain diagrams
obtained for the three different sizes. The arrows indi-
cates the moment at which the horizontal strain sud-
denly increases, that corresponds to the first vertical Figure 12. Stress-strain diagrams: Volume 1 (a); Volume 2
cracking. (b) and Volume 3 (c). The arrow indicates when first cracking
The compressive strength decreases with increasing spreads into the specimen.
the specimen size in a rather good agreement with
the experimental tests. On the other hand, the stress-
strain path in compression looks a bit stiffer than the
experimental one, especially after the cracking occurs. useful energy-related parameter for the determination
The crack pattern for the three sizes is shown of structural damage (as used by Carpinteri et al. 2003,
in Figure 13. It slightly changes varying the size, 2004, for reinforced concrete structures) by comparing
probably due to the different aspect ratio. the AE monitoring results with the values obtained on
In a previous work (Carpinteri & Lacidogna 2006), masonry elements of different sizes tested up to failure
a statistical and fractal analysis of data from labo- by means of double jacks.
ratory experiments was performed, considering the Fragmentation theories have shown that, during
multiscale aspect of cracking phenomena. The fractal microcrack propagation, energy dissipation occurs in
criterion takes into account the multiscale character a fractal domain comprised between a surface and the
of energy dissipation and the strong size effects asso- specimen volume V (Carpinteri & Pugno 2002a, b,
ciated with it. This makes it possible to introduce a 2003).

372
 5.0

log Nmax - log Cracked elements

4.0
y = 0.7431x + 0.9038
2
R = 0.9846
3.0

y = 0.6216x - 0.081
2.0 2
R = 0.893

1.0 Nmax
Cracked elements
0.0
3.9 4.0 4.1 4.2 4.3 4.4 4.5 4.6
log V/cm3

Figure 14. Volume effect on Nmax and on the number of

cracked finite elements.

600

500

Cracked elements
400
y = 0.0248x + 75.802
2
300 R = 0.93
Figure 13. Crack patterns due to flat-jack pressure in the
three specimens. 200

100

This implies that a fractal energy density (having 0

anomalous physical dimensions): 6000 8000 10000 12000 14000 16000 18000 20000

Figure 15. Linear dependency between Nmax and the num-

ber of cracked finite elements.

From Eq. (5) we get:

can be considered as the size-independent parame-
ter. In the fractal criterion of Eq. (4), Wmax = total
dissipated energy;  = fractal energy density; and
D = fractal exponent, comprised between 2 and 3. On
the other hand, during microcrack propagation, acous- from which we can obtain the structure critical number
tic emission events can be clearly detected. Since the of AE events Nmax . An energy parameter describing
energy dissipated, W , is proportional to the number of the damage level of the structure can be defined as the
AE events, N , the critical density of acoustic emission following ratio:
events, AE , can be considered as a size-independent
parameter:

N being the number of AE events currently recorded

where AE = fractal acoustic emission energy density;
and Nmax is evaluated at the peak stress, σu . Eq. (5)
predicts a volume effect on the maximum number of
AE events for a specimen tested to failure.
The extent of structural damage can be worked out
from the AE data recorded on a reference specimen
(subscript r) obtained from the structure and tested to
failure. Naturally, the fundamental assumption is that
the damage level observed in the reference specimen is
proportional to the level reached in the entire structure
before monitoring is started.
by the monitoring apparatus.
Now, we can assume that the number of AE is also
proportional to the number of Gauss points subjected
to cracking in the finite element model. Therefore, the
number of AE and the number of cracks in the finite
element model should show the same exponent with
respect to the considered volume. In fact, this is what
we can substantially observe from Figure 14.
The linear relation between the number of cracked
elements (or Gauss points) in the finite element model,
and the AE is put into evidence also in Figure 15, where
the two quantities are plotted in a direct comparison.

373
In fact, the calculated coefficient of linear regression
is equal to 0.93.
Finally, let us observe that the slope of this linear
relation depends on the discretization of the finite ele-
ment model. On the other hand, refining the mesh (e.g.
dividing by two the linear size of each element) does
not change sensibly the exponent in Figure 14.
6 CONCLUSIONS
A numerical simulation of an innovative double flat-
jack test combined with AE has been proposed. The
numerical results agree rather well with the experi-
mental evidences, both in terms of the estimated com-
pressive strength and of the crack pattern. The model
is also able to catch the decrease in the compressive
strength with increasing size.
In addition, the number of Acoustic Emissions is
put into relation with the number of Gauss points in
the finite element model where cracking takes place.
A good correlation is found between the amount of
cracking simulated numerically and the experimental
acoustic emissions for different prism sizes.
It is not possible to obtain an easy direct rela-
tion between the acoustic emission and the amount
of cracking; nevertheless, it is possible to state that
the two quantities are proportional to each other when
increasing sizes are considered.
REFERENCES
Anzani, A., Binda, L., and Mirabella Roberti, G. 2000. The
effect of heavy persistent actions into the behavior of
ancient masonry. Materials & Structures, 33:251–261.
ASTM. 1991a. Standard test method for in situ compressive
stress within solid unit masonry estimated using flat-jack
measurements. ASTM C1196-91, Philadelphia.
ASTM. 1991b. Standard test method for in situ measurement
of masonry deformability properties the using flat-jack
method. ASTM C1197-91, Philadelphia.
Binda, L., and Tiraboschi, C. 1999. Flat-jack test as a slightly
destructive technique for the diagnosis of brick and stone
masonry structures. International Journal for Restoration
of Buildings and Monuments, 449–72.
374
Brindley, B.J., Holt, J., and Palmer, I. G. 1973. Acoustic emis-
sion. III: The use of ring-down counting. Non-Destr. Test,
6(5):299–306.
Carpinteri, A., and Bocca, P. 1991. Damage and Diagnosis
of Materials and Structures, Pitagora Editrice, Bologna,
Italy.
Carpinteri, A., and Lacidogna, G. 2002. Structural monitor-
ing and diagnostics by the acoustic emission technique:
Scaling of dissipated energy in compression. Proc., 9th
Int. Congress on Sound and Vibration (ICSV9), Orlando,
Fla., Paper No. 166.
Carpinteri, A., and Lacidogna, G. 2003. Damage diagnosis
in concrete and masonry structures by acoustic emission
technique. J. Facta Univ., 3(13):755–764.
Carpinteri, A., and Lacidogna, G. 2006. Damage monitoring
of a masonry building by the acoustic emission technique.
Materials & Structures, 39:161–167.
Carpinteri, A., and Pugno, N. 2002a. Fractal fragmenta-
tion theory for shape effects of quasi-brittle materials in
compression. Mag. Concrete Res., 54(6):473–480.
Carpinteri, A., and Pugno, N. 2002b. A fractal comminution
approach to evaluate the drilling energy dissipation. Int.
J. Numer. Analyt. Meth. Geomech., 26(5):499–513.
Carpinteri, A., and Pugno, N. 2003. A multifractal comminu-
tion approach for drilling scaling laws. Powder Technol.,
131(1):93–98.
Carpinteri, A., Invernizzi, S., and Lacidogna, G. 2005. In
situ damage assessment and nonlinear modeling of an
historical masonry tower. Eng. Struct., 27:387–395.
de Witte, F.C., and Schreppers, G.J. DIANA Finite Element
Analysis User’s Manual, TNO DIANA BV, Delft, The
Netherlands.
Gregorczyk, P., and Lourenço, P.B. 2000. A review on flat-
jack testing. Engenharia Civil UM, 9:39–50.
Holroyd, T. 2000. The acoustic emission and ultrasonic
monitoring handbook, Coxmoor, Oxford, U.K.
Kaiser, J. An investigation into the occurrence of noises in
tensile tests, or a study of acoustic phenomena in ten-
sile tests. Ph. D. dissertation. Munich (FRG): Technische
Hochschule München; 1950.
Ohtsu, M. 1996. The history and development of acoustic
emission in concrete engineering. Mag. Concrete Res.,
48(177):321–330.
Pollock, A.A. 1973. Acoustic emission. II: Acoustic emission
amplitudes. Non-Destr. Test., 6(5):264–269.
Rossi, P.P. 1982. Analysis of mechanical characteristics of
brick masonry tested by means of in situ tests. In: 6th Int.
Brick/Block Masonry Conf., Rome, Italy.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Numerical simulation and monitoring of the Cathedral of

Syracuse in Sicily

A. Carpinteri, S. Invernizzi, G. Lacidogna & A. Manuello

Department of Structural Engineering and Geothecnics, Politecnico di Torino, Italy

L. Binda
Department of Structural Engineering, Politecnico di Milano, Italy

ABSTRACT: In the present paper the results from a recent monitoring campaign and numerical analysis
performed on the ancient Cathedral of Syracuse in Sicily are presented. The Acoustic Emission (AE) technique is
adopted to assess the damage pattern evolution. The localization of the propagating cracks is performed using six
synchronized AE sensors. A clear correlation between the regional seismic activity and the AE acquisition data is
shown. In fact the AE count rate presents peaks corresponding to the main seismic events. In addition, a numerical
analysis of the Cathedral is presented. The nonlinear Finite Element model is particularly refined to account for
the cracking in the most damaged pillar. Some recent seismic events in the area acted as crack propagators. The
crack occurrence obtained from the numerical analysis agrees quite well with the crack localization provided by
the AE monitoring.

1 INTRODUCTION damaged portion of a structure has been located, it

The Cathedral of Siracusa (Sicily – Italy) is located in
the higher zone of the Ortigia island. This region of
Sicily has been belonging to the UNESCO World Her-
itage List since 2005. The Cathedral of Syracuse is the
result of the transformation of the ancient Athena’s
Greek Temple (5th century B.C.), with modifica-
tions that have also been consequence of the damages
caused by earthquakes. The ancient temple had four-
teen lateral columns and six in the front. The Cathedral
incorporated many of them, as it is nowadays still vis-
ible, while other columns of the present structure were
built with the salvage remains from the temple cell.
At the present time, the structure shows an extended
damage pattern, especially in four of the nave pillars
(Binda et al. 2007).
In recent years, the authors have been working on
the development of a method for the assessment of
materials and structures based on the spontaneous
release of pressure waves originated by the evolu-
tion of damage. This monitoring technique, referred
to as Acoustic Emission (AE), is non invasive and non
destructive and therefore is ideally suited for the con-
trol of historic and monumental structures in seismic
areas (Carpinteri & Lacidogna 2006a, 2006b, 2007).
With this technique, if it is not known, the initial posi-
tion of the damage can be determined with the aid
of a multiplicity of sensors and through triangula-
tion (Carpinteri et al. 2006a, 2006b, 2007a). Once the
becomes possible to evaluate the stability of the evolv-
ing damage, which may either come gradually to a halt
or propagate at an ever faster rate. In this study, the AE
technique was used to determine the damage level in
a pillar that was part of the vertical bearing structure
of the Cathedral of Syracuse (Sicily).
Beside the experimental research, a numerical sim-
ulation based on the Finite Element Method is per-
formed, which is able to catch the main non linear
features of the mechanical behaviour such as fracture
nucleation, fracture propagation and time dependency.
The most damaged pillar has been modelled in details,
providing an accurate description of the blocks interac-
tion by means of discrete interface elements. The main
goal is to link the dissipated energy, acquired from
the acoustic emissions (AE), and the fracture localiza-
tion with the results obtained numerically (Carpinteri
et al. 2008). From the numerical model, it is possible
to observe that cracks develop due to the seismic load,
according to the experimental evidence from the AE
acquisition.
2 SYRACUSE CATHEDRAL AND STATE OF
PRESERVATION
In the 6th century AD, the 5th century B.C. Greek
temple of Athena in Syracuse, was transformed into
a Catholic Church, and successively became the
Cathedral of the City; the building was frequently

375
Figure 1. The Cathedral and its evolution.

modified along the centuries until the present config-

uration (Agnello 1950; Agnello 1996; Privitera 1863;
Russo 1991,1992).
Several styles and structural details belonging to the
different times can be recognised: (i) in the external Figure 2. Cracks pattern survey of the monitored pillar.
walls the ancient Greek columns and the filling wall
between them of the Byzantine time, (ii) the baroque
façade, (iii) the added apse and chapels. Furthermore,
being Syracuse in a seismic area, the Cathedral was
damaged, repaired or partially rebuilt several times
(Agnello 1950; Agnello 1996). Figure 1 shows the
evolution of the Cathedral plan along the centuries.
The Cathedral pillars have a peculiar interest; they
had been obtained cutting out the stonework walls of
the internal cell of the Greek temple. The pillars show
several repaired areas, replacements, but also several
cracks.
In order to evaluate their state of preservation, the
extension and the depth of the replacements and the Figure 3. The radiagram shows the presence of regularly
presence of internal defects, an investigation program spaced diffractions (about every 5–6 cm) at a depth of
was planned by the Sovrintendence of Syracuse and about 3–4 cm. The inspection proved that these pillars were
the Politecnico of Milan. As a first step, a survey of sometimes repaired with bricks masonry instead of stone.
the pillars with an accurate mapping of the superfi-
cial materials, of the defects, of the cracks and of the successfully exploited to diagnose the state of dam-
morphology was carried out. The crack pattern was age of the pillars (Binda et al. 2006). Furthermore a
classified and accurately documented and reported on monitoring of the cracks development carried out for
the geometrical survey. Figure 2 gives an example of approximately two years showed an evident trend to
this documentation referred in a damaged pillar. The increase their size in some cracks of the pillars posi-
cracks display frequently a vertical pattern due to com- tioned a the end of the central nave, which suggested
pressive stresses, whose action is often combined with a further check of the damage by acoustic emission.
the compressive-bending stresses caused by frequent
earthquakes. In some cases, the corners and part of the
3 AE MONITORING
stone blocks were expelled. The mortar traces in these
cases are trials to locally repair the damages. In the sur-
3.1 The monitored pillar
vey repaired cracks were cleared in order to evaluate
the evolution of the damages. The AE monitoring process was performed on a pil-
On the base of this detailed survey, NDT tests were lar of the Cathedral of Syracuse. The temple had 14
performed in order to investigate the depth of the dam- columns along the sides and 6 at front, and some of
age: (i) sonic and ultrasonic to find voids inside the them, belonging to the peristyle and the stylobate, can
pillars and the depth of the cracks respectively (in still be identified. In the layout of the Cathedral shown
some cases up to 40 cm), (ii) thermovision to detect the in Figure 4, all the pillars and the columns inside the
detachment of renders and repaired parts, (iii) radar to building are marked with a progressive number.
find internal cracks and inclusions (Fig. 3). Basically, the Doric columns are marked with num-
The complementarity of these NDTs already stud- bers in three ranges: 1–8; 22,23; 33–40; whereas the
ied in (Binda et al. 2000, Binda et al. 2003), could be pillars, obtained from the calcareous stone masonry of

376

Figure 4. Plan of the Syracuse Cathedral. (a) Pillars 19, 20,
30 and 31 are pointed out (b) Pillar 19 is indicated by a circle.
the temple cell, are identified with the remaining num-
bers. As said above, from the survey of the cracks, it
was determined that the pillars in the most critical con-
ditions were nos. 18, 19, 29 and 30, all of them located
near the end of the central nave (Fig. 4). These pil-
lars show an appreciable degree of deterioration, due
to the presence of added layers of plaster and con-
spicuous cracks, which in some cases seem to cut the
constituent stone blocks. Pillar no. 19, selected for the
application of the AE monitoring technique, is shown
in Figure 4 indicated by a circle.
The pillar no. 19 is one of the most damaged. This
element together with pillar no 18, is situated toward
the altar, and on this element a serious damage is
observed. It is important to stress the effects of the
1542 earthquake, which produced a great deformation
of the perimeter wall close to the pillars nos. 18 and 19.
This could justify the bad state of preservation of these
two pillars, characterized by the presence of detached
covers and deep cracks on all the prospects. Despite
the several intervention the section is reduced. The
cracks in these pillars also show a large movement
than elsewhere (Binda et al. 2007).
The pillar (save for a few strengthening works per-
formed – according to the Syracuse Superintendence
377
Figure 5. View of the four sides of the monitored pillar. In
the figure the various materials making the pillar are reported.
for Cultural Heritage – during a restoration process in
1926) was thought to be made of limestone blocks,
probably installed during the initial construction of
the temple dedicated to Athena in the 5th century B.C.
The investigation revealed instead the presence of parts
made with brick masonry ascribed to the interventions
at the beginning of the 20th century (Fig. 5). The lower
stiffness of these parts is probably the cause of the
damage developed in the stones.
3.2 AE equipment and “in situ” applications
details
Monitoring a structure by means of the AE technique
proves possible to detect the occurrence and evolu-
tion of stress-induced cracks. Cracking, in fact, is
accompanied by the emission of elastic waves which
propagate within the bulk of the material. These waves
can be received and recorded by piezoelectric (PZT)
transducers applied to the surface of the structural
elements. The signal is therefore analysed by a mea-
suring system counting the emissions that exceed a
certain voltage threshold measured in volts (V). The
leading-edge equipment adopted by the authors for the
analysis on the vertical bearing structures of the Syra-
cuse Cathedral consists of six units USAM®, that can
be synchronized for multi-channel data processing.
The most relevant parameters acquired from the sig-
nals (frequencies in a range between 50 and 800 kHz,
arrival time, amplitude, duration, number of events and
oscillations) are stored in the USAM memory and then
downloaded to a PC for a multi-channel data process-
ing (see Fig. 6). Microcracks localisation is performed
from this elaboration and the condition of the moni-
tored specimen can be determined (Carpinteri et al.
2006a, 2007a).
On each side of element 19 an evident cracking pat-
tern is observed. The AE sensors have been applied on
the middle part of the pillar as shown in Figure 7a and b.
In the Figure 7a the zones with capillar vertical cracks
are indicated by circles. The AE sensors arrangement
is represented in Figure 7a according to the scheme
 Figure 8. Differential (a) and cumulated (b) number of AE
oscillations during the monitoring time on pillar no. 19. The
Figure 6. AE signal identified by the transducer.
most relevant seismic events, with the local magnitude value,
occurred during the same period are indicated in the graph.

plot cumulative curves reflecting the count number

as measured continuously throughout the monitoring
period. This method, referred to as Ring-Down Count-
ing, is widely used for defect detection purposes (see
Fig. 6). As a first approximation, in fact, the count
number N , i.e., oscillations per unit time (differential
function) can be compared with the quantity of energy
released during the monitoring process, and the rel-
ative sums (cumulative function) may be assumed to
increase proportionately with the widening of the dam-
aged zone. Needless to say, this assumption applies
only if the damage evolves slowly (Brindley et al.
1973; Pollock 1973; Swindlehurst 1973; Carpinteri &
Lacidogna 2006a).
Figure 7. View (a) and axonometric projection (b) of the From the chart in Figure 8 it can be seen that the pil-
AE sensor applied to the pillar 19. lar is actually undergoing a deterioration process. If we
Table 1. Arrangement of the sensors applied to the pillar 19. examine the chart illustrating the differential function
of AE counts, we can see sudden increases in the oscil-
AE sensors x [mm] y [mm] z [mm] lation peaks occurring at certain intervals over time. It
should be also noted that during the monitoring period,
S1 539.7 0.00 −285 strong seismic actions were recorded in the area, within
S2 679.7 0.00 −455 a radius of ca 50 km around the city of Syracuse. Earth-
S3 434.7 0.00 335 quake data for the period were obtained from website:
S4 394.7 0.00 265 www.ct.ingv.it/Sismologia/GridSism.asp published by
S5 0.00 410 330 the Seismic Data Analysis Group of Catania (Gruppo
S6 0.00 300 −220 di Analisi Dati Sismici – INGV-CT). From the wealth
of data available, we selected the seismic events with a
local magnitude (Ml) greater than 1.2 that had during
reported in Figure 7b. The positions of applied sensors
the monitoring period. These events are illustrated in
are listed in Table 1, and the reference frame for the
Figure 8, where the relative date and magnitudo are
sensors position coordinates is shown in Figure 7b.
also given.
The chart in Figure 8, showing AE monitoring
4 RESULT OF THE MONITORING PROCESS and regional earthquake data, reveals an interest-
ing correlation between the AE activities determined
The monitoring process began at 11:00 a.m. of 19 experimentally and seismic events: the timing of the
September 2006 and ended at 12:20 p.m. of 21 January energy peaks measured by means of the AE differen-
2007. The data collected were analysed in order to tial counts is seen to coincide almost invariably with
interpret the evolution of damage and determine the seismic shocks. This correspondence seems to show
positions of AE sources within the pillar. The AE sig- how the pillar monitored behaves in a pseudo-stable
nal received by the transducers is processed by an manner when subject to the vertical loads alone, but
analyser which counts the oscillations exceeding a has a meagre capacity to respond elastically to hori-
certain voltage threshold. This makes it possible to zontal or oscillatory actions. A behaviour of this kind

378
was observed by the author during an earlier moni-
toring process carried out with the AE technique at in
site which was also subject to seismic activity (Carpin-
teri & Lacidogna 2006b, 2007). Thus, it may be stated
that whilst the AE technique finds it primary appli-
cation in non destructive tests, it is also important to
explore its potential as a monitoring tool in structures
exposed to particular environmental conditions, such
as seismic activities. As already pointed out by the
authors, in fact, the two phenomena – acoustic emis-
sions on a structural scale and seismic activities on a
territorial scale – appear really similar and mutually
correlable.
5 DETECTION OF DAMAGE IN THE
MONITORED PILLAR
5.1 Localisation of AE sources
The first stage of the localisation method consists in
recognising the data needed to identify the AE sources,
followed by the triangulation procedure. During the
first stage, the groups of signals, recorded by the var-
ious sensors, that fall into time intervals compatible
with the formation of microcracks in the volume anal-
ysed, are identified. These time intervals, of the order
of micro-seconds, are defined on the basis of the pre-
sumed speed of transmission of the waves (P) and the
mutual distance of the sensors applied to the surface
of the material. It is usual to assume that the ampli-
tude threshold of 100 µV of the non-amplified signal is
appropriate to distinguish between P-wave and S-wave
arrival times. In fact, P-waves are usually character-
ized by higher value signals. In the second stage, when
the formation of microcraks in a three-dimensional
space is analysed, the triangulation technique can be
applied if signals recorded by at least five sensors fall
into the time intervals. Thus, with this procedure it
is possible to define both the position of the micro-
cracks in the volume and the speed of transmission
of P-waves. The localisation procedure can also be
performed through numerical techniques using opti-
misation methods such as the Least Squares Method
(LSM) (Carpinteri et al. 2006a, 2007a).
In the present work, applying the localisation pro-
cedure more than 50 AE sources have been localised
with an high confidence level. Considering previous
applications of the AE technique carried out by the
authors, the approximation for elements with large
size is about ±10 mm (Carpinteri et al. 2006b). The
localised sources and the cracking pattern for pillar
no.19 are represented in Figure 9. It can be noted that
the localised sources are concentrated near the more
visible crack paths. The localisation of these source
concentration (Fig. 9) and the oscillation counting
(Fig. 8) denounce that the pillar is subject to a damag-
ing phenomenon in slow but progressive evolution.
379
Figure 9. Cracking pattern and localisation of AE sources
for pillar 19.
5.2 Time dependence of damage
The time dependence of the structural damage
observed during the monitoring period, identified by
parameter η, can also be correlated to the rate of prop-
agation of the microcracks. If we express the ratio
between the cumulative number of AE counts recorded
during the monitoring process, N , and the number
obtained at the end of the observation period, Nd , as a
function of time, t, we get the damage time dependence
on AE:
In equation (1), the values of Ed and Nd do not nec-
essarily correspond to critical conditions (Ed ≤ Emax ;
Nd ≤ Nmax ) and the td parameter must be construed as
the time during which the structure has been mon-
itored. By working out the βt exponent from the
data obtained during the observation period, we can
make a prediction as to the structure’s stability con-
ditions. If βt < 1, the damaging process slows down
and the structure evolves towards stability conditions,
in as much as energy dissipation tends to decrease;
if βt > 1 the process diverges and becomes unsta-
ble; if βt ∼
= 1 the process is metastable, i.e., though
it evolves linearly over time, it can reach indifferently
either stability or instability conditions (Carpinteri &
Lacidogna 2006b, 2007; Carpinteri et al. 2007b).
During the observation period, which lasted 121
days for the monitored pillar, the number of AE counts
was ∼ = 4300 (Fig. 8). In order to obtain indications on
the rate of the damage process, as given in Equation 1,
the data obtained with the AE technique were sub-
jected to best-fitting in the bilogarithmic plane. This
yielded a slant βt ∼= 0.98 as shown in Figure 10. The
result confirm that the damage process in the pillar
is in metastable conditions according to a quasi-linear
progression over time.
Figure 10. Evaluation of damage, βt exponent for pillar 19.

6 NUMERICAL SIMULATION

6.1 Laboratory tests and characteristic constants

Beside the experimental research, a numerical simula-
tion based on the Finite Element Method is performed,
which is able to catch the main non linear features of Figure 11. The stone block (a), and the cylindrical speci-
the mechanical behaviour such as fracture nucleation, mens used for the laboratory tests (b).
fracture propagation and time dependency. The main
goal will be to link the dissipated energy, acquired from Table 2. Average experimental values obtained from com-
the acoustic emissions (AE), and the fracture localiza- pressive tests.
tion with the results obtained numerically (Carpinteri
Specimens Peak load Peak stress Elastic mod.
et al. 2008). To this end a number of tests have [mm] [kN] (σu ) [Mpa] (E) [Mpa]
been conducted in order to determine the mechanical
properties of the Syracuse Cathedral stone. 1_a–c (Ø = 30) 7.12 10.7 1500
The ancient stone used in the construction of the 2_a–c (Ø = 60) 22.10 7.82 2015
temple was located in the area of Plemmirio where 3_a–c (Ø = 120) 96.59 8.54 2200
archaeological studies have found the presence of
various quarries of the Greek period. During the
restoration works recently carried out the technique of
while ten-node, three-side isoparametric solid pyramid
replacing stone blocks was employed for several dam-
elements have been adopted for the more compli-
aged elements. The removed elements were replaced
cated ones. Interfaces between blocks have been mod-
by other blocks consisting of the same stone. In par-
elled with triangular twelve-node quadratic interface
ticular a portion of one of the removed elements has
elements.
been used to define, through laboratory tests, the
The resulting model comprises 5081 elements and
mechanical properties of the material that since 2400
9188 nodes. Figure 12 shows the mesh of the pillar, and
years constitutes the bearing structure of the Cathe-
the details of the masonry inserts acquired by the NDT
dral (Fig. 11). Some cylindrical specimens having
analysis. The displacements at the base of the pillar
slenderness λ = 1, and diameter equal to 30, 60 and
were fixed, while the top of the pillar was elastically
120 mm respectively, were obtained from the stone
constraint, in order to account for the interaction with
block (Fig. 11a, b). Three specimens (a,b,c) for each
the surrounding structure of the Cathedral. Two main
diameter have been tested at the Fracture Mechanics
load were considered: the dead load (of the pillar and of
Laboratory of Politecnico di Torino, in this way the
the surrounding structure), and an horizontal seismic
ultimate peak stress (σu ), and the elastic modulus (E)
load provided as an horizontal ground acceleration.
are computed (see Table. 2).
Figure 13a shows a detail of the interface ele-
ments between each block of the pillar, while Fig-
6.2 The numerical model
ure 13b shows the deformed mesh under the effect
A detailed 3D model of the most damaged pillar has of the dead load. The elastic modulus of the stone
been obtained according to the geometrical survey.The block, as deduced from the experimental test, was
geometry of each block, as well as the presence of 2000 MPa, while the Poisson ratio was equal to 0.12.
masonry inserts have been considered. Fifteen-node The mean compressive strength was equal to 9 MPa.
isoparametric solid wedge elements have been used For the present sandstone, an average tensile strength
to discretize the more regular zones of the model, of 3 MPa, and fracture energy of 50 Nm were adopted.

380
Figure 12. Finite element mesh of the pillar (a). Details of
the masonry inserts (b).
Figure 13. Detail of the interface finite element mesh
between blocks (a). Dead load deformed mesh (b).
It is worth noting that, under the effect of the sole
gravity load, the maximum compressive stress do not
exceed the value of 2 MPa, which is low compared
with the strength of the sandstone. On the other hand,
due to the heterogeneity of the pillar, tensile stresses
may arise that approach the tensile strength.
381
Figure 14. Principal compressive stress (a). Deformed mesh
under the seismic acceleration (b).
Figure 15. Cracking in the central zone of the pillar when
subjected to the seismic load.
A seismic load has also been considered. If a
return period of 140 years takes place, an horizontal
acceleration of 1.288 m/s2 is applied (30% probability
occurrence in a period of 50 years). In this case, the
compressive and tensile stresses increases, as shown
in Figure 14a. The pillar deforms mainly according to
his first modal shape, as shown in Figure 14b.
Due to the horizontal acceleration, cracking can
take place in the pillar. A detail of crack nucleation
is shown in Figure 15. The crack occurrence pro-
vided by the analysis agrees quite well with the crack
localization provided by the AE recording. Cracking
correspond both to diffuse cracking in the contin-
uum elements in the sandstone blocks and opening
or sliding of the discrete interface between blocks.
7 CONCLUSIONS
The evolution of damage in a pillar made of calcareous
stone blocks that is part of the vertical bearing structure
of the Syracuse Cathedral was evaluated using the AE
technique. The data collected were analysed in order
to interpret the evolution of damage and to determine
the positions of AE sources within the pillar. From
the charts plotted for the differential and cumulative
functions of the AE signal counts it can be seen that
the pillar is actually undergoing a damage process.
Moreover, by applying the AE source localisation pro-
cedure it was possible to identify ca. 50 emission points
within the pillar. Within the stone blocks to which the
sensors had been applied, the points were seen to con-
centrate along the cracks that could be discerned more
clearly on the surface. The identification of these emis-
sion sources together with the oscillation counts shows
that the pillar is indubitably undergoing a slow but
incessant damage process.
Beside the experimental research, using a numer-
ical simulation based on the Finite Element Method,
the most damaged pillar has been modelled in details,
providing an accurate description of the blocks inter-
action by means of discrete interface elements. The
numerical analysis has been performed considering
the structure subjected to seismic horizontal loads. By
this model the crack nucleation is found in the same
positions identified through the AE monitoring. In this
way the numerical model confirms the results obtained
experimentally.
ACKNOWLEDGEMENTS
The financial support for AE application provided by
the European Union (EU) Leonardo da Vinci Pro-
gramme, ILTOF Project is gratefully acknowledged.
Special thanks for their collaboration go to the Arch.
M. Muti, to the Eng. R. Meloni and to the Arch. L.
Regalbuto of the Soprintendenza of Syracuse. The
Soprintendenza of Siracusa supported the all the other
NDT research.
REFERENCES
Agnello, G. 1996. Il Duomo di Siracusa e i Suoi Restauri, a
cura di S. L. Agnello, Ediprint, Siracusa (in Italian).
Agnello, S.L. 1950. La Rinascita Edilizia a Siracusa Dopo il
Terremoto del 1693, Archivio Storico Siciliano, serie III,
vol. IV, 1950–51, Palermo (in Italian).
382
Binda, L., Cantini, L., Condoleo, P., Saisi, A. & Zanzi, L.
2006. Investigation on the Pillars of the Syracuse
Cathedral in Sicily, 3-Day Int. Conf. Structural Faults &
Repair, M.C. Forde (Ed.), Engineering Technics Press,
Edinburgh, CD-ROM: 1–12.
Binda, L., Casolo, S., Petrini, V., Saisi, A., Sanjust, C.A. &
Zanzi, L. 2007. Evaluation of the Seismic Vulnerability
of the Syracuse Cathedral: Investigation and Modelling,
Proc. of Int. Symposium Studies on Historical Heritage,
Yildiz Technical University, Istambul, Turkey, 683–690.
Binda, L., Saisi, A. & Tiraboschi, C. 2000. Investigation
Procedures for the Diagnosis of Historic Masonries, Con-
struction and Building Materials, ed. Elsevier-Norwich
14 (4): 199–233.
Binda, L., Lualdi, M., Saisi, A. & Zanzi, L. 2003. The Com-
plementary Use of on Site Non Destructive Tests for
the Investigation of Historic Masonry Structures, Proc.
9th North American Masonry Conference, (9NAMC),
978–989.
Brindley, B.J., Holt, J. & Palmer, I.G. 1973. Acoustic
Emission-3: The Use of Ring-Down Counting, Non-
Destructive Testing 6: 299–306.
Carpinteri, A., Invernizzi, S. & Lacidogna, G. 2008 Crack-
ing simulation of brick-masonry subjected to the double
flat-jack test. Proc. of Structural Analysis of Historical
Constructions, Bath 2–4 July 2008.
Carpinteri,A. & Lacidogna, G. 2006a. Damage Monitoring of
an Historical Masonry Building by the Acoustic Emission
Technique, Materials and Structures 39: 161–167.
Carpinteri, A. & Lacidogna, G. 2006b. Structural Monitoring
and Integrity Assessment of Medieval Towers, Journal of
Structural Engineering (ASCE) 132: 1681–1690.
Carpinteri, A. & Lacidogna, G. 2007. Damage Evalua-
tion of Three Masonry Towers by Acoustic Emission,
Engineering Structures 29: 1569–1579.
Carpinteri, A., Lacidogna, G. & Manuello, A. 2007a. Dam-
age Mechanisms Interpreted by Acoustic Emission Signal
Analysis, Key Engineering Materials 347: 577–582.
Carpinteri, A., Lacidogna, G. & Niccolini, G. 2006a. Critical
Behavior in Concrete Structures and Damage Localization
by Acoustic Emission, Key Engineering Materials 312:
305–310.
Carpinteri, A., Lacidogna, G. & Paggi, M. 2006b. Acoustic
Emission Monitoring and Numerical Modelling of FRP
Delamination in RC Beams With Non-Rectangular Cross-
Section, Materials and Structures (RILEM) 40: 553–566.
Carpinteri, A., Lacidogna, G. & Pugno, N. 2007b. Struc-
tural Damage Diagnosis and Life-Time Assessment by
Acoustic Emission Monitoring, Engineering Fracture
Mechanics 74: 273–289.
Pollock, A.A. 1973. Acoustic Emission-2: Acoustic Emission
Amplitudes, Non-Destructive Testing 6: 264–269.
Privitera, S. 1863. Illustrazioni sull’AnticoTempio di Minerva
Oggi il Duomo di Siracusa, Memoria del Parroco Ser-
afino Privitera, Tipografia La Fenice Musumeci Catania
(in Italian).
Russo, S. 1991. La Cattedrale di Siracusa, Archivio Storico
Siracusano, III V (in Italian).
Russo, S. 1992, Siracusa Medievale e Moderna, Arnaldo
Lombardi, Palermo (in Italian).
Swindlehurst, W. 1973. Acoustic Emission-1: Introduction,
Non-Destructive Testing 6: 152–158.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Shaking table tests on the large scale model of Mustafa Pasha Mosque
without and with FRP

L. Krstevska, Lj. Tashkov & K. Gramatikov

University “St. Cyril and Methodius”, Skopje, Republic of Macedonia

R. Landolfo, O. Mammana, F. Portioli & F.M. Mazzolani

University of Naples “Federico II”, Naples, Italy

ABSTRACT: Shaking table tests were carried out on a large scale model of Mustafa Pasha Mosque in Skopje
within the activities of Sixth Framework Program PROHITECH – “Earthquake protection of historical buildings
by reversible mixed technologies”. The main aim of experimental investigation was to evaluate the seismic
stability of the monument after applying a reversible technology for strengthening. Beside the experimental
shaking table testing of the model, three-dimensional finite element analyses were carried out using the general
purpose software package ANSYS. The numerical models were calibrated on the basis of mechanical properties
of masonry obtained from compression and shear tests on wall specimens built with same technique used for
the original prototype. The comparison between experimental and numerical results allowed the reliability of the
implemented models to be evaluated. Finally the effectiveness of strengthening and the evolution of crack patterns
observed during each phase of the testing programme has been analyzed on the basis of the obtained results.

1 INTRODUCTION strengthening schemes based on the calculated tensile

Shaking table testing of the large scale model of the
Mustafa Pasha mosque in Skopje, (15th century) was
performed in the Laboratory of the Institute of Earth-
quake Engineering and Engineering Seismology in
Skopje in the period November-December 2006. The
main objective of the tests was to investigate experi-
mentally the effectiveness of the proposed reversible
technology for strengthening of this type of histori-
cal monuments. The performed testing was a part of
the activities within the Sixth Framework Programme
PROHITECH – “Earthquake Protection of Historical
Buildings by Reversible Mixed Technologies”.
In order to predict the response of the large scale
model during the shaking table tests performed at the
IZIIS Laboratory in Skopje, a finite element model
was implemented.
The main aim of numerical investigations was to
evaluate the peak ground accelerations to be assigned
to the shaking table during the test phases of exper-
imental programme that were devoted to the assess-
ment of the seismic strength of both the original and
strengthened structure. In particular, it was important
to control the damage on the original Minaret and
on the Mosque in order to prevent collapse before
the application of FRP strengthening. Moreover, the
numerical models were used to determine suitable FRP
stress distributions.
To assess the seismic capacity of the structure
and the effectiveness of FRP reinforcement, nonlin-
ear pushover analyses were carried out by means of
the implemented finite element model. In the follow-
ing, a detailed description of the implemented models
and of the corresponding obtained results is also
reported.
2 EXPERIMENTAL INVESTIGATION
The model of the mosque was constructed to a length
scale of 1:6 according to the “gravity forces neglected”
modeling principle, using the same materials as those
of the prototype: stone, bricks and lime mortar (Gra-
matikov et al. 2007). The experimental testing was
performed on the bi-axial seismic shaking table at the
IZIIS’ Laboratory. Following the main objective, the
testing was performed in three phases:
1) Testing of the original model for low intensity level,
to provoke small damage;
2) Testing of repaired model and strengthened minaret
with FRP, until total collapse of the minaret;
3) Testing of strengthened model with FRP and carbon
fiber bars until collapse.

383
 Figure 2. Damage to the minaret (horizontal crack) and of
the mosque after phase 1.

Figure 1. Dimensions and instrumentation of the large scale

model of Mustafa Pasha Mosque.

To investigate models’ response during seismic

action different types of transducers were placed at
characteristic points – accelerometers, linear poten-
tiometers and LVDTs, as shown in Figure 1 (Krstevska
et al. 2007).

2.1 Phase 1 – Testing of the original model

Before starting the seismic testing, the dynamic char-
acteristics of the model were defined by means of
ambient vibration method as well as by low inten-
sity random excitation within the frequency range
of 0.1–50 Hz. The accelerogram of the Montenegro
earthquake of 1979 (Petrovac N-S component) was
selected as a representative earthquake excitation.
The earthquake time histories were scaled (com-
pressed) 6 times, according to the similitude require-
ments and several tests were performed with intensity
of 0.01–0.10 g.
Under input intensity of 2% g, the first horizontal
crack appeared at the base of the minaret. In the next
tests performed under intensities of up to 10% g, dam-
age occurred on the mosque, too. The reason of this
was the frequency content of the applied excitation,
which was close to the natural frequencies of both the
minaret and the mosque.
Figure 3. Characteristic response time histories at the top
of the model for input intensity 10%g, original model.
The damaged model is presented in Figure 2, while
some representative response parameters for input
intensity of 10% g are given in Figure 3.
2.2 Strengthening of the minaret
After performing the seismic tests on the original
model of the mosque, the damaged model was repaired
by crack injection and the minaret was strengthened
by application of CFRP wrap, cut to have the corre-
sponding width and applied upon a layer of epoxy glue
(Fig. 4). The obtained strips with a width of 15 cm were

384
 Figure 5. Collapse of the upper part of the minaret.

Figure 4. Repaired model and strengthened minaret after

phase 1 testing.

placed on four sides along the length of the minaret in

vertical direction for the purpose of its stiffening.
To confine the structure, strips with a width of
10 cm were placed at four levels along the height of
the minaret in horizontal direction, while a strip with
a width of 20 cm was placed at its base.

2.3 Phase 2 – Testing of repaired model and

strengthened minaret
Before the seismic tests, the dominant frequencies of
the model were checked by random excitation.
For the minaret dominant frequency was 4.7 Hz,
while for the mosque two frequencies were dominat-
ing: f = 7.4 Hz and f = 9.6 Hz.
During this testing phase, 11 tests were performed
with input acceleration between 0.2 g to 1.5 g. The first
cracks on the minaret occurred under input intensity of
0.34 g, which indicated that the applied strengthening
enabled stiffening of the minaret and increasing of its
bending resistance. The initial cracks in the mosque
appeared under input intensity 0.42 g.
During the next tests, the cracks developed and at
input level of 0.49 g, the upper part of the minaret
totally collapsed (Fig. 5).
After reaching the max input acceleration of 1.5 g,
the mosque model was heavily damaged and the testing
was stopped.
There were cracks in the walls and also in the dome
(Fig. 6).
Figure 6. Damaged model after phase 2 testing.
The characteristic response parameters during the
seismic tests are given in Table 1, while the represen-
tative time histories for the final test with an input
intensity of 1.5g are given in Figure 7.
2.4 Strengthening of the model
After the final test of the model in phase 2 the
minaret was removed and the mosque rebuild on
damaged parts. Then the model was strengthened.
The main adopted principle in strengthening of the
model was that the methodology to be applied be
reversible and invisible. Hence, the cracks in the dam-
aged model were not repaired by injection, but a
concept was adopted that the model be strengthened
to the conditions after the preceding tests.

385
Table 1. Performed seismic tests, phase 2.

Inp. Top Top Input Top

acc acc. acc. displ. displ.
Test No (g) CH1 CH3 (mm) (dome)

1 0.2 0.65 0.4 2.24 2.3

2 0.28 0.82 0.6 3.3 3.3
3 0.34 1.7 0.7 4.0 4.0
4 0.42 – 1.0 4.7 5.0
5 0.49 – 0.88 5.48 6.4
6 0.53 – 0.9 6.1 7.5
7 0.58 – 1.0 7.0 8.9
8 0.65 – 0.96 9.1 12.4
9 1.05 – 1.1 11.4 14.7
10 1.4 – 0.7 13.9 18.8
11 1.5 – 0.54 15.3 22
Figure 8. The strengthened model ready for testing, phase 3.

course was formed whereby the tensile resistance of

the wall was improved and synchronous behaviour
of the bearing walls was achieved.
• Formation of a horizontal belt course around the
tambour by applying a CFRP wrap with a width of
10 cm.
• Formation of a horizontal belt course at the base of
the dome by use of a CFRP wrap with a width of
50 cm.
The formation of these horizontal belt courses
enabled better integrity of the tambour and the dome
base and prevented “opening” of the dome which
was the most common reason for occurrence and
prolongation of cracks in the bearing walls.
The strengthened model is presented on Figure 8.

2.5 Phase 3 – Testing of strengthened mosque

Figure 7. Response during the test with intensity 1.5 g,
phase 2.
The strengthening consisted of incorporation of
horizontal belt courses for the purpose of increasing
the integrity of the structure at those levels and pro-
viding as better as possible synchronous behaviour of
the bearing walls:
• Incorporation of carbon rods in two longitudinal
mortar joints around the four walls at two levels: the
level above the openings and at the top of the bear-
ing walls, immediately below the tambour. With the
incorporation of these carbon rods, a horizontal belt
model
Before the seismic tests, the dominant frequencies of
the model were checked.
The resonant frequency of 9.2 Hz was obtained and
it was compared to the frequency of 8.6 Hz mea-
sured after the last test of testing in phase 2. This
pointed out that strengthening had increased the res-
onant frequency of the model for about 8%, meaning
that the stiffness of the model hadn’t been recovered
completely compared to the initial state.
During this phase, 24 tests were performed
under input acceleration between 0.15 g to 1.5 g.
The accelerogram of the Petrovac earthquake (N-S
component) was scaled by 6 (compressed) in the first
15 tests. During the tests performed under input inten-
sities between 0.15 g and 0.40 g. the model’s behaviour
was stable, without provocation of large cracks. In the
next 6 tests carried out under input acceleration of
0.60–0.80 g, the dome experienced sliding at a visible
horizontal crack at its base. To provoke more intensive
response of the model, in the next tests, a time scaling

386
 Table 2. Performed seismic tests, phase 3.

Input
Scal. Input acc Top acc. displac. Top displ.
factor (g) (dome) (mm) (dome)

6 0.14 0.30 1.5 1.5

0.18 0.35 2.0 2.0
0.25 0.42 3.0 3.0
0.29 0.50 3.5 3.5
0.35 0.59 4.0 4.3
0.38 0.65 4.7 5.0
0.40 0.75 – –
0.42 0.85 6.8 7.7
0.67 1.60 7.2 8.5
0.87 1.70 10.6 14.0
0.82 1.55 10.0 14.4
Figure 9. Damage of the model after phase 3 tests accom- 0.90 1.70 11.0 16.0
plishment. 0.80 1.3 12.6 16.7
3 0.20 0.60 7.8 13.0
factor of 3 was used, producing an input acceleration 0.46 0.93 15.0 22.0
of 0.46–1.5 g. In this series of tests, many new cracks 1.20 1.10 25.0 26.8
1.5 1.0 30.0 40.0
appeared in the walls as well as in the dome, decreas- 2 0.15 0.40 8.4 12.0
ing the dominant frequency of the model to f = 4.4 Hz. 0.75 0.70 27.0 35.0
This value was more than twice lower compared to the 1.00 0.80 45.0 52.0
initially measured frequency of 9.2 Hz, thus indicating 1 0.35 0.53 58.0 75.0
a pre-collapse state of the model.
The next two tests were performed by a scaling
factor of 2, under an input acceleration 0.75–1.0 g. analyses were performed for the evaluation of struc-
Progressive cracks appeared, but still without collapse. tural capacity, both for original and reinforced model
The final test was performed by a scaling factor of (Landolfo et al. 2007).
1, under an input acceleration of 0.35 g. Heavy damage The numerical model was generated by importing
occurred in many places on the dome and around the in the FE program a three dimensional solid model of
openings as well as big cracks and inclination of one the Mosque that was implemented in a computer aided
corner of the model, including damage to the strength- design system.
ening FRP belt at the lower level of one of the walls, In order to properly evaluate the structural inter-
after which the testing was stopped. actions among the different parts, the implemented
The damaged model is presented in Fig. 9. The char- geometrical model reproduces all the main elements
acteristic response parameters during the seismic tests of the building accurately, including the openings and
are given in Table 2. The representative time histories the pendentives connecting the walls with the dome.
in the final tests are given in Fig. 10. Two separate FE models were implemented for the
As a common conclusion after the performed exper- Minaret and the Mosque, respectively (Fig. 11). In the
imental testing of the mosque model in phase 3, it last case, the symmetry of the model along the vertical
might be said that the model’s behaviour was evi- plane parallel to the direction of the input displacement
dently different in respect to that of the original model. was considered, in order to save CPU time for solving
Under tests of moderate intensity, the existing cracks non-linear equations.
were activated but during the subsequent more inten- To model the reinforcement, the outer surfaces of
sive tests, the failure mechanism was transferred to the the Mosque and Minaret were divided into different
lower zone of the bearing walls, in the direction of the areas corresponding both to the FRP sheets and bars.
excitation, where typical diagonal cracks occurred due The properties of overlapping areas were set up in
to shear stress. order to match the meshes corresponding to masonry
and FRP reinforcement elements. The whole masonry
structure was discretized with tetrahedral 3D solid ele-
3 FE ANALYSIS
ments considering a mesh size of 100.0 mm in the case
of Mosque model.
3.1 Types of analysis, geometric modeling and
SOLID45 elements were used for masonry and
meshing
SHELL181 elements were employed to model the FRP
The finite element analysis of the large scale model sheets. The shells corresponding to the FRP reinforce-
was carried out using the ANSYS software. Push-over ment were directly overlapped to the masonry bricks

387

Figure 10. Response of the model during the last test with
intensity 0.35 g, scaling factor 1, phase 3.
Figure 11. The FE model of the Mosque and Minaret
implemented for non-linear pushover analysis.
388
Table 3. Masonry material properties assumed in numerical
simulations of Minaret and Mosque.
γ E c ϕ δ
[kN/m3 ] [MPa] ν [MPa] [◦ ] [◦ ]
Minaret 19.00 16000 0.25 0.13 60.57 60.57
Mosque 19.00 680 0.25 0.09 69.13 69.13
and no interface elements were considered. The parts
of the Mosque model reinforced with bars were mod-
elled with elastic solid elements through the whole
thickness of the shear walls with Young’s modulus
equal to that of FRP material.
3.2 Material modeling
The material properties used for the two finite element
models are reported in Table 3. In particular, the elastic
parameters are referred to the values that have been
calibrated on the basis of first random vibration tests
performed on the original undamaged structure.
The strength properties were determined on the
basis of both compression and shear experimental tests
carried out on masonry wall samples.
As far as structural masonry, the elastic-perfectly
plastic Drucker-Prager material model was considered.
In the case of Mosque, cohesion c and angle of inter-
nal friction ϕ were calibrated in order to obtain values
of 0.05 MPa and 1.0 MPa for tensile (ft ) and compres-
sive strengths (fc ). For Minaret, the considered values
for ft and fc were 0.1 and 1.0 MPa. In order to assign
an associative flow rule for plastic strains, the assumed
dilatancy angle δ was equal to φ. With regard to com-
posites, an elastic material model was considered. In
particular, the adopted Young’s modulus for sheets is
equal to 240 GPa and the considered equivalent thick-
ness is 1.0 mm, according to the nominal mechanical
properties provided by the manufacturer.
3.3 Load modeling and boundary conditions
With regard to the load modelling approach imple-
mented in pushover analyses for the evaluation of
seismic capacity, a uniform and a linear acceleration
distribution along the horizontal direction was applied
to the FE models of Mosque and Minaret, respectively.
As far as the boundary conditions are concerned, full
restraints were assumed at the base of the structure in
the performed analyses.
3.4 Results of the non-linear pushover numerical
analyses
The assessment of the ultimate seismic strength of
both the original and reinforced large scale model was
Figure 12. Distribution of first principal plastic strains on Figure 13. Distribution of first principal plastic strains on
the original large scale model of the Minaret at collapse load. the reinfroced large scale model of the Minaret at collapse
load.

carried out by means of nonlinear pushover analyses

performed on the implemented FE model.
On the basis of obtained numerical results, the evo-
lution of damage distribution till collapse load on
the investigated prototype has to be ascribed to the
attainment of tensile or shear strength, while the com-
pressive resistance is never exceeded. The collapse
loads corresponding to the original and reinforced pro-
totype were determined by checking the attainment of
the maximum plastic strain calculated for masonry on
the basis of the FE models calibrated against shear
tests.
With regard to the Minaret, the results of numeri-
cal analysis show that the reinforcement increases the
ultimate strength in terms of peak acceleration at the
top from the value of 0.3 g corresponding to the orig-
inal model to 1.2 g and the collapse mechanism shifts
to the upper part without strengthening (Figs 12, 13). Figure 14. Distribution of first principal plastic strains on
As far as the Mosque is concerned and accord- the original large scale model of the Mosque at collapse load.
ing to the implemented numerical model, the original
prototype collapses with a mixed pier/spandrel mech- out plane horizontal loads. In the second step, the dam-
anism of the vertical bearing structures, that is typical age extends to the upper spandrels, between the second
of weakly coupled perforated walls (Fig 14). On the and third row of opening in the shear walls parallel to
basis of the numerical results, the evolution of collapse the direction of seismic loads. In particular, it develops
mechanism can be divided in different phases. Accord- up to the dome, among the openings in the support-
ing to the numerical model, the first diagonal tension ing polygonal drum and the ones in the shear walls.
cracks occur in the shear walls, namely in the spandrels Finally, the seismic strength of the structure is attained
between the first and second row of openings from the when the central wall at the base of the Mosque and
basement. In this phase, the damage also develops at the lateral piers collapse for shear and bending mecha-
the base of the walls perpendicular to the direction of nisms, respectively. The study of collapse mechanism,
ground motion, owing to bending stresses induced by obtained from numerical analysis on both original and

389
 1.8

1.6
original minaret strengthened minaret
1.4 Exp. Exp.
FE FE
1.2

Top Acc. [g]

1

0.8

0.6

0.4

0.2

0
0 1 2 3 4 5 6 7 8
Relative Displ. [mm]

Figure 15. Distribution of first principal plastic strains on Figure 16. Comparison between experimental and numer-
the reinforced large scale model of the Mosque at collapse ical response for the original and strengthened Minaret
load. model.

1.8
reinforced Mosque, allowed the effectiveness of FRP
reinforcement to be analyzed. The wraps around the 1.6
dome and the top of shear walls fully prevent the prop- 1.4
agation of cracks from the bottom part to the drum, as
shown in Figure 15. 1.2
Top Acc. [g]

The role played by FRP bars in the shear walls is to 1

stiffen and strengthen the spandrels forming a sort of
0.8
reinforced masonry beams at different levels able to
distribute the seismic action among the piers. 0.6
According to the numerical model of the reinforced
0.4 original mosque strengthened mosque
Mosque, in this case the collapse mechanism turns Exp. Exp.
from a mixed into a weak piers/strong spandrel type. 0.2 FE FE
0
0 1 2 3 4 5 6 7
Relative Displ. [mm]
4 COMPARISON BETWEEN EXPERIMENTAL
AND NUMERICAL RESULTS Figure 17. Comparison between experimental and numer-
ical response for the original and strengthened Mosque
The results of FE model were compared with experi- model.
mental tests. The distribution of first principal plastic
strains predicted by the numerical model fits well the
crack patterns observed on the large scale model of The comparison between experimental and numer-
the Mosque during shaking table tests. As shown in ical responses in terms of acceleration and relative
Figures 6 and 9, in the case of the original Mosque displacement at the top is shown in Figures 16 and 17.
cracks formed on the spandrels between the open- Also in this case, the numerical and experimental
ings up to the tambour, while in the reinforced model results are in good agreement.
cracks on the bearing walls were observed at the base In the case of Minaret, it should be noted that the
of the structure, according to the predicted damage experimental value of 0.2 g on the original model cor-
pattern. responds to first observed cracks and to the attainment
With this regard, it should be noted, however, that of first plastic strains in the numerical model.
sliding of the dome at tambour opening level has been However, during the testing the original Minaret did
observed before the collapse of the shear walls. not collapse for values of peak acceleration at the top
This collapse mechanisms is not predicted by up to 0.7 g and the structure exhibited a rigid block
numerical results and can be ascribed to a different behaviour with rocking.
quality of masonry at the base of the dome from other Also in the case of the strengthened Minaret the
parts of the Mosque. experimental peak accelerations at the top are higher

390
than numerical ones. Considering that a non-linear
static FE analysis was performed, this discrepancy can
be ascribed to cyclic loading induced by seismic exci-
tations and to the rigid block-type behaviour of the
Minaret observed during testing.
5 CONCLUSIONS
In this study, a numerical and experimental investiga-
tion on the seismic strength of the masonry large scale
model of the Mustafa Pasha Mosque reinforced with
FRP has been presented.
The experimental testing of Mustafa pasha large
scale model was performed to investigate the seismic
stability of the monument after applying a reversible
strengthening methodology, both for the minaret and
for the mosque.
The strengthening applied to the minaret enabled
stiffening and increasing of its bending resistance.
The mosque model’s behaviour after strengthening
was evidently different in respect to that of the original
model. Under tests of moderate intensity, the existing
cracks were activated but during the subsequent more
intensive tests, the failure mechanism was transferred
to the lower zone of the bearing walls, in the direc-
tion of the excitation, where typical diagonal cracks
occurred due to shear stress.
In order to support the experimental test set-up, to
design the FRP reinforcements and to analyze their
effects on the prototype, a finite element analysis was
performed.
In particular, the implemented finite element model
allowed the evolution of partial and global collapse
mechanisms observed during the tests to be analyzed.
The results of numerical investigation have been com-
pared with the experimental outcome. Generally, a
good agreement between the behaviour predicted by
numerical models and test results was observed, both
391
in terms of ultimate seismic capacity and collapse
mechanisms. Further studies are in progress, mainly
devoted to refine the calibration of FE model by imple-
menting non linear time history analysis with other
types of material models for masonry, such as smeared
crack. The calibrated non-linear numerical models will
represent a reliable tool both for the design optimiza-
tion of FRPs and for the evaluation of the effectiveness
of other types of strengthening on the original Mosque.
ACKNOWLEDGMENTS
It is significantly acknowledged the financial support
of the European Commission (grant No. INCO-CT-
2002-509119), for funding the research project PRO-
HITECH (Earthquake PROtection of HIstorical Build-
ings by Reversible Mixed TECHnologies), which is the
main framework of the experimental and numerical
activity presented in this paper.
REFERENCES
Gramatikov, K., Taskov, Lj., Krstevska, L. 2007. Final
report on “Shaking table testing of Mustafa-Pasha
mosque model”. FP6-PROHITECH Project, Workpack-
age 7-Experimental analysis. University “Sts.Cyril and
Methodius”, Skopje, Macedonia.
Krstevska, L., Taskov, Lj., Gramatikov, K., Landolfo, R.,
Mammana, O., Portioli, F., Mazzolani, F.M. 2007. Exper-
imental and numerical investigations on the Mustafa
Pasha Mosque large scale model. In Proceedings of Cost
Action C26 Workshop “Urban habitat constructions under
catastrophic events”, Prague 30–31.3.2007, 158–169.
Landolfo, R., Portioli, F., Mammana, O., Mazzolani, F.M.
2007. Finite element and limit analysis of the large scale
model of Mustafa Pasha Mosque in Skopje strengthened
with FRP. In Proceedings of the First Asia-Pacific Con-
ference on FRP in Structures (APFIS 2007). Hong Kong,
China, 12–14 Dec. 2007.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Integrated modeling method for dynamic behavior of ancient pagodas

Jianli Yuan
Yangzhou Univ., Yangzhou, China

Ling Yao & Shengcai Li

Department of Civil Engineering, Yangzhou Univ., Yangzhou, China

Donato Abruzzese
Department of Civil Engineering, Univ. of Rome, Tor Vergata, Rome, Italy

ABSTRACT: Dynamic behavior model is essential to the reliability evaluation and restoration scheme of
ancient pagodas. In this paper, the identification techniques and the main influence factors on the dynamic
behavior of ancient pagodas are discussed, and the modeling method integrated the predominance of parameter
forecast, sensitivity analysis, and model updating criteria is developed. The Huqiu Pagoda in Suzhou City,
a famous leaning pavilion-style masonry pagoda, was selected as a research case to present the application
characteristics of the method. A 3D finite element model of this pagoda has been constructed, and the main
structural parameters were updated according to the model updating criteria to match with the measured dynamic
characteristics of ambient vibration test to ensure the validity of the model.

Keywords: ancient pagoda; dynamic behavior; integrated modeling method; environmental random excitation;
finite element analysis; sensitivity system; model updating criteria.

1 INTRODUCTION 2 PROCESS OF THE INTEGRATED

MODELING METHOD
Chinese pagodas are outstanding representation of
ancient high-rise buildings. There are more than 3,000 The process of integrated modeling method for
ancient pagodas in China, and the most of them need to dynamic behavior of ancient pagodas is as follows
be maintained and repaired. Modeling of the dynamic (Figure 1): (1) Survey the structure, define structural
behavior is an important step in the structural relia- parameters and their variation degree, (2) Construct
bility assessment and intervention scheme of ancient
pagodas. Dynamic behavior model of a structure can
be obtained by computer analysis based on the dynam- 1
ics theory or by testing the structure on site, the key
on the model is the analysis model is equivalent to the
physical system. 2
The development of modern test and numeri-
cal modeling technology provide new approach for 5 3
construction of dynamic behavior model of ancient
pagodas. Combined with the application of ambi-
4
ent vibration technology and finite element pro-
gram ANSYS, this paper discusses the characteristics 8 6
of the integrated method and correlative influence
factors on modeling dynamic behavior of ancient N. Y.
pagodas. A study case, Huqiu Pagoda in Suzhou 7 9
City, was selected to introduce the essentials of
application. Figure 1. Flow chart of the integrated modeling method.

393
the preliminary analysis model by ANSYS, (3) Esti- Table 1. Density of main materials.
mate the dynamic eigenvalues of the structure for the
selection of proper instrument parameters, (4) Acquire Material Density
the modal parameters by ambient vibration test,
(5) Establish sensitivity system and optimize structural Brick masonry 18 (kN/m3 )
parameters, (6) Select model updating criteria, update Granite masonry 26.4 (kN/m3 )
Limestone masonry 25.6 (kN/m3 )
the analysis model, (7) Evaluate the error between
Sandstone masonry 22.4 (kN/m3 )
the analysis value and test value by Errors Evaluation China fir 4∼5 (kN/m3 )
Program, (8) Select adjustment parameters for new Pine wood 5∼6 (kN/m3 )
updating, (9) Obtain the perfect model. Black tile roof 0.9∼1.1 (kN/m2 )

3 THE MODELING TECHNIQUES AND

CORRELATIVE INFLUENCE FACTORS Table 2. Fundamental period of pavilion style masonry
pagoda.
3.1 Major factors influencing dynamic behavior
of ancient pagodas T (s)
H D d Doors/
Ancient pagodas were built long time ago, often suf- Pagoda (m) (m) (m) storey Eq.2 Test
fered various natural disasters and damages, and most
of them had been repaired. Therefore most pago- Dayan∗ 49.8 25.2 9.2 2 0.65 0.67
das have following common characteristics: complex Xiaoyan∗ 40.2 11.4 3.6 2 0.84 0.74
structure and construction type, non-uniform mate- Fawang∗ 27.8 8.4 2.7 2 0.55 0.59
rials, as well as the coexistence of various damage Liusheng∗∗ 36.1 14.3 2.0 4 0.56 0.58
conditions (Yuan Jianli, Shengcai Li, 2001). Since Zhengguo∗∗ 44.2 14.0 2.2 4 0.88 0.85
the structural parameters affect directly the dynamic Renshou∗∗ 40.2 13.8 2.2 4 0.73 0.80
Huqiu∗∗∗ 47.5 13.7 1.9 4 0.84 0.83
behavior of ancient pagodas, the dimension, construc- Wenfeng∗∗∗ 38.3 10.3 2.4 4 0.82 0.88
tion detail, damage condition and variation degree Xingfusi∗∗∗ 52.1 9.0 1.9 4 1.68 1.61
should be surveyed carefully and defined reasonably
for a good analytical model to be developed. Note: ∗ : Brick pagoda in Shanxi province; ∗∗ : Stone pagoda
According to the principle of dynamics, the motion in Fujian province; ∗∗∗ : Brick pagoda in Jiangsu province.
equation of an ancient pagoda can be expressed as
follows:
of the material strength test and refer to the specifi-
cation of code for design of masonry structures (GB
50003-2001, 2002).
where {x}, {ẋ}, {ẍ} are the structural nodal displace-
ment, velocity and acceleration vectors, respectively;
[K], [C], [M ] are, respectively, structural stiffness, 3.2 The fundamental period of ancient masonry
damping, and mass matrices; ẍo is the ground acceler- pagodas
ation. Taking [C] = α[M ] + β[K], if Rayleigh damp-
ing is adopted, the mass M and stiffness K are the main The fundamental period is a basic index of the dynamic
factors to influence dynamic behavior of the structure. behavior of pagodas. It is useful for selection of the
The mass of an ancient pagoda can be obtained from appropriate test parameter to estimate the fundamen-
the measurement of the geometrical dimensions and tal period before field test. Based on the comparison
material density of each component. The density of of theoretical value with experimental value of nine
main materials in masonry pagodas is list in Table 1. pavilion-style masonry pagodas (Table 2), a simplified
The masonry elastic modulus is one of important formula for fundamental period has been concluded:
influence factors to the stiffness of ancient pagodas.
The test and statistical data indicate that: (1) For brick
masonry the elastic modulus in direct proportion to
compressive strength of the masonry, and is related where, T = fundamental period; α1 = influence factor
to the strength degree of the mortar; (2) For stone of masonry elastic modulus, α1 equals to 1.0 for brick
masonry the elastic modulus is mostly dependent on pagoda and 1.2 for stone pagoda; α2 = influence factor
artifactitious precision of the stones and the strength of door opening ratio of the wall, α2 equals to 1.05 for
degree of the mortar. When constructing the dynamic two door holes each storey and 1.1 for four door holes
behavior model of ancient pagodas, the elastic mod- each storey; H = high from the ground to top of the
ulus of the masonry can be determined on the basis pagoda; D = diameter of inscribed circle of polygon

394

Figure 2. Sketch map of the parameters in Eq. (2).
foundation (Fig. 2); and d = thickness of the wall at
foundation.
3.3 Identification of modal parameters by ambient
vibration testing
Ambient vibration technique uses natural ground and
wind pulsations as excitation source, which is sim-
ple to conduct and does not damage the structure.
However, the excitation signal is generally faint for
the ancient pagoda with large stiffness. To obtain the
required input signal, the high sensitivity sensors and
low noise amplifiers should be selected. Furthermore,
the pagoda is situated on the scenic spot usually, to
reduce the exterior vibration disturbance the better
sampling time is that with lesser noise from tourists
and vehicles.
The excitations from ground and wind usually occur
at the same time, the output-only modal analysis is
used in the ambient technique to avoid the difficulty
of distinguishing the input signals.
According to the theory of random vibration, fre-
quency response function H (ω) can be expressed by
the following formula:
where Gff (ω) and Gyy (ω) are the auto-power spectra of
exciting force f (t) and structure response y(t), respec-
tively. When the frequency spectrum of input source
is smooth and approximately close to white noise with
a finite bandwidth, the power spectrum can be treated
as a constant C, which reduces Eq. (2) to:
395
It is now clear that the identification of natural fre-
quencies of the structure can be deduced from the
auto-power spectrum of structural response.
For output-only modal analysis, there exist classi-
cal Peak Pick (PP) method (Bendat JS, Piersol AG.
1993) and more recent Enhanced Frequency Domain
Decomposition (EFDD) method (Brincker R., et al,
2001) in the frequency domain. When the PP method
is used in analyzing testing data of ancient pagodas,
the modal frequencies can be identified according
to following criteria: (a) Peak value of auto-power
spectrum of structure response occurs with the same
frequency at each test point; (b) Coherence function is
larger between test points at every modal frequency;
(c) At every modal frequency, each test point has
approximately identical or opposite phases.
Based on random vibration theory, components of
mode shape can be determined by the value of trans-
fer function located at characteristic frequency. As to
the ambient vibration testing, transfer function is the
ratio of response at the test point to that at reference
point. Taking reference point as input and test point as
output, the mode shape can be analyzed using transfer
function between both points, which can be expressed
as follows:
where Gff and Gfy , respectively, are the auto-power
spectrum function and cross-power spectrum function
of the response signal.
The calculation of transfer function can be
expressed in the form of complex numbers:
where |H (ω)| equals to the ratio of amplitude of the
signal at test point to that at reference point; and ϕ(ω)
is the phase difference between the signal at test point
and that at reference point. The signs of mode shape
coordinate at each test point are correlated to the phase
relation of cross-power spectra: same signs for same
phases, opposite signs for opposite phases.
For ancient pagodas with small damping ratios and
well-separated modal frequencies, when ω ≈ ω̇ under
the arbitrary random excitation, the ratio between
peak value of auto-power spectrum and that of cross-
power spectrum, that is its transfer function, can be
approximated by the ratio of mode shapes.
3.4 Model updating and the reference criteria
Due to the uncertainty of structural parameters of
the ancient pagoda, dynamic characteristics predicted
by the analytical model often differ from field mea-
surements. The proper reference criteria should be
provided for the structural parameters identification absolute value of SFpi , the more the sensitivity of model
in the model updating. characteristic to structural parameter pi . A little adjust-
From the field testing we can obtain the first ment to pi will often cause a big change in the dynamic
n order modal parameters of a structure with N characteristics.
degrees of freedom, i.e., natural frequencies [ωT2 ] = According to the structural dynamics, the γ-order
diag(ω12 , ω22 , . . . , ωn2 ) and mode shapes [φT ] = diag eigenvalue λγ and eigenvector φ(γ) , should satisfy:
[{φ1 }, {φ2 }, . . . , {φn }]. The structural mass matrix [MA ]
and stiffness matrix [KA ] can be determined through
structural analysis (finite element analysis for exam-
ple). Basically, there are two sets of reference criteria By solving the above formula’s partial derivative to
for structural parameters identification (Li Guoqian, the i−parameter, the sensitivity of the eigenvalue can
Li Jie, 2002): (1) [ω2 ] = [ωT2 ], [φ] = [φT ]; and (2) be obtained:
[ω2 ] = [ωT2 ], [M ] = [MA ].
Generally, [ωT2 ] measured from the field test is accu-
rate while [φT ] is less accurate. On the other hand,
[MA ] obtained from structural analysis is compara- And the sensitivity of the eigenvector is:
tively accurate but [KA ] is less close to the actual
values. For the ancient pagodas, because absence of
the original constructing data, the field test data is
very valuable for the structural parameters identifi-
cation. To take full advantage of the information from
field test, the reference criteria (1) is usually selected
as the reference for the analysis model updating. Taking the stiffness parameter of structure as main
Many model updating techniques have been devel- study object, the sensitivity of the γ-order eigenvalue
oped in the past (Smith M J, Hutton S G. 1992, and eigenvector to stiffness kij , respectively, are:
Farhat C, Hemez F M. 1993, Renken J A. 1995,
Alvin K F. 1996, Chung Y T. 1997), however, for a
pagoda with complicated architectural details, the 3D
F.E. model consists of tens of thousands of meshing
units and the degrees of freedom usually, it is still a bur-
densome task on structural parameters identification.
To improve the effect of the model updating, the sensi-
tivity system should be constructed for selection of the
structural parameters firstly. Besides, taking the fast
analysis advantage of ANSYS program, the conven- By assessing and analyzing the sensitivity of
tional trial-error method also can be used to simplify dynamic response to the adjustment of structural
the model updating procedure. parameters, the sensitivity system of the model can be
established, which will provide reference for optimiz-
ing the structural parameter adjustment, distinguish-
3.5 Sensitivity of dynamic behavior to structural ing the influence of parameters, and improving the
parameter adjustment efficiency of model updating.
The sensitivity-based model updating procedure
has been recognized as an effective approach for
4 INTEGRATED MODELING OF DYNAMIC
improving FE models and the correlative researches
BEHAVIOR OF HUQIU PAGODA
(Zhang Dewen, Zhang Lingmi, 1992, Jung H. 1992,
Friswelli MI, and Mottershead JE. 1995, Dascotte E,
4.1 Prominent characteristics of Huqiu pagoda
et al. 1995) are helpful for the construction of the
sensitivity system of ancient pagodas. Situated on the Huqiu hill of ancient city Suzhou,
Suppose the structural parameters such as mass, the Huqiu pagoda was built between A.D. 959 and
stiffness, geometric dimensions and material char- 961 and is considered as the one with the oldest
acteristics described as pi (i = 1, 2,. . . , n), and age and the greatest inclination among existing Chi-
the eigenvalues or eigenvectors are considered nese pavilion style masonry pagodas. The pagoda is
as derivative functions of structural parameters, a seven-storey tube structure, with 47.68 meters in
the dynamic characteristics of the pagoda can height and 6100 tons in weight. Ever since its con-
be expressed as F = F(p1 , p2 , . . . , pn ). Then the struction, the groundwork of the pagoda has started to
sensitivity of F to structural parameter pi is settle unevenly, causing the pagoda to incline north-
SFpi = ∂F(p1 , p2 , . . . , pn )/∂pi , and the bigger the ward. Under the erosion by winds and rains and with

396
 Table 4. Material properties of the Huqiu pagoda.

Elastic modulus Density Poisson

Location MPa kN/m3 ratio

7th storey wall 1694∼ 2197 18.0 0.15

2nd∼6th storey wall 1386∼1794 18.0 0.15
1st storey wall 1694∼2197 18.0 0.15
RC foundation 25500 25.5 0.20
Soil groundwork 30 21.4 0.22
Clay loam 16 20.0 0.32

Figure 3. Picture and plane of the Huqiu pagoda.

Table 3. Dimensions of the Huqiu pagoda (m).

Storey No. Storey height Outer length Inner length

1 7.80 5.47 2.49

2 6.37 5.21 2.59
3 6.06 4.90 2.49
4 5.95 4.57 2.48
5 5.48 4.23 2.16
6 5.20 3.79 1.90
7 10.60 3.36 1.99

the development of uneven foundation settlement, the

pagoda has deteriorated severely. During the years of
1981 to 1986, Chinese civil engineering community
carried out an important intervention project to restore
this ancient pagoda which was then on the verge of
collapse. At conclusion of the project, the foundation
settlement and the structure inclination have been well
stabilized (Yuan Jianli, et al. 2004). As present, the
pagoda maintains a tip deflection of 2.325 meter with
an angle of inclination of 2◦ 48 . Figure 4. Physical and mathematical models.

4.2 Primary parameters and preliminary model

Figure 3 shows the picture and plane of the Huqiu
pagoda and Table 3 lists the dimensions at each storey.
Based on the actual survey data and documented
records (Suzhou Huqiu pagoda archives, 1988), mate-
rial properties and their variation range were deter-
mined (Table 4). Because various materials were used
during the different repair projects, the compressive
strength of the brick masonry with larger variation at
every storey, and the mortar of brick masonry is mainly
a mixture of clay-lime-sand, so the elastic modulus
of the pagoda is a main uncertain factor on dynamic
behavior model.
First, a 3D physical model of the Huqiu pagoda
was developed using AutoCAD software. The model
is divided by floors and areas of known damage or
retrofit of a large scale were incorporated into the
model. Next, the physical model was converted into
a 3D finite element model with the ANSYS finite
element program. A preliminary model of the Huqiu
pagoda, as shown in Figure 4, was established by fur-
ther defining element properties, element meshing,
and boundary conditions. Structural flaws of a rela-
tively smaller scale are not included in this preliminary
model. Finer meshes were used for connection and
transition areas such as door holes and walls, floor
slabs, joints of each floor. The number of elements
in the model added up to total of 73,241 (Yao Ling,
2003), and the Power Dynamics technique is adopted
for the analysis of modal parameters.

397

Figure 5. Test point locations.
4.3 Field testing of dynamic characteristics
The dynamic characteristics of the Huqiu Pagoda,
prominently its natural frequencies and mode shapes,
were tested by the ambient vibration method. The
main testing instrument used was an INV-306 intel-
ligent signal collection, process and analysis system
manufactured by Orient Vibration and Noise Technol-
ogy Research Institute of China. It uses large scale
signal collection and analysis software DASP (Data
Auto Sample and Process System), and the maxi-
mum sampling frequency of the system is 100 KHz.
The effective frequency scope of horizontal-velocity
sensors (Type 891-II) is 0.01∼100 Hz.
For the appropriate test parameters selection, the
dynamic characteristics of Huqiu pagoda were esti-
mated before the field test. According to the Eq. (2),
the natural period of Huqiu pagoda is about 0.84 sec-
ond, namely the first frequency is 1.19 HZ. Besides,
the truncated frequency of primitive signal was taken
at 20 HZ and sampling frequency at 60 HZ during the
ambient vibration testing.
Seven testing points were instrumented on the
pagoda with their relative elevations shown in Figure 5.
Eight sensors were used for the testing, one is placed
on the reference point and the other seven are placed on
the floors of the first to seventh storey. The reference
398
Figure 6. Auto-power spectrum of test point 3.
Figure 7. Transfer function of test point 3.
point was fixed on the second floor (i.e., the eighth
point), which was also the location of the sixth point.
The signals of eight sensors were collected at the same
time. All points were sampled in the X direction first,
and were sampled in the Y direction again. To satisfy
the request of higher orders frequencies to test time,
each sampling time is more than 20 minutes.
The auto-power spectrum as well as the transfer
function at every test point (shown in Figs 6 and 7
for testing-point 3, respectively) was analyzed. The
natural frequencies and mode shapes were obtained by
the software DASP using the improved EFDD tech-
nique (Ying Huaiqiao and Liu Jinming, 2002) to fit
testing data (shown in Figure 8). Tables 5 and 6 list the
natural frequencies and mode shapes of the first four
orders respectively.
4.4 Modification of modal parameters
By comparing and analyzing on the sensitivity of main
structural parameters such as the tower eaves, leaning
 Table 7. Comparison of the natural frequencies (Hz).

Frequency Analysis result Test result Error %

First 1.155 1.204 −4.05

Second 3.963 3.905 1.46
Third 7.778 7.295 6.21
Fourth 11.876 11.250 5.27

updating process further, the masonry’s modulus is

only selected as the structural modification parame-
ter on the basis that the accurate dimensions of doors
on the wall have been measured.
The elastic modulus is depended mostly on the dam-
Figure 8. Modal parameters fitting. age condition and material property of the brick and
mortar. In this investigation, based on the field test data
of the materials and observation on the damage con-
Table 5. Natural frequencies from the test (Hz). dition, the elementary elastic modulus of each storey
and their estimated adjusting range and positions were
1st Freq. 2nd Freq. 3rd Freq. 4th Freq.
determined. To convenient for modeling and analysis
1.204 3.905 7.295 11.250
of the computer, by adjusting the elementary elastic
modulus to simulate synthesis influence on the mate-
rial, damage, and repaired conditions of the masonry
to the structural stiffness parameter. Thus, the adjusted
modulus can be called as the synthesis modulus or
Table 6. Mode shapes from modal fitting.
composite modulus.
Test More than 40 elementary models that reflecting
point 1st Freq. 2nd Freq. 3rd Freq. 4th Freq. the possible combination with various material prop-
erties and damage conditions were selected for the
1 1 1 −1 −1 model updating. When the model is updated, the ref-
2 0.719 −0.248 −7.268 −0.8 erence updating criterion (1), that is, [ω2 ] = [ωT2 ] and
3 0.550 −1.098 −5.194 0.498 [φ] = [φT ], is adopted. The modification to the synthe-
4 0.415 −1.384 2.397 0.729 sis modulus follows the rule: Predicted mode shapes
5 0.183 −1.009 5.727 −0.397 are in general agreement with the measurements while
6 0.054 −0.482 3.575 −0.602 the errors in natural frequencies are within allowable
7 0 0 0 0
limit.
The updating objective limit in this investigation
is that: (1) The displacement of mode shape at every
storey is the same sign as the measured displace-
gradient, interior connecting details, door opening ment, (2) The allowable error for natural frequencies
ratio of the wall, as well as the strength and elastic is ±10%.
modulus of the masonry, the following conclusions can The model updating was carried out by analysis pro-
be drawn: (1) The wooden and tile tower eaves have gram ANSYS. And an Errors Evaluation Program has
lesser sensitivity to the dynamic modification, their been compiled for the error evaluation and the new
architectural details can be simplified as the battens parameter selection. A refined model for the dynamic
with the same mass and elastic modulus. In the case behavior of the Huqiu Pagoda was obtained from the
of Huqiu pagoda, the most of eaves were destroyed by 40 elementary models. The results show that when
environment erosion long ago, and the influence of the the synthesis modulus of the wall is 2029 MPa for
remnant brick braces to the dynamic behavior can be the first story and seventh story, and 1552 MPa for the
ignored. (2) Because the strengthened foundation can second to sixth story, respectively, the errors in the
be regarded as the fixed bearing of the analysis model, first four order natural frequencies are within ±7%,
the sensitivity of the leaning gradient of Huqiu pagoda and predicted mode shapes agree generally with the
is rather low and can be ignored; (3) The door opening measured ones. The comparison of natural frequencies
ratio of the wall and elastic modulus of the masonry between the analysis model and test result is listed in
have evidence influence to the dynamic behavior, so Table 7. The corresponding mode shapes are showed
the great attention should be paid. To simplify the in Figure 9.

399

a) The mode shapes from the test
b) The mode shapes from F.E.A.
Figure 9. Comparison of the first four order mode shapes.
5 CONCLUSIONS
The integrated modeling method for the dynamic
behavior takes advantage of ambient vibration tech-
nique and finite element analysis, which can obtain
not only the synthesis dynamic response of the whole
structure but also the contributions of individual fac-
tors such as connections, restraints, and damage con-
ditions. In the construction of the dynamic behavior
model of ancient pagodas, it is necessary to get hold
of the main characteristics of the method to improve
the application effect.
(1) Mastery of the main influence factors on dynamic
behavior of ancient pagodas and adoption of the
concise and proper finite element model, which
will redound to the parameter identification and
modification. The architectural details such as the
mast, eaves, and roofs of the pagoda can be simpli-
fied to reduce the number of elements and degrees
of freedom of the model. But the door holes, dif-
ferent structural materials, damaged positions, and
complicated connections, etc. should be reflected
400
in the model as much as possible to consider the
influence to the dynamic behavior.
(2) Environmental random excitation test technology
uses natural ground and wind pulsations as the
excitation source and is non-destructive to ancient
pagodas. But the excitation signals from ground
and wind pulsations are generally faint, for the
ancient pagoda with large stiffness, only the lower
order frequencies can be obtained distinctly, and
the identification effect on mode shapes is reduce
obviously with the increase of the frequency. In
order to obtain the satisfactory data, besides the
sensors and amplifiers with high sensitivity and
low noise should be selected, it is important to
ensure enough sampling time for each signal
record.
(3) The elastic modulus of the masonry has a remark-
able influence to the dynamic behavior of the
ancient pagodas, which can be selected as the main
modified parameter for model updating. It is con-
venient for model updating and analysis of the
computer by adjusting the elastic modulus to sim-
ulate the synthesis influence on the material vari-
ation, damage degree, and repaired conditions of
the masonry to the structural stiffness parameter.
(4) For the ancient pagodas, because absence of the
original constructing data, the field test data
is very valuable to identification of the model
parameters. To take full advantage of the informa-
tion from field test, the model updating criteria,
[ω2 ] = [ωT2 ], [φ] = [φT ], can be selected as prefer-
ential reference for modification of the analysis
model. And the rule: “Predicted mode shapes
are in general agreement with the measurements
while the errors in natural frequencies are within
allowable limit” will be helpful for obtaining the
appropriate analysis model.
(5) In order to enhance the effect of the integrated
modeling method, the link programs between the
model updating, the parameter optimizing system,
as well as the error evaluation, should be improved
further.
ACKNOWLEDGMENTS
This research was supported by the Cooperation
Project granted by the 11th Sino-Italian S&T Joint
Commission (2002, N. 47).
Suzhou administration of cultural heritage is grate-
fully acknowledged for the help in the investigation of
Huqiu pagoda.
REFERENCES
Alvin KF. 1996, Finite element model update via Bayesian
estimation and minimization of dynamic residuals,
 Proceedings of the 14th international modal analysis
conference, society for experimental mechanics, Inc,
561∼567.
Bendat JS, Piersol AG. 1993, Engineering applications of
correlation and spectral analysis. 2nd ed. USA, NewYork:
Wiley Inter science.
Brincker R, Ventura CE, Andersen P. 2001, Damping estima-
tion by frequency domain decomposition, Proceedings of
the 19th international modal analysis conference (IMAC),
Florida, USA, 441∼46.
Code for design of masonry structures (GB 50003-2001),
China Construction Industry Press, 2002 (in Chinese).
Chung YT. 1997, Mass matrix updating for model validation,
Proceedings of the 15th international modal analysis con-
ference, society for experimental mechanics, Inc, Japan,
818∼824.
Dascotte E, Strobbe J, Hua H. 1995, Sensitivity-based
model updating using multiple types of simultaneous state
variables. Proceedings of the 13th international modal
analysis conference, 1–5.
Farhat C, Hemez F M. 1993, Updating finite element dynamic
models using an element-by-element sensitivity method-
ology, AIAA Journal, 31(9): 1702∼1711.
Friswelli MI, Mottershead JE. 1995, Finite element model
updating in structural dynamics, 2nd ed. The Netherlands:
Kluwer Academic Publishers.
Jung H. 1992, Structural dynamic model updating using
eigensensitivity analysis, Ph.D. thesis. London (UK):
Imperial College of Science, Technology and Medicine.
401
Li Guoqian, Li Jie. 2002, Dynamic Detection Theory and
Application for Engineering Structure, Science Press, (in
Chinese).
Renken JA. 1995, Optimal solution selection for sensitivity-
based finite element model updating and structural dam-
age detection. AD-A306053.
Smith MJ, Hutton SG. 1992, Frequency modification using
newton’s method and inverse eigenvector updating, AIAA
Journal, 30(7): 1866∼1891.
Suzhou Huqiu pagoda archives, 1988, the archives of national
protection unit of historical and cultural heritages, (in
Chinese).
Yao Ling, 2003, Research and application on modeling tech-
nology of dynamic behavior of ancient masonry pagodas,
Dissertation of Yangzhou University, April (in Chinese).
Ying Huaiqiao and Liu Jingming, 2002, DASP – data auto
sample and process system, China orient vibration and
noise institution, Beijing (in Chinese).
Yuan Jianli, Li Shengcai, 2001, Analysis and investigation of
seismic behavior for multistory pavilion ancient pagodas
in China, Structural Studies, Repairs, and Maintenance of
Historical Buildings, WIT Press, UK, 2001(5)129–137.
Yuan Jianli, et al., 2004, Inclination control and reinforce-
ment technology applied to Huqiu Pagoda, China Civil
Engineering Journal, 37(5): 44∼49(in Chinese).
Zhang Dewen, Zhang Lingmi, 1992, Matrix transforma-
tion method for updating dynamic model. AIAA Journal,
30(5): 1440∼1443.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Monitoring and strengthening interventions on the stone tomb of

Cansignorio della Scala, Verona, Italy

G. Gaudini
Soprintendenza per i Beni Architettonici e per il Paesaggio di Verona, Vicenza e Rovigo, Verona, Italy

C. Modena & F. Casarin

Department of Structural and Transportation Engineering, University of Padova, Italy

C. Bettio & F. Lucchin

SM Engineering, Padova, Italy

ABSTRACT: The “Arche Scaligere” is the gothic monumental funerary complex of the illustrious Veronese
“Scaligeri” family. Between these monuments, the “Arca” (stone tomb) of Cansignorio della Scala is the most
sumptuously decorated. In 2005 the “Soprintendenza per i Beni Architettonici e per il Paesaggio di Verona,
Vicenza e Rovigo” (the local administrative body in charge of the conservation of the landscape, monuments and
historical structures), with the aim of studying the stone tomb of Cansignorio and in the framework of a wider
research, gave the task of designing and installing a structural monitoring system, besides the implementation of
a detailed FE numerical model, to the University of Padova. In parallel with these studies, a light strengthening
intervention was carried out in order to stabilize some critic points of the structure and to intervene in deteriorated
parts or elements, to integrate the original materials. The paper describes the different activities carried out.

1 FOREWORD

1.1 Historical notes and description

The “Scaligeri” or “della Scala” family was a dynasty
that ruled Verona, in Italy, for over a century, from
1262 to 1387. Cansignorio della Scala (1340–1375)
managed the city in a relatively peaceful period, and
adorned Verona in a way to make it call “marmorina”
(marbled) for the abundant use of ancient marbles and
roman statues.
The stone tomb of Cansignorio della Scala was built
between 1374 and 1376, by will of the same Cansigno-
rio, when he was still alive. The tomb was erected close
by the St. Maria Antica church, where the tombs of
Cangrande and Mastino the 2nd (his grandfather and
father respectively) were already built by local work-
ers. Differing from his ancestors, Cansignorio desired
a monumental tomb, where the architectural aspect
was more important than the decorative. The work was
then commissioned to Bonino da Campione, a famous
master of gothic sculpture. The monument, based on a
hexagonal plan, is adorned with sculptures and spired
tabernacles, with the overhanging equestrian statue of
Cansignorio. The tomb is surrounded by an hexagonal
Figure 1. The stone tomb of Cansignorio (on the right) and
wrought iron fence, at whose corners rise six pillars Mastino the 2nd (rear left), near the St. Maria Antica church.

403

Figure 2. The upper part of the Cansignorio stone tomb:
below the gablets with the statues of the virtues and the
tabernacles with angels; above the equestrian statue.
sustaining gothic tabernacles, containing statues of the
saint-warriors (St. George, St. Martin, St. Quirinus,
St. Sigismund, St. Valentine and St. Louis, king of
France). The tomb starts with six columns sustaining a
red marble slab on which finds place the white marble
sarcophagus, sustained by eight pillars and decorated
with bas-reliefs representing Gospel scenes. The cover
of the sarcophagus hosts a lying statue of Cansignorio,
watched over by angels.
At the second level, six further spiral columns
sustain the canopy with polylobed arches. Above
these finds place a cornice sustaining six gablets
with allegorical figures representing the virtues. At
the corners are positioned six further tabernacles
with statues of angels. The roof, corresponding to
an hexagonal pyramid made of white marble, finally
supports the massive equestrian statue of Cansignorio
(Fig. 2).
The stones used for the erection of the tomb are
the “Candoglia” white marble, the same employed
in the Milan’s cathedral, and the “Rosso di Verona”
(Verona’s red marble), besides the Pietra Gallina (a soft
limestone from Vicenza). The inner part of the roof
(above the crossed vault and behind the stone fac-
ing of the canopy) is composed by solid brickwork
masonry.
404
1.2 Past restoration works
Throughout the centuries, several repair interventions
were necessary to preserve the delicate structure of
the stone tomb, such as those carried out in the XVII,
XIX and XX centuries. In 1676 the Verona municipal-
ity adopted a resolution to execute restoration works on
the tomb, comporting strengthening interventions and
substitutions on the upper part of the monument, with-
out however intervening on the supporting elements.
Between 1827 and 1829 other restoration works were
carried out, raising arguments on the type of marble
to be used in substitutions of the deteriorated parts.
Between 1838 and 1844 the fence was restored, and
on the 24th of July 1840 a portion of the southern
gablet fell down, being subsequently restored (1846)
and lodged back in the original position. Substitu-
tions comported the use of Candoglia marble elements,
secured with iron clamps fixed with melted lead. The
sealing of the cracks was performed with filler. Main
interventions carried out were: the reconstruction of
the spires of some tabernacles; the positioning of steel
reinforcing elements on two columns of a tabernacle;
the complete reconstruction of a column and capital of
a tabernacle, and of some gablets between the spires;
the reconstruction of the tail and the left rear leg of the
horse in the equestrian statue; the substitution of the
copper tie beams of the tabernacles with saint-warriors
with new ones in iron; the sealing of the vault’s groins.
Other interventions, similar to those executed at the
half of the XIX century, were carried out between 1910
and 1914. The monument was then protected against
bombing during the two world wars. In 1919, after the
removal of the shields, some light restorations were
carried out. Then, during the positioning of the shields
of the 2nd world war, an analysis of the conditions
of the tombs was carried out, with successive light
restoration works.
2 THE INVESTIGATION ACTIVITIES
2.1 Dynamic identification
Between different Non Destructive techniques that
may be profitably used for the achievement of an
advanced knowledge of the structural layout of a his-
toric masonry building, dynamic identification proved
to be a very effective tool (Modena et al., 2001; Gentile
et al., 2004; Ramos et al., 2006), being actually the
only method able to experimentally define parameters
related to the global structural behavior. Prior to the
installation of a Structural Health Monitoring (SHM)
System, a dynamic investigation campaign took place
in August 2006. Tests were aimed at the definition of
the optimal SHM system sensors’ positioning, and at
the characterization of the dynamic properties of the
monument for FE modelling calibration purposes.
Figure 3. Identified mode shapes: (a) 1st bending N-S, 3.19 Hz; (b) 1st bending E-W, 3.24 Hz; (c) 1st torsion, 5.88 Hz;
(d) 2nd bending N-S, 12.55 Hz; (e) 2nd bending E-W, 12.88 Hz; (f) 2nd torsion, 19.42 Hz; (g) FDD method, average of the
normalized singular values of spectral density matrices of all test setups.

Following the mode shapes emerged from the FE
numerical model, sensors were placed at the first level
(in the stone slab where the sarcophagus stands), at the
second level (on the cornice above the pointed arches)
and at the top of the monument (at the foot of the
equestrian statue). A total of six sensors was employed,
considering three test setups for a total of 6 acquisition
points, recording the acceleration in orthogonal (and
parallel to the ground) directions.
The acquisition system was composed by a com-
pact unit provided with 24-bit digital acquisition cards,
connected to piezoelectric mono axial acceleration
transducers. Once fixed the transducers to the structure
in the selected positions, tests consisted in acquiring
data over a predetermined period, at a determinate
sample rate. Each test setup consisted in recording
the signal two times (65,536 points each) whit a sam-
pling frequency of 100 SPS (samples per second), with
 2.2 Monitoring
an overall setup signal recording duration of 21 51 .
For the identification of the modal parameters (natural
frequencies and corresponding mode shapes), output
only identification techniques were used (Operational
Modal Analysis). In particular, the recorded ambi-
ent vibrations were related to the wind excitation and
urban traffic.
The modal parameter extraction method selected
was the FDD – Frequency Domain Decomposition –
technique (Brincker et al. 2000) which estimates the
modes, with the assumption that the excitation is rea-
sonably random in time and in the physical space of
the structure, using a Singular Value Decomposition
(SVD) of each of the spectral density matrices. The
data series acquired at 100 SPS were processed by a
decimation of 2 (Nyquist frequency of 25 Hz), with
segment length of 2048 points and 66.67% window
overlap. Several peaks related to structural frequen-
cies were detected in the frequency domain and the
corresponding mode shapes defined (Fig. 3).
The Structural Health Monitoring System (installed
in December 2006) is aimed at the control of static

405

Figure 4. Positioning of the acceleration (left) and displace-
ment sensors (right).
and dynamic parameters related to the structural func-
tioning of the monument. The system is composed
by an acquisition unit connected to six piezoelec-
tric accelerometers, two potentiometric displacement
transducers and a temperature and relative humidity
sensor. The central unit, located at the base of the
tomb, is provided with a WiFi router for remote data
transmission.
The monitoring strategy is conceived both to collect
data at predetermined time-intervals (periodic moni-
toring, i.e. cracks opening, changes in the dynamic
response) and to automatically start to save data in case
of significant external events (such as seismic events).
Such controls will permit to appreciate possible vari-
ations in the assessed structural functioning with the
passing of time and to have a record of the dynamic
behavior of the stone tomb during severe events.
The acceleration transducers are placed in suitable
positions in relation to the mode shapes of the struc-
ture, as shown by the numerical modeling/dynamic
identification (Fig. 4, left). Four sensors are placed
on two levels for the evaluation of the vibration in
the NS and EW direction (bending modes) and in the
horizontal planes (torsion modes).
A couple of reference sensors is fixed at the base
for the record of the ground acceleration both in opera-
tional conditions (i.e. evaluation of the traffic induced
vibrations) and during seismic events. A tempera-
ture/relative humidity sensor is fixed at the intrados
of the marble slab (first level). The displacement
transducers are positioned across significant cracks
(Fig. 4, right, see also Fig. 14). The temperature, rela-
tive humidity and displacement of the selected points
(crack mouth opening) are recorded each 6 hours,
406
corresponding to 4 daily readings. Dynamic data are
collected both at fixed time intervals (each 48 hours,
approximately 22’ of recording at a sample rate of 100
Hz) and on a trigger basis (shorter records, signals
are recorded when the vibration exceeds a predefined
threshold).
No meaningful variations in terms of displacements
were reported up to November 2007 (Fig. 5a). Varia-
tions remain limited and related to the environmental
parameters, presenting maximum differences (corre-
sponding to crack mouth opening) of about 1/10th of
millimeter. No seismic events were recorded in the
monitored period. Limited shifts (max 4%) were noted
in all of the identified frequencies (see also Ramos
et al., 2007), possibly related to environmental param-
eters, as reported in Figures 5b and c (slight decrease
with the relative humidity, seasonal variations).
3 STRUCTURAL MODELS
3.1 Introduction
A detailed FE numerical model, based on a laser scan-
ner geometrical survey of the monument previously
carried out, was implemented in order to evaluate the
static and dynamic behaviour of the monument. The
evaluation of the initial results of the numerical model
(linear static and natural frequency analyses) assisted
the design phase of the strengthening intervention and
indicated the most suitable places for the sensors’posi-
tioning (dynamic identification and monitoring). The
first model was calibrated on the basis of the results
of the experimental activities, in order to be subse-
quently used to simulate the response of the monument
to different external actions.
3.2 The FE model
As a first step, linear elastic constitutive laws were
assigned to all materials in order to define the static
load pattern (self weight) and the dynamic properties
of the monument. The model is composed by approxi-
mately 49,000 brick elements and 53,600 nodes. Finite
elements’ sides are comprised between 0.10–0.15 m.
The mesh is more refined in the slender elements
(columns) and in the junctions, rougher elsewhere. The
decorative elements and statues were modelled as the
structural parts: only areas too small to be considered
significant were neglected (Fig. 6).
The linear static analysis (self weight) indicates
that compressive stresses reach their maximum values
in the columns, where stresses of about 1.0–1.5 MPa
(lower order) are found. In small areas of the upper
order of columns compressive stress peaks of 2.0 MPa
are noted. Tensile stresses generally present very low
values or close to zero. However, non negligible tensile
Figure 5. Monitoring results: (a) displacement transducers PZ1/PZ2 and environmental parameters, recorded data plotted
vs. time; dynamic parameters, identified frequencies: (b) first two bending frequencies vs. time and (c) vs. relative humidity.

Table 1. Experimental vs. numerical frequencies.

Frequency (Hz)

Mode Description Exp. FE model diff. %

1 1st bending N-S 3.19 3.25 1.88

2 1st bending E-W 3.24 3.26 0.71
3 1st torsion 5.88 5.85 0.48
4 2nd bending N-S 12.55 12.90 2.79
5 2nd bending E-W 12.88 13.30 3.26
6 2nd torsion 19.42 19.38 0.21

stresses localize at the crown of the pointed arches and

Figure 6. (a) Rendered view of the FE model, East side;
(b) corresponding mesh. The positive Y axis corresponds to
the North direction.
on the above cornices.
The natural frequency analysis indicates the elastic
dynamic characteristics of the structure (natural fre-
quencies and mode shapes). In the calibrated FE model
8 sets of materials are considered, theYoung’s modulus
ranging from 40,000 (solid stone) to 4000 MPa (brick-
work masonry), the corresponding densities from 2700
to 1900 kg/m3 . Six principal modes emerged from
the analysis (Fig. 7). A close match between the fre-
quencies/mode shapes emerged from the calibrated
numerical model and those emerged from the dynamic
investigation was found (Table 1).

407

Figure 7. 1 to 6, first six mode shapes of the numerical
model.
4 THE STRENGTHENING INTERVENTION
4.1 Introduction
In parallel with the studies previously reported, and
beneficiating from their outcomes, a light strength-
ening intervention was carried out. New structural
elements were introduced as precautionary measures,
e.g. by providing redundant confining systems, to col-
laborate with existing deteriorated elements and acting
in case of sudden structural deficiency of the original
material.
In general the stone tomb does not present indica-
tions of worrying structural problems. Interventions
mainly consisted in hooping the monument at differ-
ent levels. With reference to Figure 8, interventions
included: A) hooping the base of the canopy with
a stainless steel cable; B) hooping the capitals with
a couple of stainless steel cables; C) repair of the
junctions of existing tie beams; D) hooping the taber-
nacles with high resistance stainless steel cable. Local
interventions consisted in: (1) binding the damaged
408
Figure 8. Strengthening of the stone tomb of Cansignorio
della Scala, detail of the interventions.
supports (horse’s hooves) of the equestrian statue of
Cansignorio by means of CFRP strips; (2) strengthen-
ing a cracked capital with hoopings in high resistance
stainless steel cable.
4.2 Interventions description
Widespread cracks were noted in the cornice above the
pointed arches, and in the same arches close by their
keystones (Fig. 9a, b).
Parts of the cornices tended to separate, being this
however not a recent damaging process, since iron
clamps of previous interventions were found. A hoop-
ing device (hooping A) consisting in a stainless steel
7 mm diameter cable, connecting 6 corner steel plates
and tensioned by turnbuckles, was positioned above
the cornices. The size of the steel elements was min-
imized in order not to be invasive with respect to the
monument (Fig. 9c, d).
The existing iron tie beams, spanning between the
capitals of the upper order columns and locally dam-
aged by oxidation (Fig. 10a), were complemented with
a couple of 3 mm diameter stainless steel cables (hoop-
ing B). In fact, a strengthening intervention on the
original tie beams was not feasible without heavily
Figure 9. Hooping A, details.
Figure 10. Hooping B, details.
intervening on the columns capitals. Cables were fixed
to the capitals through stainless steel bushes, moulded
following the shape of the capitals (Fig. 10c). An even
contact between steel and stone was provided by means
of lead sheets.
The iron tie beams connecting the tabernacles of
the fence to the spiral columns, likely positioned dur-
ing the XIX c. interventions, manifested marked decay
at the connection with the original copper tie beams
anchored to the stone. The tie beam strengthening
intervention (C) aimed at the restoration of the orig-
inal elements avoiding the onset of a new oxidation
process on the copper anchoring elements. Two tita-
nium studs were placed to join the existing iron and
copper tie beams (Fig. 11a, b).
The deterioration or lack of the original iron
tie beams in the same tabernacles required the
409
Figure 11. Intervention C (a, b) and hooping D (c, d),
details.
Figure 12. Strengthening of the horse’s hooves: a) copper
bandage; b) deteriorated stone conditions; c) application of
CFRP strips; d) final appearance.
introduction of new elements (high resistance stain-
less steel 1.6 mm diameter cables – hooping D) to
restore the original layout (Fig. 11c, d). To minimize
the dimensions of the clamps, the fixing methodology
was tested (tensile strength of cable and connection)
by means of laboratory experimental activities.
The removal of the copper “bandage” provided to
the equestrian statue of Cansignorio during past inter-
ventions for the strengthening of the supports of the
statue, highlighted the presence of material decay, with
severe cracks and voids (Fig. 12a, b). Damaged sup-
ports were strengthened by the application of high
resistance CFRP strips, subsequently covered with a
plaster facing (Fig. 12c, d).

Figure 13. Strengthening of one of the stone capitals of the
spiral columns: (a) damage induced by the tie beam oxidation;
(b) intervention layout.
Figure 14. Above: a almost full-scale picture of the used dis-
placement transducers; below: displacement transducer PZ2
prior (left) and after (right) the restoration intervention.
The expansion of the iron tie beams due to mate-
rial oxidation caused the cracking of the capitals of the
upper order columns, in some cases with severe effects
(Fig. 13a). The strengthening intervention required
the sealing of the crack and the positioning of hoop-
ings on the capital, on 3 levels (1.6 mm diameter high
resistance stainless steel cable). Purposely shaped tita-
nium elements (Fig. 13b) were employed to allow the
clamping of the cable.
5 CONCLUSIONS
Extensive studies were carried out on the Arca of
Cansignorio della Scala, in Verona, in parallel with a
light structural strengthening intervention and stone
restoration activities. The research involved several
aspects, some of them not reported in the paper (e.g.
the laser scanner survey or stone characterization anal-
yses), aimed at the achievement of a complete picture
of the monument, for conservation purposes.
410
The investigation activities carried out (dynamic
identification) proposed experimental evidences for
the calibration of behavioral models of the monument.
Finite Elements models of the Arca were imple-
mented on the basis of the laser scanner survey of
the monument. Models were used to predict the static
and dynamic behavior of the building and were succes-
sively tuned on the basis of the experimental activities.
In a successive stage the models will be considered
(with material non linear properties) to assess the
response of the monument to seismic events.
The installation of a static and dynamic Structural
Health Monitoring System gives the possibility to con-
tinuously evaluate the conditions of the monument
by recording significant indicators (environmental
parameters, dynamic response, cracks opening). The
systems also allows to check the dynamic response of
the structure to traffic or seismic events. An impor-
tant aspect considered in the setup of the monitoring
system was the reduced impact on the monument,
given also the continuous attendance of tourists: sen-
sors were minimized and “camouflaged” as much as
possible (Fig. 14).
The design of the intervention was based on a almost
complete removability of the new structural elements,
positioned to compensate the lacks of the original
material. Interventions, except the consolidation of
heavily damaged parts (horse’s hooves), are based on
a mechanical assembly of metallic elements, avoid-
ing chemical connections with the original material.
New materials and structural elements were chosen in
order to maximally reduce their dimension (i.e. high
resistance stainless steel cables).
ACKNOWLEDGEMENTS
The authors would like to thanks student Manuel
Marotto for his valid collaboration in the execution
of the research activities.
REFERENCES
Bertolini V. Da Nantes a Brou e a Ginevra: influssi possibili o
irrefutabili delle Arche Scaligere. 1993, Verona (in Italian)
Brincker, R., Zhang, L., Andersen P. Modal Identifica-
tion from Ambient Responses using Frequency Domain
Decomposition, Proc. 18th International Modal Analysis
Conference, 2000, San Antonio, TX
De Maffei F. Le arche scaligere. 1955, Verona (in Italian)
Gentile C., Saisi A. Dynamic-based F.E. Model Updating to
Evaluate Damage in Masonry Towers. Proc. 4th Int. Sem-
inar on Structural Analysis of Historical Constructions,
2004, Padova, Italy
Modena C., Franchetti P., Zonta D., Menga R., Pizzigalli E.,
Ravasio F., Muti M., Meloni R. and Bordone G. Static and
Dynamic Analyses of Maniace Castle in Siracusa-Sicily.
Proc. 3rd Int. Seminar on Structural Analysis of Historical
Constructions, 2001, Guimarães, Portugal
Napione E. Le Arche Scaligere di Verona. Storia di un sepol-
cro dinastico, 2000–2003 tesi di dottorato di ricerca in
Storia dell’arte, Università degli studi di Udine (in Italian)
Napione E., Varanini G.M. «Gaspar recultor» e l’Arca di Can-
signorio della Scala, in Verona illustrata n.18. Museo di
Castelvecchio, 2005, Verona (in Italian)
Ramos L., Casarin F., Algeri C., Lourenço P.B., Modena C.
Investigation techniques carried out on the Qutb Minar,
411
New Delhi, India. Proc. 5th Int. Seminar on Structural
Analysis of Historical Constructions, 2006, New Delhi,
India
Ramos L., Marques L., Lourenço P., De Roeck G.,
Campos-Costa A., Roque J. Monitoring historical
masonry structures with operational modal analysis: two
case studies. 2nd International Operational Modal Analy-
sis Conference, 2007, Copenhagen, Denmark
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Seismic behaviour of barrel vault systems

A. Marini & G. Giardina

University of Brescia, Italy

P. Riva
University of Bergamo, Italy

E. Giuriani
University of Brescia, Italy

ABSTRACT: In this paper, the experimental study on the rocking behaviour of a full scale barrel vaulted
structure undergo cyclic horizontal loading is discussed. The study is the first part of an ongoing experimental
and theoretical research program, developed by the University of Brescia, concerning the seismic behaviour of
masonry buildings.
The scope of the paper is to provide some evidence of the rocking mechanism experienced by barrel vaulted
structures undergo horizontal loading. Understanding of the behaviour of such structural systems is fundamental
for their seismic vulnerability assessment, as well as for the correct design of possible strengthening techniques.
The structural behaviour is also investigated by means of non linear finite element analyses. Numerical results
are validated through comparison with experimental results. After validation, the FE model can be applied to
different case studies.

1 INTRODUCTION and numerically. Although the knowledge of the struc-

Earthquake hazard reduction is progressively becom-
ing one of the most important issues in the structural
rehabilitation discipline. The assessment of the seis-
mic vulnerability of historical constructions and the
design of successful strengthening techniques are dif-
ficult tasks, mainly because the structural behaviour
depends on too many parameters of uncertain defini-
tion, but also because of the little knowledge about
traditional construction techniques, historic material
properties and constraints (D’Ayala & Speranza 2002,
Binda 1999).
However, important lessons have been learned from
the analysis of the damage induced to historic con-
structions by recent earthquakes, lessons on the real
structural behaviour of engineered and non engi-
neered constructions (Lagomarsino et al. 2004), as
well as lessons on how to improve structural perfor-
mance (Borri et al. 2002, Modena et al. 2000). Recent
earthquakes also provided an opportunity to evalu-
ate previous structural interventions, based on actual
performance during a real ground motion (Guerrieri
1999).
The earthquake behaviour of masonry structures
has been widely studied experimentally, analytically
tural behaviour of masonry structures has greatly
enhanced in the last decades, little is still known
about the structural behaviour of vaulted structures
undergoing seismic action.
The survey of damaged vaulted structures shows
that seismic action induces bending of the vault ring,
as well as rocking of the abutments at the base.
The rocking of vaulted structures and transverse
diaphragm arches has been studied mainly analytically
and numerically by different authors, whereas fewer
experimental results are available.
Utilizing the principle of virtual work, focus has
been paid to the onset of failure mechanisms and to
their evolution in the large displacement field, to the
definition of structure capacity curves (Abruzzese &
Lanni 1999, Housner 1963, Lagomarsino et al. 2004),
and to the out-of-phase rocking of the abutments
(Como et al. 1991).
Furthermore, a simplified analytical method, based
on limit analysis, was proposed by Giuriani et al.
(2007) to assess the seismic vulnerability of trans-
verse arches undergoing rocking of the abutments.
The discussion focused on the different behaviour of
transverse arch systems subjected either to rest or
rocking conditions and highlighted the physics and

413
the phenomena involved in the mechanisms. It was
shown that in the at-rest condition, the arch lateral
thrust can be largely or entirely resisted by the buttress
action of the abutments, depending on their geome-
try and massiveness, thus possible ties must confine I II
only the part of the arch thrust exceeding the buttress
action (Giuriani & Gubana 1993, 1995). Conversely,
as rocking modifies the resisting mechanism of the A
abutments, increasing the arch thrust and consider- (a) (b) (c)
ably diminishing the confining buttress action, the arch
thrust must be entirely confined by the existing ties Figure 1. Definition of the experimental model geometry:
(Giuriani et al. 2007). A significant increase in the (a) reference structure, (b) unit stripe, (c) experimental model.
tie tensile force is expected during rocking. The same
conclusion was drawn by Lagomarsino et al. (2004)
0.25
by analysing the kinematics of the problem using the
principle of virtual work.
In this work, focus is paid to the experimental rock-
ing behaviour of a barrel vaulted structure undergoing

0.10

0.12
external lateral loading. The effect of seismic action

1.40
°
on the vault crown is beyond the scope of this paper 10

0.1
2
°
and will be studied in the future research.

10
The scope of the paper is to provide evidence of

1.61
the rocking mechanism experienced by barrel vaulted
3.40

51°
structures undergoing horizontal loading, which is
critical for their seismic vulnerability assessment, as

2.60
well as for the correct design of possible strengthening
2.00

techniques.
Focus is made on the modelling of the experimen-

tal test, on the choice of the geometry of the vault,
the technical and technological aspects concerning its
construction, as well as on the bench set up, the load
application system, and the instrument set up.
Furthermore, the experimental results are used to
validate a FE model. Validation of the FE model allows
widening the field of its application to different case
studies.
The study is the first part of an ongoing research
program developed by the University of Brescia, which
is focused on the role of vaults in the seismic behaviour
of masonry vaulted structures, the assessment of the
effectiveness of some retrofitting techniques, and on
the proposal of simplified analytical methods.
2 EXPERIMENTAL TEST
2.1 Modeling of the experimental specimen
In this paper the behaviour of a barrel vaulted structure
undergoing rocking is analysed. As a reference, a real
two storey building having a barrel vault at the ground
floor is considered (Fig. 1a).
Barrel vaults are 3D structures having a complex
three-dimensional stress distribution (O’Dwyer 1999).
However, in the frequent case of long barrel vaults
lacking the connections to the head walls, an esti-
mate of the structural behaviour can be obtained by
0.99
0.38 2.50 0.38
Figure 2. Geometry of the experimental model.
modelling the barrel vault as a series of adjoining
arches of unit width (Giuriani E. et al. 1999). In this
scenario, according to this simplified analytical model,
a unit stripe A was selected and analysed (Fig.1 a,b).
As a second step in the definition of the experimen-
tal model, the upper masonry wall were removed and
their weight was introduced in the model by adding
point loads at top edges of the abutments. (Fig. 1c).
The geometry of the experimental model is detailed
and shown in Figures 2–3, respectively.
As for the construction technique, the wall layout
and the layer by layer brick arrangement were derived
by historic construction treatises (Carbonara 1996,
2004, Giardina et al. 2007a; Fig. 4a). The resulting
wall cross section is a single leave three head wall.
The thickness of the vault crown varies along the
span: the vault ring is 0.12 m thick at the key and 0.25 m
at the springing (Fig. 4b).
The structure is built on two reinforced concrete
support; each abutment is fixed to the support with
vertical rebars which allow rotations at the base,

414
Figure 3. Full scale model of the barrel vaulted structure.
Figure 4. (a) Abutment brick arrangement; (b) change in the
crown thickness; (c) specimen constraint at the base; (d) detail
of the backfill; (e) Plexiglas plate constraint to the masonry.
reproducing the real structural constraints, but inhibit-
ing any shearing sliding (Fig. 4c).
An intrados tie rod confines the arch horizontal
thrust at the springing. The 12 mm tie rod, having an
initial pretension of 7 kN (equal to 75% of the vault
lateral thrust) is anchored to the abutments using steel
plates.
The backfill consists of fine sand, laterally con-
tained by means of Plexiglas plates bolted to the
masonry structure (Fig. 4d). Bolt-hole clearance
avoids any participation of the Plexiglas to the resist-
ing mechanism (Fig. 4e). Thin layers of chalk are
415
Figure 5. Vertical load application system: (a) side view of
the loading apparatus, (b) load distribution at the wall top
edge, (c) cup-springs at the base reducing the vertical bar
stiffness.
inserted in the backfill to emphasise the vault crown
deformation during loading and unloading cycles
(Fig. 4d).
Four 7 kN point loads are applied at the top edges
of the abutment to simulate the first floor masonry
weight. Transverse steel profiles spread the load along
the crown (fig. 5a,b). The load is applied trough ver-
tical rebars tightened to the concrete supports. Four
packs of cup-springs are interposed at the rebar bases,
as shown in Figures 5a–c, to reduce the stiffness of the
rebars. This was done to prevent an increase in vertical
load due to rebar elongation induced by increasing the
applied lateral displacement.
The total dead load of the specimen, including the
additional load of the vertical rebars is W = 90.7 kN.
2.2 Testing bench and lateral load application
system
The experimental test aims at investigating the
behaviour of vaulted structures undergoing earthquake-
induced rocking of the abutments. Further studies will
focus on the bending moment induced in the vault
crown by seismic action.
The seismic action is simulated by introducing hor-
izontal loads at the vault springing. Horizontal point
loads are applied by means of an electro-mechanic jack
fixed to a steel frame, as shown in Figure 6. The hori-
zontal loading system was designed in order to impose
the same force (instead of the same displacement) to
each abutment (Fig. 6). The load is equally divided
between the abutments through the pulley and rope
system shown in Figure 6. The point load is uniformly
spread at the vault springings by means of transverse
loading beams.

Figure 6. (a) Load application system; (b) perspective view
of the testing bench and the pulley system.
Figure 7. (a) Instrument layout, (b) detail of LVDT Cy
and Ey .
2.3 Instrument set up
During the loading and unloading steps, vertical and
horizontal displacement were recorded at some key
points of the structure by means of linear variable
displacement transducers (LVDT).
The drift of the abutments at different heights was
monitored by LVDT Ax , Bx , Fx , Gx , whereas LVDT
Cxy , Dy , Exy were located to detect any flexural
displacement along the vault ring (see Fig. 7a).
The vertical component of the displacements at the
vault springing and keystone were measured using
LVDT Cy , Dy and Ey . These instruments were fixed
to the vault intrados trough spherical hinges sliding
on horizontal supports (Fig. 7b), in order to reduce
interference with the global lateral displacement.
The applied load is measured by means of a load
cell placed between the jack and the transverse beam.
416
Figure 8. Applied horizontal force versus abutment top
horizontal displacement curve (LVDT Bx , Fx ).
The tension in the vertical rebars and the tie rod were
also monitored during the test.
2.4 Loading modalities
Load-controlled cyclic tests were carried out with a
loading rate of 600 N/min. The specimen was initially
subjected to loading and unloading cycles, in the same
direction. The horizontal point load was increased by
2 kN (corresponding to 2.2% of the total vertical load)
from cycle to cycle, until the vault developed a four
hinge mechanism. Later, two fully reversed cycle were
applied by repositioning the steel frame. The maxi-
mum applied load was equal to 11 kN (equal to 12%
of the total vertical load W).
It is worth noting that as a result of the increase
in the horizontal displacement, a small increase of the
axial force in the rebars occurred, changing the vertical
confinement of the specimen, despite the use of the
cup-spring systems.
Details on the experimental study can be found in
Giardina et al. (2007b).
3 EXPERIMENTAL RESULTS AND
DISCUSSION
Figure 8 shows the structural response in terms of
applied horizontal force versus abutment top horizon-
tal displacement (LVDT Bx , Fx ). Figure 9a and 9b
show the detail of the curve for positive and nega-
tive applied horizontal loads. The structural response
remains basically linear elastic until 6–7 kN (equal to
6.6% of the total vertical load) for both directions of
applied displacement.
As for the cyclic behaviour, damage is accumulated
after a few load cycles, and the global stiffness slightly
decreases.
Figure 8 shows that the structure is relatively ductile
with little dissipation capacity and a pronounced self-
centering behaviour (see the flag-type curve). Inelastic
deformations are almost negligible compared to the
maximum experienced displacements.
 Figure 10. Crack pattern for: (a) positive and (b) nega-
Figure 9. Details of the (a) positive and (b) negative applied tive applied lateral loads; (c) detail of the cracks in crack
horizontal force versus abutment top horizontal displacement pattern (b).
curve (LVDT Bx , Fx ).

The test was interrupted during the negative

load application (10 kN = 10%W, top displacement of
47 mm corresponding to a 1.5% drift) after a fur-
ther vertical crack opened from excessive compressive
strength on the four hinge mechanism.
For increasing applied lateral forces, cracks subse-
quently formed at the vault springing, where the vault
crown thickness abruptly halves, and at the abutment
bases. As a result, the global stiffness of the structure
progressively decreases (Fig. 8) and the envelope curve
shows a piecewise linear behaviour. Figure 10 shows
the evolution of the structure toward the 4 hinge mech-
anism; numbers refer to the crack onset chronology,
and to the value of the applied load triggering the crack.
It is worth noting that the mechanisms change for dif-
ferent directions of the applied loads (Fig. 10a,b). This
is evidenced by the higher position of the crack in
the left abutment base (Fig. 10b), which results in a
different rocking behaviour of the two abutments.
All cracks perfectly closed upon the first load
reversals, whereas in the final stages, for increasing
drifts, the non-cohesive backfill material poured into
the larger crack at the vault crown extrados (n.2 in Figure 11. Details of the (a) positive and (b) negative
Fig.10b), thus preventing its closing. applied horizontal force versus abutment displacement at the
vault springings (LVDT Ax , Gx ).
Figure 11 shows the structural response in terms
of applied horizontal force versus horizontal displace-
ment at the vault springings (LVDT Ax , Gx ). When vault span (Fig. 11a). Conversely, when negative hor-
positive horizontal forces are applied, a maximum izontal displacement was applied, 4 mm of relative
positive differential displacement of 0.04 mm was displacement was surveyed, resulting in the shortening
recorded, which corresponds to the lengthening of the of the vault span (Fig. 11b).

417
Figure 12. Vertical displacement at the vault intrados
springing and key sections (LVDT Cy , Ey , and Dy ) in case of
positive applied lateral loads (see Fig. 9a).
Figure 13. Sand backfill cracks due to internal tension
force.
The horizontal displacements of the abutments
induced the bending of the vault crown. The vertical
displacements at the intrados of the vault springing and
key, recorded by transducers Cy , Ey and Dy are shown
in Figure 12 for positive lateral load cycles. The maxi-
mum uplift displacement of the vault intrados is equal
to 5.6 mm at the right springing, corresponding to 1/3
of the abutment top lateral displacement; whereas a
2.5 mm maximum downward displacement, equal to
1/8 of the abutment top displacement, was recorded
at the left springing. The vault key also rose 1.9 mm.
Vault flexural deformations were emphasized by the
crack extending into backfill (Fig.13).
The tie tensile force significantly changed for
increasing applied lateral loads (Fig. 14).
When positive maximum displacements were
applied and the mechanism was that corresponding
to the crack pattern of Figure 10a, the tie tensile force
increased to 12% of its initial value (F = 0.82 kN,
Fig.14a). Note that the test was stopped in this direction
well before the cracks had completely penetrated
the element cross sections. Thus, for increasing lat-
eral displacement, a further significant increment
in the tie tension should be expected. Furthermore,
the abutments underwent a positive relative dis-
placement of L = 0.04 mm (Fig. 11a) at the vault
springings, which only partially explains the incre-
ment in the tie force. As a matter of fact, given
418
Figure 14. Tie traction force for a few cycles of: (a) positive
and (b) negative applied horizontal forces.
the tie length L = 3260 mm, the axial strain is equal
to ε = L/L = 1.23 × 10−5 , thus the increment in
the axial force is F = εESAT = 0.29 kN (where
ES = 210000 MPa, AT = 113 mm2 ). The further tie
increment indicates that following the development of
the crack pattern, the ideal arch increases its span and
the buttress action significantly reduces (Giuriani E.
et al. 2007).
When negative maximum displacements were
applied, the mechanism changed, as shown in
Figure 10b. The crack pattern showed two wide cracks
at the abutment bases located at different heights, thus
resulting in a negative relative displacement of the
abutments at the vault impost. In this case, the tie ten-
sile force reduced to approximately 50% of its initial
value (Fig. 14b).
Finally, the vertical tension in the rebars increased
during the test, as the lateral load and displacement
increased, despite the positioning of the cup-spring
packs at the bar bases. However, the maximum incre-
ment was at most equal to 15% of the initial value.
4 NUMERIC ANALYSES
Numeric analyses were carried out prior and after the
experimental tests. Pilot analyses provided some use-
ful information to guide the modelling and design
of the experimental test. On the other hand, upon
completion of the experimental test, the experimental
results were used to validate the numeric model.
 Table 1. Mechanical properties in Mesh B.

Masonry Backfill

Density 1900 kg/m3 1900 kg/m3

Young’s Modulus 4500 MPa 45 MPa
Poisson’s coefficient 0.15 0.15
Contact elements Friction coefficient 0.4

Figure 15. Mesh B. Contact elements are located aside each

white element along the vault crown and the abutments,
and along the vault-to-backfill and backfill-to-abutment
interfaces.

The model validation is necessary, in order to

extend the application of the numeric model to dif-
ferent case studies, geometries and load layouts. This
way, the FE analyses can help practicing engineers
in defining retrofitting strategies, and in testing their
effectiveness.
Different numeric models were implemented. First,
a pilot linear elastic analysis was performed for the
general understanding of the structural behaviour
(Mesh A). Once the governing mechanisms were iden-
tified, the model was further refined both in terms
of modelling assumptions and material behaviour. A
second model was developed by adopting contact ele-
ments to simulate and allow the onset and development
of cracks in the structure (Mesh B) Finally, in a third
model, non linear constitutive laws for masonry were
adopted (Mesh C). Mesh B showed the best fitting of
the numeric results and is discussed in the following. Figure 16. Numeric and experimental capacity curves:
The complete discussion concerning the numeric study (a) applied lateral load vs top displacement of the abut-
ments; (b) applied lateral load vs vault intrados vertical
can be found in Giardina et al. (2007c).
displacement.
Geometry as well as loading conditions are derived
from the experimental tests.Therefore, the loads acting
on the model are: through contact elements, which were located between
adjoining vault ring voussoirs, along the vault-to-
– dead load
backfill and backfill-to-abutments interfaces, and at
– vertical point loads at the abutment top edges
the abutment bases, according to the result of Mesh A.
– horizontal point loads applied at the springing.
Contact elements allow transferring the axial load only
The analyses is performed by incrementing the hori- when the surfaces are in contact, whereas shear trans-
zontal points loads. mission is governed by friction. Mechanical properties
Mesh A was used to identify the zones of weakness are summarized in Table 1.
where tensile stresses concentrate, i.e. where cracks Numeric results are shown in Figure 16a in terms of
might develop. The results highlighted that higher ten- lateral applied load versus left abutment top displace-
sile stresses occur along the vault ring, at the abutment ment. The numerical curve shape compares well with
base and, when horizontal loads are applied, at the vault the shape of the experimental curve. The greater stiff-
springing. ness and ultimate load are probably due to the selected
In Mesh B masonry and backfill were modelled material properties. Like the experimental curve, the
by means of 3–4 node plain strain elements (Fig. 15). numerical curve shows abrupt changes in the slope
The non linear behaviour of the structure was modelled at any crack opening (i.e. contact opening). Contact

419

Figure 17. Mesh B deformed shape at failure (Magnifica-
tion factor of 20).
opening is observed at the same locations where cracks
opened in the experimental test. It is worth noting that
the chronological order of crack development is also
the same.
Further numeric analyses showed that the crack
opening location is a function of the geometry and of
the applied vertical loads. When large vertical load is
applied to the structure (i.e. in the case of two or more
storey buildings), besides the crack at the abutment
bases, two further cracks form in the vault ring close
to the springing where the thickness of the section
abruptly decreases. On the other hand, for decreasing
vertical loads, these two cracks form in the abutments,
close to the vault springings, rather than in the vault
crown.
Numeric and experimental results also compare
well in terms of vertical displacements along the vault
crown (Fig.16b), and tensile tie force.
The deformed shape at failure is shown in Figure
17. The crack locations correspond to the experimental
mechanism surveyed during the experimental test.
5 CONCLUSIONS
In this paper, the transverse rocking mechanism of
a barrel vaulted structure was investigated by means
of a full scale experimental tests and a non-linear FE
analyses.
The most significant results of the experimental
study can be summarized as follows:
• The structure is relatively ductile in terms of
displacements. However, little energy dissipation
capacity and a pronounced self centering behaviour
are observed. Failure is reached with a four hinge
mechanism. Any possible retrofitting technique
should be respectful of this structural behaviour,
by strengthening the structure without reducing its
capacity to meet to displacement demand;
• The tie tension significantly changes during rock-
ing. The small abutment relative displacements
induced by rocking is not sufficient to explain this
increase in tie tension. The tension is likely to
increase following the loss of the abutment buttress
420
action and the increase of the ideal arch span.
This result should be carefully taken into account
in seismic vulnerability assessment, as tie rods
might experience excess in tension during an earth-
quake, thus requiring additional strengthening. On
the other hand, depending on the triggered mech-
anism, tie tension might also decrease if rocking
induced displacements shorten the vault span.
The numerical study also confirmed that the crack
opening locations are a function of the geometry and
of the applied vertical loads. When large vertical loads
are applied, two cracks form in the vault ring close
to the springing where the thickness of the section
abruptly halves. On the other hand, for decreasing ver-
tical loads, two cracks form in the abutments, close to
the vault imposts, rather than in the vault crown.
Further experimental tests will be performed on
unreinforced and reinforced masonry vaults, to eval-
uate the efficacy of some strengthening techniques,
such as: extrados spandrel walls, thin slab of clay
mortar, and FRP. Numerical and analytical models
will be improved to simulate cyclic behaviour of the
structures.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge R. Squicciarini
for his contribution in carrying out the experimental
test and L. Martinelli for his technical support. This
work was developed within the research project DPC-
ReLUIS 2005–2008, Research line n. 1: “Vulnerability
assessment and antiseismic strengthening of masonry
buildings”. The experimental specimen was built by
the students of the Scuola Edile Bresciana, within a
joined research agreement. Their work, as well as the
work of their supervisors, is gratefully acknowledged.
REFERENCES
Abruzzese D., Lanni G. 1999. Some developments on the lat-
eral strength of historical reinforced vaulted buildings. In
Structural Studies, Repairs and Mainteinance of histor-
ical buildings, VI International Conference, STREMA,
Brebbia and Jager Ed., Dresda.
Binda L., Baronio G., Gambarotta L., Lagomarsino S.,
Modena C. 1999. Masonry constructions in seismic areas
of central Italy: a multi-level approach to conservation,
8NAMC 8th North American Masonry Conf., Austin,
USA, pp. 44–55.
Borri A., Corradi M., Vignoli A. 2002. New materials
for strengthening and seismic upgrading interventions.
In International Workshop Ariadne 10, Arcchip, April
22–28 2002, Prague, Czech Republic, cd-rom: 1–24.
Carbonara G. 1996. Atlante del restauro. Torino: Utet.
Como M., Lanni G., Sacco E. 1991. Sul calcolo delle catene
di rinforzo negli edifici in muratura soggetti ad azione
 sismica. VII convegno nazionale L’ingegneria sismica in
Italia, ANIDIS, Palermo.
D’Ayala D., Speranza E. 2002. An integrated procedure for
the assessment of seismic vulnerability of historic build-
ings. In Proceedings of the 12th European Conference on
Earthquake Engineering, London, Elsevier Science, paper
n. 561 (CD-ROM).
Giardina G., Marini A., Riva P., Squicciarini R. 2007a. Anal-
isi dei meccanismi di collasso e delle tecniche di rinforzo
di archi e volte in muratura. Technical Report n. 13.
University of Brescia.
Giardina G., Marini A., Riva P., Squicciarini R. 2007b.
Allestimento e collaudo del modello sperimentale e del
banco di prova di una volta in muratura soggetta a carichi
orizzontali. Technical Report n. 14. University of Brescia.
Giardina G., MariniA., Riva P. 2007c.Analisi numerica di una
volta in muratura soggetta a carichi orizzontali. Technical
Report n. 15. University of Brescia.
Giuriani, E., Gubana, A. 1993. Recupero e consolidamento
di volte in muratura. In Atti dei Colloqui Internazion-
ali Castelli e Città fortificate, Storia Recupero, Valoriz-
zazione, Palmanova, 24–25 settembre.
Giuriani E., Gubana A. 1995. Extrados ties for structural
restoration of Vault. In Proceedings of the Fourth Inter-
national Conference on Structural Studies of Historical
Buildings Strema 95, Architectural Studies, Materials and
Analysis, Vol. 1, Ed. C. A. Brebbia, Wessex Institute of
Technology, UK.
421
Giuriani E., Marini A., Porteri C., Preti M. 2007. Meccanismo
di rocking degli archi diaframma e vulnerabilità sismica
delle chiese. Technical Report n. 7. University of Brescia.
Giuriani E., Marini A., Porteri C., Preti M. 2007. Seismic
vulnerability of churches associated to transverse arch
rocking. Submitted for possible publication to the Jour-
nal of Architectural Heritage: Conservation, Analysis and
Restoration.
Guerrieri (Ed.) 1999. Manuale per la riabilitazione e la
ricostruzione post sismica degli edifici. Regione Umbria.
DEI. Tipografia del Genio Civile.
Housner G. W. 1963. The Behaviour Of inverted pendulum
structures during earthquakes. Bulletin of the seismologi-
cal Society of America, 53 (2).
Lagomarsino S., Podestà S., Risemini S., Curti E., Parodi S.
2004. Mechanical models for the seismic vulnerability
assessment of churches. In Structural analysis of histor-
ical construction (C. Modena, P. B. Lourenço, P. Roca
Eds.), Proc. of IV Int. Seminar SAHC, Padova, Italy,
A.A. Balkema, London (UK), Vol. 2, pp. 1091–1101.
Modena C., Pineschi F., Valluzzi M. R. (Eds.) 2000. Val-
utazione della vulnerabilità sismica di alcune classi di
strutture esistenti. Sviluppo e valutazione di metodi di rin-
forzo. Roma: CNR, Gruppo Nazionale per la Difesa dai
Terremoti.
O’Dwyer D. 1999. Funicular analysis of masonry vaults. In
Computers & Structures, 73: 187–197.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Assessment of material degradation based on microcores testing

M. Skłodowski
Institute of Fundamental Technological Research Polish Academy of Sciences, Warszawa, Poland

ABSTRACT: The paper presents the results of testing of small microcore specimens using Compact Diagnostic
Test (CoDiT). Measurement of the evolution of strength and elastic properties of limestone is used to demonstrate
the applicability of microcores testing for the assessment of actual state of degradation of heritage construction
materials. The measurements are performed on artificially aged Portland Limestone, widely used as monuments
building material, e.g. in London. Microcore specimens are drilled out of samples which undergone the series of
cycles of accelerated Freeze and Thaw (F/T) ageing. Among the measured parameters there are the velocity
of ultrasonic waves and the flexural, tensile and compressive strengths of the material after various number
of F/T cycles. Results show that specimens drilled of the more aged sample are stronger than less weathered
material, which is in agreement with the results of the previous non-destructive testing of large samples done in
another laboratory. Based on the CoDiT image analysis, the effect is attributed to large material inhomogeneity
which is so important that it overpasses any possible material degradation introduced by the applied weathering
procedure.

1 INTRODUCTION application of core drills. A natural further conclusion

Direct measurements of material strength degrada-
tion are done very rarely as they are harmful to the
monument if done according to civil engineering stan-
dards. However small samples of historical materials
are, or can be, available if microcores are drilled out
of masonry during installation of sensors and other
elements, like for example scaffoldings. Testing of
small samples of historical construction materials in
the form of microcores is a new experimental field
having only a few years history. Kasal reported the
drilling of 5 mm microcores using 10 mm core drill
for evaluation of strength of wood structural mem-
bers (Kasal 2003). The method was further extended
for sampling of non-cylindrical wood specimens and
tensile strength and diametral compression loading
was studied (Kasal et al., 2003). Hydraulic portable
press for diametral compression of wood microcores
developed at the Institute of Theoretical and Applied
Mechanics, Academy of Sciences of Czech Repub-
lic, was exhibited during VII European Conference
“SAUVEUR” in June 2006, in Prague.
In parallel, a technology of microcore testing of
masonry materials has been developed, inspired by
the drilling resistance measurement method DRMS
(Tiano et al. 2000). This method makes use of dia-
mond drills with flat cutting surface and one of the
main problems is zero rotational speed (and hence,
cutting speed) in drill contact points positioned on
the drill axis. A solution to this problem might be the
was that one can use small microcore samples from
every hole drilled with a small diameter core drills,
not only in connection with the DRMS tests.
Next step in the development of small sample
testing of historical masonry materials was the manu-
facturing of the equipment for drilling long microcores
and development of the experimental procedures of
microcore testing. This development was done in 2005
at the Institute of FundamentalTechnological Research
of Polish Academy of Sciences and first reported
in (Skłodowski 2006a). In the same year the exper-
imental procedure called Compact Diagnostic Test
(CoDiT) was presented (Skłodowski 2006b) and a
comparison of the results for standard size specimens
and microcore specimens tests was done (Skłodowski
2006c).
The following report presents the results of mea-
surements performed on microcores drilled out of
samples of Portland Whitbed limestone weathered
in laboratory by Freezing & Thawing (FT) cycles.
This artificial ageing procedure was done within the
framework of European Commission research project
MCDUR by researchers from Politechnico di Torino,
DITAG – Land, Environment And Geo-Engineering
Department (Bellopede & Manfredotti 2006). They
also made non-destructive measurements of the elastic
properties of the weathered limestone. Measurements
of ultrasonic P-wave velocity, of Schmidt hammer
rebound hardness and of DRMS were used. All
three methods showed a rather unexpected result

423
that specimen which undergone 120 FT cycles had
better elastic properties than intact Portland specimens
(Morandini & Marini 2005).
Present research using microcore testing by the
Compact DiagnosticTest (CoDiT) procedure confirms
the previous results and shows that not only elastic
properties, but also strength properties of microcores
drilled out of these previously tested samples of arti-
ficially aged limestone are much higher. However, the
morphological analysis included in the CoDiT proce-
dure suggests that this “strengthening effect” should
rather be attributed to large material inhomogeneity
than to the FT ageing process.

2 THE CODIT METHOD

Compact Diagnostic Test (CoDiT) is a sequence of

experimental steps (Skłodowski 2006a) which serves
to gain of a maximum knowledge from the investiga-
tions microcores. Such a sequence of measurements
is non-destructive from the viewpoint of the monu-
ment if microcores are drilled out of masonry during
installation works.
Consecutive steps of the CoDiT are:
Figure 1. Unwrapped surface of the microcore specimen
Step 1. Draw a line to permanently mark the refer- PW6 from the sample before weathering.
ence coordinates of the material within the structure
e.g the vertical plane crossing a hole to be drilled.
a case an uniaxial cyclic compression of one half-core
Step 2. Measure the velocity of propagation of sur-
and a creep test of the other might be of a special
face wave (Rayleigh wave) along the marked the
interest.
line and perpendicularly to it.
Step 3. Drill a hole with a core drill bit, preferably
during installation works and take out a microcore.
3 MATERIAL
Step 4. Flatten the core base which resulted after its
breaking off and copy the line mark (from Step 1)
The material is Portland Whitbed Limestone used in
on it and on the cylindrical side surface of the core.
construction of historical buildings in UK. Well known
Step 5. Measure the core dimensions and weigh the
examples of such constructions are St. Paul’s Cathe-
core.
dral in London and Sydmonton Court in Newbury.
Step 6. Record an image of the side wall of the
This limestone is a fine grained Jurassic stone com-
core and analyse micro-cracks, pores and material
posed in over 92 percent of calcite and less than
inhomogeneities.
8 percent of quartz. Inside the material, shell fragments
Step 7. Perform three-point bending test of the core
are deposited among calcium carbonate forming bands
in the plane marked. Two “half-cores” will come
or are distributed randomly.
into being.
All Portland Whitbed Limestone samples were cut
Step 8. Cut the half-cores to the desired length
from the stone bed from Albion Independent quarry
and measure the velocity of propagation of the
(Portland, UK) for the research purposes of the Euro-
longitudinal wave along the cores.
pean 5Th Framework project MCDUR
Step 9. Perform uniaxial compression test using one
Figures 1–3 are microcore images showing exam-
half-core (see Step 7).
ples of three typical cases. In Figure 1 a very few
Step 10. Perform diametral compression test of the
shell fragments localised far from each other is seen.
other half-core along the originally vertical plane
Figure 2 shows large bands of hard material localised
(comp. Step 1 and Step 7).
mostly parallel to the microcore axis. In the Figure 3
Destructive tests of Steps 9 and 10 can be modified a large amount of small hard inclusions randomly
as necessary.This is especially recommended when the distributed in stone can be seen.
number of microcores available for testing is greater The CoDiT procedure reveals not only these very
than the one required for basic measurements. In such different morphological compositions of the Portland

424
 Figure 4. Samples of Portland Whitbed with drilled micro-
core specimens.

Table 1. FT procedure.

Time

Start of The saturated specimens are T0

cycle introduced into the freezing tank
Stage 1 The temperature at the centre By T0 + 2 h
Figure 2. Unwrapped surface of the microcore specimen of the monitored specimen
PW28 from the sample after 60 cycles of freezing & thawing. reaches – (10 ± 2)◦ C
Stage 2 The temperature at the centre T0 + 4 h
of the monitored specimen
remains at – (10 ± 2)◦ C
Stage 3 The specimens are introduced By T0 + 4.5 h
in the water bath and are totally
immersed
Stage 4 The temperature at the centre of
the monitored specimen
reaches (20 ± 5)◦ C By T0 + 6 h
End of The temperature at the centre of T0 + 8 h
cycle the monitored specimen remains
at (20 ± 5)◦ C

4 SPECIMENS

Specimens in the shape of microcores with the diam-

eter of 6–6.5 mm and the length of 24–30 mm are
used for measurement of several elastic and strength
parameters of the tested material.
Figure 4 illustrates how the specimens are collected
from the rectangular prismatic samples of Portland
Whitbed limestone.

Figure 3. Unwrapped surface of the microcore specimen
PW48 from the sample after 120 cycles of freezing &
thawing.
material. Mechanical tests included in the CoDiT
procedure shows the consequences of these various
material structures on elastic and strength properties
of the tested specimens.
4.1 Weathering
Weathering of Portland Whitbed samples was done by
Politechnico di Torino, DITAG – Land, Environment
And Geo-Engineering Department (Bellopede & Man-
fredotti 2006) according to the procedure presented in
Table 1.
4.2 Sampling of specimens
In the present research the microcores were drilled
out of samples originally encoded as PW052 which

425
undergone 60 FT cycles and PW128 aged by 120 FT
cycles. Microcores were drilled with their axis parallel
to the limestone bedding plane. This allows to per-
form the measurements of the P-wave measurement,
the uniaxial compression strength and the indirect ten-
sile strength with deformation localisation along the
bedding plane. Flexural strength could be measured
perpendicular to the bedding.
As the first step, a set of lines parallel to bedding
planes were marked on PW052 and PW128 samples to
allow the unique identification of the direction of loads
applied in measurement procedure. The second step
before sampling was the measurement of the velocity
of propagation of ultrasonic surface wave (Rayleigh
wave) along the marked vertical lines and perpendic-
ularly to them using edge probes (Skłodowski 2005).
After this measurement, seven long microcores were
drilled from each sample of limestone. Specimens
length was about 27 mm and their diameter between
6 and 6.5 mm.
Reference samples of non-weathered limestone are
the specimens reported in the previous author’s paper
(Skłodowski 2006c).
Examples of each group of microcore specimens
are shown in Figures 1–3 revealing the various internal
composition of the limestone. It has happed that these
various samples were used in previous FT weathering
test, each sample being frozen applying different num-
ber of cycles. Assuming that freezing process cannot
change material morphology so much, it can be sad
that the weathering processes would have lower influ-
ence on material properties than differences between
local properties of specimens used in each step of the
FT test.
5 MEASUREMENT PROCEDURE
The microcore specimens prepared according to the
above sampling method were subjected to the follow-
ing testing procedure.
At first images of a side wall of the core were
taken and combined, when necessary, into a single
image of the unwrapped side surface of the core.
Examples of such images are shown in Figures 1–3.
This allowed analysis of specimen’s morphology and
material inhomogeneity.
Next, the line marking from the microcore base was
elongated onto its cylindrical side wall to keep the track
of the specimen’s orientation during the other tests.
As the next step three-point bending test of the
core was performed to measure limestone flexural
strength (FS). Direction of loading was the same as
that of the line marked on the core. This allowed to
break the core perpendicularly to the material bedding.
Resulting two sub-cores, marked with additional let-
ters “a” and “b” (e.g bending of the specimen PW28
results in obtaining specimens PW28a & PW28b) were
426
cut to flatten their bases and to obtain a short core
of the length equal to the microcore diameter and a
long core having the length equal to twice the diam-
eter. After that, the propagation velocity of ultrasonic
P-wave along the longer sub-cores was recorded. As
the last two steps, the destructive tests of both sub-
cores were done. Smaller cores were compressed along
their diameter coinciding with the bedding plane to
measure the indirect tensile strength (ITS) of the
material and longer cores were used for uniaxial
compressive strength measurement (UCS).
Strength measurements are done using a specially
designed, hand operated, screw driven, mechanical
loading device. Recorded load-displacement curves
are based on strain gauge and LVDT sensor mea-
surements, respectively. Analog signals are amplified,
digitized online and stored in the connected PC.
6 RESULTS AND DISCUSSION
CoDiT results of limestone degradation assessment
presented here belong to three groups. The first group
are the results of optical analysis included in the
CoDiT procedure. They are presented in the form of
unwrapped cylindrical surface images of the micro-
core specimens given in Figures 1–3 and showing
specimens morphology.
The analysis of post-critical deformation and crack
formation is also performed in each mechanical test to
make sure that deformation mechanisms are correct
and hence that the strength measurements are reliable.
Examples of these deformation modes are presented in
Figure 5 where cracks developed in the Brazilian Test,
in the uniaxial compression test and in the three-point
bending test are shown for specimen PW29 after 60
FT cycles.
It can be seen in Figure 5 that for fine grained
Portland Limestone the deformation mechanisms in
all three tests are the same as in the case of testing
standard full-size specimens.
Forty two mechanical tests of microcores were
conducted including seven bending, seven diametral
compression and seven uniaxial compression exper-
iments for each of the weathered Portland samples.
Stress-displacement curves were recorded during each
mechanical test. Their examples are presented below
in Figures 6–8. They are grouped in pairs to underline
the differences recorded in the behaviour of the mate-
rial from the sample after 60 FT cycles of weathering
and that after 120 FT cycles.
Load-displacement curves for bending of Portland
Limestone presented in Figure 6 show that the spec-
imen drilled out of a more weathered sample can be
even two times more tough than the specimen from
a less aged material. Average difference in flexural
strength is also very pronounced (Fig. 9) but not as
high. It is worth to remind here that motivation for
Figure 5. Post-critical deformation modes developed in
microcore specimens tested according to CoDiT procedure.
Specimen diameter is approx. 6 mm.
Figure 6. Force [N]-displacement [mm] diagrams for bend-
ing of specimens: PW28 (after 60 FT cycles) – upper, and
PW48 (after 120 FT cycles) – bottom.
427
Figure 7. Force [N]-displacement [mm] for BrazilianTest of
specimens: PW28a (after 60 FT cycles) – upper, and PW48a
(after 120 FT cycles) – bottom.
the presented research was an unexpected high ultra-
sonic P-wave velocity, Schmidt hammer rebound hard-
ness and drilling resistance recorded during previous
research in another laboratory for the most weath-
ered Portland specimens. Three-point bending test
presented here confirms that the previous results are
meaningful even if considered strange and unexpected.
The next loading tests are done using two sub-cores
resulting from each bending tests as it can be seen in
Figure 5. Before proceeding, the halves of a micro-
core are prepared for loading by cutting their ends to
produce short and long cylindrical specimens with flat
and parallel bases. Cutting length is 6–6.5 mm for short
specimens used in Brazilian Test and about 12–13 mm
for specimens used in uniaxial compression tests.
During diametral compression of short specimens
it was possible to perform unloading-loading cycles
in some of the experiments. Such load-displacement
curves are presented in Figure 7 for the same microcore
specimens for which experimental curves recorded
during bending test are shown above. Again it can
be seen that the specimen from the most weathered
sample is much stronger than that from the less aged

Figure 8. Force [N]-displacement [mm] for uniaxial com-
pression of specimens: PW28b (after 60 FT cycles) – upper,
and PW48b (after 120 FT cycles) – bottom.
Figure 9. Average values of Portland Whitbed mechanical
properties measured for material samples artificially aged by
freezing and thawing.
material. This time the strength ratio between both
specimens is less pronounced than in the case of bend-
ing and the average values of diametral compression
ITS are even more close to each other.
428
Unloading-reloading curves can be used to get addi-
tional information about the stiffness of the specimens.
Let us observe that higher stiffness of specimen made
of more aged material shown at the bottom of Figure 7
can be seen even without thorough calculations. This
again confirms the previous results of non-destructive
tests.
Results from uniaxial compression tests of speci-
mens from the other sub-cores show the same tendency
which was observed in FS and ITS values. The strength
of specimens drilled out of the sample which under-
gone 120 FT cycles is much greater than the strength of
specimens from the sample weathered in 60 FT cycles.
Load-displacement curve in the upper part of the
Figure 8 is the example of the test which was discarded
due to an improper deformation mechanism of the
specimen and, in consequence, too low ultimate load.
This specimen was crushed in several places starting
from its upper base forming several short shear cracks
which reach about one third of the specimens height.
In the other uniaxial compression tests the pro-
portion between the recorded ultimate loads for both
weathered limestone specimens were close to 2:1
instead of 3:1 which could be suggested by the curves
presented in Figure 8.
The analysis is based on the measurement of five
parameters. Two of them are the indexes which char-
acterize the elastic properties of limestone. These are
the squared values of P-wave and R-wave velocities
VP2 and VR2 which are proportional the to modulus of
linear elasticity (Landau & Lifschitz 1986). The three
others are the measured values of FS, ITS and UCS.
Average values of these parameters are presented in
Figure 9 where ordinate axis takes the dimension of
MPa or km2 /s2 depending on the index type.
Figure 9 illustrates in a single diagram the fact
that the trends in all the measured elastic and strength
properties are the same, thus showing that this is a pre-
dominant tendency. The presented results are average
values recorded for all the tested microcore speci-
mens from the weathered samples. It is easily seen
that one sample of the limestone has much better
properties than the other one. It is also clear that
the differences are so meaningful that they overpass
any possible material degradation introduced by the
applied weathering procedure. A possible explanation
of this phenomenon can be attributed to the material
morphology and internal composition of the speci-
mens.To discuss the problem let us look at Figures 1–3.
The specimen PW48 from Figure 3 is an example of
specimens machined from the limestone sample which
undergone ageing of 120 FT cycles. Its internal struc-
ture is dense with a sparse distribution of small and
hard inclusions of shells. It resembles a good concrete
material – of course in another scale.
Specimen PW28 from Figure 2 is machined from
a sample which undergone medium ageing of 60 FT
cycles. The unwrapped surface of the specimen shows
a large shell inclusion which acts as a strengthening
element within the specimen on the one hand but also
as a stress concentrator on the other.As the result of this
inhomogeneity, a local increase of flexural strength is
observed because the shell stripe is almost parallel to
stress trajectories in bending. Also the P-wave time of
flight decreases as the wave travels the part of its dis-
tance choosing faster and easier path through the shell.
But in diametral compression and in uniaxial compres-
sion tests the specimen fracture is probably activated
along contact surfaces of oolite limestone and the shell
due to incompatibility of strains. Thus the strength
increase is smaller than in the more homogeneous
PW48 specimen.
Specimen PW6 shown in Figure 1 is machined from
an almost homogeneous sample of limestone com-
posed mostly of calcite. This sample was not exposed
to frost ageing but a relatively low strength of this
intact material results from its mineralogical compo-
sition with predominant weak calcite grains and no
strengthening effect due to almost complete lack of
shell inclusions.
7 CONCLUSIONS
The main conclusion is that the CoDiT method con-
firms the results of previous research showing that the
sample which undergone 120 cycles of FT weathering
has the best mechanical properties. However, the opti-
cal analysis included in the CoDiT procedure suggests
that this effect should be attributed to morphological
composition of the samples of limestone rather than to
the weathering processes.
Degradation of Portland Limestone due to consec-
utive freezing and thawing is much lower than the
material inhomogeneities. These inhomogeneities are
large both in the sense of morphological inhomogene-
ity and also local variations of elastic and strength
material properties.
The Compact Diagnostic Test, unlike other research
methods used for assessment of material degradation,
provides the information on the actual elastic and
strength material properties and shows that they have
usually a similar trend of qualitative correspondence.
The advantage of the CoDiT method is that the small
size of the specimen makes it possible to look at the
material structure from a relatively close perspective.
This makes it possible not only to state the mechan-
ical properties of the material, but also to see what
morphological structure gave rise to these properties.
429
ACKNOWLEDGEMENTS
The measurements described in this paper were con-
ducted using the same samples of Portland Whitbed
Limestone which were tested in Politecnico di Torino
within the EU 6 Framework Program (MCDUR,
Contract G6RD-CT2000-00266).
REFERENCES
Kasal, B. 2003. Semi-destructive method for in-situ evalua-
tion of compressive strength of wood structural member
Forest products journal 53, no. 11–12, pp. 55–58.
Kasal, B., Drdácký, M. & Jirovský, I. 2003. Semi-destructive
methods for evaluation of timber structures, in: Proceed-
ings STREMAH VIII, Series: Advances in Architecture.
Vol. 15, pp. 835–842, WIT Press, ISBN 1-85312-968-2.
Tiano, P., Filareto, C., Ponticelli, S. & Ferrari, M., and Valen-
tini, E. 2000. Drilling force measurement system, a new
standardisable methodology to determine the “superficial
hardness” of monument stones: prototype design and val-
idation. Int. Journal for the Restoration of Buildings and
Monuments, 6(2), pp. 115–132.
Skłodowski, M. 2005. Application of ultrasonic Rayleigh
wave to testing of masonry materials. In: C. Modena,
P. B. Lourenço & P. Roca (eds.), Structural Analysis of
Historical Constructions, pp. 395–400. London: Taylor &
Francis Group.
Skłodowski, M. 2006a. Compact Diagnostic Test: outline of
historical monuments testing procedure. IFTR Reports,
3/2006, Warszawa: IFTR.
Skłodowski, M., 2006b. Microcores technology for testing
of historical masonry materials. In J. Radić, V. Rajčić,
R. Žarnić (eds.), Proc. Int. Conf. HERITAGE PRO-
TECTION, constructional aspects, pp. 205–211, ISBN
953-95428-1-2
Skłodowski, M. 2006c. Quasi-non-destructive testing of
historical structural materials using microcores. In
P.B. Lourenço, P. Roca, C. Modena, S. Agrawal (Eds.),
Proc. 5th Int. Conf. Structural Analysis of Historical
Constructions, New Delhi 2006, 6–8 Nov., ISBN 972-
8692-27-7, pp. 859–865, MACMILLAN 2007.
Bellopede, R. & Manfredotti, L., 2006. In Fort, Alvarez
de Buergo, Gomez-Heras & Vazquez-Calvo (eds), Her-
itage,Weathering and Conservatio,. pp. 539–546. London.
Taylor & Francis Group, ISBN 0-415-41272-2.
Morandini, A. F. & Marini, P., 2005. unpubl. In Deliverable
10, Weathering Charts for AA Tests, Effect of the weath-
ering on stone materials: assessment of their mechanical
durability, EU Research Project G6RD-CT2000-00266.
Landau, L. D. & Lifschitz, E. M. 1986. Theory of Elasticity,
3rd rev. enl. ed. Oxford, England: Pergamon Press.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Structural performance of riveted connections in historical

metal structures

Mario D’Aniello & Luigi Fiorino

University of Naples “Federico II”, Department of Structural Engineering, Naples, Italy

Raffaele Landolfo
University of Naples “Federico II”, Department of Construction and Mathematical Methods in
Architecture, Naples, Italy

ABSTRACT: Nowadays, in Italy historical metal structures represent a significant and important part of
national architectural heritage. In fact, their intrinsic architectural and technological value represents a milestone
in the history of constructions, being an attestation of great advance of technology and innovations reached more
than two centuries ago. These structures usually show a spread damaging mainly due to the corrosion and the
structural inadequacy. In particular, among the structural deficiencies the insufficiency of riveted connections
can be easily recognized. Hence, in order to characterize the mechanical behaviour of riveted connections, an
experimental and numerical research campaign has been planned. In this activity, riveted joint specimens have
been designed to be representative of the connections of the Italian existing structures, such as roofing structure
and steel railways bridges. This paper briefly summarizes the whole research activity and underlines the crucial
themes to be deeply investigated.

1 INTRODUCTION failure of the structure, because the riveted connections

Metal structures represented since the beginnings of
the XIX century a great advance of the architecture
and one of the most important innovations in the his-
tory of constructions. However, among the variegated
constructive scenario, two main structural typologies
can be distinguished: the large span roofing structures
and the structures for bridges, mostly for railway use.
In particular, the latter cover a crucial role because
of their strategic role and their relevant economical
implications.
After several decades from their erection, this class
of constructions reveals some damages and/or struc-
tural inadequacies (Guerrieri et al. 2005). In fact, the
ancient metal structures, generally made of puddled
iron or wrought steel riveted plates, were built on the
basis of the experience on timber constructions, conse-
quently without an adequate knowledge and maturity
about typical aspects of design of steel structures,
such as local buckling phenomena and concept of
connection details. In particular, the ancient metal con-
structions present riveted connections, usually charac-
terized by inadequate capacity and improper geometry.
Moreover, these types of constructions are generally
characterized by trussed structural schemes. Hence,
the failure of connections can produce the overall
represent the weaker elements of these structures.
These considerations lead to identify the vulnera-
bility of connections as the crucial aspect concerning
the structural capacity of ancient metal structures. In
this direction, the research effort presented in this
paper is devoted to investigate the structural capac-
ity of the above-mentioned connections, in order to
provide a tool for structural safety assessment of the
historical structures built at the end of the 19th cen-
tury or the beginning of the 20th century that are still
in service. To achieve this purpose, a large number
of riveted specimens, representative of local details
of ancient connections, have been manufactured to
be tested under shear loading (lap shear tests). The
riveted joint specimens have been designed to be rep-
resentative of the connections of the above mentioned
structural typologies.
Two existing historical structures have been chosen
as benchmark to define the geometry of the riv-
eted specimens. Therefore, among the ancient roofing
structures, the Gallery Umberto I in Naples (Figure 1a)
has been assumed as study case. While the railway
bridge on “Gesso” river (Figure 1b), in Southern Italy,
has been chosen as representative of the second struc-
tural type. In detail, the experimental campaign can
be subdivided into two main branches: 1. Tests on the

431
 Figure 2. Replacement of a riveted connection of an Italian
Figure 1. The structure under examination: the roof struc- existing railway bridge.
ture of Gallery Umberto I (a); the railway bridge on Gesso
River (b).

original materials of the Gallery and of the Bridge

(Brinnel hardness test; chemical Analyses; Charpy
test; Uniaxial tensile test); 2. Lap shear tests on riveted
connections.
It is worth to notice that the planning of this exper-
imental activity required rediscovering the outdated
technology of the riveting process. In Italy, because
of its expensiveness, the riveting technology is still
implemented only by Italian railways for steel bridges
(Figure 2). In fact, the riveting process (as summa-
rized in Figure 3) is more complex than the tightening
of high strength bolts. It requires of inserting the rivet
in matching holes of the plates to be joined and sub-
sequently forming a head on the protruding end of the
shank. The holes are generally 0.1 mm greater than
the diameter of undriven rivet. The head is formed
by hot forging with a pneumatic hammer (Figures
3a,c,d). In addition to forming the head, the diam- Figure 3. The riveting process: the pneumatic hammer (a);
the rivet heating (b); the head rivet forging with pneumatic
eter of the rivet is increased, thus filling the whole
hammer (c,d).
hole and, consequently, after the process no clear-
ance can be recognized between the shank and the
joined plates. Moreover, since the rivets are installed as process the enclosed plates are drawn together with
hot rivets, the rivet is heated to approximately 900◦ C installation bolts and by the rivet equipment. As the
before being installed, so characterized by a white- rivet cools, it shrinks and squeezes the connected plates
red colour (Figure 3b). Moreover, during the riveting together, thus resulting a residual clamping force or

432
internal tension results in the rivet. The direct conse-
quence of this phenomenon is its contribution to the
slip resistance of the joint just as do high-strength
bolts. However, the clamping force in the rivet is
difficult to control, is not as great as that developed by
high strength bolts, and cannot be relied upon. Upon
cooling, the rivets shrink diametrically as well as longi-
tudinally. The amount of shank-to-hole clearance that
results also depends on how well the rivet filled the
hole prior to shrinkage.
Having lightened the rivet process put in evidence
in which terms the Installation of hot-driven rivets
involves many variables, such as the initial or driv-
ing temperature, driving time, finishing temperature,
and driving method. Hence, it is of great importance
investigating in this field.
Moreover, a numerical study of the specimens to
be tested has been carried out. FE analyses have been
carried out involving a number of complexities, such
as material and geometric nonlinearities and contact–
separation modelling. In particular, in order to properly Figure 4. The plates sampled from Dome of Umberto I
plan the experimental tests the numerical activity has Gallery, before and after the superficial smoothing down:
been primarily concerned with predicting the ultimate 6 mm thick plate (a,b); 3 mm thick plate (c,d).
load-carrying capacity. After testing, the main objec-
tives of the FE analyses will be to closely examine
the performance of the specimens and to obtain, on
the basis of the static tests performed. In particular,
the load–deformation characteristics of the connec-
tions and the nature and magnitude of stresses resulting
in the different components are studied. In this phase,
the effect of clamping force in the rivets due to the
pre-stressing of the rivet processing on the response
of the connections has been also examined.

2 EXPERIMENTAL ACTIVITY
2.1 Tests on materials
In order to characterize the material properties of
ancient riveted connections, several material spec-
imens have been sampled from the Dome of the
“Umberto I” Gallery and from a dismantled railway
bridge dated from 1894 (coeval to the Dome of the
“Umberto I” Gallery). Moreover, tests on steel plates
and steel rivets having the same nominal properties of
the riveted connection specimens have been performed
(D’Aniello et al. 2007).
In details, two plates have been sampled from the
Dome: a 6 mm thick plate sample from the main arch
of the Dome and 3 mm thick plate from the internal
ring at the Dome basement (respectively Figures 4a,b
and 4c,d). For these plates, the Brinnel hardness tests,
chemical analyses and axial tensile tests have been car-
ried out. The hardness tests showed an average value
of 158 Brinnel index for 6 mm plates (with a standard
deviation of 5,01) and 160 for 3 mm plates (with a
standard deviation of 3,62).
Figure 5. The stress-strain response of plates sampled from
Dome of Gallery Umberto I.
The chemical analyses revealed a carbon percentile
lesser than 0,1%. In fact, in the plate of 3 mm of thick-
ness the carbon content is about 0.085%, while it is
0.045%.in case of 6 mm thick plate.
The mechanical behaviour of these plates is illus-
trated in Figure 5. In particular, the tests results
revealed that the 6 mm thick plate does not show a clear
yielding, with a brittle behaviour (maximum strain of
6‰), while a more ductile behaviour was shown by
the 3 mm thick plate (maximum strain of 10%).
The plates and the rivets constituting the material
of riveted specimens, that will be the core of the ongo-
ing study, have been obtained by materials stored by
RFI and they are dated from the ’60s of last century
(hence, coeval to Bridge on Gesso River). In particu-
lar, rivets were made from bar stock by hot- forming
the manufactured head. The heads were made of the
high button-type.

433
Figure 6. Coupon sampled from a plate of riveted specimen
under testing (a); rivet coupon under testing (b).

Figure 8. Dog-bone rivet shank.

Figure 7. The stress-strain response of plates of riveted

specimens.

At this time the materials of 10 mm thick plates and

rivets having diameter of 16, 19 and 22 mm have been
already tested (Figure 6).
The stress-strain responses of steel plates are shown
in Figure 7. The average yield stress was 291 MPa
(measured at 5‰ strain), while the average ultimate
stress was 433 MPa.
In case of rivets, it was necessary to mill the rivet
shanks as shown in Figure 8, arranging the shank Figure 9. The stress-strain response of rivet specimens.
shape in a dog-bone manner. Both rivet ends have been
screwed-thread and two cylindrical threaded sleeves
have been used to fix the specimens into the test 2. Tests on steel riveted joints made up with plates and
machine. The stress-strain rivet responses are illus- rivets gained from warehouse of Italian railways
trated in Figure 9. As it can be noted it is not possible and characterized by geometric properties similar
to recognize a clear behavioural trend per rivet diam- to those of the ancient riveted connections.
eter. The average value of yield stress was 315 MPa, The former tests are of great interest. In fact, because of
the average ultimate stress was about 412 MPa, while the impossibility to sample riveted connections from
the average ultimate strain (corresponding to necking) the two study case structures (the gallery and the bridge
was 16%. above illustrated), a riveted joint sample has been taken
of a dismantling railway bridge which was built in
2.2 Tests on riveted connections the same historical period (the bridge was erected in
1894, while the Gallery in 1885). This element, shown
Several lap shear tests have been planned in order
in the original configuration in Figure 11, has been
to study the mechanical behaviour of riveted connec-
cut in order to obtain several single rivet connections
tions. In detail, it is possible to subdivide this activity
(Figure 11a) to be tested under shear loading. In par-
into two phases:
ticular, some steel plates will be welded to the original
1. Tests on ancient joints sampled by a real structure end-plates to get easy the testing (in Figure 11b the
coeval to the “Umberto I” Gallery; welded new plates are drawn in red). Moreover, the

434
Figure 10. The railway bridge (a) and the sampled riveted
joint (b, c, d).

Figure 11. Specimens obtained from the bridge riveted

joint.

trapezoidal plate has been cut in such a way to have

two dog bone samples to be tested. These specimens
will give us valuable information about the chemi-
cal composition and mechanical properties of material
and about the behaviour of an ancient single rivet
connection.
The second set of riveted specimens has been Figure 12. Manufactured riveted specimens to be tested.
designed and manufactured in order to characterize
the mechanical behaviour of ancient metallic riv-
eted connections, namely investigating on the influ-
Three specimens have been built up for every type of
ence of different geometrical properties on the shear
joint.This choice was due to the fact that similar riveted
capacity of lap shear connections. Hence, different
connections can show a different capacity response,
rivet configurations will be tested; in particular, the
because they can be affected by the hand made riveting
specimen typologies are summarized as follows:
process (Fisher & Yoshida 1969). Moreover, all speci-
1. One rivet joint; mens have been designed considering both symmetric
2. Single row rivet joint. and asymmetric lap plates, as shown in Figure 12.

435
3 PRELIMINARY NUMERICAL MODELLING

In order to plan and design the experimental activity,

several numerical analyses have been performed. The
purpose has been to estimate either the minimum and
maximum load to be applied and, on the other hand, the
out of plane displacements of asymmetric specimens.
These preliminary results will give precious informa-
tion in order to organize in a proper manner the test
setup, giving an esteem of the maximum load to be
applied by the test machine. Indeed, the knowledge
of ultimate displacements let to properly position the
measurement devices (that will be LVDT in this case).
Moreover, these initial numerical models are to be con-
sidered as the starting step of an essential theoretic
activity finalized to deepen the knowledge of riveted
lap joints’ structural behaviour. Therefore, referring to
Figure 12, four riveted specimens have been initially
analyzed. In details, they are the specimens:
a) U16-10-1; Figure 13. Modelling assumptions: implemented contact
b) S16-10-1; conditions.
c) U22-12-4;
d) S22-12-4.
Numerical analyses have been carried out by means
of the finite element program ABAQUS 6.5 The
8-node brick continuum element C3D8R, with 8 nodes
per element, 3 degrees of freedom per node and a lin-
ear interpolation function was adopted for modelling
both rivet and plates. All degree of freedom of the end
portion of the plates were restrained in order to repro-
duce the actual boundary condition during the test.
The material stress-strain relation-ships for the rivets
and plates have been obtained starting from the rele- Figure 14. Mesh thickening in the parts in contact.
vant experimental test on materials (shown in Section
2.1.). In particular, the average experimental stress- riveting. The magnitude of the residual clamping force
strain curve has been converted into the true stress-true estimated assuming an equivalent shortening showed
strain relationship, with a plateau corresponding to the the attainment of a clamping force that approaches the
ultimate strength ( fu ). The plasticity behaviour was yield load of rivets. Finally, the load pattern has been
based on the Von Mises yield surface criterion. Large simulated by applying a relative displacement between
deformation effects have been considered. the two opposite terminal ends of each connected
In particular, it has been introduced the following plates, as shown in Figure 15.
contact conditions: (a) at interface between the plates The load versus in-plane deformation behaviour of
(Figure 13a); (b) between the internal surfaces of the lap joints is summarized per investigated models in
hole and rivet shank (Figure 13b); (c) between the rivet Figure 16. This plot interestingly shows that in case of
heads and the external plate surfaces (Figure 13c,d). smaller specimens, the shear strength increase propor-
After a preliminary sensitivity study of the mesh, in tionally to the number of shear planes of the splice
order to minimize numerical discrepancy due to the (in this case two times corresponding to two shear
contact boundary conditions, it needed to thicken the planes). In case of stronger specimens, the final shear
mesh sizing, making refined partitions of plates and capacity of both symmetric and asymmetric splices
rivet as shown in Figures 14a,b. is almost the same. In fact, the failure seems to be
Moreover, the Coulomb friction model (with a fric- governed by the yielding of steel plates.
tion coefficient of 0.3) has been considered in all In details, the predicted collapse mechanism of the
contact conditions. In addition, the rivet pre-stressing specimen U16-10-1 is shown in Figure 17. As it can
has been introduced by means of an imposed rela- be observed, the shear failure concentrated in the rivet
tive shortening of the rivet shank, so simulating the shank and the effect of the flexural deformation due
shank shrinking due to the cooling process after the to the eccentricity of the applied load is evident and

436
Figure 15. Simulation of the experimental load pattern.
Figure 16. Load vs. in-plane displacement of examined
riveted specimens.
Figure 17. Predicted collapse mechanism of U16-10-1
(amplified deformed shape).
437
Figure 18. Predicted collapse mechanism of S16-10-1.
confined to the regions where plate discontinuities
occur.
Figure 18 summarizes the failure mechanism of the
symmetric splice S16-10-1. Contrary to the previous
case, the prevalent collapse mechanism is the heading
of the inner plate, essentially due to the shear load
concentration.
Figure 19 shows the failure mechanism of the sym-
metric splice U22-12-4. In this case it is evident the
plastic engagement of the connected plates due to the
out-of plate displacement induced by the secondary
bending. In this case, the collapse is mainly due to the
stress concentration in the zone of plates where the
geometrical discontinuity occurs.
Finally, Figure 20 highlights the failure mechanism
of the symmetric specimen S22-12-4. Contrary to the
previous case, no secondary bending obviously occurs
and, like the S16-10-1 specimen, the prevalent collapse
mechanism is, again, the heading of the inner plate,
essentially due to the shear load concentration.
Generally speaking, it can be affirmed that in case of
lap joints unrestrained against out-of-plane displace-
ments the joints showed considerable deformation due
to the eccentricity of the load. It is evident that the
effects of bending are mainly confined to the regions
where plate discontinuities occur. Obviously, as the
joint length increases, bending will become less pro-
nounced, and the influence on the behaviour of the
connection should decrease (similarly to that high-
lighted by Shoukry & Haisch, 1970). The influence
of bending is most pronounced in a splice with only
a single fastener in the direction of the applied load.
In such a joint the fastener is not only subjected to

Figure 19. Collapse mechanism of U22-12-4 (amplified
deformed shape).
Figure 20. Collapse mechanism of S22-12-4.
438
Figure 21. Out-of vs. in-plane displacement of examined
riveted specimens.
single shear, but a secondary tensile component may
be present as well. Furthermore, the plate material in
the direct vicinity of the splice is subjected to high
bending stresses due to the load eccentricity. Hence,
the bending tended to decrease slightly the ultimate
strength of short connections. The shear strength of
longer asymmetric lap joints seems to be less affected
by the effects of bending.
Figure 20 shows the predictive numerical response
in term of out-of plane vs. relative in-plane displace-
ment for each modelled unsymmetrical specimen.
From these preliminary numerical analyses, it seems
that the influence of bending is most pronounced in a
splice with only a single rivet in the direction of the
applied load. However, this has little influence on the
load capacity, since the material will strain-harden and
cause yielding on the gross area of the connected plate,
as it can be observed comparing the joint capacity to
the hand calculation. Moreover, it was found that the
bending stress component varies significantly within
the splice region. This may be attributed to the large
stiffness variation caused by the rivets.
4 CONCLUSION
Structural integrity assessment of historical metallic
structures can be considered one of the main issue in
the field of structural engineering. Most of these struc-
tures were built at the end of the 19th century or the
beginning of the 20th century with angles and plates
joined by rivets and made of puddled iron or wrought
steel. Many of them are still in service, but only a
few are damage free, usually due to the lack of sys-
tematic maintenance and/or to structural deficiencies.
To assess the safety level of these old constructions
under current traffic loads, the bearing capacity of their
structural members must be reliably estimated, paying
special attention to the connections. In fact, after a
study about the main causes of decay and structural
inadequacy of ancient metallic large span structures,
it was recognized that a significant source of structural
deficiency could be found in the old rivet connec-
tions. Therefore, this paper deals with experimental
and numerical works the objective of which aims at
deepening the structural behaviour of riveted connec-
tions. In fact, a wide experimental activity on historical
riveted connections has been planned and, here, it
was summarized and presented. The first experimental
results on materials have been summarized. Moreover,
the results of the predictive numerical analyses per-
formed on some representative specimens have been
shown.
ACKNOWLEDGMENTS
The following subjects are gratefully acknowledged:
– PRIN prot. 2005087058_004 “Vulnerability and
reversible consolidation techniques for historical
metal structures” for having provided its financial
support;
– PROHITEC project “Earthquake Protection of
Historical Buildings by Reversible Mixed Tech-
nologies”, co-ordinated by Prof. F.M.Mazzolani;
439
– Eng. Antonio D’Aniello, Department director of
the Italian railways (RFI), to the cooperation and
assistance in the manufacturing of riveted connec-
tion specimens and for giving the opportunity to
study the railway bridges.
REFERENCES
ABAQUS, Inc. User’s Manual. ABAQUS Standard V6.5.
ABAQUS, Inc.; 2005.
D’Aniello M., Fiorino L. & Landolfo R. (2007). “Riveted
connections in historical metal structures: analysis and
testing”. Proceedings of XXI C.T.A. Conference, Catania
(Italy) 1–3 October.
Fisher J. W. & Yoshida N. (1969) “Large Bolted and Riv-
eted Shingle Splices,” Journal of the Structural Division,
ASCE, Vol. 96, ST9, September.
Guerrieri M.R., Di Lorenzo G. & Landolfo R. (2005). Influ-
ence of atmospheric corrosion on the XIX century “iron”
structures: assessment of damage for Umberto I Gallery
in Naples. Proceedings of XX CTA Conference, Ischia
(Italy) 26–28 September.
Shoukry Z. & Haisch W. T. (1970). “Bolted Connections with
Varied Hole Diameters,”Journal of the Structural Division,
ASCE, Vol. 96, ST6, June.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Diagnostic tests and numerical simulations for the preservation of

two stone stairways in the historic centre of Genoa (Italy)

A. Brignola, A. Del Grosso, S. Podestà, S. Resemini & G. Riotto

DICAT, Department of Civil, Environmental and Architectural Engineering, University of Genoa, Italy

ABSTRACT: The study focused on the experimental investigations and numerical analyses regarding the
staircases in the interior of two historical buildings (named Caffa and Metellino) in the refurbished area of the
docks of Genoa harbour (Italy). Those stairways are made up of stone elements (pink granite of Sardinia Island)
and show a particular constructive technique that allows the monolithic steps to be linked together, providing an
overall behaviour among steps of a flight and among flights. This technique, suggested by the prescription (or
rule of thumb) of various ancient building manuals, allows the stress field to be distributed on each element and
it was modelled by means of FEM, in order to verify the effect of this technical measure relative to various load
cases. The numerical simulation, carried out through a detailed non-linear solid model of the flights of steps,
was preceded by an exhaustive diagnostic campaign (in situ and laboratory tests), useful in order to define, both
through direct and indirect tests, the mechanical parameters to be adopted. The obtained results were crucial
to preserve the staircases without performing any structural intervention that would have been useless (if not
harmful), modifying the architectonical and historical value of the original structure of the stairways.

1 INTRODUCTION analysis is needed, because no simplified models seem

The recent renewal of the dock area of Genoa har-
bour (Italy) led to the renovations of various historical
buildings used as industrial warehouses. The interven-
tions determined the need of verifying the existing
structure reliability, with reference to the new end
use and the current technical building rules, aiming
to preserve their architectonical features.
In this study, the work focused on the experimen-
tal investigations and numerical analyses regarding
the staircases in the interior of two important build-
ings (named Caffa and Metellino) in the upgraded
area. Those historical stairways are made up of stone
monolithic steps.
The constructive technique, suggested by the pre-
scription of various ancient building manuals (de
Belidor 1729, Breymann 1885) and implying a partic-
ular moulding on each step edge, leads to the structural
interaction among the elements. In this way, an overall
behaviour of the flights may be achieved.
Nevertheless, current technical building rules do
not take into account this constructive technology
(monolithic stone elements) for stairs. So, no clear ref-
erence is given to the material behaviour that has to be
assumed or safety checks which have to be satisfied.
Moreover, even if the constructive details are rec-
ommended by ancient rules of thumb and they are
effective from the static point of view, a detailed
to well reproduce the effect of this technical measure.
The numerical FEM simulation, carried out through
a detailed non-linear solid model of the flights of steps,
was preceded by an exhaustive diagnostic campaign,
useful in order to define, both through direct and indi-
rect tests, the mechanical parameters to be adopted
and the presence of crack patterns or defects in each
single step.
The obtained results were crucial to preserve the
staircases without performing any structural interven-
tion that would have been useless (if not harmful),
modifying the architectonical and historical value of
the original structure of the stairways.
2 THE HISTORICAL DOCK AREA OF GENOA
AND ITS RENEWAL
Since the 13th century, the city of Genoa (Italy) was
already a thriving maritime township in the north-
western Mediterranean Sea.
The historical dock area (“Darsena”) of the Old Port
was built from the 11th century and it was subjected
to various modifications and re-organizations during
the ages, in relation to the evolution of the port and the
town. Since the first of 14th century, the “Darsena”
was organized in order to handle the galleys of the
Genoese Republic; moreover, there are warehouses for

441
Figure 1. The historical dock area in the Old Port of Genoa.
Figure 2. “Quartieri” Caffa and Metellino (Bordoni, 1987).
the galley repair (“Darsena delle galere”) and storage
of goods (“Darsena del vino” or “delle barche”).
In the Renaissance age, during the restoration of the
naval dock yard of the Genoese Republic, a tract of the
sea water was filled in and the quarters (“Quartieri”)
grew up on the new land (Fig. 1).
These are distinct large buildings of a significant
architectonic value, due to the excellent expertise in the
design of the architects and the skilfulness in building
of the craftsmanship.
In the very inside of the docks, the Quartieri Caffa
and Metellino were built (Fig. 2). These two twin
palaces are connected to each other through a huge
glass gallery (Fig. 3), which the main part of their
appeal is ascribed to. The morphology of the two build-
ings is similar, but significant differences, which the
peculiarity of each palace is due to, may be noted. In
particular, the stairs compartments, being key-points
for the ancient end use of the buildings (storehouses)
and are object of this study, show the same struc-
tural typology, typical of Genoese architecture (Franco
1995), but they are differently located in plan (Fig. 4).
In the last decades, thanks to various funds
obtained by Genoa city in relation to important events
(Columbus celebrations in 1992, G8 meeting in 2001,
European Capital of Culture in 2004), the area of
442
Figure 3. The glass gallery between Caffa and Metellino
palaces (Bordoni, 1987).
Figure 4. Plan view of Caffa and Metellino palaces.
the historical docks was deeply modified. Nowadays,
the Faculty of Economics and the “Galata Maritime
Museum” are based in the old quarters; from 2004,
Caffa, Metellino and Tabarca buildings were restored
and they currently house a contemporary music centre
(studios, auditorium, etc.) and contemporary-art com-
plex (devoted to permanent exhibitions, artist atelier,
bookshop and other shops), as well as a civic cen-
tre, having a multifunctional room for exhibitions and
meetings, to be used by local associations.
The modification of their end use led to a com-
plete verification of the static behaviour of the pre-
viously mentioned palaces. In particular, this work
may be included into this ambit, among which the
main staircases of Caffa e Metellino palaces are under
investigation.
3 THE STAIRCASES OF CAFFA AND
METELLINO PALACES

The staircases of Caffa e Metellino palaces are made up

of stone elements (pink granite of Sardinia Island) and
show a particular constructive technique that allows
the monolithic steps to be linked together, providing
an overall behaviour among steps of a flight and among
flights.
The lack of indications, in current technical build-
ing rules, about how to model and verify the static
behaviour of stairs made up of monolithic stone
elements rules led to perform in depth analyses.
Numerical FEM simulations through a detailed
non-linear solid model of the flights of steps and
an exhaustive diagnostic campaign, aimed to char-
acterize the material mechanical parameters and to
identify the presence of defects in the stone elements,
were carried out. Besides that, an accurate historical
investigation about the ancient building technology in
case of monolithic stone stairways seemed to be very
important.
Figure 5. Monolithic stone stairs: examples of step profile –
3.1 Historical and technological aspects reprinted from Breymann (1885).
In many ancient building manuals, technical char-
acteristics and constructive details of the monolithic
stone stairways are described. Generally, they are
made up of large freestone elements. In his treatise
“Wooden, stone and brick stairs” (1885), Breymann
wrote “Using freestones, one may obtain very strong,
durable and nice stairs and also very easy to build”. In
a few words, he highlighted the main features of this
typology. In Figure 5, the original drawings are shown.
The section shapes of the step are various: start-
ing from the simplier rectangular profile, up to more
elegant ones, characterised by moulding, carving or
special intrados outline, aiming to give a more aesthet-
ical sight from below, but also to facilitate the flight
construction. Figure 6. Lateral view of the stairs: moulding.
The way through which the steps are laid one upon
the other depends on the support typology of the step
edges: they can be fixed in the masonry walls on each may be identified in the 6 cm trim shown in Figure 5,
side or they can be simply supported by beams, walls image 467.
or arches. The geometry of the staircases in the interior of
According to Breymann, in case of fixed supports Caffa and Metellino palaces is ascribable to this last
of the edge, having at least an insertion depth equal to typology. However, in the examined stairs, the super-
10–12 cm, the step overlapping could be only nec- position of the shaped steps, regards elements having
essary to hide the back of the element and ensure their edge fixed in the walls for about 20 cm (Fig. 6).
the laying clearance; the overlapping distance may So it can be presumed that those structures, rely-
be minimal (4–6 cm, as in Figure 5, image 466). If, ing on both the housing in the masonry walls and the
on the contrary, one or both the edges of the step overlapping of the steps through the moulding, show
are free-standing, the element superposition is needed a good behaviour in relation to loads applied on them.
to avoid relative displacements among them. In this According to what specified in the “Traité théorique
case, an almost 2 cm moulding has to be trimmed. The et pratique de l’art de bâtir” (Rondelet, 1802), the
shape can be various: the profiles in Figure 5, images behaviour of the open staircases, that are stairs only
464 ans 465, aim to avoid the movement of the step, supported by the moulded shape of the steps, is inde-
by means of a of 2–3 cm backing. The same scope pendent from the presence of a beam propping up

443
Table 1. Geometry of the stone step.

Dimensions

Total height H (m) 0.19

Total depth B (m) 0.39
Moulding height h (m) 0.05
Moulding depth b (m) 0.05
Total length (m) 3.4
Free span L (m) 3.0

the parapet: “The step chase along the walls and

the landing constraint [at turn of stairs] provide,
together with the moulding and the superposition, an
ingenious system that stands up well without beams
. . . [omissis] . . . It is essential observing that in stairs
without beams, when the mouldings are not sufficient,
the minimal movement can lead the step to rotate and
slide out of the mouldings if: 1st their chase in the wall
is not sound, 2nd it does not have a sufficient depth,
3rd as a consequence of the movement, the steps fail
along their span. The beams added to the stairs, in
any way they are built, have the advantage of fixing
the steps at each end, preventing them to slide out of
the mouldings, being constrained by the wall, on one
side, and by the beam on the other”.
Therefore, from what Rondelet suggested, the gran-
ite staircases of Caffa e Metellino palaces are in
agreement to the rules of thumb. Moreover, stand-
ing that the elements are fixed in the walls on both
sides, the troubles described by Rondelet should be
presumably warded off.
Nevertheless, even if the constructive details are
recommended by ancient rules of thumb, more detailed
studies are needed to enforce that they are effective
from the static point of view. Figure 7. Overall view of the staircase (a) top; (b) bottom;
(c) Geometric model: lateral view – YZ plane.

3.2 Geometry of the staircase

The analyses focus on the staircases in the interior
of Caffa and Metellino palaces. Those stairways are
made up of stone monolithic steps and the previ-
ously described constructive technique allows these
elements to be linked together. In Figure 7a-b, an
overall view of the staircase is proposed.
The geometry considered for each single step is rep-
resentative of the most part of the stone elements. In
fact, each flight is sensibly similar in terms of dimen-
sions and very low scatter was found by the survey
analysis of the complete staircases in the two palaces.
On the safe side, the most conservative hypotheses
on the geometry were assumed. In Table 1, data of the
geometric model are reported, focusing on the dimen-
sions significant for the FEM model; the nomenclature
is in Figure 7c.
The free span of the step results from the detailed in-
situ geometric survey, while its total length derives by
a realistic hypothesis (enforced by examples of similar
constructive typology in other case studies) of a 0.2 m
deep insertion in the masonry walls on each side.
This aspect was however a posteriori partially con-
firmed by the information obtained by technicians
working on the building restoration, who directly
checked the considerable chase depth of the stone step
in the walls.
4 EXPERIMENTAL CAMPAIGN
The lack of indications, in current Technical Building
Rules, about the mechanical parameter of monolithic
stone elements led to perform detailed experimental
tests.
During renewal works, some slabs, with the size
of half a step, were found. Some of them were used
directly for bending destructive test.

444
Table 2. Results of compression tests.

Sample Diameter (cm) Length (cm) fc (MPa)

Slab 1/2 7.3 14.6 143.36

Slab 2/1 7.3 14.6 109.43
Slab 3/1 7.3 14.4 129.98
Slab 3/2 7.3 14.7 133.80
Slab 3/3 7.3 14.6 138.58
C1 7.3 14.5 96.05
C3 7.3 14.6 99.39
M1 7.3 14.5 100.35
M2 7.3 14.5 110.38
Mean 117.92
Standard Deviation 18.48
5% Percentile 97.39
Figure 8. Test setup according to UNI EN 12390-6/2000.

From the others, we obtained cylindrical samples

that were used to determine uniaxial compression Table 3. Results of indirect tensile tests.
strength, uniaxial tensile strength and density (it turns Sample Diameter (cm) Length (cm) fc (MPa)
out to be 2618 kg/m3 ).
As well as these partially destructive tests, sclero- Slab 1/1A 7.3 9.2 8.66
metric, direct and indirect ultrasonic tests were carried Slab 1/1B 7.3 8.6 9.36
out (non-destructive tests). Slab 2/2A 7.3 8.5 8.31
The extensive experimental campaign aims to Slab 2/2B 7.3 8.0 9.48
obtain a statistical characterization of mechanical C 2A 7.3 10.5 7.27
parameters and, possibly, to identify local defects in C 2B 7.3 8.3 7.99
M 3A 7.3 8.6 9.04
the steps.
M 3B 7.3 7.8 8.73
Considering the age of the structure and the mate-
rial, which the stairs are made up of (not commonly Mean 8.60
used in modern constructions), this kind of analysis Standard Deviation 0.74
was considered essential to preserve these attractive 5% Percentile 7.52
buildings in their original form and, at the same time,
to ensure a suitable safety level.
In Table 3, uniaxial tensile strength, diameter and
4.1 Test set-up and data processing length for each sample are shown.
For the sake of brevity, in the following, only the tests
4.1.3 Ultrasonic tests
that were used to define numerical analysis parame-
Ultrasonic tests provide a measurement for the “time
ters are reported. Setup, data processing and results of
of flights” of ultrasonic waves in the material. Hypoth-
these tests are described.
esizing the wave paths, the velocities may be obtained.
They were used to verify the structural element homo-
4.1.1 Compression tests
geneity, localize possible reduction of material prop-
The cylindrical samples were subjected to compres-
erties (e.g. cracks or cavity) and to estimate the value
sion test to identify the uniaxial strength. The failure
of Young’s modulus.
point is the loss of proportionality between the constant
During experimental campaign, two different type
load increment and the deformation.
of ultrasonic test were performed: direct test (i.e.
In Table 2, uniaxial compression strength, diameter
receiver and emitter are placed on two opposite side
and length for each sample are reported.
of the step) and indirect test (i.e. emitter and receiver
are placed on the same side of the step).
4.1.2 Indirect tensile tests (“Brazilian” tests)
In order to estimate the ultimate tensile strength, the 4.1.3.1 Direct ultrasonic tests
so-called “Brazilian tests” were performed, according Direct tests were carried out near the landing at turn
to UNI EN 12390-6/2000. These tests are based on of stairs, where the sensor can be placed on the
the fact that a cylinder subjected to a line load on two two opposite side of the steps. The resulting velocity
opposite diameters will break due to the orthogonal was processed to obtain the p.d.f. (probability den-
tensile strength (Fig. 8). sity function), as in Figure 9, and to estimate the

445
 1907 2907 3907 4907 5907 1586 2086 2586 3086 3586 4086 4586
0.0018 0.0014
0.0016 experimental data experimental data
0.0012
0.0014
lognormal distribution d = 140 cm lognormal distribution
normal distribution 0.0010 normal distribution
0.0012
0.0010 0.0008

0.0008 0.0006
0.0006 0.0004
0.0004
0.0002
0.0002
0.0000 0.0000

1586-1818

1818-2049

2049-2280

2280-2511

2511-2742

2742-2973

2973-3205

3205-3436

3436-3667

3667-3899

3899-4130

4130-4361

4361-4593

4593-4824

4824-5055
1907-2216

2216-2524

2524-2833

2833-3141

2833-3142

3449-3757

3757-4066

4066-4374

4374-4683

4683-4991

4991-5300

5300-5608

5608-5917

5917-6225

6225-6534
ultrasonic velocity (m/s)
ultrasonic velocity (m/s)

Figure 9. Probability density function of direct ultrasonic Figure 11. Probability density function of indirect ultra-
velocity of Metellino palace. sonic velocity, sensor distance 140 cm, of Caffa palace.

For each distance, the p.d.f. of the velocity was

obtained (Fig. 11) and the extreme values were ana-
lyzed in detail. It results that the state of preservation
of the stair is essentially homogeneous.
Thanks to this wide experimental campaign, the
results of the following numerical analyses can be rea-
sonably extended to the whole stairways of both the
palaces.

5 NUMERICAL ANALYSES

In order to assess the structural safety of the ancient

stone stairways of Caffa and Metellino palaces, a set
of numerical simulations, using the code ANSYS rel.
8.0, were performed.
The numerical analyses have a double role. On the
one hand, they are useful in evaluating stress and strain
fields in case of the loads prescribed by the Italian
Figure 10. Indirect ultrasonic test setup. Technical Building Rules (2005). On the other hand,
these simulations were carried out aiming to assess
the safety coefficient (in terms of load multiplier) with
dynamic Young’s modulus that turns out to be around
reference to live loads. Those may be due, for example,
30,000 MPa (mean value) with modest variability.
to crowding.
4.1.3.2 Indirect ultrasonic tests In order to come through the computational effort
Indirect ultrasonic tests were performed on every sin- required, the FEM model simulates an assemblage of
gle step of the two stairways, in order to obtain a wide six steps. Nevertheless, this is representative of the
set of values that may represent a significant statistical overall behaviour of one stairway.
sample. In particular, aiming to study the static behaviour
The “time of flight” was obtained placing the of the steps in the middle and at the end of the stair-
receiver on the “first column” (the black one in Fig- case, it was checked that adding other elements does
ure 10) and moving the emitter in the other three not influence (conceptually and quantitatively) the
different positions on each step. behaviour.
The processing phase of these tests is a little more
complex then the direct ultrasonic ones, due to diffi-
5.1 General description of the FEM model
culty in the estimation of the wave path. For this reason,
it was decided to divide the data in three different sets, The numerical simulation, preceded by an exhaus-
one for each horizontal distance between the sensor tive diagnostic campaign and geometrical survey, is
(i.e. 70 cm, 140 cm, 210 cm). In this way, the possible particularly useful when dealing with the historical
error in the path estimation was minimized. heritage.

446
Table 4. Element types in the FEM model.
Constitutive element FEM element No.
Stone step SOLID65 33480
Interface CONTACT52 1150
In this framework, the development of very accurate
models is generally required, both in case of service
loads (for which cracks formation may represent a
condition to avoid) and ultimate limit state. In fact,
non-linear analyses may give information, by means
of incremental analysis, about the collapse, but the
models should be able to reproduce the limited ten-
sile strength and possible crushing of stone material,
the frictional behaviour on the contact surface of the
moulding.
To this aim, the modelling was carried out through
a detailed non-linear solid model of the flights of
steps, focusing on effectively describe the interface
behaviour.
Brick elements (SOLID65) are used to model the
stone material of the monolithic steps. The frictional
behavior on the contact-planes (horizontal and verti-
cal) of the moulding is simulated by CONTACT52
elements.
In Table 4, the basic information about the FEM
model, made up of 34630 elements and 41124 nodes,
are provided.
The boundary conditions of FEM model are
intended to simulate two situations, related to differ-
ent hyphotesis of interaction between the step and the
masonry walls:
– Model A: displacements along Y and Z axes are
avoided on the bottom surface of the steps, in corre-
spondence of the insertion surface of the step (0.2 m
in depth) in the wall.This condition aims to simulate
structural fixed ends (Fig. 12).
– Model B: displacements along Y and Z axes are
avoided on the bottom of the steps, in correspon-
dence of the chase of the step in the wall. A line of
nodes only (0.2 m in distance from each ends of the
step) is constrained, in order to enable the rotation,
aiming to simulate structural hinges.
The actual condition is intermediate between the
analysed ones; due to the chase depth in the walls, it
is probably better described by Model A.
Moreover, the steps at both ends of the flight need
to be realistically constrained. On the bottom one, the
vertical support of the masonry arch below (Fig. 7b)
and the horizontal constraint of the floor have to be
simulated. On the top step, instead, horizontal con-
straints are needed to model the interaction of the step
and the landing at turn of stairs, made up of stone slabs
having the same thickness. All this kinds of constraints
behave obviously as compression-only elements.
447
Figure 12. FEM model: fixed ends boundary condition.
5.1.1 Mechanical parameters
The pink granite stone mechanical characteristics are
derived by the experimental test campaign previously
described.
In the FEM simulations, in the elastic range, a lin-
ear isotropic constitutive law was assumed. This kind
of behaviour can be appropriate to satisfy the require-
ments of the safety check, serviceability and ultimate
limit state, of the Technical Rules (2005).
The values of the mechanical characteristics
assumed correspond to the average of the statisti-
cal distribution of each quantity (in case of Young’s
modulus E = 30,000 MPa, the value is derived by the
experimental tests on the stone of the Caffa palace).
Moreover, the non-linear behaviour is modelled
through the CONCRETE model (enclosed in the
ANSYS 8.0 library), whose constitutive law simulates
the limited tensile and compressive strength of stone
material. The failure surface in the stress domain is
represented by a law that depends on the hydrostatic
stress, as Mohr-Coulomb or Drucker-Prager ones.
Cracking along three orthogonal axes and crushing
are allowed and five failure-surface parameters have
to be assigned. If only two (ft and fc ) are set, the failure
surface is defined as in Willam and Warnke (1975).
The non-linear parameters in terms of strength
values are obtained as the 5% percentile of the stochas-
tic distribution of the experimental results. In the
model, the uniaxial tensile strength is set ft equal to
7.52 Mpa for and the uniaxial compressive strength fc
to 97.39 Mpa.
The stiffness parameters of the contact element
on the interface of the step moulding are assumed
to simulate the interaction effect (equivalent Young’s
modulus equal to 30,000 MPa). Being difficult to have
an estimation of these quantities, it was verified that
their variation among a pre-defined range does not
significatively affect the results.
The frictional coefficient on the contact surface is
set equal to 0.5.
5.2 Scheme of the numerical simulations Table 5. Values of maximum (σI ) and minimum (σIII )
principal stresses in Models A and B.
Two analysis phases were carried out: in the first one,
the load cases prescribed by the technical rules (ulti- MODEL A MODEL B
mate and serviceability limit state) were imposed and,
in the second one, the behaviour related to incremental σI 1.03 MPa 1.76 MPa
loads was studied. σIII 1.88 MPa 2.67 MPa
Phase 1: on the two models (A and B), the following
analyses were performed:
LC1 (ULS): 1.4 Gk + 1.5 Wk1
LC2 (SLS): Gk + Wk1
where Gk is the dead load and Wk1 is the live load
(equal to 4 kN/m2 in case of common stairs);
– Phase 2: on the two models (A and B), the analyses
carried out concerned the monotonic increase of
two load typologies (uniform pressure on every step
and transversal line load on the central step), after
the dead load assignment.
It has to be noted that, in case of Phase 1, the
behaviour is very far from the inelastic range; so, it
was verified that the action assignment in two subse-
quent stages or the concurrent loading did not lead to
different results (i.e., the effect superposition is still
correct).
In both cases, live loads were assigned to the
walking surface of the steps.

5.3 Analysis results

In the following, the main results obtained from the
numerical simulations are discussed. Special emphasis
is put on the evaluation of interaction effects among
the steps.

5.3.1 Phase 1: technical rule requirements

As previously noticed, the analyses results may be
useful in order to obtain some information about the
serviceability and structural safety of the stairways,
according to Italian Technical Building Rules.
In case of SLS, the mid-span vertical displacement
is maximum in the central steps of the flight: with
reference to Models A and B, it is equal to 0.06 mm
and 0.07 mm respectively. Those values are both much
lower than the limit displacement L/400 (7.5 mm).
In case of ULS, the stress field in Model B is higher.
The extreme values (in terms of average nodal stresses)
are reported in Table 5. Figure 13. Model B – USL stress state. Direction of
From these results, it can be highlighted that the maximum (a–b) and minimum (c–d) principal stresses.
stress state is much lower than the material strength.
This is much more significant if one considers that
those values are concentrated in very narrow areas, structural performance may be obtained (Fig. 13).
near the constraints: the safety evaluation is amply Checking the direction of minimum principal stress
conservative. σIII , in the the mid-span section, some sort of strut,
Focusing the attention on longitudinal sections of due to the transfer of compressive forces through the
steps, corrisponding to the portion near the supports steps, may be identified. This phenomenom appears,
and the mid-span one, interesting remarks about the thanks to the moulding, in both the models. Along the

448
Figure 14. Cracking pattern in case of Wk1 = 2 kN/m2 :
axonometric view from above (a) Model A; (b) Model B.
lateral section near the supports, the behaviour seems
instead to be more affected by torsional effects.
The strong interaction among the steps of one flight,
revelead by the formation of the strut in the numerical
simulations, confirms once more the static intuitions
of the ancient builders and engineers.
5.3.2 Phase 2: incremental analysis
After the dead load assignment, two load typologies
(uniform pressure on every step and transversal line
load on the central step) were monotonically increased.
5.3.2.1 Uniform pressure
With reference to uniform live load, even the nominal
action amplified more than 15 times does not cause
significant damage pattern in the steps. In case of Wk1
equal to 72 kN/m2 , only slight cracking phenomena
develop near the supports (Fig. 14).
In both the models, crushing states are not evident
and, in fact, during the loading phase, the principal
stress σIII has not reached the surface of the fail-
ure domain. The higher obtained values for σIII are
12.6 MPa (Model A) and 22.9 MPa (Model B). Even
449
Figure 15. Model A, stress state in case of Wk1 = 72 kN/m2 :
(a) maximum principal stress σI [N/m2 ]; (b) minimum prin-
cipal stress σIII [N/m2 ].
in this case, the stress peaks are located near the
constraints (Fig. 15).
Observing the failure pattern, it can be noted that,
in Model A (as Rondelet wrote), the described con-
structive technique leads these elements to be more
effectively linked together.
The force transfer through the moulding is more evi-
dent in respect to the other model, in which the rotation
of the ends of the step is allowed. In fact, the lower
steps are more loaded in compression in correspon-
dence of the mid-span point than the upper ones; this
force transfer lead the vertical planes in correspon-
dence of the fixed ends to be subjected to high tensile
stresses and cracking develops in the lower part of the
flight.
Further increase of the uniform pressure is not
investigated, because the hypothesis of fixed con-
straints at the ends of the step (modelling the effect
of the masonry walls) would have been not totally
correct. In fact, higher loads in the masonry could

Figure 16. Model A – line load. Direction of minimum
principal stress.
have led to local crushing and stiffness degradation
of the supports. This condition has not permitted the
investigation of the stair behaviour for the highly non-
linear range. Nevertheless, this kind of load seems not
to affect significantly the safety of the structure.
5.3.2.2 Line load
Assigning the transversal line load on the central step
(along X-axis), some information about the non-linear
behaviour modification is achieved.
A final value of 7.5 kN/m is reached, pointing out
slight cracking located on the central step itself, near
the supports.
Also in this case, the highly non-linear range can not
be studied, due to the unrealistic boundary conditions
which further load increase would have led to.
In Model A, the force transfer through the moulding
is less evident in respect to the uniform load (Fig. 16).
In the mid-span section, the strut, due to the transfer
of compressive forces through the steps, is not identi-
fiable anymore. Moreover, even the torsion effect near
the supports cannot be highlighted.
From the result review, one can put forward that the
direction of the minimum principal stress is almost
orthogonal to these longitudinal sections. This may be
explained through the creation of a transversal line of
thrust along each single step (especially the loaded
one). In this way, the load is directly transferred to the
supports of the step, achieving only partially the overall
behaviour of the stairway. Nevertheless, the lower steps
seem to interact with the loaded one.
450
6 CONCLUSIONS
The significance of synthesis and comparison of data
from different ambits, in order to obtain a more
detailed and more realistic overview of the problem
of safety and conservation of historical structures is
evident.
This study provided an exhaustive investigation of
the two stone stairways of Caffa and Metellino palaces,
having the aim of preserving them without performing
any structural intervention.
Combining the elements from the in situ and labo-
ratory testing, from the historical studies and from the
numerical simulations, it was demonstrated that the
original conception of the granite monolithic stairways
is amply safe, even in relation the modification of their
end use. So, it is clear that no retrofitting intervention
is needed.
On the other hand, the contribution of FEM investi-
gations played a fundamental role in achieving another
goal: by means of non-linear incremental analyses, the
structural effect of an ancient constructive technique
(moulding and superposition of the steps of a flight)
was examined and understood.
The research confirms that, in this case, the static
perception of ancient builders and engineers is essen-
tially validated and it states once more that if a larger
part of the past knowledge would have been trans-
ferred, useless (and sometimes harmful) structural
modifications could be avoided.
REFERENCES
Breymann G.A. 1885. Wooden, stone and brick stairs.
Librerie Dedalo (eds.), Rome, 2003 (in Italian).
Bordoni S., Ighina A.G., Tuscano C. 1987. Steel and building
reuse: The “Darsena” of Genoa, a development design.
Busalla, Italsider Publication (in Italian).
de Belidor B.F. 1729. The engineers’ science by de Beli-
dor with notes by Mr Navier, Italian translation by Luigi
Masieri, Milan, Truffi, 1832 (in Italian).
Franco G. 1995. Halls and staircases in the Genoese Archi-
tecture, Book of Architectural technology – Collection by
Gianni V. Galliani, Grafica KC, Genoa, 1995, p. 126 (in
Italian).
Ministry of Infrastructures and Transportation, D.M.
14/09/05 – Building Technical Rules. Official Bulletin no.
222, 23 September 2005.
Rondelet J. 1802.Traité théorique et pratique de l’art de bâtir.
Italian translation by Soresina B., Vol. II, Librerie Dedalo
(eds.), Rome, 2004.
Willam K.J., Warnke K.D. 1975. Constitutive model for
the triaxial behaviour of concrete. Proc. of International
Association for Bridge and Structural Engineering, 19,
ISMES, Bergamo, Italy, p. 174.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Evaluation of iron and steel in historic bridges

S.P. Sparks
Sparks Engineering, Inc., Round Rock, Texas, USA

ABSTRACT: This paper presents a nondestructive approach for the evaluation of wrought-iron and steel in
nineteenth and early twentieth century bridges. Many historic bridges are non-redundant structures, and require
complete evaluation of both structural capacity and the condition of the members for continued service. The
conventional approach of material characterization in older structures has relied on tensile-testing of coupons
removed from a structure to determine strength values. That approach has limitations due to large statistical
corrections that are required when using only a few samples, and because the removal of coupons from critical
members is rarely feasible, and because each class of bridge member may have distinct mechanical properties.
An approach to in-place materials characterization is presented that is based on screening for low-ductility by
examining microstructure, chemistry, and hardness.

1 INTRODUCTION

There are tens of thousands of iron and steel bridges

that are considered historically significant and which
remain in service as part of our transportation sys-
tems. For bridges with structural redundancy that have
been in continuous service, with minor corrosion dam-
age, a visual examination may be all that is required.
But where redundancy is insufficient, significant dis-
tress or damage is present, or an increase in loading is
required, a complete evaluation of the condition of the
members as well as structural capacity may be required
for continued service.
Because the geometry and sections of the main
members are relatively simple to model, material char-
acterization becomes a key issue in the structural Figure 1. Upper chord pin-and-eyebar connection in an
analysis of historic bridges. 1881 Whipple Phoenix Truss. The eyebars and compression
The determination of material strengths for historic sections are wrought iron, the joint blocks are gray cast iron,
iron and steel bridges remains problematic, partic- and the pins are wrought iron on the upper chord and steel
ularly with the need to assure ductile behavior. For on the lower chord.
members and connections that are critical to the sup-
port of the structure, special methods may be needed century bridges built of iron and steel. In addition
to qualify their integrity. to visual assessment and structural analysis, the sug-
A rigorous evaluation is necessary because the gested protocol relies primarily on a combination of
materials are of unknown quality, the critical fabrica- materials characterization and nondestructive testing.
tions were never qualified by modern standards, loads The approach is based largely on available methods; it
have increased, and the effects of decay and fatigue represents the author’s search for a rational basis for
may have reduced the capacity. On the other hand, the evaluating complex structures in which the quality of
importance of historic bridges as significant works of the materials and their properties are unknown.
history requires that our evaluation methods be the It is the thesis of this paper that in the context of his-
least invasive. torical data and the observed condition of the bridge,
This paper suggests a nondestructive protocol microstructure, hardness and chemical analysis are
for the evaluation of nineteenth and early twentieth sufficient to characterize the behavior of the material.

451
The goal is to have an approach that in most cases
eliminates the need for physical sampling and testing
of bridge members, and where sampling is used can
help reduce the ambiguity that comes from a limited
sampling program.

2 DUCTILITY

2.1 Importance of ductility

Modern engineers are accustomed to looking for high
values of tensile strength an in indicator of a “good”
material. Generally, a high-strength structural steel is
viewed as necessarily superior to mild steel. This is
because in our advanced system of specification and
quality control, the reliability of the material is rarely
an issue. Sufficient ductility is assured by the specifi-
cation, and is therefore not a variable to be considered
in modern design. When encountering a historic struc-
ture, it is common to order tests aimed at finding out
the strength, and not much attention is paid to the other
properties of the material.
It is important to consider not only the strength
of the material, but also the ductility. Ductility is
essential in fracture toughness: it assures microscopic
crack attenuation, allows the material to tolerate minor
internal defects such as inclusions, and permits redis-
tribution of stresses at extreme loads. This results in
gradual, rather than catastrophic, failure. The resis-
tance of good wrought iron to brittle fracture can be
explained by the ability of the fibrous microstructure
to attenuate crack growth by delamination and crack
branching. But it is the ductility of the ferrite matrix
that enables the fibrous character to be beneficial
rather than detrimental (Schindler).
Early steels also exhibit a wide range of ductility.
In bridges, a low-strength ductile material is preferred
over a high-strength brittle material. In fact, for both
wrought iron and early steels, high strength should not
be considered an indication of high quality (Kirkaldy
1862). Moreover:
“. . . the real strength of a material is less
in its capacity to resist than in its abil-
ity to yield. . . The inelasticity of the material
determines . . . its structural performance con-
siderably more than the separation strength.”
(Freudenthal 1950)
Gordon (1988) showed that wrought iron (e.g.
Holley’s tests performed in 1877) possessed relatively
consistent tensile strength values, while exhibiting a
wide range of ductility, as measured by reduction
of area (Figure 2). The reduction of area is a bet-
ter indicator of ductility than elongation (Kirkaldy
1862, Freudenthal 1950). Similar data were found by
Buonopane & Kelton (2007) in tests from recovered
bridge iron.
Figure 2. Tests of wrought iron (Holley 1877) based on
Gordon 1988. While a lower bound for strength is easily
determined, there is a widely scattered range of ductility as
represented by reduction of area.
2.2 The problem with coupon testing
The conventional approach for evaluating material
properties in existing structures has relied on tensile-
testing of coupons removed from the structure. In
historic truss bridges, however, tests from coupon sam-
pling are not likely to be indicative of the structure as
a whole. There are several reasons for this:
1. Samples taken from non-critical members most
likely will not be representative of the material
in the critical members. A determination of phys-
ical properties must be made for each member
class In general, each truss member class (pin, eye-
bar, rolled section, plate, lacing, rivet) should be
assumed to have different material properties. Even
if all members are the same species, different grades
are likely to have been used in the various members.
2. A small number of samples may not produce mean-
ingful results because of the large statistical correc-
tions that are required to address the uncertainty in
the data (SCI 1997).
3. The removal of coupons from critical members like
eyebars is usually not possible because of the risk
of causing structural damage.
4. Bridges were often moved or rebuilt, resulting in
the mixing of metals from different eras.
5. Materials used in past repairs will be of unknown
type and quality, and will differ from the original
materials.
In addition, removal of historic fabric should be
avoided. Damage to the original materials or character-
defining features could diminish the integrity of the
historic structure.
Conservative values for the strength of iron and
steel can be found in AASHTO Manual of Condition

452
Evaluation (AASHTO 1994) and other references,
based on date of construction. These strength values
are appropriate for use in preliminary analysis. They
are necessarily conservative, and higher values may be
justified where indicators of low-ductility are absent.
Where low-ductility is suspected or where fatigue is a
risk, lower values should be used.
3 HISTORICAL DATA
A wealth of mechanical and chemical data has been
obtained over the past century and a half for cast iron,
wrought iron, and steel. The work of Kirkaldy, Holley,
and others stand out as milestones, and much addi-
tional work has been done since. A huge benefit to our
understanding can be gained by mining the past data.
Currently a survey of wrought iron test data, mod-
ern and historic, is being prepared at University of
Manchester (O’Sullivan 2007).
4 MATERIALS CHARACTERIZATION
The materials characterization approach relies on
microstructure, chemistry, and hardness as proxy data
in lieu of tensile test data. No single proxy test is fully
diagnostic, but neither is the tensile test for that matter.
Given the immense amount of historical data obtained
for iron and steel over the years, the proper role for
proxy data is not necessarily the direct prediction of
tensile strength, per se, but as a screen for bad behav-
ior, i.e. low ductility, and to relate the specific bridge
member to existing data for known materials.
This idea is not new. Researchers were attempting to
relate chemical analysis and microstructure to material
properties in the 19th century. What we need is a set
of proxies for which many data can be obtained easily,
rather than tensile tests, which are hard to obtain in
statistically meaningful numbers.
Microstructure, chemistry, and hardness are all con-
ventional tools for understanding metals, but they
developed as laboratory methods for quality control,
and their application to field evaluation has never
been consistently applied to determining strength and
ductility.
4.1 Microstructure
There are three species of ferrous metals that may be
found in historic bridges: cast iron, wrought iron, and
steel, and within each there are several varieties. The
first thing is to determine the species and the variety
of the metal. This can be done with no damage to the
structure by the spark test (Tschorn 1963) to qualita-
tively distinguish carbon content. The spark test uses
a high-speed grinder to produce a stream of sparks,
which exhibits specific characteristics depending on
chemical composition. For practical field use, it is best
453
Figure 3. Field metallography (100×, nital etch) of 1881
wrought iron eyebar showing normal ferrite matrix and finely
distributed slag.
to have a set of samples with known material type
and chemical analysis. Polishing a small area on each
member of interest will reveal the presence or absence
of slag.
Macro-etching is a method of preparing the surface
for visual or low-magnification examination, and is
appropriate for field identification of wrought iron,
and identification of gross characteristics and flaws,
such as weld, cracks, etc. Micro-etching is the polish-
ing of the metal with a very fine grit, and etching for
observation with a microscope. Etching is done with a
suitable reagent, such as nital. Oberhoffer’s reagent can
be used to show phosphorus segregation (the copper
in these reagents will deposit on the phosphorus-free
areas first) (Gordon 2005).
Field examination of microstructure is an effec-
tive way of identifying metal types, and of screening
metal for defects in chemistry or manufacture. It is
used to reveal characteristics and flaws, grain size, and
type. Wrought iron with satisfactory microstructure
(finely distributed slag, fine equiaxed ferrite grains,
and near absence of pearlite) and chemistry (limits on
C, P, and S) will have adequate ductility to remain
in service (Gordon 2005). Steel microstructure is well
described in the ASM Handbook (ASM 2004) and cast
iron guidance is available in Walton (1971).
4.2 Chemical analysis
Chemical analysis can assist the identification of
species and give an indication of quality and consis-
tency in the structure. Chemical constituents strongly
influence the strength, ductility, and weldability of
metals. In structural steels, the main determinants are
carbon (C), manganese (Mn), sulfur (S), and phospho-
rus (P). Sauveur (1912) correlated carbon content with
strength and ductility (elongation) for hypo-eutectoid
steels (below 0.84%C). In wrought iron, these same
chemicals determine many of the characteristics, but
the quantity and distribution of slag inclusions have an
equally important role.
In-place chemical analysis is now possible with spe-
cialized optical emission spectroscopy instruments,
commonly called PMI (Positive Material Identifica-
tion). One such instrument is the Arc-Met 8000 by
Metorex. This instrument is routinely used for accu-
rate in-place analysis of steel and cast iron. Given
proper access, it is possible to obtain dozens of
field data of chemical analyses in a single day, pro-
viding an excellent statistical basis for judging the
materials.
For wrought iron, however, the use of the PMI
instrument is problematic because of the presence of
slag, and the results may not be reliable. The operator
must obtain multiple ‘burns’ at the same spot, which
requires steady support for the operator and instru-
ment. Preliminary field results have been obtained
using reference samples of known chemical content for
comparison. However, further evaluation of the pro-
cess is required before the method can be considered
useful for field analysis of wrought iron structures.
Laboratory chemical analysis can be done on very
small samples, such as drilling swarf, but obtaining
even a small sample from a critical element is risky.
Use of a ball endmill bit will produce a smooth,
concave depression, rather than a hole.
It is essential, whether in the field or in the lab, to
use a type of instrument that can be calibrated to ana-
lyze the lighter elements, especially phosphorus and
sulfur, which are the key determinants in screening
for low-ductility. For steels, it is possible to make a
conservative estimate of fracture toughness based on
carbon content and comparison with chemically simi-
lar modern metals. As carbon content increases, the
ductile-to-brittle transition becomes more dramatic,
and occurs at higher temperatures.
One very important parameter for wrought iron is
the percentage of phosphorus contained in solid solu-
tion. Phosphorus increases hardness, tensile strength,
and brittleness at low temperatures. In wrought iron,
phosphorus is partitioned between both the ferrite and
the slag (Gordon 2005). Chemical tests usually report
the total phosphorus in the sample, and do not distin-
guish between the metal and the slag. Some historic
tests of wrought iron show phosphorus content about
equally distributed between the ferrite matrix and the
slag (Higgins 1934), though Aston & Story (1936)
give a representative distribution of 0.10% in ferrite vs.
0.02% in the slag. The effects of phosphorus, and its
real distribution, is an area of valuable further research.
Table 1 suggests values for screening wrought iron
and steel for ductility-reducing chemical concentra-
tions. Chemical contents above the maximum values
shown in the table are not in themselves cause for
rejection, but rather indicate the need for further
454
Table 1. Suggested screening values for chemical analysis
of wrought iron and steel.
C P S
Wrought Iron∗ 0.10 0.30 0.04
Steel
Pins∗∗ 0.20 0.05 0.04
Eyebars∗∗∗ 0.30 0.08 0.05
∗
Aston and Story 1936; and Gordon 2004.
∗∗
Campbell, 1896.
∗∗∗
Gayler 1889; Oberg and Jones 1918.
investigation, including metallography and hardness
testing.
In steels, manganese should normally be at least
six times the sulfur content. In wrought iron, the man-
ganese content is usually less than 0.10% by weight,
though it does not seem to have an adverse effect on
ductility.
Using these criteria, the author screened historical
data from 37 laboratory tests of wrought iron for which
chemical analysis and reduction of area were avail-
able (Holley 1877, Kaufman & Roberts 1990, Frank
1974, Sparks & Badoux 1998). Of the ten samples with
reduction of area below 25% (low ductility), seven
(70%) were caught by the screen. This was done on
chemistry alone without the aid of microstructure or
hardness data. Taking reported high levels of slag as
a measure of poor microstructure, the screen would
have rejected 90% of the low ductility samples. Of
the remaining samples having sufficient ductility, the
screen only rejected about 10%.
4.3 Hardness
The field hardness survey is an inexpensive, rapid test
that serves several purposes in a structural evaluation:
• To screen for insufficient ductility
• To indicate the variability of the material within an
individual member
• To correlate the properties of one class of members
with another
• To estimate the tensile strength of the materials
Quoting Freudenthal (1950, p. 539):
“The statistical information concerning the uni-
formity of the ‘ultimate tensile strength’ of the
material obtainable from a large number of
simple and rapidly performed hardness tests is
usually more relevant . . . than the limited infor-
mation supplied by a small number of tension
tests.”
Currently, several field hardness methods are avail-
able for field use on bridges including the Ultrasonic
Contact Impedance (UCI) method, used in instruments
Table 2. Suggested screening values of Brinell Hardness
Number (BHN) for different materials.
Typical range Maximum
Wrought Iron 95–120 130∗
Steel
Pins 120–140 145
Eyebars 100–120 140
Average of three tests on an area of the same member.
∗
Local values in areas of phosphorus segregation may reach
160 BHN.
such as the Krautkramer MIC10. Regardless of the
instrument used, the results should be presented in
Brinell Hardness Number (BHN), which is closely cor-
related with strength in carbon steels. As a rule, hard-
ness values should be obtained at the same locations
as the chemical testing and metallography.
The following table suggests guidance for evalu-
ating hardness survey results, showing typical and
maximum values for wrought iron and steel. Hardness
numbers somewhat above the maximum values shown
in the table are not in themselves cause for rejection,
but rather indicate the need for further investigation,
including chemical analysis and metallography. For
example, higher values may indicate that alloy or high-
strength steels may have been used, which could be
verified by chemical analysis. It is important to repeat
that in the case of historic metals, high strength is not
necessarily an indicator of quality, per se, and is to be
considered suspect.
5 FLAW DETECTION
5.1 Critical flaw size
Inspection and nondestructive flaw detection is neces-
sary on critical bridge members for two reasons: the
original materials were never tested to modern stan-
dards for flaws, and cracks (from corrosion, ductile
failure, or fatigue) may have developed over time.
Knowledge of the critical flaw size (size of crack
which will propagate without an increase in the applied
stresses)is important when inspecting tensile-load car-
rying members. Critical flaw sizes can be derived for
various defects, and this information can be used to set
detection limits for inspection, and to develop a frac-
ture control plan. The available inspection methods
must be capable of detecting the critical flaw size.
A simplified fracture mechanics analysis can be
used to estimate the critical flaw size to establish detec-
tion limits for inspection. To estimate the fracture
toughness, KIc , a correlation to relate CVN (Charpy
V-notch impact toughness) to fracture toughness KIc
(Rolfe and Barsom 1977, ASCE 1979, Norris 1981).
In such an analysis, the stress intensity factor, K, due
455
to an applied stress at a hypothetical crack is com-
pared with the fracture toughness KIc of the material.
This approach is accepted for evaluating the fracture
toughness and critical flaw sizes in iron and steel
truss bridges (ASCE 1979). Frank (1974) examined
the fracture toughness of wrought iron as compared
to selected steels to estimate the approximate scale of
critical flaws and found that the critical flaw size for
wrought iron is often larger than that of older struc-
tural steels. The same approach has been used for pins
and eyebars using ultrasonic testing as the primary
inspection technique. (Sparks & Badoux 1998) and for
riveted connections (Keller 1995). Brühwiler (1990),
Keller (1995), Grundy (2004), and Schindler (1995)
examined fracture toughness and fatigue in riveted
wrought iron construction.
For both low-carbon steels and wrought iron
with sufficient ductility, the critical flaw sizes in
bridge members are usually large enough to permit
visual inspection and detection with ultrasonic flaw-
detection equipment (Sparks 1998 2004). Where a
material does not pass the screening for low ductility,
then the assumed critical flaw size must be adjusted
downward.
5.2 Visual inspection
Visual inspection should be the primary method used
to assess the overall integrity of the bridge and to
identify loss of section, physical damage, and previ-
ous repairs. The visual inspection serves to integrate
the findings of the detailed inspections and testing.
All elements should be inspected visually. However,
critical zones are often concealed due to the joint con-
figuration and the presence of spacing collars, bearing
plates, etc., making it necessary to use nondestructive
test methods. Knowing the extent to which the critical
zones will or will not be visible is essential to planning
the assessment.
Depending on the type of bridge, the configuration
of the connections, and the material characteristics, the
following techniques may be appropriate.
5.3 Ultrasonic testing
5.3.1 Eyebars
Because the faces of the eyebar ends are generally not
accessible for visual inspection, it is necessary to use
a technique for circumferential scanning of the eye-
bar head for cracks (Sparks & Badoux 1998). The
neck area of eyebars is a potential zone of failure due
to the possible presence of laps or welds, the stress
concentration at the neck-to-bar transition, and strains
imposed during manufacture.
An actual eyebar head of the same method of man-
ufacture and material type is the best way to calibrate
the testing technique. Known flaws (side-drilled holes
and notches) can be introduced into these calibration
 Table 3. Steel properties in 1896 Parker Truss bridge.

C Mn P S
Member BHN % % % %

Rolled Sections
Floor Beam 104 0.14 0.44 0.01 0.03
Chord Channel 105 0.15 0.48 0.01 0.04
Post Channel 94 0.12 0.44 0.01 0.03
Post Channel 100 0.14 0.47 0.02 0.02
Average 101 0.14 0.46 0.01 0.03
AISI 1012 .10–.15 .30–.60 0.04 0.05
Lower Chord
Eyebars
T10W 101 0.11 0.72 0.05 0.02
T4E 121 0.09 0.70 0.02 0.04
Figure 4. Circumferential ultrasonic testing of critical eye- T5E 108 0.13 0.74 0.04 0.04
bar on 1896 Steel Truss Bridge. Average 110 0.11 0.72 0.04 0.04
AISI 1010 .08–.13 .30–.60 0.04 0.05
pieces. In wrought iron, the sensitivity of the ultrasonic Pins
testing is reduced substantially due to inclusions. Steel L12W 146 0.19 0.50 0.02 0.03
can also contain laminations, inclusions, and coarse L1E 143 0.15 0.46 0.02 0.04
grain size that can disturb the ultrasonic signal. Prior L3E 140 0.18 0.45 0.03 0.04
L4E 135 0.18 0.48 0.03 0.04
knowledge of the microstructure is essential to obtain-
ing good results. Field trials are necessary to determine Average 141 0.18 0.47 0.02 0.04
the actual level of attenuation. The transducers should AISI 1020 .15–.20 .30–.60 0.04 0.05
generally have 2.25 MHz or higher frequency to be
Chemical analysis is from field metallurgy using an ArcMet
able to resolve the necessary discontinuities. 930. BHN is Brinnell Hardness Number. AISI numbered steel
standards are shown for comparison.
5.3.2 Pins
Whether pins are critical and require ultrasonic test- macro-etching, die penetrant, and by magnetic particle
ing should be determined by analysis. Nondestructive testing. This detailed examination will reveal signs of
testing of pins is best accomplished using longitudinal lap-weld or other anomaly in the eyebar.
wave ultrasonic transducers. In most steels and occa-
sionally in wrought iron, valid results can be obtained
by scanning from one end only. In wrought iron pins, or 6 EXAMPLE: 1896 STEEL TRUSS
those with shoulders or significant wear grooves, scan-
ning may be necessary from both ends. Test several Although it is ‘conventional wisdom’that older materi-
of the pins from both ends to determine whether valid als were highly variable, more so than modern metals,
results can be obtained from testing from one end only. the author has found very consistent data within
Calibration standards can ideally be made from member classes for an 1896, an 1887, and an 1881
actual salvaged bridge pins, if available. This is partic- bridge. Bridge builders were apparently selective in
ularly helpful for wrought iron, so that the attenuation material choices, which were selected based on the
properties can be included in the calibration. Also, a function of the member in the bridge. The following
standard having actual wear grooves or loss of section discussion is for a twelve-panel Parker through-truss
can assist in distinguishing these characteristics from bridge spanning 71m (234ft), originally constructed in
cracks. Alternatively, it is also sometimes possible to 1896.
field calibrate for pin inspection by introducing a side The initial material identification was based on
drilled hole or notch in the non-stressed area beyond visual observations, a simple spark test, and limited
the nut. field metallography. Using the spark test, the relative
carbon content of the bridge elements were estimated
in the following order, from lower to higher carbon
5.4 Dye penetrant and magnetic particle testing
content: eyebars, rolled sections, and pins. The spark
Dye penetrant and Magnetic Particle testing are estab- patterns for the eyebars were similar to wrought iron
lished test methods and are both well suited for or very low carbon steel (<0.08% C). The pins showed
revealing cracks in beam webs and flanges, around carbon indications similar to medium carbon steel.
rivet holes, and for qualifying eyebars. It is advisable The eyebars showed a visible weathering pattern on
to examine at least one representative eyebar head by the top edges resembling longitudinal ridges, similar

456
 Table 4. Estimated allowable values for
assessment: 1896 Steel Truss Bridge.
Yield stress
MPa (ksi)
Eyebars: 220 (32)
Loop Rods 207 (30)
Pins: 317 (46)
Rolled Sections 220 (32)
to what is often seen on wrought iron, where the
ridges are a result of the presence of slag filaments.
Under field metallographic examination, the eyebar
metal consisted mostly of equiaxed grains. Almost no
slag fibers were visible on the wide face, but very
fine stringers could be seen on the edges of the eye-
bars where there appeared to be approximately twelve
laminations. This laminated character is indicative of
wrought iron; however, subsequent metallurgical test-
ing, as described below, revealed that this was actually
laminated steel.
Based on their carbon contents, the example mate-
rials should exhibit distinct ductile-to-brittle transition
as a function of temperature. For the eyebars, which
are the most critical elements, this transition tem-
perature should be about −20-deg C (impact basis)
(ASM 2004). The sulfur and phosphorus contents of
all the steel elements were below the maximum values
allowed for mild structural steels. As such, good duc-
tility is expected with no significant reduction in upper
shelf fracture toughness. Furthermore, the Mn/S ratio
is sufficient to fully convert the sulfur to manganese
sulfide, a potential ductility reducer.
The pin material generally meets the specifications
for pins proposed near the turn of the century by Cun-
ningham (1896). These include limitations on sulfur
and phosphorus, and a tensile strength between 60-
ksi and 70-ksi. The field hardness numbers indicate
that the pins have a tensile strength of approximately
70-ksi. Furthermore, the pins showed a fine grain
structure without gross segregation.
For structural analysis purposes, the following yield
strengths were used (Table 4). By comparison, the pre-
scribed AASHTO allowable for pre-1905 steel is 179
MPa (26-ksi).
The eyebar, plate, and rolled sections are easily
weldable due to the low carbon content and absence
of contaminants.
7 CONCLUSIONS
There is a clear need for a nondestructive approach for
historic iron and steel bridge evaluation. In most cases,
it is not feasible or appropriate to remove sufficient
material to rigorously quantify the yield stress, tensile
strength, and ductility of the material.
457
The aim of this paper has been to illustrate an
approach that can be used to screen metals for
low ductility, based on the following key indica-
tors: microstructure, chemical analysis, and hardness.
New tools make obtaining this data much more effi-
cient. When informed by historical test data and
in combination with detailed analysis, this approach
lends confidence to the evaluation. However, a good
deal of work is needed to make such an approach
widely useful. A number of further research goals are
evident:
1. A program of trials to qualify the ductility screening
method proposed here.
2. Improve portable metals identification equipment
and calibration procedures to accurately obtain
chemical analysis of wrought iron, in situ.
3. Develop an online catalogue of wrought iron, early
steels, and cast iron micrographs with associated
tensile, ductility (reduction of area), hardness, and
chemistry
4. Expand the range of code-allowed notional strength
values to include member classes (pins, eyebars,
rivets).
5. Develop a nondestructive in-place ductility mea-
surement
6. Research possible NDT methods for characterizing
slag distribution quantitatively.
7. Establish a library of typical bridge details, along
with proper methods of stress analysis, modes
of failure, and guidance for determining stress
concentration factors and critical flaw sizes.
8. Study the effect of phosphorus content in wrought
iron and steel on ductility and fracture toughness at
low temperatures.
9. Determine the typical range of phosphorus parti-
tioning between ferrite and slag, and establish a
means of determining this in the field.
REFERENCES
American Association of State Highway and Transportation
Officials, Manual for Condition Evaluation of Bridges,
1994.
AASHTO Guide Specifications for Strength Evaluation of
Existing Steel and Concrete Bridges.
American Society of Civil Engineers, Repair and Strength-
ening of Old Steel Truss Bridges, ASCE, 1979.
American Society for Testing Materials, “Quality Standards
for Wrought Iron,” Report of Special Subcommittee X on
Research, 1934.
American Society for Testing and Materials, ASTM Standard
E 140 Hardness Conversion Tables for Metals.
ASM International 2004, ASM Metals Handbook, Tenth Edi-
tion, Volume 9. Materials Park, Ohio: ASM International.
p. 739.
Aston, James and Edward B. Story, Wrought Iron: Its Man-
ufacture, Characteristics and Applications, A. M. Beyers
Company, Pittsburgh, 1936.
Brockenbrough, Roger L., Design Guide 15: AISC Reha-
bilitation and Retrofit Guide: A Reference for Historic
Shapes and Specifications, American Institute of Steel
Construction, 2002.
Brühwiler, E. I.F.C. Smith, and M. A. Hirt 1990. “Fatigue
and Fracture of Riveted Bridge Members.” Journal of
Structural Engineering, Vol. 116(1): 198–214.
Buonopane, Stephen & Sean Kelton 2007. “Mechanical
Properties of Wrought Iron from Penns Creek Bridge
(1886)” Dept. of Civil and Environmental Engineering,
Bucknell University
Cunningham, A. C. 1896. “The Condition of Steel in Bridge
Pins,” Transactions of the American Society of Civil
Engineers, Vol. XXXVI (December).
Federal Highway Administration, Comprehensive Bridge
Safety Inspection Program, “Section 15: Inspection of
Fracture Critical Bridge Members.”
Frank, K. H., “Mechanical and Chemical Properties of
Selected Steels Used in Bridge Structures,” Report
No. FHWA-RD-75-79, Federal Highway Administration,
Washington, DC, 1974.
Freudenthal, Alfred M. 1950 The Inelastic Behavior of Engi-
neering Materials and Structures. New York: John Wiley
& Sons.
Gayler, Carl, “Wrought Iron and Steel Eye-bars,” Transac-
tions of the Association of Engineering Societies, read
January 16, 1889.
Gordon, Robert B., American Iron 1607-1900, The Johns
Hopkins University Press, Baltimore, 1996, pp. 208–211.
Gordon, Robert B., “Strength and Structure of Wrought Iron,”
Archeomaterials 2:109–137, 1988.
Gordon, Robert B., “Evaluation of Wrought Iron for Con-
tinued Service in Historic Bridges,” ASCE Journal of
Materials in Civil Engineering, July/August 2005, pp.
393–399.
Grundy, Paul, “Fatigue Assessment of Wrought Iron Struc-
tures,” Steel Construction, Volume 38, Number 1, Aus-
tralia, March 2004.
Hertzberg, Richard W., Deformation and Fracture Mechanics
of Engineering Materials, Second Edition, John Wiley &
Sons, New York, 1983.
Hartle, Raymond A. et al., Bridge Inspector’s Training Man-
ual (Washington, D.C.: Federal Highway Administration,
1990).
Holley, A. L. 1877 “The Strength of Wrought Iron as Affected
by Its Composition and By Its Reduction in Rolling,”
Trans. Am. Inst. Min. Eng., 6, pp. 101–124.
Institution of Structural Engineers, Appraisal of Existing
Structures, 2nd Edition, 1996.
Kaufmann & Roberts 1990, “Strength and Weldability Eval-
uation of the Wrought Iron Beams of the US Capitol
Terraces,” ATLSS Report No. 90-10, Lehigh University.
Keller, Andreas, et al., “Assessment of a 135Year Old Riveted
Railway Bridge,” Extending the Life Span of Structures,
Proceedings of the International Association for Bridge
and Structural Engineering Symposium, San Francisco,
1995.
Kirkaldy, David 1862. Results of an Experimental Inquiry
into the Tensile Strength and Other Properties of Various
Kinds of Wrought-iron and Steel. Glasgow.
458
Kunz, F. C., Design of Steel Bridges Theory and Practice for
the use of Civil Engineers and Students, McGraw-Hill,
1915.
Larsson, Tobias, Material and Fatigue Properties of Old
Metal Bridges, Licentiate Thesis 2006:26, Luleå Univer-
sity of Technology, Department of Civil and Environ-
mental Engineering, Division of Structural Engineering,
Sweden, 2006.
Maniar, D.R., M.D. Englehardt, and D.E. Leary 2003. Evalua-
tion and Rehabilitation of Historic Metal Truss Bridges: A
Case Study of an Off-System Historic Metal Truss Bridge
in Shackleford, County, Texas, Research Report 1741-3.
Austin: Center for Transportation Research, University of
Texas.
Norris D.M., J.E. Reaugh, and W.L. Server 1981. “A
Fracture-Toughness Correlation Based on Charpy Ini-
tiation Energy” STP743-EB. West Conshohocken, PA:
ASTM International
Oberg, Erik and Franklin D. Jones 1918. Iron and Steel, First
Edition, New York: The Industrial Press.
O’Sullivan, M. 2007 “Mechanical properties of wrought-iron
published at various times” End of first year PhD report,
University of Manchester.
Pilkey, Walter D., Peterson’s Stress Concentration Factors,
John Wiley & Sons, New York, Second Edition.
Rolfe, Stanley T., and John M. Barsom, Fracture and Fatigue
Control in Structures, Prentice-Hall, Inc., 1977.
Sauveur, Albert, The Metallography of Iron and Steel,
McGraw-Hill Book Company, New York, 1912.
Schindler, Hans-Jakob, “Toughness Evaluation and Assess-
ment of Old Bridge Steel,” Proceedings of the Interna-
tional Association for Bridge and Structural Engineering
Symposium (1995): 1313.
Sparks, Patrick 2004. Guide to Evaluating Historic Iron &
Steel Bridges. Austin: Texas Department of Transportation
Sparks, Patrick, “Wrought Iron: Properties, Evaluation, and
Reliability,” presented at the Association for Preservation
Technology Annual Conference, Banff, Alberta, Canada,
October 1999.
Sparks, Patrick & Marc Badoux 1998. “Non-Destructive
Evaluation of a Historic Wrought-Iron Truss Bridge in
New Braunfels, Texas”, APT Bulletin: Journal of the
Association for Preservation Technology, Vol. XXIX (1).
Steel Construction Institute, Appraisal of Existing Iron and
Steel Structures, SCI Publication 138, 1997.
Tilley, Graham 2002. Conservation of Bridges. London: Spon
Press.
Tschorn, Gerhart 1963. Spark Atlas of Steels: Cast-Iron, Pig-
Iron, Ferro-Alloys and Metals, New York: The MacMillan
Company.
Twelvetrees, W. N., Structural Iron and Steel. A Text Book
for Architects, Engineers, Builders, and Science Students,
Whittaker & Co., London, 1900.
Walton, Charles F. (ed.) 1971 Gray and Ductile Iron Castings
Handbook. Cleveland: Gray and Ductile Iron Founders
Society, Inc.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Diagnosis and analysis of two king-post trusses

J.M. Branco & P.J.S. Cruz

Civil Engineering Department, ISISE, University of Minho, Portugal

M. Piazza
University of Trento, Trento, Italy

ABSTRACT: An accurate geometric and mechanical evaluation of two king-post trusses, based in grading
results with data gathered from NDT as well as from mechanical evaluation, followed by full-scale tests were
performed. The trusses were reassembled in laboratory and submitted to a series of cyclic tests under symmetric
and asymmetric loading. Strengthening techniques evaluated in precedents research steps were used in a second
phase of the carrying tests.

1 INTRODUCTION Rafter (Ri1) King post (Ki0)

The work presented here has been developed within
a general research program devoted to investigate the
behavior of traditional timber trusses and to identify
suitable strengthening strategies. The study addresses
the overall behavior of full-scale timber trusses under
cyclic tests under symmetric and asymmetric load-
ing. The carrying tests were preceded by an accurate
geometric and mechanical evaluation of each tim-
ber elements of both trusses studied. Visual grading
and non-destructive tests based on local and global
evaluation were performed.
After the carrying tests on the unstrengthened
trusses, following a load procedure according to the
Service and Ultimate Limit States, the trusses were
strengthened. Strengthening techniques, based in the
use of metal devices, were evaluated in previous steps
of the research (Branco, 2005), through full-scale
tests on traditional timber connections. Strengthening
aimed to improve the carrying capacity of the trusses
and to analyze the influence of the joint stiffness in
the overall behavior, in particular, under asymmetric
loads.
2 TRUSSES STUDIED
The trusses studied have been recover from an old
building locate at Pergine Valsugana (20 km East from
Trento) which will be transformed in a school. The
construction is not very ancient, probably built in the
first part of the last century, and did not present signs
of rehabilitation or strengthening actions. The building
i = 1, Truss 1
i = 2, Truss 2 Rafter (Ri2)
Strut (Si1) Strut (Si2)
Tie-beam (Ti0)
Figure 1. Trusses geometry.
is a typical rural construction with ground floor, first
level and attic with visible timber roof structure. The
change in the use of the building results in the neces-
sity to remove the timber trusses. The trusses with
a king-post configuration have a span near to 10.5
meters.
All members are composed by one single timber
piece of Spruce (Picea Abies Karst.), presenting a
mean cross-section of 20 × 23 cm2 . All members were
marked before dismantling of the trusses necessary
for their transportation to the Materials and Structural
testing Laboratory of the University of Trento.
3 DIAGNOSIS
3.1 Geometric assessment
During the visual inspection, significant variation
in cross-section geometric properties was founded.
Because each component of the trusses is composed by
a unique timber piece, it is normal to verify significant
variability in the cross-section area of the elements, in
particular in those longer like the tie beams. To have an
adequate report of all those variabilities of the trusses
geometry, a geometric assessment was performed.

459
 193 70 50 73 evaluations methods were conducted: Sylvatest, using
ultrasonic waves, and the mechanical evaluation of the
62 60
modulus of elasticity.
212 210 75 109
3.3.1 Hardness tests
75 41 In the Turrini-Piazza tests, hardness is estimated by
195 65 80 50
measuring the load force R required to embed a 10 mm
φ steel hemispherical bit to 5 mm (Turrini & Piazza,
(a) (b) 1983). The value of R was obtained by averaging the
tests results made in the four longitudinal faces of the
Figure 2. Comparison between the geometric properties of element, in numbered sections every 80 cm wide. Each
the cross-sections: (a) supposing rectangular and (b) with test consisted in five measures taken in each tested
detailed measurements (dimensions in millimetres).
area. The result of each test was obtained by averaging
the three middle values among the five measures. It
Table 1. Visual strength grading according to UNI 11035.
means that, for each considered section: 9 for the tie
Truss Element Strength class E0,mean (MPa) beam, 5 for the rafters and 2 for the struts and king
post; twenty tests were done (four by surface). At the
1 K10 S1 12000 end, nearly 1000 hardness tests were performed. The
T10 S2 10500 correlation between R and elastic modulus in longi-
S11 S1 12000 tudinal tension E0 is expressed by the experimental
S12 S2 10500 equation:
R11 S2 10500
R12 S1 12000
2 K20 S1 12000
T20 S1 12000 where δ indicates a reduction factor, 0.5 ≤ δ ≤ 0.8, that
S21 S2 10500 depends on the defectiveness of the element.
S22 S3 9500
Because the Equation 1 is experimentally founded,
R21 S1 12000
R22 S1 12000 its reliability depends on the limits of the experi-
mentation itself: namely on the species of the tested
samples that are Silver fir (Abies Alba) and Larch
(Larix decidua), on loading conditions (700 ÷ 3000 N)
Each element, every 40 cm, was marked, named and and moisture content in the range 12%–14%. However,
the cross-section measured. Figure 2 presents the due to the large amount of hardness test performed and
cross-section A of the tie beam of the Truss 1, as since mechanical determination of the elastic modulus
example of the difference that can be obtained if an of some elements (tie-beams and rafters) is available,
inadequate measurement of the cross-section is done. it was possible to calibrate the equation for the case o
Spruce (Picea Abies Karst.). The value of δ = 0.6 was
founded. After the calibration of the δ value, it was pos-
3.2 Visual grading sible to calculate the elastic modulus of all elements
The grading criteria according to the European (EN that compose the two king post trusses.
518:1995) and the Italian national standards (UNI
3.3.2 Pilodyn
11035-1-2:2003) have been applied. The species of the
The Pilodyn method uses a steel pin of a fixed diam-
two king-post trusses analyzed is Spruce (Picea Abies
eter driven into the material by a dynamic force. The
Krast.), from North Italy, accordingly, grading criteria
depth of penetration is correlated with material density.
referred to the group ‘softwood 1’ (conifere 1) in UNI
Görlacher (1987) and Feio (2005) developed relation-
11035 §5.3 have been adopted. For visual grading, it
ships between the depth of penetration of a standard
was necessary to record the moisture content and to
pin and the density of the wood. Those correlations
calculate density.
depended on the number of measurements, species
and are affected by moisture content. For all elements
3.3 Non-destructive tests (NDT) that compose both trusses, in each section used for the
geometric assessment, the penetration depth obtained
Non-destructive tests based on local and global evalua-
by the Pilodyn in the 4 faces, was measured. Table 2
tion of the material were performed. Resisting drilling
summarized the Pilodyn tests results obtained.
obtained by means of the Resistograph® device, the
penetration depth achieved with the Pilodyn® and the 3.3.3 Resistograph
hardness Turrini-Piazza tests results are examples of The use of a small-diameter needle-like drill was intro-
techniques based on local evaluation. After, global duced by Rinn (1992). The cutting resistance of a

460
Table 2. Depth penetration means values for Pilodyn. Table 3. Sylvatest results using the direct method.

N.◦ of Density Depth w Length Time E Fb

Element measurements w (%) (kg/m3 ) (mm) Element (%) (m) (µs) Class (MPa) (MPa)

K10 20 11.7 450 16.25 K10 11,7 1,88 329 0 13200 18

T10 104 12.2 478 10.41 T10 12,2 – – – – –
S11 24 11.5 432 19.63 S11 11,5 2,23 400 1 12200 12
S12 24 11.3 410 20.42 S12 11,3 2,23 393 1 12800 16
R11 52 12.5 467 12.38 R11 12,5 5,1 850 0 15400 28
R12 48 11.9 460 11.98 R12 11,9 4,85 869 1 12200 12
K20 20 12.7 457 13.10 K20 12,7 1,87 315 0 14600 24
T20 104 12.4 432 15.68 T20 12,4 – – – – –
S21 24 12.2 384 21.42 S21 12,2 2,32 400 0 13600 20
S22 24 12.3 450 10.75 S22 12,3 2,42 459 3 – –
R21 52 12.8 479 11.54 R21 12,8 5,1 864 0 14400 28
R22 52 12.1 447 13.62 R22 12,1 5,11 824 0 15600 30

needle is recorded as a function of depth as the nee-

dle penetrates the timber. The resulting profile can be
used to determine the location and extent of voids and
variation in material density. This technique is highly
effective for quantifying the extent of deterioration in
timbers. Some researchers (Feio 2005, Frankl et al.
2006, Kotlínová et al. 2007) have tried to use the results
of the resistograph as prediction of the mechanical
properties. For that, he used the Resistance Measure
(RM), given by the quotient between the integral of the
area of the resistograph chart and the depth penetration
of the nail in the specimen:
However, the use of the RM value to estimate mechani-
cal properties is questionable (Lear, 2005). The area of
a resistance-drilling plot can be affected by multiple
parameters including drill bit sharpness and general
equipment use such as drill orientation. On a single
member, changes in the orientation of the drill with
respect to growth rings will change the calculated RM
value with each drilling.
3.3.4 Sylvatest
The ultrasound device Sylvatest® allows sending and
receiving longitudinal ultrasound waves using two
piezoelectric probes. It measures the duration of the
stress wave course in the tested sample, with regard to
its moisture content and temperature, which are also
controlled. Since there is a one-to-one relation between
this speed and the mechanical quality of timber (elas-
ticity modulus and, less significant, resistance), the
instrument displays directly the measured mechanical
quality after the evaluation of the influence of wood
moisture content in the element taken into considera-
tion. The results are given in the form of quality classes
or mechanical values, indicating the elastic modulus
and the allowable bending strength of the element
(Sylvatest, 1991).
Two measurement methods have been used: the
direct method, where the transmission of the wave
is done along the fibers and the indirect method,
where wave transmission is across the fibers. To assess
the influence of the test method on the results, both
methods were applied whenever possible. Each single
measured value represents the third reading value, to
avoid bigger variability in the test results.
Different measurement method led to different Syl-
vatest results. Even using the indirect method in two
different longitudinal faces, between the same sec-
tions, some results inconsistency has been found. This
discrepancy is more evident in the allowable bend-
ing strength, less important for the elastic modulus in
bending and disappears for the measured mechanical
quality class. Table 3 summarized the Sylvatest results
obtained using the direct method, when possible, for
all elements component the two trusses.
3.3.5 Mechanical evaluation of the modulus of
elasticity in bending
The stiffness of the elements, which is normally
expressed as MoE (modulus of elasticity) was deter-
mined by means of bending tests. MoE for timber
contains not only information about the clear wood
strength properties, but also to a large extent about
defects affecting the stiffness. Hence, results of bend-
ing tests have been used in this research to assess the
reliability of visual grading and NDT results. Static
bending tests were carried out in accordance with
the four-point loading method, with reference to EN
408:2004. The relevant static scheme is characterized

461
 Table 4. Elastic modulus (MoE) obtained by
means of four-point bending tests.
Element MoE (MPa)
T10 12980
R11 11149
R12 11367
T20 12929
R21 12646
R 22 12863
by a constant bending moment zone and a maxi-
mum shear-to-bending moment ratio. This test uses
a simply supported piece with a span of 18 times the
depth, with the loading points placed on the third of
the span. Because of the large span needed for the
correct execution of the bending tests (18 times the
depth), only the rafters and the tie-beams were tested.
Results of the four-point bending tests are reported in
Table 4.
3.3.6 Analysis and discussion of NDT results
Using UNI 11035 standards it was possible to visual
strength grade the timber trusses. In particular, a
strength class was assigned and a mean value for the
modulus of elasticity (MoE) was suggested. Three val-
ues for MoE were founded: 12000 MPa, 10500 MPa
and 9500 MPa corresponding to strength class S1, S2
and S3, respectively. For visual grading, it was nec-
essary to record the moisture content and to calculate
density.
During the visual inspection, significant variation
in cross-section geometric properties was founded.
Turrini-Piazza hardness test permits the correlation
between the load force (R) required to embed a steel
bit into the wood with MoE. However, the evaluation
of a reduction factor dependent on the defectiveness
of the element is required. This reduction factor was
calibrate (with a value of 0.6) using near 1000 tests
results and the MoE values obtained with the mechan-
ical evaluation. Apart this capability of the test, it was
tried to correlate the load force (R) with the density.
The correlation obtained is poor (r2 = 0.35) despite the
significant values used. Others researchers have tried
to correlate the Resistance Measure (RM), obtained
with the resistograph test, with density. The same
attempt was made but the correlation obtained is very
poor (r2 = 0.02). A correlation coefficient of 0.5 was
obtained for the correlation between the depth of pen-
etration measured with the Pilodyn and the density. In
conclusion, and assuming as reference values for the
MoE the results of the mechanical tests, a correlation
between all non-destructive tests able to predict this
property and that values were defined, and presented
in Figure 3.
462
13500
MoE = 0.624MoE(Turrini-Piazza) + 4655.4
13000
2
R = 0.6197
MoE EN 408:2004 (MPa)
12500
12000
11500 MoE = 0.2166MoE (Sylvatest) + 8886.5
2
11000
R = 0.1494
UNI 11035
10500 MoE = 0.2578MoE (UNI 11035) + 9357.3 Turrini-Piazza
2
R = 0.0572 Sylvatest
10000
10000 10500 11000 11500 12000 12500 13000 13500 14000 14500 15000 15500
MoE obtained with non-destructive method (MPa)
Figure 3. Correlation between MoE given by
non-destructive test and MoE given by the mechanical
evaluation.
4 CARRYING TESTS
The trusses were subjected to cyclic tests, under sym-
metric and asymmetric loading for two levels of
loads corresponding to the Service and Ultimate Limit
States. In consequences of their behavior under this
series of tests, the trusses were strengthened. Later,
the same sequence of tests was carried out over the
strengthened trusses. If, after that, the failure of the
truss was not been achieved, a final test under growing
loading was performed.
4.1 Test setup, instrumentation and procedure
The timber trusses were tested under cyclic loading
subjected to two point loads, each one directly applied
over the joint strut-rafter. Symmetric and asymmetric
tests were performed using two values of load, corre-
sponding to Service and Ultimate Limit states (SLS
and ULS, respectively). As loads, and in accordance
with the Italian code D.M. 14/9/2005, it was consid-
ered the self-weight of the original timber roof and the
snow load corresponding to the place were the trusses
were recovered, Caldonazzo Lake (500 m of altitude).
Before carrying the tests, and after the reconstruction
of the disassembled trusses, trusses were subjected to
a series of loading and unloading tests with the aim
to adjust all trusses elements together reducing to the
minimum the initial gaps between them. The first step
of the tests was the application of the load correspond-
ing to the self-weight of the timber roof (covering
material and structure plus purlins). After that, the
remaining load corresponding to each loading case
was divided in four steps.Therefore, the load target was
reached after 4 cycles of loading and unloading keep-
ing constant the minimum load applied corresponding
to the self-weight. For each test, three loading cycles
were applied, Figure 4.
The asymmetric tests were carried out with the
asymmetric values of the snow load, as considered in
the above mentioned D.M. 14/9/2005. During the tests
 60
Load target
50
25%
50%

Load (kN)
75% 40
Load

FDx
30 FSx

20
Self-weight
10
Time (a) Symmetric loading history

Figure 4. Load procedure applied in the carrying tests. 14

12

Displacement (mm)
FSx
R7 R8 FDx 10
Ai - Displacement
R5 R6 Ri - Rotation
8 A7
A6 R3 R4 A7
A1
R1 6 A6
A5 R2

A2 A3 A4 4
2
Figure 5. Instrumentation layout. 0

(b) Asymmetric response of channels A6 and A7

the main global displacement of the trusses, the rela-
tive displacements between the king post and the tie Figure 6. Asymmetric behavior of Truss 1 under symmetric
beam and the rotation of each joint were measured, loading condition (SLE).
using sixteen transducers, Figure 5.
Two hydraulic jacks, supported by a steel counter –
Displacement (mm)

10
frame fixed to the reaction slab of the laboratory, were 8 A3
used for applying the loads. As boundary conditions, 6 A5
two supports restraining the vertical displacement 4
and the right support restrained also the horizontal 2
movement were applied. 0

4.2 Evaluation of the trusses behavior Figure 7. Behaviour of the connection between the
king-post and the tie beam under symmetric loading
The response of the timber trusses under the carrying conditions.
tests performed is highly depending of the variabil-
ity, previously detected and reported, along the truss 2
members in terms of cross sections area and princi- A1
Displacement (mm)

1.6
pal moments. Moreover, it is recognized the difficulty
associated with tests over existing structures, in which 1.2
the members can present residual deformations and
signs of decay and/or the connections are not well 0.8
tight. The tests results shows that even under symmet- 0.4
ric loading conditions, the behavior of both trusses
is asymmetric. This conclusion is more evident in 0
the response of Truss 1 under symmetric loading
corresponding to SLS, Figure 6. Figure 8. Horizontal displacement recorded in the support
The connection between the king-post and the tie under symmetric loading.
beam, made by a steel rod of 20 mm diameter fixed
in the king-post, aligned in the longitudinal direction, Under asymmetric loading conditions, the asym-
passing through the tie beam, with a nut at one end, metric behavior of the trusses, already pointed out, is
fastening the two elements, performs adequately. The emphasized. Channels A6 and A7 shows asymmetric
king post suspends the tie beam, Figure 7. displacement, Figure 9.
As a consequence of the truss deformations under This asymmetric behavior of the trusses introduces
the vertical loads applied, horizontal displacement at bending stresses in the rafters and tie beam. Because
the free support was recorded, Figure 8. of the rotational stiffness offered by the king-post/tie

463
 40
35 A7
Displacement (mm)

30 A6
25
20
15
10
5
0

Figure 9. Asymmetric behavior emphasized under asym-

metric loading (Truss 1 and SLE).
Figure 11. Behavior of the king-post/tie beam connection:
(a) in plane bending of the bolt (b) out-of-plane movement
0 prevented.
-2
Displacement (mm)

-4
A2
-6
A4
-8
-10
-12

Figure 10. Distortion of the tie beam under asymmetric

loading (Truss 1 and SLS).

beam connection, the asymmetric response of the over-

all truss causes distortion in the tie beam (Figure 10).
This distortion induces important bending stresses in
the bolt of that connection (Figure 11a).
The connection between the king-post and the tie
beam presents rotational stiffness in the plane of
the truss and in the normal plane. Therefore, this
connections prevent also the out-of-plane movement
(Figure 11b) improving consequently the stability of
the substructure composed by the rafters, the king-post
and the struts.
As expected, under asymmetric loading the relative
displacement between the king-post and the tie beam,
measured in the connection between both elements,
is nearly zero. In addition, the horizontal movement
recorded in the unrestraint support is reduced.
For the case of the Ultimate Limit State (ULS), dif-
ferent behaviors were observed for both tested trusses.
While Truss 2 was able to perform all tests (sym-
metric and two asymmetric), in Truss 1 only the first
cycle of the symmetric tests was achieved. When the
maximum load corresponding to this limit state was
reached, considerable out-of-plane movements were
visible, Figure 12. To prevent the global failure of the
truss it was decided to stop the test and to strengthTruss
1 before proceeding with the experimental program.
In the case of Truss 2, the tests undertaken for SLS
were repeated for the load level corresponding to the
Figure 12. Out-of-plane movement observed in the truss 1
(unstrengthened) under ULS at the end of 1st cycle.
ULS. The main conclusion about the overall behavior
of the truss pointed out for the SLS can be extended to
ULS. Therefore, it can be concluded that, apart from
the instability observed in the case of Truss 1, the
safety of the trusses under the Ultimate Limit States
was verified. No local collapse or failure in the timber
members occurred. The bracing forces transmitted by
the purlins and the covering structure should prevent
the instability observed in the case of Truss 1.
4.3 Strengthening
The next phase in the experimental program was to
strength the trusses based in the response obtained in
first phase and to repeat the tests. Strengthening of
Truss 1 aimed essentially to repair the out plane move-
ment observed and to prevent this instability making
possible to perform the carrying tests corresponding
to the ULS. The out plane deformations were removed
and a UPN profile was bolted to the king-post. The
objective was to prevent the out plane movements,
increasing the stiffness of the connection in this direc-
tion, keeping the tie beam suspended to the king post,

464

Figure 13. Strengthening of Truss 1.
Figure 11a. The timber elements did not presented
any local failure or damage however, the connections
between the rafters and the tie beam were weakened.
In particular, the left connection, over the support with
horizontal displacement, the depth step was insuffi-
cient and the timber beyond the step was fissured.
Then, it was decided to strength this part of the tie
beam (beyond the step) with screws (12 screws M6-
200) aiming to increase its shear resistance. Moreover,
using the existing holes, an internal bolt of 20 mm
diameter was introduced at midjoint and normal to the
rafter in each rafter/tie beam connection, Figure 12b.
In the case of Truss 2, only the connections between
the rafters and the tie beam were strengthened. The
stiffness of the king-post/tie beam connections demon-
strated to be adequate to prevent the out-of-plane
instability. Rather, the timber beyond the step demon-
strated some signs of deterioration and the depth of
the steps in the rafters/tie beam connections looked
insufficient. In this last case, the same strengthening
technique used before inTruss 1 was applied: one inter-
nal bolt tightening the connection and screws beyond
the step to improve the shear resistance. However, and
as an improvement based in the failure modes observed
in Truss 1, a binding strip was used to confine the
timber, Figure 13.
4.4 Efficiency evaluation of the strengthening
The series of tests corresponding to the Service
Limit States (SLS) were repeated after strengthening
the trusses. The main objective was to analyze the
465
Figure 14. Strengthening of Truss 2. Screws and binding
strip applied in the frontal part of the step in the rafter/tie
beam connection.
influence of the joint stiffness, in this case, the stiff-
ness of the rafter/tie beam connection, in the overall
behavior of the truss in particular under asymmetric
loads. However, the main goal of the strengthening
was to increase the truss resistance: to make Truss 1
able to perform all load procedures corresponding to
ULS, and to assess the failure load of Truss 2.
Load capacity of Truss 1 was improved by the
strengthening, but only the symmetric test correspond-
ing to the ULS was accomplished. At the end of this
test, the local damages in the rafters/tie beam connec-
tions not permitted the execution of the asymmetric
tests. It was decided to perform a test until failure. It is
important to point out that the ultimate load achieved
in the failure test is lower than the load level attained
in the ULS test. Despite the strengthening measures
undertaken to improve it, the failure was caused by
lack of shear resistance of the tie beam beyond the step.
Theoretically, the screws should increase the shear
resistance but, because of the fissures already existing
before placing the screws, the timber part beyond the
step did not work as a rigid body. The screws expanded
the fissures and the rigid body became separate in
small pieces, Figure 14a. However, it is important to
point out that this behavior was also accentuated by the
previous shear failure of the rear step, Figure 14b. With
the horizontal movement of the rafter in the connection
rafter/tie beam, the joint king-post/struts was disman-
tled, Figure 14c, and important bending stresses are
introduced in the internal bolt, Figure 14d.
In the case of Truss 2, it is difficult to conclude if
the strengthening was efficient in the increase of the
load capacity of the truss as no failure was obtained in
the previous tests performed in unstrengthened condi-
tions. The strengthened truss was subjected to the full
loading history foreseen in the test procedure (SLS
and ULS, under symmetric and asymmetric loads).
After that, a loading procedure until failure was fol-
lowed. This procedure is similar to the previous one
but now, the load increment did not stop at the value
of 80 kN (maximum load corresponding to ULS).

Figure 15. Failure damages of Truss 1.
Figure 16. Shear failure of the rafter/tie beam connection
for Truss 2.
Figure 17. Double step connections between the rafters and
the tie beam of Truss 2.
The maximum load value achieved during the tests
at failure was ≈100 kN.
Again, the failure of the truss was obtained by shear
failure in the timber beyond the step in the connec-
tion rafter/tie-beam. First, the shear failure of the rear
step happened and then the shear failure of the tim-
ber beyond the frontal step took place, Figure 15. The
geometry decided which connections rafter/tie beam
failed. The failure happens in the double step connec-
tion in which the rear step presented a faulty geometry.
The rear step must be deeper than the frontal. If not,
466
the surface resisting to shear stresses is insufficient.
The double step connections of the right side (over the
fixed support) failed as a consequence of a first shear
failure in the rear step.
Once the rear step fails, the stresses concentrated in
the frontal step and the failure is reached, Figure 16.
5 CONCLUSIONS
In this work, after a detailed geometric assessment
of both trusses, the grading criteria according to
Italian standards (UNI 11035-1:2003 and UNI 11035-
2:2003) have been applied. As keys parameters for
visual grading and non-destructive evaluation, mois-
ture content and density were measured and recorded.
After, some non-destructive techniques based on local
evaluation of the material were performed. Resisting
drilling obtained with the Resistograph, the penetra-
tion depth achieved with the Pilodyn and the hardness
Turrini-Piazza tests results were correlated with the
density values. In a second phase, global evaluations
methods were used. Sylvatest, using ultrasonic waves,
and the mechanical evaluation of the modulus of elas-
ticity were performed. Some correlations between
different test results, in particular, for the prediction
of the modulus of elasticity (MoE) were established.
Assuming as reference values for the MoE the results
of the mechanical tests, the Turrini-Piazza Hardness
test show to be a sufficiently accurate non-destructive
technique in the prediction of the MoE.
The response of both trusses under the carrying
tests performed is highly depending of the variabil-
ity along the truss members in terms of cross sections
area and principal moments. Test results show that
even under symmetric loading conditions, the behav-
ior of both trusses is asymmetric. The connections
between the king post and the tie beam must ensure
the adequate stiffness to avoid the out-of-plane insta-
bility of the structure. Strengthening, realized with
metal devices, was able to improve the carrying capac-
ity of the trusses. The failure of the trusses always
happened for the shear failure in the rafter/tie beam
connection. The failure depends of the geometry of
the double step. A faulty geometry accelerates the fail-
ure. It was founded that the strengthening of wood
in the frontal step with screws could be improved by
confining this part.
ACKNOWLEDGMENTS
The first author gratefully acknowledges the Por-
tuguese Foundation for Science and Technology, for
his PhD grant SFRH/BD/18515/2004. The research
described in this paper was conducted with finan-
cial support of the Portuguese Foundation for Science
and Technology (POCI/ECM/56552/2004). This work
has been carried out with a partial financial contri-
bution of the Italian Earthquake Engineering Labora-
tory Network (RELUIS), within the research program
carried out for the Italian Agency for Emergency
Management.
REFERENCES
Branco J.M., Cruz P.J.S., Piazza M. & Varum H. 2005. Exper-
imental Analysis of Birdsmouth Joints. Report E-11/05.
DECivil, University of Minho, 49 pp.
EN 408 2004. Timber structures. Structural timber and glued
laminated timber. Determination of some physical and
mechanical properties. CEN European Committee for
Standardization.
Feio A.O. 2005. Inspection and Diagnosis of Historical
Timber Structures: NDT Correlations and Structural
Behaviour. PhD Thesis, University of Minho.
Frankl J., Kloiber M. & Bryscejn J. 2006. Non-destructive
inspection of a historical wooden structure damaged
by fire. In Engineering mechanics 2006, Svratka, zech
Republic, 2006, pp. 62–63.
467
Görlacher V.R. 1987. Zerstörungsfreie Prüfung von Holz: ein
in Situ Verfahren zur Bestimmung der Rohdichte. Holz als
Roh- und Werkstoff, Vol.45, No.7, 273–278.
Kotlínová M., Kloiber M., Vasconcelos G., Lourenço P.B. &
Branco J.M. 2007. Non-destrcutive testing of wood
structures. Report E-12/07. DECivil, University of
Minho, 63 pp.
Lear G.C. 2005. Improving the assessment of in situ
timber members with the use of non-destructive and
semi-destructive testing techniques. M.S. Thesis – North
Carolina State University, Raleigh, 137 pp.
Rinn F. 1992. Chancen und Grenzen bei der Untersuchung
von Konstruktionshölzern mit der Bohrwiderstandsmeth-
ode. Bauen mit Holz, (9).
Sylvatest 1991. Sylvates user’s manual. Cableries &
Trefileries de Cossonay SA. Switzerland. 19 p.
Turrini G. & Piazza M. 1983. Il recupero dei solai in legno.
Esperienze e realizzazioni. In Recuperare, Vol. 7.
UNI 11035-1 2003. Structural timber – Visual strength
grading for Italian structural timbers: terminology and
measurement of features. UNI – Milano, Italy.
UNI 11035-2 2003. Structural timber – Visual strength grad-
ing rules and characteristics values for Italian structural
timber population. UNI – Milano, Italy.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Non destructive and laboratory evaluation of strength of decayed wood

members in a historic construction located in Gorgan (North of Iran)

M. Madhoushi
Department of Wood & Paper Sci. and Indus., Gorgan Uni. of Agri. Sci. & Nat. Res., Gorgan, Iran

J. Eimanian
Iranian Cultural Heritage Organization, Gloestan Head Office, Gorgan, Iran

M.P. Ansell
Department of Mechanical Engineering, University of Bath, Bath, UK

ABSTRACT: The House of Bagheri is one of the most important buildings in Gorgan city (North of Iran)
which is a tourist site and the House is a Cultural Heritage Construction. Solid wood was utilized extensively
for structural members in association with other traditional materials. It has been exposed to natural environ-
mental conditions for a long time, so some faults have developed including decay of the wooden members.
The mechanical properties of decayed members have been measured in laboratory tests and the buildings have
been non-destructively evaluated. Samples from the buildings were prepared which are representative of the
whole structure. Flexure and compression tests were then performed on these samples. The results of NDE
(nondestructive evaluation) tests on the samples were correlated with those on the buildings.

1 INTRODUCTION

Gorgan is located in the North of Iran between the

Caspian sea and the Alborz mountains where the cli-
mate is mostly Mediterranean (with average rainfall of
432.1 mm and RH between 62.5% and 74%). It is con-
sidered to be a heritage city because of its history and
historic buildings some of which contain solid wood
structural elements.
This paper evaluates the structural integrity of one
of the most important historic building in Gorgan,
called the House of Bagheri which is a tourist site and
is designated as a Cultural Heritage Construction. The
strength of decayed wood members is measured by
using both a nondestructive stress wave method and a
static laboratory measurement to quantify mechanical
Figure 1. A view of the House of Bagheri.
properties.
The House was built approximately 150 years ago
with a floor area of 3000 m2 (Eimanian, 2001) and is
now under repair. Solid wood was utilized extensively The building is a collection of several main parts
for structural members including beams, columns, and yards. A small pool is located in the centre of each
rafters, roofs and floors in association with other tra- yard surrounded by buildings. Although there is a one-
ditional materials i.e. brick and cob (a mixture of storey building in this complex, it is mainly built on
clay and straw). Doors and windows are completely two-stories with a few steps between floors (Fig. 1).
wooden and are well preserved. Due to long exposure The buildings have nearly the same architectural char-
to the natural environment, some faults have developed acteristics and the plan of one building is shown in
including decay of wooden members. Figure 2.

469
Figure 2. A plan of one building of the House of Bagheri.
This building possesses a rectangular structure and
used construction materials which were essentially
brick and cob in the shear walls and solid wood in
the principal structural elements. Its main defects have
been previously reported by Madhoushi & Eimanian
(in press). Degradation caused by weathering (sunlight
and rainwater splashing), decay and insect attack have
led to damage and faults in the building. It is now
under repair and most defective structural elements
have been substituted by new members.
Today, repair and restoration of historic buildings is
a major challenge for society (Lourenco et al. 2006).
The conservation and repair of these mostly timber
buildings is important due to the unique response
of timber structural members under seismic loading
(Ceccotti et al. 2006, Tampone & Messeri 2006).
Previous studies conducted on historic wood build-
ings showed that moisture and fungal decay could be
considered as the main source of damage (Ronca &
Gubana 1998, Pasanen et al. 2000). However, it was
shown by Madhoushi & Eimainan (in press) that in
the House of Bagheri, insect and weathering damage
as well as fungal decay are the source of degradation.
The stress wave technique is one of the methods
used widely for the inspection of load-bearing tim-
ber members. Previous studies (Haines et al. 1996,
Machek et al. 2004) show that there are good relation-
ships between static MOE (Modulus of Elasticity) and
dynamic MOE. These relationships depend on species,
grain direction and existence of defects.
The fundamentals of the stress wave method and its
applications have been explained by Pellerin & Ross
(2002). This method can be applied in the evalua-
tion of members in timber structures (Ross & Pellerin
1994) and decayed wood (Ross et al. 1996). A previ-
ous investigation (Madhoushi et al. 2006) showed that
this method might be used for assessing the decay of
Iranian hardwoods and it was suggested as an in situ
evaluation technique for decayed wooden members of
historic structures.
470
Figure 3. Measuring the transmission time of stress waves.
2 MATERIALS AND METHODS
A nondestructive study was conducted using an IML®
electronic hammer in order to produce stress waves
in wooden members. Eight accessible structural ele-
ments were selected as representative of the whole
structures (since the building is under repair, access to
most of elements was not possible) and the transmis-
sion time of the stress waves were measured at three
point on each sample in order to define the area of
decay (Fig. 3). At each point, the instrument provides
three readings.
It should be noted that there was not any intention
to replace these elements with solid wood. Eight sam-
ples were removed from the structure and selected as
representative of the whole building. They were sub-
jected to static laboratory experiments to measure their
mechanical properties including flexure and compres-
sive properties parallel to grain. Finally, the data were
analyzed and compared with the NDT (nondestructive
testing) data.
3 RESULTS AND DISCUSSION
3.1 Nondestructive evaluation
The measured transmission time of stress waves indi-
cates (Table 1) that 50% of samples are totally decayed
and that 12.5% of samples are slightly decayed. Only
about 25% of the samples can be considered as sound
or nearly sound. Finally, 12.5% of samples are exten-
sively decayed and distorted and contain fractures so
that they cannot be reused. These samples are termed
“unusable”.
It is clear that the building is not in good condi-
tion and its repair is necessary. Although some parts
have been renewed without NDT inspection it seems
these parts are solid. In contrast, some parts are in their
Table 1. The measured transmission time of samples and Table 3. Comparison between stress wave NDT, laboratory
their condition. examination and visual inspection.

Transmission time (µs) TD SD UU NS S

No 1∗
2 3 Baseline (µs) Condition Stress wave NDT 50 12.5 12.5 12.5 12.5
Lab Examination 50 25 – – 25
1-h∗∗ 333 357 341 650 TD∗∗∗ Visual Inspection 25 – – 50 25
1-s 863 829 888 143 TD
2-s 634 678 618 158 TD
3-h 567 599 584 650 SD
4-h 3891 4587 3125 52 UU of investigation.Table 3 proves that there is a difference
5-h 432 489 473 585 NS between visual inspection and scientific examination.
6-s 364 313 403 520 S
7-s 741 783 718 156 TD 3.3 Comparison of the data
8-s 282 311 272 234 SD
A comparison of the data is shown in Table 3. There
∗
Place of measurements on sample. were not any unusable samples in laboratory exam-
∗∗ ination and visual inspection, because they wouldn’t
h for hardwood, s for softwood.
∗∗∗ be used again in restoration of the building. It can be
TD: total decayed, SD: slight decayed, UU: unusable,
NS: nearly sound, S: sound. seen that if we ignore the unusable samples tested using
NDT, both the stress wave NDT and lab examination
Table 2. The mechanical properties of old wood members results are very close to each other. Both methods show
measured in laboratory. that 50% of the samples are totally decayed and 25% of
samples are nearly sound. However, it is more difficult
Condition∗ to distinguish the slightly decayed by laboratory exam-
MOR MOE E in C|| ination and these results depend on the extent of decay.
∗∗
Sample 1 2 (MPa) (GPa) (GPa) This outcome is discussed in more detail by Madhoushi
et al. (2006). Visual inspection based on personal expe-
Rafter 1 NS SD 79.12 10.46 3.16 rience produces is inherently variable and results are
Rafter 2 S SD 88.73 8.9 3.43 not to be trusted.
∗∗∗ ∗∗∗
Column 1 S TD 2.44 In general, since there is good accordance between
∗∗∗ ∗∗∗
Column 2 NS TD 3.35
Door Frame 1 NS S 79.77 9.09 10.2
the results of stress wave analysis and laboratory test
Door Frame 2 TD TD 12.67 2.12 2.15 results, both methods can be used to assess the integrity
Beam 1 TD TD 22.56 3.48 2.58 of timber during repair and restoration activities. How-
Beam 2 NS S 74.8 9.15 10.68 ever, NDT is a more economic and cost-effective
method.
∗
Evaluated (1) visually by expert carpenter based on expe-
rience, (2) after examination.
∗∗
Modulus of elasticity in compression parallel to grain. 4 CONCLUSIONS
∗∗∗
was not measured.
• The House of Bagheri is a heritage wood con-
original condition without any repair or treatment, and struction building in Gorgan in which the timber
they are not in a good condition. members contain some damage and decay.
• Stress wave nondestructive method has been used in
the evaluation of the building and it showed that 50%
3.2 Static mechanical properties
of the structural elements tested are totally decayed
The static mechanical properties measured in the lab- and decay is initiated in 25% of the structural
oratory show (Table 2) that some elements do not elements tested.
possess acceptable strength in spite of the visual eval- • Laboratory testing demonstrated that some old
uation to the contrary. For example samples from members are not of acceptable strength in spite of
Rafter 1 were considered to be nearly sound elements, their good appearance and visual inspection.
whereas in fact the MOR (Modulus of Rupture or • As much of the building as possible will be retained
bending strength) and MOE and the modulus of elas- for restoration.
ticity in compression parallel to grain (E in C|| ) have • There is good correspondence between stress wave
all been reduced by decay. In contrast, Door Frame nondestructive evaluation and laboratory mechani-
1 is sound and has acceptable mechanical properties cal tests.
in spite of its grading as nearly sound. It is clear that • It is advised that the NDT method should be
visual evaluation, even by an expert carpenter, is not used in association with visual inspections in the
accurate and using NDE methods is advised in this type assessment of this heritage building.

471
REFERENCES
Ceccotti, A., Faccio, P., Nart, M., Samghaas, C. and
Simeone, P. 2006. Seismic behaviour of historic timber
frame buildings. ICOMOS Intern. Wood Committee; Proc.
15th intern. Symp., Istanbul and Rize, 18–23 September
2006.
Eimanian, J. 2001. Restoration suggestions on House of
Bagheri’, MSc, Thesis, Islamic Azad University.
Haines, D.W., Leban, J.M. & Herbe, Ch. 1996: Determination
of Young’s modulus for spruce, fir and isotropic materi-
als by the resonance flexure method with comparisons to
static flexure and other dynamic methods. Wood Science
and Technology 30 (4):253–263.
Lourenco, P.B., Luso E. & Almeida, M.G. 2006. Defects and
moisture problems in buildings from historical city cen-
tres: a case study in Portugal. Building and Environment
41 (2): 223–234.
Machek, L., Edlund, M., Sierra Alvarez, R. & Militz, H.
2004. A non-destructive approach for assessing decay in
preservative treated wood.Wood Science and Technology
37 (5):411–417.
Madhoushi, M., Ansell, M.P. & Hashemi, M. 2006. Appli-
cation of NDT stress wave method for assessing property
changes in decayed wood of Iranian beech (Fagus orien-
talis). Integrated Approach to Wood Structure, Behaviour
and Applications; Proc. ESWM and COST Action E35
meeting, Florence, 14–17 May, 2006.
Madhoushi, M. & Eimainan, J. (in press). Faults and Repairs
in House of Bagheri: A Cultural Heritage Construction
in Gorgan (North of Iran) – A Case Study. Submitted to
472
11DBMC Inter. Conf. on Durability of Building Materials
and Components, Istanbul, 11–14 May, 2008.
Pasanen, A.L., Kasanena J.P., Rautialaa S., Ikaheimoa M.,
Rantamakib, J., Kaariainenb H. & Kalliokoskia, P.
2000. Fungal growth and survival in building materials
under fluctuating moisture and temperature conditions.
International Biodeterioration & Biodegradation 46 (2):
117–127.
Pellerin, R.F. & Ross, R.J. 2002. Inspection of timber struc-
tures using stress wave timing non-destructive evaluation
tools. In R.F. Pellerin & R.J. Ross (eds), Nondestructive
Evaluation of Wood: 135–148. Forest Products Society.
Ronca, P. & Gubana, A. 1998. Mechanical characterisation of
wooden structures by means of an in situ penetration test.
Construction and Building Materials 12 (4): 233–243.
Ross, R.J. & Pellerin, R.F. 1994. Non-destructive testing for
assessing structures: A review. Gen. Tech. Rep. FPL-GTR-
70 (Rev.). Madison, WI, U.S. Department of Agriculture,
Forest Service, Forest Products Laboratory. 40 p.
Ross, R.J., De Groot, R.C., Nelson W.J. & Lebow, P.K. 1996.
Assessment of the strength of biologically degraded wood
by stress wave NDE. In C. Sjostrom (ed.), Durability of
Building Materials and Components 7 (V1): 637–644.
E & FN Spon.
Tampone, G. & Messeri, B. 2006. Compliance of the practice
of strengthening ancient timber structure in seismic areas
with the official documents on conservation. ICOMOS
intern. wood committee, Proc. 15th intern. Symp., Istanbul
and Rize, 18–23 September 2006.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

The structural behaviour of spires

E. Olosz & B. Szabó

Technical University, Cluj-N., Romania

ABSTRACT: In spite of the fact that there are plenty of examples of historic spatial roof structures – in spires,
castle towers, etc. – in the technical literature information about this kind of structure can rarely be found,
especially about their structural behaviour. The development of computerised 3D programs of analysis and
modelling helped us to improve our understanding of this. This paper offers a guide to assist in the understanding
of the structural behaviour of the load bearing structure of spires, offering a typology, followed by a detailed
presentation of these structural analyses.

1 INTRODUCTION exclusively from supporting load-bearing sub-units

To build a shelter against sun, wind or rain is one of
the most fundamental of human needs. Since the earli-
est times, people have joined together pieces of wood,
forming the first spatial structures – conical timber
huts or dwellings. Later on, the construction of tow-
ers of fortifications and churches continued the still
ongoing development of spatial roof structures.
Following technical development over centuries,
architectural styles also had their remarks on spatial
roof structures.
The authors of the present lecture represent the
PhD Studies School within the Technical University
of Cluj-N. (Transylvania), where since 1998 a number
of PhD topics addressed the scientific study of var-
ious historic roof structures of the Gothic, Baroque
and Eclectic periods. Spatial roof structures (espe-
cially spires) from the same periods are described in
this paper.
2 STUDYING SPATIAL ROOF STRUCTURES
2.1 Definitions
So that we have the same terminology, we would like
to offer a few definitions:
Spire: a tapering conical or pyramidal structure on
the top of a building, particularly a church tower with
the proportions between width and height of the spire
being between 1/1,3–1/5.
Historic roof structure: built of timber by an
empirical-intuitive load-bearing structural concepts,
without the support of designed or calculated engi-
neering theory, characterized by its support coming
(such as walls, pillars, columns), usually located on
the external face or edge of buildings [6].
Spatial system: the spatial system of a historic
roof structure is formed from different load-bearing
structural materials (hardwood or softwood timber)
connected by joints between lineal elements, so form-
ing a stable truss system. The spatial systems of a
historic roof structure can either be jointed together
with frames (particularly if the layout is rectangular),
or it is a real spatial system, which cannot be divided
into frames; to split up these roof structures into frames
would imply a forced procedure, which can cause
errors in the creation of the mechanical model [6].
2.2 Classification and components
Requirements:
a) prevention from overturning;
b) the basic horizontal framing system with secure
joints to rafters and other elements;
c) a suitable bracing system to act against wind loads.
The spires generally are four or eight sided, high,
conical roof structures. The whole height of the spire
is distributed in levels by horizontal framing systems
(less then 4,50 m one from the other). The lowest fram-
ing system is the base of the spire, connected to the
walls of the tower through wall plates. All the other ele-
ments start here, so this decides the structural concept
of the spire.
Two different systems of spatial roof structures
have been developed in Europe: one with long King-
posts (Figure 1a.) and the other with short Kingposts
(Figure 1b).

473

Figure 1. Spatial roof structures.
2.2.1 Spires with long Kingposts
The kingpost, the “spine” of the spire, is supported by
the basic framing system and supported vertically by
the angle braces or scissor braces, is the most “mys-
terious” element of the structure. Structural analysis
shows the real purpose of this element.
The basic framing system can have diagonal or
orthogonal (parallel with the sidewalls) main timber
beams. The secondary timber beams are connected
to header beams, or directly to the main beams.
(Figure 2.)
Only two beams can span across, so holding the
Kingpost at the crossing.
474
Figure 2. The horizontal framing system.
Figure 3. Prevention of overturning – Fixings.
So as to prevent overturning these structures are
anchored at five points. There is a beam built into
the walls of the tower (about 80 cm below the basic
framing system) which supports the Kingpost. So as to
hold the corners of the spire there are two solutions. In
medieval times they used a bracing system constructed
inside the tower being one or two stories across (Fig-
ure 3b1.). The other system used metal bars (diameter
φ30–32) built into the walls of the tower being about
1.5–2.0 m long (Figure 3b2). This second system was
used for both short and long Kingpost style spires.
2.2.2 Spires with short Kingposts
Usually these have eight faces. The basic framing
system can have two different forms: orthogonal con-
necting across corners (not diagonal), or polygonal
system having an open space in the middle of the spire.
(Figure 4.)
The next two or three or upper levels are constructed
with similar framing systems (Figure 1b).
Figure 4. The horizontal framing system.
Figure 5. The roof bracing system.
Preventing the spire from overturning is provided
by a similar system to that described previously, using
metal bars at the corners.
The main difference is in the bracing system acting
against wind-loads. Supported by the framing systems,
there are lower and upper purlins, connected by scissor
bracings, set out at two or more levels (Figure 5). If
the spire is the typical eight sided model, these scissor
bracings above the second level are set out alternately
in four-four sides. If the spire is four-sided, these braces
are set out on each face, but they can be missing at the
top level.
Common rafters are connected at the top to the short
Kingpost, which is supported by the last (or last two)
framing systems.
Regarding the construction of spires, we have found
a description in a nineteenth century carpenter’s book,
written in the period when a lot of spires of this type
were constructed: “The spires should not be very rigid;
they should allow a small amount of movement during
storms. For this reason the beams of the framing sys-
tems should be screw fixed to the common rafters, but
the roof bracing system’s beams should not be screw
fixed to the common rafters, rather they should finish
a few centimetres from the rafters. The basic framing
system should be well fixed into the sidewalls of the
tower. The main common rafters (spanning from the
corners) are connected to the short Kingpost, which
is secured firmly at the bottom of the framing system.
The other common rafters are connected either into the
main common rafters, or to header beams”. [2]
2.2.3 A third type of spires can be defined as a
combined structure
As the spire of Bela tower of the Mathias Church,
from Budapest (from 1896) with a long Kingpost, and
475
Figure 6. The combined roof structure of the Mathias
Church, Bela Tower, Budapest, Hungary.
a similar bracing system as described for the short
kingpost styled spires (Figure 6).
2.2.4 Other elements affecting the structural
behaviour of a spires
Additional elements affecting the performance of
spires are: the pinnacles at the four corners, the
tympanum-windows at the centre of the facades,
clock-ledges, other ornaments (especially in the case
of baroque structures). All of these elements can dis-
turb the direction and intensity of loads and so the
structural behaviour of the spire’s main structure.
2.3 Numerical (computerised) modelling
The main aspects of the structural behaviour of the spa-
tial roof structures which can be resolved by modelling
are as follows:
a) supporting arrangements – whether the structure is
supported mainly at the top of the wall, or lower
down and by the fixings or “building in” systems;
b) effectiveness of joints – whether the timber con-
nections allow the parts of the members to rotate
freely or provide restraints against rotation;
c) the role of the different components of the structure
(especially the Kingpost).
Other considerations which are not subject of this
paper:
a) consideration of the applied loads, especially the
wind-loadings – whether the loads act as the simpli-
fied European Standards states, or whether account
should be taken of the distortion effect of the
wind-loading;
b) the role of the additional elements: pinnacles, win-
dows, galleries, other ornaments (baroque roofs);
c) the effect of irregularities and missing elements to
the structural behaviour of the spires.
Figure 7. Modelling the spires with long kingpost.
2.4 Case studies
We will present the results of the structural analyses for
the two different kinds of spires (with long and short
kingpost), presenting the similarities and differences,
and also the difficulties of research and analyses of
these structures.
2.4.1 General problems of building the model
of a spire
a) Survey problems: generally the roof structures of
the spires are not accessible; the general dimen-
sions can be taken by using laser instruments but
the joints, especially the higher once are very dif-
ficult to access either because of the narrow space,
either because of highness;
b) Building the model geometrically: the model of
spires with long kingpost can be easily built using
parallel lines with the basic railing system started
from the kingpost, the beams of the horizontal lev-
els meets the common rafters having an exact joint.
(Figure 7).
In the case of the spires with short kingpost, the
horizontal railing system is not follows the com-
mon rafter’s directions. The software what we used
is based on the theory of finite elements, making
the calculations in joints of the axes and can not
deal with elements which are crossing each-other
but not in the axes. There so, finding the joints
between common rafters and upper horizontal rail-
ing systems is difficult, and the program cannot
dial with plans crossing with lines. This is why we
had to include a serial of contact elements in the
model (Figure 8). These contact elements permit to
give different rigidities for these joints.
c) Consideration of loads: Roofs must be able to with-
stand a multitude of actions. In our analyses we
considered the dead loads and wind loads, and
476
Figure 8. Modelling the spires with short kingpost – contact
elements.
we ignored the snow and ice loads – because of
the pitch of the spires which are more than 60◦ .
We also ignored the temperature fluctuation or the
actions during construction, etc.
We considered the loads according to the European
standards adapted for the spires as presented below:
The permanent actions were taken into account by
way of an average value, the time-dependent, vari-
able actions are specified by characteristic values what
were multiplied by partial safety factors during anal-
yses of the load-carrying capacity (for dead loads
this is 1.35, for wind loads is 1.50). For serviceabil-
ity analyses the partial safety factor for dead loads is
1.00, for variable loads are reduced by factors and fre-
quency values – which for wind loads are 0 = 0.60,
1 = 0.50, 2 = 0.00.
Dead loads are permanent and immovable actions
caused by the self-weight of load-bearing and non-
load-bearing components and the roof covering, what
was considered being profiled aluminium sheet-
ing, considered together with the roof decking
(0.25 kN/m2 ).
The wind load is made up of pressure, suction
and friction effects. Pressure and suction act perpen-
dicular to the surface of the building. Pressure act
on the windward side, suction on the leeward side
and friction effects could be on the parallel side of
wind load. The wind loads were considered with theirs
reference speeds taking account also to the local sur-
roundings, as follows: topography category 4 – urban
district; height above ground level z(m) = 36.0 m, the
dynamic pressure of the wind for these surroundings
is qp (z) = 0.655 kN/m2 .
In order to determine the distribution of wind pres-
sure on the surface of an eight-sided spire (as far we
had no possibility to carried it out in wind tunnel tests
 Figure 10. Stresses in the Kingpost – Nx (kN) – for
Dead-loads (1)/Wind-loads (2) for fixed joints (a)/joints with
articulations (b).

Figure 9. Distribution of wind load for eight-sided spires.

in which the wind pressure could be measured at the

surface of a model neither to make useful measure-
ments with setting sensors at the different faces of the
spires), we adapted the example given in the European
standards for hipped-roofs, as follows.

2.4.2 Results
a) Spires with long Kingpost – the roof structure of
the tower of Reformed church from Barabas.
The joints have been considered in two ways:
with free rotation and with restraint against rota-
tion, but this – as you can see it below – had
only a minor effect in the stresses of the different
elements.
The other question was the role of the different
elements (especially the Kingpost) in the structural
behaviour of the spire.
The role of the Kingpost:
In spite of the fact that the Kingpost is an
extremely slim element, with it’s cross-section very
much weakened by the joints with other elements,
the Kingpost is very important to the structure,
working for both vertical and horizontal actions
(such as roofing and wind-loads).
For dead-loads the Kingpost is working as
usual – resulting only axial stresses. What is
unusual is that wind-loads producing bending
moments in the Kingpost especially in the joints
with the compound rafters, but also at joints with
knee-braces and collars. This is why the bracing
system of the spire should not be screw fixed to
Figure 11. Bending moments in the Kingpost – My (kN m) –
for Dead-loads (1)/Wind-loads (2) for fixed joints (a)/joints
with articulations (b).
the common rafter, permitting a little movement
during rugged wind.
The other elements, such as common rafters,
compound rafters, collar beams etc. are work-
ing – as usual – in both compression or tension
and sometimes common rafters also for bending
moments.
The spires with long Kingpost are not very sen-
sitive to the rigidity of the joints. The distribution
of loads are very much depends of the rigidity of
the Kingpost, and the possibility of movements of
the bracing elements during storms.

477

Figure 12. Stresses in the compound rafters – Nx (kN) – for
Dead-loads (1)/Wind-loads (2) for fixed joints (a)/joints with
articulations (b).
Figure 13. Stresses in the short Kingpost – Nx (kN) – for
Dead-loads (1)/Wind-loads (2) for fixed joints (a)/joints with
articulations (b).
Figure 14. Stresses in the short Kingpost – My (kNm) – for
Dead-loads (1)/Wind-loads (2) for fixed joints (a)/joints with
articulations (b).
b) Spires with short Kingpost
The joints similar to the first case were consid-
ered in two different arrangements. This had an
important effects in the structural behaviour of the
spire, increasing substantially the stresses in the
elements.
478
Figure 15. Stresses in the scissor bracings – Nx (kN) – for
Dead-loads (1)/Wind-loads (2) for fixed joints (a)/joints with
articulations (b).
The Kingpost in this case is not presenting very
much stresses neither for dead-loads, nor for wind-
loads. The role of the Kingpost is much more
constructive, than supporting the structure of the
spire – so that it is an assembling element helping
to join together the common rafters.
More important role have the horizontal framing
systems and also the scissors bracings, what are
also very sensitive to the movements:
The spires with short Kingpost are very sensitive
to the rigidity of the contact elements, and rotations
in the joints between elements, producing impor-
tant redistribution of stresses. The Kingpost has a
secondary importance; in this case primary ele-
ments are the scissor bracings and the horizontal
railing systems.
3 CONCLUSIONS
In our lecture we tried to present a typology of these
structures and also the structural behaviour of these,
using the knowledge collected from the technical
literature and self-experience. There are a lot of ques-
tions what were not answered (especially concerning
the distribution of wind loads), what needs exper-
imental studies in wind-tunnel and also by farther
measurements on extant structures.
Historic spatial roof structures have considerable
historic value. It is our aim to study and create an inven-
tory of these beautifully built structures, in order to
understand their structural behaviour and to preserve
them for future generations.
ACKNOWLEDGEMENTS
The helpful comments and advice during the prepara-
tion of the paper are gratefully acknowledged: Andrew
Shepherd, Dr. Andras Vandor, Dorottya Makay, Istvan
Muka and Andras Wagner.
Additional help during the measurements and
preparation of drawings also gratefully acknowledged:
Marton Lenard.
REFERENCES
1. Bruzsa, L., Toth, E. 2007, Roof structures from A to Z
(Hungarian), Verlag Dashöfer, Budapest, Hungary
2. Diesener, H., 1897, Carpenter works (Hungarian), Keil
József Publisher, Budapest, Hungary
3. Gabor, Laszlo, 1976, Doctrines of the construction ele-
ments (Hungarian), Tankonyvkiado, Budapest, Hungary
4. Mohrmann, K. 1901, Gothic Structures’ Handbook
(German), Leipzig, Germany
479
5. Schunck, E., Oster, H.J., Barthel, R., Kiessl, K., 2003,
Roof Construction Manual. Pitched Roofs, Birkhäuser –
Publishers for Architecture, Basel-Boston-Berlin, Edi-
tion Detail Munich, Germany
6. Szabo, B. Gy. 2005, Illustrated Dictionary of His-
toric load-bearing structures, Kriterion/Utilitas Pub-
lisher, Cluj-N., Romania
7. Yeomans, D., 1992, The Architect and the Carpenter,
RIBA Heinz Gallery, London, UK
8. http://en.wikipedia.org/
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

In-situ and laboratory investigations on materials and structures for the

static improvement: “Torre Quadrata del Circo Romano” (Milan)

Susanna Bortolotto, Elisabetta Ciocchini & Fabio Zangheri

Politecnico di Milano, Dipartimento di Progettazione dell’Architettura, Laboratorio di Diagnostica per la
Conservazione e il Riuso del Costruito, Milan, Italy

Paola Marina De Marchi

Ministero per i Beni e le Attività Culturali, Direzione Regionale per i Beni Culturali e Paesaggistici della
Lombardia, Milan, Italy

ABSTRACT: The “Circo Romano” was built by the Emperor Massimiano in the last years of the 3rd century.
It was erected with the attempt to give Milan a building capable to represent the city as a capital of the Empire.
Nowadays the only witness of such an imposing group of buildings is the Tower of the “carceres”. The intervention
planned for the “Torre quadrata del circo romano”, includes the project for static improvement and the more
complex preservation project of the upper part of the building (the loggia).
In order to prepare the project it was necessary to carry out a diagnostic campaign to verify the effective condi-
tions of the building. All the retrieved information were represented on drawings identifying geometrical survey,
image rectification and stratigraphic readings referring to the historical-archival and the brickwork discontinu-
ities. In addition, there was also representation of the readings of crack networks and of out-of-plumbs; structures
and materials non-intrusive or light-intrusive diagnostics performed by means of sonic method; checking of the
status of sleepers, tie beams and chains in order to verify their functioning; mortars and bricks sampling for the
identification of the base elements; the results of petrographical and mineralogical analysis.

1 INTRODUCTION

The corpus of the survey conducted on the Torre

Quadrata del Circo Romano in Milan made it pos-
sible to delineate the status of the structure as we find
it today.
The analyses focus on two different subjects, albeit
closely linked: the historical/evolutionary aspect of the
structure and the static/dimensional aspect.
In view of this, and in agreement with the Lom-
bardy Regional Department, the appropriate analy-
ses that needed to be carried out on the structure
were established for the purpose of developing a
project to improve static functions and to re-utilize the
Torre.
From the point of view of conserving materials as
well as with reference to static aspects, the portion of
the structure that presented the most critical situation
is the summit, where even a simple observation with
the naked eye shows the complexity of the situation. Figure 1. The tower.
It is here, in fact, that deterioration can be observed
on the slim marble columns and on the wall structures as evidenced by the wood support structures put into
that support the summit section of the loggia where, place in 1981 and still judiciously left in place.
in recent years, a partial static consolidation attempt The survey operations and diagnostic examinations
was made. Unfortunately it did not serve its purpose, carried out were therefore conducted with the primary

481
purpose of verifying the Torre’s geometrics, the status
of the works already implemented and of providing
indications on the conservation status of the stone
materials and of the connecting grout in the wall
section relating to the loggia.
Figure 2. The topographic polygon.
Figure 3. Drawings representing the entity of out-of-plumbs.
482
2 SURVEY, MAPPING OF CONSTRUCTION
PHASES AND VERTICALITY CONTROL
In order to provide better support in interpreting the
results obtained and, at the same time, to provide an
additional way of looking at the situation, the stratigra-
phy survey carried out in the Nineties by the architect
Gianfranco Pertot was re-examined. The stratigra-
phy, in fact, correlates quantitative and qualitative
information, utilizing the chronological succession
of the structure construction episodes as the guiding
parameter.
The geometric findings, together with the photo-
graphic straightening, was the first operation under-
taken, by which it traced accurate maps on which,
subsequently, information acquired as a result of the
other diagnostic surveys would be mapped.
A closed topographic polygon was realized (with
laser total station) composed of seven stations in which
more than seven hundred points required for tracing
the structure geometrics were measured. Thereafter,
a detailed profile was executed, performed with the
help of direct measuring instruments such as laser dis-
tance meters, rigid meters and ribs to integrate the
missing information and to study the level of detail
of certain more significant portions of the structure.
In addition, a self-leveling laser was used to determine
the difference in level between the stairwell landings
in the internal staircase.
Seven different outlines were realized, each one cor-
responding to a landing, in addition to that for the
ground floor and for the loggia, with five sections and
four external prospects. The considerable amount of
details determined made it possible to correctly high-
light the entity of the out-of-lines, measured on the
four corners of the structure, as well as the opportu-
nity to coherently represent the deformations and the
bulging of the walls.
In order to correctly restore the plumb lines, also
taking into consideration the particular deterioration
of the bricks with which the structure was realized,
it was deemed advisable to execute the readings on
the corners from very close up (approximately every
30–40 cm for the entire height of the Torre). Suffi-
cient information thus became available to identify and
isolate any mistakes in the reading or in collimation.
The topographic bearings, used for the geometric
restitution, were also employed as references with real
coordinates for the photographic straightening pro-
cedure for the four external fronts and for the four
internal prospects on the loggia level.
Figure 4. Image rectifications.
Figure 5. Drawings representing the entity of out-of-plumbs into the loggia.
483
In order to acquire the photographic material, a dig-
ital reflex camera equipped with standard optics (focal
length 50 mm) was used and it was also equipped
with tele-optics (210 mm) based on the camera sta-
tion; wide angle optics (20 mm) were used for the
loggia shots, and the relevant deformations were
subsequently corrected by software.
The procedure for photo straightening and placing
the single images in mosaic order was realized with
dedicated software. The coordinates provided by the
topographic shots were used, as specified.
The geometric tracing and the control of verticality
made it clear that the front presenting the most dis-
placement is that facing west, on the contrary to that
facing east, which is nearly perfectly plumb.
The information determined on the external corners
of the structure is also confirmed by the measurements
carried out inside the loggia. These were effectuated
by using a high accuracy laser plumb and once again
it is the west wall that shows the most displacement
from the vertical. In this case the values are more
modest but, in fact, even the portion of the wall
analyzed is considerably less extensive than on the
external walls, where the out-of-plumb was calculated,
Figure 6. Stratigraphic analysis.
taking into consideration nearly the entire height of the
building.
The analysis of the situation concerning crevices
evidenced a modest presence of lesions and crevices,
at the same time showing the existence of no-anchoring
areas among the angles of the Torre and the loggia’s
supporting fill walls.
It is believed that the origin of this phenomena is not
attributable to structural settlement or to yielding, but
rather to the particular genesis of the entire structure.
It was constructed in many phases and over a very long
period of time and presents construction stratifications
at times not properly solved and generated by an urgent
need rather than by a single project.
On the basis of the plans realized by the photo-
graphic straightening, a stratigraphic analysis of the
structure was outlined for the four external façades.
Thanks to a high definition photographic support, sub-
ject to verifying the stratigraphy unit readings, this
analysis made it possible to restore the development
and interpretation of the numerous construction activ-
ities that followed each other during the course of time.
In fact, the stratigraphy analysis tables describe
the perimeter of the stratigraphy units (S.U.) and
of the architectural elements (A.E.). Through appro-
priate background information, they communicate
the construction activities in chronological sequence,
highlighting wall irregularities and the architectural
structure’s critical points.
484
With respect to the Torre, seven different construc-
tion phases were identified and eight different periods
of interventions were classified, to which the three
restoration phases, carried out from the second half
of the 19th century, must be added.
3 DIAGNOSTIC ANALYSIS
In addition to providing priceless information as to the
evolution dynamics of the structure, the stratigraphy
readings also constitute a fundamental basis for the
choice of homogeneous samples to analyze the char-
acteristics of the materials. At the same time these
surveys were carried out, an analysis campaign was
organized that aimed at recognizing the stone materi-
als present in the structure and at the characterization
of the connecting grout used for the walls in the loggia.
This in-depth study is justified by the evident decay
affecting the Torre’s stone elements, especially the log-
gia, where decorations can no longer be read and where
entire sections of material have disappeared.
14 samples were taken from the four façades of the
loggia to analyse the column, plinth and stone materi-
als present in the masonry in the area below the loggia’s
floor.
Samples were analysed in a laboratory with polar-
ized light microscopy (MO-LP) using traditional
methods of mineralogical analysis in “thin sections”
Figure 7. Localization of the different stone materials.
Figure 8. Characterization of the mortar beds.
that entails the use of a very thin (about 0.03 mm)
layer of stone glued between two slides. The material
is observed through a polarizing microscope and the
analysis is conducted under different types of polarized
light in order to clearly observe all the characteris-
tics of that particular sample. These characteristics,
unique to each mineral, allow specialized technicians
to recognize the materials on sight.
The results obtained confirm the visual analysis,
in that they showed the great variety of the materials
sampled, including: Pietra di Angera, Rosso di Verona,
Trachite dei Colli Euganei, Pietra di Saltrio, Pietra
Aurisina and Marmo Apuano.
The absence of consistent placement inside the
building for each litote is moreover confirmed in the
context of the elements themselves. The four columns
485
analyzed are in fact one different from the other: the
re-used stone materials originate from different urban
“quarries”.
Laboratory analyses continued, as previously men-
tioned, with the characterization of the mortar beds
used in the loggia and were conducted with X-rays
diffraction (XRD).
Diffraction allows the crystalline planes present in
the samples analyzed to be recognised and, if com-
pared with appropriate reference tables, they identify
the elementary components of the material. In brief,
the procedure uses X-rays that are directed towards
a rotating sample and then diffracted with varying
angles according to the materials that make up the sam-
ple. At the same time, a detector measures the width
of the diffraction angles and records them on file for
Figure 9. Ultrasound analysis carried out on the loggia’s columns.
subsequent analysis. This method provides qualitative
and in some cases semi-quantitative results.
The results highlighted a rather uniform situation
both in terms of the composition (high concentrations
of quartz and calcite with a significant presence of pot-
sherds) and of the state of conservation that is not good
no matter the location and morphological aspect of the
sample under study. In particular, nearly all the mor-
tars are poorly consistent with pulverization in some
sporadic cases, though, on the average they show a
uniform and compact structure.
It is thus difficult to identify deterioration problems
with respect to the exposition of the interior fronts as
well as with respect to other factors external to the
material. The results show a general state of conserva-
tion that is slightly worse as regards the mortars used to
build the loggia parapet and those used for the masonry
located over the arches and in the normal parameters
relative to angle irons.
Analyses on the stone materials continued with
ultrasound tests to control the compactness of the
elements.
Ultrasound tests are based on the relationships that
link the propagation speed of elastic waves in a given
material to the elastic properties of the material itself.
The propagation speed of these waves can be directly
linked to the physical and mechanical characteristics
of that material and, more to the point, a reduction
of the propagation speed indicates irregular (cracks,
empty spaces, detachments) or deteriorated material.
Ultrasound tests can be carried out in three ways
(by transparency, on the surface and by radial trans-
mission) according to the position of the transmitter
and receiver on the surfaces under study.
These tests were performed on the small stone
columns in the loggia and on the huge columns in
cipolin marble at the Torre entrance. They revealed
that the loggia elements are in good condition, despite
the external surface attacked by corrosive elements
and sulphatation, and that the material is compact. On
the other hand, the columns at the entrance revealed
low speed values and in some points the total absence
of propagation of the ultrasound waves, indicating a
486
situation that is worse than that seen by mere observa-
tion. The column on the left especially shows clear
signs of chipping and lamination of the crystalline
planes of the material with a progressive reduction of
the resistant section.
The Torre has undergone a number of maintenance
and restoration works over the past two centuries that
implemented a series of measures aimed at improving
the structure’s static equilibrium.
The surveys conducted therefore focused on con-
trolling the condition of these measures, and in partic-
ular the tension of the metal chains within the loggia
that were not located inside the walls and the presence
of steel strands within the masonry built (it is assumed)
during the restoration works conducted in 1981.
The check on the chain tension was conducted
through the dynamic analyses of free vibration.
Only the north and south chains (only two of the
four chains in the loggia) were assessed, inasmuch as
they were the only ones detached from the masonry. In
fact, this circumstance greatly compromises the reli-
ability of the analysis that was consequently not been
conducted on the metal elements facing east and west.
The analysis was conducted by disturbing the chain,
applying an impulse and recording the accelerations
produced with an accelerometer placed at the center of
the chain.The results acquired through a personal com-
puter showed the poor tension of the metal elements
that will undergo a tension enhancement procedure or
will be replaced during the restoration works.
The last analysis consisted in checking, (using a
magnetometer), the presence of metal consolidation
strands which according to the projects that we possess
were installed in 1981.
The theory that the documentation on paper referred
to was confirmed by the presence of markers on the
building’s façades that were painted with colored paint
which according to the other documents, represents the
actual position in which the building firm that carried
out the consolidation work had placed the strands.
Six readings were made on the external façades
(on the north and east fronts) and two on the internal
fronts (west façade), in order to verify the presence

Figure 10. Localization of the analysis carried out with
magnetometer.
of tie beams in correspondence with the paint traces
that the positive outcome of the analysis had consid-
ered reliable with respect to the actual position of the
structures.
The diagnostic analyses conducted (that consti-
tute the necessary projects to know the static condition
of the structure) were provided by the Regional Office
of the Lombardy Region to the structural engineer
who, on the basis of such data, will develop the security
implementation plan for the Torre itself.
4 CONCLUSIONS
The survey of the “Torre quadrata del Circo Romano”
has been based on the analysis of direct and indirect
sources, useful tools for the comprehension of the
building itself, taken within its historical evolution,
as well as of its structure and materials’ consistency.
Thanks to the archive research carried out, and to the
method used for dating historical buildings (stratig-
raphy, mapping and identifying of the different wall
construction techniques, chronotypology of apertures)
we were able to understand the construction of the
building in all its complexity.
Mapping of external and internal fronts led to the
identification of walls’ continuity or discontinuity and
to the recognition of toothing or non toothing of par-
titions, of dividing walls and of doors and windows.
Besides this it was also useful for the identification of
487
stratigraphic facts that suggested the “critical areas” in
which the conservation project will have to focus on.
Material data and geometrical aspects that have
been surveyed provide considerable information and
attest the building’s past history, giving, at the same
time, useful hints and indications on the methods to
apply for a correct approach to the intervention project.
As a matter of fact, all the information resulting
from this campaign, correlated with the direct mea-
surements, with the monitoring of verticality and with
the diagnostic analysis, let us achieve detailed and
exhaustive answers to the complexity of the prob-
lems that affect the building. The clearness and com-
pleteness of the preventive diagnostic outline, further
strengthened by petrographical, chemical and physi-
cal analysis of materials, has allowed for the correct
development of a specific conservation and structure
improvement project.
ACKNOWLEDGMENTS
The research “Diagnostic examination on the ‘Torre
quadrata del Circo romano’ (Milan)” has been car-
ried out for the Ministero per i Beni e le Attività
Culturali, Direzione Regionale per i Beni Culturali e
Paesaggistici della Lombardia (dott. Paola Marina De
Marchi) and Politecnico of Milan, Diagnostic Labora-
tory for Conservation and Reuse of Cultural Heritage
with the Observatory on Woodwork Preservation of the
Department for Architectural Design: prof. Susanna
Bortolotto with arch. Elisabetta Ciocchini and arch.
Fabio Zangheri.
The other competences has been divided as fol-
low: dott. Roberto Bugini, dott. Luisa Folli of CNR
Center “Gino Bozza” for the mineralogical anal-
ysis in “thin section” with polarised microscopy
(MO-LP); prof. Lucia Toniolo with dott. Davide
Gulotta of Dipartimento di Chimica, Materiali ed
Ingegneria Chimica “Giulio Natta”, Politecnico of
Milan for “Characterisation of the mortar beds” and
prof. Luigia Binda with laboratory technician Marco
Cucchi, Marco Antico of the Laboratory of Materials
Tests – Section “Masonry structures, stone materials,
mortars, diagnosis for the Cultural Heritage” of the
Department of Structural Engineering, Politecnico of
Milan for the “Experimental evaluation of the tension
of metal chains, evaluation of ultrasonic velocity on
stone columns and identification of metallic elements
hidden in masonry”.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Dynamic multi-body behaviour of historic masonry buildings models

D. D’Ayala & Y. Shi

Department of Architecture and Civil Engineering, University of Bath, Bath UK

C. Stammers
Department of Mechanical Engineering, University of Bath, Bath UK

ABSTRACT: This paper analyses a series of shaking-table tests on three 1/10-scale 3 dimensional dry masonry
models of same plan area and different height and total mass. Sinusoidal waves varying in frequencies based
on both constant amplitudes and constant accelerations are used as input. The aim is to gain better insight
into the behaviour of masonry historic structures subjected to seismic action by using the approach of multi
rigid body dynamics. The models are relatively simple. Yet a sufficient number of blocks is used to ensure that
the fundamental interaction among parts of masonry walls connected in 3D to form buildings is replicated to
reasonable accuracy. Results are presented in terms of relative displacement, frequency and energy content of
the motion of selected instrumented block. Global results in terms of deformed shape and crack pattern of the
walls are also discussed.

1 INTRODUCTION the full-scale system. Particularly he highlights the

Shaking table test on masonry structures are relatively
few in literature. Their results are also highly depen-
dent on specific material characteristics and hence of
modest generic utility. As it is seen in the brief liter-
ature review below somewhat contrasting results can
be often obtained.
A series of shaking-table tests on 24 ½-scale
two-storey masonry models (Benedetti et al., 1998)
assessed the importance of the original quality of con-
struction and the significant increase from even simple
strengthening on the lateral resistance. In pseudo-
dynamic tests conducted by Paquette et al., 2004, a
full-scale one-story unreinforced brick masonry spec-
imen with a flexible wood diaphragm and a gap at one
corner was excited with selected accelerogrammes.
The influence of discontinuous corners is stated to
be negligible during high intensity seismic excitation.
Stable combination of rocking and sliding mechanisms
was found under large deformations without signifi-
cant strength degradation. The influence of opening
ratios on local and global rocking has been studied by
Yi et al., 2006 with quasi-static tests on full-scale two-
story unreinforced masonry model. It was observed
that a large initial stiffness resulted in high initial
damage, with stiffness decreasing rapidly with the
increase of lateral drift. On the other hand, Adams
(Adams, 1996) stated the difficulty in direct compar-
ison of the results from scaled model with those from
gaps between behaviour of models in labs and of real
buildings because of differences in loading rates and
component characterizations with scaled materials.
Shaking-table tests on masonry models can be basi-
cally classified into two groups. The first group uses
only a few units of blocks (Casolo, 2000; Lourenco &
Ramos, 2004), which is easy to carry out and gives
clear behaviour of single rigid block motion. How-
ever, as it cannot simulate the interaction among large
numbers of units in real masonry walls, the results
and derived theoretical models are of limited use for
real structures. The second group uses large-scale
models(Benedetti et al., 1998; Griffith et al., 2004;
Paquette et al., 2004; Lourenço et al., 2005). In this
case, as the requirement of reasonable stress state at
joints makes very stiff components and large extra
loads, these models actually deviate largely from the
original structures in the geometrical forms, loading
conditions, stiffness and hence the dynamic character-
istics. Thus, results obtained require substantial post
analysis and interpretation to provide extrapolation
to real structure. However, both groups of tests have
shown that failure takes place in regular patterns whose
arrangements depend on structural and geometrical
characteristics of the wall layout both at material and
structural element or assembly level (Anderson, 1976;
Hendry, 1973; West et al., 1977). This brief review
explains the rational for a series of test on 1/10 scale
models of brick blocks 3D wall assemblies set in

489
dry work. The aim is to gain better insight into the need not to be addressed, the second is that results can
behaviour of masonry historic structures subjected to be more easily generalised to masonry structures with
seismic action by using the approach of multi rigid a variety of materials.
body dynamics. The models are relatively simple. Yet
a sufficient number of blocks is used to ensure that 2.2 Experimental setup
the fundamental interaction among parts of masonry
walls connected in 3D to form buildings is replicated Three series of models were built and tested on the
to reasonable accuracy. In the following sections the Instron Structural Testing Multi-Axis Shaking Table
background to this set of tests and the test set up (IST-MAST) system with 6 degrees of freedom, sited
are first presented, then the results discussed and a sim- at the Mechanical Engineering laboratory of the Uni-
ple control approach is proposed to provide a simple versity of Bath, UK. The control system used is
and effective modelling. the Labtronic 8800 Control Electronics with a Basic
8800 Console, control cards for each axis position
and SCM data acquisition boards-table accelerometer
2 SHAKING-TABLE TESTS monitoring and customer feedback. Models tested
have rectangular plan shape so as to collect informa-
2.1 Background tion on façades with different height to length ratio.
Panels perpendicular to y axis have 7 bricks in length,
The series of shaking-table tests presented here is while the facades perpendicular to the x axis have 5
part of an experimental program to identify dynamic bricks. They are built without windows and at corners
behaviour of masonry structures. Previous theoretic the superposition of bricks is equal to half the length
and experimental studies conducted by D’Ayala & resulting in competent connections. The bricks have
Speranza; 2003 Restrepo et al., 2004 and Shi & dimensions (L × W × H) 100 × 50 × 35 mm, and they
D’Ayala, 2006; Shi et al., 2008, have shown how using are regularly staggered to provide s/h = 50/35 = 1.4,
limit state analysis and pseudo-static tests is possible which from previous experimental and analytical stud-
to derive a consistent model of behaviour of cracking ies (D’Ayala, 2005, Shi et al., 2008) has shown to
and damage of historic masonry subjected to lateral provide good equivalent shear strength to the masonry.
action and how the behaviour can be correlated to a Model dimensions are shown in Figure 1. The first
relatively small numbers of geometric and structural and second series of models were built with 9 courses
parameters, without relying on stress analysis. While
results of these studies are also confirmed by in situ
observation of damage to buildings subjected to earth- x
quakes, their static nature fails to provide insight in
the damaging process and hence fails to accurately
quantify the strength and “ductility” resources that are
available during the hysteretic behaviour. It is argued y
here that notwithstanding the fact that the constituent
materials of masonry, bricks or blocks and mortar, are
not ductile, substantial dissipation of energy can take
place during the damaging process at the cracks inter-
faces due to sliding and rocking of portions relative to
each other, and hence via friction and impact.
As the processes outlined above are post material
capacity, both in tension and compression or shear, it (a)
is not essential to replicate the continuum nature of the
masonry. This is the main reason for adopting models
in dry masonry so that the emphasis is directly on the
post-elastic behaviour. It may be argued that the lack of
mortared joint would increase the dissipating capacity
of the scale model with respect to a real masonry struc-
ture, because of the possibility of each unit to randomly
move with respect to the adjacent one.This indeed does
not occur, as verified by the tests, because the grav-
ity load distribution is proportional to the real case
and hence actual relative movement does indeed occur
only along main lines of crack pattern. Neglecting the (b)
elastic phase has two advantages: the first is that in
scaling down the model the considerations on stresses Figure 1. Plan (a) and elevation (b) of the three models.

490
of bricks, while the third series had 15 courses to inves-
tigate the difference in height to length ratio further
and to study the effect of different area to mass ratio.
The roof structure is simulated by five timber beams
set a regular spacing on the top course with weights
bolted at each end to simulate the actual roof mass and
to ensure proper simulation of the constraining effect
of floor structures onto the wall. Of course the issue of
rigid or flexible diaphragm action is not addressed at
all here. However, to obtain a more even distribution of
mass and constraint, especially over the corners, series
II models have extra weight applied to the corners
for a total 20% more with respect to series I models
(Figure 2).
Besides the accelerogrammes registering the six
component of motion of the shaking table, each speci-
men is instrumented with 6 displacement transducers,
3 for each of the x and y alignment, set on the cen-
tral brick and the corner bricks on each side. As the
bricks of the top layer are less stable without mortar,
in order to represent the global shaking, displacements
of the 2nd top layer are measured. Additionally, there
is one more transducer in the middle of the 8th layer in
each direction for Model III. Details of the transducers
layout are shown in Figure 3.
Using a simple sinusoidal input with 50-cycle dura-
tion, each of the models is subjected to different group
of tests where only the amplitude to the accelera-
tion of the signal was increased, as summarised in
Table 1. One of the objectives of the tests is to clar-
ify whether for a same energy input greater levels
of damage are triggered by acceleration or increased
amplitude of the motion. A second objective is to iden-
tify whether the response is frequency sensitive, in
other words, notwithstanding the fact that the speci-
men is not a continuum, whether something akin to
natural frequency or range of frequency for which the
response is enhanced, can be identified. Hence the
input frequency varies following two principles: (1)
under constant amplitude. This is used on Model I
to identify resonant frequencies. (2) Under constant
peak acceleration. To ensure this, as the input wave
is defined in terms of frequency and amplitude, the
relationship between frequency and amplitude of two
successive input series is (fa )2 /(fb )2 = Ab /Aa , where
f is the input frequency and A is the corresponding
amplitude. For Model II Series, the effect of input

Figure 3. Displacement transducers setup.

Table 1. Group classification of shaking-table tests.

Group No. of tests Direction Constant Series

1 8 y Amplitude I
2 5 Acceleration
3 5 x Amplitude
4 5 Acceleration
5 4 y Acceleration II
6 5 x Acceleration
7 7 x, y Acceleration
8 9 y Amplitude III
Figure 2. (a) Series I and II and (b) Series III set up model 9 3 y Acceleration
setup.

491
non parallel to the structure’s walls was also consid- (a) Constant input amplitude
ered by providing sinewaves of equal amplitude and 2
displacement in the two coordinate directions. 1.8 x direction
1.6 y direction

Amplification
3 EXPERIMENTAL RESULTS ANALYSIS 1.4
1.2
Both pseudo-static tests. (Shi & D’Ayala, 2006; Shi 1
et al., 2008) and extensive on-site post earthquake 0.8
observation have shown that under lateral action,
0.6
masonry walls are more vulnerable to out of plane
than in-plane failure, even for shape ratio H /L < 1, 0.4
and slenderness ratio H /t < 11, as it is the case for 0.2
this series of tests. This was confirmed by the obser- 0
vation of both the recorded videos of the shaking and 1.6 2 2.4 2.8 3.2 3.6 4 4.4 4.8 5.2
the comparison of overall displacement of in-plane Input frequency
and out-of-plane instrument for any given motion test.
However differently from the monotonic static tests (b) Constant input acceleration
carried out on single walls with wings, the inversion 300
of motion caused by the sinusoidal wave results in a x direction
much stronger interaction at the corner between par- Dissipated energy (J) 250 y direction
allel and orthogonal walls with effects on the two sets
of walls which will be further discussed in detail in 200
the following subsection where results are presented
separately for each series of tests. 150

100
3.1 Analysis of results on Series I models
In Figure 4 maximum amplification of motion and 50
dissipated energy for series of test with increased fre- 0
quency and constant amplitude of 12 mm and constant 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6
acceleration of 1.53 m/s2 are shown for the two coordi-
nate directions of motion. The y direction relates to the Input frequency
out of plane motion of the longer façade. This shows
a peak amplification of 2 for 2.0 Hz, with a distribu- Figure 4. Amplification and energy response for constant
amplitude and acceleration for Series I model.
tion very similar to an amplification spectrum. For the
shaking in x, peak response is associated to an input
frequency of 1.9 Hz.and a value of 1.5, with a second
peak at 3.6 Hz. Assuming a simple 1 degree of free-
dom oscillator as basic model for the behaviour of the
walls would yield values of stiffness 29.5 KN/m and
15.7 KN/m respectively, contrary to expectations.
This can be explained by the fact that the floor struc-
ture afforded better friction restraint in the y direction
shaking, resulting in overall greater stiffness.
In the case of constant input acceleration, and hence
constant input energy, it is observed that for increas-
ing input frequency damage and collapse occur for a
smaller number of cycles. This phenomenon is sum-
marised clearly in Figure 4b, where the total energy
dissipated in each test is measured against the value of
input frequency. Figure 5. Damage on side panel of Series I models with
Figures 5 and 6 show that similar portion of the walls excitation in y as 2.2 Hz, 12 mm.
are excited in both directions for each peak amplifica-
tion case. In the photos the panel parallel to the motion line shows the relative sliding of blocks in the direc-
are shown. The thick lines at the edge of the panels tion of motion. It should be noted that these two crack
shows the portion of this that participate to the out of patterns relates closely to the theoretical assumptions
plane motion of the orthogonal walls while the thinner taken in D’Ayala & Speranza, 2003 to derive out of

492
 x-1.9Hz 7th cycle
2
positive peak
middle point
negative peak 1
positive peak 2
0
-18 -12 -6 0 6 12 18

Panel // y
-1

-2

-3

Figure 6. Damage on side panel of Model I with excitation

in x as 1.8 Hz, 12 mm. -4
Panel // x
y-2.0Hz 12th cycle Figure 8. Deformed shapes by excitations with constant
15
amplitude in x.
positive peak
negative peak
zero point 5 Table 2. Difference in input–output phase angles of Model I
positive peak 2 shaking in x direction.

-45 -35 -25 -15 -5

-5 5 15 25 35 Input Input
Panel // y

Frequency Amplitude Lagging Phase difference

(Hz) (mm) (s) (◦ )
-15
1.7 12 0.06 5.84
1.8 12 0.08 8.25
1.85 12 0.08 8.48
-25
1.9 12 0.06 6.53
2.2 12 0.08 10.08
2.5 12 0.1 14.32
-35
Panel // x
displacement are comparable and the overall deformed
Figure 7. Deformed shapes by excitations with constant
amplitude in y. shape is stable. It is also evident that there is modest
coupling of shaking with the orthogonal wall.
When shaking occurs in the x direction instead, lag-
plane mechanisms collapse load factors for façades. ging is much more substantial, there is no overlapping
Particularly as already outlined there the crack closer of the two subsequent cycles, but the phenomenon
to the corner will occur before the wing crack if the which is perhaps most apparent is that the movement
connection is not particularly strong, as it is the case at the corner is of the same order of magnitude than
here for lack of overweight or restraint from the floor the movement at the centre of the façade, providing a
structure directly at the corner. This lack of restraint clear measure of the torsional effect introduced earlier.
causes a torsional motion of the end blocks of each This also results in substantial coupling for the longer
side and this eventually influences greatly the overall orthogonal wall.
out-of plane movement of the wall. This is illustrated The superposition of effects of corner rotation and
in more detail by the motion profiles shown for differ- shaking of the façade has also the effect of accelerating
ent frequency inputs and different cycles in the y and the progression of damage. Thus, rotation at corners is
x direction respectively in Figures 7 and 8. more likely to influence narrower panels, and contrary
In each graph the recorded displacement of each to what expected the excitation in the x direction causes
of the six instruments is plotted for a positive peak, collapse for smaller energy input.
0 point, negative peak and subsequently positive peak The difference between input and output phase
of the input, providing basically the “modal shape” of angles ϕ are listed in Tables 2 and 3 for the shaking
the two orthogonal walls. in x and y. The further the input frequency deviates
Figure 7 shows that the long façade has relatively from the natural frequency, the larger the response
stable cycles, and although the response lags with deformation lags behind. Higher flexibility of the
respect to the input, maximum positive and negative facade also causes larger lagging phase. In Table 2, for

493
Table 3. Difference in input – output phase angles of When the maximum response amplitude exceeds the
Model I shaking in y direction. input amplitude the top starts drifting more and more
while the oscillations reduce in amplitude. When the
Input Input frequency is closer to the resonance the first oscilla-
Frequency Amplitude Lagging Phase difference
tion has the greatest amplitude and then the subsequent
(Hz) (mm) (s) (◦ )
ones are slightly smaller but associated with substan-
2.0 9.7 0.16 18.33◦ tial drifting, although overall slightly smaller than the
2.0 12 0.2 22.92◦ previous case.
2.2 12 0.2 25.21◦
2.5 6 0.16 22.92◦
2.5 12 0.18 25.78◦ 3.2 Analysis of results of Series II models
In order to further study the effect of corner restraint
and decoupling of the two orthogonal walls, extra
(a) Excitation of 1.8Hz, 12mm weights were placed directly at the edge of each wall.
20 This addition resulted in a slight increase in natural
10 frequency from 2.0 Hz to about 2.1 Hz in y direction
and from 1.9 Hz to 2.0 Hz in x direction with respect
0 to the Series I models with a substantially smaller
Displacement

-10 5 10 15 20 25 30 amplification in x, but greater in y. This confirms

the decoupling of the two motions when the cor-
-20
ners are prevented from rotating. While in the present
-30 experiment this was simply provided by increasing the
-40 frictional restraint by way of a more even distribu-
tion of applied mass, it clearly points out the necessity
-50 of ensuring the 3d integrity and box behaviour for
-60 masonry structures. As seen from comparing Fig-
Time (s) ures 4b and 10b, the overall dissipation of energy in
the Series II models is slightly lower, in agreement
(b) Excitation of 2.0Hz, 12mm with the restrained effects at the corners. The previ-
25
ous observation is also confirmed by the deformed
15 shape plotted in Figure 11a and 11b, showing, espe-
5 cially for the shaking in y direction, more regular and
lower magnitude deformations.
Displacement

-5 Further investigation of the coupling effect and the

12 14 16 18 20 22 24 26 28 30 32
-15
-25
-35
-45
-55
Time (s)
Figure 9. Time–displacement curve for Model I shaking
in y.
constant amplitude, lagging increases with input fre-
quency except for the resonance frequency of 1.9 Hz.
In Table 3 it is shown that phase lag increases with
amplitude for the same frequency and with frequency
for the same overall acceleration.
This behaviour is further highlighted by the plots of
displacement versus time for two different frequencies
for the same direction of motion and amplitude, shown
in Figure 9. For frequency below resonance the initial
cycles do not produce substantial excitation until the
phases change slightly and set off greater oscillation.
influence of rotation at corners, was investigated by
subjecting Series II models to forcing sinusoidal waves
in the direction of their plan diagonal. Maximum
displacement was set at 9 mm for a total constant accel-
eration of 1.26 m/s2 and a variable range of frequency.
In all cases instruments at midspan panels record
amplifications substantially smaller than 1 showing
that out of plane movement is modest while instru-
ment on corners show significant drift, in agreement
with substantial shear deformation of the overall panel.
As a result the energy dissipated in out of plane move-
ment is very small compared with the previous two
cases (see Figure 10b).
3.3 Analysis of Series III models
The Series III models have a H/L ratio of 1 for the
small panel and of 0.72 for the larger panel. Condition
of constraint and loading are unchanged from pre-
vious series. However smaller input amplitudes have
been chosen. Contrary to the increased geometric slen-
derness the natural frequency increases to 2.5 Hz for
shaking in the y direction, showing that the increased

494
 2.7 (a) x-2.0Hz 28th cycle
(a) in x in y 4
2.4 positive peak
negative peak
2.1 zero point 2
positive peak 2
Amplification

1.8 0
1.5 -15 -12 -9 -6 -3 0 3 6 9 12
-2
1.2

Panel // y
0.9 -4
0.6
-6
0.3
-8
0
1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6
-10
Input frequency
-12
240 Panel // x
(b) in x
210
in y (b) y-2.0Hz 16th cycle
Dissipated energy (J)

180 diagonal 4
positive peak
150 negative peak
zero point 2
120 positive peak 2
0
90
-10 -8 -6 -4 -2 0 2 4 6 8 10 12
60 -2
Panel // y

30
-4
0
1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 -6
Input frequency
-8
Figure 10. (a) amplification and (b) energy dissipation
response with different input frequencies for Series II models -10
for constant acceleration. Panel // x

Figure 11. Deformed shapes of Model II.

height also results in increased lateral stiffness due to
better frictional behaviour among the blocks.
The shaking of the façade has still an arch shape
involving only the upper 8 layers of the façade and
the 6 top of the side panels. Rotation at corners is
still visible but substantially reduced compared with
previous models. The participating mass associated to
the shaking is proportionally reduced when compared
with previous series.
Finally in order to compare the overall behaviour
of series II and III models their adimensionalised drift
for the two groups is plotted in figure 13. The unit top
drift is calculated as follows:

Unit drift = (Displacement of top layer/Input

amplitude)/Height

in order to compare shakings for different amplitudes

but overall same energy input. It can be seen that while
for the second series the value of drift increases with
frequency, for the third series this reduces in the post Figure 12. Façade view of the model at the end of the 2.5 Hz
natural frequency zone. 50 cycles motion.

495
 0.024
0.021 Model II Model III
0.018
Unit Top Drift
0.015
0.012
0.009
0.006
0.003
0 REFERENCES
1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6
Input frequency
Figure 13. Drift of Model III with constant input amplitude.
4 CONCLUSIONS
In the previous sections the results of 3 series of tests
for a total of over 50 geometry/amplitude/frequency
combinations have been presented in terms of global
response parameters. From these, it would appear
that notwithstanding the discontinuous nature of the
masonry fabric, having bricks in simple dry contact, a
behaviour that can be described in terms of natural fre-
quencies and resonance can be identified, supported
by observations in terms of both amplification and
energy dissipation. Constraint conditions at corners
are also critical to the behaviour and while they do
not seem to influence either the façade portion taking
part in the shaking or substantially the natural fre-
quency. However the amplification are substantially
reduced especially away form resonance. For frequen-
cies closer to the resonance level, the apparent greater
stiffness of the model and the fact that it dissipates
less by relative rotation at the corners, means greater
overall out of plane deformation at the centre of the
façade and hence greater damage in this area.
In summary the behaviour and hence the collapse
limits are due to the superposition of an overall motion
of the walls panel that can be reduced to beam oscilla-
tion horizontally and cantilever oscillation vertically;
however to this is superimposed a relative sliding of
bricks of subsequent courses and a rotation of some
of them around a vertical axis, mainly caused by the
staggering. This rotation is initiated at corners and
propagates. No relevant rocking of individual bricks
was observed. On the basis of these observations a
multi-body dynamic model is being developed.
496
ACKNOWLEDGEMENTS
This project is developed with support from the
Dorothy Hodgkin Award scheme and from the Royal
Society international collaboration scheme. Thanks
go to the technicians of the Structures Laboratory of
the Department of Architecture and Civil Engineering
and of the IST MAST laboratory in the Mechanical
Engineering Department of the University of Bath.
Adams, D., 1996. Effects of scale and loading rate with tests of
concrete and masonry structures. In Earthquake Spectra
12: 13–27.
Anderson, C., 1976. Lateral loading tests on concrete block
walls. In The Structural Engineer 54: 239–246.
Benedetti, D. et al., 1998. Shaking table tests on 24 sim-
ple masonry buildings. In Earthquake Engineering and
Structural Dynamics 27: 67–90.
Casolo, S., 2000. Modelling the out-of-plane seismic
behaviour of masonry walls by rigid elements. In Earth-
quake Engng Struct. Dyn. 29: 1797–1813.
Griffith, M. et al., 2004. Experimental investigation of unre-
inforced brick masonry walls in flexure. In Journal of
Structural Engineering 130: 423–432.
Hendry, A., 1973. The lateral strength of unreinforced brick-
work. In The Structural Engineer 51: 43–50.
Lourenço, P. et al., 2005. Dry joint stone masonry walls
subjected to in-plane combined loading. In Journal of
Structural Engineering 131: 1665–1673.
Lourenco, P. & Ramos, L., 2004. Characterization of cyclic
behavior of dry masonry joints. In Journal of Structural
Engineering 130: 779–786.
Paquette J. et al., 2004. Seismic testing of repaired unrein-
forced masonry building having flexible diaphragm. In
Journal of Structural Engineering 130: 1487–1496.
Restrepo-Velez, L.F. & Magenes, G., A simplified
mechanics-based procedure for the seismic risk assess-
ment of unreinforced masonry buildings, in Rose school,
118, 2003.
Shi, Y. & D’Ayala, D., 2006. Analysis of in-plane dam-
age of unreinforced masonry walls, in the 7th Inter-
national Masonry Conference. British Masonry Society,
London, UK.
Shi, Y. D’Ayala D. & Jain P., 2008. Analysis of out-of-plane
damage behaviour of unreinforced masonry walls, in the
14th International Brick and Block Masonry Conference
Sydney, Australia.
T. Yi et al., 2006. Lateral load tests on a two-story unre-
inforced masonry building. In Journal of Structural
Engineering 132: 643–652.
West, H. et al., 1977. The resistance of brickwork to lateral
loading. In The Structural Engineer 55: 411–421.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Assessment and strengthening masonry arch bridges

C.L. Brookes
Gifford, Southampton, UK

ABSTRACT: This paper describes how during the last 10 years Gifford has been applying a novel numerical
technique to accurately assess the strength of masonry arch bridges, and for weak bridges, to design repairs and
strengthening. This has involved a major development programme including full-scale laboratory tests, supple-
mentary load tests on bridges in the field, monitoring programmes and verification of the arch bridge structural
analysis which has been based on the Finite/Discrete Element technique. The advantages this technique provides
over conventional masonry arch bridge analyses, both mechanism and traditional Finite Element modelling, are
described and how through partnering an innovative assessment and strengthening service is being delivered. The
relevance of this approach to emerging serviceability limit state arch bridge assessment, which is seen as being
particularly important for Railways, is also discussed. Some practical issues relating to bridge strengthening and
working within national codes of practice are also covered.

1 INTRODUCTION
It is likely that there are at least half a million masonry
arch bridges in use throughout the world today, princi-
pally carrying road and rail. European railways alone
account for 200,000 bridges (Orbán 2004). These
bridges form a vital asset. Their replacement cost is
almost incalculable yet a world wide insatiable appetite
for economic growth is in some cases pushing their use
to the limit.
Despite being ancient in form, masonry arches
are notoriously difficult to accurately assess. At all
limit states their behaviour is complex, deriving their
overall behaviour from the interaction of individual
parts, blocks, bricks, mortar and fill. Several meth-
ods for assessing the strength of arch bridges have
become well established, but their generalised use is
limited and their application for designing strength-
ening difficult. Finite Element analysis, which has to
be non-linear to predict strength, has also been suc-
cessfully applied but the choice of tensile material
properties can be problematic as this can artificially
influence the predicted strength.
The Finite/Discrete Element method, which
involves the automatic computation of interacting
bodies is, therefore, a natural choice for representing
masonry and this type of non-homogenised structure.
Like the conventional Finite Element method, being
a generalised approach also means that, subject to
verification, any geometric form of masonry can be
simulated. Consequently, there are no restrictions to
the arch bridge form, the number of spans, rings and
piers that can be modelled. Also, unlike many simpler
strength assessment methods, there is no adherence to
predetermined failure mechanisms, for instance, a set
number and pattern of hinges.
The application of the Finite/Discrete Element
method has marked a step change in the rigour that can
now be applied to the structural analysis of masonry
arch bridges. Not only can it be used to accurately
assess strength but also to determine bridge defor-
mation, including all significant non-linear effects,
making it possible to assess behaviour at both strength
and serviceability limit states. Also, being a gener-
alised approach the behaviour of complex bridges
can be assessed where for example a concrete sad-
dle may exist, or the bridge is propped and in the
case of strengthening, retrofitted reinforcement is
introduced.
Through partnering with Cintec International who
manufacture and install a masonry anchor system,
Rockfield Software who produce the ELFEN (Rock-
field 2003) structural analysis software, Gifford have
completed over 170 bridge assessments and bridge
strengthening designs, mainly in the UK but also in
the USA, Australia and India. Known as Archtec, this
service was originally conceived for efficient, eco-
nomic and sympathetic strengthening of arches, but
the method of structural analysis can also provide
accurate strength assessment of existing bridges and
on many occasions has been used to show that bridges
do not need to be strengthened.

497
2 CONVENTIONAL ASSESSMENT
Methods of strength assessment have been categorised
(McKibbins & Melbourne 2006) as semi-empirical,
limit analysis and solid mechanics methods.
2.1 Semi-empirical methods
Most semi-empirical methods are based on the MEXE
(Military Engineering Experimental Establishment)
method which evolved from work undertaken in the
1930s for the military to rapidly assess arch bridges.
It is often still used as a first pass strength assessment
but its use is highly subjective and there are many lim-
itations. It is of little value for any detailed work such
as the design of strengthening.
2.2 Limit analysis methods
Most conventional bridge assessments are now carried
out using computerised versions of limit analysis also
known as mechanism analysis. In its simplest form
these methods consider a 2D arch comprising a series
of blocks of infinite compressive strength, which can-
not slide against each other and cannot carry tension.
A routine is used to establish the locations of hinges in
the span followed by calculations of reaction and then
vector algebra to position the resultant line of thrust.
The method produces a lower bound solution. In other
words, if a load path can be found that lays entirely
within the masonry then the modelled arch is capable
of sustaining that load even if it is not the true load path.
Limit analysis techniques have proved to be excel-
lent tools for first phase strength assessments but
several restrictions exist that are important in the
design of strengthening. The most important of these
is the inability to calculate strain and displacement.
Consequently, it is not possible to determine the distri-
bution of stress at operational load levels, it is difficult
to assess the serviceability of bridges and in the case
of strengthening, it is not possible to determine the
share of load between the existing bridge and the
strengthening.
2.3 Solid mechanics methods
The established technique used to model continuum
based phenomena in solid mechanics such as deforma-
bility is the Finite Element Method (FEM). Not sur-
prisingly this has also become the most popular solid
mechanics method used for arch bridge analysis, and
there are numerous well developed industry quality
computer programs available.
Like limit analysis most work is carried out using
2D representations, generally plane strain, but 3D shell
and solid models are used for special assessments.
Although these techniques can be good for
determining displacements, strains and stresses at
498
operational load levels they quite often become dif-
ficult to use to predict ultimate strength and damage.
This is generally because of the type of solver that
is used, normally an implicit solver, and the effort
required to ensure internal forces are in equilib-
rium with external loads, as brittle materials such as
masonry soften and redistribute load. The solution to
the equilibrium problem is normally to use a hypothet-
ical masonry tensile strength but choosing a suitable
value, large enough to achieve equilibrium conditions
are met but small enough not to influence the result,
can be a challenge.
3 THE FINITE/DISCRETE ELEMENT METHOD
3.1 Description
Numerical techniques have been devised to represent
discontinua where body or particle interaction defines
overall behaviour (Cundall 1971). Perhaps, the most
advanced technique that describes this behaviour is the
Discrete Element Method (DEM). The relatively new
Finite/Discrete Element Method (FDEM) (Munjiza
2004) is a combination of FEM and DEM and pro-
vides a more natural approach to the simulation of
many materials and structures. It has been applied to
a diverse range of engineering and scientific prob-
lems from food processing to rock blasting. Through
automated adaptive modelling, even the transition
from continua to discontinua and the fracturing and
fragmentation process can be represented.
FDEM is aimed at problems involving transient
dynamic systems comprising large numbers of
deformable bodies that interact with each other. Mod-
els involve typically thousands, but in extreme cases
millions, of separate Finite Element meshes auto-
matically interacting with each other using DEM
contact algorithms. The solution of the continuum
equations associated with FEM is well established, the
algorithms within DEM less so.
Contact detection and contact interaction lay at the
heart of DEM. Contact detection is aimed at iden-
tifying discrete elements that can potentially come
into contact with each other and eliminating those
far away from subsequent contact interaction algo-
rithms. Different algorithms have been developed for
different packing densities for example, sparse and
moving or dense and static. The chief aim here is to
reduce computing effort. Contact interaction applied
to the surfaces of discrete elements coupled through
the detection process is where interface behaviour is
calculated. Here interface laws are applied according
to the surface characteristics of the contacting discrete
elements for example, frictionless no-tension contact.
During the solution of transient dynamic problems of
even quite modest size millions of contacts will be
detected and resolved.
 Another key aspect of FDEM is that the analysis
involves all equations of motion, is therefore dynamic
and uses an explicit central difference solution scheme
(Owen 1980). This involves a time stepping procedure
that is conditionally stable, but unlike many conven-
tional FE solvers that use an implicit solution scheme,
does not involve computationally intensive matrix fac-
torisation. Solutions are achieved only through the use
of very small time steps. The critical time step size Figure 1. Part of a modelled masonry arch barrel showing
FE mesh (left) and DE mesh (right).
below which steps must remain for stability and accu-
racy is given by the time taken for a stress wave to travel
across the smallest finite element. The efficiency of
DEM contact detection and the avoidance of equilib-
rium calculations allows FDEM simulations to predict
failure, collapse and post-failure kinematic behaviour.

3.2 Application to masonry arch bridges

Masonry is a non-homogenised material, can be
regarded as a discontinuum and as such is ideally
suited to FDEM. Simply, masonry arch bridges are a
special form of masonry structure, which is an impor-
tant consideration when faced with complex bridge
arrangements.
The approach that has been developed for arch
bridges, applied using the implementation within
the Finite Element computer program ELFEN, uses
smeared masonry compressive properties and explicit
mortar shear and tensile properties. Each brick or block
unit is modelled with a separate finite element mesh
and each unit becomes a single discrete element. It
has been found that units can also be grouped together
(Brookes 1998), a blocky arrangement of four or five
bricks glued together, to improve computational effi-
ciency without any loss of accuracy. The masonry arch
is then assembled using blocky arrangements in hun-
dreds, possibly thousands of discrete elements. Other
bridge parts for example, fill, surfacing and backing
are similarly represented although the material models
are different.
FDEM arch bridge models will develop failure
mechanisms consistent with limit analysis results if
these are critical as well as providing displacements,
stresses and strains consistent with solid mechanics.
Another key aspect to the use of FDEM and
the adopted modelling approach is that representing
masonry at a fundamental scale requires only com-
monly available and basic material parameters to be
used in order to accurately characterise bridge struc-
tural behaviour. Non-linear material models are used
to define the deformable behaviour of the masonry in
compression and the fill in tension. A perfectly plas-
tic Von Mises yield criterion is generally used to cap
compressive strength, and a Rankine yield criterion
used to give a simple no-tension soil model.
The behaviour of mortar, as well as other contacting
surfaces such as masonry to fill, is included by using
interface material models. Interface models give the
Figure 2. Historic masonry ovens, FDEM simulation used
to assess residual strength of an unusual arrangement of
arches. Dark contours show areas of highest compression.
surface of discrete elements appropriate mechanical
properties. Mortar is represented differently depend-
ing on the type of construction. Historic construction
involving lime mortar joints is represented using a no-
tension Mohr-Coulomb friction relationship. Modern
masonry with cement mortar produces masonry with
some tensile strength. In these instances good pre-
dictions of masonry behaviour can only be made by
including mortar tensile strength and a fracture energy
formulation to model the development of cracking.
Generally, masonry arches are historic constructions
and do not include cement mortar.
For most types of masonry the generic material
characteristics, compressive strength, Young’s modu-
lus, mortar friction and mortar cohesion, necessary
for FDEM simulations are readily available (Hendry
1990, BD 21/01 2001, BS 5628-1 2002). They are
no more demanding to obtain than those parameters
required for conventional limit state analyses. An esti-
mate for Young’s modulus for different types of fill in
compression is similarly available.
There are no limitations to the geometric arrange-
ments of arches that can be represented with
FDEM other than those associated with computational
resource. However, models are kept as simple
as possible to reflect the confidence in material
parameters, geometric arrangement and to be reason-
ably compatible with the codes of practice through

499

Figure 3. A sequence of snap shots showing a typical FDEM
live load assessment process and this instance failing. Dark
contours show areas of highest compression.
which all design and assessment work has to be under-
taken. Hence, the large majority of simulations are 2D
and plane strain.
In assessment and design, live load is generally
applied by explicit representation of axle loads using
discrete elements. Weight is applied to these elements
and the axles moved across the span with a pre-
scribed velocity. As transient dynamic solutions are
obtained, regard has to be given to acceleration arising
from sudden movement and inertia effects. Conse-
quently, loads are applied smoothly and slowly to
ensure static responses are obtained. Permanent loads
are introduced through construction sequences, which
depending on the barrel shape, may necessitate the
use of modelled temporary false work; a process that
is always required with real arch bridges.
Although the time required to develop FDEM
bridge models exceeds that of comparable limit anal-
ysis representations, these models can still be assem-
bled in several hours. Also solution times, which are
500
Figure 4. FDEM simulation showing use of temporary
formwork to support the barrel whist dead load is applied.
Figure 5. A typical arrangement of reinforcement which in
this instance is shown installed from above.
continuously tumbling as ever faster computers
become available, are modest compared with similar
FEM representations with strength analysis completed
in around 4 hours for a typical bridge on a 3.6 GHz PC.
This includes the calculation of permanent loads and
the traverse of a single vehicle. To complete an assess-
ment or design, several axle arrangements have to be
considered to be sure the critical case is identified, so
this could take several days. With relatively small prob-
lem sizes, around 5,000–10,000 degrees of freedom,
mass scaling techniques to accelerate the solution pro-
cess are never used to obtain solutions, but are useful
to quickly check the simulation process.
4 ARCHTEC
4.1 Description
Archtec strengthening has been described as ‘Key
Hole Surgery’ for bridges because of the absence of
any major intervention to the arch barrel. Generally
construction comprises retrofitting stainless steel rein-
forcement around the circumference of the arch barrel.
The reinforcement is grouted in to holes drilled in to
the bridge with a coring rig from the road surface or,
alternatively in the case of multi-span structures, from
below.
Arches conventionally fail by the development of
four hinges leading to a mechanism. The design basis
for the strengthening is to locate the reinforcement so

Figure 6. Installation from below in a multi-span bridge
using precisely aligned diamond drilling coring equipment.
as to provide bending strength at the critical locations
thereby resisting the development of the hinges. By
providing bending resistance the arch barrel is able to
resist the critical loading conditions more efficiently
and the peak compressive stresses in the masonry
are reduced. A similar procedure is applied to more
complex arrangements including multi-span arches
although failure mechanisms and anchor positioning
requires anchors to be placed in different positions.
4.2 Benefits
Compared with conventional arch bridge strengthen-
ing such as concrete saddling and lining, the Archtec
service has several practical benefits:
1 Through the use of FDEM a good assessment of
existing strength and bridge behaviour is obtained.
This allows accurate matching of strengthening to
the loading requirements if the bridge is under
strength, thus minimising any intervention. Alter-
natively, strengthening may be avoided.
2 Strengthening is invisible which is particularly
important for historic and heritage bridges.
3 Construction is small scale and fast to implement.
4 Disruption to bridge users during strengthening is
much less than saddling.
5 A more sustainable solution with lower envi-
ronmental impact, embodied energy and carbon
emissions.
6 Because displacement and strain is predictable
assessments and strengthening designs can be
based on limits states other than purely ultimate
strength.
7 Each anchor installation provides a core of infor-
mation that can be used to confirm the materials
and internal arrangement of the bridge.
8 In many instances all these factors equate to
reduced cost.
4.3 Working with codes of practice
Archtec services have to be provided within a frame-
work which embraces as far as possible national codes
501
of practice. Unfortunately, outside of the UK, there
are few rules to help Engineers assess arch bridges.
For example live loading is almost always developed
for beam arrangements of bridges where load support
is primarily through bending, masonry strength assess-
ment is often permissible stress based, and bridge
specific earthquake rules are geared towards steel and
concrete construction. In India the railways have a code
of practice for masonry arch bridges which impose
almost arbitrary performance limits on deflection.
The use of FDEM to simulate arch bridges is
a performance based method, useful for limit state
assessment and design, but cannot be directly used for
rules that have been developed for linear, often inac-
curate, working stress approaches. In these instances
to satisfy bridge technical authorities hybrid analyses
are run along side the more realistic and reliable limit
state work. The results allow additional checks to be
made with local codes of practice.
5 VERIFICATION
The process which has been undertaken to verify the
FDEM analytical methods employed by Archtec have
included a number of key strands and evaluations as
follows:
1 Against conventional methods of arch assessment.
2 Against published data from full-scale tests of
unstrengthened arches carried out by others.
3 Against full-scale tests by TRL of bridges strength-
ened by the Archtec method which were specif-
ically commissioned as part of the verification
process.
4 Against the results obtained by monitoring bridges
in the field including before and after strengthening
comparisons.
Additionally, a philosophy of fixing material param-
eters for whole series of tests where similar masonry
construction has been employed (compressive strength
of bricks, mortar type etc.) has been adopted. This
makes it impossible to adjust an individual arch anal-
ysis to gain better correlation with tests within a series
without influencing all others. Similarly, the analysis
of Archtec strengthening follows on from verified and
fixed unstrengthened analyses.
A selection of the verification work (Brookes 2003)
and recent field trials illustrating the accuracy and
flexibility of FDEM arch simulation follows.
6 FULL-SCALE ARCH TESTS
6.1 Strengthened arches
In order to test the practical implementation of
Archtec, to validate the FDEM method of structural
analysis, to help quantify key strength parameters and

Figure 7. First Archtec full-scale test arrangement.
to illustrate the degree of strengthening that could be
archived, two full-scale tests of Archtec strengthening
were carried out at TRL. Both tests used partially ring
separated brick forms to be representative of arches
that would warrant strengthening, and were also simi-
lar to unstrengthened bridges tested in an earlier series
so that experimentally derived comparisons could be
easily made.
The anchor arrangements were configured for a sta-
tionary point load test and, therefore, were arranged
asymmetrically with respect to the span. In prac-
tice, with moving axle loads, anchor arrangements
are arranged symmetrically to reflect critical loading
positions on both sides of the span.
Comparing the load versus displacement results
obtained by FDEM simulation with those obtained
from the strengthened test shows strength predictions
to be within 2% of test results. There is also very
good stiffness correlation, displacements remaining
within approximately 5% of test values, throughout
the loading stage.
Making comparisons between the two tests,
strengthened versus unstrengthened, show the failure
load of the strengthened arch barrel has been increased
by a factor of approximately 2, the anchors delayed the
formation of hinges and added considerable strength
to the arch barrel, and the arch failed in a gradual
and a ductile manner. In practice the characteristics of
the arch barrels are improved sufficiently for intended
loading.
502
Figure 8. Comparison of load versa deflection charac-
teristics between the first Archtec test results and FDEM
predictions. Deflection is measured in the barrel at the span
position of the load.
Figure 9. Comparison of strengthened and unstrengthened
test results.
6.2 Observations relating to serviceability
No clear definition of serviceability exists for masonry
arches. Deflections and cracking behaviour is nor-
mally used to define a serviceability limit state. How-
ever, in arches these quantities are generally small and
very difficult to detect under expected service loads
and they cannot be calculated by conventional struc-
tural analysis. However, results from monotonic and
cyclic load tests have been used to derive masonry
stress limits in terms of a limiting factor of the ulti-
mate capacity below which permanent damage does
not occur from repeated loading.
Based on work done by TRL in the 1980’s, the
Highways Agency assessment standards for arches
are based on serviceability being maintained provided
applied loads do not exceed half the ultimate capacity.
Cyclic loading on bridge piers has been investigated
by British Rail Research and some progress made
in linking fatigue of brickwork with a serviceability
limit state. It was concluded that, for dry brickwork, if
applied loads do not exceed half the ultimate capacity
an infinite number of load cycles could be sustained.
However, for saturated brickwork lower load levels are
required.

Figure 10. Comparison of maximum predicted compressive
stresses between strengthened and unstrengthened bonded
rings.
Both observations of monotonic loading and cyclic
loading have led to the recommendation of a 50% rule
and are in effect stress limit based. The current Archtec
design method, which uses load factors based on the
UK Highways Agency standards, embraces the ser-
viceability limit state implicitly within the load and
material factors used at the ultimate limit state. Whilst
this method is consistent with current practice, FDEM
analysis used in the design of Archtec strengthening
enables the behaviour of the arch under serviceabil-
ity loading to be investigated in ways never before
possible.
Comparison of results from the unstrengthened
and Archtec tests show that under identical loads,
displacements are very similar. Corresponding struc-
tural analysis of the test arches predicts compressive
stresses in the Archtec strengthened arch that are lower
than the unstrengthened arch under the same load-
ing. For example, using the bridge proportions of the
Archtec tests and UK highway 40/44 tonne vehicle
axle loading, under the maximum service load the
maximum compressive stress in the masonry barrel
was reduced by approximately 15%. The reduction in
stress is due to the fact that the strengthening intro-
duces bending capacity into the arch barrel, which can
therefore resist the applied loading at the critical points
more effectively. Hence, on the basis that serviceabil-
ity can be defined by a stress limit, the reduction of
stress levels in the masonry in Archtec strengthened
bridges has a beneficial effect on serviceability.
Other aspects of bridge serviceability might be con-
cerned with specific deteriorated conditions in arch
barrels, such as loose bricks and ring separation. The
risk here is that debris falling from a bridge would
represent an unacceptable hazard. An example of an
arch barrel in a weakened condition that could develop
loose bricks as a result of partial ring separation has
been tested and used in comparison with Archtec.
Displacement results show that Archtec strengthen-
ing significantly increases the stiffness of the ring
503
separated barrel restoring it to that of the fully bonded
case (as-built condition). The implication is that strains
in the intrados have been reduced and the risk of
bricks loosening is thereby also reduced. Provided an
arch is maintained in reasonable condition the risk of
bricks loosening should be reduced compared to an
unstrengthened arch. There is also no reason to doubt
that similar trends in behaviour will occur if the inner
ring itself is in a deteriorated condition.
Bridge owners and experts in the field recognise the
desirability of further research with respect to the ser-
viceability limit state and phenomena such as masonry
fatigue. However, at the current time no specific guid-
ance or criteria exist with respect to explicit evaluation
of the serviceability state in arches.
To provide increased confidence that the service-
ability of a bridge is being improved by Archtec
strengthening additional checks have been introduced
into the design process. As a precautionary measure
in the absence of other guidance, the following addi-
tional serviceability criteria are included in the design
process:
1 Either check that stresses under the required live
loading do not exceed those in the unstrengthened
bridge under existing live loading, or alternatively
check that stresses in the strengthened bridge are
below an agreed serviceability limit state value.
2 To be sure that existing defects are not made worse,
or for that matter introduced into arch barrels by
Archtec strengthening, strains along the intrados
under the required live loading are checked to
ensure they do not exceed those in the unstrength-
ened bridge under existing live loading. Strains are
calculated over a reasonable length so that an esti-
mate of radial joint cracking, critical to loosening
of bricks, is included.
These criteria are considered very conservative and
stresses and strains beyond these limits may be quite
safe and have no adverse serviceability effects. How-
ever, further fundamental research is required to
establish appropriate limiting criteria.
7 FIELD MONITORING
Several bridge monitoring programmes have been
undertaken during the last decade to help verify FDEM
arch simulations, and for strengthened bridges, to
make before-and-after behaviour comparisons. The
most recent of these was for Indian Railways with the
first part of the programme completed in 2007 and
which involved two unstrengthened multi-span brick
arch bridges.
The principal aims of the programme were to
compare measured barrel vertical displacements and
intrados circumferential strains with those predicted
using FDEM simulations. Load tests were carried out

Figure 11. Eight span Indian Railway arch bridge, one of
two monitored. Bottom – FDEM model.
Figure 12. Typical measured and predicted static vertical
displacements for four train positions. FDEM results are
curves.
Figure 13. Typical measured and predicted dynamic (train
speed 10 km/h) vertical displacements at the crown of span 4.
using a 1000 tonne freight train. Wagon loads were
recorded at a ministry weigh station and axle loads
were recorded using calibrated arrays of strain gauges
spot welded to the tracks in several places at each
bridge. A series of static tests were carried out with
axles in designated positions. Dynamic tests were also
carried out with the train running at constant speeds of
504
10, 40 and 65 km/h. During the dynamic tests vertical
accelerations were also recorded.
Vertical displacement results were found to be
sensitive to the stiffness of the foundations, which is
essentially unknown, and also to the transverse distri-
bution of the live loading. A process involving linear
regression was used to take account of these effects.
Overall, measured and predicted results compared well
for all static load cases considered.
Vertical displacements measured during the
dynamic tests were compared with predictions and
good correlation obtained. For predictions, the engine
and two wagons traversed the bridge. Good correlation
was obtained for the first eight axles of the train.
8 CONCLUDING REMARKS
FDEM has now been used successfully for a decade
on over 170 arch bridge assessments and strengthening
projects as part of the Archtec engineering partnership,
and is now recognised as special assessment tool. Dur-
ing this period, verification of this technique has been
carried out by making comparisons with the results
of full-scale tests, with data published by others on
arch tests, with the results obtained by conventional
arch bridge assessment methods and with the results
obtained from monitoring programmes in the field.
In all instances good comparisons of strength and
stiffness have been made.
Recognising that arch bridge displacement, strain
and damage can also be predicted, and that these fac-
tors are important to bridge serviceability, further work
has been carried out to investigate in-service bridge
behaviour. This has including predicting responses
under static and dynamic live load and making compar-
isons with monitored results. However, until limiting
criteria is developed, whether strain, stress, crack or
fatigue based, and until the serviceability behaviour of
masonry arch bridges is better understood, a method
has been developed that ensures stress and strain con-
ditions when strengthening for larger loads do not
exceed those in existing arch barrels under existing
loading.
By representing the constituents of masonry arch
bridges in a natural and non-homogenised way FDEM
can provide realistic simulation of structural behaviour
for use in both special assessment and strengthening
design.
REFERENCES
Orbán Z. 2004. Assessment, reliability and maintenance
of masonry arch railway bridges in Europe. ARCH’04,
P. Roca and E. Oñate (eds). Barcelona: CIMNE.
Rockfield Software Limited. 2003. ELFEN version 3.0.4
(Elfendyn v3.3.24) Archtec version. Swansea: The Uni-
versity of Wales Swansea.
McKibbins, L.D. & Melbourne, C. et al. 2006. C656 –
Masonry arch bridges: condition appraisal and remedial
treatment. London: CIRIA.
Cundall, P.A. 1971. A computer model for simulating
progressive, large scale movement in blocky systems.
Proceedings Symposium ISRM, Nancy, France, Vol. 2,
129–136.
Munjiza, A. 2004. The combined finite-discrete element
method, Chichester: John Wiley & Sons Ltd.
Owen, D.R.J. & Hinton, E. 1980. Finite elements in plasticity,
theory and practice. Swansea: Pineridge Press Limited.
Brookes, C.L. & Mehrkar-Asl, S. 1998, Numerical Modelling
of Masonry Using Discrete Elements – Seismic Design
Practice into the Next Century, Society for Earthquake
and Civil Engineering Dynamics, London pp 131–138.
Rotterdam: Balkema.
505
Hendry, A.W. 1990. Masonry properties for assessing arch
bridges, Contractor Report 244. Crowthorne, England:
Transport and Road Research Laboratory.
BD 21/01. 2001. The Assessment of highway bridges and
structures. England: The Highways Agency: Design Man-
ual for Roads and Bridges.
BS 5628-1. 2002, Code of practice for use of masonry, Part 1:
Structural use of unreinforced masonry. England: British
Standards Institution.
Brookes, C.L. 2003. Archtec, Verification of Structural
Analysis, Rep B1660A/V10/R02 Rev C. Southamp-
ton: Gifford http://www.bridgeforum.org/dir/weblink/
B1660AV10.html
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Field tests of a timber queen-post truss and numerical analysis

J.M. Branco & P.J.S. Cruz

Civil Engineering Department, ISISE, University of Minho, Portugal

M. Piazza
University of Trento, Trento, Italy

H. Varum
University of Aveiro, Aveiro, Portugal

ABSTRACT: In-situ cyclic tests on an existing traditional timber truss were performed. The main goal of
the tests was to evaluate the overall behavior of the timber truss under symmetric and non-symmetric loading.
Moreover, the influence of the location of point loads application, without and with eccentricity relatively to
joints, was assessed. The carrying tests were preceded by a visual and non-destructive inspection aiming to
collect geometric data and to assess the decay level. The field tests results of a queen-post truss are presented
and analyzed. A numerical model was developed to reproduce the test results.

1 INTRODUCTION analysis has been developed to reproduce the test

The lack of practical, but realistic, numerical mod-
els for the simulation of the behaviour of joints in
traditional timber structures normally leads to the
replacement of old roof structures, instead of their
retrofitting to satisfy safety and serviceability require-
ments present in recent Codes and Recommendations.
Moreover, the misunderstanding of the global behav-
ior of traditional timber roof structures can result
in unacceptable stress distribution in the members,
as a result of inappropriate joints strengthening (in
terms of stiffness and/or strength). To overcome this
need, laboratory tests on scaled or full-scale speci-
mens of members, connections and trusses can be
done. However, only with field-tests the behavior in
real conditions can be evaluated.
Field tests on traditional timber trusses are not com-
mon. Researchers (Del Senno 2003; Piazza et al.,
2004; Branco et al. 2008) have preferred to transport
the full-scale specimens to laboratory. In-situ working
conditions are a barrier and the setup implementation
(measurement system and load application) is often
difficult.
The work presents the field test results of a timber
queen-post truss under symmetric and non-symmetric
loading. The influence of the number of point loads
was studied. The truss has been characterized with
regard to geometry, material properties, material decay
using non-destructive tests methods. A numerical
results.
2 TRUSS ASSESSMENT
The queen-post timber truss evaluated belongs to the
roof structure of an old warehouse of Adico indus-
try, located at Avanca (55 km South from Oporto). The
exact date of the construction is not known but the
industry exists since 1920 and some plans of the vil-
lage from 1942 already show the warehouse. Trusses
are the main elements of the roof structure, covered
with ceramic tiles, 27◦ slopes and rafters spaced 50 cm
over the purlins and the ridge. The free span of the
trusses is 11.8 m and the average distance between
their centres is 3.5 m.
The geometry of the truss is particular: the config-
uration is typical of a king-post truss, but queen-posts
were added connecting the joint strut/rafter to the
tie-beam. This is not the traditional queen-post truss
geometry, in which the king-post is substituted by a
straining beam connecting horizontally (in the upper
part) the two queen-posts, those located below the
higher purlins, and the struts connecting the bottom
part of the queen-posts to the lower purlins. Clearly,
it is an example of a timber truss with an incorrect
configuration for the span of the roof. The correct
queen-post truss geometry should have been used
or two extra posts (princess-posts) should have been

507
 Ridge (80x180) Rafter (80x200)
Purlin (80x170) Purlin (80x170)
King-post Struts
Purlin (80x190) (80x145)
(80x195)
Queen-post Queen-post
(80x150) (80x150)
Tie-beam (80x220)
11,8m
Figure 1. Truss geometry.
Figure 2. Emergence holes over the surface of sapwood.
located below the lower purlin. Point loads out of the
joints, causing bending moments in the rafters, are the
most common error detected in the preliminary survey
performed in previous steps of the research program
(Branco et al. 2006).
The truss is made of maritime pine (Pinus pinaster,
Ait.). The timber members of the truss are slender,
as characteristic of traditional Portuguese roofs struc-
tures, with cross-sections varying from 80 × 145 mm2
for the struts to 80 × 220 mm2 for the tie-beam. The
tie-beam is suspended to the posts by iron straps nailed
into the posts. Between the tie-beam and the king-post
there is a gap of 5 cm while queen-posts are in con-
tact with the tie-beam. Connections between the others
timber members are made by single step joints, in some
cases nailed, and the queen-posts/rafters connections
have a heel strap nailed (25 mm wide and 5 mm tick),
Figure 1.
Despite the apparent good condition of the timber
members of the truss, the visual inspection revealed
insects attack in the tie-beam, queen-posts and struts.
In these timber members, emergence holes over the
surface of sapwood are visible however, without signs
of active infestation, Figure 2.
To evaluate the extension of decay in the timber
truss, Pilodyn® and Resistograph® non-destructive
tests were performed. The Pilodyn 6J was used with the
aim to assess the surface hardness through the depth
penetration of the pin steel (2.5 mm) measured in each
test performed. Resistograph permits to plot profiles
(drill resistance versus penetration depth) that can be
used to determine the location and extent of voids,
allowing the calculation of the residual cross sec-
tion (since decayed wood presents lower penetration
resistance), and variation in material density.
600
500
Drill resistance
400
Purlin
(80x170)
300
200
27°
100 decay
0
0 15 30 45 60 75 90 105 120 135 150 165 180 195 210
Measurement length (mm)
Figure 3. Calculation of the cross-section reduction based
in a Resistograph test profile.
Decay
0-5%
5 - 10 %
10 - 20 %
20 - 25 %
Figure 4. Map of the decay extension in the truss.
3 TEST SETUP, INSTRUMENTATION AND
PROCEDURE
The main goal of the tests was to evaluate the overall
behavior of the timber truss selected under symmetric
and non-symmetric loading. Moreover, the influence
of the location of point loads application, without and
with eccentricity relatively to joints, was assessed.
This is the consequence of the roof geometry, which
exhibits two purlins located with an eccentricity rela-
tively to the intermediate joint of the rafter. Therefore,
firstly, joints loads were applied in the joints (F1 , F2
and F3 ) and, in a second step, loads were applied over
the purlins and ridge (F1 , F4 , F5 , F6 and F7 ).
Loading and unloading were recorded and an
attempt to measure the creep of the structure under
symmetric loading was made. The behavior of the truss
under non-symmetric loading was evaluated, in the
first scheme, only by one test and in the second with
two tests (one in each pitch side). Table 1 resumes the
tests performed.
Wood pallets suspended to the truss by four steel
cables (φ 6 mm) supported the 35 kg cement bags
used as loads. Each loading and unloading proce-
dure was divided in steps of 175 kg (5 bags). A total
load of 2625 kg (3 × 875 kg) and 2975 kg (5 × 595 kg)
was used in the first (three point loads) and second
(five point loads) schemes, respectively. The differ-
ence in the maximum load value applied between
both schemes, 350 kg, is due to the difficulty to
increase the number of bags over the pallets in the
first case. To record the deformation of the truss
during the tests, eight LVDTs (Linear Variable Dif-
ferential Transformer) and six dial gauges (DG) were
used. LVDTs were responsible for measuring the
508
Table 1. Summary of the tests performed.
875
F1

F2 F3 700

Load (kg)
525

Procedure 350
Load (kg) Load (kg)
LVDT-1
875 875 175
LVDT-3
161 LVDT-2
0
Time (min.) Time (min.) 0 5 10 15 20 25
Test Point loads Test Point loads Displacement (mm)
3F-C F1, F2 and F3 3F-N F3
Figure 6. Displacement recorded by LVDTs 1, 2 and 3
F1
during 3F-C test.
F5 F6
F4 F7
In every loading and unloading step the displace-
ment values of the LVDTs were recorded; however,
Procedure
in the case of the dial gauges, only some steps were
acquired, as results of the little variation verified.
Load (kg) Load (kg)
595 595

108 4 ANALYSIS OF THE TEST RESULTS

Time (min.) Time (min.) The behavior of the traditional timber trusses even
Test Point loads Test Point loads under symmetric loading is non-symmetric. The
5F-S F1, F4, F5, F6 and F7 response of traditional timber trusses is highly depend-
F1, F4, F5, F6 and
5F-N1 F6 and F7 5F-C ing on the variability of the timber member cross-
F7
5F-N2 F4 and F5
sections, material properties, connections, supports
and loading conditions. When using a natural and
anisotropic material like wood, it was impossible for
carpenters, sometimes in extremely difficult working
LVDT-7 LVDT-8 conditions, to avoid that variability during construc-
tion. Moreover, in the case of old constructions, due
LVDT-4 LVDT-6
DG-1
DG-3 LVDT-5 DG-4
DG-2 to decay processes and lack of maintenance, this
DG-5
LVDT-1 LVDT-2 LVDT-3
DG-6 heterogeneity is emphasized.
The field test results confirm that the truss under
investigation presents a non-symmetric behavior even
Figure 5. Instrumentation of the tests. Eight LVDTs and
six DGs.
when subjected to symmetric loading conditions,
Figure 6.
The difference observed between the displacement-
global displacement (LVDTs 1 to 3), the behavior load curves of LVDT’s 1, 2 and 3 can represent the
of the king-post/tie-beam connection (LVDT-5) and influence of the decay observed in the left queen-post.
the displacement below the purlins (LVDTs 4 to 8) The constant load rate applied during 161 minutes
also used to calculate the rotational behavior of the (see Table 1) results in a deformation increment of
joints rafter/tie-beam and rafter/strut. The values of the the truss (creep). The truss presents important plastic
LVDTs during the tests were acquired by a Data Acqui- deformations (average value of 57%) after the com-
sition System, with 8 channels, using a LabVIEW plete unloading. The king-post/tie-beam connection
program (version 8.2). Dial gauges measured the open- works effectively, i.e. the tie-beam is suspended to the
ing of the queen-post/tie-beam connections (DG 3 king-post, Figure 7.
and 4), the horizontal displacement of the rafter in The heel strap is able to suspend the tie-beam, reduc-
the rafter/tie-beam connections (DG 5 and 6) and two ing therefore the deformation of this element (see
additional points to calculate the rotation of rafter/tie- Figure 6 LVDT-2), and presents also creep, however,
beam connections (DG 1 and 2). Figure 5 shows the less than the one presented by the wood members.
instrumentation used in the tests performed. The connections between the queen-posts and the

509
 875 875

Load (kg)
700
525
700 350
DG-7
175
Load (kg)

0
525 0 2 4 6 8 10 12 14
Displacement (mm)
350
(a) Damage (b) load-displacement curve of DG-7

175 LVDT-5 Figure 9. Behavior of the left rafter/tie-beam connection

0 during 3F-C test.
0.00 0.10 0.20 0.30 0.40
Displacement (mm)
875
Figure 7. Behavior of the king-post/tie-beam connection
during 3F-C test. 700

Load (kg)
525
875
350
700
LVDT-1
(3F) DG-4 LVDT-3
175
Load (kg)

525 (3F) DG-3

(5F) DG-4
LVDT-2
350 (5F) DG-3 0
-10 -8 -6 -4 -2 0 2
175
Displacement (mm)
0
-0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2
Figure 10. Displacement recorded by LVDTs 1, 2 and 3
Relative displacement (mm)
during 3F-N test.

Figure 8. Behavior of the left rafter/tie-beam connection

during 3F-C test. Negative values are recorded when the two
elements are approaching.
tie-beams, where a heel strap suspended the tie-beam,
show different behaviors, Figure 8. Only the left con-
nections, measured by DG-3, in the second series of
tests (5F, 5 point loads), behave properly – the tie-beam
is suspended to the queen post. In the first series of
tests, both connections show plastic behaviour. There-
fore, it can be concluded that, before the tests, those
connections were dismantled. The first series of tests
were sufficient for the left connections to recover,
while the gap between both connected elements exist-
ing in the right connection was never recovered.
During the 3F-C test, significant damage was
detected on the left rafter/tie-beam connection, over
the DG-7, (Figure 9a). Damage started for a point load
level of 700 kg and with 850 kg the measurement of the
DG-7 only became stable after 30 minutes. At the end
of the loading period, the maximum relative horizon-
tal displacement between the rafter and the tie-beam
(measured by DG-7) was reached and no recover was
observed during and after the unloading procedure,
Figure 9b.
Under non-symmetric loading, as the one imposed
during the test 3F-N, distortion of the truss is observed,
in particular, in the tie-beam, as shown in Figure 10.
Non-symmetric behavior is acquired by LVDT-1 and
3 with lower values in the first LVDT as consequence
of the bigger stiffness (compression of the left queen-
post).
Dividing the total amount of load applied by more
point loads in the second tests series, from 3 to 5,
the same general conclusions about the asymmetric
behavior of the truss, even when subjected to symmet-
ric loading, can be drawn.The main difference between
the tests under 3 and 5 point loads is, in the second case,
the introduction of significant bending stresses in the
rafters.
As a consequence, greater values of rotation in the
connections are obtained (Figure 11) while the global
displacements are lower (Figure 12), when compared
with the 3 point loads case. In the first case the system
is more rigid.
Applying the loads with eccentricity relatively to
joints, the main stresses are found in the rafters, caused
by bending, and greater rotations of the rafter/struts
and rafter/tie-beam are obtained. When the point loads
are applied directly in the joints, the main stresses are
observed in queen-posts (compression), pushing the
tie-beam down. As a consequence, LVDTs 1, 2 and
3 show higher values of displacement, Figure 12. In
addition, the creep behavior observed in the case of

510
 3150 E90 orthogonal to it, plus a shear modulus, G, and a
2800 Poisson’s ratio, ν. For the material properties values
2450 the LNEC (1997) was followed.
Total Load (kg)

2100
1750
5.2 Geometric and mechanical models
1400 (5F) LVDT-4
1050 The geometry and loading of the truss permits the
(5F) LVDT-6
700 assumption of a state of plane stress for the model. The
(3F) LVDT-4
350
truss is analyzed as a frame structure, having the mem-
(3F) LVDT-6 ber an elastic behavior while a semi-rigid behavior
0
0 0.25 0.5 0.75 1 1.25 1.5
has been assumed for the connections. The semi-rigid
behavior of the connections is introduced in the model
Rotation (10-3 rad)
with the Nonlinear link elements (Nlink).
Figure 11. Comparison between rotations calculated based
on values recorded by LVDTs 4 and 6 during 3F-C and 5F-S 5.3 Loads
tests.
During the tests two types of loads were applied in
3150
the structure. Uniformly distributed loads represent-
2800
ing the self-weight of the truss member, automatically
2450
computed, and joint loads as result of the self-weight
(5F) LVDT-1 of the roof structure, transmitted to the truss by the
Total Load (kg)

2100
1750
1400
1050
700
350
0 Traditional timber joints, even without any strength-
0 5 10
Displacement (mm)
Figure 12. Comparison between the displacements
recorded by LVDTs 1, 2 and 3 during 3F-C and 5F-C tests.
three point loads (3F-C) is significantly greater but,
in this case, the influence of the damages observed
in the rafter/tie-beam connection must be taken into
account.
5 NUMERICAL ANALYSIS
The structural analysis program SAP 2000 has been
used to model the tests. It has been developed a
model involving variable dimensions beam elements,
and semi-rigid joints between the different elements,
assuming anyway a linear behavior of the material.
5.1 Material model
Timber is assumed as an orthotropic material in
the system so-called anatomic cylindrical coordinates
corresponding to the longitudinal, L, radial, R, and
transversal, T , directions of the tree trunk. Cylindri-
cal coordinates may be approximated as orthogonal,
for the material extracted from the outer region of
the trunk. The elastic modulus, for an asymmetric
model, are E0 in the direction along the fibers, and
(5F) LVDT-2
(5F) LVDT-3
purlins, and the ones applied during the different load
(3F) LVDT-1 procedures of the tests performed.
(3F) LVDT-2
(3F) LVDT-3
5.4 Semi-rigid modeling of connections
15 20 25 ening device, usually have a significant moment
capacity. Common constraint models, like hinges or
full restraint connections, indeed, cannot satisfacto-
rily describe the real behavior of these joints. The joint
behavior may be classified as semi-rigid and, being
based on friction, is influenced by the time-varying
level of compression between the joined members
(Parisi & Piazza 2000). In order to properly describe
this behavior, the elastic stiffness of each Nlink must
be defined, according to the different geometric and
mechanical features of the elements at each joint.
Candelpergher & Piazza (2001) have proposed
some expressions to define the rotation stiffness of
traditional timber connections, however, these rules
should be verified for the Portuguese traditional tim-
ber connections case. Because this calibration process
is not yet finished, only the symmetric tests will be
numerically analyzed. In the case of the symmetric
tests performed the rotation stiffness of connections
has a trivial influence in the overall behavior of the
tested truss. However, the axial stiffness of the con-
nections is crucial in the truss response (deformation
and stress distribution).
The axial stiffness (kax ) depends from the lum-
ber mechanical properties, the geometric proportions
of the connected elements and the connection angle
(skew angle):

511
where, applying the Hankinson Equation (2):
represents the wood elastic modulus in the direction
forming an angle α with the fiber.
represents the nominal notch length, where compres-
sion deformation occurred.
represents the nominal notch area, where stress was
assumed to be transmitted.
The axial stiffness of the tie-beam/posts connec-
tions (kax,hs ) has been formulated taking into account
the axial stiffness of the heel strap:
where Esteel is modulus of elasticity of steel, Aheelstrap
and lheelstrap is the cross-section and the length of the
heel strap, respectively.
6 NUMERICAL VERSUS EXPERIMENTAL
RESULTS
Numerical modeling and experimental results have
been compared in the case of the symmetric tests.
First, Step 0, the numerical model has been imple-
mented without considering the real behavior of the
truss observed in the performed tests. Then, the model
was verified and calibrate based in the tests results
(Step 1).
Assuming the axial stiffness at the connections
given by Equations 1 to 5, Step 0, the computed
values for the global displacements (LVDT 1 to 3)
represents only 33% of the test results obtained for
the first test (3F-C). The main reason for this dif-
ference is the fact that all connections, in particular
the ones between the posts and the tie-beam, showed
significant gaps between the metal devices and the
joint itself. Moreover, the deterioration and loss of
strength of the steel elements are not considered in
the calculated stiffness values. The gaps between the
metal devices and the joints are confirmed by the tests
results (see Figures 7 and 8). It is also important to
512
Figure 13. Bending moment on the truss in the 3F-C test.
point out that the plastic deformation measured in
the global displacement after the 3F-C test represents
52%, 66% and 53% of the maximum displacement,
respectively for LVDT 1, 2 and 3. In the case of the
relative displacements measured in the tie-beam/posts
connections, the residual values represents 57%, 100%
and 52% of the maximum displacement recorded at
LVDT-5, DG-3 and DG-4, respectively. Therefore, the
stiffness values calibrated for the first tests performed,
3F-C, should be only account as informative because
they report the influence of the original gaps exist-
ing between the metal devices and the joints. It is
important to note that during the first test, 3F-C, a
significant damage of the left rafter/tie-beam connec-
tions was detected (Figure 9) which influences directly
the global displacements values.
However, the calibrated model was able to repro-
duce the non symmetric response of the truss even
under symmetric loading conditions applied during the
3F-C test, Figure 13.
Table 2 gives a comparison between experimental
and numerical results (SAP 2000) for both symmetric
tests performed with regard to the displacements in
the relevant joints of the truss. The values reported
in Table 2 are the maximum displacement of the
symmetric tests performed.
Table 2 show a good fit between the values worked
out by means of the numerical model and the exper-
imental results, with exception of the LVDT-2 val-
ues, all errors reported by the numerical model are
under 10%. Experimental results for the global verti-
cal displacement of the truss under the king-post are
clearly increase by the numerical model. This inconsis-
tency can only be explained by the faulty connections
between the king-post and the tie-beam. The heel strap
must introduce a “frictional stiffness” which reduces
the mid-span deformations of the tie-beam when the
truss is loaded. Ideally, in the tie-beam/post connec-
tions it shall be used a heel strap, nailed only in the post,
suspending the tie-beam with a connection without
bending-stiffness.
The calibration process of the numeric model shows
that during the experimental campaign a stiffness
updating of the connections occurs (Table 3). This
conclusion, confirmed by the test results, is due essen-
tially to the fact of the connections were originally
dismantled (significant gaps existed between the metal
devices and the joints).
Table 2. Comparison between numerical values (Nu.) and Table 3. Axial stiffness values (kN/m) used in the numerical
tests results (Exp.). Error (Er.) express in (%). model.

Test 3F-C Test DG-5 DG-6 DG-3 DG-4 LVDT-5

Global displacements (mm) 3F-C 2900 74002 2000 2400 3200

5F-S 721155 374000 1000 31500 10200
LVDT-1 LVDT-2 LVDT-3

Exp. Nu. Er. Exp. Nu. Err. Exp. Nu. Err.

15.4 15.5 0.21 12.0 13.8 14.6 10.9 10.0 8.14 particular, the connections between the tie-beam and
the posts, recorded by DG-3, DG-4 and LVDT-5 are
Relative displacements – tie-beam/posts connections (mm) sensible to the connections rotational stiffness.

DG-3 LVDT-5 DG-4

7 CONCLUSIONS
Exp. Nu. Er. Exp. Nu. Err. Exp. Nu. Err.
The work results highlight the importance of the in
0.64 0.59 7.81 0.36 0.35 3.70 0.18 0.18 2.22 situ experimentation, to assess the global behavior of
traditional timber trusses, identify the critical areas, to
Relative displacements – rafters/tie-beam connections (mm) plan the upgrade interventions and to quantify their
effects.
DG-5 DG-6 Experimentation gave an insight of the truss behav-
ior hardly reachable otherwise. The effects of the
Exp. Nu. Er. Exp. Nu. Err. incorrect truss configuration for the roof span, the
faulty connections geometry and the existing gaps in
9.48 8.53 10.0 0.34 0.35 4.12 the joints in the overall behavior of the truss tested
were assessed.
Test 5F-S The numeric model implemented through a gen-
eral purpose and very simple FE code (SAP 2000)
Global displacements (mm)
has proved to be an effective and accurate method of
LVDT-1 LVDT-2 LVDT-3
modeling timber truss behavior, on condition that a
semi-rigid behavior of the traditional connections is
Exp. Nu. Er. Exp. Nu. Err. Exp. Nu. Err. assumed. In this work, only the symmetric tests were
modeled because the calibration process of the rotation
7.34 6.89 6.12 5.63 6.63 17.7 5.96 5.59 6.20 stiffness models for the Portuguese traditional timber
connections case is not yet finished. However, this fit-
Relative displacements – tie-beam/posts connections (mm) ting process must be accomplished in the next couple
of months.
DG-3 LVDT-5 DG-4

Exp. Nu. Er. Exp. Nu. Err. Exp. Nu. Err.

ACKNOWLEDGMENTS
0.14 0.14 2.86 0.09 0.09 7.37 0.06 0.06 0.00
Authors would like to acknowledge the cooperation of
Adico industry (Arch. Luis Dias and Fernando) and
Relative displacements – rafters/tie-beam connections (mm) José Manuel in the experimental campaign.
The first author gratefully acknowledges the Por-
DG-5 DG-6
tuguese Foundation for Science and Technology, for
Exp. Nu. Er. Exp. Nu. Err. his PhD grant SFRH/BD/18515/2004. The research
described in this paper was conducted with financial
0.05 0.05 1.82 0.10 0.10 2.00 support of the Portuguese Foundation for Science and
Technology (POCI/ECM/56552/2004).
This work has been carried out with a partial
When using the numeric models calibrated for the financial contribution of the Italian Earthquake Engi-
case of symmetric tests under non-symmetric load- neering Laboratory Network (RELUIS), within the
ing conditions, a significant discrepancy is obtained research program carried out for the Italian Agency
between the numeric and experimental results. In for Emergency Management.

513
REFERENCES
Branco J.M., Cruz P.J.S., Piazza M. & Varum H. 2006.
Portuguese Traditional Timber Roof Structures, In WCTE
2006 – World Conference on Timber Engineering, 6–10
August, Portland, Oregon, USA.
Branco J.M., Cruz P.J.S. & Piazza M. 2008. Diagnosis and
analysis of two king-post trusses, In SAHC 2008: Struc-
tural Analysis of Historical Constructions, Bath, UK,
02–04 July.
Candelpergher L. & Piazza M. 2001. Mechanics of traditional
connections with metal devices in timber roof structures.
In Proceedings of the Seventh International Conference
STREMAH 2001, Bologna, Italy.
Del Senno M. & Piazza M. 2003. Behaviour and
rehabilitation of queen post timber trusses. A case
study. In STREMAH 2003: Structural Studies, Repairs
514
and Maintenance of Heritage Architecture VIII, May,
Halkidiki, Greece.
LNEC 1997. Timber for structures – Maritime pine for struc-
tures. LNEC (eds), Ficha M2. ISSN 0873-6472, Lisbon,
12 p. (only available in Portuguese).
Parisi M.A. & Piazza M. 2000. Mechanics of plain and
retrofitted traditional timber connections. Journal of
Structural Engineering, ASCE; 126(12): 1395–403.
Piazza M., Brentari G. & Riggio M.P. 2004. Strength-
ening and control methods for old timber trusses: the
queen-post truss of the Trento theatre. In SAHC 2004:
Structural Analysis of Historical Constructions, Padova,
IT, II: 957–965.
SAP 2000. Static and Dynamic Finite Element Analysis of
Structures. Structural Analysis Program. Computers and
Structures. Inc., Advanced 9.03. California. USA.
Assessment and analytical techniques
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Modelling of vaults as equivalent diaphragms in 3D seismic analysis of

masonry buildings

Serena Cattari, Sonia Resemini & Sergio Lagomarsino

DICAT, Department of Civil, Environmental and Architectural Engineering, University of Genoa, Italy

ABSTRACT: The role of horizontal diaphragms in the seismic response of buildings is significant both in terms
of stiffening and transfer of horizontal loads to vertical elements. Their behaviour in 3D modelling is frequently
assumed (with a rough approximation) as rigid: this hypothesis may be not completely acceptable in case of exist-
ing buildings (with reference to historical masonry structures), where various archaic constructive technologies
can be found (e.g. structural brick or stone vaults). In the framework of equivalent-frame models, the adequate
definition of the equivalent elastic stiffness of various ancient floor technologies is noteworthy. In this paper, the
attention is focused on the structural brick or stone vaults, frequently built in historical heritage. Detailed FEM
elastic analyses are performed on different typologies of vaults (barrel vault, cross vault, cloister vault), in order
to correlate their axial and shear stiffness to that of an equivalent membrane. Various geometrical parameters and
boundary conditions are analysed. On the other hand, stiffness degradation, due to the non-linear behaviour of
masonry, and failure mechanisms have to be studied. In fact, damage in the vaults may substantially alter their stiff-
ening effect and the distribution of horizontal loads to vertical elements. Also in this case, detailed FEM analyses,
using a non-linear constitutive model are performed, focusing the attention on the effect of masonry pattern.

1 INTRODUCTION walls, are considered as plane stiffening elements,

The seismic response of masonry buildings is strictly
related to the achievability of its overall behaviour,
meant as the mutual cooperation of each structural
element (if properly connected). Thus, the response
depends on the whole structural system, controlled by
the in-plane behaviour of the walls, and therefore by
both connection and load transfer effect due to the
floors.
In this context, the modelling strategy based on the
idealization of the structure through an “equivalent
frame” seems very suitable for the analysis of stan-
dard masonry buildings, as also proposed in recent
international and national codes (Eurocode 8, OPCM
3431/05 Italian Decree). Having the advantage of a
reasonable computational effort, complete 3D models
of URM structures can be obtained assembling walls,
of which only in-plane response is modelled.
Recent developments in this field produced analyt-
ical tools (e.g. Galasco et al. 2004) capable to perform
non-linear static (pushover) and dynamic analyses.
This modelling approach starts from some hypotheses
on structural and seismic behaviour of URM struc-
tures: the bearing structure, both referring to vertical
and horizontal loads, is identified with walls and floors
(or vaults). The walls are bearing elements, while the
floors, apart from transferring vertical loads to the
which the horizontal actions distribution between the
walls depends on. The local flexural behaviour of the
floors and the walls out-of-plane response are not
computed.
Although it represents a crucial feature to be consid-
ered, the floor behaviour in 3D modelling is frequently
assumed (with a rough approximation) as rigid. This
hypothesis may be not completely acceptable in case of
existing buildings (e.g. historical masonry structures),
where various ancient constructive technologies can
be found (i.e. timber floors or structural brick or stone
vaults).
A proper assumption on the diaphragm stiffness
may significantly affects the overall response: in fact,
in the limit case of “infinitely” flexible floors, there
would be no load transfer from collapsed walls to
still efficient structural elements. On the contrary, in
the limit case of “infinitely” stiff floors, this resource
could be overestimated.
In this context, the adequate definition of the
equivalent elastic stiffness of various archaic floor
technologies is noteworthy. In this paper, the atten-
tion is focused on the structural brick or stone vaults,
frequently built in historical heritage, with particular
reference to their stiffening role. Thus, detailed FEM
elastic analyses are performed on different typologies
of vaults (barrel vault, cross vault, cloister vault), in

517
order to correlate their axial and shear stiffness to
that of an equivalent membrane. Various geometrical
parameters and boundary conditions are analysed.
On the other hand, damage in the vaults may
substantially alter their stiffening effect and the dis-
tribution of horizontal loads to vertical elements. All
elastic and geometrical conditions being equal, other
masonry characteristics (e.g. masonry pattern and
strength parameters) may strongly affect this feature.
In this case, detailed FEM analyses, using a non-linear
constitutive model (Gambarotta & Lagomarsino 1997)
are performed, focusing the attention on selected case
studies.
2 MORPHOLOGY AND SEISMIC ROLE OF
VAULTS
The building technique of masonry vaults continued
throughout the ages until recent times, influencing the
history of architecture and originating a multitude of
typologies and shapes (simple and complex: barrel
vault, cross vault, cloister vault, etc.).
Manifold factors affect their structural behaviour,
such as geometry, stiffness and mass distribution,
chronological succession of the building works, sub-
sequent alterations.
As an example, different masonry pattern can be
adopted on equal terms of geometry: this aspect plays a
fundamental role in a structural system where the equi-
librium is ensured by the mutual thrust of the blocks
making up the vault. In Figure 1, some examples of
pattern are shown.
As a consequence, this aspect represents a distinc-
tive building feature, strongly affecting the overall
behaviour: paying attention on the barrel vault, vary-
ing the brickwork bond, its structural response may be
interpreted as that of a three-dimensional shell (par-
allel pattern) or a set of adjacent arches (orthogonal
pattern). The brick (or stone) bond influences not only
the static behaviour, but also the damage pattern: in
fact, this latter is obviously affected by the potential
sliding planes, identified by the scheme of bed and
head joints.
Figure 1. Masonry pattern in cross vaults: parallel, orthog-
onal and oblique to the walls (herringbone bond).
518
The presence of loose fill, structural backing and
constraint boundary conditions may be considered
further factors affecting the vault behaviour.
In particular, the constraint boundary conditions are
a crucial point if the attention is focused on the vault
role in the context of the seismic overall behaviour of
a complex entity as a building is (see §3 The vault
modelling as an equivalent membrane).
From this point of view, the constraint conditions
to be modelled are function of the interlocking pro-
vided by the walls in correspondence of the vault
springing points. The constraint degree may derive
by a precise constructive purpose or by cracking or
degradation situations, which the structure may have
undergone during the ages. As an example of the first
case, lunettes in barrel vaults are often built without
connection to lateral walls: in this condition, when the
vault and the supporting walls perform out-of-phase
oscillations, they act as struts (or counterforts) if com-
pressed, but they may undergo no damage when the
wall tends to separate. The second situation may be
represented by the loss of continuity between the arch
and the end wall, e.g. in case of barrel or cross vaults,
due to crack developing near the wall.
In presence of seismic loads, another interaction
effect between vaults and supporting walls, besides
the constraint degree, is constituted by the activation
and developing of out-of-plane collapse mechanisms
of the walls. Even if the analysis of these mechanisms
is certainly very interesting, this study focuses on the
structural role of the vaults in transferring the inertial
loads to vertical walls, in the framework of the overall
seismic behaviour of the building.
There is no intention to provide information about
the detailed seismic behaviour of the vault itself, but
the aim is obtaining some knowledge about the realis-
tic stiffening effect that a vault may produce in case of
an earthquake and comparing it to that of a plane floor.
In fact, no dynamic analyses are carried out on the
vault system, but only simplified load configurations
(see next paragraphs) are studied.
3 THE VAULT MODELLING AS AN
EQUIVALENT MEMBRANE
The role of floor slab in the 3D model of a building was
previously introduced. The response depends on the
whole system and both connection and load transfer
effect due to the floors are very significant.
In this context, the floor modelling as isotropic or
orthotropic membrane elements seems to be suitable.
As proposed in Galasco (2004), the elastic element
describing the floor is identified by a principal direc-
tion, with Young’s modulus Ex , while Ey is the Young’s
modulus along the perpendicular direction, ν is the
Poisson ratio and Gx,y the shear modulus.
 The moduli of elasticity Ex and Ey represent the
normal stiffness of the membrane and each of them
structurally describes the connection degree between
the floor and the vertical wall parallel to its reference
direction, both in linear and non-linear phases. In a few
words, the normal stiffness of the floor along X-axis
provide a link between the piers of a wall parallel to
X-axis, influencing the axial force the spandrels, but it
does not control the loads transferred to a wall parallel
to Y-axis.
The term Gx,y represents instead the shear stiffness
of the floor and the horizontal force transfer among
the walls, both in linear and non-linear phases.
The evaluation of these quantities may be rather
simply identified in case of recurrent floor typologies,
ascribing it to the structural role shown by some spe-
cific elements. For example, the shear stiffness of an
r.c. floor with beams and slab is mainly ascribable
to the slab. In this example, the beam axial stiffness
leads to the definition of the Young’s modulus of the
equivalent plane element.
On the contrary, in case of vaults, beside thickness
and material properties, the stiffening contribution
strongly depends on shape and geometrical proportion
(e.g., rise-to-span ratio). So, the definition of the elas-
ticity moduli to be attributed to the equivalent plane
element may be highly arbitrary.
The proposed modelling strategy starts from the
elastic numerical simulation of vault response, in case
of pre-defined load configurations, aiming to schema-
tize the axial-only and pure shear behaviour by impos-
ing special force or displacement conditions (Fig. 2).
The scope is to establish a functional relationship
among the elastic parameters of the equivalent plane
element and the significant quantities of the vaults.
This obviously excludes the rigorous theoretical inves-
tigation of the vault behaviour as a membrane and
flexural shell.
In this way, an ideal equivalence is set between the
typology of vault examined and the plane element,
having the same planimetric dimension (squared plan
Lx Lx+∆u
y y
∆u rk
Ly
x x
ux A
uy
y y
x
F=Σfk x simulations.
Figure 2. Configurations of boundary conditions in order
to schematize the axial-only and pure shear behaviour.
519
L × L), thickness (s) and material (characterized by
the elastic moduli Ex , Ey and Gx,y ). An isotropic
plane element is accounted for if the vault typology
is symmetrical (e.g. cloister and cross vault) and an
orthotropic one if unsymmetrical (e.g. barrel vault).
In the first case, aiming to evaluate the vault equiva-
lentYoung’s modulus EV , a uniform displacement state
along x-axis (u) is imposed to one of the orthogonal
sides. The opposite one is fixed, as in Figure 2.
From the isotropic linear elastic relationships, the
numerical FEM results may be interpreted as follows.
The Young’s modulus EV is given by eq. (1).
where: E is the Young’s modulus of the plane element,
σ is the normal stress, ε the axial strain, u is the
applied displacement to one side, n is the number of
nodes on this side, rk is the reaction nodal force in the
k-th node along the analysed direction.
In order to evaluate the vault equivalent shear mod-
ulus GV , an auto-equilibrated system of forces (having
a total value F on each side) is applied. The horizontal
displacements of one edge of the model are adequately
constrained. By means of the isotropic linear elastic
relationships, eq. (2) provides the ratio between GV
and G (the shear modulus of the plane element).
where: τ is the tangential stress, γ is the shear strain, ūx
and ūy are the displacement components of the oppo-
site corner A (Fig. 2), derived by the FEM numerical
simulations.
Obviously, EV and GV are representative of the 3D
behaviour of each vault typology, summing up the
shape effect and the mechanical characterization of
the material. Further considerations about the bound-
ary condition are pointed out in §3.1 Description of
the analysed configurations.
The equivalent parameters, describing the vault
stiffness and derived by elastic analyses, represent
an overestimation of the actual properties, because
of probable existing damage of material degrada-
tion, often surveyed in ancient masonry structures.
Moreover, further damage may develop during the
seismic event. From this point of view, a preliminary
overview of this topic is given in §3.3 Non-linear FEM
3.1 Description of the analysed configurations
Numerical FEM analyses are performed usingANSYS
code (release 8.1). Vaults are modelled by means

Figure 3. Vault typologies – FEM models: (a) barrel vault;
(b) cloister vault; (c) cross vault.
of shell elements (membrane and flexural stiffness).
In both the configurations of boundary conditions
(Fig. 2), three vault typologies are analysed (bar-
rel, cloister and cross vault), parametrically varying
geometrical proportions and constraints (Fig. 3). The
analysed thickness-to-span ratios (s/L) are equal to
0.03–0.045 and 0.06. The rise-to-span ratios (f/L) vary
from 0.125 and 0.5.
Aiming only to verify the consistency of the numer-
ical results, rise-to-span ratios lower than 0.125 are
simulated, even if they provide unrealistic vault pro-
files. In case of ratio tending to 0, the convergence
to the theoretical solution of the plane membrane is
confirmed.
Special attention is put on the constraint condition
adopted. In case of the normal stiffness, two schemes
(simply supported and fixed ends) are analysed for the
barrel and cloister vault. These aim to reproduce the
interlocking provided by the walls in correspondence
of the vault springing points.
Moreover, for cross and barrel vaults, the effect of
lateral walls (Fig. 3-c), which the arch could be con-
nected to (or supported by), is simulated. Their influ-
ence is analysed parametrically varying their thickness
st (the vault thickness s being equal). The ratio st /s
varies from 0.1 to 1.
The elastic FEM simulations are performed with-
out dead loads and the presence of fill and backfill
is neglected. These are rough simplifications, but it
can be numerically demonstrated that the dead load
influences only the first steps of the analysis: as the
imposed forces (or displacements) start to prevail, the
initial stress state due to dead load is immediately and
drastically modified. These results are also confirmed
by the non-linear studies.
The proposed study has to be considered as a prelim-
inary research about this topic: these simplifications
lead in any case to stiffness estimations on the safe
side.
520
Figure 4. Ev /E ratio in case of different vault typologies.
3.2 Elastic FEM simulations
In the elastic FEM simulations, masonry is modelled
as a linear isotropic material. This assumption may be
reasonable if the aim is the study of the effects of the
vault shape on the behaviour.
As briefly described, this leads to the equivalence
to an isotropic plane element, if the vault typology is
symmetrical (e.g. cloister and cross vault), and to an
orthotropic one, if unsymmetrical (e.g. barrel vault).
In case of longitudinal stiffness of barrel vaults, the
simple theoretical solution brings to the result that the
ratio EV /E is equal to that between the area of the cir-
cular crown and the area of the rectangular transversal
section of the plane element (thickness s and span L
being equal).
The results for normal stiffness in case of various
vault typologies, varying the thickness-to-span ratios
(s/L), are shown in Figure 4.
If the rise-to-span ratio f/L ranges between 0.125
and 0.5, it can be noted that the stiffness of the barrel
vault (orthogonal to the directrix) appears to be almost
negligible.
The cloister vault seems to be less stiff than the cross
structure; for both types, the thickness s significantly
influences the behaviour. But if, in case of cross vaults,
the variation is poorly affected by the rise-to-span ratio,
for cloister vaults the difference decreases for higher
f/L values.
For the barrel vault, if lateral walls are accounted
for, the response considerably changes, varying the
thickness ratio st /s, as shown in Figure 5, and the ratio
st /L. This effect may be also noticed in case of cross
vaults, in which however the shape contribution on the
stiffness is much higher.
Even if the analyses are performed modifying the
thickness ratio st /s, being equal the mechanical param-
eters of the material, the results can be more generally
examined, interpreting the ratio st /s as quantifying the
difference in material stiffness characteristics of the
two structural elements.
As regards the analysis sensitiveness to the con-
straint degree, the results for the cloister vault are
shown (Fig. 6).
 Table 1. Numerical values of the equation coefficients ai .

Vault typology

Barrel Cloister Cross

SS F SS F SS

a1 0.01 0.05 0.17 1.41 –

a2 −0.0002 −0.0011 0.1435 0.1165 –
a3 −1.73 3.1767 −7.0867 −3.8733 –
a4 – – 0.0139 0.2336 –
a5 – – – – −0.91
Figure 5. Ev /E ratio for barrel vaults (s/l = 0.06): influence a6 – – – – −0.1567
of the lateral walls. a7 – – – – 1.9133
a8 – – – – 0.0721

SS simply supported; F fixed.

Figure 6. Cloister vault: influence of the constraint degree

(springing line).

Starting from the results of the numerical simu-
lations, by means of error minimization, analytical
relationships are defined. These equations are func-
tions of thickness-to-span (s/L) and rise-to-span (f/L)
ratios. In eq. (3), the relation in case of barrel and
cloister vaults is reported.
Figure 7. Gv /G ratio in case of different vault typologies.
Even in case of shear stiffness, from the results
of the numerical simulations, by means of error
minimization, analytical relationships are put forward.
In eq. (5), the relation in case of barrel and cloister
vaults is proposed.

The analogous equation in case of cross vaults may be

found in eq.(4). The analogous equation in case of cross vaults may be
found in eq. (6).

In Table 1, the numerical values of the equation

coefficients ai are shown.
By the result processing in case of shear stiffness
GV /G, it can be achieved that the cross vault typol-
ogy provides much less contributions in relation to the
other kind of vaults (Fig. 7).
Moreover, it shows higher sensitiveness to the
modification of thickness-to-span ratio s/L.
In case of barrel vaults, the results in Figure 7 are
related to the configuration with lateral walls (st /s
equal to 0.5).
In Table 2, the numerical values of the equation
coefficients bi are shown.
3.3 Non-linear FEM simulations
The non-linear analyses are mainly performed in order
to examine the role of some specific factors on the
stiffness decay: it is worth noting that they represent
only preliminary investigations.

521
Table 2. Numerical values of the equation coefficients bi .

Vault typology

Barrel Cloister Cross

b1 1.6433 −0.1633 −1.12

b2 −1.8191 −1.3051 −4.1766
b3 0.1133 1.9733 8.9033
b4 1.1189 1.0339 0.7273
b5 – – −34.323
b6 – – 5.6202

Figure 8. Cloister vault (f /L = 0.5; s/L = 0.045;

To describe the non-linear behaviour of masonry
structure, a non-linear constitutive model is used
(Gambarotta & Lagomarsino 1997), already imple-
mented in the ANSYS code. This constitutive model
considers a limited tensile strength for fragile mate-
rials; a damage model allows simulation of the pro-
gressive loss of stiffness, up to a predefined strength
value, and the following phase of softening, in which
the material progressively loses strength with large
increases in deformation. The model does not consider
cracks as localized phenomena, but as a state of equiva-
lent inelastic deformations. This constitutive law also
allows description of the shear behaviour according
to a Coulomb criterion and sliding in the mortar
joints. Furthermore, it allows simulation of crushing
in the masonry. In this way, the masonry material
orthotropy is considered introducing the modeling of
the bed joints.
All conditions being equal to those assumed on
the elastic analyses, non-linear FEM simulations are
performed. The attention is focused instead on the
role played by two factors: firstly, different masonry
bond types are examined (e.g. Figure 1); secondly, the
ratio between some mechanical properties of strength,
which characterize the bed joints is investigated, with
particular regard to that between the tensile strength
(σmr ) of the mortar joint and the cohesion (τmr ). The
mechanical properties assumed are representative of
standard brick masonry, characterized by regular pat-
tern and by lime mortar (friction coefficient µ = 0.6;
compressive strength of masonry fm = 6 MPa; shear
strength τmr = 0.2 MPa). The moduli E and G are
assumed according to the isotropic material hypothesis
(E = 1.8 MPa and G = 0.6 MPa).
It is worth specifying that only the early non-linear
response is examined. Thus, only the initial degra-
dation of stiffness is assessed, without analysing the
collapse state of the vault. The scope of defining drift
limits beyond which the vault cannot be considered
structurally efficient is not pursued.
In the following, the results of the axial analyses are
examined.
As an example, the ones related to the response
of the cloister vault (f/L = 0.5), characterized
τmr /σmr = 1): normal stiffness decay on varying the
masonry pattern.
Figure 9. Cloister vault (f /L = 0.5; s/L = 0.045): shear
stiffness decay, varying both the patterns and the τmr /σmr ratio.
respectively by parallel (a) and herringbone (b) pat-
terns, are illustrated in Figure 8.
It is possible to observe that the normal stiffness
quite suddenly decreases, starting from the value cor-
responding to the elastic solution. This behaviour is
much more evident for pattern b. In case of pattern a,
the coves 1 and 3 (Fig. 8) provide a greater resistance
to the damage spread: in fact, as a consequence of
the assumed non-linear constitutive law that idealizes
the masonry as a layered continuum, the head joints,
which normal tensile stress acts on, cannot open.
Even if the ratio Ev /E rather drastically decreases for
very low values of u, it is worth noting that this con-
tribution represents the minimum reliable one (e.g.,
fill and backfill are neglected). However, in existing
buildings, others structural elements, as tie-rods or
stringcourses (r.c. beams), may be often found. Con-
sidering the in-plane response of masonry walls, these
constructive details concur to confine the spandrels
and to couple the piers much more effectively than the
floor (or vault).
In the following, the results of the tangential anal-
yses are examined.
Figure 9 shows the trend of Gv /G ratio for cloister
vault, in case of the two examined bonds and for values
of τmr /σmr ratio equal to 1 and 2 respectively.

522

Figure 10. Cloister vault (f /L = 0.5; s/L = 0.045): distri-
bution of αm parameter on the vault intrados, corresponding
to three analysis steps.
In case of herringbone pattern, a much more marked
decay can be noted; it suffers to a higher degree from
the variation of τmr /σmr strength ratio (see also Fig. 10).
The bond affects not only the crack path, but also the
main failure mechanism that involves the bed joints (no
block crushing is revealed from the analyses). Some
considerations may be carried out thanks to the review
of the obtained inelastic strain (excised in this paper for
brevity) and of the damage pattern shown in Figure 10.
In particular, Figure 10 shows the value distribution
of the dimensionless damage scalar parameter αm that
describes the evolution of the mortar joint damage.
Values lower than 1 correspond to the un-cracked state,
while a value equal to 1 corresponds to the attainment
of the peak strength of the joint (both in shear or in
tension). Lastly values greater than 1 correspond to
523
the strain-softening regime. Since the adopted shell
element has different Gauss points in thickness, the
evolution of this parameter can be monitored on both
extrados and intrados. With reference to the intrados
results, as expected, in case of the pattern b, dam-
age turns out to be more pronounced near the corners,
because of cracks that propagate orthogonally to the
displacement direction. On the contrary, in case of
pattern a, damage propagates starting from the vault
keystone (less compressed and hence more vulner-
able to shear failure). The described damage scalar
parameter αm , combined with the information from
the inelastic strain components, allow to evaluate that,
in case of pattern b, the predominant failure mode is
the tensile one (crack opening), whereas, in case of
pattern a, is the shear one (shear sliding).
Further remarks, regarding the role of these
masonry bonds, can be obtained from the direction of
the tensile principal stress σI in various analysis steps
(Fig. 11).
The extent of the direction modification of the
tensile principal stress σI is influenced by the basic
hypothesis of the adopted constitutive law. In fact,
because of the masonry idealisation as a layered con-
tinuum, this component tends to be placed normally
to the head joint, which are indefinitely resistant. As
a consequence, it is fair to expect a variation of this
result if the head joints are kept into account.
Finally, in case of the cross vault (f/L = 0.5), the
results corresponding to three different patterns are
shown in Figure 12. As an example, for τmr /σmr ratio
equal to 1, similar remarks to the ones of the cloister
vault can be obtained.
Figure 13 shows the distribution of αm parameter
on the vault intrados: again the damage follows the
possible paths mapped by the masonry pattern.
4 FINAL REMARKS AND FUTURE
DEVELOPMENTS
In the paper, for some typical vault typologies, the
evaluation of an equivalent elastic stiffness is pro-
posed in order to idealize these structural elements
as isotropic or orthotropic membranes. The study is
developed in the context of the“equivalent frame”
modelling approach for masonry buildings.
Even if through simplified assumptions, the elastic
FEM simulations allow to evaluate the minima reli-
able contributions, varying the vault typologies and the
geometrical and constraint boundary conditions. On
the other hand, the non-linear FEM simulations, even
if undoubtedly no exhaustive, provided some prelimi-
nary knowledge about the sensitiveness of the stiffness
degradation in the early non-linear response.The effect
of masonry pattern and strength characterisation of the
bed joints is analysed. This work represents the first
step of further investigations.
Figure 11. Cloister vault (f /L = 0.5; s/L = 0.045;
τmr /σmr = 2): direction of the tensile principal stress σI on
the extrados vault.

0.15
Pattern a’
Pattern b’ Figure 13. Cross vault (f /L = 0.5; s/L = 0.045;
Pattern c’ τmr /σmr = 1): distribution of αm parameter on the vault
0.1 intrados corresponding to different steps of the analysis.
Gv/G

ACKNOWLEDGMENTS
0.05

The authors acknowledge the partial financial con-

0 tribution by Italian Network of Seismic Laboratories
0 0.0005 0.001 0.0015 0.002 0.0025
γ “Evaluation and reduction of seismic vulnerability of
Figure 12. Cross vault (f /L = 0.5; s/L = 0.045;
τmr /σmr = 1) – stiffness decay varying the pattern typology:
parallel (pattern a ), orthogonal (pattern b ) and oblique
(pattern c ) to the walls.
From the authors’ point of view, the effect of loose
fill and structural backing and the influence of head
joints may be interesting issues to deepen by non-linear
FEM simulations.
Finally, a further development is represented by the
definition of the drift limits beyond which the vault has
to be considered structurally collapsed. This latter goal
seems useful in order to establish the limits beyond
which the reliability of the vault, related to the load
transfer, is lost.
0.003 (RELUIS), in the frame of the 2005–2008 Project
existing masonry buildings” – Line 1.
REFERENCES
Galasco A., Lagomarsino S., Penna A., Resemini S. (2004).
Non-linear seismic analysis of masonry structures, 13th
WCEE, Vancouver, B.C., August 1–6 2004, paper no. 843.
Eurocode 8, Design of structures for earthquake resistance –
Part 3: Strengthening and repair of buildings. Pr-EN
1998-3, July 2004.
OPCM, no.3431, May 3 2005. Official Bullettin no. 107 of
May 10 2005 (in Italian).
Gambarotta, L. and Lagomarsino, S. (1997). Damage mod-
els for the seismic response of brick masonry shear
walls. Part II: the continuum model and its applications.
Earthquake Engng. Struct. Dyn., 26, 441–462.

524
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Modelling of damage accumulation in masonry subjected to

a long-term compressive load

E. Verstrynge
Department of Civil Engineering, KULeuven, Heverlee, Belgium

S. Ignoul
Triconsult NV, Lummen, Belgium

L. Schueremans & D. Van Gemert

Department of Civil Engineering, KULeuven, Heverlee, Belgium

ABSTRACT: A rheological model is used to describe the long-term behaviour of masonry. Special attention is
paid to the evolution of damage and the description of the damage parameters, based on the continuum damage
theory. The parameters of the model are derived from a test program, including three types of mortar compositions
and three kinds of tests: monotonic compression, short-term and long-term creep tests. Furthermore, a range of
stress evolutions is simulated to evaluate the model and the results are compared with experimental test data.

1 INTRODUCTION model, with a combined spring-dashpot configuration,

To describe the long-term behaviour of historical
masonry, rheological models have been used dur-
ing past researches. (Boukharov 1995, Anzani 2005)
These models are phenomenological models, based on
the stress-strain curves and strain-time behaviour of
the masonry during experimental research. In order
to link the model’s behaviour to specific types of
masonry, the model’s parameters are deduced from
masonry characteristics, recorded during short-and
long-term testing on masonry.
Under high, sustained stresses, a load condition
which typically occurs in historical masonry towers
and churches, the so-called brittle masonry exhibits
a ductile behaviour, showing an increasing visco-
elastic deformation in time. Therefore, the Maxwell
is often used to describe this creep effect. (Boukharov
1995, Papa & Talierco 2005)
Experimental research has shown that masonry
under a high persistent loading follows a deformation
pattern which can be described by the typical three-
phase creep curve. It exhibits a primary creep phase
during which the strain rate decreases in time, a sec-
ondary phase or steady-state creep, with a constant
strain rate and a tertiary phase with an increasing strain
rate, which leads to a sudden failure of the specimen
(Binda 1993, Challamel 2005). In order to enable the
model to capture the primary creep phase, a Kelvin
component is placed in series with the Maxwell com-
ponent, forming a rheological model which is able to
simulate the first two stages of the creep curve.
Furthermore, damage parameters are added to the
model in order to describe the damage accumulation
and loss of material cohesion, which will trigger the
unstable strain increase of the tertiary creep phase and
lead to failure of the masonry.

2 MODELLING OF DAMAGE
ACCUMULATION

There are different methods in describing the unsta-

ble damage accumulation during the tertiary creep
phase. Boukharov uses the Weibull cumulative distri-
Figure 1. Maxwell and Kelvin component. bution and experimentally defined strain values and

525
failure times to describe the growth of ‘hot’ crack
arrays up to a critical density. In the model presented
here, a continuum damage mechanism will be fol-
lowed, describing the overall damage throughout the
material as a continuous phenomenon.
The theory of Kachanov (Kachanov 1985, Kra-
jcinovic & Lemaitre 1987), which lies at the base
of continuum damage mechanics, poses to describe
the absence of material deterioration by means of a
field variable, the continuity. The more frequently used
damage parameter (D) can be seen as the comple-
ment of this continuity. For a completely undamaged
material D equals 0, whereas D = 1 corresponds to a
completely deteriorated material. This model has been
extended in terms of the introduction of the effective
stresses, which act upon a decreasing surface area as
a result of growing micro voids.
With s being the effective stress, which increases
when damage increases and σ the distributed stress.
For an undamaged material, s equals σ.
The damage variable, which in general is described
as a tensor, can be reduced to a scalar variable in
case of an isotropic material. The proposed model
is a time-dependent isotropic damage model, which
enables a simple scalar notation. The aim is to link the
parameters of the model to specific material proper-
ties of different types of masonry, in order to predict
the masonry’s behaviour during the experimental tests
from which the parameters were deduced as well as
to extend the model towards the prediction of real sit-
uations at which high sustained loading acts during
periods of hundreds of years. Therefore, the exper-
imental program includes short-term and long-term
creep testing on masonry wallets, built with three
types of typical mortar compositions: cement mor-
tar, hydraulic lime mortar and cement-lime mortar.
The complete set-up of the test program is discussed
elsewhere (Verstrynge 2008).
The dilatant behaviour of the masonry is simulated
by including an increasing, damage-dependent Pois-
son’s ratio. Previous studies have indicated that when
a constant Poisson’s ratio is considered, the horizontal
deformation is highly underestimated.
Two types of damage are considered: the static
damage, which influences the Young’s modulus in the
elastic domain, and the viscous damage, which influ-
ences the strain rate during steady-state creep and
triggers the tertiary creep phase.
2.1 Static damage
The static damage parameter (Ds ) decreases the
Young’s modulus of the Maxwell spring as dam-
age increases. This will result in higher strains at
526
Figure 2. Decrease of Young’s modulus due to damage
increase (from Hult 1987, p. 29).
stress increase. The amount of static damage can be
calculated as proposed by Hult, see Figure 2.
From the relation
Follows:
D = 1 − E  /E
D = 1 − E  /E
...
With E the initial, maximum Young’s modulus
The evolution of the Young’s modulus during short-
term creep test on masonry wallets (29 × 19 × 85 cm)
was calculated according to Equation 2. The tests were
performed by applying a constant load of 50% of the
masonry compressive strength (fc ) during a period of 2
hours. Subsequently, the load was increased by 5% and
again kept at a constant level. This loading scheme was
followed until failure of the specimen. In total, 9 wal-
lets were tested, of which 3 wallets were tested for each
of the three aforementioned mortar types respectively.
Figure 3 shows that the small stress increase and
related very small strain increase for every load-step
result into a rather large scatter on the evolution of the
Young’s modulus.
Therefore, extra short-term creep tests were per-
formed on masonry cores with cement mortar (diame-
ter 113 mm, height 10–18 cm), including an unloading
phase just before every stress increase in order to
enlarge the interval over which Young’s modulus was
calculated. The load was kept constant during a period
of 30 minutes. Figure 4 shows a representative result
of this test series.
A permanent decrease of theYoung’s modulus is not
found, so assuming this would result in an overestima-
tion of the elastic strain. Initially, an increasingYoung’s
modulus is found as a consequence of the closure of

Figure 3. Evolution of Young’s modulus in function of
stress at each step during a short-term creep test on a wallet
with lime-cement mortar. Stress is formulated dimensionless
(division by the maximum stress acquired during the test).
Figure 4. Evolution of Young’s modulus in function of
stress at each step during a short-term creep test on a masonry
core with cement mortar. Stress is formulated dimensionless
(division by the maximum stress acquired during the test).
the cracks oriented parallel to the plane of maximum
stress. After a maximum is reached, the static damage
increases due to the initiation and growth of cracks ori-
ented perpendicular to the plane of maximum stress.
When comparing the test specimen, there is a rather
large scatter on the stress level at which static damage
becomes important, sometimes this phenomenon even
only appears during the last load step. Therefore, an
initial Young’s modulus (calculated between 10–30%
of fc ) will be taken to simulate small stress levels and
an average of the higher values for each mortar type
will be used for the simulation of stress increases at
higher stress levels (over 50% of fc ).
527
2.2 Viscous damage
As the aim of the proposed model is to simulate long-
term deformations of masonry, viscous damage, which
governs the strain rate at constant stress levels and ini-
tiates the tertiary creep phase, has a highly important
role. The viscous damage parameter (Dv ) decreases
the relaxation time of the Maxwell dashpot, thereby
increasing the strain rate during the steady-state creep
and describing the accumulation of damage up to fail-
ure. In order to describe the evolution of Dv in time,
a twofold definition is considered: firstly, the viscous
damage parameter is a function of the stress level, as a
relation between the strain rate and the stress level has
been observed during the current test program, as well
as in literature (Challamel 2005, Anzani 2000). Sec-
ondly, a damage-rate formulation is adopted to enable
the simulation of damage increase in time when the
stress level remains constant.
To establish a relation between the amount of vis-
cous damage and the stress level, Dv was calculated for
each load step of the aforementioned short-term creep
tests on masonry wallets, according to the equation for
a Maxwell dashpot:
With εsec the strain increase during a secondary
creep phase, τM the relaxation time, EM the elastic
modulus of the Maxwell element and t the time incre-
ment over which the strain increase is calculated. For
each load step, τM (1-Dv ) is calculated according to
Equation 3. Initially, these calculated values remain
constant or increase slightly. After a few load steps, a
maximum value is found, which is set as initial value
for τM . The decrease of the relaxation time within the
next load steps is then calculated as an increase of DV .
When the stress is described in terms of percentage
of the masonry’s compressive strength (fc ), with fc the
strength during a short-term monotonic loading test,
a linear relation is found between the viscous damage
and the stress level (Fig. 5). This relation appears to be
similar for the different mortar types.
A threshold of around 70% of fc is found beneath
which no viscous damage is detected. Below this stress
level, the relaxation time (τM ) remains almost con-
stant or shows a slight increase. When the threshold is
crossed, the relaxation time decreases as the strain rate
becomes higher with each successive load step. This
linear regression curve can be described by means of
two parameters. Only the results for the short-term
tests on lime and cement mortar are presented, as the
tests on the lime-cement mortar included some unload-
ing and reloading cycles, which had a corruptive effect
on some of the results.

Figure 5. Viscous damage parameter (DV ) in function of
stress level from short-term creep tests on lime and cement
mortar wallets. Stress is formulated dimensionless.
In function of stress, DV can be formulated as
follows:
With σ ∗ being the dimensionless stress, acquired by
dividing the absolute stress by the average compressive
strength (fc ) of each mortar type, obtained during the
monotonic compression tests.
Secondly, a damage-rate formulation is used to
enable the simulation of damage increase in time
when the stress level remains constant, according to
Kachanov (Kachanov 1985, Challamel 2005):
With s being the effective stress, as specified in
Equation 1.
This simple formulation has the advantage that
only two parameters have to be determined. Damage
evolution is described using a power law equation.
Alternatively, others have used an exponential equa-
tion for damage development under constant stress
(Anzani 2005, Bodner & Chan, 1986). The two param-
eters, c and n, are determined by fitting the tertiary
creep curve. Unfortunately, experimental test data are
lacking for two out of the three mortar types as tertiary
creep was not reached or only occurred very shortly
during the short-term creep tests. Consequently, the
parameters are fitted to the results of the masonry wal-
lets built with hydraulic lime mortar, until more test
data are available from the long-term creep tests.
3 DETERMINATION OF THE PARAMETERS
The complete formula of the reological model, com-
bining a Maxwell and a Kelvin component, can be
528
presented in a simple constitutive formulation, to
describe the vertical deformation, as follows:
With M as notation for the parameters of the
Maxwell element and K for the material character-
istics related to the Kelvin element. The parameters
are calculated from monotonic compression tests and
short-term creep tests.
3.1 Experiments
As already mentioned, the test programme included
tests on masonry wallets (29 × 19 × 85 cm) with three
types of typical mortar compositions: cement mortar,
hydraulic lime mortar and a hybrid, cement-lime mor-
tar. And an extension of the test programme towards
air hardening lime mortar has been planned.
On all masonry types, three kinds of tests have been
performed (A more detailed discussion of the results
can be found in Verstrynge 2008):
– Monotonic compression tests, under a deformation
controlled environment;
– Short-term creep tests (also called accelerated
creep tests or step-by-step tests), at which stress
is increased in steps of 5% and in between kept
constant for a duration of 2 hours;
– Long-term creep tests, during which masonry wal-
lets are loaded at different constant stress levels.
The long-term tests were set up in order to assess
the strain rate of the different mortar types at different
stress levels. These tests are still ongoing. A second
goal of these tests is to assess the simulation of long-
term behaviour of masonry, by means of the proposed
model and the model parameters which are derived
from short-term tests. The model parameters were
calculated as an average for each mortar type. At
least three test specimens were made for each mor-
tar type and each type of test, which results in a total
of 27 specimens. The resulting model parameters are
summarized in Table 1.
The compressive strength of the masonry (fc )
is obtained from monotonic compression tests. The
strength of the cement mortar was remarkably low,
possibly due to bad curing conditions and absorption
of the water out of the mortar by the bricks, leading to
a weaker mortar.
The Poisson’s ratio (υ) is calculated from the com-
pression tests, without considering the last part of
the tests, where dilatation occurs and the Poisson’s
ratio suddenly increases. This dilatant behaviour of the
Table 1. Overview of the parameters used during the sim- step of the short-term creep tests, revealed a power
ulations with the rheological model. Presentation of results: law relation between the Maxwell relaxation time and
average value (± standard deviation). the time-interval during which the stress remained
constant;
Parameters of
rheological Hydraulic Cement Hybrid
model lime mortar mortar

fc [MPa] 6.3 ± 0.9 5.7 ± 0.2 7.3 ± 0.3

υ [−] 0.12 ± 0.05 0.16 ± / 0.29 ± 0.12
With α having a value of 0.7–0.8. Consequently, τM
τK [s] 300000 300000 300000 will increase in time while the stress level remains con-
EK [MPa] 1959 1764 2112 stant, due to a strengthening effect which in rheology is
τM [s] short-term 253200 491300 357200 described as a time-dependent viscosity. On the other
long-term 200e6 200e6 200e6 hand, τM will also decrease in time as a consequence
EM [MPa] 10–30% fc 2090 2390 2800 of the damage accumulation due tot stress increase and
> 50% fc 4000 4000 4000 time-dependent deformations under constant stress
DV A 0.52 0.57 0.45 level, which will lead to the tertiary creep phase.There-
B −2.24 −2.24 −2.24
c 3e−15 3e−15 3e−15 fore, the evolution of the viscosity was only assessed
n 6 6 6 during the first load step of the short-term creep test,
where the viscous damage has a negligible influence.
As Equation 8 was deduced from short-term tests, it
is not fully able yet to simulate long-term behaviour.
masonry is incorporated in the model by assuming a Therefore, a constant value will be taken for τM which
damage-dependent Poisson’s ratio, which increases as is based on experience from the ongoing long-term
the viscous damage increases. creep tests.
The Kelvin relaxation time (τK ) and the Kelvin elas-
tic constant (EK ) are derived by fitting the primary
creep phase of the short-term creep tests. These are 4 SIMULATIONS
not discussed here in detail, as their importance for
the overall simulation is limited. The constitutive equation describing the stress-strain
The initial Maxwell elastic constant (EM ) is calcu- relationship in function of time has been implemented
lated from the compression tests as secant modulus as Matlab code (The MathWorks, version R2007a).
between 10–30% of fc . When the stress level exceeds The model is evaluated in time increments, which are
50% of fc , a higher value is used during the simulation kept as small as possible. The size of the time incre-
as a consequence of crack closure and compaction of ments is chosen to be an optimum between the loss of
the material, as discussed above. This value is calcu- accuracy and the reduction of calculation time while
lated as an average value of the different load increase the size of the time increment increases. Therefore,
steps during the short-term creep tests. smaller time increments are applied for the simulation
The parameters A, B, c and n, which describe the of short-term tests. Each strain increment is com-
evolution of the viscous damage (DV ) are calculated posed of a linear strain increment, caused by the stress
as mentioned above. increase (elastic strain) and a creep strain increment,
Two different values are mentioned for the Maxwell caused by the initial stress at the beginning of the time
relaxation time (τM ), one value which is used to sim- step. This initial stress is assumed to be constant within
ulate a short-term creep test and a higher value to the interval of the specific time step.
simulate long-term behaviour. The relaxation time is
related to the “viscosity” (η) of the material, and is
4.1 A qualitative evaluation of the model
a measure for the strain rate during the steady-state
creep phase. To evaluate the general behaviour of the model, strain-
time curves are simulated for masonry at a range
of different constant stress levels. Figure 6 shows a
simulation of the strain evolution when a constant
load of respectively 75, 80 and 85% of the compres-
Initially, τM was taken to be constant with time, which sive strength is applied on the tested masonry with
resulted in very high deformations and significant hydraulic lime mortar.
underestimations of the time to failure. Comparison Failure would occur after a period of respectively 23
of the steady state phase during short-term and long- years, 3 years en 2.8 months. For a load level of 50%,
term tests revealed that the assumed viscosity of the a failure time of over a thousand years is calculated.
material is time dependent and increases in time. Cal- The three different creep phases are not clearly to
culation of the evolution of τM in time during the first distinguish, as the graph concerns a very large time

529
 85%
80%
75%
Figure 6. Simulation of strain evolution in time for masonry
with hydraulic lime mortar at a constant stress level of 75, 80
and 85% of the compressive strength.
85%
80%
75%
Figure 7. Enlargement of the simulation of the primary
creep phase for masonry with hydraulic lime mortar at a
constant stress level of 75, 80 and 85% of the compressive
strength.
span. A closer look on the primary creep phase is
shown in Figure 7, which is an enlargement of the
rectangle in Figure 6. The evolution of the viscous
damage in time is indicated in Figure 8.
Subsequently, the rheological model is used to sim-
ulate a creep failure curve for each mortar type.
Therefore, a series of constant stress levels are applied,
ranging from 72 to 90% of fc , and the related vertical
strain at failure is simulated. The obtained creep fail-
ure curves are presented in figure 9. The data point
related to the highest stress for every mortar type is
the strain at failure obtained during the experimental
compression tests. These data appear to be well in line
with the others.
The creep failure curve of the strongest mortar is
situated above the others. When a constant stress is
applied, the masonry with the weakest mortar will fail
530
85%
80%
75%
Figure 8. Evolution of the viscous damage parameter (DV )
for masonry with hydraulic lime mortar at a constant stress
level of 75, 80 and 85% of the compressive strength.
Figure 9. Simulation of creep failure curve for three differ-
ent mortar types.
sooner and have a smaller vertical creep strain at fail-
ure. This is plausible, considering that this stress level
will be closer to the compressive strength of the weak
masonry. For the stronger masonry, this stress level
is quite low, leading to higher creep strains and longer
times to failure. When the same stress level in percent-
age (relative to the compressive strength) is applied, the
weaker mortars will have larger deformations before
failure occurs.
The data point related to the lowest stress level is
the failure strain obtained at 72% of the masonry’s
compressive strength. As the analysis of the vis-
cous damage indicated that the threshold for viscous
damage was situated just below 70% of fc , an asymp-
totic strain value can be expected here. This is
in literature indicated as the viscosity limit curve
(Bazant 1988). The simulations have shown that below
70% of fc the failure time indeed increases rapidly.
Figure 10. Experimental results and simulation of
short-term creep test on hydraulic lime masonry, column 3.
Evolution of horizontal en vertical strain in time (above) and
volumetric strain (below).
The existence of the viscosity limit curve, which can
be reached without failure of the specimen, has not
been implemented here.
4.2 Comparison with experimental data
Subsequently, the model is used to simulate the short-
term creep tests from the abovementioned test pro-
gram. Two representative simulations are presented
below (Figs 10–11).
Here, the comment has to be made that some simu-
lations showed a rather poor fit to the results. This was
not due to an inadequacy of the model, but because of
the unpredictable behaviour of some specimens. After
all, masonry is a very inhomogeneous composite, with
a rather large spread on its parameters. The parame-
ters were calculated to fit a specific type of masonry
rather than one specimen whose behaviour was to be
simulated. Therefore, a stochastic description will be
studied in future analysis.
531
Figure 11. Experimental results and simulation of
short-term creep test on hydraulic lime masonry, column 5.
Evolution of horizontal en vertical strain in time (above) and
volumetric strain (below).
5 CONCLUSIONS
The long-term behaviour of masonry was described
using a rheological model. The damage increase was
modelled by means of a continuum theory and special
attention was paid to the evolution of the static and
viscous damage.
The parameters of the model were obtained from a
test program on new masonry, which included mono-
tonic compression tests and creep tests. Parameters
were calculated as average characteristics for each
specific mortar type rather than determined to fit a cer-
tain specimen. Consequently, the spread on the mean
parameter values could also be determined and con-
firms a rather large scatter. Therefore, further analysis
will have to include the spread on the determined
parameters in order to create a reliability interval for
the simulated failure time. This issue lies outside the
scoop of the present paper.
 The proposed straightforward model is able to
describe isotropic behaviour of masonry, includ-
ing monotonic stress increases and high sustained
stresses. As other phenomena influence the stress-
strain behaviour during short-term, monotonic load-
ing, this loading condition is not considered here.
A comparison with experimental data has shown a
satisfying fit between the model and the experimen-
tal results, but more experimental testing is necessary
in order to enlarge the amount of available data and
improve the fitted parameter values.
ACKNOWLEDGEMENTS
The authors express their thanks to the Flemish Fund
for Scientific Research (FWO) for the doctoral grant,
offered to Els Verstrynge.
The authors highly appreciate the collaboration with
the Politecnico di Milano and the University of Minho
in this research area.
REFERENCES
Anzani, A., Binda, L. & Roberti G.M. 2000. The effect of
heavy persistent actions into the behaviour of ancient
masonry. Materials and Structures. Vol. 33 (228):
251–261.
Anzani, A., Binda, L. & Taliercio A. 2005. Application of a
damage model tot the study of the long term behaviour
of ancient towers. Proc. 1st Canadian Conf. on effective
design of structures, Ontario, 10–13 July 2005.
Bazant, Z.P. (ed.) 1988. Mathematical modelling of creep and
shrinkage of concrete: Wiley & Sons Ltd.
532
Binda, L. & Anzani, A. 1993. The time-dependent behaviour
of masonry prisms: an interpretation. The masonry society
journal. Vol. 11 (2).
Bodner, S.R. & Chan K.S. 1986. Modeling of continuum
damage for application in elastic-viscoplastic constitutive
equations. Eng. Fracture Mech. Vol. 25 (5/6): 705–712.
Boukhavrov, G.N., Chanda M.W. & Boukharov N.G. 1995.
The three processes of brittle crystalline rock creep. Int. J.
Rock Mech. Min. Sci. & Geomech. Vol. 32 (4): 325–335.
Challamel, N., Lanos C. & Casandjian C. 2005. Creep
damage modelling for quasi-brittle materials. European
Journal of Mechanics A/solids 24: 593–613.
Hult, J. Introduction and general overview. 1987. In Con-
tinuum Damage Mechanics Theory and Applications,
D. Krajcinovic and J. Lemaitre (eds.): Springer–Verlag,
New York.
Ignoul, S., Schueremans L., Binda L., et al. 2006. Creep
behavior of masonry structures – failure prediction
based on a rheological model and laboratory tests. In
P.B. Lourenço, P. Roca, C. Modena, S. Agrawal (Eds.)
Proc. Of the 5th int. seminar on structural analysis of
historical constructions.Vol. 2: 913–920. New Delhi 2006.
Kachanov, L.M. 1985. On creep rupture time. Proc. Acad.
Sci. USSR Div. Eng. Sci. 8: 26–31
Krajcinovic, D. & Lemaitre, J. (eds). 1987 Continuum damage
mechanics: Theory and applications: Springer–Verlag,
New York.
Papa E. & Talierco A. 2005. A visco-damage model for brittle
materials under monotonic and sustained stresses. Inter-
national journal for numerical and analytical methods in
geomechanics, Vol. 29 (3):287–310.
Verstrynge, E., Ignoul, S., Schueremans, L.,Van Gemert, D. &
Wevers, M. 2008. Damage accumulation in masonry under
persistent loading evaluated by acoustic emission tech-
nique. Proc. 14th Int. Brick & Block Masonry Conference,
Sydney, 17–20 February 2008.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

A method for studying multi-body dynamics and its applications to

masonry structures analysis

Giuseppe Gariup
Departamento de Ingeniería de la Construccíon, U.P.C., Barcelona, Spain

Igino Pitacco
Dipartimento di Ingegneria Civile, Università degli Studi di Udine, Udine, Italy

Pere Roca Fabregat

Departamento de Ingeniería de la Construccíon, U.P.C., Barcelona, Spain

ABSTRACT: A multi-body approach to solid dynamics is presented. The numerical method implemented
is a discrete element method (d.e.m.) specifically intended for the analysis of masonry structures. Con-
tacts between different blocks are treated using interacting forces. Contact forces are modelled using penalty
method for the normal component and an elasto-plastic behaviour for the shear. The model considers 3
different types of damping. A first quantitative validation phase is presented together with possible future
developments.

1 INTRODUCTION

A brief review of the theoretical details and the numeri-

cal implementation of a discrete numerical method for
the analysis of blocky structures is presented together
with an initial phase of validation of the model.
Discrete element methods were first introduced to
study stability of jointed rock (Cundall 1992) and were
later applied to static and dynamic analysis of granular
material and finally to seismic analysis of masonry
structures (Mamaghani 1999).
Masonry, due to the presence of mortar or dry joints
separating stone blocks or bricks, can be considered a
discontinuous material (Lourenço 1996).
Each brick or stone block is modelled as a sin-
gle block with infinitesimal deformability. The finite
deformation of the structure, typical of a seismic
response, is therefore ascribable to the relative roto-
translation of the blocks (Fig. 1).
Experimental campaigns on loading walls and ma-
cro-blocks structures (Oliveira 2003) and observation
of seismic effects on real structures (Orduña 2003)
confirm that collapse of masonry structures is mainly
due to relative displacements of bricks or stones. Figure 1. Collapse due to roto-translation of rigid locks: (a)
The numerical method should be used at first to geometrical model; (b) degrees of freedom of the system; (c)
perform dynamic 2D and 3D analyses of structures first collapse mode (overall excessive rotation); (d) second
formed by a limited number of macro-blocks, like collapse mode (coupled effect between excessive sliding and
stone arches or loading walls, modelling each stone or rotation).

533
brick with a single block. Applicability of this method 2 CONCEPTUAL MODEL
to complex structures has been considered in the last
section of this paper. Discrete element method analyzes the interactions
The model is characterized by the following aspects: between a finite number of distinct blocks. Each block,
in the implementation used in this work, is elas-
– Formulation of the physical background for the
tically deformable, while non-linear deformation is
development of the model (Simo 1998). Choice of
concentrated in the contacts.
finite strain theory and hyper-elastic behaviour of
All the equilibrium and constitutive equations have
block material.
been obtained under the hypothesis of finite defor-
– Geometrical and finite element definition of the
mations. The deformation of a single block has been
blocks (Belytschko 2000).
considered infinitesimal but the one of the whole sys-
– Definition of the contact algorithm including the
tem, due to possible roto-translation of its parts, can be
geometrical definition of contact, contact updating
large compared to the usual finite elements problems.
through the analysis, the description of the normal
Each block or body is composed by one or more
response using a penalty method approach and the
8-node hexahedrons depending on the complexity of
description of the shear response considering stick-
the geometry. In the contact routine each body is iden-
ing and sliding and plastic behaviour of the shear
tified and interact with the others with its external
reaction (Wriggers 2002).
surface.
– Definition of damping, considering mass, stiffness
The external surface of a block is given by the
and contact damping (Zhong 1993).
external faces of the finite elements forming it (Fig. 2).
– Numerical method used to solve the system of dif-
At the beginning of the analysis all the possible pairs
ferential equations representing the f.e. problem.
of contacting faces are determined and stored in an
Choice of an explicit integration algorithm (Bathe
array. The uncertainty is introduced by the roughness
1996).
of this first check: there can be surfaces included in a
The phase of validation is focussed on two-block contact pair which are not actually touching each other
dynamic analysis. The two blocks are superposed, the but are simply close enough to trick the contact crite-
lower one serves as “ground” for the motion of the rion. During the calculation, at each time step, each
upper one and displacements are constrained on its contact pair is considered separately and is marked as
base. Different cases are modelled and analyzed with active or not-active. A contact pair is considered active
the numerical code. The results are confronted with only if one surface is penetrating into the other and
analytical ones or with the expected behaviour. Each some other criteria established to avoid singular situ-
case has been specifically chosen to test a particular ations are satisfied. This procedure has to be repeated
characteristic of the model: during the calculation every n time-steps to follow the
physical evolution of the system.
– Stability: the upper block is leaning on the lower Each active contact pair represents a contact
one at the beginning of the analysis. Gravity is the between two blocks. The procedure to calculate the
only force considered.
– Impact: the upper block is placed at a vertical dis-
tance d from the lower one at the beginning of the
analysis. Gravity is the only force considered.
– Friction: the upper block leans on the lower one
at the beginning of the analysis. Gravity is applied
together with a horizontal force. Friction is studied
varying the static friction coefficient of the contact.
– Updating of contact: the upper block leans in a
instable position on the lower one whose upper face
is inclined. The evolution of motion is given by
the overturning and successive rolling of the upper
block. Gravity is the only force considered.
– Damping tests: the upper block is inclined with
respect to the vertical and touches the lower one
with an edge at the beginning of the analysis.
Varying the damping coefficients different rocking
responses are simulated. Gravity is the only force
considered.
Figure 2. 2D representation of blocks: (a) single finite ele-
In the next sections the theoretical introduction and the ment; (b) elements forming a block; (c) block composition;
results of the numerical tests are presented. (d) block boundary.

534
forces belonging to each contact consists in the fol-
lowing:
– The larger surface is identified as the target-surface
and the other as the hitting-surface.
– If the distance between the 4 nodes of the hitting-
surface is higher than a tolerance, a regular grid of
virtual nodes is created on it.
– The position of each node, real or virtual, of the hit-
ting surface, relative to the target one, is calculated.
If the node is not penetrating, the pair hitting-
node/target-surface is considered non-active.
– If the node is penetrating the target surface, normal
and shear forces are calculated and added to the
forces acting on the two surfaces.
The normal force at contact is calculated using the
penalty method. In this approach a small penetration
is accepted during the impact of two blocks. After the
penetration, a reaction force, proportional to the pen-
etration and the penalty stiffness, orthogonal to the
target surface, starts to counteract the superposition
of the two bodies. Choosing an adequate penalty stiff-
ness the penetration can be limited to an acceptable
threshold.
The shear force has been modelled with elastoplas-
tic behaviour. Once a hitting point penetrates for the
first time in the target surface, the natural coordinates
X H 0 of the normal projection of the point onto the sur-
face are stored together with the point-surface contact
pair. At the successive time-steps has to be checked if
the point is still penetrating the surface. In this case
the new coordinates of the hitting-point X H 1 have
to be calculated. The shear force has been modelled
with elasto-plastic behaviour. At first the distance d
between X H 0 and X H 1 is calculated and multiplied
Figure 3. 2D representation of non-active and active con-
tact pair: (a) lock and bounding box (it includes the block
and permit a first rough contact test); (b) non-active contact
pair (superposition of bounding-boxes, no block interpene-
tration); (c) active contact pair (superposition of bounding-
boxes and block interpenetration).
535
by the shear stiffness K S . The value obtained is con-
fronted with the absolute value of the normal penalty
force Fn multiplied by the static friction coefficient µ.
If d·KS ≥ µ. · Fn , the shear force is still in the elastic
range, its absolute value is given by d·KS , the direction
and verse is given by the vector connecting X H 1 and
X H 0.
If d · KS < µ. · Fn , the shear force, calculated as an
elastic reaction to the displacement, exceeds the static
friction force. In this case the direction and verse of
the shear force are the ones calculated in the elastic
range while the absolute value is scaled to µ. · Fn .
The contact forces, together with the external ones,
are used to calculate the displacements and the defor-
mations of the blocks at the successive time-step.
Central difference method has been implemented to
solve the dynamic differential system given by
where u is the displacement vector, ü its second time
derivative, M the mass matrix, F ext is the vector of
external force, F c is the vector of contact forces,
F int = K · u is the vector of internal forces and K is
the stiffness matrix.
The reasons leading to the choice of an explicit
method in the integration of Equation 1 are mainly its
robustness and its simplicity, suitable characteristics
to deal with problems with many non-linearities.
In performing dynamic analysis with any code, it
is usually necessary to account for energy losses in
the physical system which are not considered in the
numerical algorithm. In this formulation the elasto-
plastic behaviour of contact shear is the only possibility
the system has to dissipate energy. Other forms of
damping have to be introduced in order to get a better
modelling of real structural dynamic behaviour.
The damped system of equations becomes
where C is the damping matrix and u̇ is the first
derivative of the displacement vector.
Three types of damping have been introduced:
– Mass damping. The damping matrix is proportional
to the mass matrix:
– Stiffness damping. The damping matrix is propor-
tional to the matrix corresponding to finite element
stiffness:
– Contact damping. The damping is proportional to
a contact stiffness matrix that takes into account a
possible normal damping at contacts:
The values of constants α, β and γ have to be
determined to adjust the model to experimental
observations.
The third form of damping seems the one with
stronger connection to the dissipative nature of the
impact. The first one has no physical meaning and it is
purely a numerical tool to dissipate energy. Stiffness
damping is related to the dissipation due to anelastic
phenomenons in the deformation of the blocks which
are normally neglectable in masonry analysis.
In the next section some examples of discrete
element analysis are presented and validated.

3 EXAMPLES

A first phase of testing and validation of the dis-

crete element numerical code has been carried out.
Numerical simulations of different cases of two-block
interactions have been completed. The material com-
posing the blocks has been considered possessing the
mechanical properties of stone:
Figure 4. Stability: (a) geometrical model; (b) forces and
constraints; (c) node numbering of block A.

The tangential or shear contact stiffness K s has been

considered equal to the normal contact stiffness or
penalty parameter K c , which has been chosen in order
to limit the penetration to an acceptable threshold:

Figure 5. Stability: oscillation in the Z component of node

10 around the equilibrium position for an undamped model.

3.1 Stability
Two cubic blocks A and B with 1.0 m side are super-
posed.
The lower one B has its base constrained. The grav-
itational force is the only force acting. Due to the
deformability of the blocks this is not an equilibrate
configuration.
The geometrical configuration, forces and con-
straints and node numbering are summarize in
Figure 4.
A tends to press B which will react deforming and
lowering its upper face. The entity of this displacement
is neglectable compared to the residual penetration
imputable to the penalty approach. During the impact
A does not encounter a physical obstacle in B, but
experiences a force representing the contact. This force
tends to contrast the interpenetration, when it reaches
a certain value it has a repulsive effect and pushes A
upward.
The equilibrium is obtained when the magnitude of
the penalty reaction d · K s equals the weight W A of A.
d is therefore the penetration that has to be tolerated
in the analysis and depends on the penalty stiffness K c .
If no damping is considered in the analysis, A will
oscillates indefinitely around the equilibrium position
with more or less the same amplitude (Fig. 5).
The analysis was performed without damping to test
the numerical stability of the method.
3.2 Impact
Two cubic blocks A and B, with side equal to 1.0 m,
are considered. The blocks are aligned in the horizontal
directions. The lower one B has its base constrained.

536

Figure 6. (a) Geometrical model; (b) forces and constraints;
(c) node numbering of block A.
In the initial configuration the bottom of block A is
placed 10 cm above the upper part of B.
The gravitational force is the only force acting.
The geometrical configuration, forces and con-
straints and node numbering are summarize in
Figure 6.
At first no damping is considered.
A should bounce on B going back periodically to
the initial configuration. At first no damping is con-
sidered. A should bounce on B going back periodically
to the initial configuration.
This configuration should test the reaction of the
contact, as any error or deviation in the calculation of
the penalty reaction would cause a visible change in
the motion of A.
Using the same geometrical configuration contact
and mass damping are included to see the effects of
the dissipation of energy.
A comparison between the undamped and a damped
solution is shown in Figure 7.
3.3 Friction
A cubic block A with side equal to 0.5 m, is placed
upon the upper face of a parallelepiped B with base
equal to 0.7 m · 4.0 m and height 1.0 m. Besides the
gravitational force, a horizontal force H is considered.
H magnitude is equal to 0.5 · WA , where WA is the
absolute value of the gravitational force acting in A,
and its direction is the same of the longer edge of
B base.
537
Figure 7. Z displacement of node 10: comparison between
a model with contact damping equal to 0.1 and an undamped
model.
Figure 8. Friction: (a) geometrical model; (b) forces and
constraints; (c) node numbering of block A.
The geometrical configuration, forces and con-
straints and node numbering are summarize in
Figure 8.
No damping is considered in this analysis.
This configuration is useful to check the behaviour
of the shear force at contact.
At first the static frictional coefficient is set equal
to 0.4. The force applied horizontally is higher than the
elastic limit of the shear force and sliding is expected.
The horizontal distance covered in the time-interval
(0, t) can be easily computed with
where a = 0.1 · g.
The analytical result can be compared with the
numerical one (Fig. 9).
Using the same configuration the static frictional
coefficient is set equal to 0.6. The elastic limit of the
shear force is therefore exceeding the applied force and
Figure 9. X displacement of node 10: comparison between
the analytical and the numerical solution.

Figure 11. Rolling: (a) geometrical model; (b) forces and

constraints; (c) node numbering of block A.

Figure 10. X displacement of node 10: elastic shear

response.

no plastic sliding should occur. The elastic sliding is

neglectable due to the high shear stiffness considered
in the model.
The elastic response is clearly depicted in Figure 10.

3.4 Contact updating

A cubic block A with side equal to 0.7 m lies onto a
45◦ inclined face (1.0 m · 4.2 m) of a hexahedron B. B
has its base constrained. The gravitational force is the
only force acting.
The geometrical configuration, forces and con-
straints and node numbering are summarize in
Figure 11.
The static frictional coefficient is set equal to 1.1.
Contact and mass damping have been considered Figure 12. Rolling: visual output of the numerical code at
in the analysis. The initial tilting couple, due to fric- instant t = 40.
tion, equals the righting one, due to W A , leading to an
instable configuration. A small perturbation, like the
3.5 Rocking and damping
one induced by the deformability of blocks or by the
penalty approach, can start the motion. A slender block A with base equal to 0.2 m · 0.7 m and
A should roll over the inclined face of B. The face height equal to 1.0 m lies with one of the longer edges
of A impacting with B is changing at each overturn of its base onto a parallelepiped B.
(Fig. 12). B has its base constrained. The shorter edge of A
This configuration is therefore useful to check the base forms a 30◦ angle with the horizontal plane. The
contact updating routine. gravitational force is the only force acting.

538

Figure 13. Rocking: (a) geometrical model; (b) forces and
constraints; (c) node numbering of block A.
Figure 14. Block A oscillation angle: comparison between
undamped and damped solutions (contact damping equal to
0.2 and 0.4).
The geometrical configuration, forces and con-
straints and node numbering are summarize in
Figure 13.
At first no damping is considered. This config-
uration should lead to a rocking behaviour of A.
The oscillations should not change amplitude during
the analysis. Using the same geometrical configu-
ration, contact damping is included. Depending on
the magnitude of damping considered, the system
should evolve with different velocities towards the
equilibrium.
A comparison between the analysis of an undamped,
a 0.2 and a 0.4 contact damping model is shown in
Figure 14.
539
4 ANALYSIS OF RESULTS
The initial phase of validation has shown a satisfactory
behaviour of the model with regards to the following
aspects:
– The method seems to be numerically stable.
– Modelling normal contact force with the penalty
method has given good qualitative results.
– The frictional behaviour of the numerical model is
the one anticipated from the elasto-plastic shear-
slip relationship used in the implementation of the
code.
– Contact updating is working properly, perceiving
changes in the contact configuration.
– Damping cause significance effects on the oscilla-
tion response of the block. Contact damping seems
to be the more suitable and manageable for struc-
tural analysis, at least when the elastic deformations
of the blocks are neglectable.
Despite the overall good behaviour of the model, a
problem has been encountered in the rocking motion
example. The block reaches the horizontal position
for the first time, for an instant ti lower than the one
obtained with the analytical solution and the difference
decreases with the refinement of the discretization of
the block.
The error introduced by the numerical method
is significative when the parallelepiped is modelled
using only 1 hexahedron and becomes tolerable when
the mesh consisted in 4 identical hexahedrons.
This behaviour is due to the undervaluation of the
moment of inertia and a resulting overvaluation of the
rotation of a block when it has not been modelled with
an adequate number of finite elements. That depends
on the fact that the algorithm has been developed con-
sidering the mass concentrated in the barycentre of the
element.
Obviously quadrupling the number of the elements
would critically increase the computational cost for
complex structures.
5 CONCLUSIONS
In this paper a discrete element method with
deformable blocks has been presented. The method
is intended for performing seismic dynamic analysis
on 3-dimensional models of masonry structures.
A problem has been encountered in the last section
when a block is modelled with an insufficient number
of hexahedral elements. It is due to the approxima-
tion introduced considering a non-consistent stiffness
matrix K.
Possible solutions are using a consistent stiffness
matrix in the calculation or using rigid blocks dynam-
ics, and thus describing roto-translations exactly.
 Calculating a consistent stiffness matrix would
increase the computational cost of the method.
Rigid-blocks analysis seems to be a viable option
considering that the deformation in bricks and stones
is negligible compared to block displacement. Never-
theless considering bodies deformable could be used
to take into account the deformations of the mortar
joints or possible damages of a block.
Future developments should explore these possi-
bilities, confronting results and choosing the most
suitable one.
A validation phase with experimental results (Peña
2005) is also needed and will be taken into account.
REFERENCES
Bathe, K. (1996). Finite element procedures. Prentice Hall.
Belytschko, T. & Liu, W. M. B. (2000). Nonlinear finite
elements for continua and structures. John Wiley & Sons.
Cundall, P. (1992). Universal distinct element code, User’s
manual. USA: ITASCA Consulting Group Inc.
540
Lourenço, P. (1996). Computational strategies for masonry
structures. Ph. D. thesis, Delft University of Technology,
Delft, The Netherlands.
Mamaghani, I.H.P. & Aydan, O. K. Y. (1999). Analysis
of masonry structures under static and dynamic load-
ing by discrete finite element method. Structural Eng.
Earthquake Eng., JSCE 16(2), 75–86.
Oliveira, D. (2003). Experimental and numerical analy-
sis of blocky masonry structures under cyclic loading.
Ph. D. thesis, Universidade do Minho, Guimares, Portugal.
Orduña, A. (2003). Seismic assessment of ancient masonry
structures by rigid blocks limit analysis. Ph. D. thesis,
Universidade do Minho, Guimares, Portugal.
Peña, F. (2005). Dinámica de bloques de piedra. Experi-
mentación y simulación numérica. Laboratorio Nacional
de Engenharia Civil, Proc. 0305/17/14483.
Simo, J.C. & Hughes, T. (1998). Computational inelasticity.
Springer.
Wriggers, P. (2002). Computational contact mechanics. John
Wiley & Sons.
Zhong, Z. (1993). Finite element procedures for contact-
impact problems. Oxford University Press.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Load-carrying capacity of multi-span masonry arch bridges

having limited ductility

G. de Felice
Department of Structures, University Roma Tre, Rome, Italy

ABSTRACT: The assessment of load-carrying capacity of masonry bridges is usually based on limit anal-
ysis methods, which implicitly assume that masonry has unlimited ductility capacity in compression; aiming
at taking into account the effective ductility capacity of masonry, an approach that makes use of beams with
fibre cross is developed. The approach is able to take into account the non-linear behaviour of masonry in
compression, including the post-peak softening branch and therefore provides a reliable estimate of the ultimate
load. A numerical investigation of one to three span masonry arch bridges under travelling concentrated load,
shows that, by parity of constitutive assumptions, the method provides the same estimate of ultimate load given
by limit analysis, as well as the same kinematics at collapse. The influence of limited ductility is then investi-
gated and the overestimate provided by classical limit analysis is highlight for the previously examined bridge
typologies.

1 INTRODUCTION assumption of no- tensile resistance, allows over-

Masonry bridges have been thoroughly studied in the
course of the last three decades, especially in Great
Britain, in United States and in Italy; several experi-
mental investigations have been carried out on existing
bridges and on scale models which have revealed some
important aspects of their structural behaviour under
service loading and under ultimate load conditions;
only a few experimental work have been carried out
on multi-span arch bridges, among which the experi-
mental investigation on scale models by Prentice and
Ponniah (1994), Melbourne et al. (1995) and Robinson
et al. (1997).
As far as modelling is concerned, a wide literature
derives from Heyman’s studies on the yield design
of masonry arches (Heyman, 1966, 1982) among
which the contribution by Boothby (1992), Gilbert and
Melbourne (1994), Clemente et al. (1995). The great
simplicity of the model together with its feasibility for
practical applications, makes the yield design based
approach one of the most suitable methods for anal-
ysis of masonry structures; however, the use of such
method, point at the question of the limited ductility
capacity of masonry. It is well known that masonry has
a brittle behaviour in traction and a strain softening in
compression, with a decrease of load capacity after
the peek load. Therefore, the yield design approach,
which implicitly assumes a plastic behaviour of the
material, may lead to an overestimate of the load capac-
ity of the structure (Salençon, 1985). The classical
coming this problem in traction, that however still
remain in compression, especially for shallow or long
span masonry bridges, which may be subjected to
high loading with a relatively weak material strength,
and therefore the possible failure due to crushing
of masonry cannot be neglected. In these cases, the
limited ductility in compression can be handled by
applying a reduction factor to masonry strength, but its
calibration still remains one of the main drawbacks for
the use of yield design-based methods. An other draw-
back to the application of yield design-based methods,
stands in the lack of control for the displacement or
strain quantities, since the problem is driven by equi-
librium and strength conditions and the kinematics is
introduced only through the virtual work equation. On
the contrary, a displacement-based prediction of the
structural behaviour may be relevant for assessment
purposes, especially under time-varying loading.
Some recent approaches have succeeded in mod-
elling the behaviour of the bridge, such as the inter-
action of filling with the arch barrel (Cavicchi and
Gambarotta 2005) and the contribution of the span-
drel walls (Fanning and Boothby 2001), in line with
what was observed in the experimental tests. Despite
the results achieved, an intrinsic difficulty remains in
three-dimensional finite element modelling, deriving
from the high dependence on of the results on input
parameters and boundary conditions not easy to deter-
mine. In any cases, full 3D non-linear approaches
would generally require a high computational effort,

541
especially for multi-span bridges, which pertain to a
more refined analysis and is not justified for initial or
intermediate-level assessment.
The aim of this paper is to investigate the influence
of a limited ductility capacity of masonry on the load-
carrying capacity of arch bridges. To this end, a non-
linear incremental analysis tool is developed, along a
research line that dates back to the well-known study of
Castigliano (1894) for the bridge over the Dora close
to Turin, and includes some recent contributions by
(Boothby 1997; Molins and Roca 1998; Brencich and
De Francesco 2004). In the present paper, aiming at
taking into account the effective behaviour of masonry
in compression, the bridge is modelled using beams
with fibre cross section. The use of beams with fibre
cross section has become current for the analysis of
reinforced concrete structures but, at the knowledge of
the author, has not yet been used for structural analysis
of masonry arch bridges.
The paper is organized as follows. In section 2
the non-linear beam model with fibre cross section
is briefly recalled, then, in section 3, a comparison of
the results provided by the present approach and those
given by limit analysis is performed for one to three-
span railway arch bridges under travelling concen-
trated load; for the comparison, masonry is considered
either with infinite compressive strength or finite com-
pressive strength with unlimited ductility. Finally, in
section 4, the influence of a limited ductility capacity is
examined, using the Kent & Park model for describing
the uniaxial constitutive behaviour of masonry, when
varying the slope of the post-peak descending linear
branch. The numerical results, which provide the ulti-
mate travelling load of the bridges, highlight the effect
of material post-peak behaviour on the load carrying
capacity, showing to what extent the assumption of
unlimited ductility would lead to an overestimate of
loading capacity.
2 FIBRE BEAM MODEL
Some recent experiments on masonry prisms sub-
jected to eccentric compression (Brencich and Gam-
barotta 2005; de Felice, 2006) have shown that the
classical hypothesis that plane section remains plane
after deformation, holds true for brick masonry even in
the non-linear range, when stress concentrations and
cracks start to develop; besides, the intrinsic difficulty
in defining the non-linear constitutive behaviour of the
beam, as for instance the dependence of flexural hys-
teretic behaviour on axial load, also due to the lack of
experimental data, suggests deriving the section force-
deformation relation from the uniaxial stress-strain
behaviour of the material. Therefore, looking for a
compromise between simplicity and accuracy, the arch
is discretised into rectilinear beams and each beam
542
is subdivided into longitudinal fibres, whose stress-
strain behaviour is defined according to the material
properties that can be easily obtained through experi-
ments. The non-linear constitutive relation of the beam
section is not given explicitly but is derived by integra-
tion of the stress-strain relation of the fibres as follows
(Spacone et al. 1996). According to the assumption
that plane section remain plain and normal to the
beam axis x during the deformation history, referring
to figure 1, for a given section generalized defor-
mations d(x) = {χz (x) χy (x) ε(x)}, comprising the two
curvatures and the axial strain at the reference
axis, the strain distribution over the section writes:
ε(x, y, z) = a(y, z)d(x), where a(y, z) = {−y z 1}. Once
the material constitutive relation is specified, the
strain distribution yields the tangent material modu-
lus E(x, y, z) and the corresponding stress σ(x, y, z).
Therefore, according to the virtual work equality, the
section stiffness matrix k(x) and the resisting forces
D(x) = {Mz (x) My (x) N (x)}, comprising the two bend-
ing moment and the axial force, that correspond to
deformation d(x) are obtained by integrating over the
cross section A(x) as:
The evaluation of the integrals (1,2) is performed
by subdividing the generic section x into n(x) fibres
and applying the mid-point integration rule:
A flexibility-based formulation is adopted for the
FE analysis, which ensures that equilibrium is strictly
satisfied within the element, and allows to overcome
the numerical instabilities arising after the onset of
strain softening; the latter point is crucial in masonry
elements, much more than in reinforced concrete ele-
ments, since due to the lack of steel, the masonry
section displays a rapidly descending branch after
the peak load. According to flexibility approach the
element stiffness matrix is obtained by inverting the
flexibility matrix: K = F −1 , where
L being the element length, and b(x) the force interpo-
lation functions matrix, which is derived form the equi-
librium of axial force and bending moments within the
element.
 Figure 3. Three span bridges.

Figure 1. Fibre beam model. Table 1. Bridge typologies characteristics.

Shallow Deep
Geometrical and Mechanical arch arch
properties bridges bridges

Span (at the intrados) [mm] 15000 15000

Arch depth [mm] 750 750
Fill in crown [mm] 750 750
Masonry spec. weight [KN/m3 ] 22 22
Fill specific weight [KN/m3 ] 24 24
Rise (at the intrados) 3750 5250
Pier height [mm] 7500 7500
Pier depth [mm] 2200 2200

Figure 2. Two span bridges with deep (up) and shallow

(down) arches.

An iteration scheme based on Newton-Raphson

method is used for solving the non-linear equation at
the element level, while at the structural level, an arch
length method with Riks iteration scheme and adaptive
solution strategy is adopted.

3 LOAD-CARRYING CAPACITY OF
MULTI-SPAN MASONRY ARCH BRIDGES: Figure 4. Load diffusion through the fill.
COMPARISON WITH LIMIT ANALYSIS

In this section, the load carrying capacity of one to
three-span arch bridges representative of European
railway network (figures 2–3) has been estimated
according to the present approach that makes use of
fibre beam elements, and the results compared with
those provided by limit analysis. The same bridge
typologies studied in (Brencich and De Francesco
2004) have been considered, having 15 mt span and
0.75 mt arch depth; the bridges are made either with
shallow (rise/span ratio equal to 0.25) or deep arches
(rise/span ratio equal to 0.35); their geometrical and
mechanical characteristics are reported in table 1.
The ultimate value of a concentrated travelling load
has been evaluated for the classical hypothesis of
no-tensile resistance and either infinite compressive
strength, or 5 MPa compressive strength with unlim-
ited ductility. As regards the application of travelling
load, a diffusion angle trough the fill 2θ = 40◦ + 40◦
was assumed, and the effective loading on the arch was
schematised as uniformly distributed (figure 4).

543

Figure 5. Load carrying capacity of a single-span shallow
arch bridge, under concentrated travelling load.
Figure 6. Load carrying capacity of a two-span deep arch
bridge, under concentrated travelling load.
Each arch was modelled by 100 non-linear beams
with cross section discretised in 100 fibres. The results
obtained with the Fibre Beam (FB) model have been
compared to those obtained with Limit Analysis (LA)
making use of the program Ring (Gilbert 2005). The
figures 5–8, give the load carrying capacity of the
bridge (one meter depth) as a function of load posi-
tion for the different bridge typologies under study.
All the analyses show an excellent agreement between
the results provided by the two approaches, by parity of
constitutive assumptions, with differences that never
exceed 4%. When neglecting the compressive strength
of masonry, an overestimate of the ultimate load of
544
Figure 7. Load carrying capacity of a two-span shallow arch
bridge under concentrated travelling load.
Figure 8. Load carrying capacity of a three-span shallow
arch bridge, under concentrated travelling load in the central
span.
about 25% and 40% is obtained for deep and shallow
arch bridges examined typologies respectively.
As regards the kinematics of the bridge close to
collapse, the formation of sufficient number of plastic
hinges that turn the bridge into a mechanism takes
place.
As shown in figure 9, the collapse mechanism of the
bridges provided by FB approach is similar to the one
provided by LA in terms of “hinges” location, apart
from a different treatment of the kinematics of joint’s
opening. In fact, while in LA approach no overlapping
of the blocks in contact across a joint is allowed, in

Figure 9. Collapse mechanisms of two and three-span arch
bridges obtained with Limit Analysis (up) and Fibre Beam
(down) approaches respectively.
FB approach the arch deforms, also in compression,
especially in plastic hinges regions, according to the
stiffness of masonry.
It appears that for shallow arch bridges the col-
lapse involves the pier and the adjacent span, while for
two span deep bridges, only the loaded arch collapses,
according to the four hinges mechanism, that does not
involve the adjacent spans. This result is clearly influ-
enced by the slenderness (the height-to-depth ratio)
of the pier, as well as by the presence of the backfill,
that acts as a restraint to the single arch mechanism,
allowing a multiple arch mechanism to take place. The
influence of the backfill is highlight in figure 10, where
a three span arch bridge with deep arches is analysed
when varying the height of the backfill form 1 to 3
meters. in the case of fibre beam approach, the back-
fill is modelled trough non-linear horizontal trusses
able to carry compression only. the results also show
a good agreement between LA and FB approaches.
4 LOAD-CARRYING CAPACITY OF
MULTI-SPAN MASONRY ARCH BRIDGES:
EFFECT OF A LIMITED DUCTILITY
The other curves in figures 5–8 provide the load
carrying capacity of the bridge according to the FB
model in the case of limited ductility; the results
545
Figure 10. Load carrying capacity of three-span deep arch
bridges, under concentrated travelling load, when varying
backfill height.
have been obtained having made recourse to the
Kent and Park (1971) relationship for modelling the
uniaxial constitutive behaviour of masonry in com-
pression. Originally formulated for confined concrete,
the Kent & Park (K&P) relationship consists of an
ascending branch represented by a second-degree
parabolic curve, a subsequent linear descending soft-
ening branch and a final constant branch. The model
is defined by three parameters, namely the peak stress
fc , the corresponding strain εc , the ultimate strain εu .
Denoting by σ̃ = σ/fc , and ε̃ = ε/εc the normalized
stress and strain respectively, the three branches take
the form:
where η = εu /εc . As shown in a previous paper (de
Felice, 2006) the K&P model is able to accurately
reproduce the effective behaviour of masonry in
compression.
Aiming at evaluating the influence of the post-peak
behaviour, a parametric study is performed, where
the different curves in figures 5–8 correspond to dif-
ferent slopes of the descending branch, from plastic
behaviour with unlimited ductility (α = 0◦ , η = ∞), to
post-peak brittle behaviour (α = 90◦ , η = 1).
A significant reduction in loading capacity is
obtained when increasing the slope of descending
branch, more pronounced for shallow arches. For
the two span arch bridges under study, a decrease in
the capacity of about 17% and 24% is found when the
material behaviour turns from perfectly plastic to post-
peak brittle, in the case of deep and shallow arches
respectively. An even stronger dependence is found
for three-span shallow arch bridges, with a reduc-
tion in load capacity of about 32%, from 385 kN/m
to 260 KN/m.
These results can be compared with those presented
in (Brencich & de Francesco 2004) since at section
level, the post-peak slope of softening branch can be
associated with an equivalent available ductility; how-
ever, as far as the overall behaviour of the bridge close
to collapse is concerned, the differences in the constitu-
tive assumptions would imply a different redistribution
of the normal force and bending moment through the
arches. It is worth noting that, while in (Brencich & de
Francesco 2004) the hinge mechanism does not takes
place since the arch collapses when the ultimate duc-
tility capacity in the critical section is reached, in the
present approach, thanks to a more accurate descrip-
tion of the behaviour of masonry in compression, the
analysis can follows the progressive flexural strength
decrease until the hinge mechanism forms.
5 CONCLUSIONS
An approach for estimating the load carrying capac-
ity of multi-span masonry arch bridges is proposed
that makes use of a beam model with fibre cross sec-
tion. A numerical investigation of one to three span
masonry arch bridges under travelling concentrated
load, shows that, by parity of constitutive assumptions
(infinite strength in compression, or finite strength
with unlimited ductility) the method provides the same
estimate of the ultimate load given by limit analysis,
as well as a similar kinematics at collapse. How-
ever, while limit analysis method implicitly assumes
a perfectly ductile behaviour, the proposed approach
is able to take into account the effective behaviour of
masonry in compression, including the post-peak soft-
ening branch. Once the limited ductility capacity is
taken into account, a significant decrease in load car-
rying capacity is found, which depends on the slope of
softening branch: a reduction between 15% and 30%
has been estimated for the bridges under study, when
the material behavior turns from perfectly plastic to
post-peak brittle. Now, since masonry cannot be con-
sidered as perfectly ductile in compression, it appears
that the use of fibre beam method provides a signif-
icant improvement in the evaluation of load carrying
capacity of masonry arch bridges.
REFERENCES
Brencich, A., De Francesco, U. 2004. Assessment of multi-
span masonry arch bridges. Part I: a simplified approach.
ASCE Journal of Bridge Engineering, 9(6): 582–590.
546
Brencich, A., Gambarotta, L. 2005. Mechanical response
of solid clay brickwork under eccentric loading. Part I:
Unreinforced masonry, Materials and Structures, 38:
257–266.
Boothby, T.E. 2001. Load rating of masonry arch bridges,
ASCE Journal of Bridge Engineering, 6(2): 79–86.
Boothby, T.E. 1997. Elastic plastic stability of jointed
masonry arches, Engineering Structures, 19(5): 345–351.
Boothby, T.E., Brown, C.B. 1992. Stability of masonry piers
and arches, ASCE Journal of Engineering Mechanics
118(2): 367–383.
Boothby, T.E., Domalik, D., Dalal, V. 1998. Service load
response of masonry arch bridges, ASCE Journal of
Structural Engineering. 124(1): 17–23.
Castigliano, C.A.P. 1879. Theorie de l’equilibre des sys-
temes elastiques et ses applications, Augusto Federico
Negro, Torino, Italy; Andrews E.S., Elastic stress in struc-
tures. Transation by Andrews, Scott Greenwood & Sons,
London, 1919.
Cavicchi, A., Gambarotta, L. 2005. Collapse Analysis of
Masonry bridges taking into account arch-fill interaction,
Engineering Structures, 27: 605–615.
Clemente, P., Occhiuzzi, A., Raithel, A. 1995. Limit
behaviour of stone arch bridges, ASCE Journal of Stru-
cural Engineering, 121: 1045–1050.
de Felice, G. 2006. Experimental investigation on his-
toric brickwork subjected to eccentric axial load, Proc.
SAHC’06 Conference, New Delhi, India.
Fanning, P.J., Boothby, T.E., Ronert, B.J. 2001. Longitudi-
nal and transverse effects in masonry arch assessment,
Construction and Building Materials, 15: 51–60.
Gilbert, M. 2005. Ring, Theory and Modelling Guide, Uni-
versity of Sheffield, Sheffield, UK.
Gilbert, M., Melbourne, C. 1994. Rigid-block analysis
of masonry structures. The Structural Engineer, 72:
356–361.
Heyman, J. 1966. The stone Skeleton. International Journal
of Solids and Structures, 2: 249–279.
Heyman, J. 1982. The masonry arch, Ellis Horwood.
Melbourne, C., Gilbert, M., Wagstaff, M. 1995. The
behaviour of multi-span masonry arch bridges. Proc.
1st International. Conference on Arch Bridges, London,
Thomas Telford.
Molins, C. Roca, P. 1998. Capacity of masonry arches and
spatial structures, ASCE Journal of Structural Engineer-
ing, 124 (6): 653–663.
Page, J. 1987. Load test to collapse on two arch bridges at
Preston, Shropshire and Prestwood, Staffordshire, Depart-
ment of Transport, TRRL Research Report 110, TRL,
Crowthorne, England.
Prentice, D.J., Ponniah, D.A. 1994. Testing of multi-span
model of masonry arch bridges. Proc. Centenary year
bridge conference, Cardiff, Elsevier Science: 169–174.
Robinson, J.I., Ponniah, D.A., Prentice, D.J. 1997. Soil pres-
sure measurements on a multi-span brick arch. Proc. 7th
Int. Conf. On Structural Faults and Repair, Eng. Technics
Press, Edinburgh: 111–119.
Salençon, J. 1985. Calcul à la Rupture et Analyse Limite,
Presses de l’ENPC, Paris, FR.
Spacone, E., Filippou, F.C., Taucer, F. 1996. Fibre beam-
column model for non-linear analysis of R/C frames:
Part I. Formulation, Earthquake Engineering and Struc-
tural Dynamics, 25: 711–725.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Assessment of irregular masonry walls

N. Augenti & A. Romano

Department of Structural Engineering, University of Naples Federico II, Naples, Italy

ABSTRACT: In masonry building, while the behaviour of regular walls (vertically and horizontally aligned
opens) was exhaustively assessed in the past, irregular walls (not aligned openings and geometry varying both
along the horizontal and vertical direction), although representing the greatest part of existing constructions,
are still object of study by many researchers. In this paper, some well-known design methods were used and
compared through applications to masonry irregular walls. Non linear static analyses on four cases-study of
bearing wall (1: regular; 2: horizontally irregular; 3: vertically irregular; 4: horizontally and vertically irregular)
allowed investigating and assessing the influence of irregularity in masonry structures. Finally, the procedure was
applied to a real wall still damaged. The following results are discussed: (i) collapse mechanisms of masonry
panels composing the wall; (ii) initial stiffness, peak load and ductile branch in the base shear versus top
displacement diagram.

1 INTRODUCTION

The greatest part of existing masonry constructions

show irregular bearing walls, that is a misalignment of
the openings in vertical or horizontal direction (Fig. 1).
Although their spread in historical town centres and
more in general in existing buildings, the assessment
of such bearing walls is still object of study by many
researchers, due to the great difficulties in determining
their behaviour.
Augenti (2006) firstly defined different kinds of Figure 1. Irregular walls.
irregularity, classifying a simple irregularity along the
length of the wall, along its height and then their
walls has been studied. The impossibility of hypothe-
combination, that is the most spread one.
sizing homogeneity, isotropy, iso-strength and elastic
The Italian national code in force, masonry section
behaviour for the material, made the adoption of FE
(Norme Tecniche 2005), provides very few instruc-
based methods extremely hard. Therefore, the need
tions regarding irregular walls. A vertical irregularity
to formulate “simplified” analyses and design pro-
is considered assuming that, once a specific analy-
cedures with minor computational efforts and wider
sis lacks, the minimum base dimension of the pier
applicability to real structures arose. In order to
has to be taken as uniform along the whole height
meet such requirements without compromising the
of the wall. This hypothesis implies a wrong distri-
deepening level of the performed analyses, some
butions of the horizontal actions among the bearing
design procedures were formulated (Tomazevic 1982;
elements, and therefore a different assessment of the
Benedetti & Tomazevic 1984). Successively, a consid-
seismic behaviour. The horizontal irregularity is not
erable number of methods, reproducing the behaviour
considered at all.
of walls and masonry structures subject to seismic and
Consequently, the research field of irregular
other kind of loads, developed and improved the POR
masonry walls is wide open and needs to be focused.
method (Tomazevic 1978).
The macro-elements methods model the open
2 MACRO-ELEMENTS METHODS masonry walls as a collection of bi-dimensional pan-
els. They can be divided into three different typologies:
Since the ‘80s, the seismic response of masonry build- (i) Pier panels (included between two openings in
ings through analyses based on macro-modelling of the vertical direction), mainly having the bearing

547
function; (ii) Spandrel panels (included between two
openings in the horizontal direction) bearing the slab
loads; (iii) Cross panels (connection between pier and
spandrel strips) carrying the slab loads and transmit-
ting stresses between pier and spandrel panels.
The most developed and successful methodologies
at the moment are:
1. RAN macro-elements method
(Raithel & Augenti 1984, 1987);
2. SAM equivalent frames method
(Calvi & Magenes 1997);
3. 3MURI equivalent frames method
(Gambarotta & Lagomarsino 1997).
In all these procedures, the following failure mech-
anisms for pier panels are taken into account: com-
posed flexural and axial load, tensile shear and sliding
shear. Two limit restraint conditions are assumed
varying from the fixed-double pendulum typology
(namely Grinter) to the cantilever one (failure of the
spandrel panels). The section reduction (induced by
cracking) allows to evaluate the stiffness degrada-
tion involving the varying of panels stress. Such a
redistribution on the reacting section implies that the
shear-displacement curve shows a bent line within
the non-linear range. Conversely, the observations
of damages produced by earthquakes show internal
cross panels free from collapse, so that no failure
mechanisms are foreseen. Moreover, such methods
allow to take into account the slabs plane stiffness, so
retrofit interventions can be designed to re-establish
the strengthening and functionality of masonry struc-
tures and their effectiveness evaluation, employing dif-
ferent materials as well. Of course, the three methods
show some differences, detailed in the following.
The RAN method allows a linear static seismic
analysis at each storey (taking into account the over-
turning moment induced by horizontal actions) and a
global non linear analysis until the pier panels failure is
reached, considering the infinite strength and stiffness
of spandrel strips (Augenti 2004).
The SAM method uses the non linear static analysis
as the preferential analysis tool and studies the whole
structure, immediately satisfying the equilibrium con-
ditions (Magenes & Della Fontana 1998; Magenes
et al. 2000). Spandrel panels have a reduced strength
with infinite ductility which allows to consider these
elements able to explicate the shear strength until the
collapse of the wall.
The 3MURI method allows to carry out linear and
non-linear analyses (Penna 2001; Cattari et al. 2005).
In this case, the method allows to model masonry
panels through non-linear masonry cantilevers char-
acterized, as the panels, by a reduced strength and a
stiffness degradation within the non-linear range.
Having already tested RAN, SAM and 3MURI
on regular walls (Augenti & Romano 2008), in this
548
1 2 n
1 H1
m Hm
B1 B2 Bn
Figure 2. Pier and Spandrel strips.
b2 b2
h1
h2
h1
h2
b1 b1
(a) (b) (c)
Figure 3. Irregularities in masonry walls: (a) vertical;
(b) horizontal; (c) horizontal and vertical.
paper the three methods have been used to assess the
behaviour of irregular masonry walls.
2.1 Regular walls
A wall is named regular when the openings are ver-
tically and horizontally aligned. Then, “n” pier strips
(with a uniform length) and “m” spandrel strips (with
a uniform height) can be defined (Fig. 2).
2.2 Irregular walls
The wall becomes irregular when the openings are
not aligned anymore and the geometry varies along
both horizontal and vertical directions. Three kinds of
irregularity can be detected:
1. vertical irregular walls (Fig. 3.a): the openings
show one different length in a pier strip at least.
The heights of the pier, cross and spandrel panels
are equal;
2. horizontal irregular walls (Fig. 3.b): the openings
show one different height in a spandrel strip at least.
The base lengths of pier, cross and spandrel panels
are equal;
3. horizontal and vertical walls (Fig. 3.c): the open-
ings show one different height and length in a
spandrel and a pier strip at least. The length and
the height of the pier, cross and spandrel panel are
all different.
3 GEOMETRICAL DISCRETIZATION
One of the main problem in irregular walls, variously
interpreted by researchers, is the panels geometry
individuation.

Figure 4. Crack pattern in irregular walls.
Figure 5. Crack patterns according to the verse of seismic
action.
In regular walls, the panels geometrical character-
istics is simply defined by the intersection of spandrel
and pier strips. For this typology of walls, the height
of the pier panels is independent from the verse of the
seismic action.
For irregular walls it depends on the kind of irregu-
larity. Irregular vertical walls (Fig. 3.a) have a uniform
pier height along the level. In this case, RAN, SAM
and 3MURI agree averaging the dimensions of the
cross and spandrel panels. When the walls are horizon-
tally irregular (Fig. 3.b) the main problem is defining
the pier panels. If SAM and 3MURI individuate the
pier panel height as the heights average constraining
the panel at the sides, the RAN method updates the
geometrical configuration according to the direction
of the horizontal force.
Such principle comes from observing the earth-
quakes damaged buildings, where the cracks in the
pier panels vary and ideally describe the geometry of
the panel (Fig. 4).
It can be stated that horizontal irregular walls show
different behaviours according to the verse of the
seismic action, as illustrated in Figure 5.
Therefore, varying the verse of the horizontal
action, the geometrical model need to be updated
adopting different heights for pier panels, that could be
stocky or slender (Fig. 6).This important consideration
has been implemented in the RAN method.
The last case of irregularity (horizontal and verti-
cal) is obviously the hardest, since the two kinds of
irregularities are coupled (Fig. 3.c).
549
H2d H3s
H1d H3d H1s H2s
(a) (b)
Figure 6. Height of pier panels according to the verse of
seismic action.
Another difference among the three methods, as
regards the geometrical discretization, is the defini-
tion of the external pier panels height. 3MURI and
SAM follow the scheme proposed by Dolce (1991).
Observing the real damaged structures, the external
cross panels result not confined at all, so that some
cracks could start from the opening and extend until
the free extreme, with a crack steeped about 30◦ respect
to the horizontal line.
4 CASE STUDY
The comparison of RAN, SAM and 3MURI has
been performed at first on a regular wall and
then some irregularities have been subsequently
added.
The case-study wall, 10.45 m length, 9.00 m high
and 0.70 m thick, measuring 5.00 m at the first level
and 4.00 m at the second one. The bearing structure is
made of yellow tuff stones and cementicious mortar
with 16 kN/m3 weight density.
Four kinds of structures are detected: (i) regular
wall (Fig. 7.a); (ii) vertical irregular wall (Fig. 7.b);
(iii) horizontal irregular wall (Fig. 7.c); iv) horizontal
and vertical wall (Fig. 7.d).
The following vertical loads were considered:
(i) self weight of the walls; (ii) slab loads
(Q = 8,10 kN/m2 ) comprising self weight and dead
loads (Gk = 6,10 kN/m2 ) and live loads (Qk =
2,00 kN/m2 ).
In Table 1 the values of the elastic and mechanical
characteristics are listed.
Finally, in the determination of the seismic hori-
zontal actions, analyses according to the prescriptions
of the Italian Code O.P.C.M. n. 3431, considering
the building of class 1 (50 years service life) in a
seismic area 3 (ag = 0.15 g) on the ground founda-
tion type D (shear wave velocity < 180 m/s, number
of impacts through Standard Penetration Test < 15,
Tip Strength through Cone Penetration Test < 70 KPa),
were conducted.
 1 2 3 4 5 1 2 3 4 5
1.1 1.2 1.3 1.4 1.5
1.1 1.2 1.3 1.4 1.5
1 1
2.1 2.3 2.5
2.1 2.3 2.5

2 2

3.1 3.2 3.3 3.4 3.5 3.1 3.2 3.3 3.4 3.5

3 3
4.1 4.3 4.5 4.1 4.3 4.5

4 4

(a) (b) Figure 8. Characteristic curves of pier panel 2.1: (a) positive
1 2 3 4 5 1 2 3 4 5 seismic verse; (b) negative seismic verse.
1.1 1.2 1.3 1.4 1.5 1.1 1.2 1.3 1.4 1.5
1 1
2.1 2.3 2.5 2.1 2.3 2.5

2 2
3.4 3.5 3.4 3.5

3.1 3.2 3.3 3.1 3.2 3.3

3 3
4.1 4.3 4.5 4.1 4.3 4.5

4 4

(c) (d)

Figure 7. Masonry walls: (a) regular; (b) vertically irregu-

lar; (c) horizontally irregular; (d) horizontally and vertically Figure 9. Characteristic curves of pier panel 2.3: (a) positive
irregular. seismic verse; (b) negative seismic verse.

Table 1. Elastic and mechanical characteristics.

Compressive characteristic strength fk
Pure shear characteristic strength fvko
Compressive design strength fd
Pure shear design strength fvdo
Normal elasticity module E
Tangential elasticity module G
4.1 Application of the RAN method
A linear static analysis was preliminarily performed.
The strength domains in the plane compression – shear
(N, V), normalized to the ultimate normal load (Nu ),
were defined for each pier panel of the wall. Elastic
limit state for composed flexural and axial load (Ve ),
non proportional elastic limit state for eccentrical axial
load (Vl ), plastic limit state for composed flexural and
axial load (Vp ), tensile shear limit state (Vt ), and slid-
ing shear limit state (Vaapp and Vaeff ) were considered.
In correspondence of the normal load applied on the
panel it is possible to read the value of the shear force.
Checking the panels should not stop at the equilibrium
condition, for panels belonging to the same level show
the same displacements under congruence conditions.
Therefore, characteristic curves for each panel, and
then for each storey, relating the shear force V applied
at the top and the dual displacement δ, were deter-
mined. Generally, the curves consist of three branches:
(i) a first straight line, in the proportional elastic field,
where sections are fully working; (ii) a second curved
line, in the non proportional elastic field, where react-
Value ing sections are reduced; (iii) a third non linear branch,
MPa in the plastic field, related to the ductility of the panel.
These curves of the whole level allow to assess the
3.0 maximum bearable shear force and stop at the dis-
0.3 placement in which the weakest panel, belonging to
1.0 each floor, is at failure. Finally, the equilibrium of the
0.1
3000
spandrel strips has been checked, with a tolerance of
1200 5% of the maximum acting stress.
In the following, the comparison of the pier walls of
level 2 in the four typologies under study is reported.
4.1.1 Pier panel 2.1
The panel reaches the failure condition for composed
flexural and axial load in both verses of the seismic
action for almost all kinds of irregularity (Fig. 8). Only
the horizontal irregular wall (and consequently the hor-
izontal and vertical wall) shows a smaller shear value
due to the different distribution of the seismic weight.
4.1.2 Pier panel 2.3
In this panel the influence of the seismic verse can
be easily detected. In the regular and vertical irregu-
lar wall, under positive horizontal actions (from left
to right) the eccentrical normal load determines the
failure. Conversely, the introduction of the horizontal
irregularity decreases the height of the pier panel from
2.8 m to 1.8 m, making it stocky and causing a shear
failure (Fig. 9.a). As expected, when the considered
seismic force has a negative verse (from right to left),
the height of the panel stay unchanged and the col-
lapse occurs in all the kinds of walls because of the
composed flexural and axial load (Fig. 9.b).

550
 (a) (b)

Figure 10. Characteristic curves of pier panel 2.5: (a) posi-

tive seismic verse; (b) negative seismic verse.

(c) (d)

Figure 12. SAM failure mechanisms: (a) regular; (b) irreg-

ular height; (c) irregular length; (d) irregular length and
height.

Figure 11. SAM model walls: (a) regular; (b) irregular

height; (c) irregular length; (d) irregular length and height.

4.1.3 Pier panel 2.5

Finally, in this panel as for the first one, the failure
occurs for eccentric axial load (Fig. 10) but the reduced
height of the panel, due to the horizontal irregularity,
influences the value of the bearable shear.

Figure 13. 3MURI model walls: (a) regular; (b) irregular

4.2 Application of the SAM method
For this numerical application the same geometri-
cal, mechanical and load models were implemented
in an available computer code, called Andilwall
(http://www.crsoft.it/andilwall/andilwall.aspx).
In Figure 11, the model views of the four walls are
given.
The failure mechanisms of the pier panels, perform-
ing a non linear static analysis, are shown in Figure 12.
The dotted lines indicate the eccentrical axial load fail-
ure and the line-point line the shear failure. Observing
level 2, it is evident that the mechanisms for panels
2.1 and 2.5 coincide with the RAN method, whilst no
failure conditions are detected for the central pier 2.3
in all the four study-cases.
height; (c) irregular length; (d) irregular length and height.
4.3 Application of the 3MURI method
The 3MURI computer code was then used (http://
www.stadata.com). A 2D view of the four walls
(Fig. 13) and a selection of the failure mechanisms
of the walls are shown (Fig. 14).
The colour green (vertically striped) means that the
panel is undamaged, the pink one (diagonally striped)
that the plastic limit state for bending and axial load
has been reached, the red one (horizontally striped)
that the collapse occurred for composed flexural and
axial load. The cross panels, in light blue (plane) are
not affected by failure mechanisms.

551

Figure 14. 3MURI failure mechanisms: (a) regular;
(b) irregular height; (c) irregular length; (d) irregular length
and height.
5 COMPARISON OF RAN, SAM AND 3MURI
Once the three analyses were completed, a comparison
of methodologies has been carried out and discussed
as follows.
Obviously, the first difference among procedures
is the geometric modelling of the panels. Extensively
reported in §2, it is opinion of the authors that a pair-
ing of the three methodologies is definitely necessary.
RAN method rightly considers the updating of the geo-
metric configuration according to the seismic verse,
whilst SAM and 3MURI rightly evaluate the external
cross panels dimension.
Regarding mechanical characteristics, the parame-
ters required by 3MURI are practically the same of
RAN whilst SAM, being mainly targeted to artifi-
cial units, requires a more refined calibration in the
perpendicular and parallel directions of load.
Another difference of methods is represented by
the collapse mechanisms of pier and spandrel panels.
The 3MURI and SAM methods did not detect shear
collapse mechanisms in any of the panels, but only
mechanisms of flexural-axial load failure or uncracked
panels. Conversely, in the application of the RAN
method, shear collapse mechanisms influenced the
panel pier 2.3 for horizontal irregular wall. The pres-
ence of this kind of collapse mechanism revealed by
the RAN method for stocky panel appears more consis-
tent with the geometry and load condition of the panel.
It has to be admitted that a general disagreement about
the failure modes is detected.
Moreover, it can be observed that, whilst SAM
and 3MURI, being implemented in computer codes,
request the use of a “close” calculus, the RAN method
552
Figure 15. Capacity curves of walls: (a) regular; (b) irreg-
ular height; (c) irregular length; (d) irregular length and
height.
allows the application through the use of spread-
sheets. This implies a simpler use of the methodology
and the possibility of adapting it to specific needs,
although it is more time consuming.
Finally, a quantitative comparison of the three
methodologies has been performed. Some non-linear
static analyses of the whole structure were carried out.
For regular walls (Fig. 15.a) the three curves show
nearly the same value of shear force at the base. The
RAN curve is stiffer than the 3MURI and SAM one.
The 3MURI ductile branch is longer than SAM’s. Once
the vertical irregularity is introduced, RAN method
provides an higher shear force and stiffness value than
3MURI and SAM. These latter show the same initial
stiffness but can not be compared in terms of peak load
and displacement (Fig. 15.b). The same considerations
on the regular wall can be made for the horizontal irreg-
ular wall (Fig. 15.c). The last irregular wall, which
considers both the horizontal and vertical irregularity,
shows the same behaviour of the vertical irregular wall
(Fig. 15.d).
In all these cases the RAN method does not provide
the same ductility value as the other two methods, since
the non linear analysis was conducted under force but
not displacement control (as SAM and 3MURI).
6 ANALYSIS OF A REAL DAMAGED WALL
Having examined the four kinds of irregularity, a real
wall in a building placed in Avellino County and hit by
the 1980 earthquake (Magnitude 6.9) was analyzed.
Since no retrofit interventions were made after the
seismic event, it was possible to carry out a detailed
geometrical, mechanical and load survey. In Figure 16
the irregular wall and the crack pattern are shown.
 Figure 17. Panel discretization: (a1) RAN method for pos-
itive forces; (a2) RAN method for negative forces; (b) SAM
method; (c) 3MURI method.

Figure 16. Real damaged wall.

Table 2. Loads on the roof structure.

Value
kN/m3

Attic slab dead load 2.0

Roof slab dead load 2.6
Attic slab live load 1.0

Figure 18. Failure panel mechanisms: (a1,a2) RAN method,

This one storey building is made by yellow tuff
masonry 30 cm thick and cementicious mortar. The
plan dimensions are 7.60 × 7.50 m2 for an height of
3.0 m. The foundation system consists of the same
walls deepening into the ground, and the roof struc-
ture consists of two slabs: an external gable roof and
an internal attic, both in reinforced concrete and holed
bricks. A weight density of 16 kN/m3 was assumed for
masonry and the following loads were deduced at the
moment of the quake (Table 2). The soil was type C
(according to O.P.C.M. n. 3431).
The panel discretization, according to the three
methods, is shown in Figure 17.Apparently, RAN fore-
sees two geometrical models for the two verses of the
horizontal action; 3MURI and SAM models give the
same model independently from the verse.
In Figure 18 the failure mechanisms of the panels
are shown.
The pier panel on the left fails for eccentrical load
according to the three methods. In fact, as illustrated in
Figure 16, horizontal cracks appeared at the superior
and inferior extremities.
The panel in the middle fails for tensile shear stress,
as also performed by the analytical estimation.The pier
shows two diagonal cracks starting from plasticized
zones. The shape of the cracks suggests that the first
wave came from the right and the second one from the
left, since this second crack stops at the intersection
with the first one.
(b) SAM method; (c) 3MURI method.
Figure 19. Capacity curves: comparison (a) Positive hori-
zontal actions; (b) Negative horizontal actions.
Finally, combined flexural and axial load provoked
the crisis of the last panel on the right, immediately
followed by shear for horizontal actions from right to
left.
Therefore, it can be stated that the methods correctly
evaluate the crisis typologies.
Then the capacity curves obtained from a non lin-
ear static analysis, according to the two verses of the
seismic actions, were compared (Fig. 19).
From these capacity curves it can be noted that RAN
and 3MURI provide almost similar stiffness and force
values only in the case of verse action from right to left
(when the central pier panel is considered in its whole
height, as really happened). Conversely, SAM method
turns from the previous ones due to the impossibility
of modelling the irregularity in a one storey building.

553
7 CONCLUSIONS
This paper deeply investigates the behaviour of
masonry irregular walls through some well-known
analysis methods. The hypothesis at the base of each,
the procedure adopted, their criteria and applica-
tive capabilities have been listed. Modelling through
macro-elements allows a remarkable reduction of the
degrees of freedom in respect to Finite Elements. For
the applicative simplicity and the plurality of offered
possibilities, macro-element methodologies are cer-
tainly a valid solution for the practitioner analysing
a simple masonry residential building.
Different kinds of simple irregularities in masonry
walls were individuated and analyzed starting from
a regular wall and then introducing the irregularities
increasing the grade of complexity. Great attention
has been paid to the geometrical discretization of
panels, according to damages observed after earth-
quakes. Then, in order to compare the numerical
analyses performed with reality, a wall damaged by
a severe earthquake in the 80’s has been considered.
Non linear analyses were performed on both the four
case-studies and the real wall: the collapse mecha-
nisms of the panels and the capacity curves have been
plotted.
Unfortunately, due to the sensitiveness of the
response depending on the used method, irregular
walls are still characterized by a great uncertainty as
regards a correct modelling.
Further researches are necessary and it is opinion
of the authors that a pairing of the procedures relying
on RAN, SAM and 3MURI methods could be a good
start.
ACKNOWLEDGMENTS
This research has been supported by ReLUIS “Rete
di Laboratori Universitari Ingegneria Sismica” in the
context of the activities of “Linea 1 – Valutazione e
Riduzione della Vulnerabilita’ di Edifici in Muratura”.
REFERENCES
Augenti, N. 2004. Il calcolo sismico degli edifici in muratura,
Torino: Utet.
Augenti, N. 2006. Seismic behaviour of irregular masonry
walls. First European Conference on Earthquake Engi-
neering and Seismology; Proc. intern. conf., Geneva, 3–8
September 2006.
554
Augenti, N. & Romano, A. 2008. Seismic design of masonry
buildings through macro-elements. XIV International
Brick and Block Masonry Conference; Proc. intern. conf.,
Sydney, 17–20 February 2008.
Benedetti, D. & Tomazevic, M. 1984. Sulla verifica sismica
di costruzioni in muratura. Ingegneria Sismica I: 9–16.
Cattari, S., Curti, E., Galasco,A. & Resemini, S. 2005.Analisi
sismica lineare e non lineare degli edfici in muratura.
Napoli: Esselibri – Simone.
Dolce, M. 1991. Schematizzazione e modellazione degli edi-
fici in muratura soggetti ad azioni sismiche. L’industria
delle Costruzioni 242: 44–57.
Gambarotta, L. & Lagomarsino, S. 1997. Damage models for
the seismic response of brick masonry shear walls. Part I:
the mortar joint model and its applications. Earthquake
Engineering and Structural Dynamics 26: 423–439.
Gambarotta, L. & Lagomarsino, S. 1997. Damage models for
the seismic response of brick masonry shear walls. Part II:
the continuum model and its applications. Earthquake
Engineering and Structural Dynamics 26: 441–462.
Magenes, G. & Calvi, G.M. 1997. In-plane seismic response
of brick masonry walls. Earthquake Engineering and
Structural Dynamics 26: 1091–1112.
Magenes, G. & Della Fontana, A. 1998. Simplified Non-
linear Seismic Analysis of Masonry Buildings. Proc. of
the British Masonry Society 8: 190–195.
Magenes, G., Bolognini, D., & Braggio, C. 2000. Metodi
semplificati per l’analisi sismica non lineare di edifici in
muratura. Roma: CNR – Gruppo Nazionale per la Difesa
dai Terremoti.
Penna, A. 2001. Una procedura a macroelementi per l’analisi
dinamica non lineare di edifici in muratura. PhD The-
sis, Politecnico di Milano, Dipartimento di Ingegneria
Strutturale, Milano.
Presidenza del Consiglio dei Ministri, O.P.C.M. n. 3431
del 3 Maggio 2005, Ulteriori modifiche ed integrazioni
all’O.P.C.M. n. 3274 del 20 marzo 2003, Roma.
Raithel, A. & Augenti, N. 1984. La verifica dei pannelli
murari. Second Conference ASS.I.R.C.O.; Proc. nat. conf.,
Ferrara, 1984.
Raithel, A. & Augenti, N. 1984. La deformazione dei pannelli
murari. Second Conference ASS.I.R.C.O.; Proc. nat. conf.,
Ferrara, 1984.
Raithel, A. & Augenti, N. 1984. Il pannello murario vinco-
lato. Second Conference ASS.I.R.C.O.; Proc. nat. conf.,
Ferrara, 1984.
Raithel, A. & Augenti, N. 1984. Il legame T-D nei pannelli
murari. Third Conference Seismic Engineering in Italy;
Proc. nat. conf., Roma, 1987.
Norme Tecniche per le costruzioni del 14. settembre 2005,
Roma.
Tomazevic, M. 1978. The computer program POR. Report
ZRMK, Institute for testing and research in materials and
structures, Ljubljana, (in sloveno).
Tomazevic, M. 1982. Lateral load distribution in masonry
buildings. VI International Brick and Block Masonry
Conference; Proc. intern. conf., Rome, 16–19 may 1982.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Geometrical and static aspects of the Cupola of Santa Maria del Fiore,
Florence (Italy)

A. Cecchi & I. Chiaverini

Department of Civil Engineering, University of Florence, Florence, Italy

A. Passerini
Leonardo Società di Ingegneria S.r.l., Florence, Italy

ABSTRACT: The purpose of this research is to clarify, in the language of differential geometry, the geometry
of the internal surface of Brunelleschi’s dome, in the Cathedral of Santa Maria del Fiore, Florence; the statics
of a Brunelleschi-like dome have also been taken into consideration. The masonry, and, in particular, the “lisca
pesce” one, together with the construction and layout technologies, have been main topics of interest for many
researchers: they will be the subjects of further research.

1 INTRODUCTION: THE DOME OF THE

CATHEDRAL OF FLORENCE

The construction of the cathedral of Firenze was begun

in the year 1296, with the works related to the exten-
sion of the ancient church of Santa Reparata: it was
designed by Arnolfo (1240,1302). The design included
a great dome, based on an octagonal base, to be erected
in the eastern end of the church.The dome is an unusual
construction of the Middle Ages, (Wittkower 1962):
Arnolfo certainly referred to the nearby octagonal bap-
tistery San Giovanni, so ancient and revered that the
Florentines believed it was built by the Romans, as
the temple of Ares, hypothesis which was not con-
firmed by excavations, that set the date of foundation Figure 1. View of the cathedral of Santa Maria del Fiore.
is between the V and IX century (Rocchi 1996).
In the second half of the fourteenth century the con- dimension was internally about 45 m, surpassing the
struction of the octagonal base was completed; for fifty greatest known, the Pantheon in Rome (43 m about,
year about the construction yard stood by, testifying the in concrete) and Hagia Sofia in Constantinople (31 m
great uncertainty about the building technique of the about, in masonry); the base itself was laying on four
dome, till Brunelleschi’s assignment in 1420. great high piles, so that the height of the top, 90 m
We should keep in mind that in that time in Italy about, and the height of the base, 60 m. about, greater
there were two other great yards open for the construc- than the Pantheon, made it practically impossible to
tion of the cathedrals of Milan and Bologna, but the erect the dome by framework, as was done in Rome.
construction of the dome of Florence was so excep- Besides, the Roman dome is a spherical revolution
tional to enable the target to appear beyond the human surface, while the Florentine dome has a much more
possibilities. complicated geometrical shape, due to its octagonal
This is understandable since the dome to be built base. Notice that the Pantheon was built in concrete, a
would be the largest ever known (Figure 1,2): its base technique which was probably lost in the Middle Ages

555

Figure 2. Plan of the cathedral of S. Maria del Fiore
(Ximenes).
and was substituted in the dome of Florence by brick
masonry. The static function of the flying buttress of
the Middle Ages cathedrals, to carry the horizontal
forces to the foundations, (Heyman 1966) was left to
the chapels, which surround the base, much more suit-
able to the classical tradition of the town, Florentia,
founded by the Romans in 59 B.C. (Davidsohn
1956).
Vasari (Vasari 1550) is the main source of infor-
mation concerning Filippo Brunelleschi’s work. He
wrote down an act, collected in the Museum of
the Cathedral, where the main dimensions of the
building were specified. We should remember the
following:
– the dome is composed by two cupolas;
– the inferior one has a variable thickness from 2.35 m
in the bottom to 1.49 m at the top and it is vaulted
“a quinto acuto negli angoli”. The function of the
superior one is: “conservalla dal umido (practical
function) e perché torni più magnifica e gonfi-
ante (esthetical function)”: its thickness varies from
0.72 m to 0.43 m;
– 24 stone ribs (pietraforte) link the two cupolas, 1 for
each of the 8 corners and 2 for each web; these ribs
are tied round by 6 hoops in pietraforte cramped by
means of iron brackets;
– further links between the two cupolas are the
“volticciole” (small vaults) chained by means of
oak beams;
– the material used is brick masonry, even if at first,
stone walls were proposed.
You can note Brunelleschi’s structural intuition,
that is to say the employment of a sandwich ribbed
structure, in order to lower the load of heavy vaults.
Nothing he wrote neither about the methods of erec-
tion of the vault, even if he used a cantilever technique
without frameworks, absolutely new for his times, nor
about the masonry “secondo sara allora consigliato
556
perchè nel murare la praticha insegnera quello ches-
sara a seguire” (Brunelleschi’s specifications), nor
about the mechanical apparatus he would have used
later to raise the heavy weights “tirare i pesi per via
di contrappesi e ruote, che un sol bue tirava quanto
avrebbero appena tirato sei paia” (Vasari 1550).
2 AUTHORS WHO DOCUMENTED THE
DOME. THE AIMS OF OUR RESEARCH
The silence of Filippo is the main cause of the discus-
sions that followed on the argument. This seems to be
common to all the Middle Ages constructions. Fitchen
speaks of “. . . the total lack of written documentation
on both the engineering structure and the erectional
procedures. . .” (Fitchen 1961).
Many authors have related the dome and his mas-
ter: among many others let’s remember Giovanni di
Gherardo da Prato (1421), Manetti (1480?), Vasari
(1550), Opera del Duomo (1691) (The Opera del
Duomo was charged by the Duke of Tuscany to write
down an essay on the stability of the Dome, threatened
by cracks), Guasti (1887), and recently Sanpaolesi
(1962), Fondelli (2004), Rocchi (1994). Sgrilli and
Ximenes have rendered the dome accurately in the
XVIII century.
Nevertheless as a result of this information neither
thorough study on the geometry of the dome, nor on
its statics is carried out.
Beyond these arguments, the topics which inter-
est many researchers, e.g. Di Pasquale (1977), Bartoli
(1994), are the methods of construction and erection of
the dome and of the masonry, particularly the special
one called “a spina pesce”.
The survey of Fondelli, based upon photogramme-
try, must be remembered for its accuracy. It started in
1968 and was carried on for many years. Fondelli’s tar-
get was to render both the outside and the inside of the
great cathedral and at the same time to link the survey
to the Italian geodetic net.
The task of this paper is to give a contribution to
the knowledge of Filippo’s work on the following four
topics:
– based on Brunelleschi’s specifications, the above
papers and our observations the geometrical shape
of the internal surface of the dome is proposed;
– the proposed shape is compared to the results of the
survey of the dome;
– the differential geometry of the surface is exposed;
– the statics of a Brunelleschi-like dome is defined
and analyzed.
Further researches will discuss the masonry, the
engineering structure, the details and the erectional
procedures.

Figure 3. Geometrical hypothesis on the dome.
3 HYPOTHESIS ON THE GEOMETRICAL
FEATURES OF THE INTERNAL SURFACE
OF THE DOME
Figure 3 shows the geometrical hypothesis, plan and
vertical section (AAIVVFIV F).
The plan of the octagon A, B, D,. . ., basis of the
dome, can be observed. C is its centre. The lines
AA,B B,D D,. . ., are the projections of eight cir-
cular arches whose ray is r: they are the rulings of the
surfaces. The octagon, represented by A ,B ,D ,. . .,
is the lantern: in section it is represented by the line
AIV FIV . According to Brunelleschi’s specifications the
arches’ centres aren’t in the centre of the octagon: they
are in the point “quinto di sesto”. E.g. AA and F F
are the projections of the circular arches AAIV , FFIV ,
whose centres are respectively the points A , F such
that, for instance, A F = 1/5 AF. Naturally eight cen-
tres form a new octagon A ,B , D ,. . .,. The points
of the webs of the dome represented in the horizontal
plan by the eight quadrilaterals AA B B, BB D D,
DD E E,. . ., belong to eight cylinders, each having
for rulings two circles: AA , BB ; BB , DD ; DD ,
EE ;. . ., and straight horizontal lines as generatrices
inclined of β = 67, 5◦ (Figure 3).
The main properties of these cylinders are the
following (Figure 3):
– a horizontal plane cuts the dome with an octagon,
e.g. A B D ,. . . .,;
– the cylindrical web. AA I I has its axis in the
A I segment. This property is common to the other
webs.
557
Figure 4. Surveyed points of the western web of the dome.
4 SURVEY OF THE DOME. COMPARISON
BETWEEN SURVEYED AND
GEOMETRICAL POINTS
The second aim of our work was to survey one web of
the dome so that the hypothesis explained in Chapter
3 can be compared.
The survey was carried out only with topographical
methods, using an electronic laser total station, a Leica
TCR 705, 5” angular precision and 2 mm.+2 ppm. lin-
ear precision and a digital camera Nikon D300 with a
12 mega pixel CCD and 50/20 interchangeable Nikkor
optics.
It is well known (Cecchi 2006) that the laser tech-
nology permits the direct survey of inaccessible points,
which is the case of this dome.
Beyond this the quick acquisition of data in a dig-
ital format in a reference frame, permits a successive
computer elaboration of the data.
In June 2007 the upper balcony of the cathedral
was reached; this one was built on the bottom of the
dome from which the best and nearest view of the
inside surface of the dome is possible. The topograph-
ical instrument was placed on the middle east side and
points of the west web were surveyed: their number is
about 700 and includes points of the bordering vaults
too . The numerical format of the surveyed points is
DXF so that AUTOCAD® immediately can place them
in a three dimensional Euclidean space (Figure 4).
This figure clearly shows how the points were sur-
veyed with zenithal constant angles: notice that the last
series refers to points along the upper lantern sides.
Then the digital coordinates can be elaborated with
the least squares method (Kraus 1998).
 Figure 6. The estimated straight line at 64◦ degree on the
Figure 5. The estimated straight line at 64◦ degree on the vertical plane projection.
horizontal plane projection. The values in the figure represent
the difference between the calculated points and the surveyed
ones.

4.1 Research of the curves connecting points

defined by constant zenithal angles
The constant zenith points were projected on the hor-
izontal plane and the line which best approximates
these points was researched.
Then with the least square method the approximat-
ing lines can be found.
Figure 4 shows these constant zenith points are not
straight lines but they differ a little from it especially
in the middle area of the web. These deviations have
been computed through projections of these points on
the horizontal and vertical planes.
The results have been shown (Figures 5,6), for the
sake of brevity, only for the angle 64◦ , where the Figure 7. The estimated round circle.
greatest deviations are noticed.
Out of Figure 5 the maximum deviation in the hori- in Figure 4, but by now it can be possible to formulate
zontal plan projection is 0.083 m, while in the vertical some hypothesis.
plane projection is 0.089 m (Figure 6). As the length of The regularity of these deviations and their increase
the chord is 10.666 m, the deviation is at about 0.8%. in correspondence of the middle web allow us to sur-
So it can be observed that this deviation is very modest mise that Filippo created it for precise reasons, perhaps
and it can be disregarded in the global geometry of the these regarding statics.
dome. A second hypothesis results from the assumption
The following step was to consider the intersection that these deviations occurred during and after the con-
of two such neighboring lines determines a point of struction of the dome because of the movements which
the curve intersection of two border web (Figure 4). were due to cracks (Opera del Duomo 1691, Fondelli
The next one was to discover the surface on which 2004), elastic deformations, thermal expansion.
these points are placed. This surface is a plane and
then, by projecting these points on this plane, the
curve which best approximates these projections was 4.2 Further research on the dimensions of the dome
studied: these curves are circles (Figure 7). The two
The following angles were considered:
surveyed arches radii are respectively 36,11 m and
36,23 m. – the angle made by the planes containing the two
The authors leave to the proposed future research of circles. This angle is 44,92◦ . Notice the analogous
Chapter 2 a study on the surveyed differences pictured geometrical angle of the octagon is 45◦ .

558
 Table 1. Surveyed dimensions. Notice: curves φ=cost are parallel horizontal lines,
whose projections on i1 , i2 are inclined β with respect
Dimensions m. to line AC; curves v=cost are translations of the ruling
circle in direction v.
AB 17,26 Deriving the vector x with respect to each Gaussian
AA 36,11
coordinate (φ , v):
BB 36,23
AF 44,94
BG 44,96
AA /AF 0,803
BB /BG 0,805
Zenithal angle AIV (lantern) 36.21◦
Through the cross vector product the unit vector
normal to the cylinder is readily obtained:

– the angles made by the projections of these two

planes with the side of the surveyed octagon. These
angles are respectively 67,30◦ and 67,79◦ . Notice:
for the geometrical octagon this angle is 67,5◦ .
From the survey more dimensions have been
retrieved and for clarity’s sake there can be identi-
fied through the analogous geometrical dimensions of and the Gaussian quantities E,G,F, that define the First
Figure 3: Fundamental Form, can be calculated:
It can be observed that 0,80 is the geometrical
coefficient for the arch “quinto di sesto”.

4.3 Conclusions of the comparison

The surveyed dimensions and, especially, the dimen-
sionless measures confirm the geometrical hypothesis
of Figure 3 and, besides, show how carefully the dome To define the Second Fundamental Form let us form
was laid out. the derivatives:
Chapter 4.1 shows that lines with constant zenithal
angle approximate well straight lines: so that the
surface is a ruled one and the ruling curves are circles.
In other words the internal surface of the dome con-
sists of eight cylindrical surfaces; two of them intersect
each other in a circular cross section. Each circle is the
ruling of the two bordering cylinders (Figure 3).
so that the quantities:

5 DIFFERENTIAL GEOMETRY OF THE

INTERNAL SURFACE OF THE DOME

Now the third goal of our research was to represent the

dome with the differential geometry methods, e.g. Do
Carmo (1976), Sokolnikoff (1951).
With reference to the Cartesian frame (O,i1 , i2 , i3 )
Figure 3, let x be a point of the cylinder AA I I, r
the radius of the circle, φ the zenithal angle, v the unit
vector of direction IA, v a scalar. Then:
define the Second Fundamental Form.
The Gaussian quantities: LN − M2 = 0 and L2 + M2
+ N2  = 0. The points of the surface are then parabolic.
It is well known that the main curvatures κ are
obtained from the solution of the following equation:

and the unit vector v:

559
Then:
κ1 and κ2 represent the main curvatures at any point
of the surface. The first one is null: it refers to the
straight lines of the cylinder, φ = cost. The other one
depends only on φ, as the lines v = cost are translations
in the direction v. This curvature is the one of the nor-
mal section ellipse, normal to the vector v. The lines
of curvatures, tangent to the principal directions, are
then the lines v=cost and the normal section ellipses,
normal to the unit vector v.
In order to find the equations of these ellipses let us
consider a plane through C parallel to the unit vectors
s and i3 (Figure 3):
With the position CA = a, its equation is obtained
(notice that s and k are parameters):
Let us consider the intersection of y with x:
Let us eliminate three of the four parameters; for
instance resolving for k, there can be obtained:
and the equation of the intersecting curve in the
reference (O,i1 , i2 , i3 ):
If the reference system is changed (C, i1 , i2 , i3 ) and
rotated in (C, s, v, i3 ):
560
Figure 8. The estimated ellipse. In blue the surveyed points,
in red the geometrical ones.
which is the equation of the ellipse, line of curvature.
For instance for k = 0, the point R is obtained, for k = 1
the point T (Figure 3).
In order to show an application of the equations let
compare points of (32) with the surveyed points of
Chapter 4.
At first let us fold the vertical plane of trace TR to
T R (Figure 3) so that the ellipse con be drawn as the
dotted line RAIVVS.
The surveyed ellipse has its minor axis 32,26 m, and
its major is 36,12 (Figure 8). Using of the letters of
Figure 3, RT = 32,26 m., T S = 36,12 m. Out of (32)
the geometrical sizes are respectively RT = 32,44 m.,
T S = 36,11.
It is evident that the difference between geometric
and surveyed dimensions is only 0.18 m, that means
the error is less than 0.01.
6 STATICS OF A BRUNELLESCHI-LIKE DOME
Let us define a Brunelleschi-like dome as an octagonal
dome, formed by ribbed cylindrical webs, intersecting
along circles, inclined 67,5◦ with respect to the cylin-
der axis. The surface of this dome is represented by
equation (32).
Let us consider the statics of this dome. Timoshenko
(1959) and Heyman (1977) consider the statics of
cylindrical shells according to the membrane theory.
According to Filippo’s specifications, Chapter 1,
this assumption could be hazardous for a practical
application: this is only one way of understanding the
main static aspects of this dome. In the FEM analysis
the correct thickness has been used.
With reference to Figure 9 and to the well known
symbols of shell structures, the equilibrium equations
are the following:
Figure 9. Equilibrium of a cylinder surface with the lines
of curvature. q is the vertical force for unit area.
Figure 10. Equilibrium along the web ribs.

Apparently these equations contain three unknowns:

the problem is then statically determinate. They are
easily integrated, if r = cost, that is to say a circle, under
two arbitrary functions, F1 (ϕ) and F2 (ϕ) to be deter-
mined from the conditions at the edges. You must also
consider the Heyman solution for a cylindrical surface
with two lateral frames: in this case he shows how the
edges of the shell are not stress free and the shear stress
Nφv is not null. So two straight edge beams in tension
are necessary for equilibrium.
In a Brunelleschi-like dome the lines v=cost are the
ellipses (32), even if they approximate circles.
You can note also that this dome has eight sym-
metry axes: axes AF, BG, DH,. . ., and axes RC, UC,
VC,. . . For example along RC it can be assumed
that Nvφ = 0, so that according to Heyman, F1 (ϕ) = 0,
while along AA , because of the rib, the tangential
stresses N//AA , although symmetric, are not null, as
described in Figure 10.
Figure 11 shows the slice AA I I cut off the dome
and the forces N//AA and N⊥AA which equilibrate
its weight. Side A I is free of forces, but after the
construction of the lantern in the real dome, weights
were applied along it.
The equilibrium of momentum with reference to
AI cannot determine the two unknowns N//AA and
N⊥AA : N⊥AA is not parallel to the line AI, so that its
contribution to momentum is different from zero.
The rib assumes and equilibrate the forces 2N//AA ,
according to the Figure 10, while the forces N⊥AA
equilibrate themselves.
Besides, the existence of eight symmetry axes
makes the structural behaviour of the dome not far
Figure 11. Slice AA I I with unit membrane forces along
the ribs.
from that of a surface of revolution. For instance
Timoshenko and Heyman present the solution for
semi-spherical domes, in the membrane theory, also
with the upper portion removed, in which, in each
point, a line of curvature is the meridian, while the
other one is obtained with a plane normal the meridian,
containing the normal m to the surface.
The elastic solution with the upper part removed
confirms the intuition of Brunelleschi on the possibil-
ity of the cantilevered erection of the dome without
frameworks.
Writing down the two equations of equilibrium,
not identically satisfied, the problem is statically
determinate: in fact, for symmetry, shears are null.
In particular from the solution, it is well known that
the forces Nv along the meridians are positive for a
latitude ϕ > 51◦ 50 .
The Brunelleschi-like dome has been modelled with
a FEM analysis with ANSYS® .

561
Figure 12. The Brunelleschi-like meshed dome.
Figure 13. The diagram of the 1ST principal stress in Pa.
Linear solution.
The virtual 3D model of the dome has been meshed
with the solid element Solid65 (Figure 12). A masonry
density of 1800 kg/m3 was assumed.
The case without the weight of the lantern has been
primarly considered, as it was during the construction.
Figure 13 shows the linear solution of the dome. In
particular it shows that part of the dome has positive
stresses, particularly in the lower part of the ribs and of
the webs, coherently with the Heyman’s both solutions
of the cylindrical shell and the spherical dome.
These positive stresses are lower than 1,5 × 105 N/
m2 : nevertheless this value is excessive for a Middle
Ages masonry. The intuition of Filippo is really sur-
prising. He used the right expedients for the engi-
neering possibilities of his time, as the “lisca pesce”
technique and reinforcements in strips of wood and
stone (Brunelleschi’s specifications).This phase, with-
out the lantern, was run across the construction; so
cracking appeared in the cupola (Opera del Duomo
1691, Fondelli 2004).
Figure 14 shows the non linear solution obtained
with a crush/cracking FEM analysis, with the value of
cracking near zero.
Confronting the Figure 13 and Figure 14 it can be
noticed how the equilibrium is possible with minor
562
Figure 14. The diagram of the 1ST principal stress in Pa.
Nonlinear solution.
Figure 15. The diagram of the 1ST principal stress in Pa with
the lantern. Linear solution.
positive stresses, but with a cracking especially diffuse
on the lower part of the dome.
Now the effect of the construction of the lantern
(5.000 kN about) on the dome can be shown.
Out of Figure 15, in the linear solution, the posi-
tive effect of the weight of the lantern on the dome
is significant: it strongly reduces the positive stresses,
especially in the lower areas of the dome.
Figure 16 represents the non linear solution , with
the value of cracking near zero, probably the actual
state of the dome.
So an other intuition of Brunelleschi can be noted:
the heavy lantern placed on the top of dome decreases
positive stresses and then increases the masonry sta-
bility.
7 CONCLUSIONS
The geometrical shape pictured in Figure 3 is con-
gruent in all its parts; the surveyed dimensions and,
especially, the dimensionless measures confirm the
geometrical hypothesis and besides show how care-
fully the dome was laid out by its builders.

Figure 16. The diagram of the 1ST principal stress in Pa with
the lantern. Nonlinear solution
Then, the equations of the differential geometry of
the dome are useful to understand and quantify the
geometrical properties of the dome.
Finally, the linear solution of the finite element
analysis fits the considerations of Timoshenko and
Heyman, but shows positive stresses, unsuitable for
the brick masonry.
The equilibrium of the dome is possible with a
diffuse cracking, especially in the lower part.
REFERENCES
ANSYS® Inc. Southpointe 275 Technology Drive Canons-
burg, PA
BARTOLI, L. 1994. Il disegno della cupola del Brunelleschi
Firenze Leo Olschki Editore
CECCHI, A. & PASSERINI, A. 2006. Survey, digital recon-
struction, finite element model of the Augustus Bridge in
Narni (Italy). 5th International Conference on Structural
Analysis of Historical Constructions, New Delhi 2006
563
DAVIDSOHN, R. 1956. Storie di Firenze. Sansoni
DI PASQUALE, S. 1977. Primo rapporto sulla Cupola di
Santa Maria del Fiore, CLUSF, Firenze
DO CARMO, M. P. 1976. Differential Geometry of Curves
and Surfaces, Prentice Hall, New Jersey
FITCHEN, J. 1961. The Construction of Gothic Cathedrals,
Oxford
FONDELLI, M. 2004. La Cupola di Santa Maria del Fiore, in
Giuseppe Rocchi Coopmans de Yoldi, S. Maria del Fiore,
ALINEA, Firenze
GIOVANNI DI GHERARDO DA PRATO, 1421. Document
conserved in the Museum of Opera del Duomo. Firenze
GUASTI, C. 1887. Santa Maria del Fiore: la costruzione
della chiesa e del campanile secondo i documenti tratti
dall’archivio dell’Opera Secolare e da quello di Stato, Tip.
M. Ricci, Firenze
HEYMAN, J. 1977. Equilibrium of Shell Structures (Oxford
Engineering Science), Oxford University Press
HEYMAN, J. 1966. The Stone Skeleton, Int. Journ. Solids
and Struct., 2
KRAUS, K. 1998. Fotogrammetria, Levrotto & Bella, Torino
MANETTI, A. 1480?. Vita di Filippo Brunelleschi
NAGHDI, P.M. 1972. The Theory of Shells and Plates.
Handbuch der Physik VI, Springer-Verlag
OPERA DEL DUOMO, 1691. Archivio di Stato. Filza 366.
Firenze
ROCCHI COOPMANS DE YOLDI, G. 1996. S. Maria del
Fiore, Università degli Studi di Firenze, Dipartimento
di Storia dell’Architettura e del Restauro delle strutture
architettoniche, Firenze
SANPAOLESI, P. 1962. Brunelleschi, G. Barbera, Firenze
SOKOLNIKOFF, I. S. 1951. Tensor Analysis – Theory and
applications, John Wiley and Sons
TIMOSHENKO, S. P. 1959. Theory of Plates and Shells, Mc
Graw Hill
VASARI, G. 1550. Le vite de’più eccellenti pittori scultori e
architetti Firenze
WITTKOWER, R. 1962. Architectural Principles in the Age
of Umanism, Alec Tiranti Ltd., London
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Some considerations on out-of-plane collapse modes of masonry walls

G. Brandonisio, E. Mele & A. De Luca

Department of Structural Engineer (DIST), University of Naples “Federico II”, Naples, Italy

ABSTRACT: In this paper the problem of out-of-plane collapse of masonry walls under vertical and horizontal
loads is addressed through the review of ancient building rules and the study of the European (Eurocode 6 and 8)
and new Italian seismic codes on unreinforced masonry buildings.
The limitations of the analysed seismic codes and the empirical geometrical rules of ancient constructors are
compared with the results of some structural analyses carried out on rectangular masonry panels with different
slenderness h/s, aspect ratio h/b (height/width) and boundary conditions. In particular, the analyses have been
carried out both with the limit analyses and non linear FEM analyses with Abaqus computer code.
The results of these parametric analyses show that the respect of the seismic codes limitations on the geomet-
rical slenderness h/s allows to implicitly have a minimum resistance to out-of-plane collapses of the masonry
walls.
A proposal is made to overcome the unjustified provisions of some codes which do not allow to have distance
among orthogonal walls larger then 7 m.

1 INTRODUCTION between the recommended h/s ratio and the aspect

Traditionally, the study of the behaviour of masonry
panels focuses the attention on the in-plane failures
and resistant mechanisms. Nevertheless, the out-of-
plane collapses of the walls are often activated both
under vertical loads and horizontal forces. These fail-
ure modes depend on many aspects, as quality of bricks
and mortar, bond pattern, geometrical characteristics
and boundary conditions of the wall.
In this paper the influence of the last two above
aspects (geometrical characteristics and restraint con-
ditions) are investigated. Particularly, the target of the
paper is to underline the relationships among: the h/s
ratio (height to thickness ratio) of the wall; the aspect
ratio h/b (height to width ratio); the edge restraint con-
ditions; and the out-of-plane strength of masonry wall
subjected to vertical loads and horizontal forces.
In this aim, in the first part of the paper the problem
of the buckling of a masonry wall under axial load is
addressed in order to find a relationship between the
wall slenderness λ (as defined in the stability theory
of columns or plates) and the load-carrying capacity.
Then, the behaviour of masonry wall under horizon-
tal actions is analyzed through both the evaluation of
the collapse multipliers and the study of geometrical
limitations, recommended both by the ancient rules of
building and modern European seismic codes (EC8’03
and NTC’07). The results are presented in a homoge-
nous approach in order to explicit the relationship
ratio h/b. Then, for several edge restraints conditions
of walls, designed according to required h/s ratio,
the relationships between the h/b ratio and the col-
lapse multipliers are illustrated and compared with the
results of a parametric nonlinear F.E.M. analysis.
The results herein obtained provide useful indica-
tions on the seismic capacity of the masonry walls.
2 WALL UNDER VERTICAL LOADS
As in to the case of a steel column under axial force,
the resistance of a masonry wall under pure axial load
depends on the slenderness λ, defined by the following
relationship:
where: ho is the buckling length of the wall in the
buckling plane considered; i is the radius of gyration
of the cross section of the wall; β is a coefficient that
depends on the type of restraint; h and s are the height
and the thickness of the wall, respectively.
From the Eq. (1) it is possible to note that the slen-
derness λ is proportional to h/s ratio, which can be
defined as geometrical slenderness of the masonry
panel.
The Figure 1 shows the relationship between the
axial resistance Nf of the wall and the slenderness λ

565
 Nf/Npl
N
h 1L 2L 3L 4L
1.0 b
s (a) (b) (c) (d)

N Figure 2. Edge restraints of the wall: (a) wall restrained at

the bottom (wall 1L); (b) wall restrained at the top and bottom
Transition Elastic (wall 2L); (c) wall restrained at the top, at the bottom and on
Crushing
region buckling one vertical edge, with the other vertical edge free (wall 3L);
(d) wall with 4 edges restrained (wall 4L).
λ 100 h/s

)
.75
4L

=0
80

(ρ2
Figure 1. Effect of slenderness on resistance of a masonry 3L

'03
5)

TC 1)
0. 7

2=
C6
wall under axial load. (ρ2=

'07
(8) '03 (ρ
6'03

:E
60 = 1)
C
E (ρ2

(5)
(3):

C6
:N
'03
C6

:E
: E

(6)
(in the chart Nf is normalized to the squash load Npl 40 (4)
(1): EC6'03 (ρ2=0.75)
of the wall). It can be observed that the diagram can (2): EC6'03 (ρ2=1) 2L
be divided in three fields, characterized by different 20 (7): NTC'07

behaviour of the wall. In fact, if the slenderness λ is

low (stocky wall), the failure would occur for compres-
sion of the material; whereas the slenderness λ is high
(slender wall), the failure would occur for elastic buck-
ling of the wall; in the case of intermediate values of
λ, the failure would occur for interaction between the
crushing and the elastic buckling of the wall (Hendry
et al. 1997).
The European (EC6’03) and Italian (NTC’07) pro-
visions for masonry structures do not refer to the
slenderness λ, but they introduce the conventional
slenderness λc defined as:
0 h/b
0 1 2 3 4 5 6
Figure 3. Comparison among the limitations of EC6’03 and
NTC’07 on h/s ratio.
– wall 3L (Fig. 2c): the value of ρ (appointed as ρ3 )
is depending on the degree of restraint ρ2 at the top
and bottom and on aspect ratio h/b:
– wall 4L (Fig. 2d): the value of ρ (appointed as ρ4 )
is again depending on the degree of restraint ρ2 at
the top and bottom and on aspect ratio h/b:

where: hef is the effective height of the wall; sef is
the effective thickness of the wall, that is usually
taken equal to the actual thickness s (for cavity walls
the EC6’03 suggests the following formula: sef =
(t31 + t32 )1/3 where t1 and t2 are the actual thickness of
the leaves; the Italian code NTC’07, instead, in every
case assumes sef = s)1 ; ρ is a reduction factor depend-
ing on the edge restraint or stiffening of the wall.
In the examined codes (EC6’03 and NTC’07) the
suggested values of the coefficient of restraint ρ are
related to the aspect ratio h/b and to the degree of the
edge restraints. Particularly, with reference to the walls
of Figure 2, the EC6’03 makes the difference among
the following three cases:
– wall 2L (Fig. 2b): the value of ρ (appointed as ρ2 )
is depending on the degree of restraint (ρ2 = 0.75
in presence of r. c. floors at the top and bottom of
the wall; ρ2 = 1.0 for walls restrained at the top and
bottom by timber floors);
1
In the following part of this paper, we will assume sef = s.
In the EC6 the slenderness λc should not be greater
than 27 when the wall is subjected mainly to vertical
loading. This limitation on λc , together with the sug-
gested limitations on ρ, may be used to establish the
maximum h/s ratio allowed by the EC6.
In Figure 3 the relationships between h/s and h/b
are shown in graphical form. In particular, the curves
(1)–(2) are associated to the limits of the wall 2L,
while the curves (3)–(4) and curves (5)–(6) are asso-
ciated to the limitations (3) and (4), for wall 3L and
4L, respectively.
The comparison among the curves (1) ÷ (6) of
Figure 3 underlines the effect of the lateral edge
restraints on the required h/s ratio. For example, in
the case of wall with aspect ratio h/b = 1, the maxi-
mum values of geometrical slenderness h/s is equal to
36 (ρ2 = 0.75) or 27 (ρ2 = 1.0), for the wall 2L; it is
equal to 38 (ρ2 = 0.75) or 30 (ρ2 = 1.0), for the wall
3L; it is equal to 56 (ρ2 = 0.75) or 54 (ρ2 = 1.0), for
the wall 4L.
Instead, the new Italian code NTC’07 does not con-
sider the case of wall 3L. In fact, in NTC’07 are

566
recommended reduction factors ρ only in the case of
wall 2L (ρ2 = 1.0) and wall 4L, for which ρ depends
only by the aspect ratio: ρ4 = f(h/b).
The maximum slenderness λc permitted in the
NTC’07 is 20; this limit, together with the recom-
mended values of ρ, allows to plot in Figure 3 the
curves (7) and (8) associated to the walls 2L and 4L,
respectively.
The comparison among the curves (1)÷(6) associ-
ated to the limits of EC6’03 and to the curves (7) and
(8) derived by the NTC’07, shows that the Italian code
is more conservative than the EC6 both for wall 2L and
wall 4L (Fig. 3). For the case of wall 4L, in Figure 4 it is
proposed the comparison between the coefficients of
restraint β and ρ, that define the relationships between
the slendernesses λ and λc and the h/s ratio (Eqs. (1)
and (2), respectively). Therefore, the diagram allows to
compare the differences between the theoretical slen-
derness λ (Eq. 1) and the conventional slenderness λc
(Eq. 2).
To define the reduction factors β of the wall 4L, it
is utilized the elastic theory of the buckling of the rect-
angular plates under compressive loads. If the wall’s
edges may be considered pinned, the critical stress σcr
of the wall 4L becomes:
where: m is the number of half waves in witch the
plate buckles in the vertical direction (m = 1, 2, . . ., n);
E·s3
D = 12·(1−ν 2 ) ; E is the modulus of elasticity; ν is the
Poisson’s ratio (in Figure 4 it is utilized E = 1100 MPa
and ν = 0.1).
The condition:
allows to define the slenderness λ in the case of wall
4L with all pinned edges:
where the reduction factor β assumes the expression:
The trends of the curve (1)÷(3) of Figure 4 show
that the factors β and ρ are equal only in the case of
EC6’03 in the hypothesis ρ2 = 1; the associated curve
(2), indeed, is situated on the bisecting line. On the
contrary, in the cases of EC6 with ρ2 = 0.75, and of
567
12·β=λ/(h/s)
1.2
(2): EC6'03 (ρ2=1)
1.0
(1): EC6'03 (ρ2=0.75)
0.8
0.6 (3): NTC'07
0.4
4L
0.2
ρ=λc/(h/s)
0
0 0.2 0.4 0.6 0.8 1.0 1.2
Figure 4. Comparison between the coefficients of restraint
β and ρ.
NTC’07, only for small values of reduction factors
(lower than 0.5) there is coincidence between β and ρ;
for higher values (greater than 0.5), in EC6 (ρ2 = 0.75)
the factor β is larger than ρ, while in the NTC’07 β is
lower than ρ.
By observing the curves plotted in Figure 4, it is
possible to found an approximate relationship between
λ and λc . In fact the curves (1), (2) and (3) suggest the
following identities:
The relationships between capacity reduction fac-
tor Nf /Npl and conventional slenderness λc , proposed
in EC6’03 and NTC’07, are plotted in Figure 5a
and b, respectively. In the diagrams are reported
several curves corresponding to different values of
eccentricity e of axial load N.
The application of Eq. (9) allows to individuate
the influence of slenderness λ on the reduction fac-
tor Nf /Npl . The Figure 5, indeed, suggests that Nf /Npl
is almost equal to 1.0 (almost no reduction) when
λ is lower than about 20 (stocky wall). When λ is
greater than about 50÷60, instead, the reduction of
resistance is significant (slender wall). Finally, when
λ is comprised between 20 and 50÷60, the wall shows
an intermediate behaviour, with interaction between
crushing and elastic buckling phenomena.
3 WALL UNDER HORIZONTAL LOADS
Also the out-of plane resistance of a wall subjected to
horizontal loads is influenced by the h/s ratio. Indeed,
since the ancient rules of the art of building masonry
structures, the h/s ratio has been properly limited in
 Nf/Npl 3.0 α (4) h/s=10 (3)
1 λc,max
e/s=0.05
2.5 4L 3L
0.8
e/s=0.10 EC6'03 α4=2.4
e/s=0.15
2.0
e/s=0.20
0.6
e/s=0.25 1.5
0.4 e/s=0.30 α3=1.2 2L
1.0 α2=0.8
e/s=0.33 (2)
0.2 0.5 1L
α1=0.1 (1)
0 λc 0.0 h/b
27
0 5 10 15 20 25 30 0 0.5 1 1.5 2 2.5 3
λ
108
0 ~ 20 ~ 40 ~ 60 ~ 80 ~ 100 ~120
(a) Figure 6. Comparison among the multipliers α of the walls
Nf/Npl 1L, 2L, 3L and 4L.
1 λc,max
When the wall is sufficiently retrained at the top
0.8 NTC'07
and bottom by the floor diaphragms (wall 2L) the
multiplier α becomes (Hendry et al. 1997):
0.6
m=6e/s=0

0.4 m=6e/s=0.5

m=6e/s=1
0.2
m=6e/s=2 m=6e/s=1.5 Instead, if the wall is stiffened on one or on two
0 λc
vertical edges (wall 3L and 4L, respectively), in the
0 5 10 15 20 25 30 hypothesis that the yield lines are slanted at 45◦ , the
λ collapse multipliers α are, respectively:
0 ~ 17 ~ 34 ~ 51 ~ 68 ~ 85 ~102
(b)

Figure 5. Reduction of resistance vs. slendernesses λc and

λ in EC6’03 (a) and NTC’07 (b).

order to reduce the vulnerability of the walls to the

out-of-plane collapses.
In this paragraph, the out-of-plane collapse mech-
anisms of a wall with different restraint conditions at
the edges are initially examined and the collapse mul-
tipliers are evaluated; then, the ancient rules of the
art and the European (EC8’03) and Italian (NTC’07)
seismic provisions for the unreinforced masonry build-
ings, which mainly consist of geometrical limitations,
are analyzed; finally the resistance of masonry walls
to horizontal loads, designed according to the above
geometrical limitations (ancient rules of the art and
seismic codes) is evaluated.
3.1 Collapse multipliers of horizontal loads
In the case of uniform distribution of horizontal loads,
the application of the principle of virtual works to the
wall 1L of Figure 2a, allows to explicit the relationship
between the geometrical h/s ratio and the out-of-
plane collapse multiplier α1 (defined as maximum
horizontal load to self weight load ratio):
The Eqs. (12) and (13) show that the multipliers α3
and α4 depend on the geometrical slenderness h/s and
on the aspect ratio h/b of the wall.
In the case of h/s = 10, the curves (1)÷(4) associ-
ated to Eqs. (10)÷(13) are plotted in Figure 6. It can be
observed the influence of both edge restraints condi-
tions and aspect ratio h/b on the out-of-plane resistance
of the wall. In fact, the diagram of Figure 6 shows
that: (i) the curves (1) and (2) (corresponding to wall
1L and 2L, respectively) do not depend on the aspect
ratio h/b; (ii) in presence of floor diaphragms at the
top of the wall (wall 2L), the multiplier α increases
of 8 times with respect to the case of wall 1L; (iii) in
presence of wall 3L and 4L with 1 or 2 stiffened ver-
tical edges, the collapse multiplier α shows a further
increase when the h/b ratio increases. For example, in
Figure 6 a comparison among the multipliers α1 , α2 ,
α3 and α4 is provided when h/b = 1.
3.2 Ancient rules of art
Regarding the stability of unreinforced masonry walls
under horizontal loads, J. B. Rondelet (1802), in his

568
30 h/s (d) h/s=12 1+(h/b)2
(a)); h/s = 10, medium stability (curve (b)); h/s = 12,
low stability (curve (c)).
4L Particularly, the curve (d) shows that the required
h/s=21 h/s ratio to guarantee the stability of the wall increases
20
with the aspect ratio h/b of the masonry panel; for
h/s=17 1L example, if h = 3.5 m, for h/b = 1 it is necessary to
(c) h/s=12 have a minimum value of h/s = 17, i.e. a wall thick-
10 (b) h/s=10 ness s = 21 cm; instead, for h/b = 1.5 it is necessary to
(a) h/s=8 have a minimum value of h/s = 21, i.e. a wall thickness
s = 17 cm.
The above Rondelet’s rules, in terms of collapse
0 h/b
1.5 multiplier α (Eqs. (10) and (13)), lead to the curves
0 1 2 3
(a)÷(d) reported in Figure 8. The comparison among
these curves shows the great increase of resistance to
Figure 7. Comparison among the rules of J. B. Rondelet.
out-of-plane loads of wall 4L (curve (d)) with respect
to wall 1L (curves (a)÷(c)). Further, the trend of curve
3.0 α 4
(d): α4= (d) shows that the multiplier α4 increases with h/b
2.5 4L 3· 1+(h/b)2 · (1-2/3·h/b) ratio; indeed, the curve (d) is asymptotic to the vertical
line h/b = 1.5. In such meaning, Rondelet writes that
2.0 the collapse of wall is impossible when the distance b
1.5 between the transversal stiffening walls is low.

1.0 (a): h/s=8, α1=0.125

(b): h/s=10, α1=0.10 1L 3.3 Seismic codes for unreinforced masonry
0.5 (c): h/s=12, α1=0.083 buildings
0.0 h/b In order to reduce the attitude to out-of-plane col-
0 0.5 1 1.5 2 2.5 3 lapse of unreinforced masonry panels, also the modern
European seismic codes (EC8’03 and NTC’07 (seis-
Figure 8. Collapse multipliers for the rules of Rondelet. mic part)) limit the h/s ratio through the conventional
slenderness λc (Eq. (2)). Particularly, in presence of
historical treatise asserts that a wall will have strong
masonry wall with natural stone units, the Eurocode 8
stability if h/s=8, medium stability if h/s = 10, low
prescribes the following limit:
stability if h/s = 12.
From the well-known study on the collapse mecha-
nisms for unreinforced masonry panels characterized
by different restraint conditions, Rondelet realizes
that, for the same degree of stability, the thickness s can
be reduced when the distance b between the transver- while, the new Italian code NTC’07 recommends the
sal wall decreases. Particularly, in the case of a wall of following limits:
one floor and unique room building, when the beams
of the roof are well fixed to the wall and there are no
roof thrust, the minimum value of the thickness s which
guarantees an adequate degree of stability for the wall,
can be calculated using the following empirical rule by
Rondelet:

The above limitations on conventional slenderness

In term of h/s ratio, the Eq. (14) can be written as
follow:
The geometrical Rondelet’s rule (15) is plotted in
Figure 7 (curve (d)) together with the limits that Ron-
delet gives for wall 1L: h/s = 8, strong stability (curve
λc , and consequently on h/s ratio, are more restrictive
with respect to the case of buildings in non-seismic
zone. In Figure 9 the comparison among the analyzed
seismic limitations are provided in term of h/s ratio.
In the cases of wall 1L and 4L, it can be observed
a substantial coincidence between the limitations of
EC8’03 and NTC’07, especially when h/b < 1 (for wall
4L); instead, for h/b > 1 the limitations prescribed by
EC8’03 are more conservative than the Italian provi-
sions ones. Finally, also for this diagram it is possible
to note the effect of edge restraints on the h/s limits.

569
 3.0 α
(5): EC8'03 (ρ2=0.75)
(9): NTC'07 (seismic zones 3 and 4)
(8): NTC'07 (seismic zones 1 and 2)
30 h/s 2.5 4L

)
(q): seismic zones 3 and 4 (λc =12)

=1
ρ2
(p): seismic zones 1 and 2 (λc =10)

3(
2=
1) 2.0
(ρ

8'0
5) '03
EC
20
): '0 3 ( ρ2=0.7 ): EC8 1.5 (n): seismic zones 3 and 4 (λc =12)
C8 (4
(6
(3): E (1): EC8'03 (ρ2=0.75);
(m): seismic zones 1 and 2 (λc =10)
NTC (seismic zones 3 and 4) 1.0
10 2L
(2): EC8'03 (ρ2=1)
0.5
(7):NTC (seismic zones 1 and 2) 2L h/b
0.0
0 h/b 0 0.5 1 1.5 2 2.5 3
0 1 2 3

Figure 11. Collapse multipliers for NTC’07.

Figure 9. Comparison among the limitations of EC8’03 and
NTC’07 (in seismic zones) on h/s ratio.
3.0 α
(h) (g)
3.0 α 2.5 4L (q)
2.5 4L 2.0 (i) (p) 3L
(l)
2.0 3L 1.5
(l) (ρ2=1)
1.5 (i) (ρ2=0.75) (h) (ρ2=1) 1.0
(g) (ρ2=0.75) 2L
1.0 0.5 (d) (f) (n)
2L (e)=(m)
0.5 (f) (ρ2=1) 0.0 h/b
(e) (ρ2=0.75) 0 0.5 1 1.5 2 2.5 3
0.0 h/b (a)
0 0.5 1 1.5 2 2.5 3 (b) 1L
(c)

Figure 10. Collapse multipliers for EC8’03. Figure 12. Comparison among the limitations of Ron-
delet, EC8’03 and NTC’07 (in seismic zones) on collapse
In terms of collapse multiplier α (Eqs. (11)÷(12)), multiplier α.
the condition (16) recommended in EC8’03 allows to
obtain the curves plotted in the diagram h/b-α of Fig- walls 2L and 4L it is possible to note a certain coin-
ure 10. Particularly in figure are reported: the curves cidence among the plotted curves; particularly, it can
(e) and (f) for the wall 2L; the curves (g) and (h) for the be observed that the Rondelet’s curve (d) is overlaid to
wall 3L; the curves (i) and (l) for the wall 4L. Each cou- the curve (q) (associated to NTC’07) and to the curve
ple of curves is associated to the condition ρ2 = 0.75 (g) of EC8’03 (wall 3L, ρ2 = 0.75) when h/b > 1 and
or ρ2 = 1 that the EC6’03 considers in the definition h/b < 0.8, respectively.
of hef .
The limitations (17) and (18) of Italian code
NTC’07 (seismic part) are plotted in the h/s-α dia- 4 F.E.M. PARAMETRIC ANALYSIS
grams of Figure 11 for wall 2L (curves (m) and (n))
and for wall 4L (curves (p) and (q)). The diagram In this paragraph, non linear analyses on rectangular
shows that in high seismicity zones (curves (m) and walls are carried out through the F.E.M. computer code
(p)), the Italian code recommends a minimum value of Abaqus 6.7-1 [Simulia, 2007]. On the basis of these
collapse multiplier α higher than the case of low seis- analysis results, some considerations about the col-
micity (curves (n) and (q)). Moreover, the curves (p) lapse multiplier α of walls 1L, 2L, 3L and 4L subjected
and (q) associated to wall 4L show a constant trend in to horizontal loads, are derived.
the range h/b = 0.5÷1, i.e. in the most common range
of aspect ratio h/b for the masonry walls. The pres-
4.1 Modelling of the masonry walls
ence of the horizontal plateau in the curves (p) and (q)
depends on the fact that the Italian code uses the same In Table 1 are reported the characteristics of the 42
expressions of the collapse mechanisms as the ones examined models. The specimens have been labelled
examined in §3.1 to define the reduction factor ρ. according to the following criteria: h-b-s Ln, where the
In Figure 12 the comparison among the limitations h, b and s are the geometrical dimensions of the wall
of Rondelet, EC8’03 and NTC’07 (in seismic zones) (in cm) and Ln (n = 1÷4) defines the edge restraint
on collapse multiplier α, is provided. In the cases of of the wall as schematized in Figure 2. Regarding the

570
Table 1. Examined models. 3.0 α
4L
Specimen 350-350-35
2.5 h/b=1; h/s=10
Specimen h/b h/s α4=2.4
2.0 3L F.E.M. analysis
350-350-45 L1 1 8 1.5 α3=1.2 Limit analysis
350-350-45 L2 1 8 1.0 α2=0.8 2L
1L
350-1400-45 L3 0.25 8 0.5
α1=0.1
350-700-45 L3 0.5 8 0.0 d [mm]
350-470-45 L3 0.75 8 0 5 10 15 20 25
350-350-45 L3 1 8 (a)
350-280-45 L3 1.25 8
350-233-45 L3 1.5 8
350-1400-45 L4 0.25 8
350-700-45 L4 0.5 8
350-470-45 L4 0.75 8
(b)
350-350-45 L4 1 8
350-280-45 L4 1.25 8
350-233-45 L4 1.5 8
350-350-35 L1 1 10
350-350-35 L2 1 10
350-1400-35 L3 0.25 10 (c)
350-700-35 L3 0.5 10
350-470-35 L3 0.75 10 Figure 13. Specimens 350-350-35: (a) comparison
350-350-35 L3 1 10 among the push-over curves and the collapse multipliers;
350-280-35 L3 1.25 10 (b) deformed shapes with stress tensor vectors of the F.E.M.
350-233-35 L3 1.5 10 models; (c) collapse mechanisms considered for limit
350-1400-35 L4 0.25 10 analyses.
350-700-35 L4 0.5 10
350-470-35 L4 0.75 10
350-350-35 L4 1 10 In order to correctly calibrate the model parameters,
350-280-35 L4 1.25 10 reference has been made to the curve fitting procedure
350-233-35 L4 1.5 10 was made by Giordano A. (2002), which utilizes the
350-350-30 L1 1 12 results of the experimental tests on masonry tuff walls.
350-350-30 L2 1 12 The applied loads are the self weight and the hor-
350-1400-30 L3 0.25 12 izontal load, which increases with monotonic low up
350-700-30 L3 0.5 12 to the end of the analysis.
350-470-30 L3 0.75 12
350-350-30 L3 1 12
350-280-30 L3 1.25 12 4.2 Results and comparisons
350-233-30 L3 1.5 12
350-1400-30 L4 0.25 12 In Figure 13a, a summary of the non linear analyses
350-700-30 L4 0.5 12 for the walls 350-350-35 is reported; particularly, the
350-470-30 L4 0.75 12 pushover curves are depicted. In this diagram, the col-
350-350-30 L4 1 12 lapse multipliers calculated through the limits analyses
350-280-30 L4 1.25 12 and already provided in Figure 6 are also reported.
350-233-30 L4 1.5 12
The comparison among the nonlinear static and limit
analyses shows that the F.E.M. results provide collapse
multipliers higher than limit analyses, because the ten-
geometry, all the walls have height h = 350 cm; the sile strength of masonry is considered in the Abaqus
width b varies between 233 and 1400 cm in order to models.
have walls with aspect ratio between 1.5 and 0.25. The Indeed, in Figure 13b and c the visualization of
thickness s is set equal to 30, 35, 45 cm, in order to have the F.E.M. deformed shapes, with stress tensor vec-
walls with h/s ratio equal to 12, 10 and 8, respectively. tors, and the hypothesized collapse mechanisms are
Four-nodes shell elements (S4R5 elements) are reported. The comparison shows a good agreement
used to model the masonry tuff walls; reduced inte- between the Abaqus and limit analyses.
gration is used for the shell elements; the number In Figure 14 the comparison among the collapse
of integration points through the thickness of shell multipliers computed by means of F.E.M. analysis and
element is equal to five. of limit analyses (Eqs. (10)÷(13)) is provided. Par-
All the examined walls have been subjected to non ticularly, the diagrams refers to walls characterized
linear analyses using a smeared cracking approach as by h/s ratio equal to 10. In almost all analyzed spec-
implemented in the computer code Abaqus. imens (Table 1) the diagrams confirm the previous

571
 6 α (4) (3) h/s=10 (Fig. 15a) and NTC’07 (Fig. 15b). The thicknesses s of
4L 3L the analyzed specimens have been obtained by Figure
5 9 curves (4) (wall 3L) and (6) (wall 4L) for EC8’03,
Abaqus 4L
4 and by Figure 9 curves (8) and (9) for NTC’07. In the
case of EC8’03, the results of F.E.M. analyses show
3 an increasing trend of the curves (Fig. 15b), while in
Abaqus 3L the case NTC’07 it possible to note a sub-horizontal
2
2L trend of the F.E.M. curves.
1 (2) Moreover, when h/b ratio is larger than one, the
0 h/b NTC’07 curves show a slight decrease of collapse mul-
0 0.5 1 1.5 2 2.5 3 tipliers; this seems to underline that the limitations of
NTC’07 give very large values of maximum h/s ratio
Figure 14. Comparison among the collapse multipliers when the wall is characterized by high values of aspect
of walls with h/s = 10 evaluated through F.E.M. and limit ratio h/b.
analyses. Finally, the comparison among the F.E.M. results
reported in the curves (a) of Figure 15 shows that the
6 α (l) (ρ2=1) (h) (ρ2=1) EC8’03 is on the safe side with respect to NTC’07; in
4L 3L fact, it can be observed that the curve “Abaqus 4L” of
5
Figure 15a is characterized by higher values of collapse
4 Abaqus 4L multipliers α than the curves “Abaqus 4L” plotted in
Figure 15b.
3
2 Abaqus 3L
5 CONCLUSIONS
1
The study of the out-of-plane collapses of masonry
0 h/b
(a) 0 0.5 1 1.5 2 2.5 3 walls subjected to vertical loads has suggested the
relationship between the slendernesses λ and λc or,
equivalently, between the coefficients of restraint β
6 α
and ρ. In term of slenderness λ, it has been observed
4L
5 that a wall characterized by λ < 20 can be defined as
stocky wall, having a low reduction of axial strength
4 Nf with respect to the squash resistance Npl of the wall;
(p): seismic zones (q): seismic zones
3 1 and 2 (λc=10) 3 and 4 (λc=12) when λ > 50÷60, instead, the wall can be considered
as slender wall. Finally, if the slenderness λ is com-
2 Abaqus 4L prised between 20 and 50÷60, the wall is characterized
1 by an intermediate behaviour, with interaction between
the crushing and the elastic buckling of the wall.
0 h/b In the second part of this paper, the issue of
0 0.5 1 1.5 2 2.5 3
(b) out-of-plane collapses of masonry walls subjected to
horizontal loads has been studied through (i) the anal-
Figure 15. Comparison among the collapse multipliers yses of the collapse mechanisms; (ii) the review of
evaluated through nonlinear static and limit analyses on ancient rules of the art; (iii) the study of the European
walls designed according to the geometrical requirements of (EC8’03) and new Italian seismic codes (NTC’07) for
EC8’03 (a) and NTC’07 (in seismic zones) (b).
unreinforced masonry buildings.
In the case of walls 3L and 4L, the comparison
observation regarding the multipliers α; in fact, the among the geometrical limitations underlines that the
values of α computed through pushover analyses are recommended h/s ratio becomes greater than the val-
generally higher than the values calculated with the ues h/s = 8, 10 or 12, that Rondelet suggests for the
limit analyses. Finally, also the curves associated to the wall 1L. For example, in the range h/b = 0.5÷1.5, the
results of nonlinear static analyses carried out on walls required geometrical slenderness h/s for the wall 4L
3L and 4L show an increasing trend with the aspect is equal to or smaller than: 13÷21 according to Ron-
ratio h/b; this confirms the beneficial effect of the ver- delet’s rule (Eq. 15); 11÷27, according to EC8’03;
tical edge restraints on the resistance of the walls to 10÷32, according to NTC’07.
out-of-plane collapses. These observations are also confirmed in the prac-
The Figure 15 shows the comparison among the tice for a wider range of aspect ratio h/b, as shown
F.E.M. results and the curves relative to collapse mul- in Figure 16 for the 8 classes of macro-elements of 10
tipliers α of the walls 3L and 4L associated to EC8’03 masonry churches studied in Brandonisio et al. (2008).

572
50 h/s
40
30
20
h/s=12
10 h/s=10
h/s=8
h/b
0 to 7 m often recurrent both in ancient treatises and in
0 1 2 3 4 5 6 the modern seismic codes. In particular, it seem that
Figure 16. h/s vs. h/b relationships for the masonry walls of
10 Italian masonry churches analyzed in Brandonisio et al.
(2008).
The study of the collapse mechanisms together
with the analyses of the geometrical requirements
derived from the ancient rules and the modern seis-
mic provisions, show the influence of the h/s, h/b and
edge restraints conditions on the out-of-plane collapse
multiplier α.
The h/b-α curves suggest a remark regarding the
“Rules for simple masonry buildings” reported in
the EC8’03 and NTC’07. In fact, both EC8’03 and
NTC’07 require a maximum spacing of 7 m between
two bracing walls; this geometrical requirement, for
the common values of storey height h = 3.5÷5 m, con-
ducts to walls characterized by aspect ratios h/b greater
than 0.5. In absence of holes, this leads to masonry
panels which resist to out-of-plane lateral loads equal
to: 87% (when ρ2 = 0.75) or 106% (when ρ2 = 1) of
their self weigh, in the case of walls designed accord-
ing to EC8’03; 120% (in seismic areas 1 and 2) or
100% (in seismic areas 3 and 4), in the case of Italian
seismic code.
In the last part of the paper, the results of a paramet-
ric nonlinear F.E.M. analysis carried out on rectangular
573
masonry walls has been presented. In the F.E.M. anal-
yses both h/s and h/b ratios and the edge restraints have
been varied. The results confirm the observation done
through the application of the limit analyses, i.e. the
respect of the seismic codes limitations on the geo-
metrical slenderness h/s allows to implicitly have a
minimum resistance to the out-of-plane collapses of
the masonry walls subjected to horizontal loads.
In the light of the previous consideration, it could
be suggested to overcome the indications concerning
the maximum spacing between the cross walls equal
a more appropriate limitation on orthogonal elements
should be given in non dimensional forms, i.e. in terms
of aspect ratio of the masonry panel: h/b ≥ 0.5.
REFERENCES
Brandonisio G., Mele E., Santaniello R., De Luca A., 2008.
Seismic safety of basilica churches: analysis of ten case
studies. VI SAHC’08, Bath, UK. 2–4 July.
Giordano A., 2000. Numerical Modelling of Masonry Struc-
tures Using the Abaqus “Concrete”. Model ABAQUS
Users’ Conference. Newport, Rhode Island.
Hendry, A. W., Sinha B. P., Davies S. R., 1997. Desig of
Masonry Strucutures, E & FN Spon.Great Britain.
NTC 2007. Norme Tecniche per le costruzioni. Draft of 6th
December 2007 (in Italian).
PrEN1996-1. Eurocode 6: Design of masonry structures –
Part 1-1: Common rules for reinforced and unreinforced
masonry structures. March 2003.
PrEN1998-1. Eurocode 8: Design of Structures for Earth-
quake Resistance. January 2003.
Rondelet, J. B., 1802.Trattato teorico e pratico dell’arte di edi-
ficare, Tomo IV - Teoria delle Costruzioni. Prima edizione
napolitana notabilmente migliorata e corretta da Raffaele
Pepe. Napoli, 1840 (in Italian).
Simulia 2007. ABAQUS Theory Manual. USA.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Innovative techniques for structural assessment: The case of the

Holy Shroud Chapel in Turin

A. De Stefano, D. Enrione & G. Ruocci

Politecnico di Torino, Turin, Italy

ABSTRACT: Environmental degradation, inevitable aging, negligence and extreme events are some of the
most frequent factors that can threaten the structural integrity of historical constructions. Hence, architectural
heritage owners and administrators urgently require cheap and effective tools for structural health assessment.
Restoration works on the Holy Shroud Chapel in Turin, Guarino Guarini’s masterpiece which was extensively
damaged by a fire in 1997, offered the authors the opportunity to test some innovative techniques that are able
to provide an exhaustive picture of the Chapel’s structural health. Dynamic tests were carried out by applying
several exciting actions and the signals were acquired to perform a modal identification of the dome in the
time-frequency domain. A stochastic multiple model approach, used to update a FE model of the Chapel, is
also presented. The PGSL algorithm, implemented in the model updating procedure, is able to define different
groups of models, each one referring to a particular damage scenario that the building is probably undergoing. The
improved knowledge about the structure is the starting point for the development of symptom-based structural
health assessments.

1 INTRODUCTION have to be diagnosed through a meticulous analy-

Italy, like many other countries throughout the world,
has inherited an extraordinary historical treasure from
the past. This architectural heritage represents a fun-
damental resource and a sign of the national cultural
background, but the high costs for its maintenance
and repairs cause concern. In the last few years, the
demand for cheap and effective tools designed for
these purposes has increased.
Researchers soon became aware of the need to intro-
duce the concept of “structural health” and to identify
some instruments and techniques to assess it in a
reliable way. This growing interest is demonstrated
by the huge number of publications and conferences
expressly devoted to these topics.
The term “structural health” stands for the condi-
tion in which the structure assures a suitable level
of safety. The definition of this safety level in turn
implies the knowledge of a set of all those parameters
which could affect it. Hence, the necessity of identify-
ing sensitive elements arises and, more in general, it is
important to acquire as much information as possible
about the construction conditions. Monitoring systems
become in this sense essential elements in strategic and
monumental structures.
The structural health assessment problem can be
seen under a different light. The structure to be moni-
tored can be considered as a “patient” whose diseases
sis of the surveyed symptoms. This symptom-based
approach has been extensively adopted and improved
in the mechanical and aerospace fields. Ancient build-
ings have to bear damage that arise during their
life-cycle from different concurring causes, such as
material quality degradation, environmental injuries,
inaccurate human interventions and modifications of
the structural body. All these situations represent
examples of so-called “damage scenarios”, i.e., the
provoking causes which lead to damage symptoms on
the construction.
Identifying damage scenarios and their probabil-
ity of occurrence requires that the existing knowledge
bases, whenever available, should be explored and
realistic simulations, capable of relating each possible
scenario to the severity of its consequences and the
meaningfulness of its symptoms should be conducted.
Dynamical tests and modal identification are power-
ful tools that can provide this required knowledge.
In addition, simulations performed with stochastically
updated models supply reliable estimations of the
residual performance a building is still able to provide
after structural changes inferred by each particular
damage scenario. This kind of information allows the
health condition of the construction to be predicted for
each damage hypothesis.
In the following chapters, a more detailed discus-
sion is given with some innovative techniques which

575
have been applied as the starting point for the develop-
ment of symptom-based structural health assessments:
the Holy Shroud Chapel is presented as an example.
2 STRUCTURAL HEALTH ASSESSMENT
TECHNIQUES
The monitoring of common structures represents a
well established and widespread practice, whereas the
situation of monumental buildings is rather different.
Only a few cases can be found referring to research
due to the strict constrains necessary to preserve the
cultural and artistic value of the buildings. Tools which
combine a suitable level of reliability with less invasive
interventions are preferred. Dynamic system identifi-
cation in this sense constitutes an optimal compromise
which is finding widespread approval among experts.
Dynamic tests are very powerful identification tech-
niques which are able to reveal global quantities such
as natural frequencies, modal shapes and damping
features of the structure under examination. These
parameters are closely related to the variations in
stiffness of the system and are therefore extremely sen-
sitive to damage and deterioration phenomena. Many
modal identification techniques are now available but
the complexity and uncertainty that affect ancient
masonry constructions drastically reduce their appli-
cation possibilities. Time-frequency domain methods
have proved to be accurate in treating non-stationary
signals and they are robust against noise. Further-
more, some instantaneous estimators, obtained as auto
and cross-correlation of bi-linear time-frequency sig-
nal transforms, allow the modal features of systems to
be reliably detected.
In the structural health assessment field FE mod-
elling also plays a very important role, even thought in
many cases the lack of numerical modelling accuracy
and a deficient correspondence to reality represent
decisive limitations. There is the risk of choosing
model parameters which do not fit real structure prop-
erties well. Model Updating (MU) provides a valuable
solution to the problem. It can be considered as the
missing link between structural identification and FE
modelling since it gives a model calibration crite-
rion on the basis of the experimental results. Model
updating techniques have undergone an exponential
growth during the last few years due to their latent
potentialities. Recently, stochastic approaches have
become more and more popular and can be distin-
guished from traditional deterministic methods due
to their capacity to deal with behaviour uncertainties
and problem complexity in a robust way. Deterministic
methods, whether direct or parametric, focus on model
calibration through the minimization of a target func-
tion obtained as the difference between experimental
results and analytical data. Since Model Updating
represents an inverse problem, the solution that is
576
obtained cannot be considered as the only possible
one. Furthermore, there are many error and uncertainty
sources that condition the final result. These should be
considered in order to obtain a consistent estimation
of the model parameters and consequently a reliable
structural health assessment.
On the basis of the remarks made so far, the authors
focused attention on the development of a model
updating method which could meet monumental build-
ing requirements:
– the treatment of errors obtained from the acquisi-
tion of experimental measures;
– the treatment of uncertainties due to construction
complexity (mechanical properties, constructive
methodologies, unknown design details, etc.);
– the management of information redundancy
obtained from many distributed sensors;
– extrapolation ability in the definition of numerical
models.
Solution accuracy is the main problem which the
authors tried to solve in a convincing way. As a unique
optimal solution for this problem cannot be found,
the authors concentrated on the possibility of generat-
ing a set of different but sufficiently reliable models.
The adopted approach is generally referred to as the
“multiple-model” due to the fact that it generates
a huge number of models to solve the problem of
the solution uniqueness. Application of the adopted
method to the case study has allowed some models
to be defined and has made them able to accurately
predict its structural behaviour. An interesting devel-
opment the authors intend putting into practice regards
the identification of possible damage scenarios for
each model obtained using the updating procedure.
The natural consequence is the estimation of the resid-
ual performance a building is still able to supply, i.e.,
its structural health.
The case of the Holy Shroud Chapel in Turin is pre-
sented in order to clarify the concepts introduced so far.
3 CASE STUDY: THE HOLY SHROUD CHAPEL
IN TURIN
3.1 The experimental campaign
The Holy Shroud Chapel in Turin (Fig. 1) is univer-
sally recognized as a outstanding example of Italian
Baroque architecture. Emanuale Filiberto di Savoia
entrusted its design to the famous Italian architect
Guarino Guarini, who built the Chapel from 1667 to
1694. Since the very beginning, the monarch’s intent
was that of housing the precious relic of Christianity
in a more prestigious seat. The architectural results
obtained by Guarini are extraordinary. The whole
building conveys the Guarini’s obsession for architec-
tonic originality and a sense of mystery which are well

Figure 1. Section of the dome.
expressed by the structural complexity and the rich-
ness of perfect shapes and theological, astronomic and
mathematical symbols. The chapel is composed of a
tambour bored by six large windows and surmounted
by three big arches which sustain the dome. On the
inside, a series of small arches overlapped and dis-
posed on six levels, creates an hexagonal geometry
which diminishes towards the top where it becomes the
circular base of a lantern. On the outside, another series
of small arches creates a complex plaiting effect and
the alternation of black marble and grey stone grants
a particular sense of dynamicity.
The recent history of the Holy Shroud Chapel has
been marked by a tragic event. A fire broke out in 1997
during some restoration works and it seriously dam-
aged the structure, producing incalculable economic
and artistic loss.
In the following months, the Sovraintendenza
dei Beni Architettonici encharged the Politecnico di
Torino with the project to carry out a general experi-
mental campaign on the materials and structure (Prof.
P. Napoli) and a dynamic test program (Prof. A. De
Stefano).
Mainly in situ investigations were performed in
order to obtain detailed knowledge of the structural
morphology in term of geometry, marble and masonry
organization, position of metal ties, etc. The investi-
gations, conducted through topographic and camera
surveys and deep extraction of samples, allowed the
experts to detect some structural elements that had
577
previously been covered over. As a result of these
investigations, a 3D computer geometrical model of
the entire building was set up and more detailed
knowledge in the structural morphology was acquired.
The mechanical properties of the materials were
measured both in laboratory and in situ tests, in the
latter case using cross-hole ultrasonic measurements.
The tests revealed the chemical degradation induced
by the heat on the surface of the elements and a reduc-
tion in the material load carrying capacity in the zones
affected by stresses caused by the effect of constrained
thermal deformations.
Local inspections provided useful information con-
cerning about the mechanical properties of the mate-
rials and structural configuration and allowed a first
preliminary estimation of extent of the damage to be
obtained. However, due to their spatial limitation, these
tests were not able to completely identify the global
behaviour of the construction. They therefore had to
be integrated with other techniques that were capa-
ble of predicting the residual structural capacity of the
building considered as a whole.
In order to achieve a complete overall image of the
Chapel’s structural health, the Research Unit to which
the authors belong, in association with the University
of Kassel (Prof. M. Link), designed a dynamic test-
ing programme, prepared a FE model and adjusted it
according to the acquired experimental knowledge.
In order to perform the dynamic experiments, 25
accelerometers were positioned on six different lev-
els to measure the response of the structure along
three different orthogonal directions: radial, vertical
and horizontal normal to the radius. The dynamic tests
were realized adopting four different exciting actions:
– environmental excitation (traffic, wind, micro-
quakes);
– impulsive excitation produced by hammering;
– impulsive excitation caused by a sphere dropped to
the ground near the foot of the building;
– wind turbulence produced by a Fire-Brigade heli-
copter flying around the top of the dome.
The obtained signals were used to identify the struc-
ture of the dome using the TFIE (Time Frequency
Instantaneous Estimators) method.
3.2 Modal identification
The results obtained from the dynamic tests performed
on the chapel provided the starting point for the subse-
quent structural identification phase. A well-designed
structural health monitoring programme cannot disre-
gard this fundamental step which supplies information
concerning the overall behaviour of the system and it
is the basis for the FE model calibration.
Many effective techniques are now available for
structural identification, but in the case of complex
monumental buildings not all are suitable to robustly
face the several uncertainties that can affect this kind
of structure. Generally, output-only methods are pre-
ferred to input-output ones since they are less invasive.
Identification techniques can also be distinguished on
the basis of the parameters that need to be identified. It
is therefore possible to choose between direct methods,
which try to determine the [M], [C] and [K] matrices
of motion equation (1):
and indirect methods working on modal parameters
such as the frequencies, damping and modal shapes.
Another classification criterion concerns the domain
in which the data are numerically treated and thus
provides three possibilities:
– frequency domain;
– time domain;
– time-frequency domain.
Frequency domain methods (for example FDD,
PSD, etc.) have gradually been replaced by time
domain ones (for example ARMAV, ERA, Random
Decrement, etc.) in order to solve problems concern-
ing frequency resolution and modal density. Methods
developed in time-frequency domain offer several
advantages which are here summarized:
– precise accuracy in parameter estimation;
– the possibility of effectively managing non-
stationary signals;
– the ability to handle moderate non-linearities;
– high robustness against noise.
The main drawback that affects both time-domain
and frequency-domain methods concerns the assump-
tion that the modal parameters do not evolve versus
time and vibration amplitude and that the input is at
least weakly stationary. Since in the civil engineer-
ing area the excitation is generally non stationary
and the presence of noise during the data acquisition
phase is unavoidable, the structural response cannot
be regarded as an unvarying-versus-time signal. These
considerations forced the authors to adopt an adap-
tive and robust identification technique, that is able to
handle these types of non-stationary excitations. The
implemented method is based on special amplitude
and phase estimators defined in the time-frequency
domain known in literature as Time-Frequency Instan-
taneous Estimators (TFIE).
The Amplitude Ratio and PH ase difference estima-
tors are expressed by equations (2) and (3), respec-
tively:
578
Figure 2. Phase difference standard deviation versus fre-
quency. (50 pts/s, 10.24s); f2 = 2.344 Hz.
where Dsi (t,f) and Dsi,sj (t,f) are the auto and cross
bi-linear time – frequency transforms of two compo-
nents of the same vibration mode belonging to signals
acquired in two different nodes of the structure. The
thus defined estimators maintain time dependency
and, in this way, they make it possible to ascertain
whether a certain frequency could be a possible vibra-
tion mode.At the modal frequency the phase difference
in (3) becomes almost constant in time and its stan-
dard deviation along the time axis approaches 0, if
the noise level is low (Fig. 2). However a downward
peak in the standard deviation of the phase difference
process, plotted versus frequency, marks an expected
modal frequency. If this occurs, the AR in equation (2)
computed at that frequency also remains almost con-
stant versus time and marks the amplitude ratio of a
modal shape.
The TFIE method has been applied to several sets
of signals, generated by the different types of excita-
tion adopted in the dynamic tests performed on the
Holy Shroud Chapel. Modal frequencies were iden-
tified using all the consistent signal pairs in all the
combinations of the following sampling parameters:
– sampling frequency: 25 or 50 pts/s
– signal length: 256, 512 and 1024 pts.
The TFIE diagrams computed from signals with
identical sampling frequencies and lengths were aver-
aged in order to make the downward peaks that mark
the modal frequencies clearly visible. The records
obtained from the excitation of the dropped sphere,
with a sampling rate of 25 pts/s and length of 1024 pts,
Table 1. Subdivision of the elements in the FE model.

Type of element Total number of elements

Shell with 3 or 4 nodes 11871

Beam 3D 1646
Truss 48
3D solid 28177
Spring 277

provided the best results. The TFIE potential is evident

as it is able to detect the two first flexional modes of
the chapel, whose modal frequencies are very close to
each other (f1 = 2.246 Hz and f2 = 2.344 Hz).
The modal shapes were identified using the same
data pairs and sampling conditions which allowed the
best results in the frequency search. The first flexional
shape is oriented in the NE-SW direction, the second
in the NW-SE direction. In general, it is convenient to
use signals in cylindrical coordinates to identify modes
where torsion prevails whilst Cartesian coordinates Figure 3. Assonometric projection of the model and sub-
division in sub-structures. Indication of the most sensitive
work better to extract modal shapes with prevailing
parameters.
translation.

3.3 The finite element model

A FE model of the Chapel was prepared adopt-
ing geometrical data of the model created using the
Rhinoceros3D programme by Arch. Abrardi and Arch.
Gallo (May 2001). Their very accurate geometrical
model, reproduced for the Sopraintendenza dei Beni
Architettonici, was simplified to build a numerical
model with ADINA.
The bottom part of the model has a lower degree
of detail than the upper one. The base of the Chapel
was modelled to have a simplified distribution of the
mass and stiffness necessary to reproduce the iteration Figure 4. The results of the parameters sensitivity analysis.
between the upper and the lower parts. The internal
provisional steel structure has not been modelled for
the sake of simplicity. The model is constituted of mode shapes leads to an ill-conditioned problem with
17334 nodes and 42019 elements, divided as reported resulting unreliable solutions.
in Table 1. For this reason, a preliminary sensitivity analysis
The model was divided into 28 sub-structures. was performed to reduce the number of parameters.
The assonometric projection of the model and its The elimination of some insensitive parameters also
subdivision in sub-structures are shown in Fig. 3. permits, on the one hand, to reduce the computational
The modal frequencies and mode shapes depend on weight and, on the other, to reduce the noise level.
various factors: The sensitivity analysis was performed modifying
the parameters one by one, in a uniform way over a
– the geometry and typology of the model;
defined range and the obtained values were recorded.
– the finite element mesh;
The results are reported in Figure 4; the shown quan-
– the mass and stiffness distribution.
tities are calculated as the mean of the ratio between the
The calibration of a numerical model cannot be con- variation of the modal properties (the first 5 frequen-
duced adopting an arbitrary number of variables. The cies) with respect to that of the parameters. The most
number of parameters is influenced by the amount sensitive parameters are the numbers 19 (3DLivello02-
of experimental information. Choosing a larger num- 2), 11 (Tamburo Esterno), 6 (Timpani) and, to less
ber of parameters than the identified frequencies and extent, 18 (3DLivello03) and 12 (Tamburo Interno).

579
All these parameters are found in the central and lower
part of the structure. These parts in fact have a greater
influence on the global behaviour of the structure,
mainly for the first modal shapes.
All the selected parameters are the Young’s Mod-
ula of different sub-structures into which the model
was previously divided. Each substructure is an axial-
symmetric structure; this fact influences the behaviour
of the structure, making it more regular in every
direction. To avoid this, a further subdivision of each
sub-structure into four different ones was made. This
subdivision leads to an increase in the number of
selected parameters from 5 to 20 and is necessary
because the data obtained in the identification stage
highlight a non axial-symmetric behaviour of the
structure.
3.4 The stochastic model updating
In order to solve the intrinsic problem of the solution
uniqueness and to treat error and uncertainty sources
a stochastic “multiple-model” approach has been
adopted to update model parameters. This method can
be divided into the three following phases:
1. generation of a huge number of models varied on
the basis of some structural parameters through the
minimization of objective functions that are able to
explore a consistent part of the solutions space;
2. selection, among all the generated models, of
those which best fit the system behaviour with a
prescribed level of accuracy;
3. analysis of the detected model properties and their
classification through clustering techniques.
The adopted method implements the PGSL algo-
rithm introduced by Smith (1998) which creates the
models and assigns them parameter values derived
from probability distributions. The principal assump-
tion is that better points are likely to be found in the
neighbourhood of families of good points. Hence, the
search is intensified in regions containing good solu-
tions. The search space is sampled by means of a
PDF defined over the entire search space. Each axis
is divided into a fixed number of intervals and a uni-
form probability distribution is initially assumed. As
the search progresses, the intervals and probabilities
are dynamically updated so that sets of points are gen-
erated with a higher probability in regions containing
good solutions.The search space is gradually narrowed
down so that convergence is achieved.
The algorithm includes four nested cycles:
1. the Sampling Cycle, where a certain number of
models is generated and analysed through the
target function, using the actual probability distri-
bution. The Best Sample is selected, memorized
in the database and then used to recalculate the
probability distribution;
580
2. the Probability Updating Cycle, in which all the
probability distributions are updated as follows:
the probability values of the sub-interval containing
the Best Sample selected in the preceding cycle and
its neighbours are increased, while, the probabilities
in the other regions are decreased.
3. the Focusing Cycle, which modifies the structure
of the variation interval of each parameter in order
to focus the search near the actual best solution.
4. the SubDomain Cycle, which works on the solution
space and increase the resolution in order to ease
the confluence.
Each operation performed by the algorithm is man-
aged by specific parameters which have to be cho-
sen carefully because some of them present a high
dependence on the considered problem.
In the case study, the Root Mean Square Error
between the identified and calculated frequencies is
utilized as the objective function which has to be min-
imized. At each step, a new parameter distribution is
created by the algorithm and new model outputs are
obtained. In order to match the numerical and exper-
imental modal shapes, the MAC function, that gives
a measure of similarity between two different modal
shapes, is calculated.
The model generation procedure has been divided
into two different phases, each one referred to a partic-
ular objective functions. Firstly, the PGSL algorithm
generates a huge number of models through the mini-
mization of a function set on the first modal frequency,
which is the best-identified in the dynamic tests. The
second objective function instead utilizes the first
three model frequencies and mode shapes matched by
means of the MAC function. This second step allows to
select only some of all generated models, while mod-
els that have a higher value than a certain threshold are
neglected.
All these selected models can be considered as rea-
sonable candidates to represent the real building, but
their large number makes necessary the application of
some data mining techniques that can be applied to
discover different types of patterns in the model data.
The PCA (Principal Component Analysis) is per-
formed to generate a new set of variables called
principal components that are linear combinations of
the original ones. The goal of the PCA is to find a
system of principal components that are sorted so that
the first components can explain most of the data vari-
ance. In this way the number of dimensions can be
reduced by choosing only the first two or three princi-
pal components. Thus the original data are represented
by a linear combination of the original parameters in a
new and lower dimensional space. The application of
the PCA and the examination of the first three prin-
cipal components supplied an important result. The
first three principal components are not sufficient to
explain most of the data variance. For this reason it

Figure 5. Percentage of the selected models which belong
to each group.
is not possible to represent all data in a 3D reference
system.
Finally, some clustering techniques have been
applied to identify and collect models into groups. The
adoption of the k-means algorithm, which sub-divides
the data into k subsets minimizing the distances within
each group and maximizing the distances between the
k different groups, allowed the definition of 9 models.
These models are the centroids of the 9 groups identi-
fied by the k-means algorithm and can be considered
as the best candidates to represent the real structure.
The obtained results are reported in Figure 5, where
the percentage of the selected models which belong to
each group is shown.
4 CONCLUSIONS
Some innovative techniques, which are able to assess
the structural health of monumental buildings, have
been presented in this article. The application of these
methods to the Holy Shroud Chapel in Turin was car-
ried out in order to identify both their potentialities
and limitations.
A brief description of the experimental campaign,
carried out to acquired more detailed knowledge of
the structural morphology, is first given. The dynamic
test programme designed by the research group the
authors belong to is also presented and the differ-
ent exciting sources adopted are listed. The results
obtained with the TFIE modal identification method
are shown in the subsequent part of the paper. The
powerfulness of this adaptive and robust tool is mainly
demonstrated by its ability to handle the non-stationary
acquired signals and to detect the first two flexional
modes of the chapel, whose frequencies are very close
to each other. The numerical model of the building,
581
its subdivision into sub-structures and the sensitiv-
ity analysis, performed in order to reduce the num-
ber of the subsequently updated parameters, are also
shown.
The last part of this study is focused on the devel-
opment of a stochastic model updating technique. The
adopted approach is known as “multiple-model” due
to the fact that it generates a huge number of models
to solve the problem of the impossibility to find an
optimal unique solution and to treat error and uncer-
tainty sources. The PGSL algorithm is implemented
in order to create these models and to assign parame-
ter values derived from their probability distributions,
progressively recalculated around those values which
minimize a target function. The application of this kind
of model updating procedure raises the problem of
information redundancy, which can be faced through
the adoption of some data mining techniques such
as the PCA and clustering. The final obtained results
are some models that can be considered as reasonable
candidates to represent the real structure.
The models obtained from the updating procedure
can be considered as the starting point for the following
phase of damage detection and structural health assess-
ment. It is now important to underline that these results
are obtained from a linear analysis while damage and
degradation effects are strictly non-linear phenom-
ena. Therefore, the subsequent numerical simulations,
applied to identify symptoms which can reveal pos-
sible damage scenarios on the construction, cannot
disregard these non-linear implications. The updated
models supplied by the multi-model approach repre-
sent a useful resource in order to estimate the health
condition of the building. Non-linear analyses per-
formed on the obtained models can reveal the effective
residual performance the construction is still able to
provide for each damage scenario.
REFERENCES
Beck, J.L., Yuen, K.V. & Katafygiotis, L.S. 2006. Unified
Probabilistic Approach for Model Updating and Damage
Detection. In Journal of Applied Mechanics, 2006, 73:
pp. 555–564.
Bonato, P., Ceravolo, R., De Stefano, A. & Molinari, F. 2000.
Use of cross Time-Frequency Estimators for the Structural
Identification in Non-Stationary Conditions and under
Unknown Excitation. In Journal of Sound and Vibration.
2000, 237: pp. 775–791.
Ceravolo, R., De Stefano, A. & Molinari, F. 2001 Develop-
ments and Comparisons on the Definitions of an Instanta-
neous Damping Estimators for Structures under Natural
Excitation. In Key Engineering Materials. 2001, 204–205:
pp. 231–240.
Kenigsbuch, R. & Halevi, Y. Model Updating in Structural
Dynamics: A Generalised Reference Basis Approach. In
Mechanical Systems and Signal Processing, 1998 12(1):
pp. 75–90.
Mottershead, J.E. & Friswell, M.I. 1993. Model Updating in
Structural Dynamics: a Survey. In Journal of Sound and
Vibration, 1993, 162: pp. 347–375.
Raphael, B. & Smith, I.F.C. 2003. A Direct Stochas-
tic Algorithm for Global Search. In Journal of
Applied Mathematics and Computation, 2003, 146:
pp. 729–758
582
Saitta, S., Raphael, B. & Smith, I.F.C. 2005. Data Mining
Techniques for Improving the Reliability of System Iden-
tification. In Advanced Engineering Informatics, 2005,
19(4): pp. 289–298.
Smith, I.F.C. 2005. Multi-Model Interpretation of Mea-
surement Data with Errors. CANSMART 2005, RMC,
Canada, 2005: pp 13–22.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

On the theory of the ellipse of elasticity as a natural discretisation

method in the design of Paderno d’Adda Bridge (Italy)

R. Ferrari & E. Rizzi

Dip.to di Progettazione e Tecnologie, Facoltà di Ingegneria, Università di Bergamo, Dalmine (BG), Italy

ABSTRACT: The Paderno d’Adda Bridge, Lombardia, northern Italy, is one of the very first great iron con-
structions designed through the practical application of the theory of the ellipse of elasticity, a graphical-analytical
method of structural analysis that was developed in the 19th century. It embeds a natural discretisation of the
structure into a series of elastic elements, treated then with standard tools of geometry of masses. In this work,
the application of such theory to the calculation of the parabolic arch of the bridge is inquired, attempting to
breathe, at the same time, the beauty of the architectonic and structural conception directly linked to that; later,
results are compared with much modern approaches that also consider now-available numerical discretisation
methods. A further, definite aim of this work is also that of trying to promote interest on the bridge, on its actual
state of conservation and future destinations. Not only it represents a true industrial monument and a living tes-
timony of the scientific and technological developments of the time but also a beautiful, effective achievement
of architecture and engineering through the methods of Strength of Materials.

1 INTRODUCTION nine bearings. Four of these supports are provided

1.1 The Paderno d’Adda Bridge
The Paderno d’Adda Bridge, also called San Miche-le
Bridge, is a metallic viaduct that crosses the Adda river
between Paderno and Calusco d’Adda to a height of
approximately 85 m from water, allowing to connect
the two provinces of Lecco and Bergamo, near Milano,
in the Lombardia region, northern Italy (SNOS 1889,
Nascè et al. 1984). At that location the river flows-
down from the exit of Lecco’s branch of Como’s lake
to the river Po through an impressive natural scenery
that even seems to have inspired celebrated paintings
by Leonardo (Fig. 1).
The main upper continuous beam, 5 m wide, is
formed by a 266 m long metallic truss supported by
Figure 1. Front view of the Paderno d’Adda Bridge,
1887–1889 (from up-stream; left bank Calusco, right bank
Paderno). A crossing train is visible inside the upper hori-
zontal continuous beam. Automotive and pedestrian traffic
runs on top.
by a marvellous doubly built-in parabolic metallic
arch of about 150 m of span and 37.5 m of rise. The
bridge shares its architectural style with similar arch
bridges built in Europe at the time (Timoshenko 1953,
Benvenuto 1981, Nascè et al. 1984), like e.g. that of
Garabit (1884, France, Eiffel and Boyer) and Maria Pia
(1887, Oporto, Eiffel and Seyrig), both doubly hinged
at the shoulders, and the Dom Luiz I (1886, Oporto,
Seyrig), doubly built-in as that of Paderno. The viaduct
was quickly constructed between 1887 and 1889 (thus
practically at the same time of the most celebrated
Tour Eiffel), to comply with the needs of the rapidly
growing industrial activities in Lombardia. It was built
by the Società Nazionale delle Officine di Savigliano
(SNOS), Cuneo, Italy, under the technical direction
of Swiss Engineer Giulio Röthlisberger (1851–1911),
the man whom the design of the bridge is normally
attributed to. He was formed at the Polytechnic of
Zürich, graduated in 1872 and got later in charge of the
Technical Office of the SNOS since 1885, for 25 years.
The bridge is still in service, with alternated one-
way automotive traffic, restricted to no heavy-weight
vehicles, and trains crossing at slow speed.
The bridge is designed through the graphical-
analytical methods of structural analysis that were
booming in the 19th century (Culmann 1880,
Timoshenko 1953, Benvenuto 1981). Specifically, it is
a remarkable application of the so-called theory of the
ellipse of elasticity (Culmann 1880, Belluzzi 1942).

583
This theory was originally conceived by Karl Culmann
(1821–1881) and then systematically developed and
applied by his pupil Wilhelm Ritter (1847–1906). It
represents a very elegant method for the analysis of
the flexural elastic response of a structure and is based
on an intrinsic discretisation of a continuous beam in
a series of elements, each with a proper elastic weight,
directly proportional to its length and inversely pro-
portional to its bending stiffness. The theory of the
ellipse of elasticity is based on the concepts of projec-
tive geometry, which lead to a correspondence between
the ellipse of elasticity of the structure and the central
ellipse of inertia of the distribution of the elastic weight
of the structure. This correspondence brings back the
problem of the determination of the flexural elastic
deformation of a beam to a problem of pure geome-
try of masses, of more convenient solution and direct
interpretation in terms of the design of the structure.
1.2 Main technical features of the bridge
The main technical features of the bridge are reported
in details in Nascè et al. (1984), which is, to our knowl-
edge, the most comprehensive publication, and one of
the very few, concerning the bridge. We rely very much
on this very valuable contribution and on the Techni-
cal Report (SNOS 1889) that was originally issued
at the time of the first try-out. Here, the essential
characteristics are reported.
The 266 m long upper flyover is made by a continu-
ous box girder with nine equally-distributed supports,
at 33.25 m distance from each other. Four of the sup-
ports are sustained by a big parabolic metallic arch; two
of them bear directly on the same arch’s masonry abut-
ments (made with Moltrasio masonry, with Baveno
granite coverings); a seventh, on the Calusco bank,
rests on a smaller masonry foundation placed between
the arch shoulder and the higher bridge supports; the
last two, in masonry work as well, are the two direct
beam bearings at its two ends, on top of the two river
banks. The four piers resting on the arch are placed
symmetrically, in between keystone, haunches and
shoulders of the arch. The inner side of the beam girder,
on which the railway is located, runs at about 255.00 m
on the sea level (osl); the rails are placed at 255.45 m
osl, the upper road at 261.75 m osl. The main verti-
cal longitudinal trussed beams of the upper continuous
girder are 6.25 m high and placed at a respective trans-
verse distance of 5.00 m, leaving a free passage for
the trains of 4.60 m. They are composed of two main
T-ribs connected by a metallic truss. The upper-level
road is 5.00 m wide and includes also two additional
cantilever sidewalks, each 1 m long, with iron parapets
1.50 m high.
The big arch is composed by two couples of sec-
ondary inclined arches. Each couple is formed by two
arches posed at a respective distance of 1 m and lay-
ing symmetrically to a mean plane inclined of about
584
±8.63◦ to the vertical. The parabolic axis of the arch
has a span of 150.00 m and rise of 37.50 m in the
inclined plane; the transverse arch’s cross section is
4.00 m high at the keystone and 8.00 m high at the
abutments (i.e. in the same 1:2 ratio between rise and
half span). The two mean inclined planes of the arches
are located at a distance of 5.096 m at the keystone and
16.346 m at the shoulders. The wall of each composing
arch is also a truss structure with two main T-ribs con-
nected by vertical and inclined bars. The two couples
of twin arches are gathered together by two transverse
brace systems located at the extrados and at the intra-
dos of the arch’s body. In essence, the resulting cross
section of the main parabolic arch supporting the hor-
izontal beam is trapezoidal, with variable, increasing
cross section from the crown to the shoulders. This,
and specifically the inclination of the twin arches, is a
beautiful key feature of Röthlisberger’s conception of
the bridge, in view of counteracting effectively wind
and transverse horizontal actions in spite of the con-
siderable slenderness of the structure. The arch cross
section at the impost is inclined of 45◦ to the hori-
zontal, so as the local tangent to the parabolic axis
of the arch to the vertical. The vertical bridge piers
that sustain the upper continuous beam are made by
eight T-section columns, linked to each other by a brace
system with horizontal bars and St. Andrew’s crosses
and, on top, by transverse beams that directly serve as
supports for the bearing devices of the upper beam.
For inspection and maintenance purposes a 1 m large
boardwalk is provided into the body of the arch and,
inside the bridge piers, a system of ladders along their
height. The bridge is a riveted wrought iron structure of
about 2600 t of metals, with near 100000 rivets just in
the arch.
1.3 Aim of this work
In this work, which in its main part largely refers to
a study developed in a Laurea Thesis (Ferrari 2006),
a detailed analysis of the SNOS Report (1889) is pre-
sented. The point of view here is the following: inquire
the application of the theory of the ellipse of elastic-
ity to the calculation of the bridge, breathe the beauty
of the architectonic and structural conception directly
linked to that, compare results with modern structural
approaches that also consider now-available numerical
discretisation methods.
After a careful review of the Report by the SNOS,
a full 3D truss Finite Element model of the arch of
the bridge has been elaborated, based also on direct
inspections of the bridge and on the screening of
the marvellous original drawings that are guarded
at the Archivio Storico Nazionale di Torino. Dif-
ferent loading conditions have been considered and
results compared with those reported in the SNOS
Report, showing the remarkable accuracy of the
adopted graphical-analytical methods and allowing to
experience the unrepeated beauty of the original anal-
ysis with respect to rather impersonal computer struc-
tural analysis. Moreover, the model that has been put
in place shows promise for possible further analyses
that could inspect other behaviours of the bridge, as for
example dynamical and inelastic, also connected to the
present and future state of conservation of the struc-
ture. These aspects are left for further developments
of the present study.
The paper is organized as follows: Section 2 pro-
vides a short account on the theory of the ellipse
of elasticity; Section 3 reports its application to the
structural analysis of the arch of the bridge; Sec-
tion 4 presents an independent validation of the orig-
inal design results with present analytical-numerical
methods.
2 ON THE THEORY OF THE ELLIPSE OF
ELASTICITY
2.1 Fundamentals
The theory of the ellipse of elasticity can be consid-
ered as a main icon of the so-called Graphical Statics,
the discipline which often characterised the resolving
approach of practical design problems during the 2nd
half of the 19th century. It represents a very elegant
and practical method for the analysis of the flexural
response of an elastic structure. It is based on an intrin-
sic discretisation of a continuous elastic problem. This
theory is basically associated to the two outstanding
figures of Culmann and Ritter, but also of people, like
Giulio Röthlisberger, that were formed at the time at
the Polytechnical Schools in Europe and that became
later structural engineers and designers and largely
contributed in the practical and effective application
of the method.
The theory is based on the following main hypothe-
ses (we refer here to the Italian text by Belluzzi 1942,
which reports results from the technical literature
of the time, basically ascribed to the two names of
Culmann and Ritter): (a) linear elastic behavior of
the material and the structure, which leads to the
proportionality between acting forces and (reversible)
displacements provoked by them (property that in turn
implies the validity of the principle of superposition
of effects); (b) existence of the ellipse of elasticity,
referred to a section of a structure; (c) correspondence
between the latter and the central ellipse of inertia of
the distribution of the so-called elastic weight of the
structure.This correspondence transforms the problem
of the determination of the elastic response of a con-
tinuous structure to a task of pure geometry of masses.
The latter can be feasibly handled by taking advantage
of the assumed discrete character of the distribution of
the elastic weight and is endowed with a visible inter-
pretation of the elastic performance of the structure,
585
Figure 2. Representation of a force R applied along line of
action r to a beam section A and the centre of rotation C of
A. C is the symmetric of C with respect to the centre S of the
ellipse of elasticity of beam AB referred to section A.
in view of its conception and design. Furthermore, the
ellipse itself may actually play the role of an hidden,
underlying, graphical construct. Indeed, the proper-
ties of projective geometry that are attached to that
allow the elastic solution of the structure even without
the explicit drawing of the ellipse itself. The methods
are also said graphical-analytical because, in practice,
main technical steps that are framed on the graphi-
cal constructions may be carried-out analytically, by
working-out formulas that arise from the inspection
of the drawings (Belluzzi 1942).
The concept of the ellipse of elasticity referred to a
section of an elastic structure is achieved by inspect-
ing the correspondence existing between the line of
action r of a force R applied to a section A of a gen-
eral, curvilinear, elastic beam (with little curvature and
continuously-varying cross section) and the centre of
rotation C of the same section (Fig. 2).
In particular, refer to a planar beam acted upon
by forces laying in the same plane and cross sections
of the beam that, during the beam’s deformation, are
assumed to remain plane and perpendicular to the geo-
metric axis, also deforming in its original plane. The
theory states that there exists an involutory relationship
between the line of action r and the centre of rotation
C of the section. Moreover, the ellipse of elasticity
is the fundamental real conic of the polarity existing
between the line of action r and the point C , which is
the symmetric of C with respect to the centre S of the
ellipse. In other words, the ellipse of elasticity can be
defined as the fundamental conic with respect to which
the lines of action r and the respective centres of rota-
tion C correspond to each other through an antipolarity
relationship.
The determination of the central ellipse of inertia of
the distribution of the elastic weight of the structure,
which coincides with the ellipse of elasticity above,
is linked first to the definition of the general concept
of elastic weight and then to the quantification of its
distribution for the structure under consideration. The
concept of elastic weight goes as follows. If on a sec-
tion A of a beam, a moment M acts in the plane which
contains the geometric axis, it causes a rotation φ of
A, around the centre S of the ellipse of elasticity. The
rotation is proportional to the applied moment M as:
where G represents the so-called elastic weight of the
beam. Thus, G can be defined as the angle of rota-
tion φ that is caused by the application of a unitary
moment M = 1; it depends on the beam’s geometrical
and physical properties; it gives a global measure of
the beam’s aptitude to deform. In case of a straight
cantilever beam of length l loaded by a moment M
at its free end, composed by a linear elastic material
with Young’s modulus E and endowed with a constant
cross section with moment of inertia J with respect to
the axis perpendicular to the beam’s plane, it turns out
that, referring to the case of flexure of de Saint Venant,
the rotation of the free end is:
This relation clearly illustrates the physical meaning of
G as the global parameter that expresses the flexural
elastic deformability of the structure.
Now, the idea arises of thinking at the structure as
the assembly of a series of discrete elastic elements of
length s, each with a proper elastic weight
such that the total elastic weight of the structure is
represented by the discrete distribution of these elas-
tic weights. It is possible to demonstrate that such
a sought distribution of G is univocally known only
for a statically-determined structure, whereas for a
statically-undetermined structure the distribution of
elastic weights is not univocally defined. This is not
surprising, due to the redundancy of equilibrium in an
hyperstatic system. Despite this, an hyperstatic struc-
ture can still be solved, via the Forces Method (with
hyperstatic quantities as unknown), through the super-
position of effects on underlying isostatic structures
and imposition of the corresponding compatibility
conditions. As the underlying isostatic structure can
also be analysed with a univocally-defined distribu-
tion of elastic weights, such distribution can also be
used to solve the original hyperstatic structure. Thus,
indirectly, its ellipse of elasticity can in essence be
determined, so the corresponding ellipse of inertia of
the distribution of elastic weights.
This allows one to write the so-called “theorems
of the theory of the ellipse of elasticity” (Belluzzi
1942), as a function of the properties of the distri-
bution of elastic weights. For example, the rotation φ
586
Figure 3. Representation of the quantities reported in Eq. (4)
for the calculation of the rotation and displacements of ter-
minal section A caused by a force Q applied to the same
section.
and horizontal and vertical displacements dx , dy of a
terminal beam section A caused by a force Q applied
to the same section along line of action q (Fig. 3), can
be nominally written as follows:

where G = G represents the total elastic weight of
the structure; Sq , Jxq , Jyq the static moment of G with
respect to q and the centrifugal moments of inertia of
G with respect to q and axis x, and q and axis y. These
parameters depend only on the distribution of elastic
weights and on the position of the applied load Q, and
can be expressed as a function of the quantities xS , yS ,
uS , uX , uY depicted in Fig. 3.
The point S in Fig. 3 represents the centre of grav-
ity of the elastic weights of the structure; the points
X,Y represent the antipoles of the reference system
axes x, y with respect to the central ellipse of inertia
of the elastic weights. It is then apparent that, once
the position of points S, X, Y and total elastic weight
G are found, the elastic response of the structure is
determined. The coordinates xS , yS , xY , yX defining
the position of these points can be evaluated by stan-
dard calculations of geometry of masses, once given
the discrete distribution of elastic weights.
2.2 Application to a doubly built-in
parabolic arch
In the SNOS Report (1889), the remarkable applica-
tion of the theory of the ellipse of elasticity to the
analysis of the arch of the bridge refers to the deter-
mination of the elastic response of a parabolic arch,
built-in at the two extremities, which is subjected to a
vertical load P (that can be put equal to 1) and acting
in an arbitrary position along the arch, at a horizontal
distance a from left extreme A (Fig. 4).
First, the position of the line of action of the reac-
tion A needs to be determined. This can be solved by a
Figure 4. Calculation of line of action LO of left reaction
A; the segments FL and VO that the left reaction A locates on
the vertical lines from A and on load P, below and above the
horizontal line from centre S are determined.
Figure 5. Calculation of line of action HK of left reaction A;
segments a, b, c to be drawn at points S,X,Y are determined
analytically. Points H,K are then located, so the direction of A.
graphical-analytical procedure. Figs. 4 and 5 represent
two ways to solve this problem (SNOS 1889, Ferrari
2006). Underlying to these constructions lay the com-
patibility conditions φ = 0, dx = 0, dy = 0 for the left
built-in constraint at terminal section A. These can be
worked-out from Eq. (4), leading to:
where A is now taking the role that force Q had in
(4) and the superposition of effects is considered with
load P, so that uS = a, uX = b, uY = c (a, b, c denote
segments used below in Fig. 5) are quantities similar to
those entering Eq. (4), but related to load P at position a
from A. They can be calculated analytically as follows,
given the distribution of elastic weights:
Once scalars uS = a, uX = b, uY = c are evaluated, the
construction in Fig. 5 determines the position of A.
587
Figure 6. Equilibrium requires that the left and right reac-
tions A and B form with load P a closed polygon of forces.
Segments V = OQ and V  = QP represent the vertical com-
ponents of A and B, with V + V  = P; segment H = TQ their
horizontal component.
Alternatively, one may also proceed as sketched in
Fig. 4, by calculating parameters µ, ν and segments
FL, VO, with same results. The last relations in (5)
also give the magnitude of reaction A, given P.
Once the position of the lines of action of left built-
in reaction A, and, by equilibrium, of right reaction
B are known, the value of their vertical components
V , V  and (common) horizontal component H can be
found as follows:
where quantities a, a , f, f are represented in Fig. 6.
Following similar arguments, it is possible to derive
the in-plane arch’s deflections dy at any position x of
the arch, for the different locations a of the load P
acting on the arch (SNOS 1989). The equation that
gives the mean parabolic line at the arch is taken as:
where f and l represent now the rise and span of the
arch. The profile of the arch is symmetric with respect
to the vertical axis x = l/2 at half span. The formulas
for dy obtained by the SNOS with this graphical-
analytical procedure correspond indeed to those that
may be obtained by the application of theVirtual Works
Principle (VWP):
where H , V , M represent the “components” of the
reaction force A (built-in moment M is positive clock-
wise) and C = EJdx/ds = cost is a quantity related to

Figure 7. Discretisation of the half arch in 14 elements
with x = cost, with indication of the geometrical quanti-
ties useful to determine the corresponding elastic weights
(quotes in m).
the bending stiffness and local inclination of the arch.
C is assumed by the designer constant along the arch,
since while J decreases from the shoulder to the key-
stone, the ratio dx/ds between horizontal projection dx
of infinitesimal element length ds of the arch and ds
itself increases (SNOS 1889).
3 STRUCTURAL ANALYSIS OF
THE BRIDGE
3.1 Explicit application of the theory of
the ellipse of elasticity to the analysis
of the arch of the bridge
Going to the explicit definition of the distribution of
elastic weights made by the SNOS (1989), all the cal-
culations reported there were developed considering
an arch which is the projection on their mean inclined
plane of one of the two couples of inclined parabolic
arches that are placed symmetrically to the vertical
longitudinal median plane of the bridge. Such arch is
placed on a plane inclined of about α = 8.63◦ to the
vertical (such that sin α = 0.15) and consists of a truss
beam with parabolic axis of 150 m of span and 37.5 m
of rise in such plane, having extrados and intrados lines
both described by parabolic functions so as to deter-
mine a cross-high of the arch of 4 m at the keystone
and of 8 m at the abutment. On the basis of this model,
the elastic weights of the structure have been calcu-
lated according to a symmetric structural discretisation
with 28 elements of different s extensions as reported
in Fig. 7.
In the Report, the procedure adopted to calculate
the elastic weights G = s/EJ is not really appar-
ent. An attempt of careful analysis is provided in
Ferrari (2006). Also, the elastic weights are actually
588
determined without the constant proportionality fac-
tor E = 17000000 t/m2 , Young’s modulus of the iron,
i.e. G  = s/J . The coordinates xS , yS , xY , yX and
total elastic weight G  = EG are finally found as:
The constant C in Eq. (9) is also evaluated in
55539000 t · m2 .
Anyway, tables are presented in which the left reac-
tion and the deflections of the arch are determined for a
unitary load located at the various elastic elements. In
practice, these influence coefficients, which are later
used for the design of the truss members, are deter-
mined by a true application of the theory of the ellipse
of elasticity as applied to the arch.
3.2 Evaluation of the loads
The SNOS Report analyses independently, one by
one, the various loadings on the arch, for subsequent
superposition of effects:
– permanent weight of the arch;
– permanent weight of the upper girder beam, of the
bridge piers and vertical actions induced by the
wind acting on the girder beam;
– accidental vertical load on the upper girder beam;
– temperature effects and compression on the arch
due to the horizontal thrust H ;
– direct horizontal wind action on the arch.
For each listed item, the SNOS reports the calculation
of the stresses in the various arch’s elements, as well
as the final value that arises by the superposition of
effects.
3.3 Dimensioning of the structural elements
The Report by the SNOS provides the calculation
of the stresses in the various bar elements of the
arch and at the stone abutments as compared to the
target admissible values that are summarised below.
This is done for the final geometries of the struc-
tural members. On the other end, the Report does not
provide specific information about the design proce-
dure that has lead, through pre-dimensioning, to such
final structural dimensions.As the overall architectural
and structural conception of the entire viaduct, these
phases seem to be linked to the engineering practice
and experience of the designer with the standards of
metallic carpentry in use at the time.
The material employed in the structural members
of the bridge is a wrought iron, with very low car-
bon percentage of about 0.01% (Nascè et al. 1984).
The admissible stresses are taken differently for each
structural component of the bridge: for the main upper
and lower arch’s ribs 6.0 kg/mm2 ; for the vertical and
 Table 1. Arch’s vertical deflections calculated at design
stage for four different loading tests (SNOS 1889, p. 71).
Negative values indicate downward displacements.

Pier I Pier II Vertex Pier III Pier IV

Deflections (m) (m) (m) (m) (m)

Test I +0.0033 −0.0010 −0.0045 −0.0108 −0.0066

Test II −0.0016 −0.0080 −0.0097 −0.0080 −0.0016
Test III −0.0040 −0.0101 −0.0094 −0.0022 +0.0027
Test IV −0.0068 −0.0064 −0.0010 +0.0035 +0.0031

Figure 8. Scheme with four test loading configurations,

with indication of the four piers resting symmetrically on
the arch (view from down-stream; Paderno left side, Calusco Table 2. Arch’s vertical deflections measured in situ at the
right side). Pier III is at half length of the upper continuous try-out for the corresponding loading tests (SNOS 1889,
beam. p. 71).

Pier I Pier II Vertex Pier III Pier IV

diagonal bars linking the inferior and superior ribs of Deflections (m) (m) (m) (m) (m)
the arch 6.0 kg/mm2 ; for the elements that compose
the transverse bridge’s brace system counteracting Test I +0.0038 +0.0001 −0.0044 −0.0106 −0.0056
Test II 0 −0.0079 −0.0080 −0.0102 −0.0012
wind 4.2 kg/mm2 . For each arch’s stone abutment, Test III −0.0016 −0.0102 −0.0073 −0.0014 +0.0025
the allowable compression stress is assumed in nearly Test IV −0.0063 −0.0058 −0.0007 +0.0042 +0.0026
31 kg/cm2 . Stresses are all found safely below these
target values.
The executive drawings, including the various cross
sections of the structural elements, the details of the with speed up to 45 km/h. Transverse oscillations at
riveted joints and part of the graphical/analytical calcu- the keystone were recorded in less than 3.142 mm and
lations are reported in 147 marvellous drawing tables. vertical deflections in less than 10 mm. According to
They are still in an excellent state of conservation and Nascè et al. (1984) a 2nd, final, try-out took place in
show the meticulous design of the complex structure June 1892, with different modalities and applied loads
of the bridge. Often, the calculus is integrated in the but with similar results.
drawing itself, which testimonies the beautiful, inti-
mate link between conception, structural analysis and
executive design. 4 VALIDATIONS OF THE REPORTS’ RESULTS
BY ANALYTICAL-NUMERICAL METHODS
3.4 Viaduct’s tests
4.1 Analysis of the elastic arch by the VWP
The first viaduct’s tests took place from 12th to 19th
May 1889. The measured deflections for loading con- In order to compare the results presented by the SNOS,
ditions conforming to those considered at design stage the hyperstatic scheme of the doubly built-in arch was
were compared to the corresponding theoretical val- solved by the Virtual Works Principle. Such analy-
ues. To this purpose, the tests were carried out in two sis allowed to notice some little inconsistencies in the
moments: first, the different road loads were obtained SNOS results. First, the reactions A and B due to P = 1
by deposition of gravel on the upper deck; second, with at distance a from left support A were evaluated and
a uniformly distributed gravel load of 3.9 t/m all over compared to the Report’s results. The match was not
the road, 6 locomotives with tender, each of 83 t of perfect, particularly for the bending moments at the
weight, corresponding to a distributed load of 5.1 t/m, extremes of the arch. The differences did not seem to
were displaced on the railway track following four be due to transcription errors, since the values reported
different loading configurations (Fig. 8). Tables 1–2 by the SNOS look coherent with formulas and tables
below list the calculated and measured vertical deflec- presented within the text. Second, further validation
tions. The good overall agreement doubtlessly shows concerning the arch’s deflections was attempted, since
the validity of the structural approach used by the they also did not seem to be calculated through the
SNOS at design stage, in species the effectiveness of reactions that are listed in the Report (Ferrari 2006).
the theory of the ellipse of elasticity as applied to the However, the final values reported by the SNOS cor-
explicit calculation of the arch of the bridge. respond, to a good degree of accuracy, to the present
A final test was also run with 3 locomotives and results obtained by the VWP (Tables 3–4). The values
30 wagons, loaded by gravel, for a total weight of the reported inTables 3–4 refer to a precise load configura-
convoy of 600 t. The train was let running 3 times, tion (1st distribution, see Section 4.2.2 andTable 5) and

589
Table 3. Bending deflections listed in SNOS (1889, p. 59). 4.2 Structural analysis of the arch by the FEM
Load Pier I Pier II Pier III Pier IV Currently, an attempt is made of building a full FE
Point (m) (m) (m) (m) model of the bridge. So far, a truss mesh of the arch of
the bridge has been developed and appropriate loading
Pier I −0.0089 −0.0029 +0.0071 +0.0042 conditions has been considered for validation of the
Pier II −0.0011 −0.0042 +0.0026 +0.0030 previous results. The FE analysis has been run with
Pier III −0.0004 −0.0003 +0.0005 +0.0001 the commercial code ABAQUS® .
Pier IV +0.0001 +0.0001 −0.0000 −0.0001
Total −0.0103 −0.0073 +0.0102 +0.0071
4.2.1 Structural model
The FE model consists of a 3D truss frame, reproduc-
ing as much as possible the actual arch geometry. It
Table 4. Bending deflections calculated by the VWP. consists of two planar parabolic trusses referring to
the in-plane geometry of the arch (see Fig. 7), placed
Load Pier I Pier II Pier III Pier IV in two inclined planes (of ±8.63◦ to the vertical). The
Point (m) (m) (m) (m) inclined planes are placed at a distance of 5.096 m from
each other at the axis of the arches at the keystone.
Pier I −0.00893 −0.00298 +0.00703 +0.00422
Pier II −0.00126 −0.00425 +0.00255 +0.00297 The truss nodes are linked to each other through a
Pier III −0.00038 −0.00033 +0.00055 +0.00016 reticular system that corresponds to the actual brac-
Pier IV +0.00006 +0.00010 −0.00004 −0.00013 ing of the arch. To each bar of the model, a cross
Total −0.01051 −0.00746 +0.01009 +0.00722 section with equivalent geometrical characteristics is
attributed (area, principal moments of inertia, torsional
stiffness). Some approximations were made as regards
to the section’s attribution to the superior and inferior
Table 5. Total vertical deflections (1st distribution) reported arch ribs, which are made with a variable number of
by SNOS (1889, p. 60) and present results by the FE model. longitudinal plates. Also, at the node junctions, there
are additional reinforcing plates, for local stiffening.
Vertical loads SNOS deflections FE deflections
Point (t) (m) (m)
In spite of this, the model has been simplified with
bars of constant average cross section. The model is
Pier I +340.6 −0.0120 −0.015903 comprised of 752 beam elements and 266 nodes. Built-
Pier II +144.0 −0.0113 −0.013847 in constraints are imposed at the nodes of the arch
Pier III −18.5 +0.0062 +0.010322 shoulders.
Pier IV +4.9 +0.0054 +0.007222
4.2.2 Obtained results
First trial loading cases considered a unitary load (1 t)
regard the vertical deflections (due to pure bending) at applied at the piers and at the keystone, in view of
the four piers due to forces acting at their locations. verifying the order-of-magnitude agreement with the
The reason of the discrepancies above may be due deflections that were listed in the Report and calculated
to the fact that the SNOS might have used in the final here by the VWP. After these preliminary checks, the
Report also data concerning a preliminary project. deflections were evaluated for five given load distribu-
Indeed, the bridge’s layout was slightly modified in tions with vertical loads acting at the four piers (SNOS
the executive project, due to new requirements on the 1889, p. 58–62).
railway trace that were posed by the Strade Ferrate A sample of these outcomes is given in Figure 9
Meridionali, after checks on the Adda’s banks (Nascè for the 1st load distribution already considered in
et al. 1984). It is therefore possible that some specific Tables 3–4, with main results summarized in Table 5.
data were referring to a previous project, while impor- Such distribution considers spans 2–3 charged by a
tant global quantities, such as the deflections caused uniformly distributed load of 9 t/m, leads to a max-
by external actions, were indeed corresponding to the imum pressure on Pier I and is somehow similar to
final one. As a matter of fact, even if this Report has Test IV considered in the try-outs (Fig. 8). The maxi-
been probably conceived to present to a general audi- mum compressive axial force at the intrados of the left
ence the main steps of the calculations, including a shoulder is found in 377.0 t, in good agreement with
very valuable account on the theory of the ellipse of the value 751.6/2 = 375.8 t calculated by the SNOS
elasticity and on its explicit application to the analysis (1889, p. 62). Agreement is also found for the hor-
of the arch, it is doubtlessly very concise and it obvi- izontal thrust H and vertical reactions V , V  at the
ously presents concepts and practical considerations shoulders, with H = 185.8 t, V = 417.2 t, V  = 53.8 t.
that may not be directly apparent to the contemporary The FE outcomes confirm, to a quite good degree of
reader. accuracy, the SNOS results, showing the true potential

590

Figure 9. FE deformed configuration of the arch for the
1st load distribution reported in Table 5 (amplification
factor = 500).
of the theory of the ellipse of elasticity as applied to
the elastic analysis of the arch of the bridge.
5 CLOSING REMARKS
An attempt to scrutinise in details the SNOS
Report (1889) has been made in view of breathing the
original conception of the design of the bridge and its
specific calculation through the elegant method of the
theory of the ellipse of elasticity. The intrinsic discreti-
sation in elastic elements is remarkable and makes a
natural connection with discretisations that can now
be provided by FE codes. The beauty of the executive
project can be totally appreciated only by the comple-
mentary screening of the 147 drawing tables of the
591
bridge. All this shows the ingenuous, effective and
beautiful approach to the design of the bridge. One
might think at this once contemplating the giant still
standing there silently, serving since almost 120 years
of duty, with an actual state of conservation that actu-
ally poses serious questions about its future survival.
We should take care of it.
REFERENCES
Belluzzi, O. 1942. Scienza delle Costruzioni. Bologna:
Zanichelli.
Benvenuto, E. 1981. La Scienza delle Costruzioni e il suo
Sviluppo Storico. Firenze: Sansoni.
Culmann, K. 1880. Traitè de Statique Graphique. Paris:
Dunod.
Ferrari, R. 2006. Sulla concezione strutturale ottocentesca
del ponte in ferro di Paderno d’Adda secondo la teoria
dell’ellisse d’elasticità. Laurea Thesis in Building Engi-
neering. Advisor E. Rizzi, Università di Bergamo, 228
pages.
Nascè, V., Zorgno, A.M., Bertolini, C., Carbone, V. I.,
Pistone, G., Roccati, R. 1984. Il ponte di Paderno: sto-
ria e struttura – Conservazione dell’architettura in ferro.
Restauro, Anno XIII, n. 73–74, 215 pages.
Società Nazionale delle Officine di Savigliano 1889.Viadotto
di Paderno sull’Adda (Ferrovia Ponte S. Pietro-Seregno).
Torino: Tip. e Lit. Camilla e Bertolero.
Timoshenko, S.P. 1953. History of Strength of Materials. New
York: McGraw-Hill.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Lower-bound analysis of masonry vaults

P. Block & J. Ochsendorf

Building Technology Program, MIT, Cambridge, MA, USA

ABSTRACT: This paper applies Thrust-Network Analysis, a three-dimensional computational method for
obtaining lower-bound solutions of masonry vaults with complex geometries. The method extends thrust-line
analysis to three-dimensional problems by finding equilibrium force networks within the vault’s geometry, repre-
senting possible paths of the compression forces. Through two case studies, this paper demonstrates the potential
of the method as a powerful tool for understanding, visualizing and exploring the equilibrium of compression-
only structures. First, an analysis of a series of groin and quadripartite rib vaults investigates the interrelationship
between different parameters and the range of possible equilibrium solutions of these vaults. A second case study
analyzes the fan vaults of King’s College Chapel in Cambridge, England.

1 INTRODUCTION sufficient friction exists. For further reading on lower-

Medieval vault builders explored three-dimensional
equilibrium, creating complex forms carefully bal-
anced in compression. The structural properties of
these sophisticated forms are still poorly understood
because of a lack of appropriate analysis methods, i.e.
methods relating stability and form.
The safety assessment of masonry structures is pri-
marily a problem of geometry and stability, rather
than of material stresses (Heyman 1995). Structures
in unreinforced masonry work in compression, and the
tensile capacity of the stone and mortar can be consid-
ered as negligible. These considerations then demand
new tools in order to understand how these structures
work and why they are able to stand for centuries. Lin-
ear elastic analysis using finite element methods is
mainly concerned with stresses, and is not appropriate
for historic structures in masonry.
1.1 Lower-bound analysis and the safe theorem
Heyman (1966) introduced the safe theorem for
masonry structures, also known as the lower bound
theorem. Put simply, a vault in unreinforced masonry
will stand if a network of compression forces in equi-
librium with the applied loads can be found which fits
within the section of the structure. This solution is a
possible lower-bound solution. Although we will never
know how exactly the masonry vault is standing, this is
not necessary. The safe theorem guarantees that as long
as we can demonstrate one way that the structure could
stand, i.e. could be in equilibrium with the external
forces, then it is safe. This approach initially neglects
sliding, which can be checked afterwards to ensure that
bound analysis for unreinforced masonry structures
see Heyman (1995) and Huerta (2001, 2004).
1.2 Thrust-line analysis and graphic statics
Thrust-line analysis is a particularly powerful method
for understanding and exploring the range of lower-
bound equilibrium solutions of compression-only sys-
tems, such as unreinforced masonry structures. It
visualizes the relative stability of these structures by
showing the paths of the resultant compressive forces
throughout the structure and, for two-dimensional
problems, suggests possible collapse mechanisms
(Ochsendorf 2002, Block et al. 2006b). However, it
is primarily a two-dimensional technique and is there-
fore most appropriate for the analysis of arches, flying
buttresses or any structure which can be reduced to a
sectional analysis.
Graphic statics can be used to compute thrust lines
(Fig. 1). The main advantage of using graphical anal-
ysis is that the funicular polygon visually represents
the forces in the system. Examples of graphic statics
for fully three-dimensional problems were performed
and demonstrated by Föppl (1892). These analyses
were difficult to perform and limited to statically
determinate problems.
2 EQUILIBRIUM ANALYSIS FOR VAULTS IN
UNREINFORCED MASONRY
2.1 Pseudo-3D equilibrium analysis methods
In order to analyze three-dimensional structures using
the same intuitive methods discussed above, the

593

Figure 1. A pseudo-3D analysis of a gothic rib vault using
graphic statics (Wolfe 1921). The web of the vault is cut into
strips which are analyzed as 2D arches. The main ribs bring
the forces from those arches down to the supports.
analyst typically must slice the structure, reducing it
to a combination of two-dimensional problems. In this
way, structural behavior is reduced to a combination of
arch actions. This process obviously does not capture
the full three-dimensional behavior of the structure and
must rely heavily on the chosen discretization. These
limitations have been the main reasons why thrust-
line analysis has not been used extensively for the
assessment of complex 3D structures.
Wolfe (1921) demonstrated how by slicing up a
structure a global pseudo-3D analysis can be done by
combining local thrust lines (Fig. 1). This methodol-
ogy was entirely manual and quickly becomes tedious
(Boothby 2001). Block et al. (2006b) proposed a
method to produce models which contain the graphical
construction but which are parametric and interac-
tive, reducing the tedious iterative nature of traditional
graphic statics. The models demonstrate the range of
possible solutions and can be used to investigate the
effects on the global stability of the vault of varying
different parameters such as buttress thickness, arch
thickness, level of fill, etc. Smars developed in his PhD
thesis (2000) computational tools to perform such a
pseudo-3D analysis in an automated fashion, starting
from the actual measured geometry of the vaults.
2.2 Fully 3D equilibrium analysis methods
The main problem with three-dimensional equilib-
rium analysis for masonry vaults is that they are
highly indeterminate structures. Antoni Gaudí’s physi-
cal form-finding process for the church of the Colonia
Guell can be used to explain this. First, before starting
594
to construct a hanging string model, Gaudi had to
decide on a suitable force pattern topology to represent
the structural action of the vaults. Then, after choos-
ing the structural logic, it is still challenging to control
or even predict the final shape, since the equilibrium
of each string influences the equilibrium of the entire
network. It is a tedious, iterative process of adjusting
and refining.
New form-finding programs which explore hanging
models in the virtual world, based on dynamic relax-
ation, such as Kilian’s CADenary tool have to deal with
the same issues (Kilian and Ochsendorf 2005). It is
very hard to control and predict how the final shape of
the compression network will look like if local changes
are being made or a string model is being assembled
and hung under gravity.
This is true for both physical and virtual string
models as for graphical methods.
In order to analyze a three-dimensional indetermi-
nate system these unknowns need to be understood
and controlled. This can be achieved by describing the
problem as an optimization problem. The equilibrium
requirements then are formulated as a set of constraints
which have to be satisfied while optimizing a certain
objective function.
O’Dwyer (1999) implemented optimization meth-
ods to investigate masonry vaults by finding possible
compression-only force networks which are entirely
contained within the boundaries of the vault. Recently,
the problem of controlling a virtual hanging strings
network has been elegantly and efficiently been imple-
mented within a similar optimization framework by
Andreu et al. (2007).A hanging string network is found
which fits within the inverted geometry of the vault to
be analyzed.
The following section briefly summarizes Thrust-
Network Analysis, a new fully three-dimensional
extension to thrust-line analysis based on projec-
tive geometry, duality theory and linear optimization
(Block and Ochsendorf 2007). Examples in Sections
4 and 5 will demonstrate its value for the analysis of
vaulted structures in unreinforced masonry.
3 THRUST NETWORK ANALYSIS
3.1 Motivation
Our goal was to develop a three-dimensional version
of thrust-line analysis similar to previous applications
using interactive graphic statics. This means that the
following features should be preserved:
– a graphical and intuitive representation of the forces
in the system; and
– an interactive exploration of the range of equi-
librium solutions bounded by a minimum and
maximum thrust.
In order to cope with the challenges of the high degree
of indeterminacy of three-dimensional problems, as
discussed in section 2.2, we want to be able to:
– identify and control the many unknowns (degrees
of freedom); and
– negotiate between the unknowns by formulating
an optimization problem with different objective
functions.
Therefore we want to explore the impact of differ-
ent assumptions about the force patterns, internal
force distributions, boundary conditions or loading
conditions.
3.2 Methodology
Thrust-network analysis extends O’Dwyer’s (1999)
work on funicular analysis of vaulted masonry struc-
tures by adding the reciprocal relationship between
the geometry and the in-plane internal forces of net-
works (Williams 1986), which was first described
by Maxwell (1864). Figure 2 demonstrates this
relationship: the internal force equilibrium of one grid
Figure 2. The two plane grids have a reciprocal relationship
as defined by Maxwell. The equilibrium of a node in one of
them is guaranteed by a closed polygon in the other and vice
versa. The labeling uses Bow’s notation (Bow 1873).
Figure 3. The input for the Thrust Network Analysis method: (a) the boundaries, the intrados and extrados of the vault; (b) a
possible force pattern  defined on the horizontal plane; (c) the reciprocal force diagram ∗ automatically produced from ;
and the weight associated per node coming from the vault’s self-weight and other imposed loads.
595
is represented by the geometry of the other grid and
vice versa.
Thrust-network analysis is developed for loading
conditions where all forces are parallel to each other,
such as gravitational loading. Note that this method
is therefore appropriate for historic structures in unre-
inforced masonry since the dominant loading is self-
weight. It is important to note that in this case the
external forces do not appear in the projection of the
system on the plane perpendicular to the direction
of the forces. This means that a plane force diagram
can be produced which represents the equilibrium in
that plane of the system independent of the externally
applied loads. As a result the force diagram is scale-
less since the external forces which typically give scale
to the force diagram are missing.
3.3 Overview of the main steps
The set-up of the program is summarized below. Block
and Ochsendorf (2007) presents details on the problem
formulation and solving procedures.
(a) Defining a solution envelope:
The compression-only solutions must lie within
given boundaries defined by an intrados and an
extrados (Fig. 3a). These put height constraints
on the nodes of the solution. These limits are
obtained from a three-dimensional model of the
actual vault.
(b) Choosing a force pattern :
In plan, a possible force pattern topology is con-
structed (Fig. 3b). The branches represent possible
load paths throughout the structure. The loaded
nodes represent the horizontal projections of cen-
troids (cf. step d). This pattern is the horizontal
projection of the final solution.
(c) Generating the reciprocal force diagram ∗ :
The reciprocal force diagram (Fig. 3c) is produced
from the force pattern such that corresponding

Figure 4. For a simple, but indeterminate, 4-bar structure,
keeping the load P, the force pattern , and the depth of the
structure the same, this image shows the effect of manip-
ulating the force diagram ∗ , i.e. changing the internal
distributions of the forces. Stretching the force diagram to
double the size in one direction is equal to doubling the forces
in that direction and thus resulting in a structure half as deep
in that direction.
branches stay parallel and nodal equilibrium in
the pattern is guaranteed by closed polygons in the
reciprocal diagram (Fig. 2). Note that the applied
loads do not appear in the force diagram because
they reduce to a single point in the horizontal pro-
jection (Fig. 4). This results in a force diagram with
an unknown scale since the relation between pat-
tern and diagram is true regardless of their relative
scales.
(d) Attributing weights:
The weights attributed to the loaded nodes come
from distributing the dead load of the 3-D tributary
area to those nodes (Fig. 3d). In addition to self
weight, other loads can be applied, such as the level
of fill.
(e) Updating the force diagram:
In the case of an indeterminate force pattern con-
taining nodes with more than three bars coming
together per node, the user can change the inter-
nal force distribution by manipulating the force
diagram (Fig. 4). This notion becomes interesting
when forces want to be attracted to certain lines in
the structure, such as along the ribs.
(f) Solving for the equilibrium solution G:
Using the geometry of the force pattern and dia-
gram, the weights applied at the nodes and the
boundary conditions, this problem can be solved
using a one-step linear optimization (LO).
The constraint equations have the following form
(matrix notation):
596
Figure 5. Possible objectives for the LO problem: (a) maxi-
mizing the load factor of an applied load; (b) maximizing the
geometric safety factor; and (c) finding the range of thrust
values, defined by a minimum and maximum thrust value.
In this equation, the unknowns are linear combina-
tions of the nodal heights z and the unknown scale of
the force diagram r. The coefficients in the equations
are functions of the connectivity matrix C, which rep-
resents the topology of the networks (Schek 1974); the
branch lengths H and H ∗ of the force pattern  and dia-
gram ∗ respectively; and p, the loading in each node.
We solve simultaneously for the nodal heights of
the solution and the scale of the force diagram. The
horizontal components of the forces in the solution G
can easily be found by measuring the lengths of the
branches in the force diagram grid and multiplying
them by the actual scale.
3.4 Applications
Before being able to solve the problem using LO, the
analyst must choose the objective function of the opti-
mization problem. Examples are (1) increasing the
load factor of an imposed load until no solution can
be found that fits within the boundaries of the structure
(Fig. 5a); (2) finding the one solution which maximizes
the geometric factor of safety to demonstrate if a struc-
ture is safe or not (Fig. 5b); or (3) finding the range
of thrust, defined by a minimum and maximum thrust
solution (Fig. 5c), to understand the capacities of the
3D vault.
The first option gives an upper-bound solution.
O’Dwyer (1999) demonstrates this for a barrel vault
with a point load. Although such an analysis is rele-
vant for bridge structures, such a loading is unlikely
for vaults inside of a building. The second option is
used by Andreu et al. (2007). If the optimization pro-
duces a result, then the safe theorem guarantees that
the vault is safe and the geometric factor of safety gives
an indication of the relative stability of the vault. The
third option is used in this paper. The range of possi-
ble thrust values gives a useful characterization of the
structural behavior of the vault. The minimum (or pas-
sive) thrust state represents the least amount this vault
can push horizontally on its neighboring elements, as
a function of its self-weight and shape. The maximum
(or active) state of thrust on the other hand represents
the largest horizontal force this vault can provide. So,
this value demonstrates how much horizontal force
this vault can safely take from its neighboring ele-
ments. For the optimization problem this means that

Figure 6. Possible thrust values at the corners for this groin
vault range from 21% to 32% of its total weight.
we want to minimize and maximize the scale of the
force diagram, resulting in globally the smallest ver-
sus the largest horizontal forces in the system or also
the deepest versus the shallowest solution which fits
within the boundaries of the structure.
The viability of the proposed method is demon-
strated through two case studies, highlighting its mul-
tiple applications. First, an analysis of a series of groin
and quadripartite rib vaults investigates the interrela-
tionship between different parameters and the range of
possible equilibrium solutions of these vaults. A sec-
ond case study looks specifically at the fan vaults of
King’s College Chapel in Cambridge, England.
4 GROIN AND QUADRIPARTITE VAULTS
Variables influencing the range of vault thrust include
the influence of the web geometry, the boundary condi-
tions, the role and effect of cross-ribs, the assumptions
for the internal force patterns, the presence of cracks
and other pathologies, and the impact of fill above the
haunches.
The groin vault shown in Figure 6 is found to have a
range of possible horizontal thrust values at the corners
which vary from 21% to 32% of the total weight of
the vault. This results from the choice of a pseudo-
3D force pattern inspired by Wolfe’s analysis (Fig. 1),
where arches are assumed to span between the ribs,
which carry the loads to the supports. The presence of
Sabouret cracks which run parallel to the edges of the
vault could justify such a pattern.
Figure 7 illustrates the relation between the cho-
sen force patterns (i.e. the horizontal projection of
the thrust network solution), the reciprocal force dia-
grams, and the shape of the 3D equilibrium solu-
tions. The image demonstrates how the force diagrams
clearly visualize the internal force distribution of the
different force path assumptions (e.g. how much more
force goes into the diagonals compared to the arches
spanning in between them). Not only the proportions
inside the force diagrams are useful, but the global
scale allows for a direct comparison of the overall mag-
nitude of the forces in the system between the different
force pattern assumptions.
597
The first force pattern and diagram (Fig. 7a) repre-
sent the minimum thrust state shown in Figure 6. The
second pattern (Fig. 7b) shows a different assumption
on how the forces could travel through the structure: all
force lines go directly to the corner supports. The force
diagram represents in a clear fashion the equilibrium of
the corner nodes and center node. A similar force pat-
tern is used in Section 5 to analyze the fan vaults. The
last pattern (Fig. 7c) shows a fully three-dimensional
force pattern. An important difference with the previ-
ous two assumptions is that forces no longer only go
to the corner supports. This network assumes that a
part of the vault is carried along the edges. As a result
the thrust at the corner supports will be reduced. The
main force lines (heavier lines in the force pattern)
are distinguished from a continuous, regular grid in
between. As can be seen in the force diagram, more
force is attracted to the main force lines compared to
the grid lines resulting in primary structural action over
these lines and three-dimensional vault action between
them. The location of these force lines can for example
be inspired by the location of ribs in the actual vault.
Figure 7d shows a three-dimensional compression-
only thrust network which fits within the groin vault’s
geometry.
5 FAN VAULTS OF KING’S COLLEGE
The fan vaults of King’s College Chapel in Cambridge,
England were constructed between 1512 and 1515
(Fig. 8a). These double-curvature vaults were first
analyzed by Heyman (1977) using membrane theory.
These vaults with complex geometries can be analyzed
using thrust-network analysis.
From available documentation (Leedy 1980), a
detailed three-dimensional model is constructed
(Fig. 8b, c). This model is used for obtaining precise
nodal height constraints and good approximations of
the weights applied at each node.
Since there are cracks between the transverse arches
and the conoid fan vaults (Leedy 1980), we can assume
that no compressive forces can be transferred between
them. The transverse arches and the fan vaults work
independently of each other. The chosen force pat-
tern should reflect this, i.e. no branches should cross
the interfaces between the fan vault and the trans-
verse arch. Figure 8d shows a thrust network which
fits within the vault’s section demonstrating that it is
stable. The force pattern follows the radial rib pattern.
The fill adds weight and alters the thrust-network
but also adds more depth to the section for the thrust-
network to travel through. This is very clear if we look
at the equilibrium of the main transverse arch (Fig. 9).
Without the level of fill, this arch would be too thin
to stand under its own weight. The level of fill causes
an increase of thrust, i.e. the horizontal component, of
Figure 7. Different possible force patterns for a groin or rib vault: (a) the diagonal ribs bring down the forces to the corner
supports, arches in the web span in between the ribs; (b) all force lines go directly to the corner supports; and, (c) primary
force lines accumulate towards the supports and a continuous 3-D mesh distributes the loads to these force lines. (d) shows
the three-dimensional thrust network, resulting from the assumptions in (c), which fit within the vault’s geometry.

Figure 8. (a) Inside view of the fan vaults of King’s Chapel in Cambridge, (b,c) A detailed model of the geometry of the
vaults and (d) a 3D thrust-network solution fitting inside the section of the vaults.

598

Figure 9. (a) The line of thrust exits the section of the trans-
verse arch if the level of fill is left out. (b) shows the stable
situation with fill.
the main arch by less than 25% compared to the case
with no fill. On the other hand, the vertical component
is more than tripled due to the added weight of the fill.
We can conclude from this that the level of fill has a
significant stabilizing effect on the buttresses which
carry the arches and the vaults.
Furthermore, the level of fill appeared to be of
crucial importance in order to find an acceptable
thrust-network, which is shown in Figure 8d. In fact,
the thrust-lines in the long direction of the vaults only
fit within the section due to the fill. Another exam-
ple illustrating the careful choice of the geometry of
the vaults and fill is the added height and weight over
the transverse ridges between two adjacent fan vaults
(Fig. 8c) which causes a kink in the thrust lines in the
short direction of the vaults such that they fit in the
very steep section. From this preliminary analysis it
seems that each stone, and even the level of fill, has
been sculpted carefully to maintain the stability of this
thin vault.
The importance of the level of fill could be inves-
tigated even further. For a given vault geometry, what
would be the optimal level of fill? What loading would
cause the thrust-network to lie as close to the middle
surface of the vault as possible? We can solve this
by using equation (1) differently. For a given choice
of force network and force distribution, we now want
to find the funicular loading which results in a given
shape (Williams 1990). This means that in Equation
(1) now C, H , H ∗ and z are known and that p and r
are the unknowns. So, the optimization process now
defines the optimal level of fill which would cause the
thrust-network to lie as close to the middle surface of
the vault as possible.
599
6 DISCUSSION
The methodology uses existing 3D drawing software as
input and output resulting in an interactive tool with a
visual representation of results and force distributions.
Analysis proceeds from an accurate 3D model of the
vault without the need for abstraction or simplifica-
tion. The computation is done in MatLAB. The number
of elements that the implementation can handle is lim-
ited to approximately 500 due to computational speed
but this could easily be improved.
For the thrust-network analysis of masonry vaults,
various parameters can be changed. (a) Different force
patterns can be chosen to compare assumptions on how
forces may be traveling through the structure. A dis-
tinction can be made between primary force lines and
secondary force lines. (b) The force diagrams can be
manipulated in order to redistribute the internal force
distributions. (c) The solution envelopes can be chosen
in order to constrain the solutions to the middle third
zone of the vault or to exclude the thickness of ribs.
(d) Different boundary conditions can be chosen. The
vault can have a continuous edge support or only cor-
ner supports. This decision can be influenced by the
curvatures of the vaults or by the existence of cracks.
(e) Level of fill or other imposed loads can easily be
integrated by adding load to affected nodes.
In addition, the loading cases do not have to be con-
strained to only gravity loads. An initial measure of the
stability of a vault under lateral acceleration can be
assessed by applying an equivalent static horizontal
force. As shown in Block et al. (2006a) using inter-
active graphic statics for 2D, this can be simulated by
tilting the model.
Future work includes the development of a more
sophisticated optimization set-up which takes into
account all possible force patterns and diagrams and
searches for the absolute minimum and maximum
thrust values for the masonry vaults.
7 CONCLUSION
This paper applied Thrust-Network Analysis, a fully
three-dimensional computational method, to obtain
lower-bound solutions for masonry vaults with com-
plex geometries.
Key elements in the proposed process are (1) force
networks, representing possible force paths through
the structure; (2) interactive reciprocal diagrams, visu-
alizing the proportional relationship of all forces in the
force network and providing a high level of control for
the user to understand and manipulate the force dis-
tributions within the system; (3) the use of envelopes
defining the solution space; and (4) linear optimiza-
tion, resulting in fast computation and visualization of
results.
 For the input of the vault geometry and the loading
conditions and for the output of the three-dimensional
results, implementations are written in existing archi-
tectural software. This allows for clear visualizations
of the results, and the smooth integration of the
form-finding process in the analysis process.
Through several examples, this paper demonstrated
the potential of thrust-network analysis as a power-
ful tool for understanding, visualizing and exploring
the equilibrium of compression-only structures such
as historic vaults in unreinforced masonry.
REFERENCES
Andreu, A., Gil, L. & Roca, P. 2007. Computational Analysis
of Masonry Structures with a Funicular Model. Journal
of Engineering Mechanics 133(4): 473–480.
Boothby, T.E. 2001. Analysis of masonry arches and vaults.
Progress in Structural Engineering and Materials 3:
246–256.
Bow, R.H. 1873. Economics of construction in relation to
frames structures. London: Spon.
Block, P. Dejong, M. & Ochsendorf, J. 2006a. As Hangs the
Flexible Line: Equilibrium of Masonry Arches. The Nexus
Network Journal 8(2): 13–24.
Block, P., Ciblac, T. & Ochsendorf, J. 2006b. Real-time limit
analysis of vaulted masonry buildings. Computers and
Structures 84(29–30): 1841–1852.
Block, P. & Ochsendorf, J. 2007. Thrust Network Analysis:
A new methodology for three-dimensional equilibrium.
Journal of the International Association for Shell and
Spatial Structures 48(3).
Cremona, L. 1879. Le Figure Reciproche nella Statica
Grafica. Milan: Ulrico Hoepli.
Föppl, A. 1892. Das Fachwerk im Raume. Leipzig: Verlag
von B.G. Teubner.
600
Heyman, J. 1966. The stone skeleton. International Journal
of Solids and Structures 2: 249–279.
Heyman, J. 1977. Equilibrium of Shell Structures. Oxford:
Clarendon Press.
Heyman, J. 1995. The Stone Skeleton: Structural engi-
neering of masonry architecture. Cambridge: Cambridge
University Press.
Huerta, S. 2001. Mechanics of masonry vaults: The equilib-
rium approach. In P.B. Lourenço, P. Roca (eds.), Historical
Constructions: 47–70, Guimarães.
Huerta, S. 2004. Arcos bóvedas y cúpulas. Geometría y equi-
librio en el cálculo tradicional de estructuras de fábrica.
Madrid: Instituto Juan de Herrera.
Leedy, W.C. 1980. FanVaulting: A Study of Form, Technology,
and Meaning. Santa Monica: Arts and Architecture Press.
Maxwell, J.C. 1864. On reciprocal figures and diagrams of
forces. Phil. Mag. Series 4(27): 250–261.
Ochsendorf, J.A. 2002. Collapse of masonry structures.
PhD. dissertation, Department of Engineering, Cambridge
University.
O’Dwyer, D.W. 1999. Funicular analysis of masonry vaults.
Computers and Structures 73(1–5): 187–197.
Schek, H.-J. 1974.The Force Density Method for formfinding
and computation of general networks. Computer Methods
in Applied Mechanics and Engineering 3: 115–134.
Smars, P. 2000. Etudes sur la stabilité des arcs et voûtes.
PhD Thesis, Department of Civil Engineering, Katholieke
Universiteit Leuven.
Ungewitter, G. 1890. Lehrbuch der gotischen Konstruktio-
nen. Leipzig: Weigel Nachfolger.
Williams, C.J.K. 1986. Defining and designing curved flex-
ible tensile surface structures. In J.A. Gregory (ed.), The
mathematics of surfaces: 143–177. Oxford: Clarendon
Press.
Williams, C.J.K. 1990. The generation of a class of structural
forms for vaults and sails. The Structural Engineer 68(12):
231–235.
Wolfe, W.S. 1921. Graphical analysis: a handbook on graphic
statics. New Cork: McGraw-Hill.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Assessment of seismic resistance of a basilica church: Modelling,

analysis and strengthening proposal

M. Betti & A. Vignoli

Dept. of Civil and Environmental Engineering, University of Florence, Italy

ABSTRACT: In this paper main results on the study of the seismic vulnerability of a basilica-type church are
presented. Static behavior and dynamic properties are evaluated using the finite element method. To this end a 3D
non-linear analysis is done that takes into account the non-linear behavior of masonry: constitutive assumptions,
characterized by elasticity, damage and friction, are made. The standard FEM modeling strategy, based on the
concepts of homogenizated material and smeared cracking constitutive law, is used. Seismic vulnerability is
evaluated using a pushover approach. Results obtained with the FEM technique are then compared with the
simplified schemes of limit analysis, and the capacity of the church to withstand lateral loads is evaluated
together with the expected demands from seismic actions. The comparison demand vs. capacity confirms the
susceptibility of this type of buildings to extensive damage and to collapse, as frequently observed. Starting
from a single case study, a contribution to the issue of modeling and analysis of monumental masonry buildings
under seismic action is provided. Eventually a retrofitting proposal is presented.

1 INTRODUCTION and the mechanical properties (Binda et al. 2000).

Church building represents a large portion of the Ital-
ian cultural heritage that have demonstrated to be
particularly susceptible to damage and prone to partial
or total collapse under earthquake loads (Lagomarsino
1999). The high seismic vulnerability of this type
of building is due both to the mechanical properties
of masonry materials, characterized by a very small
tensile strength, and to the particular configuration
(slender walls, lack of effective connections among
the structural elements, etc.).
Each monumental building is characterized by its
own history, often resulting in a composite fusion
of additions and replacement of structural elements
that gave origin to its final configuration. Besides the
dynamic behavior of ancient buildings is normally too
complicated to be interpreted with simple mechani-
cal models and the process of extrapolating analytical
procedures specifically developed for modern build-
ings is in most cases inadequate. Also, in ancient
(monumental) buildings it is sometimes quite difficult
to perform reliable quantitative strength evaluations,
due to the difficulty of gathering experimental data
on the resistance of the structural elements and even
on the mechanical properties of the on site materi-
als. Monumental historical buildings can hardly be
reduced to any standard structural scheme because of
the uncertainties affecting both the structural behavior
Therefore the study of masonry heritage buildings
remains a challenging task as many recent studies con-
firm (Lourenço & Orduña 2003, Lourenço & Oliveira
2007).
To deal with these problems a double approach,
with respect to a specific case study, is used. Firstly
a global analysis of the building is made using the
finite element technique. Specific assumptions on the
material properties and on the non linear behavior of
masonry are made to appraise the general properties
of the building’s structural response under earthquake
loads. Next step is to apply the simplified schemes of
the limit analysis on several significant architectonical
elements of the building. Some elementary macro-
elements are analyzed, to this end, with reference to
the mechanisms activated on similar buildings during
past earthquakes. Comparing the results obtained with
the global and local analyses could provide an accu-
rate description of the building response under seismic
loads. The paper suggests that comparing different
approaches for the analysis of historical buildings is
mandatory to cover the unavoidable unknowns that
always affect building materials and mechanics.
The proposed approach is applied to an emblematic
case study: the Basilica of Santa Maria all’Impruneta,
located in Impruneta nearby Firenze, (Italy). As a
first step, an in-situ survey of the building’s state
was made to obtain basic information for following

601
 7m
8m

2160
Figure 1. Aerial view of the Church case study.

41 m
analyses. In particular an in-situ survey on the geom-
etry (and on the structural details) aimed to identify
any irregularities of the structure was made.
This is necessary also to localize possible col-
lapse mechanisms that might be activated during an
earthquake. The investigation consisted in a geomet-
rical survey to identify the actual damage, a quality
check up of connections wall to wall and wall to roof,
and masonry texture characterization. This step was
accompanied by an historical survey of the building
over centuries in order to determine the original shape
and to consider the church modifications over the cen-
turies. The paper reports mainly, for brevity, results on
6m

the assessment of the seismic behavior of the church

even though the research was extended to include the
bell-tower. 32 m

Figure 2. Plan layout of the Basilica of Santa Maria.

1.1 Description of the structure
The church analyzed in this paper is the Basilica
of Santa Maria all’Impruneta (Fig. 1), located in
Impruneta near Firenze (Italy). The plan view (see
Fig. 2) shows the typical basilica layout with a single
nave and a polygonal apse.
The original Romanesque structure was built
around the XI century followed by various additions
and alterations that extensively modified the original
structure (that was quite different from the Renais-
sance profile that is seen today). It is interesting to
highlight the presence of a pronaos (dating back to
XVII century), the major architectural feature of the
church.
The last major restoration of the Church was made
in the fifties of the last century. During the Second
World War the Church was seriously damaged: on July
26–27, 1944 all the apse area was destroyed by an
air attack. After Second World War, the apse and the
timber roof were completely reconstructed using the
original on site materials.
1.2 Geometry and materials
The main dimensions of the nave were a maximum
length of about 4.0 m, a maximum width of 15.0 m
and a wall height of about 15.0 m. The masonry wall
thickness ranges between 0.70 m (nave walls) and
0.80 m (apse walls). The church roof (nave) is a timber
structure.
The walls are single-leaf and several types of
masonry weaving characterize the church wall, dif-
fering both in materials (stone, brick, etc.) and shape
(“opus incertum”, “opus mixtum”, etc.). This varie-
gated picture reflects the modifications endured during
centuries by the building. However, despite these dif-
ferences, the construction is mostly made of irregular
sandstone masonry (local stone) with thick lime mor-
tar joints. Local stone is used also for the windows and
doors jambs.
Due to budgetary reasons, a complete characteri-
zation of these materials was not performed in this

602
 Table 1. Yield criterion (main elements).

Drucker-Prager Yield criterion

Nave Apse Columns

c (cohesion) 0.1 N/mm2 0.1 N/mm2 0.5 N/mm2

η (flow angle) 15◦ 15◦ 15◦
φ (friction angle) 38◦ 38◦ 38◦

Table 2. Failure surface (main elements).

Willam and Warnke failure surface

Nave Apse Columns

Figure 3. Finite element model of the Basilica of Santa
Maria. fc (uniaxial 7.5 N/mm2 8.5 N/mm2 40 N/mm2
compressive
study. Conservative values for historic masonry have strength)
been assumed considering the Authors experience on ft (uniaxial 0.15 N/mm2 0.15 N/mm2 3.5 N/mm2
tensile strength)
similar material (Chiostrini et al. 1997, Chiostrini et al.
βc (shear transfer 0.75 0.75 0.75
1998). coeff. close crack)
βt (shear transfer 0.15 0.15 0.15
2 SEISMIC VULNERABILITY coeff. open crack)

2.1 Finite element model

Two combined approaches are applied to evaluate the
seismic vulnerability of the church. The first is a global
analysis by the FEM technique, for which a finite ele-
ment model of the building by the code ANSYS have
been built.
The masonry walls have been modeled by means of
Solid65 elements; Shell143 elements have been used to
model the main vault on the first floor in the annexes.
Beam44 elements have been used for the queen truss
of the timber roof.
The model accurately reproduces the geometry of
the structure, focusing on the variations in the wall
thickness, the irregularities and the wall connections.
The major openings in the building are reproduced.
The final 3D model consists of 27779 joints, 76895
3D Solid45 elements, 1751 2D Shell63 elements and
547 1D Beam44 elements, that correspond to 81021
degrees of freedom.
The material properties of masonry walls (Young
modulus E, Poisson coefficient υ, own weight W) are
differentiated by taking into account each area present
in the building (Fig. 3). The non-linear behavior of
the masonry elements have been reproduced assuming
a Mohr–Coulomb type failure criterion with tension
cut-off type behavior.
In particular the yield Drucker–Prager criterion
with associated flow rule have been assumed (it is
assumed then that the yield surface does not change
with progressive yielding, i.e. there is no hardening
rule). The yield criterion is joined with the Willam
and Warnke failure surface (Willam & Warnke 1975)
to introduce a cut-off on the tensile stresses.
Tables 1–2 reports the selected values needed for
the definition of the model parameters with respect
to some principal elements. Information on material
properties was already available from similar materi-
als and conservative values have been assumed from
the experience; typical values for old stone masonry
were selected for the material parameters referring on
existing tests of similar masonry and in situ testing
(Chiostrini et. al. 1992)
2.1.1 Static analysis
A preliminary static analysis has been made. The struc-
tural masonry walls have been analyzed under constant
vertical loads deriving from their own weight and from
the roof loads.
The static analysis can mainly provide information
on the interaction, in particular the stress resultant dis-
tribution, among the structural elements in the two
principal directions. Figures 4–6 reports main results
in terms of vertical stresses. In general, the stress state
induced on the church by the static loads is quite mod-
erate. The medium value of compressive strength is
about 0.6 N/mm2 . A modest value of tensile stress
appears on the top surface of the nave walls, due to
the timber roof loads; this is mainly a local effect
depending on the connection between beam and solid
elements.

603

Figure 4. Vertical stresses σz [N/mm2 ]: general view.
Figure 5. Vertical stresses σz [N/mm2 ]: nave walls detail.
The maximum compression value is found in the
columns on the facades (Fig. 6) where the stress state
reaches the value of 2.5 N/mm2 . Even if it is high, this
value is lower than the crushing limit of the column’s
material (stone). These are the critical elements of the
church as is demonstrated by the fact that they are
reinforced by a steel collar.
2.1.2 Modal analysis
The first two modes of the “as is” building involves the
translation in the two principal direction of the bell-
tower; since the paper is focused on the analysis of the
church building, these results are not reported here.
The third modal shape (Fig. 7) is the first translational
mode for the Church, and it acts on the transversal
direction of the nave. Following modal shapes (not
represented here for brevity) are a combination of
transversal vibration mode and torsional mode on the
church.
The distribution of the modal shapes shows that
the “as is” building, though characterised by very
stiff structural elements on the perimeter, displays low
604
Figure 6. Vertical stresses σz [N/mm2 ]: facade walls detail.
Figure 7. First mode on the nave T1 = 1.704 sec.
transversal and torsional stiffness, and significant out-
of-plane deformations of the elements. Furthermore
the plan deformed configurations of the structure con-
firm that the seismic loads acting along the church
transversal direction involve remarkable out of plane
deformations of the orthogonal structural elements.
2.2 Pushover analysis
The study of the seismic behavior has been made using
a non-linear static analysis method. Seismic loads are
evaluated with respect to the New Italian rule (DM
14/09/05 & OPCM 3274) using a pushover analysis
(Galasco et al. 2006, Kim & D’Amore 1999). Based
on this analysis method, the effects of the seismic loads
are evaluated through the application of two systems of
horizontal forces perpendicular to one another. These
forces, not acting simultaneously, are evaluated taking
into account two load distributions. The first load dis-
tribution is directly proportional to building’s masses
(uniform); the second load distribution is proportional
to the product of the masses for the displacements of
the corresponding first buildings modal shape. These
load distributions could be considered as two limit
states for the building’s capacity. The first distribu-
tions assume that the horizontal loads representative of
inertia seismic forces are constant with respect to the
building’s height. This means that the displacements
on the lower level of the building are overestimated,
while the opposite happens on the displacement on
the top level. On the contrary the second distributions
overvalue the displacements on top level. In this study
a conventional pushover is assumed, i.e. loads applied
on the building don’t change with progressive degra-
dation of the buildings during loading. This means
that the conventional pushover does not account for
the progressive changes in modal frequencies due to
yielding and cracking on the structure during loading.
This is a critical point for the application of conven-
tional pushover to the analysis of historic masonry
buildings, because it is predictable that the progressive
damage of the building may also lead to period elonga-
tion, and therefore to different spectral amplifications
(Antoniou & Pinho 2004). Hence the hypothesis of
invariance of static loads could cause an overestima-
tion in the analysis of masonry building especially
when a non uniform damage on the buildings or a high
level of cracking are expected. Further studies will be
devoted to the investigation of this point. However, also
in its conventional form, the pushover provides an effi-
cient alternative to expensive computational inelastic
dynamic analyses.
The seismic loads to apply to the building have
been evaluated through modal analysis with the elas-
tic response spectrum. For the case study the Class
1 spectrum as reported in the New Italian Rule (DM
14/09/05), with a ground type A corresponding to rock
or other rock-like geological formation, is assumed.
Following expressions are derived:
In eq.√(1) T is the natural period of the structure,
and η = (10/(5 + ξ) where ξ is the damping ratio.
605
Figure 8. Displacement (mm).
The peak ground acceleration (PGA) for the reference
return period is denoted by ag and for the case study
is assumed equal to 0.35 g. S is a factor depending on
the ground type that in this case (ground type A) is
equal to 1.0. Next, results concerning church behavior
are detailed. The critical load distribution is the one
acting on the y-direction (direction perpendicular to
main nave direction). This is quite expected due to
the fact that the transversal direction of the church
involves remarkable out of plane deformations of the
orthogonal structural elements.
Figure 8 reports the displacements on the transver-
sal direction at the end of the analysis; Figure 9 reports
the corresponding crack pattern. The capacity curve is
reported in Figure 10 with the corresponding behavior
of the equivalent bilinear systems. For this system it
has been obtained, by eq. (2) (see DM 14/09/05 for
more details), that the seismic displacement demand
is dmax = 89 mm against the building displacement
capacity that is equal to 88 mm.
In eq. (2)  is the so-called transformation factor, T∗
is the period of the idealized single degree of freedom
(SODF) system and SDe is the elastic displacement
response spectrum.
2.3 Mechanism analysis
The examination of the collapse and the damage con-
sequent to past earthquakes enable to foresee the
occurrence of recurrent modes and mechanism of col-
lapse in masonry church buildings (D’Ayala et al.
1997, Lagomarsino 1999, Lagomarsino & Podestà
2004). It is possible then to connect previous results
obtained by the finite element technique for a com-
parison with the results obtained by the analysis of
Figure 9. Cracking pattern (pushover, y-direction).

kN Capacity Curve
20000
18000
16000
14000
12000
10000 Figure 11. Reference case.
8000
Original MDOF system
6000
4000
Bilinear SDOF system bell-tower and the clock-tower, they are not reported
2000 here for brevity’s sake.
0
0 10 20 30 40 50 60 70 80 90
displacements (mm)

Figure 10. Capacity curve (pushover, y-direction).

simplified schemes based on the limit analysis. The
structure is considered as the assemblage of a certain
number of components depending on the structural
compound geometry and shape (e.g. the whole façade)
and on the details (e.g. quality of existing connec-
tions) whose behavior is similar to the behavior of
analogous macro-elements in other buildings, where
the collapse mechanisms can be identified. In this
study, several possible mechanisms have been taken
into account (D’Ayala & Speranza 2003), mostly out-
of-plane mechanisms. Next the results are reported
with respect to the most diffused collapse mechanisms.
To this end, several elementary macro-elements
have been identified in the church; applying the The-
orem of Virtual Work (TVW) the collapse multipliers
(α0 ) have been evaluated by eq. (3) that refers to
the simple scheme illustrated in Figure 11. After the
identification of the collapse mechanism a rigid-body
system is considered in application of the TVW. The
seismic load is assumed as an overturning force while
the gravity load is assumed as a stabilization load.
Analyses of past events have permitted to select
a group of mechanisms to analyze. These mecha-
nism concerns partial of local overturning of the main
façade, of the lateral bearing walls, of the apse. Other
mechanisms take into account the behavior of the
After the evaluation of the collapse multiplier, the
corresponding seismic spectral acceleration is evalu-
ated by (where M∗ is the mass of the idealized SODF
system and g is the acceleration of gravity):
The respect of the analyzed limit state is assured by
the respect of the following inequality (where q is the
behaviour factor assumed equal to 2.0, Z is the height
of the mass above the level of application of the seismic
action and H is the total height of the macro-element):
The step where the church is subdivided into macro-
elements is a crucial one. It consists in identifying the
architectonic elements where the seismic behavior acts
quite independent by the rest of the structure. Major
macro-elements could be searched on the façade,
on the apse or on the bell tower. For each macro-
element must be considered its typology and their

606
 Figure 15. Mechanism 4: lateral wall overturning.
Figure 12. Mechanism 1: pronaos overturning (local).

Figure 16. Mechanism 5: lateral wall overturning.

Figure 13. Mechanism 2: pronaos overturning (local).

Figure 17. Mechanism 6: lateral wall overturning.

Figure 14. Mechanism 3: pronaos overturning (global).

connection with the remaining parts of the building. reports results on the mechanisms that could be acti-
All these informations must be collected during the vated on the pronaos (Figs 12–14), in the lateral wall
in-situ survey of the church, and every structural detail of the church (Figs 15–17), in the lateral wall of the
need to be recorded (Brencich et al. 1998). This paper nave (Fig. 18) and on the apse (Fig. 19).

607
 Table 3. Collapse multiplier.

Collapse Seismic spectral

multiplier acceleration Seismic
(α0 ) (α∗0 ) demand

Mechanism 1 0.19 0.22 g 0.38 g

(pronaos)
Mechanism 2 0.25 0.29 g 0.39 g
(pronaos)
Mechanism 3 0.20 0.21 g 0.33 g
(pronaos)
Mechanism 4 0.06 0.06 g 0.39 g
(lateral wall)
Mechanism 5 0.64 0.74 g 0.39 g
(lateral wall)
Mechanism 6 0.14 0.18 g 0.29 g
Figure 18. Mechanism 7: nave wall overturning. (lateral wall)
Mechanism 7 0.07 0.10 g 0.36 g
(nave)
Mechanism 8 0.10 0.11 g 0.36 g
(aspe)

proposed. In particular a system of horizontal coun-

Figure 19. Mechanism 8: apse wall overturning.
Results from the macro-element approach, reported
in Table 3, agree with the results obtained with the
pushover analysis via the finite element technique.
The critical behavior of the church is activated when
the seismic load is acting on the transversal direction
of the nave of the church. This is mainly due to the
absence of a buttress able to counteract the seismic
action. Despite the good quality of connection between
walls and timber roof, this does not produce an effort
in this direction. Other local collapses could arise in
local areas of the church (the pronaos, lateral walls
overturning) that do not however involve the entire
structure.
3 RETROFITTING PROPOSAL
The main objective of the strengthening design is to
increase the out-of-plane strength of the lateral wall of
the nave that seems to be the critical element on the
church (because it might generate a global collapse
in the building). In order to generate a box behavior
a global reinforcement of the queen roof has been
terbracing it has been inserted between the extrados of
the timber roof and the tile covering. The steel coun-
terbracing, directly connected to the wooden beam of
the queen roof, is designed in order to be able to cre-
ate a top rigid floor that may ensure a box behavior of
the nave. A local steel reinforcement of the connection
between timber structure and masonry walls it has been
provided to avoid local failure. In addition several local
reinforcements, based on results reported in Table 3,
are inserted along the building. Among them a steel
tie is disposed along the apse perimeter to counteract
the development of mechanism 8.
4 CONCLUSIONS
The paper faces the assessment of the seismic vulner-
ability of monumental masonry buildings with respect
to a specific case study. The engineers involved in
monumental buildings assessment are asked to paid
attention to both the economical and cultural relevance
of this kind of building that must be joined with the
safety request. As a matter of fact a harder attempt
on the understanding of the historical process must
be paid.
With the aim to illustrate an integrate approach the
seismic vulnerability of a basilica-type Church has
been assessed by the combination of two integrates
approaches. First, a global analysis that refers to the
finite element technique. Next, a local approach that
use the limit analysis (kinematics approach). Concern-
ing first approach a 3D non-linear numerical model
has been created. Concerning the second approach
by the analysis of past earthquake some elementary

608
macro-elements have been identified and their vul-
nerability has been assessed. Taking into account the
results obtained with the two methods a retrofitting
proposal that offer an increase to the seismic resis-
tance of the building has been proposed. The proposed
strengthening use traditional material and techniques
in order to propose a reversible intervention.
It is believed that the results and the conclusions
obtained with respect to the static assessment and
the operative model identification procedure herein
proposed could be extrapolated for a wide variety of
masonry cultural heritage structural compound.
REFERENCES
Antoniou, S. & Pinho, R. 2004. Advantages and limita-
tions of adaptive and non adaptive force-based pushover
procedures. Journal of Earthquake Engineering 8(4):
497–522.
Binda, L., Saisi, A. & Tiraboschi, C. 2000. Investigation
procedures for the diagnosis of historic masonries. Con-
struction and Building Materials 14(4): 199–233
Brencich, G., Gambarotta, G. & Lagomarsino, S. 1998.
A macroelement approach to the three-dimensional seis-
mic analysis of masonry buildings. Proceeding. of the XI
European Conference on Earthquake Engineering, Paris,
A.A. Balkema (Abstract Volume & CD-ROM): 602.
Chiostrini, S., Galano, L. & Vignoli, A. 1998. In situ tests and
numerical simulations on structural behaviour of ancient
masonry. Proceedings Workshop on Seismic Performance
of Monuments: 197–206.
Chiostrini, S., Foraboschi, P. & Vignoli, A. 1992. Struc-
tural analysis and damage evaluation of existing masonry
building by dynamic experimentation and numerical mod-
elling damage. Proceedings of Tenth World Conference on
Earthquake Engineering, Madrid: 3481–3486.
Chiostrini S., Galano L. & Vignoli, A. 1997. Mechanical
characterization of stone masonry panels and effective-
ness of strengthening techniques. Proceeding Fourth Int.
Symposium on Computer Methods in Structural Masonry:
128–35.
609
D’Ayala, D., Spence, R., Oliveira, C. & Pomonis A. 1997.
Earthquake Loss Estimation for Europe’s Historic Town
Centres. Earthquake Spectra 13(4): 773–793.
D’Ayala, D. & Speranza, E. 2003. Definition of Col-
lapse Mechanisms and Seismic Vulnerability of Historic
Masonry Buildings. Earthquake Spectra 19(3): 479–509.
D.M. del Ministero delle Infrastrutture e dei Trasporti
14/09/2005. Norme Tecniche per le Costruzioni. G.U.
23/9/2005, n. 222, S.O. n. 159.
Galasco, A., Lagomarsino, S. & Penna, A. 2006. On the use
of pushover analysis for existing masonry buildings. First
European Conference on Earthquake Engineering and
Seismology. Geneva, Switzerland, 3–8 September 2006
(paper 1080).
Kim, S. & D’Amore, E. 1999. Push-over analysis procedures
in earthquake engineering. Earthquake Spectra 15(3):
417–434.
Lagomarsino, S. 1999. Damage survey of ancient churches:
the Umbria-Marche experience. In A. Bernardini (ed.),
Seismic Damage to Masonry Buildings, Proceeding of
the International Workshop on “Measures of seismic dam-
age to masonry buildings”, Monselice, Italy, Rotterdam:
Balkema.
Lagomarsino, S. & Podestà, S. 2004. Seismic Vulnerability of
Ancient Churches: I. Damage Assessment and Emergency
Planning. Earthquake Spectra 20(2): 377–394
Lourenço, P.B. & Orduña, A. 2003. Seismic Analysis and
Strengthening of a 17th Century Church in Azores.
Proceeding 9th North American Masonry Conference,
Clemson, USA: 932–941.
Lourenço, P.B. & Oliveira D.V. 2007. Improving the seismic
resistance of masonry buildings: Concepts for cultural
heritage and recent developments in structural analysis.
XII Convegno Nazionale ANIDIS “L’Ingegneria Sismica
in Italia”, Pisa.
O.P.C.M. 3274 del 20/3/2003. Primi elementi in materia di
criteri generali per la classificazione sismica del territorio
nazionale e di normative tecniche per le costruzioni in
zona sismica. G.U. 8/5/2003, n. 105, S.O. n. 72 e O.P.C.M.
3431 del 3/5/2005.
Willam. K.J. & Warnke, E.D. 1975. Constitutive model for
the triaxial behaviour of concrete. Proceeding Interna-
tional Association for Bridge and Structural Engineering,
Bergamo, Italy: 174–186.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Estimation of the seismic history of the city of Thessaloniki through back

analysis of its Byzantine land walls

K.-A. Stylianidis & A. Sextos

Department of Civil Engineering, Aristotle University Thessaloniki, Greece

ABSTRACT: This paper aims at developing the tools and strategy for assessing the dynamic and seismic
performance of the Byzantine Walls of the city of Thessaloniki in order to evaluate the seismic history of the city
as a whole. The particular Walls have been constructed at the end of the 4th century A.D. in the reign of Theodosius
the Great and as such, their structural integrity and record of damage reflects to a certain degree the level of
seismic forces that has developed during the centuries. Moreover, the fact that they are extending in kilometers
within the civil grid of the modern city allows the study of the role played by the local soil conditions for a
given earthquake scenario. It is worth noting that despite their relatively simple structural system, their foreseen
seismic behavior as a 3D body has not been thoroughly studied so far, primarily due to the lack of efficient
numerical tools and the high computational cost related, especially towards the study of their response in the
time domain. Along these lines, a refined dynamic analysis approach is proposed and the structural performance
of particular parts of the Walls complex is examined for a number of realistic earthquake scenarios, accounting
for the site specific soil conditions, the spatially variable nature of the incident seismic waves, as well as the
overall geotechnical/geotectonic environment of the area. Through this advanced simulation scheme, an upper
bound of the historical level of seismic forces for the city of Thessaloniki is traced (through back analysis), while
the overall refined approach can be also used as a guide for the direct assessment of the existing seismic capacity
of monuments as a whole.

1 INTRODUCTION
1.1 Scope
The impact of earthquakes on monumental heritage
is a critical issue that has attracted growing scientific
interest during the last decades. Monuments however,
are most often complex structures, whose preservation
and/or seismic strengthening heavily relies on the clear
understanding of all factors affecting their vulnerabil-
ity as well as on the accurate study of the effects of past
earthquakes. Until recently, the investigation of the
seismic performance of important and extended mon-
uments was restricted by the inherent limitations of
numerical analysis thus preventing the engineers from
the study of their dynamic response in the time domain
under realistic (recorded or artificial) ground motion
scenarios. This problem was further stressed in the
case of (Classical, Roman, Byzantine or Medieval) city
(Sea or Land) Walls whose dimensions were normally
significantly large, hence the complexity of the analyt-
ical or numerical procedures to be followed, (as well as
the subsequent computational cost) were considerably
high. Along these lines, it was deemed interesting to
focus on the city Walls of Thessaloniki, utilizing state-
of-the-art numerical tools and the experience gained
from the seismic study of other historical structures
of the Byzantine Era throughout the city, in order to
attempt to shed some light not only on the structural
history of the Walls through the centuries, but if possi-
ble, to back analyze and evaluate the reliability of the
seismic scenarios developed for the Metropolitan City
of Thessaloniki as a whole.
The Walls, still surrounding partially the old town
of Thessaloniki were initially built in 315 B.C. by
the king Kassandros and were completed at the time
of Great Theodosius (379–395 A.D.). Nowadays, the
Walls extend in kilometres within the civil grid of
the modern city but their continuity has been dis-
rupted due to partial or complete collapse at numerous
locations. Historically, due to their dynamic nature,
in the sense that they could be adapted in accor-
dance with civilian needs, thus being repaired after
sieges and following rules of economy and function-
ality where the art of war was concerned, the Walls of
Thessaloniki (as those of Constantinople, Nicaea and
others), did in fact changed considerably over the cen-
turies (Bouras, 2002) following the heavy fortification
requirements that arose. From the overall Byzantine
Walls complex which extends in kilometres within the
city, two sections (Figure 1) were chosen to be studied,

611

Figure 1. Contemporary photograph of the 4th Century
A.D. Byzantine Land Walls ofThessaloniki (North side view).
particularly: (a) a simple relatively small, essentially
stand-alone part of the Walls located approximately
in the center of the modern city and (b) the Walls
circuit in the northern part of the Byzantine fortifi-
cation. This part, which due to its size is studied more
thoroughly, forms a statically independent structural
system, extending from the beginning of the West Gate
(namely “Pyrros Gate”) to the main East Gate along the
612
Eptapygriou street, near the Trigonion Tower, inclu-
sive of the two twin gates at the East section (widened
and named thereafter by Anna Palaiologina) up to the
circle tower (Figures 2–5) constructed later. The par-
ticular Walls Section was encircling the Byzantine
acropolis (Velenis, 1998) thus separating it from the
Ancient Acropolis and it consists of numerous rectan-
gular (primarily) and triangle defensive towers. This
is essentially a monolithic and straight complex with
its main axis being parallel to the East-West direc-
tion. The superstructure was constructed of masonry
made by alternate bands of stones and bricks (Velenis,
1998). The masonry displays a mixture of construction
methods as can be seen in Figure 3.
From the two locations selected to be studied,
the small, free-standing masonry wall illustrated at
the bottom of Figure 1 was chosen on the basis
of the well controlled study environment which was
ensured by its clear geometry and structural system,
its proximity to the unique location where the 1978
earthquake was recorded and the good knowledge of
the underlying soil.
On the other hand, the extended Land Walls Sec-
tion in the upper part of the city (illustrated at the top
of Figure 1) was selected to be studied primarily for
three reasons: (a) due to its structural integrity which
has remained almost intact within the centuries thus
allowing to assume that its seismic capacity must have
not been exceeded for the last 1600 years, (b) due to
its overall length (reaching approximately 500 m) that
forms one of the most extended preserved parts of the
fortification and (c) due to the additional opportunity
that this length provides, to further investigate (in con-
trast to what is commonly performed) the potential
effect of the (typically neglected) traveling nature of
the incoming seismic waves to the dynamic response
of the system.
The scope of this research effort is therefore, to
utilize state-of-the-art knowledge and numerical tools
for the pilot study of the seismic performance of the
selected Thessaloniki city Land Walls in order to:
a) verify the good seismic performance of the specific
Walparts during the 20/6/1978 Thessaloniki severe
earthquake (Ms = 6.5).
b) assess the level of safety during future seismic
events (especially for the long circuit) through
appropriate earthquake scenarios that account for
a realistic representation of the overall seismotec-
tonic and local site conditions.
c) investigate the potential effect of ignoring spa-
tially variability of earthquake ground motion in
the analysis of long and monolithic structures.
d) if possible, to compare through back numerical
analysis, the level of the highest earthquake excita-
tion (that could have damaged the particular Wall
section but apparently has not occurred yet) with
the Maximum Credible Earthquake (MCE) defined
Figure 2. Contemporary photograph of the 4th Century
A.D. Byzantine Land Walls of Thessaloniki (North side
view).
Figure 3. Contemporary photograph of the 4th Century
A.D. Byzantine Land Walls ofThessaloniki (South side view).
through deterministic seismic hazard assessment
for the same area. It is noted that the definition
of upper limits on earthquake ground motions has
been identified as the ‘missing piece’ for seis-
mic hazard assessment for both deterministic and
probabilistic approaches (Bommer, 2002).
2 OVERVIEW OF THE ANALYSIS APPROACH
ADOPTED
2.1 Finite Element modelling of the Wall Section
Due to the complexity of the problem, it was deemed
necessary to primarily establish the overall strategy
that had to be followed for the study of the particular
city Walls Section. At first, the 3-Dimentional Finite
613
Figure 4. Axonometric view of rectangular wall between
along the East-West axis (North side view). After Velenis,
1998.
Figure 5. Contemporary photograph of the 4th Century
A.D. Byzantine Land Walls of Thessaloniki. East Gate,
Trigonion Tower (South side view).
Element model was created based on the exact struc-
tural geometry as this was extracted by the satellite
image (illustrated in Figure 1), scaled axonometric
views (Velenis, 1998, Figure 4) and on-site mea-
surements. The commercially used Finite Element
program ANSYS (ver. 10.0) is used for the static,
modal and transient response of the structure, which
is modeled with 4000 (10-node) solid elements appro-
priately refined at the locations of the gate openings.
The particular FE modelling was adopted in order to
avoid unnecessary stress concentration at locations of
ubrupt geometry change at a reasonable computational
cost, especially towards dynamic analysis in the time
domain. Although various approaches have been pro-
posed for the modelling of masonry structures (Casolo,
1998, Galasco et al., 2004, Rota et al., 2005), with
emphasis on the non-linear response of the walls, the
particular structure was studied herein as purely lin-
ear elastic, in order to permit primarily focus on the
macroscopic dynamic response of the system under
simultaneous excitation along the three principal axes.
Subsequently both local material and geometric non-
linearities (inclusive of cracking and sliding at critical
locations) were not considered.
The structure was assumed as fully fixed at its base
since geotechnical and geophysical studies have shown
that the soil conditions at the particular city loca-
tion are very stiff (Anastasiadis et al., 2001). Based
on experimental results for monuments of the Byzan-
tine Era (Manos et al., 2004, Stylianidis and Sextos,
2006) an average uniform compression strength of
fmc = 2.0 MPa was adopted for the construction mate-
rials of the whole system for simplicity, based on
the weaker brick masonry. The corresponding ten-
sile strength was set equal to fmt = 0.15 MPa. The
Young’s modulus of the masonry was taken equal
to E = 3.5 × 106 kPa (corresponding to the compos-
ite material as a whole) and the self weight equal
to γ = 22 KN/m3 while 6% Rayleigh damping was
assumed for the circular frequency range ω = 40–
200 rad/sec. It is noted that the two twin gates of
the Trigonion Tower (i.e. Annis Palaiologinis Gate)
as well as the Pyrros Gate (Velenis, 1998), have
not been modeled due to their relatively smaller
dimensions compared to the overall Wall Section
length.

2.2 Earthquake Scenario for Thessaloniki

Following the development of the Finite Element
model of the structure, an effort was made to iden-
tify a ‘reasonable’ earthquake ground motion scenario
for the earthquake under study. Along these lines, the
three components of the seismic motion recorded at the
city center (Ambraseys et al., 2000) during the earth-
quake that stroke Thessaloniki in 1978 (20/06/1978,
Ms = 6.4, PGA = 0.15 g) were taken as the refer-
ence motion and were appropriately deconvoluted
(Figure 6) to the bedrock level at the location where
the records were obtained based on geotechnical data
from the Microzoning study of Thessaloniki (Anas-
tasiadis et al., 2001). As mentioned previously, the
local site conditions can bee assumed to correspond
to rock thus classified as soil type A according to
Eurocode 8. The outcrop motion along the axis of
the Walls Section under study was then derived. As
a result, three scenarios of earthquake excitation were
developed involving the:
a) earthquake ground motion of the 1978 seismic
event corresponding to the local site conditions as
illustrated in Figure 8 and described above. (Case
A: 1978 excitation)
Figure 6. Generation of earthquake ground motion sce-
nario. 1978 earthquake record (top) and outcrop motion for
the upper city Walls of Thessaloniki (bottom).
b) the above seismic motion scaled to the level of
PGA = 0.22 g, that is specified by the Microzoning
study of the city of Thessaloniki, for the particular
site. (Case B: Potential excitation corresponding
to an average return period of 475 years)
c) the predicted seismic motion of Case B, applied
asynchronously, that is, considering the time lag
that arises from the traveling of the seismic waves
in finite velocity through the soil media. (Case C:
Asynchronous excitation corresponding to an
average return period of 475 years).
It is noted that for all the above three cases, the
Wall complex is excited simultaneously along the three
principal directions, X,Y,Z. Due to its particular fea-
ture involving the consideration of spatial variation
of ground motion, Case C is further discussed in the
following section.

614
2.3 Spatially variable earthquake ground motion
along the Land Walls under study
Development of spatially variable seismic ground
motion scenarios for the study of the dynamic response
of extended structures (primarily bridges but also dams
and pipelines) is attracting increasing scientific atten-
tion after strong evidence that not only seismic motion
may differ substantially in terms of amplitude and fre-
quency content but also that this difference may have
a detrimental effect on the displacements and stresses
of the extended structure under certain circumstances
(Sextos et al., 2003). The sources of this spatial and
temporal variations of seismic motion have been well
identified: (a) waves travel at a finite velocity, hence
they arrive at consecutive support points with a time
delay; (b) wave coherency loss in terms of gradual
reduction of the waves’ statistical dependence with
distance and frequency, due to multiple reflections,
refractions and superpositioning of the incident seis-
mic waves during propagation and (c) variation in local
site conditions that strongly affects wave amplification
and frequency content.
For the particular Wall Section, and since the soil
conditions are generally stiff, it can be assumed that
from the above sources of spatial variability, the effect
of local site conditions and the loss of wave coherency
is rather small; as a result, the asynchronous excita-
tion is estimated solely on the phase lag as is can be
defined by the distance from the reference point and
the apparent velocity of seismic waves (taken equal to
Vapp = 2000 m/sec). In order to account for the afore-
mentioned time delay at all base points (nodes) of the
500 m long Wall complex, the Finite Element model
was internally programmed to permit the subsequent
excitation of each one of the 5830 supporting points i
with a distinct 3-D ground displacement Ui,x , Ui,y , Ui,z
which is a function of space and time t and is equal to:
where xi , yi , zi are the coordinates of a base point
i, Uref ,x , Uref ,y , Uref ,z the three component seismic
motion at the reference point, Vapp is the apparent
velocity of seismic waves and xref , yref , zref the coordi-
nates of the reference point (which does not coincide
with the zero coordinates due to the slight rotation
along the z-z- axis of the left part of the Wall).
615
Figure 7. Characteristic eigenmodes of the system.
3 SEISMIC PERFORMANCE OF THE LAND
WALLS UNDER STUDY
3.1 Dynamic characteristics of the long Wall circuit
under study
The dynamic characteristics of the system were ini-
tially investigated, and as anticipated despite the height
of the Walls, the overall complex is rather stiff (i.e.
its fundamental period is equal to 0.12 sec). It is also
notable that due to the presence of the intermediate
towers, there is no clear transverse mode; on the con-
trary the system vibrates on a large number of similar
local modes having similar modal contribution to the
overall response.
Clearly, the fundamental mode is also local as it
corresponds to the east main wall which is not only
unsupported laterally (i.e. no intermediate defensive
towers exist) but also is on average 2 m higher than
other parts of the Wall circuit. The most characteristic
from these modes are illustrated in Figure 7 where it is
seen that it is only after the 25th mode that the structure
vibrates (asymmetrically) as a whole.
3.2 Seismic response of Land Walls under study in
the time domain under synchronous excitation
Having assessed the dynamic characteristics of the
system, the Walls were subjected to ground motions
derived from the records obtained at the city center
during the 1978 seismic event and were appropriately
deconvoluted to account for the different soil con-
ditions at the between the recordings and the Land
Walls site. For the particular Case A earthquake sce-
nario, the maximum tensile stress developed in time
was approximately equal to 30% of the tensile strength
(equal to 0.15 MPa) assumed for the structure. Simi-
larly, the compressive stresses developed where found
also rather low (of the order of 25% of the overall
compression strength that was taken equal to 2.0 MPa).

Figure 8. Variation of normal stresses with time for three
characteristic locations.
Apparently, the Walls remain primarily in compression
and it is only few locations around the main gates that
develop tensile forces along particular critical sections
but just for a few peaks in time.
It can be claimed therefore, that the very good per-
formance and lack of any damage of the Wall Section
under study during the 1978 earthquake, is generally
verified. This can be attributed not only to the massive
dimensions of the walls, but also to the rocky (out-
crop) foundation conditions which deamplify ground
motions (compared to sites of softer soil formations as
seen in Figure 6).
It is interesting to notice that this reduction of PGA
(compared to the record site) appears to be more sig-
nificant than the fact that outcrop motion that excites
the Walls is of higher frequency content thus closer to
the natural period of the structure (i.e. T = 0.12 sec).
Having assessed the structure for the equivalent
level of the 1978 earthquake seismic forces, the lin-
ear elastic dynamic analysis in the time domain is
repeated for the same ground motion, this time scaled
to the level of PGA = 0.22 g, that is specified by the
Microzoning study of the city of Thessaloniki, for
the particular site (Case B). As seen in Figure 8, the
variation of normal stresses with time for three char-
acteristic locations of the Walls (i.e. West Gate, East
Gate and East main Wall), neither the compression nor
the tensile strength of the structure is exceeded at any
point or time.
It is only few (two-three) peaks for each criti-
cal location that the stresses developed reach the
available tensile strength of 0.15 MPa (being 0.15,
0.12 and 0.13 MPa respectively). On the contrary, the
‘effective’ tensile stress developed (as taken approxi-
mately equal to 2/3 of the maximum) does not exceed
0.1 MPa on average for the most critical locations of
the structure. It is noted however that strictly speaking,
the critical checks should be performed at masonry
616
sections instead of local points. Nevertheless, the over-
all assessment is that seismic demand developed is
clearly inferior to the estimated capacity of the Walls.
In terms of shear stresses, the most critical loca-
tion was identified to be the North side of the East
Gate as depicted in Figure 9. Notably, both the shear
stresses developed and the shear strength are a func-
tion of time as the latter depend on the compression
stresses which also vary in time. For the assessment of
the safety factor against shear, the following equation
was used according to the provisions of Eurocode 6:
where fvk (t)is the shear strength of unreinforced
masonry in time, fvk0 is the shear strength of mortar
(taken equal to 0.1 MPa) and σz (t) is the normal com-
pression stress at the same location. As it is seen in
Figure 9, the safety factor against shear varies between
1.5 and 2.5 in the extreme case, but it can be estimated
to be clearly greater than 2.0 on average. As a result,
as was observed for the normal (tensile and compres-
sion) stresses the performance of the Wall structure is
very good also for the Case B earthquake scenario.
This observation is interesting in the sense that, as
the Microzoning-induced seismic intensity refers to
a return period of 475 years, it can be claimed that the
particular level of seismic forces predicted for T = 475
years is not adequate to produce significant damage
to the Wall structure a fact that is anticipated since the
structure is not known to have suffered extensive dam-
age over the entire period of 16 centuries. As a result,
both the satisfactory future seismic performance of
the Walls and the predictions of the seismic hazard
assessment for the Thessaloniki area are (if not explic-
itly confirmed) at least not counteracted by the refined
numerical analysis.
3.3 Seismic response of Land Walls under
study in the time domain under αsynchronous
excitation
Following the dynamic analysis of the system for
Case A and Case B earthquake scenarios, it was also
deemed interesting to investigate the potential effect of
ignoring (as it is commonly the case) the propagating
nature of the travelling waves. As already discussed,
the Walls where excited with 5830 spatially and tem-
porary variable ground displacement vectors that were
enforced to the corresponding base nodes according
to equations (1)-(3). Figure 10 illustrates the varia-
tion of normal stresses with time under asynchronous
excitation for the three characteristic and most crit-
ical locations identified (i.e. West Gate, East Gate
and East main Wall). By comparing these stresses
derived from the synchronous (Figure 8) and asyn-
chronous excitation, it can be observed that although

Figure 9. Variation of shear stresses with time. East city
Walls gate (North view). Seismic response of Land Walls
under study in the time domain under synchronous excitation.
the response can be considered as rather similar, the
extremes of the stresses are different. In particular, the
extreme tensile stress at the East Gate has increased by
50% (from 0.12 MPa to 0.19 MPa) while the extreme
tensile stress at the East Main Wall has decreased
by 40% (from 0.13 MPa to 0.08 MPa). Especially for
the middle point (on the East Gate) the maximum
stresses occur at a slightly different time (7.15 sec
instead of 7.02 sec). Such distinct response under asyn-
chronous earthquake ground motion has also been
observed in a large number of bridges (Sextos et al.,
2003) and is pronounced in the case of long and
stiff (monolithical) structures (Tubino et al., 2003)
due to the excitation of higher (primarily antisymmet-
ric) modes (Sextos et al., 2003) and the subsequent
development of additional pseudo-static forces. How-
ever, it is also noted that the ‘effective’ tensile and
compressive stresses are again lower than the esti-
mated corresponding strength, hence the comments
made regarding the satisfactory historical and future
performance of the Land Walls under study remain
valid.
In can be generally stated though, that the dif-
ferences observed between the synchronous and the
asynchronous excitation of the particular Wall Section,
even if they do not impose a threat for the structural
integrity of the system, are not negligible if one con-
siders that due to the very stiff (rock) formations at the
617
Figure 10. Variation of normal stresses with time under
asynchronous excitation for three characteristic locations.
base of the Wall circuit, the variability of earthquake
ground motion was attributed only to the time delay of
the seismic waves and not to equally if not more impor-
tant coherency loss and local site conditions variation.
As a result, the spatially variable nature of the incoming
seismic waves should not be a-priori neglected when
assessing the seismic performance of long and high
wall structures, as there is high probability that ignor-
ing asynchronous excitation, certain aspects of their
dynamic response that are related to their inherent stiff-
ness as a structural system and the potential excitation
of higher modes may be disregarded.
3.4 Seismic response of simple stand-alone
walls at different city locations
Following the numerical investigation of the seismic
performance of the long upper city Wall circuit, case
A scenario (1978 earthquake excitation) was applied
for the free standing wall of the second location under
study (Stylianidis and Sextos, 2006). Similarly to the
long Wall complex, good performance was also ver-
ified as neither loss of stability due to rocking nor
exceedance of its tensile strength was observed dur-
ing the application of the 1978 ground excitation.
Moreover, it was estimated that the particular sys-
tem studied could resist at least twice the earthquake
load imposed during the 1978 earthquake without
significant damage.
4 CONCLUSIONS
This paper is an attempt to propose a refined approach
towards the assessment of the seismic performance
of a part of the Land Byzantine Walls of the city
of Thessaloniki. Utilizing advanced numerical tools
and generating various earthquake ground motion sce-
narios, both a 500 m long section of the Walls and
a small free-standing part of the Walls at a second
location were studied in the time domain under using
 synchronous and asynchronous excitation. The main
conclusion drawn can be summarized as follows:

• It is feasible to study the dynamic and seismic

response of extended systems such as high Walls
through refined finite element models and reliable
estimates of the overall geotechnical, seismological
and structural conditions.
• For the first long Wall Section under study, its good
performance during the severe 1978 Thessaloniki
earthquake (20/06/1978, Ms = 6.4, PGA = 0.15 g)
was verified. A similarly good performance was
numerically estimated for the second free-standing
Wall where the results also indicated that the par-
ticular system studied, not only performed well
during the 1978 earthquake but also it could resist
at least twice the earthquake load imposed without
significant damage.
• It was also numerically predicted that both the par-
ticular Wall Sections should not be expected to
suffer significant damage for earthquakes with a
return period of 475 years. As a result, the satis-
factory future seismic performance of the Walls as
well as the predictions of the seismic hazard assess-
ment for the Thessaloniki area are (if not explicitly
confirmed), at least not counteracted by the refined
numerical analysis.
• The overall process adopted and proposed may be
potentially implemented in a larger scale as a means
to trace (through back analysis) the seismic history
of monuments and possibly to assess an upper bound
of the historical seismicity of a particular region
or city.
• In the framework of the process proposed, the
spatially variable nature of the incoming seismic
waves should not be a-priori neglected, since it was
shown for the long Wall circuit that (as for bridges
and other stiff and long structures) the excitation
of higher modes may impose significant pseudo-
static forces that did not develop under synchronous
excitation.
• Further investigation is certainly required with
emphasis on the potential effect of material or
geometrical non-linearities (i.e. cracking and/or
sliding/rocking at certain interfaces). Moreover, it
is evident that the reliability of the methodology
adopted is strongly dependent on the reliability in
estimating the mechanical properties of the super-
structure and foundation soil materials involved
(which can be also spatially variable), as well as on
the accuracy of the geometrical data that are used
for finite element development.

REFERENCES

Figure 11. Variation of normal stresses (σz ) with time for Ambraseys N., Smit, P., Berardi, R., Rinaldis, D., Cot-
Case A scenario (excitation with 1978 earthquake record). ton, F. and Berge. C. 2000. Dissemination of European

618
 Strong-Motion Data. CD-ROM collection. Council of
European Communities. Environment and Climate Pro-
gramme. ENV4-CT97-0397. Bruxelles. Belgium.
Anastasiadis A., Raptakis D. and Pitilakis K. 2001. Thes-
saloniki’s Detailed Microzoning: Subsurface Structure
as basis for Site Response Analysis, Pure and Applied
Geophysics – Pageoph, Vol. 158(12), 2597–2633.
ANSYS Engineering Analysis System 2005. User and Theo-
retical Manual. ANSYS, Inc., Southpointe, Canonsburg,
Pennsylvania, Version 10.0.
Bouras, C. 2002. The Economic History of Byzantium: From
the Seventh through the Fifteenth Century, Laiou, A. (Edi-
tor in-Chief), Dumbarton Oaks, Trustees for Harvard
University, Washington,D.C.
Bommer, J. 2002. Deterministic vs. Probabilistic Seismic
Hazard Assessment. An exaggerated and obstructive
dichotomy, Journal of Earthquake Engineering, 6 (special
issue no. 1), 43–73.
Casolo, S. 1998. A three-dimensional model for vulnerability
analysis of slender medieval masonry towers, Journal of
Earthquake Engineering, Vol. 2, No. 4, 487–512.
CEN. 2004. Eurocode 8: Design of structures for earthquake
resistance. EN 1998-1 General rules, seismic actions and
rules for buildings, Brussels.
CEN. 2004. Eurocode 6. Design of masonry structures. ENV
1996-1 General Rules’, Brussels.
619
Galasco, A., Lagomarsino, S., Penna, A. and Resemini.
S. 2004. Non-linear seismic analysis of masonry struc-
tures, 13th World Conference on Earthquake Engineering,
Vancouver, Canada, Paper No. 843.
Rota, M., Pecker, A., Bolognini, D. and Pinho, R. 2005.
A methodology for seismic vulnerability of masonry
bridge walls, Journal of Earthquake Engineering, Vol. 9,
Special Issue 2, 331–353.
Sextos, A., Pitilakis, K. and Kappos, A. 2003. A global
approach for dealing with spatial variability, site effects
and soil-structure-interaction for non-linear bridges: Part
1: Methodology and analytical tools, Earthquake Engi-
neering Structural Dynamics, 32, 607–629.
Stylianidis, K. and Sextos, A. 2006. Seismic response of
Byzantine Walls as a means of assessment the seismic his-
tory of the city of Thessaloniki, 1st Hellenic Conference
on rehabilitation works, Thessaloniki, CD-ROM volume,
Paper No. 084 (in Greek).
Tubino, F. Carassale, L. and Giovanni, S. 2003. Seismic
response of multi-supported structures by proper orthogo-
nal decomposition, Earthquake Engineering & Structural
Dynamics, 32, 1639–1654.
Velenis, G. 1998. The Walls of Thessaloniki: From Kasandrus
to Heraklios, University Studio Press, Thessaloniki, (in
Greek).
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Earthquake based condition monitoring of the Katholikon of Dafni

Monastery

H.P. Mouzakis
National Technical University of Athens, Greece

A. Miltiadou & N. Delinickolas

Hellenic Ministry of Culture, Athens, Greece

P. Touliatos & J. Dourakopoulos

National Technical University of Athens, Greece

ABSTRACT: The Katholikon of Dafni suffered throughout its history a large number of earthquakes that
caused many structural problems and damage to it. After the catastrophic earthquake that affected the region
of Attica during the September 1999, and in the frame of a strategic plan undertaken by the Hellenic Ministry
of Culture for the conservation and restoration of the monument including its mosaics, a seismic monitoring
system has been installed. This paper presents the monitoring system installed and the analysed data recorder
during four earthquakes. Two of them took place before the grouting injections and the other two some months
after the implementation of grouting to almost the half of the height of the perimeter walls. The comparison of
digitized data during weak earthquakes, occurred before and after grouting application show the effect to the
overall behaviour of the structure.

1 INTRODUCTION

The Katholikon of Dafni Monastery, already inscribed

in the world heritage list of UNESCO, is one of the
major-byzantine (11th c.) monuments famous world-
wide for its excellent mural mosaics (Millet. G. 1899,
Bouras, CH. 1998, Delinikolas et al. 2003). The Katho-
likon (main church of the monastery) belongs to
the octagonal type and comprises the main church,
the sanctuary, the narthex and four chapels, which
complete its orthogonal plan.
In the western part, only the perimetric walls of
an exonarthex or portico and the spiral stairway have
survived.
The central part of the main church is cross shaped
in plan, and over its square core rises the hemispherical
Figure 1. General view of the monument from N-E, metallic
dome (Figs 1 and 2), which stands on an almost cylin- buttresses are installed as urgent measures.
drical drum with 16 piers and 16 vaulted windows. The
dome and its drum are curried by eight pendentives and As one can clearly distinguish in Figures l and 2, the
eight arches (four semicircular and four squinches in exterior face of the vertical perimeter walls are built
the corners), forming an octagon and achieving in this according to the cloisonné masonry system, follow-
way the transition from circle to square. Thus twelve ing two different types of masonry. In fact, there is a
piers, laying out in a square plan, provide support to lower zone, built with large dimension stones (often
the dome together with the groin vaulted arms of the reused old material), their length placed horizontally
cross, situated at a higher level. The interior of the or vertically, in order to form crosses, and an upper
church is decorated with mural mosaics. In Figure 1 a zone constructed with smaller stones (Delinikolas
general view of the monument is presented. et al. 2003). Using radar techniques and endoscopy

621

Figure 2. Past interventions to the monument: North façade.
The different types of construction, mainly of the upper and
lower part of the monument, are shown.
(Vintzileou et al. 2004), it was shown that in both cases
a three-leaf masonry is used. As expected the width of
each of the three leaves at the lower and the upper
part is not the same due to the bigger stones used in
the external leaf of the lower part. The existence of
three-leaf masonry for the construction of the internal
vertical walls was also ascertained (Côte et al 2008),
while plain brick masonry was used in the construc-
tion of all the monolobe, bilobe and trilobe windows,
and all the vaulted roof.
The various materials and construction techniques,
the foundation and the soil condition, as well as the
past modifications and local reconstructions taken
place to remediate damages provoked due to strong
earthquakes, affect significantly the mechanical and
dynamic characteristics of the structure. The varia-
tions of temperature, settlement, creep and ambient
vibrations have also a great influence on the seismic
response of the monument. As a result, cracks occur
and divide the whole structure in interacting parts with
nonlinear behaviour.
Due to these aforementioned uncertainties in addi-
tion to the variability of the seismic excitation, even if
detailed accurate finite element models and sophis-
ticated analysis is developed, the estimation of the
seismic safety of the structure is not enhanced.
In order to increase our understanding of the seismic
structural response and to decrease the uncertainty of
the seismic action, the use of an earthquake monitor-
ing system is necessary. Moreover, the implementation
of the recorded and analyzed data will be the basis
for the further development and calibration of detailed
numerical models, which will be used for the para-
metric study of the seismic behaviour before and after
interventions.
Such monitoring systems have already installed
in Hagia Sophia and Suleimaniye in Constantinople
(Durukal, E. et al. 2003) and recorded the response
during strong earthquakes, as part of the efforts for
the determination of their earthquake worthiness.
The Katholikon of Dafni Monastery suffered severe
damage during the strong Athens’ earthquake of 1999
622
and supporting urgent measures were subsequently
undertaken (Miltiadou-Fezans et al. 2003a), together
with all the necessary investigations and studies,
including also a first series of finite element analyses
for the numerical verification of the existing damage of
the structure (Miltiadou-Fezans et al. 2003b & 2004).
This analysis helped for a better understanding of the
structural behaviour of the monument and guided the
design of the monitoring system. The monitoring sys-
tem was installed in 2003 after the implementation of
the aforementioned emergency measures and before
the application to the masonry structure of any repair
intervention. Thus, when the hydraulic lime grouting
application started (Miltiadou et al. 2008), in June
2006, the system was functioning and had already
recorded some weak earthquakes.
This paper presents the monitoring system that
was installed to record the seismic response of the
Katholikon of Daphni Monastery and the analysed
data recorder during four earthquakes. Two of them
took place before the grouting injections and the rest
some months after the first phase of grouting. During
this first phase, grouting was implemented to almost
the half of the height of the perimetric walls. The dig-
itized data during weak earthquakes which occurred
after the application of the injections confirm the effi-
ciency of grouting interventions and their effect toe
the overall behaviour of the structure.
2 MAJOR 19TH AND 20TH CENTURY
INTERVENTIONS
The monument, during its life (from 11th century
to our days) has been subjected to various modifi-
cations, it suffered numerous earthquakes, whereas
major interventions for its preservation were applied
in the 19th and 20th centuries (Delinikolas et al. 2003,
Miltiadou-Fezans et al. 2004). During the last two
centuries strong earthquakes have severely affected
the church. The 1889 earthquake (M = 6.7 R), the
1894 (M = 7.0 R), the 1914 (M = 6.0 R) the 1981
(M = 6.7 R) and the 1999 (M = 5.9 R) have to be men-
tioned. The major intervention (1890–1897) include
the following: Demolition and reconstruction of the
(heavily damaged) central dome, repairs in the lower
parts of the church, removal and partial replacement
of mosaics that were heavily damaged. Installation of
three concentric iron rings (I-beams) was carried out at
the base of the drum of the dome. After, the 1894 earth-
quake the narthex, its collapsed groined vaults and
the NE chapel were reconstructed. Between 1897 to
1907, the two stone buttresses to both sides of the north
entrance of the church were constructed and metallic
devices were installed around the two south masonry
piers aiming to their confinement. For the same pur-
pose, in 1954, an iron ring was installed in the external
face of the upper part of the drum over the windows.
Table 1. Grouting periods.
Phase Start End
I 26/6/2006 4/8/2006
II 15/12/2006 4/4/2007
The 1981 earthquake caused numerous hair cracks to
the building and damages to the mosaics. Finally, the
1999 earthquake caused severe damages to the mon-
ument. Extensive cracking of arches was recorded,
the N-E corner was severely damaged, and horizontal
cracks appeared along the perimeter of the drum (both
at its base and top). Severe damages were also recorded
at the 16 pillars of the drum (Miltiadou-Fezans A. et al.
2003b and Miltiadou-Fezans A. et al. 2004).
3 EMERGENCY SUPPORTING AND PHASES
OF STRUCTURAL RESTORATION SCHEME
Taking into account the damage of the structure and
the danger of eventual aftershocks temporary mea-
sures were urgently taken (Miltiadou-Fezans A. et al.
2003). Independent new steel elements were designed
and constructed to provide vertical support to various
arches. A system of three steel buttresses is used, at
the NE corner of the monument, where a pronounced
tilting of the external walls was noticed. This system
is able to carry horizontal seismic loads to especially
constructed concrete footings.
A first phase of structural restoration interven-
tions has been decided and implemented. It comprised
all the measures considered necessary for the repair
and strengthening of masonry elements (stitching and
deep re-pointing where necessary, systematic grouting
injection to fill voids and cracks, local reconstruc-
tions, etc). The design of a second phase of inter-
ventions (installation of ties, metallic confinements,
diaphragms, etc) is under elaboration, and will take
into consideration all the data collected from research
programs, in situ surveys during the first phase of
interventions and the monitoring system.
In the framework of the first phase of interven-
tions, a hydraulic lime based grout was selected and the
application of injections to the monument was imple-
mented in two time periods (Miltiadou-Fezans A. et al.
2007 & 2008), as shown in Table 1.
The seismic monitoring system was installed on
the monument before the beginning of restoration
works, in order to collect data before, during and
after the completion of the interventions, with the
aim to estimate the effect of the intervention mea-
sures to the overall behaviour of the structure. In fact
the results presented in this paper are based on the
elaboration of data collected from four earthquakes.
The first earthquake took place after the application of
623
0 80
PEAK ACCELLERATION IN G
0 60
0 40
0 20
0 00
10 100 1000 10000
RETURN PERIOD
Figure 3. Peak ground acceleration and return period.
the urgent supporting measures and before the begin-
ning of masonry repair, while the second one occurred
during the preparation of masonry for grouting by
stitching of cracks and local deep re-pointing. After
three and four months from the end of the first grout-
ing period, during which grouting was implemented up
to the springing level of groin vaults, two more weak
earthquakes were recorded.
4 SEISMIC HAZARD EVALUATION
The selection of the technical specifications of the
accelerometers that are installed on the monument and
on soil surface has to be based on the quantitative esti-
mation of the acceleration of the ground motion, which
is assessed by seismic hazard analysis. It was evalu-
ated that the annual mean peak horizontal acceleration
with mean frequency of two events per year is 0.059 g,
while the maximum expected peak acceleration on soil
surface is 0.80 g, as shown in Figure 3. Considering the
standard deviation of the used sample earthquake data
and in order to have a high signal to noise ratio, the
noise level was selected to be 0.001 g.
5 SEISMIC MONITORING SYSTEM
A digital dynamic monitoring system has been
installed to record the response of the core of the mon-
ument. Accelerometers and displacement meters are
used to measure acceleration and relative displace-
ment at characteristic locations. These locations were
selected in order to avoid local vibrations of indi-
vidual masonry units during earthquakes, taking into
consideration the existing cracking patterns.
The accelerometers were placed at five character-
istic levels (a) on ground level free field; (b) on the
cornice where the central arches and the squinches
are springing; (c) on the base of the drum; (d) on
the base of dome (top of the drum). The displacement
transducers were placed to record the relative displace-
ment between north and south as well as east and west
internal walls of the central part of the church.
Figure 4. Control room. Figure 6. Accelerometers located at the Drum’s Base.

Figure 7. Accelerometers located at the Cornice and

Figure 5. Accelerometers located at the Dome’s base. Ground.

The monitoring system is based on National Instru-

ments PXI and SCXI units that are controlled by a
Labview application.All signals are collected by a cen-
tral unit. A special program was developed by NTUA
to control the unit, to digitize data using pre-trigger
and post-trigger events and a voting system for the
initialisation of the recording. The resolution of A/D
converter is 16-bits, the sampling frequency is 200 Hz.
Antialising Butterworth filters with a cut-off frequency
100 Hz are connected, the clock of the system is syn-
chronised by a GPS, while there is a DC power supply
that acquire the uninterrupted function of the system
for two days. The acquisition system is placed in a con-
trol room as shown in Figure 4. Data are saved locally
and can be transmitted on line to the analysis centre
at NTUA, while it is capable to call the users when a
Figure 8. Displacement transducers.
seismic event is captured.
Fourteen accelerometers (Bernhard, B. 2002) and
eight displacement transducers were installed to can be recorded is 2 g, while the noise of the
record acceleration and relative displacement respec- system – accelerometers-filters-A/D converter is
tively. The selection of the sensors was based on 0.001 g. In Figures 5 to 8 the instrumentation setup is
the Seismic Hazard. The maximum acceleration that presented.

624
Table 2. Seismic events. accelerometers that have a frequency of 1.38 kHz and
damping coefficient ζ = 70%.
Recorded The acceleration time histories of the earthquake of
event 1 2 3 4
08/01/06 for channels 1–14 are presented in Figure 10,
Location South Kythira Penteli Fthiotis respectively. A lowpass digital filter with cutoff 25 Hz
Euboea (Attica) and a highpass with cutoff 0.7 Hz were used for data
Date 26/09/04 08/01/06 05/11/06 12/12/06 processing.
Time GMT 03:06 11:34 22:55 05:50 Total acceleration response spectrum (Nigam, N.C.
Latitude 38.31 36:21 38:05 38:62 & Jennings, P.C. 1969) on soil surface for the same
Longitude 24.18 23:41 23:87 23:31 earthquake for damping coefficient 5% are shown in
Epicenter Euboea Off the 15 km 80 km Figure 11.
eastern east of N.N.W The amplification factor between the maximum
coast of Athens of Athens.
acceleration at various locations on the structure and
the island (Penteli Malesina
of Kythira Area) Fthiotis the acceleration on the soil surface of the same
Magnitude 4.2 6.4 3.7 4.5 orientation is given in Table 3.
(Richter) The damping coefficient of the first mode is given
Epicentral 20 km 69 km 21 km 20 km in Figure 12. The natural period of the first mode was
Depth changed during earthquakes as shown in Figure 13.
Epicentral 58 km 402 km 21 km 73 km
Distance
Characteristics Weak Strong Weak Weak 8 PRESENTATION OF MAIN RESULTS
Data Source NOAIG NOAIG OASP- OASP-
NOAIG NOAIG
8.1 Earthquake 16/09/2004
The acceleration of the northern region of the base of
the dome along E – W direction is higher than the
acceleration of the southern region. The acceleration
of the west region of the base of the drum is higher
than that of the east region. Period of the longitudinal
and transversal translational modes were not changed
during the earthquake. On the cornice the acceleration
of the south – west corner along E – W direction is
higher than the acceleration of the north – east one.
The drum was mainly deformed and consequently the
acceleration on the base of the dome was increased.

8.2 Earthquake 08/01/2006

Figure 9. Epicentres of recorded earthquakes.
6 RECORDED EVENTS
Table 2 shows the seismic events recorded by the mon-
itoring system, and gives information about the earth-
quakes. The epicentres of the recorded earthquakes are
presented on the map of Figure 9.
7 PROCESSING OF STRONG MOTION DATA
The digitised signals were corrected taking in
to account the dynamic characteristics of the
The period of the monument along E – W direction
was 0.27 sec at the beginning of the earthquake and
it was found to be 0.30 at its end. The period along
N – S direction was 0.32 sec and it was changed to
0.39 sec. Consequently, the stiffness of the structure
was changed, probably due to a further deterioration
of the damaged structure, as no intervention was yet
implemented.
8.3 Earthquake 05/11/2006
This specific earthquake occurred three months after
the completion of the first period of grouting during
which injections were implemented to almost the half
of the height of the perimeter walls. Periods and damp-
ing along both main directions were found to be lower.
The dynamic characteristics of the structure did not
change during the earthquake. These results show that,
although only half of the structure has been repaired by
means of hydraulic lime grouting injections, the effect

625
 0.2

m/sec2
0.1
0
-0.1
-0.2
15 20 25 30 35 40 45 50 55 60 65 70
Seconds
0.2

m/sec2
0.1
0
-0.1
-0.2
15 20 25 30 35 40 45 50 55 60 65 70
Seconds
0.3
m/sec2

0.1
-0.1

-0.3
15 20 25 30 35 40 45 50 55 60 65 70
Seconds
1.5
m/sec2

0.5
-0.5

-1.5
15 20 25 30 35 40 45 50 55 60 65 70
Seconds
1.5
m/sec2

0.5
-0.5

-1.5
15 20 25 30 35 40 45 50 55 60 65 70
Seconds
2
m/sec2

1
0
-1
-2
15 20 25 30 35 40 45 50 55 60 65 70
Seconds

1.5
m/sec2

0.5
-0.5

-1.5
15 20 25 30 35 40 45 50 55 60 65 70
Seconds
1
0.5
m/sec2

0
-0.5
-1
15 20 25 30 35 40 45 50 55 60 65 70
Seconds
0.6
m/sec2

0.2

-0.2

-0.6
15 20 25 30 35 40 45 50 55 60 65 70
Seconds
1.5
m/sec2

0.5
-0.5

-1.5
15 20 25 30 35 40 45 50 55 60 65 70
Seconds
0.6
m/sec2

0.2

-0.2

-0.6
15 20 25 30 35 40 45 50 55 60 65 70
Seconds
1
m/sec2

0.5
0
-0.5
-1
15 20 25 30 35 40 45 50 55 60 65 70
Seconds
0.3
m/sec2

0.1

-0.1

-0.3
15 20 25 30 35 40 45 50 55 60 65 70
Seconds

Figure 10. Corrected acceleration time histories of the earthquake 08/01/06.

of grouting was clearly distinguished, as the dynamic
characteristics of the structure have changed.
8.4 Earthquake 12/12/2006
This specific earthquake occurred almost one month
after the previous one, four months after the
completion of the first period of grouting. There was
no other new element that could have affected the
characteristics of the monument except the evolution
of grout hardening. The period of the structure was the
same along the N-S direction, while along E-W direc-
tion a very slight increase was recorded during the
earthquake. The damping of the structure was higher
than that of the previous event but it remained lower
compared to the damping before the first period of
injections.

626
 0.7 Table 3. Amplification factor.
Damp. 5.0%
0.65

0.6 Amplification Factor

0.55 CH i/CH2 (S-N) ή CH i/CH3 (W-E)
Response Acceleration [m/sec2]

0.5 Direction /
0.45 channel Position 26/09/04 08/01/06 05/11/06 12/12/06
0.4

0.35 W–E/ Dome’s 8.39 4.78 6.99 9.14

0.3
Ch4 Base
0.25
W–E/ Dome’s 7.21 5.47 7.82 8.66
Ch5 Base
0.2
S–N/ Dome’s 4.20 8.30 4.20 8.69
0.15 Ch6 Base
0.1 W–E/ Drum’s 4.90 4.49 4.82 5.28
0.05 Ch7 Base
0 W–E/ Drum’s 3.53 3.96 4.22 5.65
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 Ch8 Base
Period [sec]
S–N/ Drum’s 4.35 3.14 1.50 3.11
0.75 Ch9 Base
0.7
Damp. 5.0% W–E/ Drum’s 7.00 4.39 6.91 8.29
Ch10 Base
0.65
W–E/ Cornice 2.44 2.24 – –
0.6 Ch11
0.55 S–N/ Cornice 2.70 4.43 2.42 4.46
0.5
Ch12
Response Acceleration [m/sec2]

W–E/ Cornice 3.30 0.84 – –

0.45
Ch13
0.4 S–N/ Cornice 2.36 0.05 – –
0.35 Ch14
0.3
0.25
10 00
0.2 9 00 9 30
8 68
8 00
0.15 7 75 7 61
7 00
0.1 6 00 6 92
5 84
5 00
0.05 4 80
4 00
0 3 00
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2 00
1 86
Period [sec] 1 00
0 00
26/9/2004 8/1/2006 5/11/2006 12/12/2006
1.05
1 Damp. 5.0% West - East South - North
0.95
0.9
0.85 Figure 12. Damping coefficient.
0.8
0.75 0 45
0.7
Response Acceleration [m/sec2]

0 40
0.65 0 32 0 39
0 35
Period Ti (sec)

0.6 0 30
0 30
0 27 0 26 0 26
0.55 0 25
0 20
0.5 0 20
0 22
0.45 0 15
0.4 0 10
0.35 0 05

0.3 0 00
26/9/2004 8/1/2006 5/11/2006 12/12/2006
0.25
West - East South - North
0.2
0.15
0.1
Figure 13. Variation of the period of the first mode.
0.05
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Period [sec]
were taking place, together with the stitching and deep
re-pointing of the cracks on the extrados of the vaults
Figure 11. Total acceleration Response Spectrum and this could have an effect on the dynamic character-
(a) Vertical; (b) South – North; (c) West – East direction. istics of the overall structure. It will be very interesting
to compare the existing records with the ones of future
The differences presented between the last two weak earthquakes, as from April 2007 the first phase
earthquakes can be probably attributed to the fact that of interventions has been accomplished, and injections
at that period local reconstructions of external face have been implemented until the level of the base of
of masonry of the S and W aisle of the central cross the drum.

627
9 CONCLUSIONS
The monitoring system installed to the Katholikon of
Dafni Monastery to record the seismic response of the
structure during weak earthquakes has been proved
very useful, as the analysed data collected until now
during four earthquakes give important information
for its behaviour before interventions, as well as after
application of grouting until half height of the walls.
After the urgent measures and before any interven-
tion the stiffness of the structure was decreased during
the earthquake 08/01/2006. This is totally expected,
as no remedial measures had been applied, while
the emergency measures have not been designed to
undertake seismic actions.
As a result of the implementation of grouting injec-
tions to the half height of the walls the period of
the first mode was changed. Specifically, the periods
were decreased. The damping of the monument before
restoration measures was 7.75% and 9.30% along
E – W and N – S directions respectively.After the appli-
cation of injections of hydraulic lime based grout to
the half height of the structure, damping was decreased
to 4.8% for E – W and 7.6% for N – S direction. Thus,
it was proved that the effect of grouting on the overall
behaviour of the structure can be evaluated by a seismic
monitoring system, under the condition to have earth-
quakes recorded both before and after the application
of grouting.
All these data will be used for the further analyses
of the structural behaviour of the Katholikon using
adequate methods and models, while complementary
instrumentation has now been installed to cover the
whole monument, apart from the central core already
monitored since 2003.
All the instruments are going to stay on the struc-
ture in an almost permanent way, to record the seis-
mic behaviour of this important monument of world
heritage.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the support
of the Electronic Engineer S. Asimakopoulos who has
been responsible for the installation and the opera-
tion of the system. The project has been undertaken
by the Hellenic Ministry of Culture in collaboration
with the National Technical University of Athens and
co-funded by the European Regional Development
Fund (ERDF-75%) and by National Funds (25%).
REFERENCES
Bernhard, B. 2002. Measuring with Crystals: Principles and
applications of the piezoelectric measuring technique,
pages 70.
Bouras, Ch. 1998. The Daphni monastic complex recon-
sidered, AETOS studies in Honor of C. Mango, B.G.,
Teubner, Stutgart & Leipzig.
628
Côte, Ph., Dérobert, X., Miltiadou-Fezans, A.,
Delinikolas, N., Durand, O., Alexandre, J., Kalagri, A.,
Savvidou, M., Chrysopoylos, D.., Anamaterou, L. and
Georganis, F. 2008. Application of non-destructive tech-
niques at the Katholikon of Daphni Monastery for
mapping the mosaics substrata and grouting monitor-
ing. Submitted to the 6th International Conference on
StructuralAnalysis of Historical Constructions, Bath, UK.
Delinikolas, N., Miltiadou-Fezans, A., Chorafa, E. and
Zaroyianni, E. 2003. Study on restoration of the Katho-
likon of Dafni Monastery: Phase A-Architectural and
historical Survey. Hellenic Ministry of Culture (in Greek,
with an extensive bibliography).
Durukal, E. et al. 2003. Dynamic Response of two Histori-
cal Monuments in Istanbul Deduced from the Recordings
of Kocaeli and Düzce Earthquakes. Bull.Seism.Soc.Am.,
v.93-2, pp. 694–712.
Millet, G. 1899. Le monastère de Daphni, Histoire,
Mosaiques. Monuments de l’Art Byzantin, I, Paris.
Miltiadou-Fezans, A., Tassios, T. P., Delinicolas N.,
Chorafa, E., Zarogianni, E. and Candrinos, I. 2003a.
Earthquake structural problems and urgent measures
undertaken to support the Katholikon of Dafni Monastery
in Athens, Greece, Proceedings of the 8th International
Conference STREMAH 2003 “Structural Studies, Repairs
and Maintenance of Heritage Architecture”, Halkidiki,
Greece, 7–9 May 2003.
Miltiadou-Fezans, A., Delinikolas, N., Chorafa, E. and
Zaroyianni, E. 2003b. Study on restoration of the Katho-
likon of Dafni Monastery, Phase A-Structural Survey,
Analysis and remedial measures. Hellenic Ministry of
Culture (in Greek).
Miltiadou-Fezans, A., Vintzileou, E., Delinikolas, N.,
Zaroyianni, E. and Chorafa, E. 2004. Pathology of Dafni
Monastery: survey, monitoring of cracks, interpretation
and numerical verification. Proceedings of the 4th Inter-
national seminar on StructuralAnalysis of Historical Con-
structions,Vol.2, Padova, 10–13 November 2004, Mod-
ena C., Lourenco P.B., Roca P. (ed.), Balkema Publishers,
London.
Miltiadou-Fezans, A. Kalagri, A. and Delinikolas, N. 2007.
Design of hydraulic grout and application methodology
for stone masonry structures bearing mosaics and mural
paintings: the case of the Katholikon of Dafni Monastery,
Proceedings of the International Symposium: SHH07-
Studies on Historical Heritage, Edited by Görün Arun,
Antalya, Turkey, 17–21 September 2007.
Miltiadou-Fezans, A., Kalagri, A., Delinikolas, N.,
Kakkinou, S. and Ziagrou, A. 2008. Methodology for in
situ application of hydraulic grouts on historic masonry
structures. The case of the Katholikon of Dafni Monastery.
Submitted to the 6th International Conference on Struc-
tural Analysis of Historical Constructions, Bath, UK.
Nigam, N.C. & Jennings, P.C. 1969. Calculation of Response
Spectra from Strong Motion Earthquake Records. Bul-
letin of the Seismological Society of America, Vol.59,
pp. 909–922.
Vintzileou, E., Miltiadou-Fezans, A., Palieraki, V. and
Delinikolas, N. 2004. The use of radar techniques and
endoscopy in investigating old masonry: the case of
Dafni Monastery. Proceedings of 4th International Sem-
inar on Structural Analysis of Historical Constructions.
Modena C., Lourenco P.B., Roca P. (ed.), Rotterdam,
Balkema, Vol. 2, pp. 351–360.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Continuous and discontinuous modeling of the structures in

Bayon temple, Angkor

T. Maeda & T. Yamamoto

Department of Architecture, Waseda University, Tokyo, Japan

T. Hirai
Obayashi Corporation, Tokyo, Japan

ABSTRACT: Bayon temple, Angkor, consists of dry-masonry structures made of sandstone. Strong wind in
the rainy season is suspected as one of the causes for its progressive deterioration. Since 2003, micro-tremor has
been measured at most of the tower-type structures and the continuum-equivalent elastic modulus was found
to be 1/9 to 1/27 of that of the sandstone specimen in laboratory tests. In 2006, micro-tremor was measured at
libraries in Bayon temple and the similar reduced equivalent modulus was also obtained for these frame-like
sandstone-dry-masonry structures. This extremely low elastic modulus implies the inadequacy of continuum
modeling for bearing capacity of dry-masonry structures. Then the Discontinuous Deformation Analysis was
applied to safety evaluation of the library against wind load. It was an illustrative example; however, the evaluated
safety margin for the wind velocity of 40 m/s is about two.

1 INTRODUCTION

Bayon temple, Angkor, built in late 12th century to

early 13th century consists of sandstone-dry-masonry
structures. It has been deteriorated possibly by rain,
plant intrusion, settlements, and so on. Strong wind
of more than 20 m/s was observed in the rainy season
(JSA 1995), which may vibrate the structures either
directly or via ground motion and eventually cause
damage on them.
The temple consists of a main tower, sub-towers,
and two libraries as shown in Figure 1. The main
tower is more than 30 m high, located at the top of
the man-made mound of about 15 m high. Sub-towers
stand from several different height levels of the
mound, which are interconnected via the inner and
the outer corridors. The libraries are located between
these corridors.
Vibration characteristics of the main tower and
sub-towers in Bayon temple have been studied by
micro-tremor measurements since 2003, which is a
Figure 1. Plan and section of the Bayon temple.
preliminary study for constructing analytical models
to evaluate the effects of the wind excitation. Though
the towers are merely a pile of stones, they definitely reinforced concrete structures (Sugiura et al. 2004,
have vibration modes with predominant frequencies Maeda et al. 2005, Maeda et al. 2007).
of 2 Hz to 6 Hz according to their heights, with damp- Equivalent continuum models of FEM for the tow-
ing factors of 2 to 3%, similar to modern ordinary ers were constructed by simulating experimentally

629
obtained base-fixed natural frequencies of horizontal
vibration modes with adjusted elastic constants. The
simulated equivalent Young’s modulus ranged from
1/27 to 1/9 of the laboratory test. This extremely
low modulus may be attributable to spring effects
of some material existed between the interfaces of
sandstone blocks, or to reduced contact areas at the
interfaces.
In 2006, we had a chance to measure micro-
tremor at the libraries. While the main tower and
sub-towers are basically stonewalls, the libraries are
relatively small frame-like structures. Comparing the
equivalent elastic modulus of these different types of
structures, insight on the underlying common phys-
ical phenomenon of sandstone-dry-masonry may be
found.
Our final objectives are the evaluation of bearing
capacity for these dry-masonry structures. Since the
extremely low continuum-equivalent elastic modulus
may reveal incapableness of continuum mechanics in
simulating rupture processes of dry-masonry struc-
tures, we have applied the Discontinuous Deformation
Analysis (DDA) (Shi 1993) to the evaluation of the
safety margin of the library against the wind load as
an illustrative example.

Figure 2. Appearance of the northern library (upper panel)

2 MICRO-TREMOR MEASUREMENTS AND and the southern library (lower panel).
FEM SIMULATION

2.1 Vibration characteristics of the libraries

InAugust 2006, micro-tremor measurements were car-
ried out at the northern library and the southern library
of Bayon temple (Maeda et al. 2007). They are about
4.1 m high with area of 6.6 m times 14.4 m, stand-
ing on a sandstone-covered manmade mound of 4.9 m
high, which is made of sand, gravel, and laterite, a
porous soft rock. Their appearances are similar to each
other as shown in Figure 2. The northern library was
fully restored in 2000 (JSA2000), by being disman-
tled and reassembled mostly with original materials
supplemented by some new ones, while the southern
library is now under similar restoration.
We have arranged accelerometers to capture hor-
izontal translation and rocking motion of the super-
structure as shown in Figure 3. Accelerometers are
composed of over damped oscillators, with frequency
range of 0.1 Hz to 100 Hz and the sensitivity of 1 Gal/V.
Acceleration is low-pass filtered at 30 Hz and sampled
by 0.01 s. with 20 bit AD converter.
Figure 4 shows horizontal transfer functions of roof
level to floor level for the northern library in NS
and EW directions. Clear peaks and smooth transition
of phases are observed. Peak frequencies of the first
modes are summarized in Table 1. In NS direction, the
peaks of the transfer functions for the first mode reveal Figure 3. Sensor arrangements for the library.

630
Transfer Function

25
20 L4WY
15 L4NY
10 L4SY
5 L4EY
0
0 2 4 6 8 10 12 14 16

180
Phase Lag

90 L4WY
L4NY
0
L4SY
- 90 L4EY
- 180
0 2 4 6 8 10 12 14 16

Frequency (Hz)

(a) NS direction
Transfer Function

15

L4WX
10
L4NX
5
L4SX
L4EX
0
0 2 4 6 8 10 12 14 16
Frequency (Hz)
180
Phase Lag

90 L4WX
0 L4NX
L4SX
- 90
L4EX
- 180
0 2 4 6 8 10 12 14 16
Frequency (Hz)

(b) EW direction

Figure 4. Transfer functions of the northern library in NS Figure 5. FEM model for the superstructure of the northern
and EW directions. library.

Table 1. The first mode frequencies of the libraries.

Northern library Southern library

Hz Hz

Soil-coupled NS 5.6 4.7

EW 8.3 9.0
Base-fixed NS 5.8 4.8
EW 9.4 9.2

non-uniform motion at the roof level. For the south-

ern library, similar mode shapes are identified, and the
frequencies of its first modes in horizontal directions
are also summarized in Table 1.

2.2 FEM modeling of libraries

The FEM model for the superstructure of the north-
ern library was constructed as shown in Figure 5. It
has 4,891 elements, 7,968 nodes, and 206 ton of total
mass. Continuum-equivalent Young’s modulus is eval-
uated by simulating base-fixed frequencies of the first
horizontal modes summarized in Table 1. The Young’s Figure 6. First horizontal translation mode for the northern
modulus is evaluated as 850 N/mm2 for both of NS and library in NS and EW directions.
EW directions, which is about 1/15 of 13,000 N/mm2 ,
a value obtained by the laboratory test. has larger amplitude along the north and the south
Figure 6 shows first horizontal translation modes edges compared to the east and the west edges, which
obtained by the finite element model for the north- is in accordance with the variation of peaks in transfer
ern library in NS and EW directions. The NS mode functions shown in Figure 4.

631
 Table 3. Equivalent Young’s modulus of the structures.

Young’s modulus Fraction of the

N/mm2 laboratory test

Main tower 2004 1,200 1/11

Main tower 2006 1,500 1/9
Sub tower 950 1/14
Northern library 850 1/15
Southern library 480 1/27
Laboratory test 13,000 1/1

2004 for equivalent elastic modulus. One of sub-towers

was also modeled in 2004. They all exhibit extremely
low elastic moduli compared to the one in the labo-
ratory test (Sugiura et al. 2004, Maeda et al. 2005).
Continuum equivalent Young’s modulus ranges from
480 N/mm2 to 1,500 N/mm2 ; fraction of the laboratory
value ranges 1/27 to 1/9 as summarized in Table 3.
These extremely low moduli are obtained regard-
Figure 7. First horizontal translation mode for the southern
less of structure sizes and structure types and may be
library in NS and EW directions. attributable to spring effects of some material exist-
ing between the interfaces of sandstone blocks, or to
Table 2. The first mode base-fixed frequencies of the reduced contact areas at the interfaces. Unless we can
libraries by measurements and analyses. find the reason for that, we cannot use the continuum
mechanics for evaluation of bearing capacity of the
Northern library Southern library structures in Bayon temple.
Hz Hz

Measurement NS 5.8 4.8

EW 9.4 9.2
Analysis NS 5.8 4.7
EW 10.3 8.2
The equivalent modulus for the southern library is
evaluated as 480 N/mm2 , which can explain first fre-
quencies in NS and EW directions and is about half
of that of the northern library. The first horizontal
translation modes obtained by FEM for the south-
ern library are shown in Figure 7; the mode in NS
direction exhibits similar in-plane deformation to that
of the northern library and in accordance with the
observation. Comparison of the observed base-fixed
frequencies and analytical ones are summarized in
Table 2.
2.3 Summary of equivalent elastic modulus of the
structures in the temple
We have two FEM models for the main tower, one
was made in 2004 with reference to literatures and
photos, the other was in 2006 based on the three dimen-
sional topography survey (Ikeuchi et al. 2004). Either
model was based on micro-tremor measurements in
3 AN EXAMPLE OF SAFETY EVALUATION
OF THE LIBRARY
3.1 DDA analysis
Most of studies dealing with the bearing capacity of
masonry structures are based on stresses computed
by continuum mechanics. This scheme may be taken
granted for a bonded masonry, since it implies that gen-
eration of a crack in a block or at an interface limits
the bearing capacity. On the other hand, dry-masonry
has its initial cracked state, so that the application of
this scheme is questionable.
We have been studying the applicability of the Dis-
continuous Deformation Analysis (DDA) (Shi 1993)
to dry-masonry structures. DDA is formulated as
incremental dynamic equilibrium for blocks based on
the theory of the least total potential energy with itera-
tion scheme for preventing intrusion and tensile stress
at interfaces of blocks. DDA can guarantee existence
and uniqueness of the solution and can be practical on
usual PC platforms.
We are interested in the response of masonry struc-
tures to wind load exerted horizontally on the struc-
tures. Equilibrium under the gravitational load was
first achieved, and then horizontal load was applied
gradually to let sliding occur between blocks. Figure 8

632
 Figure 9. Modeled portion of the northern library by DDA.

Figure 8. Response of series of blocks to horizontal loads

simulated by DDA.

shows the responses of a series of square blocks lined

in vertical, which is subjected to horizontal load either
at upper portion (upper panel) or at all of the blocks
(lower panel). The different collapse modes can be
simulated by DDA, i.e. sliding at the interface right
below the loaded portion for partial loading and uplift
of whole blocks for whole loading.

3.2 Response of the northern library to

horizontal load by DDA Figure 10. DDA model for the northern library.
The north-south section of the northern library includ-
ing columns and walls is simplified by plane strain
model with consideration on symmetry shown in
Figures 9–10. Friction angle is set to 30 degrees, mass 3.3 Bearing capacity of the library to the wind load
density 2.52 g/cm3 , Poisson’s ratio 0.11; no cohesion The wind load density can be evaluated by (1).
and no dynamic friction are considered. Uniformly
distributed horizontal load along the wall height is
replaced by nodal forces exerted gradually at the rate of
7,937 N/m2 /s after the settlement under gravitational
load. Figure 11 shows development of deformation and
variation of principal stress with increase of horizon-
tal load. The principal axes for the initial gravitational where D: horizontal force [N], S: area [m2 ], C: resis-
load in vertical (a) are firstly inclined by combina- tance factor, ρ: density [kg/m3 ], V : wind velocity
tion of axial force and shear force due to friction at [m/s]. For wind velocity of 40 m/s, (1) gives about
interfaces (b), and back to vertical again due to the 10,000 N/m2 , which is less than a half of the evaluated
occurrence of slide (c). The first sliding occurs at bearing capacity of the first slide at 20,117 N/m2 . Then
20,117 N/m2 of uniform load, which can be viewed the safety margin for the wind load corresponding to
as a bearing capacity of the model. wind velocity of 40 m/s is more than twice.

633

Figure 11. Deformation and principal stress in the northern
library DDA model for horizontal load.
4 DISCUSSION
The continuum-equivalent Young’s modulus for
sandstone-dry-masonry structures in Bayon temple
ranges from 1/27 to 1/9 of the laboratory value
634
regardless of sizes and types of structures, based on
micro-tremor measurements. Since the extremely low
modulus is not yet clearly explained, we cannot rely
on a continuum model to evaluate bearing capacity of
dry-masonry structures. DDA was selected to replace
FEM and tested for fundamental features by simple
problems. Finally DDA was applied to the northern
library with plane strain model. The bearing capacity
for the first slide occurrence is evaluated as about twice
larger than the wind load corresponding to 40 m/s.
ACKNOWLEDGEMENT
This paper is based on the 2006 Bachelor’s thesis
of Takeya Yamamoto and Yusuke Wako and 2006
Master’s thesis of Tomonari Hirai at the Department
of Architecture, Waseda University. We would like to
express our gratitude to the Japan government team for
safeguarding Angkor, lead by Prof. Takeshi Nakagawa
of Waseda University to provide us a chance to study
on Bayon temple. We are indebted to Prof. Naoyuki
Koshiishi of Waseda University for laboratory test data
of sandstone.
REFERENCES
Ikeuchi, K., Hasegawa, K., Nakazawa A., Takamatsu, J.,
Oishi, T. & Masuda, T., Bayon Digital Archival Project,
10th International Conference on Virtual Systems and
Multimedia (VSMM2004), November 2004.
JSA (Japanese Government Team for Safeguarding Angkor),
1995. T. Nakagawa (ed.), Annual Report on the Technical
Survey of Angkor Monument 1995.
JSA, 2000. T. Nakagawa (ed.), Report on the Conservation
and Restoration Work of the Northern Library of Bayon,
Angkor Thom, Kingdom of Cambodia, 2000.
Maeda, T., Sugiura,Y. & Hirai, T. 2005. Vibration characteris-
tics of the Bayon temple main tower, Angkor, Cambodia,
In C. A. Brebbia & A. Torpiano (eds), Structural Stud-
ies, Repairs and Maintenance of Heritage Architecture 9,
Malta 2005: 255–264, Southarnpton: WIT PRESS.
Maeda, T., Yamamoto, T., Wako, Y. & Hirai, T. 2007. Vibra-
tion characteristics and equivalent Young’s modulus of the
Northern Library and the main tower, Bayon, Cambodia,
In C. A. Brebbia (ed.), Structural Studies, Repairs and
Maintenance of Heritage Architecture 10, Prague 2007:
493–502, Southarnpton: WIT PRESS.
Sugiura, Y., Fukumoto, Y. & Maeda, T. 2004. Vibration char-
acteristics of the main tower, the Bayon temple, 21st Inter-
national congress of theoretical and applied mechanics,
2004.
Shi, G-H. 1993. Block system modeling by discontinuous
deformation analysis, In C. A. Brebbia & J. J. Connor
(eds), Topics in Engineering Volume 11, Southampton:
Computational Mechanics Publications.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Safety assessment of temple E7 in Mỹ So,n, Viet Nam

P. Condoleo, A. Taliercio & L. Binda

Department of Structural Engineering, Politecnico di Milano, Milan, Italy

ABSTRACT: The present work illustrates the results of structural analyses carried out on one of the temples
located in the archaeological area of Mỹ So,n in Central Viet Nam. The temple is a relatively well-preserved
building, hosting two square chambers covered by false vaults. A corner of the temple was destroyed by a bomb,
and some passing-through cracks were likely induced by the vegetation. The temple was built using multi-leaf
brick masonry in most of the walls, the outer leaves being more regular in texture than the inner one; also, the
bricks in the outer leaves are stuck by natural resin, which provides an excellent bonding. Finite element models
of the temple were set up, according to an accurate recently performed survey, and to drawings dating back to
the beginning of the 20th century, as far as the geometry of the nearly intact building is concerned. The models
take the layered nature of masonry into account. The masonry leaves were supposed to be either perfectly or
partially connected, allowing for a Mohr-Coulomb’s type failure condition at the interface between adjacent
leaves. Numerical analyses were carried out to evaluate the presumable stress conditions of the still intact temple
and to determine whether or not the self-weight alone might be the cause for the collapse of parts of the temple.
Also, the damages induced by bombs and vegetation were taken into account, and the current stress conditions
were determined. Some conclusions on the safety of the temple in its past and current conditions are drawn; also,
the influence of the models employed to allow for the contact between the different masonry layers is discussed.

1 INTRODUCTION The G group of temples is currently the object of

The archaeological area of Mỹ So,n is located in Central
Vietnam. The site is 15ha wide and was the main holy
place of the Champa kingdom. Between the 7th and
the 12th century, the Cham people built here more than
seventy brick masonry buildings, thirty of which are
still recognizable.
After having been abandoned and neglected, the
area was rediscovered at the beginning of the 20th cen-
tury by the French architect Henri Parmentier from
the École Française d’Extrême Orient (EFEO), who
classified and surveyed most of the buildings. The Par-
mentier expedition in 1898 found the area completely
covered by vegetation. About fifty buildings were clas-
sified as groups, each group named from A to L and
documented through geometrical and photographic
survey (Parmentier, 1904, 1918).
Nevertheless, the heaviest losses were mainly due
to American bombs in 1969. Several buildings were
completely destroyed; others were indirectly damaged
by vibrations and fragments of bomb-shells.
From 1981 to 1986, after the Viet Nam war, a
Vietnamese-Polish team lead by Hoàng Ða.o Kínhand
and Kasimierz Kwiatkowski carried out restoration
works on some groups of buildings damaged during
the Vietnam war at the end of the sixties (Kwiatkowski,
1985, 1990).
a preservation project on a chosen group of temples,
which involves Politecnico of Milan, the Institute for
Conservation of Monuments of Hanoi and UNESCO.
Since 2000 several inspections were done in Mỹ So,n,
to directly evaluate the damage state of the buildings
due to bombs and shells launched during the war and
to intensive and continuous biological attach.
The building E7, belonging to group E, was stud-
ied by the archaeologists of the Lerici Foundation,
Politecnico di Milano. E7 was also surveyed by the
Vietnamese and Italian team, and a geometrical model
was prepared. Also, tests on bricks and masonry fill-
ing were carried out. The present paper illustrates the
results of a number of structural analyses carried out
on this temple, in order to assess its structural safety
in view of a future preservation project.
2 DESCRIPTION OF TEMPLE E7
Group E is located north of groups A and G (Fig. 1).
Among its buildings it is worth mentioning E1, one
of the most ancient temples (kalan) sited in the Mỹ
So,n archaeological area, whose construction can be
dated back to the 7th century. On the contrary, E7 was
probably built later on, in the 10th century (Fig. 2).
This building is a typical example of annex to the main

635

Figure 1. The archaeological area of Mỹ So,n.
Figure 2. Plan of the group E: survey by the Vietnamese
team, 2001.
temple; its location inside the sacred area is responsible
for its designation as “south building”. The function
of E7 was probably to preserve the sacred texts and
objects used during the religious rituals (kośa grha).
When the building was discovered by Henri Par-
mentier, around 1900, its conditions were fairly good.
Nevertheless, Parmentier never carried out any archae-
ological excavations around and inside the building.
At that time, the only damaged part was the top
of the roof, which had lost its original shape. The
most relevant damage occurred during the Viet Nam
war (1969), when several parts of the main northern
entrance and of the roof were lost; moreover, the bomb-
ings severely jeopardized the stability of the structure
(Fig. 3). During an emergency intervention carried
636
Figure 3. Evolution of the plan and northern prospect from
1909 to 2001.
Figure 4. Tree growing into E7 wall.
out by the Polish architect Kazimier Kwiatkowski,
but never completed, the platform was reinforced. The
state of damage increased during the years because of
the lack of a regular maintenance, which is the main
cause of the vegetation growth: several roots pene-
trated the masonry, thus worsening the crack pattern
(Fig. 4).
During the archeological excavation inside and out-
side the building, it became necessary to carry out a
direct geometrical survey on E7. All the surfaces of the
building were thoroughly surveyed and many pictures
were taken, to serve both as document of the state of
work, and as investigation mean.
As most of the buildings of the Mỹ So,n area, E7 was
built following the typical procedures of Cham archi-
tecture. All the monuments are characterized by the
presence of four fundamental elements: the founda-
tions, the base, the central body and the roof (Fig. 5).
The internal floor level is always located at the end
of the base (cymatium). In the case of E7, the only
remainder of the original floor is a sandish filling.
The access was guaranteed by stairs, generally made
Figure 5. View of temple E7.
Figure 6. Three-dimensional model of the structure:
longitudinal cross-section.
of stone. Because of its function inside the complex,
and similarly to others in Mỹ So,n, E7 is composed by
two rooms connected by a door with a threshold, stone
jambs and a plain lintel. The cover of the two rooms
consists of two false pyramidal vaults, which are con-
nected at the top by a single false vault. Externally it
appears as a double-curvature vault. The visual double
curvature effect was obtained by simply smoothing
the external faces of the bricks: therefore, this cur-
vature has no static function. On the east and west
sides are located two small windows, with small stone
columns.
All the information gathered were instrumental in
creating a three-dimensional model of the structure,
which was drawn by means of CAD (Figs 6–7).
The structure of the foundations was investigated
by means of a pit, located on the south-east side of
E7. The pit followed the external profile of the build-
ing wall. The trench brought to light three rows of
foundation bricks, 18 cm thick, put on a preparation
layer of about 15 cm, made of small pebbles and soil
(Fig. 8). It is sreasonable to assume that, like in the
637
Figure 7. Three-dimensional model of the structure:
transversal cross-sections.
Figure 8. Foundation brick rows and preparation layer.
case of other Cham monuments, the foundations of
E7 were built inside continuous cavities, without any
foundation plate.
The masonries in elevation are composed by three
leaves. The two external leaves are characterized by a
thickness which is equal to the length of the bricks,
and are constituted by whole bricks with very thin
(micrometric) joints of organic material, whereas the
dimension of the internal leaf, as well as the type of
internal filling, varies with the maximum dimension
of the masonry. The external leaves consist of hori-
zontal and continuous layers of headers. Only in few
cases stretchers are found, to guarantee the toothing
between the external leaves and the central leaf. The
limited offset observed between the layers of headers,
Figure 9. Section of the leaves of the wall.

reduced to few centimeters, implies a weak toothing Figure 10. Elevation of the north side of temple E7.
between the external leaves and the central leaf, and
does not guarantee an adequate monolithicity to the
masonry, whose behavior tends to be similar to that of
three independents leaves (Condoleo, 2007).
Differently from the external leaf, which is homo-
geneous from the base to the roof, the internal leaf has
different characteristics from the base to the principal
body. The main difference between the two parts is rep-
resented by the dimensions. The thickness of the base
is approximately the same as that of the foundations,
with a value of approximately 110 cm. The principal
body, which rests on the base, has a smaller thickness,
with a value close to 75 cm. The internal leaf of the
base of the annexes is constituted by layers of entire
and half bricks, with irregular arrangement, and a fill-
ing of variable thickness made of clay, chamotte and
quartzitic temper (Fig. 9).
Before surveying the external and internal surfaces
of the walls it became necessary to remove the vege-
tation layers. This cleaning process made the survey
of all surface damages possible, together with a clear
identification of the missing parts. The results of the Figure 11. Crossing cracks pattern of the northern side.
survey process were reported on drawings, in order to
facilitate the damage interpretation and the detection
of its causes. For this reason, it was decided to put a timber provi-
The overall structural stability is mainly endangered sional structure. In order to avoid any modification of
by several cracks, passing through the entire wall thick- the structural behavior, the scaffolding was designed
ness. Moreover, the loss of the lintel and part of the to be passive, which means that its bearing capacity is
vault of the main entrance jeopardized its stability activated only in case of collapse or movements of the
(Figs 10–11). building itself (Fig. 12).

638
 Figure 13. Finite element mesh of the intact temple.

Figure 12. General view of the supporting structure.

3 FINITE ELEMENT MODELS

The geometric (CAD) model of the temple was defined

according to an accurate recent survey and to draw-
ings dating back to the beginning of the 20th century
(Parmentier, 1918), as far as the geometry of the nearly
intact building is concerned. The geometrical model
was later converted into a finite element model, con-
sisting of approximately 24500 4-node constant strain
tetrahedra. The total number of nodal degrees of free-
dom is nearly 23300. The FE model of the intact temple
is shown in Fig. 13. In the definition of the FE mesh,
account was taken of the heterogeneity of masonry
across the wall thickness: each leaf was individually
discretized, and the different leaves were given dif-
ferent mechanical properties. The lower base of the
model is supposed to be fully constrained to the soil
underneath.
To investigate the effect of the damages induced by
the bombs and the vegetation, part of the FE model was
suppressed and the main cracks surveyed in the exist-
ing temple (such as those shown in Fig. 11) were also
included. A mesh much more refined than that used for
the intact temple had to be defined, to capture the irreg-
ularities of the damaged model. The FE model of the
temple in its current conditions is shown in Figure 14.
Figure 14. Finite element mesh of the damaged temple.
In the analyses, the materials forming the tem-
ple were assumed to be linearly elastic and isotropic.
According to mechanical tests carried out on sam-
ples taken from similar temples (Binda et al., 2006),
the outer masonry layers, made of bricks bonded by
resin, were given an elastic modulus E of 1530 MPa,
whereas for the inner layer E = 1300 MPa. All the
layers were given a Poisson ratio ν = 0.11 and a density
ρ = 2200 kg/m3 .
Different assumptions were made regarding the
connection between the leaves of the brick walls. The
less conservative assumption is considering the leaves
as perfectly bonded (case A). The most conservative is

639
neglecting any connection, thus considering the three
leaves as fully independent (case C). As a matter of
fact, the leaves are likely to be partially connected,
according to the construction technique (see Sec. 2),
so that stresses can be transmitted from a leaf to the
neighbouring ones to a certain extent. To model this
situation, the interface between each pair of adjacent
leaves was supposed to comply with a Coulomb-type
interface (case B). The friction angle was given a value
of 30◦ , whereas the tensile strength of the interface
was neglected: the latter is likely to be a conservative
assumption.
Only the effects of the self-weight of the materi-
als were investigated in the analyses shown in Sec. 4,
failing any evidence of significant ground settlements.

4 NUMERICAL ANALYSES

The main results of the numerical analyses carried out Figure 15. Case A: contour plots of the minimum principal
on the two models of the temple are presented in the stress in the intact temple.
form of contour lines of the extreme principal stresses
plotted on the deformed FE mesh. The stress values
are expressed in Pa units. To appreciate the effect of
the different assumptions made about the degree of
connection of the wall leaves, the same magnification
factor for the displacements (=1650) was employed
in all the figures shown in the continuation of this
Section. Only the results obtained in the cases A and B
are shown: the results pertinent to case C were found
to be unrealistic, as illustrated in the continuation.

4.1 Analyses of the intact temple

The results of the analyses carried out on the model of
the temple at its presumed intact, original state are first
described. Figures 15 and 16 show the contour plots
of the minimum (compressive) and maximum (tensile)
principal stresses, respectively, obtained assuming the
leaves to be perfectly bonded (case A). The highest
compressions are attained at the base of the vertical
walls of narrowest section, and are of the order of
−0.25 MPa. The highest tensions are located in the lin-
tel of the north doorway, where they attain +0.05 MPa.
The computed stresses are quite low and compatible
with the strength of the materials forming the building.
Indeed, according to the available experimental results,
the tensile strength of the outer layers measured on
couplets of bricks bonded by resin is of the order of
0.25 MPa, whereas the compressive strength measured
perpendicularly to the joints is 11.5–12 MPa.
Figures 17 and 18 are similar in meaning to
Figures 15 and 16, but refer to the case where a
frictional Coulomb-type bonding is assumed at the
interface between the wall leaves (case B). Appar-
ently, the model deformability is much higher than
in the previous case, as a wide portion of the inter-
face between the leaves turns out to have debonded.
Figure 16. Case A: contour plots of the maximum principal
stress in the intact temple.
Contrary to case A, where stresses do not significantly
differ across the wall section from one layer to the oth-
ers, in case B the stress is usually higher in the inner
layers than in the outer one as a consequence of inter-
facial debonding. The highest compressive stresses are
of the order of −0.42 MPa in the central layer; the high-
est tensile stresses attain +0.17 MPa above the main
entrance. Despite the increase due to debonding, even
in case B the stresses estimated in the intact temple are
not found to be incompatible with the material strength
properties.
It is worth mentioning that in case C (fully debonded
leaves) tensile stresses exceeding +0.38 MPa were
computed at the base of the upper part of the temple.

640
Figure 17. Case B: contour plots of the minimum principal
stress in the intact temple.
Figure 18. Case B: contour plots of the maximum principal
stress in the intact temple.
These stresses are incompatible with the material ten-
sile strength and would imply cracking of wide parts of
the building.As the temple was rather well preserved at
the beginning of XX century, after more than thousand
years since it was built, the assumption of complete
interfacial debonding seems to be unrealistic.
4.2 Analyses of the damaged temple
The results of the analyses of the model of the temple at
its current, damaged state will be now illustrated. The
contour plots of the extreme principal stresses when
the wall leaves are supposed to be perfectly bonded
(case A) are shown in Figures 19 and 20. Comparing
these plots with the corresponding ones referred to the
641
Figure 19. Case A: contour plots of the minimum principal
stress in the damaged temple.
Figure 20. Case A: contour plots of the maximum principal
stress in the damaged temple.
intact temple (Figs 15 and 16), the increase in stress
and deformability associated with the damage suffered
by the building is apparent. In particular, tension peaks
exceeding 0.2 MPa are found, although they are quite
localized in limited regions of the damaged temple
(Fig. 20).
Much more alarming are the current stress condi-
tions of the temple if computed allowing for interfacial
debonding of the wall leaves (Figs 21 and 22). The
leaves turn out to have widely debonded, and the global
deformation mode of the temple corresponds to a sort
of punching of the deck on the lower, weakened part of
the temple. An increase in the maximum compressive
stress can be noticed (Fig. 21), with the highest values
attained at the base of the innermost brick leaf; these

Figure 21. Case B: contour plots of the minimum principal
stress in the damaged temple.
Figure 22. Case B: contour plots of the maximum principal
stress in the damaged temple.
values, however, are of the order of −0.5 MPa, that
is, compatible with the material compressive strength.
Wide regions where the maximum principal tensile
stress exceeds 0.2 MPa are found (Fig. 22). They are
mostly located at the base of the deck of the tem-
ple, where the lower walls collapsed and the deck
behaviour resembles that of a cantilever. As these high
stresses could not be borne by the material, a scaffold-
ing was duly placed in the opening caused by bombing
to prevent the temple deck from failing.
5 CONCLUDING REMARKS
According to the numerical analyses performed, the
stress in the temple E7 in Mỹ So,n archaeological site
642
was found to be compatible with the material strength
properties when the temple was still intact, unless
the excessively severe assumption of fully indepen-
dent wall leaves is made. Thus, the possibility that the
collapse of the temple might be partially due to the
self-weight alone seems to be excluded.
Nevertheless, its current state is the result of the
damage caused by the bombings during the war and the
lack of repair and maintenance during the last decades.
Therefore, the analysis was repeated to address the
damaged situation.
As very high tensile stresses were computed in the
damaged temple accounting for a limited interface
frictional bonding (case B), the building in its current
condition is not found to be safe. Provisional scaffold-
ings have already been put up, to avoid the collapse
of the upper part of the building (see Fig. 12). The
safety assessment of the damaged building is worth
being further analyzed. To this end, it might be appro-
priate employing more reliable constitutive laws, both
for the brick leaves (accounting for their brittleness)
and the interface between the leaves (accounting for its
non negligible tensile strength, coming from the partial
degree of connection of the leaves). Also, an attempt
at evaluating the local stress concentrations due to the
vegetation that took root in the temple is planned, by
introduction of a suitable pressure into some of the
main cracks.
ACKNOWLEDGMENTS
The authors are indebted to Mr Marco Crippa, who set
up the finite element models of the temple and carried
out all the numerical analyses.
REFERENCES
Binda L., Condoleo P., Cucarzi M., Lê Thành V., Pichard P.,
Hoàng Ða.o K. 2007. Preservation of temples in Mỹ
So,n (Vietnam). In P.B. Lourenço, P. Roca, C. Mod-
ena, S. Agrawal (Eds.), Structural Analysis of Historical
Constructions. New Delhi: McMillan India Ltd.
Binda L., Tedeschi C., Condoleo P. 2006. Characterisation
of materials sampled from some Mỹ So,n temples. Proc.
ICCE 7, Tehran, 8-10/5/2006, CD ROM.
Condoleo P. 2007. I templi di Mỹ So,n: indagini e progetto di
conservazione del Gruppo G, PhD thesis in Preservation
of Culture Heritage of Architecture Politecnico of Milan.
Kwiatkowski K. 1985. Recherches sur les Monuments du
Champa. Rapport de la Mission Polono-Vietnamienne
1981–1982. Warsaw.
Kwiatkowski K. 1990. Recherches sur les Monuments du
Champa. Rapport de la Mission Polono-Vietnamienne
1983–1986. Warsaw.
Parmentier M.H. 1904. Les Monuments du Cirque de Mỹ So,n.
Hanoi: F.-H. Schneider.
Parmentier H. 1918. Inventaire descriptif des monuments
čams de l’Annam, Voll. 2, E. Leroux, Paris.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Construction process, damage and structural analysis. Two case studies

R. González & F. Caballé

Veclus

J. Domenge, M. Vendrell & P. Giráldez

Universitat de Barcelona

P. Roca & J.L. González

Universitat Politècnica de Catalunya

ABSTRACT: The paper refers to the studies recently carried out on the church of Santa Maria del Mar in
Barcelona and Mallorca cathedral, built during the 14th and 15th centuries. Historical research has provided
significant insight on the construction processes, which were remarkably different in the two buildings. This
information has been considered to investigate the possible link between construction and existing deformation
and damage. As shown by these examples, insight on the construction process (and later man-caused alterations)
may be important for an accurate structural analysis.

1 INTRODUCTION developed, in the case of Mallorca cathedral, as an

The buildings analyzed, namely Mallorca cathedral
in the island of Mallorca and the church of Santa
Maria del Mar in Barcelona, show common architec-
tural features but also significant dissimilarities which,
in turn, have lead to different damage and deforma-
tion conditions. As will be later described, one of the
main differences is found in the construction processes
and their consequences for the latter condition of the
structure.
The studies were undertaken in close cooperation
between historians, structural engineers, mineralo-
gists, architects, geophysicists and other specialists.
Detailed research on available historic documents has
permitted the identification of essential issues related
to the construction process and later events (archi-
tectural alterations, earthquakes, fires) having left
their imprint on the structure. Geophysics, carried
out by means of seismic tomography, pulse radar and
dynamic testing, has been very useful in character-
izing essential structural and morphological features.
Structural analysis, carried out by means of classical
approaches (limit analysis) and advanced non-linear
numerical models, has permitted the characterization
of the structural performance under gravity, soil set-
tlements and earthquake (Clemente, 2007, Martínez,
2008).
Based on the information provided by historical and
geophysical research, a sequential analysis has been
attempt to simulate some aspects of the construction
process.
An essential issue of the diagnosis lays in the dis-
tinction of damage related to active phenomena, still
contributing to further deterioration, from that which
has resulted from past or already extinguished actions.
In some cases, stabilized damage or alterations caused
by past actions may be preserved and respected as a
sign of identity linked to the historical character of the
building. In other cases, repairing this type of dam-
age may be necessary to avoid problems related to
durability or functionality, or to improve the capacity
of the building under extraordinary actions (as earth-
quake). If repaired, historical or traditional techniques
should be preferred to alternative possibilities in order
to avoid unnecessary loss of authenticity of materials
and structure.
For obvious reasons, the ISCARSAH Recommen-
dations (2005), state that the intervention must always
address the real causes rather than “the symptoms”.
When addressing past or extinguished actions, the
need for intervention should be carefully appraised
and repair and strengthening should be limited to truly
indispensable operations.
Possible historical actions which may have con-
tributed to damage, but (in principle) are not expected
to happen again, include anthropogenic ones (wars,
fires and other types of destruction, lack of mainte-
nance and inadequate restorations) and natural ones (as

643
stabilized soil settlements). Earthquake does not lay
in this group for obvious reasons. Yet another “action”
belonging to this group, which by no means should
be neglected, is the construction process itself and its
related hazards and difficulties.

2 DESCRIPTION OF THE BUILDINGS

Santa Maria del Mar is a rare case of a Gothic church

entirely built during a short period of time spanning
53 years. Moreover, the building has not experienced
any significant architectural alteration after its con-
struction, resulting in the uniformity and architectural
purity for which the building is today acknowledged.
The overall arrangement of the structure is very sim-
ilar to that of Barcelona Cathedral, built during 13th
and 14th c., and consists of a three nave structure built
on a basilical plan.
In Barcelona Cathedral, the wish for a unitary
and diaphanous inner space, reminiscent of that of
Roman basilicas, laid to an innovative design. The
central vaults were built with longer span in both the
transverse and longitudinal directions (12 and 8 m),
compared to other contemporary buildings, and were
supported in surprisingly slender piers; the lateral
vaults spanned as much as one half of the central ones
and were built at a height close to that of the central
ones; the external walls were built along the exter-
nal perimeter of the buttresses and these were used
as partition walls between lateral chapels. In the case
Figure 1. View of the interior of Santa Maria del Mar
of Barcelona Cathedral, this combination contributed
(above) and Mallorca Cathedral (below).
successfully to create the impression of a unique and
wide, even magnificent interior space in spite of the
limited dimensions of the building. As mentioned, is half of that of the central ones, while the buttresses
the lateral vaults are almost as high (but not so high) are used to separate the lateral chapels. The building is
as the central ones; in fact, they are adequately posi- completed with the choir and the façade, whose struc-
tioned to receive the lateral thrust of the central vaults ture conforms to the uniform arrangement given to the
and carry it to the buttresses, so that structural flying perimeter of the naves. In particular, the façade is built
arches are not needed. with two large buttresses which are, in fact, needed to
Santa Maria del Mar is the only other Gothic con- counteract the longitudinal thrust produced by the first
struction having been built according to this structural square vault.
arrangement. Other Gothic cathedrals have low aisles As opposite to Santa Maria del Mar, Mallorca
(as in French Gothic) or have them as high as the Cathedral was built over a large period, spanning for
central nave (as in German late Gothic churches). 300 years (from 1306 to 1600), and was later sub-
Furthermore, the builders succeed in improving the jected to significant repairs and reconstructions. To
possibilities of the concept in terms of structural econ- some extent, the structural design of Mallorca Cathe-
omy, unity, diaphanousness and aesthetics, the latter dral seems inspired by Santa Maria del Mar and shares
soundly emanating from the geometrical proportions. with it major features, such as the search for spa-
This was attained by designing the central vaults as ciousness, the high lateral naves (although not so
almost perfectly square in plan, by giving them some high as to take the role of flying arches), the lateral
extra height (up to 32 m) and by supporting them on chapels between buttresses and the extremely slen-
austere but extremely slender octagonal piers (Fig. 1, der, solid octagonal piers. However, and because of
above). The resulting construction includes four large the large dimensions intended to the nave, the builders
vaults, spanning 13.5 m in both directions, supported also resourced to structural devices commonly used in
on 8 piers with circumscribed diameter of 1.6 m. As large Gothic buildings, such as the double battery of
in Barcelona Cathedral, the span of the lateral vaults true flying arches spanning over the aisles. In a way,

644
Figure 2. Arrangement of stone blocs in octagonal piers.
Figure 3. Diaphragmatic arches and dead weight over vaults
in Mallorca Cathedral.
the builders managed to synthesize Northern and
Southern Gothic architecture to produce a uniquely
vast and diaphanous inner space.
The central nave spans 19.9 m and reaches 43.9 m
at the vaults’ keystone. The octagonal piers have a cir-
cumscribed diameter of 1.6 or 1.7 m and a height of
22.7 m to the springing of the lateral vaults.
Other shared features are the diaphragmatic trans-
verse arches in the central naves and the dead weight
laid over the transverse arches and central vault key-
stones. In Mallorca Cathedral, the latter become par-
ticularly conspicuous (Fig. 3). Both constructions were
originally built with no pitched roof, rain water being
channeled to the gargoyles by a tile pavement shaped
over the vaults.
The filling over the vaults, however, is different in
the two buildings. In Santa Maria del Mar, the central
vault is partly filled with medieval concrete and then
with a shallow layer of ceramic pottery (Fig. 4), while
the central vaults are filled with medieval concrete up
to the tile pavement. In Mallorca, a lightweight pot-
tery fill was probably removed from the central vaults
during 18th c., while the lateral vaults are still filled
with pottery. It is remarkable that these differences are
consistent with the distinct structural role of the lateral
vaults in each case. In Santa Maria del Mar, the extra
weight provided by the full concrete filling is needed
645
Figure 4. Pottery filling over the central vaults of Santa
Maria del Mar.
to adequately counteract the horizontal thrust of the
central nave.
3 HISTORICAL INFORMATION
Construction of Santa Maria del Mar church began
on March 25, 1329. It involved, in the first place,
the formation of the full perimeter, including the but-
tresses, chapels, lateral walls, choir and façade (Fig. 5),
which were fully completed by 1350. There is evi-
dence suggesting that this phase was followed by the
construction of the arches, both the longitudinal (or
clerestory) and transverse ones, and both at the lat-
eral and central naves (Vendrell et al.). At this stage,
a normal use of the building would have been pos-
sible thanks to a provisory wooden roof supported on
the transverse arches. The construction was completed
by building the vaulted roof on the already existing
arches, starting, in each bay, with the lateral vault
membranes and then the central one. The last stone
was placed on November 3, 1383. Remarkably, this
procedure does not agree with the construction plan
(supposedly) developed in most Gothic Cathedrals,
which involved a gradual extension in the longitudinal
direction starting from the choir and finishing with the
main façade. In fact, Barcelona and Mallorca Cathe-
drals are known to have been built according to this
latter plan.
In 1379, when the last bay was almost com-
pleted, a fire destroyed the scaffoldings and caused
some damage to the stone. Another fire, provoked in
1936, caused significant damage to the piers, arches
and vault keystones. Earthquakes are known to have
occurred in several occasions, as in 1373, causing
the collapse of the upper body of one of the façade
clock-towers, and 1428, when the collapse of the rose
window killed a number of people (Fontseré, 1971).
The building has also experienced damage during
bombardments against the city in1714, (during the
Succession War) and the Spanish Civil War (1936–
1939), among other episodes.
The architectural unity and the consistency of the
construction process suggest that a fully developed
project should have been well established before 1329,
when the construction began. At this moment, not even
the choir of Barcelona Cathedral, frequently regarded
as the immediate precedent, was finished.
Construction of Mallorca Cathedral began by year
1300 starting with the presbytery (which comprises the
so-called Trinity and Royal Chapels). According to the
most widely accepted interpretation, by 1330 it was
decided to build the remaining construction according
to a three-nave plan and yet, by the mid of 14th c., it
Figure 5. Construction of Santa Maria del Mar. 1: founda-
tion; 2: perimeter; 3: arches and vaults.
646
was decided to increase the height of the vaults. The
construction of the main nave developed during the rest
of 14th and 15th c (with a major interruption from 1460
to 1570). The main façade, of noticeable Renaissance
style, was built from 1594 to 1601, when the cathedral
was consecrated.
Research on the historical books has provided sig-
nificant hints on the construction process. As shown
in Fig. 6, the construction of the nave progressed,
bay after bay, from the presbytery towards the façade
(the last part to be built). Construction of the chapels
was ahead because of the funding provided by noble
families or corporations willing them as pantheons or
gremial chapels (Domenge, 1997).
Figure 6. Construction stages of Mallorca Cathedral.
 It has been possible, at least for one of the bays
(the 4th one), to identify the process leading to its com-
plete construction (Fig. 7). Once again, it started with
the lateral chapels, followed by the piers, then one lat-
eral vault, then the other and finally the central one.
In the case of this bay, the construction of the vaults
lasted 7 years. It should be noted that during a period
of about 5 years), the lateral vaults were already push-
ing against the pier while the lateral vault was not yet
there to counteract their thrust.
The building has experienced significant prob-
lems and repairs. The 4th vault (previously discussed)
Figure 7. Mallorca Cathedral’s 4th bay construction
sequence.
647
partially collapsed 30 years after its construction.
A significant number of vaults were repaired or recon-
structed during the 17th, 18th and 19th centuries. Due
to the concerning out-of-plumb (about 1.3 m), the orig-
inal façade was taken down replaced by a new one
during the second half of 19th c. Demolition was
decided in March 1851 and hence was not connected
to the earthquake occurring in May the same year.
4 EXISTING DAMAGE
Four different types of alterations, visible in both
buildings, are here highlighted and discussed:
1- Cracking in piers. Cracks exist in a few piers
of Mallorca Cathedral and most piers of Santa
Maria del Mar (Fig. 8). Vertical or oblique cracks
have developed across the stone extending, in
some cases, to several rows. They tend to concen-
trate close to the corners of the octagonal section
(the less confined parts) and, in some cases, shape
full wedges partially or totally detached from the
core of the pier. In the case of Santa Maria del Mar,
the cracks are related to additional forms of dam-
age (black patina and crusting, superficial loss of
stone and mortar at joints) which can be clearly
associated with the severe fires experienced. The
reason for similar cracks in a few piers of Mallorca,
randomly distributed, is less clear.
2-Cracking in arches and vaults. In the lateral naves
of Santa Maria del Mar, a longitudinal crack has
developed following the keystones of the transverse
arches (Fig. 8, right). The longitudinal cracks in the
aisles are due to a differential settlement between
the pier and the buttress. The arches do actually
show the deformation and damage which should
be expected from this type of settlement (Fig. 9,
left). In a few cases, initial deformation appeared
Figure 8. Cracking in piers (Mallorca Cathedral, left, and
Santa Maria del Mar, repaired, right).
Figure 9. Aisle arches of Santa Maria del Mar. Deformation
and cracking (left). Arch deformed due to accidental removal
of centering causing expulsion of unhardened mortar (right).

due to an early or accidental removal of centering

(Fig. 9, right) related, perhaps, to the fire of 1379.
3-Cracking in walls and façade. Cracking, mostly
developed along the mortar joints, can be also rec-
ognized in the exterior or clerestory walls. It can be
linked to the out-of-plumbing experienced by the
façade in the case of Mallorca Cathedral and the
façade clock-towers in Santa Maria del Mar.
4-Deformation. The deformation of the overall
structure is perceptible in Mallorca Cathedral. The
piers show significant lateral deformation reaching,
in some cases, up to 30 cm, i.e., 1/100 of height at
the springing of the lateral vaults. Remarkably, both
the magnitude and the shape of the deformation
vary very significantly (almost randomly) among
the different bays, or even between the two halves
of a single bay (Fig. 10). Conversely, the piers of
Santa Maria del Mar show a more regular pattern
both in direction and value, with maximum values
ranging between 2 and 6 cm (8 cm in a single case),
about 1/300 of the height at the springing of lateral
vaults.
These deformations have been measured with respect
to the ideal un-deformed geometry of the structure.
For that purpose, the “ideal” profiles are reconstructed
based on the information on original geometric param-
eters (radii, positions) which can be derived (or
guessed) from the present geometry.All deviations due
to construction defects, positioning errors, deforma-
tion of centering and plastic mortar settlement (among
other possible effects) are ignored. Because of it,
real deformations cannot actually be determined and
the above measurements can only be understood as
a coarse estimation. The possibility of the deforma-
tions mentioned for Mallorca Cathedral being mostly
a consequence of errors and hazards experienced dur-
ing the construction cannot be disregarded. The same
applies to possible soil settlements estimated as a
Figure 10. Deformation of the bays of Mallorca Cathedral.
difference between architecturally related vertical ref-
erences (such as opposite imposts, capitals or arch
springings), whose unevenness might be due, at least
in part, to construction inaccuracies.
5 CONSTRUCTION PROCESS AND DAMAGE
From a theoretical point of view, structures based on
the balance of arch thrusts, as Gothic cathedrals, attain
full stability only at their final and complete con-
figuration. Moreover, adequate equilibrium requires
(again, theoretically) the simultaneous activation of
all the arches and vaults by first building the entire
system and then removing all the centering almost
at once. This is not obviously the case of the build-
ings discussed in this paper (as it is not either the
case of most similar constructions). Conversely, real
construction processes involved intermediate stages
were equilibrium was reached only thanks to auxil-
iary devices or, in a more hazardous way, by relying in
the capacity of the incomplete structure. The order in
which the structural members were built was essential
to make the entire construction viable or to limit the
construction difficulties.
The construction process has been, in no few cases,
a hazardous phase contributing by itself to initial dam-
age and deformation. Repairing this damage may have
been possible during or after the construction. Cor-
recting the deformation is difficult or impossible and,
in most cases, it has stayed as a testimony of this
initial phase.

648
 Nevertheless, Santa Maria del Mar was built fol-
lowing a procedure which would ascertain, at any
moment, an easy balance of forces. First, the con-
struction of the entire perimeter structure (including
buttresses, walls, chapels along the choir, sides and
façade) would provide a stiff system able to laterally
buttress the rest. Then, the construction of all longitu-
dinal and transverse arches (of both central and lateral
vaults) would provide the necessary lateral stability
to the slender piers. The transverse and longitudinal
(or clerestory) arches would also grant full stability
during the formation of the entire vaulted roof. In the
longitudinal direction, the stability was ultimately pro-
vided by the choir and façade buttresses, already built.
This is probably the reason why, yet at present, the
piers are showing such a limited drift in spite of their
slenderness.
The process, however, may have generated an incon-
venient side effect. The construction of the vaults
took place about 30 years after the perimeter struc-
ture had been built. As should be expected, the lateral
structure had by then already experienced most part
of its possible soil settlement. Due to the settlement
of the piers, the new members built across the aisle
(arches and vault membranes) experienced a vertical
deformation (or differential settlement). The lateral
longitudinal crack along the keystones of the trans-
verse arches may have been caused by this effect. In
fact, settlements may have been important during the
construction because the foundation soil, composed
of rubble and loose sand, is certainly deformable.
However, a later significant increment of differen-
tial settlements between lateral structure and piers
seems unlikely because the rubble existing over the
vaults (a sort of medieval lime concrete supposedly
placed in a later stage but still during the construc-
tion) does not show any crack in correspondence to
the one seen at the intrados of the vaults. More-
over, the arches appear more damaged than the vault
membrane, which does not either accompany their
significant deformation (figure 9, left); the only expla-
nation is the later construction of the vaults (after
the arches had been built) and the possible devel-
opment of most the settlement while the material of
the vault membranes was gradually placed on the
arches, but yet with the possibility of correction or
almost free deformation while the mortar had not yet
hardened
As mentioned, the construction of Mallorca Cathe-
dral followed a different path involving the subsequent
construction of the bays. The lateral vaults were built
(as in Santa Maria del Mar) before the construction of
the central ones. In this case, historical research has
not provided, so far, any hint on the way the structures
were stabilized while the central vault was not yet built
(third stage of Fig. 7). Several possibilities, however,
may be considered.
649
First, the lateral vaults could have been stabilized
by means of previously built transverse arches, as in
Santa Maria del Mar. If so, the almost 20 m span arches
would have needed some stabilizing extra weight to
resist the thrust of the lateral vaults without experi-
encing inward deformation or collapse. Extra weight
actually exists (Fig. 3), although its original purpose
might be different. Second, they could have been sta-
bilized by means of auxiliary devices such as steel or
timber ties across the lateral arches or struts across
the central one. Third, they could have been built
without any stabilizing element, the partial structure
being (precariously) stable by itself thanks mainly to
available tensile strength.
Whatever the method, there was some hazard and
chance for damage and deformation. This is consis-
tent with the fact that the lateral deformations at the
piers, as already mentioned, are very variable (almost
“random”) although large in average, and suggests
that the outcome was very sensitive to the skills and
methods used by the different builders. Moreover, dif-
ferent approaches (as the three ones mentioned) may
have been used during the certainly long and irregular
construction process.
The longitudinal stability at intermediate stages is
even more challenging as the piers had to face the
unbalanced longitudinal thrust of both the lateral and
central vaults. According to the historical information
available, a previous construction of all the clerestory
arches (as a way to stabilize the bays at intermediate
stages) should be clearly disregarded. The use of pos-
sible temporary devices (temporary ties or buttressing
walls) appears as a likely possibility.
6 STRUCTURAL ANALYSIS
Structural analysis has been carried out using limit
analysis and non-linear FEM calculations. Previous
results concerning dead loading and seismic per-
formance of Mallorca Cathedral have been already
presented (Clemente et al., 2006, Martínez et al, 2006,
Martínez, 2008). The present discussion focuses on
the performance of the structures subjected to dead
loading and the possible influence of the construc-
tion process. The non-linear FEM analyses have been
carried out using a continuous damage constitutive
model (Clemente et al. 2006). The material proper-
ties have been estimated based on laboratory tests
on cores taken from the buildings or the original
quarries. The overall stiffness (value and distribution)
was assessed by comparing experimental and numer-
ical natural frequencies, the obtained from ambient
vibration measurements (Martínez et al. 2006).
In the case of Santa Maria del Mar, the hole drilling
test was used to measure the average compression
stress at the base of two piers. The obtained value,

Figure 11. Comparison of damage parameter for an instan-
taneous analysis (above, deformation × 300) and a sequential
analysis involving two stages (below, deformation × 50).
of 3,0 MPa, agreed very well with the corresponding
numerical prediction (2,9 MPa).
In order to take into consideration the construction
process, a sequential analysis can be carried out in
which the changes experienced by the construction are
subsequently simulated and adequately superposed. In
the case of Mallorca Cathedral, a tentative simulation
of the construction process has been undertaken using
model of the typical nave bay (Clemente, 2007, and
Fig. 11). The third possibility mentioned in section 5
(the partial structure being stable by itself) has been
analyzed in detail by means of a sequential analysis
involving the following steps: (1) Construction of the
lateral nave (buttresses and lateral vaults) and (2) con-
struction of the central vault. According to the FEM
analysis (in which a small tensile strength is assumed),
the stability of the partial structure generated in step
(1) is possible, but at the cost of significant deforma-
tion and certain damage. Compared to instantaneous
analysis, the final lateral deformation attained by the
piers increases by a factor of 2.5. In turn, additional
damage appears in the upper part of the piers, lateral
vaults and upper flying arches. This damage might be
connected to some cracks and deformation still visible
today (as in the upper flying-arches), although most of
it may have been probably repaired during the con-
struction itself. The constitutive model describes this
damage as an scalar whose values range between “0”
for intact and “1” for totally damaged material.
650
7 CONCLUSIONS
As shown by the two cases investigated, the con-
struction process may have significant influence on
the initial condition of the structure. Some of the
alterations (large deformation, damage) which can be
observed in historical buildings might be attributed,
at least in some cases, to the hazards and difficulties
linked to intermediate construction phases.
As in the two examples presented, historical
documents may provide, in some cases, meaningful
information on the construction process. A detailed
inspection may also provide important clues on the
way the structure was constructed and the possible
hazards encountered by the builders. In turn, structural
analysis may contribute with a better understanding of
the structural significance and risks involved by the
intermediate stages associated to incomplete or provi-
sory configurations of the structure. For this purpose,
a sequential analysis (in which the different stages
are subsequently simulated) is preferable to a more
conventional instantaneous analysis.
Later anthropogenic actions (accidental or inten-
tional fires, architectural alterations, repairs) should
also deserve attention and should be always included
as part of the historical investigation. In fact, a signifi-
cant part of the alterations which have been recognized
in the buildings discussed in this paper are linked to the
construction process or to this second type of actions
(as the fires experienced by the structure of Santa
Maria del mar).
It turns out that some insight on the construc-
tion process may be needed for an accurate structural
analysis, while structural analysis itself constitutes a
powerful tool to better understand the construction
process.
ACKNOWLEDGEMENTS
The studies of Santa Maria del Mar and Mallorca
Cathedral have been respectively funded by the
Generalitat de Catalunya and the Spanish Ministry
of Culture. The assistance provided by the Parish of
Santa Maria del Mar and the Chapter of Mallorca
Cathedral are also gratefully acknowledged.
REFERENCES
Caselles, O., Martínez, G., Clapés, J., Roca, P., Canas , J. A.
2006. Non-destructive geophysical surveys for dynamic
characterization of Mallorca Cathedral. First European
Conference on Earthquake Engineering and Seismology.
Geneva: ECESS.
Clemente, R., Roca, P., Cervera, M. 2006. Damage
model with crack localization. Application to histor-
ical buildings. 5th Int. Conf. of Structural Analysis
 of Historical Constructions. New Delhi. New Delhi:
MacMillan India Ltd.
Clemente, R. 2007. Análisis structural de edificios históri-
cos mediante modelos localizados de fissuración.
Ph. D. dissertation. Barcelona: Universitat Politècnica de
Catalunya.
Domenge, J. 1997. L’obra de la Seu. El procés de construc-
ció de la catedral de Mallorca en el tres-cents. Palma de
Mallorca.
Fontseré, E. 1971. Recopilació de dades sísmiques de les
terres catalanes entre 1100 i 1906. Barcelona: Generalitat
de Catalunya.
ICOMOS / ISCARSAH (2005). Recommendations for the
analysis, conservation and structural restoration of archi-
tectural heritage.
651
Martínez, G., Roca., P., Caselles, O., Clapés, J., Canas, J. A.
2006. Characterization of the dynamic response of the
structure of Mallorca cathedral. 5th Int. Conf. of Struc-
tural Analysis of Historical Constructions. New Delhi:
MacMillan India Ltd.
Martínez, G., 2008. Vulnerabilidad sísmica de edificios
históricos de obra de fábrica de mediana y gran luz. Apli-
cación a la cathedral de Mallorca. Ph. D. dissertation.
Universitat Politècnica de Catalunya: Barcelona.
Vendrell, M., Giráldez, P., González R., Cavallé, F., Roca, P.
2008. Santa Maria del Mar. Estudi històric-constructiu,
materials de construcció i estabilitat structural. Barcelona:
UB-Patrimoni, Veclus, UPC.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Analysis and diagnosis of timber structures in Porto historical centre

T. Ilharco, J.M. Guedes & A. Arêde

Faculty of Engineering, University of Porto, Porto, Portugal

E. Paupério
Construction Institute, Faculty of Engineering, University of Porto, Porto, Portugal

A.G. Costa
Autonomous Section of Civil Engineering, University of Aveiro, Aveiro, Portugal

ABSTRACT: The historical centre of Porto was considered in 1996 UNESCO World Heritage due to the
outstanding universal value of the site, with urban fabric and historic buildings that bear remarkable testimony
to the development over the past thousand years of an important European city”. In order to contribute to the
maintenance of this remarkable set of buildings, with historical, architectural, constructive and heritage value,
in opposition to the continuous built heritage degradation process and the population dislocation from the centre
to the surroundings, NCREP –FEUP have been participating in many rehabilitation projects, analyzing the
conservation state of building structural elements (stone masonry walls and timber floors and roofs). The focus
of the present paper is given to the evaluation of the physical state of timber structural elements, namely to
the various instruments used on the Analysis and Diagnosis phases, in particular to the non destructive (NDT),
medium-destructive (MDT) and, in a certain context, destructive testing techniques.

1 INTRODUCTION

The inscription of the cistorical centre of Porto as

UNESCO World Heritage was based on the cultural
criterion “(iv)”, considering that the site is of “out-
standing universal value as the urban fabric and its
many historic buildings bear remarkable testimony
to the development over the past thousand years
of an important European city”, Figure 1. Hence, Figure 1. View of Porto historical centre (Google Earth).
the choice of UNESCO was based on the knowl-
edge that a remarkable set of buildings with indi-
vidual and unrepeatable characteristics transmits out-
standing testimonies about the city and its ancestor
history.
Nonetheless, all the old part of the city presents
a serious problem of degradation of its built heritage,
with a large number of buildings demanding fast inter-
vention and rehabilitation measures, together with the
social problems and the dislocation of the population
from the centre to the surroundings, Figure 2.
Therefore, it is fundamental to endeavour efforts
to preserve this remarkable set of buildings (adapt-
Figure 2. Porto Hhistorical city centre.
ing them to today comfort demands), with harmonious
architectonical solutions and decorative elements, fol- So, let’s intervene, but then a fundamental question
lowing particular constructive and typological sys- arises: should it be at any cost? In fact, the urgent need
tems, avoiding its continuous physical degradation and for the interventions can’t lead to neglectful decisions.
abandonment. On the contrary, the interventions must be based on a

653
careful analysis made by experts from different areas,
involving not only structural and architectural matters,
but also historical and social issues.
With the objective of promoting sustainable
rehabilitation interventions, following ICOMOS
(International Council on Monuments and Sites) rec-
ommendations, NCREP (Nucleus for the Conser-
vation and Rehabilitation of Buildings and Built
Heritage, a team within FEUP – Faculty of Engineer-
ing of the University of Porto) have been participating
in many rehabilitation projects, doing inspections and
structural analysis of several old buildings, in partic-
ular at the Historical Centre of Porto. The work of
NCREP consists on the evaluation of the building
structural elements, which are usually made of stone
Figure 3. Main façades of two XIX century buildings at São
masonry walls and timber floors and roofs, assessing Domingos and Lóios squares, Porto.
their physical state and, consequently, the possibility of
rehabilitation. The paper will focus on the analysis of
the timber structural elements, namely on the Inspec-
tion and Diagnosis phases, which are fundamental for
a correct structural analysis.
The importance of these two phases as support of
the intervention is emphasized in the ICOMOS text
“Principles for the preservation of Historical Timber
Buildings”: “A thorough and accurate diagnosis of
the condition and the causes of decay and structural
failure of the timber structure should precede any Figure 4. Beams and transversal bars of a timber floor. São
intervention. The diagnosis should be based on doc- Domingos building, XIX century (1856), Porto.
umentary evidence, physical inspection and analysis,
and, if necessary, measurements of physical conditions
and non-destructive testing methods”.
defining the building contour: the two façades and the
two lateral walls. The other walls are usually wooden
2 PORTO HISTORICAL CENTRE BUILDINGS type panels, sometimes reinforced by a wooden strut-
and-tie system that improves its in-plane behaviour and
2.1 Architectural and constructive aspects the connection between walls and floors. This internal
The buildings of Porto historical centre were mostly structure, together with the timber floors and the roof
built between the XVII and XIX centuries, evolving trusses, strongly contributes to the buildings efficient
from older constructions. They present a rectangular performance under vertical, but particularly under hor-
plant, with a narrow front, a long depth and a variable izontal loads. As a result, it is important to ensure a
height: an average of 3 to 4 floors and a maximum of 8 good link between these elements in order to guarantee
floors (the top floors being usually added afterwards); a good and stable global behaviour.
the architecture of the inside is quite repetitive, usually The floors that are normally supported on the
with staircases at the centre, toped by a roof skylight, lateral masonry walls consist of a set of circular sec-
one compartment turned to the front and one to the tion beams (diameters varying between 0,15 m and
back of the building, Figure 3. 0,40 m), spaced 0,60 m, and a secondary transver-
These constructions that normally belonged to the sal bar system, Figure 4. Structural elements with
popular and bourgeois branch of the population aren’t rectangular section can also be found, usually as the
exclusive from Porto and exist all over Portugal and in result of interventions performed after the beginning
different European countries. Porto uniqueness resides of the XX century. The most common species used
in the fact that, in the old part of the city the slen- on the construction of floors were chestnut, oak, pine
der houses represent almost the totality of the civil and eucalyptus. However, floors executed with other
buildings. species also exist.
The timber roofs present variable configurations,
depending on the size of the building. As the spans
2.2 Building structural scheme
were normally lower than 6 m, the structure was very
The structural schemes of these buildings are also simple, with wooden trusses constituted by two rafters
quite typical, consisting of four stone masonry walls and a tie beam. In other situations, trusses with two

654

Figure 5. Timber roof structure. António Carneiro building,
XX century (1916), Porto.
rafters, two struts, a hanging and a tie beam were used,
Figure 5. These elements have circular sections with
diameters varying between 0,10 m and 0,30 m and the
wood species used in its construction are the same of
the floors.
2.3 Common structural damages
Concerning timber structures, this type of buildings
presents well defined structural damages. In relation
to the timber floors, two main damages can be iden-
tified: shear failure at beam extremities in the stone
masonry walls, due to the reduction of section from
biotic attacks (insects and dry rot fungus) associated
to high moisture contents; bending failure at mid-span,
originated by the presence of defects, drying cracks,
reduction of section due to biotic attacks, excess of
load from changes of use, etc. This degradation, asso-
ciated to the wood fluency, is often responsible for the
excessive deformation of the floors. Besides this defor-
mation, the floors can exhibit excessive vibration,
normally caused by a deficient transversal bar sys-
tem, insufficient beam sections or excessive spacing
between beams, etc.
As for the timber elements of roof structure, besides
shear failure at the beam extremities, similar to what
happens to the floors, it is not unusual to find biotic
attacks. As a matter of fact, the roof timber elements
present a great potential of biotic attacks, as they are
more exposed to the atmospheric agents. Deforma-
tion due to fluency is also observed and, if not taken
into account, it contributes seriously to the entrance of
water in the buildings.
Although one can not generalize, some buildings
have suffered structural modifications, in particular
after the middle of the XX century, with implications
on the structure global behaviour. In the case of the
XIX century São Domingos building, the wood stairs
of the building at the level of the first floor were
removed, with the objective of extending the room
space for commercial purposes. This alteration caused
an increase of load on the floors, which, together with
the lack of the under stairs structure, caused important
deformation of the upper floors towards the buildings
central area i.e., the staircase, Figure 6.
655
Figure 6. Deformation of timber floor. São Domingos
building, XIX century (1856), Porto.
3 THE ANALYSIS PHASE
3.1 Introduction
The rehabilitation of existing constructions asks for
a methodology that includes the steps of Analysis
(involving the technical inspection), Diagnosis, Ther-
apy and Control.The first two are essential for a correct
intervention as they help defining both technologies
and materials that better suit the characteristics of the
construction (Ilharco, 2007). The results from these
phases appear on technical reports, which are the fun-
damental asset for support of the intervention options
(Therapy and Control). Simultaneously, they are an
important instrument of the decision making process,
contributing to the knowledge increase on the building
and highlighting the different approaches and interests
of each area of expertise.
The first of these phases consists on the detailed
analysis of the construction, which is obtained
throughout an exhaustive technical inspection. It
involves the analysis of the construction from a global
to a detailed level, supplying information about the
critical areas and constructive elements needing inter-
vention (Costa, 2006).
In this point, the procedures and tools of the anal-
ysis phase, such as geometrical and photographic
registries, information about the construction his-
tory and, specially, the visual inspection, will be
focused. The complementary means of diagnosis, such
as non destructive tests (NDT), will be referred in
point 4. Finally, a case of a building where the floors
would be demolished by the owner will be presented
in point 5. In this building, a campaign of tests
(non-destructive and destructive) was performed with
scientific purposes.
3.2 Previous works
The first step of a construction diagnosis process con-
sists on the gathering of information, aiming to know
its historical and architectonical context. Generally, it
is possible to obtain information from construction
licenses and plans at the municipal archives, Figure 7.
This information, together with other data about, for
instance, modifications on the building initial project,
is essential. It can reveal constructive solutions and
Figure 7. XX century (1946) refurbishment project of São
Domingos building, XIX century (1856). Porto historical
municipal archive.
structural schemes that sometimes are hardly recog-
nizable without that input. In some buildings, the
consultation of the old projects makes easier the per-
ception of structural alterations to the initial project,
contributing to the understanding of many observed
damages. The reading of old manuals that refer to
construction technologies used at the buildings time
period is another helpful instrument.
3.3 Visual inspection
The visual inspection of the constructions has the
main objective of surveying the characteristics and
the damage of the structural and non-structural ele-
ments. Data such as the dimension and materials of
the structural elements of floors, roofs and walls, the
type of wall-floor-roof connections, etc., is essential
to understand the building structural schemes and
global behaviour. The information about the structural
elements can always be completed with a campaign
of non destructive or medium destructive tests. The
collected information is normally represented on the
structural plans of the building, together with cross
sections or other useful drawings, to make easier its
reading and understanding by all the involved exper-
tises. When numerical analyses are needed, this data is
also essential to construct the models of the building.
4 DIAGNOSIS COMPLEMENTARY METHODS
NON DESTRUCTIVE TESTING
4.1 Introduction
Nowadays there are many instruments that allow a fast
and non-intrusive evaluation of the physical state of
timber elements and, at the same time, the estimation of
the physical and mechanical properties of the material.
A brief description of equipment used in non destruc-
tive testing campaigns (NDT) carried out by NCREP
is made.
4.2 Hammer, Chisel and Hygrometer
The hammer and the chisel are basic instruments that
allow a quick evaluation of the damages of timber
656
Figure 8. Use of the videoscope for the visualization of a
timber floor. Belomonte Palace, XVIII century, Porto.
elements. Throughout the comparison between the
sound originated by the impact of a hammer in a
healthy and in a deteriorated piece of wood, it is possi-
ble to distinguish the existence of rotten parts (cavities,
cracks, etc.). On the other hand, with the chisel it is pos-
sible to evaluate the superficial hardness of the wood
and its integrity.
The Hygrometer allows the estimation of the mois-
ture content of the wood, contributing to the perception
of the potential for biotic attacks. High moisture
contents may reveal deficiencies in façades or, in
particular, on roof structures. This instrument allows
technicians to act preventively in order to eliminate
water infiltrations.
On the inspection of a XIX century building tim-
ber roof, a moisture content of about 20% indicated
the existence of water infiltration responsible for the
presence of biotic attacks (termite). On an early XX
century building, the 18% moisture content found on
a timber floor, together with poor ventilation condi-
tions, was responsible for the appearance of the dry
rot fungus Serpula Lacrymans.
4.3 Videoscope
The Videoscope is an instrument that allows the
observation of hidden areas or of difficult access, Fig-
ure 8. This instrument allows us to observe timber floor
structures through small holes or gaps. For instance, in
Belomonte Palace, an imposing XVIII century build-
ing that needed urgent rehabilitation measures, the
Videoscope was a precious help, enabling the obser-
vation of some floor frameworks and the detection
of metallic reinforcements and the presence of biotic
attacks without the need to remove the floorboard.
With this visualization it was possible to conclude,
without any damage for the structure, that it wasn’t
necessary to reinforce the beams, but only to provide
a curative treatment against insects.
4.4 Resistograph
In several timber floors and roofs it has been used the
Resistograph (model E300), an instrument that relates

Figure 9. Use of the resistograph on wood stairs. São
Domingos building. XIX century (1856), Porto.
the energy spent by the penetration of a needle in a
wood element with the material resistance to drilling,
Figure 9. The fact of carrying out barely impercep-
tible drills, without any influence in the mechanical
resistance of the piece, turns this instrument into one
of the most used in timber structures survey. In order
to guarantee the adaptation to the majority of timber
structures, it has been used a 300 mm length needle
and adopted a speed of penetration of 5 cm/minute.
With Resistograph results it is possible to detect
density variations and internal defects (voids, cracks,
etc.) along the cross sections of the structural elements.
Results can be easily understood and provide interest-
ing qualitative information, as they give draw registries
of the variation of wood drilling resistance. However,
there are some significant limitations in the use of this
instrument to achieve quantitative values, since there
are experimental studies only for some species and nei-
ther all correlations between the results obtained by the
Resistograph and the mechanical or physical charac-
teristics of the wood are acceptable. Nevertheless, the
characteristic which presents better correlations is the
density and it’s possible to obtain a profile of its radial
variation along the drilling, namely, the differences of
density between the initial lumber (Spring) and the
final lumber (Fall), as well as the referred losses of
density due to degradations/voids (Botelho, 2006).
As a result of its use, it was possible, for instance,
to verify that the timber floors and stairs of a XIX cen-
tury (1856) building, located in São Domingos square,
were in excellent state of conservation, therefore not
demanding reinforcement measures. On other hand,
in the timber roof of Belomonte Palace, a building
around 300 years old, some interior voids were discov-
ered in the supports of tie beams (diameter of 0,30 m)
with around half of its cross-section. The extension
of the degradation along the elements was evaluated,
allowing estimating the length of a future prosthesis to
install (Faria, 2002). In Figure 10 illustrates the results
given by the Resistograph, where a quite long void on
the radial profile of one of the tested elements was
657
Figure 10. Results from the resistograph on a tie beam of a
wood truss of Belomonte Palace, XVIII century, Porto.
Figure 11. Pilodyn used in a timber floor. António Carneiro
building, XX century (1916), Porto.
detected. Once again, the qualitative information is
very helpful, since it estimates the extension of existent
degraded material along the cross section.
To obtain quantitative values it’s necessary to per-
form a statistical handling of the results, obtaining
a medium value, called Value of Resistograph (VR).
After that, and if applicable to the wood specie of the
element, it can be used one of the several studies that
correlates this value (VR) with physical or mechani-
cal parameters of that wood specie, such as the density
(ρ), the strength (fc,90 ) and the modulus of elasticity
(Ec,90 ), obtained in laboratory tests. Despite the poor
correlations that were usually found, in particular with
the mechanical parameters (fc,90 and Ec,90 ), there are
some acceptable correlations. For instance, Botelho
(2006) reached a correlation R2 = 70,09% between
wood density and VR for Pinus radiate.
4.5 Pilodyn
Another commonly used instrument in the survey of
timber floors and roofs is called Pilodyn and has a
functioning similar to the concrete Shmidt Hammer,
Figure 11. With this instrument it’s possible to achieve
physical parameters of the wood, such as density, from
correlations with the superficial hardness or, by other
words, with wood superficial penetration resistance.
It’s used in the detection of illness manifestations
through periodic measurements, in the establishment
of resistance classes and categories, and productivity
parameters in relation to the density between different
wood species (Feio, 2005).
In the surveyed structures, the Pilodyn allowed
finding wood defects through the reduction of the
wood resistance to the penetration of the needle.
Measurements are made on different points of struc-
tural elements (middle spam and supports) and on
different structural elements with the objective of
doing, with a large number of surveyed structures,
comparative studies about the elements integrity.
It is important to refer that, due to the short num-
ber of existing correlations for the different species,
it isn’t possible to reach quantitative values, but only
qualitative, but even so valuable information about the
conservation state of timber elements.
Most of the correlations for the values obtained by
this equipment relates the superficial hardness only
with the wood density and not with mechanical proper-
ties, which ends up being a disadvantage. On the other
hand, the majority of the experimental tests showed
that the moisture content considerably affects the depth
of penetration (Bonamini et al., 2001) and, so, it’s
important to evaluate correctly the wood moisture
content. Görlacher (1987) obtained good correlations
(R2 = 0,74–0,92) between the density and the depth
of penetration of the Pilodyn for an high number of
measurements. Furthermore, Feio (2005) found some
correlation not only for the density, but also for the
modulus of elasticity and the strength of the chestnut
wood (Castanea Sativa, Mill). However, and although
for the density the correlation is acceptable, for the
modulus of elasticity, Ec,90 , as for the strength, fc,90 ,
the estimated correlations are poor, and the author does
not advise its use as a quantitative measure. Turrini and
Piazza (1983) proposed empirical relations between
the impact force and the modulus of elasticity, adopt-
ing a factor of reduction for the modulus of elasticity
based on the visual classification of the elements: 0,8
for elements without defects and 0,5 for elements with
some knots and small degraded areas.
4.6 Seismographs
The seismographs are instruments used frequently by
NCREP in the inspection of timber floors. In partic-
ular, two 18 bits resolution seismographs have been
used, Figure 12. They include tri-axial accelerome-
ters, and an autonomous memory. The transference
of the information to a computer allows afterwards
sign analysis. Readings are done in different loca-
tions on a floor, supplying its natural frequencies.
This evaluation permits estimating the floor stiffness
and, consequently, the efficiency of its structure or its
physical and (or) material state.
4.7 Load tests
The load tests allow an in-situ realistic evaluation of
the stiffness of a structure and, in particular, of the
capacity of a structure to resist a certain load. These
are expensive tests that are used by NCREP whenever
considered convenient. The tests are normally done for
loads of about 1,5x the service load and consists on the
658
Figure 12. Seismograph used by NCREP.
application of controlled loading and unloading cycles
with continuous force and displacement monitoring. It
is important to refer that after unloading no important
residual deformations should exist on the structure.
Thus, before carrying out such a test, it’s important to
estimate, numerically or not, its behaviour, trying to
avoid surpassing the structure elastic range. In chap-
ter 5 a case of a building in which a load test was
performed is presented.
5 ANTÓNIO CARNEIRO BUILDING A CASE
STUDY
5.1 Introduction
In one of the buildings analyzed by NCREP the
possibility of performing destructive and medium-
destructive tests on the floors came up because, due to
architectonical incompatibilities, the general project
dictated its demolition. The adoption of destructive
tests is naturally to avoid, because it doesn’t allow
the tested elements to be maintained. However, this
opportunity of testing a set of floors that would be
demolished ended up being extremely important, since
it allowed the evaluation of parameters and character-
istics that will surely be helpful in similar structural
rehabilitations.
The surveyed building, built in 1916, has an area of
12,0 × 30,0 m2 and presents floors with 0,20 m diam-
eter beams, spaced 0,60 m and with 6,0 m span. Above
the beams there is a 2,5 cm thick floorboard. Perform-
ing the transversal link, there are 0,10 m diameter bars,
spaced 2,0 m.
Besides a campaign of non destructive tests carried
out with the Pilodyn and the seismographs, among
other instruments, a load test was then performed
together with a very particular test for the evaluation
of the beam-wall friction, using a hydraulic actuator.
The campaign of tests, in particular the load test
and the frequencies evaluation, had the main goal
of assessing the global floor and the single beams
behaviour when submitted to a monotonic loading his-
tory. The setup was also prepared to allow evaluating
the effectiveness of the transversal bar system and the
floorboard contribution to the stiffness and resistance
of the floor.
Figure 13. Load test on a timber floor. António Carneiro
building, XX century (1916), Porto.
5.2 Load test
The strength capacity and the stiffness of the floors
are strongly affected by factors such as the moisture
content and the malfunctioning or degradation of its
structural elements, beams and transversal bars, and
of the floorboard. The load test was carried with the
main purpose of evaluating the structural behaviour of
the floor and the contribution of its single elements to
its global resistance and stiffness, Figure 13. In order
to better understand and to be more conclusive about
the floors behaviour, the test was done on a 3,5 m wide
strip (corresponding to an assembly of 7 beams) that
was isolated from the rest of the floor. It was chosen a
well preserved floor area (beams, transversal bars and
floorboard).
With the purpose of estimating the efficacy of the
transversal bars and the floorboard on the distribution
of the loads, only the central beam was loaded. Three
reservoirs with a capacity of 600 kg each, making up a
total load of 1,80 tons were used. A flow measurement
instrument was installed at the entrance of the reser-
voirs to control and register the volume of water i.e.,
the load installed on the floor. During the loading pro-
cess, the vertical displacements of all the 7 floor beams
were monitored using LVDT’s. The data, load and dis-
placements, was acquired continuously and analysed
using an on-line homemade software.
Having in mind the objectives referred previously,
the tests were done under three conditions: (a) with
the original floor, i.e. with transversal bars and floor-
board; (b) without the transversal bars i.e., only with
the beams and the floorboard; (c) on an isolated beam.
Therefore, the load test, which is normally non destruc-
tive, ended up to have a destructive character. In order
to proceed to the sequence (a) to (c), the test remained
within the elastic behaviour range. However, with the
intention of complementing the obtained information,
a load test will be done at the Laboratory of Earth-
quake and Structural Engineering (LESE) to evaluate
the ultimate behaviour of the isolated beams.
The analysis of the results, in particular of the load-
displacement diagrams, and for the applied load range,
shows a linear behaviour, Figure 14. From the results
obtained it was possible to estimate the wood modulus
of elasticity (around 12 GPa), verify the increase of
659
Figure 14. Load-displacement diagram for two isolated
beams loaded simultaneously (the load refers to the total load
applied).
Figure 15. Seismograph used in single beam.
the floor stiffness conferred by the floorboard and the
poor effectiveness of the transverse bars to distribute
the applied loads when compared to the floorboard.
Other conclusions and analyses from the test will be
published in the future.
5.3 Seismographs
The data from the seismographs were particularly help-
ful in the evaluation of the relative importance of floor
elements. In particular, several readings were made.
Firstly, the floor with the original width was analyzed
(30,0 m); Secondly, the measurements were repeated
after the cut that isolated the 3,5 m strip from the rest
of the floor, (final area of 6,0 × 3,5 m2 ). Afterwards,
measurements were done without the transversal bars
and, finally, on an isolated beam, Figure 15. The com-
parison between the results of these four situations,
with the main natural frequency changing from 7,0 Hz
to 10,0 Hz, allowed evaluating the importance of the
floorboard, transversal bars and beams, to the floor
structural behaviour and, simultaneously, estimating
the medium values of the wood modulus of elastic-
ity. It was also verified that the values of the natural
frequencies measured in-situ were within the range of
theoretical values calculated according to EC5 (1998).
Apart from these conclusions, the measurements
using the seismographs can be used to calibrate numer-
ical models to simulate such type of timber floors
on old buildings. When needed, the simulation of the
floors can be done considering the beams and the
transversal bars as they exist on the floor, and adopting

Figure 16. Beam wall friction test using a hydraulic
actuator.
auxiliary bars, with small inertia, to simulate the floor-
board, linking the beams. Another possible model, that
tends to approach the floorboard real effect over the
whole structure, consists on considering low thick-
ness slab elements to simulate the floorboard. These
elements rest on the beams and transversal bars, but
aren’t connected to the resistant walls, as it is usually
observed in buildings (Neves, 2004).
5.4 Beam-wall friction test
To evaluate the resistance of the friction connections
between the timber beams and the masonry walls, a
destructive test was carried out. A segment of beam
was cut of to install a hydraulic actuator that would
act along the beam axis. In order to maintain the two
segments of the tested beam in the original position,
it was conceived a metallic structure with pulleys to
support it (not allowing vertical movements but free
horizontal displacements), Figure 16. To set the actu-
ator on the beam, metallic corner cupboards fastened
by M12 bolts were used. The test consisted on apply-
ing pull out and in movements of the beam in relation
to the wall. Displacement transducers were installed
at the extremities of the beam to evaluate the load-
displacement diagram of the beam-wall connection.
With the purpose of verifying any possible horizon-
tal movement of the wall during the test, transducers
were installed on the walls. The results permitted to
conclude about the limited efficacy of the connection
between the two elements, traduced by a friction force
of about 2,0 kN.
6 CONCLUSIONS
To allow the preservation of Porto Historical Centre
buildings (UNESCO World Built Heritage), NCREP-
FEUP has participated in many projects, analysing the
state of conservation of their structural elements. The
present paper refers to the work developed on the anal-
ysis of timber structures, floors and roofs, namely on
the Inspection and Diagnosis phases. It was concluded
that, with the joint use of several non destructive testing
660
techniques, it is possible to withdraw helpful informa-
tion about structural timber elements. Nevertheless,
some limitations were verified in their utilization due
to the difficulty to estimate quantitative values for
mechanical and physical parameters.
To be used in a more trustworthy way, the Resisto-
graph and the Pilodyn need experimental studies with
different wood species to create newer correlations.
Nevertheless, these instruments were very helpful in
the accomplishing of qualitative parameters, in partic-
ular in the perception of degraded areas, with voids
or cracks. In particular, the Resistograph permitted to
find voids inside structural elements that apparently
were in good state of conservation. On the other hand,
the Pilodyn allowed, through tests on different struc-
tural elements, to conclude about their relative state of
conservation through superficial hardness.
In addition, the use of seismographs on timber
floors permits the measurement of natural frequen-
cies and, consequently, the estimation of the average
modulus of elasticity. In the future, with the infor-
mation collected on different surveyed floors, it will
be possible to organize a database that will allow us
to establish correlations between the measured values
and the integrity of the floor.
Finally, it is important to refer that medium-
destructive and destructive tests, despite of being
applied only under particular conditions, can be an
important contribution for the knowledge on the mate-
rials and structures. In particular, the load test that, in
this case, had a destructive character, allowed estimat-
ing the floorboard contribution to the floor stiffness,
as well as verifying the low efficiency of the transver-
sal bars on the distribution of vertical loads. The
beam-wall friction test, only possible with the cut of
the tested beam, allowed concluding about the weak
contact conditions between the timber beam and the
masonry wall.
REFERENCES
Botelho, J. 2006 – Avaliação não destrutiva da capacidade
resistente de estruturas de Madeira de edifícios anti-
gos. Tese de Mestrado em Reabilitação do Património
Edificado. Universidade Porto – Faculdade Engenharia.
Portugal.
Costa, A., Ornelas, C., Paupério, E., Guedes, J., Ilharco, T.
2006. A Inspecção como Instrumento de Decisão. Apre-
sentação de um Caso Prático num Edifício com Valor
Patrimonial. Patorreb 2006. Porto.
ENV 1995-1-4. 1998. Eurocode 5 – Design of Timber Struc-
tures Part 1-1: General – Common rules and rules for
buildings.
Faria, A. 2002, Reabilitação de Coberturas em madeira em
edifícios históricos.A Intervenção no Património; Práticas
de Conservação e Reabilitação, Porto.
Feio, A. 2005 – Inspection and Diagnosis of Historical
Timber Structures: NDT Correlations and Structural. Tese
 de Doutoramento em Engenharia Civil, Universidade do
Minho – Escola de Engenharia. Portugal.
Google Earth, Google.
Gorlacher, R. 1987. Non destructive testing of wood: an in-
situ method for determination of density. Holz as Roh-
und Werkstoff. Vol. 45.
Ilharco, T., Romão, X., Paupério, E., Guedes, J., Costa, A.
2007. Organization of information on Built Heritage using
661
Multimedia Technologies. Euro-Mediterranean Confer-
ence Traditional Mediterranean Architecture. Barcelona.
Neves, N., Costa, A., Arêde, A. 2004. Identificação
Dinâmica e Análise do Comportamento Sísmico de um
quarteirão localizado na cidade da Horta – Ilha do Faial.
Sísmica 2004. Porto.
Turrini, G., Piazza, M. 1983. Il recupero dei solai in legno.
Esperienze e realizzazioni”. Recuperare, Vol. 7.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Baroque roof structures in Transylvania – Research and analyses

D. Makay
Irod M Ltd, – Built Heritage Conservation Research, Planning and Consulting, Cluj, Romania

B. Szabó
Technical University of Cluj, Romania

ABSTRACT: Baroque roof structures encompass important and unique historic values, within: (a) their struc-
tural conformation (trusses and longitudinal bracing systems); (b) the craftsmanship of framing and jointing
(carpenter marks, joints); (c) the authentic, historic timber material that has been preserved for centuries.
Given the high level of safety characterising Baroque roof structures, a relatively significant number of these
survive in Transylvania, though there is no through inventory, yet.
The lecture presents the typology of Transylvanian Baroque roof structures, the PhD research programme
carried out in order to analyse these structures, and through two case studies the level of safety, durability and
the unique character of these structures.
Though professional interest has increased towards historic (roof) structures, Baroque roofs in Transylvania
face a double threat. Scientific research needs to result in a guideline to be used in order to correctly
conserve/reinforce Baroque roof structures throughout Transylvania.

1 INTRODUCTION passed on from generation to generation in carpen-

Baroque roof structures developed in Continental
Europe (France, Austria, Germany, Czech Republic)
at the turn of the 16th–17th centuries became widely
used in Transylvania in the 18th century. Transylvania
is the easternmost territory where this type of conti-
nental historic roof structure can be found. In other
parts of Romania (Moldova, Walachia, etc) this type
of roof structure cannot be found.
As concerns their development in time, a number
of transitory structures can be identified between the
previously used Gothic roof structures and the Baroque
ones. Later, pure Baroque structures became widely
constructed, which persisted into the first half of the
19th century, when Eclectic roof structures came to
replace them.
Given their high level of safety, a relatively sig-
nificant number of Baroque roofs can be found in
Transylvania, though there does not exist a thorough
inventory, for the moment.
2 DEFINITION
Historic roof structures are structures constructed by
carpenters following empirical-intuitive conceptions
based on knowledge accumulated empirically, and
ter’s guilds. Researching, understanding, scientifically
describing these types of structures present a special
interest, as they incorporate historic and technological
values worth preserving for future generations.
Historic roof structures with Baroque character
include roof structures of buildings constructed in
Baroque architectural style, as well as roof structures
of ensembles built in other architectural styles:
– earlier buildings, with the roof rebuilt in the
Baroque period, or
– later buildings, with Baroque carpentry still sur-
viving within carpenter’s guilds (in Transylvania
most early 19th century buildings still have roof
structures with Baroque character).
Roof structures with Baroque character are made up
of two plane systems of linear bars, laid out in two
orthogonal directions: transversal load-bearing plane
systems (main and secondary trusses), and longitudi-
nal bracing systems, placed primarily in the rafters’
plane. The component elements (bars with rectangu-
lar or pentagonal sections) are predominantly realised
of soft wood – pine or other resinous timber.
In order to illustrate the structural conformation
of roof structures with Baroque character, a typical
Baroque roof has been developed, using the data of
over 50 such roofs surveyed. This is given below, to

663

Figure 1. Early Baroque roof, a transitory structure between
Gothic and Baroque, Franciscan church, Cluj, Romania.
Figure 2. Main truss of a typical Baroque roof structure:
1 – tie-beam 14 × 16; 2 – common rafter 12 × 16; 3 –
compound rafter 13 × 24; 4 – straining beam 13 × 20; 5 –
counterbrace 13 × 17; 6 – collar beam 13 × 15; 7 – king
strut / (double) suspension bar 2 × 12 × 17; 8 – wall plates
2 × 14 × 11; 9 – header beam 13 × 16; 10 – pentagonal eaves
purlin/pentagonal lower plate 15 × 20; 11– middle pentago-
nal purlin/upper pentagonal plate 13 × 19; 12 – longitudinal
beam of the bracing system 12 × 13; 13 – sprocket 5 × 13;
14 – diagonal of the longitudinal bracing 11 × 14.
provide a synthesis of structural sub-ensembles and
the names of various elements:
The Figures illustrate the typical Baroque roof
structure of a construction with characteristic dimen-
sions: 10 m span, 28 m length, 50.2◦ inclination angle
(6/5 height/half-span ratio). The main truss (Fig. 1)
includes the Baroque straining system made up of:
the tie-beam (1), pair of compound rafters (3); strain-
ing beam (4) and counterbraces (5), including also a
king strut/double suspension bar. The following fig-
ures give the conformation of the typical secondary
truss (Fig. 3), as well as a bay of the longitudinal
bracing system (Fig. 4).
664
Figure 3. Secondary truss of a typical Baroque roof struc-
ture. Additional element: 15 – trimmer 14 × 17.
Figure 4. Secondary truss of a typical Baroque roof
structure.
2.1 Terminology
The Anglo-Saxon (coastal) roof structures differ both
in structural conformation and elements from the
continental historic roof structures. Therefore the
English terminology of historic continental roof struc-
tures generally, and the one related to the Baroque
roof structures especially is a significant issue. The
present lecture uses the terms mainly as given in
Bálint SZABÓ’s Illustrated Dictionary of Historic
Load-Bearing Structures.
The following aspects need to be highlighted:
– The Romanian/Hungarian terminology derives
from the German one, due to the coexistence within
the Habsburg Empire (Monarchy). The 20th cen-
tury terminology used by craftsmen/builders has its
origin in the German 19th encyclopaedic books on
carpentry (which actually focused on the descrip-
tion of 19th century, eclectic/early engineered
structures, not on Baroque ones).
– Concerning Romanian (and Hungarian) terminol-
ogy, we need to mention that for almost half a
century (during the Communist Era) both timber
structures and historic buildings were neglected,
therefore their terminology has been forgotten.
– The above mentioned dictionary is trying to reflect
the structural behaviour/role of the various ele-
ments within the structure.
In order to promote international cooperation and
exchange of information, it would be necessary to
elaborate an internationally accepted English termi-
nology for non Anglo-Saxon historic roof structures.
The present lecture could present a starting point
in order to discuss the proposed terminology for roof
structures with Baroque character.
In order to clearly define roof structures with
Baroque character, contrasted to medieval (Gothic)
roof structures, the innovative elements and structural
conformation need to be shown.
2.2 Innovations of Baroque roof structures
Baroque architecture changed the volumes and ratios
of buildings, therefore in many cases it was necessary
to place the eaves height below the keystone of the
vault.
(a) This has determined one of the innovations gener-
alised within Baroque roof structures, that in most
cases secondary trusses do not possess tie beams
(various solutions within typology, criterion: 4 –
see Fig. 8).
From the point of view of structural conformation,
secondary trusses without their own tie-beams
are not self-supporting systems (contrary to the
secondary trusses of Gothic or Romanesque struc-
tures, where secondary trusses do not transfer
gravitational loads to main trusses).
(b) The invention of the Baroque straining system (see
Fig. 1) – which is actually a “false” or “partial”
straining system. It is working properly as a strain-
ing system only in the cases of secondary trusses
without individual tie-beams and without collar
beams. In the other cases the straining systems
have determinant role in stiffening the structure for
horizontal loads. (The number of Baroque strain-
ing systems is the 2 criterion of the typology).
Secondary trusses without individual collar beams
are rare. (The roof structure of the Reformed
church, Cluj, secondary truss type II, Fig. 12, has
no individual collar beam).
(c) Mansard roofs – were developed in France, within
the Baroque architectural period, and mean roofs
(rafters) with broken, uneven line (plane). This
represents the 1st criterion of the typology.
(d) The only (continental) historic roof structure in
Central-Eastern Europe containing elements with
665
Figure 5. Eaves pentagonal purlin within the main truss
joint of: tie beam (1), main/compound rafter (3) and com-
mon rafter (2). Upper (middle) pentagonal purlin within the
main truss joint of: straining beam (4), main/compound rafter
(3), collar beam (6) an upper compound rafter and common
rafter (2).
pentagonal cross section, is the one with Baroque
character. The special elements are: the eaves
pentagonal purlin and the upper (or middle) pen-
tagonal purlin (Fig. 5).
There exists a debate concerning the cor-
rect designation of these elements as “purlins”.
Both in eclectic and Anglo-Saxon roof structures
purlins transfer gravitational loads from rafters
in secondary axes (trusses) to main trusses. But
in Baroque roof structures generally secondary
trusses possess either or both tie beams and col-
lar beams, which actually minimise the transfer
of efforts from secondary trusses to main trusses,
Therefore the lower/upper pentagonal plate is also
a term frequently used for these elements.
In less important roof structures or where the
craftsmanship of the carpenters failed, either or
both pentagonal purlins can a have simple, rectan-
gular shape.
(e) The middle suspension bar/king strut (7) in
Baroque roof structures can be made up of a com-
pound section (double bars), representing the 3rd
criterion of the typology.
(f) The typical Baroque longitudinal bracing sys-
tem (Fig. 4) is one of the most innovative and
most determinant sub-ensembles of a Baroque
roof structure. It represents an efficient way of
providing longitudinal stiffness for the structure.
The Baroque longitudinal bracing system, like
bracing in Anglo-Saxon roofs, is always placed
in the plane of the rafters (Fig. 6). The various
patterns of the bracing system are given within
the 6th criterion of the typology.
(g) Unlike medieval roof structures based exclusively
on carpenters’ knowledge passed from genera-
tion to generation within the guilds, the Baroque
roof pattern was designed and drawn. There are
treaties and drawings of architects preserved, such
as Franz Ignaz Michael Neumann, The Construc-
tion Plan for theVaulting and the Roof, republished
by Hansmann (2000).
Figure 6. Double St Andrew’s cross Baroque longitudinal
bracing, Jesuit Church, Bamberg, Germany.
Baroque roofs are mostly surveyed, assessed and stud-
ied – as integrated parts of historic buildings – in order
to be repaired, conserved, reinforced or converted for
a new use.
These researches are objective-oriented and in many
cases can hardly be used for scientific purposes. The
authors promote the elaboration of a guideline for
those working in the professional survey, documen-
tation, and conservation or reinforcement planning of
historic roof structures generally, and Baroque ones
particularly.
3 TYPOLOGY
In order to provide a researchable data base, as well
as an organised overview of the roof structures with
Baroque character, a typology is a useful tool. The
typology presented within this lecture is based on the
measured survey of 50 Baroque roof structures, as well
as on visits to Baroque roof structures in Hungary,
Austria, Germany, and further roofs (not measured,
just inspected) in Transylvania.
Two main principles were followed when develop-
ing the present typology:
– It refers only to pure (mature) Baroque roof struc-
tures; it does not include hybrid or transitory
structures neither between Gothic and Baroque, nor
between Baroque and Eclectic structures (unlike
the detailed topography of roof structures in Mur-
Mürz area, Steiermark, Austria, Caston).
– It analyses the structural and geometrical confor-
mation of both structural sub-ensembles and of
elements.
This typology presents a working phase of the Transyl-
vanian Baroque Roof Structure Typology, which will
embody the results of a through inventory (phase I) on
a limited, but representative geographical area (Cluj
666
Figure 7. Roof structure of the Boys’ Dorm, Bethlen Col-
lege, Aiud (Utilitas, Transylvania Trust, 1999–2000). Code
according to the typology: B.2.1.-a(3)-II-(1).
county), to be carried out in 2008. The typology will
also be open for completion by an eventual through
Transylvanian inventory.
The typology, Figure 8 has been developed using 6
criteria in order to identify the specific characters of a
given Baroque roof structure, as follows:
(1) continuity / plane of common rafters;
(2) number and type of Baroque straining systems;
(3) type / existence of a suspension bar;
(4) horizontal force transmission system;
(5) main / compound rafter;
(6) pattern of the longitudinal bracing system.
Using this classification, all Baroque roof structures
can be identified by a code made up of 6 letters com-
bined into three groups: 1 + 2 + 3 – general + main
truss characteristics – 4 – identifying the load trans-
mission from secondary to main trusses – 5 + 6 –
describing the longitudinal bracing system.
Using the typology, the typical roof presenting ele-
ments of the Baroque roof (chapter 2) has the following
code: A.1.2(c)-b(3)-I. (1).
3.1 Classification by the continuity/plane
of rafters
All Baroque roof structures fit into one of the two
basic groups: (A) – Continuous/linear common rafters
(plane); (B) – Interrupted common rafters (mansard
roofs).
3.2 Classification by the number and type of
Baroque straining systems
Up to a 10.00 m span, in many cases Baroque straining
systems placed in main trusses are not combined with
other straining systems on upper levels (eventually just
with an upper collar beam). This represents the basic
Figure 8. Transylvanian Baroque roof structure typology, working phase 2007.

667
type of the classification (1), designated by the second
character of the code in the typology.
The most general type though is (2), where the
Baroque straining system on the first level is combined
with other straining systems on one or more upper
levels.
The third version included in the typology (3), is
rare within Transylvania, being used for large spans
and in important buildings. This subcategory includes
roofs with two Baroque straining systems placed one
above the other, combined or not with further straining
systems on upper levels.
In Austria, Germany these types are more usual than
in Transylvania/Hungary, connected to the economic
power of the area/time. Even more there can also be
identified Baroque straining systems on three levels.
This version was not included in the present typology,
as none was identified in Transylvania, yet.
In a number of special cases: large spans, or use Figure 9. Baroque roof structure of the Piarist church in
of the attic space as granary, or deposit functions, the Cluj. Code according to the typology: A.2.2(c).-c(3)-I-(1).
Baroque straining systems are additionally strength-
ened. These special cases are included in the typology
marked with ∗ , at criterion 2∗ .

3.3 Classification by the existence and type of

(compound) suspension bar
Reduced spans in a high number of cases are coupled
with the lack of suspension bars, a version identified
by the third character (1).
Suspension bars have a decisive role in limiting the
long term creeping of tie beams; therefore they are
present in all structures with a span over 10.00 m –
version identified by note (2).
Two structural solutions were identified:
(s) – hanging bar made up of one element. In
this case the joints between the tie-beam and
the suspension bar are solved using wrought iron Figure 10. The main truss of the Reformed church’s
bends; nave −15.70 m span, Cluj, Romania. Code according to the
typology: A.2.2(c).-c(3)-I-(1).
(c) – hanging bar made up of a compound ele-
ment (two bars) – joint solution: pegged/treenailed
half lap. (c) Secondary trusses with shoes of eaves purlin, with
maximum load transfer from secondary to main
trusses (compared to the other two solutions).
3.4 Classification by the transmission mechanism
of loads from secondary trusses to main ones Between two main trusses there are 2–5 secondary
trusses, usually 3.
The second group of letters describes the secondary In later structures, a frequently adopted solution is
trusses, which belong two three main groups (particu- alternating different types of secondary trusses.
lar solutions identified are not included):
(a) Secondary trusses with tie-beams – used fre-
quently in early Baroque roofs and buildings with- 3.5 Classification by the solution of compound
out vaults above the upper level. Secondary trusses rafter and elements of the longitudinal
are self-supporting, load transfer being minimal, bracing system
mainly from horizontal loads. The more common solution is that the compound rafter
(b) Secondary trusses with trimmers and header of the Baroque straining system is made of a timber
beams, the most general solution, both in buildings with large cross section, 17 × 30, having a double role
with and without vaults above the upper level. in the structure.

668

Figure 11. Secondary truss type I (with own tie-beam) of
the Reformed church’s nave – 15.70 m span, Cluj, Romania.
Pattern: MT/ST-II/ST-I/ST-II/MT.
Figure 12. Secondary truss type II (with shoes) of the
Reformed church’s nave – 15.70 m span, Cluj, Romania.
The main role of the compound rafter in the
Baroque straining system is working for axial forces
(compression from gravitational loads, compres-
sion/tension from wind loads). Due to the jointing
between the compound rafter and tie-beam, as well as
counterbrace and straining beam, transfer of bending
moment is also possible (chapter 4).
Its secondary role is serving as the inclined post
of the longitudinal bracing system, placed parallel to
the plane of rafters. This solution is marked by the
character – I – within the third group of characters.
A less used (cheaper and more rudimental) solu-
tion is having two different bars with smaller sections
placed one on the other (II).
3.6 Classification by the pattern of the
longitudinal bracing system
It refers only to the Baroque longitudinal bracing
system. The common solution with knee braces on
669
different levels is not considered a defining criterion
in classifying Baroque roof structures.
Within the analysed structures five patterns were
identified in Transylvania, which does not exclude the
existence of further patterns. The shape is marked with
a symbol showing the pattern made by the diagonals.
The number of levels on which these are placed within
the structure is given in brackets. In the great majority
of the studied Transylvanian cases, Baroque longitu-
dinal bracing systems are placed on one level (unlike
examples known from abroad – Figure 6, two levels).
The patterns are the following:
–  – the most common solution in Transylvania,
made up of a pair of ascending and a descending
diagonals, fixed in the eaves and the middle pen-
tagonal purlin, also strengthened by a longitudinal
element placed at the middle of their height.
– Y – is a similar pattern, the only difference being
that the lower end of the diagonals is fixed to the
compound rafters.
– / or \ – a very rare pattern, with only one of the
diagonals present (Fig. 1).
– X – Saint Andrew’s cross pattern, used for more
elaborate constructions, and rare in the analysed
structures, in Transylvania.
– XX – double Saint Andrew’s cross, even more rare
(a single case in more than 50 roofs analysed).
A through survey and the statistical data gathered will
serve as the basis of qualitative assessment of various
structures (rarity, structural conformation, craftsman-
ship of jointing, span, age all will be basic criteria in
rating the value of a structure).
The authors represent the Doctoral School of the
Technical University of Cluj-N. The PhD thesis of
structural engineer Dorottya Makay, supervised by
Professor Bálint Szabó, The Transylvanian Baroque
Roof Structures, aims at elaborating the rating of these
structures, as well at creating a guidelines for the
calculation of Baroque roof structures.
4 MODELLING BAROQUE ROOF
STRUCTURES – CASE STUDY: REFORMED
CHURCH, KOGALNICEANU STREET, CLUJ
In the following we shall formulate the main points
and questions raised by the modelling and calculus of
Baroque roof structures, exemplified on the 15.70 m
span, 56◦ angle roof structure of the Reformed church
in Cluj.
The safety of roof structures with Baroque char-
acter (being at least 150–200 years old) is proven by
their very existence, having survived extreme condi-
tions while possessing (in many cases a high number
of) dysfunctional joints and / or biologically decayed
or missing elements.

Figure 13. Axis 8 VM partial 3D model (of two bays).
A perfectly behaving structure (no distortions,
deflections, or any other signs of incorrect structural
behavior) is an ideal tool to develop a fair static / math-
ematic model, and in the same time to countercheck
data deriving from technical legislation.
The following questions are formulated regarding
all elements of the static model:
– Loads – especially wind loads, when a recent update
of the Romanian wind-load standard results for the
studied roof in a load 2.5–3 times greater than the
one deriving from the former standards. (Even with
the increased loads the safety factors of various
elements are over 1.00).
– Material characteristics – as new timber is charac-
terized by decreasing resistance, due to changed
technologies both in growth and processing,
standards are also giving diminishing values.
A database with characteristics of 150–200 year-old
timber would be useful.
4.1 2D modeling versus 3D models
As soft- and hardware possibilities have developed
spectacularly during the last decades, interest in 3D
modeling has also increased.
2D models raised their own set of questions in their
time, concerning the correct appreciation of load trans-
fer from secondary to main trusses, which will actually
receive correct answers when 3D models will properly
function.
Within 3D modeling, the eccentric jointing of ele-
ments is creating one of the most complicated prob-
lems, especially inAxis 8VM, where eccentric jointing
cannot be created, unlike with the software Robot.
Figure 14 shows the jointing of element-axes within
the lower main truss joint (cf. Fig. 5).
A useful tool in 3D modeling is limiting the full-
structure model to a characteristic section (minimum
670
Figure 14. Model of the 3D joint.
Figure 15. Mathematical model of the rotational capable
moment – straining beam-counterbrace joint.
two bays), that is describing the behavior of regular
elements (Fig. 13).
4.2 Bar-end conditions
One of the most debated questions concerns the inner
joints of the model.
The safest modeling, and clearly closely describing
the truss behavior, is using pin joints. Structures with
pin joints are more flexible, so if a structural checking
gives correct results with pin joints, it will surely resist
with partially fixed joints, too.
Though this approach is correct in the structural
expertise of a particular roof, the scientific approach
is the one identifying the extra safety within a Baroque
roof structure, introduced trough joints (double pegs
combined with mortice and tenon and notches) with a
capable bending moment: Figure 15.
A mathematical model of the rotational moment
capacity of the joints has also been developed. The
next step of the PhD research program is in situ test-
ing in order to control the mathematical model through
on site measurements.
Though calculation and modeling are important
steps in conserving Baroque roof structures, neither
scientific research nor the guidelines should end at
this point.
Even in the case of the studied roof structure, calcu-
lation has demonstrated that the structure is working
properly, as can be red on the structure as well, which
means that conservation should be reduced to local
repairs of the biologically damaged elements.

Figure 16. Minimum intervention carpentry repairs,
Bethlen College Aiud, ACTT 2000.
5 THE PRINCIPLE OF MINIMUM
INTERVENTIONS – CONCLUSIONS
Baroque roof structures are safe and durable. The
roof of the Boys’ Dorm of the Bethlen College, Aiud,
had by 1999 more than 50% of its trusses decayed
(one or more joints dysfunctional or one or more ele-
ments missing). Nevertheless, the ensemble was still
standing.
The British practice of minimum intervention was
introduced to Romanian practitioners first through the
ACTT 2000 project. This training programme was
jointly organised by the Transylvania Trust and the
British Institute of Historic Building Conservation.
(Since 2001 they run the Built Heritage Conservation
Training Centre within the Bánffy castle in Bontida,
Romania).
671
In the framework of carpentry workshops, in situ
repairs based on the minimum intervention principle
was carried out. (Fig. 16)
A continuity joint assures 50–70% of the capacity
of the element’s full cross section. Therefore the hand-
out of Baroque roof conservation needs to discuss the
estimation of the overall change in stiffness and resis-
tance of the structure through the introduction of a high
number of repair joints.
Baroque roof structures represent exceptional his-
toric values which need to be conserved for future
generations. Therefore their inventory and perfect
structural understanding is indispensable in order to
assure their protection, conservation and consolida-
tion, without destroying the historic fabric and values
they carry. So is expertise and craftsmanship, from
consultant to craftsmen.
REFERENCES
Caston, S.C. Ph. 1998. Dachwerkstopographie für die Steier-
mark. Ein Survey der historischen Dachwerkskonstruktio-
nen in der Mur-Mürz-Furche. Österreichische Zeitschrift
f-r Kunst und Denkmalpflege (Heft 3/4):518–536. Wien.
Hansmann, W. 2000. Barock und Rokoko: 25. Köln: DuMont.
Makay, D. 2006. The Safety of Historic Roof Structures with
Baroque Character. Historic Structures – 10th Scientific
Conference:. 136–152. Cluj-N: Utilitas.
Szabó, B. 2005. Illustrated dictionary of historic load-bearing
structures. Cluj-N: Utilitas.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

The Athens Parthenon: Analysis and interpretation of the structural

failures in the orthostate of the northern wall

E.E.Toumbakari
Directorate for the Conservation of Ancient Monuments, Ministry of Culture, Athens, Greece

ABSTRACT: An investigation on the typology of the failures recorded on the first row (orthostate) of the
Parthenon northern wall was undertaken. The analysis of the failure causes was based on the simultaneous
consideration of (a) the kinematic constraints of each block, (b) the applied system of deformations/displacements
on each block and (c) the applied system of deformations/displacements on the wall itself. Moreover, a numerical
analysis was also undertaken, and characteristic results are also reported. The effect of the marble anisotropy
(in terms of the position of the marble soft plane inside the block) was also studied. The main conclusion of the
study is that the mechanical action of the connectors could explain the observed structural pathology, whereas
rust is not a necessary condition for the structural failures to occur.

1 INTRODUCTION

The longitudinal walls of the Parthenon are composed

of 19 rows of marble blocks. The first bottom row,
called orthostate, is composed of blocks with average
dimensions 1.44 m length, 0.55 m depth and 1.15 m
height. The last row is composed of architrave with
average dimensions 2.44 m length, 0.5 m depth and
1.04 m height. The intermediate 17 rows are built with
smaller marble blocks (length 1.22 m, height 0.52 m).
The even-numbered rows are composed of two lines
of blocks, disposed with their long axis parallel to the
axis of the wall. Many blocks of the external line, fac-
ing north, are conserved, whereas the internal ones,
facing south, are lost. In the odd-numbered rows the
blocks are placed perpendicular to those of the even-
numbered rows (Fig. 1). The blocks are horizontally
connected with clamps and vertically with dowels
(Fig. 2). The main actions that affected the mechan-
ical history of the walls of the Parthenon are the fire
of 267 A.D. and the explosion of 1678 A.D. during the
siege of Athens by the Venetians. The latter caused the
collapse of the larger part of the walls. Figure 1. Structure of the walls of the cella (Orlandos 1977).
Previous investigators on the state of preservation
of the Parthenon (Korres & Bouras 1983) argue that 2 POSITION OF THE PROBLEM
the cross section of the connectors is designed so
as to have lower strength than the anchorage area. The present work summarizes the main issues of the
So failure of the connectors and not of the marble study for the structural restoration of the orthostate of
itself is expected to occur. If, however, marble failure the N.Wall, which includes the description and docu-
occasionally occurred, it must be preceded by marble mentation of the condition of the marble blocks, the
cracking due to rust and volume increase of the iron analysis & interpretation of their structural pathology
connectors. (Toumbakari 2006) together with a numerical analysis,

673
 3 RECENT OBSERVATIONS and DATA

The extent of damage reasonably permits the assump-

tion that, all those failures cannot be solely attributed
to rust. The undertaken investigation aimed to find
out whether mechanical action of the connectors
could as well generate marble failure, without rust as
prerequisite.
Before proceeding to the analysis of the failure
modes and to the numerical analysis, the following
observations need to be reported:

1. The contours and volume of the detached marble

Figure 2. Connectors at the orthostate level (Orlandos presents similarities for all blocks. This excludes
1977). There are also two dowels on the orthostate upper human intervention as a cause for fracture, because
surface, which are not shown. human intervention would have produced a variety
Table 1. Location of the marble fractures. of fracture configurations.
2. Failure concerns practically all connections of all
Dowels Dowels the available blocks. A similar case could be the
upper bedding Opisthodomos (western) area, in which original
Clamps surface surface Outside clamps and dowels of the architraves were uncov-
area of ered during the 2001–2004 restoration campaign.
No E W E W E W connectors Our assessment of their preservation, is that they
behaved very well, despite the observation of super-
1 n.e. n.e. + + 1st : n.f. + n.f.
2nd : +
ficial rust (Archives 2007, Toumbakari 2007a).
2 + n.e. + + + + + Moreover, marble failure was concentrated in areas
3 + + + + + + + with residual deformations. It is reasonable to
4 + + + + + + (C) assume that the quality of the iron of the orthostate
5 + + + + + + n.f. connectors is similar to the aforementioned. So, if
10 + + + + + + + rust cannot be excluded, it is difficult to argue that
11 + + + + + + n.f. it produced such extended mechanical effects only
12 + + + + + + + at the orthostate.
13 + + + + + + n.f. 3. The analysis (Papadimitriou et al. 2007) of the
14 + + + + + + +C
15 + + + + + + n.f.
unique ancient connector recently found in the
16 + + + + + + n.f. Northern Colonnade showed that the clamp and
lead were conserved in a good condition, rust has
Subtotal 11 10 12 12 12 12 5
Total 69 5
affected the external 2 mm of the iron whereas
the anchorage area was in an excellent state of
n.e.: not clear because of detachment of big marble volume. preservation (Toumbakari 2007b).
n.f.: no failure C: cracking +: failure. 4. In addition, rust usually creates a one-sided detach-
ment, which then provides sufficient space for the
whose main results are also discussed (Toumbakari connector to expand. Thus, marble detachment on
2007a; Filippoupolitis 2007). the other side of the connector is not necessary. The
The location of the failures at the orthostate blocks orthostate clamps’ anchorage areas are practically
is summarized in Table 1. Failure is defined as marble all characterized by detachment on both sides of the
fracture and detachment. Usually, the detached frag- connector.
ment is lost. Among the 74 failures of the orthostate 5. Available experimental results relative to the design
blocks, 69 are located at the area of the connectors and of new dowels have shown that failure always
only 5 (i.e. 7.2%) are located in areas not affected by occurs in the marble and not in the connector (Zam-
the connectors’ action. Moreover, the external (north- bas 1989). On the contrary, when new clamps (with
ern) orthostate blocks have not been affected by fire. dimensions close to the ancient ones’) were sub-
Therefore, their structural pathology can be attributed jected to tension, it was the connector that failed
to a limited set of factors, namely the explosion, the (Zambas 1994). To the best of our knowledge, there
rust and eventually earthquakes that have occurred are not available results with clamps subjected to
prior to 1687. If however the connectors were designed shear.
to fail prior to the marble, then why is the majority of 6. The failure configuration is similar to failure pat-
the observed failures located at the anchorage area? terns, already well-known in the literature (CEB

674
 1994), attributable to dowel and anchor mechanical structural elements, instrumental (or not) follow up of
action. the behaviour of the structure and structural analysis”
(ICOMOS 2004).
On the basis of the above, the following hypoth-
The direct observation of the structural pathology
esis was developed: The concentration of failures at
focused on the analysis of the type of action (tension,
the area of the connectors could be attributed to the
shear. . .) that could have produced it. It consisted in
mechanical action of the clamps and dowels, when
the definition of failure modes on the basis of similar
the bearing structure (either entablature or wall) was
and repetitive failure patterns and in the search for a
subjected to strong external (horizontal) actions.
system of deformations/ displacements at the block
Rust could have, and certainly has, occurred, but
level, which could have generated them. This sys-
it is not considered sufficient for the generation of
tem had to be compatible to the kinematic constraints
such large number of failures. Moreover, rust is not
imposed by the connectors as well as to the defor-
the unique condition for the generation of small
mations/ displacements that have been applied on the
cracks, which will later initiate fracture. Those cracks
walls. The latter, subjected to horizontal actions, could
can also be generated if the marble blocks are sub-
be described as plates with partial constraints at three
jected to tension-inducing actions, such as temperature
edges and free at the upper edge, because the roof had
changes, environmental actions and vibrations due to
already collapsed before the explosion. Consequently,
dynamic loading. Recent experimental work on mar-
three criteria for the analysis and interpretation of the
ble properties and behaviour (Vardoulakis et al. 2002)
structural pathology were defined: (a) kinematic con-
established the existence of residual deformations on
straints applied on the building block, (b) system of
marble specimen subjected to tensile load as well as
deformations/displacements applied on the block and
the non-linear behaviour of the material even at an
(c) system of deformations/displacements applied on
early stage of loading. In the study, no explanation
the wall. These criteria had to produce a system of
as to the origin of those deformations is provided.
forces, compatible to the observed pathology. During
It can, however, be assumed that they could be the
the study, it was observed that the failure configura-
result of the opening of small cracks at surfaces con-
tion also depends on the local system of argillosilicic
taining argillosilicic veins, as is empirically known to
veins as well as on the direction of the marble soft plane
practitioners.These cracks can conceptually be assimi-
inside the block. The first do not affect the failure mode
lated to “notches”, which will facilitate (but, of course,
but do affect the precise shape of the failure surface
not generate) crack propagation under the mechanical
and volume of the detached fragment. The latter permit
action of the connectors.
or hinder the development of the one or other failure
Concerning the connectors’mechanical action, it is
mode. The qualitative approach permitted the determi-
important to estimate if the developing pressure is
nation of the type of actions that were applied by the
entirely transferred on the marble walls. It is recalled
clamps and dowels on the anchorage area. It is however
that the void between the connector and the marble was
not sufficient because, if it highlights the mechanisms
originally filled with lead, which, in unconfined condi-
of failure, it still cannot differentiate between the pri-
tions, is an energy-absorbing and deformable material.
mary cause of the failures, namely mechanical action
In the case of ancient connectors, however, lead is in
or rust.
a situation of triaxial confinement. It has to be admit-
A numerical analysis was therefore necessary in
ted that this confinement might not be perfect due to
order to investigate if it is possible for the failures to
constructional imperfections, nevertheless it is certain
occur under the connectors’mechanical action. Exper-
that the deformability of the lead is limited, especially
iments are very important in this respect, and if they
in a horizontal direction, where the filling of the void
are available in the case of dowels (Zambas 1989),
is complete. Thus, doubts are cast about the deforma-
they still need to be completed in the case of clamps
bility of the system under confined conditions. Further
subjected to shear. Two 3D models were created with
experimental work is nevertheless necessary to clarify
the use of solid Finite Elements and the Sofistik
this issue.
code (Toumbakari 2007a, reworked by Filippoupolitis,
2007). They represent marble blocks with length and
4 METHODOLOGICAL APPROACH height equal to 0.6 m and depth equal to 0.5 m (which is
the average depth of the Opisthodomos architraves and
The analysis followed a dual approach, qualitative and orthostate blocks). The choice of these dimensions was
quantitative (ICOMOS 2004, Tassios 2006). The qual- justified by previous numerical analyses (Toumbakari
itative approach is defined as “an approach based 2007a), in which it was shown that the effect of the con-
on the direct observation of the structural pathology nectors does not overpass a distance of 0.25–0.30 m.
and material deterioration as well as on historic and The models are shown in Figures 13 and 14. The
archaeological research”. The quantitative approach load cases and boundary conditions are described in
is based on “the properties of the materials and Tables 2 and 3. In what follows, only the analyses of

675
Table 2. Load cases and boundary conditions of the con-
nectors’ models (s.w. = self weight).

Load
cases Clamp∗

LC 1 s.w. + vert. loads + shear force 1 kN

LC 2 s.w. + vert. loads + shear force 15 kN
LC 3 s.w. + vert. loads + shear force 29 kN
LC 4 s.w. + vert. loads + shear force 1 kN + tens.force
1 kN
LC 5 s.w. + vert. loads + shear force 15 kN + tens.force
15 kN
LC 6 s.w. + vert. loads + shear force 15 kN + tens.force
30 kN
LC 7 s.w. + vert. loads + shear force 29 kN + tens.force
30 kN
∗
Boundary conditions at surfaces which are part of the mar-
ble block: constraint of displacements & rotations in all Figure 3. Initial (dotted lines) and deformed position of the
directions. orthostate block, on which the area affected by the clamps’
pressure is highlighted.
Table 3. Load cases and boundary conditions of the con-
nectors’ models (s.w. = self weight).

Load
cases Dowel∗

LC 1 s.w. + vert. loads + shear force 1 kN (uniform

distr/tion)
LC 2 s.w. + vert. loads + shear force 5 kN (uniform
distr/tion)
LC 3 s.w. + vert. loads + shear force 14 kN (uniform
dist/ion)
LC 4 s.w. + vert. loads + shear force 1 kN (triang.
distr/tion)
LC 5 s.w. + vert. loads + shear force 5 kN (triang.
distr/tion) Figure 4. First detachment due to the clamps’ shear force.
LC 6 s.w. + vert. loads + shear force 14 kN (triang.
distr/tion)
and deformed shape of two adjacent orthostate blocks
∗ under the effect of a horizontal action are presented.
Boundary conditions at surfaces which are part of the marble
block: constraint of displacements & rotations in all direc- The system of clamps and lead (hereafter: clamps)
tions.At bedding surface: constraint for vertical displacement initially resists the applied deformation. The role of
only. lead is crucial in force transfer. Because the clamps
tend to conserve their shape, they exercise pressure on
Failure Modes I and II, which correspond to the afore- the marble with a direction opposite to the external
mentioned models, will be developed. The remaining action (grey area in Fig. 3). This could lead to a first
three failure modes will be summarily described in detachment (Fig. 4), if the shear force applied by the
this work. connector is higher than the resistance of the marble.
As the clamp is now free from the constraint imposed
by the marble wall (which detached), it is obliged to
5 FAILURE MODES OF THE ORTHOSTATE
bend, in order to follow the deformation of the wall.
BLOCKS
Inversely, the wall deformation is possible either if the
clamp is able to deform in flexion or if the marble
5.1 Failure mode I
fails. The ensuing clamp flexural deformation results
This failure mode is related to the clamps, which are in the development of pressure on the remaining mar-
located at the upper surface of the orthostate blocks ble wall. As in the first step, failure mode I occurs, if
(Fig. 3). Failure consists in the detachment of marble the applied force is higher than the shear resistance of
fragments on both, in most cases, sides of the clamps the marble. The configuration of the clamp’s area after
(Figs 5, 6). In Figure 3 both the initial (in dotted lines) complete failure is shown in Figures 5 and 6.

676

Figure 5. Typical configuration of Failure Mode I.
1. Position of
clamp
2. Missing
northern
fragment
3. Detached
and repaired
southern fragment
Figure 6. Failure of orthostate block 5 with Failure Mode I.
Alternatively, failure mode I can also occur with
simple relative horizontal displacement between the
marble blocks without rotation. In this case, a one-
sided marble detachment can occur.
Figure 6 illustrates orthostate block No 5 of the
northern wall, in which failure mode I occured. It can
be seen that, detachment on both sides of the clamp
has occurred. The northern fragment is lost, whereas
the southern was found and repaired. The shape of the
677
Figure 7. Initial position (dotted lines) of one block and
deformed position of two adjacent blocks. The position of
the southern (rear) dowel is shown on the left block, whereas
the northern (front) dowel is shown on the right block.
latter is clear, as it is not affected by the presence of
other connectors. On the contrary, in the northern side,
detachment occured at the whole marble height. This
could be explained by the simultaneous occurence of
two failure modes, I and III, due to the failure of the
adjacent (to the clamp) dowel of the upper surface.
5.2 Failure mode II
Typically, the orthostate blocks possess two dowels at
their bedding surface. They are located close to the
northern (front) and southern (rear) surfaces of the
block. Failure mode II is characterised by detachment
of the marble cover of each dowel either at a limited
height (Figs 9 , 10) or at the whole height . In this case,
failure occurs probably due to the synergistic effect of
the action of the dowels with the connectors of the
upper surface (Failure Modes I and II).
The mechanism of failure could be described as
follows: with the application of a horizontal force
(or displacement) towards the north, rotation between
adjacent blocks with or without relative displacement
takes place (Fig. 7). Both dowels of an orthostate
block develop resistance to this movement. The dowel
located close to the southern marble surface fails eas-
ily because it has a small cover against the direction of
the applied force or displacement (Fig. 8). The failure
of the dowels closer to the northern marble surface is
more difficult to interpret because it is not compatible
(it is actually opposite) to the direction of the reaction
force of the dowel.
It can, however, be assumed that the kinematic
constraints (imposed by the rear dowel of the block)
are already released, following the mechanism previ-
ously described. The constraints of the block now are:
(a) on the one side still an active clamp but not dowel,
whereas, (b) on the other side, both a clamp and a
dowel still active (Fig. 8, right block). To permit fur-
ther displacement, it is necessary that the cover of the
Figure 8. Detachment of the cover of the southern (rear) Figure 10. Failure of the northern (front) dowel.
dowel.

dowel

Figure 9. Failure of the cover of the southern (rear) dowel

of orthostate block 12.

clamp, which is located at the side of the rear dowel

(not shown in Figure 8), has failed following Mode I.
The release of both constraints on the one side of the
block result in rotation of the block around the axis of
the still active clamp and dowel of the other side. If,
failure of both clamps occurs, then rotation can occur
around the remaining dowel. In both cases rotation
could result in marble detachment towards the north
(Figure 10).
5.3 Failure mode III
Figure 11. Failure mode III of the northern dowel of the
upper surface of orthostate block 10.
the orthostate and the corresponding block of the row
above. This requirement is compatible with the defor-
mation of the whole wall as a plate. Its analysis will
be presented in another work.

Failure mode III is observed with the two dowels

of the upper surface of the orthostate. It consists in
the detachment of a marble cone towards the north
(i.e. towards the direction of the applied force or
displacement) (Fig. 11).
For this failure to occur, it is necessary that a rel-
ative displacement and rotation takes place between
5.4 Failure mode IV
This failure mode is characterized by fracture of the
block and separation in two fragments. This frac-
ture theoretically occurs in areas not influenced by
the connectors and is attributed to out- of-plane bend-
ing of the block. The position of the marble soft plane

678
 Table 5. Failure modes of the blocks of the even-numbered
rows of the Parthenon northern wall.
upper surface Failure mode
Row Surveyed
number blocks I II III IV

2nd 9 17/17 5/11 3/9 1/9

4th 7 14/14 4/9 1/7 0/7
northern (front) elevation 6th 6 7/12 2/6 0/6 0/6
8th 10 19(?)/20 1/8 0/10 0/10
Figure 12. Section of the orthostate soft plane with the 10th 10 19/19 4/10 1/10 1/10
northern elevation. Left: block 3 (vertical), right: block 4 12η 11 18/18 2/7 0/11 2/11
(horizontal). 14th 10 20/20 0/7 0/10 0/10
16th 11 20/21 1/8 0/9 3/11
Table 4. Position of marble soft plane and failure modes of Total 74 134/141 19/66 5/72 7/74
the orthostate of the Parthenon northern wall blocks
18th (archi- 4 7/7 not sur- 6/8∗ 1/4
Direction of soft plane Failure mode trave) veyed.
Section Total 78 141/148 19/66 11/80 8/78
In relation with ∗
to bedding northern The two dowel areas that did not fail, show extended
No surface elevation I II III IV cracking.

1 vert/incl1 horizontal yes yes yes no2

5.5 Failure mode V
2 vert/incl vertical yes yes yes yes
3 vert/incl vertical yes yes yes yes This failure mode is characterized by local loss of
4 vert/incl vertical yes yes yes no3 material, roughly with triangular cross-section, at the
5 vert/incl horizontal yes yes yes no front side of some orthostate bedding surfaces. It is
10 vert/incl vertical yes yes yes yes
attributed to shear, as it occurred to blocks which either
11 vert/incl horizontal yes yes yes no
12 vert/incl horizontal yes yes yes yes rotated strongly or collapsed. It is observed only in
13 vert/incl horizontal yes yes yes no the well-carved front (northern) side of the orthostate
14 vert/incl vertical yes yes yes yes4 blocks and not on the rear (southern) side, which is
15 vert/incl horizontal yes yes yes no more roughly carved.
16 vert/incl horizontal yes yes yes no
6 ON THE EFFECT OF THE POSITION OF THE
1
inclined (less than approx. 30◦ ) in relation to a vertical MARBLE SOFT PLANE
plane.
2
in conjunction to the eastern wall of the cella.
3
not fracture but cracking has occurred. Two ways of dressing the orthostate soft plane inside
4
in conjunction to the middle wall of the cella). the orthostate blocks were recorded and are described
in Table 4 and Figure 12. The investigation of the effect
of the position of the marble soft plane on the structural
pathology was carried out through the comparison of
inside the block is of fundamental importance. Indeed,
the failures recorded on the orthostate (whose soft
failure mode IV was observed only in cases, where
plane is roughly perpendicular to the bedding sur-
the soft plane direction crossed vertically with the
face, usually with inclination) (Table 3) and the marble
northern elevation, as it will be developed hereafter.
blocks of the even-numbered rows whose soft plane is
(Table 4, Fig. 12). It occurs either when not all con-
roughly parallel to the bedding surface, usually with
nectors have failed or if there are additional kinematic
inclination) (Table 5).
constraints, such as the presence of transverse walls. In
On the basis of the comparison between Tables 3
practice, even this failure mode is not always indepen-
and 5, the following conclusions may be drawn:
dent to the connectors. With the exception of orthostate
block 2 (Table 4), on the other blocks, in which this 1. The effect of the direction of the soft plane appears
failure mode occurred (blocks 3, 10, possibly 14), it clearly in failure mode III. It was observed in most
seems that the presence of weak planes close to the dowel areas of the orthostate and architrave beams.
connector area resulted in fracture which might initi- On the contrary, only 5 (out of 72) failures were
ate from the connector area but clearly continues inside recorded on the blocks of the intermediate 17 rows.
the block. This failure cannot therefore be classified 2. Similar observations are valid for failure mode II.
under modes I, II or III. If it appears practically in all orthostate dowels

679
 (a survey of the wall architraves was not possible
at the time of the study), it concerns roughly only
the 1/3 of the intermediate blocks.
3. Failure mode IV is rare in all cases. In the orthostate,
it was attributed to a specific direction of the mar-
ble weak plane inside the block (Table 4). It would
be of interest to analyze under which conditions it
appeared in the wall intermediate blocks.
4. Finally, the effect of the weak plane direction does
not only affect the possibility (or not) of generation
of a specific failure mode but its configuration as
well. Indeed, if failure mode I is recorded in prac-
tically all areas of the clamps of the orthostate, Figure 13. Model of the.
architrave and intermediate blocks, the height of
the detached fragment is significantly reduced in
the last case (the intermediate blocks).

7 RESULTS OF THE NUMERICAL

ANALYSIS

Aims of the numerical analysis were (a) the study of

failure modes I and II and (b) the determination of the
maximum compressive, tensile and shear stresses that
develop under the actions described in Tables 2 and 3.
The models are shown in Figures 13 & 14. The clamp
tensile strength was chosen equal to 55 MPa, the clamp
shear strength 29 MPa and the dowel shear strength
14 MPa (Zambas 1994, Vardoulakis et al. 2002). In
the case of the dowel analysis, two force distributions
on the marble (applied as surface distributed loads and Figure 14. Model of the.
automatically converted to loads on the corresponding
nodes) were selected, namely uniform and triangular. Table 6. Main results of the clamp analysis.
The latter results in higher corresponding stresses σx
and σy . σx [MPa] σy [Mpa] τxy [Mpa]
In the case of the clamps, in accordance to the con-
clusions of the failure modes’ analysis, mainly shear LC σc,max σt,max σc,max σt,max σc,max σt,max
(and not tensile) forces were considered. It is however
recognized that, in rather extreme situations (such as 1 −0.15 0.37 −0.10 0.35 −0.02 0.06
the explosion), significant tension can also develop 2 −2.28 4.79 −0.37 1.75 −0.17 0.85
3 −4.41 9.22 −0.74 3.37 −0.32 1.64
parallel to shear, therefore some more complex stress
4 −0.15 0.43 −0.17 0.36 −0.04 0.07
situations were also considered. 5 −2.29 5.73 −2.56 3.29 −0.69 1.11
In terms of resistances, it is well-known that mar- 6 −2.29 6.66 −5.14 6.66 −1.56 1.86
ble tensile strength values available in the literature 7 −4.42 11.08 −5.13 6.60 −1.38 2.18
present important scattering, attributable to differ-
ent test procedures as well as to marble anisotropy
(Vardoulakis et al. 2002). The values provided by Table 7. Main results of the dowel analysis.
Vardoulakis et al. (2002) were used. Precisely, the
tensile strength in the two strong marble directions σx [MPa] σy [Mpa] τxy [Mpa]
was found equal to 9.5 and 10.8 MPa respectively,
whereas the tensile strength of the marble soft plane LC σc,max σt,max σc,max σt,max σc,max σt,max
was found equal to 5.3 MPa. Consequently, tensile
stresses around 5 MPa were considered sufficient to 1 −0.21 0.37 −0.05 0.20 −0.03 0.13
2 −1.02 1.84 −0.27 0.99 −0.13 0.68
generate cracks inside the marble mass. 3 −2.87 5.14 −0.77 2.76 −0.37 1.90
The main results of the analysis of the connectors 4 −0.33 0.50 −0.05 0.23 −0.02 0.12
(Toumbakari 2007a, Filippoupolitis 2007) are summa- 5 −1.66 2.51 −0.23 1.18 −0.09 0.61
rized in Tables 6 & 7. In the case of the clamp, the 6 −4.66 7.03 −0.64 3.30 −0.27 1.71
application of a shear force equal to 15 kN (LC2) at

680
the x-direction already results in the development of a
tensile stress equal to 4.79 MPa, which is considered
sufficient to initiate cracking. A shear force equal to
the clamp shear resistance (LC3) generates a tensile
stress equal to 9.22 MPa, which is similar to the mar-
ble tensile strength in the second strong direction. The
more complex situations considered in LC5 through 7
result in the development of tensile stresses from 5 to
11 MPa in both x- and y-directions, which show that
crack generation is possible. In a similar way, in limit
state situations (LC3, LC6) the developing stresses at
the dowel area are sufficiently high to produce crack-
ing (5.14 and 7.03 MPa). These numerical results are
compatible to available experimental results on dow-
els (Zambas 1994), which have shown failure of the
dowel anchorage area and not the dowel itself.
8 CONCLUSIONS
On the basis of the analysis of the structural failures
in the orthostate of the Parthenon northern wall, the
following conclusions were drawn:
1. A methodology for the analysis and interpretation
of the structural damages was proposed, based on
the compatibility between (a) kinematic constraints
applied on the building block, (b) system of defor-
mations/displacements applied on the block and
(c) system of deformations/displacements applied
on the wall.
2. Five failure modes were recognized: (I) clamp area
failure attributable to shear, (II) dowel area failure
attributable to tension and shear, (III) dowel area
failure attributable to tension, (IV) block separation
attributable to out-of-plane bending and (V) mass
detachment at the base of the blocks, attributable to
shear with, occasionnaly, local crushing.
3. The configuration of the failure of the area of the
clamps shows that the latter work mainly in shear
and not in tension.
4. The position of the marble soft plane inside the
block was found to significantly affect the genera-
tion of the failure modes.
5. The mechanical action of the connectors could
explain the observed structural pathology. Rust is
not a necessary condition for the structural failures
to occur.
681
REFERENCES
1. Archives of the Office for Conservation of the Acropolis
Monuments, 2007 (Photographic documentation).
2. CEB (Comité Euro-international du béton) 1994. Fas-
tenings to concrete and masonry structures – State-
of-the-Art Report, Eligehausen R. et al. eds, Thomas
Telford.
3. Filippoupolitis, M. 2007. Analysis of the developing
stresses at the architraves of ancient temples under vari-
ous stress conditions – application at the Opisthodomos
of the Parthenon, Diploma Thesis in progress, under the
direction of E.-E.Toumbakari (Prof. F.Karantoni): Univ.
of Patras.
4. ICOMOS-International Scientific Committee for Anal-
ysis and Restoration of Structures of Architectural
Heritage 2004. Recommendations for the analysis, con-
servation and structural restoration of the architectural
heritage (draft).
5. Korres, M. & Bouras, Ch. 1983. Study for the restoration
of the Parthenon, vol.1. Athens: Min. of Culture.
6. Orlandos, A. 1977. The Architecture of the Parthenon.
Athens: Athens Archaeological Society, 2 vol.
7. Papadimitriou, G., Tsakiridis, P. & Skarvelis, P. 2007.
Laboratory examination of one iron clamp from block
14048 of the N.Colonnade, Techn. Report: Lab. of
Physical Metallurgy, School of Metallurgical Engg,
NTUA.
8. Tassios, T.P. 2006. Structural restoration of Monuments:
Assessment of Values, Athens: Lychnos.
9. Toumbakari, E.-E. 2006. Study for the Structural
Restoration of the Orthostates of the Northern Wall, 2
vols. Athens: Office for Conservation of the Acropolis
Monuments.
10. Toumbakari, E.-E. 2007a. Study for the Structural
Restoration of the Opisthodomos (2001–2004), 6 vols
(in progress). Athens: Office for Conservation of the
Acropolis Monuments.
11. Toumbakari, E.-E. 2007b. Study for the Structural
Restoration of the upper two rows of the entablature
of the Parthenon Northern Colonnade, 3 vols. Athens:
Office for Conservation of the Acropolis Monuments.
12. Zambas, K. 1989. Study for the restoration of the
Parthenon, vol.2b. Athens: Min. of Culture.
13. Zambas, K. 1994. Study for the restoration of the
Parthenon, vol.3b. Athens: Min. of Culture.
14. Vardoulakis, I., Kourkoulis, S., Exadaktylos, G. &
Rozakis, A. 2002. Mechanical properties and compati-
bility of natural building stones in ancient monuments:
Dionysos marble. In M.Varti-Mataranga & G.Katsikis
(eds), Building Stones in Monuments; Proc. Intern.
Conf., Athens 2002. Athens: 9 Nov. 2001: 187–210.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Structural assessment of the Winter Garden of the Royal Glasshouses

of Laeken, Belgium

L. Lauriks, I. Wouters, S. Adriaenssens, M. de Bouw & B. Verbeeck

Vrije Universiteit Brussel, Brussels, Belgium

ABSTRACT: The Winter Garden, the largest of the Royal Glasshouses, is situated at the Royal Domain in
Laeken in the northeast of Brussels, Belgium. This glasshouse was built between 1874 and 1876 in order of
king Leopold II, according to the design of architect Alphonse Balat. A two-dimensional analysis demonstrated
the basis structural behaviour under a series of symmetrical load combinations: the structure works as a cupola
with a tension and a pressure ring. The stress levels and deflections in the structure have to be evaluated in a
three-dimensional model which is still in progress. For the time, it seems that the norm values are not exceeded.
The major threat for the metal structure is corrosion. Previous interventions and a very aggressive tropical indoor
climate inflicted heavy corrosion damage. Nowadays, this elaborate plain tour de force must undergo a major
restoration.

1 BUILDING THE WINTER GARDEN
The typology of glasshouse buildings originates from
the desire for a peaceful and green neighbourhood
in the overpopulated cities during the Industrial
Revolution. The first iron glasshouse is the iron hot-
house in Hohenheim near Stuttgart (Germany) dating
back to 1789 (Kohlmaier & Von Sartory 1991). The
evolution of glass and iron production techniques in
the nineteenth century was essential for the develop-
ment and spread of the glasshouses.
The Winter Garden of Laeken is part of a major
complex of glasshouses in the Royal Domain in the
northeast of Brussels, Belgium (Figure 1). The com-
plete set of glasshouses takes up an area of 1.5 ha,
covered with 2.5 ha of glass (Goedleven 1988). All
entities are built between 1817 and 1905 in order of
king Leopold II, who reigned the country from 1865
until his death in 1909. He is known as the king who
fundamentally changed Belgium and Brussels in an
architectural as well as in an urban development man-
ner. Despite the fact that Belgium is a small country,
he wanted it to radiate grand charisma. He used his
personal influence, private funds and the profits from
the Congo colony to realize many of his ideas.
Alphonse Balat (1818–1895) was the royal architect
from Leopold’s accession until his death. He was one
of the architects who lead in the Belgian Art Nouveau
movement. Besides the well-known Museum for Old
Arts in Brussels and the Royal Palace in Laeken, he
also designed most of the glasshouses of the Royal
Domain, including the Winter Garden (Figure 1, no. 3).
The Winter Garden (1874–1876) is a significant
cultural legacy and an important artefact of the evo-
lution of structural steelwork and Art Nouveau in
Belgium. It is the first glasshouse on the Royal Domain
with a complete iron and glass covering. The Winter
Garden was meant to be a glasshouse for social events
and to this very day, it still performs that task.
For the design of the Winter Garden,Alphonse Balat
referred to the Palm House (Kew Gardens, London,
1844–1848) designed by Decimus Burton and Richard
Turner. Balat used the structural concept of the Palm
House for his Winter Garden and applied this to a
circular ground plan. The design of the Palm House
in its turn was inspired by the Great Conservatory
(1836–1840) of Joseph Paxton which was unfortu-
nately demolished in 1920. This influence is clearly
shown in Figure 2.
The structure of the Winter Garden consists of
36 arch trusses which are rotated around a central
point and form one large dome (Figure 3). The Winter
Garden can be divided into two main parts (Figure 4).
The first part consists of the dome in the middle of the
glasshouse, which is topped off with a small cupola and
a royal crown. The second part consists of a side aisle
around the middle dome. These two parts are separated
by a circular architrave on sandstone columns.

683
 As people would intuitively feel that a cupola geom-
etry (part one as described in the previous paragraph)
produces lateral thrusts,Alphonse Balat added an outer
arch (part two) to each truss of the cupola. By doing so,
he suggests the structure being composed by arches –
structures that can withstand thrusts – rather then a
cupola. The structural surplus value of the outer arches
is still being investigated. However, it is generally felt
that this (visual) addition results in a harmonious and
aesthetically more beautiful construction.

Figure 2. Historical influences in the design of the Winter

Garden (Kohlmaier & Von Sartory 1991).

Figure 1. Map of the Royal Glasshouses of Laeken, Brussels Figure 3. Building the Winter Garden, December 30th 1875
(Koppelkamm 1988; translated by the author). (Prentenkabinet of the Royal Library, s III 100752).

684
2 THE WINTER GARDEN TODAY
2.1 Maintenance of the structure
When king Leopold II passed away in 1909, the own-
ership of the Royal Glasshouses passed on to various
institutions. The land, purchased by Leopold II, and the
glasshouses constructed on that land, became prop-
erty of the Royal Donation, a semi-private institute.
The other glasshouses, situated on the piece of land
that he received to fulfil his duties as a king, are
owned by the Regie der Gebouwen, a governmental
institution. The borderline between these two adjacent
properties can be diagonally traced through the Congo
House (Figure 1, no. 4). The shared ownership of the
glasshouses does not facilitate restoration procedures.
Since the death of Leopold II, few of the Royal
Glasshouses were restored. The Winter Garden merely
received some refurbishments in the 1980’s, funded by
the Regie der Gebouwen.
Some cross-sections were reinforced by welding
new parts on to the old ones. However, the weldability
of the initial material was never investigated and there-
fore the durable efficacy of this intervention remains
questionable. Heavily corroded profiles were replaced,
yet without recording these changes in an appropri-
ate report. The whole structure was sand-blasted and
Figure 4. The structural parts of the Winter Garden (Vieren-
deel 1902; notes by the author).
Figure 5. a) corroded metal profiles (2006); b) welded plates on metal column (2007); c) missing rivets (2007); d) removing
the paint and the corrosion (2006).
685
repainted. The glass covering exists of many small flat
rectangular single glass panels. To reduce the heating
costs, all these panels were replaced by coated glass
jointed with mastic to improve the insulation qualities.
For the sake of the well-being of the perennial
plants and trees and for the visitors’ safety, restora-
tion works could not be scheduled during winter nor
during the two weeks of public opening in spring.
Therefore, all work was carried out in different stages,
which considerably complicated the work, follow up
and coordination.
2.2 Pathology of the metal structure
The metal structure of the Winter Garden is mostly
unchanged, compared to the original construction dat-
ing back to 1876. Therefore, great historical value is
attached to this building and any future restorations
have to be done with the greatest respect to the original
structure.
The major threat for the metal structure is corro-
sion. By filling the joints between the glass plates
with mastic, the ventilation of the inside air has
been dramatically reduced. This well intended inter-
vention resulted in more condensation on both the
glass and the metal structure. Together with the very
aggressive tropical indoor climate, this has lead to
increased profile corrosion on some vulnerable spots
(Figure 5-a).
During previous interventions, a series of profile
cross-sections were reinforced by welding new parts
on to the old ones, e.g. on the iron column between
the side aisle and the middle dome (Figure 5-b). The
tension ring at the bottom of this column was cut off
from the column, steel plates were welded on the col-
umn profiles and the tension ring was welded on the
added plates. The tension capacity of these welds is
unknown, as the weldability of the original iron was
never determined.
Some profile connections show missing rivets
(Figure 5-c). These connections have to be repaired,
so loads can be transferred properly.
During previous restoration works, the whole struc-
ture was sand-blasted and repainted. Nowadays, the
Figure 6. Compound profiles of the arches (2007).
Figure 7. Three-dimensional structure.
Regie der Gebouwen did again some tests with special
techniques to remove the paint and corrosion (Figure
5-d). The techniques were evaluated on the time it took
to rub a specific profile, the grade of efficiency of the
rubbing technique and the nuisance for the tropical
plants inside the glasshouse. The most efficient tech-
nique appeared to be a hammering technique where
needles remove paint and corrosion from the iron.
After cleaning the profiles, they need to be repainted
as soon as possible to avoid new corrosion.
3 STRUCTURAL ANALYSIS OF THE WINTER
GARDEN
The three-dimensional structure of the Winter Garden
is quite refined. Every single member of the metal
load-bearing structure consists of richly ornamented
compound profiles (Figures 6 & 7).
In order to understand the basic structural behaviour
of the Winter Garden under symmetric load combina-
tions, a two-dimensional model was investigated. The
historical reference to the Palm House in Kew and
the Great Conservatory in Chatsworth implies that the
structure works as a dome with a pressure ring at the
top and a tension ring at the bottom of the cupola.
686
Figure 8. Two-dimensional calculation model.
If this is the case, the outer arch of the aisle part will
merely carry minor loads and fulfils purely an aes-
thetic function. This hypothesis will be verified in the
two-dimensional structural model.
This two-dimensional model has its limits. In real-
ity, every load on one of the 36 arches is distributed
to the other arches via horizontal concentric rings as
well. As a consequence, all the arches work together
to bear the loads, especially asymmetric loads. In the
two-dimensional model, this effect is neglected for the
greater part. However, given the specific construction
and functioning of the metal structure of the Winter
Garden, it is indispensable to determine the stresses
and deflections in a three-dimensional finite elements
model.
The two-dimensional calculations were performed
on two models: one of the inner arch of the side aisle
and one of the arch of the middle dome together with
the outer arch of the side aisle and the column between
these two. In the model, the support conditions of the
structure are modelled as follows:
– the support at the bottom of the side aisle is fixed
(Figure 8-v);
– the support at the sandstone columns is fixed
(Figure 8-w);
– the connection of the outer arch of the side aisle and
the arch of the middle dome is fixed (Figure 8-y1 );
– the connection of the column and the arch of the
middle dome is hinged (Figure 8-y1 );
– the connection of the inner arch of the side aisle
and the column is hinged (Figure 8-x).
The following paragraph describes how the
three-dimensional behaviour was simulated in the
two-dimensional model. The support of the two-
dimensional arch by the horizontal concentric rings
was first modelled by sliding supports. This way,
the concentric rings act as elements without vertical
stiffness and an infinitely large horizontal stiffness.
 – design value of the tensile strength = 100 MPa as
the common historical value (de Bouw unpub-
lished);
– Young modulus = 210 kN/mm2 from Eurocode 3.
The two-dimensional calculations according to the
Eurocodes reveal that the stress levels, the instabilities
and the deflections of the structure do not exceed the
norm values. The maximum stress level appears at the
bottom of the inner arch of the side aisle and runs up
to 53 N/mm2 (which is 23% of 235 N/mm2 and 53%
of 100 N/mm2 ). The maximal buckling risk is 90%
of the buckling capacity of a profile under compres-
Figure 9. Side view of the Winter Garden (2007). sion and bending and occurs near the bottom of the
arch of the middle dome. The horizontal and vertical
deflections are less than 10% of the laid out deflec-
tion limits. Consequently, the buckling of the arches
The analysis of the model with these support reac-
is the most critical check of the metal structure. How-
tions showed that the above assumption was too much
ever, the global supporting effect against buckling of
of a simplification. To model the three-dimensional
the concentric rings can only be taken into account
behaviour in a more accurate way, the stiffness of the
in a three-dimensional model. At present this three-
concentric rings was determined and introduced into
dimensional model is still under development and no
the two-dimensional model by springs (Figure 8).
final conclusions can be made about the maximum
The two-dimensional model was then analysed
occurring stresses and deflections.
under symmetrical loading combinations made up of
self-weight and symmetrical snow load. The snow load
only applies to the structure that carries the glass cov-
ering, namely the arch of the middle dome and the
4 CONCLUSIONS
inner arch of the side aisle (Figure 9). The asym-
metric snow load, the wind load, the maintenance
The Winter Garden of the Royal Glasshouses of
load and temperature effects will be evaluated through
Laeken is part of the important Art Nouveau move-
three-dimensional calculations.
ment in Belgium. The current metal structure of the
The aim of the two-dimensional analysis consists of
Winter Garden is only minimally modified, compared
assessing whether the pressure ring at the top (Figure
to when it was built from 1874 to 1876. Currently,
8-z) and the tension ring at the bottom of the cupola
this masterpiece must undergo a far-reaching restora-
(Figure 8-x) can carry all the lateral thrusts. These
tion. The principal threat for the metal structure is the
findings would give an indication of how the struc-
corrosion of the compound profiles.
ture works. If these rings cannot withstand the lateral
In the first two-dimensional simplified model, the
thrust, the outer arch of the side aisle is needed for the
structure without the outer arch of the side aisle can
structure’s stability and therefore does not only have
withstand all the applied symmetrical loading combi-
an aesthetic function as assumed before. The pressure
nations (in terms of stress levels and horizontal and
ring is located at the boundary of the small cupola
vertical deflections). The second model, where the
and the middle dome. The tension ring is situated at
outer arch was added to the structure, shows that only
the bottom of the column. The other concentric rings
a minimum of stresses occur in the additional arch.
connecting to the arch of the middle dome (Figure 8-
The following conclusions regarding the structural
yi ) will all help carrying the lateral thrusts. The stress
behaviour of the Winter Garden can be drawn from the
levels in the structure without the outer arch of the
simplified two-dimensional analysis.
side aisle are investigated as well as the stresses in the
First, the outer arch acts as a backup structure for
outer arch itself when the complete model of the metal
the arch of the middle dome and for the inner arch of
structure is considered.
the side aisle. This conclusion reinforces the impres-
These stresses need to be compared with the mate-
sion that architect Alphonse Balat added the outer
rial properties of the used iron, which are currently
arch of the side aisle solely to ease the viewer’s mind.
unknown. These properties will be determined through
The viewer was not yet visually acquainted with the
tensile tests and a metal analysis in the course of the
very thin innovative metal constructions emerging at
next year. The current interpretation of the results of
that time.
the two-dimensional calculations is therefore based on
Secondly the stress levels and the horizontal and
the standard material properties:
vertical deflections do not exceed the limiting val-
– yield limit = 235 MPa from Eurocode 3; ues as prescribed in the Eurocodes. This conclusion

687
is however only valid for the applied symmetrical load
combinations.
Further work implies a three-dimensional finite ele-
ments analysis under asymmetric load combinations to
confirm the structural suitability of the Winter Garden
with the Eurocodes as well as material testing to obtain
the material properties of the used iron.
REFERENCES
de Bouw, M. Assessment of the metallic roof trusses in
Belgian school buildings between 1860 and 1914. Unpub-
lished PhD theses.
Goedleven, E. 1988. De Koninklijke Serres van Laken. Tielt:
Lannoo.
Kohlmaier, G. & Von Sartory, B. 1991. Houses of glass: a
nineteenth-century building type. Cambridge: MIT.
688
Koppelkamm, S. 1988. Glasshouses and Wintergardens of
the Nineteenth Century. Berlin: Granada.
Lauriks, L. 2007. Structurele herberekening van de Winter-
tuin van het complex van de Koninklijke Serres van Laken
in het kader van de restauratie. Brussel: unpublished
master thesis.
Polet, O. & Hamerlijnck, E. 2006. De Koninklijke Serres van
Laken. Brussel: Polet Photo Production.
Ranieri, L. 1990. Leopold II, Zijn urbanistische visie, zijn
monumentale bouwwerken. In Balthazar, H. & Stengers, J.
(eds.), Dynastie en cultuur in België: 173–212. Antwer-
pen: Mercatorfonds.
Vandewoude, E. 1987. De bouw van de Wintertuin te Laken,
1874–1876, een initiatief van Leopold II. In Album Carlos
Wyffels: 457–473. Brussel: Algemeen Rijksarchief.
Vierendeel, A. 1902. La construction architecturale en fonte,
fer et acier. Leuven: Uystpruyst.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Towards a systematic diagnosis of structural damage

I.A.E. de Vent, R.P.J. van Hees & G.J. Hobbelman

Delft University of Technology, Delft, The Netherlands

ABSTRACT: Traditionally, the diagnosis of damage is largely based on the intuition and experience of the
surveyor. Despite its benefits, this approach to diagnosing can be subjective, and difficult to communicate. This
paper discusses the limitations of the current intuitive approach, illustrated with the results of a questionnaire,
and suggests a method to increase its quality and efficiency. This method proposes a systematic description of
the development of structural damage, based on a uniform vocabulary, and a syntax that distinguishes between
damage patterns, the processes leading to damage, and the factors that influence course and extent of these
processes. By using this approach, it is possible to share the knowledge that is currently embedded in intuition
and experience, and enhance the retrievability of a diagnosis. The benefits are clear: the opportunity to learn
from cases, and make this knowledge available to the profession.

1 INTRODUCTION cause of damage is diagnosed with a certain accu-

The assessment of a building with damage has often
been compared to the approach to a patient in health
care. In both professions, the most important, but at
the same time most difficult phase is the diagnosis.
Despite the resemblances with medicine practice, the
field of building pathology still lacks an ‘encyclopae-
dia of diseases’. Although some atlases and expert
systems are available (e.g. MDDS), especially for stru-
ctural damage we still miss a consistent way of describ-
ing the processes leading to this damage, including a
coherent use of a well-defined terminology.
Damage is the manifestation of a lack of perfor-
mance of a building. Thus, the assessment of damage
can be regarded as an evaluation of the present and
expected loss of performance, combined with propos-
als for improvement or repair. Diagnosis is an essential
part of this assessment: its aim is not only to indicate
what has caused the damage, but, moreover, to try and
understand what initiated the damaging process (root
cause) and how this process has developed in time. In
medicine, this understanding of the course of a disease
is indispensable for finding ways to treat symptoms, to
cure a patient of his illness, and, if possible, to prevent
a disease from affecting other people as well.
The process of diagnosing damage is a deductive
process. It is based on the evaluation of the avail-
able data on symptoms and context of the damage,
and of the investigator’s knowledge on damaging pro-
cesses. This leads to the formulation of one or more
hypotheses, which, subsequently, are verified until the
racy. For a sound diagnosis, the investigator needs
thorough insight into all processes possibly leading
to damage, and also the ability to distinguish between
these processes. This is complicated by the fact that
damage often results from several causes, rather than
from just one. In its Recommendations, the ICOMOS
International Scientific Committee for Analysis and
Restoration of Structures of Architectural Heritage
(ISCARSAH) (2005) states: ‘This is why intuition and
experience are essential components in the diagnostic
process.’
A correct diagnosis of the damaging process and
its root cause is necessary for the proper assessment
of damage, and the importance of this correctness
has even more increased since we aim at minimal
interventions. Therefore, the best way to improve the
assessment of damage is by improving the accuracy
of the diagnosis. There are two ways to achieve this:
either by improving the quantity and/or quality of the
information available to the surveyor; or by improving
the qualities of the surveyor in processing (analysing,
interpreting, and evaluating) this information. Please
note that the word ‘information’ is used here for data
and knowledge in two domains. On the one hand, it
refers to data on a specific case: the symptoms and
the context of the damage. On the other hand, it also
concerns the more general knowledge of damaging
processes, which is essential for the formulation of
hypotheses.
In the past decades, many studies have contributed
to the improvement and facilitation of the diagnosis of

689
damage. We can classify these studies into five groups,
according to their main goal:
– Improvement of the data on damage;
– Improvement of the knowledge on damaging pro-
cesses;
– Improvement of the analysis of information;
– Improvement of the evaluation of information;
– Improvement of the exchange of information.
Many studies belong to the first group. Their aim
is to improve the completeness and quality of data on
the symptoms or the context of damage. These studies
have resulted in new and better techniques for on-site
investigation, for on-site and laboratory testing, and
for monitoring. Special attention has been paid to the
development of a range of non-destructive tests (Binda
et al. 2000, Binda & Saisi 2002). A good overview of
the techniques currently available is given in the doc-
umentation of the Onsiteformasonry project (2005).
Tragedies such as the sudden collapse of the civic
tower of Pavia in 1989, the Umbria-Marche earthquake
of 1997 and the Molise earthquake of 2002 have led
to a second group of research projects. This group
has aimed at extending our knowledge on specific
causes of damage. Especially the creep behaviour of
masonry and the effects of seismic actions on build-
ings have been investigated (Ignoul et al. 2006, Binda
et al. 1992).
The third group of studies has focused on the
development of techniques to help the investigator
to analyse the present behaviour of a structure and
interpret the way damage has occurred. The grow-
ing possibilities in computation have led to a range
of methods and strategies such as various methods for
numerical modelling (Lourenço 2002). For the anal-
ysis of collapse mechanisms due to seismic action,
the macro-element approach has commonly been used
(Doglioni et al. 1994). One of the applications of this
approach can be found in the digital, didactic hand-
book MEDEA (Manuale di Esercitazioni sul Danno Ed
Agibilità), which gives an overview of damage patterns
in masonry and concrete structures that have suffered
from an earthquake (Papa & Zuccaro 2004).
Furthermore, a fourth group has proposed meth-
ods to facilitate the evaluation of information. One
major aspect in this is the evaluation of the vulnera-
bility of structures, with the aim to predict their future
behaviour and the related risks (Augusti et al. 2001,
Speranza et al. 2006). The Multi-Hazard Assessment
of Vulnerability method (MHAV (historic buildings))
uses the macro-element approach to link building
typologies to their characteristic vulnerabilities to nat-
ural hazards such as earthquakes, storms, and floods.
The method evaluates the expected loss of both mate-
rial and cultural significance (D’Ayala et al. 2006).
Finally, a fifth, smaller group of studies has focused
on the improvement of information exchange. These
690
studies point out that a basic need is still to be fulfilled:
the need for better communication between all par-
ties involved in a restoration project. Suggestions have
been made to improve the exchange of data by using a
consistent terminology and standardised survey forms
(Van Balen 2001, Kelley & Sparks 2006).
The difficulties in information exchange mentioned
above are part of a larger problem: they show the
limitations that are inherent to our traditional diag-
nostic process. The development of new techniques
and methods for data collection, analysis and evalua-
tion has certainly contributed to the improvement of
diagnosis. However, less attention has been paid to the
improvement of the diagnostic process itself and to a
more efficient, practicable distribution of knowledge.
This paper discusses the limitations of the current intu-
itive approach, and suggests a method to increase its
quality and efficiency.
2 DIAGNOSING DAMAGE: LIMITATIONS OF
EXPERIENCE AND INTUITION
The current approach to diagnosing damage is, for
most experts, an intuitive one. The deduction is a men-
tal process that strongly depends on the intuition and
the experience of the investigator. These abilities are
highly valued, but the subjectivity of the approach also
brings about some inherent limitations.
Our brains are fast processors, and they are well-
trained in finding references. When diagnosing dam-
age, we use this ability to evaluate and compare a new
situation with our personal experience. It helps us to
judge and weigh different types of information, and
it makes us capable of readily formulating the most
probable hypothesis.
However, despite its benefits, this method has some
disadvantages. First of all, we select our references
more or less at random. This is a quick way to retrieve
our knowledge, but it is also subjective, irreproducible,
and sometimes inconsistent.
Then, experience can also make us biased. It could
make us jump to a conclusion by paying more attention
to facts that support our initial ideas, while we overlook
facts that are contrary. In this way, we would fail to
notice alternative hypotheses that could explain the
type of damage as well.
Moreover, our intuitive approach may keep us
unaware of the assumptions we make during the
deduction, wittingly and unwittingly. Because com-
municating this approach is difficult, the facts and
suppositions on which the diagnosis is based may
remain veiled to others. This makes it hard to judge
the reliability of a diagnosis, since the conditions under
which the deduction is valid are insufficiently clear.
To support and improve the diagnostic process, we
need to understand how the process of diagnosing
works. Therefore, a questionnaire was held among
experts. This questionnaire contained pictures of four
buildings with visible damage, and for each case the
experts were asked to answer the following questions:
– What could be the cause of this damage?
– On what information do you base this hypothesis?
– What data would you need to be able to verify your
hypothesis?
– If no initial hypothesis could be established on the
basis of the photograph(s), what would be the next
step(s) in the investigation process?
The results of this questionnaire were surprising.
On the basis of a first impression, the experts have
found different clues, made different assumptions, and
suggested a range of possible causes. In the next para-
graphs, the results of two of these cases are discussed
(Figs 1–5).
The damage in the sidewall of building A, as shown
in Figure 1, forms a typical damage pattern. Most
Figure 1. Building A, one of the cases in the questionnaire.
Figure 2. Overview of hypotheses and supporting arguments brought forward for building A.
691
experts immediately referred to the classic ‘back-of-
an-envelope’ yield line pattern, which is the failure
mode of a rectangular slab, supported around its
perimeter, and uniformly loaded perpendicular to its
plane. The crack pattern, thus, gives a clear indication
of the collapse mechanism. In accordance to the plate
bending mode, the loading must have been horizontal,
out-of-plane, and directed outwards. In this way, the
symptoms of the damage led to hypotheses such as
wind load (wind suction) and horizontal impact load
(explosion), but also hindered volume changes of a
concrete floor incorporated into the wall.
The experts used the context of the damage to
exclude some specific causes. The local character of
the damage and its distance to the ground led them to
omit the possibility of settlement as initiator. Further-
more, some experts (correctly) assumed the building
to be located in the Netherlands, which made them
conclude that an earthquake would be a less likely
cause.
Figure 2 gives an overview of the different hypothe-
ses and anti-hypotheses that were brought forward by
the participants. It also shows the arguments and the
assumptions that the experts used to frame and support
their hypotheses. It appears that, despite the typicality
of the damage pattern, the cause of damage is not so
obvious. Therefore, in response to the third question,
experts asked for more data on the properties of the
building, on its construction history, on the nature of
the location, and on the characteristics and distribution
of the damage.
The damage in building B is less evident than the
damage in building A. It can be described as an in-
plane deformation of the façade (Fig. 3). The skewing
Figure 3. Building B: overview of façade.
of the windows at the first floor, visible in Figure 4, is
accompanied by severe crack development in the lin-
tels and sills of the windows, and in the arch above the
door opening. The edges of the cracks show displace-
ments and demonstrate that the parts on either side of
the fracture have undergone a clockwise rotation.
Concerning the information on the context of the
damage, the participants particularly focused on the
construction of the building. For example, they pointed
out the difference between the timber-framed con-
struction of the sidewall, and the stonework façade. In
addition, the relatively large openings in the façade, the
apparent age of the building, and speculations on the
adjacent building on the left side and the open area on
the right side led them to the assumption that building
B may have little or no lateral stability. This assump-
tion helped them to explain the lateral displacements
in the façade.
Figure 5 shows that, in this case, both the symp-
toms and the context of the damage were used to
formulate hypotheses. These hypotheses vary from
environment-related processes such as differential set-
tlement, to overloading due to a lack of horizontal
restraint, and hindered volume changes due to cor-
rosion of the anchors. The hypothesis of differential
settlement appears to be supported by many arguments
and assumptions. Nevertheless, its root cause remains
unclear, although several options were expressed.
Summarising, the results of this questionnaire
demonstrate the benefits of our current approach. Even
on the basis of only some photographs, experts have
shown that their skill and experience make them able to
formulate a hypothesis that could explain the damage.
The use of photographs in this questionnaire did
hinder the evaluation of damage and context. Never-
theless, this method has been chosen with precisely
this goal: It forced the participants to be more aware
of the facts on which they based their hypotheses. In
this way, the responses also show the limitations of the
intuitive approach.
One of these limitations is that experience can
sometimes make experts biased. This could be over-
come by joining our experience, as the range of
692
Figure 4. Building B: Detail of damage.
alternative hypotheses resulting from the combination
of the responses shows.
Furthermore, the results emphasize that our line of
reasoning is influenced by unconscious assumptions.
For example, most experts interpreted the damage
in building A as caused by out-of-plane loading of
a slab supported around its perimeter. Apparently,
they assumed that the perimeter of the crack pattern
matches with the borders of a room behind the wall.
Yet, in reality the floor lies at the level of the horizontal
crack, which omits the possibility of wind suction and
explosions as a cause, and increases the likeliness of
hindered volume changes of the floor itself as a hypoth-
esis. It is clear that assumptions like these should be
mentioned explicitly when formulating a hypothesis.
Based on the results of the questionnaire, we can
now make some suggestions to improve the current
approach to diagnosing:
– Pay more attention to alternative hypotheses;
– Place more emphasis on explicitly mentioning the
facts and assumptions on which the diagnosis is
based;
– Find ways to join and share our experience; and
– To obtain these goals, our knowledge needs to be
accessible, retrievable, and communicable.
3 DATA ACCESS
To improve the accessibility of our knowledge on
damaging processes, we should examine how our
experience-based reference system could be supported
by a more systematic way of data cataloguing. A pos-
sibility is the use of knowledge-based systems. These
systems are not a new phenomenon: They are widely
used in health care, and also for building pathol-
ogy some systems do exist. Knowledge-based systems
consist of a framework to contain data, and a tool to
make these data accessible.This tool could be an index,
but also a deduction function: a set of rules composing
a ‘wizard’that guides a user through a list of questions,
to determine which specific information is asked for,
or to find a solution to a specific problem. In building
Figure 5. Overview of hypotheses and supporting arguments brought forward for building B.
pathology, a knowledge-based system could contain
an overview of damaging processes, including clear
selection criteria how to distinguish between them.The
benefits of such a system are obvious: It supports
the investigator in the framing of a hypothesis, while
the systematic approach makes the method unam-
biguous and, therefore, communicable. Prerequisite is
a consistent use of terminology, to make the stored
knowledge retrievable and suitable for exchange.
An example of a knowledge-based system is the
expert system and decision support tool MDDS (Mon-
ument Damage Diagnostic System, successor of the
Masonry Damage Diagnostic System). This system
aims to facilitate a minimal intervention approach,
by offering a structured, transparent and consistent
693
method for analysing and diagnosing damage. MDDS
helps to collect and order all relevant data on symptoms
and context, and supports the interpretation of these
data by offering background information on damage
types and patterns, damaging processes and methods
of testing (van Hees et al. 2005).
Until recently, MDDS mainly contained informa-
tion on damage related to the interaction between
materials and environmental factors. To fill this
gap, a module on structural damage is now being
developed. In the next section, it is explained
why a diagnostic instrument for structural damage
should be based on a clear distinction between pro-
cesses and factors, and on the use of a consistent
terminology.
Figure 6. Diagram of a structural failure mechanism.
4 DISTINGUISHING BETWEEN DAMAGES
In accordance with the suggestions made in the previ-
ous sections, an instrument for supporting the diag-
nosis of structural damage should consist of two
databases. We need a knowledge base, in which
we can store our combined knowledge on dam-
age and damaging processes. And then we need a
rule base to make this knowledge accessible and
retrievable.
The framework of both databases should provide
the diagnostic instrument with three types of use.
First, it should allow for a passive use: The knowledge
base should serve as a reference, as an encyclopae-
dia on structural damage, made accessible with a
traditional index. A more active use is possible by
supplying a search function, so that the encyclopae-
dia can also be searched by criteria other than the ones
mentioned in the index. Similarly, hyperlinks between
related articles can help improving the retrievability of
information.
These passive and active uses help to retrieve spe-
cific information. However, we may miss essential
facts, if we do not specifically ask for it. To estab-
lish a diagnosis, it is not only necessary to verify
the most probable hypothesis; we also need to be
aware of alternative hypotheses that could explain
the damage as well. The diagnostic instrument can
be particularly important to overcome this problem by
offering interactive support.
When using this interactive function, the system
should evaluate the data entered by the user, by apply-
ing the inference rules stored in the rule base. The input
data are used as arguments, from which the diagnos-
tic instrument tries to derive conclusions on the nature
and cause of the damage. In this deduction, the instru-
ment should first check the possibility that a certain
process could have led (or contributed) to the damage
under investigation. Next, it should rank all possible
processes on the probability that they could have taken
place. In this way, the interactive function can supply
694
the user with a hierarchical overview of alternative
hypotheses.
To allow for a passive, active and interactive use
of the diagnostic tool, all information on damag-
ing processes should be classified and presented in
a systematic and structured way. Therefore, all data
contained in the knowledge base need to be of uni-
form language, in vocabulary and in syntax. This asks
for clear definitions of damage, and for a logical way
of describing damaging processes. These issues are
discussed in the following two sections.
4.1 Uniform vocabulary: clear definitions of
damage and damage patterns
The consistent use of a clear terminology is a prereq-
uisite for communication without misunderstandings.
Several glossaries on damage and deterioration exist
next to each other. However, especially for databases, a
uniform vocabulary is essential. For instance, MDDS
has a built-in dictionary of damage terms. It would be
profitable to compare and combine these individual
glossaries into one.
In the process of diagnosing, we should also be
aware of the objectivity of our word choice. Our
description of damage should be independent from the
presumed cause, to prevent preliminary assumptions.
For instance, we should be cautious when using the
terms ‘material decay’and ‘structural damage’before a
diagnosis is established, as these terms imply the cause
of damage. Therefore, we need an unbiased typology
of damages. A proposal for a neutral description of
structural damage patterns can be found in Naldini
et al. (2007).
4.2 Uniform syntax: clear distinction between
processes and factors, and between symptoms
and context
Damage often results from several contributory
causes. Hence, insight into the whole process lead-
ing to damage is indispensable for diagnosing. This
Figure 7. Model for an overview of structural damaging processes; settlement-related processes discussed in more detail.
insight is particularly important when aiming at mini-
mal intervention. To be able to evaluate an intervention
on its effects on the performance of a building in time,
we need to know how the damage occurred, how it
evolved, and under what circumstances. Thus, rather
than on the root cause alone, we should focus on the
development of damaging processes.
To describe damaging processes, we need a clear
vocabulary, but also a uniform syntax. The syntax
suggested in this paper reflects the relations between
the different phases of the processes. This concept of
a structural failure mechanism has two general rules
(de Vent & Hobbelman 2007a, b).
In the first place, the structural failure mechanism
divides each damaging process into four steps: event,
impact, effect on the structure, and damage pattern
(Fig. 6). These steps describe the course of a dam-
aging process. The arrangement allows for an easy
comparison between the various processes.
Secondly, the structural failure mechanism clearly
distinguishes between processes and factors. The
damaging processes lead to damage, while the fac-
tors influence the course of a process and/or the
extent of its effects. Each factor can affect a transi-
tion between two of the steps, and is related to the
scale on which this part of the damaging process
takes place. Therefore, the factors of influence can be
separated into three main categories: environmental
variables, geometric variables, and material variables
(Fig. 6).
This systematic description will help improve the
diagnosing by giving insight into both the possibility
that a certain process has led to damage, and into the
probability that this process has taken place. In this
way, it will give insight into the time path of the occur-
rence of damage. This will help to choose an optimal
intervention.
695
5 EXAMPLE: PRELIMINARY OVERVIEW OF
STRUCTURAL FAILURE MECHANISMS
The knowledge base of the diagnostic instrument
should contain an as complete as possible description
of all damaging processes including their factors of
influence. A study of literature has resulted in a pre-
liminary overview of structural failure mechanisms.
A condensed version of this overview is shown in
Figure 7.
For clarity reasons, the structural failure mecha-
nisms have been subdivided into three main groups of
causes: differential settlement, overloading, and hin-
dered volume changes. As can be seen in the first three
columns, differential settlement of the foundations can
be caused either directly by variations in loading or in
bearing capacity, or it can be induced by differential
settlement of the soil, due to local variations in the sub-
soil or changes in the underground (e.g. construction
works).
Each category of causes has been linked to typi-
cal damage patterns, as illustrated in the last column.
The overview in Figure 7 only mentions the damage
patterns related to differential settlement. A more com-
prehensive discussion of damage patterns can be found
in Naldini et al. (2007).
This preliminary overview of structural failure
mechanisms needs further organising to make it appli-
cable in a computerised knowledge-based system.
Principles of mechanics will be used to streamline it
to a greater extent.
More work still needs to be done on the rule base,
which should contain the inference rules meant to
distinguish between processes. These rules will be
based on the factors of influence. A preliminary list
of the factors related to differential settlement is given
between the four columns of Figure 7.
6 CONCLUSIONS
Like in medicine, understanding the course of dam-
aging processes is indispensable for finding ways to
treat and prevent structural defects in buildings. The
results of the questionnaire have shown that, although
intuition and experience allow us to readily analyse
damage and formulate a hypothesis, we should be
aware that our current approach to diagnosing can be
subjective and difficult to communicate.
To improve this diagnostic process, this paper has
presented a method of describing damaging processes
in such a way that it allows for a systematic overview
of all alternative hypotheses that could be formulated
for a certain type of damage. This overview can serve
as the basis of a knowledge base.
More work on relating damages to processes is still
to be done. Important is also the formulation of infer-
ence rules, which will partly be based on objectified
results of the questionnaire.
The outcome is meant to serve as a basis for a
knowledge-based system that could be beneficial to
overcome and prevent bias, by supporting the formula-
tion of alternative hypotheses. Furthermore, it should
help to make a diagnosis more explicit, so that it is
easier to communicate. Such an expert system would
be applicable in practice, both as a learning tool for
novice surveyors, and as a support tool for experts.
Moreover, it should help in sharing knowledge among
all professionals.
REFERENCES
Augusti, G., Ciampoli, M. & Giovenale, P. 2001. Seismic
vulnerability of monumental buildings. Structural Safety
23(3): 253–274.
Binda, L., Gatti, G., Mangano, G., Poggi, C. & Sacchi
Landiani, G. 1992. The collapse of the civic tower of
Pavia: a survey of the materials and structure. Masonry
International 6(1): 11–20.
Binda, L., Saisi, A. & Tiraboschi, C. 2000. Investigation
procedures for the diagnosis of historic masonries. Con-
struction and Building Materials 14(4): 199–233.
Binda, L. & Saisi, A. 2002. State of the art of research on
historic structures in Italy. ARIADNE 11 Workshop ‘His-
toric structures’ – programme of the EC 5th Framework
activities, Praha, Dejvice, 20-26/05/2002.
D’Ayala, D., Copping, A. & Wang, H. 2006. A conceptual
model for multi-hazard assessment of the vulnerability of
historic buildings. In P.B. Lourenço et al. (eds), Proc. 5th
Int. Conf. Structural Analysis of Historical Constructions,
New Delhi, India: 121–140. Delhi: Macmillan India Ltd.
Doglioni, F., Moretti, A. & Petrini, V. 1994. Le chiese e il
terremoto. Trieste: Edizioni LINT.
Hees, R.P.J. van, Naldini, S. & Sanders, M. (2005). An expert
system for analysis of damage to plasters due to salt
696
and moisture. Proc. Sem. Soluble salts in the walls of
old buildings, Damages, processes and solutions, Lisbon:
16.1–16.11.
ICOMOS ISCARSAH 2005. Recommendations for the anal-
ysis, conservation and structural restoration of architec-
tural heritage. Istanbul: Maya Basin Yayin.
Ignoul, S., Schueremans, L., Tack, J., Swinnen, L.,
Freytons, S., Binda, L., Van Gemert, D. & Van Balen, K.
2006. Creep behaviour of masonry structures – Failure
prediction based on a rheological model and labora-
tory tests. In P.B. Lourenço et al. (eds), Proc. 5th Int.
Conf. Structural Analysis of Historical Constructions,
New Delhi, India: 913–920. Delhi: Macmillan India Ltd.
Kelley, S.J. & Sparks, S.P. 2006. The challenges of struc-
tural stabilization following the hurricane Katrina disaster.
In P.B. Lourenço et al. (eds), Proc. 5th Int. Conf. Struc-
tural Analysis of Historical Constructions; Possibilities
of Numerical and Experimental Techniques, New Delhi,
India: 261–268. Delhi: Macmillan India Ltd.
Lourenço, P.B. 2002. Computations on historic masonry
structures. Progress in Structural Engineering and Mate-
rials 4(3): 301–319.
Naldini, S., Vent, I.A.E. de, Hees, R.P.J. van & Binda, L.
2007. Definitie van constructieve schadepatronen; De
MDDS constructieve schadeatlas. In J.M. van der Veen
(ed.), Praktijkboek Instandhouding Monumenten; Deel II-
4 Buitenwanden (21): 1–20. Den Haag: Sdu Uitgevers. In
Dutch, English article in preparation.
Onsiteformasonry 2005. Results and research methodologies
of Onsiteformasonry; On-site investigation techniques
for the structural evaluation of historic masonry build-
ings. Berlin: Federal Institure for Materials Research and
Testing (BAM). Cd-rom.
Papa, F. & Zuccaro, G. 2004. MEDEA: A multimedia
and didactic handbook for seismic damage evaluation.
In M. García-Fernández & A.B. Walker (eds), Proc.
XXIX General Assembly of the European Seismological
Commission, Potsdam, Germany.
Speranza, E.,Viskovic,A. & Sepe,V. 2006. Integrated method
for the assessment of the structural vulnerability of his-
toric towers. In P.B. Lourenço et al. (eds), Proc. 5th
Int. Conf. Structural Analysis of Historical Constructions,
New Delhi, India: 651–658. Delhi: Macmillan India Ltd.
Van Balen, K.E.P. 2001. Learning from damage of masonry
structures, expert systems can help! In P.B. Lourenço &
P. Roca (eds), Historical Constructions 2001, Possibilities
of Numerical and Experimental Techniques, Proc. 3rd Int.
Seminar at Guimarães, Portugal, University of Minho:
15–28. Lda; Multicomp.
Vent, I.A.E. de & Hobbelman, G.J. 2007a. Failure mech-
anisms as a method for the assessment of structural
damage. In C.A. Brebbia (ed.), Structural Studies, Repairs
and Maintenance of Heritage Architecture X: 439–448.
Southampton: WIT Press.
Vent, I.A.E. de & Hobbelman, G.J. 2007b. Determining the-
oretical failure mechanisms for structural damage. In
A. Zingoni (ed.), Proc. 3rd Int. Conf. Structural Engineer-
ing, Mechanics and Computation: 219–220, full version
on cd-rom. Rotterdam: Millpress Science Publishers.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

From geometrical and crack survey to static analysis method: The case study
of Santa Maria del Quartiere dome in Parma (Italy)

F. Ottoni
Department of Civil Engineering and Architecture, University of Parma, Parma, Italy

ABSTRACT: Aim of this work is to show the research on the structural problems of the dome of a XVII century
church, Santa Maria del Quartiere, in Parma (Italy), as an exemplar case study on numerical analysis method
connected to geometrical and cracks survey. In order to to plan the future intervention for its consolidation, a
study of the hexagonal structure was carried out by using a finite element model which reproduced its general
geometry and various cases of load and restraint, starting from the equilibrium approach for masonry domes.

1 INTRODUCTION

1.1 The church of Santa Maria del Quartiere

The plan of the church of Santa Maria del Quartiere,
in its original drawing, dates back to the beginning
of XVII century and was plausibly attributable to the
architect Francesco Aleotti, thanks to the financial
support of Ranuccio Farnese. In 1610, Giovan Bat-
tista Magnani modified the early design by adding the
choir and the cloister on the South side. The centric
plan definition of the Church is characterized by a rig-
orous hexagonal prismatic structure, which ends with
two overlapping “pagoda” bodies. Outside, the volume
compactness is only interrupted by ribs protruding Figure 1. The hexagonal dome of Santa Maria del Quartiere.
from the large angular pillars and by the hexagonal
slender bell tower.
Inside, the large pavilion dome (frescoed by Anto- diagonals, in fact, present a maximum deviation of
nio Bernabei between 1626 and 1629) leans on the six 12 cm. Based on the characteristic of the dome, this
wide arches circumscribing the central hall. could represent an exemplar case study for the analysis
of similar pavilion domes.
1.2 The dome
2 THE CONSTRUCTION GEOMETRY
The structure seems to be concluded by the pavilion
dome, which represents the main structural element 2.1 The geometrical survey
in the whole crack outline of the building. The large
masonry dome stands out the centre of the church, A careful study on the geometry of the structure rep-
starting from the line of the last internal cornice; and resents the fundamental step for the understanding of
it rises up to approximately 9 meters. its static behaviour. Therefore, an accurate survey of
The thickness of the structure varies with its height: the monument was carried out by examining the whole
at the springer it measures about 30 cm and it gradu- church with a direct method of measurement and the
ally decreases towards the top, where it reaches 20 cm modern technique of laser-scan. The geometrical sur-
thickness. It is built above a 1,5 m high tambour. The vey of a monument represents the starting point from
hexagon at the basis presents some light irregularities, which to set up a comparative process with the past.
not too dissimilar from the ideal geometrical shape Firstly, an architectural survey has been conducted
inscribed into one 20 m diameter circle: the internal in order to acquire the dimensional data necessary

697
Figure 2. The laser-scan survey. Longitudinal section of the
church.
to the planimetric description of the monument. Dur-
ing the first phase of survey “in situ” measures were
drawn, through different integrated methodologies,
e.g. direct and instrumental survey for the interiors
and photogrammetrical outside.
Precision instruments, such as the total station, were
used in the survey of the hall and of the longitudinal
section along the major axis – from the main door to
the apse and to the room behind the choir.
Outside, a survey of the main front of the church was
made by using photogrammetry; stereoscopic metric
frames have been acquired, through the materialization
of the support points and the relative monographs.
In this phase, bi-dimensional drawings (1:50 scale),
necessary for the first crack pattern individuation,
were produced.
The first analysis of the structure unravelled the
necessity of a further acquisition of metrical and
formal data concerning the dome, and in particular
its intrados, which represents the most complex and
degraded part.
As a consequence, a complete 3D description of the
dome and the underlying structures has been carried
out in order to understand the cracks evolution. The
continuous survey of the church, by using the laser
scanner (Leica HDS 3000), with 5 different measure-
ment stations, has produced the implementation of the
first direct survey data, necessary to the definition of
a 3D model of the building.
The scanner Leica HDS 3000 uses a “by flight time”
technology, with a fixed net of 2 cm. It is important not
to reduce too much this scansion interval, when using
5 stations which could effectively cover all the cones
of shadow with a considerable superposition of the
points, mainly in the upper part of the dome.
In order to unify all the points, the method of geo-
referentiation on target was used, based on topographic
data (software Cyclone 5.6).
698
Figure 3. Series of horizontal section in the dome,
highlighting its geometry and displacements.
In order to obtain a precise definition of the dome
cracks, several transversal and horizontal sections in
the intrados have been produced, which describes all
the dome irregularities.
2.2 The cracks survey
Contextually, a crack and degradation survey has been
conducted on the monument, aimed at identifying
the main pathologies of the structure and the exact
mechanisms of fracture and degradation.
This survey showed that the church is affected by a
widespread system of cracks and degrade phenomena,
both in the interior – on the dome and the underlying
arches – and in the external masonry faces. Particu-
larly significant is the cracking pattern in the dome,
composed by passing and non-passing cracks, which
are variously distributed on each panel of the pavilion
and on the edges.
The dome is indeed the characteristic structural
element of the building and it seems to be mainly
responsible for the collapse mechanism. In order to
identify its collapse mechanisms, the crack pattern has
been analyzed, by studying the thickness of the cracks
and the degradation of the materials.
The crack pattern is mainly constituted by 6 pass-
ing cracks on the edges of the 6 panels (composing
the pavilion hexagonal dome) which develop from the
abutment limit of the dome, up to about 2/3 of the
whole height. In addition, four major cracks are evi-
dent in the centre of the panels I, III, IV and V. They
mainly develop upright, starting from the abutment

Figure 4. The cracks pattern. The six passing cracks on the
edges are evident among each panels. Other four major cracks
are in the panels I, III, IV and V.
line, then crossing the 6 windows opened in each panel,
and reaching approximately the same height of the
cracks on the edges.
Some boundary conditions, such as the recent con-
struction of buildings near the church, have produced
a number of variations in the crack pattern.
The analysis of the church cracks and deformations
has allowed to identify as a minor cause of damage the
erection of a recent A.C. building which leans against
the eastern wall of the church (Fig. 5). While it was rea-
sonable to hypothesize that the weight of this modern
building could have produced a considerable founda-
tion soil sinking, the lack of evident cracks along the
connection to the church suggests that this has not
happened.
The substantial symmetry of the crack pattern finds
some variations, even notable, in some of the panels
(Fig. 4, panels II and VI) in which the central cracks
are less large and numerous than the others. These
variation are due to the buildings joined to the church:
indeed, the buildings represent the dome thrust force
confinements.
A clear map of the cracks has unravelled a substan-
tial symmetry of the crack pattern, which seems to
confirm the typical collapse mechanism of the domes,
with a lowering of the top of the structure, mainly
because of its own weight.
The examination of the collateral structures of
the church has confirmed a typical dome collapse
mechanism: in the ambulatories around the hall some
699
Figure 5. Plan of the church and of the buildings leaned
against.
asymmetrical cracks are evident on the arches, due
to the thrust of the dome on the underlying pillars.
The same scheme is repeated in the superior corridor
around the dome.
2.3 The metrology as an analysis instrument
The comparison between geometrical survey and crack
pattern of the building has represented the first fun-
damental step for the subsequent static analysis of
the monument and, in addition, it has allowed us to
formulate some hypothesis on collapse mechanism.
Therefore, quantitative results of the survey have
been related to the unit of measurement, used at the
time of church construction, in order to investigate the
original geometry of the structure. Through an analy-
sis of measurement variance, it has been possible, in
fact, to establish a connection between the actual con-
figuration and the static movements in the monument
developed, during the centuries.
The metrology applied to this building has clarified
that it had been built by a rigorous geometry, based
upon the unit of measurement “braccio” (in Parma
it measured 54,52 cm length) and “pertica”, which
measured 6 braccia.
The identification of the original design represents a
key step in the analysis of a construction. It is a starting
point for the evaluation of its changes during the cen-
turies, and obviously of its damage evolution, through
the identification of modifications and permanence.
The plan of the church develops on hexagonal base
and it ends in the dome, which is the vertical projection
of 6 underlying arches. The plan measures 18 braccia
(corresponding to 3 pertiche) in the internal part, in
which the dome is inserted. Then, by moving in the
plan from the centre of the hall, up to the external
wall, we can find other two modules of 6 braccia (1
pertica) each.
The three consecutive hexagons can be inscribed in
as many circles, which repeat the progression 18-6-6
in the plan distribution. As it is well known, numbers 6
and 12 were fundamental in the constructive system of
the XVII century, which was based on the duodecimal
system instead of the actual decimal one.
This perfect drawing is also maintained vertically:
in fact it is possible to identify the exact classical order
proportion when examining, in the vertical section of
the building, the succession of pedestal, column, and
trabeation. All the architectonic elements are linked
by a precise proportion, based on the original unit of
measurement (Fig. 6).
The dome too is inserted in this perfect design, with
its centres lying exactly 3 braccia under its impost
plane. It is evident from the drawing the presence of
a modular repetition, both in plan and in elevation.
However, it’s very interesting to note, as revealed from
the geometrical survey, that the measurement of the
dome doesn’t fit in the ideal model of perfect mea- Figure 6. Scheme showing the transversal section of the
surement ratio. In fact, the height of the dome was dome and the underlying structures. The metrology and the
found to be 20 cm less than expected, when compared identification of proportions in order to obtain a historical
to the expected measurement. This discrepancy con- monitoring system.
cerning the height of the dome could further support
the collapse mechanism hypothesis.

3 THE STABILITY OF THE DOME

3.1 The equilibrium approach and the collapse

mechanisms
The preliminary analysis of the damage and of the
geometry of the church allowed us to state that the
observed cracks are primarily caused by the thrust
force of the dome. The collapse mechanism identified
in the present church is in fact typical of the major
dome structures, as it was clearly described by Poleni
in his study on the Vatican Dome.
“Arch is nothing else than a force caused by two
weakness” said Leonardo da Vinci, even before the
fundamental Hooke’s theory which related the shapes
of an arch and of a catenaria. Based on literature data
and, in particular on the historical approach, devel-
oped by Méry in the XVII century on the arch collapse
mechanism, it is possible to define the extreme posi- Figure 7. The Poleni’s studies on the masonry domes.
tion of the line of thrust, which determines the critical
points of the structure. Indeed, the minimum and the
maximum values of the thrust locate the position of form an arch and they work together by compression,
the cracks and the hinges in which the collapse mech- making the most of the mechanical properties of the
anism begins. For a circular arch, this position is fixed masonry.Therefore, it is possible to draw, for the dome,
at the top and at 30◦ above the arch abutment. the same line of thrust which was previously identified
This theory still holds for the domes. as the basis of the collapse theory on the arches. In this
A dome, in fact, can be imagined as composed by a way, it is easy to identify the specific equilibrium state
series of arches, which could be obtained by slicing the in compression, at which the dome is safe. This prin-
structure along meridian planes. Two opposite slices ciple has been numerically applied by mathematicians

700
and architects, following the observation of the Vatican
dome. (Fig. 7) In keeping with this, it’s interesting to
read the debate that took place in 1743 among the three
mathematicians Jacquier, Le Seur and Boscovitch, and
also the Poleni analysis untitled “Sulle colonne fesse”,
in 1743 and 1748). Starting from these observations on
the functioning of the domes, it becomes easily under-
standable the typical crack pattern of these structures,
which highlights meridian cracks.

3.2 The empiric- experimental method

In order to confirm the qualitative advanced hypoth-
esis, the structural analysis of this dome starts from
the aforementioned studies and proceeds by thor-
oughly investigation of the existing tensional states.
Indeed, the early crack analysis has to be considered
the finest instrument for the monument knowledge,
which is necessary for its conservation. The calcu-
lation procedure we use in this study follows the
“empiric-experimental method” (recommended in the
recent “Guide Lines” for the historical buildings con-
solidation, approved in Italy in 2006). The verifying
operations of the collapse mechanism, followed in this
paper, derived from historical and numerical methods,
each other integrating.
According to the historical method the dome is
investigated by using some simplifications: firstly,
starting from the symmetry of the structure, the dome
is converted to an arch, taking advantage of the
empiric-experimental knowledge on the masonry vault
and domes mechanical behaviour. Then, the thrust of
the dome is calculated by applying the equilibrium
study to one single slice of the dome. This slice is
obtained by cutting two consecutive dome panels in
their centre, and by considering the edges in the mid-
dle of them (Fig. 8). The thrust of the whole dome is
then calculated in these central edges is calculated the
thrust of the whole dome. As stated before, to calculate
this we have to consider only the weight of the dome
above the line of 30◦ degrees.
Furthermore, it has to be considered the dead weight
of each slice, applied in the centre of mass.
The reference to a simplified arch, which reaches its
stability through the solution of the three equilibrium
equations, allows to obtain the radial thrust force of
the dome (then decomposed in the two directions of
each panel) which in this case results H = 8200 Kg.
3.3 The finite element method
In the process of mechanical identification of the
structure, a first virtual model only formed by the
dome has been built up. The numerical modelling of
the dome has been performed by consecutive steps,
in order to obtain a static identification of the geo-
metrical structure of the dome and to identify the
causes underlying the observed static problems, which
Figure 8. The macro-elements calculation in the analysis of
the masonry dome structures.
emerged during the “in situ” investigation. The solid
continuum mechanics, especially the finite element
method, offers the most suitable and practical models
for skeletal structures macro-modelling.
Then, a 3D model of the dome has been built, by
using a simplified solid elaboration. A precise geo-
metrical survey restitution is very important for the
construction of the model; however the correct sim-
plification of the numerous data of the 3D survey (as
the laser-scan model) it is also of fundamental value.
In order to obtain a quite precise mesh of the struc-
ture and of the output, the finite element used for the
FEM (Abaqus) elaboration is a tetrahedron solid with
4 nodes. Based on the simple structure of the dome,
we built up a regular solid, on hexagonal base, sym-
metric, which grows in height and follows the circular
profiles of the real dome edges, as measured in the
survey. The used 3D model is constituted by a single
shell, with a variable thickness, starting from 30 cm at
the abutment up to 20 cm at the top.
Finally, the main problem is to choose the consti-
tutive law for masonry. As observed by Di Pasquale
(1984), it can be considered that traditional masonry
structures have a very low tensile strength, which
decreases with time. The masonry can also be consid-
ered as an ideal no-tension material, especially in the
case of domes, in which the prevalence of compressive
stresses is obvious.
Then, we have attributed to this 3D model the
mechanical characteristics of an elastic, homogeneous,

701
Figure 9. Visualization of the FEM results on the intact
model of the dome, Smax principals. The tensile stresses are
evident in the corners, where, in fact, the real cracks have
been actually developed.
isotropic solid. Precisely, the following values have
been applied:
Young modulus E = 10.000 kg/cm2
Poisson ratio n = 0.2
Mass density g = 1800 kg/m3.
The previous conditions have been applied both to
the intact and the cracked structure. The material is
assumed to be linear-elastic; however, the non-linear
geometry has been introduced by inserting the cracks
in the undamaged structure. In this way, the non-
linearity of the material has been concentrated in the
discontinuities inserted in the cracks, which have been
actually surveyed in the real dome.
In this phase, it’s important to compare the obtained
deformation and strengthens results with the measured
ones. The comparison with the empiric methods is
recommended for the control, based on the uncer-
tainty of the mechanical characteristics of the material
“masonry”.
The utilized finite elements have defined the struc-
ture through a mesh which has fitted the volume,
by considering its singularities. Gradually proceed-
ing from the more simplified model of the structure,
up to simulate the real critical load and restraint con-
ditions on the dome, different load cases and dome
configurations have been considered:
Case 1. Intact dome subject to the only dead weight.
The model isn’t realistic but it’s useful to simulate the
previous damage. The results pointed out the higher
tensile stresses in the zones which are actually cracked
(Fig. 9).
Case 2. Cracked dome subject to the only dead
weight. In this model the real cracks on the edges have
been inserted, with their actual measured thickness and
height. Also in this case, the model has verified the
702
Figure 10. Visualization of the FEM results on the cracked
model of the dome, Smax principals. The tensile stresses are
very high in the terminal part of the cracks, where it has been
installed a recent automatic monitoring system.
presence of tensile stresses in the central zone of each
panel, in which we can observe the other major cracks.
Case 3. Dome, symmetrically cracked, both on the
edges and on the centre of the six panels, subject to
the dead weight (Fig. 10).
Case 4. Cracked dome, as before, subject to dead
weight and to the pointed weight of the roof .
Case 5. Insertion of an encircling steel tie on the
cracked dome. The absence of tensile stresses incom-
patible with the masonry properties, has confirmed the
validity of the hypothesis which have been assumed.
3.4 Comparison on findings with finite-element
analysis and empiric-experimental method
The results of the previously described ideal model
of the dome, geometrically regularized, verified the
outcomes of the first empiric-experimental phase.
Moreover, these findings have confirmed the collapse
mechanism hypothesis, based on the observation of
the cracks.
The numerical results of the first case (Case 1)
stresses that the radial deformation on the vertical sec-
tion, under the only dead weight, is so low to justify
the use of a linear model for the whole structure. In
fact, in the undamaged configuration, the horizontal
tension stresses are higher in the edges (in which the
main cracks actually are. Fig. 9). The founded tensile
pattern matches with the actual cracks, which really
represent the main continuous monitoring system of
the structure.
The analysis on the cracked dome (Case 2) under-
lines tensile stresses on the central part of each panel.
If we image to cut the obtained model about 1/3 of the
whole height of the dome, we can observe very high
tensile stresses on the limits of the cracks. Now, these
values are controlled by a recent installed precision
monitoring system, which would give information on
the trend of the cracks, also considering the seasonal
variations.
Obviously, the presence of the cracks has gradu-
ally modified the structural behaviour of the dome.
The structure has changed from a substantial axial-
symmetric functioning (in which each parallel slice
gives a contribution to the meridian ones) to that of 6
different parts (6 panels) which are linked at the top
and which are supported by the common abutment.
The cracked model (Case 3) simulates the real
behaviour of the structure, even if it uses a linear-
elastic constitutive law for the finite elements. In fact,
the insertion of the real cracks of the dome in the Figure 11. Visualization of the FEM results on the cracked
model allows to not consider the masonry extremely model of the dome, with the insertion of the encircling tie.
low resistance to tension and shear strain: in this way Here are highlighted the axial force on the tie.
the non-linearity is concentrated in the cracks.
It’s interesting to note that the load due to the pointed
weight of the roof, applied to the cracked model of
Then, the thrust force value, which has been cal-
the dome (case 4) doesn’t provoke significant varia-
culated both by empiric-experimental method and
tions in the horizontal global thrust. This result further
by finite element analysis, has been used for the
supports the collapse mechanism hypothesis, which
dimensioning of the steel encircling tie, which has to
pointed the global thrust of the dome, due to its self
be put in work on the dome.
weight, as the main responsible of its damage.
The area of the tie section could be easily calculated
by the equation At = H/σ adm.
(where A is the area of the tie, H is the thrust force
3.5 On the linear model choice of the dome and σ adm is the still admissible tension).
The section of the tie has to be the lowest, to reduce
After these analysis, we can say that the elastic linear
the reaction of the steel to temperature variations.
model, with the empiric insertion of the cracks, could
Therefore, in order to determine the better solution
represent a good approximation of the reality, of easy
for the consolidation of the structure, the experimen-
application and control. The presence, in some models,
tation on this type of intervention suggests to compare
of tensile stresses, has evidenced the existing discrep-
the performances of different materials.
ancy, but it has also allowed to identify the fracture
In keeping whit this, we hypothesize to use a
zones.
tie constituted by an high resistance steel, e.g. type
In fact, on valuable historical buildings, it’s useful
DYWIDAG (which has a σ adm = 6500 kg/cm2). At
to know, not only the limit collapse last resistance, but
the end of the calculation we have found that the use
even the rise of the actual damage state level. Indeed, a
of this type of steel allows a significant reduction of the
linear model permits to simply simulate the behaviour
tie diameter: from 2,6 cm to 2,15 cm large. This could
of the structural organism and then to clearly identify
assure a lower dilatation during the seasonal variations.
the risk zones.
Then, in the last case analyzed by finite element
method (Case 5) we have determined the effects of an
encircling tie on the dome. In order to gain the better
4 THE EXPERIMENTAL INTERVENTION results for consolidation intervention, different type of
materials have been simulated, with different diameter
4.1 The steel encircling dimensions.
As stated before, the historical debate about theVatican Precisely, an encircling intervention has been sim-
dome cracks (during the XVII and XVIII centuries) ulated, through the insertion of a tie on the line of the
has for good clarified that a dome exercises a radial windows in the dome. The numerical results which
thrust force which has to be nullified. we have obtained stress the different reactions to
Thanks to the symmetry of the structure, the solu- the temperature variations and the different restraint
tion of this constitutive problem can be solved by conditions.
the insertion of a constraint element, which would It is in course of study an hypothesis of using fiber
neutralize this force: a tie. reinforced material, instead of the projected steel tie,

703
for the encircling of the dome. Some model have been
produced which are waiting the comparison with the
experimental results.
5 CONCLUSIONS
This analysis by numerical method has been organized
in a route from linear and quite simple model, build
with a geometrical simplification of the hexagonal
dome, up to definition of a more precise simulation.
The used linear model describes masonry through a
linear elastic constitutive equation, which assumes the
material’s lock of resistance to tension and its limited
compressive strength.
Thanks to a fusion between historical analysis,
precision geometrical surveys, experimental investi-
gations and numerical modelling, the static behaviour
of the monument has been determined, in order to
plan future intervention for its consolidation and
conservation.
ACKNOWLEDGMENTS
This research has been supported by a grant from
the Foundation Cassa di Risparmio di Parma e Pia-
cenza (Pres. Prof. C.Gabbi) to the Department of
Civil Engineering and Architecture of the University
of Parma (Research Group Resp. Prof. Ivo Iori and
Prof. C. Blasi).
704
REFERENCES
Benvenuto, E. 1981. The historical developments of science
of constructions, Sansoni (ed), The Sahara and the Nile:
21–35. Rome: Italy.
Blasi, C. Foraboschi, P. 1989. The masonry arch: a finite ele-
ment approach by no-tension-friction-elements and check
of the method. In Proc. AIMETA, IV Italian Conference
on Computational Mechanics, Padova: Italy.
Castigliano, C.A.P. 1879. Theorie de l’equilibre des systemes
elastiques et ses applications,A. F. Negro Publisher, Turin,
Italy.
Chiarugi, A., Fanelli, M., Giuseppetti G. 1983. Analysis of
a Brunelleschi-Type dome including thermal loads, Proc.
IABSE Symposium, Strenghtening of building structures
diagnosis and therapy, Venice, Italy, 169–178.
Chiarugi, A., Fanelli, M., Giuseppetti G. 1993. Diagno-
sis and Strenghtening of the Brunelleschi dome, Proc.
IABSE Symposium, Strenghtening of building structures
diagnosis and therapy, Rome, Italy, 441–448.
Couplet, P. 1729. De la pousse des voutes, Historie de
l’Academie Royale des Sciences, Paris, France.
Di Pasquale, S. 1984, Masonry solids static: theory and expe-
riences, In Proceedings of the Construction Department,
Florence University, News year VIII, n.5, Florence, Italy.
Di Pasquale, S. 1996, Art of construction between knowledge
and science , Essays Polis – Marsilio (ed), Venice, Italy.
Heyman, J. 1966. The stone skeleton Int J. Solid structures,
2, 249–279.
Heyman, J. 1980. The safety of masonry arches, Proc. Inst.
Civ. Eng. University of Cambridge, Cambridge, England.
Mery, E. 1840. Sur l’equilibre des voutes en herceu, Annales
de Ponts et Chaussess, Paris, France.
Poleni, G. 1748. Memorie istoriche della Gran Cupola del
Tempio Vaticano. Padova, Italy.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

A benchmarking study of the analysis of non-reinforced structures

applied to the structural behavior of domes

P. Morer
Department of Mechanic, University of Navarra, Tecnun, San Sebastián, Spain

R. Goñi
Department of Structural Analysis and Design, University of Navarra, Pamplona, Spain

ABSTRACT: This paper presents a benchmarking study of two methodologies used in the analysis of non-
reinforced structures: numerical and graphical methods applied to the structural behavior of domes. The structural
analysis was performed through the Finite Element Analysis software ABAQUS. The graphical methods are based
on the theory of limit analysis and they were carried out through CAD software tools Pro-Engineer and AutoCAD.
In order to test both methodologies, they were applied to the structural behavior of the Pantheon in Rome. The
results obtained through the numerical methods were compared with Mark & Hutchinson’s results. The graphical
analysis seeks to obtain and visualize the line of thrust and the Stability Factor. Results obtained were compared
with Lancaster’s results. The aim of the paper is to validate and to compare both methods.

1 INTRODUCTION of the construction process on building deformation

The Graphical Method used to study the structural
behavior of domes seeks to obtain and visualize the
line of thrust. This is a theoretical line that repre-
sents the path of the resultants of the compressive
forces throughout the structure. Heyman (Heyman
1998) has provided the theoretical basis for thrust
line analysis and has illustrated its application to
domed structures. Numerous studies (Cowan 1977,
Wolfe 1921, O’Dwyer 1999, Livesley 1992, Oppen-
hein 1992, Como 1992, Boothby 2001) have been
carried out for determining the thrust line. As early
as 1748, Poleni (Heyman 1988) analyzed St. Peter’s
dome in Rome using equilibrium methods to estimate
the internal line of thrust. Ochsendorf (Ochsendorf
2006) developed new interactive thrust-line analysis
tools using limit analysis to illustrate possible col-
lapse modes and to allow users to clearly visualize
the forces within the masonry. Although it is impossi-
ble to know the actual thrust line, since there are many
solutions, it is possible to establish its value within cer-
tain limits and to obtain the maximum and minimum
values according to the two extreme positions of the
line thrust.
The Finite Element Method can identify local areas
with of significant tension, possible cracking or dis-
tress under normal service loading conditions. With
those methods we can better understand the history of
the construction of ancient construction or the effects
and stress.
The Pantheon of Rome was chosen as a site at
which the methods were intended to be validated.
Also known as The Rotunda, it is an emblematic
building constructed two thousand years ago. Cracks
in the dome appeared a short time after its con-
struction. Mathematics, geometrical relationships and
numerical proportions were the main characteristics
of Roman construction (Huerta 2006). Roman knowl-
edge was based on experience and critical observation
of masonry building processes. Stability, equilibrium
and safety were achieved through correct geometry
and knowledge of the materials used.
The materials used in the construction, geometri-
cal data, design and structure were carried out by
Terenzio (Terenzio 1934), de Fine Licht (de Fine
Licht 1968), MacDonald (MacDonald 1976, 1982),
and Moore (Moore 1995). In 1986, Mark and Hutchin-
son (Mark & Hutchison 1986) used finite element
analysis to calculate elastic stresses in the dome of
the Pantheon. Lancaster, (Lancaster 2005) conducted
an equilibrium analysis to answer various questions
about the structure and design of the Pantheon.
This paper seeks to validate the values obtained,
comparing them with aforementioned studies. Once
the values were validated, additional studies were con-
ducted to reproduce the different stages of its construc-
tion. The results are discussed as well as the advantages
and disadvantages of both structural analysis methods

705
and the benefit of utilizing each one under different
circumstances.
2 SHORT DESCRIPTION OF THE PANTHEON
2.1 History
The first Pantheon was built byAgrippa in 25–27 BD. It
suffered severe damage in 80 AD during the great fire
of Rome. Although it was rebuilt by Emperor Domi-
tian, it was struck by lighting in 110 AD, during the
reign of Trajan. The present Pantheon was built for the
third time under Emperor Hadrian between 118–125
AD. That date is known by stamps on the bricks.
2.2 Geometry and structure of the Rotunda
The Rotunda is a circular space covered by a concrete
hemisphere. The internal height of the circular wall is
equal to the radius of the sphere of the dome. The dome
is held up by a 6-meter thick wall. The wall is made of
concrete and is covered with about 60 cm of brick. The
wall is not solid; it contains cavities, chambers and is
open towards the inside with large exedras arranged
on three levels (Lancaster 2005).
Inside the wall there is an archway of bricks, known
as a relieving arch, to support the upper wall over the
openings. The relieving arch is a semicircle of thin
bricks standing radially on end extending into the con-
crete wall. This arch distributes upper loads to the piers
during the long time it takes for the concrete to cure,
but after curing, it becomes an integral part of the wall.
This archway of bricks was only part of the wall and
did not extend into the dome. The entire structure is
resolved with arches incorporated in the wall. Exter-
nally, the lower part of the dome is hidden by the upper
storey of the cylinder, while the visible part is raised in
seven stepped rings. The first ring is 2.25 m thick. Six
other exterior step-rings present a slope 1:10 inwards.
The last ring starts at 47◦ from the axis.
2.3 Materials
The composition of The Rotunda has been documented
by de Fine Licht and MacDonald. The rotunda rests
on a ring foundation made of opus caementicium,
7.30 m wide and 4.50 m deep. The concrete is made
of travertine fragments in layers in a mortar of lime
and pozzolana. The unit weight of the mass of con-
crete is calculated to be about 2,200 kg/cm2 . The lower
level (1,310 cm) consists of alternate layers of traver-
tine fragments and fragments of tuff (the caementae) in
a mortar of lime and pozzolana. The core is faced with
a thickness of about 60 cm of brick, the unit weight
being 1,750 kg/cm2 . The concrete of the upper level
(890 cm) of the wall is lightened by changing from
706
travertine plus-tuff to brick plus-tuff, alternating lay-
ers of pieces of tuff and broken tiles or bricks also
in the same mortar. The uppermost level (840 cm) of
the wall consists of concrete, predominantly of bro-
ken bricks in mortar. The stepped rings are composed
of layers of brick fragments set in mortar. The unit
weight of the mass of concrete is calculated to be about
1,600 kg/cm2 . The tufo giallo and lightweight volcanic
scoria is used for the top of the dome, the unit weight
being 1,350 kg/cm2 .
2.4 Cracking
The dome and walls are cracked. Terenzio, during
his inspection of the Pantheon in 1930, documented
cracking in the wall and the dome. The cracks that he
documented continue up the dome to an average of
about 57◦ above the springing, according to Mark and
Hutchinson. Cowan theoretically placed this point at
37◦ degrees from the axis.
3 STRUCTURAL ANALYSIS METHODS:
STRUCTURAL BEHAVIOR OF THE DOME
A dome is one of the most difficult architectural forms
to construct. It is designed by rotating an arch 360◦
around its central vertical axis. It is a vault with double
curvature, capable of covering a large space without
any interior supports. They have great structural influ-
ence; the double curvature improves the behavior of
the compression-curvature forces. The study of their
structural behavior is based on knowledge of material
capacities, structural properties and construction tech-
niques. In ancient times, the construction was based
on geometrical rules. The relation between geometry
and structural behavior was decisive. The most impor-
tant thing was to ensure the stability and safety of the
structure.
The dome studied in this article is an unreinforced
concrete structure. For the equilibrium analysis, it is
assumed that it is a non-tension material, has virtually
infinite compressive strength and sliding failure will
not occur (Heyman 1995). The only load applied is
its own weight. This study was carried out using two
different methods: graphical and numerical.
3.1 Graphical method
In this article this is based on Wolfe’s method for
dome analysis (Wolfe 1921). He developed a graphi-
cal method based on membrane theory. He analyzed
the dome as a radial series of segments, subdivided
into voussoirs. The weight of each voussoir increases
from crown to base and rests on the center of grav-
ity cut by the axis plane. The force polygon allows
one to predict meridional and hoop forces. A recent

Figure 1. a) One segment, with a width of 45◦ , is located
between each of two adjoining exedras. b) One segment or
pier subdivided in 16 voussiors.
study at MIT applied this method to different domes,
developing computational tools to obtain the thrust
line (Wanda 2006). The thrust line is obtained based
on Lancaster’s analysis. In the Pantheon’s equilibrium
analysis, the dome is subdivided into segments of 45◦ .
The Pantheon’s floor is divided into eight main diag-
onals, where the exedras are placed. One pier, with a
width of 45◦ , is located between each of two adjoin-
ing exedras, (Figure 1a). And each segment is, in turn,
subdivided into sixteen voussoirs 5◦ each, except the
first, which is 3.5◦ where the oculus opens (Figure 1b).
3.2 Numerical method
The second structural analysis of the Pantheon was per-
formed with FEA software ABAQUS. The full model
has been meshed with C3D20RH elements with twenty
3D nodes. It was taken with an elastic modulus of
50 GPa. Different types of analyses have been carried
out and the results which illustrate the distribution of
stresses in the structure are reported in the following
sections (data is in kg/cm2 ).The dome was modeled in
the shape of a revolution and the results can therefore
be compared to Mark and Hutchinson’s results. Shells
are considered thin if their thickness is less than 5% of
the local radius of curvature, or t/R < 0,05 (Heyman
1995). In this case, 1.5/21.81 = 0.068 > 0.05, which is
why the dome is modeled as a solid 3D. These methods
give numerical information; maximum stresses and
deformations. It is necessary to know how to inter-
pret this information, which requires a great deal of
knowledge of specific computer tools.
The Pantheon’s dome is viewed as a net formed
by meridians and parallels. Internal meridional forces
increase from the crown to the base. All meridional
forces are compressive forces. Internal hoop forces act
in the latitudinal direction in a parallel ring. The hoop
707
forces allowed ring by ring construction of the dome.
For a dome of zero thickness, the membrane theory
predicts that the hoop forces change at 51◦ 49 from
compressive forces to tensile forces (Heyman 1995).
A typical failure consists of the formation of radial
cracks along its meridians that divide the dome into
segments, due to hoop forces.
3.3 Methods of comparison
In order to compare both methods, first were validated
with the results obtained by different authors (Mark
& Hutchinson 1986 and Lancaster 2005). Secondly, a
relation between each was sought. The values found
in the center of gravity are compared to in-points that
present singularities.
4 MODELS STUDIED
4.1 Base Dome: Models 1 and 2 of the dome with
different densities
The first two models analyzed are solely of the dome.
Those models are used to validate the values obtained
with the numerical method by comparing them with
the results showed by Mark and Hutchinson (Mark &
Hutchinson 1986). Subsequently, they are analyzed by
the graphical method and an assessment of the results
is obtained by comparing them to the data previously
recorded.
Two models were used. Model 1 consisted of a hemi-
sphere with an interior radius of 21.81 m and a thick-
ness of 1.5 m. According to Mark, 2,200 kg/m3 were
taken as the unit weight. It was not common among
Romans to use 2,200 kg/m3 for vaults, but we use that
unit weight to be able to compare the results obtained in
this article with their results. Model 2 included a hemi-
sphere with a unit weight of 1,350 kg/m3 in the dome.
Both have been solved with graphical and numerical
analyses.
It is assumed that the base of the dome is sup-
ported only against vertical loadings. There are no
horizontal buttresses. That simplification allows us to
analyze the dome using the shell theory. The results
obtained make it possible to know the maximum value
of circumferential tensile stress.
4.1.1 Study carried out with Abaqus
The values obtained through the graphical analysis
have been taken at the center of gravity of the volume
of the each voussoir. (Figure 2)
4.1.2 Study carried out with graphical analysis
Figure 3 shows the section of the dome used to obtain
results in the gravity center of each voussoir (volume)
and the force triangle corresponding to the centroids.

2
Figure 2. Model 2: maximum tensile stress (kg/cm ).
Figure 3. Force triangle.
The vertical line is the sum of each weight. The hori-
zontal line is an arbitrary one; the meridional forces are
parallel to segments that link two adjoining centroids.
The hoop forces are perpendicular at the meridians and
pass by the final points of the meridional forces where
they intersect with the horizontal line.
The value of the magnitude of the meridional and
hoop forces are measured and divided by the contact
area, yielding data that will be compared with previous
results.
Table 1 shows the results obtained at the base of the
dome with both methods and the values presented by
Mark and Hutchinson. Mark does not give data about
708
Table 1. Internal hoop and meridional stress, at the base of
the dome (kg/cm2 ).
Model 1 Model 2
Heavy density Light density
Hoop Meridional Hoop Meridional
forces forces forces forces
∗ ∗
Mark & H 4.9 3.3
Dome (FEA) 4.13 −3.7 2.2 −2.9
Dome (GA) 4.37 −4.6 2.68 −2.98
∗
these values were not provided by Mark & Hutchinson.
the mesh used, which could also affect the accuracy of
the results and the values of the meridional forces.
Additional information that is important to know
is to identify the point at which the internal hoop
stress changes from compression to tensile. The graph-
ical analysis shows that in voussoir number nine,
where the center of gravity is placed at 52.5◦ from
the axis of revolution, the hoop stress changes. The
numerical analysis coincides at 50◦ . These results cor-
respond approximately to the values expected for a thin
hemispherical dome [51◦ 49 ].
The vertical reactions in the base of the dome vary
from 8.5T – in the exterior of the dome- to 6.5T in the
interior.
4.2 Models 3–6: Influence of the uppermost level
and rings
The next four models deal with determining the influ-
ence in the structural behavior of different construction
levels. First, the lower and upper level are added to
the dome, secondly the uppermost and finally the
stepped-rings.
In order to evaluate the influence of the uppermost
level and the stepped-rings, four models were created.
In Model 3, the pier has been added to Model 2. The
pier has been reduced to 5.5 m and without taking the
chambers into account. In Model 4, the uppermost
level has been added to Model 3. Both are assumed
to use lightweight concrete (Figure 2). Model 5 is
a full model, with stepped rings added. The density
of heavy concrete was taken at 2,200 kg/m3 for unit
weight. Model 6 is similar to Model 5 with different
densities, 1,350, 1,600 and 1,750 kg/m3 , according to
figure 2. Models 5 and 6 deal with how to determine
the influence of the stepped-rings built over the dome.
4.2.1 Study carried out with Abaqus
Figure 4 shows the models studied withAbaqus. Model
3 assumes that the dome is embedded at the base. That
supposition modifies the values obtained. The trac-
tion zone changes, the tensile stress reaches a value
of 1.21 kg/cm2 and is placed approximately 72◦ from
Figure 4. Models 3–6 analyzed with Abaqus.

the axis (Figure 9a). The base of the dome is com-

pressed and the values of the hoop forces vary from
−0.1 kg/cm2 to −0.83 kg/cm2 at the exterior of the
Figure 5. a) Model 4 b) Model 6, analyzed with Graphical
cupola. The maximum value −0.83 kg/cm2 is taken analysis.
outside of the cupola and coincides with the line of
discontinuity.
In Model 4, the lower zone of the dome is embedded Table 2. Internal hoop stresses (kg/cm2 ).
in the uppermost level of the wall. The values of the
table are taken at the base and at three points: where
the rings start, where they end and at an intermediate
point on the interior surface of the dome. The traction
zone changes and is placed on the exterior of the upper-
most level, reaching a value of 0.6 kg/cm2 . In Figure
5b we notice a line of discontinuity where the dome
tries to open itself. At that point the value reached is
−0.1 kg/cm2 .
Models 5 and 6 show the influence of the rings
over the dome. The values obtained at the base of
the dome in model 6 hardly present differences from
model 4. The hoop forces are about 0.25 kg/cm2 and
the meridional forces about −2.7 kg/cm2 .

4.2.2 Study carried out with graphical analysis
Models 4 and 6 were analyzed with the graphical anal-
ysis method. The profile of the dome has changed. In
Model 4 the dome begins where the uppermost level
ends (Figure 5a). The embrace angle is approximately
70◦ , taken from axis.
In Model 6 the weight of each stepped ring has been
considered and the profile of the dome is taken, as
shown in Figure 5b.
The values obtained with the Graphical Analysis
method were reached with the steps indicated in 4.2
and are taken at the center of gravity of each voussoir.
4.3 Models 7–9: cracked model
The models with cracks completely change the behav-
ior of the dome. The internal hoop forces disappear.
The dome behaves like an array of arches.
Mark and Hutchinson studied the cracked dome
and modified their model by changing the boundary
conditions. The model studied in this article has been
created by dividing the base of the dome into twenty-
four meridians. In this way, the internal hoop forces
are eliminated.
4.3.1 Study carried out with Abaqus
Three models have been elaborated. In Model 7 the
dome has been divided into twenty-four segments,
with a length similar to that in Terenzio’s study.
This model is proposed with the sole purpose of
showing the cracked dome. The lower part of the
dome is divided into twenty-four slides (each fifteen
degrees). Figure 6 shows the dome deformed under its
own weight.
In Model 8 the dome has a pier at the uppermost
level, with light concrete. In Model 9 the stepped-rings
are added to Model 8.

709
Figure 6. Study of the cracked dome with Abaqus.

Table 3. Internal hoop and meridional stress at the base of

the dome and on the interior of the copula (kg/cm2 ).

Model 7 Model 8 Model 9

Light density Light density Light density
HF MF HF MF HF MF

Mark & H * ** * ** * **
FEA (base) * −16.7 * −9.4 * −2.6
FEA (1) * −4.85 * −4.25 * −1.6
FEA (2) * −3 * −3.84 * −1.58
FEA (3) * −1.15 * −1.24 * −1.52
*Hoop stress does not exist in this part of the dome.
**M & H do not give values.
HF = Hoop Forces, MF = Meridional Forces.
The results are shown in table 3. The values are
taken in three points (interior surface of the dome)
coinciding with the rings and at the base of the dome.
4.3.2 Study carried out with graphical analysis
The stability or equilibrium approach is the most
important concept to assess the safety of these struc-
tures. Equilibrium can be visualized using a line of
thrust, the theoretical line or inverted catenary, which
represents the path of the resultants of the compressive
forces throughout the structure. The forces’ line passes
through the pier up to the ground-line. At this point the
Stability Factor (SF), or Rankine factor, is obtained
(SF = b/2c, b being the pier weight). This factor was
determined by W. J. Rankine in the mid-nineteenth
century. If that factor takes a value 1, it indicates the
collapse point of the pier: a value 3 is considered safe.
That study was conducted with a Graphical Analysis
method.
Various suppositions have been made. Firstly, two
thrust lines were found for both models.
Figures 7 shows the two thrust lines found for each
supposition of the dome profile of the Pantheon. The
weight of the uppermost level is 13.10 N. Stability
Figure 7. Thrust line for profile Model 9 and Model 8. A
comparison.
factors of SF8 = 2.9 and SF9 = 3.9 were obtained for
model 8 and Model 9, respectively.
In order to know the influence of different densi-
ties used in the construction of the Pantheon, three
variables were taken for Model 9 in function of the
unit weight: a) the dome at 1,350 and 1,600 kg/m3 , b)
the dome at 1,350 kg/m3 (lighter) and c) the dome at
1,600 kg/m3 (heavier). The stability factors obtained
are: SF9 a = 3.9, SF9 b = 3.1 and SF9 c = 2.9. Those
values are compared with the values obtained by
Lancaster and are reflected in table 4.
4.3.3 Thrust line in a geometrical context
We can propose another question. What is the max-
imum height allowed for the Rotunda? What is the
relation between geometry and the thrust line? Figure
8 shows different stability values for model 9 in func-
tion of the height of the wall. That factor is taken at
the base of the pier and takes a value 1, indicating the
collapse point of the pier. The actual height is a diam-
eter of one sphere (73.75 Roman Feet) inscribed in its
interior. If the high value increases 82.35 Roman Feet
the SF is 1 and the thrust line passes out of the pier.
If the high increases 27 Roman Feet the SF value is 2,
where stability starts to be critical.
The thrust line is tangent to the extrados at point
A. That point coincides approximately with the ver-
tex of the so-called Diophantine triangle. Point A is

710
Table 4. Comparative graph between Lancaster and
Morer & Goñi
Figure 8. Relation between geometry and thrust line
(model 9a).
the theoretical point where the stress changes from
compression to tension.
4.4 Model 10: full mode with arches
Earlier models show different suppositions about
the construction of the dome. They study the influ-
ence of the different densities, the uppermost level,
the stepped-rings and the geometry of The Rotunda.
The next study has been done only with a numerical
model. This model was created by modeling the arches
built inside the wall and the lower part of the dome of
The Rotunda. We do not have more details about their
711
Figure 9. Model 10, main tensile stress.
Figure 10. Model 10, main tensile stresses, only in the
dome.
construction. Those arches distribute upper loads to
the piers during the lengthy period in which the con-
crete is curing, but after curing, it becomes an integral
part of the wall. This archway of bricks was only part
of the wall and did not extend into the dome.
The question is: what role do the arches have in
the construction of the rotunda? The big arches are
built over the exedras, discharging them of the load.
Figure 9 shows the main tensile stress. Those values
are important to predict cracking. It is clear that the
crown of each arch is supporting a greater stress. In
this model we are assuming that is a rigid model and
the dome are fixed over their supports. (Figure 9)
The maximum tensile stress is found over the
arches, in the central voussoir, with a value of
4.2 kg/cm2 . It is interesting to see the main tensile
stresses that act on the dome. Figure 10 shows that the
dome and the others part of The Rotunda are hidden.
5 DISCUSSION
5.1 Results obtained with Abaqus
On the one hand, the results found here with numeri-
cal methods correspond to the same order of the values
gathered by Mark and Hutchinson, with minor devi-
ations. The values found are fairly consistent, with
minor deviations that may be due to a large extent to
differences in the mesh and type of elements used.
Mark and Hutchinson do not point out where the
maximum values given were measured. Those results
validate our method.
On the other hand, those values have served as a
base for comparing results obtained with graphical
analyses. Then the results obtained in first Models
1–4 allow validation of the methods utilized and at the
same time, give some information about the influence
of different unit weights of the materials. The use of
lighter-weight concrete towards the crown of the dome
improves the structural behavior, reducing the internal
hoop and meridional forces.
Models 5–6 show the influence of the stepped-rings
and their different densities.
According to MacDonald, “the rings add to the load
over the critical or launch portion of the great vault and
function as buttresses, helping to bring the structure
into stability through compression.” After that study,
we can observe that the rings have been favorable for
increasing the compression value, but the dome was
working under compression before the step-rings came
into play. Their function as buttresses is not clear. It is
likely that the rings had a constructive function at the
moment when the concrete is poured and spread.
The last model evaluated, Model 10, shows how
the interior arches work with greater tensile stress in
the central voussoir. The hoop forces reach a value
of 9.4 kg/cm2 and the meridional forces a value of
−0.1 kg/cm2 . Those values indicate where and why
the dome is cracked.
5.2 Results obtained with graphical analysis
The utilization of both methods allows a comparison
of the information obtained. The graphical analysis
utilized is a simple, intuitive, fast and easy method to
implement. The three most important structural cri-
teria for a masonry structure are: resistance, rigidity
and stability. In the case of the Pantheon, the first
two criteria are irrelevant because the deformations are
very small. The third criterion, stability, is more rele-
vant. The numerical method gives information about
the first two criteria, but the graphical method gives
information about the third criterion.
Those structures work under their own weight
rather than under traction. Non-reinforced concrete
works only under compression. They are hyper-static
structures where numerous solutions exist. When the
712
structure is analyzed with the graphical method, the
objective will be to find a thrust line contained inside
the thickness of the arch. At this moment we can say
that this structure will work properly. The different
thrust lines obtained show how the use of different
densities benefits structural behavior. The whole dome
at 1,350 kg/m3 gives a stability factor lower than if
the dome were at 1,350 and the lower part at 1,600,
improving it by 20%.
The stability factor obtained for different models
varies from 2.9 to 3.9. If the dome had been built
without the third level, the uppermost level, it would
have been outside the limits of stability (Model 3,
SF = 1.38), with the uppermost level Model 8 improv-
ing and doubling this value, resulting in a valid thrust
line. This value increases to 3.9 with stepped-rings and
can be considered safe and stable. The rings are appar-
ently used for two purposes: for constructive reasons
and for improving the structural behavior of the dome,
but not as a buttress.
We assume that the dome is acting in a minimum
thrust state, as reflected by the cracks. These cracks are
irregular and narrower in shape and recorded as being
“in situ” by Alberto Terenzio in 1930. The lower part or
base of the dome is built similarly to The Rotunda, with
an archway of bricks covered by concrete. That zone
does not act as a cupola. The thrust line is tangent to
the extrados at approximately 51.7◦ and in the intrados
at 17◦ . The tangent points on the intrados coincide
with the archway. The crowns of the arches are set at
great tensile stress. The arches unload the stress on
eight piers. It can be observed that the cracks coincide
noticeably with the construction if the arches.
6 CONCLUSIONS
This paper presents a comparison between two meth-
ods of structural analysis. Those methods have been
applied to ten different models of The Pantheon in
Rome.
The results obtained have been compared to Mark’s
results. The values gathered by Mark referred only
to hoop forces. The deviations obtained were small.
These studies confirm the numerical method utilized
and offer additional information about the hoop and
meridional forces, specifically the values of the merid-
ional forces and the points of the dome where the
values were taken. Models have been elaborated with
greater precision than before.
The use of different densities in the dome improves
the stability factor and reduces the values of both hoop
and meridional forces. The stability factor increases
by 20%. The studies performed with graphical anal-
yses confirm the results obtained and give additional
information about the stability of The Rotunda and the
angle where the hoop forces change from compression
to tension. It is confirmed that the uppermost level and
stepped-rings improve the stability factor.
The cracks defined by Terenzio coincide with the
position of the archway and with the points where the
thrust line is tangent at the intrados and extrados of
the cupola. There is a close relation between the geom-
etry inherent toThe Pantheon, the thrust line calculated
for this geometry and the material used.
Lastly, it is concluded that the numeric model is
more complete and informative, but still is a theo-
retical model and is based on numerous assumptions
which can never be known. When the mesh is fine
and the elements have many nodes, it requires power-
ful computers to post-process the information given.
This method offers values for stress and deforma-
tions, but does not give information about stability.
Nevertheless, the graphical model offers information
about the stability of the dome and the position of
cracks. Cracks can be simulated with numerical analy-
sis, but require models of materials, information about
material behavior, long procedures and non-linear
analysis.
REFERENCES
Block, P., Ciblac, T., and Ochsendorf, J. 2006, “Real-time
limit analysis of vaulted masonry buildings,” Computers
and Structures,Volume 84, Issues 29–30, pp. 1,841–1,852.
Boothby,T.E., 2001, “Analysis of masonry arches and vaults.”
Prog. Struct. Engng. Mater. Vol. 3, pp. 246–256.
Como, M., 1992, “Equilibrium and collapse analysis of
masonry bodies.” Meccanica, Vol. 23, No. 3, pp. 185–194.
Cowan, H.J., 1977, “A History of Masonry and Concrete
Domes in Building Construction.” Building and Environ-
ment, Vol. 12, pp. 1–24. Pergamon, Press Printed in Great
Britain.
de Fine Licht, K., 1968, The Rotunda in Rome: A Study of
Hadrian’s Pantheon, Copenhagen.
713
Heyman, J. 1977, Equilibrium of Shell Structures, Oxford:
Oxford University Press.
Heyman, J. 1998, “Hooke’s cubico-parabolical conoid,” Pro-
ceedings Royal Society, London. pp. 39–50.
Heyman, J. The Stone Skeleton, Cambridge: Cambridge
University Press. 1995.
Heyman, J., “1988, Poleni’s Problem,” Thomas Telford Jour-
nals, Vol. 84, Issue 4, pp. 737–759.
Huerta S., 2006, “Galileo was Wrong: The Geometrical
Design of Masonry Arches,” Nexus Network Journal, Vol.
8, Num. 2, pp. 25–51.
Lancaster, Lynne C., 2005, “Concrete vaulted construction
in Imperial Rome,” Published Cambridge, UK; NewYork,
ISBN 0521842026.
Livesley, R.K. 1992, “A computational model for the limit
analysis of three-dimensional masonry structures,” Mec-
canica, Vol. 27, Issue 3, pp. 161–172.
Mark, R., and Hutchinson, P., 1986, “On the Structure of
the Roman Pantheon.” The Art Bulletin, Vol. 68, No. 1,
pp. 24–34.
Moore, D. 1995 “The Roman Pantheon: The Triumph of
Concrete,” Mangilao, Guam, University of Guam Station.
O’Dwyer, 1999, “Funicular analysis of masonry vaults”,
Computers and Structures 73, pp. 187–197.
Ochendorf, J., 2006, “Real-time limit analysis of vaulted
masonry buildings,” Computers & Structures, Vol. 84,
Issues 29–30, pp. 1841–1852.
Oppenheim, I.J., 1992, “The masonry arch as a four-link
mechanism under base Motion” Earthquake Engineering
& Structural Dynamics, Vol. 21, Issue 11 , pp. 1005–1017.
Terenzio, A., 1934, “La restauration du Pantheon de Rome,”
La conservation des monuments d’art & d’historie, Paris.
Wanda, L., 2006. “Equilibrium Analysis of Masonry,” PhD.
thesis.
William Lloyd, McD. 1976, “The Pantheon: design, meaning
and progeny.” Published Cambridge. Harvard University
Press, ISBN 0-674-65345-9.
Wolfe, W.S., 1921, “Graphical Analysis: a text book
on graphic statistics,” New York. McGraw-Hill. Book.
pp. 250–253.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Structural analysis of the Caserta Royal Palace timber roof connections

G. Fabbrocino, G. Marcari & C. Laorenza

School of Engineering, Structural and Geotechnical Dynamic Laboratory StreGa, University of Molise,
Termoli (Cb), Italy

E. Cosenza
Department of Structural Engineering, University of Naples Federico II, Naples, Italy

ABSTRACT: Preservation of historical structures often includes evaluation of timber roof trusses. According to
recent research achievements and European building design codes, connections can play a role in the response of
timber structures, but have paramount relevance for assessment of historical constructions under serviceability
and ultimate loading conditions. Whenever semi-rigid behaviour of connections is concerned, investigation
within rotational properties of connections through refined numerical modelling is required. The objective of the
present paper is to study the response under service loads of typical connections of ancient timber structures, by
using finite element-based modelling. The model is calibrated against experimental and numerical results found
in relevant literature and sensitivity analyses are carried out. An application to different connections of the roof
trusses of the Royal Palace in Caserta is presented.
Finite Element analysis seems to be able to simulate semi-rigid behaviour of joints within the elastic range.
Numerical moment-rotation diagrams are calculated for the different connections found on the reference roof
structures as a basic step for the global structural assessment under serviceability loading conditions.

1 INTRODUCTION This means that the level of analysis has to

Preservation of historical structures often includes
evaluation of timber roof trusses. Satisfactory per-
formance of existing wood truss system in terms of
both resistance to applied service loads and long-
term response is dependent on wood species, truss
proportions and timber connections (Tampone, 2007).
Generally, stress concentrations are particularly criti-
cal at joints where a component is connected to others,
such as those caused by notches (load to angle grain)
or other sudden changes in cross section.
This is confirmed by survey and inspections of tim-
ber buildings damaged after extreme natural events,
which often point out to inadequate connections as
the primary cause of damage (Derinaldis & Tampone,
2007). Common connections adopted for modern con-
structions can be designed to fit hinge requirements or
even to ensure full stiffness and strength, so that conti-
nuity can be assumed at joints. This circumstance does
not apply to existing and historical wood artefacts are
concerned.
In such cases, connection response can play a rele-
vant role both at ultimate and serviceability limit states
of historical timber structures (Branco et al. 2006;
Seo et al. 1999) and modern European building codes
(prEN 1995).
be enhanced and semi-rigid behaviour of connec-
tions properly accounted for. Obviously, this kind of
approach is by far more complex and requires a thor-
ough investigation of the specimens and advanced
numerical modelling, being of moderate interest for
practical purposes. The objective of the present paper
is to investigate the rotational behaviour under service
loading conditions of typical joints of historical timber
trusses.
Numerical analyses of the joints have been per-
formed using a standard software package used by
practical engineers (http://www.hsh.info). Orthotropic
elastic behaviour of wood was accounted, and contact
elements friction-based between the timber elements
were considered.
Reference results of the numerical simulations are
herein discussed and compared with experimental and
numerical data provided by Parisi et al. (1997) and
Parisi & Piazza (1995, 1998, 2000).
The model showed the capacity to predict the elastic
stiffness of the joints if contact elements are properly
simulated. Calibrated model is used for the numerical
investigation of the elastic behaviour of the connec-
tions of the roof system assemblage of the Royal Palace
in Caserta (Italy). It is an interesting case study involv-
ing a timber roof truss of the 18th-Century. Some

715
Figure 1a. View of the Royal Palace in Caserta, Italy.
Figure 1b. Plan view of the Royal Palace.
critical issues related to modelling of this kind of
timber truss are also discussed.
1.1 The timber roof truss of the Royal Palace in
Caserta
The Royal Palace in Caserta is one of the most
famous historical establishment of the Italian Baroque
(Caroselli 1967, Chierici 1984).
It represent a masterpiece of the creative genius
of the Italian architect Luigi Vanvitelli (1700–1773)
and it was probably the most representative monu-
mental building erected in Europe in the 18th-Century
(Figure 1a). Due to its magnificence with its surround-
ing natural landscape, the site has been included in the
World Heritage List by the world Heritage Committee
in 1997 (http://whc.unesco.org/en/list/549).
Works started in 1752 and were completed in 1847.
The building has a rectangular plan and four close
courtyards which are also rectangular, as schemati-
cally reported in Figure 1b. It is 36 metres high and has
five storeys in addition to the underground level. The
original design drawings of the building can be found
in Vanvitelli, (1756). For the construction of the roof
system, Vanvitelli has been mainly inspired to the tim-
ber roof truss of the Basilica of San Paolo Fuori le Mura
erected in Rome in the 4th Century a.C. and rebuild in
1824 after a serious and wide damage due to fire.
The timber roof of the Royal Palace is made of
trusses of Chestnut species, span of about 23 meters
are found. They are typically 3 meters spaced from
716
Figure 2. Partial view of the Royal Palace roof truss.
Figure 3. Roof truss of the Royal Palace.
the adjacent trusses. The height of the truss, measured,
from the top to the bottom of the main chord is 5.75 m.
In Figure 2 the partial view of the roof truss of the
Royal Palace in Caserta is showed.
The roof truss of the Royal Palace is a modified
version of the queen post truss type (Izenour, 1992).
A schematic drawing of such a truss layout is reported
in Figure 3. It consists of a system of two main rafters
and two secondary rafters, two horizontal chords
placed at two different levels, and inner elements
connected the top and bottom chord.
The inner elements consist of two diagonal struts
supporting the main rafters, two vertical queen posts
and a central king post. The bottom chord is connected
to both the king and the queen posts by using iron
stirrups.
The bottom chord and the main rafters have a
cross section 270 × 400 mm, while the lower rafters,
Figure 4. Connection between king post and main rafters.
Figure 5. Connection between the queen post, the lower
rafter and the top tie.
5.80 m long, have a cross section 270 × 340 mm. The
diagonal struts are 1.20 m long, with dimensions of
370 × 300 mm. The upper chord is 9.90 m long, with
a cross section of 250 × 370 mm. The angle between
the rafter and the chord is 24◦ .
Connections between timber elements are birds
mouth type, and present single or double mortise and
tenon (Figures 4, 5, 6). Timber roof trusses sit on
wood bearing supports named ‘gattone’, which are
frequently found in historical buildings (Figure 3). In
addition, the truss heels were designed without any
overhang. A binding stirrup was used to improve the
resistance of the connection between the main rafter,
the lower rafter, the bottom chord, and the ‘gattone’,
as showed in Figure 6. The main and the second rafter
were connected by iron stirrups bolted to the elements
and placed at the upper end of the secondary rafter.
A system of purlins spaced 1.35 meters and rafters
placed on the top of the purlins, was used as support
of the laths and roofing tiles.
Timber truss structure of the Royal Palace was built
according to traditional methods and old construction
717
Figure 6. Connection between the rafters and the main
chord.
Figure 7. Details of a timber truss from an ancient construc-
tion manual (Belidor, 1739).
rules that have been found by an extensive literature
review (Emy 1842, Mésange 1754; Belidor 1739).
Figure 7 reports an example of design and construction
details of a timber truss derived from ancient manuals.
It is worth noting that truss elements have a cross sec-
tion ratio B/H = 0.675 very close to the typical range
B/H = (5/6 ÷ 3/4) suggested by many authors for tim-
ber beam elements (Belidor 1739, Blondel 1771). In
addition, Blondel (1771) suggested to calculate√the
height of the cross section of the elements as H = 4 L,
where L is the length of the element measured in foot,
while B is the base equal to L/2. Conversion can be
obtained using the metric conversion chart provided
by Rondelet (1802). In this case, an underestima-
tion of the cross dimensions is observed, in fact the
cross section ratios of the elements averaged the value
B/H = 0.50.
At present time, timber trusses are characterised
by different types and stages of degradation basically
due to environmental factors, lack of maintenance,
and effects of repeated interventions. Visual inspec-
tions revealed confined damaged areas due to frequent
exposure to rapid changes in moisture content, or
to fungal attack. However, in any case the state of
conservation lead to a loss of serviceability of the
whole roof structure (Ceraldi & Russo Ermolli, 2006).
These problems were already detected in the past,
and different interventions were performed. In fact,
besides the partial replacement of the covering sys-
tem (including the purlins), the trusses were subjected
to repeated replacement of members and/or part of
members. In particular, the original heels, rafters, king
post and truss supports were frequently replaced with
new species of timber elements, which exihibited a
mechanical properties degradation due to the absence
of specific protection from moisture and biological
attacks. This circumstances lead to a variability of stiff-
ness and strength properties of wood truss members,
which influence the structural response at global level.
2 MODELING OF CONNECTIONS
2.1 Calibration of the finite element model
In the following, results of finite element (FE) anal-
yses performed on a typical birdsmouth connection
of Chestnut wood specie, are presented. In partic-
ular, calibration of FE model is performed on the
base of available data which can be found in the
relevant literature (Parisi et al. 1997, and Parisi &
Piazza 1995, 1998, 2000). In this work, the connec-
tion behaviour is simulated with the use of the FE
software package STRAUS7 (1999), and results are
investigated in terms of both force-displacement and
moment-rotation diagrams.
The geometric dimensions of the both main rafter
and chord were: length of the elements L = 1.90 m; ele-
ment cross section 190 × 225 mm and 200 × 200 mm,
skew angle equal to 30◦ ; rafter cross section notch
depth tv = 35 mm and notch length lv = 200 mm.
Two loading pattern were considered. The first one
consisted of a vertical pressure uniformly distributed
at the free end of the rafter; subsequently a trans-
verse force was applied perpendicular to the axis of
the rafter. The second one consisted of an initial axial
pressure uniformly distributed over the cross section
of the rafter; from this, a linear pressure distribution
over the cross section of the element was applied in
order to simulate a bending moment at the end of the
rafter. With reference to the axial compression stress
in the rafter, two values were accounted: fwr equal to
1.0 MPa and 1.5 MPa, respectively. These stress were
kept constant during the analysis.
718
Conventionally, transverse force, moment and rota-
tions have been assumed positive as the skew angle
was reduced. The main chord of the joint was con-
strained in the vertical displacements at the base, and
in the horizontal displacements at the vertical right-
side border. A structured mesh is used for the rafter
and the chord, whereas an irregular transition mesh is
used in the vicinity of the connection between rafter
and chord.
It is worth noting that wood exhibits anisotropic
elastic and inelastic behaviour, and the characteris-
tic inner structure which include several defects (e.g.
knots, slope grain). The use of refined non linear
orthotropic criteria is essential for detailed numerical
simulations of timber joint, since different strengths
and softening/hardening characteristics in orthogonal
directions can be accounted (Lourenço et al. 2007,
1997).
For the purposes of the paper, an elastic orthotropic
model is accounted for wood elements (Parisi & Piazza
2000; Bodig & Jayne 1982). Strengths and elastic
materials properties have been derived from available
tests given in Parisi & Piazza (2000) and represent
average values.
In order to properly fit elastic properties for the
wood members, which are characterized by different
grain directions, a local coordinate system was intro-
duced. The 1-axis and 2-axis are related to the direction
parallel and perpendicular to the grain, respectively;
the 3-axis is orthogonal to the 1–2 plane.
Therefore, the elastic parameters included the elas-
tic moduli E1 = 9200 MPa and E2 = 310 MPa par-
allel and orthogonal to the grain, respectively; the
shear modulus G12 = 580 MPa and the Poisson’s ratio
v12 = 0.4. Nonlinearities were concentrated at the
contact interfaces.
The rafter and the tie were connected by point
contact elements which allowed relative tangential
displacements of facing surfaces and limited normal
displacements. In particular, the Normal type contact
implemented in STRAUS7 was used, which provides
stiffness in compression and not in tension. In addition,
a friction coefficient was introduced in order to con-
trol the amount of lateral force that can be transferred
thorough the contact surface.
Let α the angle between the horizontal direction and
the normal of the contact, the axial elastic stiffness of
the contact elements has been estimated as EαA/L,
where A is the area of influence of each point con-
tact, L is the distance between the rafter and the chord
(1 mm) and Eα is the elastic modulus at the angle α.
The modulus Eα is calculated for each contact by tak-
ing into account the elastic moduli E1 and E2 of the
timber elements as follows (Lekhnitsii 1968):

Figure 8. Finite element mesh and boundary conditions.
In the following, the contacts are named as con-
tact#1 and contact#2. Equation (1) gives: Eα1 =
1710 MPa, and Eα2 = 320 MPa, for contact#1 and
contact#2 respectively, and for cross section dimen-
sions of 190 × 225 mm. The axial stiffness of the
inner point contacts were: Eα1A1 /L = 1943 kN/mm;
Eα2A2 /L = 728 kN/mm; for the contact points at
the end of each contact were Eα1A1 /L = 971 kN/mm;
Eα2A2 /L = 364 kN/mm.
For cross section dimensions of 200 × 200 mm,
Eα1 = 2209 MPa, and Eα2 = 416 MPa, for con-
tact#1 and contact#2, respectively, are obtained. The
axial stiffness of the inner point contacts were:
Eα1A1 /L = 1116 kN/mm; Eα2A2 /L = 558 kN/mm; for
the contact points at the end of each contact were
Eα1A1 /L = 445 kN/mm; Eα2A2 /L = 223 kN/mm.
The study included two values of the compression
stress in the rafter, fwr = 1.0 MPa and 1.5 MPa, respec-
tively. It is worth noting that self-weight of the wood
elements is considered in the analyses.
The Modified Newton-Raphson algorithm was used
in the incremental iterative solution of the nonlin-
ear problem. Figure 8 shows the finite element mesh,
including loading and boundary conditions of the
model.
2.2 Numerical results
Numerical results are showed in Figures 9–10 in terms
of nodal force-displacement diagrams with reference
to the cross section members 190 × 225 mm, and in
Figure 11 in terms of moment-rotation diagrams for a
cross section members 200 × 200 mm.
In addition, comparisons against both experimental
and numerical reference curves provided by Parisi &
Piazza (2000) are carried out. It is worth noting that
the nonlinear behaviour of the numerical response
occurred as the limit conditions of friction resistance
were attained and the rafter started to rotate. However,
numerical results have to be considered up to the elastic
limit.
As shown in Figures 9 and 10, major differences
between experimental and numerical results are found
in the case of low axial stress. For a compression stress
719
0 0.6 1.2 1.8 2.4 3
0
Force [kN]
1
Experimental curve
2 Parisi & Piazza (2000)
3 STRAUS7
7
Displacement [mm]
8
Figure 9. Comparisons between numerical and
experimental force-displacement curves for rafter
190 × 225 mm under fwr = 1.0 MPa.
0 0.6 1.2 1.8 2.4 3 3.6 4.2
0
Force [kN]
1
2 Experimental curve
3 Parisi & Piazza (2000)
4 STRAUS7
5
6
7
8
9
Displacement [mm]
10
Figure 10. Comparisons between numerical and
experimental force-displacement curves for rafter
190 × 225 mm under fwr = 1.5 MPa.
in the rafter of 1.0 MPa, the linear trend was followed
up to about a transverse load of 1.6 kN, which is about
30% higher than the experimental value (Figure 9). As
the rafter compression level increases (e.g. 1.5 MPa),
the linear trend was followed up to a transverse load
of 1.2 kN, and resulted very close to the experimental
value (Figure 10).
Again, differences between numerical and exper-
imental diagrams in terms of initial stiffness are
found for low axial compressive stress (e.g. 1.0 MPa),
because of the regularity in the numerical representa-
tion of the facing surfaces, which contrasts the reality
of hand-sawn indentations. This may explain also the
better agreement of results for higher compression val-
ues (e.g. 1.5 MPa), that partially mitigate effects of
local irregularities before that rotation of the rafter was
experienced.
Results in terms of moment vs rotation curves
for joint members are illustrated in Figure 11 for
axial stress levels in the rafter equal to fwr = 1.0 MPa
and 1.5 MPa. It can be observed that the numerical
response is in good agreement with those provided by
Parisi & Piazza (2000). Besides, the rotational elastic
stiffness is not affected by the rafter axial compres-
sion level, in accordance with the reference FE model.
 0 0.5 1 1.5 2 2.5
0 0
Moment [kNm]
0.0005
0.001
0.0015
0.002 STRAUS7 (1.0 MPa)
Parisi & Piazza (2000) - 1.0 MPa
STRAUS7 (1.5 MPa)
0.0025
Parisi & Piazza (2000) - 1.5 Mpa
Rotation [rad]
0.003
Figure 11. Comparisons of moment vs rotation curves
for joint members 200 × 200 mm for fwr = 1.0 MPa and
fwr = 1.5 MPa.
Figure 12. Contour of the compressive stress along the grain
(values in N/mm2 ). Rafter 200 × 200 mm – fwr = 1.5 MPa.
A sketch of the principal compressive stresses is plot-
ted in Figure 12, where the compressed zone are due
to the rotation mechanism of the joint.
2.3 Sensitivity analysis
A key point related to the reliability of numerical anal-
ysis is to know the influence of material data on the
structural response. A sensitivity analysis of the results
depending on the friction coefficient and the compres-
sive level fwr in the rafter is herein presented. The
model sensitivity with respect to the material parame-
ters and the skew angle have been also assessed, but not
reported herein for sake of brevity. The sensitivity of
the analysis with respect of the compressive strength
in the rafter has been illustrated in Figure 11.
A comparisons in terms of friction coefficient is
given in Figure 13 with reference to a rafter cross
section 200 × 200 mm. In particular, two values of
the friction coefficient were accounted: µ = 0.3 and
µ = 0.4.
Figure 13 illustrates results in terms of moment-
rotation plots. From Figure 11 it can be observed that
variation of fwr did not alter the response of the plain
joint in the elastic part (rotational stiffness remains
almost identical). Increase of friction factor µ results in
3 3.5 0 0.8 1.6 2.4 3.2 4 4.8 5.6
Moment [kNm]
0.0005
0.001
µ=0.3
0.0015 µ=0.4
0.002
0.0025
Rotation [rad]
0.003
Figure 13. Results of sensitivity analysis for µ = 0.3 and
µ = 0.4 for joint members 200 × 200 mm – fwr = 1.5 MPa.
an increase of the slope of the first part of the response
of about +10% (Figure 13).
2.4 Connections of the Royal Palace in Caserta
In the present section, numerical analyses of the joints
of the timber truss of the Royal Palace in Caserta are
presented. In particular, the behaviour of connections
was investigated by using the calibrate model above
discussed. The analyses accounted the joints between:
the rafters and the main chord, the king post and the
rafter, the diagonal strut and the rafter, the diagonal
strut and the king post. Due to the lack of experimental
data, the Chestnut wood class C30 is used for joint
elements in compliance to prEN 338 standard (1985).
The constitutive law of the wood elements was
reproduced according to an orthotropic elasticity
approximation. Mean values of the elastic Young’
s moduli parallel and orthogonal to the grain were
E1 = 10000 MPa and E2 = 640 MPa, respectively. In
addition the shear modulus G12 = 600 MPa and the
Poisson’s coefficient υ12 = 0.4 have been assumed.
Proper boundary conditions were considered and con-
tact points type Normal are used to model the interface
between the truss elements. The elastic stiffness of
each contact element was estimated as EA/L, in accor-
dance with Equation (1). In addition, a value of µ = 0.4
is assumed.
In this work, any interaction mechanism between
the main and secondary rafters has been accounted
for. It is worth noting that for accurate evaluation of
deformation behaviour of the joint, the partial com-
posite action between the rafters should be properly
modeled. Conversely, the effect of the metal stirrup
at the truss heel has been included in the finite ele-
ment model by using a set of linear spring elements.
The spring elements were placed diagonally with the
chord grain direction (1-axis) within the joint, and their
stiffness was derived based on formulation available
in literature (Gelfi et al. 1998). The analyses were car-
ried out under plane stress conditions and the Modified
720
 0 0.5 1 1.5 2 2.5 3 3.5
0
Moment [kNm]

0.00025

0.0005

0.00075 Diagonal strut and king post

Diagonal strut and rafter
Rafters and main chord
0.001
Main rafter and king post
Rotation [rad]
0.00125

Figure 14. Rotational response of the connection between

the rafters and the chord.

Newton-Raphson iteration method was used iterative

algorithm.
Long-term behaviour and moisture content of wood
elements were taken into account in a simplified
manner. In fact, modification factors were used to
reduce the mean elastic moduli to account for long-
term behaviour and the effects of moisture content
of wood elements, in accordance with technical lit-
erature (Giordano, 1999) and relevant building codes
(EC5, 1995). Joint between the main chord and the
rafters are characterized by a double tenon and mor-
tise, see Figure 6; the other joints have a single tenon
and mortise (Figure 4 and 5). Values of elastic stiffness
of the interface elements were calculated according to
Equation (1).
The loading scheme consisted of an initial axial
compressive force applied at the upper edge of the
main rafter, which was derived from the elastic anal-
ysis of the whole truss structure. In addition, a linear
distribution of pressure load was applied over the upper
edge of both rafters which generated a positive bending
moment in the joint. The behaviour of the connections
is reported in Figure 14 in terms of moment vs rotations
curves.
As the rafters-chord connection is considered, the
global stiffness of the joint increased after a first part
of the response. This circumstance can be related to
the loading sequence, that basically reflects the instal-
lation process. In fact, during the application of the
axial force in the rafter, the contact elements at the
joint are firstly engaged and give an initial stress and
strain distribution. The subsequent bending moment Figure 15. Contour of the compressive stress along the grain
applied at the edge of the rafter, forcing the skew (values in N/mm2 ).
angle to decrease, interacts with contact elements and
generates the observed non linear behaviour. joint; 2900 kNm/rad for diagonal strut-rafter joint;
This trend was also detected for the joints between 2800 kNm/rad for diagonal strut-king post joint.
the diagonal strut and the king post, and between the Therefore the elastic behaviour of the joint between
main rafter and the king post. With reference to the rafters and the chord was influenced significantly
the initial linear branch, an estimation of the initial by the stiffness of the metal stirrup. In fact, the mod-
stiffness can be carried out: 1700 kNm/rad for rafters- elling of the metal stirrup ensured the engagement of
chord joint; 5000 kNm/rad for main rafter-king post the contact elements between the joint elements in the

721
elastic phase, as demonstrated by deformed mesh in
Figure 15 (a).
Contours of compressive stress in the 1-axis direc-
tion are also plotted up to the elastic limit value of
rotation for each connection. It is easy to recognise a
concentration of stresses at the right end of the carving
in the chord.
3 CONCLUSIONS
The present paper dealt with the structural response
of historical wooden trusses present in historical con-
structions. In particular, stiffness and moment-rotation
performances of connections surveyed on the roof of
the Royal Palace in Caserta, Italy have been studied by
FEM numerical analyses. An advanced, but common,
numerical tool has been used to explore the capabil-
ities of general purposes program for assessment of
historical trusses under serviceability loads.
Model has been preliminarily calibrated against
numerical and experimental data available in techni-
cal literature. Satisfactory results have been found, so
that relevant information for real application have been
derived for a variety of connections found in Caserta
Royal Palace.
Obviously, results are not comprehensive, but repre-
sent an initial and crucial step for the global structural
assessment of roof trusses using advanced non linear
numerical tools.
REFERENCES
Belidor, B.F. 1739. La science des Ingènieurs dans la con-
dite des travaux de fortification et d’architecture civile.
Charles-Antoine Jombert (ed.), Paris.
Blondel, J.F. 1771. Cours d’Architecture civile (6 volumes),
Desaint (ed.), Paris.
Branco, J., Cruz, P., Piazza, M. & Varum, U. 2006.
Strengthening Techniques of Portuguese Traditional Tim-
ber Connections. Proc. Stuctural Analysis of Historical
Constructions, New Delhi 2006, P.B Lourenço, P. Roca,
C. Modena, S. Agrawal (Eds.).
Candelpergher, L. & Piazza, M. 2001. Mechanics of tra-
ditional connections with metal devices in timber roof
structures. Bodig, J. & Jayne, B.A. 1982. Mechanics of
wood and wood composites. Van Nostrand (ed.), New
York, 110–118.
Caroselli, M.R. 1967. La Reggia di Caserta: lavori, costo,
effetti della costruzione. Milano, Italy.
Ceraldi, C. & Russo Ermolli, E. 2006. Timber Coverings of
palatine Chapel in Caserta Royal Palace. Proc. Stuctural
Analysis of Historical Constructions, New Delhi 2006,
P.B. Lourenço, P. Roca, C. Modena, S. Agrawal (Eds.).
Chierici, G. 1984. La Reggia di Caserta. Ist.Poligrafico e
Zecca dello Stato-Archivi di Stato Coll. Arte Medievale e
Moderna, Roma. Proc. STREMAH 2001, 7th Interna-
tional Conference, Southampton, 415–424.
Derinaldis, P.P. & Tampone, G. 2007. The Failure of the Tim-
ber Structures Caused by Incorrect Design-Execution of
722
the Joints. Two Cases Study.Proc. ICOMOS IWC – XVI
International Symposium Material to Structure, Mechan-
ical Behaviour and Failures of the Timber Structures,
Florence, Venice and Vicenza, Italy.
Emy, A.R. 1842. Traité de l’Art de la Charpenterie. Liège.
Avanze & Comp (eds.)
Lourenço, P.B., Feio A. & Machado, J.S. 2007. Chestnut
wood in compression perpendicular to the grain: Non-
destructive correlations for new and old wood. Construc-
tion and Building Materials, 21(8), 1617–1627.
Gelfi, P., Giuliani, E. & Marini, A. 1998. Comportamento
della connessione a piolo nelle travi miste in legno e
calcestruzzo: modellazione teorica e confronti sperimen-
tali. Proc. III Workshop Italiano sulle Strutture Composte,
Ancona, Italy. (in Italian)
Giordano, G. 1999. Tecnica delle costruzioni in legno. Hoepli
Ed., Milano. (in italian)
Izenour, G.C. 1992. Roofed Theaters of Classical Antiquity.
Yale University Press, New Haven.
Lekhnitskii, S.G. 1968. Anisotropic Plates. Gordon and
Breach (eds.), New York (S.W. Tsai, T. Cheron, Transl.).
Lourenço, P. B., De Borst, R. & Rots, J. G. 1997.A plane stress
softening plasticity model for orthotropic materials. Inter-
national Journal of Numerical Methods in Engineering,
40, 4033–4057.
Mésange, M. 1754.Traité de Charpenterie et de bois de toutes
espéces, 2nd vol., Paris.
Parisi, M.A. & Piazza, M. 1995. Strutture lignee tradizionali
di copertura: modellazione e comportamento sismico. Atti
7◦ Convegno Nazionale ANIDIS, Siena, Italy.
Parisi, M.A., Piazza, M. & Modena, R. 1997. Model-
lazione e sperimentazione di giunzioni lignee di coperture
tradizionali in zona sismica. Atti 8◦ Convegno Nazionale
ANIDIS, Taormina, Italy,
Parisi, M. A. & Piazza, M. 1998. Seismic Behavior and mod-
eling of traditional timber roof structures. Proc. 11th Euro-
pean Conference on Earthquake Engineering, Rotterdam:
Balkema.
Parisi, M. A. & Piazza, M. 2000. Mechanics of plain and
retrofitted traditional timber connections. ASCE Journal
of Structural Engineering, 126(12), 1395–1403.
prEN 1995-1-1 Eurocode 5 – Design of timber structures –
Part 1-1: General – Common rules and rules for buildings.
prEN 338 (1985) – Structural timber – strength classes:
CEN/TC 124, Brussels, Belgium.
Rondelet, J.B. 1802. Trattato teorico e pratico dell’arte di
edificare. Tomo V, Caranenti (ed.), Mantova, Traduzione
italiana con note aggiunte di B. Soresina (in italian).
Seo, J.M., Choi, I.K. & Lee, J.R. 1999. Static and cyclic
behaviour of wooden frames with tenon joints under
lateral load. ASCE Journal of Structural Engineering,
125(3), 344–349.
STRAUS7 theory manual – version 2.2.3. (1999). Finite
Element Analysis System.
Tampone, G. 2007. Mechanical Failures of the Timber Struc-
tural Systems. Proc. ICOMOS IWC – XVI International
Symposium Material to Structure, Mechanical Behaviour
and Failures of the Timber Structures, Florence, Venice
and Vicenza, Italy.
Vanvitelli, L. 1756. Dichiarazione dei disegni del reale
palazzo di Caserta alle Sacre Reali Maestà di Carlo re
delle Due Sicilie e di Maria Amelia di Sassonia regina.
Milano, Il Polifilo, s.d., 128 pp.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Real-time probabilistic health monitoring of the Portogruaro Civic Tower

D. Zonta, M. Pozzi & P. Zanon

Department of Mechanical and Structural Engineering, University of Trento, Italy

G.A. Anese
Department of Public Works, Municipality of Portogruaro, Italy

A. Busetto
Busetto Engineering, Pordenone, Italy

ABSTRACT: The Civic Tower in Portogruaro is a 51 m high masonry bell-tower, originally built in the XIII
century, which today leans more than a metre out of plumb. Since 2003, the building inclination has been
continuously monitored with an optical inclinometer in an effort to see whether the tilt is still in progress. When
the monitoring started, it was thought highly unlikely that the Tower would tilt further. After three years of
monitoring and historical investigation, this idea was completely overturned. We show here how the initial view
evolved to a final awareness via a probabilistic model, based on Bayesian logic. We show how the joint use of
instrumental monitoring and historical documentation allowed timely recognition of signs of ongoing tilting and
accurate calculation not only of the mean inclination trend, but also the reliability level of this information.

1 INTRODUCTION
In September 2002 the Municipality of Portogruaro,
a town in North-Eastern Italy, asked the University of
Trento to install and operate instrumentation to mon-
itor the tilt of its Civic Tower. This building is an
ancient masonry bell-tower, of height 51 m overall,
which today shows an inclination measured as 1.197 m
at the north-east corner. Although the inclination was
and is striking, at the time there was no real specific
concern about stability because the Tower had always
been known to be leaning, as documented in local
chronicles. Also, there was no evidence of inclination
still in progress, and such an occurrence was judged
at the time to be very unlikely.
The monitoring system started recording the incli-
nation in October 2003. In September 2004, the Uni-
versity reported to the Municipality that the changes in
inclination recorded during this period were too small
to raise concern over the short-term safety of theTower.
In September 2005, after almost two years of
recording, the University told the Municipality of
Portogruaro that analysis of the data acquired to date
allowed calculation of a possible linear trend of 1.7 mm
per year. In particular, the direction of the possible
motion was very close to the maximum lean direc-
tion. Nonetheless, the leaning progression was too
small and the monitoring period too short to state with
certainty whether the trend calculated was a sign of an
ongoing process.
During the third year of monitoring, the Univer-
sity recovered some unpublished historical documents.
Among these, an original design project revealing an
elevation built in 1879, and documents reporting old
measurements of the inclination dating 1962 and 1997.
Based on this new information and on the fresh instru-
mental data, the University concluded, in the annual
monitoring report issued in September 2006, that an
increasing inclination was very likely.
This final judgment, in sharp contrast with the
initial view, might appear surprising. In reality, the
change of opinion is the result of a rigorous and quan-
titative logical analysis of the information available at
the time.
This paper follows this logical route again, showing
how the initial view evolved to final awareness via a
probabilistic model, based on Bayesian logic. In the
next Section we introduce the Tower at issue, its his-
tory and the monitoring system installed; Section 3
formulates the algorithm used to update the posterior
judgment based on the information acquired; the appli-
cation of this procedure to the Civic Tower is reported
in Section 4; finally, the outcomes are discussed at the
end of the paper.

723
Figure 1. Overview of the Tower (a); North and East elevation, cross-section and plan views at different levels.
2 THE TOWER AND ITS MONITORING
As seen today, the bell-tower of the Cathedral of
S.Andrea in Portogruaro, also known as CivicTower, is
a 59 m tall leaning campanile. The building was prob-
ably started in the XIII century. The tower was recently
surveyed by Busetto & Romanin (2001) using theodo-
lite and diastimeter, and the main geometrical features
are shown in Figure 1. From the architectural point of
view the Tower has a masonry column, a belfry and
a spire. The column has a roughly square cross sec-
tion, of size varying with height, from 7.30 m on a
side at ground level to 6.45 m at the top. The walls are
masonry infilled, with thickness varying from 1.3 m
at the base to 0.9 m at the top. There are four wooden
floors at levels 5.58 m, 12.07 m, 18.45 m and 22.74 m,
and an old masonry cross vault, now reinforced with a
concrete slab, at level 26.20 m. The main column ends
with a second cross vault, similar to that below, which
supports the floor of the belfry, at level 31.43 m. The
belfry balcony carries an octagonal tambour 5.45 m
in diameter and 4.9 m height, in turn topped by a
pyramidal spire 15.84 m in height overall.
The tower leans to the North-East. However the
angle of slant of the Tower is not constant over
the height, because the higher parts were apparently
rebuild with lower inclination in an effort to compen-
sate for a pre-existing tilt. Based on the 2001 survey,
the out of plumb respect to the north east edge of the
balcony, at level 36.62 m, is 1.197 m, corresponding to
an inclination of 35.47 ×10−3rd (equal to 2.03◦ ); while
724
the out of line of the spire is 0.31 m, corresponding to
an angle of 35.47 × 10−3rd .
Observing the outer masonry of the tower we can
see at least four different masonry textures; this is a
clear sign of successive building phases in the past.
Nevertheless, at the time monitoring started, little
documentation was available about this work. Local
chronicles report a reconstruction of the spire, in 1879.
The Municipality conserved incomplete documenta-
tion of work done by the architect Mario De Goetzen
between 1962 and 1963; this included reinforcement
of the tower with concrete ring beams and steel ties,
and consolidation of the lower level of masonry at the
North and East sides. As visible today, this consolida-
tion involved replacing the original low quality yellow
brickwork of the outer layer, with red brick of better
mechanical characteristics.
Between 2002 and 2003, the Tower underwent an
extensive experimental campaign including material
sampling, core drillings, endoscopies, flat jacks and
chemical analysis of the outer masonry foil. The results
of the investigation are reported in detail in the report
by Molteni (2003). The core drillings (labeled C1 to
C3 on Fig. 1) revealed that the masonry is infilled:
the external leafs consists of a single layer of bricks,
while the quality of the infill is extremely poor and
inhomogeneous. The flat jacks tests showed the better
qualities of the new brickwork with respect to the orig-
inal: a compression strength of 8.5 MPa was measured
at the North side (M1 on Fig. 1), against a strength of
3MPa measured at the South side (M2 on Fig. 1).
Figure 2. View of the pendulum inclinometer wire, including the network cameras (a); plan-view of the Tower highlighting
the thermocouples and plumb wire position (b).
2.1 Details of the monitoring system
Since October 2003, the tilt of the tower has been
observed continuously by a monitoring system. The
main instrument installed is an inclinometer, based
on a pendulum hanged from the ceiling of the upper
vault at level H = 29.90 m. The pendulum consists of a
1.5 mm diameter steel wire with a brass mass attached:
the mass hangs in a water tank located at ground level,
to dampen pendulum motion, as shown in Figure 2a.
The position of the pendulum is permanently recorded
by two digital network cameras, carried on a steel
frame at level 1.1 m which in turn is fastened to the
floor.
The cameras permanently acquire pictures of the
wire and transmit them though the Internet every 10
minutes to the monitoring station, physically located
at the University of Trento. Using image recogni-
tion software, these images are real-time processed
in order to calculate the position of the wire with
respect to the background. The position of the pendu-
lum is returned in the form of two coordinates, x and
y, representing the shifts in direction West-East and
North-South, respectively, with respect to the inter-
section of the axes of the two cameras. In addition
to the pendulum inclinometer, the system records the
temperature at four thermocouples, two (T1 and T2)
installed on the outer surface of the masonry, the other
two (T3 and T4) on the inside, as depicted in Figure 2b.
The temperature data is acquired through a National
Instrument Field Point device, this also remotely con-
trolled by the monitoring station at the University of
Trento.
The server processes all the data, and publishes
the state of the Tower in real time on the website
http://smartstructures.org/portogruaro. As an exam-
ple, Figure 4a shows the time history of the temper-
ature θ 1 recorded at thermocouple T1 from October
2003 to September 2007, while Figure 4b plots the out
725
of plumb in x and y directions recorded in the same
period.
2.2 Discovery of historical documentation
In late 2005, the discovery of unpublished documents
threw new light on the recent history of the Tower.
A first step was the retrieval of a geometrical sur-
vey of the Tower carried out by De Goetzen in October
1962, immediately before the restoration work. One
of the designs reports an out of plumb of 740 mm
to the East and 760 mm to the North, respect to the
North-East edge of the balcony, at level 36.74 m.
The corresponding angles are 20.14 × 10−3rd East and
20.68 × 10−3rd North. The same survey also reports
an out of plumb of the spire of 150 mm North and
150 mm East, apparently taken over a difference of
level of 22.12 m: i.e. inclinations of 7.90 × 10−3rd East
and North. During an interview with the first author
of this paper, in September 2005, De Goetzen speci-
fied that these measurements were taken using a plumb
wire, thus with a precision that can be estimated in the
order of two centimeters. In the same interview, De
Goetzen said that a many documents were collected
few years ago by a student of the University of Padova,
as part of the research work in preparation for his
Thesis.
The first author of this paper met the former stu-
dent, now a professional engineer in November 2005.
Massimo Zanet, the author of the Thesis, had some
important things to say during the interview. First:
as reported in his Thesis (Zanet 1997), he surveyed
the Tower in July 1997, and measured an out of
plumb of 770 mm to the East and 760 mm North
using the same method and reference system as De
Goetzen. These values correspond to inclinations of
20.96 × 10−3rd East and 20.68 × 10−3rd North. Sec-
ond: during his research, Zanet was able to access
and examine the original designs of the restoration
work carried out between 1877 and 1879 by the civil
engineer Antonio Bon.
Based on Zanet’s advice, we were able to retrieve
these projects in the archives of the Municipality. From
analysis of the designs, it is clear that the XIX century
work was not just simple for repair, but rather a rad-
ical modification of the existing campanile. Figure 3
reproduces a sketch from Bon’s preliminary project,
dating 1877, where the original campanile is compared
with one of the new proposals. We note that the orig-
inal Tower was only 46,85 m tall, considerably lower
than the height we see today. In the technical report
that accompanies the design, Bon mentions an existing
tilt of the original tower, without specifying its extent.
Also, there is no mention of the fact that the new spire
would have to be built with a different inclination to
that of the tower below.
Table 1 compares the inclinations surveyed in 1962
by De Goetzen, in 1997 by Zanet and in 2001 by
Busetto. Even though the precision of the first two
measurements is probably quite low, it is remarkable
to observe that only one of the measurements is smaller
than the preceding values.

3 BAYESIAN IDENTIFICATION CONCEPT

Here our sources of information are two: the instru-

Figure 3. Original restoration project of the Tower by
Antonio Bon, dating 1877.
mental data recorded by the monitoring system, and
the historical documentation acquired. The issue is
how to exploit appropriately all this knowledge in order
to warn the owner of a possible hazardous situation as
it arises. Here we present first the general paradigm of
the method; then, in the next Section, we clarify the
practical application to the specific case.
The general paradigm of the method is to try to
recognize in real-time symptoms of a specific haz-
ardous scenario Sn (in this case: leaning in progress),
from a set of instrumental measurements M, using the
principle of Bayesian statistical analysis. Bayesian the-
ory of probability originates from Bayes’ well known
essay (Bayes, 1763). Reference works on the subject
are those by Jaynes (2003) and Skilling (1998) while
many modern specialized textbooks provide the reader
with a critical review and applications of this theory to
data analysis (see for instance those of Gregory, 2005,
and Sivia, 2006).
Say the monitoring system makes use of Ns
response sensors, labeled (s1 , s2 , . . . ., sNs ), each pro-
viding values for each of k time values (t1 , t2 , . . . ., tk ).
mj (t) identifies the value obtained at time t from
sensor sj , and Mk indicates the set of values provided
by all sensors to time tk . Furthermore the structure
is equipped with a set of environmental sensors (ther-
mocouples): θ(t) indicates the vector of environmental
measurements at time t and k the complete set
up to tk .

Table 1. Historical measurement of the Tower inclination.

Body inclination∗ Spire inclination

Authors Date x [rd] y [rd] x [rd] y [rd]

De Goetzen October 1962 20.14 × 10−3 20.68 × 10−3 7.90 × 10−3 7.90 × 10−3
Zanet July 1997 20.96 × 10−3 20.68 × 10−3 – –
Busetto & Romanin January 2001 21.50 × 10−3 24.49 × 10−3 6.01 × 10−3 12.65 × 10−3
∗
At the North-East corner.

726
 In practice, we can divide the domain of the pos-
sible structural response into a mutually exclusive
and exhaustive set of scenarios (S1 , S2 , . . . ., SNd ), each
defining the structural behavior under a specific con-
dition. The structural response n rj (n p, θ(t), t) for time
t and sensor sj in scenario Sn assuming environmen-
tal condition θ(t) is controlled by a certain number of
parameters (e.g.: stiffness, degradation), represented
by vector n p. The structural response is completely
defined by specifying a scenario and values for
the environmental conditions and for the correlated
parameter set. Here, as in the Bayesian model selection
theory (Gregory, 2005, Bretthorst, 1996), the discrete
scenario can be seen as a meta-parameter which qual-
itatively identifies the type of response function (e.g.:
constant, linear, exponential) which in turn is specified
by a parameter set.
Assuming scenario Sn to be correct and after appro-
priate selection of the scenario’s parameter n p, the
observational response can be expressed as:
where ej (t) is an error that accounts for both instru-
mental noise and the unavoidable imprecision of the
model assumed. Because from an epistemic point of
view these uncertainties are random, ej is modeled as
an uncorrelated zero-mean Gaussian noise. Its stan-
dard deviation n (σe )j can be assumed independent of
time, but generally changes with the sensor. Evidently
n
(σe )j changes with sensor type, but we may also
expect a dependency, for example, on sensor position
or precision.
Once measurements Mk and k become available
from the monitoring system, Bayes’ theorem allows
calculation of the updated, or posterior, probability for
each scenario Sn , from prior probability prob(Sn ), sce-
nario likelihood PDF(Mk |Sn ) and evidence PDF(Mk ),
using the following expression:
where PDF denotes the probability density function
of a random variable. Prior probabilities assigned to
each scenario reflect the initial judgment of the evalu-
ator, independently of the outcome of monitoring. On
the contrary, likelihood computation requires detailed
analysis of the predicted structural response in each
specific scenario. First, the likelihood of a single
sample mj (t) can be expressed as a function of the
parameters n p. The simplest idea is to assume that the
measurement is normally distributed around the model
prediction. We can formally express this concept as:
727
where the notation Normal{x; µ, σ} indicates a normal
distribution with mean value µ and standard deviation
σ calculated in x. As long as errors are assumed to be
uncorrelated for each time and sensor, the likelihood
for the whole measure set Mk is obtained combining
the likelihoods of all samples for all sensors and time
intervals recorded:
The likelihood of scenario Sn is then calculated
by marginalization of parameters n p, i.e. by integrat-
ing parameter likelihood on the whole domain Dn p,
using their prior distribution PDF(n p|Sn ) as weighting
function:
As the scenario set is complete and mutually exclu-
sive, evidence of measures Mk is simply obtained by
summing the scenarios:
Bayes’ theorem also allows an estimate of the pos-
terior distribution of the corresponding parameter n p,
using:
In summary, given a fresh set of measurements
Mk , this procedure allows real-time update of the
probability of each scenario using Eqs. (2) to (7).
4 APPLICATION TO THE CASE STUDY
The procedure presented in the previous Section is
stated in a general form, and applies to any type
of monitoring problem involving structural and envi-
ronmental measurements. Below, we explain in more
detail how this procedure applies to the Portogruaro
Civic Tower. In this case, the problem is to understand
as soon as possible if the Tower is still tilting. Using
the formal approach introduced in the previous Sec-
tion, two scenarios are possible: according to the first
scenario, S1 , the Tower inclination basically does not
change with time, any shift from the mean position
being due to daily and annual temperature changes;
according to the second, S2 , the Tower tilt is increasing,
with a trend we can assume linear.
In scenario S1 the compensated inclination ϕx (t) of
the tower in x direction (i.e. east-west) is constant and
equal to 1 ϕ0,x . Conversely, in scenario S2 the compen-
sated inclination is a linear function trend wj and an
offset 2 ϕ0,x . In equations, we can write for direction x
(i.e. West-East)
and similarly for direction y (i.e. North-South).
Response measurements are in this case the out of
line components, x and y, measured at the base of the
pendulum while vector θ collects the 4 thermocouple
measurements.
According to Eq.(1), we can see the out of line mea-
sured by the pendulum as the sum of a term depending
on tower inclination, a term linearly correlated with
the temperatures recorded and a noise:
where H = 29.90 m is the reference level for the out
of line measurement, ax is the linear transformation
that correlates the temperatures to the out of plumb,
while indices 1 or 2 indicate that this vector gener-
ally assumes different values in different scenarios.
As stated in the equation, the linear relation between
temperature and wire measurements is independent of
the scenario, and vectors aj , are considered as nuisance
parameters, that is, to be marginalized to gain infor-
mation on the relevant parameters wx and wy . Random
noise e is assumed to be Gaussian with zero mean
and standard deviation n (σe )j equal to 10 mm for every
sensor and scenario: this value, much larger than that
strictly related to the instrumental noise, also takes into
account the uncertainties of the correlation model.
It is worth noting that the two scenarios have dif-
fering degrees of complexity: scenario S2 involves
parameters wx and wy , so it can follow the measure-
ments more closely. Furthermore, scenario S1 can be
regarded as a special sub-case of scenario S2 , when wx
and wy are null. The reader might argue that, because
of this, the probability of scenario S2 will always be
greater than that of S1 . Actually, this is not neces-
sarily the case. Indeed, by tuning the parameters of
scenarioS2 , we can always obtain a better fit than that
related to scenario S1 . However, according to Bayesian
logic, the ratio between the best and the average fit (the
so-called Ockham factor) plays a fundamental role too
(Gregory 2005).
To implement the Bayesian updating procedure,
we have to define quantitatively the prior knowl-
edge. As mentioned, in October 2003, the probability
of the Tower tilting further was reputed very low:
we can formalize this initial perception assuming
for the tilting scenario a prior probability equal to
728
prob(S2 ) = 1/1000, and therefore prob(S1 ) = 99.9%
for the no-trend scenario. We must also define the prior
estimate of distribution of trends wx and wy . Based
on the limited documentation available in October
2003, the only hint is that at the time of construc-
tion, say XIV century, the body Tower was reasonably
straight. Compared to Busetto’s measurements, car-
ried out in 2001, and assuming a linear trend, we can
roughly estimate average shifts of −1.5 mm·year−1 in
both directions, corresponding to inclination trends of
−40.8 × 10−6rd ·year−1 . In summary, given the scarce
information available at that time, this was assumed
as the most likely value of current inclination trend,
if any. Of course, this information is very impre-
cise: to account for this uncertainty, we can assume
a prior scatter of, say, 4 mm·year−1 , corresponding to
an angular trend of 108.9 × 10−6rd .year−1 . Note that
this large scatter simply reflects our almost total prior
ignorance of the trends: the specific value selected is
not critical, in view of the final result of the updating
process.
Based on this prior information (i.e., without con-
sidering the historical information discovered in 2005)
we can apply Eq. (7) to update our knowledge using the
data acquired real-time by the monitoring system. The
light plot on Figure 4e shows how the monitoring data
modifies the system perception of having a trend. We
can see that during the first two years of monitoring the
probability of the trend scenario is always close to zero.
Only during the third year the monitoring information
starts overturning the initial perception, to the point
that in April 2006 the data is sufficient to convince the
system that the Tower is tilting.
Similarly, the light plots of Figures 4c and 4d show
the evolution of the distributions of trends wx and wy :
we see that the trend estimates, which are very uncer-
tain during the first two years, rapidly converge to more
reliable values.
The documentation acquired in 2005 radically
changed the initial judgment on the stability of the
Tower. Anyone with common sense can understand
that this new information supports the idea that the
Tower is tilting. However, the problem is how to quan-
tify the impact of this information on the probability
of there being further tilt.
The approach we followed is to cluster the histori-
cal information in a number of separate datasets, each
associated with a specific distribution, and use recur-
sively the same Eq. (7) to update the likelihood of
Scenario 2. The first dataset identified (labelled A)
consists in the three out of plumb measurements taken
on years 1962, 1997 and 2001. In these cases the mea-
surement techniques adopted suggest we consider a
noise of 20 mm. A second set (B) includes the com-
parison of inclination measurements of the spire taken
in 1962 and 2001; in this case a scatter of 50 mm
is assumed. The third data set (C) is basically the
Figure 4. Temperature measurements at thermocouple T4 (a); out of plumb measurements (b); posterior distribution of
angular trend wx (c); posterior distribution of angular trend wy (d); posterior probability of scenario S2 (e).

notion that the spire was rebuilt vertical in 1879; this
conjecture is assumed true with a likelihood of 50%; if
true, the standard deviation of the original inclination
is assumed equal to 8 × 10−3rd .
Table 2 summarizes the outcomes of the prior infor-
mation update using recursively the three historical
data sets. With respect to the initial judgment, prob(S2 )
increases up to 3.7% when all the historical datasets are
available, while the standard deviations of the angular
trend decrease to 17.3 × 10−6rd year−1 .
Mixing this new prior knowledge with the moni-
toring data, we obtain, day by day, the posterior trend
distributions and the probability of scenario S2 , plotted
in bold in Figures 4c, 4d and 4e.
In detail, Figure 4d illustrates that, considering the
historical data, the system would theoretically become

729
Table 2. Update of prior information based on the historical information.
wx [rd year−1 ]
Authors prob(S2) mean value
Prior knowledge 0.1% −40.8 × 10−6
Considering dataset A 2.0% −30.8 × 10−6
Considering dataset B 1.3% −27.0 × 10−6
Considering dataset C 3.7% −29.2 × 10−6
* At the North-East corner.
aware of the trend as early as September 2005; there-
fore, we can say that the historical information is
roughly equivalent to 6 months of instrumental mon-
itoring. Note that, because of the consistency of the
Bayesian logic, the outcome of the updating process
is invariant with respect to the order of information
processing. Thus, the effect of finding the historical
documentation is to shift from the light plot to the bold
plot at the time this information is available, which in
our case is November 2005.
Worthy of note is that after three years of moni-
toring, the chance of continuous tilt increase of the
Tower is close to certainty, regardless of the his-
torical data assumed. In the same way the angular
trends converge to similar values after the first years
of monitoring. This means that, after a certain time,
the information acquired by the monitoring system
becomes dominant over any type of prior knowl-
edge. In detail, the trends identified at the end of
the monitoring are wx = −34.6 × 10−6rd ·year−1 and
wy = −52.1 × 10−6rd year−1 with a standard deviation
of 6.7 × 10−6rd year−1 : in terms of out of plumb respect
to level H = 29.90 m, this is to say 1.03 mm year−1
West-East and 1.56 mm year− South-North, with a
scatter of only 0.2 mm year−1 .
5 CONCLUSIONS
In the case presented, the joint use of instrumen-
tal monitoring and historical documentation allowed
timely recognition of signs of an ongoing tilt and accu-
rate calculation not only of the mean value of the
inclination trend, but also the level of reliability of this
information. The general Bayesian methodology intro-
duced deals flexibly with all the uncertainties involved
in the recognition problem: measurement noise, uncer-
tainness on the model and inaccurate prior infor-
mation. Moreover, it lets us quantitatively combine
information of completely different nature, including
incomplete datasets or unproved conjectures.
Based on this logical path, the University con-
cluded, in September 2006, that a progressive tilt
was very likely. Following these conclusions, the
Department of Public Works of the Municipality of
730
wy [rd year−1 ]
scatter mean value scatter
108.9 × 10−6 −40.8 × 10−6 108.9 × 10−6
18.0 × 10−6 −58.5 × 10−6 18.0 × 10−6
17.8 × 10−6 −57.3 × 10−6 17.8 × 10−6
17.3 × 10−6 −60.1 × 10−6 17.3 × 10−6
Portogruaro initiated an investigation campaign to ana-
lyze the state of the Tower foundations and to plan a
reinforcement to prevent any possible future risk of
collapse. The work is currently under way.
ACKNOWLEDGEMENTS
The installation and operation of the monitoring sys-
tem was funded by the Municipality of Portogruaro.
The Bayesian algorithm reported in this paper has been
developed with the financial contribution of the Ital-
ian Ministry of Education (MIUR). The authors wish
to thank all those who contributed to the success of
this work, and specifically Stefano Toffaletti, Marco
Molinari and Nicola Naldo.
REFERENCES
Bayes, T. 1763. An Essay toward solving a Problem in the
doctrine of Chances. Philos. Trans. R. Soc. London 53:
pp. 370–418.
Beck, J.L. & Au, S.-K. 2002. Bayesian Updating of Structural
Models and Reliability using Markov Chain Monte Carlo
Simulation. J. Eng. Mech. 128(4): pp. 380–391.
Beck, J.L. & Katafygiotis, L.S. 1998. Updating Models And
Their Uncertainties, I: Bayesian Statistical Framework. J.
Eng. Mech. 124(2): pp. 455–461.
Bretthorst, G.L. 1996. An introduction to model selection
using probability theory as logic. In G.R. Heidbreder (ed),
Maximum Entropy and Bayesian Methods, Springer New
York, NY.
Busetto, A. & Romanin, L. 2001. Studio preliminare per
l’individuazione del percorso atto a definire lo stato
patologico della Torre Civica, Technical report for the
Municipality of Portogruaro.
Collins, M.P. & Mitchell, D. 1991. Prestressed Concrete
Structures. Englewood Cliffs – New Jersey: Prentice Hall
Gregory, P. 2005. Bayesian Logical Data Analysis for the
Physical Sciences. Cambridge: Cambridge University
Press.
Jaynes, E.T. 2003. Probability Theory: The Logic of Science.
Cambridge, UK: Cambridge University Press.
MacKay, D.J.C. 2003. Information Theory, Inference and
Learning Algorithms. Cambridge, UK: Cambridge Uni-
versity Press.
Measures, R.M. 2001. Structural Monitoring with Fiber
Optic Technology. NY: Academic Press.
Metropolis, N., Rosenbluth, A., Rosenbluth, M., Teller, A.
& Teller, E. 1953. Equation of state calculation by
fast computing machines. J. of Chemical Physics 21:
pp. 1087–1092.
Molteni, M. 2003. Torre Civica Duomo di S. Andrea: Cam-
pagna di studi, analisi, indagini e prove sperimentali per
la valutazione dello stato patologico della torre campa-
naria: Relazione tecnica conclusiva. Technical report for
the Municipality of Portogruaro.
Papadimitriou, C., Beck, J.L. & Katafyogiotis, L.S. 1997.
Asymptotic Expansion for Reliability and Moments of
Uncertain Systems. J. Eng. Mech. 123(12): pp. 380–391.
Sivia, D.S. 2006. Data Analysis: a Bayesian Tutorial. Oxford:
Oxford University Press.
731
Skilling, J. 1998. Probabilistic Data Analysis; An intro-
ductory Guide, Journal of Microscopy 190(1–2),
pp. 28–36.
Sohn, H. & Law, K-H. 2000. Bayesian Probabilistic Dam-
age Detection of a Reinforced Concrete Bridge Column.
Earthquake Engineering and Structural Dynamics 29(8),
pp. 1131–1152.
Zanet, M. 1997. Studio storico strutturale della Torre Civica
della città di Portogruaro. Laurea Thesis, Facoltà di
Ingegneria, University of Padova.
Zonta, D., Pozzi, M., Colombo, A. & Bursi, O.S. 2006.
Industrial Design of Instrumented PRC Elements for the
Condition Monitoring of Civil Infrastructure. In: Smart
Structures and Materials 2006: Smart Sensor Monitor-
ing Systems and Applications. SPIE proc., Vol. 6167,
pp. 131–141.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Author index

Abdessemed-Foufa, A.A. 275 Bulbuk, M. 169 De Luca, A. 565, 1261

Abruzzese, D. 393 Busetto, A. 723 De Marchi, P.M. 481
Acito, M. 875 de Martino, G. 1527
Acler, E. 843 Caballé, F. 643 de Martino, R. 1527
Adam, J.M. 349 Calado, L. 817 De Matteis, G. 1225
Adekunle, T.O. 991 Campos-Costa, A. 311 De Roeck, G. 311
Adriaenssens, S. 683, 1175, 1383 Cancino, C. 1095 De Stefano, A. 575
Agneloni, E. 1007 Capozucca, R. 297 de Vent, I.A.E. 689
Alessandri, C. 1287 Carbone, I. 809 Del Grosso, A. 441
Alexandre, J. 1149 Cardani, G. 875 Delinikolas, N. 621, 1025, 1149
Algeri, C. 1215 Carocci, C.F. 1519 Delizia, F. 1527
Anamaterou, L. 1149 Carpinteri, A. 367, 375 Dewoolkar, M.M. 1369
Andreini, M. 1481 Casadei, P. 1007 Domenge, J. 643
Anese, G.A. 723 Casapulla, C. 1191 Dourakopoulos, J. 621
Ansell, M.P. 469 Casarin, F. 403 Drdacky, M. 109
Antonelli, A. 899 Casiello, S. 259 Durand, O. 1149
Antonini, M. 217 Castori, G. 921 D’Aniello, M. 431
Araiza, J.C. 1297 Cattari, S. 517 D’Ayala, D. 331, 489
Araya, M. 1399 Cecchi, A. 555 D’Ayala, D.F. 47
Arêde, A. 653 Cei, C. 1481 Dérobert, X. 1149
Askew, P. 109 Cennamo, C. 867 Díez, J. 851
Augenti, N. 547, 1475 Ceraldi, C. 915
Chesi, C. 1253 Eboli, A. 1225
Baldessari, C. 843 Chiaia, B.M. 867 Eimanian, J. 469
Baratta, A. 321, 799 Chiaverini, I. 555 Elsen, J. 965
Barthel, R. 183, 1423 Chorafa, E. 1025 Enrione, D. 575
Bartoli, G. 899 Chryssopoulos, D. 1149 Erdem, A. 1141, 1167
Beckh, M. 183 Chávez, M. 825
Bednarz, Ł. 357 Çiftçi, A. 267, 999 Fabbrocino, G. 715
Bei, G.E. 891 Ciocchini, E. 481 Fabregat, P.R. 533
Benouar, D. 275 Cizer, O. 965 Faella, G. 1391
Bertolini Cestari, C. 941 Clifton, G.A. 89 Faggiano, B. 817
Betti, M. 601, 899 Colanzi, F. 1225 Feio, A.O. 833
Bettio, C. 403 Condoleo, P. 635, 1331 Ferrari, R. 583
Binda, L. 109, 375, 635, 875, Corbi, I. 321, 799 Figeys, W. 1087
1331, 1345 Corbi, O. 303, 321, 799 Fiorino, L. 431
Blasi, C. 1271 Corradi, M. 981 Fodde, E. 1415
Blersch, D. 1277 Cosenza, E. 715 Friedman, D. 129
Block, P. 593 Costa, A.G. 653 Fuchs, C. 1181
Blondet, M. 1095 Cotta, Robert S. 227
Bocca, P. 339 Coïsson, E. 1271 Galindo, J.A. 243
Borri, A. 921, 981 Croci, G. 1445 Galván, F. 1297
Bortolotto, S. 481 Cruz, P.J.S. 459, 507 Garbin, E. 755, 763
Bozzetti, A. 1445 Curti, E. 1237 García, D. 851
Branco, J.M. 459, 507 Côte, Ph. 1149 Gariup, G. 533
Brandonisio, G. 565, 1261 Gattesco, N. 1035
Brignola, A. 441 Dalla Benetta, M. 755 Gaudini, G. 403
Brookes, C.L. 497 De Blasi, G. 199 Gemert, D.V. 1087
Bruccoleri, D. 217 de Bouw, M. 683, 1175, 1383 Georganis, F. 1149
Bucur-Horváth, I. 169 de Brito, J. 1101 Geremei, F. 1277
Budelmann, H. 1043 de Felice, G. 541, 809 Giannantoni, A. 981

733
Giardina, G. 413 Kelley, S.J. 25 Mazzolani, F.M. 383, 817, 1157,
Ginocchio, F. 1095 Kemper, J.H. 1079 1225
Giordano, A. 1391 Kim, N.H. 861 McCabe, B.P. 1357
Giorgi, L. 191 Kirizsán, I. 143, 1375 McCombie, P. 781
Giráldez, P. 643 Klemenc, I. 789 Mele, E. 565, 1261
Giuriani, E. 13, 413 Kolymbas, D. 109 Meli, R. 825
Giustina, I. 957 Korkmaz, A. 1303 Mezzi, M. 1061
González, J.L. 643 Krstevska, L. 383 Miglioli, M. 1277, 1287
González, R. 643 Kruschwitz, S. 109 Milani, E. 1277
Goñi, R. 705 Kutnyi, A. 183 Mileto, C. 1467
Gramatikov, K. 383 Kwiecień, A. 931 Miller, J.D. 1133
Grande, E. 1017 Köck, B. 235 Miltiadou, A. 621
Grazzini, A. 339 Köpp, Ch. 109 Miltiadou-Fezans, A. 71, 1025,
Gu, X.L. 1537 1149
Guadagnuolo, M. 1391 Lacidogna, G. 367, 375 Minch, M.Y. 1455
Gubana, A. 949 Laefer, D.F. 1207 Miranda, L. 875
Guedes, J. 875 Lagomarsino, S. 517 Miri, M. 909
Guedes, J.M. 653 Lakshmipriya, T. 1491 Modena, C. 403, 755, 763, 1345
Landolfo, R. 383, 431 Modi, S.M. 175
Harkness, J. 781 Lanza, S. 109 Morales, K. 1095
Hastemoglu, H. 1303 Lanzoni, L. 1277, 1287 Morer, P. 705
Hayden, N.J. 1369 Laorenza, C. 715 Mormone, V. 915
Heath, A. 781 Lauriks, L. 683 Mosseri, A. 1363
Heinemann, H.A. 151 Lawrence, R.M.H. 885 Mouzakis, H.P. 621
Hendrickx, R. 3, 973 Lee, Y.W. 861 Munari, M. 1277
Hennen, Ch. 109 Li, S. 393 Mundell, C. 781
Henriques, D.F. 1101 Li, X. 1537 Muñoz, A.M. 243
Herle, I. 109 Linscott, K. 121
Hickey, S. 1207 Lopez Izquerdo, N. 1277 Nagai, T. 1399
Hirai, T. 629 Loughlin, B. 1207 Nascosi, M. 1287
Hobbelman, G.J. 689 Lourenço, P.B. 95, 311, 739, 771, Niglio, O. 251
Hoirisch, M. 137 833, 1345 Nijland, T.G. 151
Holzer, S.M. 235 Lucchin, F. 403 Norden, J.F. 1107
Hong, S.G. 861 Lutenegger, A.J. 1079 Nunes, L. 1101
Hughes, T.G. 909 Lutman, M. 1311
Hwang, J.K. 861 Ochsendorf, J. 593
Machado, J.S. 833 Odeyale, T.O. 991
Ignoul, S. 525, 747, 1087, 1345 Macorini, L. 1035 Oliveira, B.T. 283
Ilharco, T. 653 Madhoushi, M. 469 Oliveira, D. 771
Imbimbo, M. 1017 Maeda, T. 629 Oliveira, D.V. 739, 1345
Indirli, M. 1277, 1287 Maierhofer, Ch. 109 Olosz, E. 473
Invernizzi, S. 367, 375, 941 Makay, D. 663 Orduña, A. 1297
Ivorra, S. 349 Mammana, O. 383 Ospina, Y.M. 243
Iwaki, C. 1095 Manuello, A. 375 Ottoni, F. 697
Marcari, G. 715 Ozakin, R. 1141, 1167
Jain, J. 289 Mariani, G. 1481 O’Farrell, S. 1207
Jankowski, R. 931 Mariani, M. 1123 O’Mahony, G. 1207
Jasieńko, J. 357 Marini, A. 13, 413 O’Sullivan, M. 207
Jäger, W. 1181 Marzi, T. 941
Jeong, S.J. 861 Marzo, A. 817 Pallarés, F. 349
Joo, S.J. 861 Marzullo, M. 1061 Pane, A. 1527
Masoero, E. 867 Panizza, M. 763
Kakkinou, S. 1025 Massarelli, T.M. 1461 Panão, A. 817
Kalagri, A. 1025, 1149 Matracchi, P. 191 Papadopoulos, K. 1113
Kaup, J. 1245 Matteo, J.A. 1245 Paredes, J.A. 243
Kayser, C. 1423 Maus, H. 1423 Parisi, M.A. 1253
Keersmaekers, R. 1053 Mazzanti, S. 1287 Parodi, S. 1237

734
Passerini, A. 555 Scaini, S. 867 Valletta, G. 1061
Paupério, E. 653 Schoonjans, Y. 1383 Valluzzi, M.R. 33, 755, 763
Peng, B. 1537 Schueremans, L. 525, 747, 1321, Van Balen, K. 3, 965, 973
Peña, F. 95 1345 Van Gemert, D. 525, 747, 965,
Piazza, M. 459, 507, 843, 1253 Screpanti, A. 1277 973, 1053
Pini, D. 1287 Sextos, A. 611 van Hees, R.P.J. 151, 689
Pitacco, I. 533 Seçkin, N. 267 Van Rickstal, F. 1053
Pittas, M. 1499 Shang, D.F. 1537 Vargas, J. 1095
Plizzari, G.A. 957 Shi, Y. 489 Varum, H. 507
Podestà, S. 441, 1237 Sikka, S. 1069 Vasconcelos, G. 95, 771
Popa, I. 169 Silva, R.A. 739 Vassallo, E. 259
Porter, D.W. 1369 Simonini, G. 1277, 1287 Vegas López-Manzanares, F.
Portioli, F. 383 Skłodowski, M. 423 1467
Posen, D. 1053 Smars, P. 1199 Vendrell, M. 643
Pozzi, M. 723 Sorrentino, L. 217, 1431 Verbeeck, B. 683, 1175
Preciado, A. 1297 Sparks, S.P. 451 Verstrynge, E. 525, 747, 1321,
Proença, J.M. 817 Speranza, E. 1509 1345
Puglisi, C. 1277 Speranzini, E. 981 Vignoli, A. 601
Sperbeck, S.T. 1043 Virsta, A. 109
Qing, C. 1441 Stacey, M.S. 1357 Virág, J. 169
Stammers, C. 489 Viskovic, A. 1445
Ramos, L.F. 95, 311 Stylianidis, K.-A. 611 Vitti, M. 1445, 1509
Resemini, S. 441, 517 Swailes, T. 207
Ribeiro, R.T.M. 137 Swann, J.D. 1339 Walker, P. 781, 885
Rinaldis, D. 799 Szabó, B. 143, 473, 663, 1375 Weiss, P. 789
Riotto, G. 441 Szołomicki, J.P. 1455 Wevers, M. 747
Riva, P. 413
Wouters, I. 683, 1175, 1383
Rizzi, E. 583 Taliercio, A. 635
Roca, P. 643 Tardini, C. 1253
Romanelli, F. 1277 Tashkov, Lj. 383 Yamada, S. 1399
Romano, A. 547 Teston, S. 1287 Yamamoto, T. 629
Ruocci, G. 575 Thelin, C. 121 Yao, L. 393
Russo Ermolli, E. 915 Tiraboschi, C. 1331 Yasmine, J. 1407
Russo, V. 259 Tocci, C. 1431, 1519 Yeomans, D.T. 63
Toker, S. 1303 Yergün, U. 999
Sacco, E. 1017 Tomasi, R. 843 Yuan, J. 393
Saisi, A. 1331 Tomasoni, E. 47
Sala, E. 957 Tomaževič, M. 109, 789 ˛ B. 931
Zajac,
San-José, J.T. 851 Touliatos, P. 621 Zangheri, F. 481
San-Mateos, R. 851 Toumbakari, E.E. 673 Zanon, P. 723
Santaniello, R. 1261 Tsai, P.-H. 331 Zanzi, L. 1331
Sanyal, S. 161 Tulliani, J.M. 941 Zarogianni, E. 1025
Sassu, M. 1481 Ziagou, A. 1025
Savvidou, M. 1149 Ungaro, L. 1445, 1509 Zonta, D. 723

735
STRUCTURAL ANALYSIS OF HISTORIC CONSTRUCTION
PROCEEDINGS OF THE SIXTH INTERNATIONAL CONFERENCE ON STRUCTURAL
ANALYSIS OF HISTORIC CONSTRUCTION, 2–4 JULY, BATH, UNITED KINGDOM

Structural Analysis of Historic

Construction
Preserving Safety and Significance

VOLUME 2

Editors
Dina D’Ayala & Enrico Fodde
Department of Architecture & Civil Engineering,
University of Bath, Bath, UK
CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business

© 2008 Taylor & Francis Group, London, UK

Typeset by Charon Tec Ltd (A Macmillan Company), Chennai, India

Printed and bound in Great Britain by Cromwell Press Ltd, Towbridge, Wiltshire

All rights reserved. No part of this publication or the information contained herein may be reproduced, stored
in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, by photocopying,
recording or otherwise, without written prior permission from the publishers.

Although all care is taken to ensure integrity and the quality of this publication and the information herein, no
responsibility is assumed by the publishers nor the author for any damage to the property or persons as a result
of operation or use of this publication and/or the information contained herein.

Published by: CRC Press/Balkema

P. O. Box 447, 2300 AK Leiden, The Netherlands
e-mail: Pub.NL@taylorandfrancis.com
www.crcpress.com – www.taylorandfrancis.co.uk – www.balkema.nl

ISBN Set: 978-0-415-46872-5 (Set of 2 Volumes + CD-Rom)

ISBN Vol.1: 978-0-415-48109-0
ISBN Vol.2: 978-0-415-48107-6
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Table of Contents

Preface XV

Committees XVII

Conference Sponsors XIX

VOLUME 1

Invited papers
Preservation of workmanship or workmanship for preservation 3
K. Van Balen & R. Hendrickx

Experiences from the Northern Italy 2004 earthquake: Vulnerability assessment and
strengthening of historic churches 13
E. Giuriani & A. Marini

20th century curtain walls – loss of redundancy and increase in complexity 25

S.J. Kelley

Strengthening of masonry structures with Fibre Reinforced Plastics: From modern

conception to historical building preservation 33
M.R. Valluzzi

The structural behaviour of masonry vaults: Limit state analysis with finite friction 47
D.F. D’Ayala & E. Tomasoni

Repairs to historic timber structures: Changing attitudes and knowledge 63

D.T. Yeomans

A multidisciplinary approach for the structural restoration of the Katholikon of Dafni

Monastery in Attica Greece 71
A. Miltiadou-Fezans

Structural interventions in English Cathedrals 89

G.A. Clifton

Monastery of Salzedas (Portugal): Intervention in the cloister and information management 95

P.B. Lourenço, L.F. Ramos, G. Vasconcelos & F. Peña

Cultural Heritage protection against flood – A European FP6 research project 109
Ch. Maierhofer, Ch. Köpp, S. Kruschwitz, M. Drdacky, Ch. Hennen, S. Lanza, M. Tomaževič,
D. Kolymbas, I. Herle, A. Virsta, L. Binda & P. Askew

Definition of significance and attribution of value

Structural definition and comparison of early medieval roof structures 121
C. Thelin & K. Linscott

V
Analysis of archaic fireproof floor systems 129
D. Friedman
Intervention limits in floor wood structure restoration of a historical building 137
R.T.M. Ribeiro & M. Hoirisch
The safety of historic load-bearing structures 143
B. Szabó & I. Kirizsán
Concrete: Too young for conservation? 151
H.A. Heinemann, R.P.J. van Hees, T.G. Nijland
Significance of a Mughal Mall and its role in achieving safety 161
S. Sanyal
Historical constructions – Authenticity and adaptation to the modern demands 169
I. Bucur-Horváth, I. Popa, M. Bulbuk & J. Virág
Champaner-Pavagadh – Managing conflicts – A conservation challenge 175
S.M. Modi

History and documentation

Construction and structural behavior of Vladimir Suchov’s Nigres tower 183
M. Beckh, R. Barthel & A. Kutnyi
New studies on Brunelleschi’s Dome in Florence 191
L. Giorgi & P. Matracchi
Wells Cathedral: The crossing and the central tower. A new hypothesis on the constructional
sequence and dating of the central tower 199
G. De Blasi
A study of historical test data for better informed assessment of wrought-iron structures 207
M. O’Sullivan & T. Swailes
Structural interpretation of post-earthquakes (19th century) retrofitting on the
Santa Maria degli Angeli Basilica, Assisi, Italy 217
L. Sorrentino, D. Bruccoleri & M. Antonini
Engineering survival and success: The contributions of historic structural features in the
spire and tower of Salisbury Cathedral 227
Robert S. Cotta
Structure of baroque church roofs in Bavaria 235
S.M. Holzer & B. Köck
Brick arch bridges in the High Cauca Region of Colombia: A forgotten construction tradition 243
J.A. Galindo, J.A. Paredes, A.M. Muñoz & Y.M. Ospina
From empirical engineering to science-based engineering: The assessment by three
mathematicians of the “San Pietro” Dome (1742) 251
O. Niglio
From static history to restoration issues: The Gesú Nuovo church in Naples (Italy) 259
S. Casiello, V. Russo & E. Vassallo
Building techniques of the Zeytinburnu Military Factory built in the 19th century 267
A. Çiftçi & N. Seçkin
Investigation of the 1716 Algiers (Algeria) earthquake from historical sources 275
A.A. Abdessemed-Foufa & D. Benouar
The Morro da Queimada Archaeological Park, Ouro Preto, MG – Brazil 283
B.T. Oliveira

VI
Structure as a document 289
J. Jain

Monitoring, non destructive evaluation and testing

Historic multiple-leaf masonry wall models under compression and cyclic shear loads 297
R. Capozucca
The masonry vaults: Geometry definition and possible approaches to the static analysis 303
O. Corbi
Damage identification in masonry structures with vibration measurements 311
L.F. Ramos, P.B. Lourenço, G. De Roeck & A. Campos-Costa
Stress analysis of masonry structures: Arches, walls and vaults 321
A. Baratta, I. Corbi & O. Corbi
Evaluation of historic Dieh-Dou buildings in Taiwan 331
P.-H. Tsai & D. D’Ayala
Experimental techniques for the evaluation of the long-term behaviour of masonry
strengthening materials 339
P. Bocca & A. Grazzini
Experimental and numerical studies on the belltower of Santa Justa y Rufina (Orihuela-Spain) 349
S. Ivorra, F. Pallarés & J.M. Adam
The analysis of the strains and stresses of the reinforced brick vaults 357
J. Jasieńko & Ł. Bednarz
Cracking simulation of brick-masonry elements subjected to the double flat-jack test 367
A. Carpinteri, S. Invernizzi & G. Lacidogna
Numerical simulation and monitoring of the Cathedral of Syracuse in Sicily 375
A. Carpinteri, S. Invernizzi, G. Lacidogna, A. Manuello & L. Binda
Shaking table tests on the large scale model of Mustafa Pasha Mosque without and
with FRP 383
L. Krstevska, Lj. Tashkov, K. Gramatikov, R. Landolfo, O. Mammana, F. Portioli &
F.M. Mazzolani
Integrated modeling method for dynamic behavior of ancient pagodas 393
J. Yuan, L. Yao, S. Li & D. Abruzzese
Monitoring and strengthening interventions on the stone tomb of Cansignorio della Scala,
Verona, Italy 403
G. Gaudini, C. Modena, F. Casarin, C. Bettio & F. Lucchin
Seismic behaviour of barrel vault systems 413
A. Marini, G. Giardina, P. Riva & E. Giuriani
Assessment of material degradation based on microcores testing 423
M. Skłodowski
Structural performance of riveted connections in historical metal structures 431
M. D’Aniello, L. Fiorino & R. Landolfo
Diagnostic tests and numerical simulations for the preservation of two stone stairways in the
historic centre of Genoa (Italy) 441
A. Brignola, A. Del Grosso, S. Podestà, S. Resemini & G. Riotto
Evaluation of iron and steel in historic bridges 451
S.P. Sparks

VII
Diagnosis and analysis of two king-post trusses 459
J.M. Branco, P.J.S. Cruz & M. Piazza
Non destructive and laboratory evaluation of strength of decayed wood members in a historic
construction located in Gorgan (North of Iran) 469
M. Madhoushi, J. Eimanian & M.P. Ansell
The structural behaviour of spires 473
E. Olosz & B. Szabó
In-situ and laboratory investigations on materials and structures for the static improvement:
“Torre Quadrata del Circo Romano” (Milan) 481
S. Bortolotto, E. Ciocchini, F. Zangheri & P.M. De Marchi
Dynamic multi-body behaviour of historic masonry buildings models 489
D. D’Ayala, Y. Shi & C. Stammers
Assessment and strengthening masonry arch bridges 497
C.L. Brookes
Field tests of a timber queen-post truss and numerical analysis 507
J.M. Branco, P.J.S. Cruz, M. Piazza & H. Varum

Assessment and analytical techniques

Modelling of vaults as equivalent diaphragms in 3D seismic analysis of masonry buildings 517
S. Cattari, S. Resemini & S. Lagomarsino
Modelling of damage accumulation in masonry subjected to a long-term compressive load 525
E. Verstrynge, S. Ignoul, L. Schueremans & D. Van Gemert
A method for studying multi-body dynamics and its applications to masonry structures analysis 533
G. Gariup, I. Pitacco & P.R. Fabregat
Load-carrying capacity of multi-span masonry arch bridges having limited ductility 541
G. de Felice
Assessment of irregular masonry walls 547
N. Augenti & A. Romano
Geometrical and static aspects of the Cupola of Santa Maria del Fiore, Florence (Italy) 555
A. Cecchi, I. Chiaverini & A. Passerini
Some considerations on out-of-plane collapse modes of masonry walls 565
G. Brandonisio, E. Mele & A. De Luca
Innovative techniques for structural assessment: The case of the Holy Shroud Chapel in Turin 575
A. De Stefano, D. Enrione & G. Ruocci
On the theory of the ellipse of elasticity as a natural discretisation method in the design of
Paderno d’Adda Bridge (Italy) 583
R. Ferrari & E. Rizzi
Lower-bound analysis of masonry vaults 593
P. Block & J. Ochsendorf
Assessment of seismic resistance of a basilica church: Modelling, analysis and
strengthening proposal 601
M. Betti & A. Vignoli
Estimation of the seismic history of the city of Thessaloniki through back analysis of its
Byzantine land walls 611
K.-A. Stylianidis & A. Sextos

VIII
Earthquake based condition monitoring of the Katholikon of Dafni Monastery 621
H.P. Mouzakis, A. Miltiadou, N. Delinikolas, P. Touliatos & J. Dourakopoulos
Continuous and discontinuous modeling of the structures in Bayon temple, Angkor 629
T. Maeda, T. Yamamoto & T. Hirai
Safety assessment of temple E7 in Mỹ So,n, Viet Nam 635
P. Condoleo, A. Taliercio & L. Binda
Construction process, damage and structural analysis. Two case studies 643
R. González, F. Caballé, J. Domenge, M. Vendrell, P. Giráldez, P. Roca & J.L. González
Analysis and diagnosis of timber structures of Porto historical centre 653
T. Ilharco, J.M. Guedes, A. Arêde, E. Paupério & A.G. Costa
Baroque roof structures in Transylvania – Research and analyses 663
D. Makay & B. Szabó
The Athens Parthenon: Analysis and interpretation of the structural failures of the orthostate of
the northern wall 673
E.E. Toumbakari
Structural assessment of the Winter Garden of the Royal Glasshouses of Laeken, Belgium 683
L. Lauriks, I. Wouters, S. Adriaenssens, M. de Bouw & B. Verbeeck
Towards a systematic diagnosis of structural damage 689
I.A.E. de Vent, R.P.J. van Hees & G.J. Hobbelman
From geometrical and cracks survey to static analysis method: The case study of Santa Maria
del Quartiere dome, in Parma (Italy) 697
F. Ottoni
A benchmarking study of the analysis of non-reinforced structures applied to the structural
behavior of domes 705
P. Morer & R. Goñi
Structural analysis of the Caserta Royal Palace timber roof connections 715
G. Fabbrocino, G. Marcari, C. Laorenza & E. Cosenza
Real-time Probabilistic health monitoring of the Portogruaro Civic Tower 723
D. Zonta, M. Pozzi, P. Zanon, G.A. Anese & A. Busetto

Author index 733

VOLUME 2
Experimental studies
On the strengthening of three-leaf stone masonry walls 739
R.A. Silva, D.V. Oliveira & P.B. Lourenço
Long-term behaviour of historical masonry – A quantitative acquisition of the damage evolution 747
E. Verstrynge, L. Schueremans, D. Van Gemert, S. Ignoul & M. Wevers
Experimental assessment and modelling of in-plane behaviour of timber floors 755
M.R. Valluzzi, E. Garbin, M. Dalla Benetta & C. Modena
Bond behaviour of CFRP and GFRP laminates on brick masonry 763
M. Panizza, E. Garbin, M.R. Valluzzi & C. Modena
Experimental shear behavior of stone masonry joints 771
G. Vasconcelos, P.B. Lourenço & D. Oliveira

IX
Large scale testing of drystone retaining walls 781
C. Mundell, P. McCombie, A. Heath, J. Harkness & P. Walker
Seismic isolation, strengthening of walls with CFRP strips and heritage masonry buildings 789
M. Tomaževič, I. Klemenc & P. Weiss
Experimental survey on seismic response of masonry models 799
A. Baratta, I. Corbi, O. Corbi & D. Rinaldis
Bond performance of fiber reinforced grout on brickwork specimens 809
I. Carbone & G. de Felice
Experimental analysis of rectangular shaped sleeve connectors for composite
timber-steel-concrete floors: Bending tests 817
L. Calado, J.M. Proença, A. Panão, F.M. Mazzolani, B. Faggiano & A. Marzo
Shaking table testing of a typical Mexican colonial temple 825
M. Chávez & R. Meli
Capacity of a traditional timber mortise and tenon joint 833
A.O. Feio, P.B. Lourenço & J.S. Machado
Behaviour of refurbished timber floors characterized by different in-plane stiffness 843
M. Piazza, C. Baldessari, R. Tomasi & E. Acler
Stone masonry walls: Strengthening with TRM (I) 851
J.T. San-José, D. García, R. San-Mateos & J. Díez
The effect of friction joint and Gongpo (bracket set) as an energy dissipation in Korean
traditional wooden structure 861
J.K. Hwang, S.G. Hong, N.H. Kim, Y.W. Lee, S.J. Jeong & S.J. Joo
Optimization of cutting processes in archaeological sites 867
C. Cennamo, B.M. Chiaia, E. Masoero & S. Scaini
Experimental and numerical study on the application of the flat-jack tests to masonry walls 875
M. Acito, L. Binda, G. Cardani, J. Guedes & L. Miranda
The impact of the water/lime ratio on the structural characteristics of air lime mortars 885
R.M.H. Lawrence & P. Walker
Soil mortar’s evaluation characteristics for their use in conservation practice 891
G.E. Bei
Experimental and numerical analyses for static retrofitting intervention on the “Cappella
dei Principi” in Firenze 899
A. Antonelli, G. Bartoli & M. Betti
Comparison the effect of different repair method applying on masonry arch bridges 909
M. Miri & T.G. Hughes

Use of traditional, alternative and innovative materials

Restoring of timber structures: Connections with timber pegs 915
C. Ceraldi, V. Mormone & E. Russo Ermolli
Research on composite strengthening of historical housebuilding: Retrofitting intervention
for masonry arches and vaults 921
A. Borri & G. Castori
Static and dynamic properties of a flexible joint working in cracked historical masonries 931
A. Kwiecień, B. Zaja˛c & R. Jankowski
Nanotechnologies applied to the restoration and maintenance of wooden built heritage 941
C. Bertolini Cestari, T. Marzi, S. Invernizzi & J.M. Tulliani

X
Cross laminated timber panels to strengthen wood floors 949
A. Gubana
Lime mortar with natural pozzolana: Historical issues and mechanical behavior 957
E. Sala, I. Giustina & G.A. Plizzari
Blended lime-cement mortars for conservation purposes: Microstructure and strength
development 965
O. Cizer, K. Van Balen, D. Van Gemert & J. Elsen
Assessing workability of mortar by means of rheological parameters and desorptivity 973
R. Hendrickx, K. Van Balen & D. Van Gemert
Consolidation and reinforcement of stone walls using a reinforced repointing grid 981
A. Borri, M. Corradi, E. Speranzini & A. Giannantoni
Innovative and sustainable local material in traditional African architecture – Socio cultural
dimension 991
T.O. Odeyale & T.O. Adekunle
Imported construction materials and techniques in 19th century Ottoman architecture 999
U. Yergün & A. Çiftçi
Advance composites applications on historical structures in Italy: Case studies and future
developments 1007
P. Casadei & E. Agneloni

Novel conservation engineering techniques, restoration and strengthening

FRP-strengthening of masonry structures: Effect of debonding phenomenon 1017
E. Grande, M. Imbimbo & E. Sacco
Methodology for in situ application of hydraulic grouts on historic masonry structures.
The case of the Katholikon of Dafni Monastery 1025
A. Miltiadou-Fezans, A. Kalagri, S. Kakkinou, A. Ziagou, N. Delinikolas,
E. Zarogianni & E. Chorafa
High reversibility techniques for in-plane stiffening of wooden floors 1035
N. Gattesco & L. Macorini
Prestressing of masonry as strengthening measure against earthquake loading – Experimental
and numerical investigations and influences on simulation results 1043
S.T. Sperbeck & H. Budelmann
Enhancement of geo-electrical techniques for NDT of masonry 1053
F. Van Rickstal, D. Van Gemert, R. Keersmaekers & D. Posen
Synergetic action of new and existing structures in re-using a 18th century monastery 1061
M. Mezzi, M. Marzullo & G. Valletta
Structural assessment of earthen walls and techniques of onsite consolidation and conservation 1069
S. Sikka
Preservation of historic structures using Screw-Pile foundations 1079
A.J. Lutenegger & J.H. Kemper
Strengthening of an industrial cylindrical shell damaged by a collision 1087
W. Figeys, S. Ignoul & D.V. Gemert
Experimental results on the use of mud-based grouts to repair seismic cracks on adobe walls 1095
J. Vargas, M. Blondet, C. Cancino, F. Ginocchio, C. Iwaki & K. Morales
Development of an in situ penetration test for the uptake of preservatives in applied wood 1101
D.F. Henriques, L. Nunes & J. de Brito

XI
Special structural solutions for adaptive use in grand hotel in Norwich, CT, USA 1107
J.F. Norden
The restoration of the foundation in the north side of the temple of Apollo Epikourios 1113
K. Papadopoulos
Structural consolidation methods for the Temple of Santa Maria della Consolazione in
Todi (Perugia, Italy), damaged by landslides and earthquakes 1123
M. Mariani
The diagnosis and arresting of settlement within Westminster Hall in the Houses of Parliament, London 1133
J.D. Miller
Istanbul – Fatih, Millet Library / Feyzullah Efendi Madrasah restoration 1141
R. Ozakin & A. Erdem
Application of non-destructive techniques at the Katholikon of Dafni Monastery for mapping
the mosaics substrata and grouting monitoring 1149
Ph. Côte, X. Dérobert, A. Miltiadou-Fezans, N. Delinikolas, O. Durand, J. Alexandre,
A. Kalagri, M. Savvidou, D. Chryssopoulos, L. Anamaterou & F. Georganis
The PROHITECH research project 1157
F.M. Mazzolani
The column of Marcian/Kıztaşı in Istanbul and an evaluation of its restorations 1167
A. Erdem & R. Ozakin
Restoration and strengthening strategies for 19th century iron pedestrian suspension bridges 1175
S. Adriaenssens, B. Verbeeck, I. Wouters & M. de Bouw
Reconstruction of the Sistani House at Bam Citadel after the collapse due to the earthquake 2003 1181
W. Jäger & C. Fuchs

Vulnerability to natural hazards, preparedness and retrofitting

Lower and upper bounds in closed form for out-of-plane strength of masonry structures with
frictional resistances 1191
C. Casapulla
Influence of friction and tensile resistance on the stability of masonry arches 1199
P. Smars
Unexplained blasting vulnerabilities in a historic town 1207
D.F. Laefer, B. Loughlin, S. Hickey, S. O’Farrell & G. O’Mahony
Vulnerability and seismic improvement starting from experimental investigation 1215
C. Algeri
Seismic vulnerability evaluation of the Fossanova Gothic church 1225
G. De Matteis, F. Colanzi, A. Eboli & F.M. Mazzolani
Simplified models for seismic vulnerability analysis of bell towers 1237
E. Curti, S. Parodi & S. Podestà
Guastavino dome analysis: A comparative approach for Jefferson’s Rotunda at the
University of Virginia 1245
J. Kaup & J.A. Matteo
Seismic vulnerability and preservation of timber roof structures 1253
M.A. Parisi, C. Chesi, C. Tardini & M. Piazza
Seismic safety of basilica churches: Analysis of ten case studies 1261
G. Brandonisio, E. Mele, R. Santaniello & A. De Luca

XII
The effects of temperature on historical stone masonry structures 1271
C. Blasi & E. Coïsson
A GIS platform on main natural hazards for Valparaíso city (Chile) and vulnerability studies
for some historical constructions and urban sectors 1277
M. Indirli, F. Geremei, C. Puglisi, A. Screpanti, D. Blersch, L. Lanzoni, N. Lopez Izquerdo,
E. Milani, M. Miglioli, G. Simonini, M. Munari & F. Romanelli
The construction of a GIS database for the historical center of San Giuliano di Puglia 1287
S. Mazzanti, M. Miglioli, M. Nascosi, G. Simonini, M. Indirli, L. Lanzoni, S. Teston,
C. Alessandri & D. Pini
Vulnerability assessment of churches at Colima by 3D limit analysis models 1297
A. Orduña, A. Preciado, F. Galván & J.C. Araiza
Vulnerability screening by visual inspection of churches of North-East Anatolia 1303
A. Korkmaz, S. Toker & H. Hastemoglu
Seismic resistance assessment of heritage masonry buildings in public use in Ljubljana 1311
M. Lutman

Codes, guidelines and methods for safeguarding safety and significance

Use of reliability methods for evaluating safety of historic structures 1321
L. Schueremans & E. Verstrynge
Experimental assessment of historic building safety: The case of the Isso Tower in Castelleone, Italy 1331
L. Binda, P. Condoleo, A. Saisi, C. Tiraboschi & L. Zanzi
Making-safe Sydney’s sandstone buildings within accepted conservation guidelines 1339
J.D. Swann
Long term compressive testing for masonry – Test procedure and practical experience 1345
L. Binda, L. Schueremans, E. Verstrynge, S. Ignoul, D.V. Oliveira, P.B. Lourenço & C. Modena
Safeguarding safety and significance – Saving an historic village in Bath 1357
M.S. Stacey & B.P. McCabe
Integration of structural analysis of monuments and historical constructions in engineering and
architecture studies 1363
A. Mosseri
The role of service-learning in heritage preservation and engineering education 1369
D.W. Porter, M.M. Dewoolkar & N.J. Hayden
The safety of Gothic roof structures 1375
I. Kirizsán & B. Szabó
Strengthening historic pedestrian suspension bridges: Public safety goes first!? 1383
I. Wouters, Y. Schoonjans, M. de Bouw & S. Adriaenssens
Displacement requirements in the nonlinear kinematic procedure for masonry structures 1391
A. Giordano, M. Guadagnuolo & G. Faella

Conservation practice
Relation between sub-soil and masonry structure of Angkor monument 1399
S. Yamada, M. Araya & T. Nagai
Beaufort Castle, Lebanon; conservation versus restoration project 1407
J. Yasmine
Structural faults in earthen archaeological sites in central Asia: Analysis and repair methods 1415
E. Fodde

XIII
Maria Birnbaum – Construction history, conservation history 1423
R. Barthel, H. Maus & C. Kayser
The structural strengthening of early and mid 20th century reinforced concrete diaphragms 1431
L. Sorrentino & C. Tocci
Strengthening design of Ganxi’s Former Residence 1441
C. Qing
The Trajan Markets and their Great Hall – The conservation problems and the structural
intervention for the improvement of the seismic safety 1445
G. Croci, A. Viskovic, A. Bozzetti, L.Ungaro & M. Vitti
Structural strengthening of the Dobrzyca Palace, Poland 1455
M.Y. Minch & J.P. Szołomicki
Structural analysis and restoration of the Guglia della Madonna di Bitonto 1461
T.M. Massarelli
Non conventional solutions for the consolidation of bell towers 1467
C. Mileto, F. Vegas López-Manzanares
Strengthening of a wooden covering built in the 18th century 1475
N. Augenti
Experiences of consolidation on archaeological UNESCO sites in the Sultanate of
Oman: The Fortress of Al Balid and the Citadel of Sumhuram 1481
M. Andreini, C. Cei, G. Mariani & M. Sassu
Conservation and restoration of a Ta Prohm temple 1491
T. Lakshmipriya
Restoration of a historical building for Cyprus Technical Chamber 1499
M. Pittas
The great hall of the Trajan Markets throughout the centuries 1509
L. Ungaro, M. Vitti & E. Speranza
The XIX century opera theatre of Catania: Conservation and seismic strengthening interventions 1519
C.F. Carocci & C. Tocci
House-towers in Campi Flegrei: History, conservation and re-use 1527
G. de Martino, R. de Martino, A. Pane & F. Delizia
Structural inspection and analysis of former British Consulate in Shanghai 1537
X.L. Gu, B. Peng, X. Li & D.F. Shang

Author index 1545

XIV
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Preface

The successful preservation of an historic building, complex, or city, unless assigned to museum status, depends
on its continued use and the daily care and maintenance. The possibility of continued use depends on the
adaptation of the building to modern standards and practice of living, and this might require changes of its
constructional or structural features.
Conservation engineering can be defined as the process of understanding, interpreting and managing the
architectural heritage to safely deliver it to posterity. The term ‘safely’ embodies the assumption that the bodies
and individuals responsible for the care of historic buildings work towards ensuring maximum private or public
utility vis a vis minimum loss of fabric and value.
The conflict between safety standards and conservation philosophy usually stems from the fact that not
just the standards themselves, but the practice of achieving the standards, are based on, and refer to, modern
materials, techniques and processes. An historic building is considered of value not only because of its age,
but most importantly because of its uniqueness, its deviation from the norm and hence, to a certain extent,
from what is standard. This constitutes the building’s significance or authenticity, a quality vital to conserve.
With increasing global interest in conservation worldwide it is essential to open the debate on more inclusive
definitions of significance and on more articulated concepts for safety and for acceptable and reliable techniques,
in an attempt to reconcile and integrate further the activity of all the professions involved in conservation.
It is in this framework and with these aims that the Department of Architecture and Civil Engineering of the
University of Bath set out to organize and deliver the sixth international conference on Structural Analysis of
Historic Construction to be held in Bath’s Assembly Rooms between 2 and 4 July 2008.
A grade I listed building designed in 1769 by John Wood the Younger, the Assembly Rooms are located in
the heart of the World Heritage city. Their function was to host events and gatherings and this makes it the
perfect venue for the conference, which is the last of a series that is organized every three or two years. Previous
conferences were organized in New Delhi (2006), Padova (2004), Guimaraes (2001), and Barcelona (1998 and
1995).
The event is sponsored by CINTEC International Ltd, Gifford, English Heritage, ICOMOS, IStructE, and the
University of Bath. This sponsorship has made possible the participation to the conference of authors from other
parts of the world which could have not attended otherwise. Their contribution is critical to our understanding
of practical and theoretical issues in different contexts and to the widening and inclusivity of the debate.
The joint Department of Architecture and Civil Engineering of the University of Bath has a long standing
tradition of teaching conservation and sustainability within the built environment. The MSc in Conservation
of Historic Buildings provides education on the fundamental principles of conservation and is one of the most
successful in the country. Furthermore, the Department offers the only course on Conservation Engineering in
the UK, combining structural aspects with the study and conservation of the built heritage.
The main objective of the conference is to promote debate on an international scale, by the exchange of knowl-
edge and approaches among researchers and practitioners in the different professions involved in conservation and
from diverse cultural backgrounds. The conference is structured into eleven themes: ‘Definition of Significance
and Attribution of Value’, ‘History and Documentation’, ‘Monitoring, non Destructive Evaluation and Testing’,
‘Assessment and Analytical Techniques’, ‘Experimental Studies’, ‘Use of Traditional, Alternative and Innova-
tive Materials’, ‘Novel Conservation Engineering Techniques, Restoration and Strengthening’, ‘Vulnerability to
Natural Hazards, Preparedness and Retrofitting’, ‘Codes, Guidelines and Methods for Safeguarding Safety and
Significance’, and ‘Conservation Practice’.
A total of 180 contributions are collected in the two volumes of proceedings, of which 10 are invited papers
presented by speakers of international renown to the conservation audience. The reading of these proceedings
shows that the discipline of structural analysis of historic buildings is nowadays critical to the success of many
conservation projects and the skills required are increasingly complex and more articulated. The papers collected
all show the need for multidisciplinary approach in conservation research and practice and the continually
evolving need for more sophisticated solutions where both safety and significance play a critical role. An
increasing number of papers tackle issues not covered in previous conferences, such as the structural conservation
of modern materials. It is hoped that the papers presented here can form the basis for future research and the

XV
formulation of relevant issues that need to be advanced. This we hope will constitute the basis for a lively debate
during the 3 days of the conference and will inform and stimulate the directions for future conferences.
The editors are grateful to Marion Harney for managing the organization of the conference and to the depart-
mental office for the great support. The editors are also indebted to the Advisory Committee members for
providing the opportunity and the needed guidance to organize this conference, to the Organizing Commit-
tee members for the efforts and support given to make the conference possible, and finally but crucially to all
members of the Scientific Committee which have agreed to advise in the papers’ selection and reviewing process.

March 2008
Dina D’Ayala, Enrico Fodde

XVI
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Committees

Advisory Committee
Paulo Lourenço, University of Minho, Portugal
Claudio Modena, University of Padua, Italy
Pere Roca, Politecnico de Cataluña, Spain

Organising Committee
Dina D’Ayala, University of Bath, UK
Marion Harney, University of Bath, UK
Enrico Fodde, University of Bath, UK
Alex Copping, University of Bath, UK
Michael Forsyth, University of Bath, UK
Peter Walker, University of Bath, UK
Susan Denyer, Secretary ICOMOS-UK
Geoff Clifton, Chairman, Gifford, UK
Terry Girdler, Chief Engineer, English Heritage, UK

Scientific Committee
Takayoshi Aoki, Nagoya City University, Japan
Görün Arun, Yildiz Technical University, Istanbul, Turkey
Koenraad van Balen, Katholieke Universiteit Leuven, Belgium
Luigia Binda, Politecnico di Milano, Italy
Marcial Blondet, Catholic University of Peru, Peru
Miloz Drdácký, Institute of Theoretical and Applied Mechanics, Czech Republic
Peter Elliott, Elliott & Company, UK
Xianglin Ling Gu, Tongji University, China
Mehrdad Hejazi, University of Isfahan, Iran
Peter James, Cintec International Ltd, UK
Debra Leafer, University College Dublin, Ireland
Giorgio Macchi, Università di Pavia, Italy
Christiane Maierhofer, Fed Inst Material Research and Testing (BAM), Germany
Roberto Meli, Universidad Nacional Autónoma de Mexico, Mexico
John Ochsendorf, MIT, USA
Daniel Oliveira, University of Minho, Portugal
Maurizio Piazza, Universitá Degli Studi di Trento, Italy
Jan G. Rots, Technical University of Delft, The Netherlands
Marek Sklodowski, Institute of Fundamental Technological Research, Poland
Tom Swailes, University of Manchester, UK
Miha Tomazevic, Slovenian National Building and Civil Engineering Institute, Slovenia
Maria Rosa Valluzzi, Universita’ di Padova, Italy
Elizabeth Vintzileou, National Technical University of Athens, Greece
David Yeomans, ICOMOS, UK

XVII
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Conference sponsors

XIX
Experimental studies
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

On the strengthening of three-leaf stone masonry walls

R.A. Silva, D.V. Oliveira & P.B. Lourenço

ISISE, University of Minho, Guimarães, Portugal

ABSTRACT: This paper is devoted to the experimental characterization of the structural behaviour of three-
leaf stone masonry walls. The first part of the experimental results described here was presented during the last
SAHC Conference (Oliveira et al. 2006). In total ten walls, plain and strengthened resorting to transversal tying,
injection and both techniques applied simultaneously, were tested aiming at capturing the detailed structural
behaviour. Globally, all strengthening techniques described here showed to be effective in different ways.

1 INTRODUCTION Injection is the most used technique in repair and

The multi-leaf typology is very common in historical
masonry constructions in the urban centres of Europe,
particularly three-leaf stone walls. The latter walls
are composed by two external leaves made of stone
masonry, using the most abundant stone of the region,
and a poor mortar. The inner leaf is usually made of
rubble masonry or just an infill of a very weak material
(like earth or construction residues), characterized by
a substantial presence of voids (Binda et al. 1999). The
collapse of this kind of walls is characterized by the
formation of brittle collapse mechanisms, which con-
sist essentially on the detachment of the external leaves
and the out-of-plane material expulsions, both under
compression and shear-compression loading (Valluzzi
et al. 2004, Anzani et al. 2004).
To preserve these constructions and to avoid their
continuous deterioration, intervention works are often
required. However, given their cultural importance
and historical value, such interventions must observe
certain requirements, such as the use of materials
mechanically, physically and chemically compatible
with the original ones in order to assure effectiveness
and durability of the strengthening and repair interven-
tions, see Modena (1997), ICOMOS (2001) and Binda
(2006) for further details. When these recommenda-
tions are not taken into account, the works carried
out may result in wrong and ineffective interventions
(Binda 2006). Most of the times, these problems are
related to the lack of knowledge on the material and
structure’s behaviour, which must be enlightened with
extensive research.
Currently, the most used and investigated strength-
ening techniques, due to the good accomplishment of
many of the requirements pointed above, are injection,
repointing or deep repointing and transversal tying.
retrofitting of three-leaf walls mainly because of the
simplicity of its application and the effectiveness of its
use. The aims of the injection technique are the reduc-
tion of the weakness of the internal core, by filling
the existing voids and cracks, and the improvement
of its adherence to the external leaves, by filling the
gaps between leaves. Several studies have been per-
formed in the last years concerning this technique
(Binda et al. 1994, Vintzileou et al. 1995, Toumbakari
2002, Valluzzi et al. 2004), however, special attention
with regard to material compatibility is needed, which
limits the grout selection (Binda, 2006).
A good grout selection depends mostly on the
knowledge of the wall to be injected, which determines
the properties required for the repair grout. Nowadays
the trend is to use a grout mixture mainly based on lime
with a low percentage of cement, in particular when
dealing with restoration works in historical construc-
tions. Injection is typically performed by injecting the
grout starting from the bottom part of the wall and
reaching progressively the top. Usually, for three-leaf
walls the injection pressure is very low, not exceeding
50–100 kPa to avoid the undesired detachment of the
external leaves.
The repointing or deep repointing technique con-
sists in the removal of part of the old mortar of the
joints, which is then replaced by a new mortar with
improved properties. It can be applied simultaneous
with other strengthening techniques (e.g. injection).
This technique is normally used in brick masonry
with regular joints, which allows to introduce steel
or FRP rods in the bed joints, constituting a good
strengthening solution for structures with creep prob-
lems (Valluzzi et al. 2005). Experimental works with
this technique have showed an improvement of the
compressive strength.

739
 The transversal tying technique is aimed at improv-
ing the connection among leaves, in particular between
the external ones, in order to reduce the transverse
deformation. For this purpose, stainless steel bars or
FRP bars can be used. The bars are inserted into
drilled holes through the thickness of walls and then
anchored. In case of FRP bars, the anchoring can be
achieved by using special anchoring elements (like
angle bars or connector developed on purpose) or
relying on the bond behaviour between the FRP bar
and the mortar, developed along the thickness of the
external leaf. In order to improve this last anchoring
mechanism, a local grout injection around the tie can
be applied instead (Oliveira et al, 2006). This tech-
nique can also be applied in combination with other
techniques.
Note that the aforementioned techniques try to solve
a problem of localized behavior, i.e. at the element
level, and not a global behavior problem, i.e. at level
of the structural global behavior (connections between
walls, floors, etc.) that require the application of other
strengthening techniques, which are out of the scope
of this paper.
The work presented here deals with the strength-
ening of three-leaf stone masonry walls, using dif-
ferent strengthening techniques: transversal tying by
means of GFRP bars, injection and combination of the
two previous techniques. The main objectives of this
research are: characterization of the behaviour of
three-leaf walls under different strengthening configu-
rations and development of a suitable numerical model
able to interpret and explain the structural behaviour
of such type of walls. This paper presents the exper-
imental results concerning the testing of three series
of walls (ten walls), where the three aforementioned
strengthening techniques were applied. Strengthening
was applied without any previous induced damage.
The test program is summarized in Table 1.
Information about the materials used and mechan-
ical properties of masonry components (external and
inner leaves) is also provided.
Table 1. Testing of three-leaf walls.
Wall Wall series Strengthening technique
1W1 1 U
1W2 1 U
2W1 2 U
2W2 2 T
2W3 2 T
2W4 2 T
3W1 3 U
3W2 3 I
3W3 3 I
3W4 3 T+I
U – Unstrengthened wall/T – Transversal tying/I – Injection.
740
The experimental results described in a previous
paper (Oliveira et al. 2006) are here summarized and
all new developments made are described in detail.
2 CHARACTERIZATION OF WALL
COMPONENTS
The mechanical characterization of some of the com-
ponents has been already provided in a previous paper
(Oliveira et al. 2006), but for the sake of simplicity, it
will be briefly repeated here.
2.1 Stone
A locally available granite stone was used to build
all walls. Its mechanical characterization was per-
formed in cylindrical specimens of dimensions
Ø100 × 200 mm2 , where the following average values
were obtained: compressive strength of 52.2 N/mm2 ,
Young’s modulus of 20.6 × 103 N/mm2 and Poisson’s
ratio of 0.24.
2.2 Mortar
A 1:3 binder/sand ratio and a 0.8 water/binder ratio
were selected (all ratios in weight) in order to obtain
a representative mortar composition. The binder was
composed by 25% of hydrated lime and 75% of
metakaolin (pozzolanic material). In addition, a poz-
zolanic drier (10% on binder weight) was used to
obtain a faster mortar drying and, therefore, to improve
the construction procedure of the walls.
The mechanical behaviour of mortar was assessed
using cubic specimens of 50 × 50 × 50 mm3 sampled
during the construction of the walls and tested under
compressive loading at the ages of 7, 28 and 90 days.
Average compressive strengths of 0.5 N/mm2 ,
2.9 N/mm2 and 2.2 N/mm2 were measured at the
aforementioned ages, respectively.
2.3 GFRP bar and grout
The transversal tying technique was applied by means
of GFRP bars placed transversally to the wall and
anchored along the thickness of the external leaves, by
means of injected grout.The bond strength between the
bar, the grout and the masonry was considered enough
to transmit the load from the external leaf to the GFRP
bar. Furthermore, the low stress state together with
the high tensile strength of the GFRP bar (a value
of 760 N/mm2 was provided by the manufacturer)
excluded its brittle tensile failure.
A commercial lime-based grout was used, for both
wall injection and bonding of the GFRP bars to
masonry. Its mechanical behaviour was assessed by
means of cubic specimens of 50 × 50 × 50 mm3 sam-
pled during injection. For the walls strengthened with
GFRP bars within series 2, the grout samples pro-
vided an average compressive strength of 17.6 N/mm2
and an average tensile strength of 0.3 N/mm2 . For the
walls injected within series 3, grout specimens reached
an average compressive strength of 13.7 N/mm2 (no
tensile tests were performed).
2.4 External leaf
The mechanical behaviour of the external leaf was
characterized through a set of representative stone
masonry prisms (see Oliveira at al. 2006). An average
compressive strength of 9.2 N/mm2 and a coefficient
of variation of 19% were computed for specimens
coming from the three series.
2.5 Inner leaf
Representative core specimens were also built using
granite scabblings poured into alternate layers with
mortar and avoiding any compaction, aiming at rep-
resenting the construction procedure followed for
the walls. During the construction of the walls, a
total of ten cylindrical specimens of dimensions
Ø150 × 300 mm2 were built. The specimens were
tested under uniaxial compressive loading at a dis-
placement control rate of 5 µm/s. An average com-
pressive strength of 0.3 N/mm2 and a coefficient of
variation of 45% were obtained. As expected, a very
low strength was achieved.
2.6 Injected inner leaf
The injected inner leaf was also characterized. For that,
seven prisms were sampled from the inner core of the
injected walls (3W2 and 3W3) during the dismantling
procedure, see Figure 1a. The samples were extracted
with average dimensions of 80 × 80 × 160 mm3 (h/d
ratio of 2) in order to obtain representative core
specimens.
The specimens were tested under uniaxial compres-
sive loading at a displacement control rate of 2.5 µm/s.
Figure 1. Injected inner leaf: (a) general view of a wall’s
inner leaf; (b) crack pattern of a sampled specimen after
testing.
741
An average compressive strength of 4.1 N/mm2 was
attained with a variation coefficient of 12%. The
improvement observed was of about 14 times the
strength of the plain inner leaf. The low variation
coefficient shows that injection allowed for a homog-
enization of the inner leaf strength.
In Figure 1b it is shown the crack pattern of a tested
prism, where it is possible to observe cracks contouring
the stone elements.
3 WALL TEST PROCEDURES
3.1 Wall specimens
As exposed above, the entire ten wall specimens were
built with three leaves, two granite masonry external
leaves and an inner leaf built with granite scrabblings
and mortar, without any compaction, aiming at obtain-
ing an amount of voids representative of this kind of
walls, able to allow injection. The average thickness
of each leaf was about 100 mm, which wasn’t always
possible, due to the variable size and shape of the
stones used. This feature also influenced the number of
courses, which ranged from six to eight courses. The
global dimensions adopted for all wall specimens were
600 mm long, 300 mm thick and 1100 mm high, simi-
lar to dimensions found in previous works (Vintzileou
et al. 1995, Toumbakari 2002, Valluzzi et al. 2004)
and aiming at representing 1/2 scaled models of real
walls. It is worth to mention that no stones connecting
the external leaves were used, in order to both assure
the worst conditions found in ancient buildings and
provide leaf interfaces for all specimens as similar as
possible.
3.2 Strengthening
The transversal tying technique was applied and tested
in the second and third series of walls. After drilling
two holes with a diameter of 20 mm, through the
entire wall thickness, at one third and two thirds of
specimen’s height, GFRP bars with 10 mm diame-
ter were placed and the holes were injected with the
aforementioned grout.
The injection technique was applied only in the third
series of walls. Due to the low wall thickness, injec-
tion was applied just in one side of the walls. For that,
the subsequent procedure was followed (see Figure 2):
(a) drilling of slightly inclined holes with 14 mm
diameter. Their distance varied between 100 mm and
200 mm, depending on the location of masonry joints;
(b) introduction of small plastic tubes with a diameter
of 10 mm in the holes; (c) sealing of holes and major
mortar joint absences, caused by the drilling process,
with silicone to prevent grout leakage; (d) injection
of water to verify which tubes were active and to
wet the inner leaf; (e) sealing of the transversal sides
with polyurethane foam to prevent grout leakage and
 Table 2. Summary of results of the unstrengthened walls.

fc εa,p E0 E[30–60]
Wall (N/mm2 ) (mm/m) (N/mm2 ) (N/mm2 )

1W1 2.3 6.81 3246 780

1W2 1.7 2.86 2087 1889
2W1 1.4 9.37 1422 711
3W1 2.6 3.87 1733 1351
Average 2.0 5.73 2122 1183
CV (%) 27 51 38 47

Figure 2. Wall injection procedure: (a) sealing of transversal

sides with polyurethane foam; (b) injection equipment and
(c) injection works.

without adding extra stiffness to walls and (f) prepa-

ration of the grout and injection under a low pressure
of around 0.1 N/mm2 .

3.3 Test procedure and test setup

All walls were tested under monotonic compressive
loading, using a 2 MN closed-loop servo-controlled
testing machine. The tests were performed under dis-
placement control at a displacement increment rate of
3 µm/s. In order to prevent the total collapse of the
walls, tests were stopped during the softening branch
when specimens were about to fail. Whenever pos-
sible, walls were dismantled in order to check the
efficiency of the strengthening procedure.
For the measurements of the displacements, an
internal setup and an external setup were used. The
internal setup was composed by LVDTs connected
directly to specimens and measuring vertical, hor-
izontal and transversal displacements (see Oliveira
et al. 2006 for further details). The external setup was
constituted by the control LVDT that measured the dis-
placement between the machine plates. This last setup
was used to control the test and to obtain the plot of
the post-peak force-displacement curve.
4 WALL TEST RESULTS
4.1 Plain walls
Table 2 summarizes the test results for the four
unstrengthened walls (associated with the three series)
in terms of compressive strength (fc ), peak axial strain
(εa,p ), initial Young’s modulus (E0 ) computed between
0% and 20% of the wall’s compressive strength and
Young’s modulus computed between 30% and 60%
of the wall’s compressive strength (E[30–60] ). The
computation of the Young’s modulus was performed
according to two different criteria in order to assess its
degradation with increasing stress levels.
The considerable scattering computed mainly for
the deformability parameters is essentially due to the
Figure 3. Axial stress – axial strain curves of the unstrength-
ened walls.
influence of workmanship and the variability of natural
and handmade materials.
Figure 3 represents the axial stress – axial strain
curves of the unstrengthened walls. Two distinct stiff-
ness degradation zones can be observed, which seem
to be associated to the detachment of external leaves.
However, this behaviour was not observed in wall
1W2 (see Figure 3 and Table 2), probably due to
an unexpected improved connection between leaves,
originated during the construction of the wall.
The observed failure modes of the unstrengthened
walls showed that the collapse mechanism of these
walls is governed by the out-of-plane rotation of the
external leaves. In order to evaluate this feature, the
adimensional parameter λ is now introduced. Here, λ
is given by the average value of the four rotation angle
tangents of the external leaves. This parameter can be
seen as a damage measurement of the out-of-plane
behaviour. The relationship between the λ parameter
and the axial compressive stress is given in Figure 4 for
the unstrengthened walls. This figure allows to iden-
tify the beginning of leaves separation and to better
recognize the atypical behaviour of wall 1W2.
Experiments showed also that the out-of-plane rota-
tion of the external leaves was caused by the devel-
opment of three hinges along bed joints close both

742
Figure 4. Unstrengthened walls: evolution of the λ param- Figure 6. Axial stress – axial strain curves relative to the
eter with regard to the axial stress. transversal tied walls.

Figure 5. Crack pattern of the unstrengthened wall 3W1

(The top and bottom hinges are not represented).

Table 3. Summary of results of the walls strengthened with

transversal tying.
Figure 7. Walls strengthened with the transversal tying tech-
fc εa,p E0 E[30–60] nique: evolution of the λ parameter with regard to the axial
Wall (N/mm2 ) (mm/m) (N/mm2 ) (N/mm2 ) stress.

2W2 3.3 9.35 1954 722 series, however, the hypothesis of damage induced by
2W3 2.6 5.93 1707 603 the drilling procedure cannot be totally excluded. This
2W4 3.5 9.37 1160 675
feature will be later discussed.
Average 3.1 8.21 1607 667 Table 3 also shows that the existence of the transver-
CV (%) 27 24 25 9 sal ties does not prevent the important stiffness
reduction observed previously.
to plates and the middle height of the wall. Vertical The axial stress – axial strain curves of the tied
cracks contouring the masonry stones are also visible, walls are displayed in Figure 6. For these walls it is
see Figure 5. possible to observe smooth and continuous stiffness
degradation with increasing applied load.
The evolution of the λ parameter with respect to
4.2 Walls strengthened with transversal tying
the applied stress is displayed in Figure 7 for the tied
Results concerning the three walls strengthened with walls. A sudden increase of the variation rate of λ
transversal GFRP bars are summarized in Table 3. occurs for a stress level close to the peak, while for
The presence of the ties allowed a 55% increase the unstrengthened walls, this sudden change occurred
of the average compressive strength in relation to the much sooner.
unstrengthened walls. It is believed that the lower value From above, it can be concluded that the transversal
of the Young modulus of the tied walls (using both tying technique does not prevent stiffness degradation
criteria) is due to differences associated to distinct but makes it happen in a more smooth way.

743
Figure 8. Typical crack pattern of a tied wall (wall 2W4).

Table 4. Summary of results of the injected walls.

fc εa,p E0 E[30–60]
Wall (N/mm2 ) (mm/m) (N/mm2 ) (N/mm2 ) Figure 9. Axial stress – axial strain curves concerning
injected walls.
3W2 3.9 7.00 2172 1604
3W3 3.3 6.87 1844 1114
Average 3.6 6.93 2008 1359
CV (%) 11.0 1 12 26

The crack pattern for a near-collapse condition is

showed in Figure 8 (wall 2W4), where vertical cracks
are dominant. The absence of important horizontal
cracks shows that transversal tying is able to prevent
the formation of the “middle hinge” necessary to the
development of the out-of-plane collapse mechanism.
Instead, failure happened due to localized instability
related to the detachment of some stones.

4.3 Injected walls

Table 4 summarizes the main results regarding the
injected walls. A low scattering was obtained, prob-
ably due to a homogenization effect provided by
the injection, but the reduced number of tests also
influenced it.
The injection technique allowed an increase of
the compressive strength of about 80% and 16%
when compared with the plain and the tied walls,
respectively.
Table 4 seems also to indicate that the stiffness
reduction with increasing stress is less pronounced for
the injected walls.
The axial stress – axial strain curves concerning
injected walls are displayed in Figure 9. Like in the
tied walls, also here a continuous stiffness degradation
zone was observed, with a considerable increase of the
λ parameter near peak, see Figure 10. This means that
close to the peak load the injection technique did not
prevent the detachment of the external leaves in the
same effective way as when the tied walls were tested.
The odd behaviour exhibited by wall 3W3 is most
probably due to a deficient injection procedure, lead-
ing to the existence of non-injected voids in the core,
Figure 10. Injected walls: evolution of the λ parameter with
regard to the axial stress.
as found out during the dismantling of the wall per-
formed after testing. This resulted in a premature and
unexpected local detachment of the external leaves,
with direct consequences in the compressive strength.
The crack pattern for a near-collapse condition is
showed in Figure 11 (wall 3W3), being composed
mainly by vertical cracks, despite some horizon-
tal cracks started to appear, indicating the incipient
onset of an out-of-plane mechanism, which in these
walls never reached the development achieved by the
unstrengthened walls.
The injection technique caused a more diffuse crack
distribution when compared with the crack pattern
from the previous walls, leading also to the develop-
ment of important cracks in the inner leaf, detected
during dismantling.
Failure of the injected walls was due to localized
stone cracking and instability, followed by the detach-
ment of stones. The injection technique prevented

744
Figure 11. Crack pattern of injected wall 3W3.

Table 5. Summary of the results of the wall subjected to

injection and transversal tying.

fc εap E0 E[30–60]
Figure 13. Combined strengthened wall: evolution of the
Wall (N/mm2 ) (mm/m) (N/mm2 ) (N/mm2 )
λ parameter with regard to the axial stress.
3W4 3.8 4.86 3539 1849

Figure 14. Crack pattern of wall 3W4 (combined strength-

ened wall).

further confirmed by Figure 13, where the evolution

Figure 12. Axial stress – axial strain curve relative to wall
3W4 (combined strengthened wall).
the full external leaf detachment from happen, but it
allowed partial detachments.
4.4 Combined strengthening technique
Aiming at assessing the combined use of the two previ-
ous techniques, one wall was simultaneously injected
and strengthened with transversal GFRP ties. Table 5
summarizes the obtained results. It must be noted
that the number of tested specimens is insufficient to
validate the comments provided herein.
On average terms, the compressive strength reached
is close to the value obtained for the injected walls
and slightly higher than the value obtained for the tied
walls.
Figure 12 shows the axial stress – axial strain curve
of the 3W4 wall. This figure allows to identify smooth
and continuous stiffness degradation. This feature is
of the leaves’ opening (λ parameter) with stress level
is represented. These results seem to indicate that the
simultaneous use of both strengthening techniques is
beneficial in the sense that a better global structural
behaviour was reached, namely in terms of stiffness
degradation and control of the out-of-plane movement
of the external leaves.
Figure 14 illustrates the crack pattern of wall 3W4
observed near collapse. Visible cracks are mainly ver-
tical, which may go trough some stones, with a diffuse
crack pattern distribution. The wall failure was due to
localized stone instability located in the edge of the
wall, far from the relatively localized effect of the tie
bars. During the wall dismounting huge vertical cracks
crossing the entire inner leaf were observed.
5 CONCLUDING REMARKS
The test results on three-leaf walls showed that both
strengthening techniques, applied either individually
or in a combined fashion, have led to compressive
strength enhancements. The technique that allowed the

745
biggest improvement was the combination of injection
and transversal tying, with an improvement of 90%,
followed by the injection technique with 80% increase
and finally the transversal tying with 55% increase of
compressive strength.
Due to scattering and differences among series,
it was observed that the Young’s modulus decreased
when shifting from plain to tied walls, which might be
considered an anomalous behaviour. However, when
comparing walls built in a same series, it is possible to
observe that, on average terms, strengthening usually
causes an increase of the initial stiffness, around 13%
for tied walls, 19% for injected walls and 127% for
the combined strengthened wall. It is worth to men-
tion that, due to scattering, these figures need further
support based on more experimental results.
The transversal tying technique promoted the leaves
detachment control, which caused changes in the stiff-
ness degradation evolution as well as in the crack
pattern, with the absence of major horizontal cracks.
In the same way, injection changed the behaviour of
the walls, allowing all leaves to work together till near
peak load, by increasing both the connection between
leaves and the inner leaf strength, which was noticeable
by the diffuse crack pattern and by the presence of big
cracks in the inner leaf.
The results obtained for combined strengthened
wall seem to show that this technique gathered the ben-
efits of both techniques applied separately, however
more tests are need to confirm these aspects.
Finally, it is important to refer that scattering is a
key issue when dealing with historical constructions
as well as with natural and handmade materials.
ACKNOWLEDGEMENTS
The authors would like to thank the technical staff at the
Laboratory of Civil Engineering, University of Minho,
for the help provided.
Acknowledgements are also due to the companies
Fradical, Mapei and Augusto de Oliveira Ferreira for
providing raw materials and workmanship.
The financial support provided by the Por-
tuguese Science and Technology Foundation through
746
the POCI/ECM/58987/2004 project is gratefully
acknowledged.
REFERENCES
Anzani A., Binda L., Fontana A., Pina Henriques J., 2004,
An experimental investigation on multiple-leaf stone
masonry. 13th International Brick and Block Masonry
Conference, Amsterdam, July 4–7, 10 pp., CD-ROM.
Binda L., Modena C., Baronio G., Gelmi A., 1994, Exper-
imental qualification of injection admixtures used for
repair and strengthening of stone masonry walls. 10th
International Brick/Block Masonry Conference, Calgary,
Canada, Vol. 2, pp. 539–548.
Binda L., Baronio G., Penazzi D., Palma M., Tiraboschi C.,
1999, Characterization of stone masonry walls in seismic
areas: data-base on the masonry sections and materials
investigations (in Italian), L’ingegneria Sismica in Italia,
9th National Conference, Turin, Italy, 14 pp., CD-ROM.
Binda L., 2006, The difficult choice of materials used for
the repair of brick and stone masonry walls, 1st Inter-
national Conference on Restoration of Heritage Masonry
Structures, Cairo, Egypt, CD-ROM.
ICOMOS, 2001, Recommendations for the analysis, con-
servation and structural restoration of architectural her-
itage, International Scientific Committee for Analysis and
Restoration of Structures of Architectural Heritage.
Modena C.,1997, Criteria for cautious repair of historic build-
ing. Evaluation and strengthening of existing masonry
structures, Binda L. and Modena C., Ed. RILEM.
Oliveira D.V., Lourenço P.B., Garbin E., Valluzzi M.R.,
Modena C., 2006, Experimental investigation on the struc-
tural behaviour and strengthening of three-leaf stone
masonry walls, Structural Analysis of Historical Con-
structions, New Delhi.
Toumbakari, E.E., 2002, Lime-pozzolan-cement grouts and
their structural effects on composite masonry walls.
Ph.D. Thesis, Katholieke Universiteit Leuven, Heverlee,
Belgium, pp. 364.
Valluzzi M.R., da Porto F., Modena C., 2004, Behaviour and
modeling of strengthened three-leaf stone masonry walls,
Materials and Structures, Vol. 37, pp. 184–192.
Valluzzi M.R., Binda L., Modena C., 2005, Mechanical
behaviour of historic masonry structures strengthened
by bed joints structural repointing, Construction and
Building Materials, Vol. 19, pp. 63–73.
Vintzileou E., Tassios T.P., 1995, Three-Leaf Stone Masonry
Strengthened by Injecting Cement Grouts, Journal of
Structural Engineering, pp. 848–856.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Long-term behaviour of historical masonry – A quantitative acquisition

of the damage evolution

E. Verstrynge, L. Schueremans & D. Van Gemert

Department of Civil Engineering, KULeuven, Heverlee, Belgium

S. Ignoul
Triconsult NV, Lummen, Belgium

M. Wevers
Department of Metallurgy and Materials Engineering, KULeuven, Heverlee, Belgium

ABSTRACT: Within the framework of some dramatic collapses of historical masonry structures, a test program
has been set up including short-term and long-term creep testing of masonry, composed with different types of
mortar. These experimental creep tests are very adequate to gather more insight into the described phenomenon,
but other tools are necessary in order to evaluate the damage evolution of historical masonry. Therefore, acoustic
emission detection has been used as a non destructive technique in order to evaluate the damage evolution during
experimental creep testing. This paper discusses the data obtained during the creep tests and first results of
the use of the acoustic emission technique for detecting the damage evolution within masonry under persistent
compressive load, based on laboratory tests on small specimens.

1 INTRODUCTION
During the past decades, experimental research has
been carried out to get better insight into the long-term
behaviour of historical masonry structures. The initial
motivation for the rising interest within this field of
research were a few dramatic collapses of historical
masonry buildings (Civic Tower of Pavia, 1989; Noto
Cathedral, 1996) (Binda 1992). Recent collapses of
masonry buildings in Belgium confirm the importance
of this research.
1.1 Bell tower of the Sint-Willibrordus church,
Meldert, Belgium
The bell tower of the Sint-Willibrordus church col-
lapsed on 7th July 2006 (Fig. 1). The three-leaf
masonry of the tower was composed of two outer leafs
in sandstone and an inner core with rubble masonry of
smaller sandstone chunks and large amounts of lime
mortar. The red-brown, rather porous sandstone con-
tains iron oxide, which is responsible for its colour.
The sandstone is known for having a large scatter on its
strength characteristics and easy absorption of water,
making it rather vulnerable for weathering influences.
A visual inspection of the supporting corner pil-
lars of the church tower, carried out before collapse,
showed a large amount of vertical cracks, indicating
that the masonry was heavily loaded and in bad condi-
tion. This crack pattern was found on several of the
supporting elements of the tower and a monitoring
campaign was set up, including the monitoring of a
number of cracks by means of a mechanical strain
gauge.
Within a period of two weeks, some crack openings
showed an increase of 1 mm and more and the advice
was given to close the church and to start immedi-
ate strengthening and repair actions. Just two weeks
later, while the technical staff (who was removing valu-
able sculptures before the restoration) was having a
lunch break, the tower collapsed. Fortunately, making
no casualties.
1.2 Medieval tower “Maagdentoren”,
Zichem, Belgium
The Maagdentoren partially collapsed on 1st of June,
2006 under comparable circumstances. Again, no sud-
den disturbance of the acting forces, such as an
earthquake, could be pointed out as having caused the
collapse. The tower was also composed of three-leaf
masonry in sandstone, which appeared to be in a bad
condition. A high dead load acted on the base of the
tower, especially in the area where the cross-section
of the wall was reduced by a staircase opening. It is at
this point that the wall collapse was triggered.

747

Figure 1. Bell tower of the Sint-Willibrordus church at
Meldert, Belgium, before collapse on 7th July 2006.
Figure 2. Maagden toren at Zichem, Belgium, before and
after collapse (June 2006).
Within the framework of these dramatic events, a
test program has been set up and recently extended.
During these experiments, a persistent compressive
load is applied and the deformation of the masonry
with time is evaluated. As these long-term creep tests
are difficult to carry out and require a long dura-
tion, rising up to several years, the persistent loading
of masonry is also simulated by performing acceler-
ated, short-term creep tests. The test program and the
obtained results will be briefly discussed.
748
These experimental creep tests are very adequate to
gather more insight in the described phenomenon, but
other tools are necessary in order to evaluate the dam-
age evolution of historical masonry on site. Therefore,
the acoustic emission technique has been used as a non
destructive technique for monitoring the damage evo-
lution during experimental creep testing. These tests
are discussed in the second part of the paper.
2 EXPERIMENTAL RESEARCH PROGRAM
In order to expand our knowledge on long-term
behaviour of masonry and to gather data for the fit-
ting of model parameters, a test program has been
set up, including three types of tests; the necessary
data are gathered by means of monotonic compres-
sion tests, as well as by accelerated creep tests on
masonry wallets. Additionally, long-term creep tests
were started up to enable the validation of the calcu-
lated strain and damage evolution in time, simulated
by means of a rheological model, as the parameters
used in the simulation are based on the short-term test
data. These different types of tests for assessing the
time-dependent behaviour of the masonry were pro-
posed in literature (Binda 1993, Anzani 1995). The
implementation and results of the different creep tests
are treated here briefly, the model calibration can be
found elsewhere (Verstrynge 2008).
In the ongoing experimental program, one single
type of brick is combined with 4 different types of
mortar: cement mortar, hydraulic lime mortar, air hard-
ening lime mortar and a cement-lime mortar. The
purpose is to investigate the effect of differences
in stiffness of the material on the long-term creep
behaviour. The brick type and mortar composition
were chosen to be representative for historical masonry
encountered in Belgium. For each type of mortar,
9 wallets were made, with a base of 29 by 19 cm and
a height of approximately 85 cm, which corresponds
to 14 brick layers and a joint thickness of 1 cm. The
height of the masonry wallets is at least three times
larger than the thickness to assure an uniaxial stress
distribution in the middle of the specimens.
For the monotonic compression tests (3 tests per
mortar type) as well as for the accelerated creep tests
(3 tests per mortar type), a test device, type Schenck
Trebel RM 100, was used in combination with one hor-
izontal and one vertical strain gauge (LVDT) on each
side of the specimen. For the real creep tests (3 tests per
mortar type), separate steel frames were constructed
and hydraulic jacks were used, in combination with a
mechanical strain gauge to measure the deformations
periodically. For the latter, four horizontal and six ver-
tical measuring bases were installed on each side of
the test wallet.
Figure 3. Strain-time graph of accelerated creep tests on
three hydraulic lime mortar wallets.
2.1 Monotonic compression tests
Compression tests were performed in strain-controlled
conditions to gather information on the compressive
strength of the different masonry types. Based on these
ultimate strengths, the stress path to be followed dur-
ing accelerated and long-term creep tests could be
calculated.
2.2 Accelerated creep tests
Accelerated, short-term creep tests were performed to
assess the damage accumulation in masonry under
sustained stresses. The aim is to obtain data during
a short-term test, which can be used to deduce the
parameters, necessary to simulate and assess the strain
and long-term damage evolution, for example during
the life-time of a masonry structure.
Three accelerated creep tests were performed for
each type of mortar, starting at a stress level of
50–60% of the compressive strength and increasing
with stress increments of 5% until failure of the spec-
imen. In between two stress increments, a constant
stress level was maintained during a period of at least
2 hours, but even then, failure often occurred during
the stress increment and no tertiary creep phase was
reached.
The strain-time graphs of the accelerated creep tests
on the columns with hydraulic lime mortar are pre-
sented in Figure 3. During one of the tests, the stress
level could not be kept constant at the end of the test-
ing period due to a technical problem (indicated with
arrow in Figure 3).
The time increment of two hours is relatively short
in order to reach long-term creep effects, but the tests
still indicated some interesting results. During each
load step, a primary (decreasing strain rate) and a sec-
ondary (constant creep rate) creep phase was reached.
749
Figure 4. Volumetric strain-time graph of accelerated creep
tests on three hydraulic lime mortar wallets. Positive volu-
metric strain indicates volume increase.
Only during the last load step of each test, also the
third creep phase (increasing strain rate) was reached.
At two successive steps, the strain rate of the sec-
ondary phase increased, indicating a relation between
the sustained load level and the strain rate.
The evolution of the volumetric strain initially
shows a volume decrease. During the last stages of the
tests, the volumetric strain becomes positive, indicat-
ing the dilatancy behaviour of the masonry just before
failure occurs.
2.3 Long-term creep testing
In order to assess the long-term behaviour of masonry
under persistent loading, 9 long-term creep tests were
set up. For each mortar type, respectively hydraulic
lime mortar, cement mortar and lime-cement mortar,
three masonry wallets were tested. These three wallets
were initially loaded at 50, 65 and 80% of the compres-
sive strength, obtained by averaging the results of the
monotonic compression tests. These long-term tests
are still ongoing.
3 ACOUSTIC EMISSION TECHNIQUE
The acoustic emission technique is a non-destructive
technique which detects and locates damage at the
moment of occurrence. Acoustic emissions (AE) are
high frequency transient sound waves, which are emit-
ted during local stress redistributions caused by struc-
tural changes, such as crack growth. The technique is
proposed here as a monitoring technique for the detec-
tion of damage initiation and assessment of the rate of
damage evolution during creep deformation.
When a set of several sensors is applied for the
acoustic emission measurement, localisation of the

Figure 5. Main features of an AE signal (burst type).
source of the emissions is possible by taking into con-
sideration the geometrical arrangement of the sensors
and the moment of arrival of a wave at each indi-
vidual sensor. This technique is frequently used for
detection of the location of damage within isotropic
materials. As masonry is a highly heterogenic mate-
rial, with a different propagation speed of the energy
waves in different directions, localization of damage
is very difficult and will therefore not be taken into
account here.
AnAE wave, detected by a sensor is called a “hit”. In
order to filter out the continuous low-amplitude back-
ground noise, a threshold is defined and only sound
waves passing this amplitude-threshold are detected.
Two or more hits, captured by different sensors and
originating from the same event are simply referred
to as one “event”. The main features of an AE signal
(burst type) are schematically indicated in Figure 5.
3.1 The use of acoustic emissions for damage
detection
Examples of damage assessment by AE monitoring
in brittle materials under mechanical loading can be
found in literature (Grossi 1996, Eberhardt 1997,
Colombo 2003). References describing the use of
this technique on masonry on the other hand are to
a much smaller degree available in literature, espe-
cially regarding the monitoring of damage evolution in
masonry as a consequence of long-term creep defor-
mation. Acoustic emission monitoring has been used
on masonry arch bridges (Tomor & Melbourne 2007).
A practical study of the creep phenomenon in masonry
towers combined with acoustic emission monitoring
was performed by Carpinteri (Carpinteri 2007).
The amount of hits, detected during a certain time
interval, depends on various specific boundary condi-
tions of the test setup, such as the threshold level, the
quality of the coupling between the sensor and the test
specimen, the density, coherence and speed of wave
propagation in the material and the interference of
750
surrounding test equipment to name a few. Therefore,
the necessary precautions have to be taken in order to
keep these boundary conditions as much constant as
possible. Concerning this remark, it also follows that
not the absolute amount of detected events, but rather
the change in detection level or event detection rate is
a determining factor for the assessment of the damage
accumulation. The experimental research, discussed
below, indicates that the event rate is related to the rate
of damage increase within the masonry.
3.2 AE testing on masonry cores
To gain more insight in the described creep phe-
nomena, short-term creep tests were carried out on
masonry core samples in combination with acoustic
emission (AE) detection. It concerns masonry cores,
composed of brickwork and cement mortar, which
were taken from an existing structure in order to test the
material’s compressive strength. As the drilled cores
showed a very good coherence, also creep testing could
be carried out together with AE detection in order to
test the combined use of both deformation and AE
measurements. The short-term creep tests were carried
out based on the knowledge obtained during the previ-
ously discussed test program. From the compression
tests, an average compressive strength (fc ) of 6.58 MPa
(± 1.15 MPa standard deviation) was found for the
masonry cores. The loading scheme for the acceler-
ated creep tests was set to start at a load of 20% of fc
and then the load was increased in steps of 10% at the
beginning of the test and 5% during the last load steps.
The initial load of 20% was taken to be considerably
low, to take into account the large scatter on the com-
pressive strength of the samples. By setting a low initial
load, each sample would have enough load increment
steps during the test. Each loading step was kept con-
stant for at least 20 minutes, longer time increments
were used if the strain rate was not showing a constant
or decreasing evolution after 20 minutes. By following
this scheme, each test had a duration of approximately
6–8 hours.
Acoustic emissions, the stress-strain evolution and
the evolution of the vertical strain in time could be
followed online during the test. Therefore, following
sensors were applied for each test (Fig. 6):
– Two acoustic emission sensors (range 250–
700 kHz, peak amplitude 375 kHz);
– Two LVDT’s with a range of ± 1 mm to measure the
horizontal strain (type Schaevitz LBB);
– Two LVDT’s with a range of ± 2.5 mm to measure
the vertical strain (type Schaevitz LBB).
Teflon sheets were placed between the metal plates
of the press and the specimen to immobilize fric-
tional forces. The test was carried out stress-controlled
in order to be able to keep a constant stress level,
even if deformation of the specimen would occur.
 (a) (b)

Figure 6. Accelerated creep tests (a) and cyclic accelerated

creep tests (b) on masonry core, test setup with AE sensors
and LVDT’s. Figure 7a. Stress-strain curve for an accelerated creep test
on masonry cores.
Two different types of accelerated creep tests were
performed:
– Standard tests with small stress increases, alter-
nated with periods during which the stress level
is kept constant, as explained above. They will be
called accelerated creep tests (ACT);
– Creep tests similar to the previous ones, the dif-
ference being that before every stress increase, an
unloading cycle is included. This way, it is easier
to capture the evolution of the elastic parameters
as damage increases and distinguish between the
immediate elastic deformation and time-dependent
deformations. These tests will be called cyclic
accelerated creep tests (CACT).
For the ACT (Fig. 6a), the vertically placed LVDT’s
were positioned in a cylindrical metal frame in which
the test specimen was placed. Consequently, also the Figure 7b. Strain evolution in time for an accelerated creep
deformation of the Teflon sheet and the compres- test on masonry cores.
sion of small irregularities on the specimen’s surface
were measured by the strain gauge. During the CACT load step and a secondary phase at the end of the load
(Fig. 6b), higher specimens were used, so that the ver- step. As the time increment was rather short, a clearly
tical LVDT’s could be placed on the specimen itself. pronounced secondary phase with constant strain rate
The horizontal strain gauges were always positioned was not always obtained for every step. At some load
in the metal frame (Fig. 6a-b). steps, the strain rate is still decreasing at the end of the
The stress-strain relation for a typical ACT and a step, which indicates that there is still primary creep
CACT test are presented in Figures 7–8. Horizontal present. All tests showed a stress-dependent strain rate,
strains are indicated as negative, vertical strains as as was also detected during the accelerated creep tests
positive. on the masonry wallets discussed above.
Figure 8a very clearly shows the unloading and Figure 7b shows a too large vertical strain increase
reloading cycles, from which the decrease of the elastic at the initial load step as a consequence of the compres-
modulus can be determined. From this graph, the elas- sion of the Teflon sheet when the first load is applied.
tic modulus does not appear to decrease substantially All specimens reached the tertiary creep phase very
when damage increases as the slope of the reloading fast during the last load step as a consequence of the
curves remains rather constant. short time intervals taken. Tertiary creep would proba-
The evolution of the horizontal and vertical strains bly also have been reached during the second last load
in time for both tests is indicated in figures 7b-8b. Both step if the time increment could have been longer. Due
curves show an elastic deformation at stress increase, to the time restriction (the test had to be concluded
a small primary creep phase at the beginning of each within one day) this was not possible.

751
Figure 8a. Stress-strain curve for a cyclic accelerated creep
test on masonry cores.

Figure 9. Acoustic emission measurement during acceler-

ated creep test (ACT), events indicated in time by differential
bars and cumulative (above). Load increment steps during
ACT (below).

Figure 8b. Strain evolution in time for a cyclic accelerated

creep test on masonry cores. sensors are attached to the masonry by means of a thin
metal plate which is carefully glued on the surface.
Table 1. Dimensions of the test specimen and resulting A vacuum grease is used as a couplant in between the
compressive stress during short-term creep tests. sensor and the metal plate. The preamplifier gain is set
to 49 dB and a threshold level of 38 dB is applied. The
Test diameter height fc Type of measurements were carried out with equipment from
specimen (mm) (mm) (MPa) test Vallen Systeme, type AMS3.
Figures 9–10 shows a clear relation between the
Core 1 113 100 – Failure step1 acoustic emission measurement and the stress path
Core 2 113 100 4.7 ACT followed during the accelerated creep tests as well
Core 3 113 100 5.4 ACT
Core 4 113 180 4.4 CACT
as during the cyclic accelerated creep tests. At every
Core 5 113 180 6.3 CACT stress increase, a peak in acoustic emission detection
is noticeable. After this peak value, the AE detection
drops to a much lower level. During the first steps of
constant stress, this AE detection level is very low and
Table 1 indicates the dimensions of the test spec- fades away after about half an hour. As the stress level
imen and the maximum compressive stress (fc ) increases, the AE detection during constant stress also
obtained during the short-term creep tests. increases, which results in an AE rate increase between
In parallel with the deformation measurements, the successive load steps. This is clearly visible through
acoustic emission activity was detected by means of the increasing slope of the cumulative curve and can
two AE sensors, one on each side of the specimen. The be related to the strain rate increase, noticed from the

752

Figure 10. Acoustic emission measurement during cyclic
accelerated creep test (CACT), events indicated in time by
differential bars and cumulative (above). Load increment
steps during CACT (below).
graphs which indicate the evolution of strain in time
(Figs 7b–8b).
During the CACT’s, it was noticed that a negligi-
ble amount of AE events were detected during the
reloading step until the stress level, reached during
the preceding load step, was exceeded. This phe-
nomenon is in literature described as the Kaiser effect,
named after Joseph Kaiser who first investigated this
phenomenon in the 1950’s.
To enable a closer analysis of theAE data, the instan-
taneous and the time-dependent damage detection will
be split by separating the emissions during stress
increase from the ones during constant stress inter-
vals. In terms of strain, this can be seen as the damage
occurring during elastic deformation and the damage
during time-dependent deformations. Only the CACT
is included in this analysis, as a clearer distinction can
be made between both phases for this type of test.
3.2.1 AE detection during stress increase
Figures 11–12 show results for CACT on two masonry
cores. The elastic modulus first increases, due to the
753
Figure 11. Evolution of elastic modulus and AE counts
during successive load steps for test core 4.
Figure 12. Evolution of elastic modulus and AE counts
during successive load steps for test core 5.
compaction of the material, until a maximum value is
reached and then decreases due to the damage increase.
As the number of detected counts is proportional
to the total emitted energy, and therefore related to
the total damage, the AE measurement is presented in
terms of AE counts. The detected AE counts during
stress increase appear to evolve inverse compared to
the Young’s modulus, with a pronounced increase dur-
ing the last load steps. This indicates that much more
damage is growing during these last stress increases.
As not all stress increase steps were equally large,
the AE counts indicated are calculated per unity of
stress. And to take into account the Kaiser effect, only
that part of the stress increase is considered which
exceeds the maximum stress level of the previous step.
3.2.2 AE detection during constant stress level
As the strain rate increased for the successive load
steps during both ACT and CACT, clearly showing
a stress-dependent behaviour, this behaviour is also
expected for the AE count rate. Figure 13 confirms

Figure 13. AE counts per second in function of stress level
during constant stress for two cyclic accelerated creep tests.
The stress is indicated relatively in relation to the maximum
stress.
this behaviour. During the last steps, the AE count
increases for both cyclic accelerated creep tests pre-
sented. This means that the damage accumulation rate
during constant stress depends on the load level. The
stress in Figure 13 is indicated relatively in relation to
the maximum stress obtained during the creep tests,
which is indicated as 1.
As not all constant stress intervals had an equal
duration, the AE count rate (counts per second) is indi-
cated in stead of the total amount of counts during a
period of constant stress.
During the last load step, a very high, unstable dam-
age growth is detected which leads to failure of the
specimens.
4 CONCLUSIONS
An extensive test program was set up to evaluate the
long-term behaviour of masonry under monotonic and
sustained loading. This test program involved short-
term monotonic compression and accelerated creep
tests and long-term creep tests on masonry wallets with
three types of mortar compositions. The results of this
test program are presented briefly.
These experimental creep tests are very adequate to
gather more insight into the described phenomenon,
but other tools are necessary in order to evaluate the
damage evolution of historical masonry. Therefore,
acoustic emission has been used as a non destruc-
tive technique in order to evaluate the damage evo-
lution during experimental creep testing on small
masonry cores. The tests were performed consider-
ing the knowledge gained during previous creep tests
on masonry and therefore, both accelerated creep tests
and cyclic accelerated creep tests were performed.
754
The first results of the use of the acoustic emission
technique for detecting the damage evolution within
masonry under increasing and persistent compressive
load were discussed. These results have shown that
the rate of detected acoustic emission can be related
to the rate of damage evolution. As the time incre-
ments were too short to reach secondary creep during
all the load steps, further testing is necessary includ-
ing also long-term testing in order to evaluate the
use of AE technique to monitor damage accumula-
tion at constant stress level. Even so, these first results
prove that the acoustic emission technique is a valu-
able tool in assessing damage evolution of masonry
under persistent compressive load.
ACKNOWLEDGEMENTS
The authors express their thanks to the Flemish Fund
for Scientific Research (FWO) for the doctoral grant,
offered to Els Verstrynge.
REFERENCES
Anzani, A., Binda, L., Melchiorri, G. 1995. Time dependent
damage of rubble masonry walls, 4th Int. Masonry Conf.,
London, Vol 2 (7): 341–351.
Binda, L., Gatti, G., Mangano, G., Poggi, C., Landriani, G.S.
1992. The Collapse of the Civic tower of Pavia: A survey
of the materials and structure. Masonry Int. Vol 6 (1):
633–642.
Binda, L., Anzani, A. 1993. The time-dependent behaviour of
masonry prisms: an interpretation. The masonry society
journal. Vol 11 (2): 17–34.
Carpinteri, A., Lacidogna, G. 2007. Damage evolution of
three masonry towers by acoustic emission. Engineering
structures 29: 1569–1579.
Colombo, S., Main, I.G., Forde, M.C. 2003. Assessing dam-
age of reinforced concrete beams using b-value analysis
of acoustic emission signals. Journal of materials in civil
engineering. 15 (3): 280–286.
Eberhardt, E., Stead, D., Stimpson, B., Read, R.S. 1997.
Changes in acoustic event properties with progressive
fracture damage. Int. J. Rock Mech. & Min. Sci. Vol 34
(3–4), paper nr. 071B.
Grossi, C.M., Esbert, R.M., Suarez del Rio, L.M., et al. 1997.
Acoustic emission monitoring to study sodium sulphate
crystallization in monumental porous carbonate stones.
Studies in conservation. Vol 42: 115–125.
Tomor, A.K., Melbourne, C. 2007. Monitoring masonry
arch bridge response to traffic loading using acoustic
emission techniques. Proc. of the 5th int. conference on
arch bridges, Madeira, Portugal, 12–14 september 2007:
281–288.
Verstrynge, E., Ignoul, S., Schueremans, L., Van Gemert, D.
2008. Modelling of damage accumulation in masonry sub-
jected to a long-term compressive load. Proc. of the 6th int.
seminar on structural analysis of historical constructions.
Bath 2–4 July 2008.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Experimental assessment and modelling of in-plane behaviour of

timber floors

M.R. Valluzzi, E. Garbin, M. Dalla Benetta & C. Modena

Department of Structural and Transportation Engineering, University of Padua, Italy

ABSTRACT: In the paper, the assessment of the in-plane behavior of timber floor samples is proposed by
experimental laboratory tests, aimed at characterizing the basic properties of simple construction floor typologies,
common in existing masonry buildings, for the proper selection of structural improvement solutions. On the basis
of the analysis of the testing procedures and systems proposed in literature, a specific set-up has been designed
and realized, in order to minimize undesirable and/or uncertain effects, able to simulate both monotonic and
cyclic shear behaviour. The first experimental results on unreinforced floors are discussed, together with the
calibration of a preliminary FE model.

1 INTRODUCTION 1994, Corradi et al. 2006) and numerical level

The adoption of timber for floors and roofs has been
continuously considered to realize horizontal struc-
tural components, both in old and current buildings,
thanks to their good structural and comfort perfor-
mances, and the easy production and workmanship
(Tampone 1996, Giordano 1999, Barbisan&Laner
1995). Nevertheless, especially in seismic area, due
to the low in-plane stiffness and the probable lack of
effective connections to the main walls, timber floors
and roofs have been subjected to obtrusive upgrad-
ing interventions or even overall substitutions, mainly
obtained with reinforced concrete slabs. Experiences
acquired during the on-site inspections after the most
recent earthquakes in Italy, demonstrated the high inef-
ficiency of this kind of generalized interventions, as
they caused often new and more brittle collapses (Giuf-
frè 1993, Modena 1997, Valluzzi et al. 2001, Binda
et al. 2006).
Therefore, in the last decade, many research works
have been focused on the characterization of more
compatible techniques, able to strengthen and reha-
bilitate original timber floors (Modena et al. 1998 and
2004, Giuriani 2004, Liberatore et al. 2004, Piazza
et al. 2005, Corradi et al. 2006, Gattesco et al. 2007,),
and to provide a suitable “box” behaviour (Tomaževič
1999, OPCM 3431 2005).
In this context, the study of the influence of
deformable floors on the seismic behaviour of existing
masonry buildings has not been sufficiently inves-
tigated up to now, both at experimental (Tomaževic
1991, Mezzina&Dentamaro 1997, Galasco et al.
2001, Liberatore et al. 2004, Abrams&Tena-Colunga
(Moon&Lee 1994, Cohen et al. 2001, Lee et al. 2002,
Kim&White 2004).
Moreover, only few works are available on the
role of the connections between vertical and horizon-
tal structural elements in the dynamic behaviour of
masonry buildings (Tomaževič 1999, Giuriani 2004,
Foraboschi 2005, Sorrentino et al. 2007).
In particular, some laboratory experimental cam-
paigns have been performed on full scale floor sam-
ples or elements, subjected to in-plane monotonic or
cyclic loads (Modena et al. 1998, Giuriani & Plizzari
2003, Peralta et al. 2004, Corradi et al. 2006). They
propose various set-up, related to different basis con-
ception. In Modena et al. (1998), a simple diago-
nal test on square samples of timber floor portions
(125 × 125 cm), rotated of 45◦ and positioned ver-
tically under a press, was used. This test, usually
standardized for masonry walls, is very easy to per-
form and avoids undesirable confining and frictional
effects; however, it is not completely representative
of the real floor shear behaviour, due to the lim-
ited geometrical dimensions of the sample, the way
how the load is applied, and the probable instability
of the boarding. Moreover, only monotonic test can
be easily performed. Alternatively, test set-up using
full-scale specimens laying horizontally were devel-
oped by Giuriani & Plizzari (2003) and Corradi et al.
(2006). These shear testing systems allow monotonic
and cyclic tests without instability of the boarding and
by minimizing frictional effects. Anyhow, they seem
providing a not negligible confining effect, due to the
presence of active steel frames, that force the shear
behaviour all around the floor specimen. In Peralta

755
et al. (2004), a monotonic and cyclic bending test
set-up on horizontal floor samples, treated as simple
supported or double fixed ends beam, was presented.
This procedure enables the investigation of the shear
behaviour of the floor as a part of the flexural defor-
mation, as in the beam model of Timoshenko (1934).
By this system it is possible to perform tests without
the disadvantage above, but it requires specimens in
real scale with large dimensions (about 4 × 8 m, as in
Peralta et al. 2004).
Within this context, an experimental program has
been planned at the University of Padua, in order to
evaluate the shear behaviour of original and strength-
ened timber floors, by using traditional materials. A
new test set-up has been designed and realized, in
order to perform both monotonic and cyclic tests, and
by minimizing many undesired effects and the dimen-
sional limitations. The system has been preliminary
tested with two pilot floor samples, made by five
beams and two types of boarding (raw-finished and
tongue-and-groove shaped in the thickness). Follow-
ing this preliminary phase, one of the two boarding Figure 1. Design of test set-up and structural scheme
systems will be used as basic sample for various provided by steel frame.
strengthened conditions (double-boarding, diagonal
metal stripe or timber board, etc.). The steel articulated quadrilateral was designed
In the paper, the description of the test set-up and such that out-of-plane motion at the specimen bound-
the results of the preliminary pilot samples are pre- aries (bottom, lateral and top) is prevented, as it
sented. The local behaviour measured during the tests happens in buildings. On the contrary, it allows uplift
has been used to calibrate a numerical FE model, able and in-plane deformations of the sample, by minimiz-
to reproduce the global behaviour of the floor, to be ing frictional effects. Floor specimens are inserted in
used for predictive analysis on the further planned the steel frame with the beams placed horizontally,
tests. The validation of the system is also discussed on thus allowing the free rotation of the boarding.
the basis of reliable drifts expected in masonry build- The bottom joist is fully fixed to the bottom steel
ings subjected to horizontal cyclic loads, as available reaction beam by means of bolts. The steel reaction
in literature. beam is connected to the steel basement trough three
load cells by means of mechanical sliding connections,
thus providing the three simple supports. In the hori-
2 EXPERIMENTAL PROCEDURE zontal direction, a hydraulic actuator was used, to load
the specimens at the level of the top beam.Tests are per-
2.1 Test set-up formed in displacement control, and the force applied
For real scale floor specimens, a specific testing at the top of the samples is measured through a load
machine was on purpose designed and realized, to cell. Finally, horizontal and vertical load cells apply the
adequately simulate the monotonic and cyclic shear shear and the bending moment to the floor specimen
behaviour of floors. in its plane.
Starting from the typical advantages of the vertical In Figures 1 and 2, the main characteristics of the
diagonal test, the proposed system consists in a ver- test system and some views of the experiment are
tical shear-bending test similar to the configurations given.
used for testing woodframe shearwalls (Gatto & Uang A suitable instrumentation was placed on both sides
2003). The structural scheme provides three simple of each specimen, to capture the global and local
supports applied at the floor base (Figure 1); it is equiv- shear-bending behaviour. Several transducers were
alent to a cantilever beam, which can apply bending used to record horizontal, vertical and diagonal relative
and shear deformations, as set in the Timoshenko beam displacements (Figure 3).
theory (1934). Thanks to this structural scheme, the
geometrical dimensions of the floor specimens could
2.2 Floor specimens
be almost half of those used in Peralta et al. (2004),
and about a quarter of those usually present in existing The pilot tests were performed on two timber floor
masonry houses. elements (2,2 × 2,2 m), namely F1.M and F2.M, built

756
Figure 2. View of testing machine with sample, and detail
of connection of base beam to frame.
Figure 3. Position of transducers on samples: local (a) and
global (b) quadrilateral on boarding side; global quadrilateral
on back boarding side (c).
757
Figure 4. Dimensions of floor specimens F1.M and F2, as
connected to steel frame.
Figure 5. Detail of boardings: (a) common raw-finished
(F1.M sample); (b) tongue-and-groove shaped thickness
(F2.M sample).
with components in real size. The samples reproduce
a common typology of mono-directional floors, com-
posed by simple supported beams and a transversal
boarding. In particular, five beams 12 × 14 cm in sec-
tion, and a basic boarding 13,5 cm wide and 2 cm thick,
were used (Figure 4). Spruce wood was used for the
construction of all the elements. Each board was joined
to every beam by means of 2 nails Ø2,75 × 60 mm, for
a total of 32 nails for each beam and 10 nails for each
board.
Specimen F1.M is made with common raw-
finished boarding, whereas F2.M has a tongue-and-
groove shaped connection in the boarding thickness
(Figure 5).
 Table 1. Results obtained from tested connections.

PO.F1.M.a PO.F1.M.b PO.F1.M.c Average

Fmax (N) 576 575 561 571

Fest (N) 592 592 592 592
νi,mod (mm) 0,351 0,489 0,587 0,476
Ks (N/mm) 748 539 477 588

Figure 6. Geometry of the push-out samples and view of

test.

Figure 8. Experimental behaviour of connections.

Figure 7. View of sample after test: steel connections fail

under pure shear followed by unthreading of nails at ultimate
stage.

3 EXPERIMENTAL RESULTS AND ANALYSIS

3.1 Characterization of materials and connections

Preliminary laboratory tests were performed to char-
acterize the materials of components and their connec-
tions, made by steel nails. The mechanical properties
of the spruce wood used in the floor specimens were:
455 kg/m3 as volume mass, 44 N/mm2 as compres-
sive strength, 66 N/mm2 as flexural strength, whereas
a longitudinal elastic modulus of 11000 N/mm2 was
estimated, in agreement with Giordano (1999).
The characterization of the connection was per-
formed by means of push-out tests carried out under
loading control and according to the UNI-EN 26891
standard (Figure 6). The spacing of the nails was in
agreement to the requirements of the Eurocode 5.
The moisture content of the specimens was measured
according to the UNI 9091 after their construction
and before carrying out the tests. The mean value was
around 12%.
To simulate the connection made by nails between
the beam and the basic boarding, three specimens
built with raw-finished boards where tested. The
connection behaves under pure shear, as shown in
Figure 7. According to Eurocode 5, the following
main parameters were evaluated: maximum load at
failure Fmax and maximum estimated load Fest , modi-
fied initial slip νi,mod , and corresponding slip modulus
Ks = 0,4Fest /νi,mod . Results are given in Table 1 and in
Figure 9. Deformation of specimen F1.M at maximum
displacement.
Figure 8. The connection show a good initial stiffness,
followed by a sudden lower branch, due to the yielding
of the nails at the beam-boarding interface.
3.2 Test results and analysis of floor specimens
Two preliminary monotonic tests were performed
to validate the new test set-up and investigate the
behaviour of the unstrengthened floor specimens F1.M
and F2.M, that constitute the basic floor configuration
to which the strengthening methods will be applied.
Also in this phase, the moisture content of the floor
specimens was measured; according to UNI 9091,
after their construction and before carrying out the
tests, the mean value was around 12%.
The final phase of a test is depicted in Figure 9,
whereas the load-displacement trend of the two

758
Figure 10. Comparison among load displacement curves of Figure 11. Comparison between in-plane floor and in-plane
specimens F1.M and F2.M, and schematization proposed by masonry wall deformations.
UNI EN 12512.
and an ultimate stiffness respectively 256% and 42%
Table 2. Results of shear test on floor specimens. higher than those provided by F1.M.
A first estimation of the effect of the in-plane defor-
F1.M F2.M mation of timber floors in the behaviour of masonry
subjected to lateral actions can be also suggested,
Fmax (30 mm) (N) 1047 1435
Fy,est (N) 774 901
by considering a simplified case, represented by a
Vy,est (mm) 8,60 2,67 masonry box having common dimensions of walls
Ki = tgα (N/mm) 80,93 288,29 (4 m long and 3 m high, as in Figure 11) and covered
tgβ (N/mm) 13,49 19,22 by a plain floor. The estimation of drifts (δ) normal-
tgα/tgβ – 6 15 ized to the height of the wall (H) for masonry piers
available from literature, led to the following values:
0,2 ÷ 0,3% is the common range for the achievement
samples is graphically compared in Figure 10. Both of the first cracking, 0,4% for the shear cracking,
specimens showed a failure characterized by a relative 0,5 ÷ 0,6% for the attainment of the maximum load,
rotation between the boards and the beams (Figure 9), whereas 0,8% refers to the rocking behaviour (da Porto
which involved the resistant mechanism of the moment 2005, Tomaževič et al. 1996, Tomaževič & Lutman
of the couple provided by the pair of nails fastened at 1996, Tomaževič 1999).
each beam. By dividing the shear load by the span length
The contribution of additional friction due to the (L = 4 m), and referring it to the ratio between the
tongue-and-grove connection during the relative slip maximum relative floor shear displacement (δf ) and
at the board sides, influenced the better performance of the height of the wall (H = 3 m), see box in Figure 11,
the F2.M sample, in comparison with the F1.M one. By the previous limits become: 0,23% for first cracking;
considering the conventional displacement of 30 mm 0,30% for shear; 0,45% for maximum load; and 0,60%
as reference (UNI EN 12512), an increment of strength for rocking. It can be seen how the already poor capa-
of 37% was found (Figure 10 and Table 2). bility in redistributing the horizontal seismic forces is
To analyze the results, by assuming that monotonic worsened by the loss of shear stiffness in correspon-
curves usually constitute the envelope of cyclic tests, dence of the damage zone of masonry walls, when the
for general timber structure built with metal fasteners, redistribution needs are very desirable to assure a good
the UNI EN 12512 standard can be also used (Ceccotti seismic behaviour (Figure11).
et al. 2005). Since the F2.M sample showed better performances
The results are summarized in Table 2 and repre- than the F1.M one, its results will be considered for
sented in Figure 10, where symbols are as follows: the calibration of a FE numerical model, as described
Fmax is the maximum load at 300 mm of displacement, in the following. Moreover, the boarding with tongue-
Fy,est is the estimated load at yielding, Vy,est is the dis- and-groove connection will be used as basic type for
placement correspondent to Fy,est , and α and β are the the planned strengthened cases.
angles of the slope of first and the second branch of
the rectified curve, respectively, built as described in
the box of Figure 10. 4 NUMERICAL ANALYSIS
Also in terms of yielding load the F2.M sample per-
formed better (Table 2), with an increment of 16% A simplified numerical model was developed in order
of the reference load. Moreover, it showed an initial to identify the behaviour of the plain sample F2.M.

759
Figure 12. Numerical model: (a) detail of connection
between board and beam, (b) general view of frame.
Figure 13. Comparison between experimental and numeri-
cal data for the two small diagonal transducer 3 and 4 located
on boarding side.
The beams were modelled with two-node beam ele-
ments, while for the boards four-node shell elements
were used. The two nails fastening each board at each
beam were modelled by two non-linear connection ele-
ments with six DOF. These elements were joined to the
beam ones through rigid-link elements. The frictional
effect between boards was modelled by means of non-
linear contact points, working only in compression
(Figure 12).
The beams and the boards were modelled by
using the same orthotropic elastic linear mate-
rial, by assuming the properties of spruce wood:
E1 = 11000 N/mm2 is the longitudinal elastic mod-
ulus; E2 , E3 = 367 N/mm2 are the transversal elas-
tic moduli; G12 = 687 N/mm2 is the shear modulus;
ν12,13 = 0,46, and ν21,31 = 0,03, are the Poisson coef-
ficients in the combined directions 1,2,3 (see Fig. 12
for reference).
The law derived from push-out tests (Figure 8) was
assigned to the two directions, as relative displacement
of the connections elements between beam and boards.
Finally, the steel frame was modelled as isotropic elas-
tic (elastic modulus E = 210000 N/mm2 , and Poisson
coefficient ν = 0,2).
The estimations of the model were compared with
the results of the experimental tests at different levels.
First, the load-displacement curves obtained locally
760
Figure 14. Comparison between experimental and numeri-
cal data for the two big diagonal transducer 7 and 8 located
on the boarding side.
Figure 15. Comparison between experimental and numeri-
cal data for the two big diagonal transducer 11 and 12 located
on the back boarding side.
at every diagonal transducer (see Figure 3) were ana-
lyzed: results are depicted through Figures 13 to 15.
Then, the global drift of the whole specimen was con-
sidered (Figure 16). It can be seen a good agreement
at both levels, as the numerical modelled curves are
able to reproduce the shear behaviour of the sample.
The model, sufficiently calibrated on experimental
basis, was implemented for the analysis of the whole
floor sample, as given in Figure 17.
It will be used for predictive analyses on the
strengthened configurations, planned for the prosecu-
tion of the research.
5 CONCLUSIONS
The preliminary results of an experimental campaign
aimed at characterizing the mechanical behaviour of
timber floor specimens in different strengthened con-
ditions have been presented. A new set-up has been
realized, able to perform monotonic and cyclic tests
by keeping the sample vertical, and to reduce several
problems during the test execution.

Figure 16. Comparison of experimental and numerical
results of shear behaviour of F2.M floor sample.
(a)
(b)
Figure 17. Principal stresses at 30 mm displacement:
(a) tension, (b) compression.
Monotonic tests on two plain floor samples have
evidenced the better performances of tongue-and-
grove connections among boards. The analysis of
results pointed out the need of strengthening and
stiffness of the horizontal structures, when floors are
considered restrained as in common buildings.
761
A simplified numerical model, calibrated with the
experimental results, has shown a very good agree-
ment with the experiments, both at local and global
level, and will be considered for the design and the
predictive analysis on the further samples.
The research will prosecute with monotonic and
cyclic tests on timber floor panels strengthened with
different methods (double boardings, metal diagonal
strips or diagonal board, etc.).
ACKNOWLEDGEMENTS
The authors wish to acknowledge the company “F.lli
Bozza Legnami” of Busa di Vigonza (Padova), for
supplying the wood and preparing the specimens.
Thanks also to F. Mariga and M. Novello for the data
processing, and to the technical staff of the Labora-
tory for Structural Material Testing at the Department
of Structural and Transportation Engineering of the
University of Padova. This paper has been partially
supported by RELUIS (Rete dei laboratori universi-
tari di ingegneria sismica) in the framework of the
National executive project 2005–2008 (Attuazione
Accordo di Programma Quadro DPC-Reluis del
15 Marzo 2005).
REFERENCES
Abrams D.P., Tena-Colunga A. (1994). Structural evaluation
of low-rise masonry buildings with flexible diaphragms
subjected to earthquakes, X IB2MaC, Calgary, Canada.
Barbisan U., Laner F. (1995). I solai in legno. Franco Angeli
Ed.: Milano, Italy.
Binda L., Cardani G., Saisi A., Valluzzi M.R. (2006). Vul-
nerability analysis of the historical buildings in seismic
area by a multilevel approach. Asian Journal of Civil
Engineering, Vol. 7, n. 4, pp. 343–357.
Ceccotti, A., Follesa, M., Lauriola M.P. 2005. Le strutture di
legno in zona sismica. CLUT: Torino, Italy.
Cohen G.L., Klingner R.E., Hayes J.R., Sweeney S.C. (2001).
Seismic Response of Low-Rise Masonry Buildings With
Flexible Roof Diaphragms, US Army Corps of Engineers.
Corradi, M., Speranzini, E., Borri, A., Vignoli, A. 2006. In-
plane shear reinforcement of wood beam floors with FRP.
Composites: Part B, Iss. 37, pp. 310–319.
da Porto, F. 2005. In-plane cyclic behaviour of thin layer joint
masonry walls. Ph.D. Thesis, University of Trento: Trento,
Italy.
Foraboschi P. (2005). Consolidamento statico e adeguamento
sismico. Annotazioni per le costruzioni in muratura,
L’Edilizia, n. 138, Italy.
Galasco A., Lagomarsino S., Penna A. (2001). Analisi sis-
mica non lineare a macroelementi di edifici in muratura,
ANIDIS 2001, Italy.
Gattesco, N., Marcorini, L. 2006. Strengthening and Stiff-
ening Ancient Wooden Floors with Flat Steel Profiles,
5th Structural Analysis of Historical Constructions, Proc.
intern. symp., Delhi, India, 6–8 November 2006. India:
MacMillan, pp. 405–412.
Gattesco, N., Macorini L., Benussi F. 2007. Intervento sui
solai lignei per l’adeguamento sismico di edifici storici
con tecniche caratterizzate da elevata reversibilità, ANI-
DIS 2007, Italy (CD-Rom).
Gatto, K. & Uang, C.M. 2003. Effects of Loading Protocol on
the Cyclic Response of Woodframe Shearwalls. Journal of
Structural Engineering, Vol. 129, n. 10, October 1, 2003.
ASCE, pp. 1384–1393.
Giordano, G. (1999). Tecnica delle costruzioni in legno.
Hoeply: Milan, Italy.
Giuffrè A. (1993). Sicurezza e conservazione dei centri
storici: il caso di Ortigia, Laterza: Bari, Italy.
Giuriani, E, Plizzari G.A. 2003. Studio sperimentale sul com-
portamento di solai in legno rinforzati con lastre di acciaio
per resistere alle azioni sismiche, IV Workshop Italiano
sulle Strutture Composte, Italy.
Giuriani, E. 2004. L’organizzazione degli impalcati per gli
edifici storici. L’Edilizia, Speciale Legno Strutturale, n.
134, pp. 30–43.
Kim S., White D.W. (2004). Nonlinear analysis of a one-
story low-rise masonry buildings with flexible diaphragms
subjected to seismic excitation, Engineering Structures, n.
26, Elsevier.
Lee D., Kim H., Chun M. (2002). Efficient seismic analysis
of high-rise building structures with the effect of floor
slabs, Engineering Structures, n. 24, Elsevier.
Liberatore D., Perillo G., Spera G. (2004). L’intervento a solai
deformabili per il miglioramento sismico di strutture in
muratura, ANIDIS 2004, Italy.
Mariani, M. 2004. Consolidamento delle strutture lignee con
l’acciaio, DEI srl Tipografia del Genio Civile, Roma,
Italy.
Mezzina M., Dentamaro C. (1997). Adeguamento e miglio-
ramento degli edifici in muratura ordinaria, ANIDIS 1997,
Italy.
Modena, C. 1997. Criteria for cautious repair of his-
toric building. A valuation and strengthening of existing
masonry structures, Binda L. and Modena C., Ed: RILEM.
Modena, C., Tempesta, F., Tempesta, P. 1998. Il recupero a
secco di impalcati in legno. L’edilizia 3/4, pp. 38–45.
Modena, C., Valluzzi, M.R., Garbin, E., da Porto, F. 2004.
A strengthening technique for timber floors using tra-
ditional materials, 4th Structural Analysis of Historical
Constructions, Proc. intern. symp., Padova, Italy, 10–13
November 2004. Rotterdam: Balkema. pp. 911–921.
Moon S.K., Lee, D.G. (1994). Effect of inplane floor flex-
ibility on the seismic behaviour of building structures,
Engineering Structures, vol. 16, n.2.
762
OPCM 3431 (2005). Ordinanza 3431/2005 del 3 Maggio
2005: Norme tecniche per il progetto, la valutazione e
l’adeguamento sismico degli edifici. P.C.M., Italy.
Piazza, M., Tomasi, R., Modena, R. 2005. Strutture in Legno.
Ulrico Hoepli: Milano, Italy.
Peralta, D.F., Bracci, J.M., Hueste, M.B.D. 2004. Seismic
behaviour of wood diaphragms in pre-1950s unrein-
forced masonry buildings. Journal of Structural Engi-
neering, Vol. 130, No. 12, December 1, 2004, ASCE,
pp. 2040–2050.
prEN 1995-1-1 (Final Draft, Dec. 2003): Eurocode 5. Design
of timber structures - Part 1-1: General–Common rules
and rules for buildings.
Sorrentino, L., Monti, G., Kunnath, S., Scalora, G. 2007.
Un modello meccanico semplificato accoppiato nel
piano-fuori del piano per valutare il ruolo di solai, immor-
sature, qualità muraria e muri di controvento, ANIDIS
2007, Italy (CD-Rom).
Tampone, G. 1996. Il restauro delle strutture di legno. Ulrico
Hoepli: Milano, Italy.
Timoshenko S. 1934. Theory of Elasticity. McGraw-Hill:
New York, US.
Tomaževič M. (1991). The influence of rigidity of floors on
the seismic behaviour of old stone-masonry buildings,
European earthquake engineering, Vol. 5, n. 3.
Tomaževič, M. 1999. Earth quake resistano design of
masonry buildings. Imperial College Press: London, UK.
Tomaževič, M., Lutman, M., Petkovic, L. 1996. Seismic
behaviour of masonry walls: Experimental simulation.
Journal of structural Engineering, Vol. 122, n. 9, 1996,
ASCE, pp. 1040–1047.
Tomaževič, M. & Lutman, M. 1996. Seismic behaviour
of masonry walls: Modelling of hysteretic rules. ASCE
Journal of structural Engineering, Vol. 122, n. 9, 1996,
pp. 1048–1054.
UNI EN 26891. 1991. Timber structures. Joints made with
mechanical fasteners. General principles for the determi-
nation of strength and deformation characteristics. UNI:
Milano, Italy.
UNI 9091-1. 1987. Wood. Determination of moisture content.
Electrical method. UNI: Milano, Italy.
UNI EN 12512. 2006. Timber structures. Test methods.
Cyclic testing of joints made with mechanical fasteners.
UNI: Milano, Italy.
Valluzzi M.R., Michielon E., Binda L., Modena C. (2001).
Modellazione del comportamento di edifici in muratura
sotto azioni sismiche: l’esperienza Umbria – Marche,
ANIDIS 2001, Italy (CD-Rom).
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Bond behaviour of CFRP and GFRP laminates on brick masonry

M. Panizza, E. Garbin, M.R. Valluzzi & C. Modena

Department of Structural and Transportation Engineering, University of Padova, Italy

ABSTRACT: In the last decade, innovative technologies have been developed using Fiber Reinforced Polymer
(FRP) as strengthening and repair of masonry structures. Bond of FRP to substrate is crucial for the effectiveness
of the technique especially to masonry substrate, which can have a wide variability. Few contributions are available
concerning debonding problems on masonry. In this paper the main results of an experimental campaign on the
local behaviour of externally bonded FRP, applied on clay bricks, are presented. Double-lap Push-pull Shear Tests
have been performed by using carbon and glass fiber reinforcement. Experimental results, in terms of failure load,
have been compared with predictive bond-strength models proposed in literature, mainly available for concrete.
Based on the measured strength, interface fracture energy has been calibrated. A simplified analytical model,
fitted on the experimental data, has been proposed as bond-slip law. Finally, a bilinear function, as commonly
adopted by some guidelines, has been calibrated.

1 INTRODUCTION This set-up is based on the assumption that the

The application of externally bonded textiles is a
developing technique for strengthening of masonry
structures. The bond between these products and sup-
port is a crucial aspect to clarify, as it strongly influence
the effectiveness of the intervention.
In the last decade, bond of composite laminates
on concrete substrate has been deeply investigated.
The characterization of bond behaviour has been per-
formed by means of different test set up. The most
commonly used are: Single-lap Shear Test (Chajes
et al. 1996, Täljsten 1997, Bizindavyi & Neale 1999),
Double-lap Pull-pull Shear Test (Lee et al. 1999,
Nakaba et al. 2001), Double-lap Push-pull Shear Test
(Camli & Binici 2007) and Beam-type Test (De Loren-
zis et al. 2001). Wide reviews of available strength or
bond-slip models were given by Chen & Teng (2001),
Lu et al. (2005) and Karbhari et al. (2006).
On the other hand, few investigations concerning
debonding on masonry substrate are available, such as
Aiello et al. (2005) that investigated bond on natural
stones, and Briccoli Bati et al. (2007), that tested bond
on solid clay bricks.
They adopted Double-lap Push-pull Shear Tests,
also known in literature by different names, such
as Double-shear Push Test or Near-end Supported
Double-shear Test (Yao et al. 2004). It consists in
loading in tension two reinforcement strips, symmetri-
cally connected to the support, in order to create shear
stresses at the interface; the brittle support is subjected
to compressive stresses.
applied load is equally distributed on the two strips,
but it is also particularly simple and suitable for the
usual common available test machine.
This experimental procedure has been adopted for
the present work, that aims at giving a contribution to
the characterization of the bond behaviour of carbon
and glass textiles externally bonded to clay substrate.
The results of five samples for high-strength car-
bon reinforcement and five samples for alkali-resistant
glass reinforcement are presented and discussed.
The predictions of twenty-one bond-strength models,
available in literature for concrete as parent material,
have been compared with the measured strength. The
fracture energy of the composite-clay interface has
been evaluated through the experimental failure loads.
A simplified bond-slip law has been proposed on the
basis of the data obtained from load and strain moni-
tored during the tests; furthermore, a bilinear function
has been also calibrated.
2 EXPERIMENTAL TESTS DESCRIPTION
2.1 Materials characterization
Solid clay bricks (nominal dimension 250 × 120 ×
55 mm) were used as substrate, and the MBrace©
Wet lay-up System as reinforcement. High strength
carbon fibers (CFRP) on five specimens and alkali-
resistant glass fibers (GFRP) on other five specimens
were used.

763
 Table 1. Bricks mechanical properties.

Mean cubic compressive strength 50.94 MPa

Mean direct tensile strength 2.37 MPa
Mean splitting tensile strength 3.99 MPa
Mean flexural tensile strength 5.46 MPa
Secant elastic modulus 16,100 MPa

Table 2. Reinforcement components properties.

Adhesive MBrace© Saturant Figure 1. Geometry of specimens.

Characteristic compressive strength >80 MPa
Characteristic direct tensile strength >50 MPa
Maximum tensile strain 2.5%
Tensile elastic modulus >3000 MPa
High-strength Carbon MBrace© C1-30
Equivalent thickness 0.165 mm
Characteristic direct tensile strength 3430 MPa
Maximum tensile strain 1.5%
Tensile elastic modulus 230,000 MPa
Alkali-resistant Glass MBrace© G60-AR
Equivalent thickness 0.230 mm
Characteristic direct tensile strength 1700 MPa
Maximum tensile strain 2.8%
Tensile elastic modulus 65,000 MPa

Figure 2. Test machine (left) and a specimen ready for

The main bricks properties are summarized in testing (right).
Table 1 (Cartolaro 2004). Reinforcement system
properties, obtained from producers datasheets, are
reported in Table 2.

2.2 Experimental test set-up

The specimen was made by a single clay brick with two
strips of reinforcement externally bonded (wet lay-up
system), symmetrically applied on the opposite wider
surfaces. Each strip was 50 mm wide and bonded to
the brick for 200 mm (Fig. 1).
An unbonded length from the limit of the brick,
equal to 30 mm, was imposed next to the loaded end,
in order to minimize edge effects.
An universal mechanical press, Galdabini Sun60
(maximum load 600 kN), was used as test machine
(Fig. 2).
Each strip of reinforcement, made by a single layer
of fibers and two layers of epoxy resin, was bonded,
at the loaded end, to a steel support connected to the
test machine. Brick was connected to the machine
through a steel frame, made by two plates linked by
bolts (Fig. 2).
The load was applied axially and tests were con-
trolled by a displacement rate of 0.2 mm/minute.
On the outer side of one of the two reinforcement
strip for each specimen, seven strain gauges were
applied, distributed as follows: one on the unbonded
zone, next to the loaded end of the reinforcement, and
Figure 3. Distribution of the strain-gauges.
six on the bonded one. To optimize the number of
instruments and to monitor the whole bonded region,
the strain-gauges were not equally distributed, but less
spaced near the loaded end (Fig. 3).
3 TEST RESULTS AND DISCUSSION
After test, all specimens revealed the complete detach-
ment of the reinforcement from the support. The fail-
ure involved the brick surface (Fig. 4), where curved
cracks and ripping of clay pieces were observed.
The failure loads, Pu , are given Tables 3–4 and plot-
ted in Figure 5. It is possible to note as specimens
strengthened with CFRP showed better performances
than GFRP one. The mean failure load in the first case
was around 35% higher than the latter.

764
 Figure 5. Experimental failure loads per unit width.

Figure 4. Brick and composite surfaces after the test. elastic behaviour of the composite, it was also possible
to evaluate Young’s modulus of the reinforcement for
each sample through a best fitting (Eq. 1):
Table 3. Experimental results for carbon reinforcement.

Ef Pu Pu /2bf σu
Specimen MPa N N/mm MPa

ShC1 164,419 31,884 318.8 1932 where σ = nominal tensile stress, P = axial load
ShC2 336,439 34,233 342.3 2075 on the composite strips, bf = single strip width,
ShC3 284,991 35,325 353.3 2141 t = equivalent textile thickness, ε = axial strain mea-
ShC4 277,511 39,210 392.1 2376 sured by the strain-gauge and Ef = composite elastic
ShC5 338,456 40,301 403.0 2442
modulus.
Mean value 280,363 36,191 361.9 2193 Tables 3–4 also report the composite elastic mod-
Stand. dev. 70,696 3505 ulus values, Ef , and the nominal tensile stresses at
COV 25.2% 9.7% failure, σu . The experimental mean elastic moduli
result higher than the producer’s values (22% for car-
bon reinforcement and 25% for glass one). The mean
Table 4. Experimental results for glass reinforcement. maximum stress reached by reinforcement is around
64% of tensile strength for carbon, and 68% for glass.
Ef Pu Pu /2bf σu As the reinforcement axial stiffness per unit width,
Specimen MPa N N/mm MPa Ef tf , was known for each specimen, failure loads per
unit width, P/2bf , were tabulated versus the axial stiff-
ShG1 50,934 23,380 233.8 1017 ness. Trend lines were fitted, referring to all data or to
ShG2 87,014 27,940 279.4 1215
each set (carbon and glass). The expression adopted
ShG3 80,545 27,300 273.0 1187
ShG4 102,598 26,400 264.0 1148 for the trend lines is given in Equation 2:
ShG5 84,842 28,360 283.6 1233
Mean value 81,035 26,676 266.9 1160
Stand. dev. 18,817 1985
COV 23.2% 7.4%
where c1 e c2 are regression constants, which values
are reported in Table 5. It can be observed that the
It was assumed that strain is uniform on the com- fitting of all data shows a better correlation than the
posite cross section, and that is possible to refer fitting of each single set.
mechanical properties to the dry woven: this approach By assuming the relationship between the failure
is accepted by many authors (Chen & Teng 2001) load and the square root of the axial stiffness per unit
and by some guidelines, such as Italian CNR-DT width (Eq. 7), it was possible to reduce the number
200/2004. of free parameters in Equation 2, by imposing the
Therefore, it was possible to calculate the nominal exponent value c2 equal to 0.5. Results (carbon data
tensile stress on the textile, defined as the load and set, glass data set and all data) are given in Table 5,
cross section area ratio. By coupling the strain mea- whereas Figures 6–7 compare the trend lines with the
sures (obtained from the strain-gauge on the unbonded experimental data. Regression coefficient for GFRP
region) with the load values, and by assuming the linear are slightly higher than CFRP (around 16%) and this

765
Table 5. Regression constants of load vs axial stiffness Table 6. Predictions of failure load.
(both per unit width) trend lines.
CFRP GFRP
Data set c1 c2 R2
Model Pu /2bf Error Pu /2bf Error
All data 12.876 0.310 0.876 N/mm N/mm
Glass data only 27.197 0.234 0.622
Carbon data only 35.063 0.218 0.439 Tanaka 166 −54.0% 166 −37.6%
All data (Square Root) 1.759 0.5 n.a. Hiroyuki and Wu 158 −56.2% 158 −40.6%
Glass data (Square Root) 1.953 0.5 n.a. Maeda 254 −29.7% 174 −34.9%
Carbon data (Square Root) 1.681 0.5 n.a. Khalifa 248 −31.4% 170 −36.2%
Yang 192 −47.0% 143 −46.3%
Sato 415 +14.6% 116 −56.5%
Iso 271 −25.1% 161 −39.5%
Izumo 557 +53.9% 224 −15.9%
Neubauer and R. 283 −21.8% 179 −32.7%
Chen and Teng 245 −32.2% 156 −41.6%
Monti et al. 321 −11.3% 204 −23.6%
Lu et al. Bilinear 220 −39.3% 139 −47.7%
Brosens and V. G. 359 −0.9% 228 −14.7%
CNR−DT 200 263 −27.4% 167 −37.5%
Nakaba et al. 350 −3.3% 222 −16.7%
Savoia et al. 328 −9.4% 208 −22.0%
Neubauer and R. 266 −26.4% 169 −36.6%
Dai and Ueda (1) 326 −9.9% 207 −22.5%
Figure 6. Failure loads per unit width versus reinforcement Dai and Ueda (2) 322 −11.0% 202 −24.2%
axial stiffness per unit width: experimental data and trend Lu et al. Precise 220 −39.3% 139 −47.7%
lines. Lu et al. Simplif. 220 −39.3% 139 −47.7%
Mean experim. 362 – 267 –

Yang, Sato, Iso (reported in Lu et al. 2005), express

failure load as the product of an area and a nominal
average tangential stress, τu (Eq. 3).

where bf is the reinforcement width, Le is an effec-

Figure 7. Failure loads per unit width versus reinforcement
axial stiffness per unit width: experimental data and trend
lines based on the axial stiffness square root.
could be significant for the fracture energy evaluation
as given in the following.
4 PREDICTION OF STRENGTH
Many predictive models have been developed to esti-
mate the failure load of the composite-to-concrete
bonded joint. Twenty-one of them were applied in
this study, in order to make a comparison with the
experimental results of the tests on clay substrate.
Among them, the models of Tanaka, Hiroyuki and
Wu, Maeda, Khalifa (reported in Chen & Teng 2001),
tive length and τu a tangential stress, different from
model to model. The models of Izumo (Lu et al. 2005),
Neubauer and Rostàsy, Chen and Teng (both reported
in Chen & Teng 2001) give other expressions for the
failure load. Finally, eleven ones provide an estima-
tion of the fracture energy value Gf , that has been
correlated with failure load through Equation 7. In par-
ticular, the models of Monti (Lu et al. 2005), Lu et al.
(2004), herein Lu Bilinear, Brosens and Van Gemert
(Karbhari et al. 2006) and Italian Research Council
(CNR-DT 200/2004) utilize a bilinear bond-slip law;
the models of Nakaba et al. (2001) and Savoia et al.
(2003a) adopt a Popovics curve as bond-slip law; the
models of Neubauer and Rostàsy (Lu et al. 2005),
Dai & Ueda (2003), Dai et al. (2005) and Lu et al.
(2004), Precise and Simplified models, take on other
types of bond-slip function.
The results are given in Table 7 and shown in
Figure 8. It can be seen that all the predictions, except
for the models of Sato and Izumo in case of carbon

766

Figure 8. Ratio of predicted failure loads vs mean experi-
mental value for carbon and glass reinforcement.
Table 7. Evaluation of fracture energy.
Gf from Eq. 7 Gf from S.R. fit
Reinforcement type N/mm N/mm
Carbon fibers 1.42 1.41
Glass fibers 1.91 1.91
reinforcement, underestimate the mean experimental
failure load. Moreover, all formulations provide an
estimation closer to test results in case of CFRP than
GFRP, except for the Tanaka and Hiroyuki models.
However, results show large differences from model
to model: they vary between 44% and 154% of exper-
imental mean failure load, for carbon reinforcement,
and between 43% and 85%, for glass one.
5 FRACTURE ENERGY CALIBRATION
The interface fracture energy mode II, Gf , is defined
by Equation 4 as the definite integral of the tangential
stress, τ, expressed as function of the mutual slip of
composite and substrate, s:
One of the first analytical models of the concrete-
composite bond strength was derived by Täljsten
(1996), starting both from a linear approach, based
on the beam theory, and from a non-linear approach,
related to fracture mechanics. In case of most of
the epoxy adhesives commonly used, a simplified
formulation, as reported in Equation 5, was obtained.
767
where bf = reinforcement width, Gf = interface frac-
ture energy, αT = constant value, Ec tc = axial stiffness
per unit width of the concrete substrate.
Yuan (Chen & Teng 2001) proposed a modified con-
stant value (Eq. 6) that takes into account the width (bf
and bc ) ratio of the bonded materials:
In most cases, the constant value αT , or αW , has a slight
influence on the calculation. Many authors, such as
Savoia et al. (2003a) and Dai et al. (2005), report the
following formula (Eq. 7) without any constant:
By applying these formulas to the experimental data of
this work, it emerges that taking or not into account the
parameters αT or αW leads to a difference lower than
2%. It has to be noticed that Equation 7, demonstrated
in some cases (Wu et al. 2002; Dai et al 2005), is
assumed in every case of regular interface law (Savoia
et al. 2003b).
It is considered significant also for the clay substrate
adopted in the present work.
Accordingly with Equation 7, it was possible to cal-
ibrate fracture energy Gf through the mean values of
failure load and elastic modulus; results are reported in
Table 6. The estimated value, for glass reinforcement,
is around 35% higher than carbon one.
Moreover, the fitting parameter c2 given in Table 5,
where c1 was imposed equal to 0.5 (Square Root
based fitting), allowed to evaluate Gf , as shown in
Equation 8. The results (Table 6) show no signifi-
cant difference from values obtained by means of
Equation 7.
6 CALIBRATION OF A BOND-SLIP LAW
Figure 9, from 9a to 9e, shows possible shapes
of the bond-slip function: (a) cut-off, adopted by
Neubauer and Rostàsy (Chen & Teng 2001); (b) bilin-
ear, assumed by some guidelines like fib Bulletin
14 (2001) and CNR-DT 200 (2004), and by Monti
(Lu et al. 2005), Brosens and Van Gemert (Karb-
hari et al. 2006) and Lu (Lu et al. 2004); (c) rigid
with linear softening, by Chen & Teng (2001); (d)
a single function, as the Popovics curve chosen by
Savoia et al. (2003a) and Nakaba et al. (2001), or
an exponential curve obtained by Dai et al. (2005);
(e) two different non-linear functions for ascending
and descending branch, for instance the expressions
adopted by Lu et al. (2004) or Dai & Ueda (2003).
Figure 9. Some bond-slip law shapes available in literature.

Therefore, it is commonly assumed that bond of

composite laminates exhibits a softening behaviour, Figure 10. Calibrated bond-slip laws (CFRP).
with an ascending branch followed by a descending
one, and presenting no residual stress for wider slip.
To calibrate the bond-slip law on the experimental
results, this combined approach was adopted: tan-
gential stress and interface slip points (τ–s) were
obtained from strain-gauges monitoring, while the
fracture energy value, Gf , was calculated from failure
loads trough Equation 7.
Fracture energy represents a restraint for the bond-
slip function (Eq. 4) and allows to reduce the number
of free parameters involved in the calibration process.
Equations 9–11 briefly report the main relations
(obtained from simple equilibrium and compatibil-
ity considerations) between reinforcement strain ε, Figure 11. Calibrated bond-slip laws (GFRP).
interface tangential stress τ and slip s, supposing to
disregard the slip component of the substrate, suffi-
ciently stiffer than composite. The notation x indicates where the notation i-nth indicates the strain-gauge
the coordinate along the central axis of the bonded position. The orientation of the x axis makes i increas-
region. ing from loaded end (x = 0) to free end (x = 200 mm).
Hence, it was possible to couple slip values with the
corresponding tangential stresses.
As above explained, it is assumed that the bond-slip
law should show an ascending segment and a softening
behaviour. Instead of using two different mathemati-
cal expressions for the ascending and the descending
branch, a single function was chosen. Although there
could be a slight loss of adherence to experimental
data, it reduces the required parameters making easier
the fitting process.
The proposed law, easy to integrate and derive, is
To calculate, from strain measured in discrete positions given in Equation 14.
along the reinforcement, the corresponding tangential
stress and slip values, Equations 9–10 were modified.
In the present work, the discrete formulas given in
Equations 12–13 (Valluzzi et al. 2003) were used; they
allow to manipulate data from devices not uniformly where A and B are regression constants, τ is the
spaced. interface tangential stress and s the composite slip.
Applying the calibrated fracture energy value, it was
possible to have a fitting function that depends on just
one parameter, as shown in Equation 15.

768
It is useful to rewrite the law, herein labeled UniPd Table 8. Significant values for local bond of CFRP.
curve, in a normalized form (Eq. 16).
Gf τmax s0 sf
Curve N/mm MPa mm mm

UniPd fitting 1.42 7.22 0.072 –

Bilinear fitting 1.42 7.22 0.034 0.392

where s0 = 1 / B and τmax = τ (s0 ) are the coordinates Monti et al. 1.11 5.37 0.046 0.415
of the point of maximum tangential stress. Lu et al. Bilinear 0.52 3.73 0.048 0.280
Brosens and V. G. 1.39 2.71 0.012 1.025
After the optimization of the Unipd curves, in case
CNR 0.75 7.46 0.056 0.200
of carbon reinforcement and glass one, it was possible Nakaba et al. 1.32 7.08 0.065 –
to calibrate a bilinear law, whose analytical form is Savoia et al. 1.16 7.08 0.051 –
reported in Equation 17. Neubauer and R. 0.77 5.69 0.270 –
Dai and Ueda (1) 1.15 8.58 0.103 –
Dai and Ueda (2) 1.12 6.41 0.061 –
Lu et al. Precise 0.52 3.73 0.054 –
Lu et al. Simplif. 0.52 3.73 0.048 –

where sf is the ultimate strain, related to null stress. Table 9. Significant values for local bond of GFRP.
This form is commonly proposed by some guidelines
(fib Bulletin 14 2001; CNR-DT 200 2004). Gf τmax s0 sf
Since the bilinear function depends on more param- Curve N/mm MPa mm mm
eters, to reduce them the peak tangential stress value,
obtained from the fitted Unipd curve, and the cali- UniPd fitting 1.91 6.33 0.111 –
Bilinear fitting 1.91 6.33 0.048 0.603
brated fracture energy were imposed.
Figures 10–11 show the optimized curves and the Monti et al. 1.11 5.37 0.046 0.415
experimental stress-slip data. It can be noticed that car- Lu et al. Bilinear 0.52 3.73 0.048 0.280
bon reinforcement interface seems to be slight stiffer Brosens and V. G. 1.39 2.71 0.012 1.025
than glass one. CNR 0.75 7.46 0.056 0.200
Tables 8–9 report the significant values (fracture Nakaba et al. 1.32 7.08 0.065 –
Savoia et al. 1.16 7.08 0.051 –
energy, peak tangential stress with related slip, and Neubauer and R. 0.77 5.69 0.270 –
ultimate slip) calculated by the fitting process; they Dai and Ueda (1) 1.15 7.10 0.107 –
have been compared with the values estimated through Dai and Ueda (2) 1.10 5.69 0.067 –
the eleven models based on the fracture energy predic- Lu et al. Precise 0.52 3.73 0.054 –
tion, mentioned in chapter 4. Estimated values vary Lu et al. Simplif. 0.52 3.73 0.048 –
into a quite wide range.
It has to be noticed that not all models provide
significant differences from carbon to glass reinforce-
ment; in particular, the difference between the fracture substrate. All predictions, except two in case of CFRP,
energy values experimentally calibrated, in case of underestimate the results of the tests; the models,
CFRP and GFRP, is not in agreement with most except two in case of GFRP, seem to work better for
predictions. the carbon reinforcement. However, the strength pre-
dictions vary into an wide range (between 44% and
154% of experimental mean failure load for CFRP,
7 CONCLUSIONS 43% and 85% for GFRP).
From the measured failure loads, different fracture
The bond behaviour of the composite-clay brick inter- energy values have been derived, around 35% higher
face has been investigated by means of Double-lap in case of glass reinforcement than carbon one.
Push-pull Shear Tests, for both high-strength carbon To analyze stress and slip from strain-gauges mea-
(CFRP) and alkali-resistant glass (GFRP) reinforce- surement, when instruments are not uniformly spaced,
ment. discrete equations have been used, consistent with
Far from being exhaustive, the results show a better central finite difference methods.
performance of carbon reinforcement than glass one, Moreover, a mathematical function easy to integrate
around 36% higher in the first case. and derive is proposed as bond-slip law. This function
The experimental strength has been compared with has been fitted in case of both carbon and glass rein-
twenty-one predictive models developed for concrete forcement; beside these fittings, two bilinear functions

769
have been also calibrated. The optimized functions
seem to show an interface local behaviour of CFRP
slightly stiffer than GFRP.
However, further investigations are needed, to better
inquiry the bond behaviour of composite-clay bricks
interface; this is a first step in order to take into
account, in the future, the role of the mortar joints,
characteristic of masonry structures.
Moreover, the reliability of the experimental set-up
needs to be verified; despite of its simplicity, the actual
distribution of the load should be more clarified, as it
could cause undesired bending moments inside the test
specimen.
ACKNOWLEDGEMENTS
The authors acknowledge Andrea Cartolaro and the
technical staff of the Laboratory of Material Testing
of the Department of Construction and Transportation
of the University of Padova, where tests have been
performed.
This activity has been partially supported by the
National Italian Project ReLUIS.
All materials, including fibers and adhesion system,
have been supplied by Modern Advanced Concrete
(MAC S.p.A.) of Treviso, Italy.
REFERENCES
Aiello, M.A. & Sciolti, M.S. 2005. Bond analysis of masonry
structures strengthened with CFRP sheets. Construction
and Building Materials, 20: 90–100.
Bizindavyi, L. & Neale, K.W. 1999. Transfer Lengths and
Bond Strengths for Composites Bonded to Concrete.
ASCE Journal of Composites for Construction, 3 (4):
153–160.
Briccoli Bati, S., Rovero, L., Tonietti, U. 2007. Adesione fra
blocchi in laterizio e rinforzo in CFRP (in Italian). Materi-
ali ed Approcci Innovativi per il Progetto in Zona Sismica
e la Mitigazione della Vulnerabilità delle Strutture; proc.
nat. workshop ReLUIS, Salerno, Italy.
Camli, U.S. & Binici, B. 2007. Strength of carbon fiber
reinforced polymers bonded to concrete and masonry.
Construction and Building Materials, 21: 1431–1446.
Cartolaro, A. 2004. Volte in muratura di mattoni rinforzate
con FRP. Studio sperimentale dei meccanismi resistenti
alle interfacce (in Italian). MS Dissertation, University of
Padova, Italy.
Chajes, M.J., Finch, W.W.Jr., Januszka, T.F. & Thomson, T.A.
1996. Bond and Force Transfer of Composite Material
Plates Bonded to Concrete. ACI Structural Journal, 93
(2): 295–303.
Chen, J.F. & Teng, J.G. 2001. Anchorage Strength Models for
FRP and Steel Plates Bonded to Concrete. ASCE Journal
of Structural Engineering, 127 (7): 784–791.
770
CNR-DT 200/2004. Guide for the Design and Construc-
tion of Externally Bonded FRP Systems for Strengthening
Existing Structures. National Research Council, Italy.
Dai, J.G. & Ueda, T. 2003. Local Bond stress Slip relations for
FRP Sheets-Concrete Interfaces. Fibre-Reinforced Poly-
mer Reinforcement for Concrete Structures FRPRCS-6;
proc. int. symp., Singapore.
Dai, J.G., Ueda, T. & Sato, Y. 2005. Development of the
Nonlinear Bond Stress-Slip Model of Fiber Reinforced
Plastics Sheet-Concrete Interfaces with a Simple Method.
ASCE J. of Composites for Construction, 9 (1): 52–62.
De Lorenzis, L., Miller, B. & Nanni, A. 2001. Bond of
FRP laminates to concrete. ACI Materials Journal, 98 (3):
256–264.
fib Bulletin 14 2001. Externally bonded FRP reinforcement
for RC structures – Technical Report on the design and
use of externally bonded fibre reinforced polymer rein-
forcement for reinforced concrete structures. Fédération
Internationale du Béton.
Karbhari, V.M., Niu, H. & Sikorsky, C. 2006. Review and
Comparison of Fracture Mechanics-based Bond Strength
Models for FRP-strengthened Structures. J. of Reinforced
Plastics and Composites, 25 (17): 1757–1794.
Lee, Y.J., Boothby. T.E., Bakis. C.E. & Nanni. A. 1999. Slip
Modulus of FRP Sheets Bonded to Concrete. ASCE J. of
Composites for Construction, 3(4): 161–167.
Lu, X.Z., Teng, J.G., Ye, L.P. & Jiang, J.J. 2004. Bond-
slip models for FRP sheet/plate-to-concrete interfaces.
Advanced Polymer Composites for StructuralApplications
in Construction – ACIC 2004; proc. 2nd int. conf., Surrey,
UK.
Lu, X.Z., Teng, J.G., Ye, L.P. & Jiang, J.J. 2005. Bond-
slip models for FRP sheets-plates bonded to concrete.
Engineering Structures, 27: 920–937.
Nakaba, K., Kanakubo, T., Furuta, T., Yoshizawa, H. 2001.
Bond Behavior between Fiber-Reinforced Polymer Lam-
inates and Concrete. ACI Structural Journal, 98 (3):
359–367.
Savoia, M., Ferracuti, B. & Mazzotti, C. 2003a. Non
linear bond-slip law for FRP-concrete interface. Fibre-
Reinforced Polymer Reinforcement for Concrete
Structures FRPRCS-6; proc. int. symp., Singapore.
Savoia, M., Ferracuti, B. & Mazzotti, C. 2003b. Una legge di
interfaccia non lineare per placcaggi con lamine in FRP
(in Italian). 16th AIMETA Congress of Theoretical and
Applied Mechanics; proc., Ferrara, Italy.
Täljsten, B. 1996. Strengthening of concrete prisms using the
plate bonding technique. Int. J. of Fracture, 82: 253–266.
Täljsten, B. 1997. Defining anchor lengths of steel and
CFRP plates bonded to concrete. Int. J. of Adhesion and
Adhesives, 19: 319–327.
Valluzzi, M.R., Tinazzi, D., Garbin, E. & Modena, C. 2003.
FEM modelling of CFRP strips bond behaviour for bed
joints reinforcement techniques. Computer Methods in
Structural Masonry; proc. 6th int. conf., Rome, Italy.
Wu, Z.,Yuan, H. & Niu, H. 2002. Stress Transfer and Fracture
Propagation in Different Kinds of Adhesive Joints. ASCE
Journal of Engineering Mechanics, 128 (5): 562–573.
Yao, J., Teng, J.G. & Chen, J.F. 2004. Experimental study on
FRP-to-concrete bonded joints. Composites: Part B, 36:
99–113.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Experimental shear behavior of stone masonry joints

G. Vasconcelos, P.B. Lourenço & D. Oliveira

Isise, Department of Civil Engineering, University of Minho, Guimarães, Portugal

ABSTRACT: The mechanical characterization of the shear strength properties takes a major role on the evalu-
ation of the lateral strength of masonry shear walls by means of simplified methods or when numerical analysis
based on micromodelling approach is to be followed. Thus, the present paper presents an overview of the results
obtained from direct shear tests of different types of masonry joints: dry and mortar masonry joints. Besides the
shear strength parameters, a good insight was achieved in the evaluation of the complete shear stress-shear load
displacement diagrams.

1 INTRODUCTION For pre-compression stresses above a certain level,

As reported in literature (Hamid & Drysdale 1980,
Samarasinghe & Hendry 1980), the orientation of the
mortar joints to the applied stresses takes a major role
in the ultimate strength and failure modes of masonry
under in-plane stress state. The influence of mortar
joints acting as a plan of weakness on the compos-
ite behavior of masonry is even more relevant in case
of strong unit-weak mortar joint combinations, which
are characteristic of ancient stone masonry. Two basic
failure modes can occur at the level of the unit-mortar
interface: tensile failure (mode I) associated to stresses
acting normal to joints and leading to the separation of
the interface, and shear failure (mode II) correspond-
ing to a sliding mechanism of the units or shear failure
of the mortar joint.
Although several experimental studies have been
carried out in the characterization of the bond shear
strength of unit-mortar interfaces (Atkinson Amadio
& Rajgelj 1991, Binda et al. 1997), lesser research is
available on the shear behavior of dry stacked masonry
joints, even if recent studies have been carried out on
the behavior of dry masonry joints submitted to cyclic
loading (Lourenço & Ramos 2004). On the other hand,
the features of rock joints under shear behavior can
be partly extended to dry masonry joints. The shear
behavior of rock joints has been played an impor-
tant role in the scope of rock mechanics research. In
particular, several experimental and numerical studies
pointed out the role of the surface roughness on the
cyclic shear behavior of natural rock joints (Lee et al.
2001, Huang et al. 2002).
The relation between normal and shear stresses has
a major role in the shear behavior of masonry joints,
governing its failure mode (Hamid & Drysdale 1980).
the shear strength decreases and a combined shear-
splitting failure or splitting of the units occur. In case
of shear failure along the joint by slipping of the units
along the joint, an increase of the compression normal
to bed joint leads to an increase of the shear strength.
As has been widely reported (Atkinson et al. 1989,
Riddington & Ghazali, 1990), the shear strength of
masonry under moderate normal stresses, for which
the nonlinear behavior of mortar is negligible and the
friction resistance takes the central role, can be given
by the Coulomb criterion:
where c is the shear strength at zero vertical load stress
(usually denoted by cohesion) and µ is the friction
coefficient. For dry joints the cohesion is assumed
to be zero. It should be kept in mind that the failure
envelop given by eq.1 describes only a local failure
and can not be directly related to the shear failure of
masonry walls submitted to in-plane horizontal loads
(Mann & Müller 1982, Atkinson et al. 1989, Calvi
et al.1996).
In addition to the knowledge of the mechanical
properties from masonry components, namely units
and mortar, the analysis of masonry behavior under
in-plane loading is only possible if information about
the local composite behavior and the interaction
between units and mortar is available.
Therefore, the present work deals with the mechan-
ical characterization of the shear behavior of dry and
mortar masonry joints (cohesion, friction angle and
dilatancy). In order to attain such goal, an experimen-
tal program was defined, including direct shear tests
conducted on dry and mortar masonry joints. Besides

771
 (a) (b)
Figure 1. Masonry specimens; (a) dry joints; (b) mortar
joints.
ensuring mechanical properties for numerical simula-
tions of the in-plane behavior of stone masonry wall
structures, the adopted testing program provides also
the fundamental information about the shear behavior
of two different masonry joints.
2 EXPERIMENTAL DETAILS
2.1 Test specimens and procedure
Although triplet tests have been adopted as the Euro-
pean standard method (EN1052-3 2002) to perform
shear tests in mortar joints, the shear strength proper-
ties of dry and mortar joints were obtained by means
of direct shear tests conducted on couplet specimens,
see Figure 1 (Vasconcelos 2005). In fact, in the triplet
test, the two joints do not fail at the same time and the
analysis of the experimental results is rather complex,
Lourenço et al. (2004).
The shear tests were carried out in a servocontrolled
universal testing machine CS7400S. This equipment
is composed by two independent hydraulic actua-
tors used to transmit normal and shear loads, able
to operate under force or displacement control. The
features of the testing equipment and the existing
loading platens imply that the most suitable testing
sample is composed by two units with geometry and
dimensions indicated in Figure 1 and a single dry or
mortar joint, similarly to Pluijm (1999) and Hansen
et al. (1998). The surface of the dry masonry units
adopted here is smooth resulting from sawing the
specimens, whereas the joint surface of the units of
the mortar assemblages presents enough roughness
to achieve appropriate adherence conditions and thus
more realistic masonry can be simulated. The detailed
experimental characterization of this type of granite
can be seen inVasconcelos et al. (2007).The specimens
were placed between two thick steel plates and attached
to the steel platens by steel bolts, so that shear force
could be transmitted. Thin steel sheets were attached
to the steel plates to concentrate the shear load as close
as possible in the bed joint, aiming at preventing bend-
ing moments and provide a more uniform shear stress
distribution.
772
Plan view Elevation view
Figure 2. Arrangement of the LVDTs for measuring of the
relative horizontal and vertical displacements.
In addition, two thin sheets of Teflon were interlay-
ered between the steel platens and the specimens to
minimize bending effects. In order to guarantee right
angle surfaces, the specimens were suitably ground
using a rectifying machine. The same procedure was
used in the contact surfaces between both units of the
specimens to ensure the maximum contact area in case
of dry joints. In fact, according to Hansen (1999),
the uneven stress distribution can also be attributed
to the non-uniform distribution of the material along
bed joints. In both types of specimens, when necessary,
a thin layer of glue was placed at the surface in con-
tact with the steel platens in order to provide perfectly
leveled surfaces. The confinement of the specimens
was improved for load reversal by means of a couple
of steel rods fixed to the steel plates through metal-
lic bolts. This arrangement is particularly useful in the
cyclic tests but was also used in case of monotonic
tests. The numerical assessment of the effectiveness of
the test setup was performed by Lourenço and Ramos
(2004) based on a finite element model. It was con-
cluded that although a deviation on the stresses occurs
in the zone adjacent to the steel plates, an almost uni-
form normal and shear stress distribution is achieved
in 63% of the extent of the bed joint.
In order to simulate the usual range of normal
stresses existing in ancient masonry structures three
distinct pre-compression stress levels were applied
under force control, σ = 0.5 N/mm2 , σ = 0.75 N/mm2
and σ = 1.0 N/mm2 in dry joints under monotonic and
cyclic loading. An additional pre-compression stress
level corresponding to σ = 1.25 N/mm2 was consid-
ered for the monotonic tests carried out on unit-mortar
assemblages. Three specimens were tested for each
level of pre-compression for both types of masonry
joints. The possible influence of the moisture content
on the shear response of dry masonry joins was inves-
tigated by considering dry and saturated conditions.
Although the horizontal actuator is servo-controlled,
the control was made using the horizontal LVDT adja-
cent to bed joint because a more stable response was
found. The disposition of the LVDTs for measuring
the horizontal and vertical displacements of the joint
is depicted in Figure 2.
 The relative horizontal displacement of the joint was
measured by the horizontal LVDTs placed at each side
of the specimen. Although the LVDTs were fixed to
the unit through the supports that were glued to it,
the influence of the shear deformation of the units
should be marginal in the measured final deformation
(Hansen 1999). The vertical displacements of the joint
were measured by the LVDTs placed at the opposite
corners of the specimen. The relative vertical dis-
placements were monitored in case of dry joints for
assessing the possible dilatant behavior of the joints.
Nevertheless, technical problems did not allow to mea-
sure relative vertical displacements in mortar joints, as
tests were conducted in a subsequent phase. Both shear
and normal stresses were measured and recorded by
the horizontal and vertical load cells of 22 kN capacity.
(a)
3 ANALYSIS OF RESULTS

3.1 Monotonic behavior of dry joints

The shear load-shear displacement diagrams for dis-
tinct pre-compression stress levels resulting from the
monotonic tests conducted on dry and saturated spec-
imens are displayed in Figure 3a and Figure 3b
respectively. The shear displacement is the result of
averaging the measurements recorded by both LVDTs
placed at each side of the specimen. The shear stress
is calculated by the following expression:

where H is the load in the horizontal actuator and A

is the cross area of the joint section. The normal stress (b)
is also calculated from the normal pre-compression
load, N , and is also based on the total area of the cross Figure 3. Shear stress-shear displacement diagrams in dry
section as: joints; (a) dry specimens; (b) saturated specimens.

plateau is found at peak stress, as the shear behav-

From the responses in Figure 3, no significant dif-
ferences were detected between dry and saturated
specimens and the peak shear strength was found to
increase as the normal stresses increase. By compar-
ing the peak shear stresses, lower values were obtained
in case of saturated specimens and higher scatter
was found when the maximum pre-compression level
(σ0 = 1.0 N/mm2 ) was applied.
Four stages can be identified in the shear stress-
shear displacement diagrams. The pre-peak behavior
is characterized by a linear extent for low levels of
shear stress, associated to the contact of the joint inter-
face, and by a clear non-linear stretch until peak shear
is reached. These features can be confirmed from
Figure 3, where the pre-peak part is highlighted. A
ior of dry masonry joints under monotonic loading
exhibits considerable plastic deformations associated
to the inelastic sliding.
Similarly to what has been reported in literature
(Misra 2002, Huang et al. 2002) no shear softening was
recorded after peak stress for smooth surfaces, unlike
rough rock joints that exhibit remarkable lowering of
the shear resistance as the plastic shear displacement
increases due to the roughness fracture.
3.2 Monotonic behavior of mortar joints
The shear stress-shear displacement diagrams of mor-
tar joints with respect to all levels of pre-compression
are shown in Figure 4a. The horizontal displacement
is considered as the average of the displacements

773
 (a)
(b)
Figure 4. Shear behavior of mortar joints; (a) shear
stress-shear displacement diagram; (b) pre-peak nonlinearity.
recorded by the two horizontal LVDTs located at each
side of the specimen and the shear stress is calcu-
lated according to eq. 2. The general shape of the
shear stress-shear displacement is characterized by a
sharp initial linear stretch. The peak load is rapidly
attained for very small shear displacements. Similarly
to what was reported for dry masonry joints, non-
linear deformations develop in the pre-peak regime,
see Figure 4b.
After peak load is attained there is a softening
branch corresponding to progressive reduction of the
cohesion, until reaching a constant dry-friction value.
This stabilization is followed by the development of
large plastic deformations.
As required, the shear tests were carried out with-
out significant fluctuations of vertical load (less than
774
Figure 5. Relation between total vertical displacement and
shear displacement.
2%). Note that the shear tests are conducted under hor-
izontal displacement control. This further assesses the
validation of the test setup.
In spite of the fact that the relative local vertical
displacements of the joint could not be measured due
to technical problems with the LVDTs, the total verti-
cal displacement was recorded by the internal LVDT
located inside the vertical actuator. The evolution of
this displacement with the shear displacement is dis-
played in Figure 5, where in the greater number of the
tests two distinct phases can usually be distinguished.
Firstly, the uplift of the joint is expressed by increasing
positive vertical displacements, which is particularly
remarkable for low normal stresses. The nonlinear evo-
lution of the displacements provides variable dilatancy
assuming decreasing values as the shear displacement
increases. This behavior is connected to the changes
on the interfaces due to surface wearing. It is observed
that the shear displacement associated to the maximum
value of the vertical displacement is close to the hori-
zontal displacement corresponding to the stabilization
of the shear stress.
The dilatant behavior reflects, to great extent, the
distinct shear failure modes obtained in the specimens
submitted to different normal stresses. In fact, for low
levels of pre-compression, shear failure occurs at the
unit-mortar interface along one unit face or, more fre-
quently, divided between two unit faces, see Figue 6.
For the larger normal stress level (σ = 1.25 N/mm2 ),
the failure is only localized in the mortar and a larger
amount of small mortar particles were found to be
detached. No damage was visible in the stone units
in all cases.
During the regime of pure friction the verti-
cal displacement remains constant or progressively
Figure 6. Typical Failure mode for the lower compression
level (σ = 0.5 N/mm2 ).

decreases, which is more significant as the level of pre-

compression increases, exhibiting even negative val-
ues in some specimens submitted to pre-compression
levels of σ = 0.75 N/mm2 and σ = 1.00 N/mm2 . This (a)
seems to indicate that as the pure shear develops,
the wearing of the surface joints leads to compaction
in the case of the porous lime mortar used. For the
larger pre-compression level (σ = 1.25 N/mm2 ), only
compaction of the specimen was recorded, which
is revealed by the negative values of the total ver-
tical displacements, resulting from the higher level
of degradation of the mortar joint associated to the
continuous friction.

3.2.1 Cyclic behavior of dry joints

The typical shear stress-shear displacement diagrams
obtained in direct cyclic shear tests conducted in
masonry joints of dry and saturated specimens are dis-
played in Figure 7 for the level of pre-compression
σ = 0.75 N/mm2 .
The shear behavior of dry joints during the first
cycle agrees with the monotonic diagrams exhibit-
ing nonlinearity in the pre-peak regime and post-peak
plastic deformations. Apart from the distinct values of
the peak shear stress, no significant differences in the
shape of the diagrams can be found in the range of the
tested normal stresses, which is valid for both dry and
saturated specimens.
Figure 8 shows the shear stress-shear displace-
ment diagrams for the first and last cycles of loading
corresponding to distinct levels of pre-compression.
Although minor differences are found in the shear
strength during the reversal cycles among the distinct
pre-compression levels, there seems to be a more sys-
tematic tendency for a slight increase between the first
and the last cycles as the vertical pre-compression
takes higher values. This effect is more clear in dry
specimens. Note that from the monotonic shear stress-
shear displacement diagrams, it is observed that for
a shear displacement of 0.1 mm, the maximum shear
forces were already reached.
This result is also in agreement with the findings
reported by Lee et al. (2001) concerning direct shear
tests conducted on smooth joints of granite, which
referred that the preferential degradation of quartz
(b)
Figure 7. Shear stress-shear displacement diagrams under
cyclic loading for σ = 0.75 N/mm2 ; (a) dry specimens;
(b) saturated specimens.
grains against other rock-forming minerals could pro-
vide the stick-slip on the surface of granite. When an
analysis of the normal displacement-shear displace-
ment diagrams is carried out, it is possible to observe
a more visible tendency for compaction associated to
the wearing of the joint surface.
From the normal-shear displacement diagrams,
it is possible to conclude that the values of dila-
tion/compaction are not greater than ±0.06 mm, which
shows good agreement with the values reported by
Lourenço & Ramos (2004) for dilation obtained on
sandstone smooth dry joints and with the results
pointed out by Homand et al. (2001) for hammered
granitic joints. In both cases, the degradation mecha-
nism is dominant over the phenomenon of dilatancy.
Therefore, it is reasonable to assume that this type
of smooth rock joints is non-dilatant. This result is
also enlarged to the shear behavior of saturated speci-
mens. In fact, it is well known that the dilatant behavior

775
 (a)
(b)
Figure 8. Characteristic aspects of the shear behavior of
dry joints; (a) evolution of the shear stress-shear displace-
ment diagrams between the first and the last cycle of reversal
loading; (b) compaction due to the wearing of the joint
surface.
of rock joints is highly dependent on both roughness
of the joint surface and the level of vertical pre-
compression (Amadei et al. 1998; Huang et al. 2002,
Misra 2002). As referred by these authors, in rough
joints lower dilation is obtained at high normal stresses
and for increasing shear displacements dilatancy tends
to exhibit decreasing values.
Based on the shear stress-shear displacement dia-
grams, it is observed that the shear behavior of dry
joints is characterized by an approximately constant
stiffness followed by marked nonlinearity close to the
peak load in the loading branches. On the other hand,
the stiffness of the unloading branches exhibits always
considerable high values when compared with the stiff-
ness obtained in the loading and reloading cycles.
The corrected displacement of the dry joint can be
776
Figure 9. Correction of the measured shear displacement-
dry specimens.
obtained by removing the elastic deformation of the
unit reading:
where umeas is the shear displacement given by the
horizontal LVDTs, τ is the shear stress for a given
displacement and ku is the stiffness calculated in the
unloading branches. It is possible to confirm that the
elastic deformation of the units has a minor role in
the total shear displacement of dry joint, see Figure 9.
The shear behavior of dry joints is thus charac-
terized by significant non-linear deformations in the
pre-peak stage and perfect plastic deformations after
peak stress resulting from the characteristic sliding
failure mode. The former characteristic of the shear
behavior of dry joints was already pointed out by
Lourenço & Ramos (2004). Apart from the nonlinear-
ity in the pre-peak regime, the envelop of the diagrams
is also in good agreement with the shape of the shear
stress-shear displacement diagrams indicated by Lee
et al. (2001), also for smooth sawn-cut granitic joints.
3.3 Shear strength-normal stress diagrams
Figure 10 shows the relationships between the values
of the shear strength obtained in the monotonic tests
and in the first cycle of the cyclic tests for dry and
saturated conditions as a function of the normal stress.
For both specimens, an expressive linear correlation
was attained between normal and shear stress, which
confirms the initial assumption that the shear strength
is well described by Coulomb’s friction law.
The slight decrease on the shear strength obtained
on saturated specimens is here reflected by the lower
value of the friction coefficient, being 0.65 and 0.60
the values that were achieved for dry and saturated
 (a)
(b)
Figure 10. Characteristic aspects of the shear behavior of
dry joints; (a) evolution of the shear stress-shear displace-
ment diagrams between the first and the last cycle of reversal
loading; (b) compaction due to the wearing of the joint
surface.
specimens, respectively, corresponding to a lowering
of the friction coefficient of approximately 10%. The
influence of the moisture content on the shear strength
of rock joints was also investigated by Geertsema
(2002) that pointed out a decrease on the friction coef-
ficient on saturated specimens ranging between 10◦
and 22.4◦ for mudstone. The results are obviously
not comparable because the materials are considerable
different.
The friction angle corresponding to dry joints
is slightly larger than the value pointed out by
Lourenço & Ramos (2004) for specimens composed
by sandstone sawn units tested under dry condi-
tions (µ = 0.63) and somewhat lower than the value
777
Figure 11. Relation between peak and residual shear
stresses with normal stress.
indicated by Lee et al. (2001) for sawn-cut granitic
joints (µ = 0.69). The narrow range of values for the
friction angle seems to indicate that no significant dif-
ferences for this property should be expected among
distinct types of natural stone under similar roughness
surface conditions. For the range of vertical stresses
considered, the shear strength should be more sensitive
to the roughness characteristics of the bed joint surface
than to the material properties or even mineralogical
composition.
By comparing the values of the initial and final fric-
tional resistance, it can be seen that there is a small
increase of the frictional resistance in the last cycle,
being the difference more expressive in case of dry
specimens. This result confirms the tendency for the
slight slipping of the shear stress-shear displacement
diagrams previously referred, which can be the result
of the wearing of the granitic surface. It is believed
that this effect is highlighted due to the considerable
porosity of this type of granite.
It should be stressed that in spite of the use of sawn-
cut units to characterize the mechanical shear behavior
of dry joints of masonry walls discussed later, the
experimental investigation on the shear behavior of
random rough dry joints, including the definition of
the characteristic diagrams, failure criteria and dila-
tant behavior, would characterize more realistically the
mortarless masonry joints existing in ancient masonry
construction.
The relation between the peak and residual shear
stress with the normal stress is displayed in Figure 11.
Significant correlation coefficients were obtained by
fitting linear functions to the experimental data com-
posed by peak and residual shear strength, with coef-
ficients of correlation of r 2 = 0.88, and r 2 = 0.80,
respectively. This means that for the range of pre- Table 1. Shear strength properties for different unit-mortar
compression levels tested, the peak and residual shear assemblages.
strength of the bed joint can be reasonably described
by means of the Coulomb’s friction law given by Source units mortar c (MPa) µ
eq. 1. Therefore, the linear approach provides shear
Atkinson Old clay 1:2:9 (13) 0.127 0.695
strength characteristics of the mortar joint, cohesion,
(1989) Old clay 1:2:9 (7) 0.213 0.640
c, and friction coefficient, µ.A value of cohesion about New clay 1:1.5:4.5 0.811 0.745
0.36 N/mm2 and the tangent of the friction angle, tanφ,
Amadio and Solid Cement 0.65 0.723
equal to 0.63, corresponding to a friction angle of
Rajgeli bricks Lime-
32.2◦ , were attained for the peak strength. The residual (1990) cement
shear strength can be calculated with reasonable accu-
Magenes Solid Hydraulic 0.206 0.813
racy from a friction coefficient of 0.78. This value can
(1992) bricks lime 0.081 0.652
be used for evaluation of the shear sliding resistance Lime mortar
of walls or piers submitted to seismic action failing
Binda et al. Sandstone Hydraulic 0.33 0.74
along horizontal sliding joints. The strength values,
(1994) Calcareous lime 0.58 0.58
particularly the bond strength, are greatly dependent Hydraulic
on the moisture content and porosity of the units and lime
on the strength and composition of mortar as well as on
Roberti et al. Bricks Hydraulic 0.23 0.57
the nature of the interface (Amadio & Rajgelj 1990). (1997) lime mortar
Binda et al. (1994) pointed out that when strong mortar
Lourenço Hollow Micro- 1.39 1.03
is considered, the strength of the units can also regu-
et al. (2004) bricks concrete
late the shear behavior of the joints. This yields that a
wide range of shear strength values have been pointed This study Granite Lime mortar 0.359 0.630
out for various combinations of units and mortar.
Mann & Muller (1982) indicated a mean friction
coefficient of approximately 0.65 on brick-mortar
assemblages and a cohesion ranging from 0.15 up to
0.25, depending on the mortar grade. From the results nearly 5% was recorded on the friction coefficient
of direct shear tests carried out by Pluijm (1999), the changing the moisture condition from dry to saturated.
coefficient of internal friction ranges between 0.61 Low differences were found between the peak and
and 1.17, whereas cohesion varies from 0.28 up to residual friction angle for dry masonry joints under
4.76, depending on different types of units and mortar. distinct moisture conditions. No dilatancy was found
Table 1 summarizes other results published in literature to characterize the shear behavior of dry masonry
referring to the shear strength properties for different joints.
combinations of materials. Similarly, an extended plastic branch character-
izes the residual post-peak shear behavior of mortar
masonry joints. After a reduction of approximately
25% of the peak strength, the shear stress stabilizes
4 CONCLUSIONS with considerable level of plastic deformations. The
measured total vertical displacement revealed a ten-
The experimental characterization of masonry assem- dency for the mortar joint dilation to decrease as the
blages and masonry components used in the shear normal stress increases. Only for a pre-compression of
walls discussed in Chapter 6 is addressed, focusing 0.5 N/mm2 a systematic expansive trend was verified
on the strength properties of dry masonry joints and up to peak stress. For this case, it was found that dila-
the unit-mortar interface, as well as on the compressive tancy decreases significantly as the shear displacement
properties of the masonry. increases.
A set of direct shear tests was conducted on cou-
plet specimens, either considering dry or mortar joints.
From these tests it was possible to derive the shear REFERENCES
strength properties, namely cohesion and friction coef-
ficient. Besides, the complete shear stress-shear dis- Amadio, C., Rajgelj, S. 1991. Shear behavior of brick-mortar
placement diagrams enabled a better insight into the joints, Masonry International, 5 (1), 19–22.
Atkinson, R.H., Amadei, B.P., Saeb, S., Sture, S. 1989.
shear behavior of these assemblages. An elastic per- Response of masonry bed joints in direct shear, Journal of
fectly plastic diagram was found to characterize the Structural Engineering, 115 (9), 2277–2296.
monotonic and the cyclic envelope of shear tests Binda, L., Fontana, A., Mirabella, G. 1994. Mechanical
conducted in dry masonry joints. No significant dif- behavior and stress distribution in multiple-leaf stone
ferences in the frictional behavior of dry joints under walls, 10th International Brick Block Masonry Confer-
distinct moisture contents were found. A reduction of ence, Calgary, Canada, 51–59.

778
Calvi, G.M., Kingsley, G.R., Magenes, G. – Testing masonry
structures for seismic assessment, Earthquake Spectra,
Journal of Earthquake Engineering Research Institute,
12(1), 145–162, 1996.
EN 1052-3, Methods of test for masonry: Part 3 – Determi-
nation of initial shear strength, 2002.
Geerstsema, A.J. 2002. The shear strength of planar joints in
mudstone, International Journal of Rock Mechanics and
Mining Sciences, 39, 1045–1049.
Hamid, A.A., Drysdale, R.G. 1980. Behavior of brick
masonry under combined shear and compression loading,
Proc. 2nd Canadian Masonry Conference, 314–320.
Hansen, K.F. 1999. Bending and shear tests with masonry,
SBI Bulletin 123, Danish Building Research Insti-
tute, p. 36.
Hansen, K.F., Nykänen, E., Gottfredsen, F.R. 1998. Shear
behavior of bed joints at different levels of precompres-
sion, Masonry International, 12 (2), 70–78.
Homand, F., Belem, T., Souley, M. 2001. Friction and degra-
dation of rock joint surfaces under shear loads, Interna-
tional Journal for Numerical and Analytical Methods in
Geomechanics, 25, 973–999.
Huang, T.H., Chang, C.S., Chao, C.Y. 2002. Experimental
and mathematical modeling for fracture of rock joint with
regular asperities, Engineering Fracture Mechanics, 69,
1977–1996.
Lee, H.S., Park, Y.J., Cho, T.F., You, K.H. 2001. Influence
of asperity degradation on the mechanical behavior of
rough rock joints under cyclic shear loading, International
Journal of Rock Mechanics and Mining Sciences, 38,
967–980.
779
Lourenço, P.B., Barros, J.O., Oliveira, J.T. 2004. Shear test-
ing of stack bonded masonry, Construction and Building
Materials, 18, 125–132.
Lourenço, P.B., Ramos, L.F. 2004. Characterization of cyclic
behavior of dry masonry joints, Journal of Structural
Engineering, 130 (5), 779–786.
Mann, W., Müller, H. 1982. Failure shear-stressed masonry
– an enlarged theory, tests and application to shear walls,
Proc. British Ceramic Society, 30, 223–235.
Misra, A. 2002. Effect of the asperity damage on shear behav-
ior of single fracture, Engineering Fracture Mechanics, 69,
1997–2014.
Pluijm, R.V.D.1999. Out-of-Plane bending of masonry,
behavior and strength, PhD thesis, Eindhoven University
of Technology. ISBN 90-6814-099-X.
Riddington, J.R., Ghazali, M.Z. 1990. Hypothesis for shear
failure in masonry joints, Proc. Instn. Civ. Engrs, 89,
89–102.
Samarasinghe, W., Hendry, A.W. 1980. The tensile of brick-
work under biaxial tensile and compressive stress, Proc.
7th International Symposium on Load Bearing Brickwork,
London, 129–139.
Vasconcelos, G. 2005. Experimental investigations on the
mechanics of stone masonry: characterization of granites
and behavior of stone masonry shear walls, PhD thesis,
University of Minho, Portugal.
Vasconcelos, G., Lourenço, P.B., Alves, C.A.S, Pamplona, J.
2007. Prediction of the mechanical properties of gran-
ites by ultrasonic pulse velocity and Schmidt hammer
hardness, 10th North American Masonry Conference, (in
press).
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Large scale testing of drystone retaining walls

Chris Mundell, Paul McCombie & Andrew Heath

BRE Centre for Innovative Construction Materials Dept. of Architecture & Civil Engineering,
University of Bath, UK

John Harkness
School of Civil Engineering & the Environment, University of Southampton, UK

Pete Walker
BRE Centre for Innovative Construction Materials Dept. of Architecture & Civil Engineering,
University of Bath, UK

ABSTRACT: There are numerous historic drystone retaining walls in the UK, but the analysis of these eco-
nomically important walls is restricted by the lack of experimental data. As part of an ongoing investigation into
the stability of drystone retaining structures, two full-scale walls have been constructed and tested to collapse.
Details of each test are described, including set-up, wall construction, instrumentation and data collection. Initial
findings are presented and analysed, along with the continuing aims and goals of the project.

1 INTRODUCTION understood, the mechanisms behind observed failures

Some of the oldest built structures found around the
world were constructed using drystone technology,
ranging from simple field walls to large earth retaining
structures. In the UK alone over 9000 km of drystone
retaining walls line the road and rail networks [1], most
constructed during the 19th and 20th centuries and
many still remaining perfectly stable today. Despite
their widespread use, these structures present a chal-
lenge for maintenance engineers assessing their sta-
bility. Substantially deformed walls, typically bulged,
may remain stable for generations whilst other walls
displaying only minimal movements can suddenly and
inexplicably collapse.
Drystone walling is an empirical form of con-
struction, using best practice methods formed over
generations, building on existing examples and past
experience. Historically, masons have learnt their trade
through apprenticeships. Although presently there are
several UK bodies, including The Drystone Walling
Association and British Trust for Conservation Volun-
teers, which promote professional standards of work,
there are presently no minimum standards for a prac-
ticing waller to adhere to. The task of unifying good
practice is made more difficult by regional differences
that have developed in a response to variations in local
stone and ground conditions.
Whilst the general principles for building struc-
turally safe drystone retaining walls are fairly well
of these structures are not. There are several reasons
for this, with one of the critical factors being the deficit
of scientific tests on full size retaining walls. In fact,
a series of full scale tests carried out over 170 years
ago remains almost the only data available regarding
the failure of ‘real life’ walls to date [3].
The bespoke nature of such walls must also be con-
sidered when attempting to understand the causes of a
wall failure. In addition to occurrences such as wall or
backfill settlement, pore water build up and increas-
ing loading conditions, drystone walls are equally
affected by build quality and age. For any given wall,
any combination of these factors may affect stability
to some degree, again defying any simple means of
standardisation.
2 AIM OF STUDY
The overall aim of the research project is to pro-
vide in-depth data on failure mechanisms, including
bulging, and three-dimensional effects that influence
this behaviour. Through the use of a bespoke test rig,
two full size drystone retaining walls have been tested
to failure and a further two are to be subjected to
localised surcharging, backfill settlement or wall sub-
sidence. Actions are to be applied simultaneously or
independently, in a controlled manner whilst a range of

781

Figure 1. Burgoyne’s test wall geometry[4].
instrumentation is used to monitor the walls’ response
through to their collapse.
The main aim is to have sufficient data to verify
existing models and theories, and assist with the gener-
ation of new analysis techniques, more accurate mod-
elling tools and assessment guidelines for drystone
structures. The study is funded by the Engineering &
Physical Sciences Research Council (EPSRC) and
is conducted in collaboration with the University
of Southampton. The creation of these guidelines,
together with appropriate engineering tools to back
them up, could save a substantial part of the estimated
£1 billion estimated cost of replacing the walls in the
UK alone [2].
3 RELATED RESEARCH
As previously mentioned, physical test data regard-
ing drystone retaining wall structures is sparse. The
largest reported tests to date were conducted in
1834 by Lieutenant-General Sir John Burgoyne [3],
who constructed four full size test walls in Dun
Laoghaire, Ireland. Each wall was built using the
same overall volume of square cut granite blocks,
but arranged in different sections (Fig. 1). Testing
consisted of backfilling each wall until the full reten-
tion height (6 metres) was achieved, unless premature
collapse occurred.
From this work Burgoyne proved that wall geometry
has a substantial impact on overall stability, although
the use of highly worked granite blocks perhaps
caused the walls to act more monolithically than would
otherwise be observed in more traditional drystone
structures. Regardless, his findings and observations
remain the basis for the validation of almost all of
the numerical studies carried out to date on drystone
retaining walls, despite consisting solely of dimen-
sional measurements and visual observations reported
19 years after the tests.
Work of such a physical nature was not conducted
again until 2005, when a French engineering team
782
led by Jean-Claude Morel and Boris Villemus built
and tested five large scale test walls at ENTPE,
Lyon.[5]. The walls were of various sizes ranging from
2 m–4.25 m high, up to 1.8 m thick and between 2 m
and 3 m long. Each wall was subjected to hydrostatic
forces via a PVC-lined water filled bag, in order to load
the wall using purely horizontal pressures that could
be precisely known at all times.
The main aim of Morel and Villemus’ study was to
identify the internal failure angle within the walls at
yield. As each test wall was only a short section (2–3 m
long), the wall ends were left exposed and were consid-
ered to reveal the internal actions throughout the entire
length. The tests could not proceed to collapse because
of the way pressure was applied to them, but monitor-
ing of the end faces allowed relative movements within
the wall to be measured as yielding took place. The
angles of the resulting shearing surfaces to the hori-
zontal were related to the pattern of construction of the
walls.
4 TEST SETUP
4.1 Testing rig
For the purposes of this project, a bespoke test rig
was required. The rig required the ability to impose
various conditions upon the test walls, necessary to
undermine their stability. These conditions, including
relative wall settlement, backfill compaction and sur-
charge loading, mimic the real-life situations to which
many drystone walls are subjected.
To simulate settlement of the foundations or the
backfill, the ability to raise, lower or tilt the wall
is required. To achieve these movements, the wall
itself is constructed on a 1.4 m × 4 m steel platform,
which is in turn mounted on four 20 tonne mechani-
cal screwjacks. This arrangement can produce vertical
movements of up to 400 mm, tilt up to 17◦ forwards
or backwards, as well as intermediate combinations of
these movements.
A steel frame spans the central portion of the wall
and backfill, under which a hydraulic jack is mounted.
This jack allows a patch surcharge to be applied to
almost any location behind the wall, simulating addi-
tional localised loadings of up to 200 kN (Fig. 2). The
walls were designed to be sufficiently long to minimise
the edge effects from both induced settlements and sur-
charge loading on the behaviour of the wall which was
a serious limitation of the Burgoyne experiments.
4.2 Instrumentation
To ensure as much data as possible are gathered from
each test, a combination of techniques were imple-
mented. Load cells monitor the forces acting on the
platform, and the applied patch surcharge. Simple

Figure 2. Test rig setup.
button-type compression cells are used for the screw-
jacks, with S-type tension/compression cells located
on the radial arms. Data-logging records readings at
appropriate intervals throughout the different stages
of construction and testing.
It is necessary to monitor both local movements
of the gravel and the distribution of earth pressures
within the backfill. Monitoring of movement is dif-
ficult to accomplish during testing, as there are few
non-destructive/non-invasive means to achieve accu-
rate results. For the purposes of these tests, several
layers of steel ball bearings have been laid in grids
within the backfill, with each ball carefully placed and
surveyed into position using a reflectorless Total Sta-
tion. The polished surfaces of the balls ensure that the
measuring beam is reflected back from a point on the
surface in line between the instrument and the centre
of the ball, so that a correction for the radius is all that
is needed to determine the accurate positions of the
centres. After failure of the wall, the ball bearings are
carefully recovered using metal detection equipment,
and again their positions recorded, thus revealing any
movements and aiding identification of failure planes
within the backfill.
Backfill pressures are measured using pressure
cells buried within the fill. These consist of parallel
100 mm × 100 mm steel plates separated by 2500 N
or 5000 N load cells. These are placed within the
fill either vertically or horizontally, allowing continu-
ous monitoring of lateral or vertical pressures. These
sensors are particularly useful to indicate how the sur-
charging force is distributed through the backfill. The
pressure cells are spread out through the backfill, with
the majority concentrated along the centre-line of the
wall where the loads are expected to be highest.
To monitor the wall face itself, a combination of
techniques have been used. Around 350 small targets
have been scored into the wall face in five vertical
lines. The points are relocated throughout testing using
a reflectorless Total Station. Although a slow process,
with each round of readings taking 15–20 minutes,
783
the nature of the tests allows sufficient time for the
Total Station to be used at any point up to the final
moments of collapse. Although the final failure of
the wall cannot be captured using this technique, it
is still very useful for examining bulge formation and
the movements prior to failure.
In addition to the Total Station readings, draw wire
transducers were attached to the centre line of the wall.
These allow continuous high resolution monitoring of
wall face movements during testing, facilitating con-
trol of test procedures. The transducers are anchored
to a stanchion attached to the loading platform, so
that measurements represent deformation of the wall
directly, without the need to correct for movement of
the platform.
Given the scale of variations within the construc-
tion of drystone walls, it is not possible to guarantee
that the transducers and load cells will be located on
parts of the structure where key events unfold. To
ensure detailed recording of mechanisms, this project
employs a range of visual tools to supplement the
instruments. In front of the test wall face, on a fixed
pedestal, are a pair of Nikon D40x digital SLRs, pro-
viding stereo photographs of the wall face throughout
testing. An additional camera is mounted behind the
backfill to record back face and backfill conditions in
response to the surcharging.
Southampton University used Particle Image
Velocimetry (PIV) to record movements of the cen-
treline of the face to sub-millimetre accuracy. High
resolution digital photographs were taken sighting
along the line of the wall to targets mounted on
brackets attached to a number of individual stones.
A computer program was then used to determine
changes from one image to the next.
The whole test process has also been captured on a
Canon HV20 HD camcorder. This final piece of equip-
ment ensures that the final moments of the test are
recorded in a high-definition video format. Addition-
ally, it allows any sudden movements or noises to be
recorded, information that might otherwise be missed.
5 EXPERIMENTAL TESTING
5.1 1st test wall
The first test wall was constructed in June 2007, requir-
ing approximately 35 tonnes of limestone. The wall
was 2.5 m high, tapering from 600 mm at the base
to 400 mm at the coping, with a vertical rear face
and a battered front face. This initial wall was con-
structed with two fair faces, so that the wall could be
constructed without the need to be building and back-
filling simultaneously. The wall contained three layers
of through-stones, at 0.5 m, 1 m, and 1.5 m high. As
is standard practice amongst most UK masons, these

Figure 3. Test wall 1 (completed).
through-stones protruded clearly from the face of the
wall, fig. 3.
The aggregate backfill was placed in layers of
250 mm–300 mm thick, compacted using a 1 kN
vibrating wacker plate. Through laboratory shear box
and triaxial testing, also confirmed by an in-situ plate
loading test, the angle of friction for this compacted
material was found to be 51◦ . This is very high when
compared to standard fills generally found behind
drystone walls but this aggregate type was necessary
to ensure adequate drainage. The required height of
2.2 m (the height to the bottom of the coping stones)
was achieved with no deformations greater than 2 mm
recorded during construction.
This wall was tested over the course of three days in
July 2007, with a day between each test day to allow
any relaxation or further settlement to occur. On the
first day the platform was raised up by 20 mm, to
ensure full mobilisation of friction against the back
of the wall, as would occur during settlement under
self weight of a normal fill.
On the second test day, the surcharging rig was used
to load the backfill, using a hydraulic jack situated
0.5 m from the back face of the wall along the centre-
line. The patch load was applied over a stiffened plate
400 mm × 600 mm. A force of 5 tonnes was reached
before loading ceased. It was found that due to the high
stiffness of the aggregate, this surcharge force was too
close to the back face of the wall, resulting in defor-
mations mainly towards the top of the wall. The jack
was then reset to 1 m from the back face, and a bigger,
600 × 400 mm, surcharging plate installed.
To model realistic wall movements, the wall was
rotated forwards, imitating ground settlement beneath
784
Figure 4. Test wall 1 (prior to failure).
the toe of the wall. Rotation of 2.5◦ was induced, caus-
ing the front face of the wall to appear almost vertical.
At this point, several readings were taken using the
Total Station and further experimentation was delayed.
The final day of testing involved surcharge loading,
1 m behind the initial wall position.The plate was grad-
ually loaded up to 11 tonnes, at which point the load
levelled off as the failure wedge within the backfill
began to shear. Substantial movements were recorded,
with the wall coping overhanging the toe by 500 mm
prior to failure (fig. 4).
As wall movements became more pronounced, the
load that could be sustained from the surcharging
frame steadily dropped, until the final failure mecha-
nism was produced. The wall failed through toppling,
having undergone a significant three-dimensional dis-
tortion.
5.2 2nd test wall
The second wall was constructed in September 2007,
re-using the same material from the first wall. In an
attempt to create more pronounced deformations, the
wall was reduced in section to 500 mm at the base,
tapering to 300 mm at the coping level.
This wall was of a much looser construction, espe-
cially on the rear face of the wall. Due to the com-
paratively unstable nature of this wall, backfill was
introduced as the wall was built, ensuring stability
during construction.
In an attempt to model real conditions more pre-
cisely, the backfill was not compacted for this second
test. Plate loading tests indicated the angle of friction
to be about 41◦ , which is much closer to the mate-
rial properties found behind most existing retaining
walls. In a similar manner to the first test, the second

Figure 5. Test wall 2 (prior to failure).
wall retained 2.2 m of backfill, with the top 300 mm
of coping left uncovered.
Testing was conducted in October 2007, over three
consecutive days. Test procedures similar to those of
the first tests were used, in order that a basis for
comparison of the data generated would exist.
To this end, the platform was initially raised 40 mm
in 5 mm increments. Loads on the platform were care-
fully monitored until a stable plateau was reached,
confirming full mobilisation of backfill friction on the
back of the wall.
The wall was then slowly surcharged over the course
of the next three days. The same plate was employed
in the same location (600 mm × 400 mm at a distance
of 1 m from the back face of the wall). Peak loads of
7.5 tonnes were reached, accompanied by wall defor-
mation, block sliding and clearly audible movements
within the wall. In response to backfill mobilisation
and initial surcharging a distinct bulge developed in
the lower section of the wall prior to collapse.
Failure finally occurred after the peak load had
dropped from 7.5 tonnes to 4 tonnes. The final fail-
ure mode was again overturning, however more pro-
nounced sliding and rotation was noted in the lower
courses. Overall deformations prior to collapse were
much reduced from the first test wall; however the area
of wall that failed was significantly larger.
785
5.3 Materials testing
Following tests an intact area of wall was measured,
weighed and removed to ascertain the void percentages
and densities. A void percentage of 28% was mea-
sured, giving an average wall density of 17.9 kN/m3 .
Two further test sections, each of dimensions
0.5 m × 0.5 m × 0.7 m, were constructed by the
masons to investigate this further. One was built to the
highest standard that could be achieved with this stone,
but using fill pieces of a normal size. The second was
built to a functional standard, but to represent a struc-
ture built quickly without taking care to achieve an
optimal fit between stones. Upon disassembly, the void
percentages for these two volumes were 21% and 37%,
with overall densities of 19.7 kNm−3 and 15.6 kNm−3
respectively.
Other material tests are scheduled to be carried out
within the time frame of this project, including labora-
tory investigations into the interface between the wall
and the backfill, and sliding between wall beds.
6 ANALYSIS
As the backfill was raised to the full height, and
the aggregate fully mobilised against the back of
the wall, bulging occurred; indeed the second wall
began bulging even before the platform was raised
to mobilise wall friction. Bulging consisted of move-
ments less than 5 mm, undetectable to the naked eye
on a structure of this scale, but easily detected by the
instrumentation.
These movements hint at the mechanisms within
the wall at these locations, and are probably due to the
wall adjusting to support the initial stresses through
internal block sliding and small rotations.
An obvious difference between walls 1 and 2 was the
amount of deformation prior to failure. The first test
wall moved almost 350 mm further at the coping (see
fig. 6).Although the thickness of the wall is understood
to be one of the critical reasons for this difference,
several other unforeseen reasons present themselves.
After failure of the first wall, the three-dimensional
nature of the test became apparent. Due to friction
between courses, a certain amount of tension had been
generated along the front face of the wall, allowing the
wing walls to help retain the failing central section.
Conversely, the second wall was of a much looser
construction, and during testing several vertical joints
appeared throughout the wall. These joints limited any
of the stabilising effects from the wing walls, and
hence smaller deformations were visible before the
failure conditions were reached.
The compaction of the backfill was also a critical
factor in determining the failure mode. Certainly the
high stiffness of the aggregate in the first test caused
the surcharge load to be distributed over a wider area
 2.25 4

3.5
2

3
1.75

Depth (m)
2.5

1.5

1.25
Height (m)

1.5

1
1
2 2.5 3 3.5 4 4.5 5 5.5 6
Distance (m)
1st Level Initial 1st Level Final 2nd Level Initial 2nd Level Final
0.75 3rd Level Initial 3rd Level Final 4th Level Initial 4th Level Final
5th Level Initial 5th Level Final Possible Failure Plane

0.5 Figure 7. Ball bearing movements.

0.25
0
1 1.1 1.2 1.3 1.4 1.5
Displacement (m)
1st Wall Initial 1st Wall Final
2nd Wall Final Initial Position
Figure 6. Test walls 1 & 2 total station data (horizontal scale
exaggerated).
than in the second test, leading to much higher loads
before slip planes were formed. This also had the effect
of transmitting lateral stress from the surcharge much
higher up the wall than originally desired, encouraging
toppling rather than bulging.
After dismantling the first wall, the ball bearings
within the fill were unearthed, located with the Total
Station, and then analysed for changes in position
(fig. 7). The stiffness of the fill allowed very lit-
tle movement away from the wall face; indeed the
only movements which were recorded coincided with
a failure wedge originating at the rear edge of the
surcharging plate.
The other significant impact that the angle of back-
fill friction has upon the failure is the magnitude
of active pressure exerted on the wall. The first test
with the higher friction angle has the added benefit
of a much greater vertical component upon the wall,
whereas the lower friction angle has a relatively greater
horizontal component.
It is difficult to be certain exactly how much of the
backfill friction is mobilised against the wall. How-
ever by initially raising the platform until the platform
loads peak, and given the roughness of the back of the
walls, it is reasonable to assume that it is almost fully
activated.
Both walls eventually failed via toppling. This was
to be expected for the first wall, as the combined effects
of the platform rotation and early loading near the top
of the wall succeeded in pushing the centre of gravity
1.6 1.7 1.8
very far forwards.
Although the second wall also toppled, the manner
2nd Wall Inital in which it initiated this failure was different. Block
sliding through the lower courses occurred to the point
where a significant portion of some blocks were over-
hanging the course below. This in turn allowed an
overhanging block to begin to rotate. In the final phases
of test two, specific blocks near the base of the wall
could be identified in both the photography and the
video which had clearly slid forwards over the blocks
beneath them.
As the wall is pushed further forwards, the resultant
thrust line will similarly move forwards until it is pass-
ing through the ends of these overhanging blocks. The
overturning of the wall was then triggered by the rota-
tion of one of these overhanging blocks. At this point
the thrust line will fall outside of the wall at this level
and toppling would occur. The movement of this rota-
tion along the course of stones could be seen clearly in
the video and in the sequence of stereo photographs.
7 CONCLUSIONS
Although only two of the four programmed tests have
as yet been carried out, substantial results have already
been generated. Each wall has been constructed to
emulate real walls as closely as possible, with care-
fully limited changes between successive wall tests to
facilitate meaningful comparison of the results.
An initial aim was to create the distinctive ‘belly
bulges’ found in many walls today, and both test walls
showed some bulging deformation. The reasons why
these particular failure mechanisms formed are now
better understood, highlighting issues that were not
previously considered.

786
 Though these first two wall tests do not provide data
applicable to a wide range of wall failures, the mech-
anisms observed most probably occur, and every dry
stone retaining wall that is still standing must necessar-
ily be resisting failure by these mechanisms. The tests
also provide detailed observations of the behaviour of
real walls which can be used to verify the results of
numerical modelling, including that currently being
carried out at Southampton University. This numeri-
cal modelling, once verified, will be used to investigate
other aspects that cannot be safely or effectively inves-
tigated by full scale testing, such as the effect of major
changes in pore water pressures. The tests therefore
represent a clear step towards a greater understand-
ing of these structures, and also towards the eventual
goal of formulating guidelines and accurate analysis
techniques. At the very least these tests prove beyond
all doubt that even walls such as these, designed to
deform and fail, are incredibly resilient and able to
adapt to significant loads with little affect on overall
long term stability.
ACKNOWLEDGEMENTS
The authors acknowledge support of the follow-
ing organisations and individuals, without whose
787
support this project would not have been possible:
Engineering & Physical Sciences Research Council;
Southampton University; Network Rail; Wiltshire
County Council; Gloucestershire County Council;
County Surveyors Society; Bradford Metropolitan
District Council; Cornwall County Council; Richard
Tufnell Dry-Stone Wallers; and, colleagues at the
University of Bath.
REFERENCES
1. Powrie, W., Harkness, R.M., Zhang, X. and Bush D.I.
(2002). “Deformation and failure modes of drystone
retaining walls”, Geotechnique, Vol. LII, No. 6, August,
435–446.
2. O’Reilly, M.P., Brady, K.C. and D.I. Bush, Research on
masonry-faced retaining walls, in 2nd European road
research conference. 1999: Brussels.
3. Burgoyne, J., Revetments or retaining walls. Corps of
royal engineers, 1853. 3: pp. 154–159.
4. Harkness, R.M., Powrie, W., Zhang, X., Brady, K.C. and
O’Reilly, M.P. (2000). “Numerical modeling of full-scale
test on drystone masonry retaining walls”, Geotechnique,
Vol. 50, 165–179.
5. Villemus, B., Etude des murs de soutenement en macon-
nerie de pierres seches, in L’institut national des sciences
appliquees de Lyon. 2004, ENTPE: Lyon.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Seismic isolation, strengthening of walls with CFRP strips and heritage

masonry buildings

M. Tomaževič, I. Klemenc & P. Weiss

Slovenian National Building and Civil Engineering Institute, Ljubljana, Slovenia

ABSTRACT: The efficiency of improving the seismic resistance of heritage masonry buildings by means of
seismic isolation and strengthening of structure with CFRP laminate strips has been investigated. Five models of
a simple two-storey brick masonry building with wooden floors without wall ties have been tested on the shaking
table. Besides control model, two models, isolated by either damp proof course or seismic isolators, have been
tested. Models four and five have been strengthened with CFRP laminate strips, simulating the wall ties placed
horizontally and vertically at floor levels and corners of the building, respectively. One of the CFRP strengthened
models has been placed on seismic isolators. Tests have shown that the damp proof course, unless adequately
designed, cannot be considered as seismic isolation. The isolators have also not improved the behavior in the
case of the building without wall ties. However, both models confined with CFRP strips exhibited significantly
improved seismic behavior. They did not collapse even when the accelerations of the shaking table exceeded the
accelerations, measured when tested the control models without wall ties, by three times.

1 INTRODUCTION 2 RESEARCH PROGRAM AND DESCRIPTION

OF TESTS
Besides traditional technologies, such as the tying
of walls with steel ties, strengthening the walls by 2.1 Research program
injecting cementitious grouts and applying reinforced
Typically, the walls of heritage masonry buildings are
cement coating, which have been used in the last
not tied with wall ties, what results into the separation
decades, new materials and technologies have been
of walls during earthquakes and subsequent out-of-
developed and proposed also for upgrading the seis-
plane collapse. Consequently, when retrofitting the
mic resistance of heritage buildings. Among them,
building for seismic loads, the tying of walls with steel
seismic isolation in different forms (e.g. Bailey and
ties represents the basic measure to ensure the struc-
Allen, 1988; Page, 1995; Sarrazin et al., 1996; Zhou
tural integrity and utilize the resistance capacity of
and Miao, 1996,) and strengthening the walls with
structural walls. In this regard, one of the objectives
fiber reinforced polymer laminates have been already
of the study was to investigate the possibility of omit-
studied (e.g. Schwegler, 1995;Triantafillou and Fardis,
ting the installation of wall ties by placing the structure
1997; Hamilton and Dolan, 1998; Gayevoy and Lissel,
on seismic isolators. Besides, the possibility of using a
2004; Trantafillolu, 2001). Although the requirements
simple PVC damp proof course as isolating device, has
of preservation of cultural heritage limit the applica-
also been studied. Last but not least, the idea to replace
tion of such materials and technologies, the methods
the usual steel ties with CFRP laminate strips, placed
are convenient because they require minimum inter-
both horizontally at the level of floors and vertically
vention in the existing structural system by providing
at the corners of the building, has been investigated.
substantial improvement in seismic behavior at the
same time.
2.2 Description of models
Recently, experiments to investigate some aspects
of seismic isolation and possibility of tying the walls Taking into consideration the payload capacity of
of heritage masonry buildings with CFRP laminate the simple uniaxial shaking table, installed in the
strips instead of steel ties, have been also carried out structural laboratory of the institute, experiences and
at Slovenian National Building and Civil Engineering available materials, the tests have been carried out
Institute in Ljubljana. Experiments and test results will on models constructed at 1:4 reduced scale. Five
be presented and discussed in this contribution. models with basically the same structural layout have

789
 Figure 2. Typical model during construction.
Figure 1. Scheme of laying the bricks and position of
wooden joists and lintels.

been constructed and tested, prepared for testing in

five different ways. Since the main objective of the
study was limited at obtaining basic information about
the possibility of application and efficiency of tested
strengthening method, a single room brick masonry
house with wooden floors without wall ties has been
tested. The models maintained the basic structural
characteristics of typical buildings, such as storey
height, span between the structural walls and open-
ings’ size, with outer dimensions adjusted to the size
of the platform of the shaking table. The scheme of lay-
ing the bricks and placing the wooden joists is shown in
Figure 1, whereas a typical model under construction
is shown in Figure 2.
Model M1 represented the control model with
wooden floors without wall ties. It has been con-
structed directly on the r.c. foundation slab, fixed to
the moveable platform of the shaking table, without
any specific measures to improve the seismic behavior.
Model M2 was similar, however, a simple PVC sheet
has been placed as a damp-proof course in the bed-
joint between the second and third course of masonry
units. Model M3, also similar to model M1, has been Figure 3. Position of CFRP laminate strips used to confine
the models and strengthening of walls. Transverse section.
isolated with isolators, placed between the foundation
slab and the model’s upper structure.
The effect of tying the walls with CFRP laminate slab. In addition, the piers between the openings have
strips has been studied on Models M4 and M5. The been strengthened with diagonally placed CFRP lami-
strips which simulated horizontal and vertical ties nate strips without any special provision for anchoring
(confining elements) have been bonded to masonry on at the ends. The position and dimensions of CFRP
the outer side of the walls. Horizontal strips have been strips are shown in Figures 3 and 4. In the same figure,
placed at the level of wooden floors. They have been the main dimensions of the models are also indicated.
connected together at the corners of the building with Whereas model M4 has been built directly on the foun-
steel anchor plates. Vertical strips, placed at the cor- dation slab as was the case of control model M1, model
ners, have been glued at the bottom of the walls on steel M5 has been placed on seismic isolators as was the case
angle profiles, bolted into the model’s r.c. foundation of model M3. Whereas model M3 has been placed on

790

Figure 4. Position of CFRP laminate strips used to confine
the models and strengthening of walls. Longitudinal section.
six isolators, only four isolators have been used in the
case of model M5 in order to further shift the natural
frequency of vibration of the isolated model from the
predominant frequency of the model earthquake.
Since physical modeling of CFRP laminates and
bonding properties turned out to be rather compli-
cated if impossible, the technique of simple modeling
has been used so that the strength characteristics of
brick masonry remained close to the prototype values.
Tests of model masonry walls indicated that similar-
ity in non-linear behavior and failure mechanism at
the simulated conditions of prototype loading has been
fulfilled just as well. At the same time, the distribution
of lateral stiffnesses and masses along the height of the
models was in the proportions usual for typical proto-
type buildings, hence fulfilling also the conditions of
similarity of dynamic behavior. Since all models have
been tested in equal loading conditions, their seismic
behavior can be directly compared. However, when
calculating the values of physical quantities measured
on the models to prototype, model scale factors for the
case of the simple model similarity should be taken
into consideration.
The mechanical properties of model masonry, deter-
mined by compression tests (3 specimens) as well as
cyclic lateral resistance tests at different levels of pre-
compression (7 specimens) indicate that the models
have been constructed with model masonry, which
791
Figure 5. Calibration test of seismic isolator.
had mechanical characteristic similar to the prototype
ones. The following mean values have been obtained:
– compressive strength: f = 6,1 MPa,
– modulus of elasticity: E = 1864 MPa,
– tensile strength: ft = 0,23 MPa,
– shear modulus: G = 678 MPa, and
– ductility factor: µ = 3,9.
As a damp-proofing element, a commercially avail-
able PVC sheet, 2 mm thick and cut to fit the dimen-
sions of the cross section of the walls, has been used.
The PVC sheet has been placed in the mortar bed joint
between the second and the third course of units of the
walls in the ground floor of model M2. The sliding
mechanism and friction characteristics of the damp-
proof course in relation to vertical stresses in the walls
have been determined by testing.
The isolators were 92 mm in diameter and 100 mm
high. They have been manufactured of vulcanized
rubber, 8 mm thick. In order to keep adequate stiff-
ness in vertical direction, ten pieces of 2 mm thick
steel sheets have been uniformly distributed along
the height of each isolator. Before placing and fix-
ing the isolators in the position, the deformability
characteristics of each isolator have been determined
by calibration tests (Figure 5). The measured aver-
age lateral stiffness at lateral displacement d = 10 mm
was KH = 0,0325 kN/mm, at displacement d = 30 mm,
KH = 0,0284 kN/mm, whereas the average stiffness in
vertical direction amounted to Kv = 1,64 kN/mm.
To confine the structure and tie the walls with hori-
zontally and vertically placed CFRP laminate strips, as
well as to strengthen the walls with diagonally placed
strips, readily available CFRP laminate, 1,2 mm thick,
has been cut to 2 and 3 cm wide strips. The strips
have been glued on the masonry according to the
instructions provided by the manufacturer. The tensile
Table 1. Maximum accelerations and displacements of the
shaking table, measured during individual test runs (average
values for the groups of models).

Models M1, M2 Models M4 and

and M3 M5

dmax dmax
Duration Test run amax (g) (mm) amax (g) (mm)

12 s R005 0,028 0,795 0,027 0,600

R025 0,131 3,480 0,109 2,854
R050 0,386 7,116 0,239 5,660
R075 0,505 10,827 0,380 8,509
R100 0,688 14,620 0,483 11,322
R150 – – 0,727 17,034
R200 – – 1,015 22,721
R300 – – 2,682 34,002
R350 – – 3,555 39,492

Figure 6. Typical response spectra of shaking table motion.

®
strength of material, Sika CarboDur S, in the direc-
tion of fibers amounts to 3000 MPa, and the modulus
of elasticity to 165000 MPa. Before gluing the strips,
the surface of masonry has been thoroughly cleaned
and penetrated with primer. Original epoxy adhesive
material, SikaDur, has been used to glue the strips on
the masonry.

2.3 Seismic load and instrumentation of models

The shape of the ground acceleration time history, used
to control the shaking table motion, corresponded to
the 24 seconds long strong phase of the N-S com-
ponent of the ground acceleration record, recorded at
Petrovac during the Montenegro earthquake of April
15, 1979. Maximum measured ground acceleration
was 0,43 g. The actual model earthquake, prepared for
testing the so called complete models, was 12 seconds
long, and had the same maximum ground accelera-
Figure 7. Instrumentation of models.
tion as was the case of the actual acceleration record.
The shaking table motion during test run R100 repre-
sented such an earthquake (100% of intensity). Since
the models have been made of materials with strength each test run are given in Table 1, whereas the typical
characteristics similar to the prototype (simple mod- response spectra are shown in Figure 6.
els), the actual model earthquake represented 48 sec All models have been instrumented with a set of dis-
long earthquake (St = SL = 4) with maximum ground placement transducers and accelerometers (Figure 7),
acceleration 0,11 g (Sa = 1/SL = 0, 25). fixed to the models at the level of floors. The miss-
Shaking table displacements in each successive test ing live load at the levels of floors has been modelled
run have been scaled from 5% to 350% of those by means of concrete blocks of adequate mass, which
of the model earthquake (test runs R005 to R350, have been fixed to wooden joists with steel bolts so that
respectively). All models have been tested with the the in-plane rigidity of floors has not been significantly
same sequence of seismic excitations with increased affected. In order to prevent damage to instruments
intensitiy of motion in each successive test run, the and shaking table at the moment of collapse, concrete
characteristics of the model earthquake did not influ- blocks have been loosely hanged on the crane.All mod-
ence the observations. Maximum accelerations and els have been oriented so that the direction of shaking
displacements of the shaking table motion obtained in table motion coincided with longer dimension of the

792
Figure 8. Mechanism of collapse of non-strengthened con- Figure 9. Mechanism of collapse of non-strengthened
trol model M1. model M2 with damp proof course.

model. In other words, seismic loads acted in the direc-

tion of load-bearing walls, pierced with window and
door openings.

3 TEST RESULTS

3.1 Failure mechanism

The control model M1 exhibited typical behaviour of
old masonry buildings with wooden floors without
wall ties: in the beginning of tests when subjected to
low intensity earthquake ground motion, the behaviour
was monolithic. However, with increased intensity of
shaking, vertical cracks developed in the upper part of
the model. As a result of separation of walls, the upper
storey of the model disintegrated in the subsequent test
runs and collapsed (Figure 8).
The tests of model M2 have shown that the damp-
proof course in the form of a simple PVC sheet placed
in the mortar in the bed joint cannot be considered
as seismic isolating device. Although the compressive
stresses in the walls with installed damp-proof course Figure 10. Mechanism of collapse of non-strengthened and
were low, the measurements have indicated that nei- isolated model M3.
ther sliding along the damp-proof course took place
nor rocking motion of the upper part of the building
has been observed. The walls in the upper storey disin- non-isolated models M1 and M2. However, a slight
tegrated and the storey collapsed at the same intensity difference in the sequence of damage propagation has
of excitation as was the case of control model M1 been observed. Whereas damage propagated gradually
(Figure 9). in dependence on intensity of motion in subsequent
Although improved behaviour of model M3, placed test runs in the case of models M1 and M2, the collapse
on rubber seimic isolators, has been expected, model of model M3 was sudden, without cracks occuring
M3 exhibited practically the same poor behaviour as during the previous test runs (Figure 10).

793
Figure 11. Damage to CFRP laminate strengthened, non-
isolated model M4 at the end of shaking table tests.
Figure 12. Unsignificant damage to CFRP laminate
strengthened, isolated model M5 at the end of shaking table
test.
The seismic behaviour of both models strengthened
with CFRP laminate strips, however, was signifi-
cantly improved. They did not suffer severe damage
or collapse even when subjected to ground motion
with accelerations, which by more than three times
exceeded the accelerations measured during the test-
ing of non-strengthened models (Figures 11 and 12).
Since the capacity of shaking table has been reached
794
Figure 13. Detached anchor plate caused rocking of model
M4. Vertical strip buckled at the bottom of the model.
and the output motion already distorted, tests had to
be terminated at that point. In the case of non-isolated
model M4 the anchor bolts, by means of which steel
anchor angles of vertical strips have been fixed to
the foundation slab, pulled out (Figure 13) and the
model started rocking on the foundation slab. Con-
sequently, masonry crushed at the corners and severe
cracks occurred in the lintel parts of the walls.
In the case of model M5 on seismic isolators, one
of the isolators detached (Figure 14). However, almost
no damage has been observed in the model’s walls. It
has to be noted, that also in the case of model M4 no
structural damage has been observed before the pulling
out of anchor bolts. By comparing the results of tests
of CFRP laminate strengthened and non-strengthened
model walls, it seems that, in the particular case stud-
ied, this is the result of confining the model structure
with horizontal and vertical CFRP strips, and not the
result of diagonally placed strips on the wall piers.
However, additional measurements should have been
carried out in order to confirm this observation.
The changes in dynamic characteristic of the tested
models, measured before the tests and after each sub-
sequent test run, are presented in Table 2. The values
of the first natural frequency of vibration f and coef-
ficient of equivalent viscous damping ζ (in % of
critical damping), have been determined by hitting
the model with impact hammer and analyzing the
Figure 14. Detachment of isolator at the end of tests of
model M5.

Table 2. First natural frequency of vibration f (in s−1 ) and

coefficient of equivalent viscous damping ζ (in % of critical
damping) measured on the models before the beginning of
shaking table tests and after characteristic test runs.

Before
Model test R50 R75 R150 R300
Figure 15. Response of control model M1 during test run
M1 f (s−1 ) 15,6 12,3 12,3 – –
R075.
ζ(%) 13,5 16,7 15,5 – –
M2 f (s−1 ) 19,0 15,3 13,9 – –
ζ(%) 5,5 13,4 13,8 – –
which reduced the response and prevented structural
M4 f (s−1 ) 21,2 20,6 19,5 18,9 12,6
ζ(%) 3,8 5,0 9,5 10,2 8,9 damage.
M5 f (s−1 ) 2,2 2,1 2,1 2,1 2,1 As an indication of efficiency of confining the
ζ(%) 11,6 10,2 10,1 10,0 10,3 structures wth CFRP laminate strips, measured top
acceleration and relative first storey displacement
responses of two of the tested models are shown in
Figures 15 and 16. As a measure of intensity of exci-
measured response. Fourier analysis of acceleration tation, shaking table acceleration time history is also
and displacement records has been used to obtaine plotted in each figure. In the figures, the responses of
these data. Unfortunately, dynamic characteristics of control model M1 and CFRP laminate strengthened
isolated, non-strengthened model M3 have not been model M4 to seismic excitation of the same intensity
measured. are compared (test runs R75 and R300, respectively
As expected, a trend of degradation of the first (see Table 1 for comparison of maximum shaking table
natural frequency of vibration and increase in val- accelerations).
ues of coefficient of equivalent viscous damping can Maximum values of the base shear evaluated on the
be observed with increased intensity of excitation in basis of the measured acceleration responses of the
all cases, except in the case of isolated model M5, models during each test run are compared in Table 3.
which has not suffered damage during testing. The Base shear is given in terms of the base shear coef-
differences in initial values, especially in the case of ficient BSC, i.e. the ratio between the base shear
models M1 and M2 can be mainly attributed to rela- BS developed in the model during shaking and the
tively large scattering of deformability characteristic weight of the model above the base W : BSC = BS/W .
of used model masonry materials. The increase in stiff- Base shear has been calculated as the sum of prod-
ness in the case of model M4, confined with CFRP ucts of masses, concentrated at the levels of floors
laminate strips, however, can be attributed to the effect mi and measured average maximum values of accel-
of confining the model structure horizontally and ver- erations at the same level ai : BS = mi ai . Since
tically with rigid elements. No changes in the first all models were equal, the mass concentrated at the
natural frequency of vibration and damping in the case second floor was taken as m2 = 287 kg, mass at the
of isolated model M5 are the result of seismic isolators first floor m1 = 448,4 kg, whereas the total mass of

795
 the case of model M4 without isolators developed in
the model at full intensity of shaking table motion.

4 CONCLUSIONS

Experiments have shown that a simple damp-proof

course in the form of PVC sheet installed in the mor-
tar bed joint cannot be considered as seismic isolation.
Experiments have also shown that seismic isolation
alone is not enough to improve the seismic behaviour
of heritage masonry buildings without wall ties. How-
ever, the shaking table tests of models confined with
horizontal and vertical CFRP laminate strips indi-
cated significantly improved seismic behaviour. The
CFRP laminate confined models did not collapse even
when subjected to ground accelerations which by more
than three-times exceeded accelerations causing the
collapse of the models without wall ties.
The experiments indicated the possibility of replac-
ing the commonly used steel ties by CFRP lam-
inate strips. Placed also vertically, CFRP laminate
strips additionally strengthen the structure, if prop-
erly anchored to foundation system at the ends. The
experiments also confirmed the long known fact that
Figure 16. Response of CFRP laminate strengthened model seismic isolation of rigid masonry structures repre-
M4 during test run R300. sents an efficient way to reduce seismic loads in the
case of the usual short periodic earthquakes. However,
Table 3. Maximum base shear coefficient evaluated on the isolation does not permit that the usual meaures of
the basis of the measured response of the models during seismic strengthening, such as the tying of walls with
characteristic test runs BSC = mi amax,i /W . wall ties, be omitted.
The experiments indicated the efficiency of con-
Model R50 R75 R100 R150 R200 R300 temporary technical solutions. However, they also
pointed out that technological problems need to be
M1 0,245 0,662 0,306 – – – resolved before wider application of such methods
M2 0,225 0,453 0,511 – – –
to heritage masonry buildings. In this regard, effi-
M3 0,305 0,434 0,492 – – –
M4 0,228 0,380 0,488 0,739 1,050 2,032 cient interaction between the materials with extremely
M5 0,108 0,322 0,415 0,564 0,702 0,895 different mechanical characteristics, such as CFRP
laminates and masonry is critical.

the model above the foundation slab, on the basis of ACKNOWLEDGEMENTS

which the weight of the model W was calculated, was
mtot = 856,8 kg.
From the results of tests, summarized In Table 3,
it can be clearly seen that no basic difference in the
resistance has seen obtained in the case of the models
without wall ties. Whereas control model M1 attained
its maximum resistance during test run R75, mod-
els M2 and M3 attained the maximum of somewhat
smaller value at the same intensity of motion during
test run R100. Significant improvement in resistance
(by four times) has been obtained in the case of CFRP
laminate confined model M4. In the case of CFRP
confined and isolated model M5, noticeable reduction
of seismic forces can be observed. As a result of iso-
lation, more than two times lower baser shear than in
The research presented in the paper has been carried
out within the framework of research project L2-0691
and research program P2-0274, financed by the Min-
istry of High Education, Science and Technology of
the Republic of Slovenia and co-financed by rubber
industry, Sava Company Ltd., program Construmat
from Kranj, Slovenia. CFRP laminates and bonding
materials have been given at disposition free of costs
by Sika AG, Slovenian branch from Trzin, Slovenia.
REFERENCES
Bailey, J.S., Allen, E.W. (1988). Seismic isolation retrofitting:
Salt Lake City and County building. APT Bulletin, 20 (2).

796
Gayevoy, A.V., Lissel, S.L. (2004). Monitoring of CFRP pre-
stressed masonry walls. Proceedings, 13th International
Brick and Block Masonry Conference, Vol. 2, Amsterdam.
Hamilton, H.R., Dolan, C.W. 1998. Strengthening unrein-
forced masonry with FRP. Proceedings, Structural Engi-
neering World Wide, Elsevier Science Ltd.: Paper No.
T119-5 (CD-ROM).
Page, A.W. (1995). The shear capacity of membrane type
damp-proof courses in masonry.Australian Civi Engineer-
ing Transactions, Vol. CE37, No.1.
Sarrazin, M., Moroni, M., Boroschek, R., Soto, P. 1996.
Results from a base-isolated experimental building in
Santiago-Chile. O. Proceedings, 11th World Conference
on Earthquake Engineering, Acapulco, CD ROM.
797
Schwegler, G. 1995. Masonry construction strengthened
with fiber composites in seismically endangered zones.
Proceedings, 10th European Conference on Earthquake
Engineering, Balkema.
Triantafillou, T.C., Fardis, M.N. (1997). Strengthening of
historic masonry structures with composite materials.
Materials and Structures, 30 (10).
Triantafillou, T.C. 2001. Seismic retrofitting of structures
with fibre-reinforced polymers. Progress in Structural
Engineering and Materials, 3 (1).
Zhou, X., Miao, H. 1996. Optimum design of resilience-
friction-slide base isolation system for low cost build-
ings. Proceedings, 11th World Conference on Earthquake
Engineering, Acapulco, CD ROM.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Experimental survey on seismic response of masonry models

A. Baratta, I. Corbi & O. Corbi

Department of Structural Engineering, University of Naples “Federico II”, Naples, Italy

D. Rinaldis
ENEA Casaccia Research Center, Italy

ABSTRACT: This paper is focused on the elaboration of the results recorded during some dynamic experi-
mental tests executed on a masonry arch placed on a shaking table facility at Laboratory of the ENEA Casaccia
Research Center. The experimental results are correlated to the theoretical results obtained by means of some
calculus codes, which have been elaborated for applications on masonry structures by the researchers of the
Department of Structural Engineering of the University of Naples “Federico II” and suitably modified for the
study model. The results have shown that the curves solved by the static calculus may be given an exponential
form as the dynamic curves by the tests; moreover, it is evident that the static degradation is much faster than in
the seismic response, probably thanks to kinetic energy absorption that helps the structure to resist inertia forces.

1 INTRODUCTION The scope of the present study, and, in general, of

the experimental campaign, is to characterize the own
Some interesting experimental tests have been devel- vibration frequency of existing structures and to elab-
oped at Laboratory of the ENEA Casaccia Research orate a rationale to correlate the theoretical results to
Center (New Technology Department, Rome, Italy) in situ dynamic identification and to the laboratory
on some masonry arches within a national conven- tests that have been performed. The static analyses
tion with the Authority of the Monumental Heritage have been carried on by calculus codes that have
Supervision in Benevento and Caserta Provinces as a been implemented for no-tension material structures
preliminary step finalized to the strengthening for seis- at the Department of Structural Engineering of the
mic risk of the San Rocco’s church in Benevento and of University of Naples “Federico II”.
the S. Francesco’s bell tower in Montesarchio (Campa-
nia, Italy). More in general, the study spaces from the
dynamic characterization of the considered structures, 2 APPLICATION OF THE NO-TENSION (NT)
realized by means of the monitoring of environmental STRESS ANALYSIS: THE MASONRY ARCH
and forced vibrations of the churches, to the analysis
of the seismic conditions at the site where the ancient In structural patterns of the type of arches (or vaults),
structures are built, to the tests on shaking table of the stress field can be inferred from the internal forces
selected structural elements (in real and reduced scale) on every cross section by a pattern as in Figure 1
and of some scale models of the churches. (Baratta & Corbi, 2003; 2005).
This paper is focused on the experimental tests
executed on some masonry prototypes placed on a
shaking table at the ENEA laboratory. The experi-
mental investigation consists of seismic tests acted
on some masonry arches, made of tufa bricks and
cement mortar; the prototype is solicited under quakes
of increasing intensities up to the collapse condition.
This work starts from the collection of the existing
recorded data of the laboratory tests: a pretty consistent
number of response diagrams in the time and fre-
quency domains (time history of acceleration, power
spectrum, frequency response and coherency func-
tion), produced by the sensors placed on the structure. Figure 1. Stress pattern on cross sections.

799
 Figure 3. Admissible funicular line.

Figure 2. Superposition schemes for managing equilibrium

stress fields.

The set of stress fields in equilibrium with applied

loads can be built up by a superposition scheme of the
type shown in Figure 2.
The solution of the no-tension (or NT) structural
problem is approached by the Minimum principle of
the Complementary Energy, and the procedure aims at
Figure 4. Specimen (in cm) of the masonry arch for the
identifying the redundant reactions allowing internal laboratory tests.
and external constraint compatibility.
Let Do be the definition set of the admissible NT
stress fields in equilibrium with the applied loads; the
stress field σo is found as the constrained minimum of
the Complementary Energy functional U (σ) under the
condition that the stress field is in equilibrium with the
applied loads and compressive everywhere
3 DYNAMIC EXPERIMENTAL TESTS ON
MASONRY ARCH BY MEANS OF
SHAKING TABLE

3.1 Testing model: the masonry arch

The admissibility of the stress field (Baratta & Corbi,
2003; 2005) is guaranteed by the condition that the
force funicular line is everywhere in the interior of the
arch profile (Fig. 3)
For the tests executed on the shaking table, at Labo-
ratory of the ENEA Casaccia Research Center (New
technology Department) two similar arches are built
having a circular round-headed axis and the geometry
shown in Figure 4. The arcade built in tufa bricks rests
on two piers, which continue over the imposts of the
arch in order to contain the overload imposed on the
top of the portal arch with the help of a tie-rod.

800
 The structure is made by yellow tufa bricks tied
by a poor mortar, which is a most common masonry
encountered in South Italy. The intrados profile of the
arcade is semicircular with a radius of 100 cm. The
arcade is composed by two rows of bricks determining
a masonry thickness of 20 cm; the two pillars which
support the arcade have a rectangular base 50 cm and
an height 70 cm; the depth of the whole is 100 cm. The
wing walls, continuing in height the pillars, which have
to contain the overload, are characterized by thickness
35 cm and an height 110 cm.
Moreover, some steel tie-beams fixed by means of
flexible elastic ties are placed between the structure
and the wing walls used for containing the overload,
in order to guarantee the stability during the tests. The
total weight of the masonry structure is 5,1 tons. On the
top of the portal arch an overload of material composed
by crushed tufa and lime and having a weight of 1,4 ton
is applied, in order to simulate the structural context
where the real arcade is included. Sliding is prevented Figure 5. Sensors locations on the masonry arch by
through steel profiles attached to the shaking table. Buffarini et al. (1997) modified.
The total weight of the structure plus the overload
and the steel bars is 7,0 tons. are fixed to some steel trestles integral with the shak-
ing table. A scheme of the arch with the locations of
the sensors composing the monitoring equipment is
3.2 Characteristics of the shaking table and shown in Figure 5 (Buffarini et al., 1997; Clemente
of the recording instruments et al., 1999).
As described in the above, during the tests, the seis-
mic input is transmitted to the structure by means of 3.3 Description of the laboratory tests
a shaking table; the table has the following techni-
cal characteristics: dimension of 4 × 4 m, maximum At first the tuning of the shaking table is realized in
supported weight of 10 ton, six degrees of freedom, order to check the response of the table. To this pur-
frequency range of 0–50 Hz, maximum peak accel- pose, the table is loaded with a fictitious structure
eration of 3 g, maximum velocity of 5 m/sec and simulating the testing structure both in its total weight
maximum span of 25 cm. and in its barycentre position, and the fictitious struc-
In order to evaluate the time histories of the accel- ture is subject to the selected profile, consisting in
eration and displacement some recording instruments the selected time history of the acceleration. The final
are located at some “critical” positions on the arch, objective of the tuning phase is the realization of the
which are fundamental for the determination of the test profile (Buffarini et al., 1997).
structures’ response.
The recording instruments consist of two different 3.3.1 Tests developed on the 1st portal arch
typologies of accelerometers: The first arch is tested in two phases up to the collapse.
In the first phase (Phase A) only an horizontal input is
• n. 20 piezoelectric accelerometers with feed- used; thereafter, in order to approach the collapse con-
through band of 2–15000 Hz (± 10%), and nominal dition of the structure, in the second phase (Phase B),
sensitivity of 10 pC/g, a vertical input is also added to the horizontal one. In
• n. 8 transducers of displacement (LVDT) subdi-
both the two phases the applied excitation acts in the
vided in: plane of the structure, orthogonally to its side at the
– transducers with nominal sensitivity of 0,1 Vmm-
bottom of the portal arch.
1 (± 3%), maximum displacement of ± 2 inch,
In the first phase (Phase A) the characterization of
and feed-through band of 50 Hz,
the arch having an overload on the top of the arcade
– transducers with nominal sensitivity of 0,2 V/mm
fixed to 14,1 kN is developed by means of a white
(± 3%), maximum displacement of ± 1 inch,
noise with amplitude of 0,1 g (as well known, the
and feed-through band of 50 Hz.
theoretical white noise is a disturbance with infinite
The accelerometers are directly applied on the energy uniformly distributed over the whole frequency
masonry arch, while the transducers are located in cor- range. After the initial characterization of the struc-
respondence of the external sides of the two piers, and ture, an excitation corresponding to the time history

801
Figure 6. Curve of the structural frequencies f vs the seismic
coefficient c during the laboratory tests on the first arch.

recorded in Sturno during the earthquake occurred in

Campania-Lucania on the 23rd November 1980 with
direction W-E, and increasing amplitude from 0,3 g up
to 1,5 g, is transmitted to the arch. The recorded data
are, then, filtered by means of a high-pass filter with
a cut-off frequency of 2 Hz, in order to contain the
maximum value of the displacement within the max-
imum table stroke of 12,5 cm. Thereafter, the record
is suitably scaled in order to obtain a peak excitation
of 0,1 g. After any test a white noise having a fixed
amplitude of 0,05 g, and lasting 60 sec, and finalized
to the characterization of the structure, is tested on
the arch.
The first cycle of laboratory tests was stopped at the
peak acceleration 1.5 g and caused some damages on
the structure never leading to its collapse. This unex-
pected result may be due to two major reasons: the first
is because for any increment of the damaging a reduc-
tion of the own frequencies of the total apparent modes
of the structure occurs, and the second reason is due
to a-priori filtering applied on the lower-frequencies
of the Sturno accelerogram (Fig. 6).
In the second phase (Phase B) a vertical input with
amplitude of 1,5 g is added to the horizontal input,
which continues to increase its amplitude from 1,5 g up
to the collapse condition encountered at 1,8 g (Fig. 7).
3.3.2 Tests developed on the 2nd portal arch
On the basis of the results obtained on the first arch,
a sequence of seismic tests with increasing intensity
is transmitted to the second arch, by assuming as
base signal the W-E component of the acceleration
time history recorded at Sturno during the earthquake
of Campania-Lucania (Italy) on the 23rd November
1980; moreover, a varying overload on the top of the
arcade is introduced.
A preliminary filtering at low frequencies up to
2 Hz has been introduced to avoid displacements larger
than the maximum allowable for the shaking table.
Then the accelerogram is scaled in order to obtain
Figure 7. Two phases of the arch collapse during the
laboratory tests (a) and (b).
a peak acceleration of 0,1 g, and its peak accelera-
tion is amplified of 0,1 g at any step. The subsequent
tests with increasing intensity have been staggered to
some tests for the dynamic characterization, realized
by soliciting the structure with a white noise having a
fixed peak of 0,1 g. Moreover a varying pink noise of
8–16 Hz (a signal where the energy is transmitted in the
interior of a well defined frequency band, and can be
created by high-low or band pass filtering white noise)
is introduced in the tests, probably becoming the cause
of the premature and unexpected damaging of the
structure.
For all the records a sampling step equal to 0,02 sec
is adopted and the total time of the record amounts
at 102,38 secs for the characterizations and the tests
with the pink noise, and at 61,42 sec for the seismic
tests. The time of the effective signal is respectively of
60 sec and 40 sec.
The base acceleration is always applied into the lon-
gitudinal direction, determining a plane stress state,
unless of spatial effects due to the presence of unavoid-
able executive defects.

802
 Three cycles of tests with different overload and Table 1. Scheme of the laboratory tests executed on the
excitation are developed in order to study the behavior second arch pre-consolidation.
of the second arch:
Test Solicitation Note
Cycle 1) in the first cycle an overload of 8,5 kN is
applied on the top of the portal arch, and the seis- Cl white noise Overload 8.5 kN. Tests
mic input consists of a white noise for the seismic start (cycle 1)
Pl pink noise_01
characterization of the structure, a pink noise and TI test 01
two tests of 0,1 and 0,2 g. C2 white noise
Cycle 2) In the second cycle an overload of 14,1 kN P2 pink noise_02
(equivalent to the one acting on the first arch) is T2 test 02
C3 white noise
applied and the structure is subjected to two seismic
ClA white noise Overload 14.1 kN. New
tests with peak 0,1 and 0,2 g; the cycle is then inter- start of tests (cycle 2)
rupted because of the detachment of some stones PIA pink noise_0l
at the intrados. T1A test 01
Cycle 3) In the third cycle, started after the restora- C2A white noise
P2A pink noise_02
tion of the broken bricks, an overload of 14,1 kN is T2A test 02
kept, the pink noise which is probably the cause of C3A white noise
the masonry slack is removed, and the seismic tests P3A pink noise_03 Test stop for breaking
are continued up to the peak acceleration of 0,7 g. off of some stones
C1R white noise Restoration of the
By analyzing the results, on the basis of the test broken bricks and
program shown in Table 1 (from the laboratory tests’ new start of the test
schemes by Buffarini et al., 1997 modified) and (cycle 3)
T1R test 01
Figure 8, some considerations can be outlined. C2R white noise
After the dynamic characterization (C1) developed T2R test 02
on the undamaged structure, the first test (T1) has been C3R white noise
developed with the acceleration time history scaled to T3R test 03
C4R white noise
a peak of 0,1 g. At any step the peak value is increased T4R test 04
by 0,1 g, and so on up to a maximum peak value of C5R white noise
0,7 g (T7R), about twice the real peak acceleration T5R test 05
recorded to Sturno, which is equal to 0,34 g. The test C6R white noise
T6R test 06
with a peak of 0,7 g (T7) was interrupted because of C7R white noise
the detachment of some stones at the intrados and then T7R test 07 Test stop for breaking
repeated (T7R) (Clemente et al., 1999). After any test off of some stones.
a characterization of the structure has been developed C8R white noise Characterization
following to the
by operating a white noise input (C1 to C9) plus a pink previous test
noise having an increasing peak value of 0,1 g at any C8RB white noise Characterization
step (P1 to P8). before the testing
In the first three characterizations (C1 to C3) an start
C8RATT white noise Characterization with
overload of 8,5 kN has been imposed on the top of the a scale spectrum 1/2
arch, then the tests have been interrupted and started T7RB test 07 Final test
again by a peak value of 0,1 g and an overload of C9R white noise
14,1 kN has been imposed (C1R to C9R) in order to
make tests homogeneous to the 1st arch.

3.4 Comparison between the dynamic results
in the laboratory tests
By the comparison between the curves in Figures 6
and 8, which plot the results by the dynamic tests in
laboratory on the masonry arches, some general con-
siderations can be made about the behavior of the two
arches (Fig. 9).
The laboratory tests put to evidence some differ-
ence between the two arches. Both the arches show
a typical behavior where the trend of the recorded
frequencies decreases with respect to the seismic
intensity (the coefficient “c = ap /g”) the arch has been
subjected, approximately according to an exponential
curve. Nevertheless the dynamic curve of the first arch
decreases more quickly than the curve of the second
arch. Probably this effect is due to the different pro-
gram of the shaking sequences to which the arches
have been subjected during the tests: in the tests of the
second arch an increasing pink noise has been cou-
pled to the Sturno input, with the result to anticipate
the damage of the arch.
This different behavior is pointed out in Figure 9
where the decay of the first own frequencies of the
two arches is shown. The fact is however that the trend
of both curves can be approximated by an exponential

803
Figure 8. Dynamic curve of the structural frequencies f(Hz)
vs the seismic coefficient c = ap /g during the laboratory tests
for the second arch.
Figure 9. Comparison between the dynamic curves of the
structural frequencies f vs the seismic coefficient c during
the laboratory tests on the two arches.
curve, the curve relevant to the first arch decreases
more quickly than the second one.
4 AN INSIGHT ON VULNERABILITY
FORECASTING OF A MASONRY
STRUCTURE
4.1 Procedure for vulnerability evaluation
In order to have an estimate of the seismic vulnerability
of a existing masonry structure under a seismic input,
an attempt is made to draw conclusions on the basis of
dynamic identification coupled with static analysis, in
the light of the above experimental survey.
At first, the static analysis yielding the fundamental
elements of the examined masonry structure can be
solved by considering only the geometric dimensions
and some other easily collectable data. So the trend of
some characteristic displacement parameter uc (e.g.
the maximum displacements of points of the structure
umax ) with respect to the seismic coefficient “c” (with
804
Figure 10. Trend of the seismic coefficient c vs the maxi-
mum displacements umax by the static calculus.
Figure 11. Trend of the tangential stiffness K = c(u) vs the
seismic coefficient c (b) by the static calculus.
c = ap /g) can be inferred up to the collapse condition
for c = cf (Fig. 10).
By deriving the seismic coefficient c one gets
the trend of the tangential stiffness K = K(c) = c (u)
as a function of the displacement umax and of the
coefficient c (Fig. 11).
It can be conceived that the own frequency of the
structure is a function of the type f2o = α2 Ko , where Ko
is the stiffness of the structure. Since the structure has a
NT non-linear behavior (Fig. 10), one expects that the
frequency depends on the stress intensity, that is mea-
sured by “c”. So, referring to the tangential stiffness
K(c), one grossly puts f2 (c) = α2 K(c).
After dynamic characterization on the structure in
site (e.g. by soliciting the fabric with a white noise
or other) one identifies the own frequency fo before
any earthquake strikes on the
√ building. This allows to
estimate the coefficient α = fo Ko , after calculating
the initial stiffness
√ Ko . It is found that, heuristically, the
curve f(c) = α K(c) follows a exponential proceeding
(Fig. 12). This curve is referred to in the following as
the “static damage progression”.

Figure 12. Comparison between the trend of the frequency
f vs the seismic coefficient c by the static calculus (trian-
gles) and by the seismic tests (squares), and the relevant
exponential approximations, for the first masonry arch.
Figure 13. Comparison between the trend of the frequency
f vs the seismic coefficient c by the static calculus (trian-
gles) and by the seismic tests (squares), and the relevant
exponential approximations, for the second masonry arch.
A second sentence can be assumed, i.e. that also the
progression of damage in a sequence of earthquakes
with increasing intensity is of the exponential type,
as observed in the experiments that have been sum-
marized in the above. This curve is referred to in the
following as the “dynamic damage progression”.
So the static and the dynamic curves, f∗ (c) and f(c)
respectively, can be directly compared, after having
been reported to the same initial frequency f(0) = f∗ (0)
(Figs 12–13).
4.2 Exponential damage progression
Considering that damage and consequent proneness to
seismic collapse evolves according to an exponential
law, e.g. having the form assumed in the following
805
where “c” is the intensity of the worst earthquake the
structure has suffered in the past and to which it has
survived. At the same time the own frequency “fo ”, i.e.
the easiest parameter that can be evaluated by dynamic
identification, evolves with increasing the damage of
the structure.
So, if one considers the derivative of the expression
in Eq. (4)
and the ratio
The rate of decay of the own pulsation turns out to be
constant, despite the fact that in a nonlinear structure
it is to be expected that the hazard increases with the
intensity of the ground shaking.
4.3 Comparison between the static and dynamic
results
Looking at the diagrams of the static analysis and of
laboratory tests, some observations can be outlined.
The first one is that the experimental seismic decay
of the own frequency of the arch when increasing
the seismic intensity can be modeled by means of an
exponential curve (Fig. 7).
Moreover, the static degradation of the arch,
inferred by means of the static calculus on an NT arch,
can be approximated by an exponential curve as well
(Fig. 11). The degradation is here measured by the
variations of the tangential stiffness with respect to the
increasing of the seismic component of the overload.
By the comparison of the exponential approxi-
mations of the static and dynamic curves, shown in
Figures 12–13, it is evident that the static degradation
is much faster than the dynamic one.
This effect is probably caused by the opening of
the fractures in the arch that produces an increasing
absorption of the oscillation energy as kinetic energy
at the limit for the mechanism activation, rather than
as elastic energy.
It is possible to emphasize some tentative forecast
that is possible to draw on the basis of the experimental
results.
By the observation of the diagrams, it can be con-
sidered that the two exponential curves (“static” and
“dynamic” curves) can assume a very similar form by
changing the scale of the abscissas “c”.
So, considering that the static curve obeys the
equation
Figure 14. The experimental evidence shows that dynamic Figure 15. Experimental absorption of inertia forces in
interaction with inertial forces strongly improves the seismic accordance with Eq. (12).
performance of the structure.

while the dynamic curve obeys Eq. (4), it is clear

that if the “static” c is multiplied times q∗ /q, the two
curves become very similar. If the same transforma-
tion is applied to the ordinates in the static calculus
line (Fig. 10), one obtains a very significant increase
in the seismic capacity of the structure, approximately
a seismic peak acceleration that is 5 times larger than
the limit static force.
This result would agree with test, so that it has been
necessary to rise the peak ground acceleration up to
1.8 g to bring the arch to collapse.
It is necessary to remark that the difference between
the “seismic” and the “static” curve is due to iner-
tial forces due to accelerations involved in the arch
deformation. Figure 16. Final pseudo-force/displacement curve under
Such accelerations “a” can be grossly related to the seismic action.
maximum displacements umax through the pulsation
ωo by a relation of the type

It should be expected that the difference c between

the two curves in Figure 14 can be referred to some
additional displacements umax that, according to
Eq. (11), is given by

Figure 17. Comparative seismic forecasts for both arches.

where a = g c.
This is a necessary condition for additional inertial
forces absorption. In Figure 15 the experimental values
of c are plotted versus umax as expressed in Eq. (9).
By adding the additional displacement to the abscis-
sas in Fig. 14, one gets a possible reconstruction of The procedure can be applied to both arches, yield-
the relationship between the peak acceleration and the ing comparative results of the seismic forecast as in
maximum displacement under seismic excitation Fig. 17.

806
5 CONCLUSIONS
By comparing the dynamic curves of the tested arches
(Fig. 9), which can be well approximated to an expo-
nential form, one observes that the curve relevant to
the first arch decreases more quickly than the second
one. Moreover, by the comparison of the exponen-
tial approximations of the static and dynamic curves
(Fig. 12), it is evident that the static degradation is
much faster than the dynamic one. This effect is prob-
ably caused by the opening of the fractures in the arch
that produces an increasing absorption of the oscil-
lation energy as kinetic energy, mainly due to the
progressive activation of a collapse mechanism, rather
than as elastic energy. The considerations illustrated in
Sec. 4.3 enable a transformation of the static analysis
results to produce a "seismic response expectation" as
in Fig. 16. Looking at the results of the static analysis
elaborated to yield seismic expectation (Fig. 17) the
second arch has to produce larger displacements than
the first one, to store a given amount of elastic energy.
This means that a larger part of the external energy
displayed by the earthquake has to be transformed by
the second arch in kinetic energy through the activa-
tion of a mechanism. The larger the amplitude of such
mechanism the faster the arch approaches collapse.
ACKNOWLEDGEMENTS
The laboratory tests were funded by the ENEA
Casaccia Research Center (New technology Depart-
ment, Rome, Italy) and by the Authority of the Monu-
mental Heritage Supervision in Benevento and Caserta
Provinces. The recent advancements have been per-
formed thanks to the financial support by the Depart-
ment of Civil Protection of the Italian Government,
through the RELUIS Pool (Convention n. 540 signed
07/11/2005, Research Line no. 8).
807
REFERENCES
Baratta, A. 1995. The no-tension approach for structural
analysis of masonry buildings. Proc. of the Fourth Inter-
national Masonry Conference, British Masonry Society,
London.
Baratta, A. 1998. Scale influence in the static anc dynamic
behaviour of no-tension solids. Proc. SMART 98 Confer-
ence, Varsavia.
Baratta, A. 2007. Active control technology for dynamic
shaking-table tests on scaled models of masonry struc-
tures. Proc. of Thematic Conference on Computational
Methods in Structural Dynamics and Earthquake Engi-
neering, ECCOMAS, M. Papadrakakis, D.C. Charmpis,
N.D. Lagaros, Y. Tsompanakis (eds.), Rethymno, Crete,
Greece.
Baratta,A. & Corbi O. 2003. The No Tension Model for
the Analysis of Masonry-Like Structures Strengthened by
Fiber Reinforced Polymers. Intern. Journal of Masonry
International, British Masonry Society, vol. 16 No. 3:
89–98.
Baratta, A. & Corbi O. 2005. On Variational Approaches in
NRT Continua. Intern. Journal of Solids and Structures,
Elsevier Science, Vol. 42, pp. 5307–5321.
Bintrim, J.W., Laman, J.A. & Boothby, T.E. 1998. Dynamic
testing of masonry arch bridges. In Sinopoli A. (ed.)
Arch bridges, Proc. Second International arch Bridge
Conference held in Venice: 295–303, Rotterdam, Balkem.
Brown G., Pretlove A.J., Ellick, J.C.A., Hogg, V. & and
Choo, B.S. 1995. Changes in the dynamic characteristics
of a masonry arch bridge subjected to monotonic load-
ing to failure. In Melbourne C. (ed.) Arch bridges, proc.
First Internationa Arch Bridge Conference held in Bolton,
London:375–383.
Buffarini, G., Clemente P. & Rinaldis D. 1997. Analisi speri-
mentale su tavola vibrante di un arco in muratura. Internal
Report of ENEA Casaccia Research Centre, Rome, Italy
(in italian).
Clemente, P., Buffarini, G., Rinaldis, D. & Baratta, A. 1999.
Changes in the dynamic characateristics of a masonry arch
subjected to seismic actions. Proc. of the Fourth Euro-
pean Conference on Structural Dynamics, EURODYN’
99, Prague, Czech Republic.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Bond performance of fiber reinforced grout on brickwork specimens

I. Carbone & G. de Felice

Department of Structures, University Roma Tre, Rome, Italy

ABSTRACT: This study aims at investigating the bond performance of fiber reinforcement with cement-based
matrix applied on masonry structures. The preliminary results of an experimental campaign addressed to explore
the influence of anchorage length, the effect of the discontinuity induced by mortar joint and the consequences
of fiber stiffness and roughness are presented; different specimens, made either with brick or with brickwork,
and reinforced either with carbon fibers or with steel fibers, have been tested when varying the anchorage length.
The failure modes that take place have been detected together with the corresponding debonding force and the
slip at collapse. Besides, τ-slip relationships have been deduced from experimental data.

1 INTRODUCTION in civil engineering and does not require skilled labor;

The use of fiber reinforcement is now becoming cur-
rently used in civil engineering for load capacity
enhancement of concrete and masonry structures. A
wide research (Neale 2000) has been devoted to study
the performances of fiber-reinforced polymers (FRP)
composites used for strengthening masonry structures
in the form of externally bonded surface reinforce-
ments. The reinforcement may enable masonry to
carry tensile forces, and therefore may be useful for
strengthening masonry arches, (Valluzzi et al. 2001,
Foraboschi 2004), or for enhancing seismic protec-
tion of masonry buildings, such as out-of-plane loaded
walls (Triantafillou 1998, Mosallam 2007).
Among the different types of composites available,
the use of cement-based matrix has been recently
proposed for applications on historical masonry struc-
tures (Borri et al. 2007); when compared to polymeric
matrix, cement-based matrix, generally provides a
lower strength of the bond between masonry and the
reinforcement (Huang et al. 2005). However, it fre-
quently happens that the surface of masonry is rough
and irregular and has to be regularized before the FRP
sheet is applied; the effects of surface preparation may
be responsible of a decay of the FRP-masonry bond
strength (Mazzotti et al. 2007); while for mortar based
applications, the regularizing layer is made by the same
grout to be used as the reinforcement matrix, without
any decay in the bond strength. There are some advan-
tages in using cement-based matrix, instead of resin
matrix, which are briefly recalled: (i) the mortar has
a better performance in terms of fire resistance; (ii)
it is easy to apply since the material is currently used
(iii) it allows a much higher transpirability with respect
to polymeric matrix.
In the recent years, several applications with
cement-based matrix have been proposed but the
effective mechanical properties of the bond between
masonry and reinforcement has not been carefully
investigated. Most of approaches have been devoted
to reinforced concrete rather than masonry structures.
In Huang et al. (2005), steel reinforced grout (SRG)
has been considered for strengthening reinforced con-
crete beams, through an experimental investigation on
specimens under four point bending. In Di Tommaso
et al. (2007) an experimental investigation on rein-
forced concrete beams under flexure, was undertaken
to compare the performances of C-FRP and C-FRCM.
Bond-slip tests of SRG sheets applied to concrete spec-
imens have been carried out in Matana et al. (2005).
Similar tests on brick specimens reinforced with FRP
have been undertaken in Briccoli Bati (2007), aiming
at investigating the effective bond length. However, an
experimental campaign on the performances of bond
in cement-based matrix is still lacking.
Aiming at analyzing the load-deformation behavior
and debonding of reinforcement sheets on masonry
specimens, in this paper the preliminary results of
an experimental investigation are presented of Steel
Reinforced Grout (SRG) and Carbon Fiber Rein-
forced Cement Matrix (C-FRCM) on masonry struc-
tures. Both reinforcement sheets are embedded in a
cementicious grout, while the reinforcement is made
either by steel cords (SRG) or by carbon fibers
(C-FRCM). Three different test setups have been
investigated, having different bond length, on two

809
Figure 1. Steel cords on a polyester net.
masonry supports: brick only, and brick masonry. The
purpose of the tests was to investigate the behavior of
the bond made with grout and to evaluate the ultimate
failure load due to delamination.
In section 2 the material characteristics and their
mechanical properties are briefly recalled, then in sec-
tion 3, the experimental setup is presented and, finally,
in section 4 the preliminary results of the experimental
campaign are reported.
2 MATERIALS PROPERTIES
2.1 Steel fibers
Steel reinforced grout consists of steel wires forming
cords that are put together into a fabric and embedded
in a cement-based mortar. The steel cords are obtained
form the same manufacturing process used for the
reinforcement of automobile tires, and are manufac-
tured and distributed by Hardwire, Maryland, USA.
The steel wires may have strength between 2400 and
4000 MPa, and diameter from 0.20 to 0.48 mm; they
are assembled according to different patterns, where
the surface of the cord is rough, in order to provide a
mechanical interlock with the grout, useful to enhance
the stress transfer between the wires and the matrix.
The cord used in the present experimental investiga-
tion (3SX) is made by three wires that are twisted in
a long lay length, wrapped by a fourth wire twisted at
a short length, that provides additional surface rough-
ness. A brass cover protects the wires from corrosion.
The small thickness of the cords makes it possible
to provide the reinforcement sheet with the required
flexibility, which is necessary for application. The
wiredraw allows high tensile strength, which however
is counterbalanced by a very limited ductility capac-
ity; in any case, tensile cord failure should be avoided
by a proper design of the reinforcement.
Cords are held together by knit yarns, made with
polyester, that maintain the spacing of the steel cords
as shown in figure 1. The spacing between cords may
vary between 6.8 mm and 0.4 mm according to design
purposes; in the present case an intermediate density
810
Figure 2. Carbon fibers on a polyester net.
of 12 cords per inches was selected as the most appro-
priate for avoiding steel fiber collapse, while allowing
a proper bond with the matrix between the cords.
Some preliminary tests on single cord and on 1.5
inches sheet (18 cords) have been performed in order
to estimate the stiffness, strength and ductility of the
reinforcement alone. The results are almost the same
for a single cord and a sheet, resulting in an average
tensile strength of 2480 MPa and an (apparent) elastic
modulus of 125 GPa.
2.2 Carbon fibers
Carbon fibers are distributed by Ruredil Milan, Italy;
the reinforced sheet is composed of two orthogonal
spaced strips of carbon fibers 4 mm wide, with thick-
ness 0.047 mm. and spacing of 6 mm. between each
strip (figure 2).
As for steel reinforcement, both one carbon strip
and a sheet of 1.5 inches, comprising four strips, have
been tested under traction without matrix. The tensile
resistance of the carbon fiber is about 4800 MPa and
the elastic modulus about 380 GPa; the resistance of
the sheet is about 80 kN/m.
3 EXPERIMENTAL SET-UP
The tests have been performed on brick and on
brickwork specimens as shown in figure 3.
Current bricks, with dimensions mm 55 × 120 × 230
have been used. Brick specimens (A,B) have assem-
bled by joining together two bricks and applying the
reinforcement on both external faces (figure 3)
Masonry specimens (C) have been made with
five layers of the aforementioned bricks and mortar
joints 10 mm thick. The mortar is premixed hydraulic
lime with pozzolana, with low (2 N/mm2 ) compres-
sive strength, similarly to ordinary mortar used in
widespread masonry buildings. The reinforcement is
then applied on the two opposite sides of the specimen.
The mortar used as the matrix of the reinforce-
ment is Ruredil Xmesh M25, an inorganic grout

Figure 3. Tested specimens.
with synthetic polymeric fibers added; the mechan-
ical properties of the grout are: compressive strength
38 N/mm2 , tensile strength 7.5 N/mm2 and elastic
modulus 15 GPa.
A reinforcement sheet 1.5 inches large is applied on
a layer of grout with thickness 2÷3 mm, which reg-
ularize the surface of the specimen, and is covered
with an equal layer to get a final 6÷7 mm thickness.
Aiming at avoiding fiber collapse in specimens rein-
forced with C-FRCM, two layers of reinforcement are
applied, with a grout layer of 2÷3 mm thick, interposed
between the sheets. To restrict local boundary effect, a
reinforcement band 10 mm large from the brick edge
is not glued.
An anchorage length of 300 mm have been tested
with masonry specimens (C), while on brick speci-
mens two anchorage length of 55 mm (A) and 225 mm
(B), have been tested, where the former corresponds
to one brick thickness and the latter is the maximum
anchorage length permissible with the brick specimen.
The setup of tests is shown in figure 4: a steel curved
shape anchorage draw the sheet fiber and a steel plate,
connected with a tie-bar, pull-out the specimen. The
setup is designed to give rise only to shear stress in the
support-reinforcement interface.
The experimental investigation is still ongoing; at
present time, as reported in table 1, the following
tests have been carried out on brick specimens: three
tests with short anchorage length (A) and six tests
with medium anchorage length (B) with SRG rein-
forcement; three tests with short anchorage length
(A) and two tests with medium anchorage length (B)
on C-FRCM reinforcement; while for masonry spec-
imens, only six tests with SRG reinforcement with
long anchorage (C) have been performed. All the tests
have been carried-out under displacement control with
811
Figure 4. Test setup.
Table 1. Experimental program.
Number of test
Anchorage length
Specimen SRG C-FRCM (cm)
A 3 3 5.5
B 6 2 21.5
C 6 – 30.5
a velocity of 0.01 mm/sec. The bond slip is mea-
sured with displacement transducers at both reinforced
sides of the specimen. Gauge stroke and sensitivity are
10 mm and 0.05 mm respectively. The tensile forces in
the reinforcement sheet at both sides are computed
from the deformation recorded by two strain-gauge
and then controlled by the overall force recorded by
the actuator. A set of strain-gauge with 10 mm length
are placed along the bonded reinforced sheet with a
uniform step of about 40 mm to register the defor-
mation of the fibers along the reinforcement strip. In
some of the specimens, two sets of strain-gauge have
been placed at the same distance form the edge, side
by side, in order to detect any possible difference in
deformation between the cords.
4 EXPERIMENTAL RESULTS
4.1 Failure modes and bond strength
The failure modes that take place in the experiments
are respectively peeling of the brick surface (a), crack-
ing in the grout (b) and slipping of the reinforcement
(c) as sketched in figure 5.
In the specimens reinforced with SRG, having short
(55 mm) reinforcement anchorage, the collapse occurs
with the detachment of the reinforcement from the
brick with peeling of the external surface of the brick
Figure 5. Failure modes.
Figure 6. Failure modes for short anchorage (A) with SRG
(left) and C-FRCM (right).
Figure 7. Failure modes of SRG with medium (B) anchor-
age: state of the specimen at the end of the test with the
debonded side (left) and the crack at the opposite side (right).
(type a) as shown in figure 6; in the specimens with
medium reinforcement anchorage (225 mm), the col-
lapse occurs within the grout, along a surface just
below the fiber layer (type b) as shown in figure 7.
Only at the end of the reinforcement sheet, the fail-
ure takes place with the peeling of the brick surface as
in the case of short anchorage. Similar experimental
812
Figure 8. Failure in C-FRCM specimens with medium
(B) anchorage (left), and in SRG specimens with long (C)
anchorage (right) on masonry support.
evidences were obtained in a previous experimental
work on SRG on concrete specimens (Matana et al.
2005).
The crack initially forms at the side where the
reinforcement is drawn and propagates within the
anchorage until debonding occurs suddenly. In all
cases the collapse is almost brittle. In masonry speci-
mens with long (C) anchorage, a similar behavior as in
brick specimens with medium anchorage is observed;
however, due to the discontinuity induced by mor-
tar joints, which have poor mechanical properties, the
crack propagates only as far as the first mortar joint is
reached, then debonding takes place suddenly. As for
medium anchorage, the crack forms within the grout at
the interface with the reinforcement (type b), except
for last brick, where peeling at brick-grout interface
takes place, similarly to short anchorage length.
A summary of the experimental results is reported
in table 2 where debonding strength force is reported
together with slip at collapse and failure mode that
takes place.
In figure 9, the experimental values of debonding
strength vs. anchorage length are shown and compared
with the results of the tests carried out in Matana et al.
(2005) on concrete supports. The results show that,
by parity of anchorage length, steel fiber reinforce-
ment (SRG) provides higher debonding strength, than
carbon fiber sheet (C-FRCM); as regards the behav-
ior of different supports (brick or masonry), a lower
debonding strength is observed for masonry speci-
mens since the mortar joints act as weak elements
that anticipate the complete debonding which starts
suddenly when the crack reach the first mortar joint.
In figure 10 the experimental load to slip diagram
is drawn for specimens B and C, reinforced with SRG;
as expected, a higher stiffness of the brick specimens
is observed.
Table 2. Experimental plane.

Collapse force Type of Slip at collapse

Specimen (N) collapse (mm)

A-SRG1 3995 a –
A-SRG2 3883 a –
A-SRG3 3455 a –
A-CFRCM1 2425 b –
A-CFRCM2 2175 b –
A-CFRCM3 2237 b –
B-SRG1 6368 b –
B-SRG2 6190 b 0.36
B-SRG3 6358 b 0.27
B-SRG4 7516 b –
B-SRG5 6754 b –
B-SRG6 6462 b 0.23
B-CFRCM1 3875 c 0.58
B-CFRCM2 4074 c 1.18
C-SRG1 6166 b 0.72
C-SRG2 5977 b –
C-SRG3 5589 b 0.82
C-SRG4 5438 b 0.42
C-SRG5 5309 b 0.69
C-SRG6 5599 b 0.57

Figure 11. Strain profile along reinforcement at increasing

load level for SRG specimens with setup C (top), for C-FRCM
specimens with setup B (middle), for SRG specimens with
setup B (bottom).

4.2 Strain profile

Figure 9. Debonding strength vs. anchorage length for 1
inches fiber sheet width.
Figure 10. Load applied vs. slip recorded.
The data collected from the strain gauges located on
reinforcement strip have been used to obtain strain
profile (i.e. the strain vs. sensor position relation-
ship) (figure 11): each curve in the diagram refers
to a different load level, form 0.2 to 1 time the ulti-
mate load. It was detected that, when the crack within
the grout reaches the strain gauge position, the sensor
does no more give the fiber strain, since the measure
is disturbed by crack propagation within the matrix;
therefore, the data recorded by these strain gauges
have not been represented in the diagrams. Besides,
from the strain profiles on SRG reinforced masonry
specimens (figure 11a) it can be observed that, in the
elastic range, the strain profile reaches a negligible
value between the gauges located at 167 and 233 mm;
on the other hand, the last gauge located at 295 mm
begins to measure a strain only when approaching to
ultimate load, that is when the crack at the loaded end

813
has occurred; so it can be deduced that the effective
anchorage length of SRG on masonry specimens lays
between 167 and 297 mm. In the case of brick speci-
mens, as shown by the other strain profiles in figure
11, a similar effective length is found for SRG rein-
forcement, while an effective anchorage length not
higher than 150 mm can be estimated for C-FRCM
reinforcement. As expected, a higher stiffness of the
reinforcement is the consequence of a shorter effective
anchorage length.
4.3 τ-slip relation
In specimens B and C reinforced with SRG the data
recorded by the strain gauges have been used to
determine interface τ-slip relationship as follows.
The average homogenized elastic modulus of the
reinforcement embedded within the grout can be
estimated as:
where Eg is the grout matrix elastic modulus; Ag is
the grout matrix section; Ef is the steel fiber elastic
modulus; Af is the steel fiber section and Ar = Ag + Af
is the reinforcement section. According to (Lu et al.
2005, Ferracuti et al. 2007), the average shear stress
between two consecutive strain gauges is estimated
through the following expression:
where εi , xi are respectively the deformation and the
abscissa of the i-th strain gauge and br is the rein-
forcement width. Assuming a perfect bond between
reinforcement and support, in the very last zone of the
anchorage, up to when the corresponding strain-gauge
finds a zero strain, and approximating the strain distri-
bution along the reinforcement as piecewise uniform,
the slipping between two subsequent strain-gauges can
be calculated as:
The resulting τ-slip relations are plotted in
figure 12.
5 CONCLUSIONS
The preliminary results of an experimental investi-
gation addressed to evaluate the behavior of fiber
reinforcement with cement-based matrix on masonry
specimens have been presented. The specimens have
been realized with different support (either brick or
814
Figure 12. τ-slip relationship for SRG specimens.
brick masonry), with different anchorage length, from
55 mm to 300 mm. An inorganic grout with synthetic
polymeric fibers was used as a matrix, with the rein-
forcement made either with steel (SRG) or carbon
(C-FRCM) fibers.
Some differences with respect to polymeric matrix
have been experimentally observed: (i) the collapse
generally occurs within the grout at the interface with
fiber reinforcement, rather than in the brick surface;
(ii) the effective anchorage length is higher thanks to
the shear deformation of the grout. A longer effective
anchorage (between 167 and 297 mm) was detected for
SRG and a lower one (below 150 mm) was estimated
for C-FRCM.
When dealing with historical structures, the use
of grout in stead of polymeric matrix, may be pre-
ferred for conservation purposes; in fact, the moderate
strength of the grout allows to eventually remove the
reinforcement, ensuring the reversibility of the appli-
cation; at the same time, the fact that failure usually
takes place within the grout, rather than in the support,
may be useful to preserve the integrity of masonry
structure.
Coming to the comparison between the two types
of reinforcement, the debonding strength is higher in
SRG than in C-FRCM specimens and the slip is lower,
thanks to the rough surface of the steel cords, which
enhance the stress transfer between the fibers and the
matrix. A lower debonding strength was found for
masonry specimens, rather than for brick specimens,
since the mortar joints act as weakness layers that cause
an anticipation of the collapse. The τ-slip relation-
ships for SRG have been finally estimated from the
experimental data.
REFERENCES
Borri, A. Castori, G. Casadei, P. Ebaugh, S. 2007. Rinforzo di
archi in muratura con materiali compositi innovativi. XII
Convegno L’Ingegneria Sismica in Italia ANIDIS. Pisa:
10–14 June.
Borri, A. Castori, G. Menestò, F. 2007. Interventi con
SRG sulle volte di palazzo Jacobilli in Foligno. XII
Convegno L’Ingegneria Sismica in Italia ANIDIS. Pisa:
10–14 June.
Briccoli Bati, S. Rovero, L. Tonietti, U. 2007. Adesione tra
blocchi in laterizio e rinforzo in CFRP. Workshop: Materi-
ali ed Approcci Innovativi per il Progetto in Zona Sismica
e la Mitigazione della Vulnerabilità delle Strutture. Uni-
versità degli Studi di Salerno, Consorzio Reluis. Fisciano:
12–13 February.
Di Tommaso, A. Focacci, F. Mantegazza, G. Gatti, A. 2007.
FRCM vs FRP composites to strengthen RC beams: a
comparative analysis. FRPRCS-8, University of Patras.
Patras: 16–18 July.
Ferracuti, B. Savoia, M. Mazzotto, C. 2007. Interface law
for FRP-concrete delamination. Composite Structures 80:
523–531.
Huang, X. Birman, V. Nanni, A. Tunis, G. 2005. Properties
and potential for application of Steel Reinforced Poly-
mer and Steel Reinforced Grout composites. Composites
Part B: engineering 36: 73–82.
Lu, X.Z. Teng, J.G. Ye, L.P. Jiang, J.J. 2005. Bond-slip mod-
els for FRP sheets/plates bonded to concrete. Engineering
Structures 27: 920–937.
815
Matana, M. Nanni, A. Dharan, L. Silva P. Tunis, G. 2005.
Bond performance of Steel Reinforced Polymer and
Steel Reinforced Grout. Proceedings of International
Symposium on Bond Behaviour of FRP in Structures.
International Institute for FRP in Construction.
Mazzotti, C. Savoia, M. Ferracuti, B. 2007. The effect of
surface preparation on FRP-concrete bond strength, Work-
shop: Materiali ed Approcci Innovativi per il Progetto in
Zona Sismica e la Mitigazione della Vulnerabilità delle
Strutture. Università degli Studi di Salerno, Consorzio
Reluis. Fisciano: 12–13 February.
Mosallam, A. 2007. Out-of-plane behaviour of unreinforced
brick walls strengthened with FRP composites, Compos-
ites Part B: engineering 38: 557–574.
Neale, K.W. 2000. FRPs for structural rehabilitation: a survey
of recent progress, New materials in construction, Prog.
Struct. Engng Mater. 2: 133–138.
Triantafillou, T.C. 1998. Strengthening of masonry struc-
tures using epoxy-bonded FRP laminates. ASCE Journal
of Composites for Construction 2(2): 96–104.
Valluzzi, M.R. Valdemarca, M. Modena, C. 2001. Behaviour
of brick masonry vaults strengthened by FRP lami-
nates, ASCE Journal of Composites for Construction 5(3):
163–169.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Experimental analysis of rectangular shaped sleeve connectors for

composite timber-steel-concrete floors: Bending tests

L. Calado, J.M. Proença & A. Panão

DECIVIL – Istituto Superior Tècnico de Lisboa, Lisbon, Portugal

F.M. Mazzolani, B. Faggiano & A. Marzo

DIST – University of Naples “Federico II”, Naples, Italy

ABSTRACT: The paper deals with the laboratory tests on timber-steel-concrete composite beams, equipped
with an innovative connection system. Tests are part of a more comprehensive experimental campaign, aiming at
the evaluation and optimization of the connection devices. In this paper both monotonic and cyclic bending tests
on six composite beams are presented. In particular, each specimen consists of a timber beam with rectangular
cross section, connected to a concrete reinforced slab by means of steel collars. Specimens differ for the type
of collar-timber contact and the collars spacing, in order to determine the more effective solution. The system
performance has been evaluated in terms of both maximum force-displacement relationship and accumulated
dissipated energy at failure.

1 INTRODUCTION both strength and stiffness and its consequent opti-

In the context of the restoration of ancient wooden
floors, the wood-concrete composite system repre-
sents a reliable solution for upgrading existing floor
slabs (Gelfi et al., 2002, Mungwa et al. 1999, Giuriani
et al., 1998).
Such an intervention generally consists of a con-
crete thin slab collaborating with the timber beams
by means of connectors, like screws, nails, and studs,
inserted in the beam by drilling the existing members.
In order to avoid any strength reduction of the beams,
which could be certainly caused by the perforation
for the installation of traditional devices, an innova-
tive type of connector has been conceived (Faggiano
et al., 2007). It is made up of steel collars, surround-
ing the timber beam, which are composed by two
or more parts, bolted together at appropriate folded
wings, superior wings themselves or purposely welded
studs having the function of connectors. At the inter-
face between the steel collar and the timber beam, a
rubber layer should guarantee the complete adherence
between elements. Due to the bolt tightening, the steel
sleeves also provide a transversal ringing action on
the beam.
A comprehensive research activity including both
experimental and numerical investigations is ongo-
ing in the framework of the international research
project PROHITECH (Earthquake Protection of His-
torical Buildings by Reversible Mixed Technology).
The purpose is the system evaluation in terms of
mization. In particular, experimental tests are carried
out on full-scale specimens of ancient timber beams
with circular cross sections and new timber beams
with rectangular cross sections, at the DIST (Depart-
ment of Structural Engineering) of the University of
Naples “Federico II” (Prof. F. M. Mazzolani coor-
dinator) and at the DECIVIL (Department of Civil
Engineering and Architecture) of the Superior Tech-
nical Institute in Lisbon (Prof. L. Calado coordinator),
respectively. Further experimental tests on full-scale
composite beams realized by using the studied steel
collars as connection elements are performed at the
DECIVIL, with the cooperation of the DIST. The
whole experimental investigation includes both mono-
tonic and cyclic push-out tests on single connection
devices and both monotonic and cyclic bending tests
on composite wood-concrete beams, equipped with
the sleeve connectors. Numerical analyses on single
device were carried out on several configurations,
obtained by varying the main parameters which affect
the behaviour, such as number of steel parts, both width
and thickness of collars and preloading forces in the
bolts (Faggiano et al., 2007). Results have allowed
identifying the types of collar that supplies the better
behaviour, to be used for composite system.
In this context, the paper presents the second part
of the whole experimental campaign, corresponding to
both monotonic and cyclic bending tests on composite
timber-concrete beams made of new wood and rectan-
gular cross-section, equipped with the selected collars.

817
 Figure 2. Bending tests on timber beam: static scheme [cm].

Figure 1. The typical study composite beam.

Table 1. Compressive strength for concrete

(fc,k ).

Specimen 1, 2, 3 4, 5, 6

fc,k (MPa) 38.9 37.5

fc,k mean = 38.20 MPa Figure 3. Bending tests on timber beam: comparison among
results.

Table 2. Results of bending tests.

First of all the main features of the tested device are
depicted. Therefore both monotonic and cyclic bend- Pmax δmax fb,max E
ing tests are detailed. Results and comparison of the Specimen kN M MPa GPa
analyses are discussed in terms of both force-vertical
1 8.404 0.038 28.56 12.47
displacement and force-slip relationship, in addition 2 11.116 0.041 37.78 15.29
to the accumulated dissipative energy at failure. 3 6.124 0.036 20.81 9.61
4 6.336 0.031 21.53 11.45
5 7.916 0.041 26.90 10.86
2 THE SPECIMENS FEATURES 6 8.832 0.044 30.02 11.29

The six tested composite beams are timber beams con-

nected to a concrete reinforced slab by means of eight Table 3. Mechanical properties of timber.
steel devices (Fig. 1).
Beams have rectangular cross sections 0.16 × E0,mean E0,05 E90,mean Gmean f0,k
Property GPa GPa GPa GPa MPa
0.20× 4.40 m3 and they are made of pine wood. The
concrete slab is 0.70 × 0.04 × 4.40 m3 sized and it is 
n
Ei
reinforced by steel bars class A400NR, with a 5 mm
Equation i=1n 0.67E0,mean E0,mean /30 E0,mean /16 f05 ks kv
diameter and 0.10 m spaced. Value 11.82 7.92 0.394 0.74 24.84
Preliminary tests on the component materials were
carried out in order to determine their mechanical
properties. With regard to concrete, compression tests
forces and displacements are indicated for each spec-
were performed. Since the cast was made in two differ-
imen, together with both the bending strength (fb,max )
ent occasions, two values of strength were measured,
calculated with reference to the maximum applied
one for beams n. 1, 2, 3 and the other for beams n. 4,
Force (Pmax ) and the Young modulus (E) at the origin.
5, 6. They are indicated in Table 1. According to these
The main mechanical properties of used timber,
values, concrete is classified as class C30/37.
calculated according to the UNI EN 384 (1997)
With regard to wood, bending tests were performed
provisions, are indicated in Table 3, where:
on small timber beams, 0.08 × 0.08 × 1.76 m3 .They
were subjected to concentrated loads applied with – n is the number of specimens;
a 0.05 kN/s velocity, according to the static scheme – Ei is the experimental Young modulus;
plotted in Figure 2. A transducer measures the mid- – E0,mean is the longitudinal Young modulus;
span deflection (δ). Results are shown in Figure 3 in – E0,05 is the 5-percentile modulus;
terms of Force (P) – displacement (δ) curve for each – E90,mean is the transversal modulus;
specimen. Moreover, in Table 2, maximum values of – Gmean is the shear modulus;

818
 Table 4. Characteristics of tested specimens.

Timber- Collar Collar-

Specimen steel beam Load
N. contact N. Spacing connection history

1 rubber 8 constant no monotonic

2 rubber 8 variable no monotonic
rough
Figure 4. Specimen detail: a) Lateral view of the steel collar; 3 steel 8 constant yes monotonic
b) Upper view of the steel collar; c) Whole cross section. surface
4 rubber 8 constant no monotonic
5 rubber 8 variable no monotonic
rough
– f0,k is the strength characteristic value; 6 steel 8 variable no cyclic
surface
– ks is the corrective coefficient related to the number
of specimens, which is obtained from a codified
abacus;
– kv is a coefficient depending on the type of classi-
fication, visual or mechanical, which assumes the
values 1.0 and 1.12 respectively; in the present case
the specimens are classified according to a visual
classification.
According to these values the used pine timber is
classified as C24 (UNI EN 338, 1997).
The type of connection device was chosen on the
base of a previous experimental campaign developed
at the DECIVIL of the Superior Technical Institute in
Lisbon (Faggiano et al., 2007).
The eight tested collars consist of two U-shaped
parts, 100 × 160 mm2 , 60 mm wide and 5 mm thick,
provided with triangular stiffening plates, welded at
the wings with an inclination of 60◦ (Fig. 4a). The two
parts are jointed by four bolts M8, Class 8.8, two on
each side, being the bolt tightening equal to 25 kN.
A connector type KOCO M12 is welded at the mid
upper side of the device (Fig. 4b). The composite cross
section is shown in Figure 4c. Figure 5. Lateral views (mm): a) Specimens n. 1, 3, 4; b)
Two different types of contact between timber beam Specimens n. 2, 5, 6.
and steel collar were tested: for four specimens a rub-
ber layer type 60 HRD is interposed at the interface, 3 APPLIED TESTING PROCEDURE AND
while for two specimens direct timber-steel contact TESTING SET-UP
is provided, the internal surface of the collar being
treated for improving the roughness. In the latter case, 3.1 Load histories
only for one specimen, sixteen bolts connect the collars
Bending tests were performed in both monotonic and
to the timber beam. The device provided with rubber
cyclic conditions, up to failure. The applied force-time
was conceived with the aim of allowing the complete
histories are plotted in Figures 6 and 7.
reversibility of the system. The rubber layers are 5 mm
In particular five specimens (n. 1 to 5) were sub-
thick and 60 mm wide. They are both glued to the col-
jected to a monotonic load history and only specimen
lar and laterally confined by two 5 × 3 mm2 edges,
n. 6 to a cyclic load history, considering oscillations
welded at both sides of the collar, in order to realize
around a reference load equal to 30 kN.
both a chemical and mechanical barrier to the rubber
slipping.
3.2 Testing apparatus
For all specimens eight connection devices are used;
they are placed at a variable spacing according to shear The testing apparatus consists of a steel beam, which
distribution. applies the two punctual loads to specimens, accord-
The characteristics of the six specimens are indi- ing to the load scheme, by means of a load actuator,
cated in Table 4 and shown in Figure 5. opportunely contrasted by an ad hoc frame (Fig. 8).

819
Figure 6. Force-time histories for specimens n. 1–5.
Figure 7. Force-time histories for specimen n. 6.
The specimens behaviour was monitored by six-
teen electric displacement transducers (LVDT). Two
PC were used for data acquisition and recording.
The location of transducers is dictated by the
required displacement data (Fig. 9): TR3 to TR5 mon-
itor the beam deflections (vb ); TR16 monitors the
mid-span vertical split between the timber beam and
the concrete slab (vbs ); TR6, TR8, TR10, TR12, TR14
and TR15 monitor the relative slipping between the
timber beam and the concrete slab (δbs ); TR7, TR9,
TR11 and TR13 monitor the relative slipping between
the of the timber beam and the collars (δbc ). TR1 and
TR2 are applied at the load application points.
4 EXPERIMENTAL RESULTS
The experimental results are presented in terms of load
vs absolute or relative displacement curves, according
to the aforesaid displacement parameters, measured by
a number of 3 to 16 transducers.
The sign convention used for the collar-beam slip-
ping is given in Figure 10: such as, if the top of a device
moves to the right, the transducer at its right will have
a positive displacement.
For the sake of synthesis, all diagrams related to
specimen n. 1 are plotted in Figures 11, 12, 13; in
particular in Figure 13 each diagram shows the results
820
Figure 8. Testing apparatus: a) Global view; b) Steel beam.
measured by the couple of transducers applied at each
collar location.
Test results related to all other specimens are
presented later on for comparison purposes.
All tests show at collapse cracking of the r.c. slab
and failure of the beam (Fig. 14). The ultimate load
corresponds to the attainment of the timber beam
strength.
5 COMPARISON AMONG RESULTS
5.1 General
Comparison among experimental results are presented
in terms of load vs both absolute and relative displace-
ment curves (Fig. 15).
Aim is the evaluation of the effects of the type of
beam-collar contact (rubber or rough steel), the spac-
ing between devices (constant or variable) and the
additional bolted connection between steel collar and
timber beam.

Figure 9. Transducers location.
Figure 10. Sign convention for collar-beam slipping.
Data recorded by transducers TR3, TR6 and TR7,
have been compared through the envelope curves: TR3
for the beam mid-span deflection (Fig. 15a); TR6
and TR7 for the relative beam-slab and beam-collar
slipping, respectively (Figs. 15b, c).
From a first analysis of results (Figs. 15b, c), it is
apparent that the relative displacements between the
beam-slab and the beam-collar are almost negligible.
This demonstrates the efficiency of the collar as a
connection system between the beam and the slab.
821
Figure 11. Specimen 1: a) load vs beam deflection curves;
b) load vs beam-slab split curves.
Figure 12. Specimen 1: load vs beam-slab slipping curves.
In addition, all specimens are compared in terms
of both strength and stiffness, calculated on the basis
of the load-mid span deflection curves and the rel-
ative accumulated dissipated energy up to failure. In
particular both strength and stiffness characteristic val-
ues have been calculated by using the envelope curves
of each specimen, according to the procedure shown
in Figure 16: both failure (Fu ) and yield loads (Fy ),
together with the corresponding displacements (δu ,
δy ), and both elastic (E1 ) and post elastic (E2 ) stiff-
ness, have been evaluated. Besides, the accumulated
dissipated energy up to failure (Er ) and at a load value
equal to 40 kN (E40 ) have been calculated according
to the equation in Figure 16, being F the applied load,
δ the mid-span deflection. The obtained values of the
parameters are shown in Table 5.
Figure 13. Specimen 1: load vs relative displacement
curves.

5.2 Effect of the connector spacing

For evaluating the connector spacing effect, the com-
parison among homogeneous specimens (SP) n. 1, 2, 4,
5 has been performed on the basis of the corresponding
load vs vertical displacements (Fig. 17).
It is apparent that a variable spacing of the connec-
tion devices improves the behaviour of the composite Figure 14. Failure modes.

822
 Table 5. Values of the strength and stiffness parameters.

Fu δu Fy δy E1 E2 N. Er E40
Spec. kN mm kN mm GPa GPa cycles kNm kNm

1 49 68 29 29 1,01 0,51 5 2672 1606

2 88 93 64 45 1,42 0,50 9 7114 843
3 82 73 44 21 2,11 0,74 8 5505 775
4 55 93 55 40 1,38 0,00 8 4447 1482
5 43 52 22 18 1,25 0,61 6 1664 1550
6 68 77 32 10 3,03 0,54 27 7430 856

Figure 17. Effect of connector spacing.

the timber beam (Fig. 14). In fact knots involve poorer

mechanical characteristics of the composite system,
because they create a less resistant zone in the most
stressed section. In addition, the accumulated dissipa-
tive energy up to failure related to the SP2 is almost
three times the one related to the SP1 (Table 5).

5.3 Effect of the type of beam-collar contact

Figure 15. Comparison among results: a) TR3; b) TR6;
c) TR7.
Figure 16. Strength and stiffness parameters evaluation.
beam as respect to the constant spacing (SP1 or SP4 vs
SP2): this is appraisable as a mean increment of about
70% and 20% in terms of strength and stiffness, respec-
tively. SP5 has a worst behaviour, as respect to SP2,
due to the presence of knots located at the mid-span of
For evaluating the effect of different type of collar-
beam contact, the comparison among the specimens
n. 1, 3, 6 has been performed on the basis of the cor-
responding load vs vertical displacements (Fig. 18).
It is apparent that the direct contact between the steel
rough surface and the timber beam (SP6) improves
the behaviour of the composite beam as respect to the
contact between the surfaces through the rubber layer
(SP1): this is appraisable as an increment of about 35%
in terms of strength and three times in terms of stiff-
ness. In addition, the accumulated dissipative energy
up to failure related to SP6 is almost three times the
one related to SP1 (Table 5). It is worth noticing that
in the first case collars provide some damage at the
beam contact surface.
The connection of the collar with internal rough
surface to the beam by means of a mechanical device
provides an increase of both strength (about 20%) and
stiffness (about 37%), the latter only above a load level
equal to 33 kN (SP3 vs SP6). However in this case
the collar induces a significant damage of the beam

823

Figure 18. Effect of the type of collar-beam contact.
contact surface. The accumulated dissipated energy
up to failure for SP6 is about 35% larger than for SP3.
In order to evidence the increase of both strength
and stiffness of the composite system as respect to the
simple timber beam, reference has been made to the
second moment of area. For the composite section this
has been related to a measured elastic deflection equal
to 5 mm, which corresponds to a 10 kN applied force
(Fig. 15a), according to the static scheme plotted in
Figure 2, for all the specimens. As a result, for the com-
posite and the timber beams the second moments of
area are equal to 7415 cm4 and 3413 cm4 respectively,
the increment factor being about 2.2.
6 CONCLUSIVE REMARKS
The present paper illustrates both monotonic and
cyclic bending tests carried out on timber-steel-
concrete composite beam realized by using a com-
pletely reversible connection device. The base system
is a steel collar, located astride the timber beam with or
without the interposition of a rubber layer. Collars are
made in two parts bolted together at ad hoc wings. The
function of connector is held by a steel stud welded at
the superior side of the collar. The final configuration
of the adopted collars was obtained from a previous
experimental campaign on a single connection system.
Tests illustrated concern the composite timber-
steel-concrete system with a rectangular shaped beam
cross section. The timber beam is made of new pine
wood, therefore no irregularities at the surfaces were
present, except for knots. Several configurations of
the composite system have been analysed by varying
both collar spacing and type of collar-beam contact.
Finally, the set up composite systems have undergone
both monotonic and cyclic bending tests.
The comparison among experimental results has
been performed in terms of force vs absolute and
relative displacements. It has been evidenced that:
– relative displacements between the beam-slab
and the beam-collar are almost negligible, what
824
demonstrates the efficiency of the collar as a con-
nection system;
– variable spacing of the collars (according to shear
distribution) provides a better behaviour as respect
to the constant one;
– direct contact between the steel rough surface and
the timber beam improves the behaviour of the com-
posite beam as respect to the contact between the
surfaces through the rubber layer;
– the connection of the collar with internal rough sur-
face to the beam by means of a mechanical device
provides a small increase of both strength and stiff-
ness; besides the collar induces a significant damage
of the beam contact surface, which does not justify
the choice of this solution.
– the composite system by collars provides an incre-
ment of strength and stiffness of more than two times
as respect to the timber beam.
Further investigations, both experimental and
numerical ones, are ongoing, in order to achieve
the system optimization and its structural character-
ization, which would allow to set up a analytical
formulation for design purposes.
ACKNOWLEDGEMENT
The research activity is developed within the PRO-
HITECH International project “Earthquake Protection
of Historical Buildings by Reversible Mixed Technolo-
gies”, coordinated by F.M. Mazzolani.
REFERENCES
Faggiano B., Marzo A., Formisano A., Mazzolani F.M. 2007.
Innovative steel connections for a composite wood –
steel – concrete floor: a numerical investigation. Com-
puter & Structures (Elsevier – K. J. Bathe and B.H.V.
Topping editors, ISSN 0045-7949), submitted.
Faggiano B., Marzo A., Mazzolani F.M. and Calado L. 2007.
Experimental analysis of rectangular shaped sleeve con-
nectors for composite wood-steel-concrete floors: push-
out tests. Proc. of the 9th International Conference
“Modern Building Materials, Structures andTechniques”,
Vilnius, Lithuania, 16–18 May.
Giurani, E. 2002. Timber beam floors strenghtened by means
of a collaboratine concrete slab: design criteria. Edilizia
(4): 32–40 (in Italian).
Gelfi, P.; Giurani, E. and Marini, A. 1998. Behaviour of
stud shear connectors in the composed timber-concrete
beams: theoretical modelling and experimental compar-
isons. Proc. of the 3rd Italian Workshop on the Composed
Structures, Ancona, Italy, 29–30 October (in Italian).
Mungwa, M. S., Jullien, J. F., Foudjet, A. and Hentges, G.
1999. Experimental study of a composite wood-concrete
beam with the INSA-Hilti new flexible shear connector.
Construction and Building Materials (13): 371–382.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Shaking table testing of a typical Mexican colonial temple

M. Chávez & R. Meli

Instituto de Ingeniería, Universidad Nacional Autónoma de México, México D.F.

ABSTRACT: To study the dynamic behavior of typical temples of the 17th and 18th Centuries stone masonry
construction in Mexico, a shaking table test program is being performed on a 1:8 scale model built with the
same materials and techniques than its prototype. Results of a first stage of the research in which the model was
submitted to horizontal and vertical motions of increasing intensities are presented in this paper. The test was
stopped at an intensity corresponding to a base shear force of about 40% of the total weight, causing a damage
that was considered as possible to be repaired. Vertical component of the motion showed to significantly affect
the response and to increase the damage of the model. Damage patterns were similar to those observed in actual
temples. Main features of the measured response are compared to those computed from a linear elastic, finite
element model. The degree of damage and the variation in the amplification of the response are correlated to
the variation of the fundamental vibration period of the model.

1 INTRODUCTION of damage that was considered feasible to be repaired

A shaking table test program is being performed at
the Institute of Engineering of UNAM pursuing the
main objective of providing basic information for cal-
ibration of analytical models presently available for
determination of the seismic response of ancient stone
masonry structures. Additionally, it aims at a bet-
ter understanding of the seismic response and of the
modes of failure of these buildings, as well as to eval-
uate the efficacy of typical techniques commonly used
for reducing their vulnerability to severe earthquakes.
A typical colonial temple was chosen as a proto-
type. It has one nave covered by a barrel vault; two
rather slender bell towers, and relatively light buttress-
ing (Morales 2004). Walls and buttresses are built with
a heterogeneous masonry constituted by stones agglu-
tinated by a lime-sand mortar, while vaults and bell
towers are built with bricks. Due to the size of the
shaking table, and to the maximum weight it can carry,
the model ought to be scaled down at a 1:8 geometric
scale, for which the same constituting materials and
construction techniques of the prototype could still
be used.
Summarized in this paper are the results of the first
stage of this program, consisting of the design of the
experiment, the construction of the model, and its test-
ing under increasing levels of seismic intensities, first
for the effect of the horizontal base motions in the
weak direction of the model, and then for the combined
effect of horizontal and vertical accelerations. Maxi-
mum intensity applied in this stage corresponded to
the onset of the critical modes of failure and to a level
with traditional rehabilitation techniques.
A detailed linear analysis was carried out of a finite
element model of the experimental structure, to ascer-
tain main characteristics of its dynamic response and to
determine ground motions to be applied in the tests. In
this stage of the research, comparison of measured and
computed response was made only for a linear behav-
ior, considering that it would provide a good estimation
of the seismic response and the dynamic characteris-
tics for moderate ground motions (Meli & Peña 2005).
In further stages of the research, non linear models will
be studied, aiming at reproducing measured response
for higher intensity levels (Lourenco 2002).
The seismic records selected for the horizontal and
vertical accelerations, corresponded to actual seismic
events, modified in their time scale in the order to
enhance their effects in the model.
2 DESIGN OF THE MODEL
Prototype temple is typical of central Mexico; being
this a region where strong earthquakes are not very
frequent, colonial architecture was not as robust as in
other more seismically active areas; the temple is tall
and with light buttressing (Fig. 1a). Temples of this
kind are frequently damaged by earthquakes, but their
collapse is uncommon.
In order to limit the scale factor of the model and
to ease its construction, the prototype was simplified
by eliminating two bays, including the transept and its
dome, but preserving all other characteristics (Fig 1b).

825

Figure 1. Prototype and model temples.
The model to be tested in the shaking table was reduced
at a scale 1:8, to fit the 4 by 4 m shaking table of the
Institute of Engineering at UNAM (Fig 1c), which is
biaxial (horizontal and vertical); it can support a load
up to 20 t, and can generate maximum accelerations
of 1.2 g horizontally and 2.0 g vertically; displace-
ment ranges are ±0.15 m horizontally and ±0.075 m
vertically. The model was built with the same mate-
rials as the prototype in order to reproduce as close
826
a) Plan view
b) Front view
Figure 2. Dimensions of the model (m).
as possible its mechanical properties and the same
modes of failure. It was not considered feasible to com-
ply with all requirements of the dimensional analysis
(Tomazevic 1992) without altering the dynamic behav-
ior; therefore, it was accepted that the model should
be considered as a miniaturized prototype; extrapola-
tion of its results to the prototype should be essentially
made in a qualitative way. The time scale of the seis-
mic input was reduced by the scale factor and the
acceleration was increased by this same factor; in this
fashion the dynamic stresses are maintained the same
than for the prototype, but the stresses due to the self
weight of the building are reduced by the scale factor.
Dimensions of the model are shown in Fig. 2.
Figure 3. Construction of the model.
3 CONSTRUCTION OF THE MODEL
Basic construction material of these historical build-
ings is a masonry conglomerate, constituted by stones
of different sizes, agglutinated by lime-sand mor-
tar. Light weight stones were preferred because of
ease of transportation and carving; this heterogeneous
masonry constitutes a kind of low-strength concrete.
It is lighter than normal stone masonry, and has a
greater tensile strength than brick masonry, due mainly
to the absence of weak planes constituted by the mor-
tar layers. A sandstone very commonly found in the
temples under study was used for the construction of
the model. Mortar had a lime-sand-water volumetric
ratio of 1:3:1. The vault and the bell towers were built
of clay roof tiles, which were cut to the required size.
Before undertaking the construction of the model
itself, several trials were made on small parts of build-
ing components, in order to train the labor to working
with small size stones and structural members, and
especially to enable them to obtain a distribution of
stones and amount of mortar that were close to those
found in the prototype. The volumetric ratio of mortar
to stones was kept close to 0.35.
827
Mechanical properties obtained from small walls
were as follows: for stone masonry, compressive
strength was 1.3 MPa and Young modulus 270 MPa;
for brick masonry, compressive strength was 3.9 MPa
and Young modulus 476 Mpa (Chávez 2005). The
model was built on a stiff steel frame, designed to
enable moving the model in and out the shaking
table, and to fix it to the table. Some aspects of the
construction are shown in Fig. 3.
4 MODEL INSTRUMENTATION AND TESTING
PROGRAM
The model was profusely instrumented to monitor
its global response and also local vibrations of some
critical components, as the two bell towers. Twenty
three accelerometers were placed on the table, on the
basis and on the roof of the model, as well as at
different heights of the towers. Vertical acceleration
was measured at the vault center. Six displacement
transducers were placed on stiff frames inside nave,
to measure opening and closing of the vault. Abso-
lute lateral displacements of the model were measured
 8 16

6 14
12
Acceleration (m/s2)
4

Amplitude (m/s2)
2 10

0 8
9.281775
0 5 10 15 20 25 30 35 40 45 50 55 6
-2
4
-4
2
-6
0
-8
0 0.2 0.4 0.6 0.8 1
Time (s)
Period (s)

a) Horizontal acceleration time history b) Acceleration response spectrum of horizontal input

6 16
5 14
4

Amplitude (m/s2)
12
Acceleration (m/s2)

3
2 10
1 8
9.281775
0
-1 0 6
5 10 15 20 25 30 35 40 45 50 55
-2 4
-3 2
-4
-5 0

Time (s) 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Period (s)

c) Vertical acceleration time history d) Response spectrum of vertical input

Figure 4. Basic seismic input.

by fixing displacement transducers to two stiff steel state (0.088 s), but also for the longer periods that
frames placed outside the shaking table. are expected when the structure be damaged by the
The total weight of the model was 11.96 t; an addi- earthquake.
tional weight of 2 t was evenly distributed on the vault Response to ambient vibration and to a white noise
in order to increase the inertia forces on the main shaking with small amplitude, was measured before
body of the model, as well as the vertical compressive starting tests with the selected ground motions. Initial
stresses on the walls, thus producing a slight correc- natural periods of vibration were derived from these
tion of the difference with respect to that required by tests as 0.088 s.
the dimensional analysis. When initiating the seismic tests, a malfunction-
Acceleration time-histories to be applied to the table ing of the software controlling the actuators gave rise
in the horizontal and vertical directions were derived to a strong, abrupt shaking of the table that produced
from a strong motion record obtained in 1985 near the significant damage to parts of the building: diagonal
epicenter of a large magnitude earthquake (Ms 8.1). cracking of the facade, flexural cracking on the right
The seismic records were modified in their time scale longitudinal wall and buttresses, and some disloca-
in the order to enhance their effects in the model. tion of the vault. The model was carefully repaired,
The horizontal motion was applied in the transverse first restituting its initial geometry by tensioning some
direction of the model. Their acceleration scale was external steel tie rods, and then injecting cracks with a
increased to almost reach the capacity of the table. lime grout. Natural period measured at the end of the
The time scale was reduced to produce highest spec- process were close to those of the undamaged model
tral amplitudes in the neighborhood of the fundamental (0.08 vs. 0.088 s).
period of vibration of the structure. Some modifica-
tions were made to the initial part of the vertical record 5 BEHAVIOR OF THE MODEL DURING THE
in order its maximum acceleration occur at the same SEISMIC TESTS
time than that of the horizontal record. The resulting
time histories and their acceleration response spec- Intensity of the selected seismic input was scaled down
tra for 5% damping are shown in Fig. 4. As it can to different fractions of the reference records, start-
be appreciated, the combination of acceleration his- ing with 5% and rising up to 60%. For each level
tories is very severe for the structure under study. of intensity, first only the horizontal component was
The maximum acceleration is very high in both direc- applied, and then horizontal and vertical motions were
tions; the duration of the strong motion is very large simultaneously induced.
(about 50 s); their frequency content gives rise to The description of the set of tests performed and the
large spectral amplitudes not only for the fundamental maximum acceleration applied at the shaking table for
period of the structure corresponding to its undamaged each intensity level are shown in Table 1.

828
Table 1. Test program and main results.

Maximum base acceleration

Seismic Horizontal
Test intensity Applied
No. % components m/s2 Vertical Damage description

1 5 Horizontal 0.43 — None

2 5 Horiz and vert 0.46 0.36 None
3 10 Horizontal 0.81 — None
4 10 Horiz and vert 0.77 0.73 None
5 20 Horizontal 1.4 — Small longitudinal cracks in the vault
6 40 Horizontal 3.04 — Failure at bottom of the piers of the upper body of the
bell tower.
7 40 Horiz and vert 3.23 2.6 Cracking of the left longitudinal wall, especially near
the façade. Increase of cracking of the vault.
8 60 Horizontal 4.37 — Collapse of upper part of both bell towers; upper part
of the left tower separates from the facade; increase
of wall cracking, especially in left side; descent
(5 mm) of the left frontal part of de vault.
9 60 Horiz and vert 4.25 3.57 Collapse of the lower part of the bell tower: separation
of the left tower and diagonal cracking of the façade.

– the lateral drift of the two upper bodies of the bell

towers, producing flexural hinging at the base of
their columns, and finally their overturning;
– the separation of the right tower from the facade,
as a result of its lateral vibration.
The damage was more intense on the left side of the
temple, presumably because the right side had been
injected with grout after the previously mentioned
accidental damage suffered by the model.

6 MEASURED RESPONSE

Maximum acceleration recorded at the base, and at the

Figure 5. Damage patterns for 60% of the reference motion.
The main modes of failure were:
– the lateral drift of the nave, producing flexural
cracks at the base of the longitudinal wall and their
buttresses, as well as longitudinal cracking of the
right side of vault at its groin;
– the shear failure of the façade favored by the
window and door openings;
vault and towers tops, for different intensities of the
applied motion, is shown in Table 2. As it can be seen,
for 10% of the maximum intensity, the acceleration at
the table level is amplified only 14% at the top of the
vault, indicating that the temple as a whole vibrates
almost as a rigid body; on the other hand, the ampli-
fication at the bell tower is very important, reaching
186% at the top of the tower, due to an appendix-like
response of these members. For increasing intensities,
the structural damage gives rise to a decreasing stiff-
ness of the structure and to an increasing dissipation
of energy, which is equivalent to an additional damp-
ing; the two factors have counteracting effects on the
amplification; thus in some points response increases
as amplification increases, while it decreases in others.
The maximum shear force induced at the base of the
structure, as derived from the recorded acceleration for
intensity level of 60%, was 4 t, corresponding to 33%
of total weight of the model; from the results of the
finite element model this base shear ratio was 0.35.

829
Table 2. Maximum measured accelerations for different Table 3. Maximum measured displacements for different
intensities of applied horizontal motion. intensities (mm).

Vault keystone Tower

Bell tower Bell tower

Façade Center nave base top

A E A E A E A E
Intensity
% (mm)

10 0.24 0.27 0.33 0.57 0.41 0.34 0.84 1.07

20 0.48 0.60 0.66 0.83 0.81 0.64 1.67 1.62
Base Vault Tower Amplification 40 0.97 2.24 1.32 1.97 1.63 2.20 3.34 4.65
relative to the basis
Ac1 Ac2 Ac3 Ac4 A: Analytical; E: Experimental.
Intensity Ac2 Ac3 Ac4
% (m/s2 ) (Ac1) (Ac1) (Ac1) Table 4. Effect of vertical component of table motion on the
displacements at the vault.
10 0.87 0.99 1.51 2.49 1.14 1.74 2.86
20 1.51 1.94 2.62 3.81 1.28 1.73 2.52 Horizontal-vertical
40 3.45 4.53 5.12 7.75 1.31 1.49 2.25 Horizontal component component

Spring Spring Spring Spring

Left Key Right Left Key Right
Maximum displacements measured at some rele- H V H H V H
vant points are shown in Table 3 for different intensities Intensity
of the imposed motion. As it can be appreciated, the % (mm)
displacements of the main body of the temple are
extremely small; they exceed 1 mm, only for the max- 5 0.092 0.026 0.084 0.102 0.031 0.068
imum applied intensity for which the damage was 10 0.165 0.039 0.123 0.165 0.036 0.135
20 0.420 0.079 0.288 – – –
rather severe; they were significantly amplified at the
40 1.506 0.322 0.970 1.755 0.401 1.097
top of the towers, though never exceeding a few mil- 60 4.796 0.961 2.890 5.761 1.292 3.143
limeters. It is interesting to notice that the horizontal
displacement at the center of the nave was about twice H: Horizontal Displacement; V: Vertical Displacement.
those measured at each end of the nave, thus showing
that the roof does not constitute a stiff diaphragm, but
inducing an opening and closing effect that increased
it bends horizontally producing greater displacements
lateral displacements in one side of the supporting
in its center part, which has lower lateral stiffness than
walls. Table 4 shows as the maximum vertical dis-
the two ends of the nave.
placement at the keystone of the vault, as well as the
Also shown in Table 3 are displacements obtained,
maximum lateral displacements at the springs of the
for each intensity level, from the linear elastic analysis
vault, are greater for the simultaneous application of
of finite element model reproducing the characteristics
the two components.
of the experimental model. For the two first intervals of
intensity for which the damage was not significant, a
reasonable similarity is found among computed and 7 VARIATION OF THE FUNDAMENTAL
measured displacements; for the maximum applied VIBRATION PERIOD OF THE TEMPLE
intensity, the softening of the structure by the dam-
age it suffered, gave rise to measured displacements Ratios of the spectral amplitudes of the motion mea-
largely exceeding those computed considering linear sured at the roof, related to the one measured at the
behavior. table, allow an accurate determination of the funda-
The addition of the vertical component of the mental modal frequencies of the structure, and with
motion produced a more severe condition than when less accuracy, also of the frequencies for higher modes.
only the horizontal component was applied. The main Shown in Fig. 6 are Fourier spectral ratios between the
differences were noticed in the bell towers, where the motion at the roof and at the base, for different motion
reduction of the net compressive forces on the columns intensities, whose peaks reveal the fundamental period
favored their overturning, and most importantly on of vibration.
the vault, where the vertical inertia forces produced The fundamental period of the structure, as obtained
an important variation in their thrust on the supports from the records of the response to a low intensity

830
9
W. noise Cal. H 20% 0.13
f = 9.03 Hz f = 11.15 Hz
8
Cal. H 10% 0.12
Cal. H 40% f = 12.06 Hz
f = 9.56 Hz W. noise
0.11

Period [s]
f = 12.51 Hz
7 Cal. H 5%
Cal. H 60% f = 12.46 Hz 0.1
f = 7.76 Hz
6 0.09
5 0.08
0.07
4
0 5 10 15 20 25 30 35 40 45 50 55 60 65
3
Seismic intensity [%]
2
a) Horizontal
1 0.09

0 0.08
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Frequency [Hz] 0.07

Period [s]
0.06
Figure 6. Ratios of Fourier’s spectral amplitudes measured 0.05
at the top of the vault to those at the base. 0.04
0.03
0 5 10 15 20 25 30 35 40 45 50 55 60 65
Table 5. Vibration periods derived from response to differ-
Seismic intensity [%]
ent intensities of the motion.
b) Vertical
Lateral vibration Vertical
vibration Figure 7. Variation of vibration period with the intensity of
H HV HV the motion.
Condition Period
Test of the Experimental Analytical
no. Description model (s)
7
∗ 6
Analytical 0.097 0.06
∗∗ 5
Amplitude

1 White noise Initial 0.088

4
2 Ambient Vibr. Acc 0.14
3
damage
3 Ambient Vibr. Repaired 0.08 0.04 2
4 White noise 0.08 — 0.04 1
5 5% 0.08 0.09 0.04 0
6 10% 0.08 0.09 0.04 0 5 10 15 20 25
7 20% 0.09 — Frequency Hz
8 40% 0.10 0.11 0.06
9 60% 0.13 0.14 0.08
Figure 8. Comparison of analytical and experimental trans-
10 White noise 0.11 0.07
fer functions for 40% of the maximum intensity.
∗
From modal analysis of FEM model. until the intensity reached 20% of the reference inten-
∗∗
Measured before accidental damage caused by failure of
sity. Afterwards, it increased at a growing rate; for
the control system.
the maximum applied intensity the horizontal period
increased by 62.5% and vertical one by 100%, thus
white noise, applied before starting the formal test, was
indicating that the stiffness had decreased to 38% and
0.08 s; the decrease with respect to the period of 0.088
25% of its initial value, respectively.
measured for the original model, before its acciden-
tal damage and subsequent repair, are to be attributed
to some stiffening of the structure due to the mortar 8 EVALUTION OF DAMPING
injection of some of its parts.
Periods of vibration determined from the records Damping involved in the measured response of the
of different tests are shown in Table 5; both, the fun- structure was derived from the shape of the ratio of
damental period due to the lateral drift of the nave, spectral amplitudes between the acceleration recorded
and the period corresponding to the vertical vibration the roof and the basement, around the peak at the
mode of the vault are reported in the table; the trend of fundamental frequency of vibration of the structure.
their increase with the intensity of the motion can be The widely used procedure proposed by Rinawi and
better appreciated in Fig. 7. The period increase due Clough (1992) was followed, by which the theoretical
to the unintentional damage was very large, indicating transfer function of a single degree of freedom system
severe stiffness degradation; on the other hand, when is fitted to the experimental shape (Fig. 8). Computed
this damage was repaired stiffness increased exceeding damping coefficients shown in Table 6 for different
its initial value. The initial period remained constant motion intensities; they are rather high compared to

831
 Table 6. Damping derived from response to
different intensities of the motion.
Intensity
[%] Damping
10 0.07
20 0.08
40 0.11
60 0.14
those commonly found in modern structures, indicat-
ing a large energy dissipation, which for low intensity
motion could be attributed to plastic behavior of the
mortar, and for higher intensity to sliding along the
profuse cracking of masonry members.
9 CONCLUSIONS
The model and its testing program are providing valu-
able information about the seismic response of stone
masonry construction in general, and of this kind of
colonial temple in particular. Despite of the damage
suffered by the model before its testing, the rehabili-
tation returned the model near to its initial conditions,
though with some stiffness increase in the zones
injected with grout.
The motion applied to the shaking table was par-
ticularly severe for this type of building, because of
its high spectral ordinates in a wide range of periods,
corresponding to the initial and damaged states of the
structure. Maximum induced accelerations reached
4.53 m/s2 at the roof and 7.75 m/s2 at the top of the
towers, and the maximum base shear coefficient was
almost 40%.
The amplification of the acceleration at the roof
was rather small, not exceeding 40%, nevertheless, the
acceleration at the top of bell towers was almost three
times the acceleration of the base for low intensity base
motions.
Damage patterns induced in the model corre-
sponded to those observed in actual temples hit by
intense earthquakes. Most vulnerable parts of the
structure showed to be: the bell fry suffering great
amplifications due to its flexibility; the main body
of the towers which is prone to separating from the
facade and failing by shear; the facade itself is prone
to in plane shear failure; the central part of the nave
whose lateral drift produces bending damage in walls
and buttresses, as well as dislocation of the vault. It
must be taken into account that other modes of failure
due to the ground motion in the longitudinal direction
of the temple were not included in the study.
The effect of the vertical component of the motion
showed to be rather severe. The main differences with
the case where only the horizontal component was
applied, were noticed in the bell towers where the
reduction of the net compressive forces on the columns
832
favored their overturning, and most importantly on the
vault, where the vertical inertia forces produced an
important vertical vibration causing variation in thrust
on the supports, and an opening and closing of the vault
that increased lateral displacements in one side of the
supporting walls thus affecting their stability. It can be
assumed that buildings of this kind would be particu-
larly vulnerable to earthquakes with nearby epicenters,
which show high vertical accelerations of the ground.
A good correlation was found among the amount
of damage and the corresponding period of vibration
of the structure, which constitute an effective index of
residual capacity of the structure. The gradual decrease
of frequency reflected the loss of model stiffness that
reached a factor of about two, for the maximum applied
intensity.
Lateral displacements of the model did not exceed
few millimeters; it can be assumed that due to their
high stiffness and to their low period of vibration,
displacements of structures of this kind remain quite
small until they are near to failure.
Response computed by finite element linear anal-
yses is similar to that measured in the tests, only for
rather low levels of shaking intensity. For predicting
the response in near to failure conditions, use of non
linear models is necessary.
Damage produced in this phase of the is consid-
ered as repairable; then, the next stages of the research
program will aim at evaluating the effectiveness of
some commonly used rehabilitation techniques, and
the ability of non linear methods of analysis to predict
the seismic capacity and general vulnerability of this
kind of structure.
REFERENCES
Chávez M. 2005, Estudio experimental de las propiedades
mecánicas de mamposterías de piedra natural (Experi-
mental study on mechanical properties of stone masonry),
Master Dg Thesis, UNAM, México.
Lourenco, P. B., 2002. Computations on historic masonry
structures. Progress in Structural Engineering and Mate-
rials 2002, John Wiley & Sons. Ltd.
Meli R. & Peña F. 2005, On elastic models for evaluation of
the seismic vulnerability of masonry churches, Structural
Analysis of Historical Constructions-Modena, Lourenço
& Roca (eds), November, Padova, Italy.
Morales H., 2004, Análisis de la respuesta sísmica de un tem-
plo colonial típico en cruz latina (Analysis of the seismic
response of a typical Latin cross colonial temple), Master
Dg thesis, UNAM, México.
Rinawi,A. & Clough, R. 1992, Improved amplitude fitting for
frequency and damping estimation, Proc. 10th Int. Modal
Analysis Conference. Soc. for Exp. Mechanics, Bethel,
Conn, 1992, pp. 893–898.
Tomazevic M. & Velechovosky T. 1992, Some aspects of
testing small-scale masonry building models on sim-
ple earthquake simulators, Earthquake Engineering and
Structural Dynamics, Vol. 21.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Capacity of a traditional timber mortise and tenon joint

Artur O. Feio
Madeicávado – Madeiras, S. A. University Lusíada, Famalicão, Portugal

Paulo B. Lourenço
University of Minho, Guimarães, Portugal

José S. Machado
LNEC, Lisboa, Portugal

ABSTRACT: The safety and quality of traditional timber structures mostly depends on the conditions of
their connections. A testing campaign has investigated the behaviour of mortise and tenon timber joint. The
present paper addresses the quantification of the strength capacity of wood-wood mortise and tenon joint by
physical testing of full-scale specimens. In addition, the performance of different non-destructive techniques
(NDT) for assessing global strength is also evaluated. For this three non-destructive methods (ultrasonic testing,
Resistograph and Pilodyn) are considered and the possibility of their application is discussed based on the
application of simple linear regression models. New (NCW) and old (OCW) chestnut wood, obtained from
structural elements belonging to old buildings, is used. Finally, a nonlinear model has been formulated and the
predicted behaviour compared with the behaviour observed in the full-scale experiments, in terms of failure
mode and the ultimate load. The study has shown very good agreement with the experimental values.

1 INTRODUCTION of the toe and mortise depth. According to the Euro-

In the past, timber structural design was dominated
by the carpenter know-how which in turn was based
upon his previous works or upon the works of oth-
ers both supported on timber trusses solutions that
performed reasonable well (did not fall). Although car-
penters awareness that some members were subjected
to tension and others to compression stresses is evident
from the observation of old timber structures. Joints
assure important functions as holding on the truss and
sustaining the stresses imposed on it. In traditional
timber constructions load distribution through joints
was made with empirical knowledge, transmitted and
improved through generations. The early design rules
or standards were built upon this empirical evidence.
The mortise and tenon joint was selected because it
is one of the most commonly used and a typical exam-
ple of an interlocking joint. Mortise and tenon joints
were the basic components of joint craftery in Portugal
and connect two or more linear components, form-
ing a “L” or “T” type configuration, see Figure 1. The
key problem found in these joints is the possible pre-
mature failure caused by large displacements. Unlike
most timber joints, the load-displacement behaviour
of these joints is generally very ductile.
The bearing capacity of mortise and tenon joints is
a function of the angle of the connection, and length
pean building codes, joints are of crucial importance
for the seismic design of timber structures. However,
there are no recommendations on the design codes
about general dimensions, such as length of the toe and
the mortise depth in order to avoid structural failure of
the connections, and the joints in existing structures
are based on empirical rules.
Therefore, the present study addresses such type of
joints. In particular, the objective is: (a) to investigate
the static behaviour of real scale replicates, consid-
ering both new (NCW) and old (OCW) timber con-
nections (wood-wood connections) of chestnut wood
(Castanea sativa Mill.); (b) to characterize the ultimate
strength and, the global deformation of the joint, as
well as the respective failure patterns; and (c) to verify
the impact of time and loading history in the strength
and stiffness. For these purposes, the paper presents
several correlations between mechanical properties,
density and non destructive methods.
The adopted non-destructive methods for the joints
are the Pilodyn, Resistograph and ultrasonic tests.
These in situ methods allow to assess the safety of
old structures and preserve the original fabric as much
as possible, representing a first step towards diagno-
sis, structural analysis and the definition of possible
remedial measures (Ross et al. 1997). The inspec-
tion results, combined with historical information and

833
 Table 1. Average values of density (one specimen for each
timber element).

Density (kg/m3 )

Brace Rafter Average Std. Dev. Group

J_1 584.2 602.1 593.6 25.4 NCW

J_2 584.9 544.2
J_3 604.0 605.5
Figure 1. Details of typical tenon and mortise joints. J_4 590.2 633.3
J_5 575.1 605.6 568.8 31.4 OCW
a visual survey, can also be the support of mainte- J_6 598.9 574.7
nance decisions. Non-destructive evaluation is already J_7 507.1 545.7
widely applied to the control of structural integrity, due J_8 561.7 581.8
its characteristics of reliability, simplicity and low cost.
For the purpose of numerical analysis wood is often
considered as a homogenous and isotropic material. contractor claiming that the wood has been in ser-
This is certainly not the case as: (a) wood exhibits vice for over 100 years. The OCW specimens were
anisotropic elastic and inelastic behaviour; (b) natu- made using original beams obtained from rehabilita-
ral growth characteristics such as knots, slope grain tion works carried out in the Northern of Portugal,
and other defects are always present. Defects can be using specimens with the least possible damage. The
included in numerical simulations but this requires a NCW specimens were prepared using new wood with
thorough investigation of the specimens and fine tun- minor defects.
ing, being of moderate interest for practical purposes. Attention was paid to the conditioning of the tim-
On the contrary, the usage of orthotropic failure criteria ber before and after the manufacture of the joints. The
is essential for accurate numerical simulations. conditioning was conducted in such a way that the test
Failure criteria that describe orthotropic inelastic conditions correspond in a realistic manner to ade-
behaviour offer the opportunity to perform adequate quate in situ conditions as regards the influence of
analyses of wood elements and structures, beyond the moisture content and the occurrence of gaps induced
elastic limit. This can be especially valuable in the by shrinkage.
detailed analysis of timber joints and other details with Each specimen consists of two timber elements,
complex stress distribution. with a cross section of 92 × 150 mm2 , connected with
Here, the finite element method (FEM) is adopted a mortise and tenon joint without any pins. The angle
to simulate the structural behaviour and obtain a bet- between the elements is 65◦ .
ter understanding of the failure process. Calculations
are performed using a plane stress continuum model, 3 CHARACTERIZATION OF PHYSICAL AND
which can capture different strengths and soften- MECHANICAL PROPERTIES
ing/hardening characteristics in orthogonal directions.
The failure criterion is based on multi-surface plastic- 3.1 Density
ity, comprising an anisotropic Rankine yield criterion
for tension, combined with an anisotropic Hill criterion Given the conditioning of the specimens, the average
for compression. density ρm is determined for a moisture content of
The failure criterion from Lourenço (1996) is used 12%. Table 1 presents the results for the average den-
in the analysis. The influence of compression perpen- sity organized according to two group types (wood
dicular to the grain and elastic stiffness on the response element and age). The density tests were carried out
is addressed in detail. in samples removed from the specimens ends. Even if
the sample size is very low, the NCW group presents
2 DESCRIPTION OF THE TEST SPECIMENS slightly higher values of average density (≈4%) than
OCW group, with an average of 593.6 kg/m3 for NCW
Chestnut is usually present in historical Portuguese and 568.8 kg/m3 for OCW.
buildings and all the wood used in the specimens came
3.2 Experimental tests
from the North of Portugal. The 8 specimens were
divided in two groups: New Chestnut Wood (NCW), A test set-up was built to test the specimens under
obtained from recently sawn timber, and Old Chest- compression. One hydraulic jack was used to apply
nut Wood (OCW), obtained from structural elements a compression force aligned with the rafter, with a
belonging to old buildings (date and precise origin programmed loading cycle. The system includes a sup-
unknown).The old logs were obtained from a specialist port plate with stiffeners, able to rotate and ensure

834

Figure 2. Aspects of the destructive test set-up.
verticality of the brace. The support plate includes a toe
so that the rafter does not suffer a displacement along
its axis. The brace is hold in the original alignment with
a horizontal bar, connected to a load cell. Additionally
a feed, acquisition and amplification data system was
used, to obtain and to register all the data, see Figure 2.
Displacements were measured using linear vari-
able differential transducers (LVDT), continuously
recorded until failure occurred. The measurements of
the vertical and horizontal displacements in the spec-
imens were done by two pairs of LVDT’s placed on
opposite faces of the specimens to eliminate the effect
of bending (if any).
The loading procedure consisted of the application
of 2 monotonic load stages: firstly, the load was applied
up to 50% of the estimated maximum load (determined
on the basis of the preliminary tests) and was main-
tained for 30 s.The load was then reduced to 10% of the
estimated maximum load and maintained for 30 s. This
procedure was repeated once again and, thereafter, the
load was increased until ultimate load or until a max-
imum slip of 15 mm between the two timber elements
was reached. This is based on the EN 26891 (1991)
requirements.
A constant rate of loading corresponding to about
20% of the estimate maximum load per minute was
used, in such a manner that the ultimate load or slip of
15 mm was reached in about five minutes of additional
testing time in the final loading procedure. The total
testing time is about 9 to 12 minutes. The ultimate load
of the joint (Fult,joint ) is defined as the conventional
value corresponding to a strain equal to a 2% offset in
the usual terminology.
3.2.1 Results
The results of the experimental tests are presented in
Table 2. It can be seen that the results presents a huge
835
Table 2. Average values of density (one specimen for each
timber element).
Ultimate
Force (kN) Average Std. Dev. Group
J_1 121.6 145.4 18.9 NCW
J_2 161.5
J_3 159.7
J_4 138.9
J_5 126.4 133.8 (145.5*) 27.2 (16.7*) OCW
J_6 157.1
J_7 98.5
J_8 153.0
(*) average discarding specimen J_7.
Figure 3. Typical experimental failure patterns: joint col-
lapsed in compression, with uniform distribution of damage.
scatter, ranging from an ultimate force of 98.5 kN up to
a force of 161.5 kN. Even if the number of specimens is
rather low, the average force in terms of groups NCW
and OCW exhibits a difference lower than 10%. Spec-
imen J_7 can possibly be discarded because the value
found is too low and is controlled by a local defect: the
large longitudinal crack in the rafter. In this case, the
average ultimate force values of the groups NCW and
OCW are almost the same.
The main characteristic of the adopted joint is
that the direction of the grain of the two assembled
pieces it is not coincident, forming an acute angle. The
rafter is loaded in the direction parallel to the grain,
whereas the brace is loaded at an oblique angle induc-
ing large stresses perpendicular to the grain. Due to the
anisotropic behaviour of wood, wood stressed parallel
to the grain assumes the highest values of strength.
Therefore, the rafter, stressed in compression parallel
to the grain, easily penetrates the brace. The compres-
sive damage in the brace occurred either localized at
the toe or distributed along the full contact length.
Often, out-of-plane bulging of the rafter under the
contact length was observed. In some cases, compres-
sive damage was accompanied with shear failure in the
rafter in front of the toe. Figure 3 illustrates the typ-
ical damage observed at ultimate load and gives the
experimental results in terms of ultimate force.
 240 load-displacement diagrams. It is evident as addressed
220
before, that the scatter of the OCW group is much
200
larger than the scatter of NCW group, due to J_7 spec-
180

160
imen. From the load-displacement diagrams obtained
140
the following relevant remarks can be drawn. In a
Force (kN)

Max. Displacement = 12.93 mm

120 first phase, the diagrams always start with an upward
100 curvature, exhibiting a nonlinear, non-recoverable,
80 “bedding” response, which is due to the adjustment
60 of the tenon and the mortise. In a second phase, within
40
working stress levels, the response exhibits an approxi-
20 NCW mately linear branch up to the conventional maximum
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 load, which occurred at an average displacement of
Vertical Displacement (mm) 8 mm for the NCW group and 7 mm for the OCW
(a) group. The value of the displacement associated with
240 the maximum load is lower for the NCW group joints
220 in comparison with the OCW group joints, possibly
200
indicating a slightly larger deterioration of the timber
180
of the OCW specimens. It is noted that unloading-
160
reloading cycles within working stress levels provide
140
Force (kN)

120
100
80
60
40
20
0
0 1 2 3 4 5
Vertical Displacement (mm)
(b)
Figure 4. Envelope of load-displacement diagrams for:
(a) the NCW group, and (b) the OCW group.
The specimens were executed avoiding the presence
of large defects but some small defects were present.
During the tests it was observed that the longitudinal
and radial cracks of moderate width in the rafter did not
have a considerable influence in the ultimate strength
and in the global behaviour of the joints. The longitu-
dinal cracks show the tendency to close their thickness
and the radial cracks show the tendency to open. This
effect is more salient when the cracks are close to
the joint. On the other hand, the cracks present in the
brace, namely the longitudinal ones, show a tendency
to propagate and to open during the tests. Neverthe-
less, it seems difficult to quantify the influence of these
cracks in the ultimate strength of the joint.
3.2.2 Load-displacement diagrams
The difference in the results between old and new wood
is very low, which seems in agreement with the val-
ues of density found for the sample, where the NCW
group present slightly higher values of density (≈5%)
in comparison with the OCW group.
The results of all tests in terms of load-displacement
diagrams, given by the vertical force vs. vertical
displacement, are given in Figure 4.
Figure 4a and Figure 4b show typical individ-
ual load-displacements diagrams and envelopes of
a constant stiffness, which is higher than the loading
stiffness.The justification of this behaviour is probably
Max. Displacement = 11.87 mm attributed to the nonlinear behaviour of the interface
between rafter and brace, which exhibits a closure phe-
nomenon. Finally, after the conventional maximum
OCW load the displacement increases rapidly with a much
6 7 8 9 10 11 12
lower stiffness, due essentially to the compressive
failure of the wood in the rafter around the joint.
For the purpose of a more refined numerical anal-
ysis, the true load-displacement diagrams were cor-
rected with an offset that eliminates the upward curve
related to the nonlinear behaviour of the joint previous
to full contact (joint closure).
Due to the nature of load-displacement diagrams for
these joints, choosing the linear portion of the curve
includes some subjectivity. To reduce the subjectivity
the elastic stiffness was calculated between 1/3 and 2/3
of the ultimate load. The line plotted between these two
points to visually analyze the quality of the fit to the
linear portion of the plot, indicated that the proposed
procedure is adequate.
4 DESCRIPTION OF NON-DESTRUCTIVE
TEST PROCEDURES
In order to investigate possible correlations and the
validity of using NDE as a tool to assess the joint
strength, different non-destructive techniques (NDT’s)
have been carried out. The adopted NDT’s are the Pilo-
dyn, the Resistograph and the ultrasonic tests, which
were carried out in both timber elements. Average
values were considered in all measurements and two
readings per specimen, per side, were generally made
but a third one was added if the two first readings
differed significantly. Here, it is noted that Pilodyn
and Resistograph have been carried out in samples
removed from the elements ends, in order not to affect

836
Table 3. Average results of the Resistograph and Pilodyn 180
Fult, joint = -94.13 + 0.051 x V
Tests (values in bits/mm and mm, respectively). 170
Fult, joint = -306.57 + 0.115 x V
r2 = 0.61 r2 = 0.68

160
Resistograph Pilodyn
150 Fult, joint = -116.67 + 0.055 x V

Ultimate Load (kN)

r2 = 0.70
Brace Brace Rafter Rafter Group 140

130
J_1 449.5 449.5 8.0 8.0 NCW
120
J_2 367.7 367.7 7.8 8.8
J_3 365.0 365.0 8.0 7.3 110 Rafter and Brace
Rafter
J_4 463.6 463.6 8.0 7.3 100 Joint
Linear Fit (Rafter and Brace)
J_5 391.7 391.7 8.0 8.2 OCW Linear Fit (Rafter)
90
J_6 332.0 332.0 8.0 7.3 Linear Fit (Joint)
J_7 396.6 396.6 9.0 8.8 80
3200 3400 3500 3800 4000 4200 4400 4600 4800 5000 5200
J_8 323.1 323.1 8.7 8.2
Velocity (m/s)

Figure 5. Ultrasonic pulse velocity method for all tests:

(even if marginally) the local strength of the joint, and
the ultrasonic tests have been carried out at the exact
joint location.
4.1 Pilodyn and resistograph tests
The results shown represent therefore the average of
the readings permitting to reduce the scatter related to
local measurements.
The resistographic drills were made by using the
Resistograph 3450-S. For all the specimens, a resis-
tographic measure (RM) was calculated from the dia-
gram obtained with the Resistograph (see Feio et al.,
2005), as the ratio between the integral of the area of
the diagram and the length l of the drilled perforation.
The average results are presented in Table 3.
The Pilodyn 6J can measure the penetration of a
metallic needle with 2.5 mm of diameter, which is
inversely proportional to the density of the wood,
evaluating the surface hardness or resistance to super-
ficial penetration. The average results are presented in
Table 3.
4.2 Ultrasonic tests
Given the dimensions of the wood elements and the
diameter of the transducers used (φ = 25 mm), a refer-
ence testing mesh was defined on the central mid-third
of each element.
The tests in the brace and rafter aimed at charac-
terizing the mechanical properties of the elements in
zones nearby the joint. The test across the joint tried
to evaluate in a qualitative way the effectiveness of
the assembly between the two elements. A through-
transmission technique was adopted measuring the
wave propagation velocity parallel to the grain in each
element and joint.
A Pundit/Plus device (ultrasound generator) and a
pair of cylinder-shaped transducers (150 kHz) were
used. In all tests, coupling between the transducers
and specimens was assured by a conventional hair gel,
and a constant coupling pressure was applied on top of
relation between the ultimate load and the ultrasonic pulse
velocity.
the transducers by means of a rubber spring. The trans-
ducers were fixed into a special purpose assembly jig
to ensure their alignment.
4.3 Results
The results indicated that these non-destructive tech-
niques (Pilodyn and Resistograph method) are not
good indicators for predicting the joint strength.Again,
a definitive conclusion cannot be made because the
measurements had been made in specimens ends and
not at the joint location. However, taking into account
also the previous Chapters, these non-destructive tech-
niques seem more adequate to make a preliminary
estimation of density, rather than mechanical grading
of wood.
Figure 5 illustrates the relation between the ultimate
load and the ultrasonic pulse velocity. The results show
that ultrasonic pulse velocity could be a good indica-
tor for the prediction of the ultimate load. Here, it is
noted that the results using local measurements only
in the rafter, or rafter and brace together provide bet-
ter correlations that measurements across the joint. In
the latter, also the stiffness of the joint is taken into
account, meaning that the ultrasonic pulse velocity is
much lower. The joint stiffness is a relevant parame-
ter for the estimation of deformations and, sometimes,
resistance of timber structures. A clear linear correla-
tion was found, indicating that it is possible to estimate
joint stiffness from ultrasonic testing.
5 NUMERICAL ANALYSIS
In structural mechanics, a problem is usually consid-
ered to be nonlinear if the stiffness matrix or the load
vector depends on displacements. Nonlinearities in a
structure can be typically classified as material or as
geometric nonlinearities.

837
Table 4. Adopted elastic and inelastic material properties.
Ex Ey Gxy νxy
2 2 2
800 N/mm 8500 N/mm 1500 N/mm 0.3
fc,x fc,y β γ
7 N/mm2 45 N/mm2 −1.0 3.0
Nonlinear analysis is used to trace the equilib-
rium path up to and beyond the first critical point,
at which the structure becomes unstable. There is one
algorithm commonly used in the incremental iterative
solution of nonlinear problems: the Newton-Raphson
method. The full Newton-Raphson method, with stiff-
ness matrix update in each iteration is used in the
analyses carried out in this work.
Two different finite elements were considered in the
plane stress analyses carried out in this work: contin-
uum elements (8-noded) to represent wood and line
interface elements (6-noded) to represent the inter-
face between rafter and brace. The integration schemes
used are 2 × 2 Gauss integration points for the contin-
uum elements and 3 Lobatto integration points for the
interface elements.
5.1 The adopted anisotropic failure criteria
A plane stress continuum model, which can cap-
ture different strengths and softening characteris-
tics in orthogonal directions, was formulated by
Lourenço (1996). The proposed failure criterion con-
sists of an extension of conventional formulations for
isotropic quasi-brittle materials to describe orthotropic
behaviour. It is based on multi-surface plasticity, and
wood is an example of a material for which this crite-
rion applies, having different strengths in the directions
parallel and perpendicular to the grain.
Formulations of isotropic quasi-brittle materials
behaviour consider, generally, different inelastic cri-
teria for tension and compression. In this formulation,
and in order to model orthotropic material behaviour,
a Hill yield criterion for compression and a Rankine
yield criterion for tension were adopted.
5.2 Adopted material parameters
A characteristic of the adopted model is that the tension
strength, in a given direction, must be equal or lower
to the compression strength in the same direction. This
does not hold for wood. Here, the tensile part of the
yield criterion was ignored due to the irrelevant con-
tribution of the tensile strength in the global behaviour
of the joint. This means that the yield surface reduces
to the standard Hill criterion. The adopted elastic and
inelastic materials properties used in the analyses are
detailed in Table 4.
838
220
200
180 Kinfinite KSpring, fit
160
140
Force (kN)
120 Kfit
100
80
60
40 Numerical
Experimental
20
0
0 1 2 3 4 5 6 7 8
Vertical Displacement (mm)
Figure 6. Comparison between numerical and experimental
load-displacement diagrams.
The shape of the adopted yield criterion in the
compression-compression regime, features an extreme
degree of anisotropy with a ratio fc,x /fc,y = 0.156.
6 NUMERICAL VS. EXPERIMENTAL
RESULTS
A structured mesh is used for the rafter and the brace,
whereas an irregular transition mesh is used in the
vicinity of the connection between rafter and brace.
Interface elements are also used between the rafter and
the brace. The thickness ranges from 62 mm to 93 mm.
This aims at representing the thickness of the mortise.
A preliminary analysis with an infinite stiffness
of the interface, assuming a fully rigid connec-
tion, indicated that such an assumption provided
far too stiff results. Therefore, the stiffness of the
interface elements was obtained by inverse fitting.
A first conclusion is that the stiffness of the inter-
face elements has considerable influence in the yield
strength of timber joints. In Figure 6, three dis-
tinct situations are presented: a numerical simula-
tion with infinite stiffness of the interface elements
(kinfinite = kn = ks = 109 N/mm3 ); a numerical simula-
tion with an adjusted stiffness of the interface elements
obtained by inverse fitting of the experimental results
(kfit ): kn = 6000 N/mm3 and ks = 2308 N/mm3 ; and a
numerical simulation with a spring (kspring = 106 N/m)
located in the brace to simulate the reaction cell used
in the experimental sets. The stiffness of the spring was
again obtained by inverse fitting of the experimental
results, keeping the adjusted stiffness of the interface
elements.
The numerical results, in terms of force-
displacement diagrams, with the adjusted stiffness for
the interface elements, provide very good agreement
with the experimental results both in the linear and
nonlinear parts. The influence of the experimental hor-
izontal restraint, simulated by a linear spring, is only
 220

200

180

160

140

Force (kN)
120

100

80
Experimental
60 Numerical (kn=0.5)
40 Numerical (kn=1.0)
Numerical (kn=2.0)
20

0
0 1 2 3 4 5 6 7 8

Vertical Displacement (mm)

(a)
Figure 7. Minimum principal stresses (values in N/mm2 ). 220

200

180
marginal. The usage of infinite stiffness for the inter-
160
face (rigid joint) results in an increase of the slope
140
of the first part of the response, from 30 kN/mm to
Force (kN)
120
80 kN/mm (+266.7%). The ultimate strength of the
100
joint, given by an offset of the linear stretch by 2% in
80
terms of strain values, also changes from 130 kN to Experimental
60
152 kN (+17%), once the joint becomes fully rigid. 40
Numerical (ks=0.5)
Numerical (ks=1.0)
Figure 7 shows the contour of minimum principal 20 Numerical (ks=2.0)
stresses at the end of the analysis. It is possible to 0
observe a concentration of stresses in a narrower band 0 1 2 3 4 5 6 7 8

with peak stresses at the joint (zone where the interface
elements were placed).
With this concentration of stresses one may say that
failure is clearly governed by wood crushing where, for
a late stage of the analysis, the compressive strength
of the wood in the joint is completely exhausted. This
situation is also confirmed in the experiments.
7 EFFECTS OF THE MATERIAL PARAMETERS
A strong benefit of using numerical simulations is
that parametric studies can be easily carried out and
the sensitivity of the response to the material data
can be assessed. There are a total of six key param-
eters in the present model and the effect of each
parameter on the global response will be analyzed sep-
arately. It is noted that moderate variations (±25%)
are considered for the strengths and large variations
(division/multiplication by two) are considered for the
stiffness values. These assumptions are rooted in the
fact that strength is usually better known than stiffness.
Vertical Displacement (mm)
(b)
Figure 8. Effect of the variation of parameter: (a) kn , and
(b) ks on the model response.
an offset of the linear stretch by 2%, increases from
127.2 kN to 135.0 kN (+7%).
The reduction/increase of the normal stiffness of the
interface also affects the global stiffness of the joint:
the global stiffness of the joint decreases as the normal
stiffness of the interface decreases, being more sensi-
tive to this variation when compared with the ultimate
strength. The reduction of 50% of the kn parameter,
results in a decrease of the slope of the first part of the
response, from 32 kN/mm to 26 kN/mm (−23%).
On the other hand, the multiplication by a factor
of 2 of this parameter results in an increase of the
slope of the first part of the response, from 32 kN/mm
to 41 kN/mm (+28%). Because this parameter sets
the relation between the normal traction and the nor-
mal relative displacement, the obtained results were
expected a priori.

7.1 Normal stiffness of the interface

Figure 8a shows a comparison between the results of
the variation of the kn parameter: with a reduction of
50% in kn , the ultimate strength of the joint, given by
an offset of the linear stretch by 2%, decreases from
127.2 kN to 120 kN (−6%); multiplying kn by a fac-
tor of two the ultimate strength of the joint, given by
7.2 Tangential stiffness of the interface
Figure 8b shows a comparison between the results of
the variation of the ks parameter. The ultimate strength
is insensitive to a ks variation, whereas the reduc-
tion/increase of the ks parameter affects the global
stiffness of the joint: the global stiffness of the joint

839
 220
200
180
160
140
Force (kN)
120
100
80
60
40
20
0 0
0 1 2 3 4 5 6 7
Vertical Displacement (mm)
Figure 9. Effect of the variation of the elastic modulus of
elasticity (Ex ) on the model response.
decreases as the ks parameter decreases. The reduction
of 50% of the ks parameter, results in a decrease of the
slope of the first part of the response, from 32 kN/mm
to 28 kN/mm (−14%).
On the other hand, the multiplication by a factor of
2 of this parameter results in an increase of the slope
of the first part of the response, from 32 kN/mm to
37 kN/mm (+16%).
7.3 Elastic modulus
The effect of the variation of the elastic modulus of
elasticity parallel and perpendicular to the grain was
considered individually. Figure 9 indicates that the ulti-
mate strength is almost insensitive to the variation of
the elastic modulus of elasticity for wood (± 4%).
The inclusion of the effects of the elastic modulus
of elasticity does change significantly the elastic stiff-
ness of the joint. Therefore, decreasing the parameter
E decreases the global stiffness of the joint. The reduc-
tion of 50% of the Ex parameter, results in a decrease
of the slope of the first part of the response, from
32 kN/mm to 28 kN/mm (−14%). On the other hand,
the multiplication by a factor of 2 of this parameter
results in an increase of the slope of the first part of
the response, from 32 kN/mm to 36 kN/mm (+13%).
7.4 Compressive strength
The ultimate strength and the global stiffness of the
joint are insensitive to the variation of the compressive
strength of wood in the parallel direction.
Figure 10 indicates higher sensitivity of the ultimate
strength of the joint to the variation of the compres-
sive strength of wood in direction perpendicular to
the grain, as expected: with a reduction of 50%, the
ultimate strength of the joint, given by an offset of
the linear stretch by 2‰, decreases from 130 kN to
100 kN (−30%); multiplying by a factor of 2 the ulti-
mate strength of the joint, given by an offset of the
linear stretch by 2‰, increases from 130 kN to 160 kN
220
200
180
160
140
Force (kN)
120
100
80
Experimental Experimental
60
Numerical (Ex=0.5) Numerical (fc,y=0.75)
40
Numerical (Ex=1.0) Numerical (fc,y=1.0)
Numerical (Ex=2.0) 20 Numerical (fc,y=1.25)
8 0 1 2 3 4 5 6 7 8
Vertical Displacement (mm)
Figure 10. Effect of the variation of the compressive
strength (fc,y ) on the model response.
(+23%). However, the global stiffness of the joint is
insensitive to the variation of the compressive strength
perpendicular to the grain.
8 CONCLUSIONS
Despite the wide use of mortise and tenon joints
in existing timber structures scarce information is
available for design and in situ assessment. The objec-
tive of the present study was to quantify its strength
capacity by physical testing of full-scale specimens.
Also, the performance of different NDT for assessing
global joint strength is evaluated. Finally, the adequacy
of an anisotropic failure criterion to represent the
behaviour of a traditional mortise and tenon joint was
assessed from the comparison between experimental
and numerical results.
The difference in the results for the ultimate load
between the two groups is very low, which is in agree-
ment with the values of density found for the sample.
Thus, safety assessment of new and existing timber
structures can be made with similar mechanical data.
With respect to the usage of NDT for the predic-
tion of the ultimate strength, the dispersion found
for the density, Resistograph and Pilodyn do not rec-
ommended the usage of the related parameters for
quantitative mechanical assessment. On the contrary,
ultrasonic testing provides good correlations. Novel
linear regressions have been proposed in this study.
The different failure mechanisms observed in the
experiments are well captured by the model, which is
the most important validation of any simulation. It is
striking that such excellent agreement is obtained also
in the load-displacement diagrams.
A preliminary analysis considering an infinite stiff-
ness of the interface, assuming a fully rigid connection,
indicates that such an assumption provides too stiff
results. Another conclusion is that the normal stiffness
of the interface elements has considerable influence
in the yield strength of timber joints. The numerical
840
results, in terms of force-displacement diagrams, with
the adjusted stiffness for the interface elements, pro-
vide very good agreement with the experimental
results both in the linear and nonlinear parts. The
influence of the experimental horizontal restraint,
simulated by a linear spring, is only marginal.
It has been shown that the parameters that affect
most the ultimate load are the compressive strength of
wood perpendicular to the joint and the normal stiff-
ness of the interface elements representing the contact
between rafter and brace. The tangential stiffness of
interfaces and the Young’s moduli of wood have only
very limited influence in the response. The compres-
sive strength of wood parallel to the grain has almost
no influence in the response.
841
REFERENCES
Ross, R., DeGroot, R., Nelson, W., Lebow, P., 1997 – “The
relationship between stress wave transmission characteris-
tics and the compressive strength of biologically degraded
wood”. Forest Products Journal, Vol. 47(5), pp. 89–93.
CEN; 1991 – “EN 26891 – Timber structures. Joints Made
With Mechanical Fasteners General principles for the
determination of strength and deformation characteris-
tics”. Office for Official Publications of the European
Communities. Brussels, Belgium.
Feio, A., Machado, J., Lourenço, P., 2005 – Compres-
sive behaviour and NDT correlations for chestnut wood
(Castanea sativa Mill).
Lourenço, P., 1996 – Computational strategies for masonry
structures. PhD thesis, Delft University of Technology.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Behaviour of refurbished timber floors characterized by different

in-plane stiffness

Maurizio Piazza, Christian Baldessari, Roberto Tomasi & Ermanno Acler

Department of Mechanical and Structural Engineering, University of Trento, Italy

ABSTRACT: The structural behaviour of an existing masonry building subjected to seismic action, is strongly
affected by the in-plane stiffness of the floors, and by the connections between the horizontal diaphragms and the
masonry walls. The aim of the research is to experimentally evaluate the behaviour of timber floor refurbished
using different techniques, with special regard to the in-plane stiffness. The size adopted for the specimens (5 m
span, 4 m width), is similar to the ordinary dimensions of timber floors in historical buildings in Italy. Taking
into account the size of the specimens, and the need to determine the in-plane strength and stiffness of the floor,
a special test set-up has been designed and adopted in order to allow the free in-plane deformation of the floor
itself subjected to lateral load: the load configuration applied to the floor simulates the effect of seismic action
on the floor. The experimental phase of the research aims to calibrate engineered models that can be used for
studying existing structures.

1 INTRODUCTION

The structural response of a masonry existing building

to seismic actions is strongly affected by many param-
eters, such as the plan distributions, the texture and the
quality of the masonry walls, the plan regularity, the
distribution and the size of the openings, the charac-
teristic of the floor and the interconnection between
vertical and horizontal elements.
The experience of the past earthquakes has shown
that a key role is played by the horizontal diaphragm
in the transmission of the seismic actions.
Whereas the floor is not satisfactorily connected to
the adjacent walls, or the in-plane stiffness is inad-
equate, different collapse modes can be observed
involving overturning of the walls (see Figure 1).
Masonry walls can counteract, generally, an insuffi-
cient resistance to lateral loads acting out of plane.
Whereas the horizontal diaphragm can be consid-
ered perfectly rigid, and the connections between walls Figure 1. Wall overturning modes due to inadequate
and diaphragm are correctly assured, the lateral seis- stiffness of the floor.
mic load can be fully transmitted to the walls parallel to
the horizontal action, allowing masonry to counteract walls, can imply, respectively, a significant self weight
a much higher resistance (see Figure 2). increase and a weakening of the existing masonry
The need to increase the in-plane stiffness has walls.
induced, in the past, some strengthening solutions Therefore, after the Umbria – Marche earthquake,
which recent earthquakes demonstrated to be inade- some floor refurbishment techniques has been recon-
quate or, in some cases, unfavourable. The substitution sidered: the new Italian standard code on existing
of timber floors with concrete ones, the insertion of buildings, appeared in 2003, bans the possibility to
a concrete curb “inside” the thickness of the masonry insert concrete curb in the depth of the existing

843
masonry walls, and suggests new alternative strength-
ening techniques for the horizontal diaphragm. Some
of them are presented in the next paragraphs.
2 TIMBER FLOOR TYPOLOGY AND
REFURBISHMENT TECHNIQUES
In this paper a simple supported timber floor is consid-
ered, where timber beams have a section of 18 × 18 cm,
spaced 50 cm, which is a recurrent configuration for
the floor structure in Italian historical buildings. The
deck system is composed with a simple layer of wood
planks (3 cm thick), crossly arranged and nailed to the
timber beams (Figure 3a). Starting with this config-
uration, five different reinforcement techniques were
Figure 2. Role of the diaphragm preventing the overturning
modes of masonry walls: (a) the inadequate in-plane stiffness
of the floor causes overturning of the walls perpendicular to
the seismic action; (b) a stiff diaphragm allows forces to be
transmitted to the walls parallel to the seismic action.
studied, which are described in the cited Italian stan-
dard as possible approaches in order to increase the in
plane stiffness and therefore the building robustness.
Moreover, in some cases, the existing floor can benefit
of a higher level of out-of plane strength and stiffness,
depending on the different techniques considered, as
in the case of the timber–timber or timber–concrete
composite structure.
In the first strengthening technique considered, a
second layer of wood planks (100 cm width, 3 cm
thick) is used, crossly arranged to the existing ones
and fixed by means of 10 mm steel rods, epoxy glued
into timber beams (Figure 3 b).
The application of diagonal bracing (45◦ ) on the
existing wood planks (Figure 3.c), can be done utiliz-
ing wide sheets of CFRP (50 mm wide, 1.4 mm thick),
glued to the wood by means of epoxy-based resin, or
adopting light steel plates (80 mm wide, 2 mm thick),
nailed to the planks. The mesh of diagonal bracing
applied to the floor is 705 mm for both techniques
analysed: obviously it is a function of seismic intensity.
Another strengthening technique analysed makes
use of three layers of plywood panels (21 mm thick),
arranged on the existing wood layer, and connected
to timber beams by means of 10 mm steel rods, glued
with epoxy−based resin (Figure 3.d).
A reinforced concrete slab connected to the timber
beams (Figure 3.e) is a strengthening technique which
was widely used in the past decades when restoring
timber floors. Regular concrete slab (50 mm thick)
is built on the wood planks; the slab reinforcement
is composed by a welded steel mesh (6 mm diame-
ter, mesh 200 × 200 mm). The connections between

Figure 3. Different timber floor in plane shear strengthening techniques: (a) existing simple layer of wood planks on the timber
beams; (b) second layer of wood planks crossly arranged to the existing one and fixed by means of steel studs; (c) diagonal
bracing of the existing wood planks by means of light steel plates or FRP laminae; (d) three layers of plywood panels glued
on the existing wood planks; (e) a stud-connected reinforced concrete slab (all measures in mm).

844
the timber beams and the concrete slab is obtained
by means of L shaped connectors (re-bars FeB 44 k,
16 mm diameter, 150 mm and 50 mm edges, epoxy
glued to timber beams in holes 90 mm deep).
The in-plane shear behaviour of the horizontal
diaphragm subjected to seismic action can be schemat-
ically illustrated as in Figure 5. The lateral forces must
be transmitted to the shear masonry walls parallel
to the seismic action, and therefore an adequate link
between the floor and the walls must be assured.
Moreover the in-plane deflection of the slab induces
compression and tension zones in the deck. Compres-
sion stresses can be counteracted by the wood planks,
while for the tension stresses a reinforcement element
must be on-purpose designed.
Two technological solutions were considered in the
analysis. For the floor typology depicted in Figures 3.a,
b, c, d, a steel ring curb encloses the perimeter of the
floors. The steel curb has the double role to take up the
tension stresses of the deck, and to transmit the shear
forces on the lateral walls.
Figure 4. (a) Steel curbs between the shear reinforced
planking and the masonry walls, (b) Connection between con-
crete curb and the walls, and (c) between the steel curb and
the walls.
845
In the solution proposed by Doglioni (2000), an L-
shaped profile is connected to the floor by means of
screws; both the end sides of the profile are linked to
the lateral masonry through threaded steel bars (diame-
ter varying from 20 to 30 mm), which are chemically or
mechanically connected to the masonry walls (see Fig-
ure 4). Along the floor border, additional connectors
can be placed, in order to guarantee the shear trans-
mission to lateral walls, and to prevent the possibility
of out–of–plane mechanisms.
For the concrete slab floor depicted in Figure 3e, the
concrete ring curb is guaranteed by additional steel
bars inserted along the border of concrete deck (see
Figure 4.a), in order to avoid the insertion of a concrete
curb “inside” the thickness of the masonry walls, that
can weaken significantly the existing walls.
3 EXPERIMENTAL CAMPAIGN
3.1 Test set-up
The experimental test apparatus was thoroughly
designed taking into account the specimen dimen-
sions, the boundary conditions and the load config-
uration.
The floors were built in the laboratory adopting
different specimen sizes: the first monotonic pilot
tests were performed on small size floors (1 × 2 m);
then the cyclic tests were performed on real size
floor specimens (4 × 5 m). In both test configura-
tions, real size timber elements were used: beams
(0,18 × 0,18 × 4,2 m), planks (0,2 × 0,03 × l m, with
l varying from 0,6 to 1,6 m), and the previously
described reinforcement elements. The real size
dimensions showed to be necessary in order to sim-
ulate the real contribution of the secondary elements
(planks and reinforcement elements).
Another important aspect concerning the design
of the test apparatus is the boundary conditions of
specimens.
Figure 5. Role of the steel ring curb in the transmission of
the shear force through the diaphragm.
 Instead of reproducing, in the test apparatus, the
actual boundary conditions of the floor in situ, the
adopted design approach was intended to allow free
in-plane deformation of the diaphragm. In fact the
experimental replication of the interaction between
floor and masonry wall was considered unpractical,
with many problems involved in understanding and
analysing the experimental results. On the contrary,
the adopted configuration allows an accurate measur-
ing of the deck in-plane stiffness, which is one of the
key parameters for the structural restoration design.
Therefore the floor specimen was linked to the lab-
oratory reaction floor by means of two external hinges,
according to the scheme depicted in Figure 6. The
hinges are positioned centrally, at the neutral axis level.
Particular attention was paid on the load system,
in order to reproduce the transmission of seismic
forces through the floor. During an earthquake, the
lateral forces are proportional to the vertical load
applied on the floor, which can be considered uni-
formly distributed. Therefore a uniformly distributed
horizontal action should be applied to the floor under
experimentation.
The set-up adopted for the force transmission is
depicted in Figure 7, where two levels of steel cross-
beams were adopted in order to distribute the load
of the single hydraulic jack onto four timber beams.
A preliminary numerical analysis proved this config-
uration to be able to accurately reproduce a uniformly
distributed load.
Four different load configurations were consid-
ered, from the theoretical condition of uniform load
to the real condition adopted during the test, which
can be estimated sufficiently accurate according to the
numerical results reported in Table 1. It is worth noth-
ing that the real size floor tested is made up of 11
timber beams.
Figure 6. Set up configuration of the floor adopted for the
experimental campaign.
846
The floor specimens and the steel beams form-
ing the test apparatus rested on some timber supports
fixed to the reaction floor. In order to reduce friction
effects during the tests, Teflon plate were added at the
interfaces between materials.
In the real size specimens, cyclic tests were per-
formed, therefore the loading system should work in
both directions: beside the pushing system depicted
in Figure 7, a pulling system composed by steel plates
applied to both ends of the floor beams, and connected
with pre-tensioned tie-rods, was adopted (Figure 10).
Movements, deflections, and loads were measured
at multiple locations on the diaphragm specimens
using electronic sensors of various types, connected
to a computer controlled data acquisition system.
In Figure 8, the instrumentation layout is shown. In
order to find the in-plane shear deflection of the floor,
two wire sensors were used, diagonally arranged. At
the ends of the 5 beams, not directly loaded, 5 LVDTs
transducers were used in order to record the floor in-
plane deformation.
Three other LVDTs instruments were placed for
finding lateral movements of the tested floor (orthog-
onally to the jack direction); finally, in the zones of the
slab characterised by maximum tension/compression
stresses, 4 strain gauges were used, 2 of which applied
on the metallic profile and 2 on the planks, so to find
the efficacy of the steel.
3.2 Test protocol for cyclic tests
Cyclic quasi-static tests were performed under dis-
placement control, with a loading rate varying
Figure 7. Set-up adopted for the force transmission.
Table 1. Comparison of the numerical floor deflections.
Deviation
Displacement
Load mm mm %
Uniform 25.65 – –
11 beams 23.86 1.79 7.0
6 beams 23.38 2.27 8.8
4 beams 23.54 2.11 8.2
 from 0,05 e 2 mm/s, according to the test proce-
dure described in the European standard EN 12512
(Figure 9).
In this procedure, cyclic tests are defined as a func-
tion of the yielding value, which can only experimen-
tally be determined through a preliminary monotonic
test. Such preliminary tests were conducted on the
small size specimens, with the hydraulic jack acting
only on the central timber beam.

4 EXPERIMENTAL RESULTS

4.1 Floor shear stiffness

Figure 8. Measuring set-up.
Figure 9. Test protocol according to EN 12512.
So far the test program was completed only for two
floor typologies: the so–called “original” timber floor
type, with a simple layer of wood planks (Figure 3.a)
and the timber-concrete composite floor (Figure 3.e).
The experimental results are, in the following dia-
grams, expressed in terms of force of the hydraulic jack
versus the displacement of the central timber beam.
Those diagrams can adequately represent the in-plane
stiffness of the floor: the value of stiffness was taken
from monotonic tests, according to EN 12512.
No stiffness degradation was observed for both
floor typology until the collapse, which occurs at dis-
placement level of 6 × dy for the “original” timber
floor, and 4 × dy for the timber-concrete composite
floor.
The strength capacity of both typologies proved
to be high, as can be appreciated by means of the
following simple example.

Figure 10. Global view of the set-up configuration for cyclic test.

847
 Figure 12. Experimental results for monotonic and cyclic
test on the timber – concrete composite floor (Figure 3e).
Figure 11. Experimental results for monotonic and cyclic
tests on the “original” simple plank – timber floor (Figure 3a). Table 2. Monotonic test results.

dy Fy k
Floor mm kN kN/mm
Taking into consideration a simple 3 storey masonry
building, where 60 cm masonry walls surround a floor Wood planks 30 29.7 0.99
with dimension 4 × 5 m, the lateral force at the third Concrete slab 13 112.0 8.62
storey level (height 9 m) can be estimated for a peak
ground acceleration (PGA) of 0,35 g (which corre-
sponds to the highest level, in Italian seismic standard).
This force, hereafter indicated as Feq, is the maximum Table 3. Cyclic test results.
lateral force transmitted by the horizontal diaphragm,
in this simple case. From experimental results reported d max F max d eq* F eq*
in Figures 11 and 12, the collapse of the diaphragms Floor mm kN mm kN
happened for higher values (ranging from 1,5 to
2 × Feq). Wood planks 180 203 100 133.8
These simple considerations confirm what observed Concrete slab 52 451 10 142.4
during post earthquake survey, where failure is asso-
* Seismic action on floor PGA = 0.35 g.
ciated with out-of-plane fall down of the support-
ing walls rather than with inadequate strength of
diaphragms. diaphragm can be considered as rigid). For the “origi-
In Table 3, the maximum deformation correspond- nal” timber floor illustrated in Figure 3.a, the deforma-
ing to the Feq force is reported. For the timber – con- tion value corresponding to the Feq force is 100 mm,
crete composite floor, the corresponding deformation which can be awkwardly reputed consistent with the
is 10 mm (according to the Italian standard such safety of the structure.

848
4.2 Role of the steel ring plate
Cyclic test results highlighted the role of the steel ring
curb for the specimen reported in Figure 3.a, both in
terms of stiffness and strength.
The steel ring curb increases the maximum shear
force transmitted by the floor to the walls; at the same
time, resistance is assured against tension and com-
pression forces induced in the slab by the movements
generated by the quake.
This fact can be appreciated considering the values
of the strains locally recorded in the steel L-shaped
plate and in the nearby planks, were some strain gauges
were applied. The stress values in the steel plate and
in the wood planks reported in Figure 13.b, c, were
computed considering the mechanical and geometrical
properties of the different components, and the axial
strains recorded by means of the strain gauges.
Figure 13.d reports the maximum lateral deforma-
tion of the simple plank – timber floor: the values
confirm that the floor can carry out free in-plane defor-
mation, without any lateral constraint, which was the
base assumption for the adopted set-up. Figure 14
shows the effectiveness of the steel curb in the existing
timber floor with a simple layer of wood planks.

5 CONCLUSIONS AND FUTURE WORKS

The in-plane stiffness of the floors strongly affects the

structural behaviour of an existing masonry building
subjected to seismic action. It defines the seismic dis-
tribution of forces on lateral walls and the request dis-
placement for verifying the out–of–plane mechanism
of the walls.
The real size used for the specimens proved to be
very important in order to determine the in-plane stiff-
ness of the floor and to adequately simulate the real
contribution of the secondary elements (planks and
reinforcement elements).
The tests showed also the efficiency and the con-
tribution of the steel ring curb, in terms of stiffness
and strength. The possibility to have many connec-
tors along the floor border guarantees a nearly uni-
form transmission of shear forces to lateral walls. Its
strength contribution is essential in the tension zones
of the deck. Finally the ductility of steel curb ensures
a constant strength contribution when cyclic loadings
are applied. Figure 13. Some experimental results of the cyclic tests:
The experimental campaign will continue with (a) load time pattern; (b) forces on steel curb and wood plank;
other 4 specimens, where other reinforcement tech- (c) forces on wood plank; (d) lateral displacement of the
niques are applied. At the end, it will be possible external beams in the cyclic tests.
to give a first evaluation on the effectiveness of
each strengthening techniques in order to choose the existing floors. Experimentally found in-plane stiff-
suitable reinforcing technique. ness of floors will allow to define the values of
In the next phase of the research, test results will be the parameters characterizing the behaviour of the
used to develop project procedures for strengthening structure.

849

Figure 14. Effect of the steel curb applied to the existing
deck system with a simple layer of wood planks (figure 3.a).
The final aim of the research is to develop a suitable
equivalent shell modelling of timber floors, to be easily
implemented in global numerical models of traditional
buildings. It will be possible, consequently, to choose
adequate strengthening strategies for existing floors
and to optimize the interventions for each particular
building.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the Italian
ReLUIS Consortium for financing the study, within
850
the research program carried out for the Italian Agency
for Emergency Management.
REFERENCES
Borri, A., Corradi, M., Grazini, A. 2005. A method for
flexural reinforcement of old wood beams with CFRP
materials. J Compos Part B, 36/2:143–53.
CEN, EN 1995-1-1:2004. Eurocode 5: Design of timber
structures – Part 1-1: General – Common rules and rules
for buildings.
CEN, EN 12512:2006 Timber structures – Test methods –
Cyclic testing of joints made with mechanical fasteners.
Corradi, M. Speranzini, E., Borri, A. & Vignoli, A. 2006. In-
plane shear reinforcement of wood beam floors with FRP.
J Compos Part B; 37: 310–319.
D’Ayala, D., Speranza, E. 2002. An integrated procedure
for the assessment of seismic vulnerability of historic
buildings. Proc. Of the 12th European Conference on
Earthquake Engineering, London, Elsevier Science, paper
n. 561 (CD-ROM).
Doglioni, F. (Editor) 2000, Codice di pratica (linee guida) per
la progettazione degli interventi di riparazione, miglio-
ramento sismico e restauro dei beni architettonici dan-
neggiati dal terremoto umbro-marchigiano del 1997,
Bollettino Ufficiale della Regione Marche, Ancona.
Gattesco, N., Macorini, L. 2006. Strengthening and Stiffening
Ancient Wooden Floors with Flat Steel Profiles. Structural
Analysis of Historical Constructions, New Delhi 2006 P.B.
Lourenço, P. Roca, C. Modena, S. Agrawal (Eds.).
Griffith, M.C., Magenes, G., Melis G., Picchi L. 2003. Eval-
uation of out-of-plane stability of unreinforced masonry
walls subjected to seismic excitation. Journal of Earth-
quake Engineering, Vol. 7, Special Issue 1, pp. 141–169.
Marini, A., Giuriani, E. 2006. Transformation of Wooden
Roof Pitches intoAntiseismic Shear Resistance Diaphragms.
Structural Analysis of Historical Constructions, New
Delhi 2006 P.B. Lourenço, P. Roca, C. Modena, S.Agrawal
(Eds.).
Ordinanza, P.C.M. 3431 2005. Primi elementi in materia di
criteri generali per la classificazione sismica del territorio
nazionale e di normative tecniche per le costruzioni in
zona sismica.
Piazza, M., Turrini, G. 1983. Il recupero dei solai in legno.
Recuperare, n◦ 7. (In Italian).
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Stone masonry walls: Strengthening with TRM (I)

J. T. San-José, D. García, R. San-Mateos & J. Díez

Labein – Tecnalia, Spain

ABSTRACT: The purpose of this paper is to present the state of an investigation aimed at strengthening
decayed stone masonry walls with textile reinforced lime-cement mortars (TRM). The proposed solution could
be applied by itself or jointly with other techniques of reinforcement, like sewing or injections, to guarantee the
integrity of the wall. The validation of this strengthening system is carried out by a laboratory test campaign. The
experimental study is based on the typology of the Spanish Romanesque walls, a plentiful constructive element
in this country. The validation of the solution is being done by means of a characterization of the individual
materials (stone, mortar, TRM), designing of the anchor system and construction, strengthening and testing of
1/3 scale stone walls. The wall specimens have been erected with the same configuration of the original ones.
They are three-leaf masonry walls with an inner core made of low quality mortar and rough-cut sandstone pieces
(remains from rough-shaping of the stones), poured between the two external layers, and they are deflected a
common structural damage in these elements.

1 INTRODUCTION

The use of Fibre Reinforced Polymer (FRP) is becom-

ing a common practice in the strengthening of masonry
civil structures (ACI, 2002). Furthermore, these FRP
systems seem to be a promising solution for cultural
heritage, as in some occasions they could be reversible
solutions, flexible to be applied in a wide range of
structural elements and shapes, providing clean, safe
and cost effective restoration techniques, requiring
minimum intervention in the monument.
However, these systems that are currently designed
and used for civil works (Bakis, 2002), required further
development and research as per the following aspects
have not been solved until now:
– Organic resins, used as matrix to encapsulate and
bind the fibres and to bond the FRP laminates
to ancient substrates, are incompatible with these
existing heritage materials. They are also vulnerable
to humidity and high temperatures.
– The fixing and anchorage systems are of a very
limited efficacy, as they are designed basically for
concrete substrates, thus compromising the reliabil-
ity of these systems applied to decayed substrates
of heritage structures (San José, 2006). The actual
anchorage systems in civil works imply high visual
impacts by using bolted metallic plates, steel pro-
files around the edges, high anchorage lengths
affecting annexed elements, etc.
Consequently, new ideas have to be established
for the heritage structures by combining advanced
Figure 1. Example of reinforcement core: basalt fibre
technical textile (dimensions in cm).
composites and light additional fixing systems to the
substrates, such as simple bolting, FRP mats, inorganic
mortars, etc (Casareto, 2002). One possible solution
could be the substitution of the organic binder (poly-
mer) by an inorganic mortar, lime or cement based,
compatible with the specific materials used in each
structure. Another issue is that with mortar as a matrix
of the composite strengthening system, the fibres
should be in a textile format instead of fabric.
The properties, the amount and the arrangement of
the used fibre materials have a great influence to the
characteristics of the composite. Requirements on the
fibres are: high fibre tenacity, a modulus of elastic-
ity much higher than that of the mortar matrix, small
relaxation under permanent load, a good and constant
adhesion between reinforcement and mortar, low cost
and the possibility of processing them easily on tex-
tile machinery. Alkali-resistant man-made glass fibres
(AR-glass), carbon and aramid essentially meet these
requirements for the design and fabrication of textile
reinforcements.

851
 This new structural material, namely textile rein-
forced mortar (TRM), was developed as an evolution
of short-fibres reinforced concrete, in order to obtain
very thin-structured concrete elements with a high
strength in compression as well as tension.
The results obtained from preliminary studies on
the use of TRM as a strengthening solution of unre-
inforced masonry walls (Triantafillou, 2001) can be
extended to heritage masonries where the use of epoxy
resins is prohibited. For these masonry elements, a
lower strength TRM system than for reinforced con-
crete structures will be suitable. Hence the use of lower
Figure 2. Scheme of the strengthening solution.
properties fibres than carbon, for example AR-glass or
basalt fibres can be adopted.
The need of a low cost TRM system is obvious,
since there are many thousands of heritage buildings
around the Mediterranean and many of them in coun-
tries where the cost is of major importance for the
governments to fund a heritage monuments strength-
ening project. In a TRM system, mortar’s cost is low
compared to textile cost. Hence the cost of textile is
predominant. For this reason the carbon and aramid
high cost fibres are not suitable.
The expected results of this strengthening system
Figure 3. Wall specimens.
are: an increase in ductility and tension/shear resis-
tance, a decrease of cracks by located tensile loads and design of suitable anchorage system and the construc-
an improvement in the general behaviour, especially at tion, strengthening and testing (under static loads) of
the failure moment. nine real stone walls made at 1/3 scale.
These structural tests were defined from the knowl-
2 CONCEPT OF THE PROPOSED SOLUTION edge of this type of masonry (typology, materials, etc.)
and the structural behaviour connected to the spe-
Related to the TRM application as strengthening sys- cific deficiency to eliminate. Therefore, a particular
tem on historical masonry structures, nowadays, there attention was taken in choosing the geometrical and
are in course multi-year development efforts regard- the morphological characteristics and the constituent
ing its ability to retrofit un-reinforced masonry walls materials of the walls to test, to make them as much
(URM). The objective is to establish the adaptabil- as possible representative of the Spanish Romanesque
ity of a full-compatible and highly durable structural real typology, available in situ.
strengthening system based in TRM, including:
– The strengthening core: a technical fibre textile, 3 TEST CAMPAIGN
adaptable to strengthen different structural elements
(vaults, aches, walls) and substrates (masonry, A set of nine walls was constructed with low strength
adobe bricks and timber). lime-cement mortar and sandstone units, representing
– Fixing system: based on mortars or/and in anchor- Spanish Romanesque walls and similar to walls often
age devices, in case of need. found in historical urban centres., with two external
– Conditioning and finishing mortars: based on mod- leaves and an internal core of rubble material. Both
ified or lime-cement mortars, such as a compatible of them have the external leaf made of ashlar masonry
interface between the substrate and the TRM and and the internal leaf made of rough masonry. The walls
improving its aesthetic integration. The effective- are deflected, a common structural problem in these
ness of any externally applied reinforcement is elements, having the following nominal dimensions:
highly dependant on the bond between the com- 2 m wide, 2 m high, 0,3 m thick and a 5 cm deflection,
posite and the substrate, therefore the interface from the half height to the top.
behaviour is one of the key issues in the structural
analysis. 3.1 Masonry
The validation of the strengthening system is carried Two types of stone were used for the walls. For the
out by a lab test campaign, regarding characterization ashlar masonry leaves, Sandstone1, a uniform fine
of the materials (stone, mortar, masonry andTRM), the grain sandstone rock, was used. It was received in

852
Table 1. Mechanical properties of stone.

Stone Sandstone1 Sandstone2

Density (kg/m3 ) 2090 2066

Compressive Strength (MPa) 36,20 64,60
Flexural Strength (MPa) 6,28 5,14
Elasticity Modulus E (MPa) 10.468 10.620
Shear Modulus G (MPa) 4.046 4.635
Poisson Ratio 0,3 0,15

Mortar Strength Evolution

0,35

0,30

0,25
Strength (MPa)

0,20 Figure 5. Rough masonry prisms.

0,15
Table 2. Results from stone masonry prisms test.
0,10

0,05 Material fM fM εmax E

specimen MPa MPa – MPa
0,00
0 20 40 60 80 100 120 140 160 180 200
Age (days) Dry joint ashlar masonry 3,20 13,11 0,0178 1356
PSs 3,20 13,11 0,0178 1356
Figure 4. Trend lines of the evolution of the compressive
(up) and flexural (down) strength of the used lime-cement Ashlar masonry 2,09 4,54 0,0271 177
mortar. PS1 2,09 4,54 0,0271 177
Rough masonry 0,37 1,84 0,0363 58,3
big blocks, and afterwards, it was cut in a cutting ∗
PM1 2,83 0,0435 72,9
machine of samples with a diamond disc. For the rub- PM2 0,39 1,66 0,0424 48,5
ble masonry leaves, Sandstone2, a darker rock, was PM3 0,36 1,10 0,0275 51,0
used. It was received as irregular blocks and they were PM4 0,36 1,76 0,0319 61,0
break manually by the bricklayers. Inner core 0,12 0,19 0,0192 14,7
Cores of Ø30 mm were taken from these stones and PM1 ∗
0,16 0,0265 14,6
tested, obtaining the following mechanical properties: PM2 0,12 0,20 0,0234 14,5
The mortar used in the joints and in the inner core PM3 0,13 0,18 0,0113 12,7
was designed on the basis of the low compressive PM4 0,12 0,22 0,0159 16,9
strength (<0,5 MPa) of mortar found in some decayed
∗
walls, aiming to obtain a similar behaviour. Dosage Non determined.
of this mortar was 0,5-1,5-19 (white cement – lime –
facilitate the transport and assure the flatness on the
sand).
universal test machine. Their dimensions are 50 cm
During the construction of the walls, cylindri-
wide, 40 cm high and 30 cm thick. The prims were
cal (Ø10 × 20 cm) and prismatic (4 × 4 × 16 cm) test
stored inside, in an area free of drafts, monitoring
specimens were made from samples of the different
the temperature and relative humidity of the curing
mortar mixes to check its mechanical properties, nat-
environment (18–25◦ C and HR 60–85%).
urally deviated because of the addition of water at a
Prisms were tested at the age of 150 days, with load
guess, like masons used to do. They were tested at
velocity control under uniaxial compression load, in
different ages so the evolution of the strength of the
a universal compression test machine, recording the
mortar is controlled and it will provide an estimation
development of cracks and the values of stresses and
of inner core mortar value.
displacements. In the table below, the ultimate com-
In spite of the numerous attempts to obtain the
pression stress (fM ) and the stress corresponding to
compressive strength of the stone masonry from its
the first crack is included (fM ). A deformability mod-
typology, geometry and mechanical properties of units
ulus was also calculated considering the area of the
and mortar, up to now, prisms testing is an essential
initial cross section and the strain at ultimate stress.
approach to know the behaviour of the entire walls
The following conclusions were obtained:
under compression loads.
Additionally to the walls, prismatic specimens were – Dry ashlar masonry has the highest deformabil-
casted with the same materials and configuration for ity modulus and more load capacity than the other
each of the leaves. They were built over steel plates to combinations under compression load.

853
– Ashlar masonry with bed mortar joints possesses
the most uniform behaviour, with a slope of the
strain-stress line practically constant.

3.2 Textile reinforced mortar

Textiles of TRM systems can be made of various types
of fibres. However the basic architecture of a textile
used in a TRM system is the same regardless the type
of fibres used. A bidirectional textile at 0/90 degrees
made of basalt fibres and with grid square openings of
25 mm has been adopted as preliminary design of the
TRM system in this experimental research.
The basalt fibres have excellent alkali resistance, Figure 6. Tensile text of TRM.
similar mechanical properties to glass, good fatigue
resistance and much lower cost than carbon or aramid
fibres. Basically, basalt is a natural material that is
carbonate taken into solution from the surrounding
found in volcanic rocks with a melting point of about
mortar.
1400◦ C, which can be found all over the world with
– Water vapour permeability. It is very important that
differing chemical compositions. Basalt fibres show
mortars are porous enough to allow moisture to
excellent natural adhesion to a broad range of binders,
pass through them and evaporate. This helps to keep
coating compounds and matrix materials in composite
houses dry and makes a healthy atmosphere inside.
applications. This property can be further enhanced
When the moisture level rises in the stone walls,
through optimized surface treatment.
lime plasters would allow the moisture to evaporate
The effectiveness of any externally applied rein-
easily. Damp walls that have been rendered in water-
forcement is highly dependant on the bond between
proof cement will become wetter because moisture
the composite and the substrate. Therefore, the inter-
trapped behind the render will naturally rise to new
face behaviour is one of the key issues in the structural
and higher levels within the wall. This action can
response (CEB-FIP. 2001).
also damage and weaken walls as a result of salt, dis-
The characteristics of mortars that have to be taken
solved in rising moisture, crystallizing in the mortar
into account are the workability, the rate of harden-
behind the render and breaking it up.
ing and the shrinkage for the fresh mortar, and for
the hardened mortar, the appearance, the moisture and A common mortar, used for rendering, plaster or as
air permeability, the compressive and tensile strength, a finished coat, has been adopted. The dosage is 1-0,5-
the adhesion, the ability to tolerate movements and of 5 (white cement – lime – sand), with a maximum grain
course the wetness, frost and salt resistance. size of 2 mm. This mortar has been used for years to
In this context, the most important characteristics the inner leaf of Romanesque walls, demonstrating a
of the mortars are: good water vapour permeability and enough flexibility
to allow deformations in the masonry building without
– Chemical and physical compatibility with the sub-
cracking.
strate. Mortar makes the environment in which
With this mortar and the selected technical textile
stones must live. If stones and the walls made from
of basalt fibres, previous tests were done in order
them are to remain healthy, mortars and renders
to see the compatibility between resistant core and
must be as compatible as possible with the stones
inorganic matrix. Flat specimens (fig. 6) were made
they live with. In this context, cement based mortars
for tensile test and textile reinforced mortar prisms
are not recommended for this application.
(4 × 4 × 16 cm) for bending test.
– Resilience or flexibility to allow building deforma-
Based on the test results the following comments
tion without causing substantial wall cracking. In
can be done:
this context, lime binders/mortars offer significant
advantage over other types of binders, particularly – The joint mortar-textile is highly dependent of the
the cement based ones, in that it remains slightly application procedure.
flexible, even when set and will allow the walls – Compared with glass-bitumen-coated textile, basalt
to move without developing large cracks. Another fibres grid has less geometrical stability but a bet-
important characteristic of lime mortars, when com- ter adherence fibres-mortar because of a major wet
pared to other types of mortars, is that it can effect of the yarns by the mortar.
self-repair/heal fine cracks. This occurs as a result – Ultimate failure under tension for basalt textile is
of the rainwater slowly depositing fresh calcium less ductile than for glass-bitumen-coated textile but

854
Figure 7. Basalt fibre anchor device.
Figure 8. Basalt fibre anchor device.
in any case, TRM have a more ductile failure than
FRP.
– The maximum load, both in tensile and flexural tests,
is increased by the presence of the textile. Once
the crack appears, the specimen is able to continue
holding more load.
The knowledge of the cracking process is of crucial
importance in calculating the load-bearing capacity,
the deformation behaviour and the limiting values in
designing the serviceability. The cracking distance and
the crack width are determined by the reinforcement
and bond characteristics. Cracking is determined not
only by the stress but also by the bonding action
between the textile reinforcement and the mortar
matrix.
One of the main differences between short fibre
reinforced concrete (FRC) and TRM is the design phi-
losophy at ultimate load. FRC obtains its ductility from
fibres to be pulled out of the matrix. For FRC the pull
out load has to be always lower than fibre strength.
TRM in contrast obtains its ductility from the fibre
breaking strain.
3.3 Anchor devices
For the case that used mortars will not meet the
required minimum properties, because of the low
quantity of cement component, an anchoring tech-
nique of the TRM system on the substrate will be
applied. The anchorage system has to be simple, easy
to install and low cost. Fibre anchors have already
been developed and have been used extensively in FRP
855
Figure 9. Two stages of the construction.
systems (San José, 2007). Testing prove that similar
fibre anchors can also be used with a TRM system.
At a first stage, these anchorage systems were
tested jointly with the TRM for the evaluation of bond
strength by two different experimental approaches:
pull-off test and shear test, by using concrete cube
specimens such as reference material.
Before the pull-off testing, the free external fibres
were combined together in a yarn form, which was sat-
urated with an epoxy resin to form a plate as illustrated
in fig. 8, left. Whereas, for the shear test, the free ends
of the textile were enfolded with a piece of rubber, and
then enclosed into the grips of the testing machine, as
shown in fig. 8, right.
Generally, all pull-out tested specimens failed by
rupture of the fibre anchors, while the textile and the
adhesive joints still intact with the concrete substrate.
Concerning shear test, specimens failed by fracture
of the adhesive joint at different load levels, or rup-
ture of the basalt textiles. On the contrary, no failure
was observed in any of the anchor devices employed.
The following step will be the validation and adapta-
tion of this proposed anchorage system with decayed
substrates, usually low strength mortars.
3.4 Wall specimens construction and
strengthening
Nine wall structures of 200 × 200 × 30 cm were con-
structed by expert masons in Heritage interventions,
with the same configuration of the historic original
ones (Valluzzi, 2004). They are intentionally deflected,
a common structural damage in these elements, and
they have an inner core made of low quality mortar
and rough-cut sandstone pieces (remains from rough-
shaping of the stones), poured between the two exter-
nal layers, like real three-leaf stone masonry walls.
Walls foundations were built on steel permanent
moulds (length 2 m, wide 60 cm and high 30 cm).
These metallic bases were filled in with bigger stones
than the rest of the walls units and a more competitive
mortar (1-0,5-5). The intention of it is to reproduce the
natural foundations of the real walls, maintaining their
boundary conditions.
 Table 3. Strengthening solutions.

Wall
specimen TRM Transversal tying

M1 No No
M2 1 layer No
M3 2 layers No
M4 1 layer Yes
M5 2 layers Yes
M6 No No
M7 1 layer No
M8 2 layers No
M9 2 layers Yes

Figure 10. Wall specimen. Irregular masonry leaf.

Figure 11. Details of the strengthening procedure.
Ashlar masonry leaf is made of perfect prismatic
sandstone units (18 × 9 × 9 cm) bonded by bed joints
of poor mortar. Rubble masonry leaf is made of irregu-
lar stones, with a maximum size of 35 cm, and bed and
head mortar joints. The space between both leaves was
filled up with mortar and irregular stone fragments.
The walls were built in three stages, halting their
construction to three successive heights, because of
the fall risk due to the low strength of the used mortar.
The shape of the cross section was obtained with
auxiliary guides. This way, all the walls have similar
external geometry. The first meter is perpendicular to
the ground. From that point, the wall begins a curve,
reaching a horizontal deflection of 5 cm at the top.
Connection between external leaves is obtained
through the mortar internal core and some key stones.
These are ashlar units placed tangentially to the plane
of the wall, without appearing on the irregular masonry
leaf (fig 9, left). The last 4 walls have a narrower layer
of inner filling mortar and more stones keys than the
other 5.
After five months of the building date, a finishing
mortar layer, jointly with two different strengthening
solutions, was applied to the walls:
– Transversal sewing with FRP bars. The main
scope of this technique is the improvement of the
connection between the leaves and the consequent
reduction of the transversal deformations. Six low
modulus CFRP (Ø7.5 mm) bars were used for each
wall (around 1.5 bars per m2 ), inserted into holes
(Ø12 mm) drilled approximately at 1/3 of the height
from the bottom and the top, and at 1/4 of the
width from the left and the right sides of the wall.
The holes were executed through corresponding
irregular masonry mortar joints, without crossing
completely the ashlar masonry in order to not affect
the external appearance.
– Surface treatment with TRM on one side. In some
cases, the render mortar layer at the irregular
masonry face has been reinforced in one or two lay-
ers with the technical basalt textile. Furthermore,
the TRM is anchored to the wall by means of fibre
anchor devices (around 10 per m2 ). The thickness
of the strengthening layer is less than 2 cm. In case
of application on an even surface, TRM thickness
could be just a few millimetres.
A summary of the interventions is presented in the
above table.
3.5 Test setup
The nine wall structures were tested under vertical
compression load, by means of two hydraulic jacks
and a metallic frame, under static conditions.
The frame is composed of two metallic groups of
welded steel profiles, under and below the wall. The
first group was positioned over the two jacks and it
was rigidly connected, by means of four steel rods, to
the second group at the foundation of the wall. There
are additional steel profiles between the jacks and the
top part of the wall. The load is transmitted through a
2 cm mortar layer to avoid stress concentrations.
The specimens were instrumented with 8 horizontal
and 8 vertical inductive displacement transducers and

856

Figure 12. Linear displacement transducers arrangement.
Figure 13. Strain-gauges arrangement.
various strain gauges placed on the reinforced mortar
and the visible face of some ashlar units. Two load
cells measured the applied load. A data logger was
used to register the data of the referred sensors and the
time. In addition, the crack pattern was noted manually
for each load level and a video camera registered the
deformations in the cross section of the wall.
The compression was applied by successive load
steps. The time between two steps was around 5
minutes. During this period, the deformations and
cracks were examined and registered. In this way, the
complete test took from 90 to 120 minutes.
Due to the size of the specimens, high load val-
ues and the brittleness of the masonry, careful safety
measures were adopted. Therefore, the tests were per-
formed by placing a scaffold box around the test
frame.
857
Figure 14. Buckling and disconnection of the leaves.
3.6 Expected failure modes
The ultimate load is related with the failure mode.
With this test configuration it is difficult to generate a
local crushing zone because punctual loads were not
applied. Exhaustion of the material is also improbable.
The expected structural failures are the following:
– Swelling up of the outer leaves due to the poor con-
nection and the low strength of the inner filling
mortar.
– Buckling due to insufficient union of the leaves or
eccentricities.
– Overturning of the head or the total the wall. These
phenomena can appear by two causes, mainly: hor-
izontal loads that generate a torque respect to the
starting of the walls and problems of lying of
foundations of the walls.
– Cracking by local tensile stress. In case of foun-
dations in good conditions and leaves properly
connected, the horizontal components of external
loads can develop a stress state with tensile zones in
the masonry. The low tensile strength of the struc-
tural element will be improved by the reinforced
mortar, restricting the propagation of cracks at the
same time.
3.7 Results
At the moment of writing this proceeding, the test cam-
paign is in course. Four specimens have been tested,
including the reference wall of each series. From the
test results, the following comments can be done:
– Ultimate loads correspond quantitatively with the
expected values. However, obtained loads are much
higher than the compression strength provided by
standards and traditional equations.
– Sewing with FRP rebar procedure has not been
correctly executed. The failure mode in all the struc-
tures was originated from a disconnection of the
ashlar masonry leaf from the rest of the wall. This
indicates that the length anchorage in the ashlar
units is not enough to generate a benefit. On the
 200
Strain gauges on the render mortar (West face) capacity, improvement of the failure mode, increase
of the ductility, etc).
150 – A next generation FEM model will be calibrated.
This model will consider the typology of the stone
100
walls, the connection or adherence between leaves
Load (KN)

Compressed
50
Tensioned
and the presence of reinforcements like the applied
ones.
0
-200 -150 -100 -50 0 50 100 150

-50
Microstrain 4 CONCLUSIONS
Gauge1 Gauge2 Gauge3 Gauge4 Gauge5
The effectiveness of TRM as a means of strengthening
Figure 15. Micro strains of mortar face in Wall 1. stone masonry walls is investigated in this study. The
following conclusions are stated from the concluded
works:
Ultimate axial load
250 – The FRP, in its different formats (strips, rods, sheets
or sheets), has probe to be a reasonable solution
200
for the reinforcement of masonry structures. How-
ever, actual trends are focused in a more compatible
150
technique (TRM).
Load (kN)

100
50
0 masonry have demonstrated a certain improvement
M1 M5
Wall specimens
Figure 16. Ultimate compressive load.
other hand, there are a lot of cases where architec-
tural requirements do not allow to cross or to even
partially bore the stone units. In these situations a
different approach must be carried on, placing the
rods into the mortar joints and using anchor devices.
– Strain gauges on the face covered with mortar have
revealed positive deformations in certain zones of
the wall. The presence of tensile tensions is not
easy to locate by conventional methods when the
structure is damaged or there are structural defor-
mations that affect geometry. These tensile loads
can be partially assumed by the textile of the TRM.
– The combination of surface treatment withTRM and
sewing with low modulus CFRP rebar has allowed
to increase the ultimate load a 13% for a wall of the
first typology (M5) respect the reference wall (M1).
Complete results and final conclusions of this
experimental investigation are expected at the begin-
ning of 2008. The analysis of the data will provide high
value information.
– A better understanding of the behaviour of the
masonry, specially the natural stone walls, and
the capacities of the promising strengthening tech-
niques previously described: surface treatment with
TRM and sewing with rebar (increase of the bearing
– Materials cost of the intervention is significantly
reduced. Lime-cement matrix is between 6 and 10
times cheaper than epoxy resins. Similar researches
(Triantafillou & Papanicolaou 2005) on regular
M6 M9 of the mechanical properties, even with low cost
fibre textiles.
– In order to design the anchorage solution, it is nec-
essary to consider the substrate material nature, its
conservation status, the load history, the aesthetic
requirements and the reversibility.
– By the application of TRM solution, reinforced
elements can be obtained without increasing sig-
nificantly the original weight, and even without
modifying their external appearance.
– The strengthening intervention could be done while
the structure is in service.
ACKNOWLEDGEMENTS
This research is being financed by EU Commission
through the OPERHA 517765 contract and the Basque
Government through the SISMU contract.
REFERENCES
ACI Committee 440. 2002. Guide for the design and construc-
tion of externally bonded FRP systems for strengthening
concrete structures. ACI 440.2R-02.
Bakis, C.E., Bank, L.C. & Brown, et al. 2002. Fiber-
reinforced polymer composites for construction – State-
of-the-art review. Journal of Composites for Construction:
6(2), 2–73.
Casareto, M., Oliveri, A. & Romelli, A. 2002. Strengthening
of masonry: opportunities and challenges in the use of
composites Center for Infrastructure Engineering Studies
(CIES) – University of Missouri – Rolla, USA.

858
CEB-FIP. 2001. Externally Bonded FRP Reinforcement for
RC Structures.
Monti, G. & Santini, S. 2002. Reliability-based calibration
of partial safety coefficients for fiber-reinforced plas-
tic. Journal of Composites for Construction Vol 6, N◦ 3:
162–167.
San-José, J.T., García, D., Garay, A. & Castillo, J. 2007.
Novelty FRP strengthening systems applied to different
substrates: analysis of the anchorage behaviour. FRPRCS-
8. Ed. Prof. T.C. Triantafillou, University of Patras,
Greece.
San-José, J.T., Roca, P., Prendes, P., Mieres, J.M. & Garay, A.
2006, Anchorage study of FRP strengthening systems
859
applied to different substrates. Structural Faults & Repair.
London.
Triantafillou, T.C. 2001. Seismic Retrofitting of Structures
Using FRPs. Progress in structural Engineering and
Materials, 3 (1): 57–65.
Triantafillou, T.C. & Papanicolaou, C.G. 2005. Textile Rein-
forced Mortars (TRM) versus Fiber Reinforced Polymers
(FRP) as strengthening materials of concrete structures.
FRPRCS-7, ACI SP-230: 99–118. Kansas City, USA.
Valluzzi, M.R., da Porto, F. & Modena, C. 2004. Behaviour
and modeling of strengthened three-leaf stone masonry
walls. Materials and Structures Vol 37: 184–192.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

The effect of friction joint and Gongpo (bracket set) as an energy

dissipation in Korean traditional wooden structure

J.K. Hwang
Department of Traditional Architecture, National University of Cultural Heritage, Korea

S.G. Hong
Department of Architecture, Seoul National University, Korea

N.H. Kim
Korea Bridge Design and Engineering Research Center

Y.W. Lee
Department of Architectural Engineering, Kunsan National University, Korea

S.J. Jeong
Department of Architecture, Hannam University, Korea

S.J. Joo
Senior Engineer, Korea

ABSTRACT: The main objective of this paper is to investigate structural behavior of Korean traditional wooden
structures. The main characteristics of traditional wooden structures in East Asian countries are considered as
unique techniques of joint without fastener and relying on friction between components to transfer interaction
forces.
As one of temple structures Bongjeong-sa built in the early stage of the tenth century first is considered as a
typical example of Korean traditional wooden structures. It is classified as a national treasure in Korea. It was
rebuilt several times for restoration. Its basic structural system consists of a column and lintel beam connected
by friction joint and Gongpo (bracket set). The Gongpo system is a unique system of brackets placed on top of
columns to provide additional support to beams or overhanging eaves.
A one-third scale model with the roof-dead load of 46,710 N is designed and prepared on a steel sliding table
of 5 m by 4 m to perform a series of structural tests in a laboratory. Using this model the characteristics of lateral
and twisting displacement on the top of columns of the frame and the dynamic effects of the Gongpo on the
global behavior of whole systems are investigated. It is concluded that the friction joint of column-connecting
beam serves as an energy dissipation device reducing the impact of earthquake,but the Gongpo is not efficient
to dissipate the energy as we have guessed.

1 INTRODUCTION 2000 and it was classified as the national treasure of

The main building of Bongjeong-sa(temple) located in
the Andong-city in Korea was built in the early stage of
the tenth century first. It was rebuilt several times from
first construction to now. In 1809, it was reconstructed
on a large scale and a large part of the appearance of
the present time (Fig. 1) is the result of that reconstruc-
tion. Until several years ago, the partial repair had been
advancing for roof, decorative painting and wooden
floor. The most recent total repair was performed in
Korea [1].
The main building of Bongjeong-sa(temple) was
composed by three frames in the front (13,312 mm)
and the side (8,721 mm) (Fig. 2). The column is
just placed on a foundation stone without the use of
mechanical fasteners. The frame is composed by col-
umn and connecting beam. The connecting beams are
connected by rake joints located over the columns, tak-
ing advantage of the reinforcement provided by the sur-
rounding slot (Fig. 3). The architraves which are placed

861
Figure 1. Exterior front view.
Figure 2. Plan.
Figure 3. Connecting beam and slot.
bracket sets are located on column-connecting beam
joints. All bracket sets, composed of soro (blocks) and
cheomcha (arms), are located on architraves (Fig. 4).
They are connected with one another by tie beam and
the main role of those is the one interface columns
with beams and the large percentage of the roof load
about 2,500,000 N was transferred to foundation stone
862
Figure 4. Soro(block) and cheomcha(arm).
Figure 5. Building layer.
by those. The elevation of a structure was made up of
the four building layers: base, columns and connecting
beams, bracket sets and roof (Fig. 5).
In this research a dynamic test is performed to eval-
uate vibration characteristics of the main building of
Bongjeong-sa (temple) in the laboratory. For this pur-
pose, a one-third scale model with the roof-dead load
of 46,710 N is designed and prepared on a steel sliding
table of 5 m by 4 m. From the result of the vibration
test, we evaluate natural frequency, transfer function
between base layer and damping ratio of the building.
2 EXPERIMENTAL PROGRAM
2.1 Test model and sliding table
To perform vibration test at the laboratory, a one-third
scale model of the main building which has the ele-
ments of 3119 was built. The dimensions of this test
building are 2910 mm wide and 4450 mm long. To
design this test model, Autodesk inventor 11 was used
to draw whole elements. Although the soil walls, win-
dows and doors exist between the columns, to exclude
the effect of other variables other than the joints of
columns and connecting beams, they were not con-
tained in this test model. Also, the top part of roof
was not contained in this test model and replaced by
weight. Korean red pine was used to make test model

Figure 6. The dimension of the test model.
Figure 7. The shape of the test model.
Figure 8. The steel plate of roof and sliding table.
and it kept the percentage of moisture under 17%.
The steel plate of 46,710 N was used to simulate the
weight of roof (Figs 6–8).
To perform the vibration test, the steel sliding table
of 5.5 m by 4.5 m was made. It was slid in the one-way
direction on the LM guide and designed to be able to
rotate to perform the test of orthogonal direction. It
was slid in the hydraulic dynamic actuator.
863
Figure 9. Sliding table and actuator.
Figure 10. The sensor location on architrave.
2.2 The location of sensor
The location of the sensors to measure the accelera-
tion and displacement are located in Figure 10–11. The
total numbers of sensors are sixteen. Sensor points 1,
2, 11 and 12 are located on an architrave, the top of
column and connecting layer, and measure the accel-
eration of Y-direction. Sensor points 3, 4, 9 and 10 are
located on an architrave and measure the acceleration
of X-direction. Sensors points 13 and 14 are located
on steel sliding table and measure the acceleration of
X-direction and the displacement of X-direction. Sen-
sor points 15 and 16 are attached at architrave and
measure the displacement of X-direction (Figure 11).
Sensor points 5 and 6 are located on bracket set and
measure the acceleration of X-direction. Sensor points
7 and 8 are located on the bracket set and measure the
acceleration of X-direction.
3 VIBRATION TEST
3.1 Random vibration
It is actuated the sliding table as random signal which
contains the frequency of from 1 Hz to 3 Hz to figure
out the natural frequency of the test model. In process
of the random vibration experiment, we observe that
the natural frequency of test building changes accord-
ing to the magnitude of vibration. To confirm the

Figure 11. The sensor location on bracket.
2
Force (kN)
0
- 40 - 30 - 20 - 10 0 10 20 30 40
-1
-2
-3
-4
Displacement (mm)
Figure 12. Force-lateral displacement relationship of joint.
variation of natural frequency, the random vibration
test was performed for the RMS amplitude of dif-
ferent kinds. The natural frequency of X-direction on
the connecting beam, acceleration 3 in Figure 10, and
the results of natural frequency are shown in Table 1.
According to natural frequency results, the natural
frequency of test building could confirm to change
according to the magnitude of vibration amplitude.
The variation of natural frequency could possible
when weight or stiffness varied. In case of this test
model, the weight did not change at all. So, we could
suppose that the lateral stiffness of whole frame var-
ied. The lateral stiffness of Korean traditional wooden
structure is composed for the most part by the stiffness
of column-connecting beam joints. The research result
for joint stiffness shows that the stiffness of column-
connecting beam joints is varied by restraint force and
the magnitude of lateral displacement (Fig. 12). The
stiffness of joints is the linear in the case of small dis-
placement and changes if the displacement of joint is
gradually enlarged. By this reason, the lateral stiffness
864
Table 1. The variation of natural frequency for the RMS
amplitude of random signal.
The RMS amplitude of
random signal (mm) Natural frequency (Hz)
0.2 2.606
0.3 2.576
0.5 2.258
1 1.962
Figure 13. TF from ground acc. to connecting beam acc.
of whole system changes according to the variation of
joint displacement.
The accelerations and displacements are measured
simultaneously during 180 seconds with a sampling
frequency of 100 Hz. Then, stationary portions of
records are selected to calculate the transfer function
that could evaluate about the predominant frequency
of the structure. To evaluate the variation of the pre-
dominant frequency according to the magnitude of
vibration and the effect of vibration transmission
between the layers, the transfer functions for accel-
eration between sliding table and architrave were
calculated. The results of the calculation of transfer
functions are shown in Figure 13.
The one of the main subjects of this research investi-
gates the energy dissipation characteristics of Gongpo.
To this end the transfer functions for acceleration
between sliding table and Gongpo are calculated. The
transfer functions of architrave and Gongpo are shown
in Figure 14. We can confirm that the transfer functions
of architrave are identical with the one of Gongpo.
Therefore, we can confirm that the energy dissipation
effect is in the joint of column and connecting beam
and Gongpo do not perform the role of energy dis-
sipation. Also, we can confirm that the predominant
frequencies shift to the left according to the increase of

Figure 14. TF of on connecting beam and Gongpo.
Figure 15. TF of on connecting beam for sine wave.
vibration amplitude and the value of transfer function
become smaller in Figures 13 and 14.
3.2 Sine wave vibration
To confirm exactly the shift effect of predominant fre-
quencies and the variation of transfer function value,
sine waves that the amplitude changes from 0.2 mm to
2.5 mm were actuated and transfer function was calcu-
lated. The results of transfer function by sine waves are
shown in Figure 15. In Figure 15, the predominant fre-
quencies according to the amplitude of sliding table are
shifted. If the amplitude of sliding table is larger than
1.0 mm, the transfer functions show similar values.
3.3 Impulse vibration
A free vibration experiment is performed to determine
damping ratio and verify the variation of damping ratio
by the magnitude of initial impact displacement. The
initial impact displacements of 5, 10, 15, 20, 25, 30 mm
are set to vibrate (Its duration time was 0.1 sec). The
865
Figure 16. The free vibration record of impact displace-
ment.
Table 2. Damping ratio by impact.
X-direction impact
displacement (mm) Damping ratio (%)
5 8.94
10 10.02
15 10.19
20 10.30
25 10.19
30 10.24
free vibration records of displacement at architrave,
displacement 15 in Figure 10, presented in Figure 16.
The damping ratios arecalculated by the logarithmic
decrement process and presented in Table 2. From
table 2, we observe that if the impact displacement
of sliding table larger than 10 mm, the damping ratio
shows similar values a little bit more than 10%.
4 CONCLUSION
The energy dissipation effect is in the joints of columns
and connecting beams but Gongpo do not show effi-
cient energy dissipation. The predominant frequencies
shifted to left according to the increase of vibration
amplitude and the values of transfer functions became
smaller. If the sine waves amplitude of sliding table
are larger than 1.0 mm, the transfer functions shows
similar values. If the impact displacement of sliding
table is larger than 10 mm, the damping ratio shows
similar values a little bit more than 10%.
ACKNOWLEDGEMENT
This research was supported by the Architectural
Research Division of NRICH (National Research
Institute of Cultural Heritage). The authors would like
to express their gratitude to the many researchers of
the Architectural Research Division.
REFERENCES
[4] Y.W. Lee, S.G. Hong and J.K. Hwang, Capacity of
[1] Andong-city. The report of total repair for the Main
Building of Bongjeong-sa(temple), 2004.
[2] Anil K. Chopra. Dynamics of Structures. Prentice-Hall.
2001.
866
[3] S.J. Joo, S.G. Hong, N.H. Kim, Y.W. Lee, S.J. Jeong
and J.K. Hwang, Seismic Response Characteristics of
the Main Building of Bongjeong Temple, Proceed-
ings of COSEIK Annual Conference 2007, Seoul,
Korea, 2007.
Lateral Load Resistance of Dori-direction Frame with
Jang bu-connection in Traditional Wood Structure
System, Journal of the architectural institute of Korea,
structure & construction, vol.23 n.2, 2007.02.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Optimization of cutting processes in archaeological sites

C. Cennamo
Seconda Università di Napoli, Napoli, Italy

B.M. Chiaia & E. Masoero

Politecnico di Torino, Torino, Italy

S. Scaini
DEXPLO Srl, Parma, Italy

ABSTRACT: Cutting processes in archaeological sites are critical operations which can irreparably damage
precious artefacts. The damage can consist either of irregular and wide cutting grooves and of chemical deterio-
ration triggered by flushing liquids. Therefore, an holistic approach to the optimisation of cutting operations is
an important issue which requires an accurate knowledge of the ancient base materials and theoretical models
which permit to manage the interaction between artefacts and cutting tools. According to this framework, the
results of the chemical analyses and of the mechanical tests performed on the concrete constituting an ancient
Greek Gymnasium are firstly reported in this paper. Then, an essential mechanical model of the cutting process
is proposed. It relies on an energy parameter, the “cutting strength”, which must be the target of optimisations
and can be easily post-calculated starting from data collected on site. Finally, an empirical model for the tool
wear is also proposed.

1 INTRODUCTION: HISTORICAL
BACKGROUND AND MOTIVATIONS

The archaeological complex, that was discovered in

December 2003 in Piazza Nicola Amore during the
excavation works for Line 1 of the underground rail-
way system, was identified by archaeologists as a
Greek age Gymnasium. It represents the final part of
a long path that the ancient Greek athletes, the Lam-
padodromi, ran through in honour of Parthenope, the
goddess symbol of Naples. Later, in the Roman age,
emperor August allowed the winners of his Neapoli-
tan Isolimpic Games, the Sebastà, to have their names
engraved on its arcade marbles, where they can still
be read.
The athletics path started from the present site of
Piazza Municipio (where another important archaeo-
logical site was found) and ended in the present Piazza
Figure 1. The athletics path.
NicolaAmore, across the harbour area, the commercial
and cultural centre throughout historical ages (Fig. 1).
The ruins show different types of masonry, with the building materials. In facts tuff blocks, evidently com-
typical features of that ancient age (e.g. core walls ing from Poggioreale (Naples) quarry, were widely
made in opus reticolatum technique, masonry perime- used in the foundations, road pavements and masonry
ter in opus latericium and other masonry artefacts walls. On the other hand, the good cementing proper-
made in opus quadratum and opus cementicium). ties of pozzolan permitted to obtain a mortar by mixing
Tuff and pozzolan (i.e. a volcanic sand named it with lime and sand. This mortar, mixed with clay and
after Pozzuoli, Naples) were the most widely adopted tuff aggregates, constitutes the concrete found on site.

867
Figure 2. Cement core wall in opus reticolatum (left side)
and street paved with tuff (right side).
Figure 3. Features of the Greek Gymnasium after the
removal of some blocks (left side) and one of the adopted
chain saws (right side).
Several theories in the past aimed at demonstrat-
ing (Giuffrè 1991) that the mortar looses its bonding
capacity with ageing and that the significant tensile
strength that the Greco-Roman structures undoubtedly
show, was the merit of the ancient builders’ ability of
setting stones so masterly (that means the “art rule”),
and not of the mortar adopted. These theories seems to
be partially contradicted by the ruins found in Piazza
Nicola Amore. In facts here, even if at some locations
a worse quality of mortar seems to have been used,
the mortar appears mostly homogeneous and com-
pact. This is probably due to the fact that the ruins
have been buried for centuries about 15 underground,
which prevented their contact with the atmosphere.
Since Italian Archaeological and Monumental Her-
itage Superintendence decided to obtain compact
blocks from the Greek Gymnasium inhomogeneous
masonry, cutting operations had to be performed. The
blocks had to be moved elsewhere in order to be
reassembled on site after the tunnel excavation.
The cutting operations in this archaeological site
appeared particularly critical because of the great
historical relevance of the found artifacts and the
high compactness of the ancient concrete constituting
most of them. In facts, this latter made very ener-
getic cutting processes necessary, which could produce
irregular and wide cutting grooves and, consequently,
compromise the successive reassembling operations.
Moreover, the negative effects of friction between the
adopted chain saws with impregnated diamonds bits
and the base material could have been enhanced after
868
Figure 4. Results of the thermogravimetric analyses on the
mortar specimens.
the decision to a priori prohibit the use of flushing
liquids, because of the possible risks that water could
infer to such ancient concretes.
Therefore the requirement became urgent to charac-
terize the cutting process according to every specific
material and to specific archaeological situation. The
minimization of vibrations and required energy rep-
resents the target of an optimisation that takes into
account all the mechanical parameters of saws and
masonry materials.
2 ANALYSES ON THE ANCIENT CONCRETE
2.1 Chemical analysis
The following analyses were performed at the Depart-
ment of Material and Production Engineering (Uni-
versity of Naples), on two blocks (called A and B) of
ancient concrete coming from the excavation site pre-
viously described. Each block was made of tuff stones
embedded into a mortar matrix. For each block, one
sample of tuff and one of mortar were extracted.
The mortar is made of a cementing matrix with
thin volcanic aggregates, calcium carbonate grains and
gross tuff grains, produced while cutting the bigger tuff
stones. Nevertheless, the two samples of mortar MA
and MB were extracted trying to exclude as much as
possible the presence of calcium carbonate and tuff
grains.
Thermogravimetric analyses have been performed
on the mortar samples by means of an equipment
NETZSCH STA 409 PC/4/H Luxx. The curves in
Figure 4 show the percent loss of mass as a func-
tion of temperature. The obtained curves are typical
of cementing matrixes made of a suspension of slaked
lime and pozzolan. In detail, the 4% mass loss cor-
responding to the temperature interval 120–600◦ C is
due to the decomposition of hydrate phases produced
after the reaction between pozzolan and quicklime.
On the other hand, the 6% mass loss corresponding
Table 1. Average mineralogical composition of tuff samples
TA and TB.

Smectite Feldspars Phillipsite Chabazite Analcime

% % % % %

10.4 15.6 40.0 22.5 2.0

to the temperature interval 600–800◦ C is due to the

decomposition of calcium carbonate grains, whose
presence is symptomatic of inaccurate slaking.
Moreover, acid-insoluble residue test indicates that
about the 70% in mass of each sample is composed Figure 5. Intact samples A1 and B.
by insoluble quartz and silicoaluminates, present into Table 2. Geometrical features of the specimens and per-
the pozzolana, while the remaining 30% by slaked formed tests.
lime grains and hydrate phases, produced by the reac-
tion between pozzolana and lime. Therefore, the actual Base Height
composition of the mortar is one part of quicklime Sample mm mm Test
every three parts of pozzolan. For these tests, a solution
of water and hydrochloric acid was used. A1 φ54 80.5 Ax. C.∗
Finally, X-ray diffractometric analyses were carried A2 φ54.5 75 Ax. C.
B 180 × 170 210 Ax. C.
out on the pulverized mortar and also on its acid- C1 φ54.5 31 Brazilian
insoluble residue alone, using a Philips PW1710 (rad. C2 φ54.5 33 Brazilian
CuKα) X-ray diffractometer. The results permitted to
describe the mineralogical composition of the mor- ∗
Ax. C. stands for “axial compression” test.
tar samples and confirmed the conclusions suggested
by the thermogravimetric tests. In facts, calcite was Table 3. Test results.
found as a product of the carbonatation of calcium in
excess and sanidine as the crystalline phase of poz- σu εu E Wf σu,t
zolan Moreover, the presence of chabazite indicated Sample N/mm2 ‰ N/mm2 N/mm2 N/mm2
that tuff grains were also present into the concrete.
The tuff samples, namely TA and TB, were stud- A1 6.7 6.9 ∼1500 ∼0.45 –
A2 7.8 11.9 ∼1300 ∼0.36 –
ied only by means of diffractometric analyses whose
B 4.3 16.7 ∼400 ∼0.3 –
results are collected in Table 1. This analyses permitted C1 – – – – 0.62
to classify it as Neapolitan Yellow Tuff, produced after C2 – – – – 0.65
a volcanic eruption occurred at Campi Flegrei 15,000
years ago.
mortar alone. All these tests where performed by
2.2 Mechanical tests
means of a displacement controlled testing machine.
The mechanical tests shown in this section have been For what concerns the axial compression tests, the
performed at Politecnico di Torino. Five samples of samples showed peculiar failure mechanisms. In sam-
pozzolanic mortar were used: two cylinders, namely ples A1 and A2 there was firstly debonding between
A1 and A2, with respectively one and two large clay mortar and large aggregates and than a wide vertical
inclusions at the mid of their height, one parallelepiped splitting crack. Sample A2, which looked more com-
(B) with large tuff inclusions and two squat disks (C1 pact than A1, was also affected by larger damage inside
and C2) without large inclusions. the clay aggregate which however turned out to be
Several small calcium carbonate and tuff grains stronger than the mortar.
were present into both samples. Samples A and C On the contrary, the large tuff aggregates of sample
were extracted by a unique block with good compact- B were so deformable and weak that many vertical
ness (specific weight ∼1500 kg/m3 ), differently from cracks did not cross them but bypassed up to rupture.
sample B whose quality immediately looked worse The mechanical tests’ results suggest that the
(specific weight ∼1300 kg/m3 ). strength of the pozzolanic matrix, which is generally
The axial compression tests permitted to find values comparable to that of a low resistance modern cement,
of the normal compressive strength σu , the correspond- is reduced by the widespread presence of almost pul-
ing ultimate strain εu , the Young’s modulus E and the verized tuff and calcite grains, as well as of large
rupture energy Wf of the concrete while the Brazilian clay and tuff inclusions. The negative effects of such
tests provided the indirect tensile strength σu,t of the inclusions are also more relevant for what concerns the

869
Figure 6. Stress vs. strain curves obtained from axial
compression tests.
Figure 7. Samples A1, A2 and B after rupture.
Young’s modulus. On the other hand, these inhomo-
geneities are probably also responsible for the ductile
post peak behaviour.
3 A MODEL OF THE OPTIMAL CUTTING
PROCESS
In this section, an essential model to describe the
a regime cutting process into the ancient concrete
is proposed. This model will introduce an important
parameter, the “cutting strength”, and at the end of this
section it will be clear that this parameter itself must
be the target of the optimisation of cutting processes.
Consider the cutting operation performed by the
workman with the chainsaw in Figure 8 where a
schematization of the process is contextually pre-
sented. According to this idealization, every active
tooth of the chain (black coloured in Figure 8) pen-
etrates the material with a velocity δ and follows the
rotation of the chain moving tangentially with respect
to it with a peripheral speed v.
Calling ϕ the angular velocity of the chain in radi-
ans per second, the number of rounds per second
can be simply obtained dividing it by 2π and, if L
is the perimeter of the chain measured in meters, the
tangential velocity of the tooth v is given as:
where Req is the radius of a circular chain saw with the
same perimeter.
Another important consideration is that the number
of active teeth lying on the toe (i.e. where the curvature
870
Figure 8. Cutting operation (left side) and its schematiza-
tion (right side).
radius is small) is negligible with respect to the number
of those lying on the cutting front, which is almost
straight. According to this statement, we neglected the
contribute given by the previous ones so that the whole
cutting process can be considered as performed by a
straight chain saw.Actually the model that we are going
to present can be applied also to circular saws with
large radius since, in this case, the exact trajectory of
the single tooth can be approximated with a linear one.
Looking back to Figure 8, N is the normal force
which pushes the saw into the material (usually manu-
ally applied) while T is the tensile force that the engine
provides to the chain and which is balanced by the
active teeth on one side and the workman (or a sup-
porting structure) on the other. Moreover, let us call
nd the total number of teeth and ndl the number of the
active ones.
Thus, if s is the length of the gullet between two
consecutive teeth, then (Fig. 8):
where La is the length of the chain’s portion whose
corresponding teeth are active (i.e. the cutting depth).
Now let us make the following assumptions:
– the saw and its teeth are infinitely rigid with respect
to the material;
– the material is homogeneous in terms of elastic
coefficients and cutting strength (see further for its
definition);
– the axial shortening due to wear is the same for all
the teeth.
If such conditions are respected, the tensile and
normal external forces (i.e. T and N ) can be equally
distributed between the active teeth (excluding those
lying on the toe) obtaining:

Figure 9. Motion scheme of the active teeth.
Now let us adopt the following relation between the
force fn applied to an indenter and its penetration u
into the material (Fisher-Cripps 2000)
where k and α are constants depending on the base
material, on the indenter geometry and on the tool wear
state.
If the process is optimised (this reduces vibrations
ad provides regime regularity, reducing the risk of
damage to the cut objects) then δ and v must be con-
stants. This implies that every tooth always shows the
same penetration depth u∗ given as:
and that the trajectory of the active teeth is linear.
Moreover, Figure 9 shows that an active tooth moves
for δs in the horizontal direction while spanning the
gullet between two successive teeth s in the vertical
direction. Thus the penetration u can be kept constant
(u∗ ) only if:
The consequences of the violation of the above condi-
tion will be discussed later in this section.
871
Figure 10. Material removed during the vertical advance-
ment equal to s (left side) and particular of a tooth (right
side).
Calling ts the time requested for a vertical advance-
ment of s, in the optimal condition it can be easily
obtained:
so that:
On the other hand, since in the regime process the
engine provides a constant tensile force T to the chain,
the work We (s) done by the engine to move the chain
for a length s (i.e. to remove a prism of material of
depth u∗ and height La ) is:
where η is the efficiency of the cutting process.
Thus, recalling Equation 9, the power requested
We is:
Finally, if we define the cutting strength S as the
amount of energy requested to remove a unitary vol-
ume of base material, then the energy required to make
all the active teeth advance for a length s is:
where t is the tooth width.
The introduced parameter, the cutting strength, has
the dimension of a force per unit area exactly as a
plastic strength. Nevertheless, it is actually more than
the ultimate quasi-static compressive strength since it
takes into account also the energy dissipated by fric-
tion, heating, fragmentation, milling and depends on
the shape of the fracture mechanisms induced by the

Figure 11. Schematic representation of the optimal relations
between geometrical and mechanical parameters involved in
the cutting operations.
tooth which, in turn, depend on the geometry of the cut-
ter, on the activated friction, on the flushing liquid etc.
Recalling Equations 2, 11 and 13, from a simple
energy balance we obtain:
Moreover, by considering a linear penetration law
(α = 1 in Equation 5), the regime pseudo-friction
coefficient µ∗ can be defined as:
This latter equation shows that the ratio between the
externally applied loads inducing optimal cutting con-
ditions depends on the properties of the material (S
and k) and on teeth’s width (t).
Finally, combining Equations 10, 14 and 12, the
following formula can be obtained:
which puts in relation the penetration velocity, the
power applied by the engine (and its efficiency), the
cutting strength, the number of active teeth and their
geometrical features.
A reasonable hypothesis is that the power provided
by the engine is constant during the cutting process,
which results into the inverse proportionality between
the tensile force T and the velocity of the chain v.
In this case, if the a regime normal force N is
reduced, the material tends to push the active teeth
out according to Equation 6. This reduces the teeth
872
Figure 12. Out of regime cutting conditions obtained reduc-
ing (left side) or increasing (right side) the externally applied
normal force.
penetration u and, proportionally, also the required ten-
sile force T decreases (Equation 14). Consequently,
v increases until reaching a new a regime condition
when Equation 10 is verified again (of course, accord-
ing to Equation 16 δ does not vary). Equation 10 also
states that the penetration angle θ decreases when N is
reduced and, in a limit case, a very small normal force
could also turn the cutting process into an ineffective
polishing.
On the contrary, if the optimal normal force N is
increased, also the penetration grows and the veloc-
ity of the chain decreases. Nevertheless, it is known
that a deeply penetrated tooth produces bigger frag-
ments (Chiaia 2001, Carpinteri et al. 2003) and the
consequent vibrations can induce damage to the base
material, resulting in a zig-zag groove.
These deductions suggest that the optimal cutting
process, in terms of required energy and for a constant
applied power, can not be obtained by changing the
geometrical features of the cut (except for the groove
width t) or the number of teeth. Actually, the cutting
strength can be minimized by choosing the optimal
value of the penetration u∗ or, which is the same, the
optimal value of the applied normal force.
4 BACK ANALYSIS AND ESTIMATE OF THE
CUTTING STRENGTH
The relations obtained in the previous section are
now applied to the experimental data obtained by the
actual cutting operations performed on site.These have
been done using an Hydro Maxy Power Ics chain saw
with impregnated diamond bits pivoted onto a guide
rail. The adopted machine transmits to the chain a
power ηWe = 15 hp ≈ 11.3 kW, its rotational velocity

in unloaded conditions is ϕunl ≈ 5500 rpm, its maxi-
mum cutting depth is 60 cm and nd = 47 teeth with a
thickness t = 0 .8 cm lie on the L = 170 cm long chain.
This implies a gullet length s ≈ 3.6 cm and a peripheral
speed v ≈ 156 m/s ( see Equations 2 and 1).
The average penetration velocity δ recorded in situ
was between 0.5 and 1.1 cm/s. Assuming an active

Figure 13. Cutting process with guide rail and its
schematisation.
length La = 40 cm and applying Equations 2 and 16 the
cutting strength of the material can be obtained. The
values of S found in this way vary between 600 MPa
and 250 MPa. For a modern reference concrete cut in
wet condition S ≈ 400 MPa but it must be noted that
the dry conditions increase the cutting strength since
they enhance energy dissipations due to friction, heat-
ing, etc. Comparing the values obtained for S with
the ultimate strength and the elastic modulus of the
material (Section 2.2) it can be seen that its order of
magnitude is intermediate between them. Moreover, it
is significantly higher than the rupture energy obtained
on the tested sampled (Wf ∼ 0.4). This is probably
due to a size effect i.e. these collapsed because of
one or some macroscopic cracks, while the micro-
scopic destroying effect of the teeth is more similar
to micro-fragmentation.
The variability of the cutting strength depends either
on the composition of the sample and on the differ-
ent values of the applied normal force, according to
what has been pointed out at the end of the previous
section. Unfortunately, the lack of information about
the normal forces applied by the workman eliminates
the possibility to put into evidence the correlations
between N and the cutting strength.
Nevertheless, considering the mechanics of the cut-
ting process with a saw pivoted into a guide rail
(Fig. 13), it can be reasonably hypothesized that the
force F applied by the user is around 200 N (i.e.
approximately 20 kg) which implies a normal force
N = 250 N on the active length (more exactly, accord-
ing to the proposed scheme the normal load over the
active length should grow proportionally to the depth).
Since in the 40 cm long active area there are approx-
imately ndl = 11 active teeth (Equation 2), the normal
force per tooth can be found applying Equation 4,
which indicates fn ≈ 23 N.
Moreover, assuming a rotational speed of the chain
equal to ϕ = 1000 rpm and a linear penetration law
(i.e. α = 1 in Equation 5) and applying Equation 10, we
get an a regime penetration u∗ = 0.006 − 0.014 mm.
873
Therefore, applying Equation 6 the corresponding
values of the penetration stiffness k turn out to be
comprised between 1.77 and 3.8 kN/mm.
The vertical displacement wC induced at a point C
on the surface of a semi infinite elastic continuum by
a uniformly loaded square area b × b centered in C is
given by:
where ν (assumed to be ≈0.15 for the material of the
Greek Gymnasium) is the Poisson’s modulus of the
continuum.
Fixing fn = qb2 and comparing Equations 17 and 5
and assuming u∗ = wC we obtain:
which indicates that the elastic vertical displacement
wC is equal to the penetrations previously obtained if
the equivalent loaded area is b2 = 19 mm2 . This value
is smaller than the base area Aw of the tooth actually
penetrating the material (approximately 36 mm2 ). This
underlines the importance of plastic and dynamical
effects in cutting processes, which will be investigated
in further works.
5 TOOTH WEAR
As we already pointed out in Section 3, the penetration
stiffness k depends on the mechanical properties of
the base material, on the shape of the indenter and
on its wear state. Moreover, since penetration involves
plastic flow at the indenter’s tip as a first approximation
it can be said that the force fn required to obtain a given
a regime penetration u∗ is proportional to the base area
Aw of the indenter (Glowka 1989). With this hypothesis
and for a linear penetration law (i.e. α = 1) Equation 5
implies:
where k0 is the penetration stiffness in unworn condi-
tions, C1 is a constant to be determined experimentally
and δ indicates the cutting advancement.
The rate at which Aw grows with the cutting
advancement depends on the mechanical properties of
the base material as well as on several cutting opera-
tive parameters (Nabhani 2001). Moreover, experience
shows that it tends to an asymptotic value for large
δ. Therefore, the following asymptotic law can be
proposed:
where Amaxw is the asymptotic value of the worn tooth
base area (characteristic of the specified tooth) and
C2 is another constant which incorporates the effects
of all the relevant parameters quoted above and which
can be easily evaluated by directly measuring the tooth
base area at different cut advancements. Minimizing
this constant, increases the teeth’s lifetime.
By inserting Equation 20 into Equation 19 the
following wear dependent penetration stiffness is
obtained:
Differently from C2 , C1 is hardly directly deter-
minable, since the penetration law (Equation 19)
refers to dynamic conditions. Nevertheless, if a cut-
ting process at constant power We and normal force
N is considered, Equations 21 and 5 suggest that the
optimal penetration u∗ decreases with cutting advance-
ment, while Equation 16 shows that the speed δ
remains unvaried. This implies (Equation 10) a that
the rotational speed of the chain increases with wear.
6 CONCLUSIONS
The question of coexistence among the criteria for
preservation of old items and the technological
advancement in cities, is a very serious problem for
the historical centres of ancient European towns, and
it is emphasized when situations like the one in Piazza
Nicola Amore, bring the issue to the forefront. It is
therefore necessary to pay more attention, through reg-
ulations including the adaptation of machinery and
the development of new specific technologies. In the
specific case, the dry cutting processes performed on
the archaeological site found in Naples, dealt with an
extremely heterogeneous material. A model aimed to
the optimisation of the cutting process has been pro-
posed in Section 3. This model defines the optimal
conditions that can be reached during cutting processes
and underlines how these can be determined starting
from the external forces applied by the workman and
by the tool, given an opportune description of the base
material and of the cutting machines.
As previously shown, the key expedient of the
model is the introduction of the cutting strength param-
eter (energy required to remove a unitary volume of
base material) which has dimensions of compressive
strength, but is generally higher since it takes into
account all the dissipations related to dynamics (vibra-
tions, heat, fragmentation, milling, etc.). In this way,
all the sources of complexity in the cutting process are
collected inside this parameter and the model shows
that also when the cutting strength and the tool power
are constants, different conditions can be obtained by
874
varying the external driving forces, namely the pene-
tration force provided by the workman and the tensile
force into the chain provided by the cutting tool engine.
The back analysis performed based on the on site
operations, indicates a value of the cutting strength
for the ancient concrete S = 250 ÷ 600 MPa, which
is comparable with that of a modern concrete cut in
wet conditions. Even if the poor mechanical proper-
ties of the ancient concrete would suggest a cutting
strength much lower than that of a modern concrete,
the dry conditions of the cutting processes performed
in Naples, seem to justify the obtained results, due to
the good quality of the ancient mortar.
Moreover, scale effects related to the size of the
teeth (i.e. a smaller scale involves a higher ultimate
quasi-static strength) have been put into evidence.
The small amount of experimental data coming
from the site and from laboratory tests (only a small
number of samples was sacrificed) still leave some
open issues, like the influence of dynamic effects and
the roles of topological heterogeneities into the base
material.
In conclusion, the optimisation of the cutting pro-
cess allows to reduce vibrations to a great extent and
to extract straight blocks of material from the masonry
structures, minimizing the groove thickness so that the
structures could be reassembled on situ in excellent
conditions. So, this performance could be used in case
the disassembly of a historical situ is really necessary
to the aim of its conservation, specifying every time
the mechanical model according to the characteristics
of the material under consideration.
REFERENCES
Giuffrè A. 1991. Letture sulla meccanica delle murature
storiche. Edizioni Kappa, Roma.
Wojtanowicz A.K. & Kuru E. 1993. Mathematical modelling
of PDC bit drilling process on single cutter mechan-
ics. ASME Journal of Energy Resources Technology 115:
247–256.
Fischer-CrippsA.C. 2000. Introduction to contact mechanics.
New York USA: Springer Verlag.
Chiaia B.M. 2001. Fracture mechanisms induced in a brittle
material by a hard cutting indenter. International Journal
of Solids and Structures 38: 363–371.
Carpinteri Alberto et al. 2003. Numerical analysis of inden-
tation fracture in quasi-brittle materials. Engineering
Fracture Mechanics 71: 567–577.
Glowka D.A. 1989. Use of single-cutter data in the analy-
sis of PDC bit design: Part 1 – development of a PDC
cutting force model. Journal of Petroleum Technology
33(August): 797–849.
Nabhani F. 2001. Wear mechanisms of ultra-hard cut-
ting tools materials. Journal of Materials Processing
Technology 115: 402–412.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Experimental and numerical study on the application of the flat-jack

tests to masonry walls

M. Acito, L. Binda & G. Cardani

Structural Engineering Department, Politecnico of Milan, Milan, Italy

J. Guedes & L. Miranda

Department of Civil Engineering, Faculty of Engineering, University of Porto, Portugal

ABSTRACT: Double flat-jack tests were performed on a brick masonry wall using a series of LVDT’s to
monitor the distribution of displacements in the wall above the top flat-jack. The results allowed estimating:
(i) the way the stress induced by the flat-jacks is distributed in the wall; (ii) the main direction of the joints
openings/cracks that in fact extend to the sides of the flat-jack. The analysis of these results shows that actually a
volume of wall larger than the one above the top flat-jack is directly involved in the test. A numerical simulation
of the flat-jack test using a finite element code was also performed.

1 INTRODUCTION

The double flat-jacks test is used to detect the stress-

strain behaviour of a masonry wall.
The experimental research carried out by the
authors had the aim of studying the strain diffusion in
the masonry, when the double flat-jack is performed.
The research was part of an experimental work per-
formed on two masonry walls of the Museum of Sci-
ence and Technology “Leonardo da Vinci” in Milan,
in the building of the “Cavallerizze”, the riding-stables
built by Napoleon’s Army (Fig. 1).
The diagnostic investigation was aimed to the mor-
phological and structural characterization of the Cav-
allerizze masonry walls, carried out by the DIS and
supported by the Ministry of Cultural Heritage.
The flat-jack test was calibrated for masonry by
P.P. Rossi (1982) starting with the use of a single
flat-jack for the local state of stress determination.
Furthermore the double flat-jack test was developed
in order to detect the stress-strain behaviour under
compression stress. A large number of applications
of the flat-jack test is reported in literature (Binda
et al. 1983, Ronca et al. 1997, Binda et al. 1999, Binda
et al. 2004, Binda et al. 2007). The test was codified
by ASTM C1196-1991 and recommended by RILEM
LUM.D.2-1990; now it is also recommended by the
new Italian Seismic Code since 2003, as part of the on
site investigation on existing masonry buildings.
The test with a single flat-jack, is based on the stress
relaxation due to a cut made in a compressed masonry
Figure 1. Milan: part of the museum with the riding-stables
“Cavallerizze”.
(Fig. 2a). A jack is then inserted into the cut and oil is
pumped in it until the original geometry is recovered.
The state of stress is calculated from the pressure given
to the jack.
In the double flat-jacks tests, a second cut is released
in another bedding joint, parallel to the first one and
a second jack is inserted, at a distance of about 40
to 50 cm from (Fig. 2b). The two jacks delimit a
wall specimen (WS) of appreciable size to which a
uni-axial compression stress can be applied. Displace-
ments transducers (LVDTs) applied to the WS face
provide information on vertical and lateral displace-
ments. Loading cycles can be performed at increasing
stress levels in order to determine the deformability
modulus of the masonry in its loading and unloading
phases. It is important to compare these last results
to the stress level measured with the single flat-jack,

875
 Figure 3. Scheme of the masonry texture of the
Figure 2. (a) Slot in the mortar joint before inserting the Cavallerizze.
flat-jack (350 × 250 × 4 mm); (b) Displacement measures
devices.

in order to compare the local state of stress of the

masonry in relation with its potentialities. The maxi-
mum value of stress applied during the last cycle can be
increased beyond the linear phase of the stress-strain
diagram. Nevertheless, it is usually not recommended
to arrive to or near the ultimate stress, in order to avoid
destruction of the masonry.
At the museum L. Da Vinci, the tests with dou-
ble flat-jacks were carried out beyond the linear phase
till failure. This was possible thanks to the restoration
project, which allowed the demolition of some part of
the load bearing walls.
The test was so aimed to evaluate the effect of high
stress levels in presence only of a small volume of
wall above the top flat-jacks reacting to the applied
load. A number of LVDTs was placed also outside
the compressed area between the two flat-jacks (WS)
in order to better evaluate the stress diffusion in the Figure 4. Location of the flat-jacks tests: (a) on the
surrounding masonry till failure. external wall of the building named “Siloteca”(MST-J7D);
In the following, after a brief description of the (b) on the external wall of the building named “Cavaller-
masonry typology, the experimental tests are reported, izze”(MST-J6D).
together with a numerical simulation of the flat-jack
tests using a finite element code. The mortar joints height is around 15–20 mm. The
joints are realized with lime mortar with siliceous
aggregate. The quality of the mortar determined by the
2 EXPERIMENTAL Pointing Hardness tester (Schmidt Pendulum Hammer
PM), according to Rilem Recommendation MS-D7, is
2.1 Masonry description “soft”.
The “Cavallerizze Napoleoniche” located near the
Olivetani Monastery (Fig. 1) were built during the 2.2 Location of the in-situ tests
French occupation and were used as riding-stables by
the Napoleonic Army. The history of the construction The single and double flat-jack tests were carried out
reports a traumatic event during the second world war, first on the external wall of the part of the riding-
when the stables were partially destroyed. Nowadays, stables, named “Siloteca” (MSC-J7D) and then on the
compared to the original plan, with two blocks of eight external wall of the building named “Cavallerizze”
stables each, only six stables of one of the two blocks (MSC-J6D) (Fig. 4a,b).
are remaining. The first two are used as a Museum
deposit (Siloteca) and are in quite good state of conser-
2.3 Description of the results
vation. The other four are in worst conditions, partially
roofed, partially collapsed and decayed. The masonry 2.3.1 Results of the double flat-jack test MST-J7D
face texture is well organized with regular horizontal with double flat-jacks (Siloteca)
courses as reported in figure 3. The masonry section is The double flat-jack test was carried out cutting two
realised with header bricks three leaves, one of witch mortar joints at a distance of 48 cm on the external side
is alternatively constituted by half brick. of a perimeter wall (Fig. 5). Following the masonry

876
 Table 1. Elastic parameters achieved from the test (MST-
J7D).

Load interval E secant “Poisson’s” ratio

LVDT σ [N/mm2 ] [N/mm2 ] εl/εv

Average 0.1–0.5 830 0.3

1, 2, 3, 4

Figure 7. Peak strain-height plot at each loading cycle, with

Figure 5. Displacement transducers (LVDTs) applied on reference to their distance from the top flat-jack (MST-J7D).
the external wall of the “Siloteca” subjected to double
flat-jack test (MST-J7D). Table 2. Percentage reduction rate of the strain εyy along
the vertical line y from the top flat-jack (MST-J7D).

Distance from εyy % Ratio

LVDT top jack [cm] [µm/mm] εyy /ε(02+03)/2

06 21.15 3.6 102

07 33.35 1.24 35
08 56.65 0.64 18
09 69.30 0.45 13

In particular, having as a reference the stress vs

Figure 6. Stress-Strain plot of the test with double flat-jacks.
texture, four vertical and one horizontal displacement
transducers were applied, as usual, between the two
slots (Fig. 2b).
In order to monitor the strain diffusion, additional
LVDTs were applied both on the external and internal
side of the same wall around the tested area (Fig. 5). In
figure 6, the stress-strain plot measured by LVDTs 1, 2,
3, 4 and 5 is presented. In Table 1 the Elastic modulus
and the “Poisson’s” ratio are reported, evaluated on the
envelope contour (Fig. 6).
Along the vertical line above the top flat-jack,
LVDTs 06, 07, 08 and 09 (Fig. 5) show a variation
of strain that quickly decreases from LVDTs 06 to 09
(Fig. 7).
strain curve that corresponds to the average between
LVDTs 02 and 03, it is possible, to estimate the strain
decrease in percentage along the height of the wall.
The strain peak level measured along the vertical line
y (x = 0) when the loading cycle pressure is 26 bar, put
in evidence how the effects are localized (about 70 cm
over the top flat-jack). In fact, as shown in figure 7,
the strain εyy rapidly decreases with the distance from
the top flat-jack. In table 2 the percentage strain varia-
tions (LVDTs 06, 07, 08 and 09) related to the average
strain of the WS (LVDTs 02 and 03, ε(02+03)/2 =
3.5 µm/mm) are reported.
Comparing the stress vs. strain curves below (aver-
age between LVDTs 02 and 03) and above the top
flat-jack (average between LVDTs 06 and 07), the
average deformation between LVDT 06 and 07 is prac-
tically the half of the average deformation measured
between the two flat-jacks (Fig. 8).
Along the diagonal on the right-up side of the
top flat-jack, LVDTs 11 and 13 detect an important

877
Figure 8. Stress-strain of the single transducers placed
along the vertical line above the top flat-jack (MST-J7D).
Figure 9. Peak strain position of the single transducers
placed along a diagonal line at each loading cycle (MST-J7D).
Figure 10. Peak strain-horizontal position plot of the trans-
ducers, considering their distance from the right edge top
flat-jack towards right at each loading cycle (MST-J7D).
elongation at the 11 level that decreases to a very small
value at the 13 level (Fig. 9).
Along the horizontal line above the top flat-jack,
LVDTs 06, 10, 11 and 12 show that the compression
strain decreases from the flat-jack symmetry axis (y) to
the edge. In fact, on the right edge, LVDT 11 detects an
important tensile strain which decreases to a low value
when it reaches LVDT 12 (Fig. 10). Nevertheless, the
strain at LVDT 12 is twice the one at LVTD 13. This
data, in particular the strain at LVDT 11, shows that
878
Figure 11. Displacement transducers (LVDTs) applied on
the external wall of the “Cavallerizze” subjected to double
flat-jack test (MST-J6D).
the volume of the masonry above the top flat-jack,
that is pushed during the test, is much larger than the
column included in the vertical limits of the flat-jack.
It should be noticed that LVDT 11 is positioned 15 cm
to the right of the limits of the flat-jack.
In order to verify if the pressure exceeding the ulti-
mate stress applied to the external side of the wall
could be present also on the back internal side of the
same wall, four LVDTs were vertically applied cov-
ering small (min. 11 cm) and long distances (max.
138 cm), but no sensitive measurements on that side
were recorded.
2.4 Results of the test MST-J6D with double
flat-jacks (“Cavallerizze”)
The same test with the same measurement devices
was carried out to the external wall of the build-
ing named “Cavallerizze”. Nevertheless on the basis
of the previous results some changes were adopted
with reference to the positioning of the displacement
transducers above the top flat-jack (Fig. 11).
Table 3. Elastic parameters achieved from the test (MST-
J6D).

Load interval E secant “Poisson’s” ratio

LVDT σ [N/mm2 ] [N/mm2 ] εl/εv

Average 0.1–0.5 900 0.3

1, 2, 3, 4

Figure 13. Stress-Strain plot of the test with double

flat-jacks in th WS (MST- J6D).

Figure 12. Crack below the right edge of the lower flat-jack.

In Table 3 the Elastic modulus and the transversal

dilatation coefficients (“Poisson’s” ratio) are reported,
evaluated on the envelope contour. In figure 12, some Figure 14. Collapse of the masonry below the bottom
cracks developed in the final phase of the test are flat-jack.
shown.
The point that corresponds to the decrease of the
compression strain at LVDT 01, 02, 03 and 04 when
the load is still increasing, could be defined as the
“inversing point” (Fig. 13). This point appears when
the collapse of the masonry below the bottom flat-
jack starts and when the flat-jack looses the below
stiff contrast (Fig. 14). During this phase the contact
surface, i.e. the total force applied to the wall at the bot-
tom, decreases and the “wall specimen” (WS) between
the two flat-jacks start to be unloaded. The difference
between the force applied to the top and to the bottom
of the WS is deviated to the lateral masonring (less
damaged and stiffer zones).
The unloading of the WS is not followed by the
masonry unload above the top flat-jack, for which the
compression strain is still increasing. There must be Figure 15. Peak strain-vertical position of the transducers
a more rigid slice above the top flat-jack that makes placed along a vertical line above the top flat-jack, at each
loading cycle (MST-J6D).
the maximum strain to be on LVDT 07 instead of
LVDT 06. This is perceptible in particular at the end
of the test. Then, from LVDTs 07 to LVDT 08, 09, 14 (between the two flat-jacks), it is possible, to estimate
and 15, the strain quickly decreases (the dashed line the strain decrease along the height of the wall over
corresponds to the strain for the maximum stress after the top flat-jack (Fig. 15). The same behaviour of the
the inversing point, figure 15). previous test, as shown in figure 7, is observed, and
Having as a reference the stress vs. strain curve that only the effect of a localized damage was measured by
corresponds to the average between LVDTs 02 and 03 the LVDT 07 at the end of the test.

879
Figure 16. Peak strain-horizontal position of the single
transducers placed along a diagonal line at each loading cycle
(MST-J6D).
Figure 17. Peak strain-horizontal position of the single
transducers placed along a diagonal line, at each loading cycle
(MST-J6D).
The elongation on the horizontal LVDTs 19 and 20
(Fig. 16) shows that the compressed masonry above the
top flat-jack pushes the lateral masonry that becomes
horizontally compressed.That effect is more important
closer to the flat-jack, inducing the higher compression
strain measured at LVDT 19.
Comparing the stress vs. strain curves below (aver-
age between LVDTs 02 and 03) and above (LVDTs
06 plus 07) the top flat-jack, the results are quite simi-
lar, although the masonry above presents an apparently
stiffer behaviour.
Along the diagonal line (Fig. 17) on the right-up
side of the top flat-jack, LVDTs 17, 11 and 13 detect an
elongation that decreases from the 17 to the 13. It can
be noticed that although at the beginning the elonga-
tion of LVDT 12 is slightly higher than the elongation
of LVDT 13, at the end the elongation of the first is
lower than half of the second. Nevertheless, the strain
of the two LVDTs is rather low.
The fact that there is a crack that goes up, fol-
lowing a 45◦ line, until an horizontal distance of
50 cm from the extremity of the flat-jack, will give
a column of masonry participating on the test that is
(50 + 30 + 50)/(30) = 4.3 times wider than the column
880
Figure 18. Pick Strain-height plot of the single transducers
placed along an horizontal line above the top flat-jack to the
right, at each loading cycle (MST-J6D).
Figure 19. Pick Stress-strain plot of the LVDTs 01, 02, 03
and 04 (tests MST-J7D and MST-J6D).
that contributes to the in-situ stress. This, of course,
corresponds to a very ruff simplification of the reality.
Along the horizontal line above the top flat-jack
(y = 21,5 cm), the compression strain is quite constant
inside the vertical limits of the flat-jack (LVDTs 06,
10, 11 and 12, in figure 18).
On the right side of these limits, LVDT 11 detects
an important tensile strain which decreases to almost
zero when it reaches LVDT 12. These measurements
show once again that the volume of the masonry above
the top flat-jack, pressed during the test, is much
larger than the column inside the vertical limits of the
flat-jack.
In conclusion, a comparison between the two in-
situ tests is presented in figure 19, in a stress-strain
plot of the peak stress mean values measured in the
area between the flat-jacks (WS).
The comparison between the two masonry walls
shows that the best mechanical characteristics are more
in evidence for the plastered and so better protected
masonry wall of the “Siloteca”. Furthermore, in that
test, the “inversing point” was not observed due to
interruption of the test at a pressure of 26 bar when a
damage occurred to one of the two fat jacks.
3 NUMERICAL SIMULATION
3.1 Main strategies for the numerical Modelling
Masonry is a composite structural element made
by bricks (units) and mortar (joints) with different
mechanical properties. For the detailed analysis of this
heterogeneous structure with anisotropic characteris-
tics, the use of numerical models is demanded, which
include a representation of the units, the mortar and
the unit-mortar interface.
For the analysis of global structures, usually a
numerical modelling could not distinguish brick from
mortar, assuming a single homogeneous material.
In the literature two different methods, respec-
tively named Micro-modelling and Macro-modelling
are used (Rots 1991, Guinea 2000, Lourenco 1996).
The choice of the modelling method depends not only
on the purpose of the simulation, but it’s also the result
of a compromise between the complexity and the time
spent (to construct the model and to obtain results),
and the level of accuracy required in the analysis.
In particular, when dealing with local phenomena,
such as failure modes or joint sliding, it is necessary
to refer to micro-Modelling. On the contrary, for the
analysis of the global behaviour of masonry structures,
where mean values of stress and strain are considered,
macro-Modelling is the best option.
Many authors have adopted micro-Modelling
approaches for the study of masonry failure mech-
anisms: Giambanco (2001), Lourenco & Rots
(1993), for shear-tensile behaviour, Lourenco & Hen-
riques (2006), for compression behaviour, Attard &
Chaimoon (2007), for shear-compression behaviour,
Silva et al. (2007), Costa et al. (2007), for the cyclic
load. In all these studies the difficulty of reproduc-
ing the experimental behaviour of structures made of
quasi-brittle materials, like masonry, using numerical
models is evident. In particular, when considering the
compression failure mechanisms of these structures,
characterized by the coalescence and growth of diffuse
micro-cracks, accompanied by progressive localiza-
tion of deformations and development of tension free
macro-cracks, the numerical simulation of the real
behaviour of masonry structures, is very complex.
3.2 Numerical modelling of the flat-jack test
The numerical simulation of a flat-jack test on brick
masonry to study the local mechanical phenomena
involved is a particular case where micro-models are
useful. The masonry components, bricks and mortar,
are modelled separately and, if necessary, joint mod-
els can be included to simulate the contact between
brick and mortar. However, laboratory tests on sam-
pled materials should be involved in order to allow
such a model to be properly calibrated.
This work presents the numerical simulation of an
in-situ flat-jack test until failure, on a brick masonry
881
Figure 20. (a) Non-linear model, (b) shortening strain εyy
and c) shortening strain εxx .
wall. The aim is not the research on failure modes
(to be analysed in future work), but the quantitative
comparison between the numerical and experimental
results, with reference only to the diffusion of stress
above and below the flat-jacks, contributing to a better
perception of this phenomenon. To this purpose, it was
adopted a finite element model (Fig. 20 a) that consid-
ers the single masonry components (brick and mortar),
as materials following non-linear constitutive laws, but
with different properties. This behaviour is assumed
from the in-situ response of the Wall Specimen (WS)
between the two flat-jacks. The secant modulus of
Elasticity for mortar used in the following analysis was
obtained from the composite material theory, assum-
ing an Elastic modulus for brick equal to 1300 MPa and
a modulus for masonry, considered as the composite
material, equal to 450 MPa, obtained from the exper-
imental in-situ tests, between 0,4 ÷ 1,5 MPa. In this
simplified hypothesis, waiting for the experimental
tests on the sampled masonry materials, these results
give a qualitative response.
In the following, some results of the non-linear
numerical evaluation based on a F.E.M. code, Straus7,
are reported. The shortening in vertical direction (εyy )
is shown in the strain map of figure 20b, referred to
a pressure in the flat-jack of 24 bar. The shortening in
horizontal direction (εxx ) is shown in the strain map of
figure 20c also referred to a pressure in the flat-jack
of 24 bar. The two maps show a qualitative behaviour
that suits with the in-situ measurements and confirms
the local effect induced by the test. The comparison
between the numerical and the experimental results
of strain εyy along the vertical line y direction (x = 0)
above the top flat-jack of the Siloteca walls is presented
in figure 21.
4 CONCLUSIONS
The experimental double flat-jack tests, carried out on
the brick masonry walls of the Cavallerizze building,

Figure 21. Numerical and experimental comparison of
strain εyy along a vertical line (y direction) above the top
flat-jack (MST-J7D).
have helped in analysing and understanding the stress
and strain diffusion induced during the test. As far
as the masonry area located above the top flat-jack is
concerned, some remarks could be put on evidence:
(i) the compressive stress diffusion has a slope equal
to 45◦ to 50◦ ; (ii) the compression strain along the
vertical line (y direction) rapidly decreases with the
increase of the distance from the top flat-jack (about
100 cm); (iii) the masonry transversal confining effect
(x direction at 21 cm above the top flat-jack) rapidly
extinguishes with the distance from the Y axes (about
50 cm from the edge of the flat-jack).
As far as the masonry area located between the
flat-jacks is concerned (WS), some remarks could be
put on evidence: (i) the point that corresponds to the
decrease of the compression strain in the wall speci-
men (WS) when the load is still increasing, could be
defined as the “inversing point”, the starting point of
the collapse of the masonry below the bottom flat-
jack; (ii) from the inversing point, the masonry below
the flat-jack becomes even more deformable and the
flat-jack looses the below stiff contrast, which starts
to crack (expulsion of bricks was observed); (iii) the
unloading of the WS is not followed by the unload-
ing of the masonry above the top flat-jack, for which
the compression strain still increased. The tests were
carried out until failure of the wall specimen, due to
the fact that the masonry wall will be demolished.
Normally the double flat-jack test cannot be pushed
so far. Nevertheless the results show that in general,
in brick masonry, in case of a low masonry contrast
above the wall specimen (WS), there is a stress dis-
tribution that allows to reach states of stress higher
than those previously measured with the single flat-
jack test. There is, in fact, a resource in stresses which
is given by a masonry volume much larger than the
column included in the vertical limits of the flat-
jack. Referring to the strain measured above the top
882
flat-jack, the effects extinguished at a height which
is approximately the double of the wall specimen
one (WS). Therefore these results can be eventually
transferred into guidelines for the use of the double
flat-jack test.
ACKNOWLEDGEMENTS
The research was mainly supported by the Italian Min-
istry of Cultural Heritage-Lombardia Region, then by
the Fundação para a Ciência e Tecnologia, Portugal
and the Reluis funds. The authors wish to thank the
arch. C. Di Francesco of the Ministry of Cultural
Heritage-Lombardia Region, dr. F. Galli, director of
the Museum, the architects L. Cantini, R. De Ponti and
C. Tiraboschi, the technicians M. Antico, M. Cucchi,
M. Iscandri.
REFERENCES
Rossi P.P. 1982, Analysis of mechanical characteristics of
brick masonry tested by means of in-situ tests, 6th IBMaC,
Rome, Italy.
Binda L., Rossi P.P., and Sacchi Landriani G. 1983, Diagnos-
ticAnalysis of Masonry Buildings, Proc. IABSE Symp. on
Strenghtening of Building Structures, Ed. IABSE-AIPC-
IVBH, Venezia, pp. 131–138.
Ronca P., Tiraboschi C., and Binda L. 1997. In-situ flat-
jack tests matching new mechanical interpretations, 11th
Int. Brick/Block Masonry Conf., Shanghai, China, Vol. 1,
pp. 357–366.
Binda L., and Tiraboschi C. 1999. Flat-jack Test as a Slightly
Destructive Technique for the Diagnosis of Brick and
Stone Masonry Structures, 8th Int. Conf. and Exhibition,
Structural Faults and Repair.
Binda L., Cantini L., and Tiraboschi C., 2004. Caratteriz-
zazione e classificazione di murature storiche in zona
sismica mediatne prova con martinetti piatti, XI Cong.
Naz. “L’Ingegneria Sismica in Italia”, ANIDIS, Genova,
25-29/01, 2004, CD-ROM, E4-07, ISBN 88-86281-89-7.
Binda L., Cantini L., Cardani G., Saisi A., and Tiraboschi C.
2007. Use of Flat-Jack and Sonic Tests for the Qualifi-
cation of Historic Masonry, 10th Tenth North American
Masonry Conference (10NAMC), St. Louis, Missouri.
Rots J.G. 1991. “Numerical simulation of cracking in struc-
tural masonry”- Heron.
Guinea G.V., Hssein G., Elices M., and Planas J. 2000.
Micromechanical modelling of brick-masonry fracture.
Cement and concrete Research 30, 2000 – 731–7.
Lourenco P.B. 1996. “Computing Strategies for masonry
structures” – PhD. Thesis: Delft Univ. of Technology.
Gianbanco G. et al. 2001. “Numerical analysis of masonry
structures via interface models”; Computer Methods in
Applied Mechanics and Engineering.
Lourenco P.B, and Rors J.G. 1993. “On the use of micro-
models for analysis of masonry shear walls” Proceeding
of the 2nd international symposium on computer methods
in structural masonry”.
Lourenco P.B, and Henriques J.P. 2006. “Validation of ana-
lytical and continuum numerical methods for estimating
 the compressive strength of masonry”; Computer and
Structures 84, 2006 – 1977–1989.
Attard M.M., and Chaimoon K. 2007. Modelling of un-
reinforced masonry walls under shear and compression”;
Engineering Structures 21, 2007 – 2056–2068.
Silva B., Guedes J., Arêde A., and Costa A. 2007. “Structural
assessment of a stone masonry wall using a contin-
uum damage model”, ECCOMAS Thematic Conference
883
on Computational Methods in Structural Dynamics and
Earthquake Engineering, Crete, Greece, 13–16/6, 2007.
Costa A., Silva B., Costa A., Guedes J., and Arêde A. 2006.
“Structural Behaviour of a Mansonry Wall under Hori-
zontal Cyclic Load; Experimental vs. Numerical Study”,
11th International Conference Structural Faults + Repair,
Edinburgh, Scotland, 13–15 June 2006.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

The impact of the water/lime ratio on the structural characteristics of

air lime mortars

R.M.H. Lawrence & P. Walker

BRE Centre for Innovative Construction Materials, Department of Architecture and Civil Engineering,
University of Bath, UK

ABSTRACT: Lime based mortars are now widely acknowledged to be generally superior to cement based
mortars in the repair of appropriate historic infrastructure. Increasingly the benefits of hydraulic lime mortars
are also being realised in new masonry construction. In order to standardise the expected performance of mortars,
designers will specify the type of lime, the type of filler (aggregate), the proportions of each and quantity of
water or the required workability. Limes can be non-hydraulic (calcium or dolomitic) or hydraulic (natural or
artificial).
It is well known that the water/binder ratio has a marked effect on the structural performance of cement-based
mortars. This relationship is known as Abrams’ rule, which states that when a cement mortar is fully compacted,
its strength is inversely proportional to the water/cement ratio. Abrams’ rule has also been demonstrated to apply
to hydraulic lime mortars. The reason for this is that both cement and hydraulic limes require a minimum quantity
of water to produce the chemical set resulting from the hydration of calcium silicates and calcium aluminates.
Surplus water eventually dries out, leaving micropores in the matrix which weaken the resulting set mortar.
It has generally been assumed that the same relationship applies to non-hydraulic lime (air lime) mortars.
This paper reports on results of tests conducted on air lime mortars at early stages of curing. It is known that
the form of air lime and the physical and chemical characteristics of the aggregate have a strong impact on the
structural performance of air lime mortars. Results to date show that the water/lime ratio has a minimal impact
on the structural performance of air lime mortars compared with the impact of lime and aggregate type. 91 day
compressive strengths for air lime mortars with a water/lime ratio of 0.56 (a stiff mix) are identical to those with
a water/lime ratio of 0.875 (a loose slurry). Whereas Abrams’ rule is a key consideration for designers of cement
and hydraulic lime mortars, it has been demonstrated that it requires modification in the case of air lime mortars.
A relationship between form of lime, type of aggregate, water/lime ratio, and age of mortar is proposed. The
resultant equation allows the compressive strength of air lime mortars to be predicted taking into account these
factors. The insights gained from this study will allow practitioners to more confidently design and specify air
lime mortars.

1 INTRODUCTION concrete or a cement mortar is fully compacted, its

Many of the mechanical properties of hardened cement
are associated with the physical structure of the
hydration products, viewed at the level of colloidal
dimensions [Neville, 1995]. The pore structure of
hydrated cement paste contains two distinct ranges of
pore size – gel pores of about 3nm in diameter and
capillary pores which are two or three orders of mag-
nitude larger. Cement requires sufficient water to fully
hydrate the various constituents, and water in excess
of this required amount produces capillary pores. Thus
the greater the water/cement ratio above the min-
imum required for complete hydration, the greater
the amount of capillary pores created and therefore
the higher the porosity of the hardened paste. When
strength is inversely proportional to the water/cement
ratio according to Abrams’ rule.
where w/c represents the water/cement ratio of the
mix by volume, and K1 and K2 are empirical constants.
K1 relates to aggregate strength, particle shape, size,
grading and surface texture, and K2 relates to the com-
pressive strength of the cement paste [Nagaraj & Banu,
1996].
Abrams’ rule could well be valid for hydraulic lime
mortars, since they also gain at least part of their

885
 Table 1. Water/lime ratios used [by volume]

Specimen Water/Lime
Designation ratio

Air Lime
A1 0.5
A2 0.5625
A3 0.625
A4 0.6875
A5 0.75
A6 0.875
Hydraulic Lime
Figure 1. Relationship between water/lime ratio and H1 0.375
strength (Allen et al. 2003). H2 0.4375
H3 0.5
H4 0.5625
strength through hydration products.Allen et al. [2003] H5 0.625
have shown this relationship in Figure 1.
The data presented in Figure 1 cannot, however,
be taken to be truly representative of Abram’s rule.
This is because the data are based on the compres-
sive strengths of different binder:aggregate ratios – It was found that air lime required more water in
each mortar requiring a different quantity of water to order to make a workable mix than hydraulic lime,
produce a specified flow. This means that two vari- and could accommodate more water before becom-
ables are present in the graph, and it is not clear what ing a loose slurry. This was likely to be a function of
proportion of the compressive strength is affected by the greater capacity of air lime to absorb water than
which variable. There has been conflicting evidence hydraulic lime as a result of having finer particles and
about the applicability of Abram’s rule to air lime therefore a greater surface area.
mortars. Schäfer & Hilsdorf [1993] and Winnefeld & Both mortars were de-moulded after 5 days and
Böttger [2006] present data which show that increased cured in a controlled environment of 60% RH at 20◦ C
water content in air lime mortars does not reduce until testing.
compressive strength. It has been shown that higher Compressive tests on six 50mm cubes were con-
porosity in air lime mortars allows greater access ducted after 28, 56 and 91 days from the date of
to atmospheric carbon dioxide (CO2 ), which pro- manufacture.
motes carbonation and therefore can produce greater
compressive strengths [Lanas & Alvarez, 2003].
This paper describes a systematic evaluation of the 3 RESULTS
impact of the water/lime ratio on the unconfined com-
pressive strength of air lime mortars up to 91 days after The results of compressive tests on the mortars are
manufacture. shown in Figures 2–4. Error bars are included showing
the range of results of the six tests used to produce each
data point.
2 EXPERIMENTAL These data compare well with the data produced
both by Schäfer & Hilsdorf [1993] and Winnefeld &
50 mm × 50 mm × 250 mm prisms of mortar were pre- Böttger [2006].
pared with 1 part of dry hydrated high calcium lime
(CL90) and 3 parts of silicate sand by volume using
a range of different water/lime ratios. 9 specimens 4 ANALYSIS AND DISCUSSION
of each mortar type were prepared and 2 cubes were
taken from each specimen for testing at the appropri- The data for hydraulic lime mortars appear to follow
ate time intervals. For comparison purposes a further Abrams’ rule with the relationship between strength
set of 9 prisms were prepared using 1 part NHL3.5 and water/lime ratio following an approximate hyper-
lime and 3 parts silicate sand with a range of different bolic curve. The data for the air lime mortars, apart
water/lime ratios. from the lowest water/lime ratio, show very little varia-
The amount of water added to each lime type ranged tion in compressive strength when the water/lime ratio
from the minimum quantity needed to make a workable is varied. It is conceivable that the data points for the
mortar to the amount required to make a loose slurry. air lime mortar are all to be found at the lower end of
The water/lime ratios used were as shown in Table 1. the hyperbolic curve, where there would be very little

886

Figure 2. Compressive test results on specimens 28 days
from manufacture and different water/lime ratios.
Figure 3. Compressive test results on specimens 56 days
from manufacture and different water/lime ratios.
Figure 4. Compressive test results on specimens 91 days
from manufacture and different water/lime ratios.
difference to be seen from an increase in the water/lime
ratio. The factors involved in the strengthening of
air lime mortars are different from those involved in
hydraulic lime mortars. In air lime, after an initial
strength gain achieved from the drying out of the mor-
tar, subsequent strength gain is achieved over extended
periods as a result of carbonation. In hydraulic lime
mortars, there is also an initial strength gain achieved
from drying, combined with a gain achieved through
887
an hydraulic set, which takes up to 28 days, depending
on the hydraulicity of the lime. Subsequent to this the
strength gains are due to carbonation.
Van Balen [1994] has proposed a model for car-
bonation represented by a differential equation with
a ‘sink’ term (R(w, c)) . The factors involved in the
equation include time, the porosity and diffusivity of
the material, the construction method and the presence
of cracks, and the geometrical shape of the surface
exposed to air. Carbonation depth (x) is proportional
√ root of time√(t) ± a constant (e) in the
to the square
form x = k t or x = e + k t, where k is a factor which
does not necessarily correspond to a property of the
material.
It has been shown that the compressive strength of
air lime mortars varies in proportion to the extent of
carbonation (Lawrence, 2006a), and it is therefore rea-
sonable to expect that development of compressive
strength will also be proportional to the square root
of time.
The following formula has been developed to model
the variation in compressive strength of air limes as the
water/lime ratio varies.
where fl is the compressive strength of the air lime mor-
tar, Km is an empirical constant which varies according
to the nature of the aggregate, Kl is an empirical con-
stant relating to the form of air lime, and d is the age
of mortar in days since manufacture. The factor of 150
is an empirical constant derived to provide a best fit
with actual data.
It is known that differences in the mineralogy and
granulometry of an aggregate will have a significant
impact on the compressive strength of air lime mortars
[Lawrence et al. 2006], even at a very early stage after
manufacture. Km represents this effect. This constant
will not vary for a given aggregate whatever the time
from manufacture.
As air lime mortar increases in age, so carbonation
has an increasing impact on the compressive strength
of the mortar. This effect occurs across the
√whole range
of water/lime ratios, and the expression d represents
this effect. The value of this expression will increase
as the time from manufacture increases up to the point
where carbonation is virtually complete. This expres-
sion appears to be valid up to values of 180 for d,
beyond this value – once the mortar has carbonated,
the expression would not vary.
Different forms of air lime carbonate to a greater or
lesser extent and at a greater or lesser rate, mainly
dependent on the size, shape and integrity of port-
landite crystal present in the lime. Kl represents this
effect. This constant will not vary for a given lime
whatever the time from manufacture.
Figure 5. Day 28 air lime data compared with the proposed
equation.
Figure 6. Day 56 air lime data compared with the proposed
equation.
Figure 7. Day 91 air lime data compared with the proposed
equation.
As carbonation progresses through the depth of the
mortar so the rate of carbonation will decrease. The
expression 150∗ d represents this effect.As commented
on above, this expression would become a constant
once carbonation has completed.
The proposed equation can only be considered to
be valid only up to 180 days from manufacture for air
lime mortars.
Figures 5–7 represent models using the above for-
mula for mortars at 28, 56 and 91 days, and compared
888
Figure 8. Proposed equation applied to mortars made with
different aggregates. 4SS3 = silicate sand; 4BN3 = crushed
bioclastic limestone (Ham Hill stone); 4ON3 = crushed
oolitic limestone (Stoke Ground Bathstone).
with the data shown in Figures 1–3. For these calcu-
lations, the value of Km was taken as 0.05, and that of
Kl as 20.
At 28 days the equation over estimates the com-
pressive strength by up to 25%, improving with higher
water/lime ratios. As the mortar ages, so the fit
improves, until by 91 days, apart from the lowest
water/lime ratio, the equation fits within the error bars
of the actual data.
The equation predicts a reduction in compressive
strength in line with increases in water/lime ratio
which the data do not entirely support, although exper-
imental variations may explain small increases in
strength at higher water/lime ratios. The equation does,
however, go some way towards accounting for the more
significant influence of the type of aggregate on the
compressive strength.
Figure 8 shows the proposed equation applied to dif-
ferent aggregate mortars keeping the water/lime ratio
constant, but varying the time (d). The solid lines show
the relationship between compressive strength for lime
mortars made with oolitic aggregates (green), bioclas-
tic aggregates (red) and silicate sand aggregates (blue),
and the curves predicted by the proposed equation.
Compressive strength data are taken from Lawrence
[2006b]. The water/lime ratios used to make the actual
mortars were factored in. The constant Km which rep-
resents the impact of the aggregate on the compressive
strength of the mortars was 0.420 for the oolitic mor-
tar, 0.220 for the bioclastic mortar, and 0.106 for the
silicate sand mortar.
It can be seen that up to 180 days there is a rea-
sonable correlation between actual and predicted com-
pressive strengths. The equation requires modification
to take account of the completion of carbonation, but
up to 180 days it seems to be able to predict with rea-
sonable accuracy the compressive strength of air lime
mortars as they are affected by water/lime ratio, type
of aggregate and time from manufacture. The data for
mortars made with oolitic stone shows greater early
strength than predicted by the equation. Indeed early
strength exceeds that of moderately hydraulic lime
mortars. It has been shown by Lawrence (2006a) that
no hydraulic effects are involved, and this phenomenon
is the subject of ongoing research.
5 CONCLUSIONS
The experimental data demonstrate that, unlike cemen-
titious or hydraulic lime mortars, the increase in
porosity produced by higher water content does not
result in the strength reductions predicted by Abrams’
rule. Indeed it can be seen that the choice of aggre-
gate has a significantly greater impact on the strength
characteristics of air lime mortars.
One key implication of this are that for air-lime
mortars, water/binder ratios can safely be selected to
produce a mortar with suitable workability charac-
teristics as demanded by the particular application,
rather than based on the erroneous assumption that
Abrams’ rule applies. Unlike hydraulic limes, with
air-lime mortars apart from very stiff mortars made
with water/lime ratios of 0.5 compressive strengths
are shown to be remarkably unaffected by greater
quantities of water in the initial mix.
The proposed equation will allow a first approxima-
tion of compressive strength of different air lime mor-
tars to be predicted. Continuation of the present study
will allow the creation of empirical constants for types
of lime and aggregate (Km and Kl ). More work needs to
889
be done to understand the mechanisms involved which
produce such significant differences in compressive
strength. Such an understanding will allow the formu-
lation of a more developed equation than that which
has been proposed.
REFERENCES
Allen, G., Allen, J., Elton, N., Farey, M., Holmes, S.,
Livesey, P., Radonjic, M. 2003. Hydraulic lime mor-
tar for stone, brick and masonry. Shaftesbury, Donhead
Publishing Ltd.
Lanas, J., Alvarez, J.I. 2003. Masonry repair lime-based mor-
tars: factors affecting the mechanical behaviour. Cement
and Concrete Research 33(11): 1867–1876.
Lawrence, R.M.H., Walker, P., D’Ayala, D. 2006a. Non-
hydraulic lime mortars. The influence of binder and
filler type on early strength development. Journal of
Architectural Conservation 12(2): 7–33.
Lawrence, R.M.H. 2006b A study of Carbonation in non-
hydraulic lime mortars, unpublished PhD Thesis, Univer-
sity of Bath.
Nagaraj, T.S., Banu, Z. 1996. Generalisation of Abrams’ law.
Cement and Concrete Research 26: 933–942.
Neville, A.M. 1995. Properties of Concrete. Harlow,
Longman.
Schafer, H.R., Hilsdorf, H.K. 1993. Ancient and new lime
mortars – the correlation between their composition struc-
ture and properties. In M. J. Thiel Conservation of Stone
and other Materials. London, E. & F.N. Spon: 605–612.
Van Balen, K., van Gemert, D. 1994. Modelling lime mortar
carbonation, Materials and Structures, 27: 393–398.
Winnefeld, F., Böttger, K.G. 2006. How clayey fines in aggre-
gates influence the properties of lime mortars. Materials
and Structures 39: 401–411.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Soil mortar’s evaluation characteristics for their use in

conservation practice

G.E. Bei
Aristotle University of Thessaloniki, Greece

ABSTRACT: The aim of this study was the investigation of the mechanical properties, the drying shrinkage
and the hydroscopicity of soil mortars in order to firstly understand and then classify their behaviour; for their
use as structural mortars or renderings at historical building interventions or modern building structures.
Soil mortar compositions based on traditional data and improved –stabilised mixtures were designed using
additives, admixtures and fibres. Their properties were estimated with the help of mechanical, capillary suction
in pure and salted water and shrinkage tests. Useful conclusions have been revealed, based on the tests results
and the comparisons, related to the studied parameters for the mortars behaviour. Finally, some of the studied
compositions, according the above criteria were selected for structural mortars or renderings and they were
proposed for further durability and compatibility with the substratum measurements.

1 INTRODUCTION 2 EXPERIMENTAL SET-UP

Soil was one of the main building materials in Greece 2.1 Data – Preliminary tests
from the prehistoric times until the beginning of the
It is obvious that the design methodology of appropri-
20th century, when the concrete appeared in con-
ate intervention materials on the historical buildings
struction. Less than a century time the use of soil
requires the study and the analysis of the existing mate-
in structures has been underestimated and neglected.
rials. This experimental work included the study of the
Nowadays, the environmental consequences of the
soil mortars as they were used in the past. Before the
modern materials use, such as this of concrete (CO2
begging of the production of the compositions, var-
and other toxic emulsions during its production) has
ious preliminary tests were carried out. Mortars of
led the developing world to re-establish the tradi-
different compositions have been selected from tradi-
tional structural materials. Therefore, soil is studied
tional buildings in several regions of Northern Greece
on a scientific base. Its behaviour and improving
(Kozani, Kastoria, Thessaloniki) and analysed.
ways of its properties are systematically investigated.
The results of this study showed that external soil
However, there are still limited experimental data
mortars on existing buildings were demonstrated by
(Venkatarama & Gupta 2005, 2006).
high percentages of CaO and therefore they could be
This work aims at contributing to the study of
characterised as soil-lime mortars. On the contrary,
soil as a building material evaluating soil mortar
internal mortars proved to be pure without admixtures.
compositions. Mixtures based on traditional compo-
Soil mortars used as renderings contained a 0.7 to 1.5%
sitions were examined. Improved-stabilised mixtures
of straw with a length of 2 to 5 mm or small wood sticks
were designed using additives, admixtures and fibres.
with maximum diameter 3mm and 20 mm of length.
Their properties were examined performing mechan-
ical, shrinkage during drying process, absorption and
capillary suction capacity in pure and salted water
tests. 2.2 Evaluation of soil
The tests described above made it possible for direct The above results (§2.1) were taken into account
results to be achieved for an indicative amount of spec- for the design of the new soil mortars compositions
imens, because both the production procedure and the of this project.Two different soil types have been
experimental process are time consuming. used as the basic materials. Both have been obtained

891
Table 1. Chemical analyses of soils type A and B. The magnitude of shrinkage varies with the clay
content and the type of clay (Balderrama & Teuton-
Percentage of Percentage of components ico1983). In fact, it was respected that that soil A
components soluble to 0.1 nHCl composition will shrink more than those of soil B. On
Soils A B A B the basis of mineralogical analyses, soil type A and B
were tested on the clay fraction (2 ÷ 0.2 µm). Soil A
Na2 O 1.4 2.5 0.11 0.28 had a more stable structure than B due to the higher
K2 O 2.8 1.3 0.04 0.03 percentage of illite in its composition. On the contrary,
CaO 1.99 6.2 0.70 1.80 soil B, which had less illite and a lot more of smectite,
MgO 2.63 3.63 0.33 0.44 bulked immediately after the beginning of evapora-
Fe2 O3 8.68 7.1 0.13 0.22 tion. This was observed during the drying process in
Al2 O3 18.68 19.03 0.26 0.30
almost all the specimens with soil B in the mixture.
SiO2 52.8 51.94 0.22 0.17
Weight loss 11.5 8.3 These soils can be described as Clay Loams (CL)
PH 8.5 8.9 according to Atterberg limits. Particularly, soil type A
was described as a medium plasticity clay soil with liq-
uidity limit (WL ) 38.5 and plasticity index (IP) 16.7,
while soil type B as a low plasticity clay soil, with
Table 2. Percentage of salts on soils type A and B. WL 30.1 and IP 11.74. Both types belong to A6 cat-
egory according to the American Society for Testing
Percentage of salts (%) and Materials (ASTM D423-66 & D424-59). A6 indi-
A B cates plastic soils, which usually present an important
volume change, when passing from dry state to humid
Chlorides 0.014 0.001 state. However, it should be pointed out that these clas-
Nitrates 0.0007 0.0008 sifications are not always good indicators of the soil
Sulfates 0.034 0.041 mortars behaviour. For instance, the plasticity index is
computed by means of the separate soil fraction which
is smaller than No 40 mesh and not all of the fractions.
Therefore, the plasticity index value does not neces-
Table 3. Grain size composition (%) of soils type A and B. sarily reflect the actual plasticity of the soil used for
mortars (as in our case).
Grain size composition A (%) B(%)

sand fraction SF 52.485
silt fraction MF 21.515
clay fraction CF 26.000
from special sites of brick manufacturing; soil A (red
colour), from Western Thessaloniki, soil B (brown
colour), from Eastern Thessaloniki. These two sites
are areas from where earth has been traditionally
obtained and processed since at least the12th century
AD (Theocharidou 1988).
Chemical analyses, grain size composition tests and
Atterberg limits were defined and additionally miner-
alogical analyses were realized in order for soils to be
classified (Table 1, 2).
Both samples contained low percentage of sulfates,
allowing us to foresee a fairly low susceptibility to
efflorescence. Soils A and B showed a high content in
SiO2 and Al2 O3 , in accordance with a high content in
clay minerals. Soil B was more calcareous than A, and
that both have abundant Fe+++ .
The grain size distribution of the two tested soils
showed that the clay fraction (philomorphe material)
of soil A is higher (26%) compared to B (18.39%)
(Table 3).
38.670 2.3 Requirements for the use of mortars in
42.940
conservation practice
18.390
In 1981, for the first time, a tentative list of the char-
acteristics of an ideal mortar for restoration was pre-
sented. These were easy workability, rapid and reliable
setting in both dry and wet environments, slow dry-
ing shrinkage during setting (crucial for soil mortars),
good mechanical and thermal characteristics, porosity
similar to the components of the masonry and solu-
ble salts content as low as possible (ICCROM 1981).
Renderings of soil that are directly exposed to the
environment should have high hardening speed, water
resistance, low level of microcracks during drying pro-
cess and good bonding with the substratum. Generally,
the production of the soil mortars of this work took into
account the above requirements and specimens with
porosity and mechanical characteristics similar to the
traditionals were produced.
2.4 Tested mixtures compositions
For each soil A and B, seventeen (17) compositions
were produced firstly on the basis of traditional mor-
tars (taking into account analyses of soil mortars
from existing structures §2.1) and secondly with the

892
Table 4. Soil mortar compositions for soil A. Table 5. Soil mortar compositions for soil B.

% of mixing materials % of mixing materials

code compositions (per unit weight) code compositions (per unit weight)

A soil, water (s.w.) 73% s. 27% w B soil,water(s. w.) 65% s. 35% w

A2 s.w. air entrainer 31% w + 1% air entr. B2 s.w. air entraining 26.3% w + 1% air entr.
A.2.0 s.w. water reducing 30% w + 2.5% water B.2.1 s.w.eg. cannabis 34% w + 1% cannabis
reducing B.2.2 s.w .straw 32.1% w + 1% straw
A.2.1 s.w. eg. cannabis 44.6% w + 1% cannabis B.2.3 s.w. goat hairs 32.1% w + 1% hairs
A.2.2 s.w. straw 35% w + 1% straw B.3.1 s.w. sand20% 32.5% w + 20% sand
A.2.3 s.w.goat hairs 37.5% + 1% hairs B.3.2 s.w. sand60% 22% w + 60% sand
A.3.1 s.w. sand20% 35% w + 20% sand B.4.1 s.w.1:2 lim.: poz. 39.1% w,8.3% l 16.7% p,
A.3.2 s.w. sand60% 25% w + 60% sand 75% s
A.4.1 s.w. 1:2lime: 41% w,8.3% l,16.7% B.4.2 s.w.1:4 (l:p) 35.4% w,5% l, 20%,
pozz. (l.p.) p,75% soil 75% s
A.4.2 s.w. 1:4 l:p 45% w,5% l,20% sand, B.4.3 s.w. 2:1 (l:p) 36% w,16.7% l,8.3% p,
75% soil 75% s
A.4.3 s.w. 2:1 l:p 45% w,16.7% l,8.3% p, B5 soil cement 27% w,73.4% s,14.3% c,
75% soil 12.24% sand
A5 soil cement 37.7% w,73.4% s,14.3% c, B6 1:2(l:p) + 33%sand 29.2% w,33% sand,5.5% l,
12.24% sand 11% p,50% soil
A6 1:2 l.p 33%sand 39% w,33% sand,5.5% l,11% p, B7 1:2(l:p) + 33% 33.3% w,33% sand,5.5% l,
50% soil sand + cannabis 11% p, 50%soil,
A7 1:2 l.p. 33% 50% w,33% sand,5.5% l,11% p, 1%cannabis
sand + cannabis 50% soil,1% cannabis B8 1:2(l.p.)33%sand 32% w,33% sand, 5.5% l,
A8 1:2 l.p. 33% 50% w,33% sand,5.5% l,11% p, + goat’s hairs 11% p, 50%
sand + goat’s hairs 50% soil, 1% g.hairs soil1%g.hairs
A9 A5 + cannabis 37.7% w,73.4% s,14.3% c, B9 B5 + cannabis 27% w,73.4% s,14.3% c,
12.24% sand 12.24% sand
A10 1:1.5 l.p. + 37.7% w,6.6% l,10% p, B10 1:1.5(l.p.) 33.7% w,6.6% l,10% p,
sand + cannabis 50% s,33% sand + cannabis 50% s,33% sand
A11 s.w. 30%lime 32.8% w,30% l,70% soil B11 s:w:30%lime 30% w, 30% l,70% s

agents. For this reason these products weren’t used for

addition of natural fibres (goats’ hair, egyptian can-
abis, straw), of additives (lime, pozzolan, cement,
sand) and of admixtures (water reducing, air entrain-
ing agents), in various proportions, aiming at the study
and improvement of the soil mortars properties.
Tables 4 and 5 present the soil mortar composi-
tions used for the two types of soils in percentages
per unit weight (u.w.). The water percentage in these
tables describe the water content per unit weight of
solid materials in the mixture.
Initially, soil mortar compositions with soil and
river sand were designed and they are reported here as
“pure” soil mortars (compositions A and B). Then, air
entraining products were added in the mixture.The aim
of the introduction of a bubble system with air entrain-
ing admixtures (in magnitude of 0.05 to 1.0 mm) and
the increase of the mortar volume voids has been com-
pleted in order to reduce the high drying shrinkage
which has been observed on the tests. Three different
water reducing products two of lignosoulfonic origin
and one of formaldehyde origin have been used. All the
tests showed that the workability (A.S.T.M. C230-80:
Flow test method) was reduced compared to composi-
tions of “pure” soil mortars without the water reducing
the next stabilized compositions with the additives and
the sand. The river sand content was studied in dif-
ferent proportional combinations, as well as additives
such as lime, pozzolan and cement in various contents
and their combinations with natural fibres. The same
procedure was followed for soil mortars of both soils
types.
2.5 Tests on soil mortars in wet and dry state
Firstly, 135 ± 50 mm (the diameter of the mixture
composition after 15 blows by the flow table test)
were determined as the adequate standard workabil-
ity level of soil mortars (reference A.S.T.M. C230-80).
Then, the mixture was set in a prismatic mould with
dimensions of 160 ± 0.4 × 40 ± 0.2 × 40 ± 0.5 mm.
The curing conditions inside moulds were: 90% of rel-
ative humidity and 18◦ C to 20◦ C of temperature. After
maturing in the moulds, the specimens were set in a
room where the relative humidity was 65% and the
temperature 20◦ C up to day of their mechanical tests.
The tests in dry state are related to the mechan-
ical and physical characteristics of the soil mortars.

893
The compressive and flexural strength of the speci-
mens at different ages (28, 45, 90, 180 days) were
measured. Ultrasonic test, where the dynamic modu-
lus of elasticity in specimens, was calculated, capillary
suction capacity tests in pure and salted water, drying
shrinkage (curing conditions predetermined), absorp-
tion tests were performed in order to define the open
porosity of the soil mortars. As far as shrinkage test is
concerned the results of the consequent measurements
are presented as a proportion of the initial volume of
the specimen. The first measurement at the day after
drying in the mould (0 day) was considered as the
plastic deformation (Figure 6).
3 RESULTS – DISCUSSION
The experimental investigation focused on the follow-
ing major criteria: The strength of the specimens, the
capillary suction capacity and shrinkage during drying
process.
3.1 Results of mechanical tests
The tests’ results on the Table 6 showed that strength
increased with time (columns (1), (2), (3) and (4))
although in some cases, it seemed that there was a
decrease of strength especially for specimens with
additives. Specimens of soil B presented higher
strength comparing with those of soil A. This may be
due to the lower percentage of clay in the soil B which
makes the soil less “demanding” in the mixing water
for the appropriate workability (Table 4).
Natural fibres increased strongly the flexural
strength of the soil mortars as it is shown on the Table 6.
It seemed that thin fibres with high dispersion (spe-
cific surface) were more efficient such as the egyptian
cannabis. In compositions with fibres, especially those
of egyptian cannabis, it was not possible to measure
the ultimate compressive stress of each specimen. It
was difficult to distinguish whe the specimen broke
because the load indicator on the testing apparatus
continued measuring after the first crack. The spec-
imen with the fibres continued to absorb load energy
and be deformed. Therefore, it was decided to stop the
measurement of compression test at the 50% of the
height reduction of the specimen.
In fact, the specimen was plasticly deformed until
its height, between the load plates-surfaces, became
the half (50% of slum).
Sand percentage more that 20% in the mixture of
the soil mortars seemed to decrease the strength. The
cohesion between the coarse sand and the clay min-
erals is lower when the sand percentage is remarkable
(>20%) in the mortar composition. The dynamic mod-
ulus of elasticity test detected voids and cracks in the
specimens within 60% of sand.
894
Generally, compositions improved with hydrated
lime or with a mixture of hydrated lime with poz-
zolan without fibres, showed a tendency of decrease
the strength at the age of 28 and 45 days. This may
be due to the interference of the clay material of the
soil in the absorption of the available water by lime
and pozzolan in the mixture. Nevertheless, hydrated
lime and pozzolan, have contributed to the increase
of the strength especially concerning the mixtures of
soil A which were more argillaceous while the addi-
tion of cement (composition B5) contributed better to
strength on compositions with sandy soils as soil B
(Table 4 & Table 6 compositions A.4.1, A.4.2, A.4.3,
A11 and compositions B11, B.4.1, B.4.2 and B.4.3).
The open porosity of compositions at the age of
28 days is presented in Table 6. The porosity was cal-
culated into accordance with the bibliographical data
(absorption test) and it was relative to the strength
of the tested mortars. In fact, minimum porosity cor-
responded to maximum strength (Papayianni 2001),
(Mojumbar et al. 2002).The addition of water reducing
or air entraining agents did not influence the poros-
ity of the specimens (compared to the compositions
without additives). The values of the open porosity in
the soil mortars with lime (from 34.57% to 49,39%
for compositions A.4.1, B.4.1, A.4.2, B.4.2, A.4.3 and
B.4.3) were similar to the porosity of lime mortars
(with porosity values between 30 and 40%) (Shafer et
al. 1992). The similar porosity could be an indicator of
compatibility between the tested soil and lime mortars.
3.2 Results on the capillary suction capacity
The compositions with soil B absorbs less water (3%–
14%) than those of soil A (3%–25%); this may be
due to the high content of clay materials (Figure 1
& Table 4). The use of air entrainer reduced the per-
centage of the rising water in the capillaries, even
though the porosity of the specimens with admixtures
was not higher than the free admixtures composi-
tions. It seemed a different distribution of voids, to
exist which influence negatively the capillary suction
capacity (Figure 2).
Sand addition reduced the water absorption during
suction (Figure 3). The addition of lime or pozzolan
in the compositions increased significantly the water
absorption through the capillaries (25% for A.4.1).
After the beginning of the evaporation, the weight loss
was immediate (Figure 4).The rate of water weight loss
during the evaporation was slower when cement was
added in the mixture. This seemed important because
it showed that the water content of the soil mortar sta-
bilised with lime could easily evaporate compared to
soil mortar stabilized with cement. It is well known the
facility to water evaporation of lime mortars is the main
advantage towards cement mortars for their durability
and the prevention of frost attack (Papayianni , 1994).
Table 6. Compressive/flexural strength and open porosity of specimens.

Compressive/flexural strength (mean value of 3 tested Open porosity

code specimens) (MPa) at the 28th day

days 28 45 90 180 (%)

(1) (2) (3) (4) (5)
“pure” A 2.31/0.93 2.48/1.38 3.53/1.33 4.47/2.99 21.03
B 2.18/1.13 4.87/1.35 5.70/2.03 4.99/1.19 21.74
air entraining agents A2 2.79/1.38 — — — 21.88
B2 3.74/1.59 21.99
water reducing agents A.2.0 1.58/0.62 — — — 23.16
fibres
cannabis A.2.1 9.371*/1.74 11,54/2.87 17.66*/2.38 — 22.81
B.2.1 15.89*/2.85 14.69*/3.09 16.45*/2.28 — 23.00
straw A.2.2 2.074/0.34 2.37/1.04 2.87/1.24 — 26.84
B.2.2 2.844/0.96 2.81/1.15 2.37/1.06 — 28.63
goat’s hairs A.2.3 2.97/0.69 5.1*/1.15 7.71*/1.29 — 27.80
B.2.3 6.77*/0.62 7.26*/1.37 5.21*/1.39 28.55
sand
sand (20%) A.3.1 2.46/0.89 2.09/1.03 1.75/1.22 — 21.99
B.3.1 3.21/1.32 3.54/1.26 1.81/- — 34.06
sand (60%) A.3.2 1.29/0.74 1.04/0.65 1.00/1.06 — 18.62
B.3.2 1.63/1.12 1.66/1.30 1.44/1.43 — 20.73
lime: pozzolan
1:2 A.4.1 1.24/0.36 0.85/0.40 0.80/0.41 — 41.31
B.4.1 1.99/0.68 1.80/0.72 1.61/0.90 — 38.61
1:4 A.4.2 1.02/0.31 0.99/0.30 0.67/0.41 — 34.57
B.4.2 1.23/0.41 1.41/0.51 1.44/0.77 — 38.83
2:1 A.4.3 0.98/0.37 0.97/0.62 0.86/0.45 — 49.39
B.4.3 1.30/0.55 1.64/0.68 2.20/0.47 — 39.55
lime: pozzolan : sand:fibres
lime: poz. 1:2 sand A6 0.92/0.12 0.88/0.10 0.34/0.19 — 39.09
B6 1.34/0.72 1.03/0.83 1.49/0.96 — 33.06
lime: poz. 1:2, sand, cannabis A7 4.64/0.27 4.32*/0.24 6.51/0.26 — 40.70
B7 10.0*/0.69 10.2*/0.50 11.4*/0.74 — 35.17
lime: poz. 1:2, sand, goat’s hair A8 2.87*/0.25 0.57/0.27 3.51*/0.27 — 40.47
B8 6,56*/0.39 6,01*/0.28 8.20*/0.45 — 35.58
lime:poz.1:1.5, sand, cannabis A10 1.0/0.21 0.808/0.30 1.12/0.29 — 35.49
B10 1.32/0.37 1.00/0.33 — — 38.96
cement
cement, sand A5 1.55/0.31 1.30/0.49 1.51/0.58 — 40.80
B5 3.44/1.04 4.03/1.22 2.86/0.88 — 33.76
cement, sand, cannabis A9 1.44/0.89 1.21/1.08 3.68/0.76 — 43.23
B9 14.3/2.08 15.9*/2.41 4.31(10.73*)/1.59 — 38.96
lime
lime A11 0.733/0.35 0.72/0.21 1.42/0.56 —
B11 1.936/1.36 1.87/1.26 — —

*Value which corresponds to 50% of slum.

The content of soluble salts in soil mortars should
be extremely low, if not non-existent, due to the well
known destructive effects of efflorescence. The layer-
ing structure of the clay particles of the soil in relation
to the presence of soluble salts underlies another
important consideration.
Relevant works concerning the structure of soil
stated that clay particle sediments in salt water solu-
tions tend to develop a highly flocculated structure
compared to those particles which sediment in fresh
water. Since a highly flocculated structure gener-
ates large voids, the mix will have low density that,
ultimately, will weaken the strength of the mortar
(Balderrama & Teutonico 1983). The capillary suction
capacity test of specimens in salted solution presented
the unfavorable influence of different components of
the compositions in function with the salt elevation.
The test results showed that the use of air entraining

895

Figure 1. Minimum and maximum percentage of water
absorption among the tested specimens.
Figure 2. Influence of air entraining agents at the capillary
suction capacity.
agents prevented the salt growth. The Figure 5 verifies
what was mentioned before.
3.3 Results of shrinkage test
According to the tests, the addition of fibres, sand,
pozzolan and the combination of them in the soil mor-
tar mixture significantly reduces the percentage of the
drying shrinkage compared to that of the “pure” soil
mortar. The shrinkage ratio of soil mortars appeared
an order of magnitude higher to the corresponding
lime and cement mortar ratio according the bibliog-
raphy. However, the tested soil mortars with cement
896
Figure 3. Influence of sand in the capillary suction capacity.
Figure 4. Influence of lime and pozzolan in specimens of
soil type A during the capillary suction capacity test.
and pozzolan in small amounts reduced the drying
shrinkage.
Compositions with air entraining agents restrained
significantly the linear shrinkage by 7–10%, compared
to the compositions without these admixtures (Fig-
ure 6). Fibres, as well, contributed to the reduction of
shrinkage; specifically, egyptian cannabis presented
the best results and hair the second best from 3% to
10% (comparison of specimen A with A.2.2 and A.2.3)
(Figure 7).
Sand addition acted beneficially to the reduction of
shrinkage. The addition of sand reduces the shrink-
age up to 20% (comparison of B with B.3.1 and A
with A.3.2). For example the addition of 20% of sand
resulted to 5% of shrinkage reduction for soil type
A, and 9.5% for soil type B (comparison of A and B

Figure 5. Influence of air entrainer agents in salt elevation
during capillary suction.
Figure 6. Drying shrinkage on “pure” compositions and
compositions with admixtures.
Figure 7. Maximum shrinkage of “pure” compositions, of
compositions with fibres and of compositions with sand.
with A.3.1 and B.3.1 respectively). The addition of
60% of sand resulted to 20% of shrinkage reduction
for soil type A, and to 9% for soil type B (comparison
of B with B.3.2) (Figure 7). The addition of pozzolan
in the lime-pozzolan compositions reduced the con-
traction significantly in comparison with the addition
of lime (Figure 8). The shrinkage was as high as 9%
897
Figure 8. Influence of pozzolan at the shrinkage.
when pozzolan in the lime–pozzolan composition was
added since contraction was significantly reduced in
comparison with the lime composition (Fig. 8).
4 PROPOSALS
Based on data collected from this experimental work,
a number of compositions seemed that could be used,
either as structural mortars or as renderings in restora-
tion practice. These compositions could not be final;
further experimental work and improvement should
be carried out, especially for criteria as durability and
compatibility to the materials to which they are applied
but it seems to be good indicators for further study.
The compositions proposed as structural soil mor-
tars were selected for their strength, hardening speed
and low shrinkage; whereas as renderings were
selected for their low level of microcracks, their
facility of water evaporation, their fluidity and their
hardening speed. The above mentioned criteria were
examined with mechanical, ultrasonic, shrinkage, and
capillary suction capacity tests.
It must be mentioned that the results refer only to
soil type A and soil type B as representative samples
of a richer and a poorer clay (argilaseous) material
respectively.
Compositions B.2.1, B9 and B.4.1 could be pro-
posed for further investigation for structural mortars.
More specifically, the main characteristics of the
compositions (given in order of their strength) were:
B.2.1. specimens were composed of soil, water and
egyptian cannabis in proportions of 34% water/u.w.
of solid materials, and of 1% egyptian cannabis
per u.w.. The egyptian cannabis, which was used
as fibre in the composition, increased strongly the
flexural strength. The shrinkage during drying pro-
cess was 7.6%; this percentage was relatively low
compared to the other compositions.
B9 specimens were composed of 73.4% soil, of
12.14% sand, of 14.3% cement, and 1% cannabis
 per u.w. and of 34.7% water per u.w. of the solid
materials. These specimens showed the second
highest strength between the tested compositions.
Cement at proportion lower to 15% proved to be
a good binder in soil compositions. Soil type B,
cement and water in combination with cannabis
increased strongly the flexural strength (at 2.08
MPa in 28 days). The B9 composition provides the
lower shrinkage (4.93%) without visible cracks in
the texture and high hardening speed.
B.4.1. specimens were composed of 75% soil, 8.3%
of lime and of 16.7% pozzolan per u.w. and of
39.1% water per u. w. of the solids. They showed the
highest compressive strength among compositions
stabilised by lime and pozzolan
Compositions B7 and B9 could be proposed for
additionally investigation as renderings. More specif-
ically, the main characteristics of the compositions
(given in order of their shrinkage) are:
B9 specimens showed the lowest shrinkage (4.3%)
and absorption of water (8.7%). Open porosity was
up to 38.9%, indicating indirectly that the evapo-
ration speed of B9 specimens is high, allowing the
mortar to “breath”.
B7 specimens were composed of lime and poz-
zolan in proportion 1 to 2. The use of cannabis
and sand restrained the microcraks during the dry-
ing process. The shrinkage test showed 6.3% of
contraction which was one of the lowest rates of
shrinkage among the tested specimens. The absorp-
tion of water during the capillary suction capacity
test was 7% and was one of the lowest percentages
among the tested compositions. Open porosity was
up to 35.58%.
5 CONCLUSIONS
Nevertheless, the variety of soils is very large; rela-
tively all the soils, except these with organic admix-
tures, could be used in construction when they are
modified and improved -stabilised in their compo-
sitions. The improvement of soils in situ may be
achieved using additives, sand, fibres and admixtures.
The criteria of the evaluation of different factors in
the behaviour of soil mortars are determined from
the position of the mortar in the structure (internal,
external wall or rendering). The strength, the change
in volume during drying process and the hydroscop-
icity were used in this research as a first approach
on the study of soil mortars. Their properties were
analysed by means of their mechanical characteristics,
their shrinkage and their capillary suction capacity.
898
Therefore, is very important to identity the soil
used through its mineralogical, chemical and physical
properties. If the strength improvement of the mortar is
the objective, then the addition of cement in sandy soils
or of lime – pozzolan in clayed soils may increase the
values of the mechanical characteristics. The addition
of thin fibres with high dispersion acts beneficially to
the increase of the flexural strength. At this work, rel-
atively high water resistance was obtained with sand
addition in the mixture or with the combination of lime
with pozzolan because of the high evaporation speed
that it was achieved. Finally, the combination of addi-
tive (cement, lime pozzolan) and sand decreased the
volume shrinkage.
ACKNOWLEDGMENTS
This research has been conducted at the Laboratory of
Building Materials of the Civil Engineering Depart-
ment of Aristotle University of Thessaloniki with the
help of prof. I. Papayianni and her support is gratefully
acknowledged.
REFERENCES
Balderrama, A. & Teutonico, J.M. 1983. Notes on manu-
facture of adobe blocks for the restoration of earthen
architecture. Rome: (ed.) ICCROM.
ICCROM (ed). 1981. International Symposium on Mortars,
Cement and Grouts used in the Conservation of Historic
Buildings. Rome. November.
Mojumbar S.C., et al. 2002. Thermophysical properties of
blends from Portland and sulfoaluminate – belite cements.
Acta Physica Slovaka, vol. 52.No. 5. October: 435–446.
Papayianni I. et al. 2001. The evolution of Porosity in
lime – based Mortars. 8th Euroseminar on Microscopy
to Building Materials, Athens:. 4–7 September.
Papayianni I. 1994. Durability lessons from the study of old
mortars and concrete., Proc. P.K. Mehta Symposium of
Durability of Concrete Nice:May.
Shafer J. & Hillsdorf H.K. 1992. Ancient and lime mortars.
The correlation between their compositions, structures
and properties. M.J. Thael, Proceedings Conservation of
Stone and other Materials. RILEM (ed): 605–613.
Theocharidou K. 1988. A Contribution to the study of brick
and Tile in the Byzantine and Postbyzantine Periods.
Bulletin of The Christian Archeological Society, vol. XIII,
Period IV. Athens. 97–112. (in greek).
Venkatarama R. & Gupta A. 2005. Characteristics of Cement-
Soil Mortars, Materials and Structures, 38(280):639–650.
Venkatarama R. & Gupta A. 2006. Strength and Elastic Prop-
erties of Stabilized Mud Block Masonry Using Cement-
Soil Mortars, Journal of Materials in Civil Engineering.,
vol. 18. issue 3. May/June: 472–476.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Experimental and numerical analyses for static retrofitting intervention

on the “Cappella dei Principi” in Firenze

A. Antonelli, G. Bartoli & M. Betti

Department of Civil and Environmental Engineering, University of Florence, Italy

ABSTRACT: The paper reports the results of a research concerning the analysis and the interpretation of the
cracking pattern on the “Cappella dei Principi” (the Medici’s mausoleum) in the church of San Lorenzo (Firenze).
The research was motivated by the sudden collapse of a keystone of an internal arch. Firstly, the principal results
obtained from in-situ surveys and laboratory tests on cored samples will be described. Then the numerical
analyses developed in order both to obtain the static identification of the monument and to identify the possible
causes leading to the observed static problems will be described. The numerical modeling has been performed
by different steps, from linear and quite simple models built with plane elements up to a non linear model with
three-dimensional elements. The non linear FEM, performed with respect to a series of in-situ measures, allowed
to reproduce both qualitatively and quantitatively the behavior of the structure and its static problems in the area
of the main arches covering the side apses.

1 INTRODUCTION

The “Cappella dei Principi” (Fig. 1), an octagonal

masonry building, is for importance and magnifi-
cence the second Dome in Firenze after Santa Maria
del Fiore. The original structure was conceived by
the Grand Duke Ferdinando I Medici as family mau-
soleum, and the work began in 1604 under the super-
vision of Arch. Matteo Nigetti. The entire structure
(crypt, chapel, tambour and dome) was built over about
40 years: the works finished around 1648. The shape of
the mausoleum remained untouched until 1740 when
Anna Maria Luisa dei Medici commissioned Ferdi-
nando and Giuseppe Ruggieri to modify (to renovate)
the external surface of the monument. In particular
the circular eyes of the tambour were turned to bell-
shaped windows, and a new external thick dome was
added to cover the existing one. The last intervention
on the Chapel ended in the sixties of the last century
when the marble covering of the Chapel floor, begun
in 1882, was finished.
As it is today the “Cappella dei Principi” is a compo- Figure 1. External view of the “Cappella dei Principi”.
sition of four principal elements (Fig. 1): a) the Crypt;
b) the Chapel; c) the Tambour and d) the Dome. walls thickness is about 6 meters. Four internal pillars
The Crypt is the lowest element of the structural support the depressed masonry vault roof. External
compound trough which is possible to get access to the walls are locally discontinuous because of the presence
monument (the original entrance was on the adjacent of both apses (one for each side of the octagon) and
church of San Lorenzo). The plan layout is an octagon the stairs accessing the upper Chapel level. Due to the
obtained by the intersection of two square elements presence of the apses the final section of the crypt is
(Fig. 2). The main dimensions are a diagonal length composed by a succession of strong and weak masonry
of 28.8 m and a height of about 6.0 m. The masonry walls.The strong walls, placed on the octagon vertexes,

899
Figure 2. Crypt layout.
Figure 3. Chapel layout (+12.0 m respect to the external
level).
continue up to the upper level of the monument; in
correspondence of weak walls at the upper level some
openings used as side apses are present.
The second element of the building is the Chapel
(Fig. 3). Main dimensions are a diagonal length of
29.7 m and a total height (floor level to tambour base-
ment) of about 21.5 m. The masonry walls thickness
is lower then the corresponding crypt walls, and the
pillars present on the lower level do not continue on
the Chapel. As for the Crypt, the Chapel perimeter
is composed of a succession of strong and weak ele-
ments. In particular, where at Crypt level a weak wall
is present, at Chapel level are present some openings
that correspond to the lateral apses (Fig. 1). The apses,
spanning from the Chapel up to the Tambour level, are
closed on top by a triple arch whose thickness is equal
to that of the upper tambour (Fig. 7). This arch sys-
tem is made by a triple order of non connected stone
masonry blocks with dimensions varying from 60 to
80 cm. Two additional arches are situated just below
the previous ones. The internal surface of the Chapel
is entirely covered with marble stones.
900
Figure 4. Tambour layout (+30.0 m).
Between the Chapel and the Dome a third element is
present in the building: the Tambour (Fig. 4), an addi-
tional octagonal element with wall thickness of about
4.3 m and a height equal to 11.8 m. On each side of the
Tambour openings are closed by a barrel vault. Under
each opening are present two empty spaces (accessible
only from outside, see Fig. 1), whose thickness cov-
ers the entire thickness of the tambour. Through these
empty spaces is possible to reach and inspect the triple
arch covering the lateral apses. From a structural point
of view, it is possible to observe that these openings
reduce considerably the tambour stiffness and the abil-
ity of the latter to counteract the radial Dome thrusts.
The last element of the “Cappella dei Principi” is
the Dome. This is an octagonal cloister dome whose
inner side dimension is about 11.4 and its inner diag-
onal is about 29.7 m, the overall height being about
22.6 m. The Dome structure includes three different
masonry layers (Fig. 5) having different centers. The
first layer is internal, with a thickness of about 90 cm;
the second intermediate layer is about 50 cm thick; the
last masonry layer is the external one, whose function
is to preserve the building from environmental loads,
with a thickness of about 30 cm. A cavity is present
between each layer with dimensions ranging from
1.2 m between the first two layers and 0.5 m between
the last two. At each vertex of the octagon a rib con-
nects the groin of the domes. It is interesting to observe
that on the extrados of the internal layer five levels of
annular iron ties are present, with a rectangular cross
section of about 5 cm of side dimension.
The global dimensions of the monument are then as
follow: inner diameter about 29 m (with masonry wall
thickness varying from 6 to 4.3 meters) and a total
height, with respect to the external ground, of 62 m.
On November 4th, 1999 a marble stone of the inter-
nal covering dropped down, revealing the presence
of static problems on the structure: one of the key-
stones of the arches supporting the tambour had fallen.
 Figure 8. Chapel sector’s identification.

In January 2002 a monitoring system was installed on

the Chapel, aimed to get a complete description of
the building behavior. Data from all sensors (mainly
temperature and displacement transducers) has been
logging by the system since January 2002.
Figure 5. Masonry Dome layers. At the same time some in-situ testings have been
performed on the masonry structure (after the removal
of the covering marble, whose thickness was about
25–30 cm).
In the following, the principal results obtained from
the in-situ surveys are described. Next, main results
from the numerical analyses carried out both to assess
the static safety of the monument and to understand
the possible causes of the damage are described. Sev-
eral numerical models has been setup, starting from a
simple linear model (with plane elements) up to a non
linear 3D model (with solid and gap elements). The
final non linear 3D FEM, identified with respect to a
series of results obtained from in-situ testings, is able
to reproduce the building behavior. Some strengthen-
ing hypotheses are then compared to obtain a proper
consolidation of the building.

2 THE DAMAGE SURVEY

Figure 6. Section of the “Cappella dei Principi”.
Damages in the “Cappella dei Principi” was noticed
at first in 1690 (forty years after its construction and
fifty years before the renovation works), revealing that
the Dome-Tambour system had been suffering since
the beginning from significant structural problems. As
noted before, the section of the building is character-
ized by a succession of strong (the octagon vertex) and
weak (the side apse) walls. This sequence of structural
elements with varying stiffness, coupled with the pres-
ence of major openings on the Tambour, has originated
a quite regular system of cracks in the structural com-
pound. To describe the crack pattern, in the following
reference will be made to the numbering of the eight
sides of the chapel from 1 to 8 following a counter-
clockwise direction (side 7 is the northern side, side 3
is the southern one), see Figure. 8. From an architec-
tural point of view even sectors correspond to strong
walls, while lateral apses are present in the odd sector.
Figure 7. Keystone (particular A).

901
On the sectors where on Chapel level a side apse is
present, there is a crack that develops from the upper
third of the dome, crosses the bell-window and reaches
the keystone of the apse arch. On the sectors where
the apse is not present (i.e. on the strong chapel walls)
the cracks, still developing from the Dome, reach the
bell-windows and do not go on over (Figs 9, 10). These
cracks are oriented along the typical meridian direc-
tions with the maximum opening at the windows level.
As to the cracks on the Dome it is possible to
observe that these fissures, present on the extrados,
don’t continue on the vaults intrados. In is interesting
to observe that the crack pattern found on the monu-
ment is quite similar to the cracks layout observed on
similar structures (see e.g. the Dome of Santa Maria
del Fiore described in Bartoli et al. 1996).
Figure 9. Internal crack (sectors 1 ÷ 8).
Figure 10. Internal crack (sectors 5÷8).
902
A complete cracks recognition in the Chapel was
quite difficult due to the internal marble covering that
made impossible a global survey. On the other hand,
on the external side only partial failures are visible on
the keystone in the outer masonry arches of the apse
sides. As mentioned before, the last damage occurred
on November 4th 1999 when a marble stone of the
internal covering in an lateral apse (sector #3 in Fig. 8)
dropped down revealing that one of the keystones of
the arch supporting the tambour and cover the apse
had moved (Figs 11 and 12).
Figure 11. Keystone failure (arch sector #3).
Figure 12. Keystone failure (particular).
 In order to investigate the monument, and to plan the
necessary restoration, all the covering marbles in this
sector were completely removed (the marble elements
are approximately one square meter with a thickness
of about 5 cm, fixed on a base of stone about 15 cm
thick anchored to the masonry by iron elements). This
removal permitted to plan an extensive experimental
survey.
The same damage found on the stone arch in sector
#3 was also present in the stone arch in sector #5 (the
high altar).
3 THE IN-SITU SURVEY
Both in-situ and laboratory tests have been made in
the “Cappella dei Principi” compound. The in-situ
tests were performed to assess the global structural
behavior and the local masonry characteristics; the lab-
oratory tests (crushing tests on cored samples) where
performed to evaluate the mechanical characteristics
of the masonry constituting material, mainly the ulti-
mate strength of the stone as well as its mechanical
properties (Binda et al. 2000). Results of this exper-
imental survey have been used to tune a numerical
model able to correctly reproduce the actual behav-
ior of the building (Bartoli et al. 1996). The model
has been built starting from results obtained from the
geometrical and architectural survey; in a following
phase the mechanical characteristics obtained from
experimental tests have been assigned to the materials.
In particular, two cored samples have been extracted
at the level of the tambour (Fig. 13). The first was taken
from the apse in sector #3 starting from the external
surface of the building, and it crosses the wall section.
The second one was taken inside the masonry wall, tak-
ing advantage of the presence of the windows. Even if
is quite difficult to extend results of local inspections
to the whole building, this inspection provided useful
information on the constructive techniques. The first
cored sample permitted to identify the presence of a
multi layered wall. The external face is made by stone
masonry with a thickness of about 33 cm; the other
external layer is made of stone masonry with a thick-
ness of about 75 cm; the internal filling is composed
Figure 13. Log position at tambour level.
903
by heterogeneous material (disordered stone tied by a
good mortar). The second cored sample substantially
confirmed the quality of the internal fill.
Tests with single and double flat-jacks were per-
formed on the arches covering the apse in sector #3,
in order to estimate the stress levels in the masonry
and to assess the elastic in-situ modulus. A thin single
(or double) flat-jack has been inserted in a slot sawn
into a mortar joint. After the creation of a slot in the
masonry, the compressive stresses present at that point
cause the closure of the masonry above and below
the slot. Therefore it has been possible to evaluate the
compressive stress state in the masonry by introduc-
ing a flat-jack into the slot and increasing the applied
pressure until the original distance between the points
above and below the slot was restored. The test is only
slightly destructive, as after the test is completed the
flat-jack can easily be removed and the mortar layer
restored to its original condition. The value of the
pressure p applied by the flat-jack approximately cor-
responds, taking into account some corrective factor,
to the local pressure in masonry. In particular the aver-
age compressive stress in the masonry, σ m , could be
calculated as follows:
In (1) Ka is a factor that accounts for the ratio between
the bearing area of the jack in contact with the masonry
and the bearing area of the slot; Km is the factor that
accounts for the physical characteristic of the jack
and Pf is the pressure required to restore the original
distance between the gauge points.
The double flat-jack technique has been used to
evaluate the masonry elastic modulus (Bartoli et al.
2000). The test is made by using two flat-jacks into
parallel (horizontal) slots one above the other; the pres-
sure into the flat-jacks is gradually increased in order
to induce a compressive stress state in the masonry
Figure 14. Flat-jack tests position (arches sector #3).
between the two slots. By this procedure the stress–
strain relation has been obtained by measuring the
deformation of the masonry. Tests have been repeated
in different positions.
The first investigated location was the keystone of
the stone arch covering the apse supporting the tam-
bour. In this point, after the creation of the slot where
the flat-jack had to be inserted, an opening of the
masonry above and below the slot was observed, this
meaning that in this area a tensile stress was present.
This arch was also tested at the spring (second loca-
tion) with single and double flat-jack tests. In this
area, the single flat-jack test reported a pressure of
23–25 bar corresponding to a stress of about 2.25–
2.45 N/mm2 .The third investigated location was on the
area close to the spring of the internal masonry arch
(Fig. 7). The aim of this third investigation was to iden- Figure 15. Crack opening under the bell-shaped window.
tify the functioning of this arch (i.e. to assess if this was
a structural arch or if it was just an architectural ele-
ment without structural relevance). The test resulting
in a pressure of 2.3 bar corresponding to a compressive
strength of about 0.22 N/mm2 . This low tension level
means that probably this internal arch does not play a
structural role and that the transfer of the vertical loads
from the Tambour to the ground is mainly left to the
upper stone arch.

4 THE NON LINEAR MODEL

Results from the experimental survey have been used

to identify numerical models of the structural com-
pound. Several model of increasing complexity have
been developed, based also on the results of a topo-
graphic survey performed over the entire complex.
Firstly, some simplified linear numerical models made
by shell elements have been analyzed. These mod-
els, built to have a first appraisal of the load transfer
mechanisms, aimed to investigate the contribution of
both Dome radial thrusts and Tambour stiffness on
the cracking phenomenon. At the same time, due to
the unknowns on material parameters (the investiga-
tion, due to the difficulties of gathering an extensive
experimental survey on the whole monument, was lim-
ited to the arches in sector #3), several parametric
analyses have been made where the elastic moduli of
the structural elements have been changed in a rea-
sonable range. The other unknown parameters have
been assumed by considering the typical values of his-
toric masonry. Results were then compared with those
obtained from more sophisticate numerical analyses,
again with FEM, by using solid elements. The most
interesting results of this first phase are reported in
Figure 16.
Basically, it is possible to observe a migration of
stresses from the Dome to the ribs and then to the
strong walls of the tambour. Due to this load migration
(facilitated by the bell-shaped windows openings on
Figure 16. Load pattern.
the Tambour) the stone arch over the apse on sector
#3 is characterized by an absence of normal stresses
at the crown level (as resulting also by the in-situ test
where tensile stresses have been found).
After this first phase, a more sophisticated model
by using ANSYS finite element program (Ansys Inc.
1992) has been built. Masonry elements have been
modeled by means of Solid45 elements assuming a
homogenized approach. The model accurately repro-
duces the geometry of the structure, paying attention to
any variation in wall thickness, irregularities and wall
connections. The major openings in the building are
reproduced as the building complexity asks for a global
and detailed model capable of representing both the
overall spatial configuration and the entire set of archi-
tectural elements with structural relevance. This is
particularly necessary in historic buildings where dif-
ferences between architectural or structural elements
are not always clear. The final 3D model consisted
of 107689 joints, 85771 3D Solid45 elements, corre-
sponding to 220332 degrees of freedom (Fig. 17). In
a preliminary step, the structure has been modeled by

904
Figure 17. Finite element model of the “Cappella dei Figure 18. Link10 insertion points.
Principi”.

assuming fixed restraints with respect to the ground;

next, in order to take into account the structure-soil
interaction, spring (Link8) elements have been inserted
at the building base; their stiffness has been evalu-
ated on the basis of geotechnical characteristics of the
ground.
Material properties of masonry walls (Young mod-
ulus E, Poisson coefficient υ, own weight W ) have
been differentiated taking into account each different
area present on the building.
The cracking pattern has been reproduced by insert-
ing gap elements (Link10) where cracks were present
on the buildings. These elements are non linear
elements able to transfer only compressive forces; Figure 19. Link10 properties evaluation.
an initial pretension corresponding to 0.1 N/mm2
(approximately the tensile strength of a good masonry)
was assumed. This approach has been preferred to the
smeared crack model because cracks position is quite
clear and the aim of the research was not to assess
the cracking pattern. In order to insert these non linear
elements, the nodes under the windows were dou-
bled and then re-connected by the Link10 elements
themselves (Fig. 18).
The stiffness of the elements has been estimated
taking into account the contribution due to the corre-
sponding area (Fig. 19). The structural compound has
been analyzed under constant vertical loads deriving
from the self weight and from the loads transferred by
the roof.
The final model is able to reproduce the relevant
results of the experimental survey. In particular a quite
Figure 20. Stress state at the arch spring (kg/cm2 ).
good adherence between numerical and experimental
results has been achieved on the spring area of the stone
arch (Fig. 20). The compressive stress level varies from on the first two ashlars of the arch, as seen from the
2.1 to 2.5 N/mm2 ,with a mean value close to the one experimental survey (Fig. 7).
obtained from the experimental results (ranging from The obtained model was also able to reproduce the
2.25 to 2.45 N/mm2 ). The model is also able to pre- behavior of the crack in the Chapel between the arches
dict the damage that was present on the stone arch and the upper bell-shaped windows as reported in Fig-
with accuracy. In Figure 21 tensile stresses visibly arise ure 22 (to be compared to Figure 15). On the crypt

905

Figure 21. Stress state at the arch keystone (kg/cm2 ).
Figure 22. Stress state under the bell-shaped window
(kg/cm2 ).
vaults at keystones level it is possible to observe the
development of a diffused tensile stress state varying
from 0.2 to 0.5 N/mm2 . Even if these tensile stresses
are not high enough, they could be responsible for
the cracks; this stresses are due to different vertical
vaults spring displacement. On one end these vaults
are connected with the external walls, while in the
central area they are supported with internal pillars.
Due to the fact that these pillars stop at the crypt level,
they have a different displacement with respect to the
external walls.
5 RETROFITTING PROPOSAL
The behavior observed in the monument could be
explained by several factors, the main one being the
stiffness differences among the walls of the Chapel.
Despite apparently masonry walls are quite similar
one to the other, from a structural point of view there
is a succession of strong (even sectors in Figure 8)
906
and weak walls (odd sectors in Figure 8) used as lat-
eral apses. Above the chapel these lateral apses are
closed by a complex system of arches constituting
the basis for the Tambour. This geometrical difference
probably has caused (as observed with the numerical
model) changes on the stiffness of the Tambour that
have modified the normal stress state of this archi-
tectural element. The Tambour area above the strong
walls is stronger than the corresponding one over the
weak walls. This could be considered the origin of
the switching of the stress state. Due to this deviation,
unloading on the arches has probably originated the
sliding on the keystone, also due to the specific shape
of this keystone (a parallelepiped instead of a wedge).
The retrofitting needs to take into account these phe-
nomena and it must cover these two points. Firstly, it
is necessary to create cohesion between the stone ele-
ments of the arch; secondly is important to increase the
global tensile resistance of the arch taking into account
that, probably, in the rest of its life the described phe-
nomena will be amplified. Another aspect that needs to
be considered is that the intervention must be designed
as a long term intervention. The proposals described
in this paper foresee to re-establish the arch cohesion
by means of internal steel bars that connect the three
levels of stone on the arch. At the intrados an external
covering with FRP is suggested to ensure a general
improvement of the arch functioning.
6 CONCLUSIONS
The paper presented the main results of a research
aimed to the analysis and interpretation of the crack-
ing pattern in the “Cappella dei Principi” in Firenze.
Firstly, the in-situ survey was described; then the
numerical analyses developed in order to obtain the
static assessment of the monument was illustrated.
The technique of the finite element model has been
applied, allowing to identify the static behavior of the
building. The probable causes leading to the observed
static problems have been described. The non linear
FEM, performed with respect to a series of in-situ
measures, allowed to reproduce the behavior of the
structure and the static problems in the area of the
main arches covering the side apses. It made pos-
sible to identify the most significant aspects on the
structural behavior of the buildings. One of these is
the importance of the succession of strong and weak
walls. The connection of these walls is made by stone
arches that cover the apse and create the basis for
the above tambour. These arches originally conceived
to convey the loads from the tambour to the strong
walls, tend on the contrary to weaken the structure
(as proved by the substantial absence of compressive
stresses at the arch crown).The effect of the differential
settlement is not significant for the developing of the
cracks on the Tambour-Dome systems: as a matter of
fact, the described cracks would have been developed
also assuming a fixed base model (i.e. not considering
structure-soil interaction). On the contrary, when the
latter is assumed, a development of crack on the crypt
vaults is also observable (as visible on the monument).
Results herein reported offer a first identification of
the buildings behavior. A more exhaustive interpreta-
tion of the overall building functioning will require an
interaction between several modeling strategies (i.e.
limit analysis, non-linear analysis) each of which can
offer significant elements.
Eventually a retrofitting proposal was presented
considering that engineers involved in monumental
buildings assessment are called to pay attention both
to the economic and cultural relevance of this kind of
building, next to the safety requirements.
REFERENCES
ANSYS INC. 1992. Users’s manual. Swans. analysis
systems.
Bartoli, G., Chiarugi, A. & Gusella, V. 1996. Monitoring
systems on historical buildings: the Brunelleschi Dome.
Journal of Structural EngineeringASCE 122(6): 663–673.
907
Bartoli, G., Chiostrini, S. & Innocenti S. 2000. Problems
related to the analysis of experimental data from flat-jack
tests. Atti del Convegno CICOP2000. 5th International
Congress on Restoration of Architectural Heritage.
Binda, L., Saisi, A. & Tiraboschi, C. 2000. Investigation
procedures for the diagnosis of historic masonries. Con-
struction and Building Materials 14(4): 199–233.
Chiarugi, A., Bartoli, G. & Morano S.G. 1998. The surveil-
lance of Brunelleschi’s Dome in Florence. Proceeding of II
International Seminar on Structural Analysis of Historical
Constructions, Barcelona.
Chiarugi, A., Fanelli, M. & Giuseppetti, G. 1983. Anal-
ysis of a Brunelleschi-Type Dome Including Thermal
Loads. IABSE Symposium on Strengthening of Building
Structure, Diagnosis and Therapy, Zurich: 169–178.
ICOMOS 2001. Recommendations for the analysis, conser-
vation and structural restoration of architectural heritage.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Comparison the effect of different repair method applying on

masonry arch bridges

M. Miri
Department of Civil Engineering, University of Sistan and Baluchestan

T.G. Hughes
Cardiff School of Engineering, Cardiff University

ABSTRACT: Small scale centrifuge models were used to study the behaviour of arches repaired using different
techniques. The models under test were 1/12th scale replicas of a 6 meter single span three ring arch. Two types
of arch geometry, with span/rise of 4 and 2, were studied as a shallow and deep arch geometry. The models were
tested in a centrifuge under a steady equivalent gravity of 12 g.
Two types of 2-D and 3-D arch models were studied which had the same geometry but different in the addition
of spandrel walls. The 3-D models were built with spandrel walls but without any wing walls or parapet. The
models were usually tested with fourteen passes of a rolling load and then up to the observation of first signs
of failure to enable them to be suitable for applying a repair method. The repaired models were tested using the
same procedure but up to the full failure load.
Plastic mesh reinforcement, stitching, and concrete slab on top of the soil backfill were applied as repair
methods to the 2-D arch models. Stitching of arch barrel and the barrel to the spandrels, applying partial saddle
concrete and strengthening of spandrel wall using reinforced concrete were tested in the 3-D arch models. The
results are presented and compared with each other in this paper.

1 INTRODUCTION The objective of the present study is to try to

Arch masonry bridges are used from many years
ago around the world. In some countries these struc-
tures are a significant part of their transport system.
Although these structures have a good resistance under
load and need low maintenance during their lifetime,
the traffic and load applied to them has increased
many times during recent years. Therefore repair and
strengthening of arch bridge is an important problem
for owners. On the other hand some of these structure
are a part of our historical structure and we need to have
more attention, maintenance and strengthening them.
In recent years there has been significant develop-
ment of experimental and numerical research tech-
niques into modelling repair and strengthening for
masonry arch bridges. Some of these tests were
carried out on full scale models (Sumon, S.K.,
Melbourne, C.M.,) and some using smaller scale mod-
els; work has also been undertaken looking at the
performance of actual repairs undertaken on working
full scale structures (Ashurst, D.). Arch stitching, arch
reinforcement, concrete saddles, backfill strengthen-
ing and sprayed concrete to the intrados are the most
common ways of strengthening currently used.
experimentally quantify the differences between the
behaviour of two dimensional and three dimensional
arches under loads and applying the various effects
of strengthening methods on them. The principle dif-
ference between the two models is the inclusion of
the spandrel walls in the 3-D model. The spandrel
walls are however not extended to form parapet walls.
Most current UK assessment methods do not consider
the strengthening effects of the spandrel or parapet
walls because their contribution at ultimate load is sus-
pect. However their contribution under service loading
is more assured and therefore quantification of their
effect represents a significant goal.
2 ARCH GEOMETRY AND MATERIALS
The arch models were made from brickwork formed
from scaled bricks. The bricks were 1/12 full size scale
in width and 1/6 full scale size in length and depth.
The model bricks were cut from full scale bricks using
precision indexed saws. The arch barrel was made of
three rings of bricks with a UK mortar joint type v
with a mix content of 1:3:12 (Cement: Lime: Sand) by

909
Table 1. Detail of models under test.

Parameter Dimension

Intrados span (mm) 500

Span to rise ratio (Shallow arch) 4
Span to rise ratio (Deep arch) 2
Fill depth at crown (mm) 30
Arch ring thickness (mm) 30
2-D width (mm) 345
3-D width (mm) 405
Spandrel wall thickness (mm) 30
Mortar mix (cement: lime: sand) 1:3:12
Whole brick compressive strength (N/mm2 ) 96
Backfill unit weight (kN/m3 ) 20.5
Angle of friction 53

volume. This technique has been used and established

in previous works (Baralos, P., Burroughs, P. O.)
The same bricks but with a different size were
used to build the spandrel walls. The 2-D and 3-D
arches were built with essentially the same geometry,
masonry, mortar and fills. Details of models and tests
are given in Table 1.
The deflections of the arches were measured via
rows of Linear Variable Displacement Transducers
(LVDT) located at 2%, 15%, 25%, 35%, 50%, 65%,
75%, 85% and 98% of the span. For the 2-D arches two
rows equidistant from the centreline were used for the
3-D tests the three rows were located along the cen-
treline and close to the front face and back face of the
arches. In addition vertical movements on the top of
the spandrel walls were similarly measured. Readings
were recorded at 6 second intervals throughout both
the service and ultimate load tests. The soil/masonry
normal pressures were measured by small diaphragm
pressure transducers (manufactured by Kyowa) which
were preinstalled in bricks and then laid within the
brickwork with the other brick units. Pairs of these
pressure gauges were placed across the arch usually
at 7%, 15%, 65%, 75% and 85% of the arch span. A
general view of arch model is presented in Figure 1.
3 LOAD SYSTEM AND MEASUREMENTS
3.1 Service load system
Three in-line steel rollers (equivalent to 15 tonnes on
a 2.5 m axle) simulated the service load during the
tests. These rollers were rolled on the top of the fill
from a point above one abutment to the other. Fourteen
passes of the roller were carried out in each test and
the position of the roller was recorded and referenced
to the other monitoring equipment.
3.2 Knife load system
In each model after the service load was applied
the centrifuge was stopped and the rolling load was
Figure 1. Model package, general view of model.
replaced by a 20 mm knife edge loading system across
the full width of the arch. The load was applied on
the top of the backfill at the 25% and 30% of the
span for the shallow and deep arches respectively. Both
geometries of models loaded until the observation of
the first sign of failure in model except the first 2D
model which was loaded to failure as a benchmark
test. The model was then subsequently repaired by the
applying of various strengthening methods and tested
again by the same method.
3.3 Soil/masonry interaction arch deflection
measurements
The pressures between the arch barrel and the back-
fill were measured using 6 mm diameter diaphragm
(Kyowa) pressure cells installed within special brick
units in the outer (extrados) layer of arch barrel. Two
cells were located across the arch at a number of
sections.
The arch deflections were measured normally using
three rows of displacement transducers (LVDTs). One
LVDT row was installed along the centerline of the
arch, the other two rows parallel to the first but
close to the edges (spandrel) faces. The LVDTs were
installed normal to the arch barrel to measure the radial
deflections of the arches.
4 REPAIRED MODELS
Models were repaired using different techniques after
completion of the benchmark tests. Repairs methods
were selected according to the benchmark model con-
dition. Three 2-D and three 3-D models were repaired
using four different repair methods. In the case of a

910
significant effect of the repair method in arch load
capacity, the repair was repeated to the deep arch
geometry.
4.1 Plastic mesh reinforcement
The second 2-D model (S2D-2R) was repaired using
plastic mesh reinforcement. Two mesh layers of
TENAX Promat (TENAX ) which was produced by the
TENAX UK Limited Company for ground reinforce-
ment and erosion control, were placed in the backfill.
Top layer of the backfill was removed from the arch
barrel, mesh layer was installed and the backfill was
replaced again.
4.2 Concrete slab repair
Following the initial tests, S2D-3R and D2D-1R the
arches were repaired by laying a reinforced con-
crete slab on top of the backfill. In this case, the
un-strengthened tests had a crown backfill depth of
13 mm to readily facilitate the placement of the con-
crete without overly distributing the damaged arches.
The overall depth of construction over the crown of
the arch, including the 17 mm of concrete layer in the
strengthened arches, was 30 mm.
The concrete itself was manufactured with 2.0 mm
aggregate as the coarse material, Chelford 95 silica
sand as the fine aggregate and OPC. (BS 12 1991),
with mix proportion of 1:1.8:2.8:0.6 (cement: fine:
coarse: water) by mass. Compressive strength tests
on 25 mm concrete cube samples yielded 56 N/mm2
according to British Standard (BS 1881-116). The
model concrete was nominally reinforced with a mesh
of type 304 manufactured of 0.8 mm mild steel at
20 mm centres.
4.3 Stitching the arch barrel
The first 3-D model was repaired by stitching the
arch barrel with steel bars between the barrel and
the spandrel walls joints. The patterns of holes with
an angle of 45 degree were drilled in the arch barrel
and 2.4 mm diameter stainless steel bars type 304 was
placed in them.
4.4 Partial saddle concrete
Partial saddle concrete to the extrados of the arch was
employed to repair the second 3-D arch. Strengthening
the extrados of arch barrel is an appropriate method
of repairing arches. The advantage of this method is
that it not only strengthens the arch but also improves
load distribution and ties together any cracked sections
(Department of Transport).
A 60 mm width of concrete slab was laid on both
edge of the arch barrel from one abutment to the
911
other one in the longitudinal direction. The saddle con-
crete is stitched to the spandrels walls to avoid any
disconnection between the spandrel and barrel.
Two rows of holes with a depth equivalent to one
ring were initially drilled to the arch barrel and 2 mm
steel rods were installed in them using epoxy resin.
These rods had an extra length of about 12 mm from the
hole depth through the saddle and connected the barrel
extrados to the saddle concrete. A width of 60 mm in
model scale (that means 720 mm in prototype scale),
was used which should be available as a pavement on
most of arch bridge or may be provided by restricting
a single lane of traffic to a narrow lane.
4.5 Spandrel wall strengthening
The third and fourth 3-D arches were repaired with
strengthening of their spandrel walls using a rein-
forced concrete slab. After completing the benchmark
tests, models were repaired by applying reinforced
concrete to the inner sides of the spandrel walls. The
location of the concrete was restricted to those parts
of the structure that would in normal circumstances
be readily accessible for such work. In the present
study the vertical extent of the reinforced concrete was
restricted to 100 mm in depth (about 1.2 m in the pro-
totype) this would require only limited support during
construction.
5 TESTS RESULTS
Laboratory experiments have successfully been car-
ried out on two different geometries of 1/12th scale
single span centrifuge arch models. The experiments
provided useful information on the effectiveness of the
tested repair methods on the failure mechanism and
particularly on the service and ultimate load capacity
of the arches. Final comparisons of all the repaired
arches are presented in Figures 2 and 3 for the 2-D and
3-D model tests.
The general conclusions are:
– The use of plastic mesh reinforcement, test S2D-2R,
had no significant effect on arch load capacity and
pressure distribution under applied loads.
– The relieving concrete slab on top of the backfill sig-
nificantly increased the ultimate arch load capacity
for both shallow and deep arch geometries. The load
at failure of the shallow strengthened model was
3.4 times that of the benchmark model and 2.7 of
the average benchmark. The results for the repaired
deep arch were, respectively, 3.7 and 3.2 times as
strong as the benchmark arch. The application of
the slab to the surface appears to be at least equiva-
lent to application directly to the arch intrados and
extrados.A concrete slab on top of the fill distributes
the pressure and decreases the recorded pressure on

Figure 2. Comparison between benchmark and repaired
2-D Failure load.
Figure 3. Comparison between benchmark and repaired
2-D Failure load.
the arch barrel significantly. A significant decrease
in arch barrel deflection was observed under rolling
loads.
– The first shallow 3-D arch was repaired using stain-
less steel stitching bars applied to the arch barrel
912
to connect the extrados ring and spandrel walls to
each other. Test results showed no prevention of ring
separation by the stitching bars but an increase of
50% in the ultimate arch load capacity following
application of this method.
– Applying a partial saddle concrete on part of the arch
barrel in addition to steel bar connection of the con-
crete to the spandrel increased the arch load capacity
and the stiffness of the arch. The arch load capacity
was improved by about 215% for the repaired arch,
which is comparable with applying the same con-
crete on top and beneath the barrel in the 2-D arch
model tests.
REFERENCES
Sumon, S.K. Repair and strengthening of five full scale
masonry arch bridges. in Second International Conference
on Arch Bridges. 1998. Venice, Italy.
Melbourne, C., M. Begimgil and M. Gilbert. The load test
to collapse of a 5 M span brickwork arch bridge with
tied spandrel walls. in Arch Bridges. 1995. Bolton: Tomas
Telford.
Ashurst, D. An assessment of repair and strengthening tech-
niques for brick and stone masonry arch bridges. 1992,
Transport Research Laboratory: Crowthorne.
Baralos, P. (2002). “The small-scale modelling of repair tech-
niques for masonry arch bridges using a geotechnical
centrifuge.,” PhD, University of Wales, Cardiff.
Burroughs, P. O. (2002). “A study of parameters that influence
the strength of masonry arch bridges using a geotechnical
centrifuge,” PhD, University of Wales, Cardiff.
TENAX. (2002). “Ground Reinforcement and Erosion Con-
trol, Product Guide and Price list." Tenax UK Limited.
BS 1881-116. (1983). Testing concrete – Method for determi-
nation of compressive strength of concrete cubes, British
Standard Institution, London.
Department of Transport. (1997a). “BA 16/97 – The Assess-
ment of Highway Bridges and Structures.” HMSO,
London.
Use of traditional, alternative and
innovative materials
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Restoring of timber structures: Connections with timber pegs

C. Ceraldi, V. Mormone & E. Russo Ermolli

Department C.M.M.A. University Federico II, Naples, Italy

ABSTRACT: A quite speared methodology of restoration envisages the introduction of new timber elements,
connected to the standing structure. An efficient alternative to the most used jointing technologies, which employ
metallic bolted plates, steel pins, or epoxy resin adhesives, is the use of timber pins. The design rules for pinned
joints given by Eurocode 5 are based on Johansen’s theory. The present research has the target of verifying the
applicability of this theoretical model, and of the derived practical formulas, to the case of timber pegs. At this
aim an experimental research has been done, and obtained results have been compared with theoretical ones,
showing as the different nature of the material employed for the pins needs more specific evaluations of the
“yielding moment” and of the failure mode in case of low slenderness. Furthermore, an appreciable ductility of
timber connections made with timber pegs has been detected.

1 INTRODUCTION
1.1 Employing timber pegs
In the field of ancient timber structures restoration,
the need of substituting rotten portions of structural
elements, preserving the original spatial configura-
tion without modifying the structural behaviour, is
quite frequent. In particular, in ancient trusses and
floors, taking off the end portions of rafters or beams,
damaged by biotic agents often due to high mois-
ture content inside the masonry walls, is sometimes
inevitable.
A quite speared methodology of restoration envis-
ages the introduction of new timber elements, made
of massive wood or glulam, connected to the standing
structure. This connection is generally realized with
metallic bolted plates, steel pins, or epoxy resin adhe-
sives (Mormone & Russo Ermolli 1999). Those joint-
ing technologies show some disadvantages as moisture
condensing problems in the interface between timber
and steel elements; the questionable aesthetic output
in uncovered structures; the difficult reversibility of
the intervention with resin adhesives.
An efficient alternative is the use of timber pegs in
timber to timber connections as all those handicaps
can be overcome.
An example of joints made with timber pegs has
been realized in the restoration of the timber trusses
in the Pieve of San Marino (San Marino Republic),
(Ceccotti et al. 1998). The extremities of rafters and
tie-beams, which showed a high level of decay, have
been substituted by timber prosthesis connected with
a tenon – mortise joint, reinforced by an epoxy resin
gluing and timber pegs made of Azobè wood, finally
giving to this gluing the function of granting the con-
tinuity to the connection, while timber pegs strength
has been verified in case of exceeding slipping of the
gluing or its yielding. The evaluation of the design
strength of the double shear plane dowel joint has been
made following Eurocode 5, introducing in the for-
mula calculating the bolt yield moment, the timber
peg characteristic bending stress value. At the IATF
Institute in Florence an experimental research has been
done with specimens having the same dimensions of
the real used timber elements and the same wood
essence (silver fir), making glued and bolted joints,
with the really employed configuration as well as var-
ied ones, fastened with timber pegs of Oak and Azobé,
and steel bolts. In the case of timber pegs the rupture
mode has most frequently been that of peg bending
failure.
Another experimental investigation has been done
by Menegotto (2001) at the Architectural Construction
Department of IUAV in Venice on double shear plane
bolted joints made of fir glulam, connected by tim-
ber pegs of beech wood with variable diameters (30
and 50 mm). The author has noticed that the most fre-
quent rupture mode has been due to tension failure of
timber pegs.
Actually the design rules for pinned joints given by
Eurocode 5, the normative for timber structures which
ought to be acknowledged by each European coun-
try, are based on Johansen’s theory, (Johansen 1949),
named EYM, (EuropeanYield Model), which assumes
a perfectly plastic behaviour for the connected timber
elements as well as for the metallic pins.

915
Table 1. Design formulas of bolts.
Rd = Peg Design Strength
Mode N
I fh,1,d t1 d
II 0.5fh,1,d t2 dβ

fh,1,d t1 d 4β(2+β)My,d
III 1.1 2+β
2β(1 + β) + fh,1,d dt12
−β

2β 
IV 1.1 1+β
2My,d fh,1,d d
The aim of the present research is verifying the
applicability of this theory, and of the derived design
formulas, in the case of timber pegs used as fasteners,
and it’s based on an experimental investigation made
at the Testing Laboratory on Materials and Structures
of the COMMA Department of Naples University
“Federico II”.
1.2 Design formulas based on Johansen’s Theory
Eurocode 5 (UNI ENV 1-1 – 6.2.1) gives the design
formulas to evaluate design bearing strength for fas-
tener of cylindrical shape in double shear plane bolted
joints, quoted in Table 1.
where: t1 , t2 are timber thickness; fh,1,d (fh,2,d ) is dowel
bearing strength in t1 , (t2 ); β is the ratio fh,2,d /fh,1,d ;
d: bolt diameter; My,d : yield bending moment of the
bolt.
As already said, these formulas have been derived
from Johansen theory, dated 1949, relying upon the
hypothesis of elastic – perfectly plastic behaviour of
timber and of the connecting devise.
Each evaluation of the bolt design load, for each
shear plane, corresponds to an envisaged failure mode:
while in the first two formulas the dowel bearing
strength of the jointed timber is overcome before any
plastic deformation of the bolt occurs, in the other two
the creation of one or two plastic hinges in the bended
pin takes place. Obviously the bearing strength of the
bolt in the joint is the less of the four values calculated
with the previous formulas.
The characteristic dowel bearing strength fh,0,k of
the jointed timber elements can be evaluated with the
Formula 1, (UNI ENV 1995 1-1, 6.5.1.):
where ρk is the specific gravity of timber (kg/m3 ); d
is the diameter of the bolt (mm).
The bolt yielding moment value is given by the
formula in paragraph 6.5.1.2 of the same Code:
where fu,k is the characteristic tension strength of steel.
In Formula 2 the assumed steel behaviour, which is
the same in traction and in compression, induces the
yielding limit value when the limit tension strength is
reached with linear stress distribution on the circular
cross section.

2 EXPERIMENTAL PROGRAM
At the aim of studying the applicability of this the-
ory, and of the derived design formulas, in the case
of timber pegs used as fasteners, the basic hypothesis
and also the matching of the collapse modes envis-
aged by Johansen must be verified. The definition of
the “yielding moment” of the timber connecting device
is also needed, which is employed when the joint col-
lapse passes through the creation of one or two plastic
hinges in the pin.
So an experimental research in different phases has
been carried on:
– characterization of the two different wood species
initially chosen for making the pegs, as present-
ing large specific gravity, durmast and ash, through
measurements of their density, compressive, bend-
ing and cutting shear strengths;
– characterization of timber to be used for mak-
ing the connected elements, fir glulam, through
measurements of its density and compressive
strength;
– rupture tests on double shear plane joints, on three
sample, each one constituted of three specimens,
made of three connected timber boards with vary-
ing thickness ratios, to induce the different collapse
modes envisaged by the theory. At this aim, the
connections have been designed to avoid the cut-
ting shear failure of the timber pegs, obviously
absent in the Johansen’s model with metallic bolts.
Through the readings of mutual slipping, made
employing LVDT sensors, the behaviour of the con-
nections under increasing loads has been studied,
particularly in reference to evaluation of the joint
ductility.
2.1 Characterization of timber chosen
for the pegs
For both wood species the following mechanic char-
acteristics have been evaluated: parallel to grains
compressive strength; bending strength on clear spec-
imens; cutting shear strength; the corresponding mean
and characteristic values are quoted in Table 2.
Tests have been made with an electronic board uni-
versal load machine, in displacement control, Sun 5,
produced by Galdabini, with a maximum load of
50 kN, and, for larger loads, with a oil-pressure uni-
versal load machine of 2000 kN max capacity. Tested
specimens have been conditioned in an ambient with
916
Table 2. Mean and characteristic values for ash and durmast Table 3. Mean and characteristic values of fir glulam.
timber.
Fir glulam
Ash Durmast
Properties Mean Charact.
Properties Mean Charact. Mean Charact.
Density (kg/m3 ) 447 392
Density (kg/m3 ) 652 572 680 597 Compression (MPa) 38.17 31.95
Compression (MPa) 56.29 49.42 51.01 44.31
Bending (MPa) 112.48 84.61 75.81 54.85
Shear (MPa) 31.90 23.86 26.10 20.65
has been loaded in compression to apply shear on the
two bearing cross sections.
The obtained results suggest choosing ash timber to
make the pegs, as even if its density is less than that
of durmast, its mechanical characteristics show higher
values.

2.2 Characterization of fir glulam

As the number of specimens constituting the samples
of double shear plane joints is small, using glulam
seems more appropriate, to reduce the influence of
defects.
Evaluation of density and parallel to grains com-
pressive strength, made as described in the previous
paragraph for ash and durmast timber, has been made
on a sample of 30 specimens. Mean and characteristic
values are quoted in Table 3.
Figure 1. Apparatus for cutting shear strength measure-
ment.
2.3 Tests on double shear plane joints
temperature of 20 ± 2◦ C and relative humidity of
At the aim of studying the behaviour of the designed
65 ± 5%. The load rate employed has been that induc-
timber connection, double shear plane joints have been
ing specimen failure in an interval of 300 s ± 120 s, as
tested, made of medium and lateral elements with vary-
prescribed by Code UNI EN 408. The ash and durmast
ing thickness. All the boards have the same lateral
samples are made of about 45 specimens each one.
dimensions, 200 × 300 mm, and are joined by a peg
Density of wood samples has been measured on
of 30 mm in diameter, disposed in the middle of the
clear prismatic specimens made for the following com-
boards, so respecting end distances prescribed by the
pression tests (nominal dimension: 20 × 20 × 50 mm).
Eurocode in case of steel bolts: 7D from the loaded
For each prism real measures have been taken with
side, 3D from the lateral side, where D is the bolt
a digital millimetre calliper and the weight has been
diameter.
valuated with an electronic weighting machine, with
Compression load tests have been made with the
centesimal gram precision.
electronic board universal load machine, in displace-
The test settlement (UNI EN 408) used to measure
ment control, Sun 5. The simultaneous registration of
bending strength envisages a clear span of 18 h (h being
compressive load and of the corresponding slipping of
the specimen cross section height), with load applied at
the middle board under load, has been done, with a load
2/3 of the span. Specimens have nominal dimensions
cell of 50 kN capacity and an LVDT of 50 mm of max-
of 20 × 20 × 400 mm.
imum stroke, respectively. The two transducers have
The measurement of cutting shear strength of timber
been connected to an electronic power station with 40
pegs has been made on samples constituted by cylin-
channels distributed by VISHAY MEASUREMENTS
drical turned specimens 100 mm long with a diameter
GROUP.
of 30 mm. Tests have been done employing a purposely
The three tested samples, each constituted of three
made device, constituted by two “U” steel profiles,
specimens, have been made with boards of the follow-
with round holes on the wings, 30 mm in diameter,
ing thicknesses:
allowing the insertion of the timber peg, (Fig. 1).
These profiles can slip one upon the other with a – TYPE A: t1 = 15 mm, t2 = 22.5 mm;
minimum allowance so that the two loaded peg cross – TYPE B: t1 = 30 mm, t2 = 45 mm;
sections undergo to pure shear stress. The apparatus – TYPE C: t1 = 60 mm, t2 = 90 mm.

917
 Table 4. Mean values of the ultimate load and of the
corresponding slipping for double shear plane joints.

Ultimate load Ultimate displacement

Sample N mm

Type A 24,977 9.565

Type B 28,082 10.653
Type C 40,556 14.312

Table 5. Johansen’s failure values, for double shear planes.

Sample R1 R2 R3 R4

Type A 23,094 17,320 27,498 38,860

Type B 46,188 34,642 29,476 38,860
Type C 92,376 69,282 40,934 38,860

3 ANALYSIS OF RESULTS FOR THE

DOUBLE SHEAR PLANE JOINTS

3.1 Application of Johansen’s formulas

Figure 2. Examples of the three types of tested samples.
where t1 is the thickness of the lateral boards and t2
that of the central one (Fig. 2).
In Table 4 mean values of the ultimate load and of
the corresponding slipping are quoted, for each tested
sample.
Observed failure modes for Type B and C have
always been bending failure of the pegs in the central
cross section or in the two lateral ones, respectively.
Sample of Type A, instead, hasn’t shown an uniform
failure mode, as in the first specimen the splitting of
the central board occurred while in the other two there
was the bending failure of the timber peg.
To compare the experimental results with the theoret-
ical ones obtainable employing Johansen’s formulas,
the definition of the dowel bearing strength of the fir
glulam and the yielding moment of the ash timber peg
are needed.
As no adequate experimentation has been done to
evaluate the dowel bearing strength of the fir glu-
lam, Formula 1 has been used with d = 30 mm and
ρmean = 447 kg/mc, as quoted in Table 3, obtaining
fh,m = 25.66 MPa.
The yielding moment has been evaluated employing
Formula 2, inserting the mean compressive strength of
ash, quoted in Table 2, fc,0,m = 56.29 MPa, obtaining
My,mean = 202,644 Nmm. This procedure is justified
as the transition to the plastic phase in clear timber
subjected to bending loads is characterized by the over-
coming of the parallel to grains compressive strength
elastic limit, on the compressed side, (Giordano 2005).
Using these mean values just defined, the bearing
load for each shear plane of the connecting device has
been evaluated with formulas in Table 1. In Table 5
the obtained values are quoted referring to the double
shear planes.
In Table 5 the highlighted values, for each type of
sample, correspond to the lowest ones, those which
must be taken in account for comparison with the
experimental data quoted in Table 4. The failure modes
envisaged by the theoretical prediction are:
– TYPE A, slipping of the central timber board;
– TYPE B, bending of the peg with a single plastic
hinge;
– TYPE C, bending of the peg with two plastic hinges.

918
3.2 Comparison between theoretical and
experimental results
From Tables 4–5 comparisons, the following observa-
tion can be deduced:
– for samples of type B and C, the experimental ulti-
mate load diverges of only 4% (−4.73 and +4.36
respectively) from the theoretical value and the
failure modes observed in the tests are the same
of those obtained by the application of Johansen’s
formulas;
– for sample of typeA, the experimental ultimate load
overcomes the theoretical one of 44.21%; on the
other hand the failure modes shown by tests are not
uniform and only for one specimen the splitting
of the central timber board occurs, as theoretically
predicted. In fact, experimental results point out as
more determining for specimen failure the bending
behaviour of the peg than the dowel bearing strength
of the fir timber boards. Figure 3. Bending of the timber peg.
Even if the reduced number of specimens, only
three for each sample, doesn’t allow meaning gener-
alization, some interesting observations can be done.
Failure loads given by Johansen’s formulas can be
assumed to predict experimental data, in fact even
when they show greater values, as in the case of sample
type B, the difference is negligible. The strong diver-
gence between theoretical and registered values, which
occurs in the case of sample type A, together with the
observed absence of matching of the failure modes,
reveals the inadequateness of corresponding formulas
in the particular case. In fact to R1 and R2 theoretical
values it’s associated a failure mode only due to dowel
bearing strength of the jointed boards timber, while in
two of the three tested specimens the timber peg shows
a relevant bending (Fig. 3), strong enough to give rise
to joint failure.

3.3 Ductility
The analysis of the load – slipping diagrams of
each specimen, obtained by transducers registrations,
allows some evaluations relative to the ductility of
these joints constructed employing timber pegs. On
the diagrams quoted in Figure 4, which are the most
representative of each sample, the loads corresponding
to the change of slope in the load-deformation curve
have been chosen as yield limits.
These loads values and the corresponding deforma-
tions are quoted in Table 6.
The index of static ductility D is defined as the
ratio between the ultimate displacement and the one
corresponding to the yield limit.
In the last column of Table 6 are quoted the values
of D obtained as ratios between the ultimate dis- Figure 4. Load-displacement diagrams, with yield limit
placements of Table 4 and those quoted in Table 6. punts.

919
Table 6. Values of yielding loads, displacements and the
ductility parameter D.
Load Displacement
Sample N mm D
Type A (A02) 14,700 1.880 5.09
Type B (B02) 21,303 4.156 2.56
Type C (C02) 26,641 2.935 4.88
The more or less dissipative behaviour of a joint is
generally given relating to its slenderness, defined as
the ratio between the minimum thickness of connected
elements and the diameter of the pin.
The Type B sample has the minimum thickness of
the jointed boards given by t1 which is half of the
corresponding minimum thickness of Type C sample;
consequently, as the pegs have a fixed diameter, the
two samples show a ratio of 1/2 in their slenderness.
Tests results for these samples confirm a behaviour
which is quite expected in case of steel pins: a double
value of slenderness gives an almost double value of
ductility.
The Type A sample doesn’t follows this scheme
in fact even if it has the less slenderness, an half
of that of type B joint, it shows the greatest ductil-
ity, which can be ascribed to the plastic deformation
of the timber peg, scarcely opposed by the reduced
thickness of jointed boards. This behaviour obviously
occurs because the jointed elements and the connect-
ing device are made of materials with stiffness of the
same magnitude.
4 CONCLUSIONS
This research allows some interesting observations
about the applicability of Johansen’s theory and the
derived design formulas in the case of connections
made using timber pegs.
920
First of all, obtained results show how the differ-
ent nature of the material used for the connecting
device needs a more specific definition of the “yield-
ing moment” of the pins, which must be inserted in
the formulas, if those ones have to be employed as a
design mean confidently.
Moreover, it’s been noticed that failure loads given
by Johansen’s formulas for the three tested samples of
double shear plane joints don’t diverge significantly
from the experimental ones, when the collapse mode
involves bending of the pin. Instead the theoretical and
experimental data show absence of matching in those
joints characterized by a large ratio between the diam-
eter of the timber pin and the least thickness of the
connected timber boards. So the behaviour of joints
characterized by low slenderness needs more analysis
not only from an experimental point of view, even if in
the field of restoration low ratios between the jointed
board thickness and the peg diameter difficultly occur.
Another appreciable characteristic of connections
with timber pegs, experimentally observed and quite
important in seismic zones, is the good ductility of
these joints.
REFERENCES
Ceccotti, A., Bonamini, G., Ruffini, M. & Uzielli, L. 1998.
Restauro conservativo di capriate lignee. La Pieve di San
Marino. Torino: C.L.U.T.
Giordano, G. 2005. Tecnica delle costruzioni in legno. V Ed.
Milano: HOEPLI.
Johansen, K.W. 1949. Theory of timber connections. Interna-
tional association for bridge and structural engineering
9: 249–262.
Menegotto, A. 2001. Al disopra di ogni aspettativa. Prime
indicazioni sulla possibilità di impiego di cavicchi.
Adrastea 18: 28–35.
Mormone,V. & Russo Ermolli E. 1999. Palazzo reale di
Napoli. Interventi di recupero delle coperture. Recupero
e conservazione 27: 82–87.
UNI ENV 1995-1-1 Eurocodice 5 Progettazione delle strut-
ture di legno – Regole generali e regole per gli edifici.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Research on composite strengthening of historical housebuilding:

Retrofitting intervention for masonry arches and vaults

A. Borri & G. Castori

Department of Civil and Environmental Engineering, University of Perugia, Perugia, Italy

ABSTRACT: A Seismic retrofitting of structures belonging to the architectural heritage requires meeting of
constraints which are related to preservation of artistic features. Such a philosophy was applied to the design
process of the retrofitting intervention of a masonry arch, belonging to a 17th century portico built inside a
Roman amphitheatre in the city of Spoleto (Italy), and of two masonry vaults of an ancient building in the city of
Foligno (Italy). This led to the opportunity of investigating the efficiency of an innovative composite material,
based on fine steel cords embedded in a cementitious matrix (Steel Reinforced Grout). To assure an adequate
strength against earthquake, a couple of prestressed SRG strips, in the first case, and a reticular system, made of
transversal and longitudinal prestressed SRG laminates, in the second case, have been applied on the extrados
of the arch and of the vaults, respectively.

1 INTRODUCTION improvement for historically – artistically relevant

Among the structural components in monumental
buildings, masonry arches and vaults deserve partic-
ular attention. They are very widespread in Italian
historical centers, and their preservation as part of the
cultural heritage is a very topical subject.
Because of their ages or for accidental causes (such
as earthquakes), these structures can suffer several
types of damage, so the contribution of strengthening
materials and repair techniques may be required to re-
establish their performances and to prevent the brittle
collapse of the masonry in possible future hazardous
conditions. According to this, strengthening masonry
vaults poses serious concerns because the vast majority
is of considerable architectural and historical value.
Traditional reinforcement techniques may guar-
antee an adequate increment in strength, stiffness,
and ductility, but are often short-lived and labor-
intensive, and they usually violate aesthetic require-
ments or conservation or restoration needs. Recent
earthquakes have demonstrated how such interven-
tions, based on reinforced concrete and steel rebar,
appeared extremely harmful for structures belonging
to the architectural heritage. After recognizing dam-
ages of those “seismic upgrading” following the regu-
lations, members of National Committee for Cultural
Heritage Seismic Risk Prevention claimed the princi-
ple of seismic improvement by techniques respecting
the structural system and preserving their integrity.
This way of thinking have been clearly claimed by
10.29.1996 decree explaining the mandatory seismic
structures as interventions on the structural elements
of the building in order to increase the safety mar-
gin without changing the main features of global
behaviour. The current debate about restoration and
consolidation of historical constructions assumes, in
fact, that an historical building is the primary source
of knowledge, a significant testimony in its full com-
plexity. Thus, it is essential to deal with the individual
object as a unique, unrepeatable instance, assigning
equal value, dignity, importance, and right to protec-
tion to all the components of the building and all the
material evidence contained in it. Hence, a strengthen-
ing project has to be preceded by a scientific diagnostic
approach and has to minimise the impact of the inter-
vention, by choosing the most compatible solution
with respect to the building’s current state, with the
aim of preserving it as better is possible. Therefore,
the actual approach to restoration leads to the require-
ment of new reinforcement technologies, able to work
in parallel and in cooperation with the existing struc-
tures, and moreover characterized by the fact to be
light, durable and possibly removable.
2 STRENGTHENING OF MASONRY ARCHES
AND VAULTS
Thanks to their adaptability to the changes of the geo-
metric configuration, masonry arches or vaults are able
to distribute the strain along the mortar joints, avoid-
ing the formation of significant cracks. In this way the

921
collapse mechanism does not depends by the mate-
rials’ limit strength, but it is due to the incapability
of the structure to fit the horizontal and vertical dis-
placements of the abutments. As a consequence it is
clear that similar displacements should be considered
when strengthening masonry arches introducing only
systems, which are able to realize an effective “rein-
forcement” without changing the constructive features
of the structure.
Conversely, existing methods of repair often have
been based on the idea of working over to make the
structure resistant to the seismic actions, changing, in
this way, the main features of global behavior.
According to this, it has to avoid, for instance, that
methods (i.e., concrete slab), which, in order to limit
the effects produced by the thrusts, try to eliminate
them, changing, as a result, the static behavior of the
structure, that is so reduced to a simple “ceilings”. For
the same reasons, the use of removing the spandrel
fill should be assess with care, taking into account the
geometric configuration and the cracking pattern of
the vault.
Such problems have led researchers to suggest
strengthening masonry vaults with fiber – reinforced
polymer (FRP) composites in the form of bonded
surface reinforcements. There are several advantages
related to this strengthening technique: very low
weight, corrosion immunity, high tensile strength
and low thermal expansion coefficient. Moreover the
somewhat easiness of execution of the intervention,
even in difficult operative conditions, allows a wide
range of possible applications in several situations
of damage, without considering that the possibility
of binding or wrapping structural elements made of
brittle materials (like masonry) allows, in most cases,
to avoid the collapse of the structure and so assure
the pursued safety conditions. Nevertheless, their lack
of fire resistance and their relatively high cost may
represent an obstacle for a widespread use.
According to these difficulties, beside the “tradi-
tional” FRP, the use of a new family of composite
materials, based on high strength twisted steel wires
embedded within a cementitious grout (Steel Rein-
forced Grout), is more and more considered. The core
of the project is based on the idea to combine, along
with the traditional advantages of the composite mate-
rials, the performances of a material that, using twisted
steel wires, allows the same applications of FRP mate-
rials with lesser costs. Without considering that the
use of SRG, because of the presence of a cementi-
tious grout, which replaced the polymeric resin of FRP
materials, allows to increase the fire resistance as well.
The flexural limited resistance of the masonry
vaults or arches can be overcome by introducing
“passive” reinforcing steel strip (Borri et al. 2007a),
making it able to sustain substantial bending moment
in addition to axial loads, but a better application of
922
Figure 1. Prestressing device (courtesy Eng. Giannantoni).
the method is the use of pre-stressed strips. Loading
the vault in radial direction, the SRG reinforcement
increase its compression and improve its resistance to
pressure-flexure induced by incidental loads.
The consolidation effect is realised by simply plac-
ing one or more strips alongside the extrados surface of
the vault. The strips are fixed to the masonry of the sup-
porting walls and then pre-tensioned (Figure 1). This
fact implies the transmission of radial self-equilibrated
forces between the curved strips and the arch.
As the compression resistance of the masonry is
usually high, it is possible, and not risky, to strongly
rise the axial internal load in the masonry, avoiding
the formation of the four-hinges collapse mechanism.
All the structural section of the masonry will be more
compressed as in the original state, thus postponing
the formation of the cracks. Using this technique, the
reinforcement does not interfere with the in situ mate-
rial and respect the structural behaviour of the existing
building.
The method permits a recognisable sign of con-
temporary interventions. The reinforcement so applied
works as an “active system”, which allows calibrating
the actions as it needs and, if loss of pre-stress takes
places, allows re-tensioning.
Even in case of variation of curvature along the arch
ring, the reinforcement strips act in a beneficial way,
as the mutual forces applied by the cords perpendic-
ularly to the surface are maximal exactly in the zones
where the radius of curvature is minimal. It has to
be noticed that the proposed consolidation technique
works well only if the piers are able to sustain the lat-
eral thrust induced by the arch (or vault). If they were
too weak, the structure would break in a section some-
where between the springing and the keystone. This
means that, in this case, the reinforcement have to be
prolonged till the base of the piers. They have to help
balancing load across the spans and thus they have to
be placed up to the spans. In some case shear failure
may take place in the structure and a sliding failure
mechanism can occur. The reinforcement, although
less efficient than in the flexural induced collapse case,
still bring to an increase of ultimate load.

Figure 2. Cloister Portico: plan view.
Table 1. Geometric properties of ARCH 04.
ARCH 04
Span (mm) 3310
Rise (mm) 1660
Section (mm × mm) 200 × 480
Height of the left pier (mm) 1800
Height of the right pier (mm) 1780
3 CASE OF STUDY: CLOISTER PORTICO
3.1 Investigation and field survey
The focused structure is a cloister portico, dated from
the late 18th century, affected by the Umbria-Marche
earthquake (1997). Such portico is constituted by
seven masonry arches and is built on the elliptical plan
of an old Roman amphitheatre (II B.C.) in the city of
Spoleto (Italy).
All arches have a semi-circular shape and are built
with natural stone bricks arranged in two layers bonded
together by mortar joints only. The bottom layer,
155 mm thick, is built with 70 × 155 × 310 mm bricks
laid on edge, whereas the upper layer, 35 mm thick,
consists of 35 × 155 × 310 mm bricks laid flat.
All arches have exactly the same span length
(3000 mm) and depth (480 mm), with the exception
of the fourth arch that has a span of 3310 mm.
The arch piers are also made of natural stone bricks
(70 × 155 × 310 mm) and are 480 × 480 mm square.
Each pier includes a Doric capital, 270 mm high,
whereas only the first three columns have a base
(210 mm high).
At the current state of the design process, the
retrofitting intervention involves only one of the seven
arches of the cloister portico (label 04 of Figure 2).
3.2 Retrofitting design
To increase the structural seismic strength and, con-
sequently, the safety of the structure, SRG strips,
923
Figure 3. Layout of the intervention.
Table 2. Properties of the laminate.
3SX-12
Tensile load (N/mm) 635
Elastic modulus (N/mm2 ) 210000
Ultimate strain (%) 1.2
pretensioned through an innovative prestressing
device, has been designed. Such strips will be reused
in the final intervention, when a new floor will be real-
ized at the top of the abovementioned cloister portico.
The consolidation effect is realized by simply placing
two SRG sheets alongside the extrados surface of the
arch. The sheets are fixed to the masonry of the sup-
porting piers and then tensioned. This fact implies the
transmission of radial self-equilibrated forces between
the curved sheets and the arch. The masonry of the
arch will be consequently compressed and the distinct
blocks will be helped to better support flexion, espe-
cially originated by asymmetrical conditions. Actually,
the designed prestressing stress is quite low (5% of
SRG strip ultimate strength) because the main goal
is to obtain a reinforcement able to work as an “active
system”, so that composites start working even for low
intensity seismic activity.
The prestressing device, used in such applications,
consists of three regions, comprising of two anchoring
and one loading regions, respectively (Figure 3). The
anchorage regions consists of: removable steel plates,
to anchor the SRG sheets, and a fixed steel plate,
anchored to the arch abutment, to hold the removable
steel plates and bonded sheets through steel bolts and
nuts.The loading regions consist of: two inequal angles
and a winding axis, to prestress the SRG sheets man-
ually with a dynamometric wrench, and two fixed and
as many removable plates to bond sheets through steel
bolts and nuts.
The strips used are 100 mm in width and 0.89 mm
in thickness. Manufacturer’s mechanical properties
(Hardwire 2002) of the strengthening material are
reported in Table 2.
A deflectometer has been used to register deflec-
tions. Thus it has been possible to monitor the arch
crown deflection during and after the intervention.
Figure 4. Deflectometer used to register arch crown
deflection.

3.3 Installation
In order to ensure the safety of the structure, the arch
rehabilitation started with the installation of scaffold-
ing on the intrados both of the above mentioned arch
and of the two adjacent arches. This action is neces-
sary to supply, after the removal of the filling material,
the horizontal thrust of such arches and to bear ten- Figure 5. Anchorage region.
sile stresses generated by the prestressing of the SRG
sheets.
After surface cleaning by sanding and water based
solvents and then levelling the surface of the outer
arch area, bedding bands were created using suitable
cementitious grout.
The first step in the assemblage of the prestressing
device consist of fastening the anchoring steel plates to
the arch abutment through the use of two  16 anchor-
ing rods, inserted vertically, long enough (1000 mm) to
reach the height of the pier. Each anchoring plate con-
sist of three parts: a central plate (440 × 300 × 6 mm)
with two round holes for the anchoring rods and two
lateral plates (440 × 150 × 6 mm) joined to the central
plate by a butt hinge. Anchorage of the SRG sheets is
created by bonding the ends of the sheets on the afore-
said lateral plates with a polymeric resin and by fas-
tening two removable steel plates (440 × 150 × 6 mm) Figure 6. Bonding SRG sheets to the anchoring steel plate
of the loading region.
through the use of steel bolts. The removable plates are
fixed to the lateral plates by the tightening of the nuts
in the anchorage region. High pressure must be applied Because of the dimensions of such element, which
to the SRG sheets through steel plates and ten bolts at should not permit prestressing of the two sheets simul-
each end to prevent slipping of the SRG sheets, which taneously, it becomes necessary to use an intermediate
would result in a loss of prestressing force. Test results steel plate (440 × 150 × 6 mm), placed at a position
(Borri et al. 2007b) show that the friction resulting 250 mm distant from the loading device. Using such
from the pressure was sufficient to anchor the sheets a steel plate (Figure 7), it is possible to fasten the two
during prestressing. SRG sheets, fixed to the arch abutment, to the SRG
The next step consist of creating the anchorage of sheet, fixed to the loading device. As in the previous
the SRG sheets in the loading region. cases, the anchorage is created by fixing a removable
At one end, anchorage of the SRG sheets is cre- steel plate (440 × 150 × 6 mm) to the intermediate
ated by using the same anchoring steel plates as the steel plate through the use of steel bolts.
anchorage region (Figure 6). After the anchorage was created, the desired pre-
Conversely, at the other end, the reinforcement is stressing stress can be smoothly achieved in the SRG
fixed directly to the winding axis of the loading device. sheets by tightening the winding axis manually with

924
Figure 7. Fixing SRG sheets to the loading device.
Figure 8. Arch after intervention.
a dynamometric wrench. As said, since the main goal
is to obtain a reinforcement able to work as an “active
system” and not to increase arch compression, the
sheet was prestressed only up to 5% of its ultimate
strength (3000 N). After the reinforcement was pre-
stressed and fixed to the anchoring plate through the
use of a removable plate, the sheet was cut and the
loading device was removed.
Test results showed that the mechanical device
proved to be practical and safe for prestressing SRG
sheets. In particular the deflectometer used during the
intervention has not register deflection, confirming
that the prestress load has been very low.
Figure 8 shows the details of the anchoring and load-
ing regions after the assemblage is fully completed.
925
Figure 9. First floor: longitudinal section and plan view
(courtesy Eng. Menestò).
4 CASE OF STUDY: JACOBILLI BUILDING
4.1 Investigation and field survey
Jacobilli building is a clustered complex that takes
almost half of a single-standing block in the historical
centre of Foligno (Italy). The building, that includes
various different structural nuclei affected by changes
and modifications during centuries, gained almost sta-
ble configuration around the XVIII century as a noble
house.
The building was seriously damaged by the Umbria-
Marche earthquake (1997) and may of its structures
were repaired including some composite strengthen-
ings. The present paper deals, in particular, with the
design process of the retrofitting intervention of two
masonry vaults of the building. Both vaults are located
in the building’s first level and, more in detail, in the
Music and Sacrifice room.
Both vaults are cloister vaults built with solid clay
bricks arranged in a single layer. Based on the survey,
it was determined that the length of the vault in the
Music room is 7.90 m, width is 6.75 m, while its aver-
age thickness is 120 mm. Conversely, as for the vault
in the Sacrifice room, the length is 7.14 m, width is
4.74 m and its average thickness is 120 mm.
In both cases, the analysis of the cracking pat-
tern reveals an asymmetric distribution of the cracks
(Figure 10). The main cracks are in fact concentrated
along the façade wall (west side of the Sacrifice room
and east side of the Music room). Also, both vaults
are affected by large cracks distributed along the ribs.
According to this, it would seem reasonable to assume

Figure 10. Cracking pattern: a) Sacrifice room; b) Music
room.
that the cause of such a collapse mechanism is due
to the incapability of the façade wall to supply the
horizontal thrust of the vault.
Also, at the extrados of the vault in the Music room
a significant loss of curvature near the façade wall has
been found during filling material removal. It is very
likely that such permanent deformation was caused
by the changes and modifications that affected the
building during centuries. In particular, it can be noted
that from 1712 to 1724 Giuseppe Jacobilli enlarged
the mansion reaching wath is today Via Antonietti,
incorporating other buildings in the process. As a con-
sequence, the peripheral masonry facing of the Music
room and a portion of the vault were demolished and
rebuilt (Figure 11). It seems reasonable to assume that
such an intervention modified the static behaviour of
the vault, reducing the curvature and therefore the
bearing capacity.
4.2 Retrofitting design
Both vaults were reinforced with SRG tapes at the
extrados intended to lock-out some of the most prob-
able failure mechanisms. As above mentioned, the
collapse mechanism does not depends by the mate-
rials’ limit strength, but it is due to the incapability of
926
Figure 11. Enlargement of the Music room (1712–1713).
Figure 12. Collapse mechanism: a) extrados reinforcement;
b) extrados reinforcement + anchoring.
the façade wall to supply the horizontal thrust of the
vault. As a consequence, it is clear that such a thrust
should be considered introducing only systems, which
are able to realize an effective “reinforcement” without
changing the constructive features of the vault.
A traditional solution could be the substitution of
filling material with hollow brick walls, which has
positive effects thanks to the dead load decrease.
Conversely, as the reinforcement can bear the
stresses occurring at the tensed edges, the application
of composite laminates, as externally bonded strength-
ening materials, can modify the failure mode of the
masonry vault and significantly increase the load –
carrying capacity (Figure 12a). Therefore, the brit-
tle failure of such structures, typically caused by the
formation of four (or three) hinges, can be avoided.
Depending on the position of the laminate, in fact,
the formation of the forth hinge can be prevented
(Foraboschi 2004). When the connection between
vault abutment and reinforcement is effective, their
use also prevents the formation of cylindrical hinges
in the piers and it causes high increases of inducing
mechanism activation loads (Figure 12b).
The consolidation effect is realized by simply plac-
ing a reticular system, made of transversal and longi-
tudinal prestressed SRG strips on the extrados of the
vaults. Where the transversal strips are used to resist
the horizontal thrust acting on the façade wall, whereas
the longitudinal strips are used as a connection element
between the transversal strips.

Figure 13. Layout of the interventions.
As in the previous case of study, by using the same
prestressing device, the sheets are fixed to the masonry
of the supporting piers and then tensioned.
The strips used are 100 mm in width and 0.89 mm
in thickness. Manufacturer’s mechanical properties of
the strengthening material are reported in Table 2.
Even in this case, a series of deflectometers have been
used to register deflections. Thus it has been possible
to monitor the vaults deflection during and after the
intervention.
4.3 Installation
The rehabilitation of the two vaults started with
removal of the filling material up to the haunches,
where the solid clay bricks of the structure are inserted
into the outer wall. In both cases, at the vault extrados a
horizontal wooden ring have been found during filling
material removal. This ring served to provide structural
support for the vault thrust.
After surface cleaning by sanding and water based
solvents and then levelling the surface of the outer
vault area, bedding bands were created using suitable
cementitious grout. It should be noted that, despite
927
Figure 14. Anchoring of the SRG sheets.
Figure 15. Loading region.
careful preparation, areas with abrupt variations in
curvature may occur. In these cases experimental
tests showed high degree of weakness of “traditional”
FRP sheets. Conversely, because of their higher shear
strength, the use of steel fibers may overcome such
shortcomings.
Even in this cases, the first step in the assemblage of
the prestressing device consist of fastening the anchor-
ing steel plates (220 × 150 × 6) to the vault abutment
through the use of two  16 anchoring rods inserted
diagonally, long enough to reach the height of the
springer. High pressure has been applied to the SRG
sheets through steel plates and three bolts at each end
to prevent slipping of the sheets.
After the SRG sheets have been fixed, the anchorage
of the sheets in the loading region was created. More in
detail, at one end the sheet was anchored through the
use of the same steel plates as the anchorage region;
whereas at the other end, the sheet was fixed directly
to the winding axis of the loading device.
Finally, the desired prestressing stress has been
achieved in the SRG sheets by tightening the wind-
ing axis manually with a dynamometric wrench. Even
in this cases the prestressing load was quite low. The
sheets were prestressed only up to 6% of their ulti-
mate strength (4000 N). After the reinforcement was
prestressed and fixed to the anchoring plate through
the use of a removable plate, the sheet was cut and the
loading device was removed.
Figure 16. Prestressing of the SRG sheets.
Figure 17. Stress state generated by the prestressing SRG
strips.
Figure 18. Application of steel flat plates.
Also, it can be noticed that prestressing SRG strips
permits the reinforcement to follow a regular curve
parallel to the original ideal surface of the undeformed
vault. According to this, in the Music room, connected
928
Figure 19. Vaults after intervention: Sacrifice and Music
room.
with the loss of curvature (near the façade wall) in
addiction to the reinforcement, steel flat plates, bolted
to the bricks, were used to secure the ply to the vault
extrados.
This fact implies the transmission of radial self-
equilibrated forces between the strips and the vault,
allowing to reduce deformation and to re-stabilish,
therefore, an adequate curvature.
Figure 19 shows the details of the extrados of the
vaults after the assemblage is fully completed.
Also, as it regards the Music room, because of the
low value of the prestress load, the deflectometers used
during the intervention have been register a reduction
of deformation of “only” 20% (11 mm) of the origi-
nal value. It would seem reasonable to assume that an
increase of the prestressing load should re-stabilish the
original curvature.
5 CONCLUSIONS
The operations carried out, firstly to save and then to
consolidate and restore the masonry arch of the cloister
portico built inside a Roman amphitheatre in the city
of Spoleto (Italy), and of two masonry vaults of the
Jacobilli building in the city of Foligno (Italy) have
all followed the same philosophy. To place the most
up-to-date techniques and technologies at the service
of culture, in order to respect the historic value of
the ancient buildings and to obtain adequate safety
levels, whilst changing as little as possible the orig-
inal structural conception. These technologies, never
applied before in the field of restoration, have been
studied specifically for this occasion, offering new and
interesting possibilities for the safeguard of the World
architectural heritage.
REFERENCES
Borri, A., Casadei, P., Castori, G. & Ebaugh, S. 2007a.
Experimental analysis of masonry arches strengthened
by innovative composite laminates. Proceeding of the
10th North American Masonry Conference (10NAMC),
St. Louis, Missouri, USA, June 3–6, 2007.
Borri, A., Castori, G., Giannantoni, A. & Grazini, A. 2007b.
Performance of reinforced masonry bond beams. Pro-
ceeding of the 10th North American Masonry Conference
(10NAMC), St. Louis, Missouri, USA, June 3–6, 2007.
929
Foraboschi, P. 2004. Strengthening of masonry arches with
fiber-reinforced polymer strips. Journal of Composites for
Constructions, ASCE 8(3), 2004, pp 7–16.
Hardwire llc., 2002. What is Hardwire. Product Guide
Specification. Web site: http://www.hardwirellc.com.
Huang, X., Birman, V., Nanni, A. & Tunis, G. 2005. Properties
and potential for application of steel reinforced polymer
and steel reinforced grout composites. Composites, Part
B, Vol. 36, 2005, pp 73–82.
Jurina, L. 1997. The reinforced arch: a new technique for
strengthening masonry arches and vaults using metal tie
bars. Proceedings of 16th National Congress of C.T.A.,
Ancona, Italy ,1997.
Matana, M., Galecki, G., Maerz, N. & Nanni, A. 2005. Con-
crete substrate preparation and characterization prior to
adhesion of externally bonded reinforcement. Proceed-
ings of International Symposium on Bond Behaviour of
FRP in Structures (BBFS 2005), Hong Kong, China, 2005.
Triantafillou, T.C. 1998. Strengthening of masonry structures
using epoxy-bonded FRP laminates. Journal of Compos-
ites for Constructions, ASCE 2(2), 1998, pp 96–104.
Valluzzi, M.R., Valdemarca, M. & Modena, C. 2001.
Behaviour of brick masonry vaults strengthened by FRP
laminates. Journal of Composites for Constructions,
ASCE 5(3), 2001, pp 163–169.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Static and dynamic properties of a flexible joint working in

cracked historical masonries

˛
A. Kwiecień & B. Zajac
Cracow University of Technology, Cracow, Poland

R. Jankowski
Gdansk University of Technology, Gdansk, Poland

ABSTRACT: In the paper, the static and dynamic properties of proposed flexible polymer joints are discussed.
The discussion is based on investigations on special polymer, describing its properties obtained during laboratory
and field tests. The research has allowed us to point potential advantages of the polymer application in cracked
historical masonries. In addition, the rules of the polymer application and the possible conservative mask method
of the constructed joint are presented.

1 STRESS CONCENTRATION AROUND 1.2 Influence of micro-cracks

STRUCTURAL FAULTS
1.1 Adaptation in masonry
Masonry can be defined as an association of two differ-
ent materials: brick and mortar, with complementary
mechanical properties. Bricks are elastic but brittle
elements while mortars are plastic and ductile contin-
uous binding materials. The elastic modulus of brick
is always higher than that of mortar. In the case of
masonry, there are geometric imperfections and struc-
tural faults like the vertical joints between bricks,
where stress concentrations occur around the corners
of bricks under the action of compressive forces. For
long lasting actions like the gravitational ones the duc-
tility of lime mortars protects the brick against failure
by the phenomenon known as adaptation. In the case of
short time actions of certain level the induced energy
is stored by elastic brick as potential energy. Then, by
a slow relaxation it is gradually dissipated by the plas-
tic deformation of mortar. When forces of short time
are huge, like those developed during an earthquake,
the material strength is exceeded and a fault or a crack
appears. This behavior is due to lack of time for plastic
deformations to redistribute the concentrated stress.
Under the peak of stress concentration the strains
are easily reach their limit of elasticity at the end of
crack. Further, the material in structural member is
either in the state of cracking or plastically deforming.
(Sofronie 2004, Sofronie et al. 2005).
Damage occurrence develops mechanism of energy
absorption that extensively reduces the forces acting
upon the structure. It is connected with material defects
localized in area where the principal tensile stress
(stress concentration) exceeds the tensile strength of
masonry material. Micro-cracking in masonry takes
place at low load levels and progresses following non
homogenous path which combines the two mecha-
nisms with growth and linking between micro-cracks.
It causes the reduction of the effective resisting area,
leads to an increase of damage until rupture and can
be described by the damage variable d. Micro-cracking
causes the decrease of elastic modulus and the increase
of stress in the masonry material (“effective stress”),
having to resist the external loading (Oñate et al.
1996) – Figure 1. When the damage variable d reaches
the critical value the micro-cracks start linking and the
rupture comes into being.
When the crack divides a member into two parts,
they work separately in a new stress equilibrium
and static balance being a consequence of stress
redistribution.
Such cracked masonry is characterized by a
decreased resistance to the action of additional loads,
to which the building was resistant prior to damage.
Separated parts of the cracked structure can bring into
contact at small areas in some places of the crack.
In these points new stress concentrations are formed
increasing damage. This process is especially intensive

931

Figure 1. Evolution of uniaxial tension stress-strain curve
(after Oñate et al. 1996).
under movement caused by settlements, temperature
changes and structural vibrations.
It is important that after rupture the surrounding of
the crack remains a weak zone with the micro-cracks
occurring. In this zone the strength of the masonry
material is lower than of the original one what influ-
ences the repair strength the in case of the crack
bonding.
2 CONCEPTION OF STRUCTURAL
INTERVENTION IN CRACKED MASONRY
2.1 Discussion on applying of only original
materials
The requirements of safety and use are almost per-
manently in conflict with the respect of the iconic,
historical and material integrity of the monuments and
treatment where only using of traditional methodolo-
gies is acceptable. In most countries the conservation
is controlled by officials having an influence on any
action to be undertaken imposing constraints and lim-
itations that sometimes appear unreasonable to the
engineer. On the contrary, the engineer tends to achieve
safety by means of solutions or procedures which
appear unacceptable to the officials in charge of con-
servation. It is evident that some consensus has to be
found. On one side, conservation requires the safe-
guard of the formal, material and historical integrity of
the monument, but also its survival and safe exploita-
tion (Viggiani 2006). It is important to answer the
question: do we want to have the heritage object with
only original intervention materials with the threat of
being destroyed by an earthquake or do we want to have
the safe working historical structure for the next gen-
erations, protected with small amount of non-original
intervention materials that are in compatibility with a
monument and invisible for visitors. It is believed that
the second option is suitable.
As an example of non traditional methodologies is
the steel structure inside the spires of Burgos Cathedral
932
Figure 2. The new crack at the vault surface passing around
the mortar injection, too stiff in comparison to masonry
material.
(Ortega et al. 1996), that represents very massive
intervention technique and the installation of shape
memory alloy devices in the Basilica of San Francisco
in Assisi (Castellano & Infanti 2004), representing the
group of innovative techniques.
2.2 Rules of intervention in heritage structures
Nowadays, the use of innovative techniques or par-
ticular solutions, which efficiency must however be
demonstrated, should not be in any case forbidden. On
the contrary, the possibilities and limitations of each
innovative technique should be briefly explained and
also estimated and justified by calculating their effect
in terms of variations in the global behavior of the
structure. In the first place it is necessary to allow the
structure to manifest a satisfactory global behavior,
by improving the connections between the masonry
walls and between the walls and the floors. Interven-
tions aimed at increasing the masonry strength may
be used to re-establish the original mechanical prop-
erties lost because of material decay or, alternatively,
to upgrade the masonry performance. Techniques used
must employ materials with mechanical and chemical-
physical properties similar to the original materials
(Modena et al. 2006) and neutral from the interaction
point of view.
It is highly preferred to select properly innova-
tive materials to make them work together with the
existing materials with better “compatibility” from the
mechanical point of view. This is to reduce as much as
possible high stress concentrations, which occur using
high stiffness modern bonding materials like high
strength resin or high strength mortars. It should be
noted that applying too strong strengthening material
in cracked structures (in comparison to the masonry
properties) could cause additional damages. As an
example, the new crack passing around the mortar
injection, too stiff in comparison to masonry material,
is presented in Figure 2. The crack is localized at the
repaired cloister vault of the Archeological Museum in
Cracow (Ciesielski et al. 2004). The cracks appeared
after the mistaken intervention in foundation of the
structure. The criterion in selection should concern

Figure 3. The flexible joint made of polymer (by injection),
bonding disrupted structural elements.
much more deformability than strength, especially in
seismic areas (Modena 2004).
3 THE FLEXIBLE JOINT METHOD (FJM)
3.1 Description of the method
The Flexible Joint Method (FJM) is the method of
repair using deformable elastic-plastic polymers and
reinforced polymers for bonding of damaged struc-
tures. It is dedicated to cracked masonries (of poor
quality too), especially to historical structures where
minimum intervention is required. The cracks are fill-
ing in with the special binary mixed polyurethane
mass injection, forming the flexible joints bonding the
disrupted structure elements (Fig. 3). This method is
registered in the Polish Patent Department with No.
P-368173 and was described in papers (Kwiecień et al.
2006a, b). The FJM permits safe work of the retrofitted
structure in the new stabilized state of balance. This is
the method particularly conductive to objects, in which
a redistribution of stress occurred in consequence of
damage (cracks). It permits further safe exploitation of
the object under additional static and dynamic loads.
Cracks’ filling in the previously damaged bearing
structure with specially designed polymer strengthens
the disrupted structural elements. It assures that the
damaged building regains tensile, compression and
shear resistance in place where bearing capacity was
lost. Especially, the tensile resistance, deformability
and ductility are of big importance. The deformation
of the flexible joint under load assures uniform distri-
bution of stress along the lap joint over the total contact
surface, equalises deformation and damps vibrations.
The polymer mass bonds structural faults in places
of cracks and limits development of the new stress
concentrators.
3.2 Properties of the polymers
Grouting constitutes one of the most common tech-
niques applied to historic masonries. Although grout-
ing is a non-reversible technique, it is well accepted
even for monuments of high historical and architec-
tural value, since the materials added to masonry may
933
be of the same nature as the in situ ones and because
of the positive effect of grouting on the mechanical
properties of masonry (Vintzileou 2006).
The polymers used in the FJM are flexible two-
component grouts based on polyurethane resin, hand
and machine applicable. They introduce also the pos-
itive effect on the mechanical properties of masonry,
thus should be accepted as innovative inject technique
for application in historical structures. They fulfill
requirements of the proper applicability into cracks
and the adequate rheological, physical, chemical and
mechanical properties.
Rheologically, they are injectable, have sufficient
fluidity and stability. Physically, they do not pro-
duce high temperatures and shrinkage and also have
adequate hardening time and hygroscopic properties.
They are resistant to temperature from −40◦ C to
+80◦ C (temporary up to +150◦ C) and UV radi-
ation. The polymers are insensitive to moisture and
also have long life expectance. Chemically, the poly-
mers are neutral to the masonry materials (no chemical
reactions between them, related to both durability and
mechanical properties). Mechanically, the polymers
are reducing vibration shear-resistant and permanently
elastic adhesives and also have adequate strength
and deformability characteristics. Giving an example,
for the soft flexible polymer (of the Young modulus
E = 4 MPa) described by (Kwiecień et al. 2006a, b) the
tear strength is 1.7 MPa and adhesion to the concrete
surface is 1.2 MPa with elongation of 60%.
The flexible polyurethane mass shows no signifi-
cant changes of mechanical parameters after 3 million
cycles in the fatigue test (producer data) with elonga-
tion of 10%, frequency 5 Hz and maximum stress of
1.0 MPa. Cyclic tension-compression tests under har-
monic excitation were also conducted (by the authors
of the present paper) for the soft flexible polymer,
with frequencies of 0.1, 0.5, 1, 2, and 5 Hz. The set
deformations of specimens (of 28 mm diameter and
28 mm height) were of +/− 10, 20, 30 and 40 %
with the corresponding stresses of +/− 0.16, 0.35,
0.62 and 0.97 MPa. The example results for 2 Hz fre-
quency are presented in Figure 4. The average damping
ratio (Jankowski 2003) has been calculated as equal to
D = 0.06 for all obtained data.
In addition, the field test (producer data) confirmed
that a direct fixation of rail base-plates using flexible
polyurethane mass presented only marginal changes in
flexibility (6%) after exposure to weathering and fre-
quent dynamic loading over 28 years. The differences
obtained during long-term dynamic behaviour in situ
were so minimal that the flexible joint can be expected
to last for many more years.
The polymers reach the full strength in 24 hours
after application. Depending on the actual state of
masonry the proper polymer with required properties
can be chosen. The application rules of the polymers

Figure 4. The cyclic tension-compression test for 2 Hz.
in masonry are the same as for grouts, described in
details by (Vintzileou 2006).
3.3 Adhesive aspect of the flexible joint
Adhesiveness of polymers to masonry materials is very
important matter. It is necessary to investigate mate-
rial properties of damaged masonry by the use of
diagnostic methods, e.g. micro-drilling (Skłodowski
2006), before preliminary selection of possible struc-
tural intervention methods. In the case of the use of
the FJM, the description of these properties is needed
for proper selection of polymer.
The adhesiveness between original structure materi-
als and polymer is especially important. Investigations
made by the authors, using the pull-off test (Bonaldo
et al. 2005), showed that the polymers have good
adhesion to masonry materials. The ultimate separate
strength was equal 0.7–2.5 MPa, measured in labo-
ratory condition on brick specimens with clear and
undusted surfaces (over the stress of 2.5 MPa the dam-
age in brick was observed). In cracked masonries such
934
Figure 5. Peaks of the stress concentration visible in the
photo-elastic research on a concrete specimen.
conditions are reachable only with the use of special
primers. It is believed, that kind of the primer proper
for the polymer should be matched according to the
masonry material properties and its condition. This
task is not fully recognized jet and has to be investi-
gated in co-operation with chemical engineers having
experience with materials of different kinds of heritage
masonries.
4 BEHAVIOR OF THE POLYMER JOINT
WORKING IN CRACKED MASONRIES
4.1 Reduction of the stress concentration
The stress distribution observed in micro-scale is not
regular. The stress concentrations occur at the grain
surfaces of brittle materials what was observed dur-
ing the photo-elastic research on a concrete specimen
(Dantu 1957). The peak concentrations are responsi-
ble for the micro-cracks appearance when they exceed
strength of a masonry material even when the average
stress is low (Fig. 5).
In the case of the use of flexible polymers as the
adhesive layer the peaks of stress concentration are
reduced and the uniform distribution of stress assures
the increase of material strength. It is caused by lower
amount of micro-cracks appearing in brittle material
and thus higher the elastic modulus and the material
strength (comp. Fig. 1). This phenomenon was con-
firmed by a pull-off test made on the same piece of
brick with two kinds of adhesives. In the case 2A, the
aluminum disk was bonded to the brick surface using
the hard flexible polymer (E = 600 MPa) and in the
case 2B, the stiff epoxy resin (E = 3300 MPa) was used
as adhesive. Damages were observed in brick (Fig. 6),
because of higher strength of adhesives than of the
brick.
When the aluminum disk was bonded to the brick
surface using flexible adhesive the ultimate force and
the potential energy were about two times higher than
with the stiff bond (Fig. 7). The use of the soft flexible
polymer (with E < 10 MPa) will assure much uni-
form distribution of stress and reduction of peak stress
concentrations.
Figure 6. The pull-off device and the specimens after
damage.
Figure 7. Results of the pull-off test for the flexible (2A)
and the stiff (2B) adhesive bond.
Figure 8. The investigated masonry building before damage
and just after deformation.
4.2 Damage and repair processes of the tested
masonry building
Properties of polymer flexible joints were examined
on the real masonry building presented in Figure 8.
The original structure was damaged by a caterpillar
exciting dynamic forces in the corner of the building
at the roof level.
Properties of the damaged structure were obtained
indirectly. Modes of damage showed that cracks were
caused by maximum principal stress and were passed
diagonally through bricks and mortars. The pull-off
test made on pieces of bricks, extracted from the inner
935
Figure 9. Damages of the side wall before and after rectifi-
cation of the tested masonry.
Figure 10. Cracks of the wall filled in by soft flexible
polymer.
Figure 11. A new crack appeared in the repaired masonry,
caused by the dynamic force excited at the roof level.
part of the damaged wall (Fig. 6), showed that the
tension strength of the bricks is relatively high. The
ultimate stress obtained on four specimens was 0.5–
2.3 MPa in case of the stiff adhesive and 1.1–3.3 MPa
for the flexible adhesive (compare with section 4.1).
The structure was about to collapse after the action
and had to be rectified (Fig. 9) what caused appear-
ance of new cracks. The cracked building was repaired
using the Flexible Joint Method. Cracks were cleaned
from dust and protected with special primer and then
filled in by the soft flexible polymer using injection
technique (Fig. 10). The structure was statically and
dynamically tested. At the end the destructive test was
performed on the masonry in order to examine the
ultimate work of the polymer flexible joint (Fig. 11).
Figure 12. Diagnostic dynamic exciters: the Vibrosejs of
20 T mass and the modal hammer of 5 kg mass.

4.3 Work of the flexible joint under thermal

influences (season changes) Figure 13. Transmittance of the accelerations between the
plate of Vibrosejs and the top corner of the masonry (impact).
The cracks width (Fig. 10) and the temperature in the
opened air have been measured. Measurements were
carried on for the period of one year in the temper-
ature range from −5◦ C to +28◦ C. The polymer joint
elastically changed its width in the horizontal direction
(11/H) up to 0.75 mm without the permanent displace-
ment. It has been observed that the flexible joint allows
the cracked structure for limited elastic deformation in
places of new erected dilatations, protecting it against
permanent unfavorable movement. Additionally, the
polymer counteracts formation of stress concentra-
tions when cracks close under influence of temperature
changes.

4.4 Loading of the structure with static forces
The building was loaded statically with the horizon-
tal force imposed in place where structure was excited
dynamically by caterpillar. The maximum generated
force was of 32 kN and caused permanent opening (0.4
mm) of the non bonded small crack (9/H), after over-
coming friction forces acting in the non bonded crack.
Under the same load the bonded cracks showed no
reaction (10/H and 11/H) what gives evidence of the
polymer joint strength, keeping the structure elements
together and protecting against destruction.
4.5 Diagnostic dynamic excitation of
the structure
The tested masonry structure was dynamically diag-
nosed before damage, just after damage and after
repair using the polymer flexible joints. As dynamic
exciters the Vibrosejs of 20 T mass and the modal
hammer of 5 kg mass were used (Fig. 12).
The 16-chanells system collected data from the
accelerometer system installed on the structure. There
was no test with the Vibrosejs on the damaged struc-
ture because of the real possibility of collapse. The
Vibrosejs excited vertical vibrations at the soil surface
at the distance of 15 m in front of the masonry.
The impact of the Vibrosejs plate at the soil sur-
face and the harmonic excitations with frequency
Figure 14. Maximum horizontal accelerations measured
on the top corner of the masonry (stationary harmonic
excitation).
range of 6–30 Hz were conducted. The analysis of the
building response (presented for horizontal vibrations
measured on the sensor localized at the top corner –
Fig. 12) showed dynamic characteristics of the struc-
ture. Analysis of the impulse excitation pointed that the
flexible bonded masonry damps better vibrations than
the undamaged stiff structure (Fig. 13). The response
of the building repaired using the FJM is almost 40%
lower in resonance than for the undamaged building
and favourable frequency shift is observed. It should
be noticed, that such damaged building, bonded using
the flexible polymer joint, survived the horizontal har-
monic resonance vibration measured at the top of
the building, of the 60 second duration and of the
30 cm/s2 acceleration amplitude (Fig. 14). There were
no additional damages on the repaired masonry after
the vibration test what confirm efficiency of the repair
using the FJM.
The excitation using the modal hammer allowed us
to determine the inertancy given by formula (1)

936

Figure 15. The inertancy obtained for different states of the
masonry during the dynamic test using the modal hammer.
where |Ẍ (ω)| = acceleration frequency characteristic;
|F̈(ω)| = force frequency characteristic.
The diagram presented in Figure 15 shows that after
damage the frequency shift of resonances to lower fre-
quencies is observed (stiffness degradation) and after
repair using the polymer joints a significant back shift
is visible (stiffness increase). The calculation of the
inertial forces (assuming the same level of excitation
acceleration) shows that the flexibility increase of the
masonry after damage decreases the values of inertial
forces over 3 times but they still can be too high for
the weakened structure. After the application of the
soft flexible polymer in cracks the inertial forces grow
up 1.5 times in comparison to the forces acting on
the damaged structure but the strength of the repaired
masonry is significantly upraised. More details have
been presented by Kwiecień & Zajac ˛ (2008).
4.6 Destructive test on the masonry building
The destruction made on the tested building showed
that the strength of the polymer joint is higher than
the original masonry. The destructive process went
through 3 steps. One crack, partially bonded with poly-
mer (11/H) and partially not strengthened (9/H), was
considered for observations. In the first step (after the
huge hit), a new crack appeared in the middle of the
side wall (Fig. 11) between polymer bonded cracks
and was similar to that which came into being after the
damage (Fig. 9). The new crack was localized at a cer-
tain distance from the bonded one, surrounded by the
zone where peak stress concentrations were reduced.
The observed crack opened in the non-bonded part
(9/H) of 1.54 mm and the bonded one (11/H) opened
only 0.25 mm. In the second step, (after the next huge
hit), a new crack opened more. The additional increase
of the observed crack width was of 2.80 mm in the non
bonded part (9/H) and was of 1.20 mm in the bonded
part (11/H). In the third step, the masonry structure
collapsed and the observed crack in the non-bonded
part (9/H) was totally opened, when in the bonded part
937
Figure 16. The fragment of polymer bonded wall during
destruction (with sensors 9/H and 11/H), left after collapse.
Figure 17. The cracked masonry building of poor quality
repaired using polymer joints.
(11/H) the polymer kept fast together pieces of wall
(Fig. 16). The test showed that the polymer reduces
stress concentrations and protects cracks again damage
propagation.
4.7 Repair of the old family building
The polymer flexible joint was tested on a masonry
family building localized in the small village in Poland.
The building was constructed from spider-web rub-
ble walls of poor quality in 1930. The walls consisted
from sand-stones joined with weak lime mortar. Inves-
tigated mortar crumbled away under a fingernail.
Damages in the building appeared in form of cracked
wall and were caused by settlement after a flood
(Fig. 17).
The building was repaired using the FJM. The
cracks were filled in with the soft flexible polymer
after cleaning and priming of crack spaces.
Effectiveness of the polymer joint was examined
during the window exchange. The triangle wall frag-
ment, visible in Figure 17, was hanging only on the
new constructed flexible joint for the period of sev-
eral hours. There were no notices of any fissures on
a wall plaster and on gypsum markers placed on the
joint. No cracks or fissures on structure for two years
of the exploitation have appeared. This case confirms
that a flexible joint equalizes stresses in brittle materi-
als and protects cracked masonries of the weakened
structure against appearing of new damages during
micro-movements of the separated building parts.
5 THE USE OF A POLYMER JOINT FROM THE
CONSERVATION POINT OF VIEW
5.1 Advantages of polymer joints opposite to
traditional methodologies
It is obvious that the possibilities and limitations of
each innovative technique should be briefly explained.
The Flexible Joint Method is the new approach in
retrofitting of masonries and it is necessary to make a
lot of experiments in laboratory and in natural scale to
ensure that the new method could safety work in vari-
ous cases of monuments repair. It needs proper fitting
of polymer properties and various kinds of primers
for cooperation with structural materials. The usage
of polymer joints in historical masonries have to be
weigh out individually and also estimated and justi-
fied by calculating their effect in terms of variations
in the global behavior of the structure.
The application of the innovative polymer material
can assure survival and safe exploitation of monu-
ments. Good dynamic properties of proposed polymer
and ability to dissipation of deformation energy make
this material useful to use in damaged masonries in
seismic areas. It can assure also the safeguard of the
formal, material and historical integrity. It is possi-
ble, because bonding of cracks with polymer allows
us to avoid unnecessary intervention and to ensure
acceptable safety condition. The polymer can be sim-
ply applied in cracks of width 3 ÷ 50 mm and then
covered with a material acceptable from “conserva-
tive” point of view, limiting in this way variations of
external appearance of historical structures.
The proposed method is relatively cheap and is
not time consuming. Low values of stiffness mod-
ules assure safe co-operation of the polymer with weak
materials and better “compatibility” from mechanical
point of view comparing to stiff epoxy resin or brittle
cement mortars, generating high stress concentration.
5.2 Critical discussion of using the polyurethane
mass in repair of historic buildings
The polyurethane mass proposed for application in
historical buildings as repair (bonding) mass should
be taken individually into consideration in each case.
Presented polymer can be subjected in certain range
to requirements proposed for repair mortars by Van
Balen et al. (2005).
It is obvious that polyurethane as the relatively
new bonding material does not respect the traditional
practice of the joining of historic materials and there-
fore the criticism could be provide in regard to its
use. It is also known that polyurethanes have her-
metic properties and are a barrier for vapor transport,
thus detrimental effect on the modification of hygric
properties of masonry can take place, if it is using
unconsciously. It is especially important in case of
938
hermetic vertical membranes stopping the horizontal
evaporation of the moisture from inside of a structure
and causing the negative side-effect in the long-therm
period. In such situation the drainage of the barrier has
to be done. On the other hand, the polyurethane repair
joint prevent water penetration through a cracked wall.
These advantageous properties of polyurethane mass
were exploited in renovation of joints in the stone
Monument in Latvia (Sidraba 2002). Additionally,
polyurethane durability and long-term experience was
evaluated in the climatic conditions of the Scandi-
navia for 15 years, performing good properties in
comparison to traditional materials (Sidraba 2002).
6 CONCLUSIONS
The Flexible Joint Method based on polymer flexible
joints is proposed as the new approach in retrofitting of
cracked historical masonries. The presented properties
of the flexible polymer joints, examined in laboratories
and on real cracked structures, showed their effective-
ness and advantages in the use of it as the repair method
of damaged masonries. The innovative polymer mate-
rial, bonding cracked historical structures, can assure
survival and safe exploitation of monuments and also
the safeguard of the formal, material and historical
integrity. Authors hope that the use of flexible poly-
mer joints can be accepted by conservation authorities
and can be widely used as the repair methodologies in
historical constructions.
REFERENCES
Bonaldo, E. et al. 2005. Bond characterization between con-
crete substrate and repairing SFRC using pull-off testing.
Int. Journal of Adhesion & Adhesives 25 (2005). Elsevier.
Castellano, M.G. & Infanti, S. 2004. Seismic protection of
monuments by shape memory alloy devices and shock
transmitters. 4th International Seminar on Structural
Analysis of Historical Constructions. Padova.
Ciesielski, R. et al. 2004. Application of CFRP laminates as
strengthening if cracked brick arches. 4th International
Seminar on Structural Analysis of Historical Construc-
tions. Padova.
Dantu, P. 1957 I. – Étude des contraintes dans les milieux
hétérogènes application au béton. Laboratoire Central des
Ponts et Chaussées. Publication No. 57–6. Paris.
Jankowski, R. 2003. Nonlinear rate dependent model of
high damping rubber bearing. Bulletin of Earthquake
Engineering 1: 397–403. Kluwer Academic Publishers.
Kwiecień, A. et al. 2006a. Flexible Joint Method (FJM) –
a new approach to protection and repair of cracked
masonry. 1st First European Conference on Earthquake
Engineering and Seismology (1st ECEES). Geneva.
Kwiecień, A. et al. 2006b. Repair of cracked historical
masonry structures by use of the Flexible Joint Method
(FJM) – laboratory tests. Structural Analysis of Historical
Constructions V. New Delhi.
Kwiecień, A. & Zajac ˛ B. 2008. Dynamic response of
the cracked masonry building repaired with the flex-
ible joint method – an innovative earthquake protec-
tion. 7th European Conference on Structural Dynamics.
Southampton.
Modena, C. 2004. Design approaches of investigations for the
safety and conservation of historic buildings. 4th Inter-
national Seminar on Structural Analysis of Historical
Constructions. Padova.
Modena, C. et al. 2006. Codes of practice for architectural
heritage in seismic zones. StructuralAnalysis of Historical
Constructions V. New Delhi.
Oñate, E. et al. 1996. Structural analysis and durability
assessment of historical constructions using a finite ele-
ment damage model. Structural Analysis of Historical
Constructions. CIMNE. Barcelona.
Ortega, L.M. et al. 1996. The spires of Burgos Cathedral.
Structural Analysis of Historical Constructions. CIMNE.
Barcelona.
Sidraba I. 2002. New materials for conservation of stone mon-
uments in Latvia. New materials for safeguarding cultural
heritage, ARIADNE 10 Workshop (www.itam.cas.cz/∼
arcchip/ariadne_10.shtml).
939
Skłodowski, M. Compact diagnostic test: Outline of histori-
cal monuments testing procedure. IFRT Reports 3/2006.
Warsaw.
Sofronie, R.A. 2004. Two techniques for repair and strength-
ening masonry constructions. 4th International Semi-
nar on Structural Analysis of Historical Constructions.
Padova.
Sofronie, R.A. et al. 2005. Application of reinforcing tech-
niques with polymer grids for masonry buildings, CAS-
CADE Report No. 5.
Van Balen K. et al. 2005. Introduction to requirements for
and functions and properties of repair mortars. Materials
and Structures 38 (October 2005) 781–785.
Viggiani, C. 2006. Safety and use vs. integrity of historical
constructions: conflict or synergy? Structural Analysis of
Historical Constructions V. New Delhi.
Vintzileou, E. 2006. Grouting of three-leaf stone masonry:
types of grouts, mechanical properties of masonry before
and after grouting. Structural Analysis of Historical Con-
structions V. New Delhi.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Nanotechnologies applied to the restoration and maintenance of

wooden built heritage

C. Bertolini Cestari & T. Marzi

Dipartimento di Progettazione Architettonica e Disegno Industriale, Politecnico di Torino, Torino, Italy

S. Invernizzi
Dipartimento di Ingegneria Strutturale e Geotecnica, Politecnico di Torino, Torino, Italy

J.M. Tulliani
Dipartimento di Scienza dei Materiali ed Ingegneria Chimica, Politecnico di Torino, Torino, Italy

ABSTRACT: The paper presents a preliminary experimental campaign, which aims at investigating the possible
application of carbon nanotubes to the mechanical improvement of timber structures. Different wood species are
considered, as well as different solvents or resins as a dispersing media for carbon nanotubes. The experimentation
is carried out on purpose exploiting exclusively the capillary properties of wood, in view of a possible on site
application of the technique. Although still at a preliminary stage, these first results appear to be promising, and
the research on the procedure is still in progress.

1 INTRODUCTION The development of polymers reinforced with

In contrast with other fields of civil engineering and
building recovery technology, the application of com-
posite nanomaterials employed to reinforce wooden
structures is a little known technique which requires,
albeit partially, to be fully tested before being applied
on a large scale. This holds true mainly for existing
wood components (elements); with regards to new
wooden elements, there is a much greater wealth of
experience.
The use of nanomaterials and composite materi-
als to strengthen wood was suggested seven years
ago by several American researchers who studied
the effects of carbon nanotubes-based composites on
the mechanical characteristics of reinforced wooden
elements [1].
In Italy, given the greater scope of recovery and
conservation issues, research has focussed on method-
ologies for consolidating existing ancient structures.
Experiments carried out on antique timber and existing
structures are very complicated since the mechani-
cal characteristics of the timber vary on account of
the inevitable presence of membrane defects (knots,
oblique wood fibre, lesions, etc.). However, experi-
mental results have shown a clear trend in terms of
efficiency concerning said recovery techniques.
nanoparticles is one of the most promising approaches
in the field of future engineering applications. The
unique properties of some nanoparticles (carbon black
and carbon nanotubes) and the possibility of com-
bining them with traditional reinforceing elements
(fibreglass, carbon fibre or Kevlar) have generated
an intense research program in the nanocomposites
sector [1–2].
Carbon black is made up of particles with a diam-
eter of 30 nanometres (nm) and is commonly used
as a charge to make polymers conductive in order
to avoid the accumulation of electric charges. Car-
bon nanotubes have a diameter of several nanome-
tres and their length measures several micron; they
have very good potential for improving the electri-
cal and mechanical properties of polymers, even with
0.1% weight content compared to epoxy resin [1–2].
Indeed, the resistance to traction of single wall car-
bon nanotubes or SWCNT can reach up to 600 GPa
and the elastic modulus range between 1 and 5 TPa
[1]. The difficulties lie in transferring these remark-
able mechanical, thermal and electrical properties of
nanotubes to the polymer matrix. Consequently the
correct dispersion of the nanotubes in the polymer
and timber interface, between the reinforcement and
the polymer matrix, is crucial [1–4].The dispersion

941
of carbon nanotubes in the matrix is complex given
the widespread specific surface of the nanoparticles
(1000 m2 /g or more) which tend to favour the forma-
tion of agglomerates. Different techniques for dispers-
ing said materials in solvents (acetone, ethanol. . .) in
addition to ultrasounds, mechanical shaking or a com-
bination of the two techniques, have already been tried
out [1–7].
The interface adhesion can be improved by chem-
ically functionalising the nanotubes surfaces; this
generates strong covalent-type of bonding [2, 3, 6,
7]. The bonding between nanotubes and polymers
allow the strain to be transferred from one stage to
another.
Amino groups have been used for this purpose and
double wall carbon nanotubes or DWCNT with or
without a functional group surface can be purchased
on the market (Nanocyl, Namur, Belgium) [8].
A picture of said experimental difficulties is pro-
vided by the results obtained with epoxy resins where
by adding SWCNT in the best cases the mechani-
cal resistance to bending stress is slightly increased,
although in most cases, resistance is often reduced
[5]. Recently, fabrics containing up to 39% in SWCNT
weight impregnated with epoxy resin have been pro-
duced; once again the results obtained fell below
expectations [9].
On the other hand, significant progress in mechan-
ical resistance values has been made with poly-
methyl methacrylate (PMMA), polyvynilic alcohols
and with polystyrene-based composites [5]. Promis-
ing results have also been achieved with polyurethane
resin composites with up to 10% in weight of
nanotubes [10].
Summarizing, carbon fibre nanotubes provide a
number of advantages:
– they are morphologically and chemically compati-
ble both with polymer resin, used as bonding mate-
rial, and with wood, they are anatomically similar to
strong piping bonded with a thermoplastic matrix
and equipped with strong dissipative capacity with
regards to fracturing energy.
– nanotubes allow the polymer bonding matrix to
improve it own inbuilt deformation capacities
considerably thanks to the transformation of a
homogeneous bulk with vitreous behaviour into
a micro-reticule with high level of porosity and
deformability;
– the mechanical characteristics of the resin-fibre
compound are considerable on account of the high
specific resistance of the fibres which ensure great
cohesive strength combined with high ductility.
The combination of the two produces significant
creep resistance without the composite, loosing any
deformation capacity;
– the tubular structure of nanofibres has great per-
meability to vapour potential: this is an important
942
characteristic when dealing with large surfaces
treated with glue, and especially in the case of wood
since any accumulations of humidity must be easy
to disperse in order to avoid biotic degradation.
2 MATERIALS AND METHODS
The first step of the experimentation was dedicated to
assess the efficiency of different impregnation tech-
niques.The carbon nanotubes were dispersed by means
of an ultrasonic probe for different time in ethanol and
acetone. Such solvents are supposed to act as a trans-
port medium to bring the carbon nanotubes directly
into the channels of the wood microstructure.
Specimens with section equal to 2.5 cm × 2.5 cm
were sunk in the suspensions for different times, as
shown in Figure 1. We decided to exploit only the cap-
illarity natural phenomenon instead of vacuum or other
pressure assisted impregnation techniques.This choice
was adopted in view of a future onsite application in
existing timber structures.
The following wooden species have been selected
for the analysis: fir, Douglas pine, oak and larch. Some
samples were obtained from new timber, other from
eighteenth century structures.
Nanocyl multiwall carbon nanotubes were used for
the process. For economical reasons, the series 7000,
not functionalized (Table 1), was used to set up the pro-
cess, while series 3101 functionalized with carboxyl
groups were used for the mechanical tests (Nanocyl®-
3101 series are purified to greater than 95% carbon and
then functionalized with COOH groups).
In a second step, an epoxy resin (MAPEI epojet,
[11]) was used for the dispersion in order to get a prod-
uct for mechanical improvement that could be applied
on the wood surface by painting, or act as a reinforced
glue to connect different timber part.
Epojet is a two component solvent-free epoxy adhe-
sive. The pre-measured portions (Part A = resin and
Part B = hardener) must be mixed together before
being used. Once mixed, Epojet becomes a liquid
with low viscosity very suitable for injection. The
mix ratio between part A and B is 4 to 1. This
resin has been selected because of its low Brook-
field viscosity (respectively 500 and 320 mPa.s for
part A and B).
The efficiency of the impregnation procedure has
been assessed with observation at the SEM micro-
scope. The overall mechanical improvement of the
timber specimen is going to be evaluated from compar-
ison of mechanical tests performed on un-reinforced,
impregnated and resin painted samples. The bending
strength has been evaluated by the three-points bend-
ing test carried on with displacement control. The span
of the samples was 370 mm and the load was applied
with a velocity of 12 N/s.
 Figure 2. TEM image of Nanocyl carbon nanotubes.

Table 1. Carbon nanotubes characteristics.

Method of
Property Unit Value measurement

Average diameter nanometer 9.5 TEM

Average length microns 1.5 TEM
Carbon purity % 90 TGA
Metal oxide % 10 TGA
Surface area m2 /g 250 BET

Figure 3. Detail of the three-point bending test.

3 RESULTS AND DISCUSSIONS

3.1 Mechanical characterization of un-reinforced

Figure 1. Impregnation test at different times: progression
of the solution absorption.
wood
The un-reinforced behavior of four different timber
species have been characterized, from a mechanical
point of view, with standardized three-point bend-
ing tests. A picture of the test procedure is shown in
Figure 3.
The oak old samples displayed the most fragile
behavior, while Douglas pine appear to be sensitive

943
Figure 4. Load displacement diagrams for four different
unreinforced timber species.
Figure 5. SEM image of the larch sample as such.
to local buckling phenomena. The peak load ranged
between 1340 N and 2770 N, provided that the cross
section of the samples was 25 mm times 25 mm. The
load displacement diagrams of each test is shown in
Figure 4.
A direct comparison with reinforced samples will
be provided in the future.
3.2 Carbon nanotubes dispersion in solvents
Four different suspensions have been prepared with
different nanotubes (series 7000) concentrations. Fur-
ther, each suspension has been used for impregnation
as shown in Figure 1. In order to assess the efficiency
of the impregnation the higher part of the specimen
was cut and observed with the SEM. As a reference,
also slices of not impregnated wood were observed
(Figure 5).
The first suspension was prepared by dispersing
about 0.1% carbon nanotube in ethanol, after 20 min-
utes of ultrasonication with an ultrasonic probe. A
droplet of this suspension was sampled and set upon
the SEM sample holder for observation. As illustrated
944
Figure 6. Clustering of carbon nanotubes in the first
dispersion.
Figure 7. Better dispersion of nanotubes in the second
dispersion.
in Figure 4 the dispersion was unsuccessful, since the
nanotubes remained agglomerated.
It is worth noting that the morphology of the cluster
in Figure 6 is also due to the poor purity of the adopted
carbon nanotubes as reported in Table 1. Due to this
reasons, the impregnation was not proceeded.
In the second dispersion we decided to increase
the sonication time up to two hours, and the car-
bon nanotube content (1%). The resulting suspension
(Figure 7) was more satisfying, but we could not pro-
ceed with the impregnation due to high viscosity of
the suspension (which was close to that of a paste).
The third ethanol suspension was obtained with
0.5% carbon nanotube content and two hours son-
ication. This proportion allows a good dispersion
preserving a moderate viscosity, and was chosen for
further impregnation. Unfortunately, although the cap-
illary process is active, there was no effective transport
Figure 8. SEM image of the top larch sample section: no
nanotubes appear to be penetrated the wood microstructure.
Figure 9. SEM image of the top larch sample section
impregnated with acetone solution: almost no nanotubes
appear to be penetrated the wood microstructure.
of nanotubes inside the sample, as observed with the
SEM (Figure 8).
Because of this unsuccessful trials, we decided to
change the solvent and adopt acetone in place of
ethanol.
The acetone solution of 0.5% carbon nanotube
weight content was sonicated for two hours and
the resulting dispersion was good preserving low
viscosity.
The wood sample was sunk into this suspension for
twenty-four hours. Also in this case, almost no car-
bon nanotubes are observable at the top section of the
sample from the SEM image (Figure 9).
In each case, the SEM analysis has been pushed up
to the maximum allowable resolution (10000×) but
we were not able to observe not even CNT clusters.
945
Figure 10. SEM image of CNT agglomeration in the epoxy
resin (the resin was fractured to emphasize the defects
presence).
3.3 Carbon nanotubes dispersion in epoxy resin
The main difference between epoxy resins and the sol-
vents described above pertains the higher viscosity.
Therefore, first of all it has been necessary to find the
optimal carbon nanotube content keeping constant the
two hours sonication duration.
Proved that although the dispersion in the solvent
was efficient, the overall impregnation technique was
not working satisfactorily, we moved to investigate the
dispersion using resin as a dispersing medium.
After some trials, we adopted a carbon nanotubes
weight content equal to 0.3% with respect to the resin
(A + B fraction).
We decided to disperse first the nanotubes in the
resin component B, which has the lowest viscosity.
Then the two components were mixed together during
thirty minutes mechanical stirring at room tempera-
ture. The mix was poured into a mould and let curing
for six hours at sixty degrees.
Unfortunately, this precaution was not sufficient to
avoid clustering of carbon nanotubes, as evidenced by
Figure 10, and in more detail Figure 11.
3.4 Mechanical behavior of reinforced wood
A preliminary assessment of the efficiency of the rein-
forcing procedure has been carried out in the following
way. First some of the timber samples were coated with
resin only, then other were coated with the CNT rein-
forced resin obtained after four hours sonication. The
CNT content was increased up to 0.5% with respect to
the resin.
All the coated samples were cured at room tem-
perature for seven days, according to the manufac-
turer recommendations, and to simulate an in-situ
intervention.

Figure 11. SEM detail image of CNT agglomeration in the
epoxy resin.
Figure 12. Load displacement curves for the eighteenth cen-
tury oak samples. Comparison between un-reinforced, resin
coated, and CNT reinforced resin coated.
Figure 13. Load displacement curves for the Hemlok fir
samples. Comparison between un-reinforced, resin coated,
and CNT reinforced resin coated.
The preliminary results obtained from the few sam-
ples tested till now are in some case very promising,
in other case not so unambiguous.
In the case of the eighteenth century oak samples,
we observe an increase of about 19% in the peak load
946
when only resin is used for the coating. If CNT rein-
forced resin is used for the coating, the gain in the peak
load raises up to 42%.
On the other hand, as far as the Hemlok fir samples
is concerned, an increase of about 45% is obtained in
case of resin coating, regardless the presence of a CNT
reinforcing in the resin. The influence of the CNT rein-
forcing appears even to be slightly disadvantageous,
but this could be also due to the fact that we still have
too few results to compare.
In general, the comparison must be continued
considering more samples and more timber species.
4 CONCLUSIONS
In the paper a preliminary experimental campaign is
described, which aims at investigating the possible
application of carbon nanotubes to the mechanical
improvement of timber structures. Different wood
species were considered, as well as different solvents or
resins as a dispersing media for carbon nanotubes. The
experimentation was on purpose carried out exploiting
exclusively the capillary properties of wood, in view
of a possible on site application of the technique.
As far as the present results concern, the solvent dis-
persions appears, if properly designed, effective from
the point of view of getting an optimal dispersion.
None of them, unfortunately, appear to be efficient
from the point of view of impregnation exploiting the
solely capillary phenomenon.
On the other hand, the epoxy resin dispersions are
still problematic since the aspect of carbon nanotubes
clustering has not been solved yet.
The first results concerning the assessment of the
timber retrofitting with CNT resin coating are some-
what positive, since they provide an increase in the
flexural strength which is greater or at least equal to
the one obtained with the resin alone.
These first results appear to be challenging, and the
research on the procedure is still in progress [12].
ACKNOWLEDGEMENTS
The research was carried out also with the economi-
cal support of the Italian Ministry of University and
Research, through a PRIN project co-ordinated at
national level by C. Bertolini Cestari.
REFERENCES
[1] O. Breuer, U. Sundararaj, Big return from small fibers:
A review of polymer/carbon nanotube composites,
Polymer Composites, 25 (6) (2004) 630–645
[2] F.H. Gojny, M.H.G. Wichmann, U. Köpke, B. Fiedler,
K. Schulte, Carbon nanotube-reinforced epoxy-
composites: enhanced stiffness and fracture toughness
 at low nanotube content, Composites Science and
Technology, 64 (2004) 2363–2371
[3] K.T. Lau, D. Lui, Effectiveness of using carbon nan-
otubes as nano-reinforcements for advanced composite
structures, Letters to the Editor/Carbon, 40 (2002)
1597–1617
[4] H. Miyagawa, L.T. Drzal, Thermo-physical impact
properties of epoxy nanocomposites reinforced by
single-wall carbon nanotubes, Polymer, 45 (2004)
5163–5170
[5] K. Lau, M. Lu, C. Lam, H. Cheung, F. Sheng,
H. Li, Thermal and mechanical properties of single-
walled carbon nanotube bundle-reinforced epoxy com-
posites: The role of solvent for nanotube disper-
sion, Composites Science and technology, 65 (2005)
719–725
[6] H. Miyagawa, M.J. Rich, L.T. Drzal, Thermo-physical
properties of epoxy nanocomposites reinforced by
carbon nanotubes and vapour grown carbon fibers,
Thermochimica Acta, 442 (2006) 67–73
[7] H. Chen, O. Jacobs, W. Wu, G. Rüdiger, B. Schädel,
Effect of dispersion method on tribological properties
947
of carbon nanotube reinforced epoxy resin composites,
Polymer Testing, 26 (2007) 351–360
[8] http://www.nanocyl.com
[9] Z. Wang, Z. Liang, B. Wang, C. Zhang, L. Kramer,
Processing and property investigation of single-walled
carbon nanotube (SWNT) buckypaper/epoxy resin
matrix nanocomposites, Composites: Part A, 35 (2004)
1225–1232
[10] T.J. Webster, M.C. Waid, J.L. McKenzie, R.L. Price,
J.U. Ejiofor, Nano-biotechnology: carbon nanofibres
as improved neural and orthopaedic implants, Nan-
otechnology, 15 (2004) 48–54
[11] ttp://www.mapei.it/Referenze/Multimedia/367_Epojet
_GB.pdf
[12] C. Bertolini Cestari, R. Maspoli, J. M. Tulliani,
M. Di Sivo, R. Roccati, Use of nanotechnologies in
cultural heritage for the efficiency of maintenance
systems in wooden built heritage: innovative technolo-
gies for restoration, in Nanofun-Poly. 2nd International
Symposium on Nanostructured and Functional-Based
Material and Composites, Lyon, May 2006
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Cross laminated timber panels to strengthen wood floors

A. Gubana
Department of Civil Engineering, University of Udine, Italy

ABSTRACT: Cross laminated timber panels are generally used to build walls and floor slabs of wood
houses, but they have interesting mechanical characteristics to be used for strengthening ancient wood floors
in restoration interventions. In particular they can be connected to the existing wood beams to have a floor
diaphragm effect under seismic actions. In the present paper the first results of a series of tests to determine
the panel shear behavior are presented. The experimental data show high shear resistance and huge potential for
their use.

1 INTRODUCTION used as walls and slabs in the construction of timber

Restoration interventions often require structural
strengthening and, in ancient buildings, floors are often
the part of the structure where stiffness upgrading is
more frequently necessary. Moreover if seismic resis-
tance has to be assured, a floor diaphragm behavior
has to be achieved. The importance of an effec-
tive diaphragm action in the floors of a multi-storey
masonry building is well-known in earthquake engi-
neering. Thanks to the diaphragm action, in fact, the
floors can transfer the lateral actions due to wind and
earthquake to the lateral load resisting systems.
One of the technique extensively used in Italy
for seismic rehabilitation of historical buildings is
refurbishment of existing timber floors using a con-
crete topping with a steel mesh effectively connected
to the timber joists. This type of technique ensures
an effective three-dimensional behaviour of masonry
buildings and, therefore, markedly improves the lat-
eral load resistance (Piazza & Turrini 1983, Gelfi et al.
2002, Gubana 2003).
In recent years other possible solutions were inves-
tigated with the aim to develop less invasive and
more reversible techniques (Giuriani & Plizzari 2000,
Modena et al. 1998). The attention towards preserva-
tion of cultural heritage buildings has constantly grown
and new technologies in accordance with the principles
of restoration are often preferred.
At the same time, advanced plate-like glued tim-
ber composite elements, as cross laminated timber
(Xlam), were developed by glue-lam timber facto-
ries. They consist of a sequence of layers made of
boards, which are fitted together at their small faces.
Each layer is glued to the other. They are commonly
buildings.
In retrofitting and rehabilitation intervention, on
wood floors in ancient buildings, they can be used as a
wood topping, mechanically connected to the beams,
so to create a wood-to-wood composite section. The
panels are stiff enough in their plane to give the floor
a diaphragm effect.
Few theoretical researches and experimental tests
on their structural behaviour are available in litera-
ture, as they are a recent technology (Ceccotti et al.
2006, Moosbrugger et al. 2006, Guggenberger &
Moosbrugger 2006).
In the present paper the first results of experimental
tests on shear behaviour of cross laminated panels are
presented.
2 EXPERIMENTAL TESTS
2.1 Cross laminated timber panels used in
the tests.
The panels, used in the experimental tests, have four
layers of boards: the external ones are 40 mm high,
while the two inner ones are 20 mm high, as to reach
120 mm section height. The external layers have the
boards in the direction of the maximum length of the
panel, while the two layers inside have the boards
in the perpendicular direction. The boards of the
double layer inside are arranged so that each inter-
space between two boards is covered by a board of
the other layer. The boards are 200 mm large. This
solution reduces parasite tension in the longitudi-
nal layers and has some advantages with respect to

949
Figure 1. View of the panel and principal production measures.

Figure 2. Cross and longitudinal sections of the panel.

the air impermeability. The requested length of the Table 1. Mechanical wood characteristics.
panel is reached by finger joints between the boards.
Maximum production length can grow up to 8000 mm MPa MPa
(Figs 1–2).
Compression ⊥ to the grain σc,90 = 3.23 Ec,90 = 227
Also the board height and the number of layers Tension ⊥ to the grain σt,90 = 1.49 Et,90 = 386
can vary, so that different cross section heights can Compression // to the grain σc,0 = 34.1 Ec,0 = 12148
be available.
The transversal dimension, for production technical
reasons, can reach about 1300 mm. The tests were conducted following the indications
The panels are made of spruce wood. Several tests of EN 408, EN1194 and EN 789 for compression and
were performed, at reference moisture content, to tension perpendicular to the grain and compression
check the principal mechanical characteristics. The parallel to the grain respectively.
timber is classified as GL24C, after EN 1194. Specific media data are reported in Table 1.

950
Figure 3. Overview of the shear-in plane test rig.

2.2 Description of the rig A particular device was designed to optimize the
test preparation and make easy the collocation of the
Under seismic actions the floor diaphragm is subjected
panels inside the rig.
to cyclic in plane shear tension. One of the principal
Due to the high forces necessary to reach the panel
aim of the rig design was the possibility of stressing
collapse, the length of the panels had to be reduced.
the panels with constant shear tension on the lateral
So the panels were sawn up to have a central square of
sections.
about 420–430 mm side.
This kind of rig was properly designed for shear tests
Inside of this panel portion, electronic acquisition
(Giuriani & Plizzari, 2000). It is made of an horizontal
measurement devices were placed along the diagonals
steel frame, with two lateral beams and two transversal
of an inner square of 250 mm side length, over the two
beams. A central beam is free to move longitudinally
faces of each panel (Figs 4–5).
in the centre of the frame. During each tests two panels
are stressed at the same time as described in Figure 3.
The panels lateral sections are constrained to the 2.3 Description of the tests
lateral and the central longitudinal beams. The jackets
The tests were performed using hydraulic jackets with
are placed at the end of the central movable beam,
force control.
where there is a proper contrast. The force due to
Ten couples of panels were stressed to collapse. The
pressure of the jacket against the contrast and the
first seven tests were monotonic, two tests were cyclic
transversal beam determines an axial tension action
and the last one was performed with reversed cycles
in the central beam and axial compression action in
of force. During all the tests only one of the two panels
the two lateral beams. Instability problems had to be
reached the collapse.
faced due to the high level of axial compression.
The moisture content was checked for every panel,
Two hydraulic jackets of 500 kN were used to
and the values were always around 12–13%.
perform the tests.
The transmission of the shear forces to the panels
is reached by a distributed series of dowel connectors.
3 EXPERIMENTAL DATA
This solution is similar to what can be actually built in
a building. The problem was particularly delicate, as it
3.1 Force to displacement diagrams
was necessary to avoid collapse at the interface panel-
to-rig, which was easy to happen because of high local The data direct acquired from the tests let to
pressure in timber near the dowels. draw force-to-displacement diagrams. Each test was

951
 Also in this case the collapse was less fragile with
respect to timber structure general answer.

3.2 Data elaboration

Figure 4. View of the panel and of the measurement devices
on the top side.
The acquired data needed to be elaborated to eval-
uate the τ (shear stress) – γ (shear deformation)
relationships and the modulus of shear elasticity G.
These values are necessary for numerical modeling
and design of structural interventions.
With reference to the sides of square represented in
Figure 8 it is possible to determine the average shear
deformation angle γ as a average of the deformations
α and β due to the shear stress state. For the calcula-
tion of angles α and β, deformations of the lengths of
the square sides are neglected (AB ≡ A B , AC ≡ A C ,
CD ≡ C D , BD ≡ B D ). So, from trigonometric rules
applied to AC’B triangle

it is possible to obtain:

Evaluating the undeformed diagonal length from the

sides:

Figure 5. Overview of the panel in the rig.

and taking into account that the elongations i reached
are relatively small with respect to the original length
identified with one letter, from A to L, and the two d, it is also possible to neglect the square of i ( /d
panels of each test were identified as left and right ≈ 4‰) and have:
panel with respect to the axis of the rig movable beam
(Fig. 3). For each panel four diagonals were under
control: two on the up side and two on the underside
after it is possible to determine the angle:
of panel, as indicated in Figure 4. As an example in
Figure 6 the diagrams related to the second test (Pan-
els B left and right) are reported. Is it possible to notice
that while the right panel reaches the collapse, the left
one still remains in the elastic range. There is also a In the same manner it is possible to find the value of
good correspondence between the data acquired on the β in triangle AD B
top of the panel (A1S, B1S on the left panel and A2S,
B2S on the right panel) and the data acquired at the
bottom side (A1I and B1I on the left and A2I and B2I
on the right panel). and then:
The collapsed panel showed an unexpected ductil-
ity, due to the fact that the rupture is generally not
simultaneous in all the shanks.
Always as an example, in Figure 7 the results of the
cyclic tests on Panels L are reported. The panels were Angles α e β are evaluated both on the top and on
subjected to 13 cycles : 3 from −50 to 50 kN force on the bottom face of the panel, so it is necessary to find
each of the two panels, 3 from 100 to −100 kN, 3 from the average value:
125 to −125 kN, 3 from 150 to −150 kN, and during
the 13th cycle, in the phase of inversion of force, the
collapse was finally reached at the value of −157.5 kN.

952
Figure 6. Force-to-displacement diagrams for Panels B under monotonic load.

Figure 7. Force-to-displacement diagrams for Panels L under cyclic load.

Figure 8. Deformation of the instrumented part of the panel.

Figure 9. Shear stress-to-shear deformation relationship.

In all the previous equations i is considered with its

absolute value, as its sign was already considered in
the formulation of the problem.
The shear stress was calculated as a constant average
stress over the lateral faces of the panel.
In Figure 9 the shear stress-to- shear deformation
diagram of panels B are reported.
The elasticity modulus G can be read on each
diagram as the ratio between shear stress and shear
deformation in the elastic range.
In Table 2 the collapse value of shear stress together
with the shear elasticity modulus are reported for the
ten couples of panels tested.
The average value of the shear resistance results
equal to 3.19 MPa and in any case always greater than
3.06 MPa.
The failure strength seems not to be affected by pre-
vious cycles of load. This can be explained as the panel
behavior is almost linear elastic up to the collapse.

953
Table 2. Maximum shear stress and elasticity modulus. The values obtained are more variable, but this is
to be expected, as the global deformation of the panel
Maximum shear Elasticity is influenced by the local deformation at the parallel
Panel stress MPa Modulus MPa shank interface, which is not glued. Also for this rea-
son the values obtained can be significantly less with
A 3.29 419
respect to gross wood section (the GL24C mean shear
B 3.22 331
C 3.14 327 modulus is 590 MPa after EN:1194:1999). In any case
D 3.29 346 they are much better with respect to a traditional wood
E 3.33 352 floor with no strengthening on the top.
F 3.15 327 In Figure 10 some views of the collapsed sections
G 3.06 332 are given, showing cracks inside of the shanks.
H 3.22 384
I 3.08 524
L 3.24 506
4 CONCLUSIONS

The first experimental results on cross laminated

timber panels show great potential for their use in
structural restoration of wood floors with the aim to
create wood diaphragm. In one direction the panels can
reach 8 m length, so they can cover most of the room
lengths in only one piece. The problem to connect one
side of the panel to the adjacent one can be faced with
mechanical joints.
The shear resistance is sufficiently high to resist
the shear stress expected in the floors due to seismic
actions.
The possibility offered to form the panel cross sec-
tions are wide, so it is possible to design and built
panels with a different number of layers or with shanks
of different height. A new series of tests will be per-
formed in the future on cross section less high than the
tested one, as in restoration intervention there are often
problems related to the conservation of the historical
levels of the building floors.

ACKNOWLEDGEMENTS

The author wish to thank Dr. Marco Pesamosca, who

discussed his Degree Dissertation about the “Struc-
tural Behavior of Cross Laminated Timber Panels” at
the University of Udine and collaborated successively
to the experimental tests, described in the present
paper.
Moreover the author wish to thank DiemmeLegno
s.n.c. of Pontebba (UD) – Italy, which granted the
present research work.

REFERENCES

Figure 10. View of the collapsed section of one panel (a)
and zoom view on the cracks (b).
The shear elasticity modulus were calculated taking
into account the data related to 10% and 40% of the
shear failure strength, as indicated in most of the codes
(i.e. EN:789:2005).
Ceccotti, A., Lauriola, M.P., Pinna, M. & Sandhaas,
C. 2006. SOFIE Project – Cyclic Tests on Cross –
Laminated Wooden Panels. WCTE 2006 – 9th Inter-
national Conference on Timber Engineering, Portland,
USA.
EN 408:2004. Timber structures. Structural timber and glued
laminated timber – Determination of some physical and
mechanical properties.

954
EN 789:2005. Timber structures. Test methods. Determina-
tion of mechanical properties of wood based panels.
EN 1194:1999. Timber structures. Glued laminated timber.
Strength classes and determination of some mechanical
and physical properties.
Gelfi, P., Giuriani, E. & Marini, A. 2002. Stud Shear
Connection Design for Composite Concrete Slab and
Wood Beams. Journal of Structural Engineering 128(12):
1544–1550.
Giuriani, E. & Plizzari G. 2000. Studio sperimentale sul
comportamento dei solai in legno rinforzati con lastre
in acciaio per resistere alle azioni sismiche. V workshop
Italiano sulle strutture composte, Palermo.
Gubana, A. 2003. An Analytical Approach to Design non
linear Behaviour Connections of Composite Sections.
International Journal for Restoration of Buildings and
Monuments 9(3).
955
Guggenberger, W. & Moosbrugger, T. 2006. Mechanics of
Cross Laminated Timber Plates under Uniaxial Bending.
WCTE 2006 – 9th International Conference on Timber
Engineering, Portland, USA.
Modena, C., Tempesta, P. & Tempesta F. 1998. Il recupero
a secco di impalcati in legno. Una tecnica per solai a
semplice e doppia orditura. L’edilizia 3/4.
Moosbrugger, T., Guggenberger, W. & Bogensperger, T.
2006. Cross Laminated Timber Wall Segments under
homogeneous Shear – with and without openings. WCTE
2006 – 9th International Conference on Timber Engineer-
ing, Portland, USA.
Piazza, M. & Turrini, G. 1983. Una tecnica di recupero dei
solai in legno. Recuperare 5,6,7.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Lime mortar with natural pozzolana: Historical issues and

mechanical behavior

E. Sala, I. Giustina & G.A. Plizzari

DICATA, University of Brescia, Brescia, Italy

ABSTRACT: This paper deals with the investigation of the mechanical behavior of lime mortar enriched with
natural pozzolana, widely used in historical masonry. The study aims to investigate the mechanical behavior of
natural pozzolana lime mortars, reproduced in laboratory by using different mix designs, as reported in historical
Italian architectural treatises, and natural pozzolana coming from Viterbo (Italy). To this purpose, the works of
the most important authors from the Ist century B.C. up to 1864 have been analyzed, in order to collect historical
mix designs of pozzolana lime mortars. Flexural and compression tests were performed following EN 1015-11
(2007). Mechanical tests were performed after 28 days and 60 days of curing. Penetration tests, not included in
European Standards but very suitable for practical applications, was also performed.

1 AIMS AND RESEARCH METHODOLOGY
Historical evidence on the use of mortars to meet
several needs in architecture has existed for millen-
nia. A systematic analysis of ancient mortars dating
some hundred or thousand years proved that they have
resisted decay by time and other deterioration factors.
Difference concerning their durability and the degree
of their degradation are attributed to many techno-
logical aspects that characterize each specific period,
such as the binding methods used, the type and propor-
tion of aggregates, application system as well as the
craftsmen skills.
Most of the mortars dated from the historic period
are based on the use of lime and many of them
are enriched with natural siliceous volcanic sand,
commonly called pozzolana. Referring to the Mediter-
ranean building culture and practice, it is known that,
since 1500 B.C., Greeks used a highly siliceous vol-
canic Santorini earth in wall coverings (Moropoulou,
2004). Furthermore, it is widely accepted that the
first extensive use of pozzolana was adopted by the
Romans who, since the II century B.C. (at least) used
a volcanic sand in the opus caementicium (i.e. a lime
mortar mixed with an aggregate of small stones or
rubble), whenever possible. This material, often cov-
ered by dressed stones or bricks, represented, from the
foundations to the vaulting, the very core of Roman
masonry (Lugli, 1957; Lancaster, 2005, also for fur-
ther references). Pozzolana remained at the base of
building practice during the modern ages up to the
industrialization, still inspiring, although new materi-
als and systems were developed, the basic concept of
contemporary concrete.
In the past, it was empirically clear that pozzolana
provides hydraulicity to lime mortar, enabling its set-
ting in wet conditions or underwater. Furthermore, it
was known that the addition of pozzolana was use-
ful to increase the durability of lime mortar, allowing
it to overcome the need of repeated maintenance and
accelerating the carbonation process, as it has been
also scientifically proved (Collepardi, 1993). Up to
now, however, such favorable properties of natural poz-
zolana have been subdued to relatively little scientific
investigation carried out by means of mechanical tests.
Due to the widespread use of pozzolana lime mortars in
historic masonry, wider scientific and technical issues
about mechanical properties of this kind of mortars
would be very useful in repairing and maintenance of
historic architectural heritage.
As the aim of this study was to analyze the mechan-
ical behavior of lime mortar enriched with natural
pozzolana, a new scientific investigation methodol-
ogy had to be adopted. It should be observed that it
is almost impossible to extract from historical build-
ings mortar samples without altering the mechanical
properties of the material; moreover, the required
dimensions of the samples would make this kind of
test too invasive. As a consequence, specific sam-
ples suitable for mechanical tests had to be prepared.
In this respect, the correct reproduction of the his-
toric mix designs, with their different compositions,

957
was a problem of critical importance. On this subject,
actually, in absence of convincing primary archival
sources, there is no general assessment and several
attempts on the characterization of historical mor-
tars were recently made without historical evidence.
It was therefore decided to prepare the different mix
designs following the indications traced in the archi-
tectural treaties, printed documentary sources that can
be considered as the official expression of the building
culture, generally stating what, in the past, the common
sense regarded to as “the rule of the art”. Although the
craftsmen skills and the circumstances often adapted
those guidelines to the case at hand, the mix designs
described in the treaties can be taken as a reasonably
reliable reference to reproduce historic mortars.
The works of the most important Italian authors
from Vitruvius (I cent. B.C.) to Giovanni Curioni
(1864) (and also some French authors whose treaties
gained great success in the Italian architectural culture)
were therefore analyzed in order to collect pozzolana
lime mortars historical mix designs. Following the
ratios suggested, samples were reproduced in labo-
ratory by using natural components only, and adding
Pozzolana originated from the Roman volcanic district
of the Monti Sabatini.
The different lime mortars were tested according
to European Standard EN 1015-11, after 28 days and
60 days of curing. A penetration test, not included
in European Standards but very suitable for practical
applications, was adopted as well.
2 HISTORIC PERSPECTIVE
Much of what is known about Roman architectural cul-
ture derives from Vitruvius, who in his De Architectura
(I century B.C.) discussed about building knowledge
gathering together specific information, construction
guidelines and design principles probably widespread
in Rome at the birth of the Empire. In the II book
of this treaty it can be found the first written refer-
ence to pozzolana mortars: “Est etiam genus pulveris,
quod efficit naturaliter res admirandas. Nascitur in
regionibus Baianis in agris municipiorum, quae sunt
circa Vesuvium montem. Quod commixtum cum calce
et caemento non modo ceteris aedificiis praestat fir-
mitates, sed etiam moles cum struuntur in mari, sub
aqua solidescunt.” (Vitruvio, 1997, II, VI: 132). Vitru-
vius referred to the pulvis puteolana, a sand originated
from thick deposits of reactive volcanic tuffs present
around Puteoli, the vesuvian town near Naples (today
Pozzuoli) from which the name of this material is
derived (Lugli, 1957; Adam, 1989; Lancaster, 2005).
This kind of sand, currently used in building Roman
practice, was also easily found in the surroundings
of Rome, which is situated between two well known
volcanic districts. Only two other important Roman
958
authors seem to have focused on pozzolana after Vit-
ruvio. Seneca mentioned pozzolana (62–65 A. D.),
spending on it only few words (Seneca, 2004, III,
XX, 3: 285). Plinius, instead, in his Naturalis Historia
(within 78 A. D.), wrote about pozzolana underling its
ability of hardening underwater (Plinio, 1987, XXXV,
XLVII: 1259); he also seems to be the first author
who hinted at pozzolana capacity of increasing mortar
strength with the age.
The worth of Vitruvius’ work increases if we con-
sider that in his treat the first written indications
for obtaining pozzolana lime mortars can be traced,
asserting that the ideal proportion (pozzolana:lime)
was 2:1 (as volumetric ratio) “Hae autem structurae
quae in aqua sunt futurae, viderunt sic esse facien-
dae uti portetur pulvis a regionibus quae sunt a Cumis
continuatae ad promonturium Minervae, isque mis-
ceatur uti in mortario duo ad unum respondeant”
(Vitruvio, 1997, V, II–III: 586–588). Later, many
authors referred to pozzolana in their architectural
treaties as well, but they didn’t give practical and tech-
nical information on the use of this volcanic material.
In particular, from XV century up to the end of XVIII
century, architects substantially referred to Vitruvius’
treat adding only accessory comments. Among more
than fifty treaties analyzed, there were traced only 24
mix designs pozzolana lime mortars.
During the XV and the XVI centuries, the only Ital-
ian author that considered the pozzolana for practical
purposes was Francesco di Giorgio Martini (1480–
82), who quoted Vitruvius’ pozzolana lime mortar mix
design (Martini, 1967, I, VIII: 105) but misunderstand-
ing his indications and inverting the pozzolana:lime
volumetric ratio (Table 1).
No pozzolana lime mortars mix designs can be
traced in the most important Italian Renaissance archi-
tectural treaties. Alberti (1450) widely discussed about
lime mortar and limestones, but only hinted at poz-
zolana (Alberti, 1989, II, XII: 86). Filarete (1460–64)
hinted to pozzolana too, but sending back to Vitru-
vio for further information (Filerete, 1972, III, p. 67).
Serlio (1537) spent a lot of pages of his work describ-
ing important roman buildings, but didn’t write about
materials and didn’t mention the use of pozzolana.
Vasari (1550), though not directly referring to vol-
canic material, praised the building good manufacture
in the area around Pozzuoli (Vasari, 1986, XXVII: 74).
Palladio (1570) hinted at pozzolana (Palladio, 1980,
I, IV, p. 15), focusing his attention on the hydraulic
lime used in the Venetian region around the town
of Padova (Palladio, 1980, II, IX, p. 77) and indi-
cating, as the first among the Renaissance authors
examined, the quick hardening as a peculiar quality
of pozzolana mortar. Cataneo (1567) was the first
who tried to explain the hardening process of a poz-
zolana lime mortar (Cataneo, 1985, II, IV: 269–270)
and Scamozzi (1615) dedicated an entire chapter of
Table 1. Pozzolana lime mortar mix designs traced in architectural treaties.

Volcanic
Reference Period Pozzolana Lime Sand Granite Pumice Breccia

Vitruvio (I century B.C.) 2 1

Martini F. di Giorgio 1480–82 1 2
Viviani Q. 1830 12 9 6 6
”” 1830 2 1 1
Valadier G. 1831 5/6 1/6
Quatremere de Quincy 1832 3 1
”” 1832 2 1
Cavalieri San Bertolo N. 1832 0.85 0.15
”” 1832 0.75 0.25
”” 1832 0.70 0.30
”” 1832 0.55 0.45
”” 1832 0.64 0.36
”” 1832 0.78 0.22
de Cesare F. 1855 2 1
”” 1855 6 5 3 6
”” 1855 4 4 3 9
”” 1855 57 11 8
”” 1855 3.5 6 3.5 9
Claudel J. & Laroque L. 1863 0.2 0.25 0.94
”” 1863 0.45 0.89 1.00
”” 1863 0.04 0.36 1
Curioni G. 1864 2 4 1
”” 1864 3 1
”” 1864 1 1 2

his work describing pozzolana, but without giving any
practical indication of its use (Scamozzi, 1982, II, VII,
XXII: 236–237).
During the end of the XVIII and the first half of the
XIX century, up to the Sixties, architects and engineers
turned to focus their attention on the practical uses of
pozzolana. Some authors, like Milizia (1781) (Milizia,
1991, III, I, IV: 37–38), Rondelet (1802) (Rondelet,
1834, I, I, I, III, II:125; III: 147–151; II, II: 249–250)
and Valadier (1831) (Valadier, 1992, I, II, XI: 58–59)
discussed about the geological and the chemical char-
acteristic of the volcanic material, its geologic origin
and the deposits different location. On the other hand,
Viviani (1830) (Viviani, 1830, VI- II: 125), Valadier
(1831), Quatremere De Quincy (1832) (Quatremere
de Quincy, 1842, II: 99–100), Cavalieri San Bertolo
(1832) (Cavalieri San Bertolo, 1845, II, IV: 41), De
Cesare (1855) (De Cesare, 1855, I, V: 49), Claudel-
Laroque (1863) (in Curioni, 1864, 1, VII: 133) and
Curioni (1864) (Curioni, 1864, 1, VIII: 134) proposed
different mix designs (Table 1).
The growing concern with faster settings and more
durable mortars soon led to the development and to
the spread of modern cements. Since the late XIX
century, the increasing use of Portland cement dis-
placed the pozzolana lime mortar and led to the end
of the interest of the architectural culture and practice
toward it.
3 MATERIALS AND EXPERIMENTAL
PROGRAM
3.1 Geological and chemical aspects of pozzolana
used in experiments
To evaluate the natural pozzolana contribution to the
increase of the strength in a lime mortar, specific sam-
ples of mortar were prepared in laboratory adding the
Pozzolana rossa originated from the volcanic complex
of the Monti Sabatini (Rome, Italy). As it well known,
Rome is situated between two volcanic districts, the
Monti Sabatini and the Albano Hills (Colli Albani),
which have been active for most of the last millions
years (Karner et al., 2001b).
Tephra deposits from the Monti Sabatini crop out
extensively to west of the Tiber River and to the north
of Aniene River. Three major pyroclastic flow deposits
from the Monti Sabatini crop out extensively along the
Tiber River Valley. They are, from the oldest to the
youngest, the Tufo Giallo della Via Tiberina, the Tufo
Rosso a Scorie Nere, and the Tufo Giallo di Sacrofano.
A series of ash fall (i.e. pozzolana, which was used in

959
the tests), surge and pyroclastic flow deposits separate
these eruptive units (Karner et al., 2001a).
The chemical characterization of the pozzolana
used for laboratory tests is shown in Table 2.
3.2 Mix chosen and test procedure
The 24 mortar compositions traced in the architectural
treaties analyzed (I cent. B. C. – 1864) were repro-
duced and tested in a first testing phase. By the light
of this first experimental step, a second, more punc-
tual testing phase followed, choosing only five mix
designs (Table 3) (due to synthesis requirements, only
the second testing phase results are reported and dis-
cussed herein, while whole set of results can be found
in Sala, 2008).
In reproducing the mortar specimens, both aerial
hydrated lime and hydraulic hydrated lime were used.
The new European standard for limes (EN
459 2001) defines three types of limes: natural
hydraulic lime (NHL), natural hydraulic lime with
added material (NHL-Z), and artificial hydraulic
lime (Z). In the present study, the specimens were
prepared with natural moderately hydraulic lime
NHL 3.5 (resistance after 28days: 3.5–10 MPa;
UNI EN 459-1:2001).
No indications about water quantities are given
in the mortar mix designs reported in the architec-
tural treaties (Table 1); therefore, the percentage of
water added in the mortar specimens to be tested was
determined in relation to the workability of the mixes.
Tests were carried out according to European Stan-
dard EN 1015-11 procedure in terms of preparation
and conditioning of mortar specimens to determine
the flexural and compressive strength of mortars.
Table 2. Chemical composition (%) of the pozzolana used
for laboratory tests.
SiO2 Al2 O3 FeO2 TiO2 CaO MgO K2 O Na2 O p.f.
48.7 17.1 7.7 0.8 9.6 4.2 8.7 2 0.1
The test specimens were prisms 160 × 40 × 40 mm;
three specimens were provided for each age of test.
Preparation and condition of storing specimens were
chosen in compliance with EN 1015-11.
The specimens were cast in metal moulds and
eventually placed in polyethylene bags for 2 days;
afterwards, the moulds were removed and the spec-
imens remained in the bags for the following 5
days (storage temperature of 20◦ C ± 2◦ C; relative-
humidity 95% ± 5%). After these 7 days of curing,
the samples were placed in a humidity room for
21 days (room temperature = 20◦ C ± 2◦ C; relative
humidity = 65% ± 5%).
Flexural strength of mortars was determined by
three-point-bending tests on the prism specimens. The
two halves derived from the flexural strength tests pro-
vided, for each age of test, six half prisms to be used
for the compressive strength tests.
Moreover, other tests were carried out on dif-
ferent specimens, named “sandwiches” (two bricks
and a mortar joint of 20 mm), in order to show a
relationship between the compressive strength and
a probe penetration rate. The method is based on
the measurement of the penetration depth increment
of a steel probe by means of repeated blows of a
Schmidt rebound hammer (Felicetti & Gattesco, 1998;
Giuriani & Gubana, 1993).
Specimens were tested after 28 days and 60 days of
curing.
4 EXPERIMENTAL RESULTS
4.1 Flexural and compressive strength
Test results, at the ages of 28 days and 60 days, are
shown in Table 4.
4.2 Penetration tests
Penetration tests results were obtained after 28 and
60 days of curing (Figure 1); results are reported in
Table 5.

Table 3. Mortar composition with different volumetric ratio.

Reference Period Identification Pozzolana A.H. Lime H.H. Lime Sand Water

Vitruvio (I century B.C.) V1 52.6% 26.3% 21.1%

Vitruvio (I century B.C.) V2 51.6% 25.8% 22.6%
Viviani Q. 1830 Viv1 41.5% 20.8% 20.8% 16.9%
Viviani Q. 1830 Viv2 40.6% 20.3% 20.3% 18.8%
Quatremere de Quincy 1832 Q 61.5% 20.5% 18.0%
Cavalieri San Bertolo N. 1832 Ca1 56.6% 24.5% 18.9%
Cavalieri San Bertolo N. 1832 Ca2 57.5% 24.1% 18.4%
Curioni G. 1864 Cu 60.0% 20.0% 20.0%

960
Table 4. Evolution of average flexural and compressive mix design (V2), with 2:1 pozzolana:lime volumetric
strength. ratio, shows the best results.
Vitruvius’ mix and, in particular, V2 specimen
Avg flexural Avg compressive shows the highest strength increment with time; in fact,
strength [MPa] strength [MPa]
the compressive strength of 7.16 MPa after 28 days of
Reference 28d 60d Incr. 28d 60d Incr. curing, reaches 9.49 MPa after 60 days.
Tests on Ca2 (7:3 and no sand) provide lower values
V1 1.39 1.51 8.57% 4.47 5.38 20.5% of flexural and compressive strength in comparison
V2 2.12 2.66 25.4% 7.16 9.49 32.5% with V2 specimens, therefore 2:1 seems to be the best
Viv1 2.04 2.09 2.30% 6.33 7.96 25.7% volumetric ratio for mortar composed by pozzolana
Viv2 1.38 1.20 −13.0% 4.26 4.76 11.9% and hydraulic hydrated lime.
Q 1.63 1.77 8.61% 5.46 6.78 24.2% Using aerial hydrated lime, results are little differ-
Ca1 1.47 1.52 3.72% 4.85 5.21 7.43% ent. Q mix (3:1 and no sand) shows the best mechanical
Ca2 1.59 2.03 28.1% 5.83 6.73 15.5% performance. The use of a greater pozzolana percent-
Cu 0.81 1.03 26.9% 4.07 4.22 3.79%
age in the mix is evidently due to the hydrated lime
incapacity to increase strength with time. The strength
increment is therefore ensured only by pozzolana
addition.
Viv1 and Viv2 specimens have the same poz-
zolana:lime volumetric ratio (2:1) of specimens V1
and V2, but they also have a percentage of sand. The
flexural and compressive strength results lower than
Vitruvius’ one.
During the preparation of the mortar specimens it
was observed that the medium value of the volumetric
percentage of water added in the mortars, for obtain-
ing the same workability, is 20% and that the water
increases for higher quantities of pozzolana. On the
other hand, the use of sand in the mix doesn’t seem to
Figure 1. Penetration test.
influence the water quantity.
As mentioned above, in order to verify a non-
Table 5. Evolution of average penetration tests results. destructive test method for practical uses, some pen-
etration tests were performed. Experimental results,
Number of hits/cm
plotted in Figure 3 for 28 and 60 days of curing,
Reference 28d 60d show a good direct correlation between compressive
strength and penetration test results. The trend line,
V1 5.44 6.56 obtained from the test results after 28 days of curing
V2 8.00 11.11 (Figure 3), shows a good correlation between the two
Viv1 7.67 11.33 quantities (K2 = 0.8251). This evidence demonstrates
Viv2 5.56 8.56 the possibility of using penetration tests to obtain cor-
Q 6.78 10.11 rect information about historical mortars mechanical
Ca1 5.33 8.67
characteristics with a non-destructive test.
Ca2 5.89 10.78
Cu 4.00 6.00 The second diagram in Figure 3 shows the trend
line between compressive resistances and penetration
tests after 60 days of curing. The average deviation
(K2 = 0.7066) is lower than the previous one; how-
ever, the direct correlation between the two quantities
5 DISCUSSION
is clearly evident.
The diagram in Figure 1 shows the influence of aging
on the compressive strength of different pozzolana
mortars. It can be observed that, in general, mortars 6 CONCLUSIONS
with hydraulic hydrated lime (V2, Viv2, Ca2) show
higher mechanical behavior and reach higher com- The research deals with the study of historical lime
pressive resistance values in comparison with the same mortars enriched with natural pozzolana. Historic
mortars with aerial hydrated lime (V1, Viv1, Ca1). mix designs were traced in the architectural treatises,
Referring to results from mechanical tests on mortar trough the analysis of more then fifty authors from
with pozzolana and hydraulic hydrated lime,Vitruvius’ Vitruvius (I cent. B. C.) to Curioni (1864).

961
Figure 2. Influence of aging on compressive strength.
Figure 3. Trend lines. Average Compressive Strength – Penetration tests results (28 days and 60 days).
From an historical point of view, the study of
mortars manufacture provides useful information on
the building culture and the use of building materi-
als, eventually helping in indicating different build-
ing phases and in setting conservative interventions.
From a technological point of view, mortar identi-
fication is a fundamental step in the design of the
repair materials, taking into account the necessary
compatibility between the original materials and the
new ones.
An exhaustive sampling of natural pozzolana lime
mortars was carefully accomplished in order to start
off a first study on their mechanical properties which,
up to now, were subdued to relatively little scien-
tific investigation. Flexural and compressive tests were
carried out, after 28 and 60 days of curing.
The main results of this study can be summarized
in the following.
962
– Historical analysis of architectural treaties shows a
continuous reference to Vitruvius’ indications. Up
to the end of the XVIII century, official architectural
culture seemed to be concerned with the good
properties of the pozzolana and of volcanic materi-
als but it didn’t propose new pozzolana lime mortars
compositions. Only at the beginning of the XIX
century, many authors turned to examine the prac-
tical uses of pozzolana, deepening its technical
properties and giving new mix designs.
– An exhaustive number of mortars mix designs were
selected and reproduced in laboratory in compli-
ance with UNI EN 1015-11. They were character-
ized in relation to their mechanical properties. The
tests clearly show that the compressive mechanical
strength increases with the addition of pozzolana.
– The mortars with hydraulic hydrated lime, proposed
by Vitruvio (V2; 2:1 pozzolana:lime volumetric
 ratio), showed the best mechanical properties; good
performance was also shown by Quatremere de
Quincy’s mix design (Q) (3:1), using aerial hydrated
lime.
– In general, mortars obtained with hydraulic
hydrated lime show a better mechanical behavior
than those obtained with aerial hydrated lime.
– Some further experiments were also performed in
order to verify the reliability of the penetration test,
a simple and non-destructive test that is signifi-
cant for practical use (mortar prisms are not easily
obtainable from historic buildings). The compari-
son between the compressive strength determined
from penetration tests and the one determined from
compressive tests confirmed that there is a good
correlation. Therefore, the penetration test, can be
successfully adopted for studying the mechanical
response of mortar joints in historical masonry and,
moreover, it can be easily used for in-situ tests.
ACKNOWLEDGEMENTS
Grateful acknowledgements go to Professor Alberto
Clerici, University of Brescia, for the helpful discus-
sions on the geological subjects; to Eng. Nicola Bettini
for his contribution in editing the paper.
The Authors would like to thank the companies
T.C.S. srl (Brescia, Italy) and Cava Lapillo e Poz-
zolana di Arnaldo Gentile & C. snc (Viterbo, Italy)
for providing the hydraulic hydrated lime NHL 3.5
and Pozzolana Rossa, respectively.
Finally, a special acknowledgement goes to
Mr. Domenico Caravaggi, of the Laboratory Pisa of
the University of Brescia, for his co-operation in
performing the experiments.
REFERENCES
Adam, 1989, L’arte di costruire presso i Romani: materiali e
tecniche, Milano: Longanesi
Alberti L.B., 1989, L’architettura, Milano: Il Polifilo
Averlino A. detto il Filarete, 1972, Trattato di architettura,
Milano: Il Polifilo
Cataneo P. & Barozzi G. da Vignola, 1985, Trattati, Milano:
Il Polifilo
Cavalieri San Bertolo N., 1845, Istruzioni di architettura,
statica e idraulica, Mantova: Negretti
Claudel J. & Laroque L., 1850, Pratique de l’arte de
construire, Parigi : Carilian-Goeuryet
Collepardi M., 1993, La produzione del calcestruzzo antico
e moderno, Proc. Convegno antico e moderno di Bres-
sanone, Bressanone, 6–9 July 1993: 181–192
Curioni G., 1864, L’arte di fabbricare, ossia corso completo
di istituzioni teorico-pratiche, Torino: Negro
De Cesare F., 1855, La scienza dell’architettura, Napoli:
Giovanni Pellizone
EN 196-1, 1996, Methods for testing cement. Determination
of strength
963
EN 459-1, 2002, Building lime. Definitions, specifications
and conformity criteria
Felicetti R. & Gattesco N., 1998, A penetration test to study
the mechanical response of mortar in ancient masonry
buildings, Materials and Structure, 31: 350–356
EN 1015-11, 2006, Methods of test for mortar for masonry.
Part 11: Determination of flexural and compressive
strength of hardened mortar
Giuriani E. & Gubana A., 1993, A penetration test to eval-
uated wood decay and its application to the Loggia
monument, Materials and Structures, 26: 8–14
Karner D.B. & Marra F., 2001a, Florindo F. & Boschi E.,
Pilsed uplift estimated from terrace elevations in the coast
of Rome: evidence for a new phase of volcanic activity?,
Earth and Planetary Science Letters, 188: 135–148
Karner D.B., Marra F. & Renne P.R., 2001b, The history of the
Monti Sabatini and Alban Hills volcanoes: groundwork
for assessing volcanic-tectonic hazards for Rome, Jour-
nal of Volcanology and Geothermal Research, 107, 1–3:
185–215
Lancaster L.C., 2005, Concrete Vaulted Construction in
Imperial Rome, New York: Cambridge University Press
Lugli G., 1988, La tecnica costruttiva romana con particolare
riguardo a Roma e Lazio, Roma: Giovanni Bardi
Martini F. di Giorgio, 1967,Trattati di architettura ingegneria
e arte militare, Milano: Il Polifilo
Milizia F., 1991, Principj di architettura civile,, Roma: Sapere
2000
Moropoulou A., Cakmak A., Labropoulos K.C., Van
Grieken R. & Torfs K., 2004, Accelerated microstructural
evolution of a calcium-silicate-hydrate (C-S-H) phase
in pozzolanica pastes using fine siliceus sources: com-
parison with historic pozzolanic mortar, Cement and
Concrete, 34, 1: 1–6
Palladio A., 1980, I quattro libri dell’architettura,, Milano:
Il Polifilo
Plinio il Vecchio, 1987, Plinii Naturalis Historia, Pisa:
Giardini
Quatremere de Quincy A.C., 1842, Dizionario storico di
architettura, Mantova: F.lli Negretti
Rondelet G., 1834, Trattato teorico e pratico dell’arte di
edificare, Mantova: Caranenti
Sala E., 2008, Duttilità e confinamento delle murature
storiche e caratterizzazione meccanica di malte poz-
zolaniche riportate nella trattatistica d’architettura (I sec.
a.C.-1864), Ph.D Thesis, University of Trento, Italy
Scamozzi V., 1982, L’idea dell’architettura universale,
Bologna: Arnoldo Forni
Seneca, 2004, Questioni naturali, Milano: BUR
Serlio S., 2001, L’architettura, Milano: Il Polifilo
Valadier G., 1992, L’architettura pratica, Roma: Sapere 2000
Vasari G., 1986, Le vite de’più eccellenti architetti, pittori, et
scultori italiani, da Cimabue insino a’tempi nostri,Torino:
Einaudi
Vitruvio M.P., 1997, L’architettura di Vitruvio, Gros P. (ed.),
Torino: Einaudi
Viviani Q., 1830, L’architettura diVitruvio tradotta in italiano
da Quirico Viviani, Udine: Mattiuzzi
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Blended lime-cement mortars for conservation purposes:

Microstructure and strength development

O. Cizer, K. Van Balen & D. Van Gemert

Department of Civil Engineering, Building Materials and Building Technology Division, K.U. Leuven,
Heverlee, Belgium

J. Elsen
Research Group of Applied Geology and Mineralogy, Department of Geography–Geology, K.U.Leuven,
Heverlee, Belgium

ABSTRACT: Blended lime-cement mortars are commonly used in conservation practices even though they
may show lack of adequate strength and durability for certain cement-lime compositions. This paper focuses
on understanding the hardening reactions and their influence on the strength development, microstructure and
porosity for the cement-lime mortars in various compositions. Mortars composed of 30%, 50% and 70% cement
replacement with lime hydrate and lime putty by mass were studied. Cement hydration has been found to con-
tribute to the early stage strength development while carbonation is mostly favoured after 3 days and contributes
to the strength development until 180 days. The degree of carbonation is much more pronounced with increasing
lime content and porosity of the mortars and the reaction is still in progress at 90 days. All the blended mortars
revealed lower compressive and flexural strength than that of the reference cement mortar due to their lower
cement content and higher porosity. Long-term compressive strength development has been achieved after 180
days while flexural strength for certain mortar compositions does not increase beyond 28 days. Unlike cement
mortar, the mortars blended with lime hydrate and lime putty exhibit an elastic-plastic deformation before failure
occurs, which is preferred for repair mortars to adapt to differential settlements and to allow more deformation
under critical stresses in the masonry.

1 INTRODUCTION Balen et al. 2005). In spite of this, addition of cement

Since ancient times, lime based binders such as high
calcium lime, lime with pozzolana and hydraulic lime
have been successfully used in construction and have
proven their long-term durability. Towards the end of
the 19th century, development in the building industry
and increase in the construction needs encouraged the
search for stronger mortars with more rapid harden-
ing. This eventually induced the development of the
Portland cement which became a dominant binder in
the construction industry after that period. This has
inevitably led to the replacement of lime based binders
with the cement in conservation practices, which has
resulted in serious damaging mechanisms related to
the incompatibility of cement in chemical, physi-
cal and mechanical properties when used together
with ancient building materials. There exist evidences
that stronger mortars can reduce the durability of
the masonry whereas mortars with low compressive
strength like lime mortars can increase the durability
of the masonry in the long-term (Van Hees et al., Van
to lime based mortars is still a common practice in
most of the conservation works in order to promote the
hardening reaction by cement hydration, to shorten
the construction time and to increase the strength of
the mortar. However, certain mortar compositions with
lime and cement may cause lack of strength develop-
ment and durability. The main objective of this study
is, therefore, to investigate the influence of the cement
addition on the strength development, porosity and
microstructure of the lime mortars, and to understand
the hardening reactions. This will provide fundamental
insight while defining the compatibility requirements
for repair mortars.
Literature concerning the hydration of cement in the
presence of additional lime is quite limited. Researches
have been mainly devoted on understanding the influ-
ence of lime addition on the hydration of tricalcium
aluminate (Collepardi et al. 1978, Kuzel & Pollmann
1991) as well as on porosity, microstructure and
strength development of various cement-lime mortar
compositions (Sebaibi et al. 2006, Mosquera et al.

965
2006, Arandigoyen & Alvarez 2007). Apart from Table 1. Composition of the reference cement mortar, lime
cement hydration, carbonation reaction contributes to hydrate mortar and cement-lime hydrate mortars.
the strength development of the cement-lime mortars,
which is generally neglected. Therefore, in this study Cement Lime Water
special emphasis is given on understanding hardening
Mortar (g) (g) (g) (w/b)
reactions and their influence on the strength develop-
ment, porosity and microstructure of the cement-lime Cref 450 – 202.5 0.45
mortars using lime hydrate and lime putty as a partial L – 450 392 0.87
replacement of cement in varying ratios. C7L3 315 135 234 0.52
C5L5 225 225 270 0.60
C3L7 135 315 315 0.70
2 RESEARCH PROGRAMME

2.1 Materials Table 2. Composition of the lime putty mortar and cement-
lime putty mortars.
Portland cement (CEM I 52.5 N), commercial lime
hydrate and lime putty were used as binders for mor- Lime Water Total
tars. The lime hydrate is composed of 83.22% calcium Cement putty added water
hydroxide content with a Blaine specific surface area
of 19,464 cm2 /g and a mean value of its particle size Mortar (g) (g) (g) (g) w/b
distribution of 4.40 µm. The lime putty was produced
at the laboratory through slaking quicklime (0–50 mm) LP – 1125 – 675 1.50
in 1:3 quicklime-water ratio. The putty was then stored C7LP3 315 342 94.5 301.5 0.67
in hermetically sealed containers to promote its aging C5LP5 225 569.7 51.3 396 0.88
under water. The lime putty used in the mortars was C3LP7 135 797.5 – 482.5 1.07
aged for one year and contained 60% water by mass.
It is composed of 83.47% calcium hydroxide with a
Blaine specific surface area of 21,739 cm2 /g. laboratory conditions (20◦ C, 60% R.H.) for 180 days.
The progress of hardening was studied by means of
mechanical strength test and thermal analysis. The
2.2 Mortar compositions mechanical strength tests were carried out by com-
Cement was partially replaced by lime hydrate and pressive strength and three-point bending using the
lime putty in 30%, 50% and 70% by mass of the total standard mortar prisms at 3, 7, 28, 60, 90 and 180
binder (Table 1 and Table 2). Reference cement (Cref ), days of hardening (EN 196-1). For thermal analysis the
lime hydrate (L) and lime putty (LP) mortars were mortar specimens were first gently ground in order to
studied as well for the comparison of the data. The separate sand grains from the binder fraction and then
mortars were prepared using standard quartz sand with vacuum dried at 0.025 mbar. After that, the sample
maximum grain size of 2 mm (EN 196-1). Binder/sand was sieved through 112 µm sieve to obtain the binder
ratio of the mortars is 1:3 by mass. Due to higher fraction which was ground further for thermal anal-
specific surface area of calcium hydroxide particles ysis. This analysis was carried out using a Netzsch
than cement, more water was required in cement-lime STA 409 PC DSC-TGA system in static nitrogen atmo-
hydrate and cement-lime putty mortars to reach similar sphere at a temperature range between 20-1000◦ C with
workability (125–130 mm) with the reference cement a controlled heating rate 10◦ C/min. Porosity of the
mortar, which was measured by the flow table test (EN mortars at 90 days has been studied with total porosity
1015-3). The water content increased with increasing (EN 1936) and pore size distribution using Mercury
lime content in the mortar composition. Considering Intrusion Porosimetry (Micromeritics AutoPore IV
the cement-lime putty mortars, additional water was 9500. Microstructure of the mortars was studied using
required only for the C7LP3 and C5LP5 mortars as the Philips XL 30S FEG Scanning Electron Microscope
water content available in the lime putty used was ade- (SEM) coupled with X-Ray Energy Dispersive Sys-
quate to gain the required workability for the C3LP7 tem. The analysis was carried out on the fresh fractured
mortar. surfaces coated with gold after the samples were dried
at 45◦ C.

2.3 Methods 3 RESULTS

Hardening of the mortars was studied using stan-
dard mortar prisms (40 × 40 × 160 mm) which were
prepared in accordance with the European standard
(EN 196-1). Mortars were cured at the standard
3.1 Porosity
Bulk density and total porosity of the mortars are
given in the Table 3, and their cumulative pore size

966
Table 3. Bulk density and total porosity of the mortars at
90 days.

Bulk density Total porosity

Mortar (g/cm3 ) (%)

Cref 2.19 16.78

L 1.86 28.72
LP 1.71 33.50
C7L3 2.05 18.50
C5L5 1.96 22.39
C3L7 1.87 27.21
C7LP3 1.98 19.19
C5LP5 1.81 28.23
C3LP7 1.70 35.18

Figure 2. The first derivative of the weight loss of the

reference cement mortar at 3, 7, 28 and 90 days of hardening.

recorded at temperatures between 600–800◦ C derives

Figure 1. Cumulative pore size distribution of the mortars.
distribution is given in the Figure 1. The bulk den-
sity decreases and the total porosity increases with
increasing lime content in the mortars compositions.
Cement-lime hydrate mortars show a higher volume
of pores with a diameter between 0.1 µm and 1 µm
than those of the reference cement mortar (Figure 1).
The volume of the pores within this diameter range
increase with increasing lime content in the mortar
composition.Volume of pores with a diameter less than
0.1 µm increased as well. Cement-lime putty mortars
indicated higher volume of pores than the cement-lime
hydrate mortars for the same percentage of cement
replacement.
3.2 Hardening reactions by thermal analysis
The first derivative of the weight loss (DTG) recorded
during thermal analysis of the reference cement mor-
tar is represented in Figure 2. This technique allows
following the progress of hydration and carbona-
tion reactions qualitatively. Evolution in the peaks of
hydrated phases can be seen at temperatures around
110◦ C due to the dehydration of the C-S-H phase
and around 160◦ C due to the dehydration of the AFm
phase. The peak observed at around 480◦ C indicates
the decomposition of calcium hydroxide.Another peak
from the decomposition of the calcium carbonate.
In comparison with the hydration reaction of the
reference cement mortar, DTG thermographs of the
blended mortars are represented in Figure 3. Peaks
observed are similar with those of the reference cement
mortar with some differences in the intensities.
3.3 Mechanical strength of the mortars
Compressive and flexural strength development of the
mortars is given in Figure 4 and Figure 5 respectively.
While the reference cement mortar yielded the highest
strength values at all stages, these values became lower
as cement was partially replaced with lime hydrate
and lime putty at 30%, 50% and 70% by mass respec-
tively. Cement-lime hydrate mortars indicated higher
strength values than the cement-lime putty mortars
for the same amount of replacement. Their compres-
sive strength values increased gradually until 180 days
while no considerable increase was recorded after 28
days in the flexural strength of the C7L3, C5L5, and
C7LP3 mortars.
Lime hydrate (L) and lime putty (LP) mortars
yielded the lowest strength values of all blended mor-
tars except for the C3LP7 mortar which has revealed
flexural strength values lower than that of the LP
mortar at all stages.
In Figure 6, stress-strain diagram obtained from the
compressive strength testing of the reference cement
mortar (Cref ) and cement-lime (C5L5) mortar at 90

967
 Figure 4. Compressive strength development of the
mortars.

Figure 5. Flexural strength development of the mortars.

Figure 3. The first derivative of the weight loss of the

cement-lime hydrate and cement-lime putty mortars at
3, 7, 28 and 90 days of hardening.

days is given. Although these results are influenced by

the set-up conditions of the testing machine to some
extend, stiffness and deformation of the mortars can
be estimated. The stiffness decreases with increasing
lime content in the mortar composition. In comparison
to the cement mortar, the cement-lime mortars reveal
more deformation and withstand a measurable post- Figure 6. Stress-strain behaviour of the Cref and C5L5
failure load before the failure occurs due to their plastic mortars.

968
Figure 7. SEM micrograph of the hydrated matrix of the
reference cement mortar at 90 days, indicating reticular amor-
phous and needle-like C-S-H phase where calcium hydroxide
crystals are locally embedded.
Figure 8. SEM micrograph of the hydrated matrix of
the C7L3 cement-lime hydrate mortar at 90 days, indicat-
ing a cluster of calcium hydroxide crystals surrounded by
amorphous C-S-H.
behaviour.This particular behaviour is much more pro-
nounced for the lime mortars and cement-lime mortars
with 70% lime content.
3.4 Microstructure
Partial replacement of cement by calcium hydrox-
ide has led to the modification of the microstructure
of the mortars to some extend. Reference cement
mortar is characterized by a microstructure com-
posed of reticular amorphous and needle-like C-S-H
phase where calcium hydroxide crystals are locally
embedded (Figure 7). The most distinguishing feature
between the microstructure of the reference cement
mortar and cement-lime mortars is the presence of
calcium hydroxide crystals in the latter (Figure 8).
They are present individually or as clusters embedded
969
Figure 9. SEM micrograph of the plate-like crystals, prob-
ably AFm crystals, observed in the hydrated matrix of the
cement-lime mortars.
Figure 10. SEM micrograph showing a calcium hydrox-
ide (CH) crystal, hydrated cement grain and smaller cal-
cium hydroxide having corroded edges where subsequent
formation of C-S-H is observed (highlighted within the
rectangle).
within the hydrated matrix. These calcium hydroxide
crystals indicated corroded edges where amorphous
C-S-H phase has precipitated subsequently and has
grown towards the surrounding matrix (Figure 10 and
Figure 11).
Plate-like crystals which probably belong to the
AFm phase were observed locally in the hydrated
matrix (Figure 9). With increasing lime content in the
mortar composition, less amorphous C-S-H phase is
observed and the microstructure becomes more porous
with more calcium hydroxide crystals, plate-like crys-
tals and needle-like crystals (Figure 12).
4 DISCUSSION
Partial replacement of cement with lime hydrate and
lime putty has resulted in lower compressive and

Figure 11. A detailed SEM micrograph showing the for-
mation of amorphous C-S-H on the edges of the calcium
hydroxide crystals.
Figure 12. SEM micrograph of the hydrated matrix of the
C3L7 cement-lime hydrate mortar at 90 days.
flexural strength values for the blended mortars in
comparison with the strength of the reference cement
mortar (Figure 4 and Figure 5). Strength values
become lower respectively with 30%, 50% and 70%
replacement of cement with lime hydrate and lime
putty by mass. This is related with both decreasing
cement content and increasing water/binder ratio in
the composition of these blended mortars, which has
resulted in an increase in their porosity. With decreas-
ing cement content and increasing calcium hydroxide
content, an increase in the volume of pores with diam-
eters less than 1 µm is observed (Figure 1). C3L7
and C3LP7 mortars, which have 70% replacement
of cement with lime hydrate and lime putty, revealed
total porosity values that are very close to the poros-
ity values of the L and LP mortars respectively. The
cement-lime putty mortars revealed higher porosity
than that of the cement-lime hydrate mortars for the
same amount of replacement. This can explain their
970
lower strength values than those of the cement-lime
hydrate mortars. Considering C3LP7 mortar having
70% cement replacement with lime putty, increase in
the volume of pores is much more pronounced than
the rest of the mortars especially for the pore diame-
ters between 10 µm and 0.6 µm as well as those less
than 0.3 µm.
30% replacement of cement with lime hydrate
has resulted in 38.4% reduction in the compressive
strength of the reference cement mortar while it is
56.3% reduction in case of replacement with lime putty
at 180 days. The reduction in the flexural strength is
23.6% for cement-lime hydrate mortar and 46.8% for
the cement-lime putty mortar. Further replacement of
the cement at 70% has led to extremely lower compres-
sive and flexural strength values. This is much more
pronounced for C3LP7 mortar as its flexural strength
value (0.45 N/mm2 ) is even lower than that of the LP
mortar (0.8 N/mm2 ). Therefore, the mortar composi-
tion with 30% cement and 70% lime putty does not
improve the strength of the mortar in comparison with
the LP mortar. This is not the case for C3L7 mortar
as it has higher compressive and flexural strength val-
ues than those of the L mortar and LP mortar but quite
lower than those of C5L5 and C7L3. However, its com-
pressive strength does not increase beyond 60 days as
well as its flexural strength beyond 28 days.
Thermal analysis results indicated that hardening
reactions responsible for the strength development of
the blended mortars are cement hydration and car-
bonation reaction (Figure 3). The latter is expected
due to the presence of additional calcium hydroxide
in the mortar compositions from the beginning. This
is clearly seen in the DTG thermographs with the
increase in the intensity of the calcium carbonate peak
at temperatures between 600◦ C and 800◦ C. Degree
of carbonation increases in time and with increasing
calcium hydroxide content in the mortar composi-
tion as well as increasing porosity which favours the
diffusion of the carbon dioxide into the core where
calcium hydroxide is still present to carbonate (Van
Balen & Van Gemert 1994). Cement hydration is
mostly complete within 28 days as the intensity of
the peaks of the hydrated phases does not increase
beyond that period. These phases are C-S-H and AFm
which are the same with those of the reference cement
mortar. The presence of these phases has been veri-
fied by SEM analysis. Microstructure of the reference
cement mortars is composed of reticular amorphous
and needle-like C-S-H phase where calcium hydrox-
ide crystals are locally embedded (Figure 7). Hydrated
microstructure of the blended mortars is character-
ized by clusters of calcium hydroxide crystals that
are surrounded by amorphous C-S-H phase (Figure 8)
and locally formed AFm crystals (Figure 9). These
calcium hydroxide crystals indicated corroded edges
where amorphous C-S-H phase formed subsequently
(Figure 10 and Figure 11). With increasing lime con-
tent in the mortar composition, the microstructure
becomes more porous with less amorphous C-S-H
phase and more calcium hydroxide crystals as well as
hexagonal plates and needles (Figure 12). These plate-
like and needle-like crystals are assumed to belong to
the AFm phase due to the peak observed at around
160◦ C in the DTG thermographs (Figure 3).
In the literature, hexagonal hydro-aluminate plates
were reported to be formed during the hydration of
the tri-calcium aluminate in the presence of lime
(Collepardi et al. 1978, Taylor 1990). Their forma-
tion in our mortar compositions is not clear from
the DTG thermographs as the hexagonal hydro-
aluminates reveal dehydration peaks between 100◦ C
and 200◦ C, which is the same dehydration range for the
C-S-H and AFm phases, and therefore their presence
can be masked.
With decreasing cement content in the mortars
composition less hydrated phases and more calcium
carbonate phase are observed (Figure 3). At 3 days
of hardening, hydrated phases can already be seen in
the thermographs but the peak of the calcium car-
bonate evolves in time. This shows that the cement
hydration takes place at the very early stage while the
carbonation reaction becomes effective after 3 days
and continues until 90 days. Therefore, the cement
hydration contributes to the early stage strength devel-
opment until 28 days. In the meantime, carbonation
reaction continues and contributes to the strength
development until 90 days. This is quite clear from the
compressive strength development of the mortars but
the flexural strength of all cement-lime hydrate mor-
tars and only C7LP3 mortar does not increase beyond
28 days.
In comparison to the reference cement mortar, stiff-
ness decreases with increasing lime content in the com-
position of the blended mortars, which enhances their
physical and mechanical compatibility with ancient
mortars. The cement-lime mortars indicated a rea-
sonable plastic behaviour and more deformation than
the reference cement mortar. Similar to the self-
healing ability of lime mortars, cracks occurring due
to the deformation can be healed by either dissolution-
precipitation cycles of calcite or carbonation reaction
of the calcium hydroxide that is still available in
the matrix. However, the influence of formation of
calcium carbonate crystals on the initially hydrated
phases and its consequence on the strength develop-
ment and durability of the blended mortars should be
considered well.
5 CONCLUSIONS
Hardening of the cement-lime mortars occurs as a
result of combined cement hydration and carbonation.
971
Cement hydration contributes to the early stage
strength development while carbonation is mostly
favoured after 3 days and proceeds gradually until 180
days. Degree of the carbonation is much more pro-
nounced with increasing lime content and porosity of
the mortars. The cement-lime mortars have revealed
lower compressive and flexural strength than the ref-
erence cement mortar due to lower cement content
and higher porosity of the blended mortars. Long-term
compressive strength development has been achieved
after 180 days while flexural strength for certain mor-
tar compositions does not increase beyond 28 days.
Further replacement of the cement at 70% with lime
putty does not improve the strength of the mortar
in comparison with the LP mortar while this is not
the case when lime hydrate is used. Unlike the refer-
ence cement mortar, the cement-lime mortars exhibit
an elastic-plastic deformation which enables them to
adapt to differential settlements and more deforma-
tion under critical stresses in the masonry. Considering
the mortar compositions and hardening reactions,
compatibility of the lime-cement mortars should
be further evaluated in terms of physical, mechan-
ical and durability requirements for conservation
purposes.
ACKNOWLEDGEMENT
This study is part of an ongoing research project
(OT/04/27 3E030765) conducted at the Department
of Civil Engineering at the K.U. Leuven. The research
grant offered to Ozlem Cizer by the university is grate-
fully acknowledged. Department of Metallurgy and
Materials Engineering (MTM) at the K.U.Leuven is
gratefully acknowledged for the use of the Scanning
Electron Microscope.
REFERENCES
Arandigoyen, M. & Alvarez, J.I. 2007. Pore structure and
mechanical properties of cement–lime mortars. Cement
and Concrete Research 37 (5): 767–775.
Collepardi, M., Baldini, G., Pauri, M. & Corradi, M.
1978. Tricalcium aluminate hydration in the presence of
lime, gypsum or sodium sulphate. Cement and Concrete
Research 8: 571–580.
EN 196-1. 1987. Methods of testing cement – Part 1:
Determination of strength. European Standard.
EN 1015-3. 1998. Methods of test for mortar for masonry –
Part 3: Determination of consistence of fresh mortar (by
flow table). European Standard.
EN 1936. 1999. Natural stone test method – Determination
of real density and apparent density and of total porosity
and open porosity. European Standard.
Kuzel, H. & Pöllmann, H. 1991. Hydration of C3A in the
presence of Ca(OH)2 , CaSO4 · 2H2 0 and CaCO3 . Cement
and Concrete Research 21: 885–895.
Mosquera, M.J., Silva, B., Prieto, B. & Ruiz-Herrera, E.
2006. Addition of cement to lime-based mortars: Effect on
pore structure and vapor transport. Cement and Concrete
Research 36(9): 1635–1642.
Scrivener, K.L. & Capmas, A. 1998. Calcium Aluminate
Cements. In P.C. Hewlett (ed.), Lea’s Chemistry of Cement
and Concrete: 718–724. 4th Edn. London: Arnold.
Sebaibi, Y., Dheilly, R.M., Beaudoin, B. & Quéneudec, M.
2006. The effect of various slaked limes on the microstruc-
ture of a lime–cement–sand mortar, Cement and Concrete
Research 36(5): 971–978.
Taylor, H.F.W. 1990. Cement Chemistry. London: Academic
Press Limited.
972
Van Balen, K. &Van Gemert, D. 1994. Modeling Lime Mortar
Carbonation Materials and Structures 27: 393–98.
Van Balen, K., Hayen, R. & Van Gemert, D. 2005. Weaker can
be better: Learning from the past contributes to sustainable
construction technology with lime. International Building
Lime Symposium, Proc. intern. symp., Orlando, Florida,
9–11 March. NLA Building Lime Group.
Van Hees, R.P.J., Binda, L., Papayianni, I. & Toumbakari, E.
2006. Damage analysis as a step towards compatible repair
mortars. In C. Groot, G. Ashall & J. Hughes (eds.), Char-
acterization of old materials with respect to their repair –
RILEM TC 167-COM: 107–152. RILEM publications
S.A.R.L.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Assessing workability of mortar by means of rheological parameters

and desorptivity

R. Hendrickx, K. Van Balen & D. Van Gemert

Department of Civil Engineering, Katholieke Universiteit Leuven, Heverlee, Belgium

ABSTRACT: Assessing the workability of mortars for restoration and more general purposes has been the
subject of many standards and scientific publications over the last decades. A series of mortars made with
different binders was brought to optimum water content for workability by an international group of masons.
Two methods are proposed to assess the static yield stress and water retention of these fresh mortars, two key
parameters of workability as defined by practitioners.Yield stress is measured with a vane test, originally designed
for clayey soil, and found to be related to specific surface area of the binder. The increase of yield stress with
resting time is quantified and the differences due to the effect of air entraining agent are demonstrated. Water
retention is characterised by the desorptivity of the mortars and measured in a vacuum suction experiment. The
setup is a modified version of an existing ASTM design. Both methods are complementary to assess the defined
workability parameters for different mortars and eventually can be an aid to design suitable mortars for specific
purposes.

1 INTRODUCTION guaranteed (Coussot 2005, Wallevik 2003). Existing

When selecting or designing a suitable restoration
mortar for a specific situation, the first concern is for
a material which will be compatible with the existing
fabric in appearance, composition (type of binder and
aggregate), mechanical, hygric and thermal behaviour
(Van Balen et al. 2005). These properties, which refer
to the hardened state, are interrelated in a complex
manner to the fresh properties of the mortar, such as
the rheology, water permeability and adhesion to the
substrate. Workability is a general term to cover these
fresh properties, and is considered by masons as highly
important. A mortar should be readily moulded in the
voids to be filled up, and then gain stiffness to keep
its shape after application. The kneading water should
be sucked out of the mortar by the substrate at a suffi-
cient rate to provide this stiffening, but not too fast as
to avoid difficulties in shaping the mortar body prop-
erly, and not too much as to avoid an insufficient water
content for full hydration of the hydraulic binder.
Rheology in the most broad sense describes the
deformation of matter, and more specifically of every-
thing which lies between a perfect elastic solid and an
ideal Newtonian fluid. The rheology of mortar in gen-
eral is highly complex because mortar is a granular
paste of questionable homogeneity with a solid frac-
tion far above the usual in the discipline. It applies the
laws of continuum mechanics, which is feasible if a
sufficiently large scale is considered, and homogeneity
methods to assess rheology of mortars are reviewed
in (Banfill 2005). They can be divided into rotational
methods, compressive rheology and some practice-
oriented or empirical tests methods. This paper focuses
on a rotational method, based on the principle of a
rotating vane, which induces yielding in the circum-
scribing cylindrical plane. This method is particularly
suited for materials which are sensitive to slip and seg-
regation at the cylinder surface (Lidell & Boger 1996).
Existing models that describe the rheological
behaviour of mortar all have a yield stress parameter,
and one or more parameters quantifying the stress-
strain relationship under laminar flow. It is generally
recognized that the behaviour of mortar changes with
time, and that thixotropy should be taken into account
to properly describe these phenomena (Banfill 2005).
Nevertheless, the most widely used models assume
the existence of one single steady state flow curve,
e.g. following the Bingham model: τ = τ0 + µγ̇, with
τ the shear stress, τ0 the shear stress at zero shear rate,
µ the plastic viscosity and γ̇ the shear rate. We will
show that mortar has in reality a dynamic yield stress,
which is dependent on time effects, deformation his-
tory and normal force. Thus the values obtained in our
vane test depend on resting time and the time scale of
the imposed stress or strain. Normal force is assumed
to be low in the test setup, so that the influence can
be neglected. The study of the time scale of the exper-
iment is restrained by a constant rotational velocity

973
driving mechanism, so that possible effects on yield Table 1. Description of binders.
stress are not taken into account here. The specific
apparatus used in this study was originally designed No. Description
for lab measurements of clayey soil and has recently
been proposed for use on binder pastes and mortars 1 calcic lime CL90S
2 calcic lime CL90S (EN 459-1:2001) (75%) with
(Hendrickx et al. 2006, Bauer et al. 2007).
hydraulic binder (15%) and pozzolana (10%)
The transport of water through a volume of semi- 3 calcic lime CL90S (EN 459-1:2001) with air
solid mortar slurry or paste can be described using entraining agent
the same principles as those applied for porous solids. 4 calcic lime CL90S (EN 459-1:2001) (66.7%) and
A complete modelling of flow from mortar to substrate ordinary Portland cement CEM I 42.5 R (33.3%)
using FEM could be done when the complete water 5 natural hydraulic lime NHL 5 (EN 459-1:2001)
retention curve and hygroscopic curve of both materi- 6 ordinary Portland cement CEM I 42.5 R
als are known, as well as the initial state and boundary 7 ordinary Portland cement CEM I 42.5 R with
conditions. At present, established experimental meth- admixtures: plasticizer and air entraining agent
ods do not allow the mortar to be characterised to this
extent. Alternatively, a simple sharp front model to
describe the dewatering of a slurry based on filter Table 2. Chemical properties (XRF) and specific surface
area (BET) of binders.
cake theory, is derived in (Hall and Hoff 2002) and
experimentally validated in (Collier et al. 2007). This 1 2 3 4 5 6
model presents the desorptivity R of the wet mortar at
constant capillary suction of the substrate as the main CaO (%) 96.6 75.1 96.3 75.6 51.2 63
parameter to characterise the mortar. R is determined MgO (%) 0.94 1.88 1.05 1.49 1.68 1.74
experimentally for different types of mortar using a Al2 O3 (%) 0.059 1.640 0.072 2.6 4.9 4.6
simple pressure filter used in oil industry (Carter et al. SiO2 (%) 0.144 6.2 0.19 9.6 16.6 16.7
2003). In analogy to the sorptivity of bricks, which has Fe2 O3 (%) 0.048 1.55 0.063 2.7 2.1 3.6
been accepted to be a reliable measure for its capillary CO2 (%) 2.0 10.6 2.0 3.0 19.8 2.9
absorption, desorptivity of mortar provides also more SO3 (%) 0.10 1.84 0.18 2.80 2.10 4.50
useful information than the well-known standardised SSA (m2 /g) 12.0 5.7 11.9 4.5 3.5 1.3
tests, which measure a quantity of water absorbed by
filter paper sheets (EN 413-2:2005) or filter paper
plates (EN 459-2:2001). The apparatus described in 100

ASTM C91, using vacuum suction, is adapted in the 90

experiments described below, to be used in a procedure 80

70
to measure desorptivity.
Fraction passing (%)

60
50
40
2 MATERIALS AND METHODS
30
20
2.1 Mortar composition
10
The description of the binders are listed in Table 1, with 0
a standardised denomination is possible. The chemi- 0.01 0.1 1
Sieve aperture (mm)
cal properties measured by X-ray fluorescence (XRF)
and specific surface area (SSA) measured with the Figure 1. Grading of the aggregate (Zutendaal sand).
BET method are given in Table 2. Binder 2 and 3
are prefabricated admixtured products; binder 7 is a Table 3. Mortar composition in weight ratios.
lab mix of the ordinary Portland cement (binder 6),
plasticizer (modified stearic acid in powder) and an Binder: 1 2 3 4 5 6 7
air entraining agent (fatty acid/polyglycol solution).
The dosage was 0.4% and 0.1% of the binder weight. B:A (kg/kg) 0.13 0.20 0.13 0.21 0.21 0.25 0.25
The aggregate is a medium siliceous quarry sand 0/0.5 W:B (kg/kg) 2.02 0.90 1.48 1.00 1.13 0.93 0.79
(0/1) (EN 13139:2002) from a quarry in Zutendaal,
Belgium (Figure 1).
The compositions of the batches are given in Table 3 prefabricated blended binders (binder 2 and 3), they
by their binder to aggregate weight ratio (B:A) and are in agreement with the producer’s guidelines. W:B
water to binder weight ratio (W:B). B:A values are values have been determined for optimum workability
derived from practical experience and are comparable by a panel of six masons (Hendrickx et al. in press).
to values given in various national standards. For the Mixing is done using a Hobart planetary mixer at low

974

calibrated
torque spring
vane
φ1
inner scale
knob with
carrier
ωt
φ2
outer scale
Figure 2. Working principles of the soil vane test.
speed: 1.5 min mixing, 30 s homogenising, 2 min mix-
ing. All the components are added in the beginning:
first the water, then half of the aggregate, then the
binder and at last the remaining half of the aggregate.
2.2 Yield stress experiments
Yield stress was measured with a vane apparatus for
clayey soil (ASTM D4648) (Figure 2). It measures
the maximum torque necessary for a vane (DxH),
immersed in a sample container with mortar, when
rotation at constant is initiated. Previous research has
demonstrated that container dimensions of three times
the dimensions of the vane, are sufficient to avoid any
influence of edge effects (Lidell & Boger 1996). The
mechanism is designed in a way that the inner scale
indicates φ1 , the amount of deflection over the spring
and thus the magnitude of the torque, and the inner
scale indicates φ2 , the angle over which the vane has
rotated inside the sample. From the moment that the
mortar yields and the vane starts to move quickly, φ1
remains constant and φ2 starts to increase at a rate ωt.
It can be seen that φ1 = ωt − φ2 . The torque exerted
on the vane can be expressed as: T = B · φ1 with
B a constant of the calibrated spring, indicating the
torque per unit of rotation. Hence the shear stress in
a cylindrical plane circumscribing the vane can be
calculated as T /K, with the vane constant K equal to:
975
80
φ1 (°)
φ2 (°)
60
φ1 + φ2 (°)
Angle (˚)
40
20
0
0 1 2 3 4 5 6 7 8
Time (min)
Figure 3. Typical course of an experiment at constant
ω = 9.2◦ /min (mortar with binder 7).
The assumption that shear occurs exactly in this place
is a good approximation (Lidell & Boger 1996). The
homogeneity of the material close to this shear plane
is discussed further. A typical test shows three stages
(Figure 3). In the initial stage there is an increase of
φ2 due to a small repositioning of the vane and rear-
rangement of grains in the mortar. In the second stage,
the increase of both φ1 and φ2 are constant. In the
third stage, the spring angle reaches a maximum and
the residual angle increases more, until the material
yields: this last phenomenon can not be derived from
the scale reading, due to the configuration appara-
tus. All mortars were measured with a spring constant
B = 380◦ /Nm, except binder 4, for which a vane with
B = 260◦ /Nm was used. The vane has dimensions
D = H = 45 mm. Rotation speed ω was constant and
equal to 9.2◦ /min.
2.3 Desorptivity experiments
The test setup form ASTM C91 was adapted to contain
a larger sample volume. The sample is introduced in
a perforated dish on top of a wetted filter paper. The
sample height is increased with respect to the standard-
ised version from 20 to 60 mm (Figure 4). This enables
to have sufficient data points when suction is applied
for identifying the initial linear regime of water loss
with the square root of time (Hall and Hoff 2002):
with i the amount of evacuated water, expressed in mm
water column, R the desorptivity and t time. The pres-
sure difference p1–p2 is generated by a vacuum pump,
in stead of the pressure cell used in (Carter et al. 2003).
The mass of the dish with the filter paper and sample is
measured at the beginning of the test, after wiping the
bottom of the dish quickly with a dry cloth. The vac-
uum pressure is applied to the sample via the funnel
by turning a valve. The seal between the funnel and the
dish is made with a layer of silicone, which is applied to
 p 150
binder 3 unsheared
1
binder 3 pre-sheared
binder 7b unsheared

4 cm
mortar sample binder 7b pre-sheared
100

Yield stress (Pa)

filter paper

2
50

0
0 20 40 60 80 100
Resting time (min)

Figure 5. Yield stress unsheared and pre-sheared for air

entrained mortars.

binder 1 unsheared binder 1 pre-sheared

p 150
binder 4 unsheared binder 4 pre-sheared
2 binder 5 unsheared binder 5 pre-sheared
binder 6 unsheared binder 6 pre-sheared

Figure 4. Geometry of the desorptivity test. The section of

100
Yield stress (Pa)
the dish is circular, diameter 155 mm.

the funnel and moulded in contact with the dish. Seal- 50

ing with petroleum or light grease is avoided because it
complicates the removal and accurate weighing of the
dish at different times. Measurements were done after
0
1, 4, 9, 25, 36, 49, 64 and 81 minutes, until the level 0 10 20 30 40 50 60 70 80 90 100
of the vacuum could not be sustained any more due to Resting time (min)

gas breakthrough. Gas breakthrough occurs increas-

ingly from a certain point onwards, but initially, the Figure 6. Yield stress unsheared and pre-sheared for non air
pressure drop over the sample can be kept constant by entrained mortars.
increasing the pump suction. The repeated procedure Table 4. Yield stress values after 30 and 60 min of resting
of closing the valve, removing and wiping the dish, time.
weighing, replacing, and reopening the valve was done
in 45 s. Binder: 1 3 4 5 6 7
It is assumed that the pressure, applied through the
perforations and the filter paper, is equally divided over τy vane , 98 91 78 69 64 54
the bottom surface of the sample. We further assume t = 30 s (Pa)
that the mortar’s solid fraction does not segregate dur- τy vane , 116 115 108 97 75 97
t = 60 s (Pa)
ing the time of the test and that the pressure drop over
the filter paper is negligible compared to the pres-
sure drop over the sample. The maximum pressure
difference that can be realised in this way is atmo- is observed for air entrained mortars (15 to 30% air
spheric pressure. If higher pressures are needed, use content, binders 3 and 7) and non air entrained mor-
of a pressure cell is an alternative. tars (3 to 7% air content, binders 1, 4, 5 and 6). In the
first group the increase of yield stress is limited during
the first half hour of resting time, and becomes more
important during the second half hour. The second
3 RESULTS AND DISCUSSION
group shows an adverse effect, with a large increase in
the early stage and a smaller increase after 40 min. In
3.1 Rheological experiments
both groups a large drop of yield stress is obtained by
The results of the vane test are represented as a func- pre-shearing the mortar. This means that only a part of
tion of the actual time when yielding occurs (Figure 5 the stiffening is irreversible.
and 6). By doing so, the duration of the test, which is For comparison of the different mortars, yield stress
often 10 min or more, is taken into account. The val- values after 30 and 60 min of resting time were cal-
ues range from 20 tot 80 Pa shortly after mixing, and culated by linear interpolation (Table 4). It has to
rise above 100 Pa after 1 hour. A distinct behaviour be stressed that these values are valid only for the

976
 105 10

100 9

95 8
Yield stress, t=30 min (Pa)

90 7

85 6

i (mm)
Binder 1
80 5
Binder 2
75 4 Binder 3

3 Binder 4
70
Binder 5
65 2 Binder 6
1 Binder 7
60
0 2 4 6 8 10 12 14
0
SSA (m²/g) 0 20 40 60 80 100
t0.5 (s0.5)
Figure 7. Yield stress measured with vane at 30 min of
resting time as a function of SSA of binder powder. Figure 8. Quantity of desorbed water as a function of time
for mortars with different binders.
(relatively low) rotation rate in this setup and that
3.5
strictly spoken, this method is neither rate controlled,
nor exactly stress controlled. The results are to be inter- 3
preted as a good basis for comparison: changing the
2.5
test parameters will influence the order of magnitude
of the absolute numbers. 2
y' (mm)

Segregation of mortar by rising of water towards

1.5
the surface (“bleeding”) was observed during the
tests, especially in mortars with high desorptivity (see 1
below). This implies that, for those mortars, at least
part of the yield stress increase is due to increasing 0.5

solid fraction, while for the other mortars, we can speak 0

of a real increase of internal structure.
When comparing the mortars without entrained air,
a correlation can be noted between the yield stress and
the specific surface area (SSA) of the binder (Figure 7).
The practical meaning of this correlation is that, even
when W:B is optimised for each of the mortars, the
differences in behaviour remain measurable. Different
binders have a different optimum yield stress: binders
with larger SSA are judged workable by masons even
at higher measured yield stress.
3.2 Water retention experiments
The amount of water removed from the mortar at
constant vacuum pressure increases proportionally to
the square root of time until a certain point, from
where it tends to an asymptotic value (Figure 8). This
asymptotic value corresponds to the equilibrium water
content of the material for the applied vacuum suction.
The observed behaviour can be fitted by least
squares to a linear function for the earlier part and
an exponential function for the later part. The transi-
tion point between both parts is determined by visual
assessment of the curves. The slope of the linear part
corresponds to the desorptivity R of the mortar, as
defined in equation 2. The later data points can be
fitted to an exponential equation of the form:
0 10 20 30 40 50 60
x' (t 0.5 )
Figure 9. Fitting of data to an exponential curve; data of
mortar with binder 1, vacuum pressure 200 mbar.
where (x , y ) are the coordinates relative to the transi-
tion point with the linear part. A + x = µ(ieq ) is an
estimator for ieq , the equilibrium value of i after a
long time. Figure 9 gives an example for mortar with
binder 1. From µ(ieq ), the geometry of the test and
the mortar compositions, we can calculate µ(weq ),
the estimated equilibrium water content (in kg/m3 ) of
the mortar after a long time at the applied vacuum
pressure.
Table 5 gives the measured values and calculated
parameters for the 7 tested mortar types, as well as
the values of water retention (WR) obtained for these
mortars in the standardised tests with filter paper and
filter paper plates.
For increasing pressures with the same mortar,
we observe both a higher initial increase R and a
higher final equilibrium water content weq (Figures 10
and 11). The increase of desorptivity can be empiri-
cally fitted to a power law function of the shape:
where C and n are the empirical constants. This cor-
responds to the findings of other authors (Green et al.

977
 Table 5. Numerical results of standardised water retention tests, desorptivity measurements and
regression analysis.

Binder: 1 2 3 4 5 6 7

WR filter paper (%) 93 94 95 93 85 87 88

WR filter plates (%) 83 93 92 81 66 63 78
R(mm/s0.5 ) 0.13 0.08 0.14 0.12 0.26 0.32 0.20
A(mm/s0.5 ) 3.19 1.87 2.80 3.59 5.21 5.78 5.24
B (10−3 s−0.5 ) 38.7 50.0 27.4 29.8 44.2 48.3 30.0
win (kg/m3 ) 353 178 217 292 326 313 236
µ(weq ) (kg/m3 ) 237 116 144 139 175 145 72

10 100%

8 90%

Water retention (filter plates)

7

6 80%
i (mm)

4 70%

3 200 m bar
300 m bar
2 60%
400 m bar
1 500 m bar

0 50%
0 20 40 60 80 100 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35

t0.5 (s0.5) R (mm/s0.5)

Figure 10. Quantity of desorbed water as a function of time Figure 12. Water retention measured with filter paper plates
for different vacuum pressures and mortar with binder 1. as a function of calculated desorptivity.

0.25 the transition resistance between the different sheets

confirm that this test cannot yield a physically relevant
0.2 value. The filter paper plate test on the other hand was
found to give a good indication for the desorptivity
0.15
under vacuum (Figure 12).
R (mm/s0.5)

0.1 4 CONCLUSIONS

0.05 Two important aspects of mortar workability, its rhe-

0
0 100 200 300 400
Vacuum pressure (mbar)
Figure 11. Desorptivity of mortar with binder 1 (air lime)
as a function of applied vacuum pressure, fitted with to power
law function: R = 9.75 · 10−3 · P0.5 .
1999, Carter et al. 2003). The exponent is related
in filter cake theory to the compressibility of the
filter cake, 0.50 corresponding to the condition of an
incompressible cake (Green et al. 1999).
The results of the standardised water retention tests
using filter paper sheets (EN 413-2:2005) could not be
correlated in a meaningful way to data obtained from
other experiments. Observations of the importance of
ology and water retention, have been discussed for
a series of mortars with different binders. The yield
500 600 700 stress has been measured successfully with the labora-
tory vane test for clayey soil. Results however remain
valid only for the applied rotation rate, vane dimen-
sions and spring characteristics. The influence of these
parameters is still to be investigated. The differences
in absolute value between the mortars’ yield stress and
the different behaviour in time for air entrained and non
air entrained mortars have been demonstrated.This test
has the important advantage over the flow table test or
penetration tests, that it gives a meaningful physical
quantity as a result, which can easily be interpreted
in terms of practical use. The advantage over scien-
tific rheometers is that the instrument has lower cost,
is available in most soil mechanical labs and has a
suitable torque range and dimension.

978
 The same advantages are valid for the desorptiv-
ity experiment with the vacuum pump: it starts from
a small adaptation of a standard setup which is avail-
able in a large number of building material labs. The
actual process of suction by a porous brick resembles
more to this setup, than to the technique with a pres-
sure cell used by other researchers. The disadvantage
is the rather laborious test procedure and the restric-
tion of applied suction to atmospheric pressure. The
result is more meaningful than the results of the tradi-
tional standardised tests with filter paper of filter paper
plates, because it yields a rate of water loss in stead
of one single point. Further more the data can be used
to estimate the equilibrium value at the applied vac-
uum pressure, which allows to characterize the water
retention curve of the mortar in the covered pressure
range. The weakness remains that for high desorptiv-
ity mortars, a significant inclination of the curve is
observed already after the second point, which causes
the calculated desorptivity value to be less accurate,
deviating towards the lower end. The measured des-
orptivity value is found to be inversely proportional
to the water retention value measured by filter paper
plates. This means that this simple test is a suitable
method to estimate desorptivity.
The observed differences between the mortars can
be related to the value of the SSA and to the effect of air
entraining agents in the mortar. SSA is inversely pro-
portional to desorptivity, and addition of air entraining
agent has a decreasing effect, though of less impor-
tance. There is evidence for a negative correlation
between mortar desorptivity and yield stress, through
their dependence on the SSA. A higher SSA gives
lower desorptivity and higher yield stress for mortars
without air entraining agent. Caution is needed how-
ever to interpret the results because of the different
B:A values of the mortars tested. A similar trend was
found for the mortars with air entraining agent. These
findings illustrate the important effect of air content
979
and SSA on both yield stress and desorptivity, two of
the most important aspects in assessing the workability
of mortar.
REFERENCES
Banfill P. 2005. The rheology of fresh mortar – a review,
In Cincotto, M.A., Silva, D.A., Oliveira, J., Roman, H.R
(ed.), Proceedings of the 6th Brazilian and 1st Interna-
tional Symposium on Mortar Technology, Florianopolis,
Brazil: 73–82.
Bauer E., De Sousa J.G.G., Guimarães E.A., and Silva
F.G.S. 2007. Study of the laboratory vane test on mortars,
Building and environment, 42: 86–92.
Coussot P. 2005. Rheometry of pastes, suspensions and
granular materials, Wiley, Hoboken (New Jersey).
Green K.M., Carter M.A., Hoff W.D., Wilson M.A. 1999. The
effects of lime and admixtures on the water-retaining prop-
erties of cement mortars, Cement and concrete research
29: 1746–1747.
Hendrickx R., Van Balen K., Van Gemert D., Minet J. 2006.
Workability and rheology of lime-based mortars, 15. Kol-
loquium und Workshop über Rheologische Messungen an
mineralischen Baustoffen, Regensburg, 8–9 March 2006.
(unpublished presentation)
Hendrickx R., Minet J., Van Balen K., Van Gemert D. in press.
Workability of mortars with building lime: assessment
by a panel of masons versus lab testing, 14th Interna-
tional brick and block masonry conference, Sydney, 16–20
February 2008. (accepted for publishing on CD-ROM)
Lidell, P.V. & Boger D.V. 1996. Yield stress measurements
with the vane, Journal of non-newtonian fluid mechanics,
63: 235–261.
Van Balen K., Papayianni I., Van Hees R., Binda L.,
Waldum A. 2005. Introduction to requirements for and
functions and properties of repair mortars, Materials and
structures 38, 8: 781–785.
Wallevik J.E.H. 2003. Rheology of Particle Suspensions.
Fresh Concrete, Mortar and Cement Paste with Various
Types of Lignosulfonates, PhD dissertation, Norwegian
University of Science and Technology.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Consolidation and reinforcement of stone walls using a

reinforced repointing grid

A. Borri, M. Corradi & E. Speranzini

Department of Civil and Environmental Engineering, School of Engineering, University of Perugia,
Perugia, Italy

A. Giannantoni
Servizi di Ingegneria srl, S. Eraclio di Foligno (PG), Italy

ABSTRACT: A new technique for reinforcing rubble stone masonry walls (double and triple-leaf walls), when
it is required to keep the fair-face masonry. The reinforcement technique consists of a continuous mesh made
of tiny steel cords embedded perfectly in the mortar joints after a first repointing, and anchored to the wall by
means of galvanized steel eyebolts driven into the facing. A second repointing covers the cords and the heads of
the eyebolts completely. This leads to genuine reinforced fair-face masonry in which, as already confirmed by
the first experiments, the compression, shear and flexural strength are increased, effective transverse connection
between the facings of the masonry due to the presence of the eyebolts and also the capacity to withstand tensile
stresses. The reinforcement is non-invasive and reversible, and is aimed at integrating the masonry rather than
transforming it. It is compatible with preservation of the material of which the artefact is made and is long-lasting
in view of the materials used, which are very resistant to aggression. The analysis of this reinforcement/upgrading
work has led to the formulation of practical criteria for sizing the reinforcement and to the assessment of the
strength of the reinforced panel.

1 INTRODUCTION Various different techniques are used currently to

The consolidation and strengthening of vertical
masonry elements of masonry buildings that are sub-
jected not only to their own weight but also to possible
dynamic stresses (seismic events, wind), constitute
one of the most important reinforcement works for
achieving an adequate level of safety.
This is because poor quality referred to the mechan-
ical features of the masonry (compressive strength,
shear strength, etc.), in particular in ancient buildings,
has often been the cause of collapsing or of seri-
ous damage, for example (although not only) during
seismic events.
Another element that has a particular impact
on the seismic behaviour of a masonry construc-
tion is the connection between vertical walls and
between these and the horizontal elements. If these
connections are present and effective, they can allow
the construction to respond adequately, in the form
of “box-like” behaviour, to dynamic stress, without
losing the balance of single portions.
In the absence of these connections, each single ele-
ment (a wall, floor slab, etc.) will be more vulnerable,
since it will be free to collapse separately from the rest
of the construction.
reinforce masonry constructions.Among the works for
strengthening wall panels it is possible to recall here
the following:
A) works for restoring the internal continuity of
masonry which has developed cracks, such as local
rebuilding operations on the masonry or insertions
of metal bars.
B) works connecting the various masonry elements to
one another, in addition to the “sewing” mentioned
above. It is possible to mention here circling with
strips of composite material (FRP: Fiber Rein-
forced Polymers, based on carbon fibre, glass fibre
or other materials), which is particularly effective
and not very invasive. A negative aspect is the fact
that the reinforcement remains visible, so that it is
not very suitable for fair-face walls.
C) works aimed at reinforcing the masonry. Here it is
possible to mention:
C1) injection into the masonry of mixtures (typi-
cally cement or lime-based grouts), having the
purpose of filling the empty spaces inside the
wall panels, and also of replacing the origi-
nal mortar, which is often not very solid. This
technique, however, is not very effective in the

981
 case of walls with a low index of voids (which
is a fairly frequent occurrence). Also, it is not
easy to investigate the actual spreading of the
mortar injected into the panel.
C2) ferro-cement, consisting of making two thin
reinforced concrete walls connected to one
another by means of metal connectors. This
technique is often used to reinforce irregular
or poor-quality masonrywork and is undoubt-
edly very effective from the mechanical point
of view, however it is very invasive. The inter-
nal masonry is “lost” from every point of view,
not only because it is no longer visible but
above all since it is subject in time to rapid
deterioration. Another negative aspect is that
the new wall panel is far stiffer than the orig-
inal one (and this often has a negative impact
on the structural behaviour of the building).
C3) Another technique used is that of deep
repointing of the mortar joints. This consists of
stripping the joints in the masonry by removing
the original poor-quality mortar for a depth of
several centimetres (typically 6 to 8 centime-
tres) and then repointing the joints with a good
quality mortar. If this rienforcement is carried
out on both sides of the wall facing and the
facing is not very thick, it has a good degree
of effectiveness.
If the masonry is made of bricks, it is
also possible to insert small metal bars into
the joints (reinforced repointing). These can
increase the strength of the panel considerably.
It is clear, however, that even minor irregular-
ities in the fabric of the wall will lead to the
need to cut the bricks, which is out of keep-
ing with the principles of preservation of old
buildings.
In the case of irregular masonry (stones of
different shapes and sizes giving rise to irreg-
ularly shaped joints) reinforced repointing is
not suitable, since even if the diameter of the
reinforcement rods is small, the rods cannot
follow the irregular shapes of the joints.
C4) Another recently suggested technique con-
sists basically of a system of tie-rods leading
out in the three orthogonal directions (CAM
system – Dolce et al.). The tie-rods can also
be tensioned. This technique is definitely very
effective in mechanical terms and improves
the monolithic quality and the mechanical
behaviour of the body of the wall, however
it cannot often be proposed for old buildings,
since it has a strong impact on preservation of
the old material.
The main aim of the technique suggested in this
paper is to eliminate or at least reduce the problems
referred to above.
982
The system proposed can be used both at local
level, that is to say for single walls of existing build-
ings (and also for boundary walls such as, by way of
example, town walls), and at global level, that is to
say as a system for reinforcing a masonry construc-
tion in its overall behaviour, with particular but not
sole reference to behaviour during seismic events.
The system can be used to treat masonry, whether
regular or irregular, without causing the impact and
without the invasiveness of other techniques. It is
therefore particularly suitable for fair-face walls of
buildings listed due to their historical and architectural
features.
The improvements that can be achieved do not
consist simply of improvement of the mechanical char-
acteristics of the wall treated in this way. Indeed, in
addition to strengthening the wall panel, the “rein-
forced skeleton” of the continuous mesh introduced
into the wall (and which has been given the name
of “Reticolatus”) also connects the contiguous wall
elements to one another (adjacent orthogonal walls,
horizontal elements, foundations, etc), thus forming
a genuine system for reinforcing the whole masonry
building.
The small size of the reinforcing elements and the
ease with which they can be fitted into the mortar joints
also enables widespread use, thus avoiding harmful
and dangerous concentrations of stress, such as those
that occur, for example, when metal rods are used.
2 DESCRIPTION OF THE SYSTEM
To sum up, the system consists of a continuous mesh
made of tiny cords made of high strength steel, inserted
perfectly into the mortar joints and thus embedded
in the wall in a natural way (that is to say without
intervening on the stone parts).
This reinforcing technique can be used equally on
regularly shaped walls (made of bricks or worked
stone blocks) or irregular (rubble and barely cut stone)
walls (multi leaf walls). It is in any case in the latter
case that it is often more necessary to improve the
mechanical properties of the wall, making this tech-
nique more interesting.This is because the possibilities
provided by other current techniques (repointing with-
out reinforcements or grout injection into walls that
are unsuitable since they are very compact) may not
be very effective or may be invasive and therefore
unacceptable from the point of view of preservation
(ferro-cement).
The system is based on the use of materials easy to
find on the market even if commonly used for other
purposes, that is to say:
a) High strength steel cords, which can be made from
coils to be found on the market. The coils are about
30 cm wide (Fig. 1) and variable in length from
15 m up to 1500 m, and consist of a series of cords

Figure 1. Coil of high strength steel cords.
Figure 2. (a) 3X2 cord , (b) 3SX cord.
Figure 3. Close-up showing the hooking of steel eyebolts.
arranged parallel to one another (Fig. 2) and held
together by a polyester mesh.
It is easy to pull the steel cords out from the
strip so as to use them separately for the application
suggested here. It is also possible, in any case, to
use other materials, such as composite materials
(cables or cords), provided it is possible to use a
cement or lime-based mortar as a matrix.
b) galvanized steel “eyebolts” (Fig. 3), also available
on the market, normally used in mountaineering for
providing a grip in rock faces. Basically, these are
pointed rods that can be driven into the facing, with
hooks or rings on their ends through which cables
can be passed.
As an alternative, it is also possible to use small
galvanized steel rods with threaded ends enabling the
983
Figure 4. Close-up of hooking system of threaded bars with
washers and nuts.
Figure 5. Reciprocal confinement of the “stone-
mortar-cord system.
cord to be held in place by means of a metal washer
and a locking nut (Fig. 4).
The most interesting property of the cords used in
the proposed system is the fact that the very small
size of the cords (typical average diameter 1 mm) and
their shape, formed by wrapping the single steel fil-
ament around each other helically (typically 3 or 4
filaments) give rise to high bonding and compatibility
between the cords and the mortar surrounding them.
This ensures excellent mechanical behaviour of the
“stone-mortar-cord” assembly. What is more, because
the cords are so small, they can easily be bent into
shape as required in order to pass them through the
joints between the various pieces of stone forming the
wall (Fig. 5). Cords made of composite material have
similar characteristics.
The eyebolts or threaded rods have two purposes: on
the one hand they provide “fixed points” for anchoring
the cords in the wall facing, on the other they create
connections between the two sides of the wall, thus
helping it to behave as a monolith.
The procedure for reinforcing a single wall panel
using the system suggested here is carried out in the
following stages:
• strip the mortar joints for a depth of 6–8 cm;
• wash the stripped joints;
• repoint with mortar (cement or lime-based mortar);
• insert the eyebolts or the rods, typically 4 per square
metre, arranged as regularly as possible according
to a square mesh and driving them in far enough to
involve the opposite facing, leaving the head below
the surface on which the work is being carried out
(at a depth of at least 3–4 cm);
• insert the cords into the stripped joints, passing them
trough the hooks of the eyebolts or behind the wash-
ers of the rods, proceeding in a horizontal or vertical
direction all over the facing being treated. If the
length of the single cords is insufficient, they can
be joined together with resin or, simply, overlapped
against each other for a length of about 20 cm;
• if considered necessary to further increase the
strength of the panel it is possible to apply addi-
tional cords diagonally, both in one direction and in
the other;
• repoint the joints again with mortar, so as to cover
up completely both the cords and the heads of the Figure 6. The two reinforced masonry facings enclose the
eyebolts or of the rods. inner masonry work.
Upon completing the operations described above,
a grid-like pattern (irregularly shaped) made of metal
or of composite material will have been obtained. It
will be perfectly incorporated into the wall facing but
not visible from outside, and capable of giving the
panel the mechanical characteristics (shear strength,
tensile strength and compression strength) it needs to
withstand both static and dynamic stresses.
To understand the effectiveness of this treatment, it
is possible to compare it with the ferro-cement tech-
nique. The end result of this technique is to confine
the masonry between two new thin reinforced concrete
walls (the two layers of plaster consisting of cement
mortar, reinforced with metal netting, on the two sides
of the masonry and connected to one another). With the
technique suggested here, the result is absolutely com-
parable in terms of effectiveness but the original facing
remains visible and perfectly intact, and its ability to Figure 7. Close-up of a corner area.
transpire remains unaffected. Indeed, here again there
are two new resisting walls (several centimetres thick)
connected to one another that enclose the masonry also be folded over to take in the panel at right-angles
between them (Fig. 6), but these are actually the origi- (Figs 7–8), thus connecting the vertical elements at
nal external facings which have become genuine layers right-angles to one another in a particularly effective
of reinforced masonry (reinforced wall facings) thanks manner (since they create an extensive connection).
to the grid of metal cords inserted “discreetly” but As far as concerns the bottom part of the panel (e.g.
intimately into the joints. in the foundations), the cords can be anchored effec-
Moreover, in addition to considerably increasing the tively by folding them around the pieces of stone at
tensile (and shear) strength of the masonry, this grid the bottom of the panel (Figs 7–8) or connecting them
also has the beneficial effect of confining the pieces to the foundations (if any) by means of connectors, as
of stone, that are enclosed within the circuits of the the case may be.
cords. At the top, the cords can be connected to the metal
To reinforce the whole masonry construction using or RC tie (if any), or, for example, in boundary town
the proposed system, it is possible to proceed as walls, folded over the coping so that they reach the
follows. other side of the wall.
At the ends of the wall panel, the cords can be folded As a result of these operations, each single wall was
in so that they also take in the opposite surface side of reinforced locally, and all the different elements were
the same panel. In the case of a corner panel, they can connected to one another, giving rise to a construction

984

Figure 8. Folded and anchored cords.
in which the resisting system now consisted of a
masonry reinforced to a widespread extent.
To sum up, the system presented here calls for a net-
ting structure with flexible sides consisting of the cords
mentioned above, which can be positioned accord-
ing to paths corresponding to the joints between the
elements of the wall and which can be anchored to
pre-established points of the latter by means of metal
eyebolts or transverse rods. Since the sides of the net-
ting are flexible and pass through the hooks or the rings
of the transverse elements without being an actual part
of them, it is possible to arrange them precisely along
the joints (previously stripped) of the elements forming
the masonry, however they are arranged.
3 FIELDS OF APPLICATION AND
EXPECTED BENEFITS
As already mentioned, the proposed technique is suit-
able for treating rubble stone walls when, for reasons
of preservation, the fair-face wall is required to be
retained and at the same time non-invasive reversible
and effective reinforcement is required. Typically,
therefore, some of the structural problems of old build-
ings and archaeological assets can be solved in a
manner capable of meeting the need for the high-
est safety standards combined with the requirements
relating to protection and preservation.
The proposed technique complies with the princi-
ples underlying the protection f historical buildings,
since it is:
– “compatible” with preservation of the material of
which the building is made, as it is able to adapt
and integrate perfectly into the walls;
– “long-lasting”, since the materials used have a high
resistance to chemical and physical aggression and
to weathering;
985
– intended to integrate the structure without trans-
forming it;
– non-invasive;
– reversible (or at least removable);
– compliant with the principle of the “least upgrading
work”.
It is easy to identify the possible advantages from
the mechanical point of view. They concern:
1) improved mechanical characteristics, that is to say
compressive strength and shear strength, and also
flexural strength in relation to stresses on the same
and at right-angles to the plane of the masonry;
2) the ability to connect extensively any damages, the
vertical walls to one another and the vertical walls
to the horizontal elements;
3) giving the masonry the tensile strength that, in the
case of irregular masonry (where the vertical joints
are often aligned) cannot even benefit from the
“chain effect” present in regular walls with properly
staggered orthostats and joints;
4) transverse connections between the facings of the
masonry, since the eyebolts or the transverse rods
are arranged as artificial stones capable of mak-
ing the wall panel act as a monolith. They do this
by contrasting detachment of the facings from one
another and providing adequate tensile strength
(also necessary in the presence of vertical loads
only and shear strength (necessary in the event of
stresses that tend to make the wall panel tilt over,
and therefore to make the facings slide in respect
of one another).
4 EARLY EXPERIMENTS
In order to investigate the effectiveness of the rein-
forcement technique described above, a series of tests
were planned.
After a description of the materials used, the first
results of the tests are reported on below.
4.1 Fibre characterization
The fibres used during the experiment were supplied
by Hardware LLC. One peculiarity of these materials
is their macroscopic structure. Indeed, all the fibres
are made up of high-resistance steel filaments covered
with a layer of brass to prevent oxidation of the metallic
fibres. The specifications of the cords used are shown
in Table 1.
4.2 Tests with flat jacks
In the tests with double jacks, portions of walls about
50 cm high were subjected to compression on a single
vertical axis.
Masonry walls were tested while subjected to com-
pression using two flat jacks (Fig. 9). During the test
Table 1. 3X2 cord mechanical properties. As far as concerned the normal modulus of elastic-
ity E1/3 , the results of the tests with flat jacks showed
Fibre type 3X2 cord that deep repointing of the joints on its own is not capa-
Cord diameter (mm) 0.89 ble of causing significant increases of the modulus of
Cross section area (mm2 ) 0.620 elasticity. This was probably because the mortar used
Failure tensile load (N) 1539
Young’s modulus (E) (MPa) 206842
for repointing the joints was based on hydraulic lime.
Failure stress (MPa) 2479 As far as concerns the manner in which the masonry
Elongation at failure (%) 2.1 elements between the two flat jacks failed, it was seen
that a series of vertical cracks formed between the
two flat jacks. Furthermore, there was no substiantial
14 differentiation of the type of failure between the un-
SRE 05

12
reinforced masonry (URM), the repointed masonry
SRE 04
(REP) and the masonry reinforced with metal fibres
10
(SRE). While in the cases of the unreinforced masonry
[kg/cm ]
2

REP 03
8 REP 02 and of the repointed joints the breakage occurred with
Vertical stress

6 URM 01
a small number of fairly large vertical cracks, in the
case of the masonry repointed with metal fibres a
4
larger number of smaller vertical cracks occurred. This
2 indicates an improvement in the mechanical behaviour
of the masonry.
0
0,000 0,002 0,004 0,006 0,008 0,010 The next experiments will investigate the increases
Normal strain ε
in shear strength and flexural strength. It is planned
to carry out on-site diagonal compression tests on un-
Figure 9. σ−ε diagram resulting from the tests with double
reinforced and reinforced panels, and also loading tests
flat jacks.
on panels that will first be reinforced, cut and then
Table 2. Results of the tests with flat jacks. tilted over, positioning them horizontally. They will
then be tested up to breaking point by applying vertical
Max Young’s loads.
compression modulus
Reinforcement Index stress (MPa) E1/3 (MPa)
4.3 Shear tests
Un-reinforced URM 01 0.595 480
Deep repointing REP 02 0.807 393 The diagonal test was performed on site on panels of
Deep repointing REP 03 0.857 512 1200 × 1200 mm dimension with sections of differ-
Reinforced repointing SRE 04 1.261 486 ent thickness and morphology. The load is given by
Reinforced repointing SRE 05 1.312 2416 hydraulic jacks. The test is defined by ASTM E 519-81
Standard. It is possible to calculate the characteristic
measurements were recorded the load history, maxi- strength of the masonry τk through:
mum compression stress σmax , and deflections of some
points using centesimal transducers. An equivalent
normal stiffness E1/3 was then calculated:

where P is the diagonal compressive load generated by

the hydraulic jack and A is the area of the horizontal
cross-section of the panel and τu is the ultimate shear
where ε1/3 is the normal strain corresponding to 33% strength.
of the maximum stress σmax reached. Furthermore it is possible to calculate the shear stiff-
Figure 9 shows the results for the first tests carried ness G1/2 (secant value of the modulus at 1/2 of the
out using double jacks. peak load) defined as:
On analysing the results, it is possible to state
that deep repointing of the joints, reinforced with
metal fibres, is capable of increasing significantly
the compressive strength σmax of the masonry. On
comparing these results with those concerning the non- where γ1/2 is the angular strain at 1/2 maximum
reinforced masonry and those for masonry reinforced load, τ]rmi and γi are respectively the initial shear
with repointing alone, increases in the resistance were stress (τi = 0.002 N/mm2 ) and strain values due to an
found of 116% and 50% respectively (Tab. 2). application of a pre-load.

986
Table 3. Results of diagonal compression tests.

Shear Shear
strength modulus
Reinforcement Index τk (MPa) G1/2 (MPa)

Un-reinforced DC 01 0.029 541

Deep repointing DC 02 0.039 1403
Reinforced repointing DC 03 0.063 653

The average thickness of the masonry turned out to

be about 53 cm. The material used was a roughly cut
calcareous stone (double leaf masonry wall). Of the
three panels obtained for the diagonal-compression
test, the first (DC01) was tested without strengthen-
ing, to determine its mechanical characteristics. The
remaining two were consolidated respectively with Figure 10. Typical failure modes of unreinforced masonry
deep repointing of mortar joints (DC02), and “Retico- walls: sliding shear failure (a), shear failure (b), flexural fail-
latus” technique (DC03). The panel strengthened with ure following subjecting to in-plane loading and out-of-plane
loading (c).
metal fibers shows an increase in strength τk compared
to the un-strengthened panel, of approximately 117%.
The panel reinforced with only deep repoiting of mor- t
t1 t2
tar joints did not cause an appreciable increase in terms P
of shear strength τk compared to an un-strengthened
similar panel (+34%). This shows the inefficiency of M

this strengthening method for the particular masonry

bar
texture to which it had been applied, especially due to
the high thickness of walls (Tab. 3).
The obtained value of the shear elastic modulus G1/2
was equal to 541 MPa for un-strengthened panel, while
it reached 1403 MPa in the case of strengthened panel
with deep repointing.

d
5 MODELLING AND TESTS d-x x

In order to “design” a reinforcement work of a wall ftd

panel with the proposed system, it is thought that
following vertical and horizontal loads the collapse
mechanisms may be summarised, basically, as follows:
shear, sliding-shear and flexion for actions in the plane
and out of the plane of the panel (Fig. 10).
The steel cords were arranged over an extensive
area on both outside surfaces of the panel, contributing
towards reinforcing the masonry in order to withstand
the formation of these collapsing mechanisms. This
creates a genuine reinforced wall in which the com-
pressive stress is absorbed by the stone wall and the
tensile stress by the cords.
To size /check the reinforcement elements, it is pos-
sible to use a study of the generic cross-section, apply-
ing the hypothesis of a flat cross-section guaranteed by
the monolithic character that the panel acquires follow-
ing insertion of the transverse elements, eyebolts and
threaded bars.
0.85
Figure 11. Wall subjected to out-of-plane loading.
5.1 Flexural strength
The proof with combined compressive and bending
stress, for loads both on the plane and out of the
plane of the panel, can be conducted as for any
heterogeneous cross-section.
Take a diagram of the compressive stresses of 0.85
fmd and extended to the portion of cross-section for
a depth of 60 to 80% of the distance of the neutral
axis of the compressed edge up to a maximum limit
that depends on the thickness of the compressed thin
wall and on the depth of stripping. On the average, if
the depth of stripping is 6 to 8 mm, the limit may be
estimated at 10–12 mm.

987
 barycentre of the tensite bars
P
bar
M t1
d
d-x
ftd
Figure 12. Wall subjected to in-plane loading.
With regard to flexion due to action out of the plane
of the panel, all the bars arranged on the stressed wall
contribute in the same manner to the traction effort
(Fig. 11).
In flexion on the plane, the cords in the tensile area
react differently depending on their position (Fig. 12).
Their contributions can be calculated concentrating the
area of the re-acting rods in the centre of mass of the
tensile bars.
5.2 Shear strength
For the shear check, since the formation of the resist-
ing lattice is guaranteed by the presence of the vertical
rods, the design resistance is supplied by the sum
of the shear strength of the masonry VRd,m plus the
shear strength consequent to the tensile strength of the
reinforcements VRd,t (CNR-DT 200/2004):
where:
d is the distance between the compressed edge and
the centre of mass of the tensile bars,
t is the thickness of the wall reacting to shear,
Atw is the area of the cords arranged parallel to the
shear action,
pt is the main value of the pitch of the bars,
fvd and ftd are the design shear resistance of the
masonry and the design tensile resistance of the
cords respectively,
γ Rd is a coefficient for the resistant model.
The sliding-shear mechanism becomes significant
in isolated walls where limited friction resistance is
generated due to the small axial load. In this mech-
anism, the vertical cords have an essential role in
that they prevent sliding along the horizontal mor-
tar courses of one part of the masonry in respect of
the other, when the horizontal thrust has exceeded the
friction resistance along the mortar joint.
The sliding shear strength of the reinforced panel
originates from the combination of two resisting mech-
anisms: the shear transmitted by friction from the
masonry VRd,m and the shear consequent to the tensile
t strength of the reinforcements VRd,ts .
As an alternative, to carry out a cautious check, the
t2
sliding shear strength can be assessed conservatively
using the following relationship (Tassios 1988):
x
0.85fmd
where:
Atw is the area of the cords arranged perpendicular
to the shear action, and fmd is the design compression
resistance of the masory.
6 CONCLUSIONS
The technique suggested for reinforcing masonry is
intended mainly for constructions with an irregular
(stone double or triple leaf walls) masonry texture,
such as stone walls, in order to eliminate or at least
reduce the problems of the techniques adopted for
regular walls.
The reinforcement technique consists of inserting a
continuous grid of small high strength steel cords into
the mortar joints. The nodes of the cords are secured
by means of metal rods transverse to the wall facing.
The result is that of a reinforced masonry, for which
there is an increase in compressive, shear and flex-
ural strength, and an effective transverse connection
between the leaves of the masonry.
The improvement does not concern solely the
mechanical characteristics of the masonry thus treated,
but affects the whole masonry construction, since in
addition to reinforcing the wall panel, the “skeleton”
of the continuous grid inside the masonry connects
the various contiguous masonry walls to one another,
thus forming a genuine complete reinforcement sys-
tem. Furthermore, the small size of the reinforcement
cords and the fact that they are easy to insert into the
mortar joints makes it possible to apply this treatment
on a widespread basis, that avoids dangerous concen-
trations of stress such as those that occur, for example,
when using metal rods.
The suggested system can be used both at local
level, that is to say for single wall panels of exist-
ing buildings or also for boundary town walls, and at
global level, that is to say as a system for reinforcing a
masonry construction, improving its overall behaviour.
The upgrading work is not very invasive, is
reversible and integrates the masonry rather than
replacing it. It is compatible with preservation of
the original material of the building and long-lasting
thanks to use of very durable materials and is therefore
988
particularly suitable for fair-face walls of buildings
listed due to their historical or architectural value.
The effectiveness of the proposed technique was
investigated by means of a series of tests with double
flat jacks, subjecting the masonry to vertical com-
pression and diagonal compression tests (shear tests).
Based on the results obtained, it was possible to note
an improvement of the mechanical behaviour of the
masonry and it was noted that deep repointing of joints
with metal fibres is capable of increasing compressive
strength significantly, even doubling it as compared
with the non-reinforced masonry.
REFERENCES
ASTM E 519-81. 1981. Standard Test Method for Diago-
nal Tension (Shear) in Masonry Assemblages, American
Society for Testing Materials.
Binda, L., Modena, C., Baronio, G. & Abbaneo, S. 1997.
Repair and investigation techniques for stone masonry
walls, Construction and Building Materials, Elsevier, 11,
(3), 133–142.
Cecchi, A., Milani, G. & Tralli, A. 2004. In-plane loaded
CFRP reinforced masonry walls: Mechanical characteris-
tics by homogenisation procedures, Composites Science
and Technology, (64), 13–14, 2097–2112.
Chiostrini, S. & Vignoli, A. 1994. In-situ determination of
the strength properties of masonry walls by destructive
shear and compression tests, Masonry International, 7,
(1) 87–96.
Corradi, M., Borri, A. & Vignoli, A. Strengthening tech-
niques tested on masonry structures struck by the
Umbrian-Marche earthquake of 1997–1998, Construction
989
and Building Materials, Elsevier, 16, (4), 229–239,
2002.
Corradi, M., Tedeschi, C., Binda, L. & Borri, A. Experimental
evaluation of shear and compression strength of masonry
wall before and after reinforcement: deep repointing,
Construction and Building Materials, Elsevier (in press).
Dolce, M. & Marnetto R. CAM technology, patent.
Elgawady, M.A., Lestuzzi, P. & Badoux, M. Analytical model
for the in-plane shear behavior of URM walls retrofitted
with FRP, Composites Science and Technology, Elsevier,
(66), 459–474, 2006.
Eshani, M.R. 1997. Strengthening of earthquake dam-
aged masonry structures with composite materials, Non-
metallic (FRP) Reinforcement for Concrete Structures,
Proceedings of the second international RILEM Sympo-
sium FRPRCS-2, L. Taerwe Ed., E & FN Spon, London,
England, 681–687.
Eurocode 6. 1995. Design of masonry structures – Part 1-
1: General rules for buildings – Rules for reinforced and
unreinforced masonry. ENV 1996-1-1.
Gabor, A., Bennati, A., Jacquelin, E. & Lebon, F. 2006.
Modelling approaches of the in-plane shear behaviour
of unreinforced and FRP strengthened masonry panels,
Composite Structures, Elsevier, (74), 277–288.
Tassios, P. T. 1988. Masonry mechanics , Liguori Publ.,
Napoli, (Italian Translation).
Tomazevic, M. & Alih, V. 1993. The strengthening of
stone-masonry walls by injecting the masonry-friendly
grouts. 7th European Conf. On Earthquake Engineering,
1, 10–20.
Turnsek, V. & Sheppard, P. F. 1980. The shear and flexural
resistance of masonry walls, Proc. of the research conf.
on earthquake engineering., Skopje.
Vintzileou, E. & Tassios, T.P. 1995. Three-leaf stone masonry
strengthened by injecting cement grouts, Journal of Struc-
tural Engineering, 121, (5), 848–856.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Innovative and sustainable local material in traditional African

architecture – Socio cultural dimension

T.O. Odeyale & T.O. Adekunle

Department of Architecture, Federal University of Technology, Akure, Nigeria

ABSTRACT: Local materials are the resources that can be found readily in large quantity at a particular location
or area at a certain time. It could also be referred to as materials that can be used to fabricate a finished element.
These materials however could be abundant in some area but not available in another. The availability may largely
be dependant on geographical location of the area as well as the chemical and physical components of such
materials. The paper critically appraises some local material available in south west Nigeria. The investigation
revealed such materials as laterite, textile, bamboo, mosaic, mats stones, dye, timber, tusks, snail shell, cow dung,
cowries cane and mud. These materials are cheap relative to the imported materials from outside the country.
The neglect of these readily available materials should be discouraged. The paper also suggests practical and
innovative ways for designers, architects and manufacturers which can serve as an alternative source of material
and yet sustainable; and in the long run profitable for all concern. Proper inventory, investments, packaging and
modernization can help generate much needed foreign exchange and serve as promotion of the local culture.
Close attention must be paid to the sustainable methods and means of using these materials for the good of all.
Diagrams, pictures, plates and their application of these materials are also showcased in the paper.

Keywords: Local materials, bamboo, architecture, culture, sustainability, economy, Africa.

1 INTRODUCTION in many areas limited to coastal regions, although

The indigenous technology using locally available
building materials of our African ancestors deserve to
be preserved. WestAfrica today suffers from the effects
of colonization and African Americans are in a unique
position to help ameliorate the situation. Architecture
is a field in which remarkable achievements have been
made in Africa using the indigenous technology and
materials. Examples from history abounds of the use
of local materials to achieve outstanding architectural
monuments such as
i. The ziggurat (mud and stones) 2640–2621 BC
ii. Bight of Benin (stabilized mud brick and plaster,
thatch, timber, shingles).
iii. Kofa Buka-The Kano wall (mud and vegetable mat)
iv. Centenary Hall, Ake, Abeokuta (stone, mud,
timber).
v. Tomb, Places in Egypt – The Zoster’s stepped
pyramid at Saggara.
Drawing considerations from the history of African
art and architecture where her building technology
emanated from is made problematic by the fragmen-
tary state of the evidence. Until the mid-19th century,
most European contact with sub-Saharan Africa was
the accounts of the kingdoms of Benin and Congo
provided by 16th century and 17th century traders
and missionaries from Portugal are useful exceptions.
Arab scholars are also offers some valuable infor-
mation, particularly concerning the medieval African
empires of Ghana, Mali and Songhai, but also with
regard to the East African coast. Some early for-
eign adventurers to the continent saw things his-
torical in a different light and their account vary
considerable.
There is still a persistent misconception that pre-
colonial Africans lived almost exclusively in circular
mud and thatch dwellings. Some publication since mid
19th century have associated most traditional African
with round, thatched roof, mud wall hut. However, his-
torical research indicates that square and rectangular
forms are of great antiquity and have existed in diverse
geographical regions.
Undoubtedly, some of the structures were influ-
enced by external architectural traditions. In order to
understand the aspect of African technology on build-
ing it is important to pay attention on the history of
kings and queens, princes and emperors, popes and
prelates, the story of battles, conflicts and conquests,
political maneuvers and dynastic domination.

991

Figure 1. Indigenous building of Songhai in Mali.
2 BACKGROUND ON AFRICAN INDIGENOUS
BUILDING MATERIALS
The morpho-genesis of human habitat started inside
the cave. Like the Art, the Bee and the Beaver. Man is
not contempt with God’s earth and he feels oblique
to build a better one, more suitable to his exalted
majesty. The Art build hills, the Bee makes Hives;
the Beaver constructs dams; man creates architecture.
Architecture proposes an effect on the human mind,
not merely a service to the human fame. Thus architec-
ture is the art which so disposes and adores the edifice
raised by man, for whatsoever uses, that the sight of
them may contribute to his mental health, power and
pleasures(Aluko, 1997).
As a result of this, African people look inward and
discovered those materials which were used in building
construction, and they are: laterite, sandcrete, stabi-
lized earth, burnt clay, clay brick, bamboo, raffia palm,
leaves, timber, palm-kernel shell, crash coconut shell,
animal waste, dung and others.
2.1 Early man craft
The early man craft was borne out the need to have
what identify man with his culture, belief, tradition and
environment using the materials available within his
reach. Due to this, man use materials like stone, wood
(timber), mud in creating what has a message based
on experience, inspiration, focus, norms to showcase
his talent and celebrates his creativity which formed
the craft (art) and architecture.
2.2 African traditional architecture
African traditional architecture showcases each region
of the continent utilizing various materials available
within their environment to build houses. In Tamberma
992
Figure 2. Indigenous building of northern Nigerian built
with local materials.
(Batammaliba) region of Togo and Benin, traditional
earthen round houses. Traditional architecture was
more manifested in the construction of roofs. Mud con-
struction dominated the whole of the western Sudan
and part of east Africa. Most of them were infact not
built of mud entirely but were reinforced with timber.
The range of roofs found in various parts of Africa
falls under different categories. The technology of ter-
race roofs was very common in Mauritania and upper
Niger region. These were, achieved by laying mating
or short poles arranged in herringbone pattern on top
of the roof beams and plastering the hut over with mud.
In case of the Bamileke, the house has conical
roof resting on square walls. The conical roof was
construction round two circular platforms built on a
pyramid structure which rested on top of the house
walls.Western African region, house with thatched
roofs above walls had great variations in shape, materi-
als used and construction techniques while some roofs
rested on pillars were independent of the walls below
them. Saddle back root nearly always rested on wall
plates supported on forked uprights within the walls.
The ridge pole was sometimes supported by upright
poles but usually rested between the gable ends. The
beams were occasionally used. The shape of the roof
was usually adhering in relation to the shape of the
walls. Roofs above round walls would be conical, and
roofs above rectangular walls could be saddle back
heaped on pyramid.
2.3 Rural settlement
Traditional rural dwellings are of different types; rel-
atively permanent houses grouped in villages are only
visible in agricultural societies. A typical farming vil-
lage consists of a number of family compounds along
with structures that serve the larger community. Each
family setting normally have separate buildings for
various purposes like cooking, eating, sleeping, stor-
ing food, and protecting animals from adverse effect
and external aggression at night. Based on their cre-
ativity and indigenous technological know-how using
Figure 3. Zulu beehive house built of rope lattice, thatch
and wooden strips.
Figure 4. Eastern African nomadic shelter of stick frame-
work covered with hatch.
available materials, resources and manpower within
their reach, buildings may be round, rectangular, or
semicircular. Communal structures, or holding meet-
ings and teaching children are located in a prominent
in the village.
Using available technology, geographical location,
materials availability, culture and economic, people of
Dogon in southern Mali cultivate grain on a plateau at
the top of the Bandiagra cliffs near the Niger River.
They built their villages on the steep sides of the cliffs.
Their houses are rectangular in shape built of sun-dried
mud brick and stone. The roofs are thatched, and the
dwellings rest on hedges along the cliffs. The Zulu’s in
Southern Africa, who engage in grain cultivation and
raise livestock, use their indigenous technology built
houses shaped like beehives. They built their houses
in a circular, fenced compound, and they keep their in
the middle of the compound.
Zulu buildings roof are made of thatch that covers a
framework of wooden strips and is bound together with
a rope lattice. Nomadic herders also need homes they
can easily build despite the fact that they move often
from one place to another. They build houses which
can easily actualize when their herds move to different
ground. For, example, Masai found at eastern Africa,
builds home using a framework of sticks sealed with
cattle dung.
Majority of the rural societies in Africa decorate
their buildings exterior with painted designs or with
993
Figure 5. Ndele women at work doing decorations work and
painting of various geometrical shapes in their house. Source:
www.google.com
relief (raised) patterns worked into a soft clay sur-
face. The decoration of houses is mainly to the women.
Frafra women of northern Ghana use geometric pat-
terns to decorate with patterns walls of their houses and
other buildings that communicate information about
the social status of a building’s owner. Ndebele women
of Zimbabwe and the northeastern part of South Africa
decorate and paint their houses mud walls with various
geometric patterns used on the shapes of windows, step
and other building features and everyday objects. Tra-
ditionally, Africans indigenous technology has used
natural clays as paints, but today brightly coloured
acrylic paints are popular (Jolaoso, 2001).
3 SOCIO-CULTURAL FACTORS IMPACTING
ON INNOVATIVE MATERIALS
Socio-cultural factors are primary forms, which cre-
ate ideal environment that are culturally comfortable,
spiritually meaningful and socially acceptable.
House – forms viewed from these perspectives are
usually influenced by family pattern (extended fam-
ily), lineage system (ancestral line of family), and
kinship (relation by blood).
Nigerian Traditional Architecture is an inherited
art and science from the past generation on the areas
of planning and designing with consideration to the
provision of spaces for the individual and public
interest and the construction of buildings with the
available local building materials. The development
of indigenous architecture could be viewed as a mat-
ter of circumstances because it was influenced by the
regional element of environment.
The local available building materials and building
technology of Nigerian traditional Architecture are not
durable enough to meet the present demand of 20th
century architectural demands in terms of resistance to
leading, spans of architectural spaces and many other
factors. Therefore the need to improve them to meet the
present demand is one of the achievements of study-
ing traditional architecture and indigenous technology.
Figure 6. Locally produced sun-dried brick by Timbuktu
people.
Figure 7. Palace in northern Nigeria reflecting some socio
cultural background.
Nigerian Building Research Institute (NBRI) hand is
still studying and developing many of our local build-
ing materials to meet the present challenges. Much had
been done on clay and clay products and the clay indus-
tries all over the country and many other materials
which are true witness of these efforts and achieve-
ment of Nigerian Traditional Architecture. The earth
architecture of late hand been received as a witness
to the same achievement of traditional architecture
(Madedor, & Asaolu, 1989; NBRRI , 1997).
3.1 Various building materials readily available
for housing in Nigeria
Building materials and their usage in the pre-colonial
era to achieve the monumental architectural spaces of
Nigerian traditional Architecture is the object of study
in two content. Generally, they (building materials)
were based on the natural available recourses which of
course vary from one locality to locality as the nature
had made them readily available. Therefore, available
geographical conditions especially vegetation played
a vital role in availability of these building materials
(Bello, 2001, Arayela, 1996; Madedor, et al. 1985).
These naturally occurring materials include: store,
clay, mud, timber, bamboo, thatches, spare, grasses,
palm frond, raffia palm, large, leaves, bark of trees,
local fibre of bush palm, animal waste (dung), palm
994
Figure 8. Common materials used in the construction
of African Traditional Buildings (bamboo and organic
materials).
kernel shell, broken pots and others. The uses of
these materials were based on the following factors:
availability, crude (yet real) technological know-how,
financial constraint, crude tool and equipment, trans-
portation system, the ruling class choice and the
climatic effect on such materials. However, much been
achieved in the use of these materials. These materials
are illustrated in the figure below.
3.2 Innovative technology involved
3.2.1 Timber
Timber was and is readily available and plentiful.
More so, bamboo and other forest resources like ropes,
forked wood for buildings were in abundant, therefore,
the cost of forest building materials resources were
relatively cheaper. Saving was locally done, seasoning
was done with the help of the some and craft on that
could easily handle them were readily available and
accessible (Dmochowski, 1990).
3.2.2 Stone
The use of stone as a walling material was unpopular
in the Nigerian Traditional Architecture in the pre-
colonial era due to the fact that instrument used for
cutting and the bonding materials for stone were not
readily available.
3.2.3 Clay and mud
These were the chief and commonest walling materials
of traditional architecture. The indigenous technology
on clay and mud wall, floor and roof construction
developed so much that various methods and forms
were used to produce walls, roofs and floor from these
two building materials. The technology used included
– Direct usage of mud and clay placing to conve-
nience wall thickness.
– Addition of straw (cob) to clay and mud in the pro-
duction of moulded small unit bricks and block used
for wall construction.
Figure 9. Method of wall construction with mud.
Figure 10. Method of wall construction, reinforcement and
roofing.
– Tabuli: This is a production got from broken solid
mud mixed with chalk, stone and gravel placed in
timber framework with wet mix clay and mud to
form the binding member, well rammed to form
strong consolidated wall between frameworks used
to form various shape of wall.
– Grasses like “zanna” wovened into mats and used as
cover for reinforcement got from bamboo or similar
poles to produce structural mud and clay wall.
– Sawn bamboo, raffia wovened in such a way
to form shape of required wall, whole mud and
clay were placed between to form structural wall
construction.
– In order to prevent shrinkage and reduced cracks in
mud clay walls; grasses, coconut fibres, coarse sand
were added to the mud mixed before wall placing.
– The usage of long and projected roof eaves and
spouts for mud and clay construction prevented rain
falling directly on mud wall.
– Mud and clay walls floors, and roofs constructions
were rendered with materials like indigo leave,
cattle waste, coloured mud to prevent water pen-
etration into the surface wall. The illustrations of
the process are shown above.
3.2.4 Grasses
Grasses were found in the Middle Belt and Northern
region abundantly for the purpose of construction in
995
Figure 11. Local building at riverine area built with
wild coconut timber framework and raffia palm leave roof
covering.
Nigerian Traditional Architecture. The type of grasses
available in a locality depends on the climatic region,
vegetation and closeness to water source. Toward the
South of the Middle Belt region, elephant grasses were
common; even in the Northern part of South Western,
Mid – Western and Eastern region; while in the Mid-
dle Belt and Northern region, spare grasses were very
common (Jolaoso, 2001; Bourne, 1981, Denyer, 1978,
Encarta,2005).
3.2.5 Wild coconut tree
This fibrous tree is found mainly in the savannah region
(Middle Belt) of Nigeria. It produced good structural
member for root, wall lintel, ceiling and bridges con-
struction. It is also used as overlay for toilet pits.
Unlike timber, it is a termite resistance structural and
don’t easily decay unless subjected to long period of
dampness (Dmochowski, 1990).
3.2.6 Bamboo
This pole – like structure was mostly used in con-
struction structural members of wall, roofs, floors and
ceiling so also in fence construction which was com-
monly found in South west part of Nigeria (Bassey,
1989).
3.2.7 Leaves and back of trees
In the rainforest, some deciduous trees have broad
and large leaves (like Gbodogi leaves and “Ebi–eni”)
and bark that could be used as roof cover. Also, the
raffia palm leave that found in the most river banks
were made into thatch roof cover called “Bambu”. The
barks of such tree were used as ridges to root while
leaves and thatches were used as roof cover. Some
leaves could be as well be used for wall finishes e.g.
Indigo (Dmochoswki, 1987; Jolaoso, 1991).
3.2.8 Animal waste
The animal waste like dungs and bones were used as
finishes to wall in the Nigerian traditional architecture.
Cowries were used for decorations of walls and floors.
Though, building technology of Nigerian Traditional
Table 1. Types of Building examined during study survey. Table 2. Bar Chart showing the determining factors of inno-
vative building material in some selected cities of Nigeria.
30
30
25 25

20 20

15
15
10
10
5
5
0
cost avail taste durability tech-know climate
0
resd inst resort religion hotel other

The chart reveals that many of the buildings examined in the

study are mostly residential, followed by religious building Table 3a. Mostly used indigenous building materials for
and lastly resort and institutional buildings. walls.

Percentage Cummulative
Labels Frequency % percentage %
Architecture suffered a little setback because of “the
relative fragility of building materials and the simplic- Mud /laterite 48 60 60.0
ity of the working tools”. But it is worth of note that Timber 7 8.75 68.75
Nigerian Traditional Architecture left behind “tech- Stone 20 25 93.75
nological perfection” which is a result of a thorough Wattle 3 3.75 97.5
Bamboo 2 2.5 100.0
understanding the qualities of the materials used, and
of skill developed and improved by many generations
of builders great and small (Fadamiro, et al.1996).
Table 3b. Bar Chart interpretation of mostly used indige-
nous building materials for walls.
4 RESEARCH FINDINGS, DATA COLLECTION
AND DISCUSSIONS 60

The target of this data collection is to come out with
reasonable and reliable findings which will be of great
help as regards the technological influence on Nigerian
indigenous (traditional) building materials in terms of
innovation, availability, cost- effective, climatic suit-
ability, mass production, improved products, social
and legal acceptability, technical know-how, durabil-
ity as well as the occupation and economic activities,
craft activities, religious beliefs and practice, security,
compound and housing unit.
The research methodology adopted was by study-
ing various modern/ traditional buildings built with
local materials found in some selected towns and cities
in Nigeria and the effect of innovative technology on
these materials and houseform using questionnaire.
4.1 Data Presentation, sample size and
techniques
A sum of 90 questionnaires were prepared and dis-
tributed for administration at various locations in some
selected towns and cities. The buildings focus are res-
idential buildings, traditional buildings, public and
institutional buildings. Questionnaire of close ended
type was used, all questionnaires were distributed and
administered and 80 retrieved after proper follow-up
50
40
30
20
10
0
laterite timber stone wattle bamboo
from the respondents but 10 were not retrieved. There-
fore, this data analysis will make use of 80 question-
naires retrieved as 100%. Random sampling technique
was adopted for this project work and residents in
different locations were interviewed to uncover their
opinion about innovative technological influence on
indigenous building materials in their modern housing
design.
Table 2 corroborates the known fact that cost is a
deciding factor in the adoption of new technology and
use of innovative materials. Several of the respondents
believe that innovative material should be made avail-
able at affordable price at the open market. Table 3 a &
b suggested that laterite/ mub block accounted for
the 60% of the material used in indigenous buildings.
Bamboo is the least used.

996
Table 4. Socio-cultural factors impacting on the adoption
of innovative local material.
Percentage Cummulative
Labels Frequency % percentage %
Political/social 8 10 10
Status
Religion 10 12.5 22.5
Occupation 11 13.75 36.25
Leisure activity 31 38.75 75
Family 20 25 100
Organisation
5 CONCLUSIONS
Also, there are several existing evidences that indige-
nous materials such as mud, bamboo, stone, grasses,
leaves, reeds, cow dungs among others were used for
construction of various buildings in Nigeria. As these
materials are readily available in abundance, the usage
has been on decline in some time past due to flair for
imported building materials which are produced with
disregarded for the climatic suitability of Nigeria and
high cost implication.
The fact that most people believe that living in
brick, timber or stone houses symbolize abject poverty
which has created hatred for these materials. Also the
quest for modernization without considering our back-
ground did not help us to look inward and develop our
indigenous building materials rather than searching for
foreign materials.
In view of this social impression that causes the
decline in the utilization and with cognizance to these
materials’ numerous advantages, cost benefits, this
research work provide vivid information on avail-
able sources of building materials, their classification,
properties, advantages and their uses as well as the pro-
duction of some building components using identified
local materials.
5.1 Suggested recommendations
The encouragement of the use of local building mate-
rials produced modern technology should be a col-
lective responsibility of the government, individual
and private organization, for instance, the following
recommendations will be of great help:
(i) That the Government should ban or place restric-
tion on some imported building materials. This
will allow the growth of local building industries
and encourage people in using locally produced
building materials.
(ii) That the Government should encourage the local
building materials industries by creating con-
ducive atmosphere for operation right form the
registration process to the grant as well as granting
997
of holiday tax in some cases when it is highly
needed.
(iii) That government should encourage the exporta-
tion of the locally produced building materials
to find places in the world market and to gener-
ate enough capital to enhance the development
of home based manufacturers.
(iv) That government and private bodies/ organiza-
tions should be encouraged to establish indus-
tries that produce local building materials.
(v) Some planning authorities who hitherto restrict
people from using local building materials in
the cities should be made to relax their strin-
gent condition thereby encouraging developer
to commence their development with the use of
local building materials.
(vi) That the Government (at all levels) should
always ensure that all related research findings
are implemented and not for record purposes
alone. This may be achieved by establishing
a well funded, research findings’ implementa-
tion commission, which would be responsible
for demonstrating such findings in the form of
mass housing (construction) units.Thus, serving
as an architectural/structural model that would
help in no small way in encouraging and pro-
moting; the acceptance of the use of locally
sourced materials with the corresponding and
appropriate indigenous technology.
(vii) That public enlightenment campaign or dissem-
ination of information should be encouraged
on the developed local building materials and
their applications to the masses, as better sub-
stitutes to the expensive and relatively scarce
conventional materials, possibly through the
existing rural development programme of the
Government.
(viii) That the curriculum of technological, architec-
ture, engineering institutions should be reviewed
to reflect the study of the application of locally
sourced building materials.
(ix) That the Government, Architects, Engineers and
other relate professionals should lay examples
by building their respective houses with these
materials as a substitute to the conventional
ones.
REFERENCES
ALUKO, O.O. 1997. Nigerian Traditional Architecture,
A Case Study of South-west Nigeria. A Lecture Notebook,
Department of Architecture, Rufus Giwa Polytechnic,
Ondo State (Unpublished). pp. 1–10, 20–25.
ARAYELA, O. 1996. The Politics of Housing the Masses
in Nigeria – A Retrospective Introspection. A Journal
of the Association of Architectural Educators in Nigeria
(AARCHE). Vol. 1, No 3. pp. 29–33.
BASSEY, N. 1989. What are Local Building Materials?
A Paper presented at the Mineral Seminar of the Nigerian
Institute of Architects, Edo State Chapter.
BELLO, Y.O. 2001. Elements of Creativity in Traditional
Architecture; A Descriptive Study of Theory of Architec-
ture. A Journal of AARCHE (Association of Architectural
Educators in Nigeria). Vol. 1, No 6. pp. 95–13.
BOURNE, L.S. 1981. The Geography of Housing. Edward
Arnold (Publishers) Limited, London. pp. 13–21.
DENYER, S. 1978. African Traditional Architecture. Heine-
mann Press, London. pp. 155–171.
DMOCHOSWKI, Z.R. 1987. An Introduction of Nigerian
Traditional Architecture, Southwest and Central Nigeria.
Vol. II. Ethnolographical in Association with National
Commission of Museums and Monument. pp. 20–70,
110–288.
DMOCHOWSKI, Z.R. 1990. An Introduction of Nigeria
Traditional Architecture, South- Eastern Nigeria, the Ibo
speaking Area. Vol. III. Ethnographical in Association
with National Commission of Museums and Monument.
pp. 1–36.
FADAMIRO, J.A., & OGUNSEMI, D.R. 1996. Fundamental
of Building Design, Construction and Materials. Fancy
Publications Limited, Ile-Ife, Nigeria. p. 5.
998
JOLAOSO, B.A. 1991. A Study of the Use of Local Build-
ing Materials for Low Cost Housing in Western Nigeria.
Unpublished, B.Sc Terminal Essay. Ahmadu Bello Uni-
versity, Zaria.
JOLAOSO, B.A. 2001. Housing and Indigenous Building
Technology; An Introduction. Desi-GCA Publications,
Abeokuta, Nigeria. pp. 116, 123–131.
MADEDOR, A.O. & ASAOLU, E.G. 1989. Technology and
Standards for the Production of Local Building Materials
in Nigeria. NBRRI Information Pamphlet. pp. 12–15.
MADEDOR, A.O. & OMOGE, G.N. 1985. Survey of Brick-
making Industries in Nigeria. A proceed from a Semi-
nar on Clay Bricks and Blocks: Research Development
and Current Practice in Nigeria; NBRRI Information
Pamphlet. pp. 7–22.
Microsoft Encarta Encyclopedia. 2005. Information on
African Housing and Nigeria.
NBRRI 1997. Spotlight: The Nigerian Building and Research
Institute, Lagos. In Housing Today. July/ August, Vol. I,
No. 1, pp. 10–17.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Imported construction materials and techniques in 19th century

Ottoman architecture

Uzay Yergün & Aynur Çiftçi

Department of Architectural Restoration, Faculty of Architecture, Yıldız Technical University, Istanbul, Turkey

ABSTRACT: European societies created the model for the modernization period that began in the Ottoman
state in the eighteenth century and continued during the so-called “westernization” period of the nineteenth
century. The construction of new types of buildings such as military buildings, western style palaces turned
into a reform movement which took western architecture as its model. Due to its conservative structure, the
Ottoman State could not completely subscribe to the intellectual movements in the west so it remained tied to the
industrial goods that Europe manufactured. For this, after the first quarter of the nineteenth century solid bricks,
tiles, metal beams and similar building materials were imported from countries like Britain, France, Belgium,
Italy, Germany and Austria. Many buildings of the period that remain have been registered as essential cultural
assets have faced various problems. The interventions undertaken have damaged construction techniques and
the originality of the imported materials.

1 INTRODUCTION public purposes were built with western techniques.

With the repercussions of the French Revolution, Euro-
pean societies developed in a new direction from the
end of the seventeenth century. By the mid-eighteenth
century, a traditional economy based on agriculture
and handicrafts developed into a modern manufactur-
ing economy and there began the “Industrial Age” in
Europe. The Ottoman State’s “Westernization” period
beginning in the eighteenth century and continu-
ing into the nineteenth century was influenced by
developments taking place in Europe.
In an attempt to catch up with the modernizing
changes taking place in Europe, The Ottoman State
first established close relations with European States
to reorganize its military corps, then the lifestyle in the
palace and its immediate circle was influenced which
consequently led to conceptual changes in architec-
ture. With the proclamation of the Gülhane Edict in
1839, the Ottoman State officially inaugurated the
Tanzimat or “Re-ordering” era, and it is the turn-
ing point for modernization and architecture in the
Ottoman State.
The Gülhane Edict sped up the reform process and it
was felt that İstanbul needed to change its image from
“Oriental” city to “Western” capital. While traditional
techniques were used for the predominantly wooden
dwellings and solid mason monumental buildings,
structures for military, administrative, official and
The use of solid masonry blocks and timber frame-
work in traditional Ottoman architecture did not suit
the principles of western architectural design, there-
fore modern building materials and techniques were
used for these new structures.
With the Gülhane Edict and the establishment of the
Tanzimat, the years between 1840–1860 and 1860–
1876, the steps taken toward modernization may be
characterized by a “process of industrial strength-
ening” taking place in the Ottoman State. However
for both internal and external reasons, the newly
established factories and institutions were not produc-
tive enough and the Ottoman State continued to be
dependent on European technical aid and manufac-
tured goods for building (Önsoy, 1988). In order to
acquire the Ottoman market for the sale of manufac-
tured goods, European countries endeavored to estab-
lish treaties of alliance and trade with the Ottoman
State with the result that from the beginning of the
nineteenth century countries like England, France,
Belgium, Italy, Germany and Austria were exporting a
variety of industrially produced tools besides machine-
made bricks, metal supports made with advanced
metallurgical techniques and clay tiles etc. . . for build-
ing and interior decoration to the Ottoman State.
The names and addresses of local producers and the
companies and offices importing building materials
for Ottoman use are published in the Ottoman trade

999

Figure 1. Traditional Ottoman brick (24 × 24 × 3.5 cm).
annuals of the nineteenth century such as “Apel Gül-
benkian”, “P. Gülbenkian & Cie”, “S. Capou & J.
Pelletan”.
2 IMPORTED CONSTRUCTION MATERIALS
2.1 Imported bricks
Bricks used in Ottoman architecture before the indus-
trial age fell into three groups: “tuğla-i carşu” (full-
sized bricks), “tuğla-i miri carşu” (half-sized bricks)
and “tuğla-i harci carşu” (outsized bricks) denot-
ing the three different sizes in which bricks were
produced (Arseven, 1965). The dimensions of full-
sized bricks were 24 × 24, 28 × 28, 30 × 30 cm in
length and 3, 3.5, 4 and 4.5 cm thick (Fig. 1). Since
the Byzantine period, the most important produc-
tion sites were located on the shores of Büyükdere
and along the Golden Horn in Piripaşa and Kırkağaç
(İnciciyan, 1976).
With the technological advancements taking place
in nineteenth century Europe, brick production
develops analogously as its physical attributes are
highly resistant and all of the bricks industrially
produced were of standardized compactness and den-
sity. Industrially produced modern bricks measur-
ing 6 − 8 × 10 − 12 × 21 − 23 cm were the preferred
building material because they were inexpensive, prac-
tical to use and fire resistant compared to wood and
could support solid wall structures.
The Russian Embassy was the first building to be
constructed in İstanbul with modern bricks in 1838 and
it was highly appreciated by the reformers of the Tanzi-
mat era. The Swiss architect G.T. Fossati was hired for
the project and after the completion of the Russian
Embassy, in 1841, he was entrusted with the construc-
tion of the Military Hospital at the Imperial Gate and
this was the first Ottoman structure made with modern
bricks completed in 1843. The Balyan family had been
the palace architects until that time and it is significant
1000
Figure 2. Imported brick from Marseille, France
(6.5 × 10.5 × 21.5 cm).
that a European architect was hired for an imperial
project (Can, 1993). Many European architects besides
G.T. Fossati, such as W.J. Smith, G. Fossati, M.A.A.
Bourgeois, G.B. Barborini and others as well as Euro-
pean educated architects of the Balyan family were
entrusted with building projects (Yergün, 2002).
These architects were instrumental for the develop-
ment of imported bricks and modern building tech-
niques in the Ottoman State during the 1840s as
Ottoman traditional building techniques were replaced
by European techniques and materials. The tran-
sition from wooden buildings to modern masonry
techniques was achieved in secular architecture with
modern bricks through legal settlements and various
incentives.
Research has shown that the bricks used in nine-
teenth century buildings in İstanbul generally measur-
ing 6 − 7 × 10 − 12 × 22 − 24 cm dimensions came
from Saint-Henri and Saint André, two villages
near Marseille, France. In the mid-nineteenth cen-
tury these two villages were internationally known
as brick production centers. Bricks stamped with
“GUICHARD FRERES St HENRY MARSEILLE”
(6.5 × 10.5 × 21.5 cm) are proof of this (Fig. 2). It is
well documented that the Çırağan Palace (1863) was
built with bricks imported from Marseille measuring
6.5 × 11 × 23 cm.
The Darülfünun (İstanbul University) dated 1845–
1854 and Princess Adile’s Palace dated 1876 used
bricks measuring 5 − 6 × 14 − 15 × 30 − 32 cm and
documentation shows that they came from Livorno,
Italy (Fig. 3). Certainly, transporting bricks from Mar-
seille and Livorno to İstanbul by ship was also easy
and fast.
The “FRATELLI ALLATINI SALONICCO” com-
pany was another important producer of bricks for
buildings in İstanbul was located in Salonica (now
Greece) (Fig. 4). The Allatini family was of Jewish
origin and had emigrated from Livorno to Salonica
in 1802 and set up the company in 1836. It was this
Figure 3. Imported brick from Livorno, Italy
(5,5 × 15 × 30 cm).
Figure 4. Imported brick from Salonico, Greece (Sandalcı,
1997).
same company that owned “Roller mill” and “pottery”
and later became the “Ottoman Industrial and Com-
mercial Company of Thessaloniki S.A.” in 1897. This
company also was trading in other realms and in 1926
was transferred to Greek businessmen. In 1964, the
“Allatini S.A.” and “Ceramics Allatini S.A.” compa-
nies separated from the parent company and continues
to produce ceramics today.
In the last quarter of the nineteenth century, the
Ottoman state began producing bricks and tiles indus-
trially. Karaağaç in 1874 and Büyükdere, Feriköy and
Hasköy in 1876 became production centers in İstanbul
while outside of İstanbul in the towns of Mürefte and
Eceabat brick and tile factories were set up in this
period (Mori, 1906; Sandalcı, 1997; Can & Girardelli
1996).
2.2 Imported tiles
Studies on some of the Ottoman buildings using locally
“alaturka” produced tiles show that these are of differ-
ent dimensions and shapes than those tiles imported
from Marseille, Salonica and Livorno. The producer’s
name and sometimes the date and images are stamped
onto the tiles.
The roof tiles covering the mid-nineteenth cen-
tury Taşkışla building are stamped with “ARNAUD
ETIENNE Cıe 1890 MARSEILLE St HENRI”,
“SACCOMAN FRERES St HENRI MARSEILLE”
1001
Figure 5. Imported tile from Marseille, France.
Figure 6. Imported tile from Marseille, France.
(stamped with the image of a light house),
“GUICHARD CARVIN & Cıe SEON St ANDRE
MARSEILLE” (stamped with bee image that is still
used by the Lafarge company) and also tiles pro-
duced by “GUICHARD & PIERRE FRERES SEON
St HENRY MARSEILLE” as well as “FRATELLI
ALLATINI SALONICCO”. A total of five differ-
ent companies’ tiles were used for this single struc-
ture. Dated to 1900–1901 the Topkapı Palace Police
Station’s roof tiles were imported from two com-
panies: “GRANDE ECAILLE POUR TOITURE,
BREVETES S.G.D.G. St. HENRY-MARSEILLE,
ROUX-FRERES” (stamped with a heart image) and
“FRATELLI ALLATINI SALONICCO” (Figs 5–8).
The Marble Mansion (1860) used tiles from the
“GUICHARD FRERES SEON St HENRY MAR-
SEILLE” company measuring 42 × 25 × 4.7 cm and
a stamped glass tile was found that belonged to a
building dating to the end of the nineteenth cen-
tury produced by the “SOCIETE GENERALE DES
TUILERIES DE MARSEILLE ET Cıe MARSEILLE
FRANCE” measuring 42 × 25 × 4.7 cm. In the Cibali
Tobacco Factory also a stamped glass tile was found
(Figs 9–10). In İstanbul, the Büyükdere factory pro-
duction costs were high and the quality of the tiles was

Figure 7. Imported tile from Marseille, France.
Figure 8. Imported tile from Salonico, Greece.
Figure 9. Imported glass tile from Marseille, France.
poor. For these reasons tiles and bricks from Livorno
were considered better as well (Önsoy, 1988).
2.3 Imported metal supports
Iron was wrought from the fourteenth century on and
by the mid-eighteenth century cast iron technique was
used for shaping metal but after the Industrial Revolu-
tion, steel replaced iron (Ökten, 1995). In Ottoman
architecture, iron was not used on its own to bear
weight but rather as a supporting component (Tanyeli,
1002
Figure 10. Imported glass tile.
1990). From the last quarter of the nineteenth century
on, imported metal supports of varying shapes and
sizes were used in buildings of many floors having
rooms of large dimensions in İstanbul. Examples of
these found in İstanbul were imported primarily from
England and Belgium after trade agreements were
made between these countries and the Ottoman State
in 1838.
The English engineer Sir William Fairbairn (1789–
1874) came to İstanbul in 1839 and he made various
orders for the Unkapanı bakery and flour mills. In
1841, Fairbairn made a three-story building for milling
corn entirely of cast iron and wrought iron including
the supports, the walls and the roof structure and exhib-
ited it in his factory. This was the first building of
this type in England and became a prototype for iron
churches, houses and storage spaces.
According to Batur, this structure was disassem-
bled and brought to İstanbul in 1841 by ship. The
building measured 8.10 × 15.50 m was three stories,
only the foundations and the wall supporting the inner
machinery were masonry. The floors and the roofing
were supported by cast iron and beams (Batur & Batur,
1981). This structure no longer stands today.
Since 1855s, especially after 1870 imported metal
products from Belgium came to İstanbul and other
eastern ports by regular steam ship routes from the
Port of Anvers. The “Feshane-i Amire Factory” was
set up in order to manufacture fezes and broadcloth
for the soldiers of the Ottoman army in the last quar-
ter of the nineteenth century. The metal roofing of the
weaving hall of the factory was supported by 274 cast
iron columns measuring 25 cm in diameter, bearing
the stamp of “PROVIDENCE MARCHIENNE BEL-
GIQUE” (Figs 11–12). “Les Forges de la Providence in
Marchienne-au-Pont” factory was established in 1838
and with the convergence of the Cockerill-Ougrée fac-
tory in 1966 became one of Belgium’s most important
industries. The English industrialist, John Cockerill
(1790–1840) established a company in the region of
Wallonia in 1817 and this company exists today as
“Cockerill Maintenance&Ingénierie” since 2004.
Figure 11. Imported cast-iron columns of the “Feshane-i
Amire” (Imperial Fez Factory).

Figure 13. Cast-iron columns of the “Maçka Silahhanesi”

(Maçka Armory).

Figure 12. Cast-iron column of the “Feshane-i Amire”

(Imperial Fez Factory), imported from Belgium.

The Maçka Armory was built in 1874–1875. The

weight bearing system consisting of cast-iron columns
with bases and capitals and beams are believed to have
been imported (Fig. 13). They did not survive since
the flooring of each storey was replaced by concrete
in 1955 (Çiftçi, 2004). Cast-iron columns are found
also in other factories and the commercial buildings
of this period (Figs 14–15).
From the last quarter of the nineteenth century
on, a construction technique known as jack arch
began to be used for horizontal supports. The jack
arch is formed by brick vaults between “I” shaped
steel beams which are placed on the short sides of Figure 14. Cast-iron columns of a commercial building in
the buildings (Fig. 16). The cross section of the “I” İstanbul (Maçka Armory).
shaped beams of the buildings studied varies between
5.5 − 9.5 × 16.5 − 22.5 cm.
The first buildings in İstanbul to use the jack (1874–76) (Yergün, 2002). The “I” shaped beams in
arch, as far as can be determined, are the German the “Mekteb-i Tıbbiye-i Þahane” (a Military Medical
Embassy (1874–77), the German Hospital (1874–78), School) dated 1895–1900 are stamped with “BUR-
the Europe Passage (1874) and the “Cité de Péra” BACH 22 NP” (Fig. 17). The “Burbach İron and

1003

Figure 15. Cast-iron columns of the “Cibali Tütün Fab-
rikası”(Cibali Tobacco Factory).
Figure 16. Jack arch construction technology (Yergün,
2002).
Steelworks” was first founded in 1856 by Luxem-
burger industrialists as the “Société en participation
des Forges de Sarrebruck”. Other metal elements of the
building came from the Belgian factory of “Cockerill”
(Batur, 1994).
An international competition was launched in
1892 for the architectural elements of the Saint
Stephen Church designed by Hovsep Aznavur. The
R. Ph. Waagner Company of Vienna, Austria won the
competition and produced the iron construction parts
1004
Figure 17. Roof structure and jack arch construction of the
“Mekteb-i Tıbbiye-i Şahane” (Military Medical School).
Figure 18. The Saint Stephen Church, iron architectural
elements from Vienne, Austria.
for the project in 1893. These parts were transported
by ship to İstanbul along the Danube River to the
Black Sea and construction was completed in April
of 1896 (Kuruyazıcı & Tapan, 1998). “R. Ph. Waagner
Vienne” is written on the side of some of the pilaster
bases while “BURBACH 30 NP”, “BURBACH 16
NP” stamps are found on the beams (Figs 18–20). This
building still exists as the sole example of its kind in
İstanbul. The Waagner Company was founded in 1854
and merged with the Biró Company in 1905. The com-
pany continues to operate today as the “Waagner-Biró
Stah”.
Large numbers of arms and tools were imported
from the German Krupp company founded in 1811.
Steel products of this company were used for the
construction of the Ottoman Railroads and railroad
buildings especially in the last quarter of the nine-
teenth century (Kösebay, 2007). The company took
the name “Thyssen-Krupp AG” in 1999 continues to
manufacture goods.

Figure 19. İron pilaster stamped with “R. Ph. Waagner
Vienne”.
Figure 20. Imported steel beams from Burbach factory.
3 CONCLUSION
A forceful change occurred in design philosophy and
construction techniques in Ottoman architecture in
the nineteenth century due to improved economic
and political relations between the Ottoman State and
European countries and the introduction of imported
materials such as modern bricks and metal supports.
However, some of the buildings from this period have
collapsed without ever being thoroughly documented
from an architectural viewpoint. Many of the exist-
ing structures dating to this period are on the brink of
disaster and require immediate attention even though
they have been registered as cultural assets.
1005
Some of the interventions carried out with the
aim of protecting them have further damaged both
the construction technique and the original imported
materials.The original bricks are disappearing because
the walls have been opened up and additions have
been made of solid masonry. The most wide scale
incursions occur from removing the metal elements
which support the flooring and the roof structures,
and then pouring concrete flooring. The registration
of these structures is also lacking. Many of the orig-
inal elements were removed before the building was
sufficiently documented. It has been noted that when
the roofs were being repaired, even tiles that were in
good condition were replaced with new ones without
any documentation. The imported bricks and all of the
metal elements are the architectural elements that bear
witness to construction techniques used at the time and
they need to be documented and preserved because
they reflect an era in Ottoman architecture.
Generally, the construction materials are not con-
served as required in the buildings that have been
observed. The two main causes of the decay are the
materials conditions and external causes. Due to the
conditions and locations of the materials like bricks,
tiles and metal supports can be conserved by clean-
ing methods and consolidation. If the material has lost
its feature as a structural element due to the inter-
nal causes, the material should be replaced. The new
bricks, tiles or metal elements must have the same
shape, same dimensions and same constitutional fea-
tures as to keep the harmony with the original ones.
In addition to this, the new materials should be dis-
tinguished from the originals by its texture, color
etc. according to the restoration and documentation
criterion.
REFERENCES
Arseven, C.E. 1965. Sanat Ansiklopedisi, İstanbul: Milli
Eğitim Bakanlığı Yayınları.
Batur, A. & Batur, S. 1981. İstanbul’da XIX. Yüzyıl Sanayi
Yapılarından Fabrika-i Hümayunlar”. I. UluslararasıTürk-
İslam Bilim ve Teknoloji Tarihi Kongresi, 14–18 Eylül
1981: 331–342. İstanbul: İstanbul Teknik Üniversitesi
Mimarlık Fakültesi Yayını.
Batur, A. 1994. Mekteb-i Tıbbiye-i Şahane Binası. Dün-
den Bugüne İstanbul Ansiklopedisi. 5: 377–379. İstanbul:
Kültür Bakanlığı ve Tarih Vakfı Ortak Yayını.
Can, C. 1993. İstanbul’da 19. Yüzyıl Batılı ve Levanten
MimarlarınYapıları ve Koruma Sorunları. İstanbul:Yıldız
Teknik Üniversitesi Fen Bilimleri Enstitüsü Doktora Tezi
(yayınlanmamış).
Can, C. & Girardelli, P. 1996. Beyoğlu’nda Bir Latin Anıtı.
Yaşayan Çizgiler-Sant’Antonio Kilisesi Mimari Çizimler
Sergisi: 13–71. İstanbul: Yapı Kredi Kültür Merkezi.
Çiftçi, A. 2004. 19. Yüzyılda Osmanlı Devleti’nde Askerî L’Indicateur Ottoman Annuaire-Almanach Commerce, 1881,
Mimarî ve İstanbul’da İnşa Edilen Askerî Yapılar. Ýstan- Constantinople: Cervati Frères & D. Fatzea.
bul: Yıldız Teknik Üniversitesi Fen Bilimleri Enstitüsü
Doktora Tezi (yayınlanmamış).
İnciciyan, P.Ğ. 1976. XVIII. Asırda İstanbul, İstanbul: Fetih WEB-SİTE
Cemiyeti İstanbul Enstitüsü Yayınları.
Kösebay, Y. 2007. Anadolu Demiryolu Çevresinde Gelişen
Mimari ve Korunması. İstanbul: İstanbul Teknik Üniver-
http://www.cmi.be
sitesi Fen Bilimleri Enstitüsü Doktora Tezi http://www.cockerill-sambre.com
(yayınlanmamış). http://www.crwflags.com
Kuruyazıcı, H. & Tapan, M. 1998.Sveti Stefan Bulgar Kilis- http://www.saarstahl.com
esi, Bir Yapı Monografisi, İstanbul: Yapı Kredi Yayınları. http://www.hebig.org
Mori, A. 1906. Gli Italiania Constantinopoli. Modena. http://www.lafargeroofing.co.za
Önsoy, R. 1988. Tanzimat Dönemi Osmanlı Sanayii ve Sanay- http://www.allatini.gr
ileşme Politikası, Ankara: Türkiye İş Bankası Yayınları. http://www.british history.ac.uk
Ökten, S. 1995.Yapı Mühendisliği ve Sanayi Devrimi. Ekrem http://www.answers.com
Hakkı Ayverdi Hatıra Kitabı: 127–136. İstanbul.
Ratier,Y. 1998. La Terre de Marseille: Tuiles, Briques et Car-
http://www.waagner-biro.at
reaux, Marseille: Chambre de Commerce et d’Industrie. http://www.thyssenkrupp.com
Sandalcı, M. 1997. Tuğlalar. Tombak Dergisi, 17: 50–58.
İstanbul: Horhor Yayıncılık.
Tanyeli, G. 1990. Osmanlı Mimarlığında Demirin Strük-
türel Kullanımı (15–18 y.y.). İstanbul: İstanbul Teknik
Üniversitesi Fen Bilimleri Enstitüsü Doktora Tezi
(yayınlanmamış).
Yergün, U. 2002. Batılılaşma Dönemi Mimarisinde, Yapım
Teknolojisindeki Değişim ve Gelişim. İstanbul: Yıldız
Teknik Üniversitesi Fen Bilimleri Enstitüsü Doktora Tezi
(yayınlanmamış).
Osmanlı Başkenti’nde Belçika, İlk Adımlardan Güzel Çağa
Kadar, Belçika SarayınınYüzüncüYıldönümü 1900–2000,
2001. İstanbul: Belçika Başkonsolosluğu Yayını.

1006
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Advance composites applications on historical structures in Italy:

Case studies and future developments

P. Casadei
Consultant Engineer, R&D
TEC.INN. S.r.l., Technical & Commercial Division, Milano, Italy

E. Agneloni
TEC.INN. S.r.l., Head Quarters, San Mariano, Perugia, Italy

ABSTRACT: Due to their light weight, high stiffness-to-weight and strength-to-weight ratios, and potentially
high resistance to environmental degradation, resulting in lower life-cycle costs, advanced composites mate-
rials, commonly known with the acronym FRP, are increasingly being considered for use in civil engineering
applications, ranging from the retrofit and rehabilitation of buildings and bridges to the restoration and strength-
ening of historical-monumental masonry structures. Thanks to the recently published Italian FRP guidelines,
CNR-DT 200/2004, by the Italian National Research Council, there has been an increasingly demand of their
implementation for general strengthening and retrofitting applications as well as seismic mitigation of structures
and infrastructures. This paper aims to provide a general overview of the Italian state of the art implementation
of FRP materials in the last 15 years on historical structures, through different case studies, and presenting the
potentials of a new family of retrofitting materials made of ultra-high tensile strength steel filaments arranged
in cords and knitted similarly to FRP sheets.

1 INTRODUCTION

After an initial phase of experimentation and research,

validated by extremely positive results, the use of
advanced composite materials in the civil engineer-
ing industry, is, now days, a well established real-
ity. Currently the most retrofit and strengthening
projects, regard buildings, industries, infrastructures
and, particularly in seismic areas, masonry historical
structures.
In the last years, the technological innovation in the
aforementioned sector has been characterized by an
important acceleration. The boost of such technology
started in the 90’s, when innovative fiber materials,
thanks to their very competitive weight to strength
ratios, high durability and ease of installation, repre-
sented the most ideal substitute of the widely applied
technique of “Béton Plaque”.
1.1 Fiber Reinforced Polymers Materials – FRP
Fiber Reinforced Polymer (FRP) materials are com-
posite materials consisting of high strength fibers in
a polymer matrix. The fibers in an FRP composite
are the main load-carrying element and exhibit very
high strength and stiffness when pulled in tension. An
Figure 1. Béton Plaque strengthening examples on RC
structures.
FRP laminate will typically consist of several million
of these thin, thread-like fibers. The polymer matrix
(sometimes referred to as the resin) protects the fibers
from damage, ensures that the fibers remain aligned,
and allows loads to be distributed among many of the
individual fibers in the composite. There are a variety
of fiber types and resins that may be used to create
an FRP composite. Fibers are selected based on the
strength, stiffness, and durability required for the spe-
cific application, and the resins are selected based on
the environment the FRP will be exposed to as well as
the method by which the FRP is being manufactured.
Among several possibilities, the fiber types that
are typically used in the construction industry are
carbon, glass, and aramid. In selecting the type of

1007

Figure 2. Example of available advanced fiber materials for
manual lay-up applications: carbon, glass and aramid sheets
in different width.
Figure 3. Example of using aramid bars for anchoring
AFRP sheets used for retrofitting the natural staircase of the
Koch Palace in Rome.
fiber to be used for an application, there are a few
things to consider. Glass FRP is excellent for seis-
mic upgrades where the seismic loads only temporarily
engage the FRP. In cases where stresses are sustained
in the FRP (such as in bending and shear strengthen-
ing), glass FRP should be avoided because of creep
rupture effects. Carbon is much more suitable in these
applications. Similarly in exterior applications, car-
bon FRP will be much more durable. Aramid fibers
are indeed mostly used now days in masonry applica-
tions to realize connection between the strengthening
layer, commonly of carbon or glass fibers, and the main
structural members thanks to their better performance
towards concentrated shear stresses that make them
the most suitable for this type of application.
FRP systems in the civil repair industry are used to
strengthen existing structures/infrastructures. Struc-
tures may need strengthening due to deterioration,
design/construction errors, a change in use or loading,
or for a seismic upgrade. In particular this last issue
has become of high priority in Italy and in all seismic
countries, where new and updated seismic codes have
come into place.
FRP essentially works as reinforcement in con-
crete/masonry/timber and provides strength where
concrete/masonry are weakest – in tension. FRP may
be used on beam or slab soffits to provide additional
1008
Figure 4. Hardwire material.
flexural strength, on the sides of beams to provide addi-
tional shear strength, or wrapped around columns to
provide confinement and additional ductility (a pri-
mary concern in seismic upgrades). Among many
other applications, concrete and masonry walls may be
strengthened to better resist seismic and wind loads,
concrete pipes may be lined with FRP to resist higher
internal pressures, and silos and tanks may be strength-
ened to resist higher pressures. On historic structures
FRP are mainly used to strengthen masonry domes
and vaults providing a link among the several portion
of the elements without adding any additional mass
and most importantly avoiding the most critical fail-
ure modes of such type of structures, contrasting the
formation of hinges and retarding/avoiding premature
collapse.
1.2 Steel Reinforced Polymer – SRP & Steel
Reinforced Grout – SRG materials
Next to such consolidated retrofitting technology, is
nowadays available also another family of composite
materials similar to FRP, made of fine ultra-high ten-
sile steel filaments, twisted to form cords and tailored
similarly to unidirectional (carbon, aramid or glass)
fiber sheets in order to then be easily installed on site
with the manual lay up technique. The steel cords used
are identical to those used for making the reinforce-
ment of automotive tires, and manufactured to obtain
the shape of the fabric tape prior to impregnation
(Hardwire, 2002).
The twisting of the wires allows some mechanical
interlock between the cords and the matrix, and may
also induce an overall ductile behavior upon stretching.
The cords are also coated with either brass or zinc mak-
ing the material potentially free of any corrosion and
suitable for different kind of environmental exposure.
Characterization work, including durability studies, is
currently in progress as necessary for implementation
in future design guidelines.
The great advantage of such materials is that they
can be either impregnated using epoxies, then called
SRP (Steel Reinforced Polymer), or using mortars,
then being called SRG (Steel Reinforced Grout). The
choice of the resin is based upon type of substrate
where bonding the material and also upon the density
Figure 5. Two types of available cords.
Figure 6. Available sheet densities for each kind of cord.
Figure 7. Material installation examples.
of the tape to be installed: higher density sheets may be
impregnated only with resin matrices, while low and
medium densities with either type of matrix.
Such materials present the same advantages of FRP
in terms of lightness (this is true mostly for low and
medium densities tape), strength and ease of installa-
tion, but with several additional advantages that have
made them very attractive for masonry retrofitting.
The possibility of impregnating them with cement
or hydraulic mortars allows a much easier installation
for skilled workers and also a considerable reduction of
primary material costs with respect to FRP. Secondly,
being such sheets made of steel cords, they have a con-
siderable shear strength, not present in FRP, making
them particularly attractive for uneven surfaces and
also for mechanical anchors, prohibitive with typical
FRP strengthening solutions.
1009
Figure 8. Most common failure mechanisms on masonry
structures for which FRP have proven to be successful in
retrofit.
1.3 Design guidelines and jobsite procedures
It is evident though that the installation of these mate-
rials cannot be handed into everybody’s hands: it is of
critical importance that experienced and well trained
engineers and contractors need to be always hired when
dealing with these innovative materials. Nowadays the
publication of international guidelines, among which
there is also the recently published Italian document
published by the National Research Council CNR-DT
200/2004 (also available in English) “Guide for the
Design and Construction of Externally Bonded FRP
Systems for Strengthening Existing Structures”, makes
this sector of the civil industry safer from extemporary
designers and contractors. In any case though, man-
aging these materials both on the design as well as
installation phase cannot be reduced to simply follow-
ing a design guide line. A critical role in the field of
restoration and consolidation is based on preliminary
non-destructive and static load tests to acquire all the
necessary information about the structure, prior the
design and the installation of these materials. In order
to assure the quality of the execution, non-destructive
tests shall also be performed once the job is complete
to verify that all critical parameters, such as bond to
the substrate, fiber alignment among the most signif-
icant ones, are all within the recommended values.
This approach not only allows verifying the quality
of the contractor job but in many cases has also been
necessary to guarantee the owner of the efficiency of
the technology chosen. Many are the jobs where a pilot
study has been conducted over a small portion of the
structure following this procedure: first a static load
test is performed to evaluate the current behavior of the
structure under loads (deformations, degrees of fixity,
vibrations, strains, etc), then the FRP is installed and
finally a last static load test is performed in order to
compare the behavior of the structure before and after
strengthening. In many cases the direct comparison
of the load-vs-deflection curve before and after the
FRP installation is the key parameter that validates
the technology.
Figure 9. The masonry arch bridge of St.Peter in Cariano.
1.4 Italian situation
In Italy, as well as in many other seismic countries,
the use of advanced composite materials for strength-
ening and retrofitting existing structures has been
heavily introduced in the market by recent earthquakes
and natural disasters. In particular the earthquake that
interested the central Italian regions of Umbria and
Marche in 1997–’98 has heavily pushed engineers
and contractors on the use of the FRP technology
thanks to the aforementioned qualities. In particular, to
strengthen structures against seismic/dynamic loads,
FRPs are particularly efficient because in front of very
performing mechanical properties they have very low Figure 10. NDT testing prior and after FRP strengthening.
weights and consequently they do not add any mass
to the structure making them ideal for this type of
2 FRP APPLICATIONS ON MASONRY
retrofits.
HISTORICAL INFRASTRUCTURES
Nowadays the mainstream market of these mate-
rials in Italy is for masonry historical structures
2.1 CFRP strengthening of a masonry arch bridge
where their implementation varies from: increas-
located in San Pietro in Cariano – Verona (IT)
ing the capacity of panels, arches, or vault; wrap-
ping of columns to enhance their compressive The retrofit concerned the strengthening of a solid
strength and ductility; reducing thrust forces in thrust- brick masonry arch bridge located in San Pietro in
ing structures; transforming non-structural members Cariano (Verona). The change of the traffic loads in
into structural members by increasing their stiffness recent years and the lack of maintenance have caused
and strength; strengthening and stiffening horizon- on the bridge a remarkable crack state needed to be
tal non-thrusting structures; wrapping buildings at urgently addressed.
floors and roof locations to improve vertical and The work sequence has been as follows: first of
horizontal strength to lateral loads, typically seismic all before any FRP installation, all cracks have been
actions. filled with tixotropic mortar carefully removing any
TEC.INN. S.r.l., founded in Perugia-Italy in 1986 falling parts of masonry and leveling with mortar.Then
has been pioneer in Italy in the research, installa- the entire soffit of the arch has been hand-blasted to
tion and development of the FRP technology with prepare the substrate for an ideal bond. Once this initial
over 1000 retrofit-consolidation-strengthening appli- phase has been completed the substrate was ready for
cations all over Italy and few also abroad, vary- FRP installation.
ing from reinforced and pre-stressed concrete to CFRP sheets have been installed longitudinally and
historical-monumental masonry structures. Specialty transversally contributing to an overall increase of the
contractor on innovative materials, TEC.INN. S.r.l. arch capacity and to provide a transversal contribution
is able to offer, in Italy and worldwide, advanced of the entire structure subjected to concentrated loads.
retrofitting and strengthening solutions, using FRP Before and after strengthening, an exhaustive test-
systems and innovative materials, to public and private ing campaign has been conducted, in order to provide
owners, consulting engineering firms and contrac- all necessary information to design the CFRP retrofit
tors, starting from in-situ diagnostic phase final- solution. Firstly laboratory tests have been performed
ized to support designers of the retrofitting system, on samples for material characterization. Then on site
till the installation phase and/or knowledge trans- tests, such video endoscopy and thermography have
fer to local contractors of advanced composite FRP been conducted to evaluate the crack state and over-
technology. all structure. Once the FRP installation was completed,

1010
Figure 11. The “Maggiore Fountain” in Perugia.
pull-off tests were conducted to evaluate bond strength
with the substrate. Finally in situ load tests using
preloaded trucks carefully positioned on the road-
way simulating the most possible loading conditions
have been performed before and after strengthening
to determine structural improvement. Vertical move-
ments of the bridge and strain deformations of the
CFRP have been monitored, demonstrating the success
of the strengthening.
3 FRP-SRP & SRG APPLICATIONS ON
MASONRY HISTORICAL STRUCTURES
3.1 FRP retrofitting and consolidation of the
“Maggiore Fountain” in Perugia.
The “Maggiore Fountain” in Perugia was built and dec-
orated between 1277 and 1278. The fountain consists
of three basins laid one upon the other and on the third
bronze basin that was cast by Rosso Padellaio of Peru-
gia, is placed a group of three women. On the other two
lower basins there are symbolic bas-reliefs and statues
which testify the strength of Perugia at the end of the
thirteenth century.
The restoration has brought the artistic work back
to its ancient splendor, eliminating all the invasive
and obsolete retrofitting works carried out during
the previous restorations. The lower basin in fact has
undergone a considerable damage in consequence of
very invasive retrofitting works that had caused the
crack of the marble panels. Therefore the retrofitting
had the aim to strengthen, border and join the mar-
ble panels inside the lower basin of the fountain. The
restoration has been preceded by the catalogue of a vol-
ume containing all the information – both geometrical
and photographic – about each single panel, includ-
ing the performing of strengthening solution and the
suitable experimental not destructive testing (NDT)
in order to illustrate the map of cracks before the
restoration.
The retrofit was subdivided into two stages: first the
strengthening of each single panel through the applica-
tion of GFRP strips (unidirectional glass fiber sheets)
lengthwise and crosswise to restore the loss of conti-
nuity because of the cracks created by changes of tem-
perature, structural settling and unsuitable previous
1011
Figure 12. FRP strengthening of the marble panels of the
fountain.
Figure 13. Two views of the Royal Palace in Milan.
restorations; second joining contiguous marble pan-
els through unidirectional GFRP. GFRP strips have
been prearranged in order to respect the original idea
of panels not rigidly fixed, so as to avoid, in case of
earthquake shocks, a possible falling apart.
3.2 Strengthening of the masonry vaults and of the
timber floors of the “Royal Palace” in Milan
using two different FRP technologies
The structure that characterizes the buildings of “Royal
Palace” is very complex and different, showing the
long and difficult evolution that it has undergone
dug the centuries with retrofitting jobs and following
reconstructions. “Royal Palace” in fact presents dif-
ferent structural schemes corresponding to different
constructive systems and ages.
While the vertical structures basically consist of
load bearing, the walls horizontal one presents a
remarkable variety of structural systems: we can notice
masonry vaults, timber floors with single or double
beam layers, steel joists’ floors with masonry vaults
insisting on them, floors made by clay bricks and RC
joists and reinforced concrete beams even with large
spans.
The study of the current state of art and appropriate
structural understanding of all its elements represent
the main “start” to make the most suitable decisions
for each single case. Consequently a lot of NDT inves-
tigations have been done to describe the state of art of
the structures, aiming to characterize the units of the
building and the static details of each single floor sys-
tem. The main NDT campaign consisted of endoscopy
tests integrated with sound testing, flat jacks and
small localized sample material removal for material

Figure 14. Retrofit phases of the timber beams.
Figure 15. GFRP strengthening of masonry vaults and
anchoring the sheets using aramid bars epoxied into the
perimeter walls.
characterization. Where endoscopy could not be used,
geo-radar, laser profilometer and thermography has
been adopted following the initial phase of knowledge
about the structure, the most suitable retrofit solution
was chosen.
The structural retrofit aimed to respond to a change
of the loading conditions of an entire floor of the
Palace, due to a change in destination of use of it.
The overall retrofit job consisted of three main phases.
The first one consisted of the local realization of
members such as staircases or lifts needed to assure
the accessibility of the rooms and the safe standards
imposed by current laws. The second one concerned
the complete reconstruction of structural elements,
above all of floors, that do not have sufficient static per-
formance nor allow an economic and easy restoration.
This was realized by installing glass FRP poltruded
profiles on top of the existing timber beams without
removing the existing timber joists. The GFRP “I” pro-
files were bonded on the beam after that it was partially
reconstructed, by creating a perfectly horizontal pro-
file with epoxy resins especially designed for timber
reconstruction.
1012
Figure 16. The Saint Biagio’s temple in Montepulciano.
Finally, the third one, concerned the strengthen-
ing of the masonry vaults, that did not have adequate
structural performance according to the new loading
scenario, or simply because they needed to be pre-
served because of their architectonical-monumental
value.
Retrofit was realized using GFRP sheets on the
extrados of the vaults and aramid bars installed on the
lateral masonry walls for anchoring the extremities of
the GFRP sheets on them.
3.3 Strengthening natural stone columns using
pretensioned aramid bars of the Saint Biagio
Temple in Montepulciano-Siena
The temple, built in the late 1400 and beginning of
1500 with a perimetral base shaped as a greek cross
with the transept and the principal nave intersepting
in the centre, is constructed entirely with travertine
natural stone.
In the recent years, due to multiple structural rea-
sons, the main columns started presenting vertical
cracks indicating that they were subjected to vertical
loads that induced splitting and cracks in the traver-
tine. Because the column are hidden within the main
structural walls, they do not allow easy structural
retrofit solutions; in addition to this, because the tem-
ple does not present frescoes on the natural stones
there was not the possibility of hiding standard retrofit
solutions.
Due to all these considerations the retrofit was
designed in order to be the least invasive possible
and also to be an “active” consolidation solution.
These considerations called for retrofitting the traver-
tine columns with pretensioned small diameter aramid
bars inserted into small holes drilled in the columns
and injected with fluid resin. Operatively the retrofit
consisted of drilling holes into the travertine with pre-
cise diamond blades, then injecting the hole with epoxy
resin and immediately inserting the aramid bar in order
to anchor it on the column. Once the resin cured, pre-
tensioning of bar was applied using a dynamometric
spanner to control the stress induced in the bar. Once
pretensioning was completed, injection to fully fill the
hole was realized. Finally to mask the hole, a coloured
plaster was applied on the surface of the small hole.

Figure 17. Retrofit phases of the travertine columns.
Figure 18. The SS.Trinity Convent in Orvieto: Overview
and frescos detail.
3.4 FRP retrofitting of the vaults and SRP
strengthening of the ring curb beams of the
“SS.Trinity Convent” in Orvieto
The monastery of the SS.Trinity was built west of the
city of Orvieto, in the valley of the river Paglia, in
the early 1034, and then annexed to the monastery of
the Benedictine nuns.
In the centuries the structure has been subjected to
several structural changes such as extensions, raisings
and more, nevertheless several restorations aiming to
solve the degradation acting on the structure. It was
only in the year 1929, after being sold to private clients
that the convent was declared an historical and artistic
heritage structure. Unfortunately the monastery was
yet left with no use till the early 2007, when the cityhall
of Orvieto decided to restore the structure and promote
it under an intense project of restoration, aiming to
respect and yet promote its historic and artistic value.
The project aimed to realize a high-class hotel and
conference centre using its large rooms and open-
sky cloister. In order to do so, there was a need
to consolidate the entire structure, with the goal of
maintaining its historic view yet reaching the current
code standards for centres like the one wanted by the
cityhall.
1013
Figure 19. Retrofit of the masonry double curvature vaults
with unidirectional GFRP sheets.
Figure 20. Strengthening of the ring corbel masonry beam
with unidirectional high strength fiber sheets.
The design called for retrofitting all the superstruc-
tures, in particular the masonry vaults, floors and
roofs in order to provide strength without increasing
the masses, providing so a consolidation solution that
could withstand the new Italian seismic code.
It was then chosen to consolidate the vaults using
unidirectional glass fibre sheets (painted in red in
Figure 19).
In order to absorb possible lateral loads coming
from seismic shocks (Orvieto is indeed in a highly
seismic area classified as “Zone 2”, being 1 the most
critical), the project called for realizing corbel ring
beams with masonry reinforced with unidirectional
high strength steel sheets (Hardwire® 3X2-medium
density tape). Previous tests conducted at the Univer-
sity of Perugia, showed how this solution could be
particularly effective since it provided the same level
of strength of an identical reinforced concrete beam,
commonly and wrongly so used in the past for this kind
of retrofit, with the big advantage of having a highly
reduced self weight of the retrofit solution.
For this particular strengthening solution it was cho-
sen to adopt an epoxy matrix to impregnate the medium
density 3X2 Hardwire sheets, to ensure best adhesion
possible between the masonry bricks and create the
ring beam of the desired strength and stiffness.

Figure 21. The Contucci’s Palace in Montepulciano-Siena:
Overview and frescos detail.
Figure 22. FEM model of the vault with the cut trough.
Figure 23. SRP detail installation on the vaults.
3.5 SRP strengthening of the vaults of “Contucci
Palace” in Montepulciano-Siena
The Contucci’s Palace is an exclusive historic private
residence, located in the heart of Tuscany, in one of
the most magnificent and most admired renaissance
squares, Piazza Grande in Montepulciano.
The retrofit was needed in order to consolidate the
vaults of the first and noble floors, necessary because
the owners wanted to install an elevator between the
two floors, consequently cutting trough the vaults.
In order to decide any kind of retrofit solution, a
finite element model was created with commercial
FEM routines, in order to simulate the behaviour and
stress concentrations of the two vaults before and after
the cut-through.
Once the stresses involved were known, the retrofit
was designed in order to absorb the increment of
stresses generated by the cut, creating strengthened
strips along the sides of the hole, like a sort of supports
for the vault weakened by the cut. The strengthen-
ing material chosen was the medium density Hardwire
3X2 sheet, impregnated with epoxy resin. The reason
why epoxies were chosen was generated by the need
of creating stiff strips around the hole.
1014
Once the strengthening of the entire vault was com-
pleted, and the entire vault was shored, the cut was
realized.At this point the strips of material reaching the
hole where anchored on it, by creating sort of stirrups
that could anchor the material and prevent any vertical
movement around the hole due to the loss-continuity
of material induced by the cut.
4 CONCLUSIONS
The reported case studies have shown the large field
of possible cases where FRP and more recently SRP-
SRG materials have been largely introduced and
successfully applied. In recent years thanks to the
introduction of national as well as international guide-
lines, that not only regulated the analytical design of
section strengthened with these materials, but also
their installation, advanced composite materials have
entered massively the civil construction industry and
we should expect that in recent years even more
projects will involve their use, not only for retrofitting
and strengthening existing structures, but also for
designing new ones. Relying on experienced contrac-
tors and engineers remains though the key issue for a
continuous success of this industry.
REFERENCES
Borri, A., Castori, G., and Grazini, A., (2007). “Cordoli
sommitali in muratura armata con SRG” XII Convegno
Nazionale L’Ingegneria Sismica in Italia (ANIDIS), Pisa,
Italia, 10–14 Giugno, 2007.
Borri, A., Castori, G., Casadei, P., and Ebaugh, S. (2007).
“Rinforzo di Archi in Muratura con Materiali Compositi
Innovativi” XII Convegno Nazionale L’Ingegneria Sismica
in Italia (ANIDIS), Pisa, Italia, 10–14 Giugno, 2007.
Borri, A., Castori, G., Casadei, P., and Ebaugh, S. (2007).
“Research on Composite Strengthening of Masonry
Arches” Eighth International Symposium on Fiber Rein-
forced Polymer Reinforcement of Reinforced Concrete
Structures (FRPRCS-8), Patras, Greece, July 16–18, 2007.
Casadei, P., Nanni, A., Alkhrdaji, T., and Thomas, J., (2005).
“Behavior of Double-T Prestressed Beams Strengthened
With Steel Reinforced Polymer”, Advances in Structural
Engineering an International Journal (ASE), Vol. 8, No. 4,
pp. 427–442.
CNR-DT 200/2004, 2004: “Istruzioni per la Progettazione,
l’Esecuzione e il Controllo di Interventi di Consolida-
mento Statico mediante l’utilizzo di Compositi Fibror-
inforzati” Published by the National Research Council
(CNR), Roma, pp. 164 (also available in english).
“Compositi FRP: Linee guida per il rinforzo strutturale”
published by NCT Global Media Editore, 2002, ISBN
88-900892-0-2.
Hardwire LLC, 2002,“What is Hardwire,” www.hardwirellc.
com, Pocomoke City, MD.
TEC.INN. S.r.l., “Innovative Technologies” San Mariano-
Perugia, Italy, www.tecinn.com.
Novel conservation engineering techniques,
restoration and strengthening
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

FRP-strengthening of masonry structures: Effect of

debonding phenomenon

E. Grande, M. Imbimbo & E. Sacco

Department of Mechanics, Structures and Environment, University of Cassino, Italy

ABSTRACT: The present paper shows and discusses some aspects concerning the delamination phenomenon
of Fibre Reinforced Plastic (FRP) materials fixed on masonry support. With this aim, in the first part of the
paper the results of an experimental campaign conducted by the authors concerning the effect of the bond length
on the delamination phenomenon of CFRP glued on clay bricks are discussed. Moreover, in order to underline
further aspects concerning the decohesion mechanism of FRP-reinforcements, in the second part some numerical
analyses are also presented.

1 INTRODUCTION This aspect has been widely investigated by several

Masonry structures constitute a significant part of
historical constructions in Europe and particularly in
Italy. Many of these structures are structurally defi-
cient for current or safe use in the light of the new
seismic code regulations. This involves the need to
retrofit and upgrade the masonry structures through
strengthening techniques.
A quite recent strengthening approach consists
in applying FRP strips on the external surface of
the structure adopting different configurations and
anchorage modalities; the FRP strips are arranged
in order to give an external resisting system which
enhances the global capacity of the structure. This
technique is used particularly to preserve historic
buildings and monuments because the lightweight
FRP strips or sheets are easy to handle and can be
also applied into restricted spaces with little risk of
damaging the support material. In addition, compos-
ites can be made to fit irregular geometries and can be
cut or trimmed in-situ.
In recent years, large investments have been
concentrated in order to investigate both the modali-
ties of application and the efficacy of FRP-reinforcing
systems. Several developments have been performed
and some guidelines for the strengthened and the con-
servation of existing structures have been formulated
(CNR-DT200/2006).
Several aspects concerning the behaviour of FRP-
strengthened structures have emerged from the
performed studies.Among these, the decohoesion phe-
nomenon between the support and the FRP sheets
represents the most important one because it is
responsible of the efficacy of the reinforcing system.
authors particularly with reference to RC structures.
Only in the last few years some experimental tests have
been developed also to study the nature of the bond
between composite and masonry supports (Aiello and
Sciolti, 2005; Casareto et al., 2003; Barbi et al.,
2004). These tests have evidenced several differences
between the FRP-strengthened masonry and the FRP-
strengthened concrete structures, particularly in terms
of bond strength and detachment mechanisms. This
involves the need to further analyze the experimental
behaviour of masonry elements strengthened by FRP
and provide the possible improvement for the standard
code formulation.
The present paper shows and discusses some aspects
concerning the delamination phenomenon of fibre
reinforced plastic (FRP) materials attached to masonry
support. To this purpose, in the first part of the paper
are discussed the results of an experimental cam-
paign conducted by the authors concerning the effect
of the bond length on the delamination phenomenon
of CFRP glued on clay bricks. In addition, to better
understand further aspects concerning the decohe-
sion mechanism of FRP-reinforcements, in the second
part of the paper are also presented some numerical
analyses.
2 EXPERIMENTAL PROGRAM
2.1 Test specimens
The program involves eight bonding tests (single shear
pushing tests) performed on clay bricks reinforced
by carbon-fibre-reinforced-plastic laminates (CFRP)

1017
 Table 1a. Tests results (compressive strength).

40 mm
Dimensions (mm)

80 mm

120 mm
250 mm

160 mm
Specimen L B H Section (mm2 ) fb (MPa)

1 55 56 55 3080 41.5
2 55 55 54 3025 35.7
3 56 55 55 3080 40.6
4 55 55 55 3025 34.7
5 55 55 54 3025 36.6
120 mm 6 55 55 55 3025 41.7
legend: strain-gauge on the unbonded part
Average value 38.5
strain-gauge on the bonded part

Figure 1. Geometry of the tested specimens.

Table 1b. Tests results (tensile strength).

Dimensions (mm)

Specimen L B H Section (mm2 ) fv (MPa)

1 55 255 117 14025 3.84

2 55 255 117 14025 3.15
3 55 255 117 14025 4.45
4 55 255 117 14025 2.45
5 55 255 117 14025 3.75
6 55 255 117 14025 3.02
Average value 3.44

Commercial available clay bricks, of dimensions

Figure 2. (a) Test for determining the compressive strength,
(b) Test for determining the tensile strength, (c) Test for deter-
mining the elastic modulus and (d) Test for determining the
shear strength.
with different bond lengths. The specimens were char-
acterized by four different values of the bond length Lb
and the same type of support. Figure 1 shows the details
of the specimens (scheme 1: Lb = 40 mm; scheme
2: Lb = 80 mm; scheme 3: Lb = 120 mm; scheme 4:
Lb = 160 mm).
equal to 250 × 120 × 55 mm3 , are used. Standard
tests for determining the secant elastic modulus and
the compressive and tensile strengths are performed
according to UNI EN 8942/3, UNI 8942/3 and UNI
6556 respectively (Figure 2a, b, c). In particular, the
compressive tests have been performed considering
specimens of dimensions 55 × 55 × 55 mm3 obtained
by cutting the brick; the tests for the elastic modu-
lus have been performed on specimens of dimensions
50 × 50 × 150 mm3 , obtained by cutting the brick; the
tests for the tensile strength have been performed con-
sidering the entire brick; the tests for the shear strength
have been performed on the brick providing four ver-
tical cuts. The tests were performed using a universal
testing machine (Gabaldini SUN) and deducing the
applied load (for all the tests) and the strain values
(only for the specimens used for evaluating the secant
modulus).
The results are reported in Tables 1a, b and c. Addi-
tional shear tests (Figure 2d) were performed and the
results are also reported in Table 1d.
The CFRP composite used is a unidirectional lam-
inate strip of type Sika Carbodour S512 characterized
by the properties reported in Table 2. The adhesive
used is of type Sikadur 30.

1018
Table 1c. Test results (elastic modulus).

Dimensions (mm)

Specimen L B H Section (mm2 ) Eb (MPa)

1 50 50 150 2500 13850

2 50 50 150 2500 16618
3 50 50 150 2500 15369
Average value 15279

Table 1d. Tests results (shear strength).

Dimensions (mm)

Load
Specimen Cut height (mm) (N) τb (MPa)

1 36.0 4524 1.14

2 36.5 2100 0.52
3 37.0 3679 0.90 Figure 3. Test set-up.
4 36.5 3878 0.97
Average value 0.88 Table 3. Bond strength and failure mode.

Scheme Lb (mm) Fu (kN) Failure mode

Table 2. Characteristics of the strengthening system.
Series 1 1 40 6.32 a+b
2 80 6.90 a+b
FRP Laminates (Sika Carbodur S512) 3 120 10.28 a
Average value of the tensile strength ftfm 3100 MPa 4 160 8.34 a
Minimum value of the elastic modulus Ef 160000 MPa
Ultimate deformation εf 1.7% Series 2 1 40 6.57 a+b
Thickness tf 1.2 mm 2 80 7.09 a+b
Width bf 25 mm 3 120 10.40 a
4 160 9.57 a
Adhesive material (sikadur 30)
Average value of the tensile strength fta 25 MPa
(+15◦ ) 2.3 Test results
Average value of the shear’s modulus Ga 2694 MPa
Average value of the elastic modulus Ea 12800 MPa The results of the tests in terms of bond strength and
failure modes are summarised in Table 3.
The first, indicated as “a” in the table, is shown in
Figure 4.a and is characterised by the detachment of a
2.2 Experimental procedure
thin and uniform layer of the brick material. The sec-
The specimens were tested by using the test device ond, indicated as “b” in the table, is shown in Figure 4.b
reported in Figure 3. and occurs with the removal of a considerable part of
In the test set up the bricks are fixed on the steel the brick material near the unloaded end of the rein-
plate while the free end portion of the CFRP lami- forcement. It was observed that the type of the failure
nate is gripped to the special device connected to the mode was particularly influenced by the length Lb of
hydraulic jack. All the specimens were instrumented the bonded part of the FRP-reinforcement. In fact, the
by electrical strain gauges applied along the middle schemes with the smallest dimensions of Lb (schemes
longitudinal section of the reinforcement. In partic- 1 and 2) showed the failure mode “b” whilst the oth-
ular, five strain gauges were used in the case of the ers (schemes 3 and 4) primarily showed the failure
scheme 1 and eight strain gauges were used in the mode “a”.
others schemes (Figure 1). The experimental values of the bond strength
The signals from the gauges and the load cell were reported in Table 3 point out the following consider-
recorded by a National Instrument data acquisition ations. First, the variation of the bond strength from
system with 16 channels and processed through the scheme 1 (Lb = 40 mm) to scheme 2 (Lb = 80 mm)
labview software (2006). is small. Second, the bond strength increases by 50%

1019
Figure 4. (a) Decohesion mechanisms of specimens and
(b) Decohesion mechanisms of specimens.

if the bond length increases from 40 mm (scheme 1)

to 120 mm (scheme 3). These first results suggest that
the bond length of scheme 1 could be lower than the
optimal length of the reinforcement since an increase
of the bond length (from 40 mm to 120 mm) enhances
the bond strength. Third, the bond strength and the fail-
ure mode do not significantly change by increasing
the bond length from 120 mm (scheme 3) to 160 mm
(scheme 4). This implies that the bond length of the
scheme 3 could be close to the optimal bond length of
the reinforcement; in fact, further increase of the bond
length, for example scheme 4, does not affect neither
the bond strength nor the failure debonding mode.
In order to analyse the transfer mechanism of the
applied tensile load from the FRP reinforcement to
the brick support, in Figure 5 are reported the strain
paths occurred along the reinforcement of the speci-
mens (series 2) at various load levels. From the plots
it is possible to notice a greater slope of the curves in
the vicinity of the loaded end of the plate. This could
suggest the fact that the load transfer mechanism pri-
marily occurs along a limited length of the plate next
to the loaded end.

3 NUMERICAL ANALYSES
In order to propose a simple numerical model based
on the experimental data and able to capture the global
behaviour of CFRP-laminates glued on masonry
blocks, some numerical analyses have been performed.
3.1 Modelling approach
Regarding the modelling approach adopted by the
authors for simulating the response of the examined
specimens, a 2D finite element model (Figure 6) has
been developed for all the examined specimens (i.e.
varying the bond length Lb of the reinforcement)
using the code DIANA9.1 (2000). In the models, both
Figure 5. Strain profiles of FRP laminates.
the brick, the FRP-laminate and the FRP/brick layer
have been modelled by adopting the following finite
elements:
• clay bricks: four-node quadrilateral isoparamet-
ric plane stress elements (labelled in the code as
Q8MEM) based on linear interpolation and Gauss
integration.
• FRP-laminates: beam elements characterized only
by the axial deformation (also called truss-element
and labelled in the code as L2TRU).

1020

Figure 6. Numerical model.
• FRP/brick layer: interface elements (labelled in the
code as L8IF) placed between L2TRU and Q8MEM
elements. The basic variables for the interfaces ele-
ments are the nodal displacements. The derived
values are the relative displacements and the trac-
tions used by code to describe a relation between
the tractions and the relative displacements across
the interface.
3.2 Modelling of material behaviour
From the experimental tests it was observed that the
delamination phenomenon involved only the brick,
whilst it did not affect neither the FRP laminates nor
the FRP/brick interface. On the basis of this observa-
tion and in order to simplify the numerical model, it
has been adopted a nonlinear behaviour only for the
brick. On the contrary, a linear-elastic behaviour has
been assumed for the FRP laminates and the FRP/brick
interface. The mechanical properties of the bricks, the
FRP and the adhesive material have been described
in the previous section 2. The mechanical properties
of the FRP/brick layer (i.e. the tangential stiffness
kt and the normal stiffness kn ) have been obtained
through an homogenization of the brick and adhesive
properties using the following relationships:
where:
Ea (see table 2), Eb (see table 1.c) are the elastic’s
moduli of the adhesive and the brick respectively; hi is
the thickness of the interface layer between the brick
and the FRP, assumed equal to 1.0 mm, as observed by
tests; Ga (see table 2), Gb (see table 4) are the shear’s
moduli of the adhesive and the brick respectively. The
properties of the elements used in numerical model are
summarised in Table 4.
In order to model the nonlinear behaviour of the clay
brick, an orthotropic elasto-plastic continuum model
with the Hill (1947) yield condition and characterized
by an elastic-perfectly plastic behaviour in tension and
compression has been considered.
1021
Table 4. Material properties used for the numerical models.
Element Mechanical properties
FRP-laminates Young’s modulus: EF = 160000 MPa
(L2TRU) Poisson’s ratio: νF = 0.3
clay brick Young’s modulus: Eb = 15279 MPa
(Q8MEM) Poisson’s ratio: νb = 0.2
Shear modulus: Gb = 6366 MPa
interface layer Tangential stiffness: kt = 4670 N/mm3
(L8IF) Normal stiffness: kn = 78891 N/mm3
Table 5. Parameters characterizing the yield domain.
Mechanical parameters Values
Compressive strength in x (y) direction 38.5 MPa
Tensile strength in x (y) direction 3.44 MPa
Shear strength 0.88 MPa
The adopted yield condition is an extension of the
von Mises failure criterion and is characterized by the
following formulation (see also DIANA User’s Guide;
DIANA, 2000):
where σ is the stress vector, σ ∗ (κ) is the reference
yield strength as a function of the internal state vari-
able κ, and P is the projection matrix which depends
on the yield strengths in the x and y directions, and
on the yield strength in shear. An associate flow rule
is adopted for the yield criterion. The parameters
reported in Table 5, selected for characterizing the
yield domain, have been deduced by the experimen-
tal tests performed on the clay bricks and described in
section 2.
3.3 Results
Two types of analyses have been performed: a linear-
elastic analysis where all the elements are elastic and
an incremental nonlinear elastic analysis where the non
linearity has been adopted only for the brick elements.
The first type of analysis has been performed with
the aim of capturing the elastic behaviour of speci-
mens. In Figure 7 have been compared the displace-
ments of the FRP deduced by the experimental tests
(circular symbols) and those obtained by the numer-
ical analyses (continue line). In particular, the figure
refers to the scheme 4 of the examined specimens con-
sidering a low value of the applied force (assumed
equal to 20% of the maximum force sustained by the
reinforcement) in order to examine the behaviour of
Figure 7. Comparison between numerical and experimental
values of FRP-displacements.

the specimen before delamination or damage of the

support, which could be considered elastic.
From the Figure it is possible to observe a good
agreement between the numerical and the experimen-
tal values of the FRP displacements particularly in the
zone of FRP-laminate near to the unloaded end. On the
contrary, the experimental values of the part of strips
near to the loaded end are greater than the numeri-
cal ones. This effect is probably due to the presence
of some damages which arise for low external loads
already and local effects orthogonal to the FRP strips
which arise near the vertical edge of bricks, do not
captured by a 2D model.
The second type of analysis has been performed
in order to examine the capability of the numerical
model to provide the maximum load sustained by the
reinforcement.
In Figure 8 are reported the curves in terms of
applied force vs. horizontal displacement of the FRP
obtained from the numerical analyses. In the same fig-
ures are reported the results obtained from the tests
(triangular symbols). In particular, the displacement
has been measured at the section located at 65 mm
from the free edge of the brick. The choice of the sec-
tion is due to the fact that it represents the location of
a strain gauge for all the specimens.
Figure 8 shows that the theoretical curves are in
good agreement with the experimental results. This
means that since the numerical model is based only on Figure 8. (a) Push-over analysis of the specimens:
the nonlinear behaviour of brick, it could be assumed scheme 1, (b) Push-over analysis of the specimens:
that the support is the main responsible of the nonlinear scheme 2, (c) Push-over analysis of the specimens: scheme 3
response of the tested specimens. Moreover from the and (d) Push-over analysis of the specimens: scheme 4.
plots it is also evidenced an increase of the maximum
load sustained by the reinforcement by increasing the the examined specimens particularly for the schemes
bond length. 2 and 3.
In Figure 9 is shown the comparison between the It is also important to notice that, the bond strength
maximum load obtained by the numerical analyses is not linear dependent on the bond length.
(triangular symbols) and the corresponding values In Figure 10 are reported the variation of the bond
obtained by experimental tests (vertical bars). From strength Fmax deduced from the experimental tests
the figure it is possible to notice that the adopted model and that obtained from the numerical analyses. In the
furnishes a good estimation of the bond resistance of same figure the dotted line indicates a variation of

1022

Figure 9. Maximum force sustained by the reinforcement.
Figure 10. Bond strength variation.
the bond strength linearly proportional to the bond
length of the reinforcement. From the figure it is clear
that the variation of the bond strength with the bond
length is strongly far from the linear variation and is
characterised by a smaller slope.
This is a further aspect that confirms the presence
of different interaction mechanisms between the FRP-
reinforcing and the support depending on the bond
length. Indeed, the increase of the bond length of the
FRP not only produces an increase of the zone which
resists to external loads but produces a variation in
terms of stress transfer between the reinforcement and
the support through the interface layer.
4 CONCLUSIONS
In this paper are discussed the results obtained by an
experimental campaign and a numerical study devoted
to analyse the delamination mechanisms of CFRP
laminates glued on clay bricks.
1023
From the experimental results, the influence of the
bond length on the bond strength and the delamination
mechanism has been examined in detail. In particular,
it was observed that after the detachment of the FRP
laminate only the brick support was damaged.
On the basis of the experimental evidence, a 2D
numerical model of the tested specimens has been
developed adopting an elasto-plastic behaviour for the
brick response.
The good agreement between the numerical and the
experimental results both in terms of strength and non-
linear behaviour have confirmed the reliability of the
proposed model.
It is clear that a more comprehensive study of the
bonding response would require the evaluation of a
fracture energy as defined in the theoretical formu-
lations reported in the codes. This means that first,
the numerical model would also provide the maximum
deformation in the brick and the post-peak behaviour,
and second, the experimental tests would be performed
by a displacement control procedure.
REFERENCES
Aiello, M.A. & Sciolti, S.M. 2005. Bond analysis of masonry
structures strengthened with CFRP sheets. Construction
and Building Materials. Vol.20: pp. 90–100
Casareto, M., Olivieri, A., Romelli, A. & Lagomarsino,
S. 2003. Bond behaviour of FRP laminates adherent to
masonry. In: Proceedings of the international conference
advancing with composites, Milano, Italy
Barbi, L., Briccoli Bati, S. & Ranocchiai, G. 2004. Anal-
isi sperimentale di campioni in mura-tura fasciati con
composito CFRP. in Atti del II Convegno Nazionale “Mec-
canica delle struttu-re in muratura rinforzate con FRP –
materials”, Venezia, Italy (in Italian).
CNR-DT200/2006. 2006 Guide for the design and
construction of externally bonded FRP systems for
strengthening existing structures. Materials, RC and
PC structures, masonry structures. National Research
Council, Rome-CNR
National Instruments LabVIEW 2006. Graphical Develop-
ment Platform for Design, Control and Test.
DIANA, 2000. Displacement analysis finite element
software. Version 9.1, TNO-Building Division, Delf,
The Netherlands.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Methodology for in situ application of hydraulic grouts on historic masonry

structures. The case of the Katholikon of Dafni Monastery

A. Miltiadou-Fezans, A. Kalagri, S. Kakkinou & A. Ziagou

Directorate for Technical Research on Restoration, Hellenic Ministry of Culture

N. Delinikolas
1st Ephorate of Byzantine Antiquities, Hellenic Ministry of Culture

E. Zarogianni & E. Chorafa

Directorate for Restoration of Byzantine and Post-Byzantine Monuments, Hellenic Ministry of Culture

ABSTRACT: Grouting can be a durable and mechanically efficient intervention technique, not only when
the grout composition is suitably designed, but also when the technique is correctly implemented following
an adequate methodology. Such a methodology has been developed for the in situ application of grout in the
Katholikon of Dafni Monastery in Athens. Hereby are presented the main features of this methodology. Apart
from the information regarding the masonry preparation for grouting, the necessary quality control of the grout
prepared at the worksite and the injection process, the importance of collecting specific data during the injection
process, which allow monitoring the movement of the grout into the masonry and estimating the grout volume
consumed in the masonry, is underlined.

1 INTRODUCTION mechanical properties of masonry elements, before

The Byzantine monastery of Dafni, already inscribed
in the world heritage list of UNESCO, is one of
the most important monuments of middle Byzantine
period, famous worldwide for the excellent mosaics of
its 11th century Katholikon (Delinikolas et al. 2003).
All the structures of the monastic complex and espe-
cially the Katholikon (main church of the Monastery)
have suffered severe damage during the September
1999 earthquake that affected the region of Attica.
Numerous cracks have appeared both on the three leaf
stone masonry walls and the vaulted roof, as well as
on the mural mosaics situated on their internal faces
(Miltiadou et al. 2004).
On the basis of the results of a series of research
programs, investigations and the structural restoration
study realized by the competent authorities of the Hel-
lenic Ministry of Culture, the application of adequate
hydraulic grouts injections was decided for the repair
and strengthening of the Katholikon (Miltiadou et al.
2003) as well as for the in situ conservation of its mural
mosaics (Chryssopoulos et al. 2003).
Before the application of this specific technique to
the Katholikon of Dafni Monastery, extended experi-
mental investigations were carried out in order (a) to
design adequate grout compositions, (b) to assess the
and after grouting, and finally (c) to evaluate the effect
of grouting on their mechanical properties (Vintzileou
et al. 2004, 2006, Miltiadou et al. 2006, Kalagri et al.
2007, Miltiadou et al. 2007).
However, the proper design of a grout composition
and its application cannot ensure on their own the suc-
cessful completion of the grouting intervention in situ.
Evidently, particular care has to be taken of the ade-
quate execution of the whole intervention on a daily
basis. To this end a specific application methodology
has been developed, based on the experience gained
from applications to various monuments and, in par-
ticular, to the Columns of the Opisthodomos of the
Parthenon of the Athens Acropolis (Miltiadou et al.
2005). This methodology has been proposed by the
structural restoration study (Miltiadou et al. 2003),
later tested on the aforementioned wallettes (Milti-
adou et al. 2006) and finally applied in situ, during
the structural restoration works.
The methodology comprises specific instructions
for a) the preparation of masonry for grouting and the
survey of exit and entrance tubes, b) the mixing proce-
dure, the equipment and the quality control of the grout
prepared at the worksite, c) the injection process and
the data to be collected, and d) the elaboration of these
data in order to monitor the grout movement in the

1025
interior of the structure, and the volume consumed per
masonry area. The need for the efficiency of the grout-
ing intervention to be controlled, after the completion
of the works, by means of adequate NDT and seismic
monitoring instrumentation, is also underlined.
In the present paper the research process followed
for the selection of the adequate grout will be first
briefly summarized, as it is in direct conjunction with
this paper, and then the most important elements of
the applied injection grouting methodology will be
presented and commented upon.
2 SELECTION OF THE GROUT COMPOSITION
The design of high injectability grouts was applied
on the basis of the methodology proposed by Milti-
adou & Tassios (2006). First, were taken into account
the performance requirements deriving from the struc-
tural restoration study (Miltiadou et al. 2003), and
then the following target values were set for the basic
mechanical properties of the grouted masonry: ten-
sile strength approximately double that one of the
masonry before grouting, and compressive strength
approximately equal to 3.0 MPa.
On the basis of the available literature (Vintzileou
& Tassios 1995, Tassios 2004), it was estimated that
the compressive strength of the grout at the age of
six months should lie between 6MPa and 10MPa; a
grout flexural strength of the order of 2-3MPa was
required. In addition, the physical-chemical properties
of the raw materials should be selected in a way that
the durability of the structure and its precious mosaics
would not be jeopardized. Finally, the grouts should
have high injectability capacity, so that, under low
pressure (∼0.075 MPa), they enter and fill fine voids
and cracks, with a nominal minimum width (Wnom )
equal to two tenths of millimeter.
According to literature (Miltiadou 1990,
Toumbakari 2002, Valluzzi 2000) and previous inves-
tigations carried out by the Directorate for Technical
Research on Restoration (DTRR, Hellenic Ministry
of Culture), two main categories of grouts could sat-
isfy injectability, strength and durability requirements:
(i) ternary grouts composed of a low cement content
(30%), lime 25% and pozzolan 45%, and (ii) hydraulic
lime – based grouts.
Thus, various grout mixtures, belonging to the
above two categories, were designed and tested (in
order to assess their physical, chemical and mechani-
cal properties) at the laboratory of DTRR. The use of
white Danish cement in the ternary grout was chosen,
due to its fineness, low alkali content and high sulphate
resistance. As far as the hydraulic lime – based grouts
are concerned, various types of natural hydraulic limes
(classified by EN 459 as NHL2, NHL3.5, NHL3.5-Z
1026
and NHL5), were tested in the laboratory of DTRR,
and grout formulations, with or without superplastizer,
were examined. The main results of the research are
reported in Kalagri et al. (2007).
In order to determine the injectability characteris-
tics, the penetrability, fluidity and stability of the sus-
pensions were fully examined in various water/solids
ratios, with or without additives. The grouts were pre-
pared by using an ultrasound dispersion mixer, assisted
by a mechanical device of low turbulence. The stan-
dardized sand column test method (NF P18-891, pr.
EN 1771), was used to check the penetrability and flu-
idity, along with the standard apparatus for testing the
fluidity (NF P18-358) and stability (NF P18-359) of
the suspensions. In each case, a time limit of 50 sec for
the sand column penetrability test (T36 ); an efflux time
of 500 ml of grout (td=4.7 ) shorter than 45 sec (Marsh
cone d = 4.7 mm fluidity test), and a maximum accept-
able limit of 5% for the bleeding were set initially
(Miltiadou 1990) for the laboratory investigations.
The compositions presenting satisfactory injectabil-
ity capacity were further tested to evaluate their
behavior to salt decay and estimate their mechani-
cal characteristics (compressive and flexural strength).
Six alternative grout formulations presenting similar
injectability, were injected at low pressure into twenty
eight cylindrical specimens, simulating the infill mate-
rial of three-leaf stone masonry, which were then
subjected to compression in different hardening ages
(Kalagri et al. 2007). After comparative evaluation of
the results, two grout compositions (the ternary grout
and a natural hydraulic lime NHL5-based grout) ful-
filled simultaneously the injectability, the strength and
durability requirements. Therefore, they were selected
to be applied to six three-leaf stone wallettes, simulat-
ing the masonry of the upper parts of the monument,
subjected either to compression or to diagonal com-
pression up to their maximum resistance, before and
after grouting (Vintzileou et al. 2006, Miltiadou et al.
2006).
The substantial (compressive and tensile) strength
enhancement of wallettes, the rather ductile behaviour
under diagonal compression (compared to that of
masonry grouted with the ternary grout), the physico-
chemical properties that ensure a durable intervention
and contribute to the protection of mosaics led to the
selection of the natural hydraulic lime- based grout for
the application in the Katholikon of Dafni Monastery.
Given that these hydraulic limes are characterized by a
relatively high percentage of available lime, the addi-
tion of pozzolan, in adequate proportion, is expected
to have beneficial effect. Thus, in order to improve the
hydraulic lime based grout, the addition of fine nat-
ural pozzolan (dmax < 75 µm) in various proportions
was investigated. The addition of a small percentage
of pozzolan (10%) was decided, on the basis of addi-
tional data deriving from porosity measurements, salt
Table 1. Composition and injectability characteristics of the
selected grout measured in the laboratory and in situ at the
first pilot preparation.

GROUT COMPOSITION
NHL5 (St Astier) 90%
Pozzolan 10%
Superplasticizer (1), (2) 1%
Water (1) 80%
GROUT PROPERTIES In lab In situ
T36 (sec) – Sand column 19–22
1.25/2.50 mm (voids
∼0.2–0.4 mm)
Bleeding <1% 1%
Apparent viscosity – td=4.5 (sec) In lab In situ
0 min after mixing 21 22
60 min after mixing (agitated) 23 25
Apparent density (gr/cm3 ) In lab In situ Figure 1. Historical pathology of the external east façade of
0 min after mixing 1.5050 1.4978 the Katholikon of Dafni Monastery. Uncolored joints still pre-
60 min after mixing (agitated) 1.4986 1.4870 serve the old Byzantine pointing mortars (Delinikolas et al.
2003).
(1) % of the solid phase of the grout.
(2) superplasticizer based on polycarboxylic ether.

durability tests and from in situ trials of the grout

formulation. (Miltiadou et al. 2007). The mix pro-
portions and the injectability characteristics of the
selected grout are presented in Table 1.
It has to be noted here that, due to the importance
of the monument, and on the basis of the results of
the laboratory and in situ pilot tests, using the work-
site materials and equipment (Table 1), the following
stricter limits were set for the quality control of the
grout in situ. For the fluidity test, a maximum efflux
time (td=4.7 ) of the order of 25 sec (fluidity test: 500 ml
of grout, Marsh cone d = 4.7 mm) has been considered,
as this limit covers even the measurements one hour
after mixing. As far as the bleeding is concerned, a
maximum acceptable limit of 3% was set, as this value
gives the possibility for covering the small differences
due to site conditions. Finally, the apparent density
values measured in situ had to be 1.50 gr/cm3 , with an
acceptable variation of ± 1% for the measurements on
grouts after mixing, and ±5% for the measurements on
grout selected from various exits during the injection
process (see also §4).
3 APPLICATION METHODOLOGY
The application of grouts comprises the preparation of
the mix to be injected, the execution of the injections
and, finally, cleaning and finishing the surface of the
grouted masonry.
The in situ application of injection grouting tech-
niques in the Katholikon of Dafni Monastery had to
be implemented without removing the old pointing
mortars that have survived on the external façades,
Figure 2. Fissured mosaic, situated in the upper part of the
sanctuary.
and, of course, without removing the mural mosaics
decorating its interior. Figure 1 presents the historical
pathology of the external east façade (Delinikolas et al.
2003). In all the uncolored joints, the old byzantine
pointing mortars are still preserved in situ. In Figure 2
a fissured mosaic in the upper part of the internal east
façade is shown. In order to protect these valuable ele-
ments, the injections had to be realized using entrance
and exit tubes of small diameter in these areas, while
the pressures developed had to be low, and absolutely
controlled, so that not to exceed the 1 atm. Further-
more, a methodology of monitoring the movement of
the grout in the masonry had to be applied in order to
ensure that the grout reached and filled the numerous
cracks and fissures. To achieve these goals, a spe-
cific application methodology has been proposed by
the relevant study (Miltiadou et al. 2003), to be fol-
lowed during the structural restoration works in the

1027
Katholikon of Dafni Monastery. The methodology has
been tested and further refined during the injections
of the aforementioned wallettes, and then applied to
the whole injection project for the repair of the mon-
ument’s masonries. The most important aspects of
this methodology are presented below, together with
a synthesis of indicative quality control data, concern-
ing its in situ application in the Katholikon of Dafni
Monastery.
3.1 Preparation of the masonry
a) After the removal of the renderings from the interior
of the building, all the areas that did not conserve old
mortars were prepared for deep re-pointing, while all
the deteriorated old mortars were consolidated in situ
by competent Conservators. Stone or brick or titanium
stitches were installed in order to key the areas with
severe cracks, dislocations or local reconstructions.
During this first step, a refinement of the detailed
survey (Delinikolas et al. 2003), presenting the joints
conserving old mortars in the external façade of the
monument, was made, and the possible positions for
drilling holes for the installation of grouting tubes were
clearly identified, as well as their possible respective
size, ranging between 1 to 10 mm.
b) A similar and more detailed procedure has been
undertaken by the competent Conservators for the sur-
vey and documentation of the state of preservation
of the mural mosaics including: historical and recent
pathology, detachments of substrata presented on man-
ual sonic maps or GPR maps, etc (Chryssopoulos et al.
2003, Côte et al. 2004, 2008). In this way the location
of adequate areas for the installation of grouting fine
tubes on the mosaics was defined.
c) On both faces of masonry and at adequate dis-
tances, holes were drilled in mortar joints, in order to
form a kind of grid. Given the fact that the thickness of
the walls is ∼80 cm, the distance of consecutive holes
has to be between 0.5 to 1.0 m. The nodes of the inte-
rior grid should be arranged at mid-distances of those
of the exterior grid, both horizontally and vertically.
Half of the holes should be located at a depth equal to
1/3 of the masonry thickness, and the rest of them to 1/2
of the masonry thickness. All the above arrangements
are necessary because they provide the possibility, not
only to inject the grout, but also to ensure the air exit
and the control of the grout overflow. Thus, the holes
had to be realized in such a way so as to reach (i) the
interface between external leaves and the infill mate-
rial, (ii) the infill material of the three leaf masonry
and (iii) the interior of the path of the crack.
In areas conserving old mosaics or frescoes or old
mortars, a denser grid of holes was formed on the inter-
nal façade of the masonry (when it was not decorated).
In case of decorated internal areas, it has to be noted
1028
Figure 3. View of the external masonry (a) after installation
of grout tubes, and (b) after their numbering.
that it is essential to drill the maximum possible num-
ber of holes, and install tubes (mainly tubes of small
diameter, used as exits) in order to avoid further dam-
ages, due to grout pressure or uncontrollable leakage
of the grout, and follow the grout path from this face
of the structure.
d) Subsequently, re-pointing or sealing of cracks has
taken place. Simultaneously, transparent plastic tubes
of a diameter of 1–10 mm (depending on the already
predefined diameter of the holes) were installed into
the drilled holes, and sealed to the wall using the same
mortar as that for re-pointing (usually a lime-pozzolan
based mortar). Both on the masonry and the mosaics,
all kind of cracks, including the small and fine ones
were sealed, in order to allow efficient injection, and
avoid uncontrollable leakage. On the mosaics, fine
tubes of adequate diameters were placed at two depths,
1.5 cm and 3 cm, in order to reach and control the
eventual masonry grout flow behind the substrata and
tesserae, respectively. In cracked areas, when possible,
tubes were installed to a certain depth in order to cre-
ate exits for the masonry grout. The protruding part
of all kinds of transparent plastic tubes should be of
approximately 0.5 m.The inner part of the tubes should
be perforated at two positions, while the edge of the
tube should be cut inclined (∼45◦ ), to allow the grout
flow through masonry, even if the edges of tubes are
in contact with an in situ material. In Figures 3 and 4
examples of preparation for grouting of masonry and
mosaics are shown respectively.
Plastic tubes reaching different depths in masonry
should be adequately marked and reported in drawings
or sketches. They should also be numbered consecu-
tively from bottom to top of masonry. In the case of the
Katholikon of Dafni Monastery, due to its large dimen-
sions and in order to avoid confusion resulting from
this way of numbering (tubes with numbers consisting
of many digits), a predefined way of tube numbering
was followed.

Figure 4. View of a mosaic during the preparation for
grouting.
A special code was adopted, comprising letters,
relating to the region of the masonry, accompanied
by numbers in an ascending sequence. The numbering
and surveying of grouting tubes is extremely useful, as
it is absolutely necessary to record all the entrances and
corresponding exits during grouting, together with the
volume consumed per entrance, and the pressure vari-
ations at the entrance of the grout, in order to be sure
that the damaged areas of the monument have been
reached and filled with the grout. Finally, perforated
plastic bags were placed on the edge of each tube, in
order to protect the wall from grout leakage, and avoid
extensive surface cleaning, thus protecting the patina
of external facades.
3.2 Worksite equipment
The main characteristics of the grouting equipment
used in the worksite were in accordance with the
requirements prescribed in the study (Miltiadou et al.
2003) and consisted of the following units: a high tur-
bulence colloidal mixer, an agitator, a grouting pump
equipped with control system and manometer, all the
necessary grout lines, a grout recorder, nozzles of
adequate diameter, proportionate to the ones of the
entrance tubes. A manometer and a triode valve were
installed near the nozzle, in order to control continu-
ously the pressure in the entrance of the wall, and be
able to take immediately corrective measures in case
of a sudden overpressure. Further description of the
worksite equipment is beyond the scope of this paper.
However, it is very important to underline the impor-
tance of using a grout recorder when injecting masonry
structures. This equipment permits to record automat-
ically and in function with time, the pressure and the
flow rate, as well as, when suitably programmed, the
grout volume consumed during a given period of time
or even per each injection entrance. It has to be noted
that this recorder has to be specially manufactured for
1029
masonry structures (due to the relatively low flow rates
and pressure values).
3.3 Grout preparation and injection process
In general, before mixing starts, all materials have to
be separately weighted and be ready to be introduced in
a high turbulence colloidal mixer (1,500–2,000 rpm),
as the mixing time is rather short to allow for delays.
In this case, the water and the superplasticizer were
placed first into the mixer and then the solid materials
followed, in the order of increasing grain size (poz-
zolan, hydraulic lime). The mixing time was in total 4
minutes (2 min/solid constituent). As the mixing pro-
cedure is a very important parameter for the production
of a stable and fluid grout, the accurate timekeep-
ing was underlined. Another parameter that affects the
injectability characteristics of the grout is the ambi-
ent temperature. It has to be mentioned that the whole
grouting procedure was stopped, when this temper-
ature reached extreme values (<10◦ C and >35◦ C).
After mixing, the grout was discharged into an agitator,
so as to ensure a continuous process of injection.
Before starting the grouting application on the mon-
ument, a trial grouting preparation was performed, in
order to check the skills of the technical staff, the ade-
quacy of the equipment, as well as the injectability
characteristics of the grout. In this way, it was ensured
that the grout prepared in situ presented the same
characteristics with those of the grout prepared and
investigated in the laboratory (Table 1).
Injections started from the bottom (proceeding
along the length of the element to be grouted) to the
top. When grouting was in process in one entrance
tube, the grout supply should not be interrupted unless
pressure at the nozzle reached the maximum accepted
value of 0.5–1 atm. Interruption of the injection before
exhausting the potential of a tube to consume more
grout, may result in failure of filling all the intercon-
nected voids. As a matter of fact, if one tries to repeat
injection from an entrance abandoned even a few min-
utes earlier, injection is rarely feasible, as the entrance
has already been blocked. During injection, pressure
was constantly checked at the entrance to the masonry.
Its upper limit should remain within the range of
0.5–1.0 atm. Only in lower areas, without any old mor-
tar or mosaics or frescoes, the pressure was allowed
to reach 1.5 atm, while in the areas bearing heavily
damaged mosaics, the lower upper limit of 0.5 atm
was kept. In cases of sudden overpressure, the use of
the triode valve permitted the immediate deviation of
the grout to a plastic vessel, thus avoiding harming
the masonry or the connections of the pipes of the
system. During the whole procedure, competent Con-
servators were present, when grouting was applied in
areas adjacent to mosaics or frescoes.
After the inflow and/or outflow of the grout had
been terminated, the tubes were firmly tied and fixed
upwards, since grout should be kept under pressure
inside masonry until its hardening is completed. In
some cases, when overflow of grout had taken place at
locations that no tubes had been installed, the leakage
was stopped with the help of pozzolan powder, or an
adequate paste (pozzolan paste, clay paste or paper
pulp, etc.). Cement or lime was absolutely avoided.
After that, the masonry was immediately cleaned.

4 GROUTING QUALITY CONTROL

4.1 Grout quality control tests

Fluidity, stability and apparent density of the mixture
were tested at the worksite at least twice per day, in Figure 5. Apparent viscosity values of grout measured in
order to ensure the quality of the intervention. All the situ after mixing, twice per day during the whole grouting
tests results were noted on a specially prepared daily period.
calendar, together with the total number of batches
prepared.
4.1.1 Fluidity – Apparent viscosity test
The efflux time of 100 ml, 500 ml and 1000 ml of
grout through a Marsh cone having a nozzle-diameter
equal to 4.7 mm was measured. The total efflux time
of 500 ml of grout had to be of the order of 25 sec,
according to the limits set for this specific case (see
also § 2). In Figure 5 the average values of grout appar-
ent viscosity measured in situ after mixing (twice per
day, during the whole grouting period, 20 weeks in
total), are presented. These values were very close
to the values of the laboratory and pilot tests, and
lower than 25 sec. The small variations were daily
assessed, together with the bleeding and density results
Figure 6. Bleeding values of grout measured in situ after
were linked to worksite conditions (temperature, new mixing, twice per day during the whole grouting period.
delivery of superplastisizer, etc).

4.1.2 Stability – Bleeding test

The grout was sealed into three transparent volumetric
cylinders of 100 ml (diameter 25 mm, height 25 cm)
and after remaining in a shadowed place for three
hours, bleeding water volume was measured.The grout
bleeding was given by the average value of the three
measurements, and had to be <3% and in no case
>5%, according to the limits set for this specific case
(see also § 2). The measured bleeding variations were
considered generally acceptable, given the fact that the
grout was prepared in the worksite (Fig. 6), and tak-
ing into account that, the first six weeks, the ambient
temperature was relatively high due to summertime.

4.1.3 Stability – Apparent density test

After mixing, several measurements of apparent den-
sity versus time were made. Apparent density was
determined by weighting known grout volume (50 ml),
collected from the same position of a 2000 cc volumet-
ric tube (usually at the 2/3 of its height). The apparent
density values measured in situ had to be 1.50 gr/cm3 ,
with an acceptable variation of ±1%. As shown in
Figure 7. Density values measured in situ after mixing twice
per day during the whole grouting period.
Figure 7, all density values measured in situ during
the whole project satisfied this limit.
Furthermore, except for the aforementioned qual-
ity tests of stability and fluidity, of the grout after

1030
Table 2. Apparent density values of grout outflows. Table 3. Evolution of strength of the in situ prepared grout.

Grout Density (gr/cm3 ) Age (days)

No. of No. of 28 90 180 270

entrance exit After Variation
tube tube mixing Outflow (%) Flexural strength (MPa) 1.01 2.80 2.41 2.20
Compressive strength (MPa) 2.02 5.26 6.04 6.33
AE74 AE78 1.4885 1.5254 2.48
AM18 AE178 1.5056 1.5298 1.61
AM288 AM276 1.5013 1.5126 0.75
BE68 BE65 1.4936 1.5024 0.59
BE86 BE3 1.4858 1.5015 1.06
BE517 BE493 1.5022 1.5240 1.45
NE36 NE43 1.5023 1.5298 1.83
KP405 KP385 1.4989 1.5143 1.03
KP404 KP328 1.4983 1.5196 1.42
TH431A TH422 1.5048 1.5152 0.69
D45/1155 X260 1.5048 1.5068 0.13
T358 T361 1.4982 1.5045 0.42
K251 K260 1.5038 1.5036 −0.01
K116 K109 1.4809 1.4953 0.97

mixing, the apparent density of the grout collected

from selected exit positions was also controlled, where
it was possible (42 grout outflows were checked). The
acceptable density variation was set at <5%.
This experimental procedure is essential for con-
trolling the stability, fluidity and penetrability of the
grout. Table 2 presents indicative density values of
grout after mixing, and of grout selected from corre-
sponding exits. It has to be noted that even in the cases
the exit of grout had been taken place at a distance
of two or three meters, the variation of density was
retained at <5%. This has shown that the grout has
retained its properties during its movement through
the porous masonry materials, thus confirming both
the quality of the grout (e.g. segregation or water loss
due to absorption, etc) and of its application.
4.1.4 Mechanical properties test
Flexural and compressive strengths of the grout were
tested at the age of 28, 90, 180 days etc, on prismatic
specimens (40 × 40 × 160 mm), which were prepared
at the worksite at certain time intervals. Table 3 sum-
marizes the results of mechanical properties on the
specimens tested until now. Taking into account that
the specimens have been taken and cured in the work-
site the very first days, the strength values are consid-
ered satisfactory, compared to those of the laboratory
study (Miltiadou et al. 2007).
4.2 Application quality control
In order to control better the injection process and
to ensure the quality of the intervention, during the
whole project, entrances and corresponding exits of
Figure 8. External east façade – Schematic presentation of
grouting consumption.
the grout are recorded on a time basis, together with
any change of pressure at the entrance to the wall.
All these data are noted on specially designed daily
calendars, with reference to the area of grouting
application. These manually collected data and those
collected by the Grout Recorder (pressure, flow rate
and grout volume consumed) are then combined, and
the grout volume corresponding to the various groups
of entrances and exits is determined, together with the
total volume of the grout consumed during the cor-
responding day. All the recorded data are organized
in tables and after appropriate processing the volume
of the grout consumed into the masonry is extracted.
Then the data collected are reported on the aforemen-
tioned series of drawings presenting the positions and
numbers of the grouting tubes. Thus a good estimation
of the grout movement and consumption in relation to
the various regions of the structure can be achieved.
The aforementioned procedure was followed during
the whole project in the Katholikon.
In Figures 8 and 9 the corresponding drawings for
the internal and external façade of the east wall of the
monument are presented. One can easily recognize the
areas with high grouting consumption, and those with
a lower one. The tubes with no consumption are also
noted. Thus for this specific wall the volume of the
grout consumption was estimated to reach 6.5% of the
total volume of the wall. In the case of the west wall of
the monument, which was reconstructed during past
restoration interventions (Delinikolas et al. 2003), the

1031

Figure 9. Internal east façade – Schematic presentation of
grouting consumption.
estimated grout consumption attaining 2.5% of its total
volume. The elaboration of data is still in progress.
Furthermore, such drawings, consist the “as built-
drawings” of this non visible intervention, and give the
possibility of an overall assessment of the grouting
intervention. In this way, grouting process is better
monitored during its application and the quality of the
intervention is ensured.
The effectiveness of the grouting interventions
should be tested also a certain period after the accom-
plishment of the works, by applying appropriate non-
destructive techniques, in combination, if permitted,
with semi-destructive techniques. In earthquake prone
areas, the comparison, before and after grouting, of
the records of a seismic monitoring system installed
on the structure can also give important information
about the grouting effect.
In the case of the Katholikon of Dafni Monastery
sonic transmission measurements and sonic tomogra-
phies were undertaken in selected areas. The results
obtained until now confirmed the efficiency of grout-
ing (Côte et al. 2004, 2008), thus proving that both
the grout and its application were adequately designed
and implemented. Finally, the seismic monitoring
data (before and during grouting), have also shown
the grouting effect on the overall behaviour of the
monument (Mouzakis et al. 2008).
5 CONCLUSIONS
The holistic design of grouts based on rational criteria,
laboratory and in situ pilot tests, led to an optimum
composition, exhibiting the adequate stability, fluidity
and injectability characteristics, as proven by the in
situ quality control tests.
The proposed grouting application methodology
gives the possibility for a more rational and fully
controlled implementation of injections, not only to
ordinary structures, but also to important monuments
1032
bearing mosaics, frescoes and old mortars, that have
to be preserved in situ.
The whole procedure and quality controls can be
easily applied by qualified scientific and technical per-
sonnel, and allow for corrective measures to be taken
during the project. Moreover, the collection and elab-
oration of all the proposed data during the works is
absolutely necessary for an overall assessment of the
grouting application.
This assessment should be further supported, by
means of non destructive techniques especially in case
of important historic structures.
ACKNOWLEDGEMENTS
The structural restoration works were realized by the
Directorate for Restoration of Byzantine and Byzan-
tine Monuments. The quality controls and the collec-
tion and elaboration of in situ data were undertaken
by the Directorate for Technical Research on Restora-
tion. The project was included in the Operational
Program “CULTURE”. It was co-funded by the Euro-
pean Regional Development Fund (ERDF-75%) and
by National Funds (25%).
REFERENCES
Chryssopoulos D., Anamaterou L., Georganis F. 2003. Docu-
mentation study for the mosaics of the Katholikon of Dafni
Monastery after the 1999 earthquake. Directorate for the
Conservation of Ancient Monuments, Hellenic Ministry
of Culture (in Greek).
Côte Ph., Dérobert X., Miltiadou-Fezans A., Delinikolas N.
2004. Mosaic-grouting monitoring by ground-penetrating
radar. Proc. Int. Sem. on Struct. Anal. of Hist. Const. Proc.,
Padova (IT), nov.
Côte Ph., Dérobert X., Miltiadou-Fezans A., Delinikolas
N. Durand O., Alexandre J. Kalagri A., Savvidou M.,
Chryssopoylos D., Anamaterou L, Georganis, F. 2008.
Application of non-destructive techniques at the Katho-
likon of Daphni Monastery for mapping the mosaics
substrata and grouting monitoring. Submitted to the 6th
intern. conf. on StructuralAnalysis of Historical Construc-
tions, Bath, UK.
Delinikolas N., Miltiadou-FezansA., Chorafa E., Zaroyianni E.
2003. Study on restoration of the Katholikon of Dafni
Monastery: Phase A – Architectural and historical Survey.
Hellenic Ministry of Culture (in Greek).
Kalagri A., Miltiadou–Fezans A., Vintzileou E. 2007. Design
and evaluation of hydraulic lime grouts for the strengthen-
ing of stone masonry historic structures. Proc. int. symp.
on Studies on Historical Heritage, Antalya, Turkey, 16–21
September, 2007.
Miltiadou A. 1990. Etude des coulis hydrauliques pour la
réparation et le renforcement des structures et des monu-
ments historiques en maçonerie. PhD Th., ENPC Pub. in
1991 by LCPC ISSN 1161-028X, Paris, France.
Miltiadou-Fezans A., Delinikolas N., Chorafa E.,
Zaroyianni E. 2003. Study on restoration of the Katholikon
 of Dafni Monastery, Phase A – Structural Survey, Analysis
and remedial measures. Hellenic Ministry of Culture (in
Greek).
Miltiadou-FezansA.,Vintzileou E., Delinikolas N., Zaroyianni
E., Chorafa E. 2004. Pathology of the Dafni Monastery:
survey, monitoring of cracks, interpretation and numerical
verification. Proc. 4th inter. Sem. on Structural Analy-
sis of Historical Constructions, Vol. 2, Padova, 10–13
November.
Miltiadou-Fezans A., Papakonstantinou E., Zambas K.,
Panou A., Frantzikinaki K. 2005. Design and application
of hydraulic grouts of high injectability for the struc-
tural restoration of the column drums of the Parthenon
Opisthodomos., Proceedings of STREMA IX, WIT Press,
pp. 461–471.
Miltiadou-Fezans A., Tassios T.P. 2006. New rational criteria
for the holistic design of hydraulic grouts. 1st Conference
of the Society for Research and Promotion of Scientific
Restoration of Monuments (in Greek).
Miltiadou-Fezans A., Vintzileou, E., Papadopoulou, E., Kala-
gri, A. 2006. Mechanical properties of three-leaf stone
masonry after grouting. Proc. 5th inter. Sem. on Structural
Analysis of Historical Constructions, New Delhi, India.
Miltiadou-FezansA., KalagriA., Delinikolas N. 2007. Design
of hydraulic grout and application methodology for stone
masonry structures bearing mosaics and mural paintings:
the case of the Katholikon of Dafni Monastery. Proc. int.
symp. on Studies on Historical Heritage, Antalya, Turkey,
16–21 September, 2007.
1033
Mouzakis Ch., Miltiadou-Fezans A., Touliatios P.,
Delinikolas N., Dourakopoulos J. 2008. Earthquake
based condition monitoring of the Katholikon of Dafni
Monastery. Submitted to the 6th intern. conf. on Structural
Analysis of Historical Constructions, Bath, UK.
Tassios T.P. 2004, Rehabilitation of three-leaf masonry. In
Evoluzione nella sperimentazione per le costruzioni, Sem-
inario Internazionale 26 Settembre–3Ottobre, CIAS
Toumbakari E.E. 2002. Lime-Pozzolan-Cement grouts and
their structural effects on composite masonry walls. PhD.
Thesis, Department of Civil Engineering, KULeuven.
Valluzzi M.-R. 2000. Comportamento meccanico di murature
storiche consolidate con materiali e tecniche a base di
calce. PhD Th., Univ. of Trieste.
Vintzileou E., Tassios T.P.1995. Three leaf stone masonry
strengthened by injecting cement grouts. Journal of Struc-
tural Engineering, ASCE, Vol. 121, No 5.
Vintzileou E., Miltiadou-Fezans A., Palieraki V.,
Delinikolas N. 2004. The use of radar techniques and
endoscopy in investigating old masonry: the case of Dafni
Monastery. Proc. of 4th intern. sem. on Structural Analysis
of Historical Constructions. Modena C., Lourenco P.B.,
Roca P. (ed.), Rotterdam, Balkema, Vol. 2, pp. 351–360.
Vintzileou E., Miltiadou-Fezans A., Vrouva A., Anagnos-
topoulou S. 2006. Mechanical properties of three-leaf
stone masonry. Proc. 5th intern. Conf. on Structural
Analysis of Historical Constructions, New Delhi, India.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

High reversibility technique for in-plane stiffening of wooden floors

N. Gattesco
Department of Architectural and Urban Design, University of Trieste, Italy

L. Macorini
Department of Civil and Environmental Engineering, University of Trieste, Italy

ABSTRACT: A simple technique to increase the in-plane stiffness of ancient wooden floors is herein proposed
and discussed. The technique consists of using some nail plates for connecting longitudinally the timber boards
so as to prevent the slip among them. In such a way a stiff horizontal diaphragm can be obtained, which is
able both to restrain the masonry walls against the out-of-plane displacements caused by seismic excitation
and to transfer horizontal forces to the shear walls. Such a technique is also characterized by high reversibility
and low invasiveness, as requested by the conservation requirements for monuments and historical buildings.
Experimental tests both on connection samples and on a real size strengthened floor were carried out. The
former allowed the load-slip relationship of the nail plate to be determined, whereas the latter enabled to check
the effectiveness of the proposed technique. The results evidence good in-plane stiffness for the strengthened
floor, which is more than one order of magnitude higher than that of the unstiffend floor.

1 INTRODUCTION 2001, Giuriani 2006). This solution assures, through a

In ancient masonry buildings the floors and the roofs
are mainly made of timber joists and wooden planks.
They normally have large flexural deformability and
scarce in-plane stiffness. This poor structural perfor-
mance strongly influences the seismic behaviour of the
historical masonry buildings which is often not ade-
quate and it has led, even in the recent past, to heavy
damage, building collapse and losses of human life.
In order to guarantee an effective structural
behaviour in the case of seismic action, the floor sys-
tems should prevent the out of plane displacements of
the masonry walls and distribute the seismic forces to
all the shear walls. This is essential to avoid the early
collapse of some masonry piers affected by dangerous
out-of plane bending moments.
Therefore the ancient wooden floors have very
often to be strengthened and stiffened. Moreover, if
the building has an historical value, the intervention
techniques have to respect the conservation require-
ments (Venice Charter 1964), so these solutions have
to be characterized by low invasiveness and high
reversibility.
The current most employed stiffening and strength-
ening technique corresponds to the use of a rein-
forced concrete slab connected to the timber decking
(Piazza & Turrini 1983, Ronca et al. 1991, Gattesco
composite action, increased flexural stiffness and
strength and it allows the floor to behave as a stiff
diaphragm. However this strengthening technique has
some drawbacks (Gattesco & Macorini 2006), in par-
ticular it is frequently refused by the Cultural and
Environmental Assets Service because it is consid-
ered with low reversibility and high invasiveness and
therefore it can not be used in the case of buildings of
historical value.
In order to solve this critical drawback, some “dry”
strengthening techniques have been investigated in
the last years. Even though some solutions employ-
ing planks, timber panels, steel sheets and steel plates
(Giuriani 2004, Modena et al. 2005, Gattesco &
Macorini 2006, Gattesco et al. 2007) have already
been studied and proposed, the problem concerning the
strengthening and stiffening of ancient wooden floors
has yet to be completely solved. Nowadays there is a
need to carry out experimental analyses to investigate
the structural behaviour of wooden floors reinforced
by means of different strengthening techniques, in
order to provide the designers with simple formula-
tions for determining the effectiveness of the different
strengthening solutions.
The aim of the research work is to investigate a
new “dry” technique for in-plane stiffening of ancient
wooden floors, focusing on the evaluation of its

1035
Figure 1. Wooden boards connected with nail plates.

effectiveness in terms of stiffness increase with respect

to unstiffened floors.

2 STIFFENING TECHNIQUE

The proposed technique to increase the in-plane stiff-

ness of the wooden floors of historical buildings Figure 2. Stiffening technique: (a) schematic view of the
concerns the use of punched metal plates (gang nails) floor, (b) detail of the connection to the masonry.
to connect longitudinally the boards (Fig. 1), so to
prevent the slip due to shear forces.
Different fasteners may be used according to the
application technique: a) nail plates embedded into the
wooden board using special devices with a hydraulic
jack, b) knuckle nail plates applied by hammering in
the “knuckle nail”. The former technique requests an
appropriate device able to force the nail plate up to
achieve the complete penetration of the nails in the
wooden boards; a hydraulic jack is needed which is
arranged on a special steel frame. The latter appli-
cation technique is simpler but a lower efficiency of
each fastener may be obtained. In the study the first
application technique was considered.
In order to provide the floor with adequate in-plane
stiffness and resistance a perimetric L-shaped steel Figure 3. Push-out specimen for tests on fasteners.
profile is used (Fig. 2a). In such a way the steel profiles
perpendicular to horizontal action constitute the two to transfer the horizontal forces to the shear walls
chords needed to resist bending moment, the stiffened (shear-effect) (Gattesco & Del Piccolo 1997).
wooden floor supports shear forces and the steel pro-
files parallel to horizontal action transfer the in-plane
forces to the shear walls. 3 TESTS ON NAIL PLATE FASTENERS
The perimetric L-shaped steel profiles are con-
nected to the floor joists and to the planks through Some push-out tests were carried out to determine the
16 mm screws, so to avoid the relative slip among load-slip relationship of the connection, which is nec-
them, which is due to shear forces acting in the deck essary to design the strengthening intervention for the
plane. The steel profiles are also effectively connected wooden floor.
to the masonry walls by means of 16 mm driven dowels A nail plate 75 × 250 mm, 1.5 mm thick with 80
injected with cement grout (Fig. 2b). These connectors nails 16 mm length was considered in the study.
allow both to prevent the out-of-plane displacements The specimens were arranged by using three boards,
of the masonry walls at floor level (tie-effect), that 450 mm length, 23 mm thick, jointed with two
are caused by seismic excitation of the building, and nail plates (Fig. 3). In order to avoid unexpected

1036
 24
20
16
Shear load [kN]
12
8 stone masonry walls (50 cm thick).
sample C2
sample C3
4
sample C4
0 tre. A boarding with an average thickness of 23 mm is
0 2 4 6 8 10
slip [mm]
Figure 4. Experimental load-slip curves of the single fas-
tener (nail plate 75 × 250 mm, 1.5 mm thick).
Table 1. Experimental values of stiffness in service kps and
maximum shear load Fpu for a single nail plate.
Sample C2 Sample C3 Sample C4
kps (kN/mm) 9.31 8.91 9.16
Fpu (kN) 19.8 19.0 21.0
out-of-plane displacements of the specimen, due to the
eccentricity of the fastener with respect to the mid-
plane of the boards, a second group of three boards
jointed with two nail plates was connected to the first
one by interposing two wooden listels. In such a way
a symmetric test sample was obtained.
The specimens were provided with two displace-
ment transducers to measure the slip among central
and lateral boards. The load was applied by means of a
hydraulic testing machine at a rate of 0.1 kN/s. In each
test the results of each single nail plate were achieved
by averaging the results of the whole specimen.
Three equal samples, identified as C2, C3 and
C4, were tested. In Figure 4 the experimental curves
concerning the single fastener shear load against the
relative slip among the boards are plotted. The Figure
points out very close curves. The values of the stiff-
ness in service (kps ), evaluated according to EN 12512
(2001), and the maximum shear load of a single nail
plate 75 × 250 × 1.5 mm (Fpu ) are reported in Table 1.
The average stiffness is equal to 9.13 kN/mm, while the
average maximum shear resistance is equal to 19.9 kN.
4 FLOOR IN-PLANE STIFFNESS
In order to point out the enhanced structural perfor-
mance achieved by using the proposed strengthening
1037
technique, a simple numerical investigation was car-
ried out. In the numerical analyses the in-plane stiff-
ness of ancient wooden floors, strengthened with
different techniques, was assessed and compared. The
floor of an existing ancient masonry building (19th
century water mill building) was assumed as refer-
ence for the study. The main block of the building is
rectangular (7.90 × 14.20 m) and has three floors. The
vertical structure corresponds to perimetric multi-leaf
The considered floor (Fig. 2a) consists of parallel
timber joists (Eastern Alps spruce), with cross section
170 × 200 mm, net length 6.9 m set at 550 mm on cen-
connected to the beams extrados through a couple of
3 × 70 mm nails at each board-joist intersection. The
nails of each couple are spaced 150 mm.
For the same floor three different strengthen-
ing techniques were considered: a) nail plates
(75 × 250 mm) to connect boards against longitudi-
nal slip (fastener spaced 550 mm), b) diagonal steel
strips (60 × 5 mm) to prevent in-plane shear defor-
mations (Gattesco & Macorini 2006), c) reinforced
concrete slab (50 mm thick) connected to timber joists
through steel dowels (Piazza & Turrini 1983). In the
former two cases a steel perimetric angle element
(100 × 200 × 10 mm), adequately connected to the
timber members and to the masonry walls, is also pro-
vided. The diagonal steel strips (case b) are welded
to the perimetric steel angles. The concrete slab is
anchored to the masonry walls through driven dowels
injected with cement grout.
For the numerical simulation of the floors
behaviour, a grid model of beam elements crossing
perpendicularly was considered. In such a model the
first order of elements represents the timber joists,
whereas the perpendicular elements represent the
boards. The grid mesh was 550 × 200 mm, which
corresponds to the space among beams and boards,
respectively.
The relative horizontal translations among joists
and boards in correspondence of their intersections
were partly prevented by using linear springs that sim-
ulate the shear stiffness of nails. The torsional stiffness
of beam elements was assumed negligible; the relative
rotation at intersections between joists and boards was
partly prevented through a rotational spring able to
simulate the effect of the couple of nails that connect
the boards to the joists.
The translational and rotational stiffnesses of linear
and rotational springs were evaluated on the basis of
the shear stiffness of the single nail, which was deter-
mined according to EN 1995-1-1 (2004). This stiffness
was obtained through the relation
where ρk is the wood density in kg/m3 and d is the
nail diameter. The density of the timber of the joists
and boards was assumed equal to 400 kg/m3 . Hence
the stiffness of the linear springs, representing the
shear stiffness of two nails 3 mm diameter, is equal to
Kh = 2.4 kN/mm. The stiffness of the rotational spring,
for a nail spacing a = 150 mm, is equal to

The perimetric L-shaped steel elements were also mod-

elled as beam elements. In the first strengthening
technique (a) the effect of the nail plates that pre-
clude the longitudinal slip among boards was modelled
by using shear springs that connect longitudinally the
beam elements representing the boards. For the stiff-
ness was assumed the average value achieved through
experimental tests (Section 2 − kps = 9.13 kN/mm). In
the second strengthening technique (b) the diagonal
steel strips were modelled as truss elements. Whereas
in the floor strengthened by means of a concrete slab
(c), four nodes “plane stress” finite elements were
used. The in-plane stiffening contribution of timber
elements was neglected because it is very low with
respect to that of the concrete slab.
In the numerical analysis linear elastic behaviour
was assumed for materials and for spring elements.
For timber a longitudinal elastic modulus equal to
Figure 5. Numerical models for wooden floors: (a) stiffened
11000 MPa and a shear modulus equal to 690 MPa
with nail plates, (b) stiffened with diagonal steel strips.
were considered, whereas the elastic modulus for steel
and concrete were assumed equal to 207000 MPa and
30000 MPa, respectively.
Due to symmetry, half of the floor was considered Table 2. In-plane stiffness parameters of analyzed floors.
(6.9 × 6.6 m) in the numerical study. The structural
models for the floor and the loading schemes for the Displacement Stiffness Stiffness
Floor type (mm) (N/mm) ratio
first two strengthening techniques (a, b) are shown in
Figures 5a, b. Not strengthened 71.80 139 1.0
The structural response of the four floor models, With nail plates 3.63 2755 19.8
not stiffened, stiffened with nail plates, stiffened with With steel diagonals 1.52 6579 47.3
diagonal steel strips, stiffened with concrete slab, was With concrete slab 0.05 188679 1357
evaluated by applying a load F = 10 kN, distributed
on all nodes of the mid-span section of the floor. All
displacements in the loading direction of the nodes
in correspondence of the extremity of the floor were
restrained. Moreover all the displacements perpen- which is the ratio between the stiffness of the rein-
dicular to the loading direction of all the nodes of forced floor and the stiffness of the unstiffened floor.
the mid-span section were also restrained in order to By analysing these results it can be pointed out that
simulate the effect of the second half of the floor. both the “dry” strengthening techniques (nail plates,
The results of the numerical simulations are diagonal steel strips) provide a significant increase of
reported in Table 2. In particular the displacement of the floor in-plane stiffness that is from 20 to 50 times
the nodes at mid-span section in the direction of the higher than that of the unstiffened floor. Even though
load and the shear stiffness of the floor are shown. The these enhanced stiffnesses are considerably lower than
shear stiffness is simply evaluated dividing the applied that guaranteed by the concrete slab, they are ade-
force by the mid-span displacement. In the last column quate to prevent the out-of-plane deformations of the
of the Table is reported the stiffness increment ratio, masonry walls at floor levels (Gattesco et al. 2007).

1038
5 SHEAR TEST ON A FULL SCALE FLOOR

A full scale wooden floor stiffened through the tech-

nique described in Section 2 was subjected to a system
of in-plane loading so to point out its shear behaviour.

5.1 Specimen details

A wooden floor similar to those than can be found in
many ancient buildings was built in the Laboratory
for Testing Materials and Structures of the Univer-
sity of Trieste. The timber joists, with a cross section
160 mm wide and 180 mm depth, were taken from an
old building, whereas new boards, 23 mm thick, were
used for the deck.The floor specimen, 4.0 × 4.0 m, was
made of seven parallel timber joists, equally spaced,
and 21 equal boards, which were connected to the
joists through couples of nails (3 × 70 mm) at each
board-joist intersection.
Figure 6. Full scale floor specimen.
The boards were connected one another by means
of six nail plates (75 × 250 mm, 1.5 mm thick) equally
spaced along the board length. The nails of the fasten-
ers were forced to penetrate completely in the boards
by using a special steel frame and a hydraulic jack.
In the floor perimeter U-shaped steel elements were
arranged so as to simulate the perimetric L-shaped
steel elements used to connect the floor to masonry
walls (Fig. 2). These elements were connected to the
timber members through 16 mm screws. Finally the
steel elements were pin jointed in the corners, so to
allow free relative rotation (Fig. 6).
The plane of the specimen was arranged in ver-
tical; the bottom steel element was anchored to the
stiff concrete floor of the Laboratory (concrete plate
1.0 m thick) through a couple of large diameter bolts.
At the top the specimen was restrained against out-
of-plane displacements by means of two couples of
guides provided with a thin layer of PTFE (polyte-
trafluoroethylene) so to reduce friction during the test.
The guides were supported by two cantilevers fixed on
stiff steel columns.

5.2 Experimental apparatus

In the test a cyclic horizontal force was applied
at the top of the floor specimen so to subject the
boarding system to shear. The horizontal force was
applied through an electro-mechanic actuator (500 kN,
±300 mm stroke), governed via software through an
electronic unit system. The actuator is mounted on a
strong steel frame, which is effectively anchored to the
stiff concrete floor of the laboratory. In Fig. 7 a view
of the experimental apparatus is shown.
The load applied was measured by means of a
load cell mounted on the actuator. The relative dis-
placements between some significant points of the
specimen were surveyed through seven linear variable
Figure 7. Experimental apparatus view.
displacement transducers (LVDT). In particular the
displacement at the top of the specimen (S1) was mea-
sured through a rotative transducer with maximum
elongation of 500 mm. Two rotative tansducers were
used to survey the diagonal length variation of the
specimen (S6 ÷ S7). Two linear transducers (25 mm
maximum elongation) were used to check the vertical
displacements of the bottom part of the specimen (S2
right side, S3 left side), and another linear transducer
was also used to check the horizontal displacement of

1039
 70 60
(a)
60 50
50 40
40 30
30 20

Shear load [kN]

20
Load [kN]

10
10
0
0
-10
-10
-20
-20
-30
-30
-40 -40
-50 -50
3.0 m from the bottom
-60 -60
-40 -30 -20 -10 0 10 20 30 40 50 60 70 -1.2 -0.8 -0.4 0 0.4 0.8 1.2
displacement [mm] slip [mm]
60
(b)
Figure 8. Load-displacement relationship of the specimen 50
tested. 40
30
the bottom part of the specimen (S0). Moreover three
20
Shear load [kN]

linear transducers were employed to survey the slip

between adjoining couples of boards at three quotes 10
from the bottom of the specimen: 1.0 m (S3), 2.0 m 0
(S4) and 3.0 m (S5). All transducers were connected -10
to the same unit system that governs the test. -20
The test was carried out by controlling the hori- -30
zontal displacement at the top of the specimen. The -40
displacement was varied at a rate of 1 mm/min and it -50
was changed gradually between couples of opposite 2.0 m from the bottom
-60
values of the applied load so to consider the actual -1.2 -0.8 -0.4 0 0.4 0.8 1.2
cyclic behaviour during earthquake excitation. slip [mm]

5.3 Test results 60

(c)
50
The results of the experimental test are summarized
40
in Figures 8 and 9. In the first figure the relationship
30
between the applied load and the horizontal displace-
ment at the top of the specimen (S1 − S0) is displayed. 20
Shear load [kN]

As it can be noted four loading cycles were performed 10

between the loads ±15 kN, ±25 kN, ±40 kN, ±50 kN. 0
Then the floor damage became significant so that the -10
displacement was increased monotonically up to the -20
failure of the specimen. The maximum load was equal -30
to 60.7 kN and the maximum displacement was equal -40
to 63.0 mm. -50
The collapse of the floor occurred by longitudinal 1.0 m from the bottom
-60
shear failure of some boards, whereas the connection -1.2 -0.8 -0.4 0 0.4 0.8 1.2
did not show appreciable damage. Such a result was slip [mm]
due to the presence of some longitudinal cracks in the
boards used in the specimen, which propagated when Figure 9. Load-slip relationship: (a) level 3.0 m, (b) level
the specimen was loaded. The propagation of cracks 2.0 m, (c) level 1.0 m.
caused also a reduction of the in-plane stiffness, which
became more pronounced than that awaited because
of the actual nonlinear load-slip relationship of the In Figure 9 the curves representing the shear load
fasteners (Fig. 4). The hysteresis evidences that some against the slip between two adjacent boards are
plastic deformations and damage occur even for low shown. The plots refer to boards at different levels
values of the load. from the bottom of the specimen: 3.0 m (Fig. 9a), 2.0 m

1040
Table 3. Comparison between experimental and numerical
results.
Displacement Secant stiffness
Load Experimental Numerical Experimental Numerical
(kN) (mm) (mm) (N/mm) (N/mm)
15 3.85 7.20 3896
25 8.33 12.00 3000
2146
40 17.20 19.20 2325
50 25.30 24.00 1976
(Fig. 9b), 1.0 m (Fig. 9c). Even though all the curves
have a quite similar shape, the curve plotted in Fig-
ure 9b is steeper than the others. This reveals a greater
stiffness of the connection at level 2.0 m.
However the average stiffness differs slightly from
the values achieved in the push-out tests presented in
Section 3.
5.4 Comparison with the numerical results
The experimental test was also simulated using a com-
mercial numerical procedure for structural analyses.
The specimen was modelled as a grid of perpendicu-
lar beam elements, which simulate the timber joists,
the boards and the perimetric steel elements. As in Sec-
tion 4, the connection between boards and joists (cou-
ple of nails) was modelled through linear springs char-
acterized by translational stiffness equal to 2.4 kN/mm
and rotational stiffness equal to 13.5 kNm. Finally the
connection between boards (nail plates) was modelled
through shear springs having stiffness equal to the
average value obtained from tests (Section 3).
The displacements of the nodes at the bottom of the
specimen were restrained. The load was applied at the
top of the sample equally distributed among the nodes
of the top section.
The numerical simulation allowed the in-plane stiff-
ness of the specimen to be assessed. In Table 3 the
results of the numerical simulation are compared to
the experimental outcomes. In particular the displace-
ments and the secant stiffness relating to some values
of the applied load are reported.
The value of the numerical stiffness is in quite
good agreement with the experimental values of the
secant stiffness. Actually, for low values of the load
the experimental stiffness are appreciably greater than
the numerical one because of some friction actions
in the experimental system. For loads greater than
50 kN, the damage occurred in the boards (propagation
of existing cracks) caused a significant progressive
reduction of the floor in-plane stiffness.
The theoretical stiffness of the unstrengthened floor
is equal to 128 N/mm, which is more that one order
of magnitude lower than that obtained in the test
1041
for the floor stiffened with the proposed technique
(Section 2).
6 CONCLUSIONS
In the research work a technique to increase the in-
plane stiffness of the wooden floors of ancient masonry
buildings is presented and discussed. This technique
aims at preventing the longitudinal slip among the
boards of the deck by using some nail plates. The
technique is characterized by low invasiveness and
high reversibility, which are the requirements for the
conservation of historical buildings.
Some numerical simulations were carried out on
the shear behaviour of a typical wooden floor, orig-
inal and stiffened through different techniques. The
results achieved allowed both to point out the improved
structural performance guaranteed by the proposed
technique and to compare it with some other strength-
ening solutions. The in-plane stiffness of the floors
strengthened through the proposed technique is nearly
20 times greater than that of the unstiffened floor.
The experimental test carried out on a full scale floor
specimen (4.0 × 4.0 m) allowed the effective shear
stiffness to be determined. Some pre-existing cracks
in the boards of the floor deck caused a progressive
stiffness reduction because of the crack propagation
by increasing the load. Moreover the cracks caused a
premature collapse with the occurrence of complete
shear longitudinal splitting of the boards, while the
nail plate fasteners did not show any damage.
The numerical simulation of the experimental test
evidenced a good agreement with experimental results
till the pre-existing crack propagation became signif-
icant. However the in-plane secant stiffness achieved
through the experimental test is more than one order
of magnitude greater than the unstiffened floor theo-
retical stiffness.
Even though these are some early results of an
ongoing research project, it can be concluded that
the proposed strengthening technique allows to ade-
quately stiffen existing wooden floors. So they can
preclude dangerous out of plane deformations of the
masonry walls at floor levels when the building is
subjected to seismic actions.
ACKNOWLEDGEMENTS
The study was supported by a grant for research
projects of the Italian Government (PRIN 2006). The
nail plates used for tests were kindly offered by
Wolf System, Campo di Trens, Bolzano, Italy. The
Authors wish to thank Prof. Fausto Benussi for his
useful advices. The testing staff (Dr. Franco Trevisan,
Dr. Andrea Cernigoi, Mr. Piergiulio Corazza) is grate-
fully acknowledged.
REFERENCES
EN 1995-1-1, 2004 Eurocode 5. Design of Timber Struc-
tures. Part 1-1: General Rules and Rules for Buildings,
CEN European Committee for Standardization, Bruxelles.
EN 12512, 2001 Timber Structures. – Test methods. – Cyclic
testing of joints made with mechanical fasteners. CEN
European Committee for Standardization, Bruxelles.
Gattesco, N. 2001. Experimental Study on Different Dowel
Techniques for Shear Transfer in Wood-Concrete Com-
posite Beams. In A. Shing (ed.), Creative Systems
in Structural and Construction Engineering: 487–492.
Rotterdam: Balkema.
Gattesco, N. & Del Piccolo M. 1997 Shear transfer between
concrete members and stone masonry walls through
driven dowels. European Earthquake Engineering, No. 3,
3–17.
Gattesco, N. & Macorini, L. 2006. Strengthening and stiff-
ening ancient wooden floors with flat steel profiles. In
P.B. Lourenco, P. Roca, C. Modena & S. Agrawal (eds.),
Structural Analysis of Historical Constructions: 405–412.
Macmillan India ltd, New Delhi.
Gattesco, N., Macorini, L. & Benussi, F. 2007. Strengthening
of wooden floors for seismic rehabilitation of historical
buildings through high reversibility techniques, ANIDIS
2007, XII Convegno “L’Ingegneria Sismica in Italia”,
Pisa, 10–14 June 2007.
1042
Giuriani, E. 2004. L’organizzazione degli impalcati per gli
edifici storici. L’Edilizia, Speciale Legno Strutturale,
n. 134, 30–43.
Giuriani, E. 2006. Rinforzo di solai in legno mediante soletta
collaborante di calcestruzzo. In L. Dezi & N. Gattesco
(eds), Strutture Composte – Nuove costruzioni, recu-
pero, ponti, International Centre for Mechanical Sciences,
CISM, Udine, Italy, ISBN 88-85137-20-2, 311–322.
Modena, C., Valluzzi, M.R., Garbin, E. & da Porto, F.
2004. A strengthening technique for timber floors using
traditional materials. Proceedings of the Fourth Inter-
national Seminar on Structural Analysis of Historical
Constructions, Padova, Italy, 10–13 Nov. 2004, 911–921.
Piazza, M. & Turrini, G. 1983. Una tecnica di recupero statico
dei solai in legno. Recuperare, Vol. 5, 6, 7, Milano.
Ronca, P., Gelfi, P. & Giuriani, E. 1991. The Behavior of
a Wood-Concrete Composite Beam under Cyclic and
Long Term Loads. Structural Repair and Maintenance
of Historical Buildings, 2nd International Conference
STREMAH 91, Seville, Spain, Vol. 1, 263–275.
Venice Charter, 1964. International Charter for the Conser-
vation and Restoration of Monuments and Sites. Second
International Congress of Architects and Technicians of
Historic Monuments.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Prestressing of masonry as strengthening measure against earthquake

loading – Experimental and numerical investigations and influences on
simulation results

S.T. Sperbeck & H. Budelmann

Institute of Building Materials, Concrete Construction and Fire Protection, TU Braunschweig, Braunschweig,
Germany

ABSTRACT: Local vertical prestressing is considered as strengthening measure to reduce the vulnerability of
masonry structures in case of earthquake action. The shear strength and the ductility of masonry is improved. This
is shown in static cyclic tests. Static and dynamic behaviour is investigate for different masonry walls. To avoid an
excessive amount of experimental tests, this paper is focused on numerical simulations and impacts on the results
such as boundary conditions and different means to model prestressing. Previously, brief comparison of simula-
tion techniques and material models is given. The task is mainly completed by means of a suitable material model
for macro modelling of masonry. The researched parameters which influence the numerical results are discussed.
Case studies are carried out to investigate the parameters. The importance to model the tendons, especially for
nonlinear dynamic analyses, is shown. The impact of prestressing on the dynamic behaviour is investigated.

1 INTRODUCTION prestressed masonry structures. Firstly, some observed

The resistance of masonry structures against earth-
quake action is limited by its low shear strength.
Vertical prestressing is considered in order to improve
the shear capacity, and the ductility of masonry. Static
and static cyclic tests have already shown the suit-
ability of this method (Budelmann et al. 2004). More
research for different masonry constructions is rec-
ommended. In addition the dynamic behaviour should
be investigate, before using the strengthening method
against earthquake action. The large quantity of nec-
essary experimental tests, especially dynamic ones
with shaking tables, is very expensive. The final
goal is to provide possibilities, based on numeri-
cal methods, to investigate the dynamic behaviour of
parameters, which influence the numerical results are
investigated and discussed.
2 STATIC CYCLIC EXPERIMENTAL TEST OF
PRESTRESSED SHEAR WALLS
The experimental tests of prestressed shear walls
briefly summarised below are described in detail in
Budelmann et al. (2004). Four walls were tested, with
dimensions and extra loads like walls used for stiff-
ening of buildings with three floors. All walls had
a height of 2.5 m and thickness of 0.175 m. In these
static cyclic tests only the ground floors were consid-
ered as they are most critical under seismic action.

W1 W2 W3 W4

Figure 1. Complete experimental set-up for wall 1, wall 2, wall 3 and wall 4.

1043
For all walls, two tendons (strands) have been used of cracks. 2xP0 is the sum of the initial prestressing
for vertical prestressing. The complete experimental forces of two tendons. The dead load of the wall and
set-up is depicted in Figure 1 on the left for wall 1. upper stories is expressed by G+F. In the last column,
Furthermore, another position of the tendons closer to the forces in the tendons after reaching the ultimate
the middle was investigated in wall 2. The influence of loading point (2xP0 )u are given. A significant decrease
slenderness is taken into account by means of wall 3 was observed. Moreover, it is listed where stone failure
and wall 4. With wall 4, the floor slab is supported only (partial collapse of the wall) occurred, or if maxi-
on one end. The wall properties are listed in Table 1. mal displacement of the testing equipment (vmax ) was
A horizontal static cyclic displacement was applied reached.
in the centre of the concrete slab. The crack patterns
are shown in Figure 2. The diagonal crosses are very
typical of earthquake damage. The load displacement 3 NUMERICAL SIMULATION TECHNIQUES
curves are displayed in Figure 3. Wall 2 had the most AND MATERIAL MODELS
useful behaviour. The shear capacity and the ductil-
ity are very high. The area enclosed by the hysteresis 3.1 Theoretical background
shows the energy dissipation, which is very good for In the following, it is briefly discussed which sim-
wall 2 and wall 3. Measured values like the displace- ulation techniques and material models are suitable
ment u and the horizontal load H are summarised in to investigate the dynamic behaviour of prestressed
Table 2 for all walls. The index u means the ultimate masonry. In case of cyclic and dynamic loading,
point of loading, whereas cr indicates the occurrence a degradation of stiffness and strength occur. The
dynamic behaviour is effected strongly by the stiffness
Table 1. Properties of the walls. which is an important consideration.
Usually, the accurate micro modelling is applied
Distance Distance to investigate masonry in detail. The material model
of tendon of by Oliveira (2003) includes degradation for this
Length Height Thickness to edge tendons Support modelling strategy. It is based on the plasticity theory,
m m m m m described by Lourenço (1996) and Rots (1997). Due to
the high calculation effort, the material model is very
W1 2.5 0.25 2.00 on 2 time consuming in the case of dynamic simulations
W2 2.5 2.5 0.175 0.625 1.25 sides of large structures. More suitable for such simulations
W3 2.5 0.25 0.75
is the material model of Gambarotta & Lagomarsino
W4 2.5 0.25 0.75 on 1 side
(1997), which is based on fracture mechanics and

Figure 2. Crack pattern of wall 1, wall 2, wall 3 and wall 4.

Figure 3. Horizontal load displacement diagrams (hysteresis) for wall 1–4.

1044
macro modelling. Also, this model is able to describe detail in Ötes & Löring (2006) as well as other liter-
the post-peak behaviour and the dynamic behaviour. ature. Two extreme cases are given. For case BC 1,
Damage modelling is used in which the inelastic the top of the wall is constrained, because of that it
strains are described by means of two internal damage stays horizontal. Mainly shear loading occurs. For case
variables which express the damage evolution in the BC 2, the top of the wall is free and can rotate. Thus,
bricks and mortar. The model of Gambarotta & Lago- the wall behaves like a cantilever, and bending loading
marsino (1997) works efficiently also in the case of occurs. Below it is designated as BC 2. Figure 4 shows
dynamic nonlinear analyses and probabilistic analyses walls with these boundary conditions and shapes.
of large scale masonry structures. In reality, the support of the wall is between these
extreme cases, depending on the behaviour of the floor
slabs. To predict the behaviour realistically, beams
4 IMPACTS ON NUMERICAL RESULTS
should be modelled, as depicted in Figure 4 in middle.
Below, the impacts are explained in more detail.
It is important to know which parameters have a
significant influence on the numerical results. The
findings of a literature review and of simulations 4.2 Case study of static analyses
made for this contribution are discussed as follows.
To gain deeper insight, a case study was performed in
Due to the great impact of boundary conditions, the
which the following parameters were varied to inves-
different conditions are separately explained in the
tigate their impacts. Four variations of slenderness
next subchapter.
S (0.5, 1, 2, 3) are made. For the different slender-
ness values, different heights of 1.25 m, 2.5 m, 5 m
and 7.5 m are obtained. The boundary conditions on
4.1 Boundary conditions of shear walls the top of the wall BC 1 (constrained), modelled by
means of a ridge L-framework, and BC 2 (free) are
The boundary conditions influence the behaviour and used. Furthermore, for BC 2 walls with tendons close
failure mechanisms strongly. This is explained in more to the middle are investigated. Two means to model
prestressing are applied (external forces and tendons),
Table 2. Loading and results of the experimental tests. illustrated in Figure 5.
Its outcome are the impacts on: the change of pre-
2xP0 G+F Hcr Ucr Hu Uu Type (2xP0)u stressing forces in the tendons, the restoring forces,
kN kN kN mm kN mm of fail. kN the rotation of the top, the shear capacity, the ductil-
ity as well as the suitability of the tendons location.
W1 360 245 +217 3 +109 17 vmax 260 The subsequent values are fixed for all variations of
−227 −105 the models: the width of the walls is 2.5 m, thick-
W2 356 245 +223 6 +89 23 Stone 128
ness 0.175 m, prestressing force per each tendon is
−220 −94
189 kN, and vertical load of upper stories is 197.3 kN.
W3 352 110 +110 7 +60 23 vmax 260
Furthermore, the material parameters are the same for
−114 −60
W4 274 110 +101 11 +34 20 Stone 140
all models as listed in Table 3.
−100 −16 Below, some important results of this case study
are briefly summarised. In general, the means to sim-

Figure 4. Left: BC 1 (constrained), middle: real boundary conditions, right: BC 2 (free).

1045
ulate prestressing is important for BC 2, especially
when the tendons are close to the edges. For BC 1 this
phenomenon can often be neglected. Restoring forces
occur and can be simulated when the prestressing is
modelled by means of tendons. The restoring forces
have to be divided into horizontal and vertical compo-
nents. The last one is important only for BC 2. Here,
the vertical movement of the corners during the top
rotation leads to changes of tendons length. The result
is a variation of the prestressing forces in the tendons,
which decrease in the tendons on the lower corner and
Figure 5. Investigated walls in dependency of the slenderness, boundary conditions, means to model the prestressing and
position of tendons.
1046
increase in the tendons on the upper corner. In the case
of low walls, the top rotation leads to significant dif-
ferences in the prestressing forces. These differences
decrease as the walls become higher as illustrated in
Figure 6. The reason is simple. An equal change of
length of a long and short bar gives high stresses in
the smaller bar, but only small stresses in the longer.
For models with tendons close to the edges, the rota-
tion of the top edge is smaller, but the tendons have to
be modelled to notice this effect. If only external forces
are used to model prestressing, no significant differ-
ence can be observed. Of course, the shear capacity and higher than 5 mm. Figure 8 depicts greater stiffness
the ductility depend decisively on the slenderness and and shear capacities for walls with BC 1 than for
boundary conditions, as already noted and described BC 2, but smaller ductilities. Especially for slender
above. walls the differences in ductility are significant. It is
To get information about the shear capacities and found in case of BC 2, when the tendons are modelled,
ductilities it is necessary to apply a horizontal loading the shear capacity is higher (Fig. 7 left) and the duc-
tility is a bit smaller. This means of modelling leads
Table 3. Material parameters used for the case study. to smaller horizontal displacement. Figure 7 left also
shows higher stiffness for walls modelled with ten-
Symbol Variable Value dons. If the tendons are placed close to the middle,
no significant difference can be observed, as shown
µ Friction coefficient 0.8 in Figure 7 right. The post-peak behaviour of mod-
σmr Tensile strength 0.15 N/mm2 els with tendons near the edges is more useful, where
of mortar joints
higher forces can be applied. Obviously, the tendons
τmr Shear strength of 0.20 N/mm2
mortar joints
carry tensile loads after tensile failure has occurred in
σbr Compressive 3.5 N/mm2 the masonry wall. If the tendons are close to the mid-
strength of masonry dle, only a small improvement regarding the post-peak
τbr Shear strength 1.5 N/mm2 behaviour can be observed. A comparison of Figure 7
of masonry left and right shows this. The findings of this case
η Poissons ratio 0.1 study regarding the impact of slenderness cannot be
ρ Density 2000 kg/m3 generalised since the width of the walls is constant.The
E Young’s modulus 2000 N/mm2 height of the wall has an important influence, because
βm Softening mortar 0.6 it is equal to the basic length of the tendons. Small
βb Softening brick 0.4 changes of length lead to large differences of the forces
cmt Inelastic deformation 1.0
parameter for mortar
inside short tendons. For long tendons, much higher
cb t Inelastic deformation 1.0 differences in length are necessary to reach significant
parameter for brick changes of such inner forces. As shown, many reasons
exist to model tendons, such as the tendon forces that

Figure 6. Forces in tendons in dependency on the horizontal top displacement of walls for BC 2 and tendons close to the
edges.

1047
change during static horizontal loading. This impact
is significant for BC 2 and compact walls. The shear
capacity depends considerably on the vertical loads.
Another important reason for modelling the tendons,
is the decreasing of the prestressing force in the ten-
dons during static cyclic and seismic loading. This
was observed in experimental tests (Budelmann et al.
2004), probably caused by the reduction in height of
bed joints due to slipping in the joints. The static sim-
ulations show horizontal and vertical restoring forces
lead to smaller horizontal displacement and smaller
rotations as well as to increased stiffness (Fig. 7 left).

Figure 7. Left: Horizontal load displacement diagram for BC 2 and tendons close to the edges, right: horizontal load
displacement diagram for BC 2 and tendons close to the middle.

Figure 8. Horizontal load displacement diagram for models with tendons for BC 1 and for BC 2 for different tendon positions
(edges and middle).

1048
4.3 Case Study of nonlinear dynamic analyses vertical load of 197.3 kN. The variations ‘external
forces’ and ‘tendons’ have an additional prestress-
The increasing of stiffness, as mentioned above, affect
ing force of 189 kN per each tendon. The results
the dynamic behaviour. A small case study was per-
for the vibration behaviour regarding the horizon-
formed to investigate further. The vibration behaviour
tal top displacement are depicted in Figure 9. For
of prestressed walls modelled with external forces,
all of the wall models and load levels, the biggest
tendons, and for walls without prestressing, is cal-
value for the roof displacement is reached after 0.12 s.
culated in nonlinear dynamic simulations. A ground
The highest displacement reaches 34 mm for the non-
displacement is applied as an impulse for all men-
prestressed wall. As expected prestressing leads to a
tioned simulations below. The time for the impulse
reduction of the vibration amplitude. For the mod-
is 0.12 s to move ground and return back to the orig-
els with external forces the displacement amounts to
inal position. The three different load functions are
32 mm and for the models with tendons 32.4 mm. The
time dependent as identified in Figure 9 with dashed
observed difference between the maximum horizontal
lines. Table 4 gives a short overview of the carried
roof displacement of ‘tendons’ and ‘external forces’
out dynamic analyses, the level of the applied dis-
are smaller for lower load levels.
placement, and whether it was possible to receive
The horizontal displacement of the prestressed
convergence.
walls is insignificantly smaller than the prestressed
For these models the same values for width, thick-
walls for lower vertical load levels (5 mm and 10 mm).
ness and material parameters (Table 3) are used as
Nevertheless, the mortar damage is much less for
for the static case study above. All have a static
prestressed walls, than for the non-prestressed wall
(Fig. 10). This indicates that vertical prestressing can
be a useful strengthening measure also for dynamic
Table 4. Overview of the carried out analyses with impulse
loading. More detailed investigations are still in
loading.
progress.
Means to model prestressing The dynamic behaviour is different for all consid-
ered walls, as periods vary significantly.The walls with
Impulse External Without tendons vibrates faster. This means they are stiffer.
displacement forces Tendons prestressing Observations of experimental tests lead to a probable
reason. Ötes et al. (2002) observe that in the range
5 mm Yes Yes Yes of high horizontal loading, the stiffness of the wall
10 mm Yes Yes Yes
17 mm Yes Yes Yes
is mainly affected due to the spring properties of the
tendons after the occurrence of gaping joints. This

Figure 9. Horizontal displacement of the top time-dependent for modelling of prestressing by means of external forces,
tendons, and without prestressing in case of BC 2.

1049
Figure 10. Mortar damage for impulse loading of 10 mm, left: external forces, middle: tendons, right: without prestressing.

Figure 11. Mortar damage for impulse loading of 17 mm, left: external forces, middle: tendons, right: without prestressing.

would explain also the bigger difference between ‘ten-
dons’ and ‘external forces’ in case of higher shaking
levels. Also, presented static simulations show that
the post-peak behaviour varies with the prestressing
is modelled as ‘external forces’ or ‘tendons’ (Fig. 7
left). Modelling of tendons leads to higher resistances.
In general, the walls vibrates slower, when the ground
acceleration is higher. A reason is higher damage of
the walls leading to smaller stiffness, and lower fre-
quencies. For a higher load level (horizontal ground
displacement of 17 mm) also the prestressed walls are
significantly damaged (Fig. 11). Here, the difference
between the two means to model prestressing become
larger, but not important. In all these simulations the
mortar damage, as well as the brick damage, is a bit
higher for tendons, than for external forces.

1050
5 CONCLUSIONS
The static cyclic experimental tests on internal pre-
stressed shear walls indicate the functionality of
this method to strengthen masonry against earth-
quake loading. Useful material models were discussed
and used for numerical investigations. Many rea-
sons to model the tendons were pointed out, e.g.
the restoring force and the impact on the dynamic
behaviour. Deeper investigations especially regarding
the dynamic behaviour of prestressed walls are recom-
mended. This has been done by means of first dynam-
ical nonlinear simulations. The mentioned important
factors have to be considered carefully in further
simulations.
ACKNOWLEDGEMENTS
We would like to thank Prof. Lagomarsino and
Dr. Calderini from the Università degli Studi di Genova
for their permission and support in including their
material model into our research. Also we have to
express deepest gratitude to Prof. Bartoli from the
Università degli Studi di Firenze.
REFERENCES
Budelmann, H., Gunkler, E., Husemann, U. & Becke, A.
2004. Rationell hergestellte Wände aus vorgespanntem
1051
großformatigem MW mit hohem Erdbebenwiderstand,
Abschlussbericht, iBMB der TU Braunschweig, Germany
Gambarotta, L. & Lagomarsino, S. 1997. Damage Models for
the seismic response of brick masonry shear walls, Part I
and II, Earthquake Engineering and Structural Dynamics,
Vol. 26, 1997: 423–462
Lourenço, P.B. 1996. Computational strategies for masonry
structures, Ph.D. Thesis, Delft University of Technology,
Netherlands
Oliveira, D.V. 2003. Experimental and numerical analysis
of blocky masonry structures under cyclic loading, Ph.D.
Thesis, University of Minho, Portugal
Ötes, A., Löring, S. & Elsch, B. 2002. Erhöhung der Schub-
tragfähigkeit von KS-Wänden unter Erdbebenlasten durch
schlaf bewehrte Beton-stützen in Formsteinen bzw. durch
Vorspann-ung der Wand, Forschungsvereinigung Kalk-
Sand e.V.
Ötes, A. & Löring, S. 2006. Zum Tragverhalten von Mauerw-
erksbauten unter Erdbebenbelastung, Bautechnik, No. 83,
Heft 2, 2006: 125–138
Rots, J.G. 1997. Structural Masonry: An Experimental
/Numerical Basis for Practical Design Rules, Balkema,
Rotterdam, Netherlands, ISBN 90 5410 680 8
Verfmeltfoort, A.Th. & Raijmakers, T.M.J. 1993. Deforma-
tion controlled meso shear tests on masonry piers, Part 2,
Draft report, TU Eindhoven, dept. BKO, Netherlands.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Enhancement of geo-electrical techniques for NDT of masonry

F. Van Rickstal & D. Van Gemert

Department of Civil Engineering, Building Materials Division, Katholieke Universiteit Leuven, Belgium

R. Keersmaekers
Department of Architecture, Provincial University College Limburg, Hasselt, Belgium
Department of Civil Engineering, Building Materials Division, Katholieke Universiteit
Leuven, Belgium

D. Posen
Department of Architecture, Provincial University College Limburg, Hasselt, Belgium

ABSTRACT: A geophysical resistivity measuring device was modified to perform automatic monitoring of
historical masonry structures before, during and after grout injection for consolidation purposes. The obtained
image is called a geo-electrical tomography. The technique was used to evaluate the deteriorated masonry of the
recently partly collapsed Maagdentoren in Zichem, Belgium. The results of these measurements are discussed.
An important improvement is the adaptation of the “cross-borehole” measurement technique for masonry. With
this technique, a constant resolution in depth is achieved. Laboratory tests, using an adaptation of the “cross-
borehole” technique, enabled to show the resistivity distribution changing during the injection of a test specimen
with a hydraulic grout.

1 INTRODUCTION GENERAL PURPOSE OF
THE RESEARCH PROJECT
With a grant of the Flemish government (IWT-Institute
for the Promotion of Innovation by Science and Tech-
nology in Flanders) and the support of a group of
Belgian companies involved in restoration activities,
a research project started in September 2005, aim-
ing to improve control methods for grout injections
in masonry. The project was developed within a part-
nership between the Provincial University College
Limburg, Department of Architecture, Hasselt (B) and
the Katholieke Universiteit Leuven, Department of
Civil Engineering, Building Materials Division (B).
The main objectives of the project are:
1.1 Inventory of knowledge
Due to the wide spectrum of possible applications of
grout injections, a lot of practical experience is present,
but this knowledge is partitioned, spread throughout
the building practice and not readily accessible. The
knowledge of injection techniques is fragmented and
the structured inventory of all this knowledge is there-
fore very important. This classification will not only
be the summary of existing techniques in the build-
ing practice, but it will also be a well-argued decision
matrix, which points at the correct grout mixture and
evaluation sequence, given a specific grouting prob-
lem and boundary conditions. This structuring forms
the first part of the research project. Specific expe-
riences with grout injections of a group of Belgian
companies involved in restoration activities will be
studied, and expert aspects will be derived out of these
experiences.
1.2 Integration of non-destructive techniques
Secondly the implementation of non-destructive tech-
niques in the field of grout injections will be stud-
ied, not only for the evaluation of the quality of
the deteriorated or injected massive (masonry or
concrete) before and after injection. The aim is
to develop the NDT-technology that should enable
on-line surveying of grout injections. The bene-
fit would be the possibility to alter the injection
parameters during the injection itself so that the
desired injection result can be obtained. Geo-electrical
survey of masonry already proved to be a reli-
able non-destructive tool for this purpose (Janssens
1993; Keersmaekers 2003; Keersmaekers et al. 2004;
Keersmaekers et al. May 2006; Keersmaekers et al.
Nov 2006; Marchisio et al. 2002; Van Rickstal and

1053
Vanhellemont 2002; Van Rickstal et al. 2003;
Van Rickstal et al. 2006; Venderickx 2000).
1.3 Test program
The last objective of the project, indispensable as a sup-
port for the first goal, is a thorough investigation of the
physical, physical-chemical, chemical and mechani-
cal properties of different grout compositions and the
relation between the different parameters responsible
for these properties. Special attention is given hereby
to the bending tensile strength of mineral grout mix-
tures, being the most representative for the bonding
properties of the grout.
2 BASICS OF GEO-ELECTRICAL
MEASUREMENTS
In the past, geo-electrical resistivity measure-
ments were already used successfully in restoration
projects by the Reyntjens Laboratory (Janssens 1993;
Venderickx 2000). The interpretation of the geo-
electrical measurements was based on the apparent
resistivity value. The word “apparent resistivity” is
used because this value is the resistivity value that
would hold for a fully homogeneous and isotropic
material. In reality this is not the case. Walls have a
heterogeneous resistivity distribution.
In the recent past new developments were made in
geophysics within the field of geo-electrical survey of
soils. This was possible thanks to the growing effec-
tiveness and power of numerical methods. Software
was developed which can invert the measured apparent
resistivity values to a distribution of absolute resistiv-
ity values of the subsoil. (Dey and Morrison 1979;
Geotomo 2002; Lines and Treitel 1984; Loke and
Dahlin 1997; Loke 2002) The resistivity measurement
remains the same, but the sequence is different and
the number of measurements is larger than previously
common on masonry structures.
The technique is known in geophysics as geo-
electrical tomography. Figure 1 shows the necessary
measurements for a Wenner-alfa electrode-configu-
ration with 20 electrodes. Each of these electrodes,
mostly stainless steel nails, can be used to inject cur-
rent (current electrodes C1, C2) or to measure potential
(potential electrodes P1, P2). Conducting the first
row of measurements, a distance “a” (n = 1) between
C1P1, P1P2, P2C2 is kept constant, leading to 17
(20–3) measurements on this particular row. Raising
the electrode distance to “2a” (n = 2) gives a bigger
influenced zone of the masonry (i.e. more profound
penetration of the potential field), resulting in an
apparent resistivity value representative for a larger
and deeper zone of the masonry. Conventionally, this
apparent resistivity value is allocated to a physical
1054
Figure 1. 2D geo-electrical tomography; Wenner-alfa
electrode-configuration; subsequent measurements for
building a pseudo-section.
point, centrally located between the four electrodes
and on a depth equal to the average depth of the
influenced zone of the potential field. Note that the
second row of measurements (with an electrode dis-
tance “2a”) has only 14 (20–2 × 3) measurements. The
number of measurements decreases with augmenting
electrode spacing (i.e. higher penetration depth). In
this way apparent resistivity values are gathered over
the depth of the masonry structure and a “pseudo-
section”, which is the graphical representation of the
measured apparent resistivity values, is built.
Unfortunately, the 2D pseudo-section does not
enable direct interpretation of the apparent resistiv-
ity values. No conclusions about the real resistivity
distribution in the substrate (internal structure of the
masonry) can be drawn based on apparent resistiv-
ity values. To build the real resistivity distribution
it is necessary to numerically invert or transform
the pseudo-section values into real resistivity val-
ues. Forward modelling enables the calculation of the
apparent resistivity values (pseudo-section), starting
from a given resistivity distribution. The idea is to
construct a numerical model with a resistivity distri-
bution whose calculated pseudo-section corresponds
with the measured pseudo-section. The section of the
wall is therefore numerically divided into blocks or
elements, where every element is given a resistiv-
ity value. A least square approximation between the
calculated and measured pseudo-sections determines
how the model parameters (i.e. the resistivity values
of the model blocks) must change in the next iteration,
resulting in a better correlation between calculated and
measured pseudo-sections.
In practice, a measuring campaign starts by posi-
tioning the electrodes. Stainless steel nails are used
for masonry structures. The electrodes are connected
to the automatic scanning module which switches
the nails as current or potential electrode. In this
research project, the geo-electrical measuring device
Terrameter SAS1000 was used.

Figure 2. After the collapse on June 1st, 2006.
3 CASE STUDY: MAAGDENTOREN
ZICHEM (B)
3.1 History
The “Maagdentoren” or “Tower of the Virgins“ is
located at the bank of the Demer river and is a 26
meters high, 15 meters wide ferruginous sandstone
tower. The walls have a thickness of 4,2 meters at the
base and 1,8 meters at the top. Over history, numerous
reparations, renovations and restorations were done.
This exceptional piece of medieval archi tecture was
built in 1380. The original function is unclear, but it
can be assumed that the tower served multifunctional
purposes. It was used residentially, military and pres-
tigiously. The tower was partly destroyed during the
Eighty Years’War around the 16e century. The upper
level and the entrance were demolished.
A series of necessary repairs were executed in 1863
and 1905. After that, the tower deteriorated for the rest
of the century, resulting in a partial collapse of a part of
the outside parament in 1995. The biggest catastrophe
occurred on June 1st, 2006, when a huge part of the
outside wall collapsed (Figure 2).
Figure 3 shows the location on the tower where
a dipole-dipole electrode configuration was used to
perform a geo-electrical tomography (white line).
The survey line of the tomography was deliberately
1055
partially placed over a visually good looking part of the
wall and a bad looking zone. This visually bad looking
zone is caused by the demolition of an historic defence
platform that was situated there (Figure 3).
3.2 Geo-electrical measurements
Figure 4 shows the inverted section of the measured
survey line, using the dipole-dipole electrode config-
uration. Remark that the inversion result does not lead
to the same conclusions as would be drawn from a
visual inspection of the surface only. A dark colour
represents a high electrical resistivity, meaning a dete-
riorated zone in the masonry. From the outside visually
good looking zones of the masonry show high resis-
tivity values on a depth of 30 to 40 centimetres on the
tomography.This depth corresponds with the thickness
of the outer leaf of the wall. These high resistivity val-
ues continue over a thickness of 50 to 60 centimetres.
After that the resistivity values decrease, meaning less
deteriorated and less eroded masonry. This conclusion
was verified with an endoscopic (borehole 1) exami-
nation. The following damage mechanism can explain
this phenomenon. The ferruginous sandstone of the
outer leaf absorbs rainwater like a sponge. This water
penetrates the outer leaf by capillary forces, wetting
the inner core of the three leaf masonry and dissolving
lime from the mortar. When the wall dries, leaching of
the dissolved lime from the inner core takes place.
After several centuries, this led to a severely deterio-
rated zone of approximately 50 centimetres thickness
behind the outer leaf. Deeper in the wall, the masonry
is less deteriorated and shows better cohesion (lower
resistivity values). The masonry that visually looks
the worst (zone of the defence platform), shows lower
resistivity values on the tomography (indicating a bet-
ter cohesion) compared with the adjacent masonry. An
exception can be observed at borehole 2, where a more
deteriorated zone is seen on the tomography. This zone
was verified by the endoscope. Borehole 3 confirmed
the overall better state of the masonry. The explana-
tion for the fact that the bad looking masonry from the
outside is less deteriorated and has a better cohesion
than the good looking masonry is the following. At this
place there was a defence platform witch protected the
now exposed masonry for centuries for deterioration.
Around the tower, evidence of foundations of three
of these defence platforms were discovered during
archaeological surveys conducted in the eighties of
the past century. When these platforms were removed
is not clear until the present day (Oost and Geyskens
1982).
3.3 Conclusions for the case study
Based on the geo-electrical tomography and the endo-
scopic examination of some boreholes, it can be
concluded that the overall state of the “Tower of the
Virgins” is seriously deteriorated for a significant part
of its thickness. Keeping in mind the collapse on
the first of June 2006, a consolidation injection with
an appropriate hydraulic grout is urgently needed to
preserve this monument for the next generations.
4 LABORATORY TESTS: CROSS-BOREHOLE
Geo-electrical tomographies, as described above, have
the disadvantage that the resolution decreases with the
penetration depth of the measurements. Nevertheless,

Figure 3. Position of the survey line on the tower; Insert: detail with in white the position of the electrodes and in black the
position of the historic defence platform, removed in an unknown past.

Figure 4. Above: position of the electrodes on the wall (white line); Below: inversion result of dipole-dipole measurements
and the location of the boreholes for the endoscopic survey.

1056
very interesting conclusions can be drawn based on
these tomographies as was clearly shown in the case
study above. In the research project, another exist-
ing technique was adapted for masonry, based on a
two sided approach of the wall, whereby on both
sides electrodes are placed. In geophysics this tech-
nique is referred to as “cross-borehole”. Theoretical
research and numerical simulations showed that this
approach leads to a constant resolution over the depth
of the wall (Bernaerts and Moens, 2006; Stroobant and
Verkest, 2007). Figure 5 shows the result of a success-
ful cross-borehole measurement, without going into
the technical aspects. A replica of a three leaf masonry
wall was built in the laboratory, containing a piece of
insulation simulating a void in the wall. The inversion
result (Figure 5 Right) clearly shows that the insulation
Figure 5. (a) Position of insulation as a simulation of a
void; (b) Position of the electrodes; (c) Inversion result of the
cross-borehole survey, with the high resistivity at location of
insulation block.
is detected as a high resistivity within the graphical
representation of the resistivity distribution.
5 LABORATORY TESTS: ON-LINE
MONITORING OF A GROUT INJECTION
The cross-borehole technique seemed suitable for on-
line follow up of hydraulic grout injections.This means
that it can be used during injections in a way that
an image can be constructed showing the grout flow
in the masonry. This would be a useful appliance to
evaluate the effectiveness of the injection. Also the
detection of grout leaching in unwanted directions
and the incomplete filling of voids could be detected
in real time.
To accomplish this, a customized sophisticated data
logger had to be built. This was necessary because the
system must be capable of reading, filtering and pro-
cessing the data fast enough to evaluate the injection
in real time. A Labview-based (National Instruments)
system was built using 48 analogue input channels
which can be sampled parallel with each other. The
principle is as follows. On one side of the wall 3 parallel
rows of 16 current electrodes are placed and connected
with the automatic switching module. This is the active
side of the wall. Likewise, on the backside of the wall
3 parallel rows of potential electrodes are placed and
connected to the data logger. This is called the pas-
sive side and here the response to the injected currents
is measured. In this way it is possible to measure 9
cross borehole tomographies in 90 to 120 seconds (see
Figure 6: Left). After inversion the nine 2D-images
are combined to a single 3D-image of the resistivity
distribution (Stroobant and Verkest, 2007).

Figure 6. (a) 9 cross-borehole tomographies used to construct a 3D image of the inversion; (b) 3 × 16 current electrodes, the
potential electrodes (3 × 16) at the backside; (c) Position of the volumes of gravel to be injected and the result of the inversion
3D-inversion result at the end of the injection; (d) Injected volumes, excavated after hardening of the grout.

1057
 A laboratory test was built in order to test new
procedures for on-line follow up of grout injections.
Three zones with high resistivity (representing dete-
riorated, little cohesive masonry) were placed in a
volume of sand. This was done by inserting three
volumes of tennis court gravel (different sizes) in
the sand volume (Figure 6: Middle). Gravel consists
of crushed bricks, an easily injectable material. A
reference measurement was done before the injec-
tion started. Then, every gravel volume was gradually
injected with a hydraulic grout. Simultaneously, the set
of 9 cross borehole tomographies was measured every
two minutes. Drawing of the relative difference charts,
which gives graphic representation of the difference in
resistivity between two subsequent 3D-tomographies,
clearly shows the gradual filling of the different gravel
volumes. The good electrical conductivity of the grout
leads to local resistivity drops creating a big contrast
with the former high resistivity of the gravel. This
provides an excellent condition to detect the injected
area on a relative difference chart. Figure 6 (Middle
right) gives the 3D relative difference chart between
the last measurement (at the end of the injection of the
three volumes) and the reference measurement before
injection. Notice how clearly visible the injected zones
appear on the image. The three volumes were exca-
vated after hardening of the grout (Figure 6: Right).
A good correlation between these volumes and the
inversion result was observed.
Drawing of the relative difference charts, which
gives graphic representation of the difference in
resistivity between two subsequent 3D-tomographies,
clearly shows the gradual filling of the different gravel
volumes. The good electrical conductivity of the grout
leads to local resistivity drops creating a big contrast
with the former high resistivity of the gravel. This
provides an excellent condition to detect the injected
area on a relative difference chart. Figure 6 (Middle
right) gives the 3D relative difference chart between
the last measurement (at the end of the injection of the
three volumes) and the reference measurement before
injection. Notice how clearly visible the injected zones
appear on the image. The three volumes were exca-
vated after hardening of the grout (Figure 6: Right).
A good correlation between these volumes and the
inversion result was observed.
The above technique seems very suitable for on-
line follow up of grout injections. Conventional geo-
electrical tomographies (like in the case of the Tower
of the Virgins, or cross-borehole), used to determine
the internal structure of the masonry, can not always be
interpreted unambiguously.Therefore, the conclusions
based on conventional topographies should always be
verified with destructive tests, like endoscopic survey
or core drilling. This disadvantage disappears when the
technique is used for on-line follow up of grout injec-
tions. When composing a difference chart, relative to
1058
a reference measurement before injection, the contrast
is caused by an external influence (the injected grout).
This definitely simplifies the interpretation, because
the difference in resistivity can only be caused by the
added grout.
6 CONCLUSIONS
With a grant of the Flemish Government (IWT)
and the support of a group of Belgian companies
involved in restoration activities, a research project
was started in September 2005, aiming to control
grout injections in masonry in a more reliable way.
One of the objectives was the implementation of non-
destructive techniques in the field of grout injections.
The project showed the feasibility of performing geo-
electrical tomographies (one side approach and two
sided approach by means of cross-borehole) for the
evaluation of the quality of deteriorated or injected
masonry before and after injection. An adapted new
NDT-technology has been developed for on-line sur-
veying of grout injections.The benefit is the possibility
to adapt the injection parameters during the injection
so that the desired injection result can be obtained.
ACKNOWLEDGMENTS
The authors thank the master thesis students
K. Bernaerts, M. Moens, D. Verkest and W. Stroobant,
for their contribution to the discussed research pro-
gram and their work in the laboratory.
The grant of the Institute for the Promotion of Inno-
vation by Science and Technology in Flanders (IWT)
and the support of the Belgian companies to the project
are gratefully acknowledged.
REFERENCES
Bernaerts K., Moens M., 2006 Onderzoek van geo-
elektrische meettechnieken voor niet destructief onder-
zoek van historische metselwerkmassieven. Master
Thesis. KULeuven. Department of Civil Engineering.
Leuven
Dey A., Morrison H.F., 1979a, Resistivity modelling for
arbitrary shaped two-dimensional structures. Geophysical
Prospecting 27.
Goetomo., 2002. Manual of the software packages
Res2Dmod and Res2Dinv version 3.5. Download from
http://www.geoelectrical.com/.
Janssens H. 1993., Geo-elektrische kontrole van consolidatie-
injectie bij metselwerk. PhD thesis. KULeuven.
Keersmaekers R. 2003. De geo-elektrische methode toegepast
op metselwerkstructuren: implementatie van recente
ontwikkelingen. Master thesis. KULeuven. Department of
Civil Engineering. Download http://www.kuleuven.ac.be/
bwk/materials/Publications/.
Keersmaekers R., Van Rickstal F., Van Gemert D., Geo-
electrical techniques as a non-destructive appliance for
restoration purposes. Proceedings of 4th international
seminar on structural analysis of historical constructions –
2004. Padova. Italy.
Keersmaekers R., Schueremans L., Van Rickstal F., Van
Gemert D., 2006. Development of an appropriate grout
for the consolidation of the column foundations in Our
Lady’s Basilica at Tongeren (Belgium). Proceedings of
the 8th CANMET/ACI international conference on recent
advances in concrete technology –Montreal. Canada.
Keersmaekers R., Schueremans L., Van Rickstal F., Van
Gemert D., Knapen M., Posen D., 2006. NDT-control of
injection of an appropriate grout mixture for the consolida-
tion of the columns foundations of Our Lady’s Basilica at
Tongeren (Belgium). Proceedings of the 5th international
seminar on structural analysis of historical constructions –
New Delhi. India.
Lines L.R., Treitel A., 1984. A review of least-squares
inversion and its application to geophysical problems.
Geophysical Prospecting 32.
Loke M.H., Dahlin T., 1997. A combined Gauss-Newton
inversion method for the interpretation of apparent resis-
tivity pseudo-sections. Proceedings of the 3rd meeting of
the Environmental and Engineering Geophysics Society –
European Section. Aarhus. Denmark.
Marchisio M. et al., 2002. Non-destructive testing on
masonry structures: a series of different methodologies
1059
applied on the cathedral of Lucca. Proceedings of the 8th
Meeting EEGS – ES. Aveiro.
Oost T., Geyskens L., 1982. Overzicht van de opgravingen
1981–1982 aan de Maagdentoren te Zichem.
Stroobant W., Verkest D., 2007. Verfijning en afijking van
geo-elektrische kaarten voor NDT van monumentaal met-
selwerk. Master Thesis. KULeuven. Department of Civil
Engineering. Leuven.
Van Rickstal F., Vanhellemont Y., 2002. Niet-destructief
opzoeken van scheuren in metselwerk en pleisters.
Technieken en case studies. Wetenschappelijk-technische
groep voor aanbevelingen inzake bouwrenovatie en mon-
umentenzorg. Studiedag. Academiezaal Sint-Truiden.
Van Rickstal F., Keersmaekers R., Van Gemert D., 2003.
Geo-electrical investigation of masonry walls: develop-
ments and case studies. Department of Civil Engineering.
KULeuven. Belgium. Proceedings of the 6th meeting
of the Materials Science and Restoration Society –
Karlsruhe. Germany.
Van Rickstal F., Keersmaekers R., Van Gemert D., 2006.
Consolidation of ancient masonry. assessment. develop-
ments in grouting. 22. Internationales WTA Kolloquium
Bauinstandsetzen-Bauphysik. HAWK. Hildesheim. WTA
Almanach ‘Restoration and Building-Physics’.
ED. J. Gaenssmantel. ISBN 3-937066-05-5.
Venderickx K., Evaluatie van geo-elektrische metingen op
metselwerkstructuren. PhD thesis. KULeuven juni 2000.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Synergetic action of new and existing structures in re-using a 18th century

monastery

M. Mezzi
Department of Civil and Environmental Engineering, University of Perugia, Perugia, Italy

M. Marzullo
Studio Arch. Marzullo, Terracina, Italy

G. Valletta
Department of Construction, University of the Republic of Uruguay, Montevideo, Uruguay

ABSTRACT: The project concerns the restoration of a ruined conventual building of the 18th century in
Southern Italy, aimed at the insertion of a school complex. A new project was carried out providing for the
building of new steel and wooden light structures inserted within the fields defined by the in-plan alignments
of the existing walls. The steel structure foundations allow for a minimum interference with the expected
underground finding of graves and burials. The masonry walls will be strengthened by mortar injection and tied
by the insertion of twin steel beams at the floor levels, joined across the wall thickness by means of passing-
through injected bars. A particular solution has been conceived based on the synergetic interaction between
new and existing structure. Special connection systems between new structures and old masonry walls allow
for a triple function: avoiding the interaction in the vertical load bearing capacity, avoiding collapse mechanism
orthogonal to the wall plane through the horizontal bracing effect of the steel frame, using the horizontal load
bearing capacity of the retrofitted and strengthened walls in their plane. The adopted structural and constructive
solutions are presented. The restoration works are being currently carried out.

1 INTRODUCTION the saved elements. In this case the restoration of the

This paper is aimed at illustrating a project concerning
the restoration for the reuse of a three hundred years
old ruined monastery, located in a seismic prone area
of the Southern Italy.
When facing a ruined building, both the ruin derives
from a prolonged lack in maintenance, as in the present
case, or from other accidental factors, like earthquakes
or fires, three alternatives are present for its recovering.
The basic one consists of the romantic solution of
leaving the time takes its course. The effects are only
mitigated by means of minimum preservation works
or protective external defences. The ruin is considered
in its archaeological worth, as witness of historical
significance, without architectural utility.
The first of the two operational alternatives consists
of the restoration of the ruin by the reconstitution of a
copy of the original building. The result is pursued by
means of the substitution of the decayed elements and
the volumetric, structural, architectural and decorative
integration of the lost portions: the lost fragments are
rebuilt on the model of the original ones, as result-
ing from the existing documentation and the survey of
building is approved by an accurate historical research,
aiming at recovering the original building researching
a continuity with the past and, paradoxical, also erasing
the signs of the time not contemporary to the original
“facies”.
The second design alternative concerns the restora-
tion of the ruins through the research of the contrast
with the original building. In this case the existing
fragments are re-connected by means of new elements
resulting complementary but physically and visually
separated, thanks to a differentiation of the mate-
rial and shape. The restoration is carried out through
the building of elements independent from the orig-
inal ones, guaranteeing a rigorous protection of the
construction authenticity.
While in the first solution it is necessary to avoid the
simple recovering of the ruins to their original status,
in the second case a recover of the ruin as a scene shall
be avoided.
Between these two alternative hypotheses, a differ-
ent solution, the so called “critical restoration”, has
grown in the last decades: it corresponds to the respect
of the superimposed layers. The methodology does

1061
not have univocal pre-defined solutions, but adapts
itself, case by case, recognising the knowledge (data
from archives, analyses of the materials in situ, direct
investigations) and the special and unique nature of
the single building.

2 CASE STUDY BUILDING: HISTORICAL

INFORMATION AND EXISTING
SITUATION

The project concerns the ruins of the ancient

Monastery of Salemme in Sanza, a small village
located in the interior part of Campania, a region of Figure 1. General plan of the complex.
the central-southern Italy. The monastery is located
in a seismic area characterized by medium seismicity
according to the Italian hazard zoning.
The monastic complex of the Saints Joseph, Peter
and Paul was founded by Don Giuseppe Teutonico, son
of Gaetano, from Salerno, and Agata Vittoria Barzel-
loni, from Sanza (as it results from the “Archivio
Barzelloni”) probably before the year 1720. This is
the date of the altar sculpted by Andrea Carrara for a
chapel and now located in the St.Vito’s chapel. The
naming of “Salemme” is probably due to the fact
that it was used by monks of the Franciscan Order
of Jerusalem.
Andrea Carrara, from Padula, also sculpted on 1742
the busts of the Saints Joseph, Peter and Paul, now
located in the church of Santa Maria Assunta (resulting
from the “Catasto Onciario” of 1753).
In the second half of the 19th century, one hundred
years since the foundation, the complex was already
abandoned and the chapel was used as burial, like the
“extra-moenia” chapel of Santa Maria del Popolo.
The ruins include two small buildings tied together
by a perimeter wall closing a large area probably
assigned for agricultural use. The first building (A),
more complex, was used for the monastic life: it
consists of a prolonged portion and a number of rect-
angular rooms, one of which seems to be a small
church. It is a 2-stories building with a 3-stories por-
tion. The second building (B) consists of two volumes:
a low volume, probably covered by a roof and used for
recovering animals and a second volume, next to the
previous one, showing the ruins of a 3-stories small
tower, covered by a roof, with a top decoration. Figure 1
shows a general plan of the complex where the two
buildings, A (the upper one) and B, can be identified.
Figure 2 shows some sections of the existing build- Figure 2. Sections of the existing buildings.
ings. Figure 3 reports the southern lateral view of the
complex. it appears as ruins (Figures 5, 6, 7). Failed portions
The walls are made of multi-leaves stone masonry occupy the internal spaces. All the roofs and wooden
(Figure 4). The exterior leaves are made of stone blocks floors totally collapsed. Only some portions of the
roughly shaped, while the interior part consists of many masonry walls stand, while large portions, especially
little blocks. the crowning/coping, collapsed. The lack of protec-
The building has undergone strong decay phenom- tion from the environmental and atmospheric agents
ena in the last decades, due to the neglect status, and caused the decay of the residual walls, particularly the

1062
Figure 3. South view of the complex.

Figure 5. Porch of building A.

Figure 4. Detail of the multi-leaves stone walls.

Figure 6. External view of building A.

lack of compactness, due to the loose of mortar. Also

all the wooden architraves are lacking. The general
stability of the walls is unsafe.

3 PROJECT FOR RE-USING

The Venice Charter (1964) stated that “The conser-

vation of monuments is always facilitated by making
use of them for some socially useful purpose. Such
use is therefore desirable but it must not change the
lay-out or decoration of the building”. In the present
case, the rehabilitation project is a part of a larger
project concerning the realization of the Centre for
the Environmental Education of the “Cilento e Vallo Figure 7. Statical criticity local situations.

1063
di Diano” National Park. Within this project a num-
ber of different initiatives are provided, supporting a
unique territorial complex in which the protection of
the environment and the social and economic devel-
opment of the activities related to the development
of the park will harmonically coexist. Two areas were
pointed out for pursuing the project goals: one, related
to the urban activities, consists of an existing build-
ing previously hosting the Rangers’ barracks; another
one consists of the ruins of the so called Monastery
of Salemme - Santa Maria delle Stelle, located at the
slopes of the park’s mountains. Figure 8. Initial project. Plan of buildings B and C.
The first initiative of the project to be put into effect
concerns the construction of the Centre for the Envi-
ronmental Education, composed by a school of jour-
nalism and by a number of information and learning
activities, in the ancient Monastery of Salemme. The
recovery of the ruins will allow for an information desk
of the visitors, a didactic approach before the visits,
a formation school of environmental journalism. The
project solution hypothesizes a use of the Monastery
even apart from the ordinary institutional activities:
the didactic laboratory will allow holiday visits to the
complex also to the inhabitants of the territory around, Figure 9. Initial project. Plan of building A.
a small restaurant will allow a more comfortable stay-
ing. Moreover the complex will include a new building
(called C) devoted to a conference hall, of about 100
places, promoting the fruition of the building over the
use as school and information centre.

4 INITIAL CONCEPT SOLUTION FOR THE

RETROFITTING

A preliminary design hypothesis provided the con-

struction of r/c structures sticking to the existing
masonry walls and relevant demolition and reconstruc-
tion works (Figures 8, 9, 10). This kind of work does
not appear appropriate with the aim of the conservation
of the historical portions of the building.
Since the Athen Charter (1931) it was recom-
mended a “judicious use of all the resources at the
disposal of modern technique and more especially of
reinforced concrete” and specified that the “work of
consolidation should whenever possible be concealed
in order that the aspect and character of the restored
monument may be preserved”, moreover it was recom-
mended the adoption “in cases where their use makes
it possible to avoid the dangers of dismantling and
reinstating the portions to be preserved”.
The preservation principle is also reminded in
the cited Venice Charter (1964): “The process of
restoration is a highly specialized operation. Its aim is Figure 10. Initial project. Sections of building A.
to preserve and reveal the aesthetic and historic value
of the monument and is based on respect for original distinct from the architectural composition and must
material and authentic documents. It must stop at the bear a contemporary stamp.”
point where conjecture begins, and in this case more- According to this principle, the preliminary project
over any extra work which is indispensable must be was considered not feasible and therefore rejected

1064
by the superintendence authority on the historical
heritage.
5 CONCEPT SOLUTION OF THE FINAL
RETROFITTING DESIGN
A completely new concept solution has been con-
ceived. The building restoration provides for a number
of works allowing for obtaining a significant increase
of the static and seismic safety of the building con-
sidering the present situation of failure and decay.
The typical provided works for each class of statical
problems are presented in a following paragraph.
The seismic enhancement of the existing structures
has taken into account both the Codice dei Beni Cul-
turali (2005) and the Linee Guida (2006), that is the
most recent Italian guidelines concerning the preser-
vation of the cultural heritage, with particular attention
to the constructions located in seismic areas. The pri-
mary goal offered by the provided works consists of
the reduction or elimination of the main vulnerability
causes and can be resumed in the following items:
• increasing the strength of the existing stone masonry
walls;
• tying the building in correspondence to the levels
of intermediate floors and roof;
• conferring a suitable flexural strength to the
architraves.
The seismic enhancement is completed thanks to
the interaction of the enhanced old structures with the
new ones that guarantees:
• the behaviour of the floors as horizontal diaphragms;
• the sustain against the overturning mechanisms (1st
mode mechanisms in general);
• the reduction of the wall length included between
successive orthogonal walls;
• the effective connections between walls and floors.
A particular solution has been conceived based on
the synergetic interaction between new and existing
structure. The new project was carried out provid-
ing for the building of new steel and wooden light
structures to be inserted within the fields defined by
the in-plan alignments of the existing walls. The steel
structure foundations allow for a minimum interfer-
ence with the expected underground finding of graves
and burials. The masonry walls will be strengthened
by mortar injection and by the insertion of twin steel
beams at the floor levels, across the wall thickness
by means of passing-through injected bars. Special
connection systems between new structures and old
masonry walls allow for a triple function: to avoid
the interaction in the vertical load bearing capacity,
to avoid collapse mechanism orthogonal to the wall
plane through the horizontal bracing effect of the steel
frame, to use the horizontal load bearing capacity of the
1065
retrofitted and strengthened walls in their plane. Suit-
able technological solutions have been adopted allow-
ing for the compatibility with the existing materials
while evidencing the separation with the existent.
6 DETAILS OF THE DESIGNED SOLUTION
6.1 Consolidation and strengthening of stone
masonry
The following works are provided for the consolida-
tion of the existing stone masonry walls and their
strengthening.
Demolishing of unstable portions. The situation
concerns some vertical cantilevered walls and some
unstable architraves.
Deep refilling of mortar joints. This work concerns
all the walls in a generalised way and consists of taking
away and restoring the mortar of the external portion
of the joints among the stone blocks.
Injection of the elevation walls. All the walls are
concerned by a lack of link and compactness in the
depth. To overcome the lack of mortar, a generalized
injection with fluid no-shrinkage lime mortar is pro-
vided. Ultrasonic tests will be carried out before and
after the injection for evaluating its effectiveness in
increasing the compactness.
Rebuilding/Adjusting the top portions and ruins.
The top portions are rebuilt with masonry having
the same characteristics (blocks and texture) of the
existing one. The insertion of reinforcing steel bars
is provided, so that the top portions will include a
reinforced masonry tying beams.
Drilled steel connectors. Many L (angular) and
T (intermediate) connections among walls are com-
pletely or partially lacking in continuity and do not
give an effective solidarity and connection. In all these
cases is provided the tying of the orthogonal walls
by means of bars inserted within holes drilled in the
masonry and injected with the same mortar used for
the masonry injection.
Existing foundation strengthening. The existing
masonry foundations will be strengthened by means
of deep injection, slightly reinforced, performed in the
foundation and bottom part of the masonry (Figure 11).
Top tying beam in reinforced masonry. The actual
situation is characterised by the lacking of longitudinal
link at the top of the masonry walls. Moreover, the
most part of the walls are completely lacking of the top
portion, already collapsed.A reinforced masonry tying
beams is provided at the top of all the walls (Figure 12).
Intermediate steel tying beams. At the levels of the
intermediate floors of the internal new structure lon-
gitudinal steel tying beams will be built consisting of
two U-shaped steel beams, coupled through the wall
thickness by means of anchoring bars injected in holes
drilled in the masonry (Figure 13).
A multiple structural effect results:
• obtaining a longitudinal continuity of the walls at the
intermediate level (at the walls’ top, the reinforced
masonry tying beam is provided);
• creating a connection system with the new structure
for the mutual bracing against horizontal actions;
• tying effect at the floor level contributing to equili-
brate some static horizontal actions (e.g. the pushes
from the arches).
Closing of cracks. Many cracks are present, some-
times passing through and in some cases having
Figure 11. Consolidation of foundations.
Figure 12. Top tying beam in reinforced masonry.
Figure 13. Intermediate steel tying beams.
1066
relevant width. Different kinds of work are provided
for the different situations:
• removing and repositioning of blocks along the
cracks;
• mortar injection and tying effect given by the tying
beams;
• connection by means of slightly reinforced injected
holes.
Re-making up architraves. Almost all the open-
ing are lacking of their architraves. They will be
re-made up according to a flexural resistant arrange-
ment consisting of steel beam covered by wood boards
underneath.
Re-making up failed openings. In some cases the
openings, not only are lacking in architraves, but are
failed along their total perimeter. They will be restored
rebuilding the stone masonry of the horizontal and
vertical sides.
Closing niches and flues. The walls integrity will
be pursued by the filling of the voids by effec-
tively interconnected stone masonry having the same
characteristics of the existing ones.
Re-opening pre-existing openings. Many old open-
ings are present, closed with non-continuised infill
masonry tables. If the opening shall be re-opened, the
infill will be removed and the opening will be sub-
jected to the works provided for the failed ones. If the
opening shall remain closed, the same works provided
for the crack closing will be applied.
Arches of the western porch and relevant ones.
Figures 15 and 16 respectively show the situation of the
porch of building A and that one of an internal arch. It
is assumed that the insertion of the intermediate tying
beams gives to the masonry the capacity to sustain the
horizontal pushing forces of the arches. In some cases
specific tying bars are provided at the base of the arch.
6.2 New steel structures
Considering the current status of the structures,
that is particularly sensitive and critical due to the
constructive typology, nature of the site, historical
Figure 14. Solution for the new architraves.
destination, decay status, other witnesses not yet inves-
tigated, the new structures were designed as made of
steel frames with wooden horizontal structures.
This typology, which guarantees punctual founda-
tions, longer spans and more “connections” with the
“existing”, allows to reduce the risk to compromise the
historical witnesses. In each internal area of the exist-
ing building, individuated in-plan by perimeter walls,
only four base supports are present (at the corners), so
allowing for the preservation of the ancient structures,
Figure 15. Arches of the porch of building A.
Figure 16. Typical arch at the interior of the building.
Figure 17. Plan of 1st story of building A.
1067
both those already pointed out in the old chapel and
the other ones still hidden.
The framed structure consists of columns and beams
having standard I-shaped section. The floors are made
of wood with single-aligned or double-aligned beams.
The horizontal bracing is given by crossed tying bars
located in all the rectangles of the structural grid.
Figure 17 show a plan of the building A with the inter-
nal structures. Figure 18 show a section of the same
building with relevant details.
6.3 Synergetic interaction between new and
existing
The new and the consolidated existing structures
perform together in a synergetic way allowing the
enhanced performance of the building against the
seismic attacks.
Special tools (Figure 18) are arranged for connect-
ing the floor beams of the new framed structure with
the steel tying beams running at the interior of the
stone walls at floor level and with the top reinforced
masonry tying beams. They allows for free mutual ver-
tical displacements, while the lateral displacement are
blocked by the bolted connection.
The complex of the new and existing structures per-
form in different ways with respect to the lateral and
vertical actions. The vertical loads of the new struc-
tures, both permanent and live, are carried by the steel
frame and transferred at the base by its columns. The
masonry walls carry their dead load. The behaviour
against the lateral loads is synergetic.
According to the scheme reported in Figure 19, the
seismic actions on the walls, orthogonal to their plane,
are supported by the horizontal elements of the steel
frame, thanks to the connectors.
The lateral actions of the steel frame are partially
supported by light vertical bracing system, included
in the steel frame, and partially transferred to the wall
parallel to the lateral actions (Figure 20).
Figure 18. Section and details of building A.
Figure 19. Transfer mechanism of seismic forces to the steel
structure.
Figure 20. Transfer mechanism of the seismic forces to the
masonry.
7 CONCLUSIONS
A restoration project aimed at the re-use of an old
monastery, at present in a ruined condition, has been
carried out providing for new technological solutions
allowing for the preservation of the existing structures
even decayed.
1068
The design solution provides for the insertion of
new steel structure with wooden floors at the inte-
rior of the existing stone walls, suitably restored and
integrated.
The two systems, the existing one and the new one,
work in a synergetic way with respect to the vertical
loads, but, most of all, against the forces induced by
the seismic action.
The new structures supply the missing perfor-
mances: diaphragm behaviour of the floors and capac-
ity against the overturning; the existing structures
contribute to sustain the seismic forces, thanks to their
high in-plane stiffness and strength.
The adopted solution show the possibility offered
by the synergetic contribution of the existing and new
structures, while evidencing the separation with the
existent construction.
ACKNOWLEDGEMENTS
The Comune of Sanza is the owner of the buildings.
Arch. Mauro Marzullo is in charge of the project. Soc.
Cooperativa EdilAtellana is the contractor.
REFERENCES
The Venice Charter “International charter for the conserva-
tion and restoration of monuments and sites”. 1964. 2nd
International Congress of Architects and Technicians of
Historic Monuments. Venice.
The Athens Charter for the Restoration of Historic Monu-
ments. 1931. 1st International Congress of Architects and
Technicians of Historic Monuments. Athens.
Codice dei beni culturali e del paesaggio. 2005. D.L.
22/1/2004, n. 42. G.U. n. 45 24/2/2004 (in Italian).
Linee Guida per la valutazione e riduzione del rischio sis-
mico del patrimonio culturale con riferimento alle norme
tecniche per le costruzioni. 2006. Dip. della Protezione
Civile e il Dip. per i Beni Culturali e Paesaggistici (in
Italian).
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Structural assessment of earthen walls and techniques for onsite

consolidation and conservation

Sandeep Sikka
Superstructures (Architects + Engineers), New York, USA

ABSTRACT: The areas of Spiti and Kinnaur district in the Western Himalayan region have experienced some
significantly strong earthquakes in the last few decades. Recent seismic tremors have caused some serious struc-
tural damage to the buildings in the region especially to the un-reinforced historic adobe structures. The deforma-
tion and bulges resulted in the adobe structures, as a resistance to the seismic vibrations, are the vulnerable areas,
most likely to collapse in an event of any future horizontal force. These historic structures, which were originally
designed for arid climate, are now facing problem with increased precipitation and regular rainfall for the last few
years. The paper puts forward various repair techniques carried out on the site to consolidate load-bearing walls
with cracks and research data on upgraded component used for the protection or repair work for these building.

1 INTRODUCTION

The Spiti and Kinnaur region in the Northern Indian

State of Himachal Pradesh has some remarkable
ancient Buddhist temples made of Adobe. Constructed
between the 10th and 15th century AD [1], these
temples preserve some of the earliest artistic her-
itage of the Tibetan Buddhism in the form of mural
paintings, polychromed clay sculptures and decorative
wooden ceiling members. This arid region in the West-
ern Himalayas have witnessed a gradual development
of Buddhist temples architecture from simple single
storey buildings constructed on a relatively flat land to
a complex maze of multistory fortress on the mountain
top, over a period of 500 years. Standing on a highly Figure 1. Nako Temple Complex, Himachal Pradesh, India.
seismic zone and having survived for several centuries
the buildings are undoubtedly a living evidence of
out to stabilize and repair the two load bearing temple
highly engineered structures. These historically well-
walls at the Nako temple complex in Kinnaur, have
crafted temples structures originally designed for arid
been discussed as a case study in his paper.
climates today lie susceptible to an annual precipita-
tion of 200 mm–400 mm due to the globally changing
climatic patterns. Recent seismic tremors have caused 2 NAKO TEMPLE COMPLEX
some serious structural damage to these un-reinforced
historic adobe structures buildings. The deformation, The great translator Lotsawa Rinchen-Zangpo (958–
bulges and cracks in the load bearing adobe wall, 1055 A.D.) reportedly founded the historic Buddhist
as a resistance to the seismic vibrations and mois- complex at Nako in the Northern Indian state of
ture, are the vulnerable areas, most likely to collapse Himachal Pradesh. The Buddhist temple complex at
in an event of any future movement in the building. Nako is an arrangement of four dedicated unicellu-
A detailed assessment of the condition of structures lar temples along a common courtyard. The site of
after the recent earthquakes with a methodology and the temple complex gradually slopes from the east
results of much needed on-site research on reinforce- to the north clockwise in the four directions with the
ment of structural components, followed by techniques northern edge lying at a difference of about 3 meters
and methods of repairs which was and can be carried from the common courtyard of the complex. The

1069
four temples in elevation rise to a height of about
5 to 6 meters from their respective entrance lev-
els on the exterior. The two lower temples Lotsawa
Lhakhang dedicated to the great translator Rinchen-
bzangpo and Karjung Lhakhang ‘the white temple’
face east lie on a relatively flat ground are mentioned
in this paper as above ground structures. The other
two temple Gongma Lhakhang and the Gya-pag-pai
Lhakhang facing west are mentioned as below ground
or semi-subterranean structures because the ground
level outside these temples is much higher than the
interior floor level.This historic temple complex which
houses invaluable artwork, murals, polychrome clay
sculptures and decorative wooden ceilings, survived
for more than 800 years, is showing alarming signs of
deterioration.
NakoTemple Complex was nominated and included
in the 100 most endangered sites of the world by WMF
in 2001. Since then efforts have been made every year
under Nako Preservation Project [2, 3] to protect and
stabilize the structure and its artwork.
In the year 2006 and 2007 Institute of Conser-
vation and Restoration, University of Applied Arts,
Vienna,Austria carried out a conservation and research
campaign [4] at the temple complex with limited
funds from the University and Austrian Development
Agency (ADA). The conservation project team mostly
focused on the following tasks.
2.1 Art work conservation
Task 1: The restoration of damaged clay sculptures
inside the Lotsawa Lhakhang,
Task 2: Stabilization and cleaning of wall paintings
Task 3: Stabilization and cleaning decorative
wooden ceiling panels inside the Gongma
Lhakhang.
2.2 Structural conservation
Task 4: Structural assessment and stabilization of load
bearing wall at Gya-pag-pai Lhakhang and Karjung
Lhakhang.
This paper is concerned with the structural issues
and describes the research and repairs for structural
conservation work carried out on the two temples
(Gya-pag-pai Lhakhang – Semi subterranean struc-
ture and Karjung Lhakhang – Above ground structure)
to accomplish Task 4.
3 ARCHITECTURE OF BUDDHIST TEMPLE
COMPLEX AT NAKO: MATERIALS AND
METHOD OF CONSTRUCTION
The design of Buddhist temples are simple, rectangu-
lar geometric spaces with carefully designed structural
members unveil an outcome of years of trials, wisdom
1070
and extreme climatic conditions and natural disas-
ters. The walls of these historic structures are made of
(Adobe) a sun-dried large sized mud brick laid in mud
mortar with the foundation in rubble stone masonry,
generally rests on a stable solid ground1 . The thick-
ness of the walls varies from 2½ feet to about three
feet. Vertical measurements shows that sometimes the
outer faces of the walls are slanted so that the wall
thickness is wider at the base and gradually tapers to
the top, providing extra stability to these tall and flex-
ible structures. Due to the cold climatic conditions for
most time of the year, the openings in the walls of these
temples are minimum possible and contribute to less
than 5% of the total wall surface in a rectangular space
and are located at the center of the wall. Besides the
wooden doorframes such punctures in the walls are
reinforced with thick vertical and horizontal tensile
wooden members connected to each other with flex-
ible joint. Series of wooden lintels laid next to each
other along the thickness of the walls are anchored
deep and extend into the masonry on both sides of the
openings like additional horizontal tie members. The
load bearing walls of the temples are reinforced with
a wooden framework of horizontal wall ties, with a
cross section of 6 to 8 in width and 4 to 5 in height,
as ring beams around the building, installed externally
and internally flushed with the surface of the wall.
The ring beams tie the entire structure together with
each beam running at a distance of approximately 1½
or 2 meters to the other. The exterior horizontal wall
ties were joined to each other at the corner externally
with wooden vertical ties, now missing in most of the
temples. The arrangement thus prevents any outward
movement during any seismic vibration and these wall
ties, along with the wooden lintels disrupt structural
cracks that could otherwise extend the full height of
the wall eventually causing total collapse. Strength of
the adobe and the mortar joints vary all along the wall
hence differential loading. The horizontal ties there-
fore redistribute the load evenly through out the wall.
In addition to these horizontal members, the load-
bearing walls of the temples in Nako are reinforced at
the corners with buttresses at the corners. These but-
tresses were added later to support the masonry, after
the building was struck by an earthquake separation of
corner walls. The buttresses are random rubble stones
stacked one over the other against the corners of the
building forming a pyramid. The roofs of the temples
are flat and are made of mud laid in various layers and
compacted. About 7 inches of compacted mud rest on
2 inches thick rectangular wooden panels or a mesh of
willow twigs, with a layer of local shrubs or birch bark
sandwiched between the two for waterproofing. These
1
Trial holes dug in the past next to the outer wall revealed
that the foundations generally rests on either rocky outcrops
or solid ground.
are in turn supported over wooden rafters and beams,
which are further supported directly on load bearing
mud walls and wooden columns.
4 STRUCTURAL ASSESSMENT
4.1 Methodology of diagnosis
With budget limitations, inaccessibility to testing lab-
oratory, investigation equipments and limited time
frame, the investigation methods selected for the
structural assessment of two temples were primarily:
• Visual inspection
• Study of previous data on documentation of wall
sections, strength and properties of masonry com-
ponents from published and unpublished reports [2,
3, 5, 6, 8]
• Load calculation through study models.
4.2 Semi-subterranean structure
The temple of Gya-Pag-Pai, which is, mentioned as
semi-subterranean structures because the ground level
outside the building is higher than the inner floor
level was affected the most and has been under enor-
mous structural stress after the earthquake of 1975.
A detailed documentation of the walls in the past [3,
4] has revealed that there is tremendous movement in
the upper portion of the load bearing walls. The lateral
outwards horizontal movements in walls have resulted
in large-scale bulges and cracks in the masonry.
Increased moisture due to recent changes in the cli-
mate has further caused enormous seepage into the
interiors and has also disturbed the structural integrity
of the temples. The roof of Gya-Pag-Pai due to the
shortage of funds has been covered temporarily, by
the locals, with corrugated galvanized iron sheets. The
focus of the structural work, as mentioned in the task
4, was to assess and address immediate threat to the
structures. Survey of temple walls and measurement
of moisture inside the masonry through conductivity
meter revealed that there is large-scale ingress of water
and moisture from the surrounding ground.
Due to inadequate length of the wooden spouts
added recently on the adjacent building’s roof and in
absence of any proper drainage facility from the roof
of the Gya-Pag-Pai, the rainwater and water from the
melting snow drain next to the load bearing walls at the
rear of the temple. Consequently the water seeps into
the building walls and foundation causing enormous
damage to the water-soluble un-reinforced masonry.
In addition to this the site of the temple complex
slopes down along the east-west direction and con-
sequently the excess water draining from the village
on the higher level towards the rear of the temple con-
tributes excessive moisture to the ground adjacent to
the load-bearing wall. As a result the soil outside the
1071
Figure 2. Horizontal crack in the load bearing wall of the
semi-subterranean structure.
building swells after gaining moisture, which exerts
horizontal pressure on the load bearing walls. The rear
wall of the Gya-Pag-Pai Lhakhang is unable to resist
the horizontal shear stress along the floor level outside,
resulting in a major horizontal structural crack. (See
figure 2).
Due to inadequate drainage the excessive mois-
ture accumulated outside, gradually seeps into the
wall causing serious damage to the masonry, plas-
ter and painted surface inside. Problems of salts and
other agents of chemical, environmental and biolog-
ical decay followed water seepage causing severe
damage to the 800-year-old structure. The compacted
mud flooring of this semi subterranean structure was
replaced with impermeable cement concrete flooring
in the year 1996 AD, has aggravated the problem of
moisture many times. The Cement concrete flooring
prevents the effective evaporation of water rising from
the ground that eventually evaporates from the walls.
Constant drying and wetting of the painted area causes
deterioration and loss of binding strength. The inte-
rior plaster decomposed into a powdery mass, has
started leaving the surface, revealing the masonry
below. There is also a drastic change in the humid-
ity and level of moisture content in the walls after the
insertion of cement flooring that have not only affected
the strength of the plaster but has also affected the
strength of the mud blocks. Another intervention made
recently to the structure was cladding of the base of
porous earthen walls with impermeable stone slabs on
the exterior, with intension to avoid basal erosion from
water and wind, resulted in accumulation of moisture
inside the masonry. The moist adobe walls with hori-
zontal structural cracks and deteriorated base are not
only vulnerable to any seismic vibrations but are also
a major threat to human life and artwork stored inside
the building in an event of collapse.
Figure 3. Thrust created by the expanding ground due to moisture gain next to the wall results in a horizontal crack.
4.3 Above ground structures
Of all the temples in Nako the condition of above
Ground Temple Structure Karjung Lhakhang is the
worst. During the earthquake of 1975AD, the walls
of the Karjung have developed major, vertical and
diagonal structural cracks on all sides of the building
resulting in overall destabilization of the structure.
Outward movement in the walls due to excessive
load from the roof and thrust due to seismic vibra-
tions, have caused separation at the corners, resulting
in vertical cracks. This above ground structure is tem-
porarily covered with galvanized iron sheet above the
damaged roof with no proper gutters, leader or roof
drainage system to canalize the water. Consequently
water from rain and melting snow draining along the
slope of the sheets, eventually runs along the exterior
surface of the wall eroding the plaster and the masonry
forming large crevasses. These crevasses thus formed
are the most vulnerable areas in an event if a seismic
vibration or building movement in the future. Diag-
nostic investigation through visual inspection revealed
that is large-scale damage to the upper portion of the
outer Wythe of wall partly due to the earthquake and
partly due to water seepage that had been repaired by
the local craftsmen with big stone pieces. The mud
blocks at the upper portion of the wall have deterio-
rated due to the ingress of water, resulting in the loss
of strength. These mud blocks below are unable to
take the load of big stones resulting in an outward
bulge. These areas have now become major threat to
the rest of the wall below and over all stability of the
structure.
1072
5 MITIGATION AND REMEDIAL MEASURES
On-site visual inspection of the load-bearing wall of
the two 12th-century Buddhist temples at Nako reveled
some intersecting conditions, which required judicious
research for development of components for repairs to
avoid further deterioration of masonry and eventual
collapse.
5.1 Methodology for development of remedial
measures
The methodology of repairing semi-subterranean and
above ground structures requires detail understanding
of its materials, its chemical and physical characteris-
tic, their present condition and future behaviour and
vulnerability to during any seismic event, unforeseen
loading and exposure to moisture. Due to remoteness
of the temple complex, limited funds and inacces-
sibility to lab and issue of sustainability of repairs,
mentioned previously, it was decided to find locally
available materials for repairs. Another reason for
choosing local materials was to maintain the character
of the building.
5.1.1 Experimental program for improvement of
Adobe block
For the production of adobe for the repair of the exter-
nal Wythe of the structural walls, it is essential to study
their behaviour and strength against the increased load
and the changing climatic conditions. To improve the
strength of the adobe and their resistance to water, it
Figure 4. The samples were tested for their Flexural
Strength through an apparatus designed specifically for test-
ing the adobe in the field; using compressed earth block units Figure 5. Results of strength test conducted on four samples
of 7 kg. of adobe on site.

Table 1. Composition and weight of sample adobe blocks.

was decided to conduct an experiment with various
samples of adobe made with the locally available soil Sample no. Composition of blocks Wt. (Kg.)
with additives and stabilizers to impart extra strength
S1 Sieved local soil 23
and resistance against moisture.
SC Sieved local soil mixed 23
with 10% local Clay (9:1)
5.1.2 On-site strength test SL 10% Hydrated Hydraulic Lime 24
An apparatus designed specifically for measuring the and sieved local soil
strength of the adobe in the field was used for deter- SCC 6% Cement and local sieved 7
mining the 3-point flexural strength (See figure 4). soil (compressed)
The 3-point loading was applied using series of com-
pressed earth blocks each weighing 7Kg. The particle
size analysis (see figure 7) of the local soil and clay 5.1.3 On site erosion test
indicated that the percentage of sand content in the soil The water through a spray gun of pressure 50 K pa
was far more than the required quantity. Therefore to was thrown directly at the centre of each sample at a
increase the strength it is essential to increase the quan- distance of 600 mm.
tity of the binder in appropriate proportion. The mix Each sample was exposed to equal water pressure
thus prepared with various binders like cement, lime for about 180 seconds. It was studied that the spray
and locally available clay for extra strength were then jet used in this test has flow rates much higher than
put to test. Four kinds of adobe samples were then pre- the combined effect of rain and wind documented in
pared with relatively dry mixes, wet compacted inside this region. The test indicates the resistance of adobe
the moulds and were left to dry for four weeks. to the direct effects of fine spray of water. The ero-
sion was carefully measured perpendicular from the
5.1.2.1 Results and discussion original surface to the eroded pit with a measuring
It was observed that the addition of 10% local clay scale.
or 10% lime as a binder significantly increases the
strength of the adobe. Further tests with varied propor- 5.1.3.1 Results and discussion
tion of clay and lime can be tested before the manu- Addition of local clay 10% by volume increases
facture of adobe for conservation work. The stabilizers the resistance of adobe against water erosion. Addi-
and additives, which perform well for increasing the tion of chopped straw also significantly increases the
strength, may or may not perform well against water cohesion and resistance to water.
erosion. To generate a solution for the stabilization
of the adobe with appropriate stabilizing materials
5.2 Experimental program for development of
against the increased precipitation, a scientific study
stabilized sacrificial external wall plaster
of the samples of the various adobe formulated from
the local soil and clay with different stabilizers were To generate a solution for the stabilization of the locally
tested on site for their resistance against moisture. available soil with appropriate stabilizing materials
These blocks were then tested on site for their perfor- against the changing climatic conditions, a scientific
mance against general weather conditions and water. study and technical examination was carried out at
The final mix design should also impart strength, as Nako of the samples of the various plasters mixes for-
they have to take the load of the roof at the same time. mulated from the local soil and clay with different

1073

Figure 6. Result of spray erosion test conducted on six
stabilized adobe samples.
Figure 7. Particle size analysis of local soil and local clay.
stabilizers. It was also studied that the additives for
stabilization of plaster cause no adverse effects on the
historic structure as they render a coat of water proof-
ing layer externally and at the same time keep the walls
porous.
These mixes were then tested on site for their perfor-
mance against general weather conditions and water.
The mix finally designed should also impart strength
by consolidation to the otherwise fragile surface of the
historic monasteries.
To determine the exact mix of the stabilizer the first
step undertaken was to understand the composition of
the local soil and clay and their behaviour.
The data obtained from the laboratory analysis of
the local soil samples helped in understanding the
physical nature and behaviour of the existing soil and
the local clay in the Nako Temple Complex, Kinnaur.
The following results thus obtained can be read.
The plasticity index indicated a low cohesive
strength when plotted against the quantity of clay
1074
present in the soil sample. To increase the binding
properties and inter-particular binding strength of the
local soil it is required that their elementary proper-
ties are modified so that their particles remain firmly
united and unaffected by moisture variation. Addition
of lime will certainly change its chemical properties
in which the clay of the soil reacts with lime to form
a relatively strong binder. This hypothesis was then
tested on site at Nako temple complex with other
locally available stabilizers into different plaster sam-
ples. The samples were then monitored for two weeks
and finally tested and documented for their durability
against water pressure.
5.2.1 Composition and application of plaster
samples
It was decided that the plaster samples should be
applied and tested on the ancillary structure of the
Temple Complex before any trials with the historic
fabric.
The wall surface selected for application was even,
completely dry and faced no problem of seepage or
dampness. The surface received adequate sunlight,
natural rain and wind. It was also exposed to the tem-
perature variations throughout the day. The first step
towards surface preparation for the application of the
plaster included scraping with a wire brush to create
undulations on the surface, so that the new plaster can
adhere to the wall. The following step involved wet-
ting of the wall surface properly with water spray at a
very low pressure before application. This was done to
avoid suction of water into the wall from the new plas-
ter mix. The various plaster mixes were then prepared
in a mixing tray with a trowel taking into consideration
the percentage of water the minimum of which should
be able to stick to trowel and spread evenly on the wall
surface. Fifteen samples of the plaster mixes (<6 mm
thick) with different local materials were then applied
each on Feet Square of area of the wall. These fif-
teen plaster samples [7] were monitored for two weeks
and then tried against the wet erosion test. The water
through a spray gun of pressure 50 Kpa was thrown
directly at the centre of each sample at a distance of
600 mm. After starting from the bottommost plaster
sample; each sample was exposed to equal water pres-
sure. It was observed that the spray jet used in this
test has flow rates much higher than the rain obtained
in this region. The process of erosion was carefully
documented for the failure of plaster samples against
continuous spray for three minutes. It was observed
that the finer particles (comprising the soil) washed
out of the plaster mixes creating initial pits on the
surface and finally resulting in absolute failure. The
plaster samples with lime as a binder and a stabilizer
could resist water pressure for a longer time. Perme-
ability of the water into the plaster was also tested on
site, simply by digging the wet plaster and measuring
Figure 8. Fifteen plaster samples tested on-site for water
erosion.
Figure 9. Results of water spray erosion test.
the depth of water penetration till the substrate. It was
found that the plaster mixes stabilized with lime had a
penetration of water <6 mm into the surface.
5.2.2 Results and discussion
The results obtained from the above mentioned tests
could be interpolated and studied in the following
graph. The intention of the exercise was to generate
a sustainable solution against the changing climatic
conditions that can be used as a sacrificial coat for
structural adobe walls.
The selected mixes as shown in the graph (P8, P9,
P10) after the tests are not only stable against mois-
ture but at the same time porous enough to facilitate
evaporation of moisture which is trapped behind so
that the internal painted surface on the other side is
not affected. The strategies thus derived after interpre-
tation of the above results for the surface protection
of these historic earth structures against the changing
climate are not only practical and economical, but at
the same time address the re-strengthening of thousand
years old local traditions.
1075
Figure 10. Eight feet deep drain dug at a distance of 2 meter
from the walls of semi subterranean structures.
5.3 Structural stabilization of load bearing wall for
semi subterranean structures
With budgetary constrains it was not possible to con-
solidate the masonry walls and replace appropriate
roof from Gya-pag-pai therefore it was decided to
address the structural and water seepage problems in
mentioned in task 4, by providing appropriate drainage
system both for the existing roofs and for the surround-
ing ground. With the help of the local craftsmen and the
village community who shared their experience, opin-
ion and labour happily with the conservators, it was
possible to design a state of the art drainage system for
the two historic structures. The drainage system, which
was similar to a French drain commonly, used for effec-
tive drainage of historic buildings in most part of the
world was designed and detailed carefully with local
materials and technology. A detailed sketch explaining
the section of the drainage is illustrated below (Fig.14).
A drain 2.5 meters wide placed exactly 1.5 meters
away from the buildings was dug along the load
bearing walls of the two semi-subterranean temples.
The depth of the drain was kept 8 feet (about 2.4
meters) that was decided according to the internal floor
level. The bottom surface of the drain was kept at least
2’ below the internal floor level of the temples so that
all the water is drained out without any possible danger
of seepage into the surrounding area.
The perforated terra cotta pipe laid at the bottom of
the drain is protected with steel bars laid across and
Figure 11. Perforated clay pipe laid at the bottom of the drain sloped away from the site.

above the pipe. The drain was then covered with stones
of descending size till the ground level for effective
evaporation. The drainage had outlet points for effec-
tive cleaning and the slope of the drain was designed
as per levels of the surrounding site to effectively drain
water away from the building complex. This state of
the art inception of the drainage system was tested and
monitored for a year for its effectiveness by checking
the moisture levels in the ground outside the building
and also inside the masonry. Repairs to the horizontal
cracks will proceed as the moisture subsides.
5.4 Stitching of diagonal and vertical cracks in
above ground structures
The cracks and crevasses developed in theKarjung
Lhakhang are the areas of the potential weakness in
the load bearing walls need to be repaired and sta-
bilized to provide structural continuity. The crevasses
are mostly along the cracks and in such case stitching
would be a sympathetic alternative to strengthen and
retain most of the original masonry.
Stitching adobe buildings have been developed, dis-
cussed and published before by many professionals.
Techniques of stitching a crack in the load bearing
wall with mural paintings on the inner face in a highly
seismic zone is certainly a challenge and have not been
tried or been discussed before.
As part of the research and development of a sus- Figure 12. The structural cracks have been raked and the
tainable strategy for repair at Nako, it was decided to masonry is reinforced again with thin slate tiles.

1076
do an example as trial with local material and tech-
niques that could be monitored for its effectiveness on
the bases of its performance. Severely weather portion
of the load-bearing wall at the rear of this above ground
structure was selected for making repairs. The surface
of the wall which is about 2 –6 thick was chased to
a depth of about 1 –3 into the wall, with a pointed
Slater Hammer about 10 to 20 cms on both side of the
cracks on the weathered surface to form a space for
resting the stitch material. Generally masons or con-
servators tend to widen the crack to effectively rest the
repair material. The wider cracks require substantial
material to hold on to the parent masonry to provide
enough flexibility as well as a strong adhesion to react
appropriately to any building movement.
Hence, widening of crack was avoided and the stitch
material was inserted into the masonry by making
small slits into the adobe blocks in zigzag manner
alternately to bridge the crack to achieve extra strength.
Study of ancient adobe blocks at Nako temple
complex indicate the use of varied size of course
aggregates and slate stones to gain extra reinforce-
ment and strength. Thin Slate tiles are locally available,
lightweight and have been use extensively for the con-
struction of traditional houses. It is well known, learnt
and documented by several locals and professional in
the Himalayan region that if the slates tiles are well laid
horizontally in layers inside a masonry wall, performs
well during an earthquake. According to the survey
conducted in 2003 the stone masonry structures with
such horizontal band of slate tiles were affected less
during earthquakes of 1975 in the region as compared
to similar stone masonry structures without such band.
Slates tile and local fiber reinforced lime-stabilized
mud-mortar as filler was therefore selected as stitch
material for cracks.The tiles were laid into the slits hor-
izontally and voids in-between were then filled with
mortar. This was repeated up to the top of the crack.
Addition of fibers in the mortar mix reduced shrink-
age and created better cohesion. The stitched portion
of the wall was then rendered with lime stabilized mud
plaster discussed above. A stabilized soil with lime and
jaggery grout previously tested in the lab for its per-
formance was injected from the inner painted surface
to fill up rest of the void along the remaining width of
the wall.
5.5 Removal and replacement of outer Wythe
of wall
With diminishing knowledge of traditional mud bricks
and repair techniques and sudden earthquake of
1975 resulted in some serious interventions in terms
of masonry repairs. Most of the upper courses of
the masonry near the roof, externally, have been
replaced with big stone blocks that have eventually
detached or are budging outwards due to excessive
load on the walls. Replacement of stone with improved
1077
adobe blocks, tested through onsite research, discussed
above, will be used for structural repairs to the outer
Wythe of the Karjung Lhakhang’s load bearing wall.
Finally drain pipes and gutters were designed and
installed at the roof level of all the temples. The corru-
gated sheets covering the roof were fitted with gutters
molded with flat galvanized sheets. It is proposed to
remove the cement flooring inside the temple to relive
the masonry and flooring from trapped moisture.
6 CONCLUSION
There are no ready-made answers, no ideal material
and no ideal solution to the complex problems these
earth structures are facing. For sustainable repair and
conservation of these historical construction there is a
need to understand the structure, materials and causes
of damage in order to employ the appropriate remedia-
tion. With limited budget, inaccessibility to laboratory
and treatment materials and distance from the civiliza-
tion, in the middle of the high Himalayan Mountains
more than 3500 feet above sea level, it was essential
to find locally available materials and develop sus-
tainable repair and maintenance techniques to through
on site research which can be carried out as an when
required by the local craftsmen to stabilize 800 year
old structures against further seismic vibrations and
changing climate.
REFERENCES
1. Luczanits, C. 2004. Buddhist sculptures in clay.
Early Western Himalayan Art, late 10th to early 13th
centuries. Chicago: Serindia Publications.
2. Khosla, R. 2003–2005. Report on Nako Preservation
Project, Phases-I, II, III and IV. RKDS: New Delhi
(Unpublished).
3. http://athene.gis.univie.ac.at/project/nako
4. http://www.dieangewandte.at/konservierung/stories/
story Reader$66
5. Sikka, S. 2003. Research report on Conservation of
historic earthen structures in the Western Himalayas-
Museum of archaeology and anthropology. Univer-
sity of Cambridge: England (Unpublished).
6. Sikka, S. 2002. Conservation of earth structures in the
Western Himalayas- MA Dissertation Bournemouth
University: England (Unpublished).
7. Sikka, S. 2003. A research on Stabilization of Plaster
Samples for conservation of historic earthen struc-
tures in the Western Himalayas. Terra 2003, 9th inter-
national conference on the study and conservation of
earthen architecture. Preprints Yazd.
8. Bogin, S. (2005). A technical study of the early
Buddhist wall paintings at Nako, Himachal Pradesh,
India. Zeitschrift für Kunsttechnologie und Kon-
servierung. 19(2): 199–229.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Preservation of historic structures using Screw-Pile foundations

Alan J. Lutenegger
Department of Civil & Environmental Engineering, University of Massachusetts, Amherst, Ma. USA

James H. Kemper
Hubbell Power Systems, Centralia, Mo. USA

ABSTRACT: The structural integrity of historic buildings is often compromised over the life of the structure as
a result of foundation deformations over time. Foundation deformations can be created by a number of problems
related to the subsurface conditions, groundwater removal, softening of the bearing soils, or simply from long-
term soil consolidation. Often times, such deformations produce large and usually differential settlements that
lead to a condition of imminent collapse or unsafe conditions and place the structure in jeopardy. The preservation
of these structures requires innovative foundation repair approaches that will not unnecessarily disrespect the
historic nature of the structure. This paper describes the preservation of several historic structures by foundation
underpinning using helical Screw-Pile foundations. Five case histories are presented where helical Screw-Piles
were used to preserve the foundations of structures ranging in age and use from the stone gate house of an 11th
century castle to a late 19th century university brick classroom. The cases illustrate the economic and effective
use of helical Screw-Piles to retrofit the foundations with minimal intrusion to the structures. For each case, a
brief historical review of the structure is presented and a summary of the foundation problems and site conditions
are given along with a description of the Screw-Piles selected for the repair work.

1 INTRODUCTION limited access, sensitive status of the historic fabric

Historic structures, like all structures, can undergo set-
tlements over their life that are generally produced
from two primary sources. Settlement can occur as a
result of the slow long-term consolidation process that
occurs if a structure has been built on soft soils such
as clays or organic soils. Alternatively, settlements can
occur if subsurface changes have occurred at a site for
example by lowering the water table or removing sup-
port from under or adjacent to a structure by recent
excavations. No matter the cause, the result of set-
tlements is generally structural distress that must be
addressed before additional problems arise or for that
matter a partial or complete failure occurs. Structural
intervention to provide an engineering solution in these
cases generally involves some form of underpinning
although other methods, such as soil extraction which
was successfully used to stabilize the bell tower at Pisa
(Burland et al. 2003).
The foundation restoration of historic structures can
present some unique challenges to the engineer to
devise appropriate remedial support techniques. These
challenges often include such issues as: unknown
existing foundation conditions, unknown or unex-
pected soil and groundwater conditions, heavy loads,
of the structure, (Ciancia, et al. 2006) but can also
include; maintaining adequate support during restora-
tion, sensitivity to specific construction activities, such
as vibration or noise, and the desire to create the least
amount of intrusion to the structure or disrupt the cur-
rent use of the structure. Historic structures can be
particularly sensitive to structural intervention to sta-
bilize the foundation as in some cases, the structural
connection between the foundation and the overlying
structure may also be weakened or may simply be less
sound from the original construction techniques used
during construction.
A number of examples may be cited in which foun-
dation repairs have been made to historic masonry
structures using a wide variety of techniques, includ-
ing, drilled micropiles, jacked minipiles, jet grout-
ing, chemical grouting and conventional piles (e.g.,
Dooff et al. 1995; Avellan and Lange 1997; Chartres
1997; Rodriguez Ortiz and Monteverde 1997; Niehoff
2004). Five case histories are presented in this
paper that illustrate the application of Helical Screw-
Piles for the preservation of historic masonry struc-
tures The cases represent applications in a wide
range of soil conditions and a wide range of loads
and demonstrate that Screw-Pile present a unique

1079
Figure 1. Typical geometry of Screw Pile used by Mitchell
(1836–1880).
approach for engineers who are faced with such
problems.
2 SCREW-PILE FOUNDATIONS
Screw-Pile foundations are a product of mid 19th
century Victorian engineering. Most historian give
credit to the use of Screw-Piles as viable founda-
tions to Alexander Mitchell, a Scottish self-taught
engineer and contractor who initially developed iron
Screw-Piles for use as foundations for offshore light-
house around the coasts of England. Mitchell (1848)
described the mechanics of the functionality of Screw-
Piles as:
“Whether this broad spiral flange, or “Ground
Screw,” as it may be termed, be applied …to sup-
port a superincumbent weight, or be employed
…to resist an upward strain, its holding power
entirely depends upon the area of its disc, the
nature of the ground into which it is inserted,
and the depth to which it is forced beneath the
surface.”
Figure 1 shows the geometry of a typical Screw-
Pile used by Mitchell at the Maplin Sands Lighthouse
in the Thames estuary. All of the Screw-Piles used by
Mitchell were fabricated by a combination of cast-iron
plates and wrought-iron shafts, both of which could
be readily manufactured in England during this time
period. In nearly every case of Screw-Pile installation
before about 1900, the Screw-piles were installed by
manual labor using a large capstan attached to the top
of the shaft and turned by as many as 40 men. No verti-
cal load was applied to assist the installation, the pitch
of the screw blade was sufficient to produce advance
of the crew with rotation of the shaft.
During the mid to late 19th century, Screw-Piles
were used throughout the world to provide support
1080
Figure 2. Configuration of modern Screw-Piles: (a) single
helix; (b) multi-helix; (c) grouted shaft.
for offshore and onshore lighthouses, ocean front
piers, bridges, slope stabilization and underpinning.
Duckham (1889) described the use of Screw-Pile foun-
dations to underpin the town hall at Great Yarmouth
which had undergone several inches of settlement.
Since the early 1950s, the Screw-Pile has taken on
a different form, to solve a different kind of problem.
A.B. Chance recognized that a helical screw element
could be used to provide tension support for electrical
transmission tower and other tall structures and devel-
oped a screw anchor industry that still is in use today.
Since about the 1970s however, Screw-Pile founda-
tions have returned to applications similar to those
in which they were originally developed by Mitchell
and others. Modern Screw-Piles have a variety of
geometries and applications and include both single
helical and multi-helix configurations on both square
shafts and round shafts. With the advent of modern
hydraulics in the construction industry which largely
took off after the 2nd World War, Screw-Piles can now
be installed with ease using a variety of commonly
available construction equipment.
Modern Screw-Piles have a variety of names, such
as Helical Pier, Screw Anchor, or Helical Pile but they
are all generally fabricated as one of three general con-
figurations as shown in Figure 2. They may consist of
a single helical element, a multi-helix lead section or
a helical lead section with a grout column. The cen-
tral shaft may be square or round. If a grout column is
used, an appropriate amount of time is allowed for the
grout to achieve sufficient strength before load trans-
fer. Grout columns may be used to provide additional
load capacity, working in combination with the lead
helical section, or they may be used to simply trans-
fer the load to the lead helical section, as in the case
of short Screw-Piles end-bearing on rock or other hard
material. The grout column typically consists of a Port-
land cement based neat grout or some similar material.

Figure 3. Sequence of underpinning existing foundations
with Screw-Piles.
Crocker (2005) previously described the potential for
using helical piers in historic settings.
The Screw-Piles are generally attached to existing
foundations using a foundation bracket that can be
bolted to the existing foundation and then attached to
the shaft of the Screw-Pile. In some cases, strength-
ening of the existing foundations, which often consist
of stone or other masonry, may be necessary using
a thread bar or a grouted bar before installing the
Screw-Pile support system. This may be particularly
important given the age of most of the structures that
predate the use of concrete for foundations.
3 SEQUENCE OF FOUNDATION
UNDERPINNING USING SCREW-PILES
When using Screw-Piles for underpinning, it is first
necessary to make a small excavation to expose the
base of the existing foundation. The Screw-Pile is then
installed using hydraulic torque motor to the required
design depth or a sufficient depth as determined by the
installation torque to develop sufficient load capacity.
A foundation “L” bracket is then attached to the shaft
of the pile and to the existing foundation to provide a
1081
structural connection between the two and provide for
load transfer.
If appropriate, given the condition of the structure
and the degree of distress, a small amount of lifting
may also be performed using hydraulic jacking sys-
tem to provide vertical adjustment to the structure and
preloading of the screw-pile. This load is then locked
off and the final position of the foundation bracket
is achieved. If needed, the existing foundation may
also be reinforced using either thread bars or grouted
bars inserted into the existing foundation blocks or a
small section of reinforced concrete may be cast in
place and tied to the existing foundation. The under-
pinning sequence is illustrated in Figure 3. Note that
Figure 3 illustrates that the Screw-Pile can include a
grout column around the central shaft or it may be
installed without a grout column, depending on the
soil conditions and the design of the pile for bearing
capacity.
4 CASE HISTORIES
Table 1 gives a summary of the five historic masonry
structures described in this paper. It should be remem-
bered that what may be considered as an historically
significant structure may vary depending on the cur-
rent usage of a particular structure, the age, location
and setting of the structure in context to the surround-
ing environment. A description of each of the cases
will be given describing the state of distress and the
intervention applied.
4.1 Case no.1 – Mark Twain Hotel, Hannibal,
Missouri
The Mark Twain Hotel in Hannibal, Missouri is a four
story brick building built in 1906. The existing struc-
ture was founded on timber piles. The Hotel was being
converted to senior housing and required the addition
of two new four story stairways and the underpinning
of an existing stairway. It was decided not to disrupt
the existing structure by attaching the new stairways
to the existing building by the new construction. Soils
at the site consist of low plasticity silts and silty clays in
the upper 9.1 m which overly loose to medium dense
alluvial sand and gravel extending to about 18.3 m.
The ground water was located between 4.6 and 6.1 m
below ground surface.
Screw-Piles with grouted shafts (Pulldown
Micropiles) were installed to the sand and gravel to
provide support. Triple helix lead sections (8/10/12)
were used with either a 38.1 or 44.4 mm square shaft to
depths ranging from 5.8 to 16.8 m although most were
between 9.1 and 16.8 m in length. A 127 mm diameter
grout column was used. A total of 21 screw-piles were
installed using either heavy duty foundation brackets
 Table 1. Summary of case histories.
Case
no. Structure
1 Mark Twain Hotel
2 Ralls County Courthouse
3 Our Lady of Lourdes Church
4 Forest Lodge, Hulne Park
5 Engineering Building East
University of Missouri
Figure 4. New construction bracket attached to Screw-Pile
for new foundation or extension of existing foundation.
in the case of the existing stairwell or new construction
brackets as shown in Figure 4 for the new construction.
4.2 Case no.2 – Ralls County Courthouse, New
London, Missouri
The Ralls County Courthouse, located in New London,
Missouri, is a limestone structure built in 1857 with a
limestone addition added in 1936. Because of the set-
tlement of the addition away from the original structure
had become a concern it was necessary to underpin the
addition to arrest further settlement. Based on test bor-
ings performed at the site, the soils consisted of silty
clay and clay with a layer of limestone present at a
depth of about 5.2 m. No groundwater was encoun-
tered in test borings at the time of drilling. Estimated
settlements at the time of repair were on the order of
44.4 mm.
A total of 57 screw-piles were installed to depths
ranging from 2.4 to 7.6 m below existing foundations
1082
Age of
Location structure
Hannibal, Missouri 1906
New London, Missouri 1857/1936
Minneapolis, Minnesota 1854–1857
Alnwick, Northumberland 1853
Columbia, Missouri 1892
Figure 5. Heavy duty foundation bracket attached to
Screw-Pile and existing foundation.
to support loads ranging from 50 kips to 86 kips.
Both double and triple helix Screw-Piles with 44.4 mm
square shafts were installed around the perimeter of
the structure to develop the required loads. While most
piles were installed along the footings from the outside,
a few piles were required inside the structure using a
hand held portable torque head. Piles were attached
to the existing foundations which were located at a
depth of about 2.7 m below ground surface using a
heavy duty foundation bracket assembly, as shown
in Figure 5. Piles were spaced approximately 1.5 m
apart around the outside of the structure. Installation
torque was monitored on each pile. A small excavation
was required at each location to allow the foundation
bracket to be attached to the existing foundation.
4.3 Case no.3 – Our Lady of Lourdes Church,
Minneapolis, Minnesota
Our Lady of Lourdes Church in Minneapolis,
Minnesota is the oldest continuously used church in
the city of Minneapolis and was built between 1854
Figure 6. Heavy duty foundation bracket attached to exist- Figure 7. Foundation bracket and leveling jack used at
ing foundation. Forest Lodge.

and 1857.in the Greek temple style. The building is

built of native limestone and was designated a U.S. his-
toric landmark in 1934. Soils at the site consist of about
2.5 m of sandy fill overlying dense gravely sand. At the
time of drilling no water table was encountered in the
upper 10 m. During installation of new water service to
the church, the existing foundation at one location was
undermined which caused some settlement of a wall.
Screw-Piles were used to underpin the founda-
tions to prevent additional subsidence. Triple-helix
screw-piles (8/10/12) with 44.4 mm square shafts were
installed to a depth of about 4.6 m below existing
foundation levels and were attached using heavy duty
foundation brackets from the outside of the structure to
provide load capacity of 120 kips each. Figure 6 shows
a typical heavy duty foundation bracket attached to an
existing foundation.

4.4 Case no.4 – Forest Lodge, Hulne Park, Alnwick,

Northumberland, U.K.
Forest Lodge is the gate house into Hulne Park, which
is part of the Duke of Northumberland’s estate in
Northumberland, U.K. The Lodge was built in 1853 of
local stone. Over the years, the north wing of the gate
house had experienced excessive settlement. To pre-
vent additional settlement, Screw-Piles with grouted
were installed. The soil at the site consisted predomi-
nantly of miscellaneous fill overlying bedrock at about
7.5 meters below the ground surface.
Screw-Piles consisted of twin helix lead sections
(10/12) with a 38.1 mm square shaft with a grouted
column as shown in Figure 8. Displacement plates
were used to produce a 100 mm grout column. The
piles were installed to bedrock which varied between
6 and 8 meters below existing foundations so that they
would act in end bearing. Work was performed both
inside the cellar of the gate house which included a
Figure 8. Screw-Pile with grouted shaft used for
underpinning.
vaulted ceiling with a kitchen above using a hand-held
portable hydraulic torque head. A total of fourteen
screw-piles were installed to support the structure.
Foundation brackets were then attached to the shaft
of the screw-piles and then bolted to the existing
foundation stones.
Figures 9 and 10 show some of the Screw-Piles and
foundation brackets installed at the Forest Lodge Gate
House during the underpinning and restoration work.

1083

Figure 9. Installing Screw-Piles at Forest Lodge.
Figure 10. Corner of Forest Lodge Gate House after instal-
lation of Screw-Piles and foundation brackets.
4.5 Case no.5 – Engineering Building East
University of Missouri, Columbia, Missouri
Engineering Building East is located on the University
of Missouri campus in Columbia, Missouri and is a
two story brick structure with a central bell tower built
in 1892. Because of movements of the building that
were thought to be the result of expansive soils in the
area and changes in the building drainage system the
building and developed some severe distress. Soils at
the site consisted of silty clays, sandy silt in the upper
zone with weathered limestone rock at depths between
6.1 and 11.0 m deep. The structure had a basement.
One of the issues imposed during the restoration work
was the need for the building to remain in use during
the foundation repair.
Repair of the existing foundation required installa-
tion of screw-piles both inside and outside the structure
in areas of limited access. Twin helix Screw-Piles with
44.4 mm square shafts were used spaced on 1.2 m cen-
ters along the building walls and 0.6 to 0.9 m centers
1084
beneath the bell tower. A skid loader was used to install
the Screw-Piles on the outer perimeter while a portable
hand held hydraulic torque head was used inside,
where the work space was limited. Screw-Piles were
attached to a heavy duty foundation bracket attached
to the existing masonry rubble foundation which was
then encased in a new cast-in-place concrete block to
tie the existing foundation to the Screw-Piles.
Table 2 gives a summary of the different configu-
rations of Screw-Piles used for each of the cases.
5 ADVANTAGES OF SCREW-PILES FOR
STRUCTURAL REPAIR/RESTORATION
The use of Screw-Piles for the structural restoration
and underpinning of historic structures has a number
of advantages that in most cases cannot be realized
by most traditional underpinning techniques. One of
the highest priorities for this type of work is preserv-
ing the historic fabric of the existing structure during
construction activities.
5.1 Minimal disruption to existing structure
The installation of Screw-Piles is generally minimally
invasive work involving slow rotation using a high
torque hydraulic torque head. There are no soil cut-
tings to dispose of from installation of the Screw-Piles.
Soil from the excavation to expose the existing foun-
dation may be carefully stockpiled and used as backfill
at the completion of the repair work. In some cases,
the structure may still be used while the work is in
progress.
5.2 Flexibility of geometry
As indicated with cases presented, Screw-Piles are
available in a variety of configuration, including sin-
gle and multi-helix lead sections with a variety of
diameters. This allows considerable flexibility for the
designer to select an appropriate geometry for a spe-
cific project. Increasing the helix diameter and/or
number of helices or modifying the installation length
to achieve the required load capacity is generally very
easy. In cases where increased bending stiffness is
needed to support eccentric loading, the use of the
grouted shaft may be desirable. Even when a grouted
shaft is used, the additional equipment needed to mix
and place the grout is minimal.
5.3 Minimal construction vibration
There are essentially no vibrations produced by the
hydraulic plant during installation of the Screw-Piles.
This may be particularly important in situations where
the existing structure may be sensitive to construction
vibrations. Most Screw-Piles for restoration work may
Table 2. Screw-Piles used in case histories.
Case Screw-Pile
no. configuration Engineer
1 Grouted shaft triple-helix (8/10/12) Klinger & Associates, P.C., Hannibal, Mo.
2 Square shaft double- and triple-helix Klinger & Associates, P.C., Hannibal, Mo.
3 Square shaft triple-helix (8/10/12) Meyer Borgman & Johnson, Inc.
4 Grouted shaft double-helix (10/12) Patrick Parson Ltd.
5 Square shaft double-helix Structural Engineering Association,
Kansas City, Mo.
be installed with small light duty construction equip-
ment such as a rubber track mini excavator or skid
steer or portable hand held equipment.
5.4 Installation in high groundwater
The installation of Screw-Piles is unaffected by high
ground water condition since no excavation is required
below the foundation level. Of course is water is above
the foundation level, directing may be required as in
most other replacement methods.
5.5 Construction in confined space
As demonstrated by the description of several of the
cases described in this paper and many other projects,
Screw-Piles may be installed in areas of limited access
or low head room. The portable equipment can eas-
ily be managed by a single operator. Short extension
section may be use inside to install the Screw-Piles
to the required depth. The power unit used to operate
the portable torque motor can be placed outside and
hydraulic hosed can be routed through a window or
small opening in the structure.
5.6 Field verification of capacity for QC
During installation the hydraulic torque can be mon-
itored to provide a record of the installation torque.
This attribute of Screw-Piles allow for field verifica-
tion of the soil conditions at each pile location and
for verification of load capacity. This is particularly
important in regard to historic structures where the soil
conditions under the structure are often unknown and
difficult to determine, especially at interior locations.
The installation torque record provides an excellent
quality control tool and should be included as a part
of every project.
5.7 Rapid installation and construction
In most soil conditions, Screw-Piles may be installed
very quickly. A typical installation time using
conventional construction equipment such as a skid
1085
Foundation
contractor
Tri-State Concrete Correction, Co.
Tri-State Concrete Correction, Co.
Atlas Foundation Co.
Subsurface Technologies, Ltd.
Kidwell Construction and
Pro Services.
steer or mini excavator is about 1 hour for a 10 m
length of pile. Only a small amount of additional time
is needed when using a grouted shaft Screw-Pile.
5.8 Instant capacity/load transfer
In most soils Screw-Piles allow for load transfer to
essentially take place as soon as the pile has been
installed. This can be important in some cases where
emergency repairs are needed before additional dam-
age can occur. In the case of grouted shaft Screw-Piles,
sufficient time, usually 5 to 7 days may be required
for the grout to attain sufficient strength before being
loaded.
6 CONCLUSIONS
Several case histories have been presented describing
the use of helical Screw-Piles for providing structural
repair and restoration of the integrity of the founda-
tion of historic structures. The primary advantages of
Screw-Piles in this application, which often involves
sensitive structures, is the minimal disruption to the
historic fabric of the existing structure, the minimal
intrusion to the site and surroundings, minimal noise
and vibrations that can accompany other more inva-
sive methods and the speed at which the work may be
performed. The methods described in this paper may
be applied to other structures to achieve a cost effec-
tive intervention for preserving the structural stability
of historic structures that have undergone or are con-
tinuing to exhibit additional movements because of
subsurface conditions.
ACKNOWLEDGEMENTS
The authors wish to gratefully acknowledge the assis-
tance of a number of individuals who provided
details and important project information on these
projects; Ron King of Tri-State Concrete Construction
Co. Quincy, Illinois; Don Bobbit, P.E. of Cen-
tralia, Missouri; Mike Moeller, Structural Anchor
Supply/Atlas Foundation Company; and Duncan
Robson, Subsurface Technologies Ltd, Alnwick,
Northumberland.
REFERENCES
Avellan, K., and Lange, K., 1997. St.-John’s Chirch,
Tartu, Estonia – Underpinning of Tower. Geotechnical Engi-
neering for the Preservation of Monuments and Historic
Sites: 521–525.
Burland, J.B., Jamiolkowski, M., and Viggiani, C., 2003.
The Stabilization of the Leaning Tower of Pisa. Soils and
Foundations, 43 (5): 63–80.
Chartres, F.R.D., 1997. Monitoring the Installation of Piles
Beneath the Round Tower, Windsor Castle, U.K. Geotech-
nical Engineering for the Preservation of Monuments and
Historic Sites: 349–356.
Ciancia, A.J., Biesiadecki, G.L., and Ladd, B.E., 2006.
Challenges of Underpinning Two Historic Landmark
Buildings. Practice Periodical on Structural Design and
Construction, ASCE, 11(3): 142–148.
1086
Crocker, E.E., 2005. Non-invasive Underpinning Technolo-
gies in Historic Settings. Proceedings of the 4th Inter-
national Seminar on Structural Analysis of Historical
Constructions: 821–828.
Drooff, E.R., Furth, A.J., and Scarborough, J.A., 1995.
Jet Grouting to Support Historic Structures. Foundation
Upgrading and Repair for Infrastructure Improvement,
ASCE: 42–55.
Duckham, F.E., 1889. Underpinning Great Yarmouth Town
Hall. Minutes of the Proceedings of the Institution of Civil
Engineers, 98: 372–374.
Mitchell, A., 1848. On Submarine Foundations; particularly
Screw-Pile and Moorings, Civil Engineer and Architects
Journal, 12: 108–149.
Niehoff, J.W., 2004. Stabilization of a Historic Struc-
ture Founded on an Unstable Mesa. Proceedings of
the 5th International Conference on Case Histories in
Geotechnical Engineering, paper no. 1.05, 6 pp.
Rodriguez Ortiz, J,M., and Monteverde, P.R., 1997. A Review
of Recent Spanish Interventions in Monuments. Geotech-
nical Engineering for the Preservation of Monuments and
Historic Sites: 499–508.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Strengthening of an industrial cylindrical shell damaged by a collision

Wine Figeys, Sven Ignoul & Dionys Van Gemert

Department of civil engineering, KULeuven, Belgium
Triconsult N.V., Lummen, Belgium

ABSTRACT: This case-study concerns an old industrial reinforced concrete shell structure at the port of
Antwerp. The building serves as a pilot for stowing of shipped steel and wooden products. A fork-lift hit one of
the supporting columns. The impact of the collision caused the concrete to crush and the internal reinforcement
to shift considerably. Some of the internal reinforcement yielded or broke. Temporary supports had to be placed
to avoid collapse of the total structure. The stress situation is analyzed. A FEM model was built to gain insight
into the stress distribution in the shell structure. In comparison with the undamaged situation, membrane forces,
shear forces and moments did strongly increase which explains the actual damage. A solution is worked out to
repair and strengthen the damaged part of the structure. The FEM-analysis demonstrates that a compression arch
will arise in the top of the barrel shell to span the displaced support. This requires a tension member at the bottom
side of the concrete barrel. Therefore, externally bonded reinforcement is applied to increase the tensile capacity
of the perimeter beam. The web reinforcement has yielded or is broken so that it is replaced by external CFRP
reinforcement. After strengthening the structure, a monitoring system is installed. Strain gauges are glued on
several laminates. These measurements give feed back on the structural behaviour of the strengthened structure
and the applicability of the FEM used. This case-study illustrates the load distribution mechanisms in the shell
structure, the assessment of the actual condition and an appropriate intervention based on external reinforcement.

1 INTRODUCTION

An old industrial reinforced concrete shell structure

(1955) at the port of Antwerp in Belgium is used as
a pilot for stowing of shipped steel and timber prod-
ucts, Figure 1. This shell needs a strengthening after a
collision of a fork-lift with a supporting column.
Restoration is generally based on a sequence of
anamnesis, analysis, diagnosis, therapy and control
(ICOMOS 2003), (Schueremans 2003):
• the anamnesis. Significant and objective infor-
mation of the building is collected. Data can be
gathered from literature, direct visual observation
and field research;
• a structural analysis is performed;
• the causes of the damage and/or decay are deter- Figure 1. The shell construction.
mined in the diagnosis;
• if necessary, one or more therapies are proposed to 2 ANAMNESIS
repair and upgrade the building. In this case study,
the technique of externally bonded reinforcement is The reinforced concrete structure consists of a
applied; sequence of 25 cylindrical barrel shells, supported by
• the last step is control whereby checks are carried reinforced concrete columns. The span of each shell
out during and after intervention. is 15 m. Each shell is supported by 3 columns at both
sides. The thickness of the shell decreases from 13 cm
These basic steps are used as a guideline through at the abutment till 8 cm at mid span. In the mid part
the strengthening process. of each shell, a longitudinal roof light is located. In

1087
Figure 2. Damaged column, December 2004.

Figure 4. Recently damaged column, November 2007.

Figure 3. Damaged column in the past.

December 2004, a fork-lift hit one of the 6 meter high

concrete columns, Figure 2. This accident was not the
first one, Figure 3. In November 2007 a new accident
occured, 4.

2.1 Visual inspection

The settlement of the column and the crack formation Figure 5. Observed crack formation and measured crack
widths [in mm].
were studied (Figeys and Van Gemert 2006). Mea-
surements showed a horizontal displacement of about
14 cm at the impact location of the damaged column,
Figure 2. The corresponding vertical (downward) dis- the support and run under an angle of 45◦ to the top
placement of the shell support equaled approximately of the shell, Figure 5. Due to the large displacements,
8 cm. This differential settlement resulted in severe part of the internal reinforcement yielded.
cracking of the supported shell structure. Fortunately,
only the two neighboring half parts of the shell were
damaged. The roof lights in the top of the shell avoid
2.2 Field research
force transfer to the other half of the shell.
At different places, the full concrete shell cross- The concrete quality is determined experimentally
section is cracked. The maximum crack opening was by means of pull-off-tests (CEN pr EN 1542 1998).
almost 20 mm, Figures 5 and 6. The cracks start from A mean tensile strength of 3.8 MPa is measured.

1088

lb = 15.9 mm. The strain in the steel bars can be esti-
Figure 6. Crack above the damaged column with a maxi-
mum crack opening is 19.8 mm.
3 ANALYSIS AND DIAGNOSIS
The stress situation in the concrete shell is studied
to determine the necessity and the amount of the
strengthening. Therefore the concrete quality and the
loads are determined.
3.1 Material characteristics
3.1.1 Concrete
The compressive strength and the Young’s modulus is
determined from the measured mean tensile strength
according to Eurocode 2, (Eurocode 2 1999). The char-
acteristic tensile strength of the concrete, fctk , is derived
from the results of the pull-off-tests:
with
fctk the characteristic tensile strength
fctm the mean measured tensile strength = 3.8 N/mm2
σ the dispersion = 0.7N/mm2
The compressive strength, fck , is estimated using the
inverse of formula 2 of Eurocode 2, 1999). The charac-
teristic compressive strength, fck , equals 46.0 N/mm2 .
The design value found with Equation 3, taking into
account a partial safety factor of 1.5.
The Young’s modulus can be determined using
formula 4 and equals 35.9 kN/mm2 :
1089
3.1.2 Internal reinforcement
The reinforcement of the concrete shell exists of main
steel bars and a web. The location of the bars (φ15 mm)
correspond with the position on plan. The distance
between the bars of the web (φ5 mm) is 150 mm in
the longitudinal direction and 200 mm in the other.
Close to the column, the mean crack opening is
mated by this elongation. Taking the anchorage length
as the original length, the strain is equal to Equation 5.
The anchorage length can be calculated according to
Eurocode 2, Equation 6.
with
φ diameter of internal reinforcement
fyd design yielding strength of internal reinforcement
fbd design adhesion strength of concrete, 1.7 N/mm2
The strain in the main reinforcement equaled 1.54%
and in the web 4.62%. The strain of the main rein-
forcement is high. The bars yielded. However the strain
remained below the ultimate stress level (steel BE 22:
18%). The strain in the web is higher, close to the ulti-
mate stress level according to NBN 24-304 (6%). At
the main crack, the web reinforcement had yielded and
was possibly broken. Visual inspection confirmed the
broken web bars.
3.2 Determination of loads
The design value of the load is calculated according
to Eurocode 0 (Eurocode 0 2002) and 1 (Eurocode
1 1999). Besides of the self-weight of the concrete,
also loads from wind (Eurocode 1 1995) and snow
(Eurocode 1 2003) are taken into account.
The density of the concrete is determined by mea-
suring the weights of small cubes of broken concrete
of the column. The mean density equals 2268 kg/m3 ,
the characteristic value is 2419 kg/m3 .
The snow load, Equation 7, is counted as an enlarge-
ment of the self weight. As the wind load has an
opposite and thus positive effect on the shell structure,
it is neglected.
with
µi shape factor, equal to 1.08
sk characteristic snow load in Antwerp
Ce exposure coefficient, equal to 1.0
Ct thermal coefficient, equal to 1.0.
 Figure 8. Calculation of the membrane forces in the Y-axis
Figure 7. Schematic view of reinforced concrete shell
in the shell before the column settlement.
construction.

The increased density of the concrete is calculated

in the ultimate limit state, Equation 10, and equals
40.76 kN/mm3 .

3.3 FEM-analysis
A FEM-model is built to analyse the stress distribution
in the shell structure (Ansys 2005). As the damaged
column is located in the central part of the building,
edge effects can be neglected. Only the two neighbor-
ing shells are included in the FEM-model, the damaged
column being located at the outer left part, Figure 7.
Symmetry boundary conditions are applied to the left
and right edge of the model.
The shell structure is modeled with the 8-node
quadrilateral SHELL93 shell element, which has 6 Figure 9. Original reinforcement located at high stress
degrees of freedom at each node: translation in the zones.
X, Y and Z direction and rotations around the X,
Y and Z-axis. The beams around the roof lights are
modeled by means of 2-node BEAM4 and BEAM44 The original reinforcement plan (1955) show that
elements, which have also 6 degrees of freedom at the internal main reinforcement is bundled in the mid
each node. These node displacements are translation span of the beam and spreads out to the supports. This
in three directions and rotation around the three axes. main reinforcement take care of the tensile stresses in
Membrane forces, shear forces and moments are the shell, Figure 9.
calculated (Figeys and Van Gemert 2006). The original In comparison with the undamaged situation, cal-
situation is studied, Figure 7. Tensile forces in longi- culations show that the membrane forces, shear
tudinal direction are present at mid span of the edge forces and moments strongly increase at the imposed
beam (with a maximum of 289 kN/m) and in the shell deformations. In particular, the membrane tensile
near the central column (150 kN/m), Figure 8. forces in the longitudinal direction of the barrel

1090
Figure 10. Calculation of the membrane forces in theY-axis
in the shell after the column settlement.
shells (longitudinal axis, Figures 8 and 10) increased
extremely.
This caused the actual damage state and, more-
over, changed the structural system.The damaged shell
seems to be supported by the two outer columns only.
High tensile stresses (with a peak of 4820 kN/m2 )
appear in the edge beam at the missing support (i.e.
damaged column). In the top of the shell compressive
stresses in the longitudinal direction are introduced
by the arch action. The available reinforcement in the
edge beam at the support is insufficient, which led to
the observed cracking of the concrete.
4 THERAPY
A method to repair and strengthen the damaged part
of the structure was studied and executed. The FEM-
analysis demonstrates that, due to the vertical settle-
ment of the support, a compressive arch developed in
the shell, that spans the displaced support. In order
to restore as much as possible the initial stress-strain
situation, the structure is jacked up, Figure 11.
However, it is not possible to neutralize the total
vertical displacement. The plastic deformation of the
internal reinforcement can not be recovered and chip-
ping off of concrete would arise at excessive lifting.
An additional tension member at the bottom edge of
the concrete barrel should take up the induced tensile
forces. For that purpose, externally bonded reinforce-
ment is applied to increase the tensile capacity of the
longitudinal beam.
The tensile force in the edge beam can be calcu-
lated as:
1091
Figure 11. Jack up of the shell structure.
with
TSd Design value of tensile force in the bottom of
the barrel at one side
L Span without damaged column, 30.0 m
f Height of the shell, 2.95 m
p Load per meter along the longitudinal axis,
that equals:
θ Half of the opening angle of the shell, 43◦
R Radius of the shell trajectory, 10.9 m
d Mean thickness of the shell, 0.10 m
ρd Design value of load, including proper weight,
wind loading, snow loading 40.76 kN/m3 .
The tensile force present at the beam for each half
shell equals TSd = 1285 kN . Taking into account the
original reinforcement (9 × φ15 mm, TRd = 325 kN),
five CFRP laminates 100 × 1.2 mm2 are additionally
needed to take of this tensile force, Equation 13.
A safety factor γM ,CFRP of 1.5 is applied.
Also the internal reinforcement mesh near the col-
umn is damaged: yielding and cracking are observed.
However, because of the crack formation, the calcu-
lated stress peaks at the imposed deformations are
reduced. Therefore, it is safe to take the stress level of
the undamaged situation into account. The stress level,
calculated with the FEM-model, is compared with the
amount of undamaged reinforcement. In the longitu-
dinal as well as in the transversal direction additional
reinforcement is needed.
The tensile force in the longitudinal direction equals
about 150 kN/m near the column and decreases to
Figure 12. Proposed strengthening of the shell structure.

Figure 14. Strengthened shell construction.

Figure 13. Section of the strengthened structure.

0 kN/m near the roof light. Also in the transversal

direction, additional reinforcement is needed. Tak-
ing into account the maximal transferable load and
the minimal needed anchorage length calculated from
the pure shear model described by Brosens (Brosens
2001), the lay out presented in Figures 12 and 13, is
chosen.
The strengthened shell construction is shown in
Figure 14. After reconstruction of the damaged part of
the column, Figure 15, the temporary supports were
removed.

Figure 15. New reinforcement and concrete for damaged

5 CONTROL
A monitoring system is installed to follow up the
stress situation in the external reinforcement. Strain
gauges are glued on several laminates, Figure 16. After
repairing of the column, the temporary supports are
removed. During this action, the different strain gauges
are measured. The strain gauges are measured at differ-
ent moments: before removing the supports (17.7◦ C),
after removing (20.5◦ C), after 3 days (16.5◦ C) and
after 31 days (22.6◦ C). Measurement of the temper-
ature allows correcting the measured values to the
effective stresses in the laminate. These stresses are
part of the column.
presented in Figure 16. In Table 1, the stress levels
after one month are given.
As the concrete column is replaced and the struc-
ture is jacked up, stresses as predicted with the original
structure are expected, Figures 17 and 18. Close to the
column, a peak tensile force of 150 kN/m is present. In
the first meter near the column, 3 steel bars φ15 mm
an 5 CFRP (1.2 × 100 mm2 ) are present to take these
tensile force. In the ultimate limit state a stress level of
182 N/mm2 is expected. Taking into account the real

1092
Figure 16. Location of the different strain gauge.

Table 1. Stress level in laminates on different

locations.

Location Stress [MPa]

1 −24.4
2 −36.5
3 65.8
4 12.8
5 −73.0
6 14.5
7 −78.5
8 −12.0
9 19.7
10 11.7 Figure 17. Stress level on different location derived from
11 11.4 measured strains.
12 −40.3

+: compression; −: tension.
situation, only the mean density acting on the struc-
ture (ρd = 22.68 kN /m3 instead of 40.76 kN /m3 ), the
stress level will be reduced to about 90 N/mm2 . These
tensile stresses can be observed in these locations
(5 and 7), except in location 6.
Strain gauges 1, 2, 3 and 4 are located further on the
laminates. In these zone the tensile stress decreases.
Except for location 3, the tensile stress levels in these
locations decreased. Immediately after removing the
supports, a relatively large compressive stress is found
in location 3. This location is situated at the bottom
CFRP laminate, two meters from the concrete column.
The measuring point is close to the injected crack with
the largest opening of about 20 mm. After removing
of the supports, the original situation is approached
in which compressive stresses will appear in the bot-
tom of the edge beam in the zone of the crack. These
stresses will compress the epoxy resin in the crack. The
laminate is compressed as well, which can be observed
in the compressive stress curves.
Location 8 and 9 are situated at the upper part of
the shell, Figure 16. The tensile stresses decrease into
the upper part to change in small compressive stresses
at the top, as observed at location 9.
Three strain gauges measure the strain in the trans-
verse direction. According to the calculations only
small tensile stresses are observed, Figure 18.
6 CONCLUSIONS
Externally bonded CFRP-laminates are used to
strengthen and repair a damaged shell structure. FEM
analyses are used to determine stresses and strains in
both the undamaged and damaged shell. This FEM-
model allowed to identify the formation of a com-
pressive arch in the damaged shell structure and to
determine the required additional reinforcement, to be
applied by means of externally bonded CFRP lami-
nates on the edge beam and on the shell structure in
the vicinity of the damaged column. Monitoring of
stresses in the laminates during different phases of the

1093

0: before removing of the supports A: after removing of the
supports
Figure 18. Stress level on different location derived from
measured strains.
repair showed the effectiveness of the repair procedure.
As it was possible to jack up the shell and to restore the
support, stress levels as predicted for the undamaged
situation are observed.
1094
ACKNOWLEDGEMENT
The authors would like to thank the Flemish Insti-
tute for Promotion of Scientific and Technological
Research in the Industry (IWT – Vlaams Instituut voor
de Bevordering van Wetenschappelijk-Technologisch
Onderzoek in de Industrie) for their financial support.
REFERENCES
Ansys, I. (2005). Ansys version 10.0.
Brosens, K. (2001). Anchorage of externally bonded steel
plates and CFRP laminates for strengthening of concrete
element. Ph. D. thesis, Katholieke Universiteit Leuven,
Leuven.
CEN pr EN 1542 (1998). Products and systems for the
protection and repair of concrete structures – test methods-
measurements of bond strength by pull-off.
Eurocode 0 (2002). Basis of design: Design principles.
Eurocode 1 (1995). Basis of design and actions on structures-
part 2–4: Actions on structures – wind actions.
Eurocode 1 (1999). Basis of design and actions on structures-
part 1 – 1: Actions on structures – densities, self-weight
and imposed loads.
Eurocode 1 (2003). Eurocode 1 – basis of design and actions
on structures – part 1–3: General actions – snow loads.
Eurocode 2 (1999). Design of concrete constructions – part
1–1 (in dutch).
Figeys, W. and D. Van Gemert (2006). Report d/00413/05
dd. 13-03-2006. shell repair. hybrid solution (in Dutch).
Technical report, Triconsult N.V.
ICOMOS (2003). Recommendations for the analysis, con-
servation and structural restoration of architectural her-
itage. International Scientific Committee for Analysis and
Restoration of Structures of Architectural Heritage.
Schueremans, L. (2003). Evaluation of bearing capacity and
design of consolidation and strengthening by means of
injections. In WTATagung, Leuven.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Experimental results on the use of mud-based grouts to repair

seismic cracks on adobe walls

Julio Vargas & Marcial Blondet

Pontificia Universidad Católica del Peru

Claudia Cancino
Getty Conservation Institute

Francisco Ginocchio, Carlos Iwaki & Kathya Morales

Pontificia Universidad Católica del Peru

ABSTRACT: This paper presents the main results obtained during an ongoing experimental test program
performed to explore the possibility of repairing structural cracks on adobe walls through the injection of mud-
based grouts. The structural objective was to attempt to recover the original strength of severely cracked adobe
masonry elements by repairing them with mud-based grout injections. Masonry strength was first estimated
with indirect tension tests of adobe sandwiches and later measured with diagonal compression tests of adobe
masonry small walls. It was decided to study the use of grouts prepared with sieved soil alone or stabilized with
lime, cement, and gypsum. The test results showed that grout injections based on the original soil used to build
the walls are effective in restoring completely the original strength of adobe walls which had endured significant
seismic damage. It seems, therefore, that a new criterion for the conservation of earthen buildings in seismic
areas could be formulated, using compatible materials for the restitution of the original seismic strength of the
earthen structural elements.

Keywords: Adobe, conservation, repair, cracking, grouts.

1 BACKGROUND cement, lime or gypsum (Vargas 1979, Mattone et al.

In seismic areas, it is crucial to be alert to the presence
of cracks in the walls of adobe buildings, as they are the
most important structural elements. Structural cracks
may significantly reduce the stiffness and strength of
the walls, and thus increase the risk of collapse during
the occurrence of new earthquakes, with the conse-
quent loss of lives and serious material damage to the
cultural heritage.
Traditional methods of repairing earthen buildings
are based on criteria of increased resistance. More
recently, repair methods have been developed that are
based on criteria of stability, by the use of tension-
resistant reinforcement (Tolles et al. 2000, 2003).
These criteria are complementary, and must take into
account the recommendations of the International
Conservation Charts that are universally accepted.
This project aims to study the feasibility of repair-
ing cracked earthen walls through injections of liquid
grouts made of only mud and of mud stabilized with
2005).
2 TESTING PROGRAM AND EXPERIMENTAL
METHODS
One hundred fifty indirect tension tests were per-
formed, using adobe “sandwiches” made by joining
two adobe blocks with a layer of mortar made of mud
sifted to remove particles larger than 2 mm. Batches
of three similar sandwiches were tested. Each batch
was made from mortars of 10 mixes of mud with and
without additives, and of 5 nominal thicknesses (2 to
10 mm). Mortars were studied of only soil; soil stabi-
lized with cement in 5%, 7% and 10% ratios, soil sta-
bilized with lime (5%, 7% and 10%) and soil stabilized
with gypsum (5%, 10% and 20%). Finally, a series of
diagonal compression tests were conducted on small
adobe masonry wall specimens. Fifteen specimens
were cracked under diagonal compression, repaired

1095
Figure 1. Device developed to perform indirect tension test
on adobe sandwiches.
by injecting grouts of only soil and soil stabilized with
cement, lime or gypsum, and tested again in order to
measure the effectiveness of the repair procedure.
2.1 Initial considerations
Adobe masonry consists of a series of sun-dried blocks
joined by mud mortar. The strength of the masonry
results in large part, therefore, from the adhesion and
resistance of the mortar. There are many alternatives
for improving the quality of the mortars in order to
attain stronger masonry (Vargas 1979). The joints are
generally the critical areas of the adobe wall, as cracks
in the mortar due to drying shrinkage reduce the adher-
ence of the mud mortar to the adobe blocks. It is
possible that a similar phenomenon occurs with grouts
injected into the cracks in the walls. The quality of con-
struction also has a great influence on the strength of
adobe walls.
It seems reasonable to assume that the strength of
adobe masonry buildings with walls that have been
repaired with injections of grouts of stabilized soil is
due to a combination of the strength of the grout and
of the original mortar. While it is possible to make the
strength and adhesion of the grout surpass those of the
original mortar, this does not necessarily mean that
the repaired wall will achieve a greater strength than
the original, because when faced with a significant
loading, the wall would simply fail in a different area
than that which had been repaired.
2.2 Indirect tension test
The indirect tension test consists of vertical compres-
sion of a “sandwich” of two adobe blocks joined by
mortar, forming a vertical and centered joint (Fig. 1).
The force applied causes fairly uniform tension
stresses at the mortar-block interface (Fig. 2). The
tensile strength of a sandwich is expressed by σt = α
(P/A), where P is the breaking force, A is the area of the
1096
Figure 2. Distribution of normal horizontal stresses on
adobe sandwich subjects to indirect tension test. The area
directly below the load is in tension. The sandwich opens at
the mortar-block interface.
mortar interface (A = lh, where h is the height and l, the
length of the sandwich) and α is a dimensionless factor
to estimate the maximum tension stress at the interface.
A large number of mortar mixes were tested using
this simple and inexpensive test. Although test results
have a large dispersion, there is a significant correla-
tion to more sophisticated tests, such as the diagonal
compression test (Vargas et al. 1986). Both the blocks
and the mortars were prepared with soil from the PUCP
campus, which has a high percentage of materials with
low plasticity. For the study of mortars, the moisture
content and the maximum size of particles were varied.
2.3 Diagonal compression test
In this test, a square sample of masonry is subjected to
a compressive force in two opposing corners (Fig. 3).
The behavior of the specimens is representative of
the seismic behavior of adobe masonry (Vargas et al.
1986), but the test is elaborate and costly, so it has
been used only to test the most efficient grouts. The
distribution of force is similar to that which occurs in
the indirect tension samples and so, the type of failure
expected is a crack between the points of application of
the force, generated by the dominant tension stresses.
After testing, the failed specimens were transported
to a repair area, and were placed in their “natural” posi-
tion (horizontal rows) to have gravity acting as it does
in real walls. After being repaired through grout injec-
tion, and drying for 3 to 4 weeks, depending on the
weather (Fig. 4), the samples were moved to the labo-
ratory to be tested again under diagonal compression.
The failure load of the repaired samples was com-
pared with that of the original samples to assess the

Figure 3. Diagonal compression test of adobe masonry.
Figure 4. Repairing wall specimens through injection of
mud-based grout.
degree of strength restoration achieved by the repair.
It is important to note that while in the indirect tension
tests, the mortars were of different mixes, in this case,
mortars were always made of mud and the grouts were
made of different mixes.
2.4 Injection process
The process of repairing by injecting mud grouts was
conducted in the following steps:
1. Seal the crack faces with gypsum. Also place,
passing through the gypsum seal, 3-mm diameter
1097
Figure 5. Results of indirect tension tests. Tension stress in
the mortar-adobe interface in sandwiches with earthen mor-
tar with 25% moisture. The thinner mortars tend to improve
tension strength at the interface. The curve corresponds to a
linear adjustment using the least squares method.
plastic tubes in order to form the holes through
which the grout will be injected. Once the gyp-
sum is hardened, the tubes are removed to leave the
holes open.
2. Inject water into the holes. This procedure is to pre-
vent the fine material on the inner surface of the
cracks from insulating the injected grout. It also
seeks to provide better lubrication for the injection
of the grout and increases the moisture in the crack
walls, decreases the rate of drying and reduces the
formation of micro-cracks in the filling material.
3. Immediately inject the grout from the bottom up,
through the holes. Grout is injected through a hole
until the material has reached the level of the next
higher hole and begins to escape.
4. Remove the gypsum seal and retouch the outer sur-
face of the injected fissure to achieve an acceptable
finishing.
The equipment used is a discarded cylindrical injec-
tor, which was originally sold in hardware stores
to place putty for glass, or silicon. In the process
of selecting injection equipment, more sophisticated
pressure equipment, pressurized water pumps and
other possibilities were ruled out.
3 EXPERIMENTAL RESULTS
3.1 Mud mortars and grouts without stabilizers
The results of indirect tension tests in sandwiches
revealed that thinner mortar obtained greater tension
strength, as shown in Fig. 5. The values obtained were
similar than those reported for adobe masonry (Vargas
et al. 1986). It was difficult to make sandwiches with
mortar less than 2 mm thick, as it failed to adhere to
the blocks.
To try to understand the influence of the thickness
of the mortar on the strength of the adobe masonry,
 Table 1. Strength recovery in repaired wall sections with
grouts of sifted soil.

Force of fracture (kN)

Wall Recovery
section Initial Repaired of resistance

MDNR-1 7.40 8.43 114%

MDNR-2 6.20 4.44 72%
MDNR-3 8.93 9.63 108%
MDNR-4 8.44 3.65 43%
MDNR-5 6.83 9.17 134%

Figure 6. Maximum tension stress in the interface of sand-

wiches with natural soil mortars +25% moisture and sifted
soil mortars +30% moisture. Sifted soil mortars are, on
average, more resistant.

it was decided to make a series of sandwiches built

with decreasing mortar thicknesses. After 48 hours,
they were opened and the number and width of the
cracks in the mortar were measured. It was found
that although the number of cracks was similar in all
specimens, in the thinner mortars the fissures were
finer. It seems, then, that the greater strength in thinner
mortars is due to the reduced thickness of the micro-
cracks. Eight specimens made with 5 mm thick mortar
were also built and tested in diagonal compression.
These specimens were, on average, 150% stronger than
traditional walls made with mortars 15 to 25 mm thick.
These results indicate that it is possible to build
stronger adobe masonry by reducing the thickness of
the mortar and, in the walls repaired by injection, it
could be expected that the repair of thinner fissures
would lead to greater strength recovery.
To produce finer and more diluted mortars and
grouts, the earth was sifted with a #10 (2 mm) sieve,
leaving large sand particles in the soil, which helped to
reduce microcracking due to drying shrinkage (Vargas
et al. 1986). Thirty to 40% moisture was used in
the mixtures. Figure 6 shows the indirect tension test
results. Mortars made of sifted soil with high mois-
ture levels can be placed in thin layers (2 to 5 mm),
to fabricate stronger sandwiches than those made with
natural soil mortars.
Five 150 mm thick adobe masonry wall specimens
with 15 mm-thick mortars were made and tested under
diagonal compression. The specimens were subse-
quently repaired with natural soil grouts sifted through
a #10 mesh and with nearly 40% moisture, and
retested. Table 1 shows the results. In most of the
wall sections, the original strength increased. In the
two cases that failed to recover the strength, errors in
repairing had been previously observed. These results
indicate that it is possible to successfully repair cracks
from 3 to 5 mm thick in adobe walls by carefully
injecting grouts of mud without stabilizers.
Figure 7. Maximum tension stress in the mortar-adobe
interface in indirect tension tests of sandwiches with mortars
of sifted soil stabilized with gypsum. Only the mortars with
20% gypsum achieve greater strength than those of sifted soil.
In general, diagonal compression testing produces
one or two large cracks, without visible cracking
nearby. Only visible cracks were repaired. In cases
of effective repair, the cracking of the repaired wall
specimens followed a path distinct from the origi-
nal. When the repair was not well done, cracking of
the repaired specimen occurred along the path of the
original, poorly repaired crack.
3.2 Mortars and grouts of stabilized mud
To study the effect of the addition of cement, lime or
gypsum to mortars and grouts, several series of indirect
tension tests were conducted, using sandwiches with
mortar prepared with percentages of stabilizers rang-
ing from 5% to 20% of the total weight of the soil. The
strength of these sandwiches was compared with that
obtained in sandwiches with mortar of natural sifted
mud and about 35% moisture.
The results with the gypsum stabilizer in 5%,
10% and 20% percentages were generally better
than those obtained with lime and cement. How-
ever, only the gypsum in percentages of around 20%
improved the strength obtained with mortars of natu-
ral sifted soil, as shown in Fig. 7. This result could be
attributed to the fact that the gypsum being in contact
with moisture in the mixture increases the volume,

1098
Table 2. Strength restoration in wall specimens repaired the cracked adobe walls. Good quality of the injection
with grouts of sifted soil stabilized with gypsum in 10%, process is crucial to achieving effective repair.
20% and 30% proportions. Stabilizing mud mortars by adding cement or lime
did not produce good results. Gypsum was the best
Force of fracture (kN)
Gypsum Wall % Strength
of the stabilizers studied. However, its use is ques-
content specimen Initial Repaired recovery tionable, since gypsum increases the difficulty of the
repair.
10% MDNR-8 7.62 6.59 86% A thinner mortar increased adobe masonry strength.
MDNR-9 9.11 10.25 113% This implies that by grouting thinner cracks, a more
MDNR-11 7.40 3.59 49 % effective repair may be obtained.
MDNR-12 7.37 5.96 81% Further research is necessary on repairing thinner
20 % MDNR-7 9.46 12.79 135% structural cracks and thicker walls, in order to cover
MDNR-13 5.55 11.21 202% all the cases presented in historical adobe monuments.
MDNR-15 10.44 6.37 61% New types of grouts and more efficient injection
30 % MDNR-10 5.43 9.27 171%
techniques must continue to be explored. Emphasis
MDNR-14 8.32 8.63 104%
MDNR-20 13.22 15.14 115% should be placed on the use of natural and processed
additives through the study of their mechanical and
physical-chemical properties.
which reduces the drying shrinkage of the earthen
mortar, and therefore reduces the micro-cracking and ACKNOWLEDGEMENTS
increases the strength. The increased volume of the
mortars stabilized with gypsum has been confirmed by A previous version of this paper has been presented,
measuring the volumetric shrinkage of 20% gypsum in Spanish, at the Adobe USA2007 conference, held
pastes (Mattone et al. 2005). in El Rito, New Mexico in May 2007. The authors are
The strength of the sandwiches prepared with sta- grateful to the Academic Research Directorate of the
bilized soil with lime or cement, in 5%, 7% and 10% PUCP for providing funding, without which it would
percentages, was generally lower than that obtained have been impossible to carry out this project. They
with sifted soil. Therefore, only adobe wall specimens would also like to thank the staff of the Laboratory
repaired with injection of grouts of sifted soil and sifted of Structures of the Department of Engineering, espe-
soil stabilized with gypsum were subject to diagonal cially its director, Eng. Gladys Villa-Garcia; Dr. Stefan
compression tests. Simon, for his ideas and suggestions; and Mary Hardy,
Soil from the PUCP was used in manufacturing and Project Specialist of the Getty Conservation Institute,
repairing the diagonal compression specimens. The for their efforts, recommendations and information in
cracked specimens were repaired with grouts of sifted support of the working group.
soil with 10%, 20% and 30% gypsum. The percentage
of moisture used, from 35% to 40%, is the minimum
needed to practice injection of the grout. Injections of REFERENCES
grouts stabilized with gypsum were more difficult to
do than those of only soil, because the gypsum makes Mattone R, Pasero G, Rivotti A, and Tosco V. 2005. Uso de
the grouts harden faster. Table 2 shows the degree of productos naturales para mejorar el comportamiento al
strength restoration obtained. agua de revoques a base de tierra. IV Seminario Ibero-
Americano de Construcción de Tierra. Portugal.
The results show that in order to increase the Tolles L, Kimbro E, Webster F, and Ginell W. 2000. Seis-
strength of the adobe masonry walls repaired with soil, mic Stabilization of Historic Adobe Structures. Getty
it is necessary to add gypsum in percentages between Conservation Institute Scientific Program Reports. Los
20% and 30%. However, it is possible that the relative Angeles.
difficulty of the process of repair using grouts with Tolles L, Kimbro E, and Ginell W. 2003. Planning and
gypsum could affect the quality of the repair. Engineering Guidelines for the Seismic Retrofitting of
Historical Adobe Structures. Getty Conservation Institute
Scientific Program Reports. Los Angeles.
4 CONCLUSIONS AND RECOMMENDATIONS Vargas J. 1979. Albañilería de Adobe con Variaciones de
Mortero. Pontificia Universidad Católica del Perú, Publi-
The proposal to repair structural cracks in historic cación DI-79-02. Lima, Perú.
Vargas J, Bariola J, Blondet M, and Mehta P K. 1986. “Seismic
adobe buildings appears to be a viable alternative, Strength of Adobe Masonry”. Materials and Structures,
based on criteria of resistance, to return the building Vol. 19, No 112, July–August.
to its original, uncracked condition. WJE Engineers, Architects, Material Scientists. 2004. Ran-
Repair by injection of grouts made of soil without cho San Andres Castro Adobe Rehabilitation. California
stabilizers was effective for recovering the strength of State Parks.

1099
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Development of an in situ penetration test for the uptake of preservatives in

applied wood

D.F. Henriques
Instituto Superior de Engenharia de Lisboa, Lisboa, Portugal
Laboratório Nacional de Engenharia Civil, Lisboa, Portugal

L. Nunes
Laboratório Nacional de Engenharia Civil, Lisboa, Portugal

J. de Brito
Instituto Superior Técnico, Lisboa, Portugal

ABSTRACT: This paper concerns the study of biocides application in old timber structures of maritime pine
(Pinus pinaster Ait.), previously impregnated with other products. A method was developed in laboratory to
determine in situ the penetration depth of a product applied superficially. As initial treatment, three traditional
products for sawn timber for buildings were used and, for new treatment, two newer, more environmentally benign
products were used. Their ability to penetrate the pre-treated surfaces was evaluated after 1, 2 and 3 applications
at 24 hours intervals and the results obtained are presented. Finally, the applicability of the developed test to the
in-situ evaluation of timber structures is also discussed.

1 INTRODUCTION manufacturers voluntarily withdraw CCA products

Among the anomalies most currently observed in
buildings timber structures those due to the presence of
xylotrophic organisms stand out, such as rot fungi and
subterranean termites, as well as dry-wood termites
and wood-worms. The first two occur in timber with
high moisture content and the second two in dry timber.
Both the maintenance of a good conservation state
of the timber and the curative treatment when there is
an infection or infestation normally resort to the use of
chemical preservation products. These should be dully
registered after application.
As for the type of products it can be referred that
until the early 90’s of the XX century the use of
CCA (chromated copper arsenate) compounds was
generalized, as well as active substances such as
pentachlorophenol, copper, tin or lindane applied in
organic solvents (LOSP). Although some of these for-
mulations were very efficient in extending the life
of wood, the health hazard of operators and the risk
of environmental impact in soil and landscape has
to be seriously questioned (Savluchinske-Feio et al.,
2007). Therefore, in the last decades severe restric-
tions to the use of many of the substances mentioned
above were imposed in Europe (Directive 2003/2/CE)
and for instance in the USA, the wood preservative
from the market.
Also frequently applied there are some traditional
products based in ancient practices, such as linseed
oil, adequate as a water repellent finishing, and used
motor oil that though with very arguable effectiveness,
is often used in Portugal in the preventive treatment of
wood.
In this context present maintenance interventions
in ancient buildings will necessarily lead to the use
of newer and more environmentally benign products
for existing timber. However, the application of a
curative/preventive measures is often impaired by the
presence of previous treatments or finishes not always
well documented and difficult to characterize.
Within the development of a PhD project concern-
ing the rehabilitation of degraded timber structural
elements due to biological agents, through labora-
tory and in situ analysis, the need arise to develop
a penetration test that would allow a swift evalua-
tion of the possibility to re-treat with a certain new
wood preservative with minimum disturbance to the
structure under rehabilitation.
Three types of products were considered as rep-
resentative of previous treatments of old structures,
namely, oil-borne preservatives (in the present case
the “traditional” used motor oil), light organic solvent

1101
preservative (LOSP) using white spirit as the solvent
carrier to deliver the actives into timber and arsenic
copper combinations such as chromate copper arsenate
(CCA). The later was not considered relevant under the
present study as it usually performs well when applied
in roof timber structures and re-treatments are seldom
needed.
This paper describes the test developed and presents
the results of its calibration for untreated maritime pine
(Pinus pinaster, Ait.) and for maritime pine treated
with either an oil carrier formulation or two typical
LOSP formulations, one that has disappeared from
the market due to the active ingredients restrictions
already mentioned and one that has been marketed for
a very long time.
Two products were chosen for the re-treatment, one
of them a water borne product (boron based) with bet-
ter toxicological and environmental characteristics and
a second product (LOSP) of common use at the present
time, easily obtained for the curative and preventive
treatment of timber structures. Their ability to pene-
trate the pre-treated surfaces was evaluated after 1, 2
and 3 applications at 24 hours intervals and the results
obtained are presented.
These products and their sequence of application
are represented schematically in Figure 1. In Table 1
the active ingredients and the solvents of each of the
products applied are presented.

2 WOOD TREATMENT

In historical terms, industrial preservation of timber

Figure 1. Schematic representation of the application
sequence of the preservatives.
started in 1838 when John Bethell registered a patent
for timber treatment with creosote in autoclave, using
the so-called Bethell of filled-cell process, which is
even today the most frequently used preventive in
depth treatments even if coupled with other products.
Much later the empty-cell impregnation process came
along using creosote (Rueping in 1902 and Lowry in
1906), a method still used with the same product 90
years later.

Table 1. Description of the products used in the test.

Product Type Active ingredients (a.i.) Solvent

Initial treatment products

A LOSP 5,0% – Pentachlorophenol
0,9% – Lindane White spirit
0,5% – Diclofluanide
B Oil-borne Used motor oil – unknown
a.i. unknown
C LOSP 13,75% – Copper naftanate White spirit
New treatment products
Water – Water
White spirit – White spirit
D Aqueous Sodium oxide 14,70%
Solution Boric oxide 67,10% Water
C LOSP Cypermethrin 0,17%
I.P.B.C 0,15%
Propiconazole 0,15% White spirit
Tebuconazole 0,15%

1102
 Creosote remained until the 30’s of last century
as the only recognized timber preservative product.
In 1931 pentachlorophenol was patented and in 1933
the first product containing copper salts, chromium
and arsenates, commonly known as CCA, followed
suit. The domination of these three substances was
only slightly shaken by the introduction in the 50’s
of boron in treatments by immersion – diffusion and
especially of products in light organic solvent, some
of which were 10 years later top sellers, mostly in the
market segment so-called “general public” due to the
versatility and ease of application (Nunes, 2001).
The last two decades of the XX century dramatically
changed the panorama of an area that seemed to be
consolidated, environmental concerns, disposal issues
and general public perceptions questioned the use of
traditional active substances (Goodell et al., 2003).
Research related with the search of new active raw
materials for timber treatment has been a field of
intense work in the last years. Copper (e.g. alkaline
copper quat and copper azole) stood out as fungi-
cide agent, applied in formulations without chromium
and arsenate, both in organic solvent and in aqueous
basis. Products based on copper are therefore posi-
tioned as the predominant solution in the near future
(Humar et al., 2007) for the preventive long term treat-
ment of wood. However, rot fungi that are resistant to
copper have been known for some time and most of
the environmental concerns that were expressed for
other active ingredients may also be called against
copper use.
The use of boron should also be considered when-
ever the risk of leaching is not a conditioning factor,
since it possesses good fungicide and insecticide char-
acteristics (Nunes, 1997). Laboratory and field tests
have demonstrated that timber treatments with boron
are very effective in the prevention of brown and white
rot as well as insect attack (Jorge et al., 2004) and
wood modification together with boron might be able
to help the recognized problem of leaching (Kartal
et al., 2007).
New multi-components biocide systems to inhibit
the growth of moulds, rot fungi and termites are
also under development and show promising results
(Clausen & Yang, 2007).
The use of natural products, such as rosin deriva-
tives and other extractives and modified wood, are
also possibilities to take into account (Rowell, 2006;
Savluchinske-Feio et al., 2007).
3 STUDY DESCRIPTION
3.1 Objective
The objective of the present study is the develop-
ment of a method to be used in situ to determine
the impregnation depth achieved by a new generation
1103
biocide product, when applied on timber from an
ancient building treated with a non-documented prod-
uct difficult to characterize without extensive and
eventually expensive analysis.
3.2 Development
For laboratorial simulation of real situations, con-
trolled quantities of the products chosen for the initial
treatment were applied in a first step and later on new
products of prevention and/or treatment were similarly
applied.
To quantify in situ the penetration depth, a method
was devised, very little intrusive and easy to apply to
any building timber structure. It consists on drilling
holes parallel to the face under treatment, at distances
from it successively bigger, in which cotton swabs tes-
timonies are inserted before the application of the new
treatment product. The testimonies absorb the liquid
impregnated into the timber when it reaches the depth
that corresponds to its position.
4 MATERIALS AND METHODS
4.1 Test specimens
For this purpose, 50 × 7 × 5 cm specimens were used,
cut from dry maritime pine (Pinus pinaster Ait.),
with various ring dispositions, from pith and sap,
with some knots and some sapstain. The aim was
simulate non-exhaustively the diverse real situations
that can be found in ancient buildings timber struc-
tures. These specimens were divided in four groups,
to three of which a pre-treatment was applied. The
remaining one, without treatment, was used for control
purposes.
The preparation of the test specimens occurred five
and a half months after the application of the pre-
treatment and according to the procedure described
next. Four holes (diameter = 0,45 cm) were drilled
parallel to the face under treatment and at distances
successively greater from it (0,5, 1,0, 1,5, and 2,0 cm),
as shown in Figures 3 and 4. Six plastic rings were
then glued over the larger face that had fewer defects,
as represented in Figures 2 and 4, to limit the applica-
tion of the new product. A transparent epoxy adhesive
was used, uniformly distributed in the border and
around the ring, in order to confine the absorption of
the product to the area inside the ring. The compat-
ibility between the glue and the solvents was tested
experimentally.
After the preparation, the specimens were left to
stabilize for 1 week in a conditioned chamber with
constant relative humidity (65 ± 5%) and temperature
(20 ± 2◦ C). The moisture content just prior to the tests
was as presented in Table 2. This value was obtained
using a needle humidimeter.
 Table 3. Average absorption of initial treatment.

Initial treatment Average absorption (g/m2 )

products ± standard deviation

A 225 ± 37
B 218 ± 47
C 246 ± 22
Figure 2. Dimensional representation of the test specimen
and location of the rings for the introduction of the new
products.
Table 4. Densities and quantities of new treatment products
applied.

Quantity applied

Density at Absolute volume Mass per unit

Product 20◦ C (kg/m3 ) (ml) area (g/m2 )

Water 1000 1.0 455

White spirit 670 1.0 305
D 1000 1.0 455
Figure 3. Dimensional representation of the holes location. E 800 1.0 364

is unknown. Table 1 presented the active substances

in percentage of mass and the solvents of each of the
products used.
The pre-treatment products were applied with a
brush on one of the larger faces of each specimen
(7 × 50 cm). From each of the 4 initial samples four
specimens with the highest levels of absorption of
the three products referred were chosen (Table 3).
Therefore, 16 specimens were subjected to testing.

4.3 New treatment products

Figure 4. Detail of a specimen prior to being tested. To verify the method under development two products
(D and E) were used, as described above, as well as
Table 2. Test specimens and moisture content before the their solvents, water and white spirit.
new treatment. To make the presence of the products and solvents
more easily recognizable a dye was added before treat-
Test specimens and moisture content ment. After various preliminary selection tests, it was
decided to use Neutral Red as a marker for water and
0A 12.5% 1D 14.0% 2A 12.3% 3D 14.0% the boron-based product. For organic solvents, the dark
0B 14.0% 1G 12.5% 2D 13.2% 3E 14.2%
0D 13.3% 1H 11.5% 2H 12.0% 3H 12.5%
blue dye Ceresblau was selected. The concentrations,
0E 12.2% 1I 14.2% 2I 13.0% 3J 12.5% measured in mass, were as follows:
– Water + 0.5% Neutral Red;
– Water + 5.0% Product D + 0.5% Neutral Red;
4.2 Initial treatment products – White spirit + 0.5% Ceresblau;
As referred three types of products were used for the – Product D + 0.5% Ceresblau.
initial treatment, considered representative to simulate With the help of a pipette, 1 ml of product was
real situations. Products A and C are both light organic applied over the area defined by the interior of the pre-
solvent preservatives (LOSP) and product B is an oil- glued rings, measuring 22 ± 0.8 cm2 . The density of
borne product obtained from an auto repair-shop and is each product and the quantity introduced are presented
a mix of various used motor oils, whose composition in Table 4.

1104
 Untreated wood
2,50

Penetration medium (cm)

2,00
1st application
1,50
2° application

1,00 3° application

0,50

0,00
Water White Spirit Product D Product E
New treatment
(a)
Initial treatment with Product A - LOSP
2,50

Penetration medium (cm)

2,00
1st application
1,50
2° application

1,00 3° application

0,50
Figure 5. Sequence of actions relative to the second appli-
cation of colored white spirit over product B: (a) application 0,00
Water White Spirit Product D Product E
of new product; (b) covering with a membrane; (c) after 24 h,
New treatment
visual checking that the product was totally absorbed by the
(b)
timber; (d) testimonies reading.
Initial treatment with Product B - Oil
2,50
4.4 Methodology
Penetration medium (cm)

2,00
The new treatment products were applied in three 1st application
1,50
coats with 24 h intervals between them. Before the 2° application

3° application
application, the testimonies were positioned within 1,00

the specimens. After the application, the rings were 0,50

covered with a membrane in order to stop the loss of 0,00
product by evaporation. Water White Spirit Product D Product E

Six replicas of application were performed for each New treatment

of the new treatment products for each situation of pre- (c)

treatment, randomly distributed over the specimens. A Initial treatment with Product C - LOSP
total of 6 × 4 × 4 = 96 product application locations
2,50
Penetration medium (cm)

were tested for each of the three coats. 2,00

After each period of 24 h the testimonies were 1,50

1st application

2° application
removed and the respective presence or absence of
1,00 3° application
color registered.
In Figure 5 (from a. to d.) the sequence of non- 0,50

destructive laboratory test actions is illustrated. 0,00

Water White Spirit Product D Product E

5 RESULTS New treatment

5.1 Results from the non-destructive analysis
Readings of the testimonies were visual for water,
white spirit and Product E. In the case of product D,
the presence of boron was also verified by a colorimet-
ric qualitative method as the testimonies were sprayed
with curcumin reagent which is considered adequate
for boron detection (Nunes, 1997). The reagent con-
sisted of 0.25 g of curcumin and 10 g of salicylic acid
dissolved in 10 ml of concentrated hydrochloric acid
diluted with 100 moles of ethanol.
In the situations where the dyed product involved
the cotton of the testimony in at least ¾ of its cross-
section it was considered that it had reached the axis
of the hole, i.e. the depth of 0.5, 1.0, 1.5, or 2.0 cm.
(d)
Figure 6. Results obtained using the non-destructive anal-
ysis for the various situations studied: (a) Test specimens
untreated; (b) pre-treatment with product A; (c) pre-treatment
with product B; (d) pre-treatment with product C.
In the situations where less than ¾ of the cross-
section were involved by the product, including those
in which there were only traces, it was considered that
the product was probably not well distributed around
the hole and therefore the depth considered was the
one over the hole, i.e. 0.25, 0.75, 1.25, or 1.75 cm.
The results were read for the three coats of the
six replicas for each of 16 situations considered.

1105

Figure 7. Results obtained using the destructive analysis for
the various situations studied: (a) water over pre-treatment
with product B; (b) white spirit over pre-treatment with prod-
uct C; (c) product D over pre-treatment with product C; (d)
product E over pre-treatment with product C.
The average absorption depths obtained for each coat
are presented in the graphs of Figure 6.
5.2 Results from the destructive analysis
To check the results obtained previously some of the
specimens were cut longitudinally in half and in a quar-
ter of the cross-section. It was then possible to register
through visual analysis the penetration depth of the
two products in organic solvent and in water.
It was confirmed that the products were uniformly
absorbed by the timber, leading to a continuous surface
of penetration in the cut cross-section, for the white
spirit and Product E, and an irregular surface for the
water and Product D.
The two products in organic solvent penetrated in
some cases more than 2.5 cm as illustrated in Figure 7.
6 DISCUSSION
The objective of the method under development, in lab-
oratory, is to be used, in future, to determine in situ the
penetration depth of a product applied over an already
treated surface.
The main results obtained, so far, can be summa-
rized as follows:
• The penetration of the organic solvent preservative
and the organic solvent alone was good, indepen-
dently of the product used as initial treatment;
• The same was observed for the water borne product,
the initial treatment does not seem to influence the
results.
• The penetration of the organic solvents is, as
expected, much better (around × 1.5) than the water
borne ones and would most probably be adequate for
a curative/preventive treatment;
1106
• Also, for the organic solvents alone each application
seems to increase the depth at which the product
penetrates (around 0,5 cm);
• The number of applications does not seem to espe-
cially influence the final result, for water and the
water borne product.
The applicability of the developed test to the in-
situ evaluation of timber structures was not yet fully
studied but the results obtained are quite promising as
a good picture of the distribution of the products inside
the wood was obtained just by the insertion of easy to
obtain cotton swabs in pre-bored small holes.
ACKNOWLEDGMENT
The authors acknowledge the contribution of Sofia
Simões to the work presented.
REFERENCES
Clausen, C.A., Yang, V. 2007. Protecting wood from mould,
decay, and termites with multi-component biocide sys-
tems. International Biodeterioration & Biodegradation.
9(1): 20–24.
Directive 2003/2/CE, 2003 – “Restrictions to marketing and
use of arsenates”. Official Journal L4, Brussels.
Goodell, B., Nicholas, D.D., Schultz, T.P. 2003. Introduction
to wood deterioration and preservation. In Wood Deterio-
ration and Preservation. Advances in our changing world.
Ed. by Goodell, B.; Nicholas, D.D.; Schultz, T.P. ACS
Symposium Series 845, Washington, DC. pp. 2–7.
Humar, M., Bucar, B., Pohleven, F. 2007. Brown-rot decay
of copper impregnated wood. International Biodeteriora-
tion & Biodegradation 58(1): 20–24.
Jorge, F.C; Nunes, L; Botelho, C. 2004. Boron in wood preser-
vation: Problems, challenges and proposed solutions. An
overview on recent research. Journal from the Faculty of
Science and Technology (University Fernando Pessoa) 1:
3–15.
Kartal, S.N., Hwang, W-J., Yamamoto, A., Tanaka, M.,
Matsumura, K., Imamura, Y. 2007. Wood modification
with a commercial silicon emulsion: effects on boron
release and decay and termite resistance. International
Biodeterioration & Biodegradation 60: 189–196.
Nunes, L. 1997. The effect of boron-based wood preservatives
on subterranean termites. PhD Thesis. Imperial College
of London, 292 pp.
Nunes, L. 2001. Preservation of construction timber. Present
situation and future perspectives (in Portuguese). LNEC,
COM 79, Lisbon.
Rowell, R. 2006. Chemical modification: a non-toxic
approach to wood preservation. 2nd International Con-
ference on Environmentally-Compatible Forest Products,
Fernando, Pessoa University, Oporto, Portugal, 20–22
September 2006. pp. 227–237.
Savlunchinske-Feio, S., Nunes, L., Pereira, P.T., Silva, A.M.,
Roseiro, J.C., Gigante, B., Curto, M.J.M. 2007. Activity of
dehydroabietic acid derivatives against wood contaminant
fungi. Journal of Microbiological Methods 70: 465–470.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Special structural solutions for adaptive use in grand hotel in

Norwich, CT USA

J.F. Norden
Gibble Norden Champion Brown, Consulting Engineers, Inc. Old Saybrook, USA

ABSTRACT: The Wauregan Hotel (c. 1855) in Norwich, CT, USA was restored in 2005–6 and adaptively
reused for new affordable housing and retail. The structure was unique with sloping timber joists to accommodate
early indoor baths with full mortar setting beds. Floors were leveled and strengthened by inserting engineered
wood members (Laminated Veneer Lumber). The exterior walls contained narrow fireboxes and stove flues where
central boilers could not go at that time. These flues needed to be grouted solid as they represented weakened areas
in these bearing and shear walls. Special construction shoring was designed to allow environmental abatement
while providing safe working floors for the selective demolition and abatement crews. Modern Codes (IBC
2003) required a full seismic upgrade and an interactive structural analysis was performed using 1850’s brick
bearing walls in conjunction with interspersed reinforced concrete block walls. Special diaphragm to shear wall
ties were devised and over 1500 installed.

1 INTRODUCTION
The Wauregan Hotel was constructed in downtown
Norwich, Connecticut, from 1853 to 1855. Local busi-
nessmen, manufacturers, and railroad and steamboat
interests formed a corporation for the purpose of cre-
ating a first-class hotel, designed to accommodate
travelers between New York and Boston, and those
visiting Norwich on business or pleasure. The upper
two floors of the five story building were rented to
permanent, year round residents.
The president of the corporation was William Buck-
ingham, later Republican governor of Connecticut
during the U.S. Civil War. Local architect Evan Bur-
dick designed the building in the Italianate style.
The hotel contained rooms for public meetings. Plan-
ning meetings for commemorative events, such as the
200th anniversary of the founding of Norwich, polit-
ical meeting, and public receptions were held at the
hotel. The most famous guest was Abraham Lincoln,
on March 9, 1860. Lincoln spoke at the Norwich city
hall that evening, and later held forth in the hotel
parlor. Other distinguished visitors included President
Ulysses S. Grant and Booker T. Washington, in whose
honor receptions were given. Gypsy Rose Lee stayed
at the Wauregan while giving performances at the
summer stock theater in town.
The Clarendon Annex, with a second floor dining
room/ballroom, was added as a north, integral wing,
in 1906. The hotel lobby, entrance, and storefronts
were remodeled about the same time. About 1950, the
ornate Victorian window trim was removed to “mod-
ernize” the building during the post WWII modern
image era in the USA. Construction of the interstate
highway system, and motels and hotels by the high-
way exits, was a factor in the decline of the Wauregan
as a hotel in the second half of the 20th century. By
the 1980’s, the building was deteriorating. A fire in
1989 destroyed the boiler room, in a rear addition, and
the Wauregan was condemned. In 1999, the City of
Norwich announced its intention to demolish the hotel.
The Connecticut Historical Commission, in a decision
on December 1, 2001, voted to oppose the demoli-
tion, as the Wauregan was a contributing structure
to the Downtown Norwich National Register Historic
District. The entire complex became a major restora-
tion and development project by Becker & Becker
Associates and has transformed the building into 70
apartment units with retail space on the first floor
(Plummer, D. 2007).
Significant structural challenges had to be met due
to the devastating water and environmental damage
sustained during the vacancy period. Modern Code
requirements came into play adding complexity to the
safe structural design required for these upgrades. This
paper will highlight some of these unique challenges;
their solution allowing this historic structure to be
saved and progressively transformed to a downtown
focal piece and impetus for future urban development
in Norwich.

1107
2 GENERAL DESIGN REQUIREMENTS coefficient, R, for URM walls is 1.5 as opposed to a
higher value of R = 5.5 for special reinforced masonry
Generally, the building was going from a commercial shear walls which we used in numerous locations as
hotel use to a new multi-family residential use. This replacement walls. As can be seen in the base seismic
change in consort with the major level of renovation shear (V) equation.
triggered the requirement for a structural upgrade to V = 1.25DS W/R, a smaller R value produces a large,
current code standards. Structurally, this meant com- more conservative seismic design base shear. Even so,
plying with Chapter 16 “Structural Design” of the 2003 the high wind loads offset by the lower floor mass (W)
International Building Code (IBC). Both gravity floor produced a higher wind base shear for design purposes.
and roof loads needed to meet the current code require-
ments as well as an analysis and upgrade for lateral
loads caused by either wind or earthquake.
3 STRUCTURAL CHALLENGES
The building’s structural systems were comprised of
sawn timber joists of local pine species for both the low
Due to water damage and severe winter climate of New
sloping roofs and interior floors.These were bearing on
England, the building’s unoccupied years and lack of
the exterior, multi-wyeth brick walls and interior stud
any maintenance caused areas of major deterioration.
corridor or room demising wall partitions. Original
In particular, one area was in extremely bad condition
wood lath and lime plaster was the predominant wall
and collapse in 2004 was imminent. The Wauregan’s
finish material even on the exterior walls which were
rear, reentrant corner five store rear walls were in dan-
wood furred and then plastered.
ger of collapsing. The roof’s valley terminated at the
Gravity live loads for hotel use and modern apart-
corner and water cascaded down these corner walls
ment use were almost equivalent, and most of the
causing excessive damage. In order to save the floor
existing floor joists which were not water damaged
framing in these walls, special two-way heavy timber
were found to be capable of sustaining the necessary
shoring was designed and erected in a tiered fashion.
superimposed loads required for structural upgrade.
This was completed in September of 2004 and over
Where members had minimal damage, new “sister”
the upcoming winter the walls did collapse, exposing
timbers could be added and fastened to the existing
the buildings interior to the weather. The floors were
timber to provide the necessary reinforcing. In the
saved, however, due to the prior shoring being timely
case of major damage of floor timbers and joists, new
installed (See Fig. 1).
members were inserted to replace these excessively
This shoring remained in place until the floor could
damaged ones. Stronger, modern laminated veneer
be adequately reinforced and the new exterior wall
lumber (LVL) members or engineered I joists were
chosen as their depth could replicate the original
timbers and provide added strength and serviceability.
The building’s lateral loads were originally taken
by the mass brick exterior walls as a “box system”
with some interior masonry walls being used to act
as bearing and shearwalls. These augmented the exte-
rior walls and added strength due to the building’s ell
shape. The design basic wind speed of 120 mph for
Norwich produced the greater building design lateral
loading as opposed to the seismic lateral requirement.
Connecticut is in a moderate seismic zone, with the
State’s ground accelerations per Code analysis pro-
ducing less forces in the building’s lateral resisting
structural elements. Two factors contributed to this
result. First, the building’s height (5 stories for the
Wauregan and 6 stories for the Clarendon Annex) pre-
sented a substantial area for wind loads, and second,
the light weight wood floor did not contribute sub-
stantial mass in the earthquake’s lateral load formulas.
The wind’s total base shear was thus in excess of that
produced by seismic analysis. It is worth noting here
that due to the re-use of the building’s numerous orig-
inal unreinforced masonry brick walls (URM’s), the
most conservative seismic design factors were used
in the analysis. The seismic response modification Figure 1. Collapsed corner walls.

1108
reconstructed. Where possible, new reinforced con- Demolition. This of course uncovered numerous struc-
crete or hollow, grouted reinforced concrete masonry tural issues including faults in the major shearwalls
units (CMU) were installed for the necessary shear- due to old fireplaces and, more notably, old unlined
wall strengthening at the building’s lower levels (See flues. These flues were voids in the building shear-
Fig. 2). The rear of the building was being covered walls that were unanticipated during the earlier design
with an Exterior Insulated and Finish System (EIFS) phase. These voids had to be grouted from above, but
and stuccoed a brick color to match the existing fin- the old soot/tar inhibited the bond of grout to brick.
ished brick. This was a cost savings effort applied to a To bond these two together to act as a single shear-
building’s rarely viewed exterior walls. wall unit, new 9 mm Helifix stainless steel self drilling
Again, excessive water had damaged much of the rod anchors were inserted through the brick and void
building’s original wood deck floor including sup- to provide a “shear-lug” anchor system. These were
porting joists. As this original timber deck would not
provide the necessary floor diaphragm to resist loads
due to wind forces, it required replacement or reinforc-
ing. A new, lightweight flowable gypsum floor fill was
specified to be used in floor areas where the old sags or
deflection exceeded a value of 1 in. (2.5 cm). In order
to achieve the required diaphragm design values and
provide a sufficient substrate to receive the gypsum
floor fill, a new layer of 3/4 in. (1.9 cm) plywood was
applied over the original decking. Where the original
decking was rotted and removed, new in-kind decking
was placed prior to the new plywood installation (See
Fig. 3).
As necessitated per Code, prescriptive for seismic
requirements, all floor diaphragms must be attached
to transfer their in plane loads to the building’s shear-
walls. Generally, 1850’s construction was very weak
on this requirement and no original connections were Figure 3. Damaged floor decking.
found. With all new plywood, this diaphragm was then
attached to the brick or CMU shear walls with clip
angles, screws and epoxy inserts to the masonry. Some
1500 such clips were required to provide this vital
positive attachment (See Fig. 4).
The project developer chose not to have the plas-
ter finishes removed prior to bidding the project due
to initial lack of project funding at that point. This
left the finish removal as part of the General Con-
tractor’s job during construction, Phase I – Selective

Figure 2. Repair collapsed corner. Figure 4. Seismic diaphragm ties.

1109
 Figure 6. Trussed ballroom joists.
Figure 5. Flue/grout anchor detail.

drilled at a 22.5◦ angle to insure solid brick would be

penetrated as opposed to being in a more weakened
lime mortar joint. Over 2000 of these re-anchors were
employed successfully and economically to solve this
complex problem (See Fig. 5).
An unusual find during plaster removal were the
long span joists of the second floor ballroom in the
c. 1906 Clarendon ballroom. These timber joists were
2 in (5 cm) × 14 in (36 cm) spaced at 18 in. (46 cm)
on center. They were spanning approximately 22 ft.
(6.70 m) to a central steel carrying beam. The origi-
nal builders or designers knew the heavy live load of
the ballroom floor. The current code calls for 100 psf
(490 kg/m2 ) for this assembly use. As the original Figure 7. Restructured front façade.
plain wood joists could not safely carry this loading
the original design incorporated two 7/8 in. (2.22 cm)
diameter steel rods in an end-trussed condition as
added strengthening pieces. These “queen posts” rods 1906 Clarendon building structures. This façade con-
would come into play when live load was added above, struction was present along the entire east and south
go into tension and then deliver this force to the joist elevations facing Main and Broad streets. The archi-
ends at the bend-up points near the ends of each joist tect wished to keep this façade and show the original
(See Fig. 6). wrought iron decorative columns as well.
This trussed rod apparatus was generally confined These columns at the lowest level below the long
to large, long span industrial timber girders. It was brick shearwall/bearing walls above represented a lat-
quite unusual to find these in a repetitive floor joist eral seismic “soft-story”. The soft-story concept in
application. lateral load design is one in which the more rigid,
upper shearwalls attempt to deliver their shear loads
to the column tops which are unbraced and only fixed
at the top and bottom to the building’s structure. This
4 LATERAL LOAD CHALLENGE large lateral load deforms these columns into an “S”
shape whereupon these columns go into an incipient
The Wauregan Hotel and Clarendon Annex each had buckling mode and fail. In order to prevent this fail-
grade level glass store fronts supported on original, ure mode and still allow the architectural store front
ornate wrought iron columns. Between the columns mode, a new deep beam-wide column moment frame
were low rise brick arches used to span this unsup- was introduced just behind the cast iron store front.
ported distance and carry the multi-story brick bearing New steel columns were placed to the rear sides of
wall above. Again, no consideration to lateral sta- the cast iron columns and new steel headers were
bility was apparent in both the 1855 Wauregan and erected between them to support the overloaded low

1110

Figure 8. Restored façade of the Wauregan Hotel.
(Camardella, P.2006)
rise arches. Poured concrete was added between the
low rise arch void and the steel beams to ensure com-
plete vertical load transfer (See Fig. 7). Note each
of the described elements in Figure 7 including the
new concrete girder behind the top of the store front
windows.
1111
5 CONCLUSIONS
The Wauregan Hotel restoration and adaptive use
project is successful proof that innovative structural
engineering solutions can be employed to save build-
ings that might otherwise be considered hopeless and
ready for demolition. With a thorough inspection and
understanding of each of the structure’s important
components, solutions can be obtained to save and
even improve the functional capabilities. In the case
of the Wauregan project, this was especially acute
as numerous problems arose during the construction
phase which had to be dealt with in a timely fashion
to keep the project on schedule. On site presence dur-
ing this phase was practically on a daily basis by the
structural engineer, and complete cooperation of both
the architect and contractor(s) was tantamount to each
successful solution.
In the final analysis each of the challenging and
often unique solutions when put together went into
making the Wauregan adaptive use project a great suc-
cess. Today it stands as a “poster child” for the preser-
vation community stating that patience and teamwork
can turn the one time wrecking ball project into a new
and vibrant facility restored to its original grandeur
ready and for new life (See Fig. 8).
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

The restoration of the foundation in the north side of the temple of

Apollo Epikourios

K. Papadopoulos
Technical Team for the Restoration of the Temple of Apollo Epikourios, Bassai, Greece

ABSTRACT: One of the main objectives of the ongoing restoration programme of the classical Greek temple of
Apollo Epikourios is the repair and strengthening of its foundation. In this paper the restoration of the foundation
in the north side of the temple (where the work has been completed) is presented. In that part of the monument,
foundation failure was caused, mainly, from the existence of a clay layer, beneath the foundation top course
(the euthynteria). Consequently, the restoration involved removing the clay material, and repositioning it at a
later stage, after stabilizing it with lime and cement. Thus, a durable base was created for the resetting of the
restored euthynteria that followed. The work included, also, the built of small, non-visible walls in the north-west
corner, for retaining the subeuthynteria course. The restoration of the temple north foundation was carried out
with respect to the ancient constructions, and did not alter significantly the original structural character of the
monument.

1 INTRODUCTION its stability, and they endanger its safety, in case of a

The temple of Apollo Epikourios at Bassai, in the
mid-west Peloponnesus, is a Doric temple, con-
structed in the last quarter of the 5th century B.C.
It has 6 × 15 columns in its perimeter and measures
16.14 × 39.84 [m] at the foundation top course. It is
built, mainly, of the white-light gray local limestone.
The temple is one of the best preserved buildings of
classical antiquity (Fig. 1). However, it presents many
severe structural damages: foundation deformations,
stone-blocks fractures in the steps, and significant tilts
of the columns. Furthermore, its material has dete-
riorated extensively due to the extreme weathering
conditions that prevail during winter time in the area
(the temple’s altitude is 1130 m a.s.l.). All those prob-
lems of the monument have diminished significantly
Figure 1. The temple of Apollo Epikourios, before the
beginning of the restoration work, from NW.
future earthquake.
In 1987 the monument was covered with a tem-
porary enclosed protective canopy. The canopy was
erected to shield the monument from damage from the
environment, for as long as required to materialize the
restoration plans that were being formed.
The problems of the temple and the programme for
its conservation and restoration were presented at the
1st International Meeting for the Conservation of the
monument in 1995, where it was widely acknowledge
the need for extensive intervention.
The restoration work began in 2001, and focused on
the north side of the temple, the object of the first phase
of the intervention. It consists in removing architraves
and columns (in full height), dismantling the steps
and the top course of the foundation (the euthynteria),
strengthening the foundation, structurally restoring
and conserving the stone-blocks, resetting the euthyn-
teria and the steps, using new clamps made of titanium,
and reinstating the columns and the architraves. Hav-
ing already restored the foundation and repositioned
the blocks of the euthynteria and the first two steps,
the work is currently directed towards restoring and
repositioning the third step (the stylobate) and reinstat-
ing the six north columns to the monument in vertical
positions.
In this paper the studies and actions (experimental
and numerical investigations, intervention practices)
related to the restoration of the foundation in the north

1113

Figure 2. The foundation in the east side of the temple (cross
section very close to the north side, Svolopoulos 1995).
side of the temple of Apollo Epikourios (where the
work has been completed) are presented.
2 THE FOUNDATION IN THE NORTH SIDE
OF THE TEMPLE
Apollo Epikourios temple is built facing north, on a
natural slope that inclines gently to the west-southwest.
The slope required the dressing down of bedrock in
the east side of the building and in its west side the
addition of an artificial fill and the construction of
walls to retain the fill.
The temple foundation consists of two rectangu-
lar systems: the external, under the colonnade in the
perimeter of the building, and the internal, under the
walls of the cella (Fig. 2). Along most part of the north
side of the temple (as along the east side) the bedrock
is very close to the superstructure. Hence, the foun-
dation in that area of the monument is formed solely
by one layer of stone-blocks, the euthynteria. Between
the euthynteria and the bedrock the ancient builders
had placed a layer consisting of clay fill mixed with
stone flakes, of varying thickness (from zero to few
tens of centimeters). This layer seems to serve as a
convenient way to give each euthynteria block a firm
and continuous contact with the rough bedrock surface
below (Fig. 2).
The bedrock at the north side of the monument
drops about midway from the east, at which point
a single subeuthynteria course begins (Fig. 3). This
course progressively becomes thicker, as it proceeds to
the north-west corner. A layer of clay separates also the
euthynteria from the subeuthynteria stones, and the
subeuthynteria from the rock.
The dismantling, for restoration purposes, of the
north side of the monument allowed for the first time
an extensive part of its foundation to be fully unveiled.
Firstly, it allowed to uncover the euthynteria blocks
(Fig. 3); then (and after the excavation of the clay layer)
1114
Figure 3. The euthynteria stone-blocks in the north side of
the temple (isometric sketch).
Figure 4. The subeuthynteria stones in the north-west corner
of the temple (down view).
it was possible to reveal the subethynteria course in the
northwest corner, five subfoundation stones scattered
at the bedrock, a cavity in the bedrock with a group
of few stones inside, and the bedrock upper surface at
the rest of the excavated area.
The north euthynteria of the temple consists of rect-
angular shaped blocks in two rows, which were laid as
headers (Fig. 3). Twenty-one blocks are in the façade
row, with varying width and with common length and
depth about 0.68 m and 0.28 m, respectively. A pair of
double-lengthened blocks (EU-N16, EU-N19) appears
in the west part. The corners terminated with nearly
square shaped blocks measuring 1.15 × 1.12 [m]. The
sizes of the back row blocks vary significantly; con-
sequently, the vertical joints run in random patterns.
Clamps connected the façade blocks only in the top
face, close to the north edge.
The subeuhtynteria course in the temple’s north side
begins below the sixteenth from the east euthynteria
façade block (EU-N16), (Fig. 3). It consists of irregular
sized cut stones (Fig. 4), some of which (most prob-
ably) came from an earlier monumental building and

Figure 5. The stones inside the bedrock cavity and the
reused, reshaped block in the north limit of the cavity (view
from SE).
were reshaped for new positions. An oversized stone
has been placed in the corner, obviously for stability
reasons, measuring about 2.10 × 3.05 × 0.40 [m]. The
top surfaces of these stones have been shaved down to
form a continuous horizontal bedding for the euthyn-
teria. No clamps secured the subethynteria stones.
The cavity found in the bedrock below euthynteria
(Fig. 5) is roughly semi-spherical in shape, with depth
reaching 85 cm and volume about 2.5 m3 . Its center is
beneath the back face of the sixth from the east euthyn-
teria façade block (Fig. 7). Inside the cavity a group of
stones were found, placed one on top of the other, occu-
pying the northern third of the cavity. The stones are
of various sizes and irregular shapes. The rest volume
cavity was filled with clay. Although the stones have
settled, the highest altitude of the top stone is close to
the altitude of the proximate bedrock, indicating the
initial group elevation. It is obvious that the structural
function of the stones is to form a subfoundation ‘pil-
lar’ for transmitting the building loads to the rock. In
the north limit of the cavity, parallel to transverse axe
of the temple a reused block is laid, based partly on
rock and partly on clay.
3 DIAGNOSIS OF DAMAGES IN THE
NORTH SIDE OF THE TEMPLE
The failure of the foundation of Apollo Epikourios
temple has been recognized as one of the most severe
structural problems of the monument. This is because
the foundation deformations are considered as the
main factor that caused the damages of the build-
ing (Fig. 6). Furthermore, before the enclosure of the
monument with the canopy, the deformation of its
foundation was a phenomenon in progress.
In previous study (Papantopoulos 1995), it was sug-
gested that in the east part of the temple foundation
failure occurred, mainly, because some clay material
was washed away from rainwater, evoking the above
euthynteria blocks to settle. In the west part failure
1115
Figure 6. Sketch presenting, in exaggeration, the defor-
mations in the foundation of the temple (Papantonopoulos
1995).
Figure 7. The top stone of the group inside the bedrock
cavity, the proximate subfoundation reused block and the
neighbouring euthynteria blocks (down view). The black
arrows indicate the inclinations of the stone, the fragments
of the reused block, and the blocks. The white arrow shows
the direction of column N2 tilt. All arrows are pointing at the
lower parts of the cavity.
was caused by the gradual slide of the artificial fill
down the slopping rock, which dragged along the
above foundation. Those assessments were based on
measurements of settlements in different parts of the
monument and observations about the foundation con-
struction, which were made in trenches excavated in
few selective positions around and inside the building.
As the restoration work progressed, the full uncover
of the north foundation of the monument provided new
data, which study further enforce the aforementioned
assessments, confirming the necessity of the extensive
intervention. The new data came from: (a) the docu-
mentation of the deformed positions of the euthynteria
blocks; (b) the detail measurement of the clay layer
thickness; (c) the discovery of the bedrock cavity and
the group of stones inside; and (d) the documentation
of the deformations of the subeuthynteria stones.
More specifically, the data showed that: (i) The
foundation stone members based directly on bedrock
or on thin clay layer presented limited settlements.

Figure 8. The large corner subeuthynteria stone (right) and
its side stone to the east (left). View from north. It is obvi-
ous that the stones have settled, inclined toward the slope in
the west, and that their cracks are at the contact points with
the rock.
Figure 9. Left: The numerical model. Right: North (up) and
south (down) views of the model lowermost part.
(ii) The stones based on thick clay have settled signif-
icantly, inclined in every case toward the side where
the below clay layer was thicker. This was very clear
in the euthynteria blocks placed above the bedrock
cavity (Figs 7, 9 right). (iii) The blocks based partly
on bedrock and partly on clay were fractured across
the base condition differentiation line, with their frag-
ments based on clay suffering the larger settlements.
The latter was quite obvious in the reused block placed
beside the group of subfoundation stones (Fig. 7),
and in the north stones of the subeuthynteria course
(Fig. 8).
Thus, the data strongly indicate that, during the cen-
turies, a gradual reduction of the clay layer led to the
settlements of the foundation stone members. The rea-
sons for the clay layer diminishing were the washing
away of material from rainwater (which started pass-
ing into the temple, after the collapse of the roof),
and the creeping flow of clay toward lower parts of
the bedrock, under the loads of the building and with
the decisive cooperation of rainwater. The gradual clay
escape was toward the lowermost parts of the cavity
(Fig. 7), through the fissures and the discontinuities
of the bedrock, and down the slope in the west side
(Fig. 8).
1116
It should be noted that, although small creeping set-
tlements are certain to have take place in some parts
of the bedrock during the monument’s lifetime, they
are evidences which denote that no significant local
collapse in the bedrock has occurred. The evidences
are the existence of the group of stones inside the cav-
ity and the fact that the altitudes of the top surfaces
of the subeuthynteria stones are close to the altitude
of the top course of the western ancient retaining wall
and to the altitude of the bedrock in the central part of
the north side of the building.
Based on the above assumptions, the course of
events that led to the damages of the temple in its north
side can be described as follows: The north façade,
as the whole building, remained intact for many cen-
turies, until the collapse of its roof, which, probably,
took place due to the biological decay of the wooden
supporting elements (historical data report that the
temple was whole with none noticeable deforma-
tion, six centuries after its construction!). The absence
of cover allowed rainwater inside the building, onto
the blocks of the steps. As a consequence, the rain-
water chemical action caused material deterioration,
diminishing the surfaces of the blocks, opening the
gaps in their interfaces. This fact permitted rainwa-
ter to penetrate further, until, eventually, it was able
to reach the clay layer beneath euthynteria. Then,
rainwater started to wash away clay material and to
make easier the creeping flow of clay toward lower
parts of the bedrock, causing: (i) reduction of vary-
ing magnitude to the clay layer thickness; (ii) sunk
of the subfounadtion stones; and (iii) slide of the
subeuthynteria course. Those damaging for the temple
foundation slow natural phenomena accelerated when
pillagers, searching for the metal clamps, removed the
intercolumnar slabs and broke the façade blocks of
the steps.
The deformations in the supporting system of
the euthynteria generated non-uniform settlements
and irregular displacements of its blocks. (However,
because almost all the blocks remain fully based on
clay, they did not fracture). Naturally, the euthynteria
subsidence led to simultaneous deformations in the
blocks of the steps, which caused loss of full contact
in the blocks interfaces and subsequent modification
of the blocks loading condition from compression to
tension. The latter alteration induced material failure
in the weakness points of the blocks; therefore, fur-
ther settlements occurred in the steps, causing columns
inclinations and entablature deformation. The first
cracks in the blocks of the steps led to rearrangement
of their contacts and consequential creation of new
point or line loadings, which caused new fractures and
so on.
It is worth noticing that as long as the temple
remained intact, earthquakes did not damage it signif-
icantly. Only after human intervention (the breakage
of the blocks and the looting of the clamps by the
pillagers), strong earthquakes probably disintegrated
further the building, by opening the blocks joints and
their fragments gaps or/and by creating instantaneous
or residual stresses concentrations that led to new
blocks fractures.
4 NUMERICAL REPRODUCTION OF THE
DAMAGES IN A PART OF THE TEMPLE
4.1 The numerical model
In order to further enforce the assessment that foun-
dation failure is responsible for most of the damages
in the north side of the temple of Apollo Epikourios,
a numerical analysis was conducted with the aim to
reproduce the damages in part of the monument, as
a consequence of the euthynteria deformations. For
that purpose, with the use of the software Abaqus, a
3-dimentional model was created, which included the
second from the east column of the north side, and
the blocks of the steps and of the euthynteria that are
beneath the column (Fig. 9 left). In other words, in
the analysis it was simulated the part of the temple
which is over the cavity of the bedrock, without the
entablature.
The model’s geometry definition was based on
actual measurements of the temple stone blocks. The
model was formatted in such a way so that all the
blocks (except those of the euthynteria) were simu-
lated in the state and position that they were in the
initial, undamaged condition of the temple; meaning
as distinct members, in full contact, with no relative
displacement, and in vertical or horizontal positions.
On the contrary, the euthynteria blocks were sim-
ulated inclined and displaced (Fig. 9 right), so as
to represent the documented in-situ settled positions
of the corresponding temple blocks. The aim was to
include in the analysis the deformations of the tem-
ple foundation, for finding out what damages will
appear in the superstructure, as a consequence of the
deformed foundation. In order to succeed that, the
euthynteria blocks were fully constrained, while the
rest of the model was free to deform nonlinearly in
any direction under its own weight.
Gravity loading was the only load applied to the
model; it was ramped up over 0.3 sec, and then it was
maintained constant. The dynamic analysis total time
(1.5 sec) was sufficient for the full appearance of the
damages in the model. For simplicity purposes, the
model discretization took place using 8-nodes hexa-
hedra elements that resulted to the cross sections of
the column shaft to be polygonal, instead of circular
with 20 flutes, as in reality.
The material behaviour of the blocks of the three
steps was simulated using a non-linear model, suit-
able for quasi-brittle materials which elastic behaviour
is isotropic and linear and its main failure mechanisms
1117
are cracking in tension and crushing in compression.
The properties adopted in the analysis were based
on experimental study about the temple limestone
material (Papantonopoulos 1995) and are for the mod-
ulus of elasticity 80.1 GPa, for the Poisson’s ratio 0.35,
for the tensile strength 1.4 MPa and for the compres-
sive strength 38.5 MPa. The material behaviour of
the rest blocks was simulated using a linear model,
with modulus of elasticity 80.1 GPa and Poisson’s
ratio 0.35.
The interactive behaviour of the model members
normal to their interfaces was determined with the use
of a ‘hard’contact model that allows when two surfaces
are in contact, any pressure to be transmitted between
them and, when the surfaces separate, it reduces the
contact pressure to zero. In the tangential direction of
the interfaces, a classical friction model was used, with
friction coefficient of 0.75 (value derived from tests,
Papantonopoulos 1995).
4.2 The uncertainties in the analysis
It should be noted that the numerical analysis contains
uncertainties related to the simplifying assumptions
made in the simulating process, as well to the complex
parameters that influence the dynamic behaviour of
the temple. The main points that render the numerical
predictions as a rough estimation of the behaviour of
the simulated temple part are:
The limestone of the monument is anisotropic, due
to the layers and the discontinuities of its structure;
instead, in the analysis it was simulated as a isotropic,
continuous material. The state of the temple when
foundation failure occurred and cracks appeared in the
blocks, is unknown. Hence, it is uncertain if the model
should have included stone members from the entab-
lature, for more accurate simulation of the dead loads
that caused material failures. Furthermore, dynamic
actions due to seismic events are not included in the
analysis; although those actions may have partly con-
tributed to the monument geometrical deformations or
material failures. Finally, elements representatives of
the metal clamps in the blocks of the first two steps
were not included in the numerical model.
4.3 Numerical results – conclusion
The numerical analysis resulted, as it was expected,
both significant displacements and rotations of all the
model members and material failures in the blocks
of the steps. The deformed model is presented in
Figure 10. It is quite obvious that the steps have set-
tled significantly, and that the column, carried along
by the steps, has subsided and inclined toward SE,
that is toward the position of the bedrock cavity in the
monument (Fig. 7). Specifically, the analysis predicted
maximum column deviation from the vertical position

Figure 10. The deformed model, from east (left) and north
(right). It is clear the intensive column inclination toward the
position of the cavity in the bedrock.
Figure 11. Up: The blocks fractures (as documented in-situ)
in the first (left) and in the second (right) step. Down: the
relative numerical results.
by 23.6 cm to the south and by 7.7 cm to the east. In-
situ the column was measured tilting 26.5 cm to the
south and 5.0 cm to the east.
The computed material failures in the blocks of the
first two steps and the respective documented blocks
fractures are shown in Figure 11. The comparison indi-
cates that the majority of the blocks fractures were
reproduced by the analysis at the same or nearby
positions.
In conclusion, it can be formulated that the numeri-
cal simulation resulted damages for the monument part
that is above the bedrock cavity (as a consequence of
foundation failure) very similar to the ones observed
in-situ. Thus, the analysis argues for the assessment
that most of the temple damages were caused by foun-
dation failure, and, by extension, for the necessity of
the large-scale intervention.
1118
5 THE BASIC MATERIAL USED IN THE
RESTORATION OF THE FOUNDATION
5.1 The reasons for the material selection
As previously mentioned, the problems in the north
foundation of the temple of Apollo Epikourios were
caused, mainly, because rainwater gradually washed
away material from the clay layer. Hence, one of the
basic objectives of the foundation restoration was the
replacement of clay with a more durable material,
capable to arrest the damaging phenomenon of mate-
rial loss in the layer beneath euthynteria. In order to
attain that objective, it was decided to reuse the clay
material, stabilized with lime and cement.The decision
was based on the following reasons:
Through stabilization clay develops sufficient
strength which assures durability against weathering
actions, and, consequently, the safe transfer of the
building loads to the bedrock. In parallel, stabilized
clay is a material similar to the one placed by the
ancient builders; thus its use in the restoration work
ensures minimum repair, without altering significantly
the original structural character of the temple foun-
dation. Furthermore, stabilized clay does not develop
very strong adhesive bond with rock, hence, its imple-
mentation is considered almost entirely reversible.
Additional reasons which render stabilized clay as suit-
able material for the restoration work are the facts that
it remains plastic for a few hours after its prepara-
tion (therefore, it facilitates the resetting of the blocks
by allowing microadjustments in the blocks positions
or/and in its top surface) and that it should be com-
pacted for developing the highest possible strength
(thus, its compaction secures that no voids are left in
the rough bedrock surface).
5.2 Basic material properties
The exact composition of the stabilized material used
in the restoration of the temple foundation was derived
from an experimental study, carried out at the Labo-
ratory of Road Construction of the National Technical
University of Athens, in two stages (Kolias 2003,
Papadopoulos & Karahalios 2003). The basic scope of
the study was to investigate various components pro-
portion of stabilized clay, concluding about the most
suitable mixture for use in the restoration work. The
criterion for the mixture selection was the development
of compressive strength above 2.0 MPa, requirement
which, according to the international practice of road
construction, insures that the stabilized clay material
will not suffer damages due to weathering actions.
The constituents proportion (by mass) of the
selected mixture is: 37.6% clay (from the layer beneath
euthynteria), 37.6 limestone sand and gravels (from
crushing the stone flakes from the layer beneath
euthynteria and stones from the surrounding area
of the monument), 3.0% lime, 7.5% cement, and
14.3% water. The basic experimental results about the
mechanical properties of the stabilized clay mixture
which, eventually, was used in the temple restoration
work are the following: The material developed, at
the age of 28 days, compressive strength 2.156 MPa
(average value from testing 8 cylinder specimens,
d/h = 7/14 [cm]) and, at the age of 63 days, 3.204 MPa
(from tests on 3 cylinder specimens); its indirect ten-
sile strength was also measured at the age of 28 and
63 days, and it was found 0.341 MPa and 0.651 MPa,
respectively (the latter values were deduced from
‘Brazilian tests’ on 8 and 2 specimens).
5.3 Test implementation
Before the use of stabilized clay in the restoration work
of the temple, its behaviour was tested under real load-
ing conditions in a structure that simulated a small part
of the building foundation. More specifically, a layer
of stabilized clay, 20 cm thick, was constructed on the
bedrock by the monument’s east side. On the stabilized
layer two stones were laid, within the first hour from
the layer construction.
The stones were in shape and in size similar to the
temple euthynteria blocks (Fig. 12 left). On top of the
stones, large blocks were placed (Fig. 12 rigth) for fur-
ther loading of the stabilized layer. The loading was
applied for over a year, gradually increasing, achiev-
ing at its peak, 90 kPa compressive stress at the layer
(value larger than the loading which the stabilized layer
will bear in the restored state of the monument); during
that time, periodic altitudes measurements were taken,
which showed that the stabilized layer settled, about,
1 mm.
The diminutive layer settlement, in combination
with observations made during the preparation of the
stabilized material and the laying of the stones (such
as they were no voids between the stones and the layer,
Fig. 12 left), led to the overall conclusion that stabi-
lized clay exhibited satisfactory behaviour in its test
implementation and it can be used in the restoration
work of the temple.
Figure 12. Left: The two stones placed on the stabilized clay
layer. Rigth: Loads upon the stabilized layer.
1119
6 THE RESTORATION OF THE
FOUNDATION
6.1 Consolidation of the subfoundation stones
and the subeuthynteria course
For the consolidation of the subfoundation stones
found in the north side of Apollo Epikourios temple,
the five scattered stones and the top stone of the
group inside the cavity were removed and structurally
restored; meaning the fragments of the stones were
connected using threaded titanium bars and white
cement paste, according to the technique developed
in the restoration work of the Acropolis monuments
(Zambas 1988, Vintzileou & Papadopoulos 2001).
Then, the stones were conserved and repositioned.
In the cases where the stones did not fully based on
bedrock, the void between the stones and the rock
was filled with stabilized clay. The stones were reposi-
tioned horizontally, with altitudes resulted from obser-
vations about the altitude variation of the proximate
bedrock surfaces.
The consolidation of the group of stones inside the
bedrock cavity was accomplished with the construc-
tion of a wall in the remaining empty cavity space.
The wall was built with stones from the local area and
stabilized clay as binding material.
For the consolidation of the subeuthynteria course
in the north-west corner of the monument, trenches
1.0–1.5 m wide were excavated across the north and
south limits of the course, both in its external and inter-
nal sides. It should be noted that for the excavation of
the west external trench a part of the artificial fill, con-
sisting of clay and stones of irregular shape and size,
had to be disturbed; however, the trench width was kept
the minimum necessary, and the removed stones were
replaced in nearby positions. The trenches extended
down to the bedrock, and underneath the course for a
width of 0.50 m.
In the trenches, walls were built to support and
retain the course (Fig. 13). The walls were built with
some of the stones removed from the ancient structure,
Figure 13. The consolidation of the subeuthynteria course.
Cross section parallel to the transverse axe of the temple.

Figure 14. The restored foundation in the north side of the
temple, from NW.
stones from the local area distinguished by indica-
tive sign, and stabilized clay as binding material. The
remained empty spaces of the trenches were filled with
clay and the rest removed stones.
6.2 Resetting of the euthynteria blocks
The resetting of the euthynteria blocks in the north side
of the monument followed the consolidation of the
subfoundation stones and the subeuthynteria course
and took place simultaneously with the construction
of the stabilized clay layer, which was interjected
between the blocks and the bedrock. All the blocks
were structurally restored and conserved before they
were repositioned.
First repositioned block was the eleventh from east
of the façade row (EU-N11).This block is based almost
entirely on bedrock and it is from the euthynteria part
which presented the minimum deformations. Thus, the
position and the altitude of the block before the inter-
vention were considered as the least deviated from
the initial ones; therefore, block EU-N11 was selected
as the key stone of the resetting. After block EU-
N11 the resetting continued towards the two ends of
the north side, simultaneously. All the blocks were
repositioned in horizontal positions, with top face alti-
tude 1130.119 m (value equal to the altitude of block
EU-N11).
A brief description of the blocks resetting procedure
is presented below: Initially, the block is placed on
wooden wedges in its four corners, fact which allows
microadjustments of the block position; then, after
the approximate finding of the block resetting posi-
tion a load is placed temporary on top of the block,
for securing its position. Afterwards, stabilized clay
is inserted in the gap between the block bottom face
and the bedrock, and it is compacted. Finally, the last
corrections of the block position are made, using topo-
graphical instruments for the position checking. The
next day the wedges and the load are removed and the
resetting continues to the adjacent block. Care is taken
1120
so that the joints of the stabilized layer are not in the
same plane with the vertical joints of the blocks.
For the restoration of the foundation in the north side
of the temple 45 euthynteria blocks were restored and
repositioned (Fig. 14). The restoration of the temple
north foundation was completed with the reconnection
of the euthynteria façade blocks using new titanium
clamps. The new clamps are designed to improve the
monument seismic behaviour, without inducing blocks
fractures (Papadopoulos 2007).
7 CONCLUSIONS
Although Apollo Epikourios temple is one of the
best preserved buildings of classical antiquity, it faces
many worrying structural problems, one of which is
foundation failure. In the north side of the temple
foundation deformations were caused, mainly, from
the existence of a clay layer beneath the foundation
top course (the euthynteria). New data, revealed in the
course of the dismantling (for restoration purposes) of
the temple north side and subsequent damages diag-
nosis further enforced the aforementioned assessment,
confirming the necessity of the large-scale interven-
tion. Consequently, the restoration of the monument
foundation involved removing the clay material and
repositioning it at a later stage, after stabilizing it with
lime and cement. Thus, a strong and durable base was
created for the resetting of the structurally restored and
conserved euthynteria that followed. The exact com-
position of the stabilized material was derived from an
experimental investigation and, prior to its implemen-
tation on the monument, its behaviour was checked
under real loading conditions, in a structure that sim-
ulated a small part of the foundation. The restoration
work in the north foundation of the temple included,
also, consolidation of the subfoundation stones and the
subeuthynteria course in the north-west corner, which
was carried out with the construction of supporting
and retaining walls, using stabilized clay and stones
from the local area.
The restoration of the temple north foundation was
carried out with respect to the ancient constructions,
and did not alter significantly the original structural
character of the monument. In order to attain the latter
objective, the restoration study was undertaken from
an approach based on the one hand, on an attempt to
understand the monument’s structural behaviour and
its repair needs, and on the other hand, on numerical
and experimental analyses. Furthermore, although the
study and its consequent intervention practices were
aiming at the restoration of a small part of the tem-
ple foundation, they can also be applied when the
intervention expands to the remaining parts of the
monument.
REFERENCES
Kolias, S. 2003. Stabilization of the clay layer in the foun-
dation of Apollo Epikourios temple. Athens: National
Technical University of Athens (in Greek).
Papadopoulos, K. & Karahalios, A. 2003. The material for
the restoration of the foundation in the north pteron of
Apollo Epikourios temple. Athens: Committee for the
Preservation of the Temple of Apollo Epikourios (in
Greek).
Papadopoulos, K. 2007. The restoration study of the con-
nections between the stone-blocks in the steps of the
temple of Apollo Epikourios. In P.B. Lourenço, P. Roca
C. Modena & S. Agrawal (eds), Structural analysis of his-
torical constructions; Proc. 5th intern. symp., New Delhi,
6–8 November 2006. New Delhi: Macmillan.
Papantonopoulos, C. 1995. Temple of Apollo Epikourios.
Structural restoration study. Athens: Committee for the
1121
Preservation of the Temple of Apollo Epikourios (in
Greek, with English summary).
Svolopoulos, D. 1995. Temple of Apollo Epikourios. Archi-
tectural study. Athens: Committee for the Preservation of
the Temple of Apollo Epikourios (in Greek, with English
summary).
Vintzileou, E. & Papadopoulos, K. 2001. Dowel action of tita-
nium bars connecting marble elements. In R. Eligehausen
(ed.), Connection between steel and concrete; Proc.
intern. symp., Stuttgart, 10–12 September 2001.Bagneux:
RILEM.
Zambas, C. 1988. Principals for the structural restoration of
the Acropolis monuments. In P. Marinos and G.C. Koukis
(eds), The engineering geology of ancient works, mon-
uments and historical sites; Proc. intern. symp., Athens,
19–12 September 1988. Rotterdam: Balkema.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Structural consolidation methods for the Temple of Santa Maria della

Consolazione in Todi (Perugia, Italy), damaged by landslides and
earthquakes

M. Mariani
Studio per le Ricerche Applicate – Arch. Ing. Massimo Mariani, Perugia, Italy

ABSTRACT: Between the years 1990 and 2001, restoration and strengthening operations were carried out on
the Temple in two distinct phases: firstly, in order to face the historical foundational difficulties of this edifice,
which had begun as far back as at the completion of its construction (spread roughly over a century, from 1508
to 1606); secondly, in order to repair the damages to its higher structure caused by the earthquakes in Umbria
and Marche, which had begun in 1997. After an ample description of the building, the geology of the Todi Hill
is defined, followed by the diagnostic and analytical phases of the project (1985–1987), the project itself (1987)
and the restoration and strengthening operations (1990). Subsequently, we describe the preliminary studies, the
project and the conservation and repair work carried out on the monument after it was damaged by earthquakes
beginning in 1997.

1 INTRODUCTION

All of the studies from the 19th century to date which

concern the architecture of Humanism, proclaim the
Temple of Santa Maria della Consolazione in Todi as
the exemplary image of the “ideal Church”, as it is a
definite expression of the solid and complex cultural
and religious content which characterized the architec-
tural ideals of that time. It represents these models as
it is an image of the transition from the Middle-Ages
to the Modern Ages, a period dense with values and
boundaries (Fig. 1).
It is a unique one-off building which, as Leon
Battista Alberti describes in his De re aedificatoria,
embodies all the ideal features of the “Temple”: a mon-
ument “which, on account of its purpose to serve and
be an expression of the divine, requires more than any
other ‘ingenuity, ability, diligence”’ so that “nothing
could possibly be imagined with a more ornate appear-
ance (...), so that the visitors are struck by amazement
and admiration at the sight of such worthy things, and Figure 1. Church of Santa Maria della Consolazione in Todi
can barely refrain from exclaiming: what we see here is (Perugia, Italy).
truly a place worthy of God”. These idealizations rep-
resent precisely the Consolazione in Todi in its strong the hours in the days and the four seasons cycle, are
connection to the “truth” and to the “justice” of the conjured up.
cosmic organization which are expressed, above all, in
the precise correspondence of the cross’s axes with the 2 HISTORY
four cardinal points, seen as spatial as well as temporal
cornerstones of the world, through which the directions History ascribes the Temple to Bramante, but
ordering the cosmos, the sun’s path, the passing of many researchers, not having found any documented

1123
reference about his presence, have come to the con-
clusion that it is the fruit of a worthy conception of
little-known architects and builders.
De Angelis D’Ossat, contrary to this theory, reports
that Bramante accepted the project commission and
carried it out, but, “being unable to go there person-
ally... sent Ventura Vitoni da Pistoia in his place”.
Todi tradition, as early as 1574, and as testified by
the Apostolic Visitor Pietro Camaiani when the works
on the Temple were not yet finished, also documents
that the Consolazione Temple “was designed by Bra-
mante”, and this conviction remained constant until
1872, when Adamo Rossi discovered the name of Cola
di Matteuccio da Caprarola in the construction records,
whereas he never found the name of Bramante during
his research.
Cola appears in 1499 next to Antonio da Sangallo
the Elder, working for Pope Alexander VI as a papal
fortress builder. In Todi, in the construction of the
Consolazione, he appears from 1508 in the records
as a magistro contractor, a construction conductor a
cotimo or coptimarius or murator, although from May
1509 until the end of his involvement in 1515, he is
called architettore. Cola was also contractor of the Figure 2. North-South section with identification of the
papal fortress in Civitacastellana designed by Antonio main vertical crack.
da Sangallo the Elder, who was most likely assisted by
Bramante as sottoarchitettore for Pope Alexander VI.
Reports from those times also tell of the collapse of • 1524/1563 – Filippo di Salvi da Meli, architect and
a cross vault in the Foligno Duomo, built by Cola as sculptor;
contractor, for which Bramante’s advice was sought • 1565 – Giovandomenico Berzugli da Carrara, archi-
first. Cola da Caprarola is therefore an architect- tect;
contractor, certainly connected to the Roman Sangallo • 1565 – Jacopo Barozzi da Vignola, architect;
and Bramante circle. • 1584 – Guglielmo Portoghese, architect;
This seems to confirm Jurgen Zanker’s theory, • 1584/1587 – Valentino Martelli, architect;
which vehemently rules out any intervention by • 1584/1594/1597 – Ippolito Scalza di Orvieto.
Bramante, assuming that “the Consolazione archi-
Even from the first decade of the 17th century, just
tecture is the result of various consecutive projects,
after its building was completed, up until today, the
which were drawn up or modified during construction,
stability of the Temple’s structure has often been put
depending on the opinion of the various architects who
to the test and has revealed its hardship in cracks and
followed each other.” And he concludes: “there is no
structural element collapses, which were mainly due to
single architect thanks to whom the Consolazione was
differential foundation subsiding or to seismic activity,
built, but there are several of them.”
as was the case during the most recent earthquake in
It is indeed documented that from 1508 – the year
1997 (Fig. 2).
in which the works began – onwards, several architects
In 1638, the tie-beam which closed the eastern apse
took part in the construction in a way that was anything
broke; the same tie-beam broke again another two
but marginal; and it could not have been otherwise
times in the following centuries, and in 1670 an inves-
because the construction was finished in 1606, that
tigation was carried out, which consisted of installing
is to say almost a century later. From the documents
topographic reference points with the intention of mea-
collected, here is the list of the architects who took part
suring the relative movement between the construction
in the construction of the Temple:
and the natural slope; this verification was necessary
because a part of the monument is built on settled
• 1508 – Cola da Caprarola, architect; ground built with the aim of creating a square facing
• 1509 – Gabriele di Giovanni da Como, construction the valley.
master; After subsequent interventions and continual
• 1515 – Giovandomenico da Pavia, architect; episodes causing instability leading to the complete
• 1518 – Baldassarre Peruzzi, architect; exposure of the foundations of the southern apse,
• 1516/1525 – Ambrogio da Milano, architect; which had already undergone under-foundation work

1124
in 1792, a semicircular gravity wall, designed by the
engineer Luigi Poletti, architect for the Papal State,
was built to the South towards the valley between 1836
and 1860, aimed at contrasting the thrust of the square’s
protuberance.
In 1926, following serious damage to the drum, the
eastern apse underwent consolidation work, consisting
mainly of creating an under-foundation to support the
southeastern pilaster.
In 1953 a Technical-Scientific Committee was set
up which had the task of examining the episodes of
instability which had caused a large crack to open up
along the median of the semi-cupola above the eastern
section, and the consequent damage to the external
walls.
The Committee had the task of suggesting pos-
sible remedies to finally put an end to the ongoing
dynamism.
The resulting consolidation work mainly consisted
of the injection of a cement-like mixture into the foun-
dations, with the intention of making them monolithic.
3 THE BUILDING
The church, which has a Greek cross plan, stands on
a structural system resting on four large pilasters at
the top of a square, on the sides of which the apses
are placed: three of these are polygonal and one, the
one in the north, is semicircular with its centre slightly
moved towards the geometric centre of the church.
The square formation, brought about by the
pilasters, is also found in the terrace which can be
reached by a stairway set in the shaft to the left of the
choir stalls.
The pilasters, by means of large arches, sustain the
drum and the cupola; the latter has a small lantern
above it, also covered by a cupola. Twenty double
embrasure windows, all surmounted by a tympanum,
open up between one pilaster and another. This same
scheme is repeated, with the same number of square
windows, beneath the impost of the apsidal cupola.
The drum contains a sequence of alternating windows
and niches.
The Church’s masonry consists of excellent qual-
ity calcareous stone originating from the Titignano
quarry, near Todi, and partly also from the demolition
of the medieval fortress nearby.
The Temple is set in a wide square, almost com-
pletely made with an embankment supported by a
gravity wall to the south, built, as already mentioned,
by the engineer Luigi Poletti between 1836 and 1860
(Fig. 3).
The building is about 50 m high and seen on a loca-
tion plan, it can be contained within a circumference
of a diameter of about 53 m. The apse walls are about
2 m thick, and the pilasters’ imprint covers a surface
1125
Figure 3. Southern gravity wall project (Luigi Poletti 1836).
of about 25 square meters. The apses and the pilasters
are built at different depths; the former, about 20 cm
wider on each side compared to the higher section, are
2 m below the ground level, whereas the latter are 4 m
below (without considering the under-foundation work
carried out on the north-eastern pilaster at a depth of
16.50 m, and on the eastern apse).
The weight of the structure is about 17,000 tons,
composed of 6250 tons for the columns and the pen-
dentives, 1870 tons for each polygonal apse, 2030 tons
for the altar’s circular apse, 1740 tons for the drum and
1370 tons for the large cupola and the smaller cupola.
The building transmits, on average, 0.95 MPa from
the pilasters and 0.43 MPa from the apse walls to the
foundation grounds.
The pilasters and the external walls are composed of
limestone block masonry held together by lime mortar
on the outside, and masonry with not regular filler on
the inside.
4 GEOLOGY OF THE TODI HILL
The geologic formation of the Todi hill is due to
the sedimentation of lacustrine Pliopleistocene ori-
gin in the ancient Tiber Basin, which extended to
the Umbrian Valley, the Tiber Valley and other Minor
Valleys. The hill, originating at the south-western
edge of this basin, is made up of distinct lithological
sequences, at the lower level, in two complexes:
• basic clayey complex: grey-blue silty-sandy clays,
locally marly, with an average slope of a few degrees
towards north-east;
• peak conglomerate complex: shingle and sand with
a slimy matrix and lenticular position.
On average the soil is plastic (Ip = 20% ÷ 30%) and
strongly over-consolidated.
These materials have irregular surfaces of tectonic
origin and deposits of thin layers of sand.
The piezometric level of the aquifer layer inside the
town as well as on the hillside, in a modified and
disrupted surface environment, is not always paral-
lel to the external soil profile. This variability is due
to the existence of soils with different positions and
different permeability levels: high permeability in the
detritus, in the sand levels and in the conglomerates
(K = 10−2 ÷ 10−3 cm s−1 ) and low permeability in the
silt (K = 10−4 cm sec−1 ) and, even more so, in the clays
(K = 10−5 ÷ 10−7 cm sec−1 ).
The undrained (cu ) and drained (c ) cohesion lies
within the following values:
This geologically young sedimentary composition,
which has always been influenced by morphogenetic
factors, is prone to continuous modifications due to
natural deterioration, constantly increased by damag-
ing anthropic contributions.
The consolidation interventions, financed by subsi-
dies deriving from the Law 545/87 and the Regional
Council Resolution 548/98, were carried out on all the
hill’s slopes subject to landslides threatening the town
boundaries, which have always been prone to upheaval,
nowadays still visible in the deficits and cracks in some
of its structural parts. The landslides are mainly due
to two predominant factors: the continuous incision
in the ditches and the alteration of the clays and silts
below the stratum. The incision in the ditches is due
to the dynamic action of the meteoric waters and the
chemical aggression of substances found in the sew-
ers which deteriorates the organic components of the
clays and silts. The continuous deepening of these inci-
sions on the hill extends to bordering areas, causing
the retrogressive loss towards the hill of the hillside
stability.
It was only in the 1980s that the reasons for the con-
stant deterioration of the clays and silts of the Todi hill
below the stratum and the lessening of their mechani-
cal capacities were revealed in laboratory experiments.
Research in the last twenty years has led to new knowl-
edge about these soils with regard to the stability of the
slopes and in particular to the swelling and alteration
processes of the surface layers.
It is precisely the Todi clay that was the sub-
ject of an experimental study programme undertaken
at the geotechnical laboratory of the Structural and
Geotechnical Engineering Department of La Sapienza
University in Rome (Calabresi, Esu, Pane, Scarpelli,
Rampello).
From the analysis performed on undisturbed soil
samples immersed in the same water as the stratum
on site, it appeared that the swelling of the material
was independent from the history of the applied loads,
which occurs “in more time than the consolidation pro-
cess, considering the same tension level”, and that the
inter-particle binding intensity in the Todi clay is not
very strong.
1126
It may thus be deduced that the presence of ver-
tical and sub-vertical discontinuities (cracks) in the
soil structure represents a dominant element for the
spreading of the swelling and for the lessening of its
mechanical capacities.
The swelling obtained in the laboratory is quite rep-
resentative of the actual swelling that occurs on site,
and it is accompanied by a considerable reduction of
the effective cohesion, mostly due to the increase in
water content.
Taking into account the fact that clay cohesion
(and silt cohesion) depends on particle cementing,
on electrostatic forces which they exchange amongst
each other individually, and on the binding between
atoms, one becomes even more convinced that, sur-
passing a certain distance limit between one particle
and another during the swelling of the composition, a
polarity inversion sets in and repulsive forces begin to
develop which cause, as already mentioned, a lessen-
ing of the cohesion. In summary, the persistence of the
over consolidated clays and silts in Todi in a “wet-dry”
situation due to the alternation of the stratum because
of season variations, causes the constant lessening of
their mechanical properties, and in particular of the
effective cohesion in the wet-dry area.
The Umbrians, the Etruscans and the Romans set-
tled in this beautiful but unsafe – because of possible
landslides and earthquakes – landscape, and with time
they enlarged the town, adapting it to the hill’s config-
uration and expanding it with terracing that not only
filled furrows and ditches, but also acted as a boundary
wall.
This settlement, as also happened on other hills of
sedimentary origin of the Tiber basin (Perugia, Mon-
tone, Ilci, Monte Castello di Vibio, and so on), was
certainly advantaged by the easy drawing of water from
the stratum which was not very deep, thanks to wells
and drainage shafts.
5 DIAGNOSTIC PHASE
An initial study and research phase aimed at examining
the Temple’s stability was carried out by the Turin Poly-
technic in the context of the agreement entered into
with the Umbrian Monuments and Fine Arts Office
No 1072 of 1985, rep. 2961.
It lasted over two years and included the carrying
out of topographic and geo-mechanical measurements,
drilling, laboratory tests, numerical analysis and the
setting up of instruments to measure the gradient of
the slope to check the static behaviour of the struc-
tural elements and their interaction with the foundation
ground.
The researchers who undertook the study were Prof.
Bruno Astori and Prof. Roberto Chiabrando for the
topographic part and Prof. Gian Paolo Giani for the
geotechnical part. Later on, in 1990, the consolida-
tion of the Consolazione Temple was planned as the
funds connected to the Special Act for the Todi Hill
and the Orvieto Rock were made available (Act No
545/87). On that occasion, further geognostic inves-
tigations were carried out which completed the ones
carried out by the Turin Polytechnic, and inclinometers
were installed on the square and just below the square’s
support wall, on which a lengthy monitoring was car-
ried out, the results of which were always connected
to the atmospheric precipitations occurring.
5.1 Geognostic investigations and geotechnical
laboratory tests (Turin Polytechnic)
Three geognostic drillings were carried out in addition
to those which had already been done in the inves-
tigations for the “Definitive consolidation of the Todi
Hill and Orvieto Rock” (Law 280/78) and to those pre-
liminary to the undersigned’s project for the Temple’s
consolidation.
The geotechnical laboratory investigations carried
out by the Turin Polytechnic consisted in: granulom-
etry, Atterberg limits, humidity and volume weight,
edometric tests, three-axial, isotropically stabilised
and undrained tests.
The geotechnical parameters obtained were:
• OCR consolidation degree;
• RR recompression ratio;
• CR pressure ratio;
• SR drain ratio;
• Ed load deformability module;
• Es drain deformability module;
• 
σvp maximum and effective pressure that has
strained the land in its history;

• σvo present effective vertical pressure;
• K compression module (load and drain);
• p average effective pressure.
Further investigations, preliminary to the planning
of the consolidation of the depth upheaval, in addi-
tion to the thorough research made by the Turin
Polytechnic, made it possible to identify a shifting
landslide with its detaching edge immediately above
the temple, involving Poletti’s wall which supported
the square obtained from an embankment, and the
entire structure.
The modelling of the upheaval, based on the inves-
tigations, apart from small differences, coincided with
the schematization of the past.
5.2 Topographic precision monitoring (Turin
Polytechnic 1985–1987)
It was precisely the analysis of the documents regard-
ing the Temple’s history from the 17th century to date,
1127
as proof of the recurrence of the instability episodes
concerning the elevation structures, the foundations
and southern side protuberance on which it partially
rests, which led the researchers to set up a programme
of experimental measurements, which in essence were:
• the monitoring of the movements of the structures
and the terrain of the foundation;
• the definition of the loads affecting the pilasters and
the apse walls;
• the variation in the opening of the existing cracks;
• the mechanical characterization of the materials
making up the structures and the terrain.
Static analyses were also carried using numerical
modelling, aimed at reconstructing the history of the
foundation’s terrain and at calculating the differential
subsiding and the rotations of the pilasters’ foundation
base, in order to be able to trace the reason behind the
drops measured. Tests were also carried out to check
the stability of the whole southern side of the Todi
Hill, which has always been subject to landslides due
to erosion at the foot of the Naia Torrent slope, at the
Arnada Torrent inlet.
The topographic survey carried out by the Turin
Polytechnic mainly concerned altrimetic monitoring,
with the creation of a wire-mesh installed around the
Church. Twelve points (which became fourteen in the
summer of 1987) were fixed to monitor the Church’s
possible differential movements in relation to the sur-
rounding slope, as well as the differential movements
of the various parts making up the structure Particu-
lar attention was given to the southern area, where a
benchmark was fixed near the support wall below the
Temple.
Another three benchmarks were placed on the
northern side (originally there was only one, but as
anomalies in the behaviour of the wall were noted,
another two benchmarks were added). All the external
points were linked to each other by means of a closed
polygon hinged to the same point where it converged
and the high precision geometric levelling closed ring
was bound, inside the Church.
Another fundamental operation concerned the mea-
suring of the pilasters’ drop variation using the four
brackets cemented from 1953 (for the same purpose),
placed at a height of approx. 12 m above the pilaster
base. A wire was fixed to each of these, with a 5 kg
weight at its extremity immersed in a basin full of oil
to reduce oscillations (Fig. 4).
The large crack on the drum and on the semi-
cupola of the eastern apse, which also cut through the
portal, was put under control by means of a milles-
imal crack measuring device. During the two years
of Research, seven measurement campaigns were car-
ried out in addition to the initial one (December
1985), which served to determine the initial reference
measurements.
Figure 4. Historic development of the drops.

5.3 Mechanical characterization of the building

structure with non-destructive measurements
on site (Turin Polytechnic – 1985)
The state of the vertical strain in the pilasters and in
the Church’s outside walls was registered in eighteen
measurements with flat jacks. The cuts on the build-
ing structure were made at a height of about 1.5 m
from the ground for those inside the church, and at a
height of about 2.0 m from the ground for those on the
outside walls. The deformability module stabilized on
average values between 22,000 MPa and 28,000 MPa.
Figure 4 contains the deformability module table and
the corresponding histogram (from the ISMES report
12/87).
The vertical strain measured with the flat jack tech-
nique was in accordance with the results of the load
testing.
5.4 Stress-deformations finite element analysis
(Turin Polytechnic – 1987)
The finite element method (FEM) was defined with the
aim of reproducing the tensional state of the Church’s
foundation terrains and of determining the subsiding in
the apse wall and the pilaster, as well as the rotations at
the pilasters’ foundation base. The three-dimensional,
100 m high, 105 m long FEM model, made up of 969
nodal points and 579 iso elements, formed a paral-
lelepiped which frames the foundation terrain, the
foundations and the semi-circular gravity wall. The
FEM simulation was carried out using the following
steps:
• loading and drainage of the overconsolidated ter-
rain, in accordance with the results of the edo-
metric tests, in order to reconstruct the history of
the tensional state before the construction of the
Church;
Figure 5. Formation of the upheaval and tension state of the
round behind the bulkhead.
• construction of the protuberance and of the grav-
ity wall, excavation of the foundation’s terrain and
construction of the foundation masonry;
• application of loads equalling the weight of the
Church to the pilasters’ foundations and to the apse
walls.
The comparison between the results of the finite ele-
ment analysis and the tension values measured in the
pilasters led to the identification of vertical subsiding
gradients with the same direction, having considered
the drop values existing in 1953, the drop variation
values between 1953 and 1985, and the drop variation
value definition between 1985 and 1987.
5.5 Finite element analysis for calculating the
thrusts on the bulkhead
The project for the consolidation of the Temple, carried
out with funding by the State through Law No 545 of
29th December 1987, required Finite Element Analy-
sis to identify the tensional state of the ground, once
the sliding area had been located, for the investigation
of strains on the future upheaval contrast structures
(Fig. 5).
The test results were compared with the strains
resulting from the calculation made with the “Janbu”
method, assuming that the layer is altered and in a

1128
movement without discontinuity. In the calculation,
the unstable terrain was modelled using flat elements
with 3 or 4 nodes with two leeway degrees for each
junction (translation according to Y and Z) and a level
state of deformation.
The mesh of the elements in the proximity of the
estimated position of the contrast bulkhead was inten-
tionally thickened to collect the values of the tensions
acting on it with greater precision. The model was
bound along the sliding area and to the area which
was in contact with the bulkhead of piles. The weight
which the Temple puts on the ground in upheaval was
not taken into account, because at that stage it had
already been decided to carry out an underpinning with
small diameter piles over the entire foundation area. A Figure 6. Planimetry of the operation on the upheaval. Bulk-
supposition, also used for calculating the micropiles, heads with large diameter piles and depth reinforcement with
was that the whole weight of the building be entirely small diameter piles.
transferred to the ground below the sliding area. The
comparison between the results of the finite element
analysis on the consolidated system and those of the
stability test has shown a substantial qualitative and
quantitative coherence of the two methods used.

6 CONSOLIDATION INTERVENTIONS

The Consolazione Temple was the subject-matter of

two designs carried out by the undersigned for consol-
idation interventions:
a) on the upheaval of the foundation, Law No 545 of
29th December 1987;
b) on the damage caused by the earthquakes which
occurred on 26th September 1997 and in the follow-
ing days: Umbrian Regional Council Resolution
No 548 of 25th September 1998.
Figure 7. Small diameter piles on the foundations and front
6.1 Consolidation of the upheaval of the foundation bulkheads with large diameter piles.

It consists of the following works:

• creating an under-foundation for the church using
232 piles with a small diameter (Ø 150 mm),
obtained by directly perforating the foundations of
the pilasters and the apses for a total of 5800 m;
• contrast the dynamism of the shifting land with two
bulkheads (Figs. 6, 7) composed of 257 piles with
a large diameter (Ø 600 mm): the first was carried
out under the Tiberina main road, and the second
behind Luigi Poletti’s support wall, which is below
the road, for a total of 6024 m. Both bulkheads were
made of piles of reinforced concrete piling brought
together and bound to one another by means of a
reinforced stiffening slab in order to obtain their
total cooperation.
This whole structural support system obtained in
this way was designed to support the ground thrusts
and those caused by the Temple up to a possible
undermining of the valley equal to a height of 7 meters
due to hydro-geological upheaval on the hillside,
consolidated with the subsidies of Law 545/87.
6.2 Consolidation of damages caused by
earthquakes starting on 26th September 1997
The intervention has planned the following works
(Figs 8 and 9):
• internal binding at the cupola base height by means
of a stainless steel band AISI 304 (300 × 15 mm)
which can be soldered, band “A”;
• external binding at the height of the drum base
by means of a stainless steel band AISI 304
(220 × 15 mm) which can be soldered, band “B”;
• internal binding at the height of the apses’ base
by means of a stainless steel band AISI 304
(300 × 15 mm) which can be soldered, band “C”;

1129

Figure 8. Stainless band disposition “A” – “B” – “C”.
Figure 9. Stainless band disposition “A” – “B” – “C”.
• saturation of the internal and external cracks with
hydraulic, sulphate-resistant mortar without alka-
line water-soluble salts and with a minimum com-
pression resistance (Rc ≥ 10 MPa).
6.3 Internal bindings
For the placement of the internal bindings, it was
necessary to mount scaffoldings inside the Temple
in order to be able to reach both of the altitudes
where the works were to take place. The internal
1130
Figure 10. Detail of the internal binding.
bindings were joined to the masonry by means of rein-
forced perforations (Ø = 35 mm) with stainless steel
threaded rods (Ø = 28 mm), the holes being filled with
high-viscosity epoxy resins (Fig. 10).
Four unthreading traction tests were made on the
anchorages until the load N = C L 5 kg cm−2 was
reached, in which:
C = hole circumference; L = hole length.
The longitudinal positioning of the holes on the
bands was performed with quincunx in order to avoid
twists during the nut screwing and draught phase.
Moreover a reinforcement stainless steel AISI 304
plate (100 × 100 mm).with a hole (Ø = 32 mm) was
soldered in correspondence to each hole.
All the soldering performed was Class I with a
continuous wire and it was finished off with acid to
eliminate the contact scorch marks.
The spaces between masonry and band were sealed
with sulphate-resistant hydraulic mortar sediment,
without alkaline water-soluble salts and with a min-
imal compression resistance Rc ≥ 10 MPa.
6.4 External binding
For the placement of the external bindings, it was not
necessary to mount scaffoldings; the area where the
works were to be performed is in fact accessible by
a staircase. All the materials were taken up to the
necessary altitude with the aid of a lift.
The external binding was carried out at the height
of the drum base and subjected to tension by four
temporary tensioners (Fig. 11) with a final force of
traction of at least F = 300 kg cm−2 . In order to aid the
band’s tension, “sliding pads” were used, made of neo-
prene parallelepipeds and PTIFE (TEFLON) plates,
vulcanized with neoprene (s = 1.5 mm) and stuck with
non-aggressive substances to the wall as well as to the
band, to help it to slide more easily during the traction
phase on the part of the tensioners. The tensioning of
the band was carried out in the summer, so that the

Figure 11. Tensioner apparatus for the external “B” band.
The sketch shows also the junction element.
band’s connection elements for the restitution of con-
tinuity at the tensioners’ level could be soldered when
the thermal dilatation had already occurred.
Even in this case, the spaces between the masonry
and the band, once positioned, were sealed with
1131
sulphate-resistant hydraulic mortar sediment, without
alkaline water-soluble salts and with a minimal com-
pression resistance Rc ≥ 10 MPa.
6.5 Conclusions
The structural consolidation methods thought up to
save this important example of Italian Renaissance
architecture, which has always been under threat from
landslides and earthquakes, are two-fold. The two
approaches used belong to scientific fields that are
close to each other but distinct: geo-technology and
seismology. Specific and evolved technology belong-
ing to the two sectors has been applied, especially
that which is in no way invasive, and that respects the
magnificent building in question.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

The diagnosis and arresting of settlement within Westminster Hall

in the Houses of Parliament, London

J.D. Miller
MA CEng MICE MIStructE Membre CNISF, Technical Director, Gifford

ABSTRACT: This paper considers the necessary pre-requisites for a successful conservation project in terms of
the profile of the consultancy team, its relationship to the client and curatorial representatives and the functioning
of the contract for execution of the works. It also draws out a number of examples of interest from the project at
Westminster Hall:
• Issues of document research and archiving
• The nature of the initial comprehensive diagnostic investigations
• The safe application of heavy ground engineering techniques in the context of materials of very high
archaeological significance
• The conservation philosophies applied to the work
• The conclusions of the post-project review.

1 INTRODUCTION abroad. Three hundred years later Richard II increased

the height of the walls and re-roofed the building with
Westminster Hall is an extraordinary building. It has thirteen huge hammer beam trusses, built by Hugh
witnessed social change as English monarchical power Herland. He also reworked the interior, making his
has evolved into British parliamentary democracy. It own mark in carvings in oak and stone with images
has been the seat of the Exchequer, the home of the that convey a divine connection with royal power.
highest courts and some of greatest trials in the land, it Westminster Hall was the seat of royal power. Coro-
survived a devastating fire of 1834 that destroyed The nations of monarchs from were held here, and the
Palace of Westminster and it became a classic exam- Kings presided over courts and hosted lavish banquets.
ple of modern roof restoration in the 1920’s. Its very Later, highly respected architects including William
size and location defies the designers of the present Kent and Sir John Soane added structures in and
to achieve in materials and engineering what the mas- around the Hall and as the function of the Hall changed
ters of building achieved 900 years ago. Now, locked these were subsequently removed (Figure 1).
within the security cordon of the Houses of Parliament,
it remains perhaps one of the country’s best-kept her-
itage secrets, on view only to those who queue to tour
the Palace of Westminster.
The desire to improve the visitor’s experience and
security arrangements at the Palace of Westminster
lead to the development of a new Visitor Reception
Building and this afforded the opportunity to address
a long term problem of settlement within Westminster
Hall.
The Hall was built for William Rufus, son of
William the Conqueror, in 1097–1099. He chose to
site the building close to where the ancient River
Tyburn joined the Thames in the former inter-tidal
zone. The walls of the Hall themselves are founded
on Terrace gravels at some 4 m depth. The Hall that he
erected was immense by medieval standards, measur- Figure 1. Westminster Hall looking South before Barry’s
ing 20.7 m × 73.2 m; it was a sign of power at home and transformation.

1133
 Of particular interest to us is the history of the
floor of the Hall. It appears that the original medieval
masons took steps to address the poor ground they
found when they laid the earth floor by capping the
ground with a layer of compacted clay. Richard II laid
a Purbeck stone floor.
Then between 1834–7 Robert Smirke, architect
for the British Museum date and Somerset House,
undertook extensive excavations and archaeological
recording of the ground and earlier floor levels as part
of his restoration of the Hall and the building of the
floor we see largely unchanged today. He built it using
some 500 Crosland HillYork flagstones each weighing
around 750 kg, supported, as we confirmed during this
project, on approximately two-courses of brick sleeper
walls around the edges of each flag. These walls were
supported on a 400 mm thick mass lime concrete slab.
Sir Charles Barry, with input from Pugin, won an
architectural competition for designs to rebuild the
Houses of Parliament. His vision was for Westminster
Hall to become the entrance to the Houses of Parlia-
ment. In 1852 he transformed the Hall by removing
the arched window from the south wall of the Hall to
create an opening to the new St Stephens porch. He
added a magnificent processional accent steps from
Westminster Hall to St Stephens Porch thus creating a
staged setting and continuous movement of people in
the previous static layout of the Hall. He constructed
the steps by supporting Hopton Wood stone treads on
a series of closely spaced brick walls and piers with
arched openings to allow access across the under the
stair space.
Westminster Hall is Grade I listed and part of the
UNESCO Westminster World Heritage Site.
2 BRIEF
The substantial structural movement apparent in
Smirke’s floor and Barry’s staircase are the subject of
this paper.
However, this project was undertaken in parallel
with a separate project to build a new Visitors’ Recep-
tion Building. Visitors were to, in future, enter the
Palace of Westminster through the North doors of
Westminster Hall, walk through the Hall, up the impos-
ing South Steps and into St Stephen’s porch and the
Houses of Parliament, a route that recognizes Barry’s
vision. Therefore, the condition of the Hall became
more pressing to the Parliamentary Works Services
Directorate and there was a rare opportunity to close
the Hall to the public and carry out repair works. Upon
visiting Westminster Hall, approximately 180 mm set-
tlement was immediately evident in the lower flight of
the South steps and in the floor close to the steps.
The brief for this work was to diagnose the cause of
the settlement to the floor and steps, consider options
for repair and recommend treatment for the problem.
1134
An OJEU notice was published by the Parliamentary
Works Services Directorate towards the end of 2004
and Gifford were appointed to carry out the commis-
sion, commencing in January 2005. The brief invited
best conservation practice because it required diagno-
sis of the cause and options to be presented, not just
a solution. The brief also included new railings to the
steps and a new plant space beneath the steps but these
aspects are not discussed in this paper.
3 RESEARCH
From the outset, considerable time was spent by
the project engineer and archaeologist in the early
stages of the project researching the archives with the
Palace Archivist, and the Institution of Civil Engi-
neers, amongst other records. These provided initial
background knowledge of the site and history of
Westminster Hall and then, as the project progressed,
the archaeological team was able to search for specific
information to inform, interpret and explain findings
from the site work. So the project gradually added to
the body of information about the site. Archaeological
research examined the site from times when inter-tidal
sediments were deposited, research continued through
the history of the construction methods and materials
to the social and political use of the Hall, all of which
were pertinent when piecing together the value and
significance of the Hall and interpreting geotechnical
data.
Research also yielded invaluable details of set-
tlement monitoring work from the 1940’s, drawings
showing some construction details of the nineteenth
century steps and proposed repairs in the 1960’s. As
consultants we were in the rare position of having
movement monitoring of sorts for a period of some 60
years, about half the life of the present steps. Amongst
other information, the early research gave four sets of
data on floor settlement over a total period of around
60 years. It was clear from this that the movement of
the steps showed no signs of stopping, with an addi-
tional 16 mm of settlement in the centre of the floor
being recorded between 1992 and 2004 (Figure 2).
Mid height of Lower Steps
3.700
3.650
3.600
Mar-89
3.550
Oct-89
Apr-91
Mar-92
Jan-04
3.500
16mm settlement
3.450
between 1992 and
0.000 5.000 10.000 15.000 20.000 25.000
Distance fromeastedge of hall (m)
Figure 2. Deflected profile of the steps from East to West.
This historical monitoring was assessed and suggested the flags were carefully cut free using hand tools and
historical and ongoing settlement of approximately where necessary, after careful reflection a small disc
1 mm a year. cutter. It was found important to completely free the
joints otherwise spalling started to occur at the edge of
the flag. A single stone flag was successfully freed and
4 INVESTIGATIONS
relaid in the middle of the area of settlement, reveal-
ing Smirke’s construction of dwarf sleeper walls and
However, although one or two boreholes were avail-
oversite concrete slab below (Figure 3).
able for areas close by, very little engineering data
Once it had been established that the flags could
or calibrated properties on the ground conditions was
safely be lifted, ten flag stones were selected at loca-
uncovered by research for the locations beneath the
tions across the Hall, concentrating on transects across
steps and floor. It was clear that to both correctly diag-
the areas of settlement but with one chosen at the
nose the movement and to support any discussions
North end, remote from the steps. At each flag loca-
with the curatorial bodies with regard to intervention,
tion two cores were taken through the 400 mm concrete
a major ground investigation needed to be carried out,
oversite slab to enable Dynamic Probe Penetrometer
and that the success of this would be key to the quality
testing and allow the retrieval of MOSTAP thin-walled
of decisions made thereafter.
samples for visual analysis and archaeological record-
It was also important to gain as much as infor-
ing. The Penetrometer was mounted on a portable rig
mation as we could through non destructive testing,
specially made for the project and surcharged with 4
so a two stage approach was developed; firstly com-
tonnes of dismountable kentledge, which permitted
missioning a suite of geophysical tests to determine
penetrometer readings to be taken down to a depth
the floor construction and as much about the ground
of around 7 m, at which point the kentledge lifted on
below as we could, across the whole of Westminster
the underlying material, presumed to be stiff clays and
Hall and secondly targeted lifting of flags to confirm
gravels. Ground water level was monitored and found
the construction and intrusive geotechnical boreholes.
to vary by just 55 mm compared with tidal variation
A suite of non destructive tests comprising geophys-
or between 4 and 6 m. Samples of the stone flags were
ical radar, microgravity and resistivity was commis-
taken for petrographical analysis to identify the quarry
sioned. Whilst the specification and execution of such
from which the flags had been won. Interestingly the
non-intrusive investigations was an essential part of
the initial work in the context of the Hall, at the time
they were interpreted they were of marginal value in
defining sub-floor and deeper features of stratigraphy.
With developments in processing of geophysical test-
ing there may still be more information to be gleaned
from these results.
Faced with the value of the historic fabric within the
Hall it quickly became apparent that any work, includ-
ing investigations, would require significant intrusion
into the building. Thus the design of such investiga-
tions would require considerable thought. To make
an appropriate diagnosis of the causes of settlement
and to design an appropriate solution we needed t
a robust theorem as to the cause of settlement, as
there was much speculation based on fragments of
evidence. It was essential to have a comprehensive
reliable geotechnical data for diagnosis.
Fundamental to this was the assumption that data
could actually be obtained. Any investigation would
require the successful lifting and relaying of floor flags
within the Hall, something that no-one working on the
staff in the Palace had witnessed before. The 1.6 m
square flags have tight or minimal joints and there
were concerns that lifting individual slabs might not
be achieved without major dismantling from the edges
of the Hall, or that flags might be dropped, cracked or
not be re-laid satisfactorily Equipment was therefore
assembled for a trial lift one weekend in the pres-
ence of the Conservation Architect. The joints between Figure 3. Trial lifting of the stone flags.

1135

Figure 4. Dynamic Probe and MOSTAP sampling rig.
stone masons’ visual identification of the stone source
was confirmed by the laboratory testing, so it is always
worth asking Samples of the concrete oversite slab
were also taken and tested for strength and composi-
tion. Work was carried out at weekends through late
May and June 2005 whilst the Houses were not sitting,
at a rate of one or two locations per day (Figure 4).
Under the steps it was observed that in places the
brick support had settled and so the steps were in
effect hanging. Beneath the lower flights of the steps
headroom was very restricted and it was not possi-
ble to undertake the dynamic probing and thin walled
sampling. Instead we examined some pre-existing
excavations in the concrete slab. These revealed infor-
mation about the steps foundations and some stone
fragments of archaeological interest. We also carried
out PANDA tests, which in essence con sisted of a
man striking a probe into the ground and measuring
physical resistance.
Archaeological attendance on the investigations
yielded some exciting finds in both building recording
and examination of the recovered cores. The lifting of
slabs showed evidence of the trackway support of the
mobile access scaffolding used by Sir Frank Baines in
the repairs of the roof in the 1920’s, with a fragment
of newspaper dating from 1922 being found. The soil
cores showed evidence of a rammed chalk floor in the
Hall, the stratification giving some indication that this
had been built up over centuries, possibly as previous
settlement had taken place.
5 DIAGNOSIS OF SETTLEMENT
The quality of data obtained from this investigation
was consistently high, allowing the classification of
soils and the soil strength and stiffness to be measured
directly in-situ with the probe and correlated through
the testing shear strength of recovered soil samples
1136
using established conversions with design parame-
ters being determined in-situ and correlated to shear
strength using established conversions. The samples
were examined by geotechnical engineers and archae-
ologists together, collaborating to enhance their under-
standing and produce a single interpretation of the
ground: there is only one reality, not a ‘geotechnical’
interpretation and an ‘archaeological’ interpretation.
The investigation revealed that overlying the London
Clay formation is a layer of Terrace Gravels, then a suc-
cession of soft and loose alluvial deposits, then made
ground that formed the dry land for Westminster Hall
and then a sequence of earth floors.
The results showed that material at a depth of
between roughly 1 m and 3 m below the floor at the
base of the steps comprised very weak alluvial clays
with some organic content and varying in thickness. It
is this local area of weak clay alluvium that is gradually
compressing, and causing settlement, the problem is
exacerbated by the additional load due to a large stone
candelabra support structure of considerable weight.
Further to the North, this material became firmer,
giving way to terrace silty sands and sandy gravels
(Figure 5).
The diagnosis was, therefore, that the floor and steps
were likely to be settling because of the weak and
organic materials present locally below, with constant
internal shearing of the soil resulting in continuous
movement.
6 DESIGN APPROACH AND CONSTRAINTS
Thus it had been established that the basis of the prob-
lem was essentially a ground engineering issue with
possibly very significant archaeological impact, and so
strong representation was made by specialists within
the company in both these disciplines.
Alongside the conservation structural engineer and
project director these formed a tight group of four indi-
viduals who remained close to the project throughout.
The project director also adopted the separate role of
lead consultant. The internal team structure for the
project was a key factor in delivering what we believe
to be a highly successful project.
Once the cause of settlement was understood we
held team workshops to brainstorm different options
for remedial solutions. This approach meant that
we were working as one multi-disciplinary team.
Each discipline, structural engineer, archaeologist and
geotechnical engineer had a good understanding of the
problem, the brief and the site and an appreciation of
the other disciplines, each was able to offer possible
solutions and assess the advantages and disadvantage
of schemes. This offered an integrated approach that
was very different from three specialists each work-
ing to a limited brief or a solution produced by one
discipline to be commented on by the others.
Figure 5. Geotechnical section showing layer of very weak alluvial clays.
The key issues and priorities of each aspect
project discipline were captured, including operational
requirements peculiar to the Houses of Parliament
and noted project constraints against which we could
assess the suitability of possible solutions.
7 OPTIONS AND DESIGN OF PERMANENT
WORKS
The full Gifford project team met to consider the
data and final options. Any option appraisal for con-
servation works should start with the option and
implications of doing nothing; of leaving the build-
ing as found. Although the brief had indicated that
works were anticipated, this was the assumed starting
point for the scheme to develop the correct conserva-
tion approach. We also needed to assess the necessary
extent of the repair works. We considered that it was
not necessary to carry out major intervention to all
flights of steps, just the lower flight, where settlement
was worse, and we selected an area where settlement
of the floor was greater than 20 mm, in both cases our
intent was to minimise the interventions.
Different options were considered for the correction
of the steps and for the floor, but these can be broadly
categorized in to the following approaches each with
their advantages and disadvantages:
1. Do nothing and monitor further.
It was evident that the movement of steps and
floor was continuing and not showing signs of
abating and that further settlement would exacer-
bate uneven floor surfaces (a real concern given
the future increase in footfall), and also result in
an increasing risk of damage to the elements of
the building fabric, such as the candelabra support
structure and railings. ‘Do nothing’ can only be
judged to be sound conservation practice if it does
1137
not make the situation worse. Although there would
be no impact on the below ground archaeological
remains, the movement would not be arrested and
the brief would not be met.
2. Dismantle and re-level the lower steps and floor
without any foundation strengthening, in effect
correcting the superstructure or symptoms with-
out addressing the substructure and cause. This
approach has the advantage of appearing to solve
the problem by lifting and packing stones to create
a level finish with out interfering with the below
ground archaeological potential. If the movement
continues at between 1–1.5 mm a year, the flags
could be lifted say every 70 years and an acceptable
level finish maintained. Indeed, in 70 years time
new technologies could make other approaches
viable. However, it does not address the mecha-
nism of settlement. In effect this is the approach
that appears to have been taken throughout the life
of the Hall to date; there is stratigraphic evidence
of the earth floors being re-laid and documentary
evidence that a number of stone floors have been
re-laid, presumably because they were no longer
satisfactory. In the past, however, there has not been
the option to address the underlying cause of settle-
ment and every time the flags are lifted and re-laid,
there is a risk of damage.
3. Dismantle and re-level the lower steps and floor,
and strengthen the ground beneath the floor and
steps to permanently arrest settlement. Within
this approach a number of different underpin-
ning or ground strengthening options were consid-
ered. Under the steps these included mini piling
or trench underpinning. Under the floor ground
strengthening by compaction grouting or soil frac-
ture grouting were considered. These solutions
addressed the mechanism of movement but clearly
had more impact on the archaeological remains, in

Figure 6. Rig for insertion of compaction grouting tubes.
line with the outlined constraints above, the mit-
igation strategy ensured that these impacts were
limited, observable and measurable.
The chosen approach, presented in the final sketch
plan submitted to the Client, was the third of these.
The lower flight of steps and brick wall supports was
dismantled down to over site slab level and mini piles
installed with ground beams to create solid founda-
tions to the steps. Under the floor initially it was
proposed to strengthen the ground using dry soil mix-
ing whereby a dry lime-cement mix is mixed into
the weak layer of alluvial material to stiffen it. How-
ever, this presented practical problems and would have
damaged the archaeological potential in this stratum.
The final design strengthened the deep ground strata
below the floor by carefully injecting columns of stiff
grout under pressure in a controlled method to pro-
vide lateral compaction to the deep strata, known as
compaction grouting (Figure 6). The archaeological
remains, located principally in the top 1–3 m of mate-
rial were not extensively damaged by injected grout as
it was injected here at hydrostatic pressure, in effect
creating piles passing through the layer. So the impact
on the archaeological potential was limited and known.
The depth and pressure of the grout columns were care-
fully controlled to reduce the increase in stiffness of the
ground adjacent to the unstrengthened area of floor.
1138
In addition, the oversite slab was strengthened.
Under the steps, new reinforced ground beams above
the piles were cast within the depth of the existing
oversite slab, reinforced with stainless steel bars to
give a design life of nominally 500 years (comparable
with cathedrals) to support the brick crosswalls. The
steps were rebuilt on new crosswalls. The oversite slab
beneath the floor flags and dwarf walls was reinforced
with a grid of stainless steel bars, in effect it was post
reinforced to spread the load to the point supports at
the grouting locations The area of strengthened floor
was dowelled into the area of unstrengthened floor to
ensure continuity across the floor.
8 IMPLEMENTATION AND DESIGN
DEVELOPMENT
8.1 Protection to the floor
The Hall was closed to the public in January 2006 and
site work commenced with the protection of the flags
by covering with polythene throughout the Hall and
then laying 75 mm thick close-boarded timber protec-
tion overlaid with plywood boards to form a trackway
along the centre of the building from the North door off
New Palace Yard. The weight of plant and the stacking
of storage materials was restricted to avoid overload-
ing the flags, which were only supported around their
perimeter by dwarf walls.
8.2 Lifting the flags
Dismantling of stonework was assisted by a suc-
tion lifting device slung from of a 7 tonne capacity
tele-handler. One of conditions of planning approval
was the safe, protected storage of the stonework
once removed from site. Flags were carefully labelled
to an agreed method and taken to a bonded ware-
house in East London, where they were held until
reconstruction commenced.
8.3 Archaeology
Another planning condition, and an integral part of
the solution was the implementation archaeological
mitigation strategy. Two test pits were identified and
discussed in detail with the ConservationArchitect and
English Heritage. One measured 2.5 m × 3.5 m by 2 m
deep, and was located close to the suspected loca-
tion of a medieval hearths for the great fires, lit to
heat the Hall. The other was sited close to the West-
ern candelabra pillar, at a location where fragments
of medieval masonry had been found during earlier
inspections of the sub-step void. These excavations
were programmed to be excavated in advance of main
groundwork operations, whilst dismantling was still

Figure 7. View towards South end of the Hall during
archaeological mitigation work.
taking place elsewhere around the Hall, and so did not
result in delay to the overall programme (Figure 7).
The pits were excavated as part of the main contract
work, with the archaeological field work being under-
taken as a named subcontractor. The first pit revealed
detail on the progression of nearly a thousand years
of rammed chalk floors, but no evidence was found of
the hearths. However, the pit by the candelabra support
revealed a succession of masonry fragments identified
as parts of the King’s Table that stood at the end of the
Hall in medieval times. The find was of great signif-
icance and was widely reported in the national press
and archaeological journals.
The excavation for ground beams supporting the
lower landing of the steps exposed a small area of
earlier flooring. This was recorded and the ground
beam redesigned locally to step up and over the fea-
ture, which was covered with a layer of inert sand and
so preserved in-situ.
8.4 Concrete slab
The concrete oversite slab was a very weak lime based
concrete and the lack of binder meant that when the
concrete was cored loose aggregate became jammed
1139
in the drill, slowing the progress of drilling. Once this
became evident, an early review of the design of the
strengthening of the slab was undertaken and as a con-
sequence, instead of drilling and coring stainless steel
bars and grouting them in place the method of strength-
ening the concrete was changed. Instead slots were
chased in to the slab and bars laid into the slots and
grouted in place.
The intention was to keep as much of the sleeper
walls as possible, however sections needed to be cut
out in order to insert the reinforcement bars. The walls
were rebuilt in concrete for speed and ease but also to
clearly differentiate the modern work.
8.5 Relaying of the stonework
Only the lower flight of step were dismantled and
relevelled so the treads had to be relaid to match the
deflected shape of the remaining steps and then relev-
elled across the landing to a horizontal edge at the
top of the lower landing, This presented difficulties as
the stones needed to be laid with falls in two direc-
tions. Consultation with the stone mason and the use
of lines to gain approval of levels, ensured that this was
successfully achieved. The lower flight was then kept
level. When stones were cracked a number of repair
methods including doweling stone inserts, repointing
of cracks with lime and in places repairing cracks with
epoxy resin, and piecing in carefully matched stones
produced a good result.
Overall a partnership approach to solving design
and implementation issues was adopted and this
resulted in good progress on site and completion on
time.
9 POST-PROJECT REVIEW
The project offered a rare privilege to work in one of
the most important and stunningly beautiful historic
buildings in the world. The work was completed on
time, on budget, with no damage to the historic fab-
ric, it looks fabulous and met with the approval of
English Heritage and the Client. Like any successful
project there are always a number of indefinable rea-
sons why things worked but we can also learn a number
of lessons that are transferable to similar projects:
1. A good geotechnical investigation is essential so
that decisions can be based on facts.
2. Technical decisions were influenced by client
requirements and good conservation principles. We
did not always give way to client demands but main-
tained best practice which ultimately fulfilled the
client’s aspirations.
3. There is sometimes the need for brave, but not reck-
less, decisions. For example to lift a significant
 number of flagstones to undertake geotechnical
investigation.
4. A good brief is very helpful, in this case requiring
cause, options and solution to be presented.
5. The design team worked as a tripartite multi-
disciplinary team. Each discipline shared in the
aspirations of the project goals and there was a
strong desire to understand and to compromise but
also to stand firm when required. There was mutual
respect across the team and healthy argument.
6. During the design process there were a number of
previews, reviews and combined site visits. Com-
munication was face to face not through emails and
minutes.
7. There was realistic fee for the work.
8. The subcontractors were very good and their ideas
fed into design development. Ultimately, however
tight the specification and supervision, the qual-
ity of the finished stonework is done to the master
mason executing the works.
10 CONCLUSIONS
The design and construction team have repaired a
very long term settlement problem in an historic
1140
environment using carefully chosen engineering tech-
nology and employing the least intrusive methods
possible, but nevertheless methods that are of our
time. The interventions were based on a very good
understanding of the causes of damage. We have under-
taken extensive recording of structures and have taken
the opportunity to gain a greater understanding of
Westminster Hall than existed before the project.
Client :
Parliamentary Works Services Directorate
Lead Consultant, Structural Conservation Engineer,
Archaeological Consultant:
Gifford
Contractor:
Verry Construction Limited
Stonework Subcontractor: Stonewest Limited
Ground Engineering Subcontractor: Keller Founda-
tions Limited
Archaeological Contractor: Museum of London
Archaeological Service.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Istanbul – Fatih, Millet Library / Feyzullah Efendi Madrasah restoration

R. Ozakin & A. Erdem

Yildiz Technical University, Istanbul, Turkey

ABSTRACT: The Feyzullah Efendi Madrasah, built in Fatih, Istanbul in 1700, is a small complex with a
library, school, fountain for ablutions and two other fountains. In the 1999 earthquake, the madrasah was severely
damaged and cracks appeared in its foundation, main walls and domes. The restoration of the building was carried
out between 2000 and 2007, the ground under the building was firmed up, foundations, domes and walls were
consolidated, the cracks in the walls and domes were stiffened using an injection method. This complex was
used first as a madrasah / faculty and then as a library. In this paper, the damage caused by the earthquake and
the problems of the new function given to the building is investigated. As for the restoration, the repair methods
to be applied were examined while estimating to what extent the historical value of the building was preserved.

1 INTRODUCTION librarian Ali Emiri’s book collection that was mostly

The Feyzullah Efendi Madrasah in the Sofular quar-
ter of Istanbul’s Fatih District is at the intersection of
Macar Kardeşler Street and Feyzi Pasha Street (Fig-
ure 1). In 1700 it was built as a Daru’l Hadis / Hadith
Sciences Faculty by Şeyhülislam Seyyid Feyzullah
Efendi, the teacher of Sultan Mustafa II. This small
complex was made up of the madrasah, library, school,
fountain for ablutions and two other fountains and the
architect of the it is unknown. The school was pulled
down during town planning works in 1912 (Ahunbay,
1994).
The Feyzullah Efendi Medrese was converted into
a library with the donation in 1916 of the historian and
composed of handwritten manuscripts and began to
provide services as the “Public Library.” In 1993, it
gained the identity of a research library. The library
was enriched over time and now there are nearly
70,000 Ottoman language books, both handwritten
and printed. The only handwritten copy of the book
“Divan-ı Lügat’üt Türk,” the oldest known Turkish
dictionary, by Kaşgarlı Mahmut is held in this library.
The complex was registered as “a cultural property
that has to be protected” in 1995 and in 2001 was
taken under the category “No. 1 Group Old Work”.
Today it continues its function of being a library under
the administration of the Culture Ministry, Libraries
General Directorate.

2 ARCHITECTURAL CHARACTERISTICS

The Feyzullah Efendi Complex consists of two sepa-

Figure 1. Millet Library, Macar Kardeşler facade of the
main building, 2007.
rate buildings on the sides of a rectangular courtyard
(Figure 2). The main building borders the northeast
side of the courtyard and the madrasah building bor-
ders the southeast and southwest sides of the courtyard.
The main building is separated into education and wor-
ship functions and the madrasah rooms are assigned
as student accommodation.
The main building that sits on a low basement con-
sists of a classroom/small mosque and library that are
located on both sides of the entrance like an iwan (it
is a half open space that the courtyard side is open)
in the middle (Figures 3 and 4). On the open side of
the rectangular iwan that provides the connection with

1141
Figure 2. A general view of the Millet Library (Archive
of the Protection Council of the Natural and Cultural
Properties).
Figure 3. Entrance facade of the main building, 2007.
Figure 4. Millet Library, plan (Ahunbay, 1994).
1142
Figure 5. A view towards the “mihrap” of the mosque, main
building, 2007.
the courtyard is a staircase with 11 steps. Together
with the classroom / mosque and the library, in the
same dimensions there were square planned spaces
(Figure 5). These spaces are covered with domes each
with pendentives 7.5 meters in width and 1.40 meters
in height. As for the entrance / iwan, there is a compli-
cated covering made up of domes and different vaults
that sits on six columns and the walls. Above of the
staircase is covered with a cross vault, in front of the
classroom / mosque and library with the mirror vaults
and the parts that overlook the courtyard are covered
with domes with pendentives (Figure 6). The covering
of the north part of the iwan that is open to the street
has three windows and is organized as a dome in the
center and two cross vaults on the sides. In the center of
the iwan is a conical dome with a stalactite (Ahunbay,
1994), (Figure 7).
The madrasah building with its “L” shaped plan is
made up of 10 square medrese rooms and in front of the
rooms is a half-open portico/revak section (Figure 8).
In the façade of the madrasah complex that faces
Feyzullah Pasha Street, there are the entrance door into
the courtyard of the complex and two fountains. There
is an inscription above the entrance door to the court-
yard with a segmental arch and dome. Today, because
the level of Feyzullah Pasha Street has been raised,
the fountains and the courtyard’s entrance door have

Figure 6. A view from the “eyvan”, main building, 2007.
Figure 7. In the center of the “eyvan”, a dome with stalactite,
2007.
remained in a hole. In the courtyard, there is hexag-
onal fountain for ablutions / şadırvan with a marble
reservoir with a wooden dome supported by six mar-
ble columns and a well made from limestone/küfeki.
The facades of the buildings that make up the com-
plex are dominated by almaşık wall mesh that is made
of rough stone and brick in turns but in some places
the cut stone wall mesh can be seen. The almaşık
1143
Figure 8. A view of the courtyard, “sadirvan” and madrasah,
2007.
walls of the main building were built with painstak-
ing workmanship according to the almaşık walls of
the madrasah building. The brick domes are covered
with lead sheets.
3 REPAIRS THAT THE COMPLEX HAVE
UNDERGONE
The complex was affected by the earthquakes, fires
and town planning works that Istanbul has experi-
enced since the 18th century (especially the 1894
earthquake) and therefore, it has been repaired at var-
ious times. Certain information about these repairs
could not be obtained. However, in 1911, it is said
that the water system was repaired and, in 1913, the
lead sheets were replaced. The school building found
inside the complex was pulled down in 1912. Demol-
ishing the madrasah building that was in ruins in 1914
was considered (Kütükoğlu, 2000). However, through
the efforts of the wife of the French ambassador, who
was a member of the Commission for the Protection
of Ancient Works, this proposal was abandoned. Fol-
lowing the librarian Ali Emiri donating his collection,
the madrasah complex was repaired in 1916 and turned
into a library (Ahunbay, 1994, Kütükoğlu, 2000). After
this date there are no records of any further restoration
work. However, in 1983 some basic repairs were made
and the courtyard entrance that had been closed for
years was again made functional (Ahunbay, 1994).
On 17 August 1999 during the Marmara earthquake
the complex suffered significant damage and became
unusable. This complex that was providing services
under the name of the Millet Library and was attached
to the Culture Ministry was closed and repair work
was begun in 2000.

Figure 9. The dome crack of the main building, before
restoration (Archive of the Istanbul Survey and Monuments
Directorate).
4 DAMAGE/PROBLEMS CREATED IN THE
MADRASAH AFTER THE 1999
EARTHQUAKE
The Marmara earthquake resulted in significant dam-
age to the building. In both structures consisting of
masonry walls, a series of cracks and deformations in
the surface materials were seen. Removing the plas-
ter from the surface of the buildings, revealed deep
cracks in the arches, domes and in the bearer walls,
particularly in the madrasah building, this damage was
determined to negatively affect the support system, and
the bearer walls had moved from their plane (Akcan,
2001), (Figure 9). Also it was determined that these
cracks in the walls and domes might possibly extend
as the foundations (Akcan, 2001).
In addition to the earthquake damage, there were
areas of discoloration on the marble surfaces resulting
from the effects of air pollution and the salts that may
or may not be soluble. The paintings within the domes
were damaged by exposure to moisture. repair work
was begun in 2000.
5 RESTORATION METHODS
Under the supervision of the Culture Ministry’s
Istanbul Survey and Monuments Directorate, the Fatih
Millet Library repair was carried out by the Ak-Alay
Construction Company, beginning with the madrasah
building in December 2000. The restoration work
begun in 2000 was completed in 2007. It was aimed to
re-open the complex for service with all kinds equip-
ment such as heat, moisture, security, fire, etc. in 2008
as the “Research and Specialization Library.”
5.1 Problems that appeared during restoration
In the consolidation project it was envisioned that first
the ground would be firmed and the repair started from
1144
Figure 10. The consolidation method of bar-mat on the
dome, main building, restoration stage.
the madrasah building. The foundations, walls, domes
and portico arches of the badly damaged building were
strengthened with metal construction reinforcement
and the foundations were converted into a raft foun-
dation system (Figure 10). However, the dimensions
and places of the material that the metal construction
would create and the relations with the main walls
were not clarified and the behavior of this system
in future earthquakes was not investigated (Ahunbay,
2001). The plaster at the repair stage while being rasps,
the cracks which occurred in the dome and walls were
investigated; however, it was not investigated whether
or not the historical value of the plaster would carry
them. In addition the wooden posts that carried the
temporary roof constructed to protect the building
were seated without taking into account any measures
for cracks (Ahunbay, 2001) and this was one reason
why the damage to the building increased.
In 2002 the repair of the madrasah building was
mostly finished; however, the fine construction works
and the floor coverings were yet to be completed.
As for the main building only the firming of the
soil had been carried out. The Protection Commission
requested a statement outlining whether it was pos-
sible to strengthen and reinforce the building using
traditional methods and materials, apart from the
metal reinforcement method suggested in the project.
Therefore, a group of restoration experts from Mimar
Sinan University examined the building and prepared a
report which stated that“the main building has charac-
teristics that differ from those of the madrasah building
such as architectural elements, height, the place sunk
into the floor, the delicacy coefficient (the building
height that is on the short side of the plan of the build-
ing). Also, because of the large number of decorations
in the main building, it must be repaired with a different
method from that used in the madrasah. Determining
what is necessary in what dimensions for the metal
reinforcement, there must be a scientific investigation
into whether or not the characteristics of the binding
mortar would be lost.” Following the report, a three-
dimensional finite-element model was created by the
Figure 11. The application of the raft foundation for
madrasah.
University team. Static and dynamic analysis of the
F-E model showed that undesirable tensile stresses
were happened in the building and that the tensile limit
values were exceeded. Since the applications could not
be carried out to strengthen and stiffen the building
with special construction and materials it was noted
that the use of steel material as reinforcement was
appropriate (Eriç, 2002). However despite this rec-
ommendation, the Protection Commission reached the
conclusion that the main building should be repaired
using traditional methods.
5.2 Work done on the levels of restoration
The restoration work as noted above began first on the
madrasah part. During the restoration work the soil was
firmed up with the injection method cement and the
foundations were consolidated and the cracks present
in the supporting walls and domes were repaired in
the main building and the madrasah. The façade was
cleaned, the lead roof covering was replaced and the
damage to the inner space decorations, including the
paintings and the golden leaf, was repaired.
Foundation strengthening: The foundations of
the madrasah building and the main building were
strengthened using stainless steel. The floors of the
madrasah rooms were dug down nearly 3.50 meters
before cement was poured on the holes in order while
the hardcore layer, rubble concrete and iron reinforcing
bars was put in place (Figure 11). So it was provided
that the madrasah rooms would sit on a raft foundation
extending nearly 2.00 meters below and 1.50 meters
above the surface level.
Strengthening the walls, arches and domes: Splic-
ing was applied to the cracks on the walls and domes
of the madrasah rooms and on the portico arches
and domes. At the same time strengthening work was
undertaken with the injection method (Figure 12). In
this method, the cracks are filled by injecting into the
cracks a grout made of brick powder, limestone/küfeki
powder, marble powder, sand and hydraulic lime. To
further stabilize the structure, injection holes 24 mm
1145
Figure 12. Injection method for the consolidation of the
cracks, restoration stage (Archive of the İstanbul Survey and
Monuments Directorate).
in diameter were opened at a depth of 20–50 cm in
the masonry wall surface at intervals on average every
50 cm. and in the brick domes on an average of 20 cm;
the domes and walls were stabilized as the injected
grout was pumped through these holes.
In the damaged walls of the madrasah building, in
the portico arches and domes, a metal construction
was used, consisting of hoops and sashes (Figure 13).
In this application in order for there to be the perpen-
dicular elements within and without, they were placed
in the masonry walls, tying each one to the other at
specific intervals. The channels opened in the walls
of the madrasah rooms while NPU 120 profiles were
being mounted were surrounded with thin plates and
while the steel matting was being laid the walls were
strengthened. In the repair of the damaged domes,
metal elements were aligned horizontally and perpen-
dicularly on interior and exterior surfaces of the domes.
The madrasah and portico domes and the arches that
carried the domes were strengthened with steel matting
that was laid inside and outside.
Strengthening the stones and joints : After it was
demonstrated that the main walls of the main building
had been damaged, they were strengthened in part by
filling the cracks with an injected grout. Prior to this
work being carried out, a portion of the stones that
were damaged were removed and the missing parts
Figure 13. The application of NPU 120 profiles of the
madrasah walls, restoration stage.
Figure 14. The application of bar-mat on the madrasah
walls, restoration stage.
were reconstructed using new stones. Khorasan mortar
was used in the joint intervals (Figure 14).
Strengthening the spike eaves/brick eaves : The
spike eaves that had broken loose or rotted were
removed. That is the bricks that were necessary for the
spike eaves were obtained by putting Khorasan mortar
in rectangular molds 5 × 15 × 30 cm and drying them
under the sun. The new bricks were mounted with the
help of the Khorasan mortar.
Lead covering: In order to replace the lead plate, the
previous lead covering had to be removed. Following
1146
Figure 15. Laying of the lead boards on the dome, restora-
tion stage (Archive of The Protection Council of the Natural
and Cultural Properties).
the work of strengthening the domes and renewing
the spike eaves, a mud mortar made of soil, hay and
water was spread over the brick dome was spread over
the dome 5 cm thick, providing heat insulation and
protecting the lead plates. After the mud plaster had
dried, the new lead plates were installed, covering the
brick dome, beginning at the edge of the dome and
extending up to the finials on the top (Figure 15).
Marble cleaning: First the dirt on the surface
of the marble was softened using a water gun and
paper dough, prepared with ammonia bicarbonate, was
applied over a softened plate. In order to minimize
the air contact with the chemical it was sealed with
a gelatin covering before the cleaning work was com-
pleted with brush and water. The ammonia bicarbonate
was re-applied locally on any badly soiled sections.
Painting Repair: All the painted decorations on the
walls of the main building, on the domes and vaults
were repaired in accordance with the data obtained
from the research. After the existing decorations were
documented with photographs, rasp work was carried
out in order to reveal the main motifs (Figure 16).
Following the rasp work the main motifs were found
beneath 6–8 layers of dye and these motifs were trans-
ferred to translucent paper. The original designs on the
surfaces that had earlier been strengthened and plas-
tered with a lime based mortar were transferred back
by means of opening holes with a needle and closing
them up with coal powder. These designs were then
colored with metal oxide paints.
6 CONCLUSIONS
Feyzullah Efendi Madrasah is a small complex of inter-
national significance that has survived until the present
day. In the repair of this complex the application of
a “consolidation” method that is a minimal interven-
tion, has yielded positive results. For consolidation

Figure 16. Detail of a main motif, before painting repair
(Archive of the İstanbul Survey and Monuments Directorate).
in the main building, traditional methods and mate-
rials were used and, in the madrasah building, steel
and concrete were used. The new materials used were
concealed within the bearer walls and the plaster. It
could not be escaped that the structural fabric of the
building that had been damaged in earthquakes, and
still could be, had to be strengthened using materials
and methods. In addition, concealing the interven-
tion provided integrity to the restoration. However,
in restorations of this type, priority has to be given
to methods that will result in the least damage to the
building. In strengthening the madrasah building, the
dimensions and places of the materials that were to
make up the metal construction did not clarify the rela-
tions between the main walls. Also, the approach paid
insufficient attention to protecting its inherent value,
such as authenticity and historical evidence. The repair
of the main building was carried out with greater care
through the pressure of the Protection Commission
and the decorations on the inner spaces in particu-
lar were protected as a valuable historical document
for the structure and the paintings were repaired with
painstaking care.
Protecting historical buildings and giving them a
new function is a complex challenge. The new func-
tion must not destroy the mass and spatial integrity
of the old structure. The arrangements to be made
and the new materials to be used must meet mod-
ern standards. The new function must protect and
enrich the old building’s spatial quality. There is no
role in the modern educational system for the Feyzul-
lah Efendi Medrese’s educational function from the
Ottoman period. Therefore a change in function was
1147
made and it was seen as appropriate that the building
should continue its library function that it has fulfilled
since 1916. The main library building was separated
from the collection of historical handwritten books
and resident research experts; the madrasah rooms
have been designated for administration and as a book
depot; and the portico section for reference readers
and visiting researchers. The facades that overlook the
courtyard of the portico section of madrasah and the
iwan in the main building were enclosed in order to
gain enclosed space and because of the difficulty of
walking in the half-enclosed areas they were enclosed
with glass panes in aluminum frames. In addition the
portico section was sectioned off with glass framed
with aluminum again to create the reading room. Due
to these changes the main form of the madrasah with
the closed space/ the room-half open space / the portico
and courtyard/open space relationship was spoiled.
Using the madrasah rooms for a book depot is also not
the correct approach to maintaining the spatial quality
of the original layout.
In conclusion the new function given to the building
is a burden on it and detracts from the spatial integrity
of the historic building. In this building the museum
function is a usage that is more appropriate and does
not spoil the basic form. However, despite these issues,
what is pleasing is that the window borders are partially
concealed in the back of the arch and the tie rods,
using glass panes in the undivided and allowing these
additions to be distinguished and removed.
REFERENCES
Ahunbay, Z. 1994. “Feyzullah Efendi Medresesi”, İstanbul
Ansiklopedisi. Tarih Vakfı Publication, Vol.3: 308–309.
Ahunbay, Z. & Diren, Ş. 2001. The Report on the repair of
Faih Millet Library, 16.05.2001.
Akcan, S., et. al. 2001. The Determination Report of Ankara
Survey and Monuments Directorate, 29. 01. 2001.
Dwight, H.G. 1915. Constantinople Old and New, Long-
mans, Green & Co., London, 69–70.
Eriç, M. et al. 2002. The Technical Report of Millet Library
of Mimar Sinan University, 08.07.2002.
Goodwin, G. 1971.A History of OttomanArchitecture, United
States of America – Baltimore, 1971.
Kütükoğlu, S.M. 1978. Darü’l-Hilafeti’l-Aliyye Medresesi
ve Kuruluş Arefesinde İstanbul Medreseleri, İstanbul,
İstanbul Üniversitesi Edebiyat Fakültesi Publication
157–158.
Kütükoğlu, S.M. 2000. XX.Asra Erişen İstanbul Medreseleri:
Ankara Türk Tarih Kurumu Publication, 264–265.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

Application of non-destructive techniques at the Katholikon of Dafni

Monastery for mapping the mosaics substrata and grouting monitoring

Ph. Côte & X. Dérobert

LCPC, Bouguenais cedex, France

A. Miltiadou-Fezans & N. Delinikolas

Hellenic Ministry of Culture, Athens, Greece

O. Durand & J. Alexandre

LCPC, Bouguenais cedex, France

A. Kalagri, M. Savvidou, D. Chryssopoulos, L. Anamaterou & F. Georganis

Hellenic Ministry of Culture, Athens, Greece

ABSTRACT: The Katholikon of Dafni Monastery, one of the most important monuments of the middle-
byzantine period situated near Athens, has suffered severe damages due to the 1999 earthquake. Several studies
have been engaged, in order to collect all relevant data of the internal masonry structure and that of the
mosaics substrata by applying non-destructive techniques, such as Ground Penetrating Radar (GPR) and Seismic
techniques including tomographies. In some areas these techniques have been used both before and after grouting.

1 INTRODUCTION complementary, by virtue of their ability to provide dif-

The Katholikon of Dafni Monastery, situated near
Athens, is an 11th century world Heritage Monu-
ment, famous for its excellent mosaics (Delinikolas
et al. 2003). Due to the 1999 Athens earthquake both
the masonry structure and the mosaics have suffered
severe damages (Miltiadou-Fezans et al. 2004). On the
basis of the results of a series of research programs,
investigations and detailed studies undertaken by the
competent authorities of the Hellenic Ministry of
Culture, the application of adequate hydraulic grouts
injections was decided for the repair and strengthening
of its load bearing structure and for the in situ con-
servation of the detached mural mosaics (Miltiadou-
Fezans et al. 2003).The masonry grouting was initiated
in June of 2006 and completed inApril 2007, while that
of the mural mosaics is still in progress.
In this framework, several studies have been
engaged, in order to collect all relevant data for the
internal masonry structure and that of the mosaics
substrata by non-destructive techniques, such as GPR
(Côte et al. 2004, Vintzeleou et al. 2004) and seis-
mic techniques. In some areas these techniques have
been already used both before and after grouting.
These techniques have been selected as they are fully
ferent information in the structural diagnostic process
(MacCann et al. 1988).
On the basis of a preliminary study (Côte et al.
2004), high frequency ground-penetrating radar map-
ping of more than 50 mosaics, related to their very near
bearing-structure, has been realized, in order to locate
doubtful zones including delaminations, changes of
substrata mortar (i.e. recent repairs) or buried hetero-
geneities. The results of this work, in conjunction with
the manual sonic classification realized by the com-
petent Conservators, can be very useful for a more
detailed and less subjective evaluation of mosaics
substrata pathology.
Furthermore, the possibility of using non destruc-
tive procedures to monitor the movement of the grout,
in real time, during injection was also investigated.
The main objective was to develop a methodology
for detecting, during the injection process, the loca-
tion of any change behind the mosaic’s first or second
substrata, which could be assimilated to void fill-
ing, so that to be able to take corrective measures if
necessary (complementary supporting of areas with
extended delaminations etc). The first results of this
approach realized on a model were very promising
(Côte et al. 2004). In this paper the results of similar

1149
measurements, conducted first on a test wall and after-
wards on the Katholikon itself, are presented and
commented upon.
Investigations have been also realized to control the
grouting effect into the masonry walls, using two seis-
mic techniques. The objective herein was to propose a
simple survey methodology which gives information
for the mechanical characterization of the materials
inside the structure.This approach includes some sonic
2D travel time tomographies that present the advantage
to describe a section of the structure, and may illus-
trate the nature and the geometry of its interior (stone
blocks, fillings, . . . ). A first campaign using this tech-
nique has been undertaken in non injected and semi-
injected areas, as well as in injected ones. A second
campaign is still in progress comprising the repeti-
tion of the tomographies after the completion of the
injections. Furthermore, these few tomographies per-
mit the understanding and calibration of the behaviour
of exhaustive Vp transmission measurements that can
be practiced over the major part of the structure.
2 RADAR TECHNIQUES
2.1 Mosaics mapping
In 2005, an exhaustive radar survey has been realized
on the fifty main mosaics of the Katholikon of Dafni
Monastery. The principle of the radar technique, which
is particularly adapted to masonry structures (Maier-
hofer & Leipold 2001, Valle et al. 1999), is based
on the transmission of an electromagnetic pulse, at
very high frequencies, through the examined struc-
ture. Successive echoes caused by reflections at the
inner layered structures are recorded through time sig-
nals and eventually the juxtaposition of these signals
enables the construction of a time–section (GPR pro-
file), which can be related to the internal geometry of
the investigated structure.
For this specific application a 1.5GHz central-
frequency antenna has been used. The applied proce-
dure has already been tested by Côte et al. (2004).
For reasons of comparison the selected parameters
remained constant during the execution of the radar
survey. Thanks to specific parallel radar profiles, maps
can be constructed from chosen trenches of the struc-
ture. The gray-color scale of these maps is related to the
amplitude of the echoes in that trench. In this specific
case, maps were constructed, representing the energy
of the first part of the radar surface wave.
As the wavelengths of the radar technique are about
7 cm in the mosaics, echoes from the substrata are
mixed with the surface echo, therefore the resolution is
not enough to distinguish one from the other. As con-
sequence, the radar parameter was fixed (
t ∼ 1 ns) completely detached or/and the nature of mortar used
in order to map the first few centimeters including the
mosaic itself and its substrata.
1150
Figure 1. Manual-sonic map of “Saint-Nicolas, constructed
by the mosaic’s conservators”. Damaged zones are in gray.
The objective of the current study is to correlate
this radar gray-scale maps to a level of detachment
or heterogeneity by comparing them to the manual-
sonic maps and the pathology ones, constructed by the
Conservators of the mosaics. Although, it has to be
noted that this comparison has certain limitations, due
mainly to the fact that GPR maps are constructed based
on a multi-valued gray-scale, while the manual-sonic
maps consist of two values (bad or good), deriving
from subjective human-expert senses.
Apart from the aforementioned maps, all the infor-
mation selected by the Conservators concerning the
nature of the substrata mortar has been taken into
account in order to investigate any possible existence
of GPR data-similarities among the mosaics exhibiting
the same substrata. Four main types of mortars were
identified: (i) the original lime based mortar (Byzan-
tine period), preserved only in very few mosaics, (ii)
the Novo-type mortar (late 19th century interven-
tions), having as binder natural hydraulic lime, used
in the vast majority of the mosaics, (iii) the cement-
based mortars (1950’s interventions), used in a small
percentage of the mosaics and (iv) the cement-lime
based mortars (1960’s interventions), found as well
in a small percentage of the mosaics (Chryssopoulos
et al. 2003). The coexistence of all types of mortars is
possible in cases of great size mosaics, due to previous
interventions.
The two following examples show the significance
of such non-destructive technique, which can be auto-
mated in order to obtain comparable information. For
the mosaic “Saint-Nicolas” (Figs 1–2), radar results
indicate that the head and the left shoulder and arm
correspond to a sound area, confirmed also by the
manual-sonic map. The rest of the mosaic is either
for the substrata is utterly different from the one of the
sound area. This result is also confirmed by the fact
Figure 2. Sub-surface radar map of “Saint-Nicolas”. The
damaged zones are in dark.
Figure 3. Manual-sonic and sub-surface radar maps of
“Saint-Orestes”. Damaged zones are in dark.
that, the substrata mortar of the sound part is a cement-
based one, whereas the rest of the mosaic conserves
the initial lime based mortar.
Similar results were observed in the case of the
mosaic “Saint-Orestes”, located on a masonry arch,
(Fig. 3), where once again radar information shows
clearly the existence of two areas: a sound basis and an
upper part which is either completely unstuck from the
masonry or/and the mortar is of completely different
nature. The results are again completely coherent to the
mortar composition of the mosaic substrata. The lower
sound part is a cement based mortar. The upper part
conserves the initial lime mortar used by the Byzantine
artists.
In the case of mosaics having great size, (over
few squared meters) and mainly one type of sub-
strata mortar, correlation can be more complex, as for
the example of the “Birth of Virgin” (Figs 4–5). The
1151
Figure 4. Manual-sonic map of the “Birth of Virgin”. The
more light the color is, the more sound the mosaic is. The
black arrows correspond to the sections of radar monitoring
during the masonry grouting.
Figure 5. Near-surface radar map of the “Birth of Virgin”
mosaic (done in 2005).
manual-sonic map constructed by the mosaic’s con-
servators indicates a quite extended problematic area,
while the radar maps present a more complicated sit-
uation including different levels of detachment or het-
erogeneity. The Novo mortar (based on hydraulic lime)
has been mainly used for the substrata of the mosaic.
From the above presented results it seems that when
the mortars used are quite different in nature, as in
case of lime and cement mortars, the radar profiles
can reveal this difference in a quite obvious way and
this is in accordance with the manual sonic maps. On
the contrary when the Novo mortar, based on hydraulic
lime, is used in the most of the surface the results
are more complicated, than the ones presented by the
manual sonic maps.

2.2 Grouting monitoring applied on a wall model

A test wall, enclosing purposely voids and cavities
at specific locations has been constructed by the
Directorate for Conservation of Ancient and Modern
Monuments. Furthermore two square mortar struc-
tures (60 × 60 × 5 cm) have been attached to this wall,
simulating mortar used for wall mosaics substrata. In
both cases empty spaces have been left in between the
mortar structure and the wall. The scope of the whole
experiment was to make a test of grouting procedure,
using three different hydraulic grouts compositions
designed in purpose, and to examine the long term
behavior of grouts hardening in site conditions. The
wall has been injected in collaboration with the Direc-
torate for Technical Research on Restoration in two
phases. In the first phase a lower and medium zone
was injected using two different grout compositions.
In the 2nd phase the upper zone has been injected. Dur-
ing this injection application (2nd Phase of grouting) a
radar system (1.5 GHz antenna and survey wheel) has
been also used, in order to record the progress of grout-
ing behind the one of the mortar structures (Fig. 6a).
In Figure 6b a schematic presentation of all the
sides of the wall and the injection entrances and exits
is given. In Figure 6c the two selected 44-cm verti-
cal GPR paths are noted. The grouting procedure was
realized at six stages, in order to allow the gradual Figure 6. (a) Test masonry wall with the grouting devices
collection of GPR data after the completion of each on the left and the radar team on the right, (b) Schematic
representation of the test wall – 1st and 2nd Phase of grouting,
grouting stage. The selected setting parameters of the
(c) Mortar structure attached to the test wall, simulating the
radar system remained constant and same with those mosaics substrata – GPR Profiles, (d) Application of the 5th
applied on reference GPR profiles, conducted prior injection of the 2nd phase, situated close to the tested area.
the initiation of grouting. For each injection, the loca-
tion of entrances and exits and the quantity of grout
consumed per entrance were noted.
The first four injections were performed at locations
far away from the tested area, whereas the last two
injections were performed in the vicinity of the two
vertical GPR profiles (Fig. 6d).
As the injection was just applied, the grout contain-
ing 80% of water has been too conductive for radar.
Thus, we have studied the amplitude of the surface
wave itself. Figure 7 shows the evolution of the ampli-
tude of the surface echo at the GPR path A after the
completion of each grout injection.
Figure 7. Amplitudes of the surface echoes, obtained from
During the first three injections the signals successive radar profiles at GPR path A, after each elemen-
remained stable, possibly due to the fact that grouting tary injection.
had been applied far away from the tested area. After
the completion of the fourth injection the appearance grout. During this injection, moisture was detected on
of a void (A1) at about [16–27 cm], possibly filled, is the wall surface itself. The presence of this second
observed. A second void (A2) at about [35–44 cm] is void and perhaps its filling is confirmed after the sixth
also detected after the completion of the fifth injec- injection.
tion, which is conducted directly at the tested area and Similar results are obtained regarding the GPR
corresponds to the highest amount of the consumed Path B (Fig. 8). One void at about [28–35 cm]

1152
 Figure 9. Subtraction of profiles on the “Juda’s Betrayal”
mosaic after the 4th injection.
Figure 8. Amplitudes of the surface echoes, obtained from
successive radar profiles at GPR path B, after each elemen-
tary injection.

is also detected after the completion of the fifth

injection.

2.3 Grouting monitoring applied on mosaics

In March 2007, based on the experience acquired from
the wall model, two mosaics of the Katholikon of
Dafni Monastery, Juda’s Betrayal and Birth of Vir-
gin, have been monitored during grouting by means
of GPR measurements, in order to detect the progress
of grouting in real conditions. Several GPR profiles
were performed on two selected linear sections located
on each one of the two examined mosaics after the Figure 10. Localization of the tested sections (between
completion of a series of sequential grout injections scotch papers) of the mosaic “Birth of Virgin”.
(Figs 9–11). The setting parameters for all profiles
were remained constant and same with those selected
for the reference profile conducted for both mosaics
prior the initiation of grouting.
In order to detect and localize any change caused
by the grouting effect, the reference profile of each
mosaic was subtracted from all the corresponding GPR
profiles. In order to ensure accuracy regarding the
location and start position of the successive profiles,
a survey wheel was used.
The main information obtained from these two Figure 11. Subtraction of profiles on the “Birth of Virgin”
mosaic after the 2nd injection.
grouting monitoring are the followings.
Figure 9 reveals that a void has been filled after Through these two examples of radar monitor-
the 4th injection, which is located inside the masonry ing during injection, we can consider that the radar
behind the “Juda’s Betrayal” mosaic, at the beginning technique could be a useful tool for surveying the
of one of the profile. This is the only important change injection in real time. However, the application of such
detected on this section. a technique in practice revealed two main limitations.
Regarding the “Birth of Virgin” mosaic (Fig. 10), First of all, it must be noted that the resolution
after the 2nd injection, a void filling, just behind the of the radar detection depends on the wavelength (of
mosaic, between 32 to 50 cm of the right section, has the order of few centimeters), whereas the size and
been detected, as presented in Figure 11. This area moreover the thickness of the probable defects, can
corresponds to a detached zone according to the man- be millimetric. Secondly, another important aspect is
ual sonic map of the conservators (see Fig. 4). Along the selection of the appropriate linear section for the
this linear section no other notable change has been performance of the survey, as it is practically diffi-
detected. Concerning the left section, no significant cult to monitor the whole mosaic (inability to ensure
change was detected. This result was expected, as the an untrammeled movement of the survey wheel on the
examined area was located in a sound area according mosaic, due to the presence of a great number of plastic
to the manual sonic map. tubes used for the performance of the grout injection).

1153
Thus it is not evident that the grout will really pass
behind the area that is monitored.
3 SONIC TECHNIQUES
3.1 Sonic tomography
Aim of the complementary seismic tomography cam-
paign was the construction, with a non-invasive fash-
ion, of a map revealing the internal mechanical prop-
erties of the masonry walls. By means of measuring
travel times of the compression wave between source
and receiver points located on opposite sides of the
walls, it is possible to derive a compression wave
velocity map. In the case of an a priori homoge-
neous material, the appearance of a lower-velocity
zone indicates that the material has been weathered
locally. Seismic transmission tomography based on
travel times is more sensitive to zones of micro-
cracking, than it is to isolated cracks, especially if the
micro-cracks are not closed and if the material has been
damaged. In the case of a homogeneous medium, the
difference in travel times, both with and without an
isolated crack, might very well be of the same order
of magnitude as the level of accuracy in the selected
times. Seismic amplitudes or rising times are more
sensitive to cracking than travel times, but remain
substantially more difficult to implement within real
investigation processes (Spathis et al. 1983).
At the Katholikon of Dafni Monastery some tomo-
graphies were performed at different zones, suitably
selected for further investigation regarding the grout-
ing process.
During an initial series of experiments, an optimum
distance between the source and receiver points was
determined, so that the information contained on the
tomography maps would be sufficient enough to per-
form the same diagnostic evaluation to the one of a
denser map. It was shown that a distance of about
10 cm, which leads to a total number of about 400
rays, could be a good compromise.
A multi-channel acquisition system (10 bits,
1 MHz/channel) was used to collect and store the
seismic signals (Fig. 12). Since the shortest source-
receiver travel times amount to some ms, the sampling
frequency was set at 1 MHz, in order to ensure acquir-
ing a sufficient number of points for picking the arrival
times. The source consisted of a hammer coupled with
an accelerometer, while the receivers were accelerom-
eters attached to the wall. Both the receiver and source
signals were recorded for all possible source-receiver
combinations. The time pickings were subsequently
carried out in the laboratory. These arrival times and
coordinates were then fed into the inversion process.
The tomography algorithm employed in this
paper is known as “RAI-2D”; it has already led
to numerous applications in both soil surveying
(Abraham et al. 1998) and structural NDT (Côte &
1154
Figure 12. Seismic acquisition system.
Abraham 1995). RAI-2D has been inspired by
the Simultaneous Iterative Reconstruction Technique
(SIRT - Gilbert 1972). The field of investigation is
discretized into a mesh of points, on which the seis-
mic velocity gets defined. One of the key features
of RAI-2D pertains to its influence zones, which as
opposed to a fixed block-discretization grid are used
when selecting rays to calculate slowness at a given
grid point. The size of these influence zones con-
stitutes one of the adjustment parameters that allow
RAI-2D to adapt its convergence process to the type
of actual data. This algorithm is also characterized by
its use of circular analytical rays. The level of accu-
racy for civil engineering purposes inherent in this
simple and rapid inversion technique, which has been
widely tested, using both synthetic and field data, is
similar to that provided by other more complex meth-
ods based on matrix inversions and sophisticated ray
path computations.
It is recommended to include certain complemen-
tary information with the final velocity map in order
to guarantee the quality of the survey and facilitate its
interpretation.
First of all, algorithm convergence should be
tracked from a statistical point of view (mean residual,
standard deviation). Furthermore, the residual statis-
tics of each source and receiver should be checked,
so as to eliminate those sources and/or receivers dis-
playing out-of-scale statistical values. Secondly, since
both the precision and resolution of the velocity map
are linked to the ray coverage, the plot of rays should at
least be studied. For instance, in zones with few rays,
the velocity value is less precise than in zones with
well-distributed and large numbers of rays.
A horizontal tomography over a length of about 3 m
for a 0.8 m thick wall is presented in Figure 13. The
convergence of the algorithm is reached after 50 iter-
ations. Not any a priori model is implemented in this

Figure 13. Tomographical reconstruction in a grouted area.
Figure 14. Tomographical reconstruction in a non grouted
area.
process. The real edges of the wall are figured by the
black lines.
This image illustrates quite clearly the inner
building-up of this wall. Three leaves are observed.
On both external parts of the wall the high veloci-
ties are associated to stone masonry, while the lower
velocities at inner part are associated to the infill mate-
rial. It is quite interesting to notice that the geometry
of the stone facings is well presented. The thickness
of the stones can be estimated. The linear arrangement
of the inner face of these facing stones is obtained
without any geometrical control of the solution, nei-
ther with any a priori model. The mean velocity of the
stones is about 4200 m/s while the mean velocity of
the filling materials is about 1500 m/s. This part of the
wall has already been grouted.
The tomography illustrated in Figure 14 has been
realized in a non grouted part of the structure. The
process is identical to the one of the previous case.
Here the investigated length of the wall is about 2 m
for a thickness of 0.6 m. This time, the pickings of first
arrivals for each signal are much more difficult.
The attenuation of the signals is very strong. The
early beginning of each signal is quite soft and then
the accuracy of the picking decreases. Nevertheless,
1155
the reconstruction algorithm converges to a three-leaf
medium (again without a priori information). This
time, the inner filling materials are associated with
very low velocities (500 m/s).
The comparison of these two cases permits to
conclude that the grouting survey of such masonry
structures may be achieved by the observation of the
velocities of their inner part. Adapted tomographical
algorithms are able to reconstruct the inner geometry
of such structures.
These tomographies can be realized over some
particular areas, while a simple punctual time trans-
mission set can consider the whole structure by a
convenient scanning.
3.2 Sonic transmission testing
Some simple sonic transmission tests have been real-
ized at Dafni Monastery. The acquisition devices are
quite simpler than the ones applied to tomographical
measurements. A couple of accelerometers are linked
to an acquisition card or to a simple oscilloscope. A
hammer equipped with a shock accelerometer consti-
tutes the source, whereas a second accelerometer is the
receiver. The measurement principle is based on posi-
tioning the accelerometers face to face on both sides
of the wall and knocking the wall in the vicinity of one
of the accelerometers, which is used as a trigger. The
first arrival time can be then picked on the signal from
the other sensor. The whole set is then displaced and a
scan of wall or of a part of wall can be performed.
The example illustrated in Figure 15 corresponds
to a set of measurements conducted over a vertical
line along a wall which has been grouted up to the
limit indicated by the dotted line. Beyond this line, the
behavior of the sonic measurements changes radically.
Two other measurement sets (over horizontal lines)
are plotted in Figure 16. They have been obtained at
two levels of the same wall. The upper level has not
yet been grouted. The difference in the behavior of the
mean transmission velocities is quite obvious.
Such measurements are relatively easy to be per-
formed and interpreted. Large parts of structures can
be then observed. Some indicators can be deduced for
the qualification of the homogeneity of the structure
before any treatment. In case of complex behavior,
a complementary tomographical arrangement can be
performed. After grouting, the same measurements
can be repeated in order to evaluate the efficiency of
the intervention.
4 CONCLUSIONS
High frequency Ground Penetrating Radar (GPR)
mapping of more than 50 mosaics, related to their
very near bearing-structure, has been realized, in
order to locate doubtful zones including delamina-
tions, changes of structures (i.e. recent repairs) or

Figure 15. Sonic transmission velocities across a wall
partially grouted.
Figure 16. Sonic transmission velocities across a similar
wall grouted (red) or not yet grouted (purple).
buried heterogeneities. The results of this work in con-
junction with the manual sonic classification realized
by the competent conservators were very useful for a
detailed evaluation of mosaic substrata.
Sonic tomographies permit the understanding and
calibration of the behavior of exhaustive Vp transmis-
sion measurements that can be practiced over some
areas of the structure. These investigations are com-
plementary to simple transmission investigations that
need simply to locate corresponding measurement
grids on both sides of the wall. At each node, the
transmission time is measured. The map of these mean
velocities for an entire wall may be constructed before
and after a grouting process. Their comparison leads
to a verification of the homogeneity of the wall after
the grouting process.
The co-evaluation of NDT results can be a useful
tool for the assessment of the efficiency of the grouting
intervention.
ACKNOWLEDGMENT
The scientific and technical personnel of the Direc-
torate for Technical Research on Restoration, of The
1156
Directorate for the Conservation of Ancient and Mod-
ern Monuments and the 1st Ephorate of Byzantine
Antiquities of the Hellenic Ministry of Culture, that
has assisted the in situ measurements is gratefully
acknowledged. The project has been realized in the
framework of a research program realized and co-
financed by the Directorate for Technical Research on
Restoration of the Hellenic Ministry of Culture and
the Laboratoire Central des Ponts et Chaussées.
REFERENCES
Abraham, O., Ben Slimane, K., Côte Ph. 1998. Seismic
tomography: factoring anisotropy into iterative geomet-
ric reconstruction algorithms. Int. J. Rock Mech. Mining
Sci. 35 (1), 31–41.
Côte, Ph., Abraham, O. 1995. Seismic Tomography in Civil
Engineering. NDT-CE, Berlin, 459–466.
Côte, Ph., Dérobert, X., Miltiadou-Fezans,A., Delinikolas, N.
2004. Mosaic-grouting monitoring by ground-penetrating
radar. Proc. Int. Sem. on Struct. Anal. of Hist. Const. Proc.,
Padova (IT), nov 2004.
Chryssopoulos, D., Anamaterou, L., Georganis, F. 2003.
Documentation study for the mosaics of the Katholikon of
Dafni’s Monastery after the 1999 earthquake. Directorate
for the Conservation of Ancient Monuments, Hellenic
Ministry of Culture (in Greek).
Delinikolas, N., Miltiadou-Fezans, A., Chorafa E.,
Zaroyianni, E. 2003. Restoration Study of the Katholikon
of Dafni Monastery, Phase A-Architectural and Historical
Survey, Hellenic Ministry of Culture (in Greek).
Gilbert, P. 1972. Iterative methods for the three dimensional
reconstruction of an object from projections. J. Theor.
Biol., 105–117.
MacCann, D.M., Jackson, P.D., Fenning, P.J. 1988. Compar-
ison of the seismic and ground probing radar methods in
geological surveying. IEE Proc. 4 , 380–391, part F.
Maierhofer, C., Leipold, S. 2001. Radar investigation of
masonry structures. NDT&E Int., 34: 139–147.
Miltiadou-Fezans, A., Delinikolas, N., Chorafa, E.,
Zaroyianni, E., 2003. Restoration Study of the Katholikon
of Dafni Monastery, Phase A Structural Survey, Analysis
and remedial measures, Hellenic Ministry of Culture (in
Greek).
Miltiadou-Fezans, A., Vintzileou, E., Delinicolas, N.,
Zaroyianni, E., Chorafa E. 2004. Pathology of Dafni
Monastery: Survey, monitoring of cracks, Interpretation
and numerical verification, Proc. Int. Sem. on Struct.Anal.
of Hist. Const. Proc., Padova, Nov 2004.
Spathis, A.T., Blair, D.P., Grant, J.R. 1983. Seismic pulse
assessment of the changing rock mass condition induced
by mining. Int. J. Rock Mech. Mining Sci. 22 (5) , 303–312.
Valle, S., Zanzi, L., Rocca, F. 1999. Radar tomography for
NDT: comparison of techniques. Journ. Appl. Geophys.,
41: 259–269.
Vintzileou, E., Miltiadou-Fezans, A., Palieraki, V.,
Delinikolas, N. 2004. The use of radar techniques and
endoscopy in investigating old masonry: The case of Dafni
Monastery. In Modena C., Lourenco P.B., Roca P.(ed.)
Proceedings of the 4th International Seminar on Struc-
tural Analysis of Historical Constructions., Rotterdam,
Balkema, Vol. 2, pp. 351–360
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

The PROHITECH research project

F.M. Mazzolani
Department of Structural Engineering, University of Naples “Federico II”, Naples, Italy

ABSTRACT: The PROHITECH research project is framed within the INCO thematic areas, devoted to “Pro-
tection and conservation of cultural heritage” in the Mediterranean area. The main subject of the research is
the seismic protection of historical and monumental buildings, and the main objective consists in developing
sustainable methodologies for the use of Reversible Mixed Technologies (RMTs) in the seismic protection of
the existing constructions. RMTs exploit the peculiarities of innovative materials and special devices, allowing
ease of removal when necessary. Furthermore, an optimization of the global behaviour under seismic actions is
achieved by the combined use of different materials and techniques. The endpoint of the research is a proposal
of codification for the use of such technologies in the seismic protection of existing constructions.

1 INTRODUCTION methods must be not only reliable and durable, but

The Mediterranean and Balkan area is greatly exposed
to seismic hazard. Consequently, its cultural heritage
is strongly susceptible to undergo severe damage
or even collapse due to earthquake. The construc-
tions mostly exposed to seismic risk are the his-
torical and monumental ones, since in many cases
they are not endowed with basic anti-seismic fea-
tures and/or no seismic retrofit has been applied to
them. If the latest earthquakes occurred in this area
are considered (Friuli-Italy, 1976; Vrancia-Romania,
1977; Campania and Basilicata-Italy, 1980; Spitak-
Armenia, 1988; Banat-Romania, 1991; Erzincam-
Turkey, 1992; Dniar-Turkey, 1995; Umbria-Italy,
1997; Adana-Turkey, 1998; Iznit and Duzce-Turkey,
1999; Athens-Greece, 1999; Afyon-Turkey, 2002;
Bingol-Turkey, 2003; Bourmedes-Algeria, 2003; Al
Hoceima-Morocco, 2004, to mention the most impor-
tant, only), the extremely unsatisfactory degree of
seismic protection is clearly apparent. Degradation in
material quality, lack of appropriate maintenance and,
above all, absence of elementary anti-seismic provi-
sions are the clear reasons of the very large number of
the collapses, particularly in old masonry structures,
occurred during earthquakes.
The extreme seismic vulnerability of the histori-
cal constructions is confirmed by this evidence and,
consequently, urgent strategies for the seismic pro-
tection of the cultural heritage are strongly required.
Considering the construction as a system, the objec-
tive is improving its global performance, rather than
providing solutions to specific structural or archi-
tectural problems, requiring the set-up of new tech-
nological systems. Moreover, the new intervention
also, if required, easy to monitor and remove, the lat-
ter aspect corresponding to the widely shared policy
of safeguarding existing buildings from inappropriate
restoration interventions, with particular reference to
historical and monumental constructions. At the same
time, modern constructional systems have provided
good seismic performances, strongly limiting dam-
age and completely avoiding collapse. Consequently,
a slow but continuous increasing in the sensitivity
to the use of more advanced technologies in the
earthquake protection policy has started. The excel-
lent performances of innovative materials have been
acknowledged and the potential advantages of using
special techniques for seismic resistant structures has
been recognized, in a step by step process. Although
initially referred to new buildings, this trend represents
an important study field in seismic rehabilitation of
existing buildings, with particular interest for histori-
cal and monumental constructions (Mazzolani 2005a,
b, c, 2006a, b, c, 2007a, b).
2 THE PARTNERSHIP
The scientific activity of PROHITECH project is sub-
divided into four parts, aiming at producing four main
deliverables, to be developed since October 1, 2004 to
September 30, 2008. The workplan is based on twelve
scientific workpackages, plus three management
workpackages. A number of sixteen scientific work-
package deliverables is foreseen. Sixteen academic
institutions, coming from twelve Countries mostly
belonging to the South European and Mediterranean
area, are involved in the research programme (Fig. 1).

1157

Figure 1. The PROHITECH partner countries.
The partner Countries are: Algeria (AL), Belgium
(B), Egypt (EG), Macedonia (MK), Greece (GR),
Israel (ISR), Italy (I), Morocco (M), Portugal (P),
Romania (RO), Slovenia (SL), Turkey (TR).
In the following the partner institutions and the
relative responsible persons are indicated. UNINA:
University of Naples “Federico II”-Engineering Fac-
ulty (F.M. Mazzolani, project general coordinator);
B: University of Liège (J. Jaspart); MK: Univer-
sity of Skopje (K. Gramatikov); GR: Technical
University of Athens (I. Vayas); NA-ARC: Univer-
sity of Naples “Federico II”-Architecture Faculty
(R. Landolfo); P: Technical University of Lisbon
(L. Calado); ROPUT: Polytechnical University of
Timisoara (D. Dubina); ROTUB: Technical Univer-
sity of Bucharest (D. Lungu); SL: University of
Ljubljana (D. Beg); TR: Boğziçi University of Istanbul
(G. Altay Askar); ISR: Technion Israel Institute of
Technology, Haifa (A.V. Rutenberg); EG: Engineer-
ing Centre for Archeology and Environment, Faculty
of Engineering, Cairo University (M. El Zahabi); M:
Moroccan National Scientific and Technical Research
Centre, Rabat (A. Iben Brahim); SUN: Second Uni-
versity of Naples (A. Mandara); AL: University of
Science and Technology “H. Boumedien” of Algier,
Civil Engineering Faculty (M. Chemrouk); UNICH:
University of Chieti/Pescara (G. De Matteis, project
technical coordinator).
3 THE BASIC ISSUES
Within the technical field of seismic rehabilitation,
two aspects are receiving an increasing attention by
engineers and researchers, namely:
– Preservation of Structural Integrity of existing
buildings under severe/exceptional seismic actions;
– Improvement of building seismic performance by
means of RMTs.
Both these aspects are closely interrelated each
other, in the sense that the application of RMTs is,
in some cases, the only tool to achieve a satisfying
level of Structural Integrity under severe earthquake
actions. The concept of Structural Integrity relies on
1158
the necessity to ensure seismic protection against col-
lapse also in case of destroying events. In this view, it
can be properly framed within the advanced concept
of Performance Based Design (PBD). As well known,
the PBD is a new way to approach the structural design
against seismic actions, having the purpose to ensure a
proper degree of structural reliability under any speci-
fied working conditions, including both serviceability
and ultimate limit states.
Till now, the PBD has been applied to new structures
only, which can be easily designed complying with
relevant behavioural thresholds set by PBD it-self. No
applications exist in the field of existing constructions,
yet. In particular, neither criteria nor methodologies
are available for achieving a satisfying design level
against strong intensity earthquakes. This is indirectly
confirmed by most of national seismic codifications,
which, as a matter of fact, allow to avoid a rigorous
seismic retrofit in case of historical constructions. This
approach, of course, tends to preserve the monumen-
tal value of the construction, but at the same time is
not adequate to protect against severe earthquakes. It
is evident how this aspect deserves great attention not
only in the perspective of saving human lives, but also
at the light of preserving inestimable buildings from
destruction. The use of innovative materials and Mixed
Technologies is the most appropriate answer for ensur-
ing an adequate performance, and hence the Structural
Integrity, under strong seismic actions.
RMTs are based on the integration of structural
members of different materials and/or construction
methods into a single constructional organism. The
basic feature of RMTs is that their application should
be always completely recoverable, that is reversible,
if required. This is considered as an essential design
requirement in order to prevent historical and monu-
mental buildings from unsuitable rehabilitation oper-
ations. The main aim of RMTs is the best exploitation
of material and technology features, in order to opti-
mize the structural behaviour under any condition,
including very severe limit states produced by strong
seismic actions. This practice, initially concerned with
new technologically advanced buildings, is now being
looked up with increasing interest also in the field
of structural rehabilitation, due to the large possibili-
ties of structural optimization and, hence, performance
maximization, both achieved thanks to mixed tech-
nologies. In few words, the use of RMTs would
involve the best exploitation of each material and/or
technology used in the intervention, providing the
best performance from both technical and economical
point of view.
4 THE RESEARCH WORKPLAN
It is planned to achieve the above objectives through
the creation of twelve scientific workpackages dealing
Figure 2. Workplan of the PROHITECH project.
Figure 3. Interconnection among WPs, project deliverables
and main deliverables.
with sixteen deliverables, which are aimed at the pro-
duction of four main deliverables, representing the
final output of the four parts which the research plan
has been subdivided into. The final workplan is shown
in Figure 2, whereas the interconnections among the
WPs and the four Parts, leading to the achievement of
the project goals, are shown in Figure 3. Details on
the project workpackages can be found in Mazzolani
(2005 a, b, c, 2006 a, b, 2007a).
5 MAIN OBTAINED RESULTS
5.1 First main deliverable
The first main deliverable is the synthesis of the results
of the first year activity within the project part R1.
Since this part represents a sort of state-of-the-art in
1159
the field of the intervention strategies for the protection
of buildings, all partner Countries have provided their
own contribution, leading to the definition of a com-
mon background, which represents the starting point
for the development of the innovative contribution of
the project (Mazzolani et al. 2005a).
The contents of the volume, composed of five
chapters, essentially derive from the material col-
lected within the first four workpackages. In particular,
the contents of Chapters 1 and 3 are mainly based on
the output of WP3, the contents of Chapter 2 synthe-
size the output of WP2, the contents of Chapter 4 derive
from WP1 activity and the contents of Chapter 5 are
based on WP4.
In Chapter 1 (authors F.M. Mazzolani and G. De
Matteis, the general and technical coordinators of
the project, respectively) information on seismic haz-
ard in the PROHITECH Euro-Mediterranean Coun-
tries is synthesized, including seismicity maps, main
historical events, macroseismic intensity maps, seis-
mic zonation maps and information on acceleration
response spectra specific to different soil conditions.
The study is organized by subregions and by Coun-
tries. Thus, three main subregions have been identified
on the basis of the geographical proximity and com-
mon geodynamical environment of the considered
Countries: a Western PROHITECH region, which
includes Algeria, Morocco and Portugal; a Central
PROHITECH region, in which Belgium, Greece, Italy,
Macedonia, Romania, and Slovenia are grouped; an
Eastern PROHITECH region, including Egypt, Israel
and Turkey.
In Chapter 2 (authors G. Altay Askar, WP2 leader,
and H. Luş) the damage assessment methodologies
widespread in the PROHITECH Countries are illus-
trated. The chapter presents an overview of build-
ing typologies and construction practices, with some
emphasis on the development of seismic design pro-
visions, and it shows both quick and detailed damage
assessment methods in use in various Countries. Fur-
thermore, the damage patterns observed in previous
earthquakes, the results of post-earthquake damage
surveys, together with earthquake history and perfor-
mance of some buildings representative of the cultural
heritage of the Mediterranean area, are collected.
 Chapter 3 (author A. Iben Brahim, WP3 leader),
is devoted to the seismic vulnerability assessment
of historical buildings in the PROHITECH Coun-
tries. In particular, the current approaches to damage
assessment and the definition of the vulnerabilities of
structural types are presented, together with the anal-
ysis of seismic risk related to the historical building
heritage of the Mediterranean area, on the basis of the
damage data reported in Chapter 2.
In Chapter 4 (authors A. Mandara, WP1 leader, and
F.M. Mazzolani) the existing techniques used in the
PROHITECH Countries for the protection of build-
ings against the seismic action, with reference to the
different structural types, are described.
Finally, Chapter 5 (authors D. Lungu, WP4 leader,
and C. Arion) illustrates the intervention strategies for
seismic protection of buildings, with special attention
towards both the technological and policy aspects of
the problem.
5.2 Project deliverables D1-to-D4
Besides the main deliverable D-I, the output of Part R1
is completed by four project deliverables, as previously
mentioned. Each project deliverable, whose contents
come from the partner contributions, has been edited
by the leader of the respective workpackage, and is
made of two parts. The first part is the text of the
deliverable, written by the WP leader on the basis of the
partner contributions. The second part is an appendix
collecting all the contribution documents written by
the partners.
Project deliverable D1 (Mandara 2005) is an
overview of the traditional technological systems used
in the Euro-Mediterranean area for the rehabilitation
of historical buildings. Project deliverable D2 (Altay
Askar 2005) deals with the assessment of the seismic
damage in historical constructions. Project deliverable
D3 (Iben Brahim 2005) is focused on the risk and
vulnerability assessment of the building historical her-
itage. Project deliverable D4 (Lungu 2005) deals with
the intervention strategies for the seismic protection
of historical buildings.
5.3 Second main deliverable
The main deliverable D-II deals with the elements
required for a correct choice of both materials and
technologies to be used in seismic rehabilitation
(Mazzolani et al. 2006). It merges the outcome of the
activity of WP5 (Innovative Materials and Techniques
for Seismic Protection, leader L. Calado) and WP6
(Set-up of Advanced Reversible Mixed Technologies
for Seismic Protection, leader D. Beg).
D-II is composed of four Chapters. Chapter 1
(authors F.M. Mazzolani and L. Calado) is an introduc-
tion to the RMTs, where the main aspects of seismic
protection based on the use of special devices and
1160
systems are underlined. Moreover a general classifi-
cation of the innovative devices for seismic protection
is briefly presented.
The contents of Chapter 2 (authors L. Calado,
J.M. Proença, P. Skuber and M. Esposto) mainly de-
rive from the WP5 final report. Innovative materials
are presented considering their basic principles, struc-
tural features, fields of application, experimental tests
and numerical models, design criteria and codification
issues, structural applications and economic aspects.
The Chapter is concluded by a comparison and eval-
uation among the considered materials, based on the
output of WP6 final report.
Chapter 3 (authors L. Calado, A. Panão and
L. Pavlovčič) deals with the innovative devices which
can be adopted for mixed reversible technologies. The
structure of the Chapter is based on WP5 final report,
while the contents come from both WP5 and WP6 final
reports. Also in this case, every device is presented by
considering the aspects mentioned in Chapter 2.
Chapter 4 (authors D. Beg and G. De Matteis) is
devoted to the seismic protection systems. Its con-tents
come from WP6 final report and deal with the seismic
protection systems for masonry buildings, reinforced
concrete frames, brick or stone masonry structures,
like towers, domes and vaults, and stone temples.
At the end of the Chapter a general over-view on
the seismic protection with active control system is
presented.
5.4 Project deliverables D5-to-D7
The project deliverables D5, D6 and D7 are related to
the activities developed within WP5 and WP6.
Project deliverable D5 (Calado 2006) is focused
on innovative materials, including new metals and
metal-based intervention techniques, in order to select
suitable materials for creating both strengthening sys-
tems and special devices aimed at the optimization of
the structural behaviour.
Project deliverables D6 and D7 (Beg 2006), as
specific output of WP6, contemporary represent the
complement and completion of the work performed
in WP4 and WP5, by providing the information nec-
essary to the proper use of innovative materials and
mixed technologies in strengthening interventions, as
well as the definition of special systems for seismic
protection to be applied to existing buildings.
6 ACTIVITY IN PROGRESS
6.1 General
The activities actually in progress are those related to
the project Parts R3 and R4 (WP7-to-WP12). Part R3,
finalized to the set-up of adequate calculation methods
for RMTs used in seismic protection, deals with the
experimental and numerical analyses which represent
Figure 4. R.c. building for full scale tests in Naples (Italy).
Figure 5. Interventions on the full scale r.c. building:
(a) first repairing by FRPs; (b) second repairing by BRBs.
two fields strongly interconnected each other and
undoubtedly fundamental for the project, they being
expected to yield the necessary tools for performing
practical design calculations. Part R4 is related to the
validation of the innovative solutions and interven-
tions, to the focus on ad-hoc selected study cases and
to the development of adequate design guidelines.
6.2 Part R3: Experimental and numerical
research
WP7 (Experimental analysis, leader K. Gramatikov,
co-leader L. Taskov) is entirely devoted to the lab-
oratory activity of PROHITECH, useful for giving
an experimental contribution to the assessment and
set-up of new mixed techniques for repairing and
strengthening of historical buildings and monuments
belonging to the Cultural Heritage of the Mediter-
ranean basin. The experimental activity in progress is
developed at five different levels: full scale building,
large scale models, sub-systems, full devices, reduced
scale devices.
The full scale cyclic tests have been performed on
one r.c. building in the Bagnoli area in Naples (Fig. 4).
This test represents a unique occasion of knowledge,
since the studied building is not an “ad-hoc” built
model, but it is a “real” construction, representative
of a large part of the present building stock in many
Countries during 20th Century.
The full scale building has been strongly damaged
by applying a seismic input and then it has been
repaired by means of FRP bars in the mortar joints
of the external walls (Fig. 5a). Successively it has
been damaged again, and an intervention by means
of buckling restrained braces (BRBs) has been carried
out (Fig. 5b).
1161
Figure 6. (a) Three-columns model of Greek temple;
(b) Mustafa Pasha Mosque model during tests.
Figure 7. Steel-wood-concrete composite floor system:
(a) push-out tests (Naples); (b) beam tests (Lisbon).
The programme of large scale tests on shaking table
comprehends the models of a Gothic Abbey (Fos-
sanova, Italy), of an Ancient Greek Temple, of the
Byzantine St. Nikola Church (Psacha, Kriva Palanka,
Macedonia) and of Mustafa Pasha Mosque (Skopje,
Macedonia).
Two three-columns models of a Greek temple
(Fig. 6a) have been tested in Athens (Greece), where
the mechanical properties of marble have been experi-
mentally determined too. The Mustafa Pasha Mosque
model (Fig. 6b) has been tested at the IZIIS Laboratory
in Skopje, Macedonia (Krstevska et al. 2007, 2008),
where the Gothic Abbey and the Byzantine Church
models will be also tested.
Tests on sub-systems are related to both full scale
and reduced scale specimens.
Full scale tests involve innovative steel-wood-
concrete composite structures and timber frames
retrofitted by shear panels and/or dissipative beam-
to-column connections. The experimental campaign
on steel-wood-concrete composite structures, set-up
and designed at the University of Naples “Federico II”
(Italy), has been performed in Lisbon (Portugal) and
in Naples (Italy). Monotonic push-out tests for study-
ing the connections behaviour and both mono-tonic
and cyclic tests on composite beam and floor system
(Fig. 7) have been carried out.
The set-up of metal panels, useful for retrofitting
interventions, has been already experienced on a full
scale r.c. building in Naples (Fig. 8a). Furthermore,

Figure 8. (a) Metal shear panel experienced on full scale r.c.
building; (b) tests on bracing type aluminium shear panels.
Figure 9. Dissipative beam-to-column wooden connection:
(a) experimental tests; (b) component parts of the connection.
an experimental campaign on bracing type pure alu-
minium shear panels (Fig. 8b) is currently ongoing in
Naples.
Experimental studies on timber frames retrofitting
have been carried out in Naples (Italy) and in Istanbul
(Turkey). The behaviour of a dissipative connection of
a beam-to-column wooden joint has been evaluated by
means of “ad hoc” tests (Fig. 9).
Reduced scale tests on masonry walls retrofitted
by steel/aluminium plates or by a steel wire mesh
have been carried out in Timisoara (Romania). The
behaviour of r.c. panels is experimentally analysed in
Bucharest (Romania). Testing on buckling and bend-
ing behaviour of iron/steel elements reinforced by FRP
is in progress in Liège (Belgium).
Finally, tests on DC 90 dampers and on titanium
clamps have been carried out in Ljubljana (Slovenia),
while tests on magneto-rheological devices are going
on in Aversa (Italy).
WP8 (Numerical analysis, leader R. Landolfo) is
aimed at the set-up of numerical models for describing
the behaviour of both structural materials and special
devices on the basis of experimental tests. Numeri-
cal procedures are then applied to the seismic analysis
of upgraded constructions in order to investigate the
influence of main variables by means of parametric
dynamic analyses. The first part of work within WP8
has dealt with a series of preliminary numerical studies
1162
Figure 10. FEM model of Fossanova Gothic Abbey.
Figure 11. FEM model of an ancient Greek temple.
Figure 12. FEM model of St. Nikola Church.
concerning a benchmark activity, by which the part-
ners have created a sort of collective background in the
field of numerical analyses. The second part of WP8
activity has concerned the pre-analyses of some of the
experimental tests planned within the project. Figure
10 shows the FEM model of the Gothic Abbey in Fos-
sanova, whereas Figure 11 shows the collapse mode
given by the numerical model of the Greek temple
subassemblage. The FEM model of St. Nikola Church
is plotted in Figure 12, whereas Figure 13 represents
the vibration modes of the FEM model of the Mustafa
Pasha Mosque (Landolfo et al. 2007), used to predict
the behaviour of the large scale model and to design
the consolidation system.
On the bases of the experimental and numerical
analyses, the calculation methods for reversible mixed
technologies used in seismic protection will be set-up
within the WP9.

Figure 13. FEM model of Mustafa Pasha Mosque.
The results of Part R3 will compose the third main
deliverable, which will be organized in three main
parts: (1) Experimental activity (referring to the dif-
ferent types of experiments); (2) Numerical activity
(dealing with the simulation of experimental tests
by numerical models and comparisons of results);
(3) Calculation models (referred to the identification
of simplified models to be used as design tools in
structural restoration).
The output of WP7 and WP8 will be collected in D8
and D9 project deliverables, respectively. The output
of WP9 will lead to two project deliverables: D10,
dealing with analytical models for special materials
and devices for the seismic structural control, and D11,
dealing with simplified models for the seismic analysis
of historical constructions.
6.3 Part R4: Set-up of codification rules
The activity related to Part R4 will be completed at the
end of the project. The results of the work performed
during the whole project period will be finalized to
codification proposals. Contemporary, the selected
innovative materials and techniques, the developed
reversible mixed technologies, together with the calcu-
lation methods, will be subjected to extensive valida-
tion studies. The data obtained by WP5 and WP6 will
be analysed and compared with traditional materials
and technologies. Firstly, a cost-to-benefit evaluation
will be performed to outline the scope and the effec-
tiveness of new materials and/or technologies. As a
second step, the performance of the proposed solu-
tions will be assessed in terms of strength, including
low-cycle fatigue resistance, ductility and dissipation
capacity compared with conventional solutions. The
results from the experimental (WP7) and numeri-
cal (WP8) analyses will be extensively used. Special
conclusions will be drawn on the applicability of dif-
ferent innovative solutions depending on the specific
conditions and design requirements.
Special attention will be paid to the validation
of the proposed calculation models on the basis of
experimental and numerical results, so as to prove
their reliability in the perspective of their inclusion
in the Design Guidelines (WP12). For the specific
1163
case of monumental constructions, a feasibility study
will be performed to assess the applicability of var-
ious innovative solutions to different types of build-
ings. Together with the considerations based on the
structural performance, additional architectural and
historical–cultural aspects will be accounted for in the
context of the possible interventions.
Within WP10, the validation criteria for restoration
interventions are synthesized in the project deliverable
D12, while project deliverable D13 deals with the per-
formance assessment of the new technologies, which
are compared to the traditional ones.
The knowledge gained within the project will be
conveniently applied to some selected study cases
(WP11), consisting in historical buildings belong-
ing to the heritage of Mediterranean Countries, in
order to perform feasibility analyses of seismic pro-
tection interventions by means of RMTs. The selected
study cases are indicated hereafter. The Mustafa Pasha
Mosque of Skopje and the St. Nikola Church in Psacha,
both in Macedonia, as well as the Gothic Abbey in
Fossanova (Italy) and the Greek Temple in Athens, cor-
respond to the project large scale models (WP7). The
Royal Palace of Naples and the Gallery “Umberto I”,
both in Naples (Italy), are characterized by wooden
and iron structures, respectively. The Koletti building
in Athens (Greece) is the only r.c. structure consid-
ered, while the Beylerbeyi Palace in Istanbul (Turkey)
is interesting for studying the interaction between tim-
ber and masonry elements. The Medina of Salé in
Morocco has been also selected for considering the
complex behaviour of undiversified building blocks.
The design of relevant application solutions based
on RMTs is also foreseen together with the critical
evaluation of collected examples of real restoration
cases, belonging to the current experience in each
partner Country. This activity is framed in WP11, lead-
ing to the preparation of the corresponding project
deliverable (D14). The interconnection between the
study cases and the development of design guidelines
(WP12) is evident, since the first step can be consid-
ered as a benchmark for the set-up of the codification
rules. As a result of the WP12 activity, two project
deliverables will be realized: a proposal of codification
rules for the design of seismic protection interventions
by using RMTs (D15); a manual for the actual imple-
mentation of the procedures proposed and developed
within the project (D16).
The final document of the whole project will com-
prise the fourth main deliverable, downstream of the
fourth year activity, containing a set of recommen-
dations elaborated on the basis of all collected data
and results. The proposal will comply with the most
up-to-dated codification issues in the field of seismic
design, say the Performance Based Design, and will
share the same global layout, language and philosophy
as Structural Eurocodes issued by CEN.
7 EXPECTED RESULTS
The PROHITECH project is an important opportunity
to develop knowledge and technology in the field of the
seismic protection of the Euro-Mediterranean cultural
heritage (Mazzolani 2007a, b). The innovative char-
acter of the technical solutions proposed for seismic
retrofitting is mainly based on the concept of RMTs.
The main expected results of the research activity are
the following.
The basis for the assessment of an up-to-dated state-
of-the-art concerning advanced systems of seismic
protection of existing constructions will be set-up,
in harmony with the specific demand of all Euro-
Mediterranean Countries for a more comprehensive
framing of anti-seismic rehabilitation.
Both conscience and knowledge about “new” mate-
rials and technologies as a suitable alternative to
“traditional” solutions will increase, since the last ones
are proved to be often inadequate to provide a satisfy-
ing seismic performance, in particular when applied
to historical and monumental constructions.
The adoption of materials and systems which
are reversible, recyclable, environmentally friendly,
and economically sustainable will be supported. The
present degree of knowledge on the application
of these materials and systems is not particularly
advanced, neither codified in any form, hence it is
expected that the research activity carried out within
PROHI-TECH project will yield significant innova-
tion in seismic protection practice.
New and up-to-dated information on the problem
of seismic protection will be disseminated, thanks
to the participation of acknowledged institutions, be-
longing to both Europe and Mediterranean basin, all of
them widely experienced in the field of seismic design
and with an ongoing significant research activity in
such area.
Young engineers and architects, as well as
researchers involved in seismic design, will have the
opportunity for a qualified training and research activ-
ity, aimed at an enrichment of existing skills in the field
of structural engineering.
Design and constructional rules for interventions
based on advanced and innovative technologies will be
set-up. This is expected to recall to a greater interest
of both construction industry and practicing engineers,
so as to have a remarkable impact on the current anti-
seismic rehabilitation practice.
Information on the ongoing activity is available in
the web site www.prohitech.unina.it.
REFERENCES
Altay Askar, G. (ed) 2005. PROHITECH project deliverable
D2 Assessment of earthquake-induced structural damage
in historical buildings of the Mediterranean area.
1164
Beg, D. (ed) 2006. PROHITECH project deliverable D6-7
Development of reinforcement procedures for structural
elements based on the use of reversible mixed technolo-
gies, Set-up of seismic protection systems based on the
dissipation of seismic input energy.
Calado, L. (ed) 2006. PROHITECH project deliverable D5
Identification of innovative materials and special devices
to be used for reversible mixed technologies in structural
rehabilitation.
Iben Brahim, A. (ed) 2005. PROHITECH project deliverable
D3 Assessment of seismic risk maps and evaluation of
seismic vulnerability of historical building heritage in the
Mediterranean area.
Krstevska, L., Taskov, Lj., Gramatikov, K., Landolfo, R.,
Mammana, O., Portioli, F., Mazzolani, F.M. 2007. Experi-
mental and numerical investigations on the Mustafa Pasha
Mosque large scale model. COST C26 Workshop, Prague,
Czech Republic.
Krstevska, L., Taskov, Lj., Gramatikov, K., Landolfo, R.,
Mammana, O., Portioli, F., Mazzolani, F.M. 2008. Shak-
ing table tests on the large scale model of Mustafa
Pasha Mosque without and with FRP. Proc. of SAHC08
Conference, Bath, UK.
Landolfo, R., Portioli, F., Mammana, O., Mazzolani, F.M.
2007. Finite element and limit analysis of the large scale
model of Mustafa Pasha Mosque in Skopje strength-
ened with FRP. Proc. of APFIS 2007 Conf., Hong Kong,
China.
Lungu, D. (ed) 2005. PROHITECH project deliverable D4
Definition of methodologies for seismic up-grading of
constructions based on both strengthening of structural
elements and control of the seismic response.
Mandara, A. (ed) 2005. PROHITECH project deliverable D1
Overview of traditional technological systems adopted for
seismic rehabilitation of historical buildings in European
and Mediterranean Countries.
Mazzolani, F.M. 2005a. Earthquake protection of historical
buildings by reversible mixed technologies: the PROHI-
TECH project (oral presentation). Symp. on Seismic prot.
of the cultural heritage by innovative techniques, Rome,
Italy.
Mazzolani, F.M. 2005b. Earthquake protection of historical
buildings by reversible mixed technologies: the PROHI-
TECH project. Proc. of the Symposium on Damage and
repair of historical and monumental buildings, Venice,
Italy.
Mazzolani, F.M. 2005c. Earthquake protection of historical
buildings by reversible mixed technologies: the PROHI-
TECH project. 2nd H & mH International Conference
on “Vulnerability of 20th Century Cultural Heritage to
Hazards and Prevention Measures”, KOS, Greece.
Mazzolani, F.M. 2006a. Earthquake protection of histori-
cal buildings by reversible mixed technologies. Proc. of
STESSA 2006 Conference, Yokohama, Japan.
Mazzolani, F.M. 2006b. Earthquake protection of historical
buildings by reversible mixed technologies: the PROHI-
TECH project. Proc. of the 7th European Conference
”SAUVEUR” Safeguarded Cultural Heritage, Prague,
Czech Republic.
Mazzolani, F.M. 2006c. The research project PROHITECH:
state of advancement. In The Construction Aspect of Built
Heritage Protection, Dubrovnik, Croatia.
Mazzolani, F.M. 2007a. Earthquake protection of historical
buildings. Invited lecture, Reluis workshop, Salerno, Italy.
Mazzolani, F.M. 2007b. Earthquake protection of historical
buildings. COST C26 Workshop, Prague, Czech Republic.
Mazzolani, F.M., De Matteis, G., Mandara, A.,
Altay Askar, G., Lungu, D. (eds) 2005a. PROHITECH
first main deliverable D-I Assessment of intervention
strategies for the seismic protection of historical building
heritage in the Mediterranean basin.
Mazzolani, F.M., Gramatikov, K., Dumova-Jovanoska, E.,
Milutinovic, Z., Taskov, L. 2005b. Earthquake protec-
tion of historical buildings by reversible mixed tech-
nologies: the PROHITECH project. MASE Macedonian
Association of Structural Engineers.
Mazzolani, F.M., De Matteis, G., Calado, L. Beg, D.
(eds) 2006. PROHITECH second main deliverable
D-II Reversible mixed technologies for seismic protection:
guide to material and technology selection.
Out-put of the PROHITECH project meetings: 1st general
meeting in Anacapri (Italy) on 5th–6th November 2004;
2nd general meeting in Azores (Portugal) on 2nd–3rd
June 2005; WP1-to-WP4 joint meeting in Naples (Italy)
1165
on 14th–15th July 2005; 1st editorial meeting on 29th
August 2005 in Ohrid (Macedonia) and 1st Seminar on
30th August 2005; WP5-to-WP9 joint meeting in Naples
(Italy) on 6th–7th October 2005; 3rd general meeting
in Crete (Greece) on 24th–25th November 2005; WP8
special meeting in Naples (Italy) on 23rd–24th February
2006; WP5-WP6 joint meeting in Ljubljana (Slovenia)
on 9th–11th March 2006; 4th general meeting in Istanbul
(Turkey) on 6th–7th April 2006 and 2nd Seminar on 8th
April 2006; WP11 “Iron meeting” in Naples (Italy) on 8th
June 2006; WP11 “Rabat meeting” in Rabat (Morocco)
on 9th–11th July 2006; 5th general meeting in Poiana-
Brasov (Romania) on 18th–19th September 2006; WP7
meetings in Skopje (Macedonia) on 13th–14th November
2006; WP8 meeting in Timisoara (Romania) on 8th–
9th December 2006; WP10-WP11 meeting in Athens
(Greece) on 8th–9th March 2007; 6th general meeting in
Liège (Belgium) on 26th–27th April 2007; WP10-WP12
meeting in Haifa (Israel) on 5th–6th September 2007.
 Structural Analysis of Historic Construction – D’Ayala & Fodde (eds)
© 2008 Taylor & Francis Group, London, ISBN 978-0-415-46872-5

The column of Marcian/Kıztaşı in Istanbul and an evaluation of its

restorations

A. Erdem & R. Ozakin

Yildiz Technical University, Istanbul, Turkey

ABSTRACT: The Column of Marcian, or Kıztaşı, is one of the magnificent monuments that were constructed
in the name of Roman Emperor Marcian in Istanbul. It was built in his honor around 450–452. The monument
consists of a square marble plinth block seated on a three-stepped platform and marble blocks on a single granite
column topped with a marble column capital. Through the intervening years, the Marcian Column has been
damaged by earthquakes, fires and repairs. With the 1999 earthquake, the column became slanted on its axis;
the iron rings on the bases and those on the body itself were dislocated; the capital was deformed and hairline
cracks were found in the column’s body. In 2005, the Marcian Column was restored by the Istanbul Metropolitan
Municipality. This article will examine the old and new restorations and will utilize contemporary restoration
criteria to evaluate the last intervention.

1 INTRODUCTION square on Constantinople’s fourth hill (Muller-Weiner,

In the past, “vow and memorial columns” served
as some of the most important elements within the
monumental appearance of Constantinople/Istanbul.
Among these magnificent columns that were espe-
cially erected in the forums in honor of the emperors
or as a memorial for a victory are the following:
the Column of Constantine, the Column of the Goths,
the Column of Arcadius, the Column of Justinian
and the Column of Marcian (Yerasimos, 2000).
From the time of its erection, the name of the Col-
umn of Marcian, or Kıztaşı as it is called, has been
eponymous to the name of the area in which it is
located. The vicinity of the monument was called the
Square of Marcian in the Roman and Byzantine peri-
ods and was named the Kıztaşı District in the Ottoman
period (Yalcın, 1994). Today the column is located
on Kıztaşı Street in the Sofalar Quarter of the Fatih
District.
2 BRIEF HISTORY OF THE MONUMENT
According to the inscription it bears, the Column of
Marcian was erected by the city governor Tatianus
in honor of the Roman Emperor Marcian (450–452).
The erected date of the column cannot be known
accurately, but it is estimated that it was constructed
around 450–452 when Prefect Tatianus was serving
as governor and its location was in the middle of a
2007).
Because the Column of Marcian stood in the garden
of a private house in the Ottoman period, travelers were
unaware of its existence until the middle of the 16th
century. This column was first mentioned in 1540 by
P. Gyllius in his travel notes. Drawings of the Column
of Marcian that Ottoman traveler Evliya Celebi spoke
of in 1634 (Yalcın, 1994) were made by Spon in 1679
and by Flachat in 1766. This column was not very well
known until the beginning of the 20th century when
it was discovered after the major 1908 fire (Muller-
Weiner, 2007). After the fire, new plans for the region
called for the area around the Column of Marcian to be
situated in the middle of a square created where streets
intersected (Yücel, 1969).
Although initially named after the Emperor Mar-
cian, this monument was known in the Ottoman period
as Kıztaşı (Girl’s Stone) or Kızdirek (Girl’s Column).
It is thought that the monument was named the Kıztaşı
because of the Nike reliefs found on the column plinth.
However this name is also thought to have been related
to the Ottoman slave market that was set up in this
area (Ousterhout & Basgelen, 2005) or the name of
another Kıztası used in the construction of Süley-
maniye Mosque was given to the Column of Marcian
(Yerasimos, 2000).
In the Republican period in 1998 the Column of
Marcian was registered as “a cultural property that has
to be protected.” In 2001 in recognition of the impor-
tance of its universal value, it was included within the
framework of an “old work of the first category”.

1167
Figure 1. The Column of Marcianus/Kıztaşı, at the begin-
ning of the 20th century (Eken, 1992).

Figure 2. Kıztaşı, east façade after the restoration, 2007.

3 ARCHITECTURAL CHARACTERISTICS OF
THE MONUMENT

In the past, the Column of Marcian was constructed

as a monument with a marble plinth that sits on a
stepped marble platform, a single piece of granite
with a marble capital, marble block, the marble upper
plinth and on the uppermost the emperor’s statue.
Today the three-stepped platform part of this monu-
ment remains partially below ground and the emperor’s
statue has been lost (Figures 1–2). Prior to 2005 the
platform was buried underground by the road work
that was carried out during the 20th century. How and
when the emperor’s statue disappeared is not known
(Muller-Weiner, 2007).
The entire surfaces of the monument’s prismatic
lower plinth that was made of Marmara marble
is decorated. The rectangular plinth whose width
and length are close to each other (175 × 195 cm)
is 237 cm high. There is a round wreath decora-
tion of laurel leaves in which is found the mono-
gram of Jesus, the letters I and X in the center
on the east side of the column plinth. Although
the sources write that the same decoration was also
found on the west and south façades, these decora-
tions today have been almost erased (Figures 3–4).
On the north façade a better quality decoration and
inscription are found. On this façade are the symmet-
ric forms of two opposing angels or Nikes (goddess Figure 3. Kıztaşı, west façade of the column plinth
of victory) holding a round medallion or crown motif before the restoration, 2005 (photograph by Üveysuna, D.
(Figures 5–6). While the medallion and the angel on & Demirel, A.).

1168
Figure 4. Kıztaşı, west façade of the column plinth after
the restoration, 2007.

Figure 6. Kıztaşı, north façade of the column plinth after

the restoration, 2007.

Figure 5. Kıztaşı, north façade of the column plinth after

the restoration, 2007.

the right are almost complete, the angel on the left

side is missing her head. In the inscription above this
decoration on the north façade is the statement: “This
column was erected for the Emperor Marcianus by
the Prefect Tatianus”. While it has been said that in
the past the inscription was comprised of bronze let-
ters (Yücel, 1969), only remaining today are the holes
through which the letters were held.
A single granite column that is nearly 110 cm in
diameter and 870 cm in height rises on the lower
plinth of the column. Between this column and the
lower plinth is a marble bracelet 50 cm in height and
on the top of the column there is a marble capi-
tal of Corinthian design (Figures 7–8). The capital, Figure 7. Kıztaşı, west façade of the column capital, marble
decorated with sharp-pointed acanthus leaves and its block and upper plinth, before the restoration (Akşit & Akşit,
corners voluted, has been destroyed in fires. A block 2001a).

1169

Figure 8. Kıztaşı, west façade of the column capital, marble
block and upper plinth, after the restoration.
of marble at the 160 cm height is decorated like a sec-
ond capital on top of the column capital. The corners
of this block are decorated with open-winged eagle
motifs. At the uppermost of the monument there is a
marble upper plinth 60 centimeters high that is thought
to have been where the emperor’s statue was. The total
height of the monument together with the platform
is approximately 17 meters and without the platform
15.30 meters (Bilge, 1972).
4 REPAIRS MADE TO THE MONUMENT
The Column of Marcian, which ranks as one of Istan-
bul’s oldest monuments, has passed through many
earthquakes and fires and has withstood many rebel-
lions and wars. In spite of all these disasters the fact
that it has managed to survive to our day was secured
through repairs carried out at various times. Some
sources provide us with definite information and clues
related to these repairs.
1170
4.1 Ottoman period repairs
In the photographs of the Marcian Monument taken at
the beginning of the 20th century, it is seen that the
monument had had hoops put on it at various levels.
This repair must have been carried out on the parts of
the monument that were damaged in the 1894 earth-
quake and the 1908 Fatih fire. In this photograph,
the marble stepped platform is still at street level. On
the lower plinth of the column, there are steel hoops
at the column bracelet and on the body. The lower
plinth has been encircled with four hoops, the bracelet
with two and the column body with three. One cannot
understand, however, if the metal cramp and chains
that are on the column today–but are not seen in the
photograph–were put on during this repair or in repairs
carried out later. The photograph also does not show
the statue that is said to have been on the uppermost
of the monument, so it could not have been still extant
at this date.
4.2 Republic period repairs
At the start of the 1970s the Kıztaşı was repaired
because its cracks and the pieces falling off the marble
column capital had been posing a danger. Corrosion of
the cramps and mortises found on the capital and their
decomposition were the reason for the damage on the
monument. In this repair the rusty and corroded met-
als were removed and then the marble pieces that had
separated from the capital were cleaned and reattached
with Akemi brand adhesives and the cracked parts of
the capital and the body were filled in. In addition,
because the corroded hoop bolts had fallen out, they
were replaced and the loosened hoops were screwed
back together. The other metal parts and chains, for
which no objections were raised for their remaining,
were painted as a barrier to corrosion.Additional hoop-
ing was carried out on the lower plinth part and the
cracks were filled with cement mortar. Because ade-
quate funding could not be located for this repair at
that time, permanent measures could not be carried
out with the result that the aim of that repair was to
use basic interventions to keep the monument standing
(Bilge, 1972).
Five or six years after this repair, the damage to
the Kıztaşı began to increase and the metal hoops
again became warped. After the major earthquake
that occurred in 1999, the monument further deteri-
orated and began to pose a danger to surroundings. In
2005 restoration was started on the Kıztaşı under the
supervision of the Istanbul Metropolitan Municipal-
ity; however, because the restoration was not being
carried out under the supervision of the Istanbul
Restoration and Conservation Center and architects
Derya Akşit and Ferit Akşit, who were the restora-
tion project authors, it was stopped in 2006. Attempts
were then made to finish this restoration according
to the technical reports that had been prepared by
experts.
5 THE DAMAGE TO KIZTAŞI PRIOR TO THE
2005 RESTORATION
In the survey measurements carried out for the repair
it was determined that the monument was not in an
upright position. It had begun listing, 7 cm to the south
and 16 cm to the east from its axis (Akşit & Akşit,
2001a). It was determined that, because there was
an asymmetric distribution of the load created due
to the disintegration of the marble upper plinth on
which undoubtedly the statue rested, the column was
in weak condition vis-à-vis lateral movements and it
had become dangerous for the vicinity (Babuş, 2001).
Also, the rusting and decomposing of the cramps and
chains on the capital were the reason for the cracks
and broken off parts in the marble. The hoops at the
level of the lower plinth had dispersed because the bolts
had snapped off. Aside from these structural problems,
other problems were also noted: the dirtying of the
Kıztaşı granite body by time and atmospheric condi-
tions, the cleavages on the north façade and formation
of a crust, corrosion and stain on the metal parts, cracks
on the body, the damage on the relief decorations and
the breaks in the corners of the eagle motifs. In addi-
tion, the cement mortar that had been used in previous
repairs had produced salt and caused the materials to
behave differently.
6 2005 RESTORATION PROJECT DECISIONS
In the restoration project, the suggestions given by
the Istanbul Restoration and Conservation Center were
adhered to for cleaning, repairing and protecting the
stone and metal materials in the Kıztaşı (Gökce,
2002).
6.1 Basic principles approved of for the restoration
project
– Today the monument is located at a crossroads with
heavy traffic flow and must be partially protected
from the traffic by being isolated within a circular
square.
– The stepped platform of the monument that remains
underground must be uncovered with an excavation
supervised by the Istanbul Archaeological Muse-
– The writing in the inscription section should be re-
mounted on the marble column using bronze letters.
– The rusted and disintegrated metal parts must be
replaced with materials that are bronze or rustproof.
– The capital with its eagle figures has to be fixed
into place or the original should be removed to a
museum for protection and a copy placed on top
of the column. In the case that a new capital and
new upper plinth are used they should be produced
by pouring a mixture of white aggregate and white
cement strengthened with fiber mesh into special
moulds.
– The monument inclination should be measured
periodically for 12 months to determine whether
there is an increase in incline from its axis. If
the inclination increases, action has to be taken
accordingly.
– In order for the repair work to be done in connec-
tion with Kıztaşı, first the monument’s resistance
against lateral forces has to be increased and a sup-
portive skeleton must be established with a trans-
verse support against it from the ground (Akşit &
Akşit, 2001b).
6.2 Repair methods recommended in the
restoration project
– Cleaning must be carried out without damaging the
stone in order to remove both the dirty layer on the
façade and the cement mortar supplements. Squirt-
ing water on the monument will soften the dirt and
then a brush with a plastic claw brush and water
should be used to eliminate the dirt. For dirt that
cannot be removed, a mixture of mashed paper and
10% ammonia carbonate should be applied onto
the surface of the stone. This should be covered
with plastic sheeting and allowed to wait for 2–3
hours and then the process should be repeated until
the desired result is attained. During this work care
must be taken that metal parts do not come into
contact with the solution.
– Filling for the ditch and cavity stone surfaces: Mor-
tar including lime and mortar that suits the aspect
of the stone must be used. In joint fills the filling
material used in prior repairs that include cement
must be carefully removed and replaced with mor-
tar that is a mixture of hydraulic lime mortar and
marble powder.
– Filling cracks: epoxy that is not affected by water
and is especially designed for bonding (Araldit
AY103 – Hardaner 956) ha