Challenging Glass 3 Conference On Architectural and Structural Applications of Glass 1nbsped 9781614990611 9781614990604 Compress

Challenging Glass 3
Copyright © 2012. IOS Press, Incorporated. All rights reserved.
Challenging Glass 3 : Conference on Architectural and Structural Applications of Glass, edited by F. Bos, et al., IOS Press, Incorporated,
Conference on Architectural and Structural Applications of Glass
Faculty of Civil Engineering and Geosciences, Delft University of Technology
June 2012
www.challengingglass.com
ISBN 978-1-61499-060-4 (print)
ISBN 978-1-61499-061-1 (online)
Challenging Glass 3
Conference on Architectural and Structural Applications of Glass
Faculty of Civil Engineering and Geosciences
Delft University of Technology
June 2012
Bos, Louter, Nijsse, Veer (Eds.)
IOS Press
iii
© 2012 The authors and IOS Press. All rights reserved.
ISBN 978-1-61499-060-4 (print)

ISBN 978-1-61499-061-1 (online)
Published by IOS Press under the imprint Delft University Press
PRINTED IN THE NETHERLANDS
Publisher
IOS Press BV
Nieuwe Hemweg 6b
1013 BG Amsterdam
The Netherlands
tel: +31-20-688 3355
fax: +31-20-687 0019
email: info@iospress.nl
www.iospress.nl
Legal notice
The publisher is not responsible for the use which might be made of the following
information.
Cover credits
Background: Andreas Keller
Photo ribbon, from back left to front right: Erick van Egeraat, Rogier van der Heide,
Tim Macfarlane, Erick van Egeraat, David Sunberg (provided by JCDA), JCDA, SOM
(2x), Timothy Hurstley (courtesy of Israel Museum; provided by JCDA)
Editors
Bos, Louter, Nijsse, Veer
Organizing Committee Scientific Committee

Freek Bos Rob Nijsse, chair (NL)
Christian Louter Fred Veer, secretary (NL)

Jan Belis (B)
Fabrice Bernard (FR)
Freek Bos (NL)
Paulo Cruz (PT)
Martina Eliasova (CZ)
Jean-Paul Lebet (CH)
Christian Louter (CH)
Jens Schneider (D)
Geralt Siebert (D)
Holger Techen (D)
Bernhard Weller (D)
Frank Wellershoff (D)
iv
Foreword
Dear reader,
Once again, glass engineering, research and design specialists gather in the historical
town of Delft, for the third edition of the Challenging Glass Conference (CGC3) at the
TU. The proceedings of this event lie before you. Relative to the two previous occasions,
the conference has made a considerable growth spurt and now features some 90 papers,
distributed over seven topics, as well as five key note addresses.
Fascinating contributions can be found under each topic. ‘Curved & Bended Glass’ has
grown significantly, ‘Laminates & Composite Designs’ as well as ‘Projects & Case
Studies’ are traditionally well represented. Under the topic ‘Joints, Fixings &
Adhesives’, it is clear that structural adhesive bonds in glass construction are on the rise
and here to stay.
But a quarter of all papers is still related to the topic ‘Strength, Stability & Safety’.
Clearly, there is, even now, much to learn about the basics of glass design. We urge the
authors to grab every opportunity for discussion and debate: there is still a considerable
way to go before we will have reached a common understanding of the nature of glass
failure and how we should translate that into design guidelines.
Preparing these proceedings has been a privilege, and we trust any participant in
Challenging Glass 3 will move on with new ideas and inspiration to apply in his/her
own work. We would like to express our gratitude to the key note speakers as well as to
the other presenters and authors. Furthermore, we thank the Scientific Committee
members, the supporting organizations, our main sponsor Glas Trösch, as well as our
other sponsors, and, of course, all participants.
Welcome to Challenging Glass!
Wishing you a stimulating and enjoyable conference,
Freek Bos, Christian Louter

Joint Chairmen of the Organizing Committee
Rob Nijsse, Fred Veer

Chairman and Secretary of the Scientific Committee
v
Sponsors, Support & Partners
Platinum Main Sponsor

Glas Trösch
Gold Sponsors
APG International
Dow Corning
Eckersley O’Callaghan
Silver Sponsors
DC Mat
Glasimpex
GlasStress Ltd.
Groep Leroi / Lerobel
Octatube
Scheldebouw
Van Noordenne Groep / Vindico
Supporting Organizations
IABSE Dutch Group
Kenniscentrum Glas
COST Action TU0905 Structural Glass
Host & Partners

TU Delft
EPF Lausanne
Witteveen+Bos
U-Dispuut
vi
Contents
Keynote papers 1
ADVERTISEMENT: GLAS TRÖSCH 2
Light in the Public Realm 3

James F. Carpenter
Emotion 7
Erick van Egeraat
Fusing Design, Innovation and Light 9

Rogier van der Heide
Engineering Invention in Glass Architecture 11

Tim Macfarlane
Case Study 1 World Trade Center – Podium Wall Design Development 17

Christoph Timm
Projects & Case studies 39

ADVERTISEMENT: ECKERSLEY O’CALLAGHAN 40
The Glass Screens of the Japan Post Tower 41

Lucio Blandini, Werner Sobek
The Glass Sphinx: A Massive Stacked Glass Sculpture 47

Freek Bos, Tom van der Heijden, Pieter Schreurs
The Apple Glass Cube: Version 2.0 57

James O'Callaghan, Charles Bostick
Future Application of Structural Use of Glass 67

Thomas Henriksen
A Laminated Glass Wall Will Protect Warnemünde From High Water 75

Frank Heyder, Franziska Paulu
Lincoln Center Canopies – Performance in Glass 83

Jan Knippers, Jochen Riederer, Matthias Oppe
Project for the Eiffel Tower: Constructive Geometry 93

Nicolas Leduc, Jacques Raynaud, Niccolo Baldassini
vii
Challenging Glass 3
Challenges in the Design, Fabrication and Installation of Glass Structures

Comprising of Super Jumbo Glass Sheets 101
Peter Lenk, Harriet Lambert
Glass Walls Carrying the Roof and Withstanding the Wind Load on the
Façade: Conservatory of the Museum in Dordrecht and Raaks Glass Cube in
Haarlem. 111
Rob Nijsse
Inclined Glass Fins for the King Abdulaziz Center for World Culture 121
Silvia Prandelli, Damian Rogan
Design of Suspended Glass Ceiling Structure in High Sesimic Hazard Zones 129
Toru Takeuchi, Kenichi Sugizaki, and Koichi Yasuda
Designing a Glass Pavillion to Protect an Ancient Greek Temple 139

Fred Veer, Phaedra Oikonomopoulou, Regina Bokel
A True All-Glass Staircase 151

Ernst Wälchli, Bruno Kassnel-Henneberg
Two Lines – Arup with David Chipperfield Architects 157

Felix Weber
Torre Iberdrola, Bilbao, Spain 167

Axel Zemborain
Joints, Fixings & Adhesives 175

ADVERTISEMENT: DOW CORNING 176
Experimental Investigation of Unconventional Canopy Prototypes, Suspended

by Adhesive Bonds 177

Jan Belis, Arno Van Hulle, Dieter Callewaert, Jonas Dispersyn, Bas Out
Connecting Through Reinforcement – Experimental Analysis of a Glass

Connection Using Perforated Steel Plates 187
Paulo Carvalho, Paulo J. S. Cruz, Frederic Veer
Determination of Adhesives Properties for Non-linear Numerical Simulation

of Structural Steel-Glass Connections 195
Vincent Dias, Oliver Hechler, Christoph Odenbreit
Shear Capacity in Adhesive Glass Joints 209

Maria Fröling, Kent Persson, Oskar Larsson
Experimental and Numerical Analysis of Edge Seal Spacers of Insulated

Glass Units for Structural Sealant Glazing Applications 221
Anneliese Hagl
viii
Contents
Tensile Loading of Silicone Point Supports – Revisited 235

Anneliese Hagl, Oliver Dieterich, Andreas Wolf, Sigurd Sitte
Investigation of Stress-Whitening in Transparent Structural Silicone Adhesive 249

Anneliese Hagl, Andreas Wolf, Sigurd Sitte
Designing a Glass Bearing Connection with a Probability to EN1990 CC2 259

Ron Kruijs
Influence of Various Factors on Mechanical Properties of Adhesive Joint in

Glass Structures 267
Klára Machalická, Martina Eliášova
Seismic Behaviour of Point Supported Glass Panels 281

Luís Martins, Raimundo Delgado, Rui Camposinho, Tiago Silva
The Mechanical Performance of Adhesives for a Steel-Glass Composite

Façade System 293
Shelton Nhamoinesu, Mauro Overend
Load Carrying Behaviour of Metal Inserts Embedded in Laminated Glass 307

Kerstin Puller, Werner Sobek
Direct Glass Fabrication – New Applications of Glass with Additive

Processes 315
Lisa Rammig
Analytical Solutions for Detail Problems in Structural Glazing 323

Matthias Seel, Geralt Siebert
Glazing with Countersunk Point Fittings 335

Geralt Siebert, Tobias Herrmann
Reduction of Edge Effect in Adhesive Joints of Glass Details 349

Olena Soroka, Yurii Rodichev, Alexander Shabetia
Strength, Stability & Safety 359

ADVERTISEMENT: DCMAT 360
ADVERTISEMENT: VAN NOORDENNE GROEP 361
ADVERTISEMENT: GLASSTRESS LTD 362
Improvement of Quality of Tempered Glass with Numerical Modeling 363

Antti Aronen, Reijo Karvinen
Analytical Approaches for Buckling Verification of In-plane Loaded

Laminated Glass Columns and Panels 373
Claudio Amadio, Chiara Bedon
ix
Challenging Glass 3
Contact Damage Near the Supporting Pillars in Vacuum Glazing Units 387
Minxi Bao, Jiang Yang, Xiaogen Liu, Yiwang Bao
Towards a European Structural Glass Network: COST Action TU0905 397

Jan Belis, Jürgen Neugebauer, Jens Schneider, Mauro Overend, Danijel
Mocibob
How to Model Failure in Load-Bearing Glass Elements? A Discussion Based

on Analytical, Numerical and Experimental Considerations 405
Fabrice Bernard, Bouazza Fahsi, Baghdad Krour
New Expressions for the Effective Thickness of Laminated Glass 419

Laura Galuppi, Gianni Royer-Carfagni
A Study of Polishing Performance of Glass Using Fluid Jet Polishing 431

Anders Jönsson, Pia Lindahl, Johan Fredin, Christina Stålhandske
Design of Arches Strengthened with Cables for Glass Roofs 441

Wim Kamerling
Design of Cable Strenghtened Beams for Glass Structures 455

Wim Kamerling
The Design of an All Glass Roof to EN1990 467

Ron Kruijs
Stability Analysis of Structural Glass Systems 475

Peter Lenk, Franklin Lancaster
Numerical Simulation of Façade / Window Glazing Fracture Under Impact

Loading 489
Christoph Müller de Vries
Glazing Under Long Term Loads Induced by Water 501

Arthur Pistora, Barbara Siebert
Energetic Approach of Elastic Strain Energy of Thermally Tempered Glass 509

Stefan Reich, Bernhard Weller, Nora Dietrich, Stephan Pfefferkorn
Surface Cracked Layer and Damage Resistance of Glass Under Contact

Loading 523
Yurii Rodichev, Yurii Yevplov, Helen Soroka, Frederic Veer
Surface Defects and Statistical Characteristics of Glass Strength 535

Yurii Rodichev, Yurii Yevplov, Helen Soroka, Frederic Veer, Nikolay
Tregubov, Vladimir Polivyany
Load-Bearing Capacity of Thin Film Photovoltaic Modules 553

Jens Schneider, Johannes Kuntsche, Jonas Kleuderlein
x
Contents
Stress Distribution of Tension Structural Glass Plates 565

Tomas Serafinaviius, Jörg Hildebrand, Gintas Šauiuvnas
Polishing Glass with Fluid Jet Technology 579

Christina Stålhandske, Maria Lang, Marita Jansson, Kent Persson, Anders
Jönsson
Influence of the Load History on the Edge Strength of Glass with Cut Edge
Finishing 587
Marc Vandebroek, Jan Belis, Christian Louter
The Relation Between Fracture Morphology and Failure Strength in

Annealed Glass 597
Frederic Veer, Yurii Rodichev
Strength of Flat Glass Subjected to Thermal and Mechanical Loads 607

Vladimir Zubkov, Nadezhda Kondratieva
Laminates & Composite Designs 619

ADVERTISEMENT: LEROBEL 620
ADVERTISEMENT: GLASIMPEX 621
ADVERTISEMENT: SCHELDEBOUW 622
Laminated Glass Panels in Combination with Timber Frame as a Shear Wall

in Earthquake Resistant Building Design 623
David Antolinc, Roko Žarni, Franci epon, Vlatka Raji, Mislav Stepinac
Anchoring Triangular Glass Panels to Prevent Collapse 633

Graham Dodd, Leroy Reed
Through-Cracked Tensile Delamination Tests with Photoelastic

Measurements 641
Daniele Ferretti, Marco Rossi, Gianni Royer-Carfagni
Composite Structures Made of Glass and Ultra High Performance Concrete -

Current Research 653
Bernhard Freytag, Gerhard Santner, Lutz Sparowitz, Franz Götschl
The Influence of Parameter Setting on the Mechanical Properties of PVB in

Lamination Processes 669
Maurizio Froli, Leonardo Lani
Rectangular Plate Finite Element for Triplex Laminated Glass 679

Ivelin Ivanov, Dimitar Velchev, Nikolay Georgiev, Ivo Ivanov
Reinforced Glass Beams Composed of Annealed, Heat-Strengthened and

Fully Tempered Glass 691
Christian Louter, Jan Belis, Freek Bos, Fred Veer
xi
Challenging Glass 3
The Applicability Evaluation of Hybrid Glass Structure Using Pre-stressing

as Structural Member 703
Naoya Miyasato, Akira Okada, Masao Sitoh, Keisuke Nomoto
Structural Design of Composite Steel-Glass Elements 715

Michal Netusil, Martina Eliasova
In-plane Loaded Glass-Steel Elements 725

Geralt Siebert, Andreas Haese
Load-Bearing Behaviour of Non-Monolithic Glass Laminate Beams for Wide

Spans 735
Erich Trösch, Ernst Wälchli, Thomas Baumgärtner
Experimental Investigations on Continuous Glass-GFRP Beams; Preliminary

Non-linear Numerical Modelling 745
Luís Valarinho, João R. Correia, Fernando Branco, José Sena-Cruz
Design of Glass-Polycarbonate Composite Panels 759

Thorsten Weimar
Further Research About the Short and Long-Term Breakage Behaviour of

Hybrid Glass-Steel Elements 769
Bernhard Weller, Philipp Krampe, Stefanie Retsch
Curved & Bended Glass 783

ADVERTISEMENT: OCTATUBE 784
Durability of Cold Bent Insulating-Glass Units 785

Keith Besserud, Matt Bergers, Andrew J. Black, Lawrence Donald Carbary,
Arkadiusz Mazurek, Donald Misson, Kenneth Rubis
Application of Thermally Curved Glass in the Building Industry 819

Michael Elstner, Markus Kramer
Cold-Bent Single Curved Glass; Opportunities and Challenges in Freeform

Facades 829
Mark Feijen, Ivo Vrouwe, Peter Thun
Quality Control of Bent Heat-Strengthened and Fully Tempered Glass by the

Application of Photoelasticity 837
Markus Feldmann, Pietro di Biase, Ruth Kasper
Layout Strategies and Optimisation of Joint Patterns in Full Glass Shells 845
Thiemo Fildhuth, Sebastian Lippert, Jan Knippers
xii
Contents
Special Aspects in Planning of Thermally Curved Glass: Structural design,

Constructional Hints 857
Markus Kramer, Michael Elstner
Curved Glass in Four-Point Bending 865

Bernhard Weller, Michael Engelmann, Philipp Krampe, Stefan Reich
Architectural Design & Lighting 879
Glassolutions LED-in-Glass Outdoor at Oskomera Group HQ 881

Rens Demarteau, Esther Hebly, Paul Roman, Rino Messere
Special Glazing 885

Christian Eckhardt
Architectural Aspect of Structural Design of Glass Façades / Glass Skin

Applications 891
Aleksandra Krstic-Furundzic, Tatjana Kosic, Jefto Terzovic
Cones Made of Glass 901

Juergen Neugebauer
Green Houses – Why Can´t They Be Built as Our Ancestors Did? 907
Holger Techen, Matthias Michel
Glass in Façades 915

ADVERTISEMENT: APG INTERNATIONAL 916
Novel Testing Techniques for Building Glass in Service Conditions 917

Yiwang Bao, Xiaogen Liu, Yan Qiu, D.T.Wan, Minxi Bao, JiangYang
Fire Resistance without Fire Resistant Glazing 925

Csilla Csoke, Johan Koudijs
Building Design Might Soon Be Driven by Energy Requirements Only 933

Christof Erban
Ceramic Digital Printing - Customizing Glass Façade Design 943

Bernd Hoffmann
Behaviour of Electrochromic Glass in the Mediterrean Area 957

Gianraffaele Loddo, Daniela Ludoni, Marco Pittaluga, Gian Piero Cossu
Building Integrated Photovoltaics 971

Barbara Siebert
Design Methods and Structural Components of Blast Enhanced Facades 981

Frank Wellershoff
xiii
Challenging Glass 3
Indices 997
Author Index 999
Advertisement Index 1007

xiv
Challenging Glass 3
Key Notes
1
This page intentionally left blank
Challenging Glass 3 – Conference on Architectural and Structural Applications of Glass
Bos, Louter, Nijsse, Veer (Eds.), IOS Press, 2012.
doi:10.3233/978-1-61499-061-1-3
Light in the Public Realm

James F. Carpenter
James Carpenter Design Associates Inc, 145 Hudson Street, 4th Floor, New York, NY
10013, USA, www.jcdainc.com, info@jcdainc.com
1. Preview
The material properties of glass and its ongoing technological development offer
particular opportunities for daylighting and lighting. Our own approach is to consider
the possibility of approaching glass as an opportunity for the reestablishment of
experiential light in public realm. Electrical lighting at night is one area where quantity
is suppressing the possibilities of presenting subtle qualities of light.
Figure 1: Israel Museum, Jerusalem, Israel. Photo: Timothy Hurstley (courtesy of Israel Museum).
Today, in keeping with our prioritizing of light in the public realm, we continue to
explore many materials, but glassy materials such as glass continue to be of particular
interest for their specific ability to unravel and reveal the density of information
3
Challenging Glass 3
inherent in light. Though we used to focus on the structure of transparency when talking
about glass a decade ago, we always placed transparency within a spectrum of
characteristics. Today, the transmission, reflection, refraction and absorption of light
still defines our approach to glass. Integrating structure within the glass, the use of
oversize panels, coatings and other developments serve these interests for us, be it in
daytime or nighttime conditions.
There are many steps toward changing prevailing attitudes towards overabundant
electric lighting in the public realm, many that due to city, state, federal and national
codes seem intractable. However, is there a place for the glass industry in the discussion
of public attitudes and eventual changes in practice? Can glass become a material that
demonstrates the opportunities for lower, more diverse and better deployed levels of
light in the public realm? We believe glass is rich with potential – just as we have
always explored the depth of light information that can be revealed within glass, so we
believe that glass as it is used in the public realm, be it curtain walls or future
applications emerging from thin glass films and computer technology, can be designed
to enhance a denser and darker environment with subtle qualities of light that can
provide enough illumination, wayfinding and safety while enhancing our human
experience with a powerful sense of nature, both day and night.
Figure 2: Dichroic Light Field, New York, 1994.
4
Light in the Public Realm
Figure 3: 7 World Trade Center Podium and Envelope, New York, 2002-2006. Photo: David Sundberg.
Figure 4: Ice Falls, Hearst Tower, New York. Photo: Andreas Keller.
5
doi:10.3233/978-1-61499-061-1-7
Emotion
Erick van Egeraat
Designed by Erick van Egeraat, Rotterdam, NL
1. Preview
Glass is one of the true marvels discovered by men. The mere heating and cooling of
ever-present silica allows for a transformation and opportunity which is unparalleled.
To me, glass incorporates an enduring quality I strive for in each of my projects. Not as
a mere opening within a structure but an integral part of Architecture.
Glass and Light are a one, without the one the other cannot shimmer, shine, reflect or
capture. If architecture is gravity, then glass is the illusionist allowing light(ness) and
interaction. We have been pushing the boundaries of glass throughout the centuries.
And like the glass in the Gothic Cathedrals it tells a story for those who wish to listen
and observe. The opportunities to tell this story as an Architect are endless: etching,
coloring, sandblasting deforming of glass, to name just a few, make it one of the most
versatile materials. This richness in treatment and effect make it still the favorite in the
development of our contemporary world.
To use Glass as an Architectural asset requires determination, skill and a love for detail.
As our Gothic predecessors it still is much a process of ‘Learning by doing ‘. Prototypes
and tests are still very much a part of making Glass work. At our offices in Rotterdam,
Moscow and Budapest you can still witness these samples from which our Architecture
was derived.
Beside today’s incorporated technology, Glass to me remains a primarily emotional

element. It invites light to structure defining our comfort zone and well-being.
2. Re-creation
For our design of the new Leipzig University we use Glass to re-create an emotional
reminder. In the sixties of the 20th century the East-German regime at the time
destroyed one of the oldest churches in the region. This Pauline-church at the time was
believed to be the center of the critics of the regime. This symbol of freedom will now
be re-created as reminiscence to it. The Gothic arches and columns will be made of out
of ceramics and Glass, allowing light to shimmer, shine and reflect and at night
transform into lighting elements.
Our world is covered with Glass which we mostly take for granted. Let’s aim for a
world in which we can enjoy the richness Glass can offer.
7
Challenging Glass 3
Figures 1a, b: Office Tower, Amsterdam.
Figure 2: City Hall, Alphen a/d Rijn. Figure 3: New Leipzig University.
Figure 4: Extension of InHolland University, Rotterdam.
8
doi:10.3233/978-1-61499-061-1-9
Fusing Design, Innovation, and Light

Rogier van der Heide
Philips Lighting, Rotterdam, NL
1. Preview
Rogier van der Heide is Vice President and Chief Design Officer of Philips Lighting. In
that role it is his mission to inspire architects, interior designers, artists and lighting
designers around the world with products and solutions that unleash one's creative
freedom. As a thought leader and designer himself, Rogier has completed many projects
on all continents. He specializes in the creation of spatial experiences that are
imaginative, colorful and stimulating.
Figure 1: Buckminster Fuller’s Fly’s Eye Dome.
9
Challenging Glass 3
Figure 2: Force Field, London.

Figure 3: Galleria West Façade, Seoul.
10
doi:10.3233/978-1-61499-061-1-11
Engineering Invention in Glass Architecture

Tim Macfarlane
Glass Light and Special Structures Limited, London, UK
The introduction of new structural configurations within the field of glass

engineering is the topic of this address. The generation of these design ideas and the
process by which they are realized is particular to the production of architecture and
although there is considerable cross fertilization with the world of product design
there are some fundamental differences in the two processes that highlight in what
circumstances architectural invention can best thrive.
Keywords: Engineering history, invention, glass structures
1. Architecture and the Engineering Consultant

At around the end of the 19th Century structural engineers were not part of the architects
design team. In the field of building construction the engineer was challenged with
inventing uses for the newly emerging construction materials which at that time were
reinforced concrete and steel. As a result of this effort a number of patented frame
systems appeared on the market which were used in the construction of factory sheds
and other simple building types. The development of these patented frames was in every
way similar to the way in which products for the mass market are introduced: a large
amount of design and testing followed by the production of a large number of units. The
larger the number, the greater the return.
The architect would have been acutely aware of these new building technologies and
although the architectural critic Ruskin was in this period declaring that the introduction
of ferrous metals into the fabric of architecture would be a corruption of its very soul
there were a number of ‘modern’ architects who could see that reinforced concrete and
steel would be the construction materials of the future.
The problem was that the means of design and production was in the hands of the
contractors and the engineers who worked for them. The relationship between the
contractor, the architect and the client then as now meant that access to the contractors
design knowledge to allow the architect to fit the structure to the architecture was
limited.
The contractors focus is on building safe economic structures and building the same
solution more than once gives him the opportunity to reduce risk and increase profit.
His natural tendency is therefore to avoid embarking on something he has not done
before and which during the bidding period he has insufficient time to investigate. A
further disincentive is that he is often only one of a number of contractors and there is
no guarantee that any effort he might make to understand the full implications of a new
design idea will be rewarded.
11
Challenging Glass 3
Architects, when designing a work of ‘Architecture’ are in almost all instances

proposing a bespoke solution within which their control of the construction materials of
the building is critical.
In order to bridge this gap between the architect’s desire for a bespoke solution and the
contractor’s desire to reduce risk the engineer was invited to be a consultant to the
design team. This meant that the architect could use the engineers skill to develop
bespoke solutions during the design process and thereby relieve the contractor of the
design responsibility for the proposed structural configuration and thereby his risk. This
new partnership gave rise to the creation of Associations of Consulting Engineers in
many countries in Europe between 1900 and 1910. The independence of the engineer
from the commercial world meant that proprietary technical knowledge could be
brought into the public domain and ultimately codes of practice and standards for
designing in these materials came into existence
2. The Emergence of Glass as a Structural Element

The two structurally significant inventions in the manufacture of glass which happened
in the middle of the 20th Century were the float line process for making glass in a
continuous ribbon and a method for strengthening glass by tempering. In addition to
these two major developments, the improvement of laminating techniques, which had
first been invented in the late 19th century, allowed the production of thicker glass
panels with significantly enhanced fail safe performance. These three techniques offered
architects the opportunity of developing a new design vocabulary for glass which,
however, required a comprehensive understanding of its structural behavior.
Up until the mid 1980’s Pilkington, the company that patented the float line process,
controlled the way that glass manufactured by their process was used in building
structures. Glass processors and installers were bound by Pilkington’s rules as the
company held patent rights on all the float line machines. This made it very difficult for
an architect to investigate new uses for glass as fabricators were unable to provide glass
which did not conform to Pilkington’s strict guidelines.
After the patent rights ran out in the mid 1980’s there was an opportunity for architects
to begin to think of new uses for this fundamentally traditional material. They found,
however, that as much as they were able to imagine a floor made of glass, they were
unable to persuade fabricators to design and build these structures.
3. The Era of the Structural Glass Consultant

The structural design of glass was not a part of the academic training of a Structural
engineer until recently and in the mid 1980’s there were very few sources for
determining the physical properties of glass and no codes that a practicing engineer
could consult. Information of this type was jealously guarded by the glass
manufacturers who could use their technical expertise to offer proprietary systems.
Architects reasonably thought that structural consultants should be able to help them
develop original design ideas and engineers were encouraged by a number of ambitious
architects during this period to start thinking about how to design glass structures.
12
The following examples indicate the progress we have made since 1988 as glass
engineering consultants. And although there are still no current codes for the structural
design of glass, there are now a number of extremely useful handbooks to consult which
means more engineers can now offer this support to the architect
4. Inventions in the use of Structural Glass

Our progression in understanding the structural properties and behavior of glass started
with the design of a simple staircase tread. Through subsequent projects over the
following 25 years, we were able introduce radical ideas that expanded the vocabulary
of glass architecture. The following projects illustrate the structural glass inventions that
we introduced during that period.
Figure 1: Glass and acrylic treads, 1988.

Figure 2: Laminated glass floor plates.
13
Challenging Glass 3
Figure 3: Simple glass beams and columns.

Figure 4: Bolted glass beams.
14
Figure 5: Friction grip laminated glass beams and columns.

Figure 6: Parallel cable supported glass walls.
15
Challenging Glass 3
Figure 7: Two side supported glass treads using ionoplast laminates.

16
doi:10.3233/978-1-61499-061-1-17
1 World Trade Center – Podium Wall

Design Development
Christoph Timm
Skidmore, Owings & Merrill LLP, New York, USA, christoph.timm@som.com,
www.som.com
On the new 1 World Trade Center Tower an exterior permeable façade treatment
was to be designed for the lower 20 floors cladding a concrete shear wall and
mechanical louvers. Engineering and design had to be balanced with the client’s
desire to streamline maintenance issues for the tallest office building in the western
hemisphere. The scope of the podium façade treatment includes over 12.000 m2 in
total area with more than 4.000 glass fins each free spanning 4m in height between
attachments. Various glass fin-metal connection details were evaluated for aesthetic
and performance criteria.
Keywords: Laminated Glass, Glass Metal Connection, LED Lighting, Aero Elastic
Testing, 1 World Trade Center
1. General
Upon completion 1 World Trade Center will be the tallest building in the western
hemisphere with a symbolic 1776’ (541m) in height and 104 floors. Its outer skin will
have a total of about 88.000 m2 (945,000 sf), comprised of 75.000 m2 (807,000 sf) of
tower office curtain wall starting at the first office floor at Level 20, podium wall below
12.200 m2 (132,000 sf) and cable net wall systems at each of the 4 entrances on all
sides of the 62,48m x 62,48m (205’ x 205’) square building.
This paper focuses on the expedited development of the podium wall design process.
Particular emphasis is given to selection of the glass fin make up and engineering of the
glass/ metal attachment to the podium wall units, its components, and the choice of
glass in the context of aesthetics and performance.
1.1. Project Delivery Method

The podium section of World Trade Center One was previously designed with a
prismatic glass enclosure that had been using the traditional Design-Bid-Build
Delivery Method. When the contractor who had been awarded the scope failed to
deliver his work in a timely and acceptable manner the decision was made to abandon
the original scheme and develop an alternate design under an expedited schedule.
Therefore the Design Assist Method was chosen for delivery and the new podium
design developed to a level typical for a mid construction document phase in the US.
Bid documents included engineered key details and illustrations showing the
architectural intent for the entire scope of work.
17
Challenging Glass 3
In May 2011 the construction manager Tishman Construction engaged Benson

Industries Inc., the curtain wall contractor delivering the tower office curtain wall to
assist the design team in developing the new scheme. By the beginning of the summer
the client decided to bid the podium wall scope of work and bidding guidelines
prevented Benson Industries to be further involved in the development in order to stay
eligible for bidding. APG International (APG) took over Benson’s role in assisting the
design team with further development, providing in depth technical assistance, cost
estimation services and eventually a full size proof of concept mock-up that convinced
the client and design team that the new scheme was feasible and met all aesthetic
expectations. At the end of 2012 Permasteelisa North America Corp. (PNA) was
awarded the contract for the podium wall through an RFQ and RFP process. They
immediately went to work together with the design team, construction manager and
client.
Figure 1: Rendering World Trade Center One [1]
18
1 World Trade Center One – Podium Wall Design Development
1.2. Building Ventilation Requirement Air intake, Exhaust

All office floors of World Trade Center One are enclosed in a glass envelope without
major louver interruptions. Therefore the building “breathes” below and above the
office section with extensive louver areas. Figure 2 illustrates the extent of the podium
ventilation zone on World Trade Center that wraps around the entire building. It shows
the unclad podium portion of the building being visually bifurcated into a lower part
that consists of concrete shear walls extending up to the top of the cable net entrance
portals and an upper part that is comprised of recessed mechanical louvers set back 6’
behind the façade system, structural steel lattice beams and columns. This 6’ zone
between permeable podium wall façade and mechanical louvers is used as a plenum for
air to mix and freely move in and out of the building. All intake louvers are located on
the North and South Façade and all air exhaust louvers positioned on the East and West
Façade. Baffles at the corners between the plenums prevent air from short circuiting and
ensures that exhaust air is not simply being reused as fresh air. Due to tenant changes
and potentially changing mechanical requirements extend of the active mechanical
louvers will change over time but will not be perceptible through the decorative face.
Figure 3 demonstrates the unifying character of the decorative podium wall as it screens
the lower concrete shear walls and the mechanically active upper section of the podium.
The podium wall is an aesthetic device and not a thermal envelope and not watertight.
Figure 2: Construction Photo, South East Podium Corner (June 2011)
19
Challenging Glass 3
Figure 3: Rendering of Finished Podium Wall South East Corner, note the north pool waterfall of the World
Trade Center Memorial in the foreground of the image
2. System description
2.1. General
The 4,06 m x 1,52 m (13’-4” x 5’-0”) grid of the podium wall is an extension of the
unitized office tower curtain wall grid above. The podium itself is vertically divided into
14 tiers; each full tier encompassing 40 typical podium wall panels and two half corner
units per façade. At the base of the wall there is a short tier that includes a wall panel
with stone cladding below that meets the plaza on the east and west side of the building
and the street sidewalk on the north and south.
Figure 4 illustrates the repetitive nature of the design with over 90% of the panels being
identical except for the angle of the glass fins as discussed later in the paper. The
efficiency of this system translated directly into cost reduction.
20
Figure 4: Podium Wall Panel Types
2.2. Unitized Podium Wall Components

The podium wall panels are in principal a curtain wall system without the requirements
for air and water tightness. In fact, the opposite is required as the panels will have to
allow for free air movement to ventilate the building. The unitized panels will arrive
prefabricated at the site, ready to be hung off of previously installed anchor assemblies
on the face of the building.
Each panel is comprised of a series of layers that as an ensemble screens the inside of
the building but allows air to travel through. The outermost layer of the design are two
3,96m x 0,60m (13’ x 2’) laminated glass fins that cantilever off each of the typical
podium wall panels. The fins are dynamically oriented in varying angles across the
façade with a more closed appearance at the top and bottom tiers of the podium and a
more open configuration for maximized air throughput at the mid section. Structurally
siliconed to a stainless steel extrusion receiver shoe at the top and bottom of the glass,
each fin assembly is bolted to a stainless steel outrigger that fixes the angle. At the
center of the unit between the fins is a vertical channel located that houses the LED
lighting system which is serviced from the exterior of the building only. Behind the
glass fins are horizontal stainless steel slats attached to the unit framing adding visual
surface to the assembly and a notion of weaving.
At the back of the painted aluminum panel framing an aluminum perforated sheet metal
screen is mounted and serves 3 functions: It screens the plenum particularly where the
glass fins are more open, acts as a projection surface when lit at night and prevents
birds from entering the plenum.
21
Challenging Glass 3
Figure 5: Explosion Axon of Typical Wall Panel
2.3. Varying fin angles

The primary design intent for the glass fins is to create ever-changing reflections of the
sky and context, and refract light in infinite angles - different in medium, but very
similar in concept to the waterfalls at the World Trade Center Memorial site nearby.
Throughout the day and season the play of light and shadow will animate the façade as
well as the observer’s changing vantage points. This dynamic character of the façade is
amplified by the changing of the angle that the glass fins form with the building face.
While a set of fins parallel to the building face always defines the corners and volume of
the podium on every one of the 14 tiers, all other fins change angles in a carefully
composed way. See Figure 6, 7 and 8.
22
Figure 6: Plan Detail showing Corner Panel and two typ. Panels with Varying Angles Tier 1-14
Figure 7: Dynamic Glass Fin Angles vary with Tiers Figure 8: Close up Perspective Diagram at Corner
2.4. Ventilation Performance of Podium Wall Units

Air traveling through the podium wall panels finds its way through the various layers of
the panel - glass fins, slats, perforated metal, framing, etc., with all components causing
some level of resistance (pressure loss). This resistance and its impact on the
mechanical system had to be evaluated.
Initial desk top calculations performed by Rowan Williams Davies and Irwin Inc.
(RWDI) indicated acceptable pressure loss in the plenums but further verification was
required with computational fluid dynamics modeling (CFD). In an effort to develop
accurate input data for the CFD model, a full scale mockup with all relevant layers of
the podium wall system was constructed and pressure loss across the system measured
at varying fin angles (Figure 9). Data for the various fin angles under intake and exhaust
mode were then plugged into the CFD analysis (Figure 10).
23
Challenging Glass 3
Figure 9: Wind Tunnel Testing of Podium Wall Assembly
The objective of this task was to estimate the average static pressure differences across
the plenums of each facade. Based on a review of the mechanical and architectural
drawings, it was anticipated that the performance of the intakes on the south facade
would reflect the worst case conditions.
CFD analysis showed that the expected static pressure differences across the facades
would be in a range well within the limits that would allow the mechanical system to
function as engineered.
Figure 10: CFD Analysis shows Correlations w/ Mechanical Louver [5]
2.5. Glass selection

The idea for the glass fins on the podium wall is to match the visual reflectivity of the
tower curtain wall above to the largest degree possible in an effort to visually unify the
building. Knowing that different lighting conditions on the office floors and mechanical
floors would make the glass appear differently, the design team focused on matching
reflectivity of both glass assemblies as a means to visually tying them together.
The Viracon VRE-54 Insulated Glass Units on the tower curtain wall with a reflectivity
of 33% the target, the team selected Interpane’s Ipasol Bright White coating due to its
extremely high light transmittance of 58%. Typically used in solar-control applications,
its unique characteristics result in a reflective and bright light appearance. As a
monolithic glass lite, Ipasol Bright White’s has a reflectivity of about 36% which
24
decreases by half if laminated into a safety glass assembly with a clear lite. To counter
this effect and to maintain the high reflectivity, the final glass assembly will feature the
Interpane coating on the #2 and #3 surface resulting in a reflectivity matching almost
exactly the 33% of the office curtain wall above the podium.
Besides the aesthetic appearance characteristics, the Bright White coating also features
the technical advantage of not requiring an edge deletion due to absence of silver in the
coating design that could degrade when exposed to humidity. Edge deletion would
create a vertical line and visual frame destroying the homogenous and scale-less
appearance of the fin’s glass surface.
Figure 11: Reflectivity of Tower Glass vs. Podium Glass
25
Challenging Glass 3
Figure 12: Surface Characteristics of Glass Fins
Studies with the reflective glass were undertaken and it became clear that under certain
lighting conditions the glass would come close to matching the tower curtain wall
appearance. With a change of lighting conditions and a different vantage point however,
the fins tended to lose their presence and amplified the horizontal slats at the back of the
podium wall due to their reflective nature. To counter this effect and add luminosity to
the assembly, an acid etch finish was added to the back surface of the glass. This was
for important aesthetic considerations as described and was not required structurally.
Figure 13: Renderings, W/O and W/ Acid Etched Back Layer
26
Figure 14: Surface Characteristics of Glass Fins
2.6. Structural analysis of glass make up

A series of varying glass assembly make-ups were studied for their structural
performance and feasibility on the system. The following base criteria were established:
x Structural deflection limits for glass fins < l/100 typically as the glass fins only
serve as decorative components of the façade
x Maximum acceptable deflections within dimensional boundaries of the units
illustrated in Figure 15, the fins cannot hit each other or have contact with the
stainless steel slats behind
x Design wind load +/- 2,87 kPa (+/- 60 psf)
x 1.52mm Ionoplast interlayers
x All glass to be heat strengthened
Based upon finite element analysis (FEA) Glass make up GL-01 and GL-03 were found
to meet all criteria and were specified in the bid documents (see Figure 16, 17)
Figure 15: Dimensional Boundaries for Glass Fins [2]
27
Challenging Glass 3
Figure 16: Structural Analysis Glass Make up Option GL-01 [2]
28
Figure 17: Structural Analysis Glass Make up Option GL-03A with Stainless Steel Embed [2]
29
Challenging Glass 3
2.7. Glass Fin – Podium Wall Unit Attachment

Initially two schemes were developed for the attachment of the glass fin assembly to the
curtain wall unit. As the base bid scheme a traditional “U” shaped stainless steel
extrusion clamps the glass at its horizontal edges from both sides with structural silicone
in between.
Figure 18: Base Scheme Attachment Glass – Podium Wall Unit – Stainless Steel U- Extrusion
Figure 19: Alternate Scheme Attachment Glass – Podium Wall Unit – Stainless Steel Embed
The alternate scheme also attaches only at the top and bottom of the glass fins, but
instead of the clamping action of the shoe, a stainless steel fitting is embedded into the
glass during the lamination process in the autoclave. This integrated approach may have
reduced the appearance of solid metal that was deemed distracting from the glass itself.
During the bidding process none of the three invited bidders wanted to entertain the
alternate scheme with the embedded stainless steel fitting and therefore the more
traditional shoe approach was selected with the potential to revisit the precise
attachment detail during the design assist process.
30
Figure 20: Three Variation Schemes: “J”, “L” and “U” Stainless Steel Extrusion [3]
31
Challenging Glass 3
After award of the contract Permasteelisa North America investigated this attachment
detail further together with the design team. Variations of the Shoe shaped extrusions
were investigated but eventually discarded in favor of the “U” extrusion (Figure 20).
2.8. Aeroelastic calculations and modeling

During the design development phase the issue of the glass fins starting to vibrate under
certain wind loading conditions related to the natural frequency of the assembly was
examined. Two independent computer finite element studies for the natural frequencies
of the glass fin assembly were undertaken to determine if vortex shedding, flutter and
galloping concerns could be ruled out for the design. Both studies returned favorable
results as the frequencies were calculated to be higher and the assembly to be stiffer
than what is generally considered as critical (2HZ). However, during the shop drawing
phase of the project the computer calculations will be confirmed by the contractor via
two 1:10 physical aeroelastic section models that will be tested in a wind tunnel setup
under expected wind loading conditions using material representing stiffness and
distributed weight of the full scale assembly.
Figure 21: FEA Study Natural Frequency Mode 1-3 [4]
32
3. Lighting
Developing a design scheme for the podium wall that would look engaging and
dynamic throughout the diurnal cycle cannot focus on daylight hours only. The same
design intents that apply to daytime, also apply to night time hours with its play of
(ambient) light and shadow. Therefore an (artificial) lighting scheme was developed that
would allow the base of 1 World Trade Center evolve into an ever changing light
choreography as the sun sets. The final lighting program is still under development, but
the system being proposed has a high degree of flexibility and programmability. Ideas
similar to giving resident artists a canvas to paint on with light at the podium have been
floated but no decision made on how to put to work a lighting system that can be
controlled in real time to a pixel size of 0,3 m x 1,52 m (1’ x 5’) in color.
The genesis of the overall podium wall unit design can to a degree be traced back to an
initial design decision to illuminate only from the inside. In fact the physical light
source itself is not visible at all in the podium wall unit and only its light effect
mysteriously illuminates the base of the building like a lantern. The only place to locate
the fixture symmetrically and economically was the center of each unit, dividing it
vertically into two sides with two sets of stainless steel slats and two glass fins, one on
each side. Evenness of the lighting was a primary goal and a max. 1:4 luminance ratio
across the perforated screen as the receiving screen specified. Furthermore the light
fixtures will be serviced from the front due to the lack of access from the back at the
lower podium section at the shear walls. For the purpose of access, a vertical snap cap
can be removed at each of the 3 stacked light fixtures per panel.
Figure 22: Plan Detail Lighting in Podium Wall Unit
33
Challenging Glass 3
Figure 23: Full Scale Lighting Mock-up Testing and Evenness Verification
The advent of LED lighting technology has revolutionized the possibilities of

illuminating buildings particularly in situations where allocated space for the fixture is
limited and performance requirements for optics are challenging. Both criteria are
prevalent on this podium wall, making LEDs the light source of choice in this
application. The custom LED fixture specifications on the project were developed by
Brandston Partnership Inc. (BPI) as Lighting Designer with Lumenpulse as a fabricator.
During the bidding process Color Kinetics/ Phillips was subcontracted to develop the
specifications further into a final fixture.
Figure 24: APG Mock-up At Night time
34
Figure 25: At Daytime close up

Figure 26: Suspended 150’ in the Air
35
Challenging Glass 3
4. Full Scale Proof of Concept Mock-up at APG

With the design and engineering task for the bid documents finalized, a visual proof of
concept mock-up was constructed by APG with the help of Lumenpulse. For glass
comparison purposes, two additional tower wall units were also added above the
podium wall. To evaluate the design at various heights off the ground, the mock-up was
built around a (12.5’ x 7.5’ x 33’) structural steel cage that could be elevated by truck
crane. Furthermore the mock-up was constructed with a glass fin attachment detail
allowing for various fin angles as depicted e.g. in Figure 25/26.
The mock-up was extremely helpful to the design team as it influenced further decision
making in many ways. It verified the design concept and gave key decision makers a
comfort level needed to proceed with this unusual design.
5. Outlook
With World Trade Center One about to top out in the summer 2012, fabrication and
installation of the podium wall panels will get on the way at about the same time if
performance testing goes according to plan. Installation of the panels is planned to last
about a year for the over 2000 typical panels with the last leave out panels to be
installed 2014/ 2015.
Figure 27: World Trade Center One, 20 April 2012
36
6. Acknowledgements
The success of any project depends largely on the expertise and enthusiasm of those
who direct it: First and foremost, I would like to thank Roger Duffy (FAIA), TJ
Gottesdiener (FAIA), Ken Lewis and Nicholas Holt for their leadership and guidance
heading the SOM efforts. I would like to take this opportunity to express my gratitude
to team mates Frank Mahan, Christian Kotzamanis, Andrea H. Wong and Scott
Matthews of BPI who have all made this design process memorable. Special thanks go
to Nicole Dosso, Donald Marmen and Benjamin Reich whose contribution was
instrumental for converting a design idea into a soon to be reality. Thanks also go to
structural engineers Charles Besjak and Dmitri Jajich of SOM for volunteering their
time at a critical junction of the design process.
Architects do not build their design; this huge task is left to the construction manager
and contractors who literally put it together. Everyone stepped up along the way to the
final design. Thank you Tishman Construction for managing the process and APG
International with Ed Zaucha, Dirk Schulte and Bob Unruh whose dedication to the
project was extraordinary. Thank you Permasteelisa with Alberto De Gobbi and team
for adding refinement to the overall design together with Phillips Color Kinetics and
Interpane. There was never a moment where I felt this project was treated as business as
usual!
Finally, I would like to acknowledge the ownership on the project, The Port Authority
of New York and New Jersey and the Durst organization for challenging us in a healthy
and productive way and allowing us to work on this once in a lifetime project.
7. References
[1] Rendering by Dbox
[2] Schulte, Dirk, Structural Analysis Podium Wall, APG International, 08-07-2011
[3] Drawings and renderings by Pemasteelisa North America Corp.
[4] Studies by Wacker Ingenieure
[5] Studies by RWDI Consulting Engineers
37
Challenging Glass 3
Projects & Case Studies

39
doi:10.3233/978-1-61499-061-1-41
The Glass Screens of the Japan Post Tower

Lucio Blandini, Werner Sobek
Werner Sobek Stuttgart, Germany, www.wernersobek.com
The Japan Post Tower is a 200 m tall building, rising on the site of the former
Tokyo Central Post Office in the Marunouchi business district, Tokyo. The
complex, designed by Helmut Jahn, integrates in the base area the restored historic
façade of the old Post Office. On the north side, the façade of the highrise is
characterized by three asymmetrical folds and by four glass screens framing the
unitized glass façade. Such screens have to reach a high degree of transparency,
while still being able to withstand tornados (up to 8.5 kN/m2) and high earthquake
loads (up to 1.2 g).
Keywords: Glass Screens, Highrise Buildings

Figure 1: Japan Post Tower - Architectural Rendering (© Murphy/Jahn Architects, Chicago)
41
Challenging Glass 3
1. Introduction
The redevelopment of the historic Tokyo central post office, designed by Helmut Jahn,
Chicago, is a 200 m high tower rising out of the renovated historic building. The tower
main façade, located on the north side, is characterized by three asymmetrical folds as
well as by four almost dematerialized glass screens, which frame the unitized glass
façade. The two external fold lines are inclined, and the middle fold line is vertical; thus
the façade surface is defined by a vertical plane (on the outer parts on the left and right
sides) as well as by two spatially inclined planes in the central part. The 31 m high top
screen is the most challenging high-transparency screen due to its size, the loadings and
the fold geometry. Further screens are the wings on the two sides (5.4 m wide and
129 m high) and a 6 m high screen enclosing the unitized glass façade at the bottom.
Figure 2: Japan Post Tower – North Facade (© MJS, Tokyo)
42
2. The Top Screen

The top screen (61 m wide and 31 m high) is designed to withstand tornado loadings up
to 8,5 kN/m2 and earthquake accelerations up to 120% of the earth gravital acceleration.
Nevertheless, it reaches a high degree of transparency, due to the fact that forces are
mainly transferred to the main steel structure (see figure 3, red profiles) by means of
tension and compression members, thus allowing for minimal profile sizes.
The glass panels (made of 2 x 12 mm laminated fully tempered glass) are 1,80 m wide
and have different heights varying between 4.55 m and 4.90 m. The glass dead load is
carried by means of welded tapered T profiles, which are hung at the top. The wind
loads are transferred at every level from the vertical T profiles to stainless steel tapered
wind needles ending with spherical hinges. The façade steel structure is braced by a
discrete number of diagonal needles and by a slender hollow steel profile running
horizontally parallel to the façade surface, in order to increase the facade capacity of
withstanding earthquake forces. Double hinge endings have been designed and
engineered for the nodes where two needles, a straight one and a diagonal one, converge.
All the arising vertical and horizontal forces are transferred by the needles and by the
tapered T-profiles to the main steel frame, which is set on a vertical plane located
behind the folded façade. The distance between the folded façade surface and the main
steel frame varies between 1.1 m in the middle and 3.6 m on the sides.
Figure 3: Top Screen - Steel Structure Rendering (© Werner Sobek Stuttgart)
43
Challenging Glass 3
Figure 4: Top Screen – View from outside (© MJS, Tokyo)

Figure 5: Top Screen – View from inside (© MJS, Tokyo)
44
3. The Wings
The wings are placed on the outer left and right sides of the north façade between level
8 and level 36. They are 5,40 m wide and have an overall height of 129 m. They are
subdivided in blocks of 5 to 6 levels; structurally speaking, these blocks are completely
independent of each other.
The horizontal loads are transferred from the glass panel to the T Profile in the same
way as in the top screen. These forces are transferred at every level from the T Profile to
the slab by means of a horizontal welded hollow steel profile, hinged at one end and
supported horizontally at 2/3 of the span by means of a tapered wind needle. The
vertical forces are transferred by the T profiles every 5 to 6 levels to three vertical
inclined tension rods. A fourth vertical tension rod secures the system at the bottom in
case of uplifting earthquake forces.
Tension rods
Hinged support
Figure 6: Wings - Steel Structure Rendering (© Werner Sobek Stuttgart)
Figure 7: Wings – View from below (© MJS, Tokyo)
45
Challenging Glass 3
4. The Bottom Screen

The bottom screen is 61 m wide and 6 m high. Horizontal loads are transferred by the
glass panels to the T profiles. These are hung on top and horizontally supported by a
tapered wind needle at the bottom. Diagonal needles brace the system horizontally. A
cable connects all the T profile, thus stabilizing the structure horizontally. The cable is
anchored on the two ends of the bottom screen to a stiff system made of two diagonal
and two horizontal wind needles and by a horizontal profile connecting them.
Figure 8: Bottom Screen - Steel Structure Rendering (© Werner Sobek Stuttgart)

Figure 9: Bottom Screen - View from outside (© MJS, Tokyo)
5. Conclusions
The four glass screens of the Japan Post Tower Tokyo demonstrate impressively how
light and transparent glass structures can be designed for regions with very high seismic
and wind loads.
46
doi:10.3233/978-1-61499-061-1-47
The Glass Sphinx: A Massive Stacked

Glass Sculpture
Freek Bos
Witteveen+Bos Consulting Engineers, the Netherlands, f.bos@witteveenbos.nl,
www.witteveenbos.nl
Tom van der Heijden, Pieter Schreurs
Scheuten Absoluut Glastechniek, The Netherlands, tvanderheijden@scheuten.com,
www.scheuten.nl
The refurbishment of the Meuse river boulevard in Venlo instigated Scheuten Glass
to donate a giant-sized, 6 metre high version of the stacked glass statue the Sphinx,
which had originally been made as a 80 cm sculpture to commemorate the city’s
650th anniversary back in 1993. Many hurdles had to be taken to, starting with the
preliminary feasibility study, on to adhesive selection, joint design, glass selection,
cutting methods, glass sheet layout, and final construction procedure.
Keywords: Stacked Glass, Adhesive Tape, Construction, Manufacturing
1. Introduction
In 1993, the city of Venlo, in the South Eastern province of Limburg, the Netherlands,
celebrated its 650th anniversary. For the occasion, the Venlo-based glass company
Scheuten had local sculptor Fons Schobbers produce a piece solely out of glass. The
result was the Sphinx, Figure 1a and b, a massive stacked glass object, in a free,
asymmetrical gate-like shape, approximately 80 cm high. The large round head on one
side stands on a slender leg, while the lower side across rests on a broad leg, with an
sort of arch connecting both sides.
Figure 1a and b: The original Sphinx sculpture.
47
Challenging Glass 3
Now, with the refurbished Meuse river boulevard completed and the decennial Floriade
World Horticultural Expo putting Venlo in the spotlight, Scheuten offered the city a
larger-than-life version of the statue. With a height of 6 m and a weight of well over
100,000 kg, the Glass Sphinx is a giant copy of the original. Figure 2 gives an indication
of the size.
Figure 2: Impression of the large Sphinx.
The Sphinx has been constructed on a prominent spot on a perpendicular corner of the
Meuse rive dike just south of the new yacht harbour and refurbished river front. It is
well visible from a major entrance road into the city, the historical city centre as well as
the new city district ‘De Kazerne’ on the other side of the river. The object is placed on
a pile supported concrete block foundation that cantilevers out of the dike by
approximately 0.5 m. It will be constructed in the summer of 2012.
Witteveen+Bos consulting engineers conducted a feasibility study to construct the

sculpture completely in stacked glass. Scheuten Absoluut Glastechniek took care of
detail engineering, manufacturing and construction.
2. Feasibility Study
2.1. General
A review of similar objects showed the Sphinx would be the largest of its sort in weight,
albeit not being the highest. Figures 3 and 4 show similar projects: the Police Memorial
in London [1] and the Angel sculpture in Zwolle (the Netherlands) [2]. These sculptures
show two different methods to obtain stability. The Police Memorial relies on vertical
prestress induced by five prestress rods. There is no intermediary material between the
glass sheets. Alternatively, the glass sheets of the Angel are connected by 3M VHB
48
The Glass Sphinx: A Massive Stacked Glass Sculpture
adhesive foil. This is more suitable for geometrically irregular shapes, but raises some
questions regarding durability, contamination, and crushing of the adhesive.
Figure 3: Police Memorial, London. Figure 4: Angel statue, Zwolle, the Netherlands
Potential problems that were to be expected included intrusion of water (freezing causes
glass breakage), dimensional inaccuracies in the seams, and deviations in glass
thickness of different glass sheets (EN 572-2 [3] for float glass allows ± 0.3 mm).
Literature [4] also reports differences in glass thickness across the width of the ribbon of
float glass (thin in the middle, thicker to the sides), measuring 0.3 to 0.5 mm.
Settlements, thermal expansion differences, and concrete shrinkage should also be

accounted for. Section 3 discusses how all of these issues have been addressed.
2.2. Preliminary FE-Analysis

Additional to the general feasibility study, an explorative FE analysis was conducted.
Three models of the Sphinx were analysed: a massive model (1), a layered model (2)
and a massive hollow model (3), Figures 5a, b, c. The first model served to gain a quick
insight in the structural behaviour and order of magnitude of stress. The effect of the
soft interlayers was investigated with the second model. In this model, layers of 10 cm
were applied (instead of 10 mm), to avoid long modeling and analysis times. For the
interlayer thickness, a tenfold was also applied. In the third model, a hollow sphere was
introduced in the Sphinx head. The aim was to reduce overall weight (material cost and
foundation loading), make the weight distribution between both legs more balanced and
also to create more interesting visual effects (a solid head would absorb a lot of light).
Wind loading was applied to all models as well as uneven settlement underneath the
legs. A temperature load analysis was made on model 1. The effect of step-by-step
stacking of the glass on stress distribution was analyzed using model 3.
The stresses due to wind loading remain extremely low, well beneath 1.0 MPa (Table 1).
In the layered model, the stresses were significantly lower than in the other two as the
soft interlayers allow for more distribution of the stresses.
49
Challenging Glass 3
Figures 5a, b, c: FE models for feasibility study: massive, layered, and massive hollow.
Table 1: Maximum tensile stresses in FE Models of feasibility study due to wind loading.
LC Description Massive Layered Hollow
1 Self weight [MPa] 0.11 0.16 0.11
2 Self weight + wind dir. 1 [MPa] 0.25 0.20 0.36
Uneven settlements may cause much more severe stresses. Table 2 gives the maximum
stress per length of uneven vertical settlement between the legs. Assuming an allowable
permanent stress of 8.0 MPa, it also gives the estimated allowable uneven settlement. It
ranges from 2.80 mm for model 1 to 8.79 mm for model 3. However, with a sound
foundation design, such uneven settlements should be avoidable.
Table 2: Uneven vertical settlement.

LC Description Massive Layered Hollow
Under slender Max tensile Stress/length

6 2.86 0.91 2.82
leg [MPa/mm]
Estimated max allowable
2.80 8.79 2.84
uneven settlement [mm]
The stacking process does increase the maximum stresses relatively significantly.
However, in absolute terms, they still remain very low.
For the thermal analysis a rather extreme situation was modeled as a starting point: a
temperature rise in the object from -20 ºC to + 80 ºC in 8 hours. This leads to a
maximum stress of 23 MPa. It is remarkable that even after 8 hours (28,800 s), only the
outer approximately 20 cm of the sculpture has risen above 0 ºC. The inner 40 cm (in a
120 cm wide section) remains at -20 º C (Figures 6a, b and 7). This means that in reality
the Sphinx will never come close to having that core temperature, but will rather have
one close to the local annual average, of about 10 ºC. When it is furthermore considered
50
that the actual Sphinx will be much more flexible (segmentation and layers) than the
massive model 1, there seems to be no reason to expect thermal breakages.
The results of the feasibility study were promising enough to continue work on the
development of the Sphinx.
Figure 6a, b: Temperature and stress distribution in Sphinx after 8 hours.
90
80
0sec
70 5.000sec
10.000sec
60
15.000sec
50 20.000sec
40 25.000sec
28.800sec
Temperatuur [ 0C]
30
20
10
0
2,900 3,100 3,300 3,500 3,700 3,900 4,100
-10
-20
-30
Y-Coordinaat [mm]
Figure 6a (top) and b (bottom): Temperature distribution in Sphinx after 8 hours.
3. Sphinx Design
3.1. Glass
For the Sphinx a transition glass is used, which is obtained when a float line oven is
switching from normal glass to extra clear low iron glass (Scheuten Super White). The
transition glass is almost color neutral. A very clear glass was selected as a normal iron-
containing glass would result in an almost opaque sculpture because of its size.
51
Challenging Glass 3
Furthermore, a darker glass would lead to higher thermal stresses as it would result in a
higher temperature gradient in the object.
3.2. Adhesive
AFTC Silver Tape 8502 was selected as adhesive for the sculpture. This is a transparent
modified acrylate tape with a nominal thickness of 0.25 mm. It consists of 100 %
adhesive, i.e. it has no carrier or filler. Table 3 provides some properties, taken from the
product data sheet [5].
Table 3: Some properties of selected adhesive foil, AFTC Silver Tape 8502 [4].
Property Code Value Unit
Thickness - 0.25 mm
Thickness tolerance - 0.1 mm
Density - 840 kg/m3
Peel adhesion ASTM D3330 2.6 N/mm
Tensile strength ASTM D897 0.75 MPa
Shear strength ASTM D3654 0.72 MPa
Solvent resistance - Excellent -
UV resistance - Excellent -
Temperature resistance, long term - 100 ºC
Temperature resistance, short term - 160 ºC
Several tests have been carried out to determine suitability of the adhesive. Standard
conformity testing according to ASTM D412, D897, D3654, D1002 and D3330 is
performed on the adhesive. Additionally, adhesive glass-glass connections were
subjected to heating at 80 ºC and visually inspected, but no defects were encountered.
Also, no sideways flow of the adhesive was detected, thus relieving worries that surplus
adhesive might spill out of the seams of the Sphinx.
The adhesive is applied to one side of the glass sheets (the smaller one respective to the
one it connects to), in the workshop using a custom made roller construction. Adhesive
foil rolls of 900 and 500 mm wide were used. The protective cover is removed on site.
The glass surface is covered completely by adhesive foil to avoid air inclusions,
contamination, or water (vapour). During curing, the adhesive slightly expands, thus
filling possible small gaps caused by deviations in glass thickness and driving bubbles
out.
3.3. Foundation-to-Sphinx Joints

The Sphinx will be placed on a concrete foundation block covered in unpolished black
stone slabs, similar to ones that have been used in the new yacht harbor. The use of
black stone makes the issue of differential thermal movements between foundation and
sculpture more important. These are accommodated by designing the connection of the
broader leg to the foundation as a sliding joint. Figure 8 shows the detail.
52
Cover plate (stainless steel)

Filler plate (stainless steel)
Top plate (stainless steel)
Teflon strips
Base plate (stainless steel)
Grout filling
M16 positioning bolts

Concrete foundation
Figure 8: Sliding joint between foundation and broad Sphinx leg.
The joint consists of a 20 mm thick stainless steel plate positioned on the concrete
foundation block using M16 bolts. The space underneath is filled with grout poured
through holes in the plate. 10 cm wide strips of Teflon are applied to the top of the plate.
Another stainless steel plate is placed on the Teflon strips and connected to the lower
plate by bolts in slots, so that they can slide relative to one another. A filler plate comes
next and then, finally a top plate is applied, 5 cm wider than the footprint of the glass.
From there on, the stacking of adhesive foil and glass sheets starts. The connection
under the slim leg is similar, but without the teflon strips and slots, so that it can not
slide.
3.4. Glass Sheet Layout

The size of the glass sheets that could be applied was limited for two reasons. Firstly,
small pieces introduce flexibility and dilations. Secondly, the sculpture is constructed by
hand, so the sheets can not be too heavy. Thus, each layer ended up consisting of 6 to 16
sheets.
A major concern in determining the glass sheet layout was avoiding over- and
underpressure in the seams between the sheets, as particularly underpressure attracts

moisture, dust and dirt, which could make the sculpture dirty on the inside. This has
been solved by interconnecting (almost) all seams through vertical channels and two
small openings, one in each leg (with a filter). In each layer, the glass sheets are rotated
with regard to the previous layer, to assure structural integrity.
As introduced in the FE study, a 2 m diameter hollow sphere was introduced into the
Sphinx head. This reduces the overall weight by some 16 %, from 111,813 kg to 93,725
kg.
The seams of the Sphinx are sealed from the outside with a clear MS polymer. This
sealant is more durable than clear silicone, and -more importantly- is compatible with
the acrylic adhesive foil. If water nevertheless penetrates through a seal, glass breakage
by stand still water freezing is avoided by designing the seams so that they grow wider
towards the Sphinx interior, from 8 to 25 mm. This also allows for some liberty in
dimensional accuracy of the sheets.
53
Challenging Glass 3
Figures 9a – e show the sheet layout of subsequent layers 334 through 338.
Figure 9a - d: Glass sheet layout of layers 334 - 337.
54
Figure 9e: Glass sheet layout of layer 338.
4. Manufacturing and Construction
4.1. Glass production

The glass for the Sphinx was produced in a single batch (some 7-8 hours of production
time) by her sister company f glass in Osterweddingen, Germany. Thus, all glass
sheets have an identical thickness, which is essential to avoid problems when stacking
layers that consist of multiple sheets of glass. The thickness across the ribbon was
measured. However, contrary to what has been reported in literature [4], no significant
differences were found. If they were even there, they were smaller than the adhesive
layer thickness (i.e. < 0.25 mm).
4.2. Cutting the glass panels

For shaping the glass panels, ‘ordinary’ glass cutting was selected. The artist preferred
the ragged, but clear glass edge that results over smoother translucent edge that would
result from laser- or waterjet cutting. However, cutting glass in these shapes proved to
be a challenge. Normally, plateaus are being carved in an optimal pattern by machine,
supplied with cutting oil and also broken by machine, in a fast pace.
However, the irregular curved sheets for the Sphinx had to be broken by hand after
being carved by machine. Personnel had to be trained specifically to do this properly.
Carefully breaking the glass costs a lot of time, otherwise a running crack will leave the
applied carve line, and go straight. Breaking glass slowly, on the other hand, causes the
cutting oil to dry out too soon. The cutting oil was therefore adjusted to stay fluid longer.
Also, additional cutting oil was applied to the carves during the breaking process.
4.3. Construction
All together, the glass sheets of the Sphinx provide an unsolvable puzzle. Therefore, all
sheets are numbered by layer and position. To obtain the Sphinx shape out of these
individual pieces on the building site, a number of vertical wood fiber boards are
erected. Their edges have been sawed to follow the Sphinx contour, derived from a
digital 3D model. The boards are positioned radially around two center points. In
between the boards, working floors have been constructed. The whole building site was
covered by a 20 x 15 x 7 m tent so that construction could take place free of weather
influences, humidity and contamination.
55
Challenging Glass 3
The construction procedure was as follows. Adhesive foil is pre-applied to one side of
each glass sheet, on the side where it will meet a larger sheet (i.e. the smaller sheet has
the foil applied). At the site, all sheets of layer y are initially placed on layer x without
removing the protective covering of the adhesive layer. Thus, the right position of each
sheet can be determined. Subsequently, markers are used to fix the position of each
sheet. Each sheet is then lifted individually, the protective adhesive covering is removed,
and the sheet is replaced at the right position using the markers. This process is executed
very carefully as sheet can not be removed after it is glued to the adhesive foil.
The size of the seams will make it possible to accommodate dimensional deviations,
which, due to the cutting process can easily vary between + and -3 mm.
Finally, the sculpture is brushed with glass powder (glass grinded to dust). The reason
for this is that the edges of the adhesive foil remain sticky and do not solidify. This
would attract dirt. The glass powder adheres to the adhesive foil edge and protects it
from moisture while simultaneously avoiding contamination.
Scaling a 80 cm art piece to a 6 m sculpture introduced a host of expected and

unexpected challenges. But it results, finally, in a fascinating landmark for the city of
Venlo.
5. References
[1] Smith, A., Mass Glass Structures, Challenging Glass 2, Delft, the Netherlands, May 2010.
[2] Nijsse, R., Glass walls carrying the roof and withstanding the wind load on the façade: Conservatory of
the Museum in Dordrecht and Raaks Glass Cube in Haarlem, Challenging Glass 3 (this conference),
Delft, the Netherlands, June 2012.
[3] NEN-EN 572-2: Glass in Building – Basic soda lime silicate glass products – Part 2: Float glass, July
2004.
[4] Wurm, J., Stacking of Glass –Structures and Sculptures, Glass Performance Days, June 2007.
[] AFTC High Performance Tapes, Silver Tape, Industrial 85 serie, Product Information 04.2010.
56
doi:10.3233/978-1-61499-061-1-57
The Apple Glass Cube: Version 2.0

James O'Callaghan
Eckersley O’Callaghan, United Kingdom, james@eocsd.com,
www.eckersleyocallaghan.com
Charles Bostick
seele sedak GmbH&Co.KG, Germany, Charles.Bostick@seele.com, www.seele.com
This paper describes engineering and fabrication innovation behind the rebuilding
of the Apple 5th Avenue Glass Cube in New York City. The original of which was
completed in 2006 and as a result of a number of years of glass fabrication and
connection innovation, with which the authors have been instrumental, a proposal
for a replacement structure embracing all the technology and techniques mastered
was enacted.
Keywords: Structural Glass, Connections, Lamination, Fabrication innovations
1. Introduction
The subject of the paper is renovation through innovation with structural glass.
In 2005 the companies to which the authors are associated were intrinsically involved in
the design and construction of a structural glass cube for Apple Inc at their high profile
5th Avenue store location. The design concept at the time was to find an enticing way
for people to be encouraged to shop below ground, a retailing challenge. This challenge
was addressed with architecture and solved by the architects Bohlin Cywinski Jackson
through the proposal of an all glass entrance building so iconic that it would encourage
curiosity and hence people to investigate its purpose. Ultimately over the following
years the products developed and sold by Apple have been so popular that the original
architectural intention has been overtaken by shear demand. However, this does not
diminish the success of the structure and the interest it has captured from New Yorkers
and tourists alike.
2. Why
Over the last five years Apple, EOC and seele sedak have developed a number of design
and fabrication technologies as a result of a drive to develop the language of glass
within the Apple retail environments. These developments have been both in the
fabrication techniques and the methods of connecting glass together. Significantly, the
production of much larger panels of glass has led to laminated and tempered panels now
being achievable up to 15m x 3.6m. The concept of metallic inserts being laminated in
the glass build up has also been advanced further to the extent that more sophisticated
details have been developed that facilitate entirely embedded joints. These particular
techniques were pursued and partially applied on other projects but it was decided that
these advancements in glass technologies would be most notably honed and illustrated
should there be an opportunity to rebuild the 5th Avenue glass cube. Hence, it was with
57
Challenging Glass 3
these concepts in mind that in 2010 both Apple and the architect began to investigate
what result these advancements would have on the design of the glass cube.
3. Concept
The original concept of a glass cube structure 10m in dimension was unaltered for the
revised version. Given the size parameters by which the glass panels are achievable it
was concluded that the dimensional split of each side into six could now be done in
three. In addition to this the length to which panels could now be fabricated resulted in
the need for only one over the height of the cube. Therefore it was possible to reduce the
number of panels per side from 18 to 3. Multiplying this theory over the entire cube
resulted in reducing the façade and roof glass panels from 109 to 20 in total, less than
20% of the original number required. A significant consequence of the reduced panels is
the reduced number of connections required, it being only necessary to have three per
vertical joint. This represented a massive reduction in fittings, which when combined
with the reduction in panels and subsequent joints resulted in a dramatically more
transparent glass structure.
The concept was quickly modeled and rendered for presentation and discussion with
Apple.
4. The Decision
The compelling model and renderings clearly illustrated that the advancements in glass
design and fabrication that Apple have supported over the last 5 years. The crystal clear
structure bereft of joints and fittings was a clear demonstration of the original goal of an
all glass cube. Ideally if it had been possible when the first cube was conceived it would
have been made from only 5 pieces of glass and hence this latest incarnation was
drawing closer that that ideal.
Apple and the design team deliberated as to whether the original cube should be
replaced and this debate called in to question the appropriateness of replacing an already
iconic glass structure with one that was to the untrained eye, much the same. There were
those who preferred the original because the increased amount of detail defined it more
clearly and related it better to the building it sits in front of. Ultimately it came down to
the CEO, Steve Jobs, to make the call and his thought process was typically crystal clear.
He made the point that if we had shown him the version we could now create when we
first designed it in 2005, he would have chosen it back then, so why would he not
endorse it now.
5. Existing Conditions
The existing conditions were familiar to us having created them in 2005. This was a
major advantage when reanalyzing the support structure for the new cube. At the time
of the original build we had not anticipated this latest move and as such the structure
supporting the cube had been carefully design to accommodate the point loads from 5
fins. The total structural weight of the new cube is actually marginally heavier because
the envelope glass is thicker given its larger span between fins. Therefore this combined
with the concentration of load from 3 rather than 5 fins caused a slight over stress in the
existing steel and a deflection greater than desired for the unbalanced loading conditions.
58
Apple Glass Cube: Version 2.0
A result of this being that the existing structure had to be reinforced for its new
conditions and a more complex approach to supporting the fins and glass panels needed
to be devised so as to even out these deflections over the width of the cube glass.
6. Detailed Design and Modeling

The structure of the cube is similar in nature to its original relative in that the overall
stability of the structure is maintained by the stiffness of the sidewalls. The main
difference other than the fewer elements is the fact that the roof now is only two single
spanning glass beams rather than a structural grillages as originally devised. The glass
fins support the glass roof beams at the front and rear of the cube, with secondary roof
beams spanning from the sides and between the primary roof beams (see Figure 1). The
roof plate consists of three cold bent laminated panels which achieve the minimum run
off for water and maintenance. The roof panels are bent to a subtle but complex two
dimensional curve embracing a technique that seele sedak has developed and mastered.
As with all glass structures beyond the concept design we pay particular attention to the
modeling of the connections in order to be confident of their behavior under different
loading conditions (see Figure 2). This is typically done by extracting the local
conditions from the global modeling in to detailed solid models. These give a more
accurate reflection of how the glass, interlayers and fittings interact.
Figure 1: Global Model Figure 2: Local Connection Model
7. Connection Advancements
The reduction of the number of glass units in the cube’s facades meant fewer glass fins
– two per side instead of five – which led to fewer façade to fin units connections,
which begged the question how could they “disappear” altogether. One of the key
reasons why the rebuilding of the cube was considered was the opportunity to
demonstrate the connection advancements that we had developed in concept and
worked with seele sedak to realize. We were very keen to take the idea of embedded
fittings to the next stage whereby they were used structurally and in an façade/building.
seele sedak had, by this time, already carried out a lot of testing on laminated inserts and
felt confident that the concept of using them to connect planes of glass together would
be successful. The challenge was to ensure that they connection of the planes of glass
with their respective laminated inserts could be achieved without seeing bolts of screws
59
Challenging Glass 3
from any visual aspect. We developed a detail that hollowed out the insert allowing a
metal tab to rotate info the insert from having been aligned with the vertical joint. The
rotation could be done through the joint itself and then once secured could be siliconed
over to cover the mechanics of the connection. This detail has worked perfectly and has
resulted in there being no fittings on any side of the cube protruding from the face of the
glass itself. This adds real magic to the structure and results in perfectly reflective and
flat surfaces on each face of the structure (see Figure 3).
Figure 3: Insert Fitting Figure 4: Roof Structure
8. Comparison
In 2005-2006, Eckersley O’Callaghan structural design and seele sedak designed and
built the first glass cube (see Figure 5) – a 10x10x10 meter self supporting all glass
structure with laminated glass columns that doubled as vertical beams to take wind load
upon which a glass beam roof grid sat, braced by the façade and roof panes against
racking and twisting. Fabrication of the laminated columns, beams and façade panes in
2005 stretched the capabilities of glass processing technology in size and quality.
Figure 5: Original Cube Structural Density Figure 6: New Cube Structural Density
In 2010-2011, Eckersley O’Callaghan structural design and seele sedak were asked to
rebuild the glass cube with larger units, fewer columns and beams, per the latest glazing
technology (see Figure 6). Between the two cubes, lay 5 years of development during
60
which more than forty all glass structures of all sorts were executed, each project
pushing the limits of glass construction a bit farther. The advance in technology can be
seen in part in the following comparison of the two glass cubes (see Table 1):
Table 1
Building Part Cube 1 Cube 2
columns 5 per elevation x 4 = 20 2 per elevation x 4 = 8
façade panels (incl. doors) 72 12 + 2 doors + 2 side lights = 16
roof beams 25 @ 3.3m + 10@ 1.6m = 35 2 @ 10m + 7 @ 3.3m = 9
roof panels 36 3
entrance canopy 1 1
Subtotals 109 panels 20 panels
20 fin columns 8 fin columns
35 beams 7 beams
Total 164 glass units 35 glass units
Less than 22% of the number of glass units were used in the new Cube.
This list however does not explain why less units were necessary.
Nor does it explain how it was possible to use a lesser number of glass units.
Following are the various building parts and their technological advances.
8.1. Larger Façade Panels

The original cube had 10m x 10m façades using glass units 6 across and 3 high of 3x
SGP laminated glass in approximately 1.66m wide x 3.33m high unit size. When the
original cube was fabricated there was no autoclave for laminating glass in lengths of
more than around 8.4meters. In addition, the use of smaller panes allowed more
movement and less stringent deflection requirements for the overall structure. For the
10m high glass fin columns, the use of an autoclave from an experimental aircraft
company where carbon fiber wings were laminated had to be arranged, a not so perfect
cobbled manufacturing solution. The reason for the formation of seele sedak in the year
2005 was to fill this manufacturing void. An autoclave specifically made for laminating
glass in maximum 12 meter lengths and soon after another autoclave for 15 meters
lengths have been the heart of seele sedak from the beginning. Five years of experience
laminating all sorts of glass, different interlayers, extra jumbo sizes and extra thick
laminations provided the basis for the engineering and fabrication of the Cube 2
oversize façade panels in 3.280m width x 10.30m height. This was a long process of
testing, experience in doing oversize laminations, adjusting processes and slowly
improving quality, the results of which were a defined process for maximum high
quality, almost bubble free, large size laminated glass that eventually was given the
brand name, “glascobond”®.
61
Challenging Glass 3
Figure 7: Jumbo glass
Figure 8: “Woven” glass beam grid Figure 9: Simple glass beam structure
8.2. Longer, fewer roof beams

The structural concept of the first roof was of a “woven” glass beam grid or “grillage”
that sat upon the glass fin columns with pin connections leaving the bracing of the cube
against parallel racking and twisting to the façades (see Figure 8). The “grillage” was
achieved with shorter 3.30m long beams connected to the middle of a traverse beam of
similar length with moment stiff connections achieved thru specially milled stainless
steel “saddle” connectors. This lent an economy as well as a bit of redundancy
providing safety thru many small glass members, no great single spans and little roof
load per member - still a monumental achievement at the time to do a 10m span.
The new cube actually achieves the span with a simple 10m spanning beam braced
laterally at the third points with shorter glass members (see Figure 9). The advance here
is the experience in fabricating and having tested a glass beam of this length with the
proper end connections that can carry more than 30 square meters of roof area safely.
62
8.3. Larger roof panels

On top of the “grillage” glass beam roof structure sat, in six by six rows, 36 laminated
glass units, each 1.66m x 1.66m, with corner patch fittings (see Figure 8). What few
people realize is that the roof was actually a pyramid with minimal slope in order drain
rainwater. A pyramid normally has visible, diagonally running, corners but in the
original cube these were made ‘invisible’ by heat bending the units at the diagonals with
a fold to create the pyramid. In the new cube the new structural concept and large unit
manufacturing capability demanded that there be only three roof units, each
approximately 3.3 meters wide and 10.0 meters long (see Figure 9). A heat bent
pyramid shape was out of the question due to the units’ size so, cold bent lamination
was used to give the roof a domed shape – this means that the panels have a double
curved form – a first in that size of unit.
9. Fabrication
The glass units in the New Cube are all larger which means that structurally they have
to span more. However the existing steel did not allow much more glass weight, ergo
not a lot more glass thickness. So the new glass build up selected was 3x 12mm fully
tempered glass as opposed to the original 3x10mm heat-strengthened units. A new
tempering oven installed in the early part of 2011 allowed seele sedak for the first time
to temper the glass in house allowing better control over the end result. The façade
panes at 3.28 meters (!) wide were a bit of a challenge but the real hurdle was the
tempering of the outside roof panels that hang over the façade units, just ever so slightly,
meaning that they needed to be more than 3.28 meters in width. After trials, the
fabricating machines decided the final width would be 3.295 meters.
In addition, the lamination of the fittings into the units, as well as keeping the laminated
units bubble free and flat, posed problems that eventually thru trial and test were solved.
10. Installation
With the development of large glass unit fabrication came the need for large glass
installation technology. The first large glass units requiring new expertise in installation
were fabricated for the Apple Store in Sydney, Australia. Large size vacuum packs that
can handle the dimensions but, also the several tonnes weight of such units were
specially designed and made. The installation crews practiced on one of the first 13.5m
x 3.05m three times laminated units – picking it up in its horizontal position from an
imaginary flat bed truck, raising it and tilting it upright (see Figure 10). Since then,
large glass units for facades and fins have been installed in the retail store on Broadway
in Manhattan as well as in two stores in Shanghai, one in Hong Kong and a few others.
The new Cube presented a challenge in that installation was only allowed at night.
Cranes could not be left standing during the day limiting their size – the set up and take
down times of a large crane would not allow time for installing glass. In addition the
store was not to be closed so that two construction enclosures had to be built – one to
the inside and one to the outside, the outside one being operable, opening up its top at
night to allow material access by crane only from above (see Figure 11). As usual a
“template” scaffold was installed. This has pre-surveyed and positioned guides for all
large glass members so each unit, weighing from one to two tonnes, can be simply
dropped in place – a necessity when the construction tolerances have to be plus/minus 1
millimeter! During construction, it appeared that we may have been taunting the gods.
Right before the first fins were to be installed, there was an earthquake on the East
63
Challenging Glass 3
Coast causing a hasty survey check of the template scaffold to make sure it was still
aligned. After a good portion of glass was installed there came tail winds from a
hurricane that closed the construction site for a few days and finally, just days before
opening, a snow storm dropped the outside temperatures, slowing down silicon jointing
work. In the end it all was done, on time, beautifully.
Figure 10: Practicing lifting + turning Figure 11: Placement of the large roof units
64
11. Cube Version 2.0

The new Apple store on 5th Avenue was unveiled on November 4th, 2011, after the store
that never closes, closed for a day, to a welcome reception (see Figure 12).
Figure 12: Completed
12. Acknowledgements
We would like to thank the people on the project teams at the architects, Cywinski
Bowlin Jackson, the seele Group partner in Manhattan, seele Inc., as well as our
colleagues at Eckersley O’Callaghan and seele sedak without whom both of the cube
structures would not have been possible. It almost goes without saying that Apple
Corporation, through their vision, backed with their commitment to excellence, and
contributions in time and money as the instigator of both cubes earns our deepest
gratitude. The cube has become a symbol of Apple to the point that it is Apple – a well
earned badge.
65
doi:10.3233/978-1-61499-061-1-67
Future Application of
Structural Use of Glass
Thomas Henriksen
Waagner Biro, thomas.henriksen@waagner-biro.at
The structural use of glass has been explored extensively over the last decade,
especially with Apple stores where the boundaries for their application have been
advanced significantly. This paper reviews the history and also the state of the art in
the structural use of glass. The aim of this paper is to discuss a way forward in
relation to glass as a structural material, so that future innovations are not driven
solely by the ability to produce and laminate large pieces of glass. Future
challenges lie in having a unified understanding of glass as a structural material,
high-lighting current limitations when designing with structural glass in relation to
current codes in different parts of the world. A continued dialogue on the wider
applications of structural glass will advance the state-of the art beyond its current
range of use.
Keywords: Glass, code of practice, Residual Stress
1. Introduction
Contemporary structural use of glass dates back to the Victorian greenhouses [1], in
particular the Bicton greenhouse for it’s minimalistic and transparent structure. The
greenhouse was erected in approximately 1830 and is the one of the few preserved
structures, where the glass is part of the structural system. The glass is located in
between thin steel elements. The interaction between the glass and the steel ensures
structural stability. The green house is shaped as a half dome and is then supported to
the north by a brick wall. In Figure 1 the Bicton greenhouse viewed from outside and
from inside.
Figure 1a, Figure 1b. Bicton green house.
67
Challenging Glass 3
The many small glass rectangles allow sufficient redundancy in the half dome if some
pieces should break. This is one of the reasons why it still remains intact today. The
question which arises is; why do similar contemporary glass structures not exist today?
One answer could be that these structures were developed based on trial and error, using
a very labour intensive construction method. With increased requirements from safety
regulations and the necessary detailed documentation of the structural system it has not
been possible so far to build similar structures today. For more simple structural
systems it has been possible to advance the structural use of glass. This has enabled an
increase in the structural use of glass over the last two decades. Early pioneering work
in the structural use of glass has been done by Rob Nijsse [2] and Duwhurst MacFarlane
[3]. The most significant development is evident in the many Apple flagship stores
around the world engineered by Eckersley O’callaghan [4]. Figures 2a and 2b show the
Apple glass staircases in Amsterdam and in Hamburg, respectively.
Figure 2a. Apple store in Hamburg Figure 2b Apple store in Amsterdam.

Photo EOC
One of the greatest challenges which the designer face when realising such glass
structures is local building control, and acquiring the required building permits, since
the local building control often has limited experience in dealing with structural use of
glass. At the same time the lack of a unified code of practice for structural use of glass
is also globally recognised. The building controls are therefore often reluctant to give
the approval without asking for excessive 1:1 glass tests to be performed. The Apple
stores have to some degree helped with this situation since all the stores are situated in a
public domain; the performance requirements for the glass structures were more
onerous. It is therefore to some extent now easier to convince the building control of a
new glass project by referring to existing Apple stores. The presence of other Apple
stores, worldwide have ensured meaningful discussion with local building control in
that city and that experience forms the main topic of this paper. Is it possible to find a
way forward towards a unified global design method for the structural use of glass?
This paper will discuss the outstanding issues in terms of stress in the material, safety
factors and combinations of loads over different time durations.
68
Future Application of Structural Use of Glass
2. Design methodology for structural use of glass

The current design methodology for the structural use of glass is mostly represented by
3 different approaches; American, European and Australian. The American code [5] and
the Australian code [6] mainly refer to glass as an infill panel in a primary structural
system and currently not intended for the structural use of glass, but in Europe there is
no consensus for a Eurocode for glass, as is the case for steel and concrete. A future
code of practice for structural use of glass is planned. There exists a draft version for an
EN Standard [7] that addresses glass as primary structure; however this standard has not
been released in its current version. Widely recognised are the German TRLV [8] and
the TRAV [9] for determining the allowable stresses in glass, but again these do not
address glass as a primary structural element and refer to infill panels. On this basis it is
understandable that building controls have difficulties in accessing structural
calculations for structural use of glass, and demand 1:1 testing. It is important to state
that it is outside the scope of this paper to critically appraise the approaches in the above
mentioned design recommendations, but as stated in section 1, to highlight the
remaining issues that need to be addressed to archive a coherent design methodology for
the structural use of glass.
2.1. Stresses in heat treated glass

Stresses in heat treated glass, toughened (fully tempered) or heat strengthened have in
the past been one of the main points for discussion, since the residual compression
surface stress (RCSS) in heat treated glass could vary depending on the thickness of the
glass, the tempering line, operation of the tempering line and the ambient temperature of
the air used in the quench. In the past the recognised quality assessment of the residual
stress in a thermally toughened glass was the destructive testing of a 360mm x 1100mm
glass sample [10]. Current research by Tallin University of Technology [11], shows that
RCSS in a thermally toughened glass may be measured without having to perform a
destructive test. This makes it possible to determine the actual RCSS of a heat treated
glass pane, and it is then possible to lower the safety factor, which is added to the
assumed characteristic RCSS. The safety factor is added to compensate for the
uncertainty of the actual RCSS in the heat treated glass today, without measuring the
RCSS.
It is suggested by the Author to make it mandatory to perform stress check of all glass
panels used as primary structural elements, and that the glass panels have the measured
residual stress printed on the glass.
2.2. Safety factors for glass

Safety factors exist to reduce any risk of material failure and such factors are usually
added to the material by reducing the characteristic stress to an allowable design stress.
It is common to link the glass codes to the same country’s code of practice, e.g. the
prEN 13474 [7] is linked to the Eurocode – Basics of structural design [12]. Designing
with glass according to the Eurocode means that a safety factor is added both to the
glass, a material safety factor and a similar safety factor is added to the load. The safety
factor on the material is necessary because of the determination of the strength of the
annealed glass. If the strength of a heat treated glass pane was determined separately
from the residual strength then it would probably be possible to reduce the safety factor
which is added to the material side of glass. The strength of annealed glass is currently
69
Challenging Glass 3
the cause of much debate in advancing towards a European structural glass code. The
topic has been discussed by Fred Veer [13].
To progress from the issue of the assessment of the strength of annealed glass is the
Author of this article suggestion to ignore the annealed strength and allow the glass to
work within the RCSS alone. This would, in theory, mean that the surface cracks in the
glass would never open, except from accidental loads, which are not accounted for in
the assessment of the RCSS, but by laminating several glass plys together. This could
be a short term solution to the assessment of the strength for annealed glass in relation
to structural design with heat treated glass. With the possibility to measure the exact
residual strength of a heat treated glass as mentioned in section 2.1, it would be possible
to lower the safety factor currently added to heat treated glass.
2.3. Time dependent strength of the glass

The strength of annealed glass is dependent on the time duration of the load. In the most
simple cases this is because the surface cracks in the glass open slowly over time and
thereby reduce the strength for long term loads [14][15][16]. The time dependent
reduction factor for glass can be described by [17]:
1

kmod 0,663t 16
Where t is the load duration in hours. In the prEN13474 [7] the factor kmod has a
maximum value of kmod = 1 and a minimum value of kmod = 0.25. Interestingly enough,
the kmod factor does not comply for short term loads, wind gusts under 1.2 sec, blast
loads, or impact loads. The corresponding strength of annealed glass for time durations
of less than 3 seconds is not determined. Jens Schneider have investigated the strength
of glass for impact loads [18], but besides his research work very limited information
exists for short term loading with a duration of less than 3 seconds.
The time dependent strength of the glass is valid area for debate. If the glass is heat
treated the issue of static crack propagation over time can be neglected, since the RCSS
prevents the surface cracks from propagating. However, in the case of impact loads the
residual strength from the tempering process might not be sufficient to withstand the
magnitude of the load and it is necessary to take the additional annealed strength into
account.
2.4. Combined loads with different load duration
Load combinations of different load duration (wind load, snow load etc.) are normally
governed by the code of practice from the different code systems. Usually different
safety factors are added to the loads, dependent of which load is the governing load case.
For most engineering materials this is not problematic since the strength of the materials
are not dependent on the load durations unless it is a dynamic load. For glass; the load
durations as mention above have a significant impact on the strength. Therefore it is
important to understand how to combine loads of different duration, and at the same
time determine the correct corresponding allowable stress in the glass. This topic has
been discussed frequently over the last decades and is evident in the different revisions
70
to prEn 13474 [7]. Mauro Overend [19] suggested in 2010 the use of a stress-history
interaction equation. The equation takes into account short term loads, (wind loads),
medium term loads, (snow loads), and long term loads, (dead load). This argument is
based on the long and medium term loads causing static crack growth in the surface
cracks, thereby weakening the glass. In the Cost C13 JM Andersen [20] presented
results showed that, loads with different load durations acting together, cannot directly
be linked, and that especially glass under wind loads did not have a time dependent
reduction in the strength. However it also stipulated that the number of tests were
limited, and that further enquires was necessary. It is the author’s opinion that these
considerations only apply to annealed glass where no RCSS are in place to force the
surface cracks to remain closed. If heat treated glass is used, it can be assumed, for the
purposes of design, that the strength is not dependent on the load duration, if the heat
treated glass is designed in a way that the residual stress is not overcome in bending.
This would ensure that the surface cracks always remain closed. This would be valid for
most design cases. But there still exist situations where this design assumption cannot
be fulfilled and the additional annealed strength has to be considered. This was the case
for the design of the work of art by Olafur Eliasson ‘Your Rainbow Panorama’ in
Aarhus, Denmark [21], shown in Figure 3a & 3b.
Figure 3a and 3b. “Your Rainbow Panorama” Work of Art by Olafur Oliasson, Aarhus, Denmark. Photo SOE.
In this work the glass walls are the primary structural elements. In this instance the issue
of combining loads with different load duration became relevant. The local municipality
requested that the load bearing glass walls be dimensioned for; self-weight, wind load,
snow loads and an impact load according to En12600 [22] acting at the same time. All
the loads have different time durations. The sizing of the glass would be possible
according to the TRAV [9] if the impact load was not considered. But the coincident
impact load adds an additional tensile stress in the glass pane so the allowable stress in
the glass is exceeded. Normally in this situation it is necessary to perform a 1:1 test to
prove that the glass has sufficient capacity to accommodate the loads with different time
durations but it should be possible to design with glass without having to do such tests
each time.
71
Challenging Glass 3
3. Designing with structural use of glass

Designing the structural use of glass is challenging given the lack of guidelines or codes
of practice. Local building control often follows guidelines that are not up to date
compared to the current research for contemporary built glass structures. Besides the
necessary structural requirements, the safety aspects of glass structures also need to be
considered. This aspect is rarely covered in the guidelines. This is mostly because safety
is linked to local governmental building regulations. The topic has been discussed by
Stephan Niderehe [23] and Freek Bos [24]. The safety aspect often leads to additional
glass layers being added to ensure redundancy in the structure if one layer should fail.
Today most glass fins and beams are designed with a minimum of 3 layers of glass
where 1 of the 3 layers ensures redundancy in the structure, which means that only 2
layers is necessary structurally.
When designing with glass most of the design possibilities are governed by the
limitations to the manufacture of glass. Post fracture integrity of glass structures is also
important to minimise the risk of failure. Post fracture integrity can be ensured by
laminating heat strengthened glass with toughened glass. The difference in break-
patterns will ensure that the panel will be self-supported until temporary measures are in
place.
The size of glazing has long been governed by the maximum dimensions of “Jumbo”
sized glass, 6m x 3.21m. This is still the most economical size of glass and any building
grid conforming to this size has less waste glass from the cutting process. Additionally
most Post-processing plants are only equipped to handle these sizes or smaller. If the
sizes are above the standard size of 6m x 3.3m, which is the case in most Apple stores,
then far fewer suppliers is available. The overall size of individual glass sheets becomes
an issue when considering the control of the continuity of the residual stress in the heat-
treated glass.
Figure 4a. Police Memorial, Fosters & Partners. Figure 4b. Glass Bench intended for Plantation place.
London.
72
If glass with holes and notches are used in structural design it is important to follow the
guidelines in terms of edge distance to ensure that that the parts get cooled correctly in
the quench. Laminated glass is always used when designing with glass for structural
use. But the understanding of the interlayers’ ability to transfer shear loads under
different temperatures is not fully covered by any guidelines. Much research has been
undertaken in Germany on this topic; the works of Wellershof [25] can be used as a
reference for design purposes. The research only covers pvb as an interlayer limited but
additional research has been undertaken for Sentry glass interlayer.
There are many more aspect which are important to understand when designing for
structural use of glass dependent on the application. With lack of extensive experience
in the topic it is recommended to undertake a 1:1 test of the glass structure to ensure
capacity and the post breakage behaviour.
4. Conclusion
In this paper the current development regarding the structural use of glass has been
discussed. A design methodology for the structural use of glass which alone utilizes the
residual strength in heat treated glass, has been suggested. This would minimise some of
the issues in understanding the brittle behaviour of glass. This, together with a clear
redundancy scheme, would enable a code of practice for structural use of glass to be
drafted. It is clear that workmanship is still an important issue in maintaining the
robustness of glass structures. Also understanding the stress distribution in glass
structures, especially when glass elements are mechanically connected via bolted
connections, is necessary. The emphasis needs focus on the quality checks of the
residual stress in the glass, but as already mentioned will depend on existing equipment
capable of reliably measuring this residual stress. It is necessary to conduct further
research into understanding the issues mentioned in section 2, to be able to advance
towards a comprehensive code of practice.
5. Acknowledgements
The Author would like to thank Dr. Stephen Lo and Dr Stephen Morse for reviewing
this paper.
73
Challenging Glass 3
6. References
[1] Koppelkamm, Stefan, Glasshouses and Wintergardens of the Nineteenth Century, Rizzoli International
Publications, INC, NY, USA, 1981.
[2] Nijsse, Rob, Glass in Structures. Elements, Concepts, Designs, Birkhäuser, Basel, Switzerland, 2003.
[3] Structural use of Glass in Buildings, p 75. , IstructE, Dec 1999
[4] O’Callaghan James. Thinking big with Structural Glass, GPD proceedings 2009, Finland
[5] ASTM 1300-9a, Standard practice for determining load resistance for glass in buildings, June 2009.
[6] AS 1288-2006, Glass in Buildings-Selection and Installation. Jan 2006
[7] prEN14373-3, June 2008, Glass in building - Determination of the strength of glass panes- Part 3:
General method of calculation and determination of strength of glass by testing.
[8] TRLV, Techniches Regeln für die Vervendung von linienförmig gelagerten Verglasungen. Techniche
Report, Mitteilungen des Deutschen Instituts für Bautechnick, Berlin, 2006
[9] TRAV, Technische Regeln für die Verwendung von absturzsichernden Verglasungen, Fassung Januar
2003.
[10] EN12150-1:2000, Glass in buildings - Thermally toughened Soda lime Silicate Safety glass.
[11] Anton, J, et al. On the inhomogeneity of Resiudal Stress in Tempered Glass Panels, GPD, Finland, 2011
[12] EN 1991-2005, Action on Structure.
[13] Veer, Fred, et Al. The strength of architectural glass, Challenging Glass 1, 2008
[14] Griffith, A.A, The Phenomena of Rapture and flow in Solids, Philosophical Transactions of the Royal
Society of London. Series A, Containing Papers of Mathematical or Physical Character, Vol. 221 (1921)
pp. 163-198.
[15] Charles, R.J. Static fatigue of glass. I, J. Appl Phys., 1958, 29, 1549
[16] Charles, R.J. Static fatigue of glass. II, J. Appl Phys., 1958, 29, 1554
[17] Brown, W.G., A practicable formulation for the strength of glass and its application to large plates,
Publication No. NRC 14372, National Research Council of Ottawa, 1971.
[18] Schneider, J. ‘Festikeit und Bemessung punktgelagerter Gläser und stossbeanspruchter Gläser“PhD.
Thesis, TUDarmstadt, Institut für statik, 2001.
[19] Overend, Mauro, Recent development in design methods for glass structures, The Structural Engineer,
IStructE-Journal-issue-88, London, UK, 2010.
[20] Andresen J.M. et A, Proposal for a Code Calibration Procedure, EU COST C13, Glass and Interactive
Building Envelops IOS Press, 2007
[21] Henriksen T: ARoS, Your Rainbow Panorama, .GPD proceedings, Finland, 2009.
[22] EN12600:2002, Glass in buildings – Pendulum test – Impact test method and classification for flat glass.
[23] S. Niderehe, Glass structures and their structural reliability – a discussion, Challenging Glass 1, 2008
[24] Bos, F, Safety Concepts in Structural Glass Engineering, Towards an Integrated Approach, PhD
TUDelft, Dec 2009.
[25] Wellershoff, F. Bemessungsschubmodulwerte für Verbundglasscheiben, Stahlbau 76, 2007, Heft 3.
74
doi:10.3233/978-1-61499-061-1-75
A Laminated Glass Wall Will Protect

Warnemünde From High Water
Frank Heyder, Franziska Paulu
hpl-Ingenieure, Germany, heyder@hpl-ingenieure.de
Warnemünde is a former fishing village on the Baltic coast, now part of the city of
Rostock. A new flood protection wall is due to be built along a river in an
architecturally sensitive inner-city area. Transparent or movable solutions are
necessary, while the barrier must withstand water, flotsam, ice and the impact of
boats of up to two tons weight. Here the optimum solution is a 4-layer laminated
glass wall. The article describes the research required to establish realistic impact
loads (via transient-dynamic finite element analysis), the safety concept and the
applied design criteria for glass sections.
Keywords: flood prevention walls, laminated glass, flotsam and boat impact,
transient-dynamic finite element analysis
1. Planning objectives
Flood protection structures at the Baltic Sea are required to protect against wind-
induced seasonal flooding rather than tidal floods. The typical flood season is winter.
An existing concrete flood protection wall is no longer considered stable or sufficiently
high, due to an increase of predicted peak water level. The client brief for the new wall
was challenging: The flood barrier must not hinder the current usage of the quay for
boat moorings, and must not disturb the view from adjacent houses to the river and vice
versa. After a flood warning there is very little time and only a limited workforce
available to close temporary openings or to erect mobile walls, and therefore mobile
elements must be reduced to a minimum or made fully automatic. All solutions must
also function in winter under freezing conditions or after heavy snowfall. There is only
a narrow strip of land available for any flood prevention construction between quay and
street. The waterfront architecture is of historic interest and must not be spoiled by
technical constructions.
2. Feasibility studies, alternatives, comparison

The following solutions have been considered: moving walls (flaps, miter gates,
elevating walls) as a permanent mechanical solution; removable walls, which are
erected only when a storm flood has been forecast, but with a permanent sub
construction and coupling points in the pavement; and, finally, rigid walls. All three
solutions have drawbacks: moving walls may fail to work in severe winter conditions,
and need plenty of maintenance. The removable walls require more time and many
workers for erection. Rigid walls are less complex and require comparatively little
maintenance, but can greatly disturb the surrounding architecture and block riverside
views if not transparent. Thus, the optimum solution is a combination of all these
75
Challenging Glass 3
alternatives: Rigid concrete for the base, rigid-transparent (glass) in the upper part of the
wall where transparency matters most, and mobile walls for openings, giving access to
moorings and the quay.
Figure 1: Proposed solution with a glass-topped wall, left [2]
3. Optimum solution
The optimum solution for the flood protection wall is an integrated approach which
solves both the technical issues and the objectives for maintaining Warnemünde’s
architectural unity. The technical and optical functions go hand in hand: The glass
balustrade protects the historic city against high water, while ensuring an unobstructed
view in both directions from the small fishing houses to the water. The wide quay is not
only designed for mooring, but also creates a new space for tourists and inhabitants – a
pedestrian zone at the water’s edge which serves as a small harbour for recreational and
fishing boats. The formerly dreary embankment becomes an attractive part of the town.
The concrete lower section at the water’s edge is constructed as a sheet pile wall. The
upper section is a glass-steel-construction, which not only protects against flooding, but
also serves as a transparent balustrade for pedestrians. Gates at various stations offer
barrier-free access via stairs or ramps to the lower section.
Figure 2: Future view from the river Alter Strom to the storm flood protection wall [2]
76
A Laminated Glass Wall Will Protect Warnemünde From High Water
The laminated glass wall consists of 4 layers: 2 inner layers of heat-strengthened glass
(TVG) and 2 outer layer of fully tempered glass (ESG), with 3 layers of foils in-
between. Additional exterior foils create a no-scratch coating. The glass construction is
shown in Figure 5. The posts (Figure 3) and the handrail bar (Figure 4) are made of
stainless steel, elastically embedded in the concrete construction.
Figure 3: Cross section of lower beam Figure 4: Handrail design
Figure 5: Cross section of laminated glass

Figure 6: Cross section of flood protection wall as optimum solution, compared with current situation
77
Challenging Glass 3
4. Structural design
4.1. Safety Concept

The proposed wall is one element of a wider storm flood safety scheme for the entire
coastline of Mecklenburg-Vorpommern, and for Rostock-Warnemünde in particular
[1,2]. A special safety concept was required for the laminated glass wall due to the
special fracture behaviour and stability of glass structures. The concept assumes that
when any glass section shows signs of distress, such as cracking, the distressed section
will remain watertight for sufficient time to allow a temporary stoplog to be installed. A
first step was to determine the behaviour of the laminated glass for typical loads (wind,
hydrostatic, flotsam as static force) according to codes of practice to ensure the usual
level of building safety. In a second step, a probabilistic risk assessment investigated the
probability of glass failure under the impact of larger flotsam and abandoned boats, for
which no codes of practice exist. This allowed an estimation of the annual costs for
glass replacement and the number of stoplogs and workers required to cope with heavy
flooding. The proof of water tightness of broken laminated glass is part of the
experiment described in Section 5.
4.2. Loads
To calculate the impact loads which the laminated glass walls will have to bear, the
effect of insufficiently moored boats striking glass sections was investigated. First, the
100 boats currently moored at the quay were listed and classified. Eight typical
categories of boats where modelled in Strand7 FEA-Software to calculate the typical
stiffness of the boat hull. The mass was taken from known examples. The average
impact velocity and wind loads were calculated using design parameters typical to boat
construction.
Table 1: Loads of various boat categories

Boat category Stiffness [N/m] Velocity [m/s] Mass [kg]
Heavy boats
1 with build-up 1.8 x 106 2,50 2000
without build-up (wood) 2,00 2000

7
2 with build-up 3.5 x 10 2,50 2000
without build-up (steel) 2,00 2000
6
3 with build-up 1.0 x 10 2,50 2000
without build-up (composite) 2,00 2000
Lighter boats
4 with build-up 1.8 x 106 2,20 1000
without build-up (wood) 1,70 1000
78
4.3. Calculation method

The numeric simulation of impact was calculated using an FE-model in Strand7. All
elements (glass wall, steel posts and bars) are plate elements. The posts are elastically
embedded in the wall’s foundation.
Figure 7: Model laminated glass wall with elastic posts in the ground
and handrail as crossbar with Impact 1 situation
The calculations were carried out as a nonlinear transient-dynamic FE analysis for a

hard impact. The model has three different impact situations:
x Impact 1: Impact in the centre of the glass section (area 50 cm x 50 cm)

x Impact 2: Impact at the top of the post
x Impact 3: Impact in the middle of handrail (crossbar)
Each impact is analysed under the load of all boot categories 1-4 with and without
build-up. The analysis of Impact 1 shows the maximal glass stresses, the Impact 2 and 3
the maximal internal forces in the steel sections.
The following stress limits for impact loads according to TRAV[4], based on [3] and [6],
were used to assess the breaking probability of the glass:
RD = 170 N/mm² (ESG)

RD = 120 N/mm² (TVG)
The 4-layer VSG-glass sections are modelled as isotropic plate elements with full
composite effect, acting like a full cross section. This is justified by the extremely short
loading time during impact and by the typically low temperatures during winter storm
flooding. Wellershoff describes the dependence of G (shear modulus) of the composite
interlayer (PVB, SGP) vs. load duration and temperature in [5].
79
Challenging Glass 3
4.4. Conclusion
Glass failure is only associated with impact by a category “boat 2 with build-up”, for
which stress limits are exceeded. However, as the number of boats in category “boat 2
with build-up” is limited to 8 % of all boats, there is only a low probability that a wall
section will in fact be breached. For all other categories, the loads and corresponding
stresses are below critical level and do not result in glass failure. The steel-frame
construction resists all kinds of boat impact.
Figure 8: Glass stress from boat impact (category “boat 2 with build-up”)
Figure 9: Stress on steel elements from boat impact (category “boat 2 with build-up”)
5. Experimental verification
In an experiment undertaken at Dresden’s Technical University, Prof. Weller’s team
investigated the load bearing capacity and the deformation of broken laminated glass
sections under hydrostatic water pressure (1.10 m water column to ground-level glass
section) resulting from storm flooding. In the experiment, two different composite
constructions were investigated, one with PVB foil and the other with SGP interlayers.
During the whole experiment duration of 9 hours, no relevant deformation could be
detected, and the two composite constructions did not display any differences in terms
of deformation [7]. However, in an additional test with a free falling compact mass, the
80
SGP glass turned out to have a significant better resistance against local penetration
than the PVB glass.
Figure 10: Water filling of the specimen [7]
6. Outlook
The project is now fully designed and in its authority approval phase. Detailed design
and the tender is planned for 2013. The construction phase is scheduled to run from
2014 to 2016. As the pictures show, the Warnemünde flood protection wall is a highly
versatile solution, suitable for many locations worldwide with similar design
requirements. Hopefully, the Warnemünde project will become a model for the
successful application of laminated glass in flood protection walls in visually sensitive
inner-city areas.
Figure 11: Future riverside view [2]
81
Challenging Glass 3
7. Abbreviations
Table 2: Abbreviations
Abbr. in UK in USA in Germany
Float Float glass Annealed glass (AN) Floatglas
Spiegelglas
TVG Heat-strengthened glass Heat-strengthened glass teilvorgespanntes Glas (TVG)
(HS)
ESG Toughened glass Fully tempered Einscheiben-Sicherheitsglas
glass(FT) (ESG)
ESG-H toughened glass fully tempered glass Einscheiben-Sicherheitsglas
with heat-soak test with heat soak test mit Heißlagerungstest
PVB Polyvinyl butyral Polyvinyl butyral Polyvinylbutyral
SGP Sentry glas (plus) Sentry glas ( plus) Sentry glas (plus)
TRAV see reference [4]
Strand7 Finite Element Software package
8. Acknowledgements
The project’s client, the Staatliche Amt für Landwirtschaft und Umwelt Mittleres
Mecklenburg, Dezernatsgruppe Küste, was fully confident that the glass wall would
prove to be the right solution for the project, long before the laboratory tests had been
carried out by TU Dresden, Institut für Baukonstruktionen. The authors are grateful to
both these institutes for their confidence and professional advice.
9. References
[1] Regelwerk Küstenschutz Mecklenburg-Vorpommern, Übersichtsheft, published by Ministerium für
Landwirtschaft, Umwelt und Verbraucherschutz, Schwerin, Germany, 2009.
[2] Sturmflutschutz Warnemünde, Alter Strom, Süd, folder, published by Staatliches Amt für
Landwirtschaft und Umwelt Mittleres Mecklenburg, www.sturmflutschutz-warnemuende.de, Rostock,
Germany, 2010.
[3] Rück, R, Voelker, G.E.: Untersuchung von 4-seitig linienförmig gelagerten Scheiben bei Stoßbelastung,
Stuttgart, Germany, 1999.
[4] Technische Regeln für die Verwendung von linienförmig gelagerten Verglasungen (TRLV), German
code of practice, 2006
[5] Wellershoff, Frank: Bemessungsschubmodulwerte für Verbundglasscheiben, in: Stahlbau 76 (3/2007),
p. 177ff, Berlin, Germany, 2007.
[6] Wörner, J.-D, Schneider, Jens: Abschlußbericht zur experimentellen und rechnerischen Bestimmung der
dynamischen Belastung von Verglasungen durch weichen Stoß, Darmstadt, Germany, 2000.
[7] TU Dresden, Fakultät Bauingenieurwesen, Institut für Bauingenieurwesen, Prof. Dr.-Ing. Bernhard
Weller: Prüfbericht Nr. 2010/246 Bauvorhaben: Sturmflutschutz Rostock-Warnemünde, Bauteil:
Sturmflutschutzwand aus Glas; Prüfung: Experimenteller Nachweis unter statischem Wasserdruck,
(10.12.2010).
82
doi:10.3233/978-1-61499-061-1-83
Lincoln Center Canopies

- Performance in Glass
Jan Knippers, Jochen Riederer, Matthias Oppe
Knippers Helbig Advanced Engineering, Germany, info@knippershelbig.com
This paper concerns two cantilevering steel-glass structures recently built in New
York, USA. Each structure consists of two outward tilting primary steel beams with
a length of 27m, a central column, and a glass surface providing lateral stability and
weather protection. Twelve panels of four-layer laminated safety glass with
dimensions of 2,3×4,4m are connected to the underside of the beams. The glass
panels act as the only bracing system of the structure; no additional bracing is
required. The connection between glass and steel is made by custom-built point
fittings with four fittings per glass panel connected by a two-part injection mortar.
At the column kink a ‘glass knee’ of two frameless four-layer laminated safety
glass panels is used for even load distribution.
Keywords: glass structure, structural glazing, SGP-interlayer, canopy, roof,

cantilever, steel-glass-structure, New York
1. General Introduction
The Lincoln Center for the Performing Arts was built in the 1950s and is the most
important cultural venue in New York. As part of a renewal of the whole complex, the
architects Diller Scofidio + Renfro of New York designed the new Josie Robertson
Plaza and the entrance from Columbus Avenue / Broadway (Fig. 1).
Vivian Beaumont
Theater
W62nd Street
W65th Street
David H. Koch Theater Josie Robertson Avery Fisher Hall

Plaza
Canopy I Canopy II
ay
adw
Columbus Ave Bro
W63rd St
Dante Park
Figure 1: Plan View ´Lincoln Center´
83
Challenging Glass 3
The architects developed two large cantilevering glass canopies to provide weather
protection for the visitors arriving from Broadway streetside. The two structures each
consist of two 27m long hollow section steel beams that cantilever 12m towards the
street and rest on two steel columns. The glass panels are connected to the underside of
the beams to provide lateral stability to the system. This made it possible to construct
the canopies without any additional bracing elements. The steel columns are also linked
together by two glass panes, the ‘glass knees’ at the column kink. Due to this sculptural
geometry a distinctive and at the same time light and transparent structure has been
developed (Fig. 2).
Figure 2: Canopies and Lincoln Center, New York (Photo: Keller Fotografie)
2. Construction
The two canopies comprise of two slightly tilting 27m long welded steel hollow
sections. They are fixed to the existing buildings by a circular hollow cross beam at one
end. Together with two supporting column legs the structure rests on these four points.
The beams cantilever out with a length of 11,6m (Fig. 3) towards the adjacent street.
Twelve glass panels with dimensions of 2,3×4,4m are connected to the underside of the
beams. The point fittings connecting the panels to the steel beams were developed using
high strength steel. In order to enable correct load transfer between the glass panels and
the steel structure a two-part injection mortar was used to rigidly connect the glass to
the steel point fitting. During the development of the structure it was therefore possible
to eliminate any additional horizontal bracing system apart from the glass panels. Each
panel has a slight incline for drainage of the roof surface without requiring an additional
gutter. Due to the fact that each steel beam is slightly tilted and also not parallel to each
other, each single point fitting had to be custom built. A central steel column built of
two welded hollow steel sections supports the roof of the canopy. The steel column has
84
Lincoln Center Canopies – Performance in Glass
a kink that creates a Y-shape in elevation. The column is founded on the basement level,
supports a pedestrian ramp, the roof, and has a total height of 8m.
Figure 3: Canopy Avery Fisher Hall (Photo: Keller Fotografie)
The two column legs are connected at approximately mid-height to allow for an even
horizontal load distribution. This connection is situated at the kink and uses two
structural laminated glass panels.
3. Glass Supporting Steel
3.1. Structural system and Development of details

The structure was developed in close cooperation between the structural engineer, the
construction company, and the architects and client. The overall concept was progressed
in several stages, beginning with the development of the structural glazing and
proceeding to the detailed design of the point fittings and the connection detail between
the beams and columns. By including local key decision maker a very commanding
time schedule could successfully be achieved.
3.2. Design Criteria

The canopies were designed using regular load cases such as snow and wind loading.
Wind tunnel tests with two different models had to be performed due to the extensive
cantilever and the unusual design. For the determination of the local characteristic wind
speeds a surrounding area model was used, with adjacent buildings modelled in 1:350
scale (Fig. 4a). A second model built at a larger scale (Fig. 4b) helped determine the
pressure distribution on the roof surface and the steel structure.
85
Challenging Glass 3
Figure 4a and b: Wind tunnel test - 1:350 Context Model- and 1:60 Detail model
(Photos: Wacker Ingenieure)
With this analysis the static and dynamic effects of the wind loads could be correctly
determined. The canopies are also designed for maintenance loading. Additionally, a
temperature load case was applied in the structural calculations to account for variable
heating of the dark steel beams and the transparent roof surface, as well as the varying
coefficients of thermal expansion of the different materials.
3.3. Steel Structure

The two main criteria for the geometry and detailing of the steel structure (Fig. 5 and 6)
were to create a sculptural but simple structure where the beams rest on the columns
without any visible connection details, and to connect the various steel members using
bolted connections to prevent any site welding due to cost reasons. These goals were
achieved using countersunk screws or providing steel pockets at the connection details.
After construction the details were covered, sealed and painted. The beams and columns
are welded hollow box sections prefabricated as variable depth custom members.
1
27,24 m
15,61 m 11,63 m
4,41 m
3 2 6
4
7,95 m
3,54 m
1 existing concrete structure

2 steel girder
3 suspended glass
4 pedestrian ramp
5 steel column
6 glass knee
Figure 5: Structure - Elevations
86
27,24 m
15,61 m 11,63 m
4,42 m
1,41 m
2,33 m
Figure 6: Structure – Plan View
The slightly tilted primary beams are rectangular hollow sections that vary in depth
from 300×55mm to 700×180mm. The sections have flange thicknesses of 50mm and
web thicknesses of 10-25mm. The two column legs are also variable depth custom -
profiles using steel plates of 460×185mm at the column foot and 800×185mm at the
column head. The heavy plate thicknesses of 40mm for the columns were determined
using dynamic analysis of the overall structure. The structure is designed to American
steel codes.
3.4. Suspended Roof Diaphragm

The roof surface consists of twelve planar glass panels with dimensions of
2324×4420mm. The laminated safety glass consists of three 15mm thick upper and one
8mm thick lower fully tempered glass panes. The panes are laminated together using a
1,52mm thick Sentry Glass Plus (SGP) laminating foil.
2
4420mm
2324 mm
1 point fitting
2 glass cutout at column
Figure 7: Top view glass panel with four point fittings / Detail – Analysis model – stress distribution
(Figure: LFK Ingenieure)
The lowest panel has a silkscreen dot pattern for sun protection. The two glass panels
next to the columns are cut-out to allow the columns to run through the roof surface.
Each glass panel is connected by four point fittings. Four holes are drilled through the
glass at each fitting with a diameter of 46mm (Fig. 7 left). Each glass panel weighs
roughly 1,5 tonnes. After the structure is carefully erected to avoid any lock-in forces in
the elements, the gap between the glass panel and the connecting bolt is grouted using
injection mortar Hilti Hit HY 70. The grout allows the transfer of horizontal forces
87
Challenging Glass 3
between the steel beam and the glass panels. The analysis of the glass panels used
linear-elastic finite element methods. For this a volumetric model with three 15mm
layers and a 1,52mm SGP- interlayer was constructed. (Fig 7 right)
The lower 8mm glass pane is applied as self-weight only because it is not directly
connected to the point fitting and therefore does not provide any load bearing function.
For the SGP- interlayer the applied shear modulus depends on the loading duration (see
Table 1).
Table 1: Consideration of material properties SGP-interlayer

Loading Duration, Shear modulus
Load Case SGP interlayer [MPa]
G, long, self-weight 2
G, long, snow 15
G, short, wind 80
The load combinations were derived from common loadings such as self-weight, snow,
wind, temperature and live load, and also from earthquake loading. Additionally the
restraint forces due to lateral stability are applied as point forces within the glass hole.
Finally, additional accidental load cases assuming a broken upper or lower glass pane
were investigated. The analysis was undertaken in accordance to American Standard
ASTM E 1300-04 [1]. For the accidental load combinations the permissible stress of the
glass panes was increased by 50%.
3.5. Glass Knee

The so-called “glass knee” connects the two steel column legs to create one composite
steel-glass column (Fig. 8).
4
2
314 mm
2
3
800 mm
1
314 mm
185mm 236 mm 185mm
1 glass knee LG - 4x12mm FT 1

2 pin
3 steel column 3
4 continuous support - injection mortar
394 mm
Figure 8: Detail - Glass Knee (Figure: seele GmbH)
88
It consists of two coupling glass panels with dimensions of 311×800mm. The build-up
of the glass is derived from the maximum moment induced by the column legs. Each
glass panel comprises of laminated safety glass with four 12mm fully tempered glass
panes, each connected by an intermediate 1,52mm SGP-interlayer. The composite
action between the column legs and the glass panes is provided by connecting the glass
to the legs using four corner bolts as point fixings. Additionally a partial line support in
the glass corners is provided through injecting mortar in these locations.
3.6. Point Fittings

The point fitting was developed in close cooperation with Seele Sedak GmbH with the
goal to achieve an even and smooth underside of the roof without any protruding point
fitting or bolts (Fig. 9). Therefore, each point fitting is embedded within the lower glass
pane of 8mm and covered by a cover plate. The adjacent bore holes are always linked
by one point fitting. The detail is very important for the overall structural behavior it has
to accommodate loads due to the bracing system as well as lock-in loads due to the
varying thermal expansion of the materials glass and steel. Each point fitting was
custom-built due to the variable geometry. To allow for a slim and elegant detail high
strength stainless steel type 1.4462 was used.
7 4 5
3
1 steel girder 6
2 point fitting
3 pin Ø 40mm
4 silicone sleeve
5 laminated glass 3x15mm FT + 8mm FT
6 cover plate
7 injection mortar Hilti HIT HY 70
Figure 9: Point Fitting – Vertical Section (Figure: seele GmbH)
The canopies are laterally restrained by the active connection between glass pane and
steel structure. However, all loads could not be transferred if the connection would have
been completely rigid. Therefore the detail was developed in an iterative process. First
the glass panes were mounted without introducing any constraint forces and grouting
took place only after all glass panes were in place. By doing so, the constraint forces
due to self-weight were eliminated. In a second step a system was developed that
provided enough stability for the structure while keeping the restraint forces in an
acceptable limit. Silicone bushings of 2mm thickness where placed around each fixing
89
Challenging Glass 3
steel bolt in the glass hole. These act as springs in the structural system. Additionally
various bolt diameters and the effects on the structure were investigated. By selecting
the correct bolt diameter and silicon bushing material with the proper hardness the
forces from all load combinations that have to be transferred by the glass could be
determined. Finally a dynamic analysis on the overall structural system was performed
to determine the effect of the eigenvalues on the applied wind loading. The steel-glass-
system was optimised in various iterative steps and the assumptions verified by material
tests for the silicone bushing (Tab. 2).
Table 2: Lateral Stability System

System Spring stiffness Diameter Eigenvalues Max. Force
Silicone bushing Bolt System FK [kN]
[kN/mm] d [mm] f [Hz]
Model 1 10 20 1,2 0,3
Model 2 10 40 1,3 1,5
Model 3 100 20 1,6 14
Model 4 100 40 1,9 19
Sleeve 2mm, 80 Shore 44 / 941 40 1,7 17

1
upper and lower bound of the measured silicone bush stiffness
After the final system with a 2mm thick silicone bushing of shore hardness 80 was
chosen, the structural system was checked again with an upper and lower bound for the
spring stiffness. This took into account the varying bush hardness due to temperature
difference and the slip between the bush and the steel bolt.
3.7. Performance Criteria

The beams were pre-cambered to minimise the deflection of serviceability loading. The
self-weight deflections at the cantilever are approximately 60mm vertical and 95mm
horizontal. Thus, the beams were fabricated using stencil plates. The camber in
longitudinal beam direction was constructed by partial heating. Due to the tilted shape
of the beams this had to be performed in three-dimensions for the vertical and horizontal
direction.
4. Fabrication and Erection
4.1. Fabrication Tolerances

The general construction tolerances of +/- 3 to 5 mm for steel structures had to be
exceeded significantly due to architectural demands for the appearance of the cantilever
structure. A tolerance limit of +/- 1,5mm was established for the steel structure. For the
glass panes, the steel bolts of the point fittings had to be placed exactly so that the
silicone bush could be fitted and the injection mortar could achieve a minimum
thickness of 3mm.
90
4.2. Shipping and Erection

Shipping and construction of the structure was been a logistical challenge (Fig. 10). The
whole structure with an overall weight of 40 tonnes for each canopy had to be
prefabricated in Europe, transported to North America by ship, and driven to Manhattan
in the heart of New York by heavy goods transport. For this the steel structure was
prefabricated in two pieces. One piece consisted of the columns including the pre-
mounted glass knee. The second piece consisted of the complete steel roof structure. On
site the two pieces were connected at the column head. Because of the large
prefabricated elements the construction time could be considerably shortened to keep
interruption of the traffic to a minimum.
Figure 10: Shipping of steel structure (Photo: seele GmbH)
5. Credits
x Client: Lincoln Center for the Performing Arts, Inc., New York, USA
x Architect: Diller Scofidio + Renfro, New York, USA
x Conceptual Design: Dewhurst Macfarlane, New York, USA

x Design Development- and Construction Documents: Knippers Helbig GmbH,
Stuttgart, Germany
x Glass Design: Knippers Helbig GmbH, Stuttgart with LFK Ingenieure GmbH,
Lauffen, Germany
x Windtunnel Tests: Wacker Ingenieure, Birkenfeld
x Fabrication and Erection: Seele Sedak GmbH & Co. KG, Gersthofen, Germany
x Completion: 12.2009
6. References
[1] ASTM International: ASTM E 1300-04: Standard Practice for Determining Load Resistance of Glass in
Buildings, 2004
91
doi:10.3233/978-1-61499-061-1-93
Project for the Eiffel Tower:

Constructive Geometry
Nicolas Leduc, Jacques Raynaud, Niccolo Baldassini
RFR, France, www.rfr-group.com, nicolas.leduc@rfr-group.com,
jacques.raynaud@rfr-group.com, niccolo.baldassini@rfr-group.com
The three pavilion facades for the renovation of the Eiffel Tower’s first floor
constitute a new challenge in the field of transparent skins with free forms: this
smooth double curved surface adapts to the high-performance thermal constraints
of a of a unique public interior space. In avoiding the standard facetted solution, the
developed project retains the full integrity of the initial Architectural proposal while
guaranteeing economic feasibility. Thanks to the optimization of a parametric
model, the variable double-curved surface of the transparent skin can be
approximated from a quadrangular cylindrical panel surface. This process implies
zero torsion at structural nodes and permits the fabrication of structural elements
from developable curved sheets by welding thus enabling a standard compact detail.
Keywords: Eiffel Tower, Cylindrization, Discretization, Freeform geometry,

Geometrical optimization
1. Project introduction
For the third time since its construction, the first floor of the Eiffel Tower will undergo
renovation. The competition organized by the SETE (Eiffel Tower Operating Company)
was won by a partnership led by the architecture agency Moatti & Rivière and the
general contractor BATEG by providing an appropriate response to current uses, while
offering a new way of experiencing the tower’s particular space and emptiness.
Figure 1: Renovation project of the Eiffel Tower’s first floor
93
Challenging Glass 3
The project is the planned replacement of the Gustave Eiffel and Ferrié pavilions with
two new pavilions housing an auditorium, a shop and exhibition spaces along with the
façade renovation of ‘restaurant 58’ to render this harmonious with the new pavilions.
The three lifts which give access to the first floor will be equipped with new shelters.
Visitors will be able to experience the 57 meters of void under their feet with a 1.5metre
wide glazed floor which is to run around the perimeter of the centre of the tower. This
experience will be reinforced by a fully glazed 2.5-metre-high balustrade inclined
towards the void. All of the new structures echo the Eiffel Tower geometry in that the
façades of the pavilions and shelters as well as the balustrades follow the inclined
direction of the columns. This distorsion is particularly tangible for the façades of the
pavilions which are oriented towards the central floor opening and have a double-curved
surface.
RFR are collaborating with Moatti & Rivière Architects for the technical design of the
project’s glass works: glass floors, balustrades, pavilions and shelter façades. In this
article we will focus on the design of the main feature of the project: the inclined double
curved façades of the two pavilions. Moreover, from the technical point of view, the
free geometry of these façades faces a dual challenge from the perspectives of a glazed
skin complying with thermal constraints and those of the support structure.
Figure 2: Curved façades of pavilions
2. Morphogenesis
The façade design is the result of a process where geometrical optimization is developed
according to the architectural intent thus guaranteeing the aesthetic goals aimed for by
the architect.
In a façade with relatively modest spans, the constructive question of the skin is in the
core of the challenge.
2.1. Skin
Several constructive strategies of freeform glazed skins have been ruled out. While the
fabrication of double-curved panels was rendered impossible due to economic
feasibility reasons, the use of flat panels, whether triangular or quadrangular, was
deemed inappropriate. This was especially true for a curved surface where the reading
of the double curvature was a priority.
94
Project for the Eiffel Tower: Constructive Geometry
Figure 3: Double-curved Figure 4: Flat triangular Figure 5: Flat Figure 6: Cylindrical

quadrangular panel panels quadrangular panels quadrangular panels
The adopted solution preserves the curved aspect of the façade while guaranteeing
economic feasibility thanks to the use of industrially produced cylindrical panels.
Today’s bending machines enable the fabrication of cylindrical panels with variable
radii (See §4.1.).
This solution implies two types of discontinuity between two consecutive panels: a
discontinuity in position and in tangency which can be minimized with geometrical
optimization.
2.2. Jointing setting-out

The minimization of divergences is a major challenge from an architectural point of
view as well as from a technical one (fixation of panel, watertightness, etc.). The
orientation of the jointing setting-out along the lines of curvature is minimizing the
divergences. Indeed, since generatrixes of cylinders are aligned with the direction of
maximum curvature, it is preferable that the jointing layout adopts the same direction in
order to reduce the discretization length which can be described as the maximal length
of the panel in the generatrixes’ direction.
If the principal maximum curvature is small, the cylinder orientation and therefore the
jointing layout orientation are less constrained, therefore a greater freedom in the
jointing layout can be implemented in the most planar zones.
d1
a
d2> d1
Figure 7: Divergences implied by cylindrical panels Figure 8: Influence of the jointing layout
orientation in relation with curvature lines
95
Challenging Glass 3
2.3. Structure
The principal curvature directions are also well-adapted to the geometry of curved sheet
mullions: the geodesic torsion being null along those curves, the envelope of the
tangential planes and the normal surfaces are developable and always perpendicular
between them, which permits a much higher level of constructability (See §4.2).
Figure 9: Developable normal Figure 10 : Developable tangential Figure 11 : Normal &

surfaces surfaces tangential surfaces
3. Project process
Placing a parametric model of the surface which integrates the architectural intention at
the heart of the design process creates a synergy between architectural constraints and
geometric and technical aspects.
The variations of the surface parameters, such as the central rise or the tangency of the
upper edge, allow rapid surface modification and the generation of families of solutions.
Figure 12: Central rise parameter Figure 13 : Tangency parameter
For each set of parameters, the principal curvature lines as well as the corresponding
jointing layout are generated. In order to regulate the panels’ dimension, the jointing
layout is modified and does not follow exactly the principal curvature lines. This has a
very minimal impact on divergences between panels because this decorrelation is
operated in the most planar zones.
Figure 14: Example of jointing layouts for three parameter sets
96
The choice within this solution matrix is made according to architectural criteria and
technical performances (evaluation of divergences, panels’ size, etc.).
Thanks to algorithms of global optimization, the radii and positions of cylindrical panels
are finally optimized in order to minimize the divergences in position and tangency in
adjacent panels (Eigensatz & Schiftner [1]). This optimization process has been carried
out with Evolute and is one of the key issues of the European research project which is
currently being conjointly pursued by RFR, the University of Vienna and Evolute
(IAPP). The research aims at finding solutions of how to build “Architectural Freeform
Structures from Single-Curved Panels” [2].
The quality of the output from the first optimization phase is very satisfactory and
results in divergences in position of lower than 2mm. This led to a second optimization
phase with new architectural constraints: cylinders are re-oriented to concentrate
tangency discontinuities according to vertical lines to underline the ascensional
character, particular to the Eiffel Tower.
Figure 15: First optimization phase Figure 16: Smoothing of cylinders according to the
vertical direction
4. Materialisation
4.1. Glass panels

The panels are insulated glazed units which are composed of an external toughened
laminated sheet for public protection as the glass overhangs the public space and an
internal a single sheet of tempered glass.
The traditional method for fabricating curved glass is to heat flat float glass in an oven
over a heat resistant mould, very often made of steel, and let it bend under gravity
loading.
This was not appropriate to the project for two reasons: firstly because the geometry of
each individual panel requires a specific mould, which is very expensive when there is
no repetitivity, and secondly because it is very difficult to create reliable heat treatments
of the glass with this bending technique.
Glass bending machine techniques are becoming standard practice and are able to bend
the glass and provide heat treatments of good quality at the same time. These machines
are derived from industrial tempering machines which have been altered in order to
97
Challenging Glass 3
bend while tempering. Instead of being bent by moulding, the glass is bent via computer
controlled mechanisms which include air blowers and rollers.
These systems provide good quality tempered bent glass but have two important
limitations which are the consequence of their complexity: firstly the bending shape is
limited to circular cylinders and secondly the bending direction is fixed.
The first limitation has to be controlled by the cylinders’ optimization and the second
limitation imposes a panel size constraint which is dependent on the angle between the
ruling of the surface and the edges of the panel.
Figure 17: Traditional bending on a steel mold Figure 18: Modern bending and tempering machine
without mold
4.2. Structure
The 8-meter-high façade of the Eiffel pavilion is supported against its perimeter to
transfer all loads to the rest of the pavilion or to the existing structure of the Eiffel
Tower. Permanent dead loads are mainly transferred by vertically orientated mullions in
the central part of the facade and then by cross beams. Mullions are subjected to
bending in order to take wind loads. In order to reduce deflections generated by the
wind, the base of mullions are rigidly connected to the lower edge beam.
Figure 19: Moment diagram
To maximize the slenderness of the mullions, we chose a rectangular section whose

strong inertia is normal to the reference surface and geometric axis follows the curved
glass jointing layout. We have previously observed that the normal and the tangential
surfaces leaning against a principal curvature line are developable. The orthogonal
offset of developable surfaces keeping its developability (Pottmann et al. [3]), the four
98
sheets of built-up girders are therefore developable: webs are obtained by orthogonal
offset of median normal surfaces and flanges by orthogonal offset of median tangential
surfaces.
The mullion is therefore obtained by the forming of four faces by simple bending
without torsion from sheets cut in plan and then assembling by welding. Figure 20
shows the process for a built-up girder whose curvature has deliberately been
accentuated.
Figure 20: Tri-dimensional Figure 21: Unfolding of Figure 22: Physical model
modelization sheets
4.3. Tightness
Main curvature lines have one more interesting feature: normal surfaces leaning against
principal curvature lines are not only developable, but also bisectors to two adjacent
cylindrical panels. Thanks to this feature, the discretization angle is better distributed on
both sides of the mullion. Furthermore, because divergences in position and angle are
limited, it is possible to adapt a standard detail despite the geometric complexity of the
façade.
Figure 23: Tightness detail on mullion
The pressure plate is fabricated on a principle similar to the one of mullions’ flanges:
this is a developable surface formed by bending cut-in-plan sheet without torsion.
99
Challenging Glass 3
5. Conclusion
The refurbishment project for the first floor of the Eiffel Tower which will be
inaugurated in 2013 is a successful application of cylindrical discretization principles of
double-curved complex forms using double glazing.
A construction feasibility approach to the glazing and structural elements helped

finalize the jointing setting out thus controlling building costs. Mullions, which are free
forms in space, are realized thanks to the assembly of developable surfaces. Panels are
portions of cylinders shaped by industrial forming methods. The minimization of the
divergences allows the use of ‘off-the-shelf’ glazing fixation products possible.
This design process, resulting from a technical and productive reflection, intended for a
unique public interior space with high-performance thermal constraints, allowed the
architectural intentions to become a reality in such a prestigious historical site.
Figure 24: Execution geometry
6. Acknowledgements
We would like to thank the teams we worked with for this project, particularly Alain
Moatti and Marie-Pierre Guerin at Moatti & Rivière, Alexander Schiftner and Michael
Eigensatz at Evolute and Alex Beeputh at RFR.
References
[1] Michael Eigensatz and Alexander Schiftner, Case Studies in Optimization of Glass-panelized
Architectural Freeform Designs, Proceedings Glass Performance Days Conference, Finland, 2011.
[2] Research on the panelisation of free-form structures by single-curved panels is being carried out by a
consortium consisting of Vienna University of Technology, RFR and Evolute, and is funded by the EU
through project ARC (IAPP Project 230520).
[3] Helmut Pottmann, Andreas Asperl, Michael Hofer, Axel Kilian, Architectural Geometry, Bentley
Institute Press, 2007.
100
doi:10.3233/978-1-61499-061-1-101
Challenges in the Design, Fabrication and

Installation of Glass Structures Comprising
of Super Jumbo Glass Sheets
Peter Lenk, Harriet Lambert
Eckersley O’Callaghan, UK, info@eckersleyocallaghan.com
The recent works at Apple Stores in Hamburg, Germany and SoHo, New York,
USA have incorporated the installation of new internal glass structures: 12m
straight stairs; 10m balustrades forming the atrium perimeter of the upper storey;
and 5m bridges. The design of each project elegantly balances structural glass
elements with connecting metal fittings through the choice of super jumbo glass
panels and inserted laminated connections. An overview of these projects is
presented focusing on challenges which have arisen during design, fabrication and
installation. These include the global stability and dynamic behavior of the stair and
bridge; matching the limited tolerances of large glass panels with movements of
existing structure; and dealing with restraints imposed by building authorities.
Keywords: Structural Glass, Dynamic Analysis, Super Jumbo Glass Panels,

Laminated Fittings
1. Introduction
The new glass structures installed in Hamburg and in SoHo, New York have a similar
format. This is apparent in the images of Figures 1a and b. Each flat glass structural
element is a single glass panel. The only exception to this is the floor edge guardrail
(balustrade) at SoHo, which is divided into two panels between the stairs and bridge.
Figures 1a and b: The new internal glass structures for the Apple Stores in Hamburg and SoHo, New York.
1.1. Stair
The glass stringers are the primary structural elements of the stair. They span as beams
between the lower and upper floor levels. The glass stringers comprise of five sheets of
12mm thick, flat, fully tempered glass, laminated with DuPont™ SentryGlas®
interlayers. They measure approximately 12m in length and 1.6m in height. The stringer
101
Challenging Glass 3
supports at lower floor level are hidden within the floor build-up. They are vertically
supported on bearing blocks and laterally supported between mild steel plates. The
stringer supports at upper floor level are partially visible and partially hidden. They are
visibly supported on two full-penetration bolts. These connect to hidden rocking
brackets which provide vertical support and out-of-plane rotational stiffness.
Glass treads span 2.3m between stringers as simply supported beams. The treads and
landings comprise of five sheets of flat glass, laminated with DuPont™ SentryGlas®
interlayers. The treads are 8mm / 8mm / 19mm / 8mm / 8mm all annealed to allow post-
lamination polishing for a crisp edge.
Each handrail connects to each stringer at eight or nine locations and continues beyond
the stringer to connect to the main structure on the lower floor and to balustrade
elements on the upper floor. Connections in glass elements use metal alloy components
and resin.
Figure 2a, b and c: The bridge, staircase and balustrade of Apple Store Hamburg.
1.2. Bridge
The glass guardrails are the primary structural elements of the bridges. They span as
simply supported beams, connecting to the main floor edge beams. The guardrail panel
supports are similar to the stringer supports at upper floor level.
Glass beams span between the guardrails. Three beams are used at Hamburg; whilst
only two beams are used at SoHo. The beams connect to the guardrails at four points:
two points vertically and all four points laterally. The bridge walkway is supported on
the main structural floor and the glass beams through direct bearing. Each handrail
connects to the bridge guardrail at locations aligned with the beams and continues,
connecting to the floor edge guardrail (balustrade) on each side of the bridge.
1.3. Balustrade
The balustrade comprises of glass panels point-fixed to the face of the 2nd floor slab.
The pairs of point fittings supporting the balustrade panels are located at approximately
1.5m centers and align with the handrail fittings above. All point fittings provide lateral
restraint. Two fittings per panel provide vertical restraint.
102
Comprising of Super Jumbo Glass Sheets
2. Glass Design Challenges
2.1. Global structural behavior

The first step in the design of the glass elements is to study the global behavior of each
glass structure. Global finite element models of each of the glass structures were created.
In these models, glass panels are modeled using plate elements with calculated effective
thicknesses and connections are modeled using beam elements with rigid links allowing
load share between nodes in the region of each fitting. These models are used to gather
results related to global behavior, such as structural deflections and reaction forces at
connections.
A significant design challenge was the lateral flexibility of the staircase stringers which
arises due to the extensive free-spanning length of these elements. The upper and lower
floor level stringer supports are designed with some lateral rigidity to increase the
stiffness of the system. Despite this, the lateral stiffness of the system remains
inadequate if all glass-glass connections are designed with full rotational releases.
Several structural systems were investigated to address this. The addition of diagonal
bracing rods below tread level and the addition of vertical risers between treads were
both considered. Both of these approaches are effective in increasing the lateral stiffness.
However, both were rejected by the client on aesthetic grounds. Designing the treads as
fixed struts between stringer panels (forming a Vierendeel truss) is very effective in
increasing the lateral stiffness. However, stress concentrations develop in the glass
elements at the rigid connections which are too high for a glass element, of reasonable
thickness, to resist. The approach which offered the best solution was to design the tread
connections as semi-rigid and to increase the thickness of the stringer panels. The
design of connections is controlled to achieve an optimum balance between the lateral
stiffness of the system and the magnitude of stress in the glass. The design of the
connections also accounts for manufacturing accuracies and installation tolerances.
In the bridge, a lateral stability structural system is used which is similar to the staircase,
as described above. The shorter span and horizontal form of the bridge (with a single
continuous walkway panel) means that it is inherently stiffer laterally than the staircase.
The bridge guardrails are thinner in build-up than the staircase stringers.
2.2. Connection design

The next step in the design of the glass elements is to study the local behavior of the
glass and fittings in the region of glass-glass and glass-structure connections. Having
discussed some aspects of the connection design in the above section, it is apparent that
the global design and local design stages are not entirely independent. The two
processes interact and iterations in the design are required. Detailed local finite element
analysis models are used. These use three-dimensional brick elements and cover a
region of approximately 300mm in all directions from the center of the connection
being investigated.
All of the fittings in these projects are innovative designs; having evolved from the
fitting designs used in previous Apple Stores; they are used for the first time in these
projects. The majority of the fittings used are laminated insert fittings: such as at the
connection between the bridge beams and bridge guardrails. Some fittings use through-
bolts and rocking bracket bearings; such as at upper floor level staircase stringer
103
Challenging Glass 3
connections. Several design challenges were faced at this stage. Two are presented
below which relate to the magnitude of forces transferred: firstly, at the bridge beam to
bridge guardrail fittings; and secondly, at the upper floor staircase stringer connections.
Figures 3a and b: The bridge beam to guardrail laminated insert Figure 4: Local FEA stringer connection
fitting: vertical and horizontal sections model - bolt hole stress contour plot.
The bridge beams connect to the guardrails via two laminated insert fittings at each end.
These fittings penetrate two plies of the three-ply laminated glass guardrails, and
continue into the central ply of the five-ply laminated glass beams. See figures 3a and b.
One fitting per pair transfers vertical and horizontal loads, whilst the other fitting only
transfers horizontal loads. At SoHo, the bridge has two beams. The loads transferred in
each fitting for the SoHo bridge are easily accommodated by the glass. However, the
stress developed in some fitting components is close to design capacity. To carry out an
accurate analysis of these components it is necessary to account for non-linear material
behavior, allowing for plastic deformation of some metal parts. It was decided that
physical testing of the fitting is the most appropriate design assessment method for these
components.
Two fitting specimens, manufactured by TriPyramid Structures, Inc. were tested. They
were fixed to mild steel plates in a testing rig. A shear force across the fitting was
increased in steps until breakage. The minimum breakage load was found to be 54.5kN.
This is 3.7 times greater than the applied load.
The connection at the top of the staircase transfers high reaction loads. Unfortunately,
the capacity of the minimalistic and neat laminated insert fitting is insufficient at this
location. The use of through-bolts allows for more uniform distribution of load through
the element thickness, reducing peak stresses. By connecting to a hidden rocking
bracket, load-share between two through-bolts is achieved. Analyses accounted for the
possibility of the breakage of both outer glass plies and for the possibility that forces
may not distribute evenly between the two bolts – allowing for an extreme case of a
70:30 load share. See figure 4.
2.3. Failsafe Design

The failsafe condition is achieved through redundancy within the laminated glass sheets.
Failure of an outer ply of any single glass panel will not result in successive failure of
the remaining plies. Any glass panel with a broken ply will continue to function under
full loading for a short period of time, and will remain in position without causing
successive failure under self-weight for a minimum of one day.
104
Figure 5: Residual capacity test of stair tread Figure 6: Impact test of balustrade
A series of impact and residual capacity tests were carried out by Labor fur Stahl- und
Leichtmetallbau GmbH which confirmed the analytical results. See figures 5 and 6. In
the impact tests, the glass is allowed to break but the connections must not fail, the glass
element must not be penetrated by the impactor and there is to be no dangerous debris
falling from the test specimen. All components passed this series of tests.
2.4. Dynamic Analysis

The staircase structure has a relatively low stiffness to mass ratio which means it is
likely to have a relatively low natural frequency. There is a significant potential that
such a structure will be dynamically excited by human-induced loading. An assessment
of the dynamic behavior of the staircase is conducted to ensure that the structure does
not trigger human discomfort in normal use. As the stair is more flexible in the lateral
direction, this study focuses on the lateral dynamic behavior.
A significant amount of research into human-induced vibration of structures has been

carried out in recent years. Valuable information can be found in [2], [3], [4] and [8].
From these resources, the following can be stated:
x Human induced loading varies in magnitude and in range of frequencies.

x According to [1], ‘about 10% of the vertical loading, which is about 4% of

pedestrian’s weight, works laterally when people walk’.
x People ascend and descend staircases at a mean frequency of 2Hz.
x Major lateral load components are associated with frequencies of 1Hz and 3Hz.
Figure 7: Harmonic load components (Fourier amplitudes) [1]
105
Challenging Glass 3
In Figure 7, the harmonic load components induced by a 60kg person are presented.
This highlights the significance of the two major lateral load components associated
with frequencies of 1Hz and 3Hz.
A modal analysis was carried out to calculate the natural frequencies of the structure
using Strand7 FEA software. A 10% live load mass was applied in addition to the self-
weight mass. There is little guidance available on the damping properties of glass
structures. This topic is discussed further in [7]. Analyses were carried out with 1% and
5% critical damping to cover a range of values for this fundamental dynamic property.
A parametric study was carried out to investigate the sensitivity of the dynamic
behavior to variations in connection stiffness. The study concluded that within the
limited range of stiffness for the semi-rigid connection design, the variation in dynamic
behavior was minor.
A harmonic analysis was carried out with a lateral 70N force oscillating sinusoidally at
mid-span of the stair. This force was identified as appropriate for two 90kg people
walking simultaneously on the stair. The results are presented in Figures 8a and b.
Distinct peaks in acceleration are apparent at particular frequencies which correspond
with the natural frequencies identified during the modal analysis. The maximum
acceleration is located at 4m which corresponds with the 1st mode shape.
Figures 8a and b: Results of harmonic analysis. Lateral acceleration of the stringer panel at 1m spaced
positions over a frequency range 0Hz to 20Hz for models with (a) 1% damping and (b) 5% damping. Note the
difference in magnitude of node accelerations along the vertical axis for the two graphs.
Figures 9a and b: Results of time history analysis. Lateral acceleration of the stringer at 2m-spaced positions
(a) at tread level, and (b) at handrail level.
106
A time history analysis was carried out for two 90kg people walking simultaneously up
and down the stair. Results are presented in figures 9a and b. It is apparent that lateral
accelerations at tread level are greater than at handrail level. The maximum peak lateral
acceleration experienced by the stair, with 1% critical damping, is 1.7m/s2 at tread level
and 0.5 m/s2 at handrail level. This result corresponds well with the harmonic analysis
presented in figure 8a.
Finite element software incorporates fast Fourier transformation signal post-processing

methods providing further information on the dynamic behavior and distribution of
energy across the frequency range of the analyzed structure. In figures 10a and b, power
spectral density plots are presented which have been post-processed from the results of
the time-history analyses presented in figures 9a and b.
Figures 10a and b: Power spectral density plots processed from figures 8a and b respectively.
A set of reference curves describing the human perception of continuous vibration is

presented in [5] see figure 11. Human perception of a vibration depends on its
frequency, its acceleration and its direction relative to the human body. In these curves,
vibration is rated relative to the vibration at the threshold of human perception. A
recommended performance target for indoor bridges is R<24, according to [6]. That is
to say the maximum acceptable level of vibration (r.m.s. acceleration) is 24 times the
level of vibration at the threshold of human perception. This performance criterion is

applicable to the internal stair designed here.
Acceleration in these curves is presented as an r.m.s. value. This value can be calculated
using the following formula:
1 T
a rms
T ³ 0
a t2 dt (1)
The maximum lateral r.m.s. acceleration for the stair has been calculated as 0.15m/s2.
Referring to figure 11 it can be seen that at a frequency of 2.5Hz, this r.m.s. acceleration
has a response factor of approximately 25. This is accepted as fulfilling the
recommended performance stated above. Indeed, use of the stair since installation has
confirmed that its dynamic behavior is acceptable and does not trigger human
discomfort. Nonetheless, it would be beneficial to conduct on-site dynamic
measurements to develop understanding of dynamic behavior of glass structures.
107
Challenging Glass 3
Figure 11: Building vibration curves Figure 12: Time dependent deflection study of RC structure
in the horizontal direction [5] supporting balustrade
2.5. Support Structure Movement Analysis

The Hamburg Store has a reinforced concrete superstructure. The potential for problems
arising from differential deflections after installation of the glass elements is significant
due to the scale of the super jumbo panels and the spans of the superstructure elements.
A series of detailed time history movement analyses of critical areas of the main
structure were carried out. These accounted for the sequence of installation, varying live
loads during store use, concrete cracking and concrete creep. One example is presented
in figure 12.
The portion of structure in this example covers the length of one balustrade panel. The
balustrade panels are manufactured such that the fittings have some tolerance allowing
for misalignment during installation. Furthermore, vertical movement is required in
lateral restraint fittings to accommodate differential movement between the flexible
reinforced concrete structure and rigid glass panel. The movement analyses were carried
out to establish the magnitude of the differential movements after installation and hence
the required movement capacity of these fittings.
From the analyses the rotation of panels could also be established. Rotation of
neighboring panels tends to open or close the gap between them. The required gap
magnitude could be verified as a result of the study.
108
3. Other Project Challenges
3.1. Building Authority’s Acceptance

Regulations imposed on the structural use of glass vary significantly around the world.
It is critical to understand the design approval process for a project at an early stage as it
may impact significantly on the construction schedule. In Germany there is a published
list of accepted building materials and construction forms. Only a narrow range of
innovative structural glass applications are listed. In instances where the desired
material or construction form is not listed, special approval verification is required. In
the case of the Hamburg Store, individual approval (Zustimmung im Einzelfall) was
required.
As part of this process, the local building authority reviewed the initial design and
prescribed certain design parameters and minimum requirements. These included
limiting the fully tempered glass design strength to 30MPa; specifying various post-
breakage scenarios and performance criteria; evaluating vibration behavior through
dynamic analyses; using three-dimensional finite element analysis models at
connections; verifying material properties of the structural interlayer DuPont™
SentryGlas®; and, verifying structural safety through soft and hard body impact tests
and residual strength capacity tests.
3.2. Fabrication and Installation Challenges

At the time of writing, the glass structures for the SoHo Store refurbishment are in
fabrication. Therefore this section focuses on the challenges presented at Hamburg.
Figure 13a and b: Installation of the bridge, staircase and balustrade of Apple Store Hamburg.
The design of fittings presented the glass contractor, Seele Sedak, with a number of
challenges. For instance, the use of laminated insert fittings restricts the ease with which
construction tolerances are accommodated. With standard site-installed bolts, holes in
the glass can be over-sized. With laminated insert fittings, the construction tolerances
have to be accommodated within the fittings. Some fittings need to accommodate
significant differential movements during the life of the structure (as presented earlier).
Furthermore, the rigidity of some fittings needs to be accurately controlled (as presented
earlier). These challenges were overcome with a development process involving the
consideration of different solutions, eliminating some and fine-tuning others.
109
Challenging Glass 3
The stair stringer panels use the largest single sheets of glass Seele has ever fabricated.
Furthermore, these sheets are not rectangular. At each step in the fabrication process,
new equipment and methods needed to be considered. Each stringer panel contains 76
insert fittings. To fabricate the panel, it was necessary for sheets of glass, with
accurately positioned holes, to be laid over all of these fittings simultaneously. Several
different assembly methods were designed and tested to develop the best solution.
The stringer panels are 12.3m in length and weigh 4 tons. The site location meant that
access was limited. The duration permitted for panel delivery was restricted and could
only take place at night. Once at site, there was no crane to aid installation. To install
such a large and heavy panel with tight installation tolerances was a challenge that was
achieved by an experienced and highly skilled installation crew.
4. Conclusion
The engineering challenges presented by the Apple Stores in Hamburg and SoHo, New
York were evident from the outset. The design team embraced these projects and their
inherent challenges, exercising substantial efforts in research and development to solve
the problems faced whilst maintaining the aesthetic goals. The extensive skills and
experience of the glass contractors, Seele Sedak, teamed with the glass consultants,
Eckersley O’Callaghan, were key to the success of the projects. Innovative design
solutions have transpired allowing the realization of elegantly balanced glass structures
whilst extending the envelope of capability in glass engineering. Furthermore, the work
carried out has identified areas where further research and development will
significantly aid further advancements in the industry.
5. Acknowledgements
Apple Computers Inc., Bohlin Cywinski Jackson, Seele Sedak GmbH & Co.KG,
Tripyramid Inc.
6. References
[1] Bauchman, H; Amman, W; Structure Induced by Man and Machines, Structural Engineering Documents
3e, IABSE – AIPC – IVBH, Zürich, Switzerland, 1987.

[2] Kerr, S. C.; Bishop, N. W. M.; Human induced loading on flexible staircases, Engineering Structures 23
2001, pp. 37-45.
[3] Kerr, S. C., Human induced loading on flexible staircases, submission for doctorate of philosophy,
Mechanical Engineering Department, University of London, UK, 1998.
[4] Newland, D. E., Vibration of the London Millennium Footbridge: Part 1 Cause & Part 2 Cure, web
article, Department of Engineering, University of Cambridge, UK
http://www2.eng.cam.ac.uk/~den/ICSV9_06.htm & http://www2.eng.cam.ac.uk/~den/ICSV9_04.htm
[5] British Standards Institution, Guide to evaluation of human exposure to vibration in buildings (1 Hz to
80 Hz), BS 6472:1992, BSI, London, UK, 1992.
[6] Willford, M. R.; Young, P.; A Design Guide for Footfall Induced Vibration of Structures, The Concrete
Centre, CCIP-016, Surrey, UK, 2006.
[7] Lenk, P.; Coult, G., Damping of the Glass Structures and Components, Proceedings of the second
Challenging Glass Conference (CGC2), Delft University of Technology, the Netherlands, 2010, pp. 341-
350.
[8] Kala, J.; Bajer, M..; Barnat, J.; Smutny, J; Experimental approach of the single pedestrian-induced
excitation, Slovak Journal of Civil Engineering 2010/4, pp. 14-20.
110
doi:10.3233/978-1-61499-061-1-111
Glass Walls Carrying the Roof and

Withstanding the Wind Load on the Façade:
Conservatory of the Museum in Dordrecht
and Raaks Glass Cube in Haarlem
Rob Nijsse
Technical University of Delft, the Netherlands, r.nijsse@tudelft.nl
In this paper three special glass structures are treated. Two are buildings; one is a
surprise: a work of Art. The material glass plays a dominating structural role in
them. The structural ins and outs of each type of building are considered with a
special interest in the safety that these glass structures have to provide in order to be
accepted by the authorities as structures that comply with the official standards that
have to be respected. The three realized structures are as well a proof that glass
slowly is turned into a trustworthy building material.
Keywords: Glass, Conservatory, Cube, Glass Angel.
1. General
Since 1980 all over the world experimental glass structures are designed and, important,
starting to get built. The architectural interest in glass is understandable, it is transparent,
water does not affect it and it needs no painting or protective layers. The thorn in the
side of glass is its brittle behavior under tensile stresses. Only recently thanks to
lamination and toughening of the glass we are able to make safe glass structures that
have the tendency to be ductile instead of fragile. As noted before all over the world
(safe) glass structures are being built and in this paper three special buildings/ structures
are presented. First an all glass Cube, 7 x 7 x 7 meter, in Haarlem (NL) that acts as
entrance building to a underground parking. Second a 30 x 4 x 3 meter all glass
extension to a historic building housing a museum in Dordrecht (NL). In this
conservatory the walls of glass carry the glass roof. The last but not the least structure is
an 4 meter high all glass statue of the archangel Michael that now stands on the church
square of Zwolle (NL). This statue is formed from water jet glass panels that are glued
on top of each other by a 3M gluing tape. All this three examples give a good insight
into the structural possibilities of glass as structural trustworthy building material.
111
Challenging Glass 3
2. A Glass cube in Haarlem (NL).
2.1. General.
A part of the old city center of Haarlem (NL) came free due to the demolishing of an old
factory. Under this square an underground parking garage was planned. This
underground garage requires an entrance building; for this purpose an all glass cube was
designed by the architect Kraayvanger Urbis. ABT/ Rob Nijsse was invited to work out
the structure for this all glass cube.
The challenge with all glass structure is to reduce the number of steel connections to an
absolute minimum. So the following elements were selected to compose the all, well as
all as possible, Haarlem Glass Cube. For the roof six glass panels supported by beams
that rest on the glass fins in the façade. For each of the for walls or facades of the cube,
again, six glass panels, standing on top of each other for gravity and supported by glass
fins for the wind. In figure 3 the overall structure of the Haarlem Glass Cube is sketched.
The glass cube has been constructed from March to May 2010.
Figure 1: Structure of the Glass Cube
2.2. Roof structure of the Haarlem all glass cube.

When loaded vertically by dead load, snow or wind the vertical glass panels of the roof
are loaded in an unfavourable way. This means that they will have to be thick to resist
bending. That is why they have to be two times laminated 12 mm thick glass, heat
strengthened. Also each panel of 2.35 x 2.35 meter had to be supported on the four
edges of each panel. This could be done with a glass beam but we decided to do it with,
a slender as possible, steel hollow section 70 x 140 mm, t= 6 mm. To improve the
stiffness a steel suspension cable D= 8 mm was put under the steel beam. To emphasize
on the glass character of the Haarlem cube an all glass rod D=30 mm, l= 400 mm was
used to the steel cable separated from the steel beam. The plane of the roof also has to
provide lateral stability by forming a stiff plain. For this purpose the connections
between glass panels are detailed in such a way that horizontal stability forces can be
transported by them, see figure 3.
112
Glass Walls Carrying the Roof and Withstanding the Wind Load on the Façade: Conservatory of the Museum
in Dordrecht and Raaks Glass Cube in Haarlem
Figure 2: The realised glass cube on the Raaks plein in Haarlem.
113
Challenging Glass 3
2.3. Facade structure of the Haarlem all glass cube.

The Haarlem cube has four facades of glass; three made all out of glass and one with a
door; a glass door in a steel frame inside one of the standard 2.35 x 2.35 meter panel of
the facades. Two large glass fins in each facade spanning from floor to roof take care of
the wind loads on the facade. The glass fins are 7050 mm long each and are made out of
one piece; this means that they have been made in China.
A big length indeed; but with a width of 450 mm the deformation by wind load is
reduced to only 3.3. mm. The glass composition of the fins is 3 times 12 mm, heat
strengthened glass. Since the wind load can both pull and push against the facade the
free edge of the glass fin can be loaded in compression, hence the danger of instability
or partial buckling is present. Therefore we decided to attach one cable system at the
rear of the glass fin to reduce the buckling length from 7.05 meter to the half; 3.525
meter.
Due to the support of the glass fins the glass panels of 2.35 x 2.35 meter have two
supported edges and can be made out of heat strengthened glass of one layer of 10 mm
and other layer of 8 mm in laminated configuration.
Besides taking up the wind load perpendicular to the facade the four faces of the cube
also have to take up the stability forces summoned by the wind load on the cube in
general. This meant that we had to build in the possibility to create the occurrence of
compression-diagonals in the plain of the glass facades. This aspect required connecting
details that could transport these compression forces without flaws. Since the
compression forces are in the plain of the glass panels we incorporated in the point
where four glass panels meet a steel cross thick 6 mm in the 12 mm (=theoretical size)
joint to make sure that the joint fill at this point would not be too fluffy and is able to
transport relatively large stresses over this location. A steel plate was mounted over
each cross point of the various panels that the facades are made from to fix/ clamp the
glass panels to the glass fins. See figure 3 and 4.
This system of connections and glass panel that provide stability was calculated with a
finite element program showing the so called stress trajectories. These trajectories
indicate the flow of stresses through the plane of the façade.
114
Figure 3, the flow of stability stresses through an all glass façade

(with steel crosses in the joint where four glass panels meet).
3. A Glass Conservatory in Dordrecht (NL).
3.1. General.
Making extensions to historic monuments is a controversial subject. Questions of style
and authenticity regarding the monument are hard to answer in a acceptable way.
Unless…..the structure is made out glass for glass is transparent and the view on the
historic facades of the monument is not blocked or altered. So when during the
renovation of the 16th century building of the Museum of Dordrecht the wish was made
for a conservatory adjacent to the restaurant inside the building the architect Dirk Jan
Postel came with the proposal to make an all glass conservatory. The monument
commission, responsible for keeping the heritage as good as possible, accepted this
extension to the monument. It only implied that the use of steel elements, like beams
and columns, was prohibitive. Not an easy task to make a safe and sound structure with
the help of this traditionally accepted structural elements.
115
Challenging Glass 3
Figure 4: The glass conservatory to the monument of the Dordrecht Museum.
3.2. Structural concept of the all glass conservatory.

A conservatory is basically a roof and a façade wall. Since the conservatory had to made
out of glass the roof was simply a panel of insulated glass and for the façade wall the
same choice was made. In order to make a safe structure out of these two elements
optimal use of the existing building had to be made. The monumental thick brick walls
carry the roof load and provide lateral and horizontal stability to the all glass
conservatory. The singular aspect of this choice for roof and wall was that the glass
façade wall had to carry the weight of the glass roof. Glass walls that carry the wind
load are a well-known feature but glass wall carrying a roof is a seldom shown
phenomenon. But ABT experimented one time with this, (no) coincidentally with
architect Dirk Jan Postel as well, in look out room at a holiday home far away in the
Burgundy woods in France by loading the wooden roof on the four glass walls of the
room. In Dordrecht things were more complicated, the span of the roof, about 4 meter
implied that glass beams had to make. These glass beams could be easily mounted
against the brick walls of the monument but attaching the glass beams to the glass
116
façade wall is different subject. After studying a lot of possibilities the choice was made
for a bolted connection of an almost transparent appearance. Only the essential steel
members were made. In this connection the load of the roof was transferred to the
structure of the all glass façade wall. There the next structural challenge had to be faced
since the glass wall was loaded by the roof on top and loading by the wind horizontally
to its plane. This challenge was tackled by carefully calculating and controlling the
buckling of the glass panes loaded on top by the roof and loaded transversal by the wind.
A critical combination since the buckling due to the load on top is very much negatively
influenced by the horizontal bending action of the wind. At the end we were able to
prove by finite element computer calculations that a two times 10 mm thick glass panel
(the inside panel of the insulated glass panel) could carry the roof load and the wind
with sufficient safety.
4. An all Glass Statue; the archangel Michael in Zwolle (NL).

The official spiritual guardian of the city of Zwolle in the Netherlands is the biblical
archangel Michael. He is pictured on the heraldic shield of the city and stands as a metal
statue on top of the roof of the main church of Zwolle. Dutch artist Herman Lamers was
asked by the municipality of Zwolle to design a modern version of a statue of the
archangel Michael to be placed on the market square before the church. Herman Lamers
decided that a three meter high glass angel would be the perfect way to depict a modern
Michael. But how to realize this in an economic way? For casting a three meter high
glass angel in a mould with molten glass is a very difficult and costly way. After
studying numerous methods we decided to make the glass angel by stacking panels of 8
mm glass. With the help of a computer generated image of the glass angel we were able
to make cross sections of the statue at each level. This information was send to the glass
producer that cut out the correct shape from a rectangular sheet of glass. Next critical
point was how to connect the about 370 glass panels on top of each other to make an
integrated glass object. After analyzing various possible types of glue we came to the
choice for 3M double side tape. This type of tape is also resistant to the outside
conditions; rain, frost etc. that the glass statue of Michael has to cope with. Executing
this gigantic stacking puzzle was quite a challenge to Herman Lamers and his team but
the completed the difficult task perfectly. A very nice side effect was obtained by
stacking also the left over glass panels after cutting out the shape of the archangel
Michael. In this way a glass angel inside a glass box was obtained. Two statues of the
glass archangel, a positive and a negative, for the prize of one.
117
Challenging Glass 3
Figure 5: The glass Archangel Michael of Zwolle.
118
Figure 6: The negative, glass statue of the archangel Michael.
119
Challenging Glass 3
5. Discussions and conclusions

The three described projects, the Glass cube, the Glass conservatory to a monument and,
especially, the glass angel prove that glass structures are more and more accepted in the
modern structural society. The big advances of glass; its transparency and its non-
detainable properties (no rust etc.), combined with a structural safety concept will lead
to a new type of building that is decoration to modern society. Also the fact that glass is
easily and already massively recycled is a important sustainable fact. The glass angel
proves that gluing of glass is the ideal way to connect in a safe and adequate way
individual glass elements. The point of the durability of the gluing action is however
still a much debated issue.
6. References
[1] Nijsse, Rob, Glass in Structures. Elements, Concepts, Designs, Birkhäuser, Basel, Switzerland, 2003.
120
doi:10.3233/978-1-61499-061-1-121
Inclined Glass Fins for the King Abdulaziz

Center for World Culture
Silvia Prandelli, Damian Rogan
Buro Happold Ltd, UK, www.burohappold.com, silvia.prandelli@burohappold.com,
damian.rogan@burohappold.com
King Abdulaziz Center for World Culture is a new development in Dhahran, Saudi
Arabia designed by Snøhetta architects where five complex geometry buildings,
called Pebbles, accommodate different civic and cultural functions. Penetrating the
ground level podium are three open-air spaces called Oases, each with a glazed wall
leading into the basement areas. Each Oasis façade is fully glazed, up to 15m tall
and constituted by faceted double glazed panels tilting in and out of plane, all with
different shape. The architectural intent called for a fully transparent design with
flush glass. Buro Happold proposed a solution using glass fins as secondary support
and a toggle system for retaining the glass. Buckling of the slender fins is resisted
by horizontal glass beams. The team was able to build on previous experience in
considering buckling and restraint issues for tall the slender fins, but the fins’ 15°
lateral inclination presented a significant stability challenge. The adopted design
connects each fin to one side of the double glazed units, so each set of fins and
glazing units acts a stable L-shaped structural unit. Unique connection details were
developed to achieve the necessary structural behaviour. Detailing around access
doors and movements of supporting structures posed further complications which
have had to be resolved in an architecturally sensitive manner. The glass fins and
the façade panels have been analysed both individually and in a global model to
verify the stresses and deflections were within the limits in all the different
conditions. This paper presents the design criteria, the findings and the solutions
adopted to achieve the architects’ vision.
Keywords: Glass, Fin, Complex geometry, Façade, Faceted glass, Toggle glazing
1. Introduction
King Abdulaziz Center for World Culture is a new initiative of the Saudi Aramco oil
company to promote cultural development within the Kingdom. Following an invited
architectural design competition in 2007, the Norwegian architectural firm Snøhetta was
selected to design this prestigious cultural facility.
Located in Dhahran in the Eastern Province, the Cultural Center will provide for a wide
range of activities serving the local population and becoming a cultural landmark on a
regional, national and global horizon. When completed the project will contain some
45,000m2 of diverse cultural facilities, including an auditorium, cinema, library,
exhibition hall, museum and archive.
The Center is the company’s gift to the country at their 75th anniversary and is to be
built in an extremely demanding environment. High temperature, frequent sand and dust
storms, and condensation will take their toll on the façade of this building.
121
Challenging Glass 3
The project is made up of five pebble-shaped buildings, each of largely solid wall
construction and wrapped in stainless steel tubes. These sit on a common podium level
which unites the masterplan. Three open-air, landscaped Oases are sunken into the
podium to allow light into the basement museum areas. The largest Oasis is 75m x 40m
and the smallest is 37 x 27 m.
Figure 1: Architect’s rendering of the masterplan of the Center
2. Architectural intent for the Oasis facades

The Oasis façades all follow an irregular ellipsoidal shape in plan and are inclined in
section. The irregular shape of the walls is formed of flat glass panels facetted in plan
and supported by laminated glass fins with a maximum height of 11m and typical
spacing of 2m. The glass is all double-glazed and has high-performance coatings to
satisfy the environmental performance criteria. An architectural feature of the Oasis
façades is that the glazing joints are inclined in elevation, meaning that the glass fins are
inclined both parallel and normal to the glass surface. The geometry of the Oases is
summarized in the table below:
122
Inclined Glass Fins for the King Abdulaziz Center for World Culture
Table 1: Oases geometry
Lower Deck: Vertical fins

Fins tilted parallel to the glass (up to 13°)
Length up to 3.4 m
Upper Deck: Fins tilted parallel to glass (up to

17°)
Fins tilted perpendicular to the glass –
outwards (up to 16°)
Administration Oasis Length up to 5.4m
Fins tilted parallel to glass (up to 13°)
Fins tilted perpendicular to the glass – inwards

(up to 23°)
Length up to 11 m
Children’s Oasis
Vertical fins: length up to 11 m
Great Hall Oasis
It was the architects’ intention to have the glass fins manufactured in one element as
well as the double glazed units. However, as both elements had a maximum height of
11m, it was decided that these should be divided into three glass pieces to be more cost
effective and improve procurement options, since only a single fabricator was identified
who could produce DGUs as single elements up to the 11m height required.
3.1. Calculation approach

The tilt parallel and normal to the double glazed units caused different issues in the fin
sizing. The following parameters had to be checked for each fin:
x Deflections had to be within the limits of the selected retention system

x Short-term stress from wind loads
x Long-term stress from self weight due to the inclination
x Lateral torsional buckling
123
Challenging Glass 3
The structure was analysed in Autodesk Robot FEA software, with both local models of
worst-case fins and global models to verify the behaviour of the overall structure.
Allowable stresses were taken from the IStructE Structural Use of Glass in Buildings
[1], and allowable deflections were taken from the CWCT [2] and industry guidance.
The glass fin design was verified against both to Australian Standard AS 1288 [3] and
the design method suggested by Luible in the IABSE guide Structural Use of Glass [4].
The latter guide was crucial as it provided a method for considering the buckling
behaviour of the fin under long-term load, when the laminated glass construction would
exhibit non-composite behaviour.
3.2. Structural strategy

The earliest structural concept for the fin wall considered each fin to be toggle-fixed to
the DGUs, thus having only a nominal lateral restraint against buckling. Architectural
restrictions dictated a maximum depth of 500m for the tallest fins, and hand calculations
showed that this depth could satisfy the deflection criteria. However, more detailed
calculations quickly showed that minor-axis bending due to the inclination of the fins of
up to 13° in the glazing plane would cause excessive stress and long-term sag.
It was realised that a mechanism must be found for supporting each fin laterally, both to
resist bending under self-weight and to provide stability against buckling. It was
decided that the strategy for doing this would be to join each fin to an adjacent glazing
panel, thus forming an L-shaped unit.
Figure 2: Fin and glazing paired to form a stable L-shaped unit – model and deflection results under wind
pressure
With a strategy in mind, a common approach could be developed for each Oasis, so that
the same detail for supporting the glass fins and connecting the glass panels to the fins
124
could be used everywhere. But while the concept of the L-shaped unit solved the
challenge of stabilising the inclined fin, further challenges had to be overcome to turn
the concept into a workable design.
Firstly, the glazing would provide support only to the front edge of the fin, leaving the
back edge only indirectly supported against its self-weight. Study models were analysed
in Robot using a reduced long-term/high temperature modulus for the laminate material.
These showed that the rear edge of the fin would still sag to over 20 mm, putting a large
shear stress on the laminate and resulting in high tensile stresses on the minor axis of
the fin. To solve this, a horizontal element would be introduced behind each panel joint,
providing a torsional restraint to the vertical fin while also supporting the horizontal
edges of the glazing panels. In order to reduce the visual impact of these members,
horizontal glass fins were adopted.
Secondly, the lean of the fin placed a lateral force on the glazing panel that had to be
resisted and transferred to the supporting structure. Since the glazing panel was broken
into three separate DGUs vertically stacked, instability could happen if the panels were
merely mounted on top of one another. To prevent this instability the units were
mechanically fastened to one another, distributing their own weight through the setting
blocks.
Lastly, these structural requirements had to be realised in a set of details that preserved
the architectural requirement for transparency and also dealt with the complexities of
the adjoining structures: notably the supporting podium slab, which experienced large
deflections in service; interfaces with the free-form building structures; and interfaces
with the doors in the Oasis façades.
3.3. Findings and solutions

Having identified responses to the principal challenges, the proposed solution was
analysed in a comprehensive FEA models to finalise the design.
Detailed models were produced of the worst-case L-shaped fin and panel on each Oasis,
and global models were produced to represent each Oasis. The global models
considered static loads and were served to verify the reactions onto the supporting
structures and the deflections that had to be accommodated in the glass retention system.
Meanwhile, the detailed models used a combination of static and non-linear buckling
analyses to determine the member sizes and connection forces.
Figure 3: Max deflections on the Administration Oasis upper deck under wind pressure
125
Challenging Glass 3
The static analyses were carried out for both long- and short-term loading. The long-
term model used a low modulus for the PVB laminate, with an equivalent thickness
determined using the method of ASTM E-1300 [5]. The short-term analysis considered
wind loading and considered a greater equivalent stiffness for the load duration. The
stresses derived were then combined in ratio to the long- and short-term allowable
stresses of the IStructE guide. From these models, a fin build-up of 4x12mm laminated
toughened sheets was proposed.
Finally, the buckling analysis was carried out to ensure that an adequate safety factor
was ensured. While the buckling analysis considered only the short-term loads and the
higher short-term stiffness of the fins, lateral loads were included to represent
imperfections and the possible lateral sway of the system. The analysis proved a safety
factor of 4. The FEA was then checked against the buckling formulas of AS 1288 and
the Luible method, each altered to account for the additional lateral forces on the fins.
These likewise proved the design to have an adequate safety factor against buckling.
3.4. Façade details

The architectural aesthetic for the Oasis façades did not lend itself to exposed steel
bolted connections. However, as the system relied on the transfer of permanent loads
between elements of a relatively high magnitude, bolted connections were felt to be
unavoidable. The challenge, therefore, was to realise the connections in a discrete
manner.
The major detail to be resolved was the one that would enable the fin and glazing panel
to be considered as a single L-shaped unit. A stainless steel U-shaped channel was SSG-
bonded to the front edge of the inclined fin, while L-shaped carrier profiles were bonded
to the vertical edges of the DGUs. The carrier profiles were regularly bolted to the steel
channel to transfer wind loads and the self-weight of the glazing units onto the fin, and
the lateral load of the fin onto the glazing units. At the joints between stacked DGUs,
thin, flat splice plates were then bolted to the rear of the carrier profiles. These
transferred shear and axial loads induced by the lateral load of the leaning fins and form
a continuous unit.
Figure 4: Typical horizontal section – Children’s Oasis
126
4. Contractor’s design
The final design and construction of the Oases was awarded to Seele Austria, thanks to
the extensive documentation provided at the tender stage.
A similar structural approach was proposed which showed base supported fins and
horizontal glass fins at glass joints. The vertical glass fins proposed were constituted of
two layers of 15mm low iron tempered heat soak tested glass with a 1,52mm PVB foil.
Figure 5: CGI of the connection at the horizontal fin
Tests were also carried out to verify the load capacity of the bolted connections at the
vertical fins through pulling tests, the mechanical strength of the glass through bending
tests and the quality of the toughening through fragmentation.
Figure 6: Schematic setup of the test for load capacity
An average load capacity of 214 kN has been found for laminated samples (15+15mm).
127
Challenging Glass 3
Figure 7: Origin of the fracture during the pulling test on one of the laminated samples
5. Conclusions
This demanding project underwent through many different design iterations from the
concept stage to the tender submission. The involvement of Buro Happold Façade
Engineering had been an essential link between the architects and the construction
industry throughout the scheme development. The details presented for the tender
submission were the result of a successful teamwork which could bridge over different
branches of the façade knowledge, including structural and architectural design,
building physics and budget costing. The facades of the Oases are currently under
construction by Seele, based the current design on Buro Happold solution presented in
this paper.
6. Acknowledgments
The authors would like to thank Saudi Aramco Oil Company for their support, Seele
Austria (http://www.seele.com/) for the availability to share their knowledge and

suggestions during the design phase and beyond. A special thank also goes to Peter
French and Margrethe Lund from Snøhetta architects for their endless creativity.
7. References
[1] Institution of Structural Engineers, Structural Use of Glass in Buildings, 1999
[2] Centre for window and cladding technology, Standard for systemised building envelope, 2005
[3] Australian Standard 1288, Glass in Buildings, 2006
[4] International Association of Bridge and Structural Engineering, Structural Engineering Document 10:
Structural Use of Glass, 2008
[5] American Society for Testing and Materials, Standard Practice for Determining Load Resistance of
Glass in Buildings - E 1300, 2009
128
doi:10.3233/978-1-61499-061-1-129
Design of Suspended Glass Ceiling

Structure in High Seismic Hazard Zones
Toru Takeuchi, Kenichi Sugizaki, and Koichi Yasuda
Tokyo Institute of Technology, Sat-Consultant, Yasuda artier, Japan,
This paper briefly reports on the structural design of the large (26 m × 24 m)
sashless glass ceiling that covers the entrance stairs of the underground shopping
mall in front of Kawasaki Station, located in Kawasaki City, one of the largest
satellite cities in Tokyo. The glass ceiling is totally suspended from an L-shaped
gate composed of spatial truss frames. Because these structures were constructed in
a heavy seismic zone, they were carefully designed to withstand well during
earthquakes whose maximum ground acceleration reaches 350 cm/sec2, yet use a
minimum number of effective elements to maintain transparency of the structure.
On March 11. 2011, the Great East Japan Earthquake hit this structure; however,
none of the components including the glass panels were damaged.
Keywords: Glass ceiling, Suspended, Steel structure, Seismic design
1. Combination Roofs for Kawasaki Station

Kawasaki is one of the largest satellite cities in Tokyo, and more than 185,000
commuters pass through the Kawasaki Japan Railway station. More than half of these
commuters exit the station and descend to the underground shopping mall en route to
downtown because the ground level is covered by a bus terminal and obstructed by a
private railway (Fig. 1). The architect proposed a stacked pair of bright and transparent
roofs for the entrance stairs leading from the gate of Kawasaki Station to the
underground mall as shown n Figs. 2 and 3. The lower roof has a sashless glass ceiling
of 26m × 24m that covers the entrance stairs of the underground mall, and around
100,000 commuters pass under this glass ceiling everyday (Fig. 4).
Figure 1: Facilities in front of JR Kawasaki Station Figure 2: The Combination Roof *
129
Challenging Glass 3
Figure 3: Suspended Lower Glass Roof * Figure 4: Glass Ceiling over Entrance Stairs *
This glass ceiling structure is totally suspended from the upper roof structure, and the
structure allows sunlight into the underground mall owing to its transparency.
Constructed in a heavy seismic zone, these structures are required to be carefully to
withstand an earthquake with a maximum acceleration of up to 350 cm/sec2 at ground
level. The detailed structural design, including the seismic design, is described below.
2. Structural System
The structural system of the combination roof is shown in Figs. 5 and 6. The upper roof
covers a 40m × 25m area, and the structure comprises lines of L-shaped steel-tube truss
columns and beams. These are connected to each other by tube members, forming a
space-framed wall and roof. The tubular members used range from 356 mm in diameter
and 16 mm thick to 216 mm in diameter and 8.2 mm thick. The edge of the roof is
supported by a line of circular steel pillars each with a cross section that is 406 mm in
diameter and 19 mm thick. The lower roof structure supporting the glass ceiling forms
another L-shaped frame comprising plate beams and columns. The vertical weight of
the lower roof is totally supported by the upper roof, suspended by 42 high-strength
steel bars. The lower roof has several diagonal members fixing it to the ground to resist
against the horizontal forces produced by earthquakes. Consequently, the glass ceiling
structure is dependent on the upper roof to handle the vertical weight; however, it was
independently designed to withstand seismic forces.
130
Design of Suspended Glass Ceiling Structure in Heavy Seismic Zone
Pillars
S
W
E N
Figure 5: Structural System for the Combination Roof Figure 6: Structural Elements
Upper Roof
Suspension Rods
Lower Roof (Glass Ceiling)
Entrance Stairs
Foundation
Underground Mall
Figure 7: Section View of the Combination Roof

and Underground Mall Figure 8: Structural Detail of Upper Roof
A section view of the combination roof is shown in Fig. 7. The foundation of the roof
sits on the roof of the underground mall structure, and various service utilities such as
electric lines pass between them. Therefore, the foundation had to be designed with a
very limited height (400mm above stairs) and limited reaction forces. Figure 8 shows
the structural details of the upper roof. Various types of casted block were employed for
the joints of the tube sections, together with ordinary welding branch joints. To
minimize the foundation height, exposed column base systems with prestressed anchor
bolts were employed.
3. Glass Ceiling Structure and Seismic Design

The structural system of the lower roof structure for the glass ceiling is shown in Figs. 9
and 10.Toughened glass panels of 1.8 m×1.8 m are supported by a point support system
[1]
of steel structures. The steel structure comprises plate beams that are 300mm high by
40 mm thick. These beams are, connected to each other with tapered steel "glass
131
Challenging Glass 3
Figure 9: Structural Detail of Glass Ceiling

Figure 10: Detail of Glass Supporter
S
E
W
N
Figure 11: Entrance View of Glass Ceiling Figure 12: Bottom View of Glass Ceiling
132
supporters" designed in the shape of airplane wing in mind. The shear stiffness of the
roof membrane is secured against movement by the roof braces of high-strength steel
rods. Because the wind pressure on the roof can lift it up the roof, tie-down anchor and
reversal parabola-shaped strings were provided to keep the structure in place. Diagonal
tubes and braces were also provided for the north-south (NS) direction to provide
resistance to the horizontal forces produced by earthquakes. Fig. 11 shows the view
from the north. Stiffness and strength against movement in the east-west (EW) direction
are secured by L-shaped beam-column frames. Fig. 12 shows a drawing of the view
from the bottom of the glass ceiling.
The high-strength steel materials used for tensile members and their connections are
summarized in Table 1. PC-bars have yield strength over 930 MPa, and employed the
rolled thread to keep the high performance against fatigue failure. Fig.13 (a) shows the
fatigue test results of high-strength bars with various materials and threads. Rolled
thread has much higher performance than machine threads, even used for high-strength
PC-bars. In this structure, PC bars with rolled thread are mainly used for tensile
members, and designed to keep the stress less than 600MPa against seismic and wind
loads. Fig.13 (a) indicates that the fatigue capacity in this amplitude exceeds 105 cycles.
The joints for these tensile members were designed using folk-ends and pins with high-
strength steel material whose yield strength are over 500Mpa and 750Mpa respectively.
Table 1: Steel Material and Mechanical Properties of Tensile Members

Member Steel Yield Strength Ultimate Strength Elongation
Type Material (N/mm2) (N/mm2) (%)
PC-bar PC-B 930 1080 5
Tie-rod HT690 440 690 20
Folk-ends SCM440H 500 700 15
Pin SCM440H 750 850 15
㻭㼟
1000 㻭A
㻠㻡㻌
45
PC-bar (Rolled Thread) 㻭㼝

Stress Amplitude (N/mm2)
㻼
㼞
㻭㼠
㻾
Tie-rod (Bumped Thread)

PC-bar (Machined Thread)
㼑
SM490-B
SM490-R
SM490-C
PC-R K3.➼⣭
Level
PC-C
HT690-B
HT690-C
100
1000 104 105 106 107
Fatigue◚᩿ᅇᩘ
Failure Cycles
(a) Fatigue Performance of High-strength Steel Rods (b) Design and Tests of Folk-end Joints
Figure 13: Performances of Tensile members and Their Joints
133
Challenging Glass 3
They are designed to satisfy their maximum strength exceeding the yield strength of
connected rods with the following conditions [2].
A45V yb (1)
!N y
2
where, A45:diaonal section of folk-end in Fig. 13 (b), Vyb: yield strength of folk-end, Ny:
yield strength of connected rods. Eq. (1) gives similar results as BS 5950, and the effect
of the pin eccentricity can be adequately evaluated. Tensile and fatigue tests were
carried out including these folk-end connections, and it was confirmed that the fractures
were decided by rods themselves.
For seismic design, the Japanese structural design code requires structural safety against
an earthquake with a 500-year return period; such an earthquake may have up to 1.0 G
of horizontal acceleration in an elastic manner. In plastic design, the design shear force
can be reduced depending on the structure's ductility. To confirm the collapse
mechanism and ductility of the system, pushover analyses were carried out for both the
upper and lower roof structures. The dynamic natural period of the upper roof was
estimated as 0.87 sec in the EW direction and 0.51 sec in the NS direction, and that of
the lower roof was as 0.53 sec. in both directions. Because each roof was confirmed to
move independently, seismic design for the two was carried out separately. Results of
the push-over analyses for the glass ceiling structure in both directions are shown in Fig.
14 (a), (b). In the NS direction, the structure performed in an almost elastic manner until
a shear force rate of 0.7 G with a spare safety factor of 2 was attended; i.e., until the
beams buckled. In the EW direction, ductile collapse was confirmed for the plastic
hinges at the column and beam ends around 1.2 G. Similarly, the upper roof structure
2,000 BeamBuckling
NSdirection Brace
1,800 Buckling
1,600
EWdirection
1,400 Beam
Buckling
ShearForce(kN)
Column
ỈᖹຊQ(kN)
1,200
Yieldhinge
CollapseMechanism Column
1,000 Yieldhinge
1.0G(853kN) NSDirection
800
600
EstimatedUltimateStrength
400
200
RequirementinJPNcode
Column
0 Beam
Yieldhinge
0 100 200 300 400 500 600
StoryDrift(mm)
ኚ఩ 㹢 (mm) EWDirection
(a) Load-Displacement Relationship (b) Collapse Mechanism of Glass Ceiling Structure

Figure 14: Push-over Analysis of Glass Ceiling Structure
showed ultimate strengths of 0.88 G in the EW direction and 1.68 G in the NS direction.
Both satisfied the code requirements with large margins.
134
4. Construction
Because construction of the roofs had to be carried out above the daily commuters,
assembly plans were discussed very carefully. The stairs could be closed only three
hours per day between 1 and 4 a.m. Also, the existing roof needed to remain in place
until the new roof was completed, to avoid the entry of rain into the underground areas.
To solve these problems, a sliding construction method was employed, as shown in Figs.
15 to 17.
The upper roof was built in three parts, then assembled and finished at the southern end
of the final location. It was then shifted northward to its final place. Last, the old roof
was demolished and replaced with the glass ceiling. The mock-up scene at the fabricator
is shown in Fig. 18, and the completed glass ceiling structure is shown in Fig. 19. The
glass ceiling structure is estimated to have a damping ratio of 3-5 % owing to the
sealing materials between the glass panels [3],[4], while these glass panels need to transfer
part of shear forces produced by earthquakes. To avoid the fracture at glass supporting
points by possible seismic shear forces transferred along deformations of the structure,
the site fillet welding at each point was increased from 9 mm to 15 mm as in Fig. 20,
whose strength exceeds those of glass seals.
Assembly SlidingArea
60tf yard
RafterCrane
SlidingConstruction
JR
Station
OldRoof
Underground DailyCommuters
Mall
Figure 15: Sliding Construction of Upper Roof Figure 16: View after the First Slide Operation
Figure 17: Sliding Operation Figure 18: Mock-up of Glass Supporter
135
Challenging Glass 3
Figure 19: Completion of Glass Assembly
Site Fillet Site Fillet

Weld 15mm Weld 15mm
Figure 20: Details of Glass Supporting Points
136
Figure 21 Night View of the Combination Roof *
5. Conclusions
On March 11, 2011, Kawasaki Station was shaken by the Great East Japan Earthquake
with a maximum acceleration of around 150 cm/s2. The collapse of concert hall ceiling
in the other side of the station suffered a large collapse. The combination roof over the
stairs deformed largely at the upper roof; however, the glass ceiling was fixed to the
ground and no glass panels or associated elements were damaged. The structure met
seismic performance expectations, while simultaneously having an elegant, transparent
appearance and providing bright sunlight to commuters.
6. Acknowledgements
Overall project design for East Kawasaki Station Redevelopment was lead by Kawasaki
City Machizukuri-div., and Nikkenn Sekkei Civil Engineering Ltd. Construction was
carried out by Kajima-Goyo-Tekken-Sigeta JV. The photographs marked * in this paper

were taken by Mamoru Ishiguro.
7. References
[1] T.Takeuchi, H.Kitamura, K.Harada, K.Higuchi, A.Harada, M.Iwata 㸸 A Practical Design and
Construction of Tension Rod Supported Glazing, IASS-ASCE International Symposium 1994.4 Atlanta,
pp.684-693
[2] T.Ono, T.Takeuchi, M.Iwata, K.Ishida: Fatigue and Ultimate Strength of Tension Rods and Their
Connections, J.Struct. Constr. Eng., AIJ, No.537, pp.91-98, 2000.11 (in Japanese)
[3] T.Takeuchi, E.Tanaka, T.Hisada, F.Katano: Dynamic Properties of Sealing Material in Glass Walls,
J.Struct. Constr. Eng., AIJ, No.625, pp.481-488, 2008.5 (in Japanese)
[4] T.Takeuchi, Eiji Tanaka, Takashi Hisada: An Experimental Study on In-plane Dynamic Characteristics of
Glass Wall, J.Struct. Constr. Eng., AIJ, Vol.74, No.643, pp.1621-1630, 2009.3 (in Japanese)
137
doi:10.3233/978-1-61499-061-1-139
Designing a Glass Pavilion to Protect an

Ancient Greek Temple
Fred Veer, Phaedra Oikonomopoulou, Regina Bokel
Delft University of Technology, f.a.veer@tudelft.nl
The ancient temple of Apollo Epicurius at Bassae in the Peleponnese in Greece is a

unique world heritage monument. The limestone used however is very susceptible
to the elements (ice, rain, acid rain). Currently the temple is protected by a canopy.
A case study has been made to build a glass pavilion around the temple. This needs
to meet some very strict criteria. First it needs to be as invisible as possible.
Secondly as columns cannot be placed inside of the monument glass beams with a
24 m span are required. Lastly the climate inside has to be kept comfortable,
preferably using passive systems only.
Keywords: glass structural design, glass building physics design
1. Introduction
The ancient Greek temple of Apollo Epicurius in Western Peloponnese, built in the late
5th century BC (around 420 BC), is one of the most important Greek monuments of the
Classic era. It was built mainly of local limestone and it is said by Pausanias (VIII 41.7-
8) to have been designed by Iktinos, the same architect as Parthenon. The temple was
the first Greek monument to be listed by UNESCO as world heritage in 1986.
At the present, it is difficult to appreciate the beauty of the temple, due to an opaque
canopy which completely encases it and obstructs any exterior views of the monument,
prohibiting visitors from realizing its unique architecture. This protective canopy was
erected in 1987 as a temporary solution since it was essential to protect the specific
temple from further weathering; the extreme weather conditions that prevail during
wintertime in the area have caused the extensive deterioration of the limestone, leading
to severe erosion problems and structural damages of the temple. Since 2000, the temple
is under extensive restoration works in order to repair and strengthen its foundation and
restore the damage of the architectural members. Nonetheless, even after the completion
of the restoration works, the monument, if uncovered and exposed, will remain subject
to severe corrosion because of the combination of the building material and the
environmental conditions.
139
Challenging Glass 3
Figure 1: Temple in the 1950’s Figure 2: Canopy enclosing the temple
In this work, a novel, completely transparent shelter is proposed so as to ensure the

protection of the monument from the elements, and at the same time, to make it publicly
visible by following minimal intrusive architecture principles. In order to minimize the
visual obstructions, not only the cladding, but also the load bearing structure of the
shelter, is proposed to be made of glass. The design of such a protective shell is not only
an architectural challenge - where material and structural innovation should be
harmonized with the ancient character and heritage of the monument- but also a
challenge in terms of structure; since such a large glass beam span has not been realized
in the past. Furthermore, it is a challenge in terms of thermal performance -regarding the
high sun radiation intensity and temperature values that appear in Greece during
summer.
2. Structural elements
To build the proposed structure, glass structural elements need to be used. Using the
basis provided by Louter, [1], reinforced glass beams and columns can be created. It has
been demonstrated by Trösch et al. [2], that large beams are possible using modern
autoclave techniques. It is assumed that beams of 26 m length and columns of 16.5 m
height can be laminated using sentry glass plus (SGP) with reinforcement all around the
elements, extending the principle that was first shown in [3]. The safe design method
philosophy developed by Bos, [4], was used in analyzing the risks and determining
factors of safety.
Furthermore, a large double glazing element with dimensions 6*3 m is proposed

composed of plates of 8.8.2 PVB laminated glass using a thicker stainless spacer that
also acts as reinforcement.
Using these as structural elements a relatively light-weight structure can be made where
the cladding is actually a skin that stabilizes the structure.
It is assumed that the elements can be joined together along the whole length or at
enough points to allow adequate load transfer.
140
Designing a Glass Pavilion to Protect an Ancient Greek Temple
3. Architectural design
Aim of the new shelter is to function as a protective shell for the temple against the
elements, especially rain, ice and wind, which can extensively damage the monument.
At the same time, following the principles of the Venice Charter for the Conservation
and Restoration of Monuments and Sites, it intends to exhibit the ancient temple, by
placing the monument into full view, re-connecting it to the surrounding area1. The
special importance and architectural value of the temple of Apollo Epicurius, demand a
shelter that will not stand out, but contrariwise, be blended into the environment so that
a unified space appears. The presence of the shelter should actually be a form of
absence. The key for this is minimal design and maximum transparency: the overall
design takes advantage of the “keep it simple” principle, and both the cladding and the
load bearing structure are suggested to be made of glass.
Nonetheless, there are many more factors that were considered and incorporated into the
design for the achievement of a sustainable, long-lasting and efficient architectural
engineering solution. The main factors that lead to the final shape, form and overall
dimensions of the shelter are presented below:
x In pursuance of maximum transparency, both cladding and structure are

made of glass. For the cladding, flat glass panels should be used and not
curved ones, to achieve the minimization of visual distortions due to reflections.
The load bearing structure should be the minimum required while the corners
should be left empty, free of structure, in order to optimize the visual
connection with the monument.
x The shape and form of the shelter should be simple, so that the visitors’
attention will not be distracted from the monument. Following the proportions
and dimensions of the temple, the shelter has an orthogonal shape in plan.
x The overall dimensions of the shelter in plan are the result of two parameters:
(a) the incorporation of the girder crane into the shelter2 and (b) the restrictions
regarding the foundation location: Underneath the west side of the temple,
ancient supporting walls extend, which are considered part of the monument
and should not be damaged, unless imperative (see fig.5). In response to all the
above, the shelter is designed in an offset distance from the temple of

approximately 5 meters, resulting to an overall plan view dimension of the
shelter of about 50 x 25 meters.
x Crucial for the final shape of the shelter was its sustainability in terms of self-
maintenance, reflected primarily on the self-cleaning ability of the glazing;
one of the main problems the new shelter has to face is the cleaning of the
1
Venice Charter (1964), article 13: ‘Additions cannot be allowed except in so far as they do not
detract from the interesting parts of the building, its traditional setting, the balance of its
composition and its relation with its surroundings’.
2
It was decided that the glass shelter should be installed on site before the finishing of the
restoration program, which is expected to last a few decades more. Thus, the existing girder crane,
required for dismantling and moving the architectural members of the temple, should be
incorporated to the design.
141
Challenging Glass 3
glass surface from the dust. The less accessible area is the top surface and
accordingly the most difficult to be cleaned. Nevertheless, with a roof
inclination of approximately 10%, the glass roof can be self-cleaned by
rainwater.
Figure 3: Render of the proposed shelter
Regarding all the above, the proposed shelter design is as seen in figures 3 and 4. The
shelter is extended towards the west side, providing a loading/unloading area for the
existing girder crane, and a more spacious interior for visitors all the same. A one-sided
inclined roof covers the shelter, corresponding to the natural slope of the bedrock
underneath the temple.
Figure 4: Render of the proposed solution, front facade
142
The facades’ division module was based on two crucial aspects: the glass size
restrictions due to the manufacturing processes3 and the minimization of visual barriers
between the monument and the external environment. More specifically:
x The new shelter corresponds to the architectural proportions of the monument

by applying the exterior columns repetition module to the division of the
panels. More specifically, the facades’ widthwise division is equal to the
biaxial distance of the peripheral columns of the ancient temple (2.67 meters).
By displacing this division module by half the distance between two columns,
the panel’s joints –which are the only element conceived by the human eye as a
visible- are always located in the middle of the biaxial distance; consequently,
there are no visual obstacles between the vertical architectural elements and the
environment (see fig.5).
x For minimizing the optical boundaries and the structural connections the
structural frames are placed behind the panels’ joints.
x The horizontal division is as follows: The lower panels have a height of 7
meters, almost equal to the sum of the height of the columns and the base of
the temple together. A row of significantly shorter, openable panels is added
on the top of the facades in order to allow for the natural ventilation of the
interior space.
Figure 5: Section of the proposed shelter. The façade module of the shelter corresponds to the axial distance of
the columns. This section also indicates the incorporation of the girder crane and the proposed foundation
location of the new shelter.
3
Although glass is manufactured in a continuous ribbon of approximately 3.5 m wide, it is cut to
a standard final size of 6x3.21 m. Longer panels can be produced on demand but the
manufacturing cost is then significantly increased.
143
Challenging Glass 3
At various points in the architectural design models were made using the ANSYS 12
FEM software package. This led to the following basic structural design given in figure
6. Considering the nature of the structure full 3-D modeling was too time consuming.
The structure was modeled using shell 281 elements as flat box. The actual shape being
stiffer, the flat box gives a conservative result.
There were several critical steps in the progression of the structural design:
x Dimensioning the portal structure for gravity

x Dimensioning the portal as part of an infinite series of portals using the skin for
stability allowing for gravity, wind and snow
x Buckling resistance
x Validating the whole structure under static and variable loads
x Validating the whole structure against earthquakes
The most important parts will be discussed below in more depth.

Figure 6: Structural design
4.1. Dimensioning the portal as part of a repetitive series of portals using the skin for
stability allowing for gravity, wind and snow.
An assumption was made based on hand calculations that the columns were to be
composed of 3*10 mm stainless steel reinforced SGP laminate. For the beams two
beams of 3*10 mm stainless steel reinforced SGP laminate are used, one on each side of
the columns. The columns and beams are assumed to be made of large single glass
plates and not spliced. The plates are composed of 2 plates each of 2 sheets of PVB
laminated glass 8 mm thick with a 40 mm spacer, giving a total thickness of 32 mm of
glass.
144
The basic model is shown in figure 7. A wind force of 2000N/m2 was added to the right
together with a snow load of 1500N /m2 on the roof. The resulting x stresses (the
bending stresses in the beam are shown in figure 8.
Figure 7: Model of portal Figure 8: Bending stress in beam
4.2. Buckling resistance

Since the structure is designed as a mechanical whole, the skin supports the columns
and prevents sideways displacement. The beams are supported similarly in the
compression side by the roof panels. Although the individual columns would buckle
under the normal roof load, the overall structure prevents buckling. ANSYS modeling
also suggests that the structure would work in this way to prevent buckling. As this is
dependent on the detailing more explicit calculations and experimental testing are
required when the connection detailing is finished.
4.3. Validating the whole structure under static and variable loads
The whole structure has been modeled by repeating in equal distances the portals and
adding a front and back façade. Winds of 2000N/m2 have been put on the front and one
side façade with an added snow load of 1500 N/m2 on the roof. Figure 9 shows the
stresses in the beams. The stresses are summarized in table 1.
145
Challenging Glass 3
Figure 9: Whole structure under wind. Gravity and snow loading
Table 1: Peak stresses

Location Direction in model Peak stress
beam x 8 MPa (tension)
column y 12 MPa (compressive)
plate Stress intensity 4 MPa (tension)
This implies that with a safety factor of 4 heat strengthened glass is needed for the
beams and columns and annealed for the plates. This would also be preferable for the
long term safety as annealed glass is sensitive to stress corrosion, [6], and the beams and
columns should be made of pre-stressed glass. Heat strengthened glass in the main
structural members with annealed in the skin will also ensure with the use of stainless
steel reinforcement that the structure is sufficiently damage tolerant.
4.4. Validating the whole structure against earthquakes

The whole model was recalculated using a continuous steel support ring under the glass.
Under each column an elastomeric bearing was modeled which rested on the continuous
steel ring. Figure 10 shows the model. The bottom ring was assumed to move in the
horizontal plane only. Data from an earlier earthquake, [6], were used to provide the
displacements and velocities. The resultant stresses in the structure were modeled and
analyzed.
146
Figure 10: Finite element model used for earth quake calculations
Although the approach is limited by the use of two dimensional shell elements and a 3-
D approach is needed to model the elastomeric bearings more correctly the results
suggest that the initial dimensioning does not need to be changed significantly. The
glass box can move as a whole on the support and should survive the earthquake.
More detailed calculations with a 3-D model are in progress.
5. Building physics aspects

The aim is to minimize the overall energy demand through the optimum combination of
material properties and passive systems.
The shelter functions as a protection shield from the nature’s elements and it can be
described as a semi-outdoor space, an area which while ‘still being exposed to the
outdoor environment in most respects, include man-made structures that moderate the
effects of the outdoor conditions’, [7]. Accordingly, the temperature range, conceived as
comfortable by the occupants is wider. It was determined that for the shelter this range
lies between 13-29 Celsius degrees. This range also enhances the further maintenance of
the temple, since wide outdoor temperature fluctuations can damage the stone.
The meteorological data of the temple’s specific location indicate a warm climate with
many hours of sun exposure. Thus, the main measures concerning the shelter were taken
towards reducing the risk of overheating4.
4
Since the days presenting very cold temperatures are few, it is expected that the shelter can then
be heated- if needed- through mechanical systems.
147
Challenging Glass 3
In short:
x Sun-shading is applied in the form of gradient glass fritting, as indicated in
figure 11. Fritting creates an effective UV shield that helps controlling solar
gain and glare and enhances diffuse light, which is the most appropriate light
for anaglyphs such as in the temple.
x In order to further reduce the solar gain to the radiant transmittance, double
glazing with an external high reflective pane (Pilkington Suncool Silver 50/30)
was applied in the roof and with a low-e glass pane (Pilkington Optitherm) for
the facades.
x Stack ventilation is used for the recycling of air inside the shelter through
natural ventilation. This is achieved by an inlet at the bottom of the west façade
and an outlet on the top of the east façade, as seen in figure 11. The air flow
rate is accelerated due to the roof slope. The first thermal simulation results via
the Design Builder software are shown in figure 12.
Figure 11: Passive strategies for achieving thermal comfort inside the shelter.
148
Figure 12: Climate results in summer.
6. Discussion
The study shows that large transparent enclosures for monuments or other objects that
need to be protected are possible. The proposed structure minimally detracts from the
ancient temple, allowing it to be seen from far away.
In addition the monument is protected from the weather while access is controlled. The
complex requirements of the Venice charter can be met while preserving the building
for posterity.
An unusual and untried structural solution has been used based on pioneering research.
This has led to an all glass structure of uncommon slenderness. The structural scheme is
dependent on an integrated column, beam and skin structure working together.

Although such structural schemes are common in airplanes and ships, they are not
usually used in architecture. The reason why this scheme has been selected is to
provide the best answer to the conflict between cost, transparency and safety.
To make this structural scheme workable joints are needed that visually do not intrude
and allow force transfer and provide structural safety. The reinforced glass solution
chosen will only be safe if the joints are as safe as the components.
Buckling is one worry. Preliminary analysis suggests that the proposed structure should
be safe but there is a need to validate the system once the design of the joints is finished.
Similarly earthquake resistance is important. Not only should the building be safe for
visitors, but also the enclosed monument should not be at risk if the glass structure fails.
A headline ‘2500 year old temple demolished by falling glass structure’ is not the type
of advertisement the structural glass industry needs.
149
Challenging Glass 3
The building physics results show that by incorporating existing technologies the need
for protecting the ancient structure in winter can be combined with the need to protect
the workers and visitors in summer.
Although several aspects need to be worked on, a structure such as proposed, essentially
based on existing technology could be realized within several years.
7. Conclusions
From the results the authors conclude that:
x The proposed structure can be made

x The proposed slender dimensions are possible if the skin is part of the load
bearing structure
x The details for the connection need to be developed to allow this.
x The proposed structure is sufficiently cool in summer and warm in winter to
allow people to work there and to prevent frost damage to the monument.
8. References
[1] Louter C, 2011, ‘Fragile yet Ductile Structural Aspects of Reinforced Glass Beams’, TU Delft, the
Netherlands.
[2] Trösch E, Wälchli E, Baumgärtner T, Load-bearing Behaviour of Non-monolithic Glass Laminate
Beams for Wide Spans, Challenging Glass 3 proceedings (this conference), TU Delft, 2012
[3] Designing and Planning the World’s Biggest Experimental Glass Structure,F.P.Bos et al., Proceedings
Glass processing days , 2005, Tampere, Finland
[4] Bos, F.P., Safety Concepts in Structural Glass Engineering, PhD thesis TU Delft, 2009
[5] Veer, FA & Rodichev, YM (2010). The strength of glass, hidden damage. In FP Bos, PC Louter & FA
Veer (Eds.), Challenging glass 2 : conference on architectural and structural applications of glass (pp.
395-404). s.l.: s.n.. (TUD)
[6] Papadopoulos K, Bintzilaou E, 2008, 'The Seismic Response of the Columns in the Perimeter of the
Temple of Apollo Epikourios', Third National Conference on Earthquake Engineering and Engineering
Seismology, Athens
[7] Spagnolo, de Dear, 2003, ‘A field study of thermal comfort in outdoor and semi-outdoor’, Building and
Environment, Vol. 38
150
doi:10.3233/978-1-61499-061-1-151
A True All-Glass Staircase

Ernst Wälchli
Wälchli + Partner AG, e.waelchli@dwing.ch
Bruno Kassnel-Henneberg
Glas Trösch, info@glastroesch.ch, www.glastroesch.ch
Glas Trösch, a well-known glass manufacturer and glass finisher, supplies high-
quality glass solutions for all sectors of the construction industry as well as for the
automobile industry. An all-glass staircase is used to show that, as an innovative
glass producer, Glas Trösch has embraced the trend towards large façade
components. The example clearly illustrates that large, complex structural all-glass
components can feature astonishingly simple yet visually attractive details, if the
solutions chosen are appropriate for working with glass as the main material.
Keywords: structural use of glass, large glass panels, glass lamination, full glass
stair
1. General
Glas Trösch is a company that is involved in the entire spectrum of glass applications in
construction from the production and finishing of glass right through to the
implementation of projects. Supplying glass to end consumers and metal construction
companies is naturally the company’s core competency. Glas Trösch is renowned for its
cost-effective fabrication of high-quality insulation glass, glass laminates and special
types of glass for both interior and exterior applications.
Triple insulating glazing with standard dimensions of 7.5 m x 3.2 m and the
manufacture of pre-stressed and laminated glass panels with dimensions of 9 m x 3.2 m
clearly demonstrate that Glas Trösch has embraced the current trend towards
increasingly large glass formats. Continuous expansion of the fabrication to include
formats measuring 12 metres in size is indicative of the company’s ambition to place
itself at the forefront of this area of development.
Special solutions with regard to glass laminates in particular are regarded as challenges
and – in collaboration with renowned metal construction companies throughout Europe
– frequently result in unique glass structures. The lamination of load-bearing glass
panels with a range of interlayers – very often in combination with lamination of textiles
and other materials – is successfully implemented on an ongoing basis. Furthermore, the
fabrication section is equipped with cutting-edge machinery that allows float glass with
a width of 2 mm to be fabricated fully automatically with dimensions of 6 m x 3.2 m.
151
Challenging Glass 3
Figure 1: special lamination solutions
In addition to conventional glass fabrication, the company offers system solutions in the
metal-glass construction sector. A wide variety of awning and staircase systems as well
as stair and façade systems are available.
The movement toward all-glass constructions was already embraced by Glas Trösch
some time ago. One of the most successful and – due to the enormous glass dimensions
– impressive examples can be seen in the company’s own reception area at Glas Trösch
AG Swisslamex in Bützberg, the headquarters of the group of companies.
2. A true all glass staircase

The staircase has floor plan dimensions of 5.50 m x 8 m and is almost 12 m tall. It is
designed with load-bearing glass beams and glass fins. The building envelope is made
of structural insulating glazing units with a total thickness of 74 mm (2 x 12 mm
laminated safety glass outside, 28 mm spacing and 2 x 10 mm laminated safety glass
inside), a weight of just under 3 tons each and a length of 9 m. These units transfer the
wind loading to the glass beams and fins into the building substructure.
152
Figure 2: true all-glass staircase – stair and envelope structure (Hans Ege, artege, Luzern)
The same glass built up is used for the insulated glass roof units. The structural concept
for the design of all insulated glass panels takes into consideration a complete failure of
the outer layer under full load without any risk of a progressive failure. This ensures the
overall structural integrity even under extreme load assumptions as this is usually
performed for structural full glass envelopes.
153
Challenging Glass 3
On the inside, the all-glass stair using standard steps (Swissstep) spans over three floors
of the office building, including the side glass stringers, rails and glass landings.
Figure 3: entrance floor plan

Figure 4: entrance elevations
154
2.1. Envelope structure

The actual static load-bearing structure consists of a so-called asymmetric frame with a
height of nearly 12 m and a width of 5.50 m. It is supported on a concrete wall on one
side. The bottom fixings of the facade fins are supported on a concrete foundation.
Figure 5: Principle of asymmetric frame
The zero moment point is located at a height of 9 m. There a hinge is provided to

transfer only shear forces. The entire glass frame is made from 4 x 12 mm annealed
laminated safety glass with a width of 75 cm and a length of 9 m laminated with PVB.
The lateral stability of the glass fins and beams is generated by the insulated glass units
by connecting the elements using structural silicone sealants.
Solely for the loading condition of wind suction on the facade, the inner edge of the
glass fin is compressed and exhibits inclination. However, the compressive stress values
are small and additional tipping stabilisation is not required. The entire construction is
structurally bonded with SG silicone.
This structure demonstrates that large scale load bearing glass structures can be realized
without using complex connection techniques as laminated inserts or composite steel-
glass trusses. This makes things much easier in many aspects and therefore should be
regarded as an option more often. The simplicity also enables us to use annealed glass
for the insulated units, fins and also for the glass beams. This is a big advantage in
respect to costs and visual appearance of the glass structure.
155
Challenging Glass 3
The main structural details as the hinged connection in the glass fin or the top support of
the beams are designed in a way to minimize local stress concentrations. This is
required due to the use of annealed glass. In addition to this the designers wanted to
visibly show the structural concept by emphasizing the hinged connections. The total
elimination of metal in all the details also became possible as the friction which is
activated by the dead load is taken into consideration for the detail design and for the
transfer of the out of plane stabilization forces.
Figure 6: hinged full glass connection details
2.2. Assembly
Figure 7: assembly of façade panels 8.5m height
156
doi:10.3233/978-1-61499-061-1-157
Two Lines
Arup with David Chipperfield Architects
Felix Weber
Arup, United Kingdom, Felix.Weber@arup.com
For the 2011 London Design Festival, David Chipperfield Architects and Arup
worked closely together to realise Two Lines, a free standing glass pavilion that
showcases Sefar Architecture Vision, a metal coated fabric laminated between two
glass panels. David Chipperfield Architects wanted the pavilion not only to be a
landmark on London’s South Bank next to the Royal Festival Hall, but also to
invite the public to walk into it, interact and experience the unique appearance of
the glass with the shimmering translucent mesh that constantly changes with
orientation, view angle and light conditions. The design features two interlocking
forms (“Lines”) that are similar in shape but with opposite orientation. One Line
features copper and one aluminium coated mesh, each Line consists of fourteen free
standing glass wall panels, orientated on a rectangular grid and supporting five
glass roof panels. The different orientation of the glass walls not only creates an
interesting space that guides the view of the visitor, the arrangement also stabilises
the 3.6m tall structure. The glass panels are laminated with DuPont Sentry Glass
that encloses a 50% open mesh that has a copper or aluminium coating on one side
and is black from the other. The design of the connectors supports the Architectural
concept: reduced linear stainless steel connectors where electrochemically coloured
to match the colour of the mesh. The bonded connectors were factory bonded to the
glass, this reduced the onsite installation time. The project was designed and
realised within less than five months with the combined effort of the Architect,
Engineer, the Material Suppliers and specialist contractor Bellapart.
Keywords: Structural Glass, Laminated Glass, Coated Mesh, Glass Connector

1. General
The glass pavilion “Two Lines” was commissioned by the London Design Festival for
the Size and Matter program 2011. Size and Matter has taken place since 2007, each
year leading designers and architects are invited to experiment with innovative building
products or processes. The installations are showcased for typically one month in public
places in London. Arup has been sponsoring the Festival with free engineering advice
since the start of Size and Matter. Previous years involved architects like Shigeru Ban,
Zaha Hadid and Amanda Levete experimenting with paper tubes, Corian and black
polished precast concrete. In 2010 Clemens Weisshaar and Reed Kram used eight
robotic arms from Audi’s production line fitted with LED lights to trace messages into
London’s night sky.
For Two Lines, David Chipperfield Architects were asked to design a pavilion next to
the Royal Festival Hall on London’s South Bank, using Sefar Architectural Vision, a
very fine black polymer fabric that is coated with metal particles on one or both sides.
The fine, shimmering fabric is laminated with SentryGlas between two sheets of glass
157
Challenging Glass 3
to create a strong and robust product that is suitable for structural glass applications.
The mesh gives the glass a unique appearance and increases its solar performance.
2. Solar control and colour option on structural glass
2.1. Body tinted glass

Body tinted glass is the simplest way of providing solar protection and still used in
automotive glass. The number of colours is limited. The body tint absorbs light; the
glass has a low reflection. Glass colours other than neutral grey transmit light which
might not be desirable for architectural use. Sizes depend on float width, maximum size
for handling and lift off capacity of the float line.
2.2. Laminated XIR film

XIR or California Film is a sputtered polyester film that is laminated between panes of
glass. Different films offer different amounts of solar protection. The film works in
reflection and absorption and is more effective than a body tint. It can be used in single
glazing and to a certain extent in double glazed curved glass. Commonly PVB is used
for lamination, we are not aware of any application with SentryGlas, the film is limited
in width to 2.4m but unlimited in length.
2.3. Glass coatings

Glass Coatings have become the most common way to apply solar protection to glass.
On-line applied pyrolytic coatings are generally less selective than sputtered coatings
and come on float glass. The sputtering process is off-line and can be applied all sorts of
glass as long as it is flat. Coatings with high selectivity meet the Architect’s demand for
highest possible visual clarity of the glass, masking techniques can be used to create
decorative patterns. Robust coatings can be heat treated, they offer more flexibility in
the production process and can be used for roller bent and mould bent glass, however
anisotropy might become more obvious on coated and toughened panels. Sputtering
lines come with very high cost, the number is limited and they are generally owned by
glass producers rather than glass processors. Sizes are currently available up to 12m in
length at full float width. Silica based “spray on” coatings, which are available as anti
reflection treatment for photovoltaic panels, have not as yet made their way on
architectural glass.
2.4. Ceramic Frit

Ceramic frit is printed onto the glass and molten onto the surface under heat in the
toughening furnace. Frit offers no selectivity, Double sided frit features two different
colours printed on top of each other, a common strategy is to have a white frit on the
outside to reflect off solar energy and black from the inside to allow a good view out
through the glass. A frit on the external glass face reduces the reflection off the glass
surface. Screen or offset printing machines are commonly found in glass processors’
factories. The Engineer needs to take into account that the fused frit weakens the glass
surface, for example EN 12150 states 75MPa compared to 120MPa for toughened glass.
158
Two Lines - Arup with David Chipperfield Architects
2.5. Laminated metal mesh

Laminating metal sheets or mesh into glass panels is popular for internal decorations
and has also been used externally. The wires reflect light in more than one direction
creating a shimmering appearance rather than the mirror like reflection of glass coatings.
As the internal and external appearance is similar, double vision is not achievable. The
different thermal expansion coefficients of the glass and the metal as well as the
protection of the metal around the glass edges against corrosion need to be carefully
considered for external use to avoid corrosion problems or delaminating. Meshes in the
cavity of double glazed units carry a lower risk.
2.6. Laminated metal coated polymer mesh

The metal coated mesh “Architecture Vision” is a very fine polymer mesh with a metal
coating. The coating can be applied on one or two sides, one side coated mesh offers
good vision through when seen from the one (black) side and appear as reflective metal
fabric when seen from the other (metal coated) side.
Different metals are available as coating; corrosive metals require protective layers
when used in an external environment. The mesh is cut to size and laminated with
ionomer interlayer (Sentry Glass) in between two panels of glass. The mesh needs to be
set back from the edge of the glass to protect it within the laminate.
The maximum width of the mesh is approximately 1.6m with unlimited length, suiting
standard office modules of 1.5m. When using mesh in structural glass assemblies,
especially with dense fabrics, a reduction of the shear stiffness of the interlayer needs to
be considered.
Depending on the density of the mesh, the shimmer metal side of the panels varies from
semi translucent to near opaque. The mesh does not have any selectivity, but the light
reflection leads to less heating up of the glass compared to a ceramic frit.
3. The pavilion
3.1. Design intent

David Chipperfield Architects are not an office that is much known for ambitious
designs with structural glass but more for the use of decorative glass products in their
facades that play with reflection and transmission of light. For example, a metallic mesh
in glass was used for the entire facade of the Des Moines Public Library scheme to
create a solid appearance of the building from the outside and allowing the visitors good
view out into the adjacent park during the day. At night the lit interior becomes visible.
For the art pavilion the design team was looking for a design that would allow the
display of different metal coatings as well as the “twin view” effect. The design should
also display the structural potential of the glass laminate. It was important to limit the
number of glass panels to keep the scheme affordable.
Different designs where discussed and explored in models and renderings. The final
design features two meandering formations, the “Two Lines”, that are similar in shape
but with opposite orientation, one features copper and the other aluminium. Each
159
Challenging Glass 3
formation consist of fourteen free standing glass wall panels, orientated on a rectangular
grid and linked by five glass roof panels.
The glass structure is organised on a rectangular grid of 1.2 x 1.2m and is 3.6m tall. The
design appears very calm and rational on plan, but the different orientation of the double
vision walls creates many different situations. A 50% open fabric was chosen by the
Architect to give the glass a semi transparent appearance.
The September date of the festival and the location on the South Bank promised a lot of
public attention. The aim was to invite the public to walk in between the glass walls,
explore the interesting visual effects and views on the surrounding Royal Festival Hall,
Hayward Gallery and London Eye created by the varying reflection and transmission of
the glass.
The pavilion sits on a shallow plinth providing a levelled base and foundation that does
not disturb recently paved square. When the pavilion is reconstructed at another location
following the Festival, the base structure is intended to be hidden in the ground.
Figure 1: The pavilion, view from the Royal Festival Hall (London Design Festival).
160
3.2. Structure
The wall panels supporting the roof are orientated in two directions providing lateral
stability. The 3.6m height of the pavilion was dictated by the maximum span for a 2 x
12mm glass laminate to withstand the design wind loads and control their natural
frequency. 12mm is the maximum available thickness for heat strengthened glass and
heat strengthened was preferred for robustness. The fabric in the glass is encapsulated
between two layers of 0.89mm thick SentryGlas interlayer and set back by 10mm from
the glass edge to minimise the risk of corrosion of the metal layers. The roof panels are
acting mainly as a plate and made of thinner glass due to their small span.
The connection design had to respect the architectural concept and be simple enough to
be provided by competent manufacturer. Reduced linear stainless steel connectors made
of flat and folded plates were developed. The connectors can be attached to the glass by
the glass processor under factory controlled conditions, allowing steel to steel
connections on site.
For the top connectors two L-profiles are structurally silicone bonded along the top
glass edge of the wall panels with a 60mm wide bead. Flat plates are bonded to the
underside of the roof glass panels. A link plate transfers in plane loads between two
adjacent roof panels.
Figure 2: Top connector.
The base connectors transfer significantly higher forces when the glass panels act as
shear walls. For bending around in the weak axis of the panels they act as hinges.
Grouted bolt connections near the glass corners connect the glass panels and the two
steel plates. Pig nose bolts cover the holes for injection of the grout. The steel plates
receive the anchors to the concrete base.
David Chipperfield Architects wanted to express a conceptual link between the metal
connectors and the metal coated mesh in the glass. The stainless steel is
electrochemically coloured to the colours of the adjacent mesh.
Chemical colouring works by light interference. A chemical treatment increases the

oxide layer and by controlling the oxide thickness the eye sees different colours.
161
Challenging Glass 3
Figure 3: Bottom connector.
3.3. Analysis
The wall panels where assessed for maximum wind load and horizontal imposed loads
according to BS6399 and for their natural frequency. Seasonal reduction factors were
ignored as they would have not provided a large reduction of wind pressure due to the
date of the installation and would have been unfavourable because of the anticipated re-
erection of the pavilion at another location after the festival. The worst wind load case is
under 45° when there are no sheltering effects between the rows of wall panels. The
roof needed to be split into 5 separate panels for production and installation purposes,
the structural silicone joints soften the plate and lead to an uneven load distribution
between the wall panels. An analysis model of the roof was used to assess the stress in
the silicone joints of the connectors and the shear forces in the wall panels that defined
the forces in the glass bolts of the base connector.
The natural frequency of the walls was limited to 4Hz to make it difficult to accelerate
the structure by hand and give the structure a “solid feeling”.
Bellapart, the specialist glass contractor, undertook separate calculations to inform their
construction drawings.
4. Realisation
4.1. The specialist contractor

Having designed the scheme for the pavilion, the design team was faced with what
proved to be most challenging task - finding a specialist contractor for the pavilion.
162
The sourcing of the different materials was progressing well. SEFAR used their industry
contacts to line up glass processors BGT and Glass Trösch for the production of the
glass panels. DuPont donated their interlayer material SentryGlas. Christian Pohl GmbH
from Cologne became the manufacturer of the stainless steel, INOX provided the
chemical colouring and Keltbray offered to build the concrete base plate.
However, finding a partner to take on the design responsibility, provide production

drawings and build the pavilion on limited budget was another challenge. The scale of
the pavilion proved to be difficult, it was too large for smaller local British Contractors,
but too small to be of interest for any of the large ones. The installation time in mid
September is typically a time when specialist glass contractors are fully occupied.
The design team approached several Contractors but received few positive responses. At
the beginning of July, 2011, the Festival could appoint Bellapart SAU, Spain, for the
final engineering and installation. At the time of the appointment, the program allowed
Bellapart less than three weeks to produce calculations and submit the production
drawings to the selected manufacturers. Due to this tight program, any changes that
Bellapart proposed to the original design intent required a very quick resolution and
turnaround with the Architect and Design Team.
4.2. Installation
In summer 2011 the South Bank celebrated the sixtieth anniversary of the Festival of
Britain and the site between Belvedere Road and the Royal Festival Hall became very
busy.
The concrete plinth was cast in the week before the glass installation and had to cure in
the middle of a crowded food market. Bellapart’s glass installation program require the
Monday after the casting of the slab to set out and level the base steel connectors,
followed by one day installation for each of the glass structures followed by the press
opening on Thursday morning. The glass for each part was programmed to arrive in the
morning before its installation.
The installation concept proved to work well; delays were only caused by delayed glass
deliveries and the fact that one of the glass processors hade misread the fabrication
drawings and bonded the connectors to the wrong side of the roof glass. But the
bonding concept proved to have good flexibility. The connectors could be cut off the
glass and bonded to the glass in their right position under controlled conditions near the
site the following weekend.
4.3. Festival
The pavilion appeared to be well received by the public, especially during the typically
busy summer weekends on the South Bank. People of all age walked in between the
glass panels, played with the unusual visual effects or sat on the floor to relax or enjoy
food from the close market. No attempt of vandalism was observed by the security of
the South Bank other than a group of skaters damaging the edge of the concrete plinth.
The London Design Festival and the project team were pleased to see the pavilion being
well received and fulfilling the intent of the Festival: “bringing the design to public
spaces in London and reaching people who don't really think they know what design is”.
163
Challenging Glass 3
Figures 4, 5: The pavilion during the London Design Festival.

Figures 6: The pavilion at night (©Susan_Smart).
164
5. Acknowledgements
Two Lines was a Size+Matter commission for the London Design Festival.
Size+Matter was supported by the Southbank Centre.
Two Lines by David Chipperfield Artichtects was supported by:
Sefar, Arup, Bellapart, BGT, Dupont SentryGlas, Glas Trösch, John Morgan studio,
Keltbray, Pohl, Zumtobel
The realisation of the pavilion would have been impossible without the commitment of
the design team, the material suppliers and the installation contractor, a special thank
you to Rouven Seidler.
6. References
[1] Wurm, Jan, Structural Glass Fin Façade St. Davids Hall, Cardiff, Challenging Glass 2, 2010.
165
doi:10.3233/978-1-61499-061-1-167
Torre Iberdrola, Bilbao, Spain

Axel Zemborain
Pelli Clarke Pelli Architects, USA, azemborain@pcparch.com, www.pcparch.com
After winning an international competition held in 1993 to design the Master Plan
for the Abandoibarra area of Bilbao, Pelli Clarke Pelli Architects designed its main
compositional element: Torre Iberdrola. The tower was conceived as a softly
rounded triangle in plan that tapers upwards towards the sky. Access to the building
is through a sculpturally shaped glass lobby which encloses a large landscaped area.
The tower is clad with a double-skin façade that follows the geometry of tilted
cylinders, in order to standardize the curtain wall panels and reduce costs. The
lobby is enclosed by a warped glass wall that creatively solves the challenges posed
by its complex 3D geometry by using cold-bent glass and hybrid glass-and-steel
structural fins.
Keywords: Glass, Conference, Bilbao, Iberdrola, Cold-bent glass, Pelli Clarke

Pelli
1. The Abandoibarra Master Plan

Torre Iberdrola is the central element and the culmination of a large scale intervention
that was meant to become one of the most successful urban renovations in recent times.
For nearly a century the area of Abandoibarra was the industrial center and port of
Bilbao. The area is located immediately adjacent to the Ensanche, the nineteen century
extension of the old city that is nowadays the most vibrant area of Bilbao. The port
created a true barrier between the city and its river, the Nervión, accentuated by the
presence of a high speed highway and a drop in elevation of nearly six meters (fig. 1).
When an economic crisis led to the closure of numerous industrial plants and the port
was relocated to a deeper river basin area closer to the nearby ocean, a unique
opportunity was created to reconnect the city with its river. The first stage encompassed
the construction of the Guggenheim museum by Frank Gehry and the Palace of
Congress and Opera House by Federico Soriano and Dolores Palacios at both ends of
the site. In 1993 Bilbao Ría 2000, the non-profit entity created to recuperate deteriorated
areas of the city, organized an international competition to develop the remaining 35Ha
of the area of Abandoibarra. Pelli Clarke Pelli Architects, in conjunction with Balmori
Associates and Eugenio Aguinaga submitted the winning entry (fig. 2).
Our master plan extended the city grid of the Ensanche, resolved the abrupt change in
elevation with a gently sloped park to facilitate pedestrian access to the river, and
substituted the existing highway with a tree-lined low-velocity street that connects the
Museum with the Opera House, where also runs a new tramway. We also created a new
border to the existing Doña Casilda Park. In all, 70% of the area was dedicated to open
landscaped space. At the axis of Elcano Street, which connects this new neighborhood
167
Challenging Glass 3
to the main city square, Plaza Moyúa, the Master Plan created a new public space, Plaza
Euskadi, and located the focal point of the composition: a 165m high emblematic
building, Torre Iberdrola.
Figure 1: Abandoibarra before intervention.

Figure2: The Abandoibarra Master Plan.
2. The building
Torre Iberdrola’s massing is a triangular prism with gently curved sides which tapers
toward the sky. If the lines of its edges were to continue upwards, they would meet at
the symbolic height of 1000m (fig. 3a). Its forty one floors are divided in three sectors:
the lowest five floors are prepared to house either offices or a hotel, the intermediate
twenty one levels are leased as tenant office space and the upper eight floors
accommodate Iberdrola’s headquarters. The rest of the floors are dedicated to a
restaurant, business center and mechanical rooms.
168
The tower is recessed from the street and therefore it is accessed through an elongated,
sculpturally shaped lobby that reaches the sidewalk surrounding Plaza Euskadi. Inside
the lobby, two landscaped areas bring the surrounding park into the building. The
ground floor also contains an auditorium for 220 people.
In close collaboration with the façade engineers, Buro Happold’s London office, the
Design team I had the privilege to lead developed ingenious solutions for both the tower
and the lobby. The curtain wall was further developed in conjunction with the contractor,
the joint-venture between Félix Constructions, from Switzerland, who was in charge of
the tower cladding, and Bellapart, from Spain, who developed the lobby enclosure. The
final product is testimony of this collaborative effort and the common pursuit of a high
standard of quality.
Figure 3a and b: Conceptual massing and Torre Iberdrola seen from Plaza Euskadi.
3. Tower cladding
Early in the process we decided to clad the tower with a double-skin façade that would
be not only beautiful but also greatly contribute to the sustainability of the building. We
envisioned a taut, simple glass skin, very transparent but with a hint of reflectivity to
produce a light and dynamic effect, while maximizing the possibility of daylight
harvesting.
At first sight, the obvious solution to clad the tapering shape of the tower seemed to be
to follow three conical surfaces. However, this would have meant that the curtain wall
panels would have had to be different at every floor, and their width would decrease
169
Challenging Glass 3
towards the top of the building, creating a denser wall at the least desirable place, where
the main offices of our Clients would be.
We therefore approached the problem from a different perspective. We considered that

if we used cylindrical surfaces, instead of conical ones, the majority of the panels in the
building could have identical rectangular shapes, with the obvious advantages from the
fabrication and cost points of view. But in order to achieve the tapering shape sought,
we would have to tilt these cylindrical surfaces. As a result, the stack joints of the
curtain wall panels would have to follow a gentle curve along the façade, rather than
being perfectly horizontal as in any standard installation (fig. 4a).
We solved the geometrical discrepancy between the arched sills of the tilted façades and
the horizontal slabs by treating the inner and outer walls of the double skin
independently. The exterior wall, which acts as the water barrier, tilts following the
building massing, while the inner wall remains vertical. This inner wall consists of a
series of operable windows that allows maintenance of the resulting cavity. It also acts
as the fire and smoke boundary of the building, which allowed us to provide the solid
1m fire separation required by code while creating a completely glazed exterior with
minimal transoms as intended (fig.4b).
Figure 4a and b: Tilted cylinder concept and double-skin façade.
The spandrel was designed as angled metal panels that absorb the geometrical
differences between the two walls, forming a shape that we called “fishmouth”. As the
stack joints of the exterior panels arch towards the center of each façade from a low
point at the corners, the angles of the fishmouth vary from module to module creating a
warped surface (fig. 5a and b).
Automatic motorized blinds inside the wall catch the heat that enters the cavity, which
is then mechanically removed. That heat can be later recovered to pre-treat the fresh air
in winter time. This double skin solution was instrumental to the energy efficiency of
170
the building, which is expected to achieve a LEED Platinum certification from the
United States Green Building Council.
Figure 5a and b: Fishmouth detail at corner and at center of façade.
At each of the tower corners, where two tilted cylinders come together, the curtain wall
panels absorb their geometric differences. Each of these panels is custom-made for its
specific position since their overall dimensions vary and the corner angles gradually
decrease from floor to floor. We purposefully avoided having a vertical joint at the
corner in order to achieve very sharp and controlled edges, which resulted in rather
large corner units. Therefore, we had to develop profiles that allowed variable angles
not only for the stack joint between panels but also for the corner mullion.
The cavity between skins is divided every four modules with a low-iron glass fin, to
control the mechanical ventilation and provide an acoustical barrier. The inner and outer
vertical mullions always align at the ceiling level. But these fins are only perfectly
straight at the center of the façade, because as we approach the corners the difference in
geometry of these two skins creates a shift at the bottom of these fins that forced us to
cold-bent them to follow the resulting warped geometry.
The selection of the glass is absolutely critical for a building with these characteristics,
both for aesthetic and performance reasons. Our design intent called for a very
transparent glass with no hint of green tones and a very slight reflectivity. We selected a
low-iron, high-performance low E glass from AGC, the Stopray Vision 60T that has a
light transmission of 60, a Reflectance of 14 and a Solar Factor G of 0.44. This glass
allows for maximum transparency without compromising the thermal performance of
the double wall. The inner glasses are low-iron from Guardian.
171
Challenging Glass 3
4. Lobby cladding
We faced another set of challenges in designing the enclosure of the main entrance
lobby. Our design intent was to create a sculptural, luminous and unique space that
would offer a warm welcome to the visitors entering the building. The enclosure we
created is an undulating wall inclined towards the top at varying angles, producing a
complex warped surface (fig. 6a, b, c and d).
The roof of the lobby is supported by a couple of three dimensional trusses, tied to a
perimeter concrete beam at the second level of the tower and spanning more than 23m
until they rest on a portal immediately behind the entrance to the building. From there, a
15m cantilevered canopy reaches the sidewalk curb. The roof is clad with zinc
composite panels and features a large skylight in its middle section, which continues
along the canopy.
The 66m long storefront walls are supported by a series of structural fins, each one tilted
at a different angle and about one meter apart. They transfer the entire gravity load of
the storefront to the floor through pinned connections and they distribute the horizontal
loads to the roof trusses and the floor slab. These fins reach a maximum height of 10m
with an additional 5m cantilevered section above the roof line.
Figure 6a, b, c and d: Lobby structure and enclosure.
172
Since our intention was to create a very bright and transparent space, we first
investigated the possibility of using simple glass fins to provide lateral support to the
walls. But due to the dramatic height of the space, early calculations indicated that they
would have to be about 1000mm deep and built in sections joined by bolted connections.
Neither of those conditions was acceptable to us from a design point of view.
Therefore, along with Buro Happold, we developed a hybrid solution for the structural
fins that satisfied our design intent: 500mm I-beams composed by steel flanges and a
web of glass laminated with a SentryGlas Plus layer. High strength bolts provide the
compression force that allows the transferring of loads between steel and glass by
means of friction. In order to avoid seepage of the interlayer due to those forces, a
composite material with exactly the same thickness of the SentryGlas Plus was inserted
in the areas surrounding the bolts. Also, between the steel and glass, a material with a
high friction coefficient was incorporated to further guarantee the transfer of forces. We
carefully placed and detailed the bolts (their nut side covered with custom made caps)
so that their alignment produces interesting geometrical patterns that add yet another
level of detail and complexity to the wall (fig. 7 and 8b).
Figure 7: Hybrid steel and glass fin.

Figure 8a and b: Lobby wall, exterior and interior views.
173
Challenging Glass 3
The storefront is composed by insulated glass panels, each of a different size and shape
and cold-bent to follow the warping of this complex surface. The glass is mechanically
fastened at the top and bottom and supported by a toggle system along the vertical edges,
in addition to structural glazing on all four sides. We limited the maximum deformation
of each cold-bent glass panel to 45mm to maintain the permanent loads created by
warping the glass within acceptable levels (fig. 8a).
We used extra clear low-iron glass for all the components to minimize their visual
impact. The DGU glass is very similar to the one used at the tower, double laminated
AGC Stopray Vision 50T instead of the 60T of the tower. We also added a fine white
ceramic frit pattern on the second face, which fades from 0% at the pedestrian level up
to around 80% at the very top. This frit helps reduce glare and thermal gain, but also
conceals the louvers of the mechanical room located behind the upper cantilevered part
of the wall.
5. Conclusions
The design of Torre Iberdrola gave us the rare opportunity to work at a multitude of
scales, from determining the ideal location and shape of the tower as a key component
of a large scale urban development, to the meticulous attention to detail of the curtain
wall components. The teamwork approach maintained during the design and
construction phases, along with the open-mindedness and support of our Client, allowed
us to explore, develop and apply innovative techniques that include the use of cold-bent
glass and hybrid steel and glass structural members. The result is a state-of-the-art new
icon for a city that feels proud of attracting projects designed by some of the most
renowned contemporary architects in the world.
6. Acknowledgements
Client: Torre Iberdrola A.I.E.
Architect of Record: Ortiz / León
Façade Engineer: Buro Happold
Structural and MEP Engineers: Buro Happold and IDOM
Construction Management: Iberdrola Ingeniería

Façade Contractor: Félix Constructions/Bellapart
Pelli Clarke Pelli Architects’ Design Team: César Pelli, Fred W. Clarke, William Butler,
Susana La Porta Drago, Ileana Dumitriu, Rose Martin, Rodrigo Giménez, Axel
Zemborain
7. References
[1] www.pcparch.com
[2] www.torreiberdrola.es
[3] www.visitavirtualtorreiberdrola.com/#/home/
174
Challenging Glass 3
Joints, Fixings & Adhesives

175
doi:10.3233/978-1-61499-061-1-177
Experimental Investigation of
Unconventional Canopy Prototypes,
Suspended by Adhesive Bonds
Jan Belis, Arno Van Hulle, Dieter Callewaert, Jonas Dispersyn
Ghent University - LMO, Belgium, jan.belis@UGent.be, www.lmo.UGent.be
Bas Out
Delft University of Technology, The Netherlands
In this contribution the performance of an adhesively bonded glass to metal

connection with a carefully selected epoxy adhesive is investigated experimentally.
Deliberately pushing the limits of conventional safety concepts for the sake of the
experiment, laminated glass canopies were suspended on stainless steel local point-
fixings, adhesively bonded to the upper glass sheet only. Subsequently, the
suspended system was subjected to a ball-drop test, a uniform load and a hammer
impact. Different glass types and canopy arrangements have been tested in this way.
Depending on the design parameters, several compositions demonstrated
surprisingly good residual structural safety: even after 24 hours of full loading in
broken conditions, the heat-strengthened canopy configurations did not collapse.
Keywords: Glass, Adhesive bond, Epoxy, Experiment, Impact, Residual capacity
1. Introduction
In recent years several building projects and research programs have demonstrated that
truly challenging glass structures can be made when glass components are connected to
metal devices by adhesive bonds [1-9]. However, in practice it is not always evident to
select strong and durable adhesives for specific requirements [10-12]. Moreover, design
regulations tend to be rather conservative, and not very stimulating to explore the full
potential of adhesive connections.
To help designers to select potential adhesives as a function of specific environmental

exposures and loading conditions, researchers of Ghent University and Delft University
of Technology have conducted an extensive experimental program on adhesives for
structural applications with glass and metal connectors. The project yielded an adhesive
selection tool based on performance criteria, as well as a proposal of specific adhesives
which performed outstandingly well throughout the testing phases on relatively small
specimens relevant for specific applications [11, 12].
More specifically, during the project five typical applications with specific boundary
conditions, exposures and requirements have been selected in dialogue with practicing
and industrial partners – also known as the project’s Steering Committee - involved in
the project. For each application, a number of prototypes has been built to investigate
technological issues such as surface pre-treatment and practical adhesive application in
177
Challenging Glass 3
a “real” building component. Subsequently, the successfully built prototypes were

subjected to specific experimental tests, selected in function of the expected exposures
and actions typical for that specific application.
2. Test case
2.1. Background and general requirements

The application which is subject of this article is a point-fixed canopy which should be
resistant against full exposure to external weathering conditions. In such case, the
adhesive connections are constantly loaded by permanent actions such as the own
weight of the laminated glass, and occasionally loaded by variable actions such as snow
and wind.
However, the authors want to emphasize that the case presented here is very
unconventional and will usually be conceived as being unsafe, because of the absence of
a secondary load path in case of failure and because of the significant risk of injuries
(after all it concerns overhead glazing!). In addition, uncertainties with respect to quality
control and in general a limited trust in adhesive-only connections in buildings will
hinder the application of such a canopy.
Nonetheless, this prototype was built on explicit demand of several members of the
project’s Steering Committee, interested in pushing the limits of adhesive technology
and deliberately willing to experimentally check its behaviour and performance in
applications beyond common practice.
2.2. Canopy configuration and materials

The basis of the prototype design was a standard canopy produced by GSC
StarConstruct (see Figure 1). The standard canopy had conventional bolted point-fixings,
connected through boreholes to fully tempered glass. However, in the prototype the first
were replaced by point-fixings which were adhesively connected to the upper surface of
the glass without any holes. Advantages of the adhesive connection compared to the
bolted version are a reduction of stress peaks, possible cost reduction (e.g. due to
absence of hole drilling operations) and a smooth glass surface.
Figure 1: 3D Drawing of GSC StarConstruct standard suspended glass canopy configuration.
178
Experimental Investigation of Unconventional Canopy Prototypes, Suspended by Adhesive Bonds
Each canopy consisted of two laminated glass plates with a surface of 1m x 1m each
and a thickness of 8 mm for both glass layers individually. To verify the influence of
different glass strengths and fracture patterns, two glass types were tested: fully
tempered glass (FTG) and heat-strengthened glass (HSG). In total, three heat-
strengthened panels and two fully tempered panels have been destructively tested.
Furthermore, to reduce deformations of the laminate after glass breakage and to reduce
corresponding peel stresses in the adhesive, SentryGlas® was chosen as the interlayer
material.
Suspensions and connector elements were made of stainless steel AISI 316, and their
geometry was based on the geometry of the standard canopy’s parts. More specifically,
four circular connectors with a diameter of 57 mm were used on each laminate. Due to
the spatial arrangement of the canopy suspensions, both normal forces and shear forces
were induced in the adhesive (Figure 2a). Moreover, due to the specific geometry of the
connector, a force eccentricity additionally caused a bending moment and
corresponding peel stresses on the adhesive (Figure 2b). However, the average stresses
in the connection were more than ten times smaller than the stress levels expected in the
worst loading case.
Fx
F
e
Fy
Fx
a b
Figure 2: Force components acting on the adhesive. a) Normal and shear forces due to overall canopy
geometry; b) Connector eccentricity e causing bending moments Fx.e on adhesive bond.
Apart from a high strength and stiffness along with a good resistance to humidity, UV
radiation, relatively high and low serviceability temperatures, major requirements for
the adhesive are a good resistance against creep and impact. For this reason we selected
a 3M Scotch Weld 9323 epoxy, as this particular adhesive had demonstrated one of the
highest strengths and good performance under static loading and different exposures
during the preceding broad screening and the extensive testing program executed in
earlier phases of the research project. The specific tests performed on adhesives and the
selection tools used are beyond the scope of this article, but the interested reader can
find more details in literature [11, 12].
179
Challenging Glass 3
3. Assessment of the canopy and its adhesively bonded suspensions
3.1. Testing procedure

The tests on and assessment of the canopies followed a fixed schedule consisting of
different steps, based on testing procedures for normal, non-accessible overhead glazing
described by Haldimann et al. [13] and Siebert [14]. The full testing procedure was
executed at room temperature in the lab.
Firstly, each glass plate was subjected to a uniformly distributed static load of 1.5
kN/m² for at least 24 hours.
Subsequently, the load was removed and a steel ball with a mass of 4 kg was dropped
on the middle of one of the suspended laminated glass plates from a height of 3 m
relative to the glass upper surface level.
Finally, the complete canopy was (more or less uniformly) loaded by a static load of 0.5
kN/m². In case any of the glass layers in the canopy was not broken yet by the impact,
they were broken at that moment by hitting it with a hammer-driven centre punch.
The test is considered to be successful if the structural integrity of the canopy is

guaranteed for at least 24 hours. More specifically, the connections should not fail, the
panels may not be penetrated and no dangerous glass fragments may fall down after 24
hours of static loading of a fully broken test specimen.
Additionally, an equivalent procedure without ball-drop is also considered: in this case

both plates need to be broken by means of a hammer-driven centre punch after they
have been loaded by a static load of 0.5 kN/m². The evaluation of the results in this case
is identical to the procedure described above. Both version of tests have been executed
during our experiments, more specifically three ball-drop tests (on 1 FTG and 2 HSG
panel, respectively) and two tests with only centre punch (on 1 FTG and 1 HSG panel,
respectively).
4. Results and discussion
4.1. 24 Hours initial static loading test

Application of a uniformly distributed static load of 1.5 kN/m² for at least 24 hours did
not cause any damage to any test specimen (see Figure 3). The observed maximum
deflections in the centre of the laminate were in the order of magnitude of 2 mm and
were mainly due to the elastic elongation of the stainless steel tension rods and to a
lesser extent to the bending deformation of the laminated panels.
180
Figure 3: Canopy subjected to a uniformly distributed static load of 1.5 kN/m² for (at least) 24 hours.
4.2. Ball drop test

All ball-drop tests caused the lower glass layer of the impacted laminate to break. In
contrast, the upper glass layer always remained undamaged, even if it had to resist the
direct contact of the steel impactor. This observation stands for both the fully tempered
glass panels and the heat-strengthened ones. At the one hand side, the broken lower
glass sheets shows general similarities between both glass types, such as the radial
fracture pattern with a very fine mesh in the impact centre, a clearly more coarse mesh
in an intermediate zone and finally a finer mesh again in the outer glass area. However,
on the other hand side there are also noticeable differences between both glass types in
terms of absolute density of the fracture mesh, as was expected (see Figure 4). As for
the adhesive connections, no damage whatsoever could be observed after the impact.
a b
Figure 4: Fracture pattern of the lower glass plate in the laminate due to ball-drop test. a) heat-strengthened
glass; b) fully tempered glass.
181
Challenging Glass 3
4.3. Residual load-bearing capacity of heat-strengthened plates

As the adhesive connection did not require any boreholes, heat-strengthened glass is
principally a valid glass type option, especially as it is known to combine a relatively
high strength with a relatively coarse (so favourable in terms of residual stiffness)
fracture pattern. After loading the impacted canopy with a distributed load of 0.5 kN/m²
and breaking the remaining glass plates with a centre punch and a hammer, the canopy
with heat-strengthened laminated glass entirely maintained its structural integrity after
24 hours of load duration: the adhesive supports did not fail, the impact body did not
penetrate the laminated glass panel and no dangerous glass fragments fell down.
Consequently, it passed the test. Figure 5 illustrates the overall appearance of the
canopy after 24 hours of load duration. Hence, significant deformation of the fully
broken panel is visible in Figure 5 as well. The latter is due to randomly coinciding
fractures in the lower and upper panel, as indicated in Figure 6.
Figure 5: Final appearance after full glass breakage and 24h of static load on suspended canopy made with
heat-strengthened glass panels. The visible deformation of the front panel is due to coinciding fracture lines in
upper and lower glass sheets of the laminate (see also Figure 6).
182
Figure 6: Coinciding fracture lines in upper (red line) and lower (yellow line) glass sheets of the laminate.
4.4. Residual load-bearing capacity of fully tempered glass plates

Repeating the same procedure for the canopy with fully tempered laminated glass,
surprisingly it initially resisted quite well: the fully broken laminate kept its integrity
and the adhesive connections did not fail.
Unfortunately, after approximately 12 hours of static loading in fully broken condition,

the fully tempered glass panels collapsed. However, close inspection revealed that the
adhesive did not fail. Instead, the adhered glass cullets had split more or less parallel to
the outer glass surface, leaving a thin layer of glass adhered to the adhesive but losing
its suspensions this way (see Figure 7).
183
Challenging Glass 3
Figure 7: Detail of stainless steel connector after collapse of the fully tempered laminated panel. A thin layer
of broken glass is still adhered to the adhesive, which in itself was not damaged.
4.5. Discussion
The better performance of the heat-strengthened glass compared to the fully tempered
glass can be easily understood. Indeed, due to the dense crack pattern of the latter, 1) the
overall residual stiffness of the broken laminate is more reduced compared to the heat-
strengthened glass (see also comments below) and 2) a sudden volume expansion upon
fracture of the fully tempered glass causes immediate stresses which are transferred to
the relatively stiff epoxy adhesive.
Nonetheless, both glass types performed better than expected. However, to enable such
a favourable post-breakage performance, the overall stiffness of the fully broken

laminate is of major importance. In detail, angular displacements caused by a more
flexible broken laminate will cause unfavourable peel stresses on the adhesive and/or on
the interlayer, and splitting stresses on the glass, which may initiate a premature
collapse. Consequently, the choice made here to use SG as interlayer material, well-
known because of its relatively high stiffness, is believed to play an important role in
the achievement of the good results discussed above. In other words: it is possible that
the post-breakage performance of the canopy would be significantly less in case a more
flexible interlayer was used, or in case the test was performed at e.g. a higher
serviceability temperature.
184
5. Conclusions
Based on a large experimental investigation of adhesives for structural glass to metal
connections, 3M Scotch Weld 9323 epoxy was selected to build experimental
prototypes of a glass canopy suspended by adhesively bonded point-fixings. The main
conclusions of the experimental impact assessment at room temperature of the
prototypes are the following:
x 24 Hours of static loading in the unbroken state only caused limited elastic
deflection of the canopies, but no damage;
x Ball-drop impact caused breakage of only the lower sheet of the impacted
canopy panels, but never caused visible damage to the adhesive bonds;
x The tested heat-strengthened canopy panels met the test requirements as the
connections did not fail, the panels were not penetrated and no dangerous glass
fragments fell down even after 24 hours of static loading in a fully broken state.
However, it should be noticed that worse results may be possible if the post-
breakage stiffness of the laminate is reduced, for instance when a more flexible
interlayer is used or when the temperature during the test is increased;
x The canopies provided with fully tempered glass did not meet the requirements
of the standard, although they also resisted to the applied static load for about
12 hours after breakage of all glass sheets. Collapse was due to thin layer
failure of the glass, and not to failure of the adhesive bond.
6. Acknowledgements
The authors would like to acknowledge the support of IWT Flanders (TeTra Grant
090170 – Bouwen met glas en adhesieven). In addition, the support throughout the
project of the following companies and professional organisations is acknowledged (in
alphabetical order): 3M, BBRI, Blasco Ltd., Bureau Bouwtechniek bvba, Clusta, Delo,
De Witte Aluminiumconstructies nv, Dow Corning Seneffe, DuPont De Nemours, de
Federatie van Aluminiumconstructeurs vzw, Geert Verbeeken, Glastec bvba, GSC
StarConstruct nv, HQ Bonding, Huntsman, Kuraray Europe GmbH, groep Leroi -
Lerobel, Lord Adhesives, Miniflat nv/sa, Oosterlinck Consulting & Development,
Permasteelisa - Scheldebouw B.V., Proviron Industries nv, Sadef nv, Scabro, Sika, Seco,
SIGU, Soudal and Viba. Finally, the authors acknowledge the work of dr. Freek Bos,
who cooperated as a project engineer during the early phases of the project, Nathan Van
Den Bossche from the department of Architecture and Urban Planning, Warren Tanghe
and Joris Decoene.
Finally, the authors would like to acknowledge COST Action TU0905 “Structural glass
– novel design methods and next generation products”, which provided excellent
networking and feedback opportunities in the field of Structural Glass.
185
Challenging Glass 3
7. References
[1] Wellershoff, Frank; Sedlacek, Gerhard, Glued Connection for New Steel Glass Structures, Proceedings
of Glass Processing Days, Tampere, Finland, 2005, p. 340-343.
[2] O’Callaghan, James, A Case Study of the Apple Computer Stores - Glass Structures 2001 – 2005,
Proceedings of Glass Processing Days, Tampere, Finland, 2005, p. 406-409.
[3] Peters, Stefan; Fuchs, Andreas; Knippers, Jan; Behling, Stefan, Ganzglastreppe mit transparenten SGP
Klebeverbindungen - Konstruktion und statische Berechnung, Stahlbau, 76-3, 2007, pp. 151-156.
[4] Belis, Jan; De Visscher, Koen; Callewaert, Dieter; Van Impe, Rudy, Laminating metal-to-glass:
preliminary results of case-study, Proceedings of Glass Performance Days, Tampere, Finland, 2009, p.
191-193.
[5] Belis, Jan; Callewaert, Dieter; Delincé, Didier; Van Impe, Rudy, Experimental failure investigation of a
hybrid glass/steel beam, Engineering Failure Analysis, 16-4, 2009, p. 1163-1173.
[6] Wolf, Andreas; Hirai Kazuo, Evaluation of transparent structural silicone film adhesive for bonded
point-fixed glazing. Proceedings of Engineered Transparency Conference 2010, Düsseldorf, Germany,
2010, p. 167-177.
[7] Louter, Christian, Fragile yet Ductile: Structural Aspects of Reinforced Glass Beams, Delft University
of Technology, Delft, 2011.
[8] Netusil, Michal, Hybrid steel-glass beams, Czech Technical University in Prague, Prague, 2011.
[9] Van Hulle, Arno; Belis, Jan; Callewaert, Dieter; Scheerlinck, Lies; Out, Bas, Development of Structural
Adhesive Point-fixings, Proceedings of Glass Performance Days, Tampere, Finland, 2011, p. 661-664.
[10] Overend, Mauro; Jin, Qian; Watson, James, The selection and performance of adhesives for a steel–
glass connection, International Journal of Adhesion & Adhesives, 31, 2011, pp. 587–597.
[11] Belis, Jan; Callewaert, Dieter; Van Hulle, Arno, Bouwen met glas en adhesieven – Praktische gids voor
ontwerper en uitvoerder (in Dutch – English version is being prepared), Ghent University, Ghent,
Belgium, 2011. ISBN 978 90 382 1902 8.
[12] Belis, Jan; Van Hulle, Arno; Out, Bas; Bos, Freek; Callewaert, Dieter; Poulis, Hans, Broad screening of
adhesives for glass-metal bonds, Proceedings of Glass Performance Days, Tampere, Finland, 2011, p.
286-289.
[13] Haldimann, Matthias; Luible, Andreas; Overend, Mauro, Structural use of glass, International
Association for Bridge and Structural Engineering, Zürich, 2008.
[14] Siebert, Barbara, Safety aspects of point-fixed glass constructions, Proceedings of Glass Performance
Days, Tampere, Finland, 2007, p. 432-436.
186
doi:10.3233/978-1-61499-061-1-187
Connecting Through Reinforcement -

Experimental Analysis of a Glass
Connection Using Perforated Steel Plates
Paulo Carvalho, Paulo J. S. Cruz
School of Architecture, University of Minho, Portugal,
paulo.carvalho@arquitectura.uminho.pt, www.arquitectura.uminho.pt
Frederic Veer
Faculty of Architecture, Delft University of Technology, Delft, The Netherlands,
f.a.veer@tudelft.nl, www.glass.bk.tudelft.nl
Reinforced glass comprising laminated glass with a steel perforated plate is a

feasible way of increasing the residual strength of glass, counteracting its
brittleness. This important feature may be expanded by extending the reinforcement
outwards of the glass panels to create a connection. It is proposed connect it
mechanically to another steel plate while remaining adhesively connected to the
glass. Preliminary tests showed the validity of the concept, which depends on
adequate bending stiffness, to prevent buckling on the compression zone until
tensile strength is reached on the steel perforated plates mechanically connected.
Keywords: Glass, Connection, Steel perforated plates, Laminated, Bending

strength
1. Introduction
The brittleness of glass is still a challenge to its structural use. Although its theoretical
strength is extremely high it decreases abruptly in practice due to small flaws at the
glass surface and small defects at the edges as a consequence of cutting and grinding.
This dependence on surface quality is usually dealt with by strengthening glass with
heat treatments. This process induces compressive stresses in the glass surface
preventing crack growth, effectively increasing the glass tensile strength. On the
meanwhile, the amount of energy released at failure results in the glass disintegrating
into small fragments threatening the integrity of the component or structure. This leads
to a contradiction that the desired high tensile strength (to increase size and reduce
thickness) is inversely proportional to the residual strength required for structural safety.
To overcome this limitation an embedded reinforcement can be added during the
lamination process capable of increasing the residual strength even for fully tempered
laminated glass [1]. The reinforcement to be added to the laminate must have high
tensile strength, good young’s modulus and high deformation capacity. Only if this
meets the necessary criteria the transfer of forces from the failed glass to the
reinforcement is effective. It must also have low thickness and a certain amount of
perforations, not only to guarantee a certain transparency of the compound element but
also to be suitably embedded in between two interlayers during lamination.
187
Challenging Glass 3
Several materials fulfill the conditions: fiber fabrics, wire meshes and perforated metal
plates. Among all, the perforated metal plates differ from the others by presenting
certain shear resistance even when extremely thin (1mm). Extending it outwards of the
glass panel opens the opportunity to develop a connection solution that optimizes the
concept of laminated glass reinforced with steel perforated plates. This paper focuses on
the preliminary tests conducted to study the bending behaviour of the connection.
2. Concept
The brittleness of glass calls for the development of connection solutions that avoid
stress concentrations. According to the geometry and load transfer mechanism it is
possible to avoid excessive stress and perform a uniform flow of forces. The progress
made in recent years in the chemical industry resulted on the development of reliable
adhesive solutions for glass. The ionomer interlayers (SentryGlas Plus), stiffer and
stronger than the previous available PVB interlayers opened new possibilities in the
connection design. Several recent state-of-the-art glass structures used its full
possibilities [2] and proposed innovative embedded connection by introducing metallic
elements in between two laminated glasses. This type of connection relies on the
adhesive resistance of the contact surface between metal and interlayer to transfer loads.
Besides the non-quantifiable aesthetic value, achievable by discrete detailing and
integration within the laminated glass layers it allows a more gradual load transfer
combining adhesive and contact mechanisms.
Based on the two concepts: reinforcement and embedded connection, the starting point
of this research project was to investigate how a steel perforated plate laminated in-
between two glasses could contribute to a connection mechanism. An effective glass
connection must transfer three types of forces: axial forces (tensile and compressive),
shear forces and bending moments. Joining two perforated steel plates by extending
them outwards of the glass panels and mechanically fixing them creates a load path. The
steel perforated plate is adhesively connected to the glass and mechanically connected
to the other steel plate. It's slenderness (1mm) restricts its contribution to axial tensile
loads and the tensile forces of the bending moment. The other part of the bending load,
as well as the axial compressive and bending forces, must be transferred via contact on
the glass edges. This can only be achieved with the application of a suitable
intermediate load transfer layer to accommodate stress concentrations. The goal is to
develop a frameless edge connection for a slim and autonomous load bearing glass
system. In terms of visual appearance, a smooth transition between reinforced glass
elements is intended in order to optimize its transparency.
In previous tests performed by the author, the adhesive resistance of the interlayer and
the steel perforated plates (R5T8) with 1mm thickness was studied [3]. Each small-scale
specimen was laminated with two interlayers with the same thickness (1,52mm) at each
side of the steel plates. Results showed behaviour dependent on the use of PVB or SGP.
The specimens laminated with PVB presented an adhesion resistance inferior to the
yield stress strength of the steel perforated plates (around 7kN) on the two embedment
depths of 20 and 40 mm, which resulted in the sliding of the perforated steel plates
causing stretch-out of the interlayer. In comparison the SGP laminated specimens
always had superior adhesion resistance compared to the tensile strength of the steel
perforated plates (around 10kN) causing the steel to fail, hole by hole until total release.
188
Connecting Through Reinforcement - Experimental Analysis of a Glass Connection Using Perforated Steel
Plates
In both types of interlayer an anchoring effect was observed that increased the ductility.
In the case of PVB specimens this was caused by the interlayer stretching-out, which
influences the slip mechanism. For the SGP specimens this was caused by earlier
elongation of the perforated plate. The superior adhesive strength of the SGP interlayer,
its high stiffness and strength together with a lower propensity to degrade by adverse
environmental conditions (humidity and heat) encourages its use in the investigation of
this connection system.
3. Preliminary experimental tests

In order to understand the bending behaviour of the semi-embedded steel perforated
plates when superimposed and mechanically connected using bolts, a series of four
point bending tests were performed, using four different types of assembly: a) simply
bolted in four points, b) bolted in four points with steel joint cover, c) bolted in four
points with aluminium "L" shape bolted in two points in the compressive zone and d)
bolted in three points with nylon bar bolted in two points in the compressive zone.
3.1. Specimen description

The four point bending test specimens consists of a beam with total size of 1047mm x
100mm that is composed by two symmetrical parts connected at the centre. Each part is
made of two 10mm thick rectangular shaped pieces of float glass with dimensions 500
mm x 100 mm. These are laminated using two ionomer interlayers (SGP) with 1,52mm
thickness. An AISI 304 stainless steel perforated plate is semi-embedded in between. It
has 1mm thickness and the perforation pattern is R5T8, which means that it has 5mm
diameter holes that are disposed diagonally at an angle of 60º, distancing 8mm from
centre to centre. This type of pattern has 35,4% of the surface perforated. The metal
insertion has the height of the glass but it is only 78mm wide. Its embedment placement
is 40 mm inside and 38 mm outside (see figure 1).
Figure 1: Four point bending test specimen simply bolted in four points
189
Challenging Glass 3
The two-piece beam specimen is mechanically connected at the middle using M5

screws and nuts that are inserted through the plate’s perforation. On the first assembly
type four screws vertically displaced were used (see figure 2a). Three other options of
assembly were tested in which complementary elements were added to increase the total
stiffness: one very similar to the first but with a 3mm thick steel joint cover at each side
of the beam (see figure 2b); another that consisted on the addition of a 1 mm thick
10x10 mm aluminium "L" shape with the length of the gap between the two glass
laminates (47mm) on the upper (compression) part of the beam (see figure 3c); and the
last a 10mm x 10mm nylon bar also with the length of the gap and placed at the upper
(compression) part of the beam (see figure 3d).
Figure 2: Four types of connection tested: a) simply bolted in four points, b) bolted in four points with steel
joint cover, c) bolted in four points with aluminium "L" shape bolted in two points, d) bolted in three points
with nylon bar bolted in two points
190
Plates
3.2. Test setup

The four point bending test specimens were supported on two lower points distancing
800 mm from each other. The load was applied on two points distancing 200 mm from
the centre. Additional steel support frames were used with distributed points of contact
to prevent buckling (see figure 3). The specimens were loaded using a constant
displacement rate of 2 mm / minute. For the instrumentation of the models a charge cell
of AEP Transducers, with maximum capacity of 50kN and a displacement transducer
(LVDT) were used. The average room temperature during the tests was 18ºC.
Figure 3: Setup of the four point bending test

Figure 4: Simply bolted in four points specimen before testing
191
Challenging Glass 3
3.3. Test results

Figure 5 shows the load-displacement diagrams of the four point bending tests.
Figure 5: Load-displacement diagram of the four types of assembly
The four point bending test with the specimen simply bolted in four points showed a
non-linear but constant increase of the load until almost 3kN when the compression
installed in the upper zone of the perforate steel plates reached a limit above which the
plates started to buckle. This happened at a displacement of about 5mm extending at
least until 35mm, when the test was interrupted with around 1kN of remaining load.
The test of the specimen with two joint covers also shows non-linear load-displacement
behaviour until a maximum of around 5kN that occurred at a displacement of 15,5 mm.
It was observed that both perforated steel plates buckled due to the compressive stresses
on the non-covered areas close to the glass edge. When a steel plate touched the glass, a
local stress concentration occurs and the glass cracks at the point of contact resulting in
a slow decrease of the remaining load.
The specimen with two aluminium "L" shapes in the compression zone, presents a
similar behaviour compared to the previous ones, characterized by a non-linear growth
with the maximum load increasing lightly to 6kN at 7,9 mm of displacement, when the
"L" starts to buckle reducing its resistance. After this point, the load decreases very
slowly until just 5kN already with a displacement of 33 mm. Then one of the "L" shape
slid away from the glass edge and the load suddenly dropped significantly.
Lastly, the specimen with two nylon bars at the upper zone presents a continuous linear
increase in applied load until around 4kN at 3mm of displacement. Afterwards, the
bending stiffness continued to grow from 4kN until around 6.5kN allowing some
deformation of the nylon. At this point, the yield stress strain of the perforated steel
plate was reached and it started to elongate in an almost linear manner until a maximum
load of around 7.8kN when due to the buckling of the perforated steel plate in the
compression zone, deformed the nylon causing it to slide from the glass edge, voiding
the compression stress.
192
Plates
3.4. Discussion
The preliminary tests described in this paper illustrate a step-by-step approach to
optimizing the bending efficiency achievable with the concept under investigation. The
bending strength of the steel perforated plates overlapped and simply screwed was
investigated and it clearly showed its inability to resist due to its high slenderness,
which causes buckling in the compression zone, drastically reducing allowable load.
With the help of two joint covers to stabilize, the result improved but proved to increase
the risks as it caused glass breakage after a stress peak due to metal to glass contact. The
need to maximize the buckling resistance in the compression zone led to the addition of
two "L" shapes screwed through the perforated steel plate. Although it improved the
bending stiffness of the connection it wasn't enough to prevent buckling in the
compression zone. The addition of the nylon bars firmly screwed through the steel
perforated plate, due to its superior contact surface on the glass edges allowed an
interesting performance of the connection. The total stiffness improved considerably,
resulting in such an increase in compressive load on the nylon bars that the tensile stress
in the lower part of the steel perforated plates exceeded the yield stress. The
considerable deformation allowed by the nylon bars (see figure 5), allowed them to
accommodate some buckling deformation of the steel perforated plate, prevented any
further increase in load, the buckling of the steel forcing the nylon bar to slid away from
the glass edge.
Figure 6: Deformation of nylon bars under compressive load at the final stage of the test
4. Conclusions
The advantages of reinforcing laminated glass with a steel perforated plate may be
expanded by extending it outwards of the glass panels to create a connection.
The load transfer path thus created is based on an adhesive connection to the glass and a
mechanical connection to a similar steel perforated plate. The preliminary tests
performed clearly show the validity of the concept. When sufficient bending stiffness is
acquired, adding two additional elements firmly screwed with sufficient stiffness and
contact area with the glass edges, it activates the compressive stresses at the upper zone,
forcing the tensile strength on the mechanically connected steel perforated plates to be
reached.
193
Challenging Glass 3
5. Acknowledgements
The authors gratefully acknowledge the material support of Vicer, Vidraria Central de
Ermesinde and the technical support of Eurico Silva and Conceição Casal.
The first author wishes to thank the Portuguese Foundation of Science and Technology
for the PhD grant SFRH/BD/62061/2009.
6. References
[1] Feirabend, Steffen, Reinforced Laminated Glass, Proceedings of Challenging Glass, Conference on
Architectural and Structural Applications of Glass, Delft, Netherlands, 2008.
[2] O’Callaghan, James, A Case Study of the Apple Computer Stores - Glass Structures 2001 - 2005,
Proceedings of the 9th International Conference on Architectural and Automotive Glass (Glass
Processing Days), Tampere, Finland, 2005.
[3] Cruz, Paulo J.S.; Veer, Fred; Carvalho, Paulo, Perforated steel plate to laminated glass adhesive
properties, Proceedings of the 12th International Conference on Architectural and Automotive Glass
(Glass Performance Days), Tampere, Finland, 2011.
194
doi:10.3233/978-1-61499-061-1-195
Determination of Adhesives Properties for

Non-linear Numerical Simulation of
Structural Steel-Glass Connections
Vincent Dias, Oliver Hechler, Christoph Odenbreit
ArcelorMittal Chair of Steel and Façade Engineering, University of Luxembourg,
Luxembourg, vincent.dias@uni.lu, wwwen.uni.lu
In the construction of façades, connections between glass and glass as well as other
materials with varying properties, e.g. metals, are currently designed with, inter alia,
the application of adhesives. The related analytical evaluation of the load bearing
characteristics of these kinds of assemblies requires the correct determination of the
adhesives’ properties; knowledge in this field is still incomplete. Therefore an
extensive test campaign on the load bearing and displacement behaviour of
common structural silicone and UV-acrylic was carried out with the aim, to identify
appropriate models to describe their properties under quasi-static loading and for
their implementation in numerical analyses. The test program, comparisons
between the different adhesive materials and a summarised outcome will be
presented in this paper.
Keywords: Structural glass connection, Adhesive material laws, joint design,

numerical simulation
1. Introduction
Thanks to the research results of the last years, the use of glass as façade element is
taking an entirely new scale. In spite of its brittleness, glass is more and more employed
as structural load bearing element. In the consequence, the connection properties face
enhanced requirements concerning ductility and load bearing capacity. The structural
bonding is often achieved with adhesives like silicone, acrylic or epoxide resin.
The advantages of adhesive connections are:

x The distribution of the stress along the jointed parts is nearly uniform and
without inducing stress concentration peaks;
x Adherents with dissimilar mechanical and thermal properties can be bonded;
x Some adhesives have the capacity to transfer high loads between the jointed
elements.
However, difficulties arise during the design phase of the bonded assembly due to a lack
of comprehensive knowledge on the adhesives’ material behaviour. First of all, data
from simple small-scale tension tests implemented in standard constitutive models are
generally insufficient for reliable numerical analyses. Thus additional sets of large scale,
time consuming, experimental tests are commonly demanded to prove the feasibility of
the structural connections. Secondly, the structural performance of the adhesive
195
Challenging Glass 3
materials is influenced by environmental factors such as temperature, humidity and/or

wind velocity, which must be considered in design.
In the consequence, an extensive test campaign on the tension, compression and shear
behaviour including the influence of environmental factors has been defined for two
adhesives, namely structural silicone and UV-acrylic. The test campaign was carried out
at the ArcelorMittal Chair of Steel and Façade Engineering at the University of
Luxembourg. Tests results, conclusions and derived constitutive material laws for
numerical simulations are presented in this paper.
2. Adhesives and parameters of influence
2.1. Selection of the adhesives

The various adhesives on the market can be classified in three main categories, i.e. the
elastomers, the thermosets and the thermoplastics [1]. While the elastomers are softer
and exhibit a large range of deformations (up to 400%) under little shear forces, the
thermosets are stiffer and reach a high shearing stress with minor deformations [2].
For structural application the adhesive should both, incorporate a high bearing capacity
and possess a certain ductility to compensate differential thermal expansion of the
adherents which may cause failure of the brittle glass panel.
To cover a common market range, a representative elastomer and thermoset have been
selected for experimental test series:
x The two-component silicone (elastomer) Dow Corning DC993, meeting the

requirements according to the ETAG 002 [3] for the realization of structural
sealant glazing systems.
x The Bohle UV-acrylic MV760, which has, compared to other acrylics, the
highest shear strength and the better stability after immersion in a façade
cleaning product at high temperatures [4].
2.2. Parameters of influence

The influences of three environmental factors concerning the material's ageing have
been investigated: temperature variations between -20 and +65 °C, humidity and UV-
radiation. The tests have been carried out according to EN ISO 11431 [5]. The
procedure covers a period of three weeks in which the samples are placed in water for 5
hours a day and in an oven under UV-radiation for the rest of the time. Further tests on
acrylic under elevated temperatures will be carried out to determine the “in service”
properties, like for example an expected loss of stiffness.
The experimental investigations presented in this paper have been performed under
quasi-static loading with varying loading rates. Aim has been to evaluate and apprehend
the material's dependency on these rates. To account for the change in the stress-strain
behaviour of silicone elastomers under repeated loading [6], cyclic tests will be carried
out in the near future. Their results will complete the material characterization.
196
Determination of Adhesives Properties for Non-linear Numerical Simulations of Structural Steel-Glass
Connections
3. Uniaxial tension tests
3.1. Samples preparation

Dog bone samples with dimensions according to ISO 527-2 [7] are generally used for
tension testing. The manufacturing of these samples depends on the adhesive curing
process and viscosity. They are either produced by casting in a mould or by a machining
procedure (with cutting dies).
The silicone samples were produced by sucking the high-viscous silicone in a Teflon
mould by using a vacuum pump at ambient temperature and humidity. As the selected
silicone was two-component, the curing process started in the mould, although the
mould was totally airtight. The samples were taken out of the mould 24 hours after the
injection and were stored during one month at ambient temperature and humidity. The
so-formed elements were exempt of any edge-flaws and air bubbles.
For the acrylic samples, the curing was done thanks to an UV-lamp which should be
applied at 90° to the bonding surface and as close to the acrylic as possible. For that
reason, the employed mould was fully transparent and composed by five layers
symmetrically disposed (centre part in Teflon covered by polyethylene films of 0.1mm
and Plexiglas plates), see Figure 1. The acrylic with medium viscosity was injected
through the upper hole of the assembled system and flowed down slowly under its self-
weight to the bottom of the shape. Bubbles of air were automatically rejected at the
upper hole. As for the silicone samples, the acrylic dog-bone, see Figure 2, were stored
during one month at ambient temperature and humidity, away from light radiation.
Figure 1: Curing system for acrylic under UV- Figure 2: Acrylic (left) and silicone (right) dog-bone
lamp samples.
After the storing period of one month, half of the test specimens of silicone, as well as
acrylic, were subjected to artificial ageing as presented in section 2.2 before to be finally
tested.
197
Challenging Glass 3
3.2. Tests
All the dog-bone specimens were tested on an Instron mobile hydraulic piston
presenting self-closing wedge clamps with a large number of teeth at their surface to
avoid any sliding, see Figure 3. While the silicone samples were directly installed in the
clamping devices, the acrylics were additionally equipped with covering aluminium
plates fixed to the samples with steel centering pins, see Figure 4.
The design of the test campaign was decided with the help of the Taguchi method [8].
As evoked previously, the tested factors were the artificial ageing according to the EN
11431 standard and the loading rate, which was different for the both considered
adhesives, see Table 1. To have a better variance, two tests series were conducted.
Figure 3: Tensile dog-bone test on silicone Figure 4: Modified acrylic sample
Table 1: Taguchi design of tension tests on dog-bone samples

Factors
Standard Test No
A = ageing B = loading rate
1 1 1
2 1 2
3 2 1
4 2 2
Level 1 No Ageing 50mm/min

Level 2 Ageing 250 /500*mm/min
*
The maximum loading rate for the acrylic tension tests was 250mm/min while 500mm/min for silicone
samples
198
Connections
3.3. Results
The recorded force-displacement data were transformed in stress-strain curves by
dividing the force by the tested section area and the displacement by the initial testing
length of 60mm.
The behaviour of the silicone was fully non-linear with a full-recovery of the
dimensions after unloading, which is characteristic of hyperelastic materials. The results
for the silicone samples are presented in Figure 5. The curves are converging which
means that the considered influent factors, i.e. the artificial ageing and the loading rate,
have no significant influence on the silicone properties. The tension-stiffness at the
origin was calculated according to ETAG 002, which consists in a linearization of the
curves by a conversion of the strain data with a determination of the tangent stiffness at
the origin. A variance analysis was carried out on these calculated values confirming
then insignificance of the loading rate and the artificial ageing on the test results.
Stressstraindiagram
1.20
1.00
0.80
(MPa)
S04 50mm/min NA
0.60 S05 50mm/min NA
S19 50mm/min NA
S02 50mm/min A
0.40 S12 50mm/min A
S13 50mm/min A
S01 500mm/min NA
0.20 S16 500mm/min NA
S06 500mm/min A
S08 500mm/min A
0.00
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60

()
Figure 5: Stress-strain curves for silicone dog-bone samples under uniaxial quasi-static tension loading
(NA=non-aged, A=aged)
For the acrylic, two types of curves were observed according to the influencing
parameters:
x For the non-aged specimens, a ductile material behaviour was only observed
under a low loading rate of 50mm/min, see Figure 6. The maximum value at
the inflection point corresponds to the beginning of the constriction phenomena
which propagates in the sample without variation of the applied force. Useful
values for design are the stress and strain at the inflection point, the elastic
modulus and the limit of elasticity, corresponding to the first slope of the
curve.
199
Challenging Glass 3
x A more brittle failure was observed for the aged specimens and/or for the high
loading rates, see Figures 7a and b. Bilinear approximation of the curves could
be envisaged. The most important values to be recorded are the Young’s
Modulus, the limit of elasticity and the stress-strain data at breakage. ISO 527
proposes to choose the Young’s Modulus as chord modulus between the strain
values of 0.05% and 0.25%.
The averages of the characteristic points are given in Table 2, curve type I
corresponding to ductile behaviour while curve type II to brittle material.
A05 50mm/min NA
12
0,054;
10 10,85
8
(Mpa)
4 max
2 A05 50mm/min
0
0 0.5 1 1.5
()
Figure 6: Stress-strain diagram for non-aged acrylic samples under low loading rate - Type I
A21 50mm/minA A19 250mm/min A

15 15
10 10
(Mpa)
(Mpa)
5 5
0 0
0 0.01 0.02 0.03 0 0.01 0.02 0.03
() ()
Figures 7a and b: Examples of stress-strain diagram for acrylic samples-Type II
200
Connections
Table 2: Characteristic values of the acrylic material under uniaxial tension
Chord- Stress at Strain at
Factor Curve modulus break/constriction break/constriction
combination type
(MPa) (MPa) (-)
50mm/min-NA I 1033.88 10.41 0.056
50mm/min-A II 1181.76 10.59 0.023
250mm/min-NA II 1495.16 13.67 0.020
250mm/min-A II 1383.34 13.11 0.020
4. Uniaxial compression tests
By numerical simulations on silicone it has been shown, that the uniaxial test data in
tension were clearly insufficient to fully describe its hyperelastic material behaviour.
Therefore a set of uniaxial quasi-static compression tests was carried out.

A Teflon mould stemming from the cut of a circular standard silicone cartridge was
prepared. The silicone was poured inside and the free extremities pressed with two
Teflon plates to remove the surplus. The diameter of the sample was 45mm with a
height of 20mm, see Figure 8. A variance analysis on the weight of the manufactured
samples allowed proving the exactness and reproducibility of this method.
Figure 8: Silicone sample for Figures 9a and b: Silicone installed in the set-up and tested under
compression tests compression
As for the tension specimens, the silicone samples for compression were stored for one
month at ambient temperature and humidity, shielded against light radiation. After this
period, half of the samples were subjected to the artificial ageing as described in section
2.2.
4.2. Tests
An Instron mobile hydraulic piston with a system composed by two fully parallel steel
plates was assembled, see Figures 9a and b. A Teflon spray was pulverized on the upper
and lower steel surfaces to minimize friction between the sample and the support
(“barrel” effect). The Poisson’s ratio was evaluated thanks to an image-processing
program, using the video recording of the tests.
201
Challenging Glass 3
The design of the test program was carried out with the help of the Taguchi method;
artificial ageing and the loading rate have again been identified as governing test
variables. Compared to the tension test campaign, only the loading rates changed, i.e.
for this test type the minimal value was 5mm/min and the maximal 50mm/min.
4.3. Results
The results showed non-linearity and the stability of the silicone behaviour towards
artificial ageing and the loading rate, see Figure 10.
Stressstraindiagram
7
6
SC055mm/minNA
5
SC205mm/minNA
4 SC025mm/minA
(MPa)
SC065mm/minA
3
SC1550mm/minNA
2 SC1850mm/minNA
SC1050mm/minA
1
SC0150mm/minA
0
0 0.2 0.4 0.6 0.8
()
Figure 10: Stress-strain curves for silicone samples under uniaxial quasi-static compression loading
(NA=non-aged, A=aged)
After a deformation in compression of more than 70% of the samples height, the
samples recovered completely their initial shapes during unloading. For all the samples,
the determination of the Poisson’s ratio led to a value close to 0.5; thus proved the
incompressible hyperelastic material behaviour of silicone.
5. Shear push-out tests

The shear properties of the adhesives are essential to apprehend the behaviour of
structural steel-glass connections. In the majority of building constructions, like the
structural sealant glazing systems, the adhesive is subjected to shear forces arising from
the self-weight of the glazing element (long-term action), temperature variations
(middle-term action) and the transfer of wind loads to the sub-structure (short-term
action).

The global geometry of the assembly was corresponding to the double-lap joint samples
investigated by Hart Smith [9]. The specimens were composed by two outer fully-
tempered glass plates with dimensions of 240x200x12mm and an inner steel plate with
dimensions of 200x200x12mm. The adhesive was disposed symmetrically on both sides
at the steel-glass interfaces. In this way, peel stresses were avoided and only pure shear
was considered. All the tests were realized by compression to avoid tensile stresses
crucial for the brittle glass elements.
202
Connections
Again, the Taguchi method was used to define the test program matrix for the two
adhesives. Four influent factors were evaluated for silicone, i.e. the artificial ageing, the
bonding area, the adhesive thickness and the loading rate; for the acrylic three
influencing factors have been identified, i.e. the artificial ageing, the bonding area and
the loading rate.
The silicone test specimens were manufactured in two different ways:
x Procedure 1 is based on an injection system to make four bands of silicone of

(160x30mm) with two bonding thicknesses of 3mm and 6mm. A self-designed
injector adaptable to the cartridge, see Figure 11, allowed the introduction of
the adhesive in the pre-assembled element. The appropriate dimensions of the
joint were obtained with the help of Teflon spacers, see Figure 12.
x Procedure 2 covers a simple pouring process to realize samples fully glued at
the steel-glass interfaces, 2x (160x170mm), for the two thicknesses of 3mm
and 6mm. With this method the different parts were added horizontally step by
step beginning with the application of the adhesive on a first glass plate, then
by arranging the steel plate on the first part with a calibrated system composed
by Teflon spacers. The specimens were finally subjected to compression to
evacuate the surplus.
Figure 11: Injector for silicone samples Figure 12: Pre-assembly for injection of the 4 strips
For the production of the acrylic specimens, polyethylene films of 0.1mm were used as
template spacers to obtain a constant thickness and the desired shape of the adhesive. As
for the pouring process of the silicone samples, each layers were brought step by step on
a calibrated piece, see Figure 13. For the acrylic test campaign, specimens in analogy to
the silicone specimens (glass-steel-glass) and assemblies made fully with glass elements
(i.e. inner steel plate was replaced by a glass plate) were produced.
Figures 13: Calibrated piece for acrylic push-out specimens preparation
203
Challenging Glass 3
The specimens were stored for one month at ambient temperature, humidity and
shielded from light. Again half of the specimens were subjected to the artificial ageing,
as described in section 2.2.
5.2. Tests
All the tests were carried out on a large Instron hydraulic press of 630kN maximum
capacity. Steel-glass contact was avoided with a Teflon spacer at the bottom support,
although allowing the lateral expansion of the glass, see Figures 14 and 15. As the
rotation of the glass plates around their supported edges was undesired, a polyamide
spacer was inserted between the two glass plates to keep the gap constant.
Figure 14: Push-out test set-up Figure 15: Spacer avoiding rotation of the glass plastes
Finally six displacement sensors were placed on this assembly:

x Two vertical sensors, measuring the relative displacement between the steel
plate and a glass plate (rotation of the steel plate also estimated).
x Two horizontal sensors, positioned perpendicularly and at the top of the glass
surfaces to ensure that no “opening” of the sample occurred.
x Two lateral sensors, parallel to the surface of glass plate to measure the
expansion of this element.
The type of failure, either adhesive, cohesive or in the adherend was documented.
5.3. Results
For all the silicone tests, the “opening”, which means changes in the horizontal distance
between the glass plate and the steel sheet, and lateral displacement in the glass were
insignificant. The shear stresses were calculated by dividing the load by the net bonding
area measured after breakage with an image-processing program. The shear strains were
obtained by dividing the relative displacement by the thickness of the silicone layer. The
shear stress-strain diagram is shown in Figure 16. The behaviour is non-linear and for
all curves in the same range for their first part up to a strain of about 0.75. At present,
further additional tests must be carried out to complete the Taguchi design of
experiments (containing 16 tests) and to finalise the study on the influence of the four
considered factors.
204
Connections
Shearstressstraindiagram
1.8
1.6
GPO15NA4B33
1.4
GPO13NA4B390
Shearstress (MPa)
1.2 GPO23NA4B63
GPO24NA4B690
1 GPO16NA2B33
0.8 GPO14NA2B390
GPO18NA2B63
0.6 GPO02A2B33
0.4 GPO04A2B390
GPO05A2B63
0.2
0
0 0.5 strain () 1 1.5
Figure 16: Stress-strain curves for silicone samples under quasi-static simple shear loading
(NA=non-aged, A=aged; 4B=4strips, 2B=2strips; 3/6=thickness; 3/90= loading rate in mm/min)
The failure of the silicone push-out specimens was 100% cohesive which demonstrated
the sufficient adhesion between steel and glass materials.
For the acrylic specimens fully made with glass elements, failure of the system always
corresponded to the breakage of the glass adherend, see Figure 17. For the push-out
specimens with an inner steel plate, the failure was always noticed as adhesive failure to
the steel surface, see Figure 18. Comparing the maximum shear stress for the same
adhesive joint dimensions for both specimens, similar values were found with an
average of approximately of 3.4MPa.
Figure 17: Adherend failure for fully-glass push- Figure 18: Adhesive failure for steel-glass push-out
out tests tests
The relative displacements between the inner element and the outer glass plate were not
measurable with the connected displacement sensors and so shear stress-strain diagram
could not be plotted. Further enhanced tests will be design and conducted to determine a
constitutive material law for the acrylic under pure shear.
6. Numerical simulations
Basic numerical simulations on tension, compression and shear were conducted with
material constitute models already existing in the finite element software Abaqus®. The
most common incompressible hyperelastic models to represent the behaviour of silicone
205
Challenging Glass 3
are the Mooney-Rivlin, Yeoh, Neo-Hooke, Ogden and the polynomial forms. The
numerical results were compared to the experimental test results.
As the results of the numerical simulations with these material laws implemented did
not show a sufficient concordance with the experimental tests, an own model based on
the strain energy potential (1) was developed and implemented via a subroutine in the
software code, see Figure 19, bottom diagram.
D D
U A ( I 1 3) B ( I 1 3) 2 C ( I 2 3) (1)
(I 2 3 E ) E
with the following coefficients determined for the results of the test carried out so far:
Table 3: Coefficients of the own-developed material law (1)

A B C D E
0.259101 -0.0100315 0.0444932 0.00902896 0.186184
The results of the numerical analysis with the developed material law implemented are
in concordance with the experimental test result and, thus, the developed material law is
recommended for detailed design of adhesive joints made from silicone.
Figure 19: Comparison of numerical simulations with self-developed law to experimental data
206
Connections
7. Summary and future works
The stability of the silicone regarding the quasi-static loading rate and the ageing was
demonstrated and a hyperelastic material law was developed. This law was verified
successfully via the simulation of the double-lap joint test assembly. Changes could be
brought to this law after the realization of the cyclic tests which will take the Mullins
effect into consideration.
The stress-strain behaviour of acrylic is clearly governed by the loading rate and the
ageing. Numerical simulations of the push-out tests based on the tensile properties of
this adhesive showed that the material law is still not sufficient. Further enhanced tests
on the acrylic adhesive will be undertaken.
Long-term properties are still missing for the both adhesives and should be investigated
to evaluate the effect of permanent actions.
8. References
[1] Haldimann, M.; Luible, A.; Overend, M., Structural Use of Glass, IABSE, ETH Zürich, 2008
[2] Weller, B.; Kothe, M.; Nicklisch, F.; Schadow, T.; Tasche, S.; Vogt, I.; Wünsch, J., Kleben im
konstruktiven Glasbau, Stahlbau-Kalender 2011, Berlin: Ernst & Sohn, p.596
[3] ETAG 002 Guideline for European Technical Approval for SSGS, EOTA, 2001
[4] Puller, K.; Sobek, W., Glass-Steel Connections Using Acrylate Adhesives, Challenging Glass,
Conference on Architectural and Structural Applications of Glass, TU Delft, Netherlands, 2008
[5] EN ISO 11431 – Jointing Products: Determination of Adhesion/Cohesion Properties of Sealants after
Exposure to Heat, Water and Artificial Light through Glass. CEN, 2002
[6] Saad, P., Modélisation et identification du comportement non linéaire des cales en caoutchouc, Thesis,
France, 2003
[7] EN ISO 527-2: Determination of Tensile Properties of Plastics. CEN, 1996
[8] Vigier, Michel, Pratique des Plans d’Expériences – Méthodologie Taguchi et Compléments, Les
Editions d’Organisation, Paris, 1988
[9] Hart-Smith, L.J., Adhesive-bonded Double-Lap Joints, NASA Technical Report, USA, 1973
207
doi:10.3233/978-1-61499-061-1-209
Shear-Capacity in Adhesive Glass Joints

Maria Fröling, Kent Persson and Oskar Larsson
Lund University, Sweden, maria.froling@construction.lth.se
The shear-capacity of adhesive glass-joints was investigated. Various stiff and soft
adhesives were tested in a short-term load-case. The tests were conducted with
small specimens in order to achieve a homogenous state of stress. The results of the
tests were used in order to determine the material models of the adhesives. Finite
element analysis of the test set-up was used for the determination of the material
models. Large-scale tests were conducted to verify the material models from the
tests of the small specimens. It could be concluded that with further validation, a
combination of small-specimen tests and finite element simulations may allow for
the determination of joint behavior for any joint size.
Keywords: Glass, Adhesive Joint, FEM, Shear-capacity
1. Introduction
Recently, there has been an increasing interest in using glass as a structural material.
When constructing with glass, it is often necessary to connect different structural
elements. The most common technique used for joining glass elements is to use bolted
joints. The use of bolted joints leads to stress concentrations in the glass. Glass is a
brittle material, which makes it sensitive to stress concentrations.
An alternative is to use adhesive joints to connect the glass elements. In general,

adhesive joints are capable of distributing the stress over the surface of the joint so that
stress concentrations are avoided. With many adhesives it is also possible to keep the
transparency of the glass at the joint.
Adhesive joints are normally designed to be loaded in a state of shear rather than in a
state of tension. This paper investigates the shear-capacity of a set of common adhesives
in a short-term load-case. The adhesive products are chosen so that a wide span of
different adhesive characteristics is obtained. The shear-capacity of the adhesives is
tested in pure shear. From the experimental data, material models for the adhesives are
determined. The material models are valid for a short-term load-case. The material
models are determined through a finite element analysis of the complete test set-up.
Later, the material models are verified through large-scale tests. Further information
about the work could be found in [1].
2. Shear-Capacity Tests of Small Specimens

The main purpose of the tests was to determine the shear-capacity of various adhesives
for connecting glass. For this reason, it is important that the tests create a situation as
close as possible to a state of pure shear. In the tests, small specimens were used. Small
specimens ensure a relatively homogenous state of stress and it is easier to ensure
209
Challenging Glass 3
fracture in the adhesive and not in the glass. The shear-capacity was evaluated for a
short-term load, i.e. a load that was applied with a fairly high rate with the aim of
causing failure in the adhesive. A constant shear strain rate at approximately 3 % per
second was chosen for the tests.
2.1. Testing Equipment

To obtain a situation of pure shear in the adhesive, the test equipment was designed
with the following characteristics. Firstly, all loads were applied centrically to avoid
eccentricity that may cause tensile and compressive stresses to arise in the adhesive.
Secondly, the adhesive had to be be able to freely expand/shrink in the direction
perpendicular to the direction of the shear forces to avoid stresses caused by
constraining the material strains.
A schematic drawing of the testing equipment is displayed in Figure 1. It consisted of

two steel-parts that transmit the forces from the testing machine to the specimen. For
stiff glues, the test equipment was loaded by compressing the steel-parts. For the softer
adhesives the test equipment was loaded with tensile forces in order to allow the large
deformations in the joints.
The specimens in the tests consisted of two pieces of glass with dimensions 20 × 20
mm2 joined together with an adhesive layer. Two different specimens were used in the
tests. Specimen 1 had an adhesive layer that fully covered the surface of the glass-parts
and this specimen type was used for the softer adhesives. Specimen 2 had an adhesive
layer of dimensions 5 × 20 mm2. It was used to test the stiffer glues in order to reduce
the applied force needed to conduct the test. The geometries of the two specimens are
shown in Figure 2.
2.2. Measurements
A MTS testing machine was used to apply force to the steel-parts. Data regarding the
applied force and displacements were collected every 0.5 s. In order to obtain the shear-
capacity for the adhesives, avg,u, the ultimate shear force obtained from the
measurements were divided by the initial surface area of the adhesive.
3. Tested Adhesives
The tested adhesives can be grouped into softer adhesives and stiffer adhesives. The
softer adhesives contained four types of silicone based adhesives, three types of SMP
(Silyl Modified Polymer) based adhesives and Bostik Multifog 2640. The stiff
adhesives consisted of polyurethane adhesive, HBM Rapid Adhesive X 60, strong
epoxy adhesive and UV-hardening glass-glue. Information about the adhesives above
can be found in [2]-[5].
Test specimens had different thickness of the adhesive layer. For the silicone glues, the
thickness was 6 mm, for the SMP based adhesives and Bostik Multifog 2640, the
thickness was 2 and 0.3 mm. The polyurethane glue and HBM Rapid adhesive X 60 had
an adhesive thickness of 0.2 mm, whereas the strong epoxy and the UV-hardening glue
had 0.3 mm as thickness of the adhesive
210
Figure 1: The test equipment used in the shear-capacity tests, tensile load.
Figure 2: Drawing of the two different types of specimens.
4. Finite Element Modeling of the Small-scale Tests

A finite element model of the entire test arrangement was developed, Figure 3, where
the geometry of each of the joints was modeled. The steel-parts were modeled as linear
elastic materials with the material parameters E = 210 GPa and = 0.3, where E denotes
the modulus of elasticity and denotes the Poisson’s ratio. Glass was modeled as linear
elastic with the material parameters E = 70 GPa and = 0.23.
211
Challenging Glass 3
Figure 3: Finite element model of the test arrangement.
The evaluation of the results consisted of plotting the measured data of the shear-force
versus the deformation of the test series. The data was fitted to a polynomial curve using
the least-squares’ method and compared with the data extracted from the finite element
simulations. For each adhesive, different material models were tested until a satisfying
agreement was obtained. For the softer adhesives, the hyperelastic material models Neo-
Hooke and Mooney-Rivlin were tested whereas the stiffer adhesives were modeled as
linear elastic, all with = 0.25. From the respective matching material model, data on
the shear-deformation were extracted and relationships between shear-stress and shear-
strain were established.
An initial shear modulus, G, was calculated from the shear-stress versus shear-strain
diagrams and an ultimate shear-stress, avg,u, was determined as the maximum value of
the shear-stress for the increment closest to the average of the maximum load capacity
of each adhesive.
5. Results from the Small-scale Tests

For brevity, results for one softer adhesive and one stiffer adhesive are presented. In
Figure 4, experimental results and simulation results on force versus deformation for
one SMP based adhesive is displayed (top). In the same figure, results from finite
element simulations on average shear-stress versus shear-strain for the same adhesive
are shown (bottom). The finite element results and the experimental curves coincide
initially and overall the numerical and experimental results show close agreement. The
material model for the adhesive was determined with good accuracy. The lower graph
has a value of the maximum average shear-stress close to the experimentally obtained
value of around 2.3 MPa.
In Figure 5, the corresponding graphs are shown for the polyurethane glue. There was a
perfect agreement between simulations and measurements for this case. The
experimental maximum value for the average shear-stress of the polyurethane adhesive
was 3.8 MPa.
212
Figure 4: Results from the FE-evaluation of SMP based adhesive 3, 2 mm specimens.

Top: experimental results (circles) and FE-results (line).
Bottom: Average shear-stress versus shear-strain extracted from the FE-model.
As a summary of further results, diagrams showing average shear-stress versus shear-

strain are displayed in Figures 6 and 7 for softer adhesives of a certain layer thickness
and for stiffer adhesives respectively. From the graphs it is clear that there were
differences in mechanical behavior between the adhesives within each group.
From Figure 6 it could be observed that Bostik Multifog 2640 was the softest adhesive
and also had the lowest ultimate shear-strength of this group of adhesives. Among the
three other adhesives of this group, Figure 6 shows that different adhesives could have
similar stiffness but different ultimate strengths and vice versa.
213
Challenging Glass 3
Figure 5: Results from the FE-evaluation of the polyurethane glue.

Top: experimental results (circles) and FE-results (line).
Bottom: average shear-stress versus shear-strain extracted from the FE-model.
From Figure 7, it could be seen that the epoxy adhesive was the stiffest adhesive and
had the highest ultimate load. For the other stiff adhesives, the stiffness was quite equal
but there were differences in ultimate shear-strength.
Results showing the mechanical characteristics of the two groups of adhesives (softer, 2
mm specimens and stiffer) are shown in Table 1. In general, the stiffer adhesives had
greater stiffness (G) and ultimate shear-strength (avg,u) than the softer adhesives.
The parameters of the material models for the softer adhesives are shown in Table 2. C10,
C01 and D1 are parameters of the material models. For the stiffer adhesives, the material
model parameters are displayed in Table 3.
214
Figure 6: Shear-stress versus shear-strain for the 2 mm specimens of the softer adhesives.
Bostik Multifog 2640: (x), SMP based adhesive 1: (+), 2: (o), 3: (triangle).
Figure 7: Shear-stress versus shear-strain for the stiffer adhesives, 0.2-0.3 mm specimens. Polyurethan glue:
(x), HBM Rapid Adhesive X 60: (+), strong epoxy: (o), UV-hardening glass-glue: (triangle).
Table 1: Mechanical characteristics of adhesives.

Quantity Softer Adhesives (2 mm Specimens) Stiffer Adhesives
G [MPa] 0.5 - 1.2 83 - 500
avg,u [MPa] 1.3 – 2.3 4 - 20
u [%] 200 – 300 4 - 10
215
Challenging Glass 3
Table 2: Parameters of material models for softer adhesives.

Adhesive Material Model C10 C01 D1
Silicone Based 1 Mooney-Rivlin 51·103 77·103 -
3
Silicone Based 2 Neo-Hooke 93·10 - -
3
Silicone Based 3 Neo-Hooke 247·10 - -
3 3
Silicone Based 4 Mooney-Rivlin 10·10 190·10 -
3 3
SMP Based 1 Mooney-Rivlin 200·10 330·10 -
3 3
3 3
3 3
Bostik Multifog 2640 Mooney-Rivlin 100·10 150·10 -
Table 3: Parameters of material models for stiffer adhesives.

Adhesive Material Model E [MPa]
Polyurethane Glue Linear Elastic 200 0.25
HBM Rapid Adhesive X 60 Linear Elastic 320 0.25
Strong Epoxy Linear Elastic 1500 0.25
UV-hardening Glue Linear Elastic 300 0.25
6. Large-scale Testing
A large-scale experimental test was made with the aim of determining the shear-
capacity of an adhesive joint in a large dimension glass beam.
Five adhesives from the small-scale tests were chosen to be tested in the large-scale
tests. The five adhesives were chosen considering the results from the finite element
simulations of the corresponding test set-up, see below. The two strongest SMP based
adhesives and the stiffer adhesives polyurethane glue, UV-hardening glass-glue and the
strong epoxy adhesive were chosen. When performing the tests, the deformation speed
was kept constant at 10 mm/min.
6.1. Testing Equipment

The test performed was a four-point bending test of a beam that consisted of three
beams made of flat-glass joined together by two adhesive joints. The test arrangement is
shown in Figure 8 and creates a symmetrical beam in order to obtain pure shear-
stresses in the joint. The beam had a span of three meters and every flat-glass element
had dimensions 250×2000 mm and a width of 12 mm. The adhesive joints had
dimensions 250×250 mm.
6.2. Measurements
Data on load and deformation of the mid-point of the beam were collected.
216
7. Finite Element Modeling of the Large-scale Tests

A finite element model was made of the large-scale test set-up. In the tests, the stiffer
adhesives were simulated with a thickness of 0.2 mm and the softer adhesives with a
thickness of 2 mm. The glass elements were modeled with a modulus of elasticity of 70
P/2 P/2
250
A A
250
7 6
5 4 3
1200
A-A
2000
Adhesive layer
12
B B
3000
Figure 8: Test arrangement for large-scale tests.
GPa and a Poisson’s ratio of 0.23. The adhesives were modeled according to the
material models from the tests with the small specimens. Load-deformation graphs were
determined for the tested adhesives, and maximum load values were determined and
were set equal to the load that corresponded to the maximum shear-stress, max, of each
adhesive.
8. Results from the Large-scale Tests

Figure 9 displays the principal pattern of shear-stresses in the joints of the large-scale
tests. In the stiffer adhesives stress-concentrations occurred at the corners of the joints.
In the softer adhesives, the stresses were more evenly distributed.
In Figure 10, a graph showing load versus deformation for the third SMP based
adhesive is shown. The measured data show a less stiff behavior than the data from the
simulations. The measured data end at a lower load because of failure of the glass beam
before the ultimate load was reached.
A load versus deformation diagram for the polyurethane adhesive is displayed in Figure
11. After an initial deviation, the stiffness of the measured data compares well with the
stiffness of the results from the finite element simulations.
Table 4 summarizes the ultimate loads and deformations for a selection of the softer
adhesives. The results are obtained from the finite element simulations. The
corresponding results for the stiffer adhesives are shown in Table 5.
217
Challenging Glass 3
Figure 9: Principal pattern of the shear-stresses in the joints of the large-scale tests.
Left: softer adhesives. Right: stiffer adhesives.
Figure 10: Comparison of load versus deformation diagrams for FE-simulations and experimental tests for the
third SMP based adhesive. Top: FE-simulations. Bottom: experiments. The different points of measurements
are marked: 4 (o), 5 (+). See Figure 8.
Table 4: Ultimate loads (P = P/2 + P/2) and deformations at the mid-point of the beam (4) for softer adhesives.
Adhesive Ultimate Load [kN] Ultimate Deformation [mm]
Bostik Multifog 2640 28.8 53
SMP Based Adhesive 1 49.3 51
218
Figure 11: Comparison of load versus deformation diagrams for FE-simulations and experimental tests for the
polyurethane adhesive. Top: experiments. Bottom: FE-simulations. The different points of measurements are
marked: 4 (o), 5 (+). See Figure 8.
Table 5: Ultimate loads (P = P/2 + P/2) and deformations at the mid-point of the beam (4) for stiffer adhesives.
Adhesive Ultimate Load [kN] Ultimate Deformation [mm]
Strong Epoxy 30.3 10.0
Polyurethane Glue 10.3 3.5
UV-hardening Glue 22.3 7.5
HBM Rapid Adhesive X 60 20.3 7.0
Overall, the softer adhesives had ultimate loads ranging between 28 and 50 kN and the
stiffer adhesives had ultimate loads in the interval 10-30 kN.
From the finite element simulations it was apparent that all the beams of the large-scale
tests would suffer failure in the glass of the beam. Thus, no ultimate load of the joints
could be measured by means of experimental testing. Comparing Figure 9 and the
results of tables 4 and 5, the apparent stress concentrations of the stiffer adhesives
reduced their ultimate load compared to the softer adhesives.
9. Conclusions
It can be concluded that the test method suggested for the small-scale tests is a
functioning method for evaluating the shear-capacity of the adhesives. The method
creates a state close to pure shear.
Material models valid for the short-term load-case could be determined with close
accuracy between experiments and simulations for the group of stiffer adhesives. For
the SMP based adhesives, the correlation between simulations and experiments is
shown to be good for the small-scale tests. For the large-scale tests, there are deviations
219
Challenging Glass 3
between simulated and experimental results. An explanation of this deviation is that

these adhesives were not fully hardened at the moment of the experiments. Further
research is necessary to validate the material models for the SMP based adhesives for a
large-scale joint.
In the small-scale tests stiffer adhesives give a stronger joint than softer adhesives. Thus,
stiff adhesives may be used for small joints. From the large-scale tests it is concluded
that for a large joint, a more soft adhesive may give the strongest joint.
With further validation, the methodology presented may be used to predict the
mechanical behavior of any joint size through a combination of small specimen tests
and finite element modeling.
10. Future Work

Further tests are conducted to investigate the shear-capacity of and to determine
material models for a complementary set of adhesives. Initially, the tests are made for a
short-term load-case. Later, the tests will be performed for a long-term load case and the
creep parameters of the adhesives will be determined.
The support from The Swedish Research Council FORMAS, Glasbranschföreningen
and Svensk Planglasförening is gratefully acknowledged.
12. References
[1] Larsson, Oskar, Shear Capacity in Adhesive Glass-Joints, Master’s Thesis, Division of Structural
Mechanics at Lund University of Technology, Sweden, 2008.
[2] Product sheets of different Casco adhesives, Akzo Nobel Bygglim AB, Casco, Stockholm, Sweden,
www.casco.se
[3] Product sheets of X 60 adhesive, Hottinger Baldwin Messtechnik, Darmstadt, Germany, www.hbm.com
[4] Product sheets of different Bostik adhesives, Bostik AB, Helsingborg, Sweden, www.bostik.se
[5] www.sika.com
220
doi:10.3233/978-1-61499-061-1-221
Experimental and Numerical Analysis of

Edge Seal Spacers of Insulated Glass Units
for Structural Sealant Glazing Applications
Anneliese Hagl
A. Hagl Ing. GmbH, Germany, www.a-hagl-ingenieure.de
For applications beyond the European Technical Guideline ETAG 002 for
structural sealant glazing (SSG), a detailed knowledge of the mechanical
characteristics of all components is required. While the physical properties of glass,
steel and aluminum and silicone adhesives are quite well known or in the focus of
current research activities, edge seal spacers of insulated glass units have not been
investigated yet in detail. Edge seal spacers show a high level of design variations
and thus their behavior differ significantly with respect to the various designs. This
paper focuses on experimental results of compressive tests of edge seal spacers
which build the basis for the derivation of numerical models allowing to analyze
SSG designs and applications beyond ETAG 002.
Keywords: Edge Seal, ETAG 002, Insulated Glass, Spacer, SSG,
1. Introduction
Usually, structural sealant glazing (SSG) designs are covered by the European guideline
ETAG 002 [1]. As the bond between glass and supporting structure shows complex
mechanical performance characteristics, simplifying assumptions are derived for sizing
neglecting also the characteristics of the edge seals and their impact on insulating glass
units. Thus a detailed knowledge of bond and edge seal material behavior is not
required. Nevertheless, progress demonstrated that in-detail studies lead to improved

physical insight which might be important for applications beyond ETAG 002 as for
example cold bending applications [2].
In order to extend numerical procedures to insulated glass units the adequate modeling
treatment of structural edge seals is of major importance. Edge seals ensure the air
tightness of the insulating glass under operating conditions and provide spacing between
the glass panes. Spacers are the core components of edge seals differing significantly in
layout. The design of the spacers might be based either on metals or polymers or on
combinations enclosed by butyl strips. Regarding metallic designs, stainless steel or
aluminum are first choices for material. For polymers, additional distinction is made
between foam and homogenous material. Thus, mechanical properties of the spacers are
expected to differ significantly.
As first step towards understanding of edge seal behavior, compressive tests were
performed of spacer specimens of various designs. The baseline specimens consist
usually of pieces of approximately 50 mm length. This paper presents experimental
221
Challenging Glass 3
results showing significant differences in the structural performance of the various

designs. Detailed finite element studies of the tests were performed in order to check for
applicability of numerical procedures for predictive capabilities of structural edge seals.
Beyond the experimental results, focus in the paper is put on the derivation of simplified
mechanical models for application on global structural SSG models. Outlook of these
research activities is to improve SSG sizing towards higher loads taking into account
edge seal behavior.
2. Insulated Glass Units and Edge Seal Spacers

Insulated glass units consist of two or more glass panes with the volume(s) between the
panes filled by air or gas in order to decrease thermal conductivity for improved
insulation properties. In addition to the glass panes the insulated glass units are
composed of a combination of edge seals and edge seal spacers for mechanical integrity
with the edge seals typically based on silicone adhesives and the spacers as dedicated
sub-components, see Figure 1. This combination acts in parallel as sketched in Figure 2
when the insulating glass units are exposed to wind loads or similar. The stiffness
properties of both glass panes and spacers determine the load share between spacers and
edge seal. Other load cases especially related to insulating glass units are climate loads
due to the expansion of the insulating fluid between the panes in case of temperature
increase and installation loads due to different atmospheric conditions (pressure
altitudes) between manufacturing site and installation site. The following table presents
the characteristics of the edge seal loading for these load cases assuming constant
pressures (isobaric case) or constant volumes (isochoric case) of the insulating fluid as
extreme cases:
Table 1: Insulating glass unit load cases.
Insulating fluid volume Insulating fluid pressure Resulting edge

Specimen
(isobaric case) (isochoric case) seal loading
High
Increasing volume Increasing pressure Tensile loading
temperature
Low Compression
Decreasing volume Decreasing pressure
temperature loading
High pressure
Increasing volume No change Tensile loading
altitude
Low pressure Compression
Decreasing volume No change
altitude loading
While the isobaric case is related to infinitely flexible glass units, the isochoric case
represents infinitely stiff glass units. Due to flexibility of the glass panes and the other
insulating glass components neither the pure isobaric case not the isochoric case are
expected in reality. Instead a combination of volume and pressure changes will occur.
222
Experimental and Numerical Analysis of Edge Seal Spacers of Insulated Glass Units for Structural Sealant
Glazing Applications
Figure 1 : Sketch of insulated glass unit used in SSG design [2].
Figure 2 : Load schemes of edge seals of insulated glass units.
Edge seal spacers are used in insulating glass units to fix the distance between the glass
panes of the unit. Typically the spacers are composed of a body consisting of either
metals or polymers or combinations of both plus adhesive strips on each contact surface
of the glass panes based on butyl material. Figure 3 presents typical edge seal spacer
designs while Figure 4 shows the functional environment of an edge seal spacer.
Inside insulating Towards edge

glass unit sealant
(inboard) (outboard)
Figure 3 : Edge seal spacer designs: Left side visible inside insulating glass unit.
The spacer is attached to the glass panes by an inner seal made of butyl material.
Depending on the spacer design the butyl has to be added to the spacer or is directly
integrated. Thus loads transferred from the glass panes to the spacers have to pass the
butyl strips. E.g. for compressive or tensile loads, the edge seal spacer assembly can be
mechanically interpreted as three flexible elements put in series. Thus the effective
stiffness of the edge seal spacer assembly can be approximated by calculating the
harmonic mean of the butyl and spacer stiffness values.
223
Challenging Glass 3
While design procedures are available for the edge sealant in case of SSG designs by
the guideline ETAG 002, the impact of the spacer assembly on the global mechanical
behavior is typically neglected in design codes. Furthermore, although detailed thermal
analysis can be found in the advertisements of the manufacturers for advanced spacer
designs related mechanical properties are never indicated. Therefore, several spacer
designs were experimentally investigated in order to get an overview of the mechanical
behavior of these components.
3. Experimental Activities
3.1. General Approach

Load cases of interest for the edge seal spacers are tensile and compressive loading and
shear loading. Regarding these load cases, the compressive load case is selected as most
interesting one as – comparing to the tensile load case – no adhesive failure is expected
and compared to the shear load case, tensile and compressive load cases are assumed to
be more complex in their mechanical behavior. In order to reduce buckling two test
articles are combined into one specimen as shown in Figure 5. Please note that in the
insulating glass unit application case the edge seal spacer is stabilized by the edge
sealant.
Figure 4 : Elements of edge seal spacers. Figure 5 : Test set-up for edge seal spacer testing.
Beyond understanding of the spacer behavior and insight into deformation patterns,
quantitative objective of the test campaign is the derivation of stiffness parameters for
the various spacer designs. In this first exploration phase monotonous load histories
were applied. Based on experience of the first test campaign more complex load
histories will be designed e.g. in view of tensile loading and of cyclic loading.
3.2. Component Testing: Butyl Strips

While the linear-elastic material description of metallic parts such as aluminum and
steel is straight forward using well-known values for the Young’s modulus and the
Poisson’s ratio, the adequate description of the material behavior of the butyl strip is
less obvious. Therefore, butyl strips of 3.5 mm width were separately tested under
compressive loads on 50 mm strip length in a similar manner as the edge seal spacers.
Figure 6 presents the results in terms of loads versus displacements. The curve shows a
progressive behavior i.e. stiffness increases with loads. While for low loads, the
stiffness will probably be mainly related to the butyl strip, for very high loading and
thus very high stiffness, the flexibility of the test set-up has to be considered for
224
quantitative evaluation as well. After unloading and inspecting the specimen, it can be
seen that the butyl strip deforms mainly in a plastic manner.
As first step the material characteristics obtained by the load curve are mapped into
elastic material laws – either linear or hyperelastic. This kind of black box approach –
relating displacement inputs to load outputs – is adequate as long as the load scheme –
compressive monotonous loading – is not changed. Future steps might result in the
application of plastic or elasto-plastic material laws for the Butyl strips. For the current
numerical studies, it is assumed that the deformation range of the Butyl strips do not
exceed a range of 5% - 10% engineering strain. Thus the material laws for the Butyl
strip were adjusted within these intervals.
1000
750
Load - N
500
250
Displacement - mm
0
0 0.25 0.5 0.75
Figure 6 : Experimental results of butyl strip testing.
3.3. Testing of D-shape Aluminum Edge Seal Spacers

Next, the various edge seal spacers were tested in a similar manner. Figure 7 shows the
design of the first test series featuring D-shaped aluminum bodies of three different
heights. The specimens were delivered without butyl strip and with heights of 13.5 mm
(labeled small size in the paper), 15.5 mm (labeled mid size in the paper) and 17.5 mm
(labeled large size in the paper). It is assumed that the nominal spacing amounts to
14 mm, 16 mm and 18 mm respectively leading to nominal butyl strip thickness values
of 0.5 mm in total for both ends. The butyl strips used for the derivation of material
laws showed a thickness of approximately 0.9 mm. These strips were also applied to the
D-shape spacers within this test campaign.
225
Challenging Glass 3
Figure 8 presents the mechanical characteristics of the D-shape aluminum spacers. The
figure left shows the distributed loads, i.e. the loads per length spacer, for a
representative specimen. The decrease of loads for large displacements is due to a kind
of buckling. It is expected that the real loading range is significantly below this load
peak. Therefore, results for the three designs are presented in the figure right. The
following statements can be drawn from this figure: First the results rank in highest
stiffness for the mid spacer size, immediately followed by the small spacer size and
finally the large spacer design with significantly lower levels. This is slightly
inconsistent to what is expected at first glance from theory. As the smallest spacer
theoretically features the highest stiffness, the expected ranking is small spacer – mid
spacer – large spacer.
Figure 7 : D-shape aluminum spacers. of heights 13.5 mm, 15.5 mm, 17.5 mm
-8 -10
Distributed Load - N/mm
01_1-13.5
01_2-13.5
-6 -8 03_3-15.5
03_8-15.5
03_9-15.5
-6 05_1-17.5
-4 05_2-17.5
-4
-2
-2
0 0
0 -0.5 -1 -1.5 -2 -2.5 0 -0.2 -0.4 -0.6 -0.8 -1

Displacement - mm Displacement - mm
Figure 8: Mechanical characteristics of D-shape aluminum spacers.
226
3.4. Testing of D-shape Steel Edge Seal Spacers

In accordance to the Aluminum spacers the D-shape steel spacers were tested and
evaluated in a similar manner. While Figure 9 shows a photograph of these spacers,
Figure 10 presents the experimental results in terms of load curves. The steel spacers are
slightly stiffer compared to the aluminum spacers and again dependency on size is not
consistent with expectations from a theoretical point of view.
Figure 9: D-shape steel spacers of heights 13.5 mm, 17.5 mm, 15.5 mm
-25 -10
02_1-13.5
-20 -8 04_1-15.5
04_2-15.5
-15 -6 06_1-17.5
06_2-17.5
-10 -4
-5 -2
0 0
0 -0.5 -1 -1.5 -2 -2.5 0 -0.2 -0.4 -0.6 -0.8 -1
Figure 10: Mechanical characteristics of D-shape aluminum spacers.
227
Challenging Glass 3
3.5. Testing of Composed Edge Seal Spacers

Spacers composed of polymer and metallic parts are gaining more interest in view of
low thermal conductivity where pure metallic designs show dis-advantages due to the
high thermal conductivity especially of aluminum. Figure 11 presents such a spacer
design for low conductivity composed of a polymer body and a steel surface sheet for
same heights as before. Figure 12 displays the related load curves leading to the
conclusion that the order of magnitude is similar to the pure metallic spacer designs.
Again, slight inconsistencies are visible with respect to the ranking of the stiffness
characteristics versus spacer size.
polymer
metallic part
Figure 11: Two-material spacers of heights 13.5 mm, 15.5 mm, 17.5 mm.
-20 -10
07_1-14
-8 07_2-14
-15 08_1-16
-6 08_2-16
-10 09_1-18
-4 09_2-18
-5
-2
0 0
0 -0.3 -0.5 -0.8 -1 -1.3 -1.5 0 -0.2 -0.4 -0.6 -0.8 -1
Figure 12: Mechanical characteristics of two-material spacers.
3.6. Comparison of Edge Seal Spacers 15.5mm Height

In order to check for design varieties, edge seal spacers of different shapes from other
manufacturers were tested as well. Figure 13 presents C-shape steel spacers, T-shape
aluminum spacers and a D-shape spacer composed of polymer and steel for a spacer
height of 15.5 mm. In Figure 14 results of 15.5 mm height spacers of steel (C-shape,
label S), aluminum (T-shape, label A) and the two-material D-shape spacer (label Co)
are presented. It can be seen that with respect to steel and aluminum spacers, the
stiffness is lower compared to the spacers presented in the sections before. Nevertheless,
the order of magnitude is the same. Regarding a comparison with the D-shape two-
material spacer, the mechanical characteristics are quite similar although the Young’s
modulus of the materials differs significantly.
228
polymer
body
metal
sheet
Figure 13 : C-shape steel spacer, T-shape aluminum spacer, D-shape two-part spacer design.
-15 -4

T-04_1-A T-04_1-A
T-05_1-Co T-04_2-A
-3
-10 T-03_2-S T-05_1-Co
T-03_1-S
-2
T-03_2-S
-5
-1
0 0
0 -1 -2 -3 -4 0 -0.2 -0.4 -0.6 -0.8 -1
Figure 14: Mechanical characteristics of different spacer designs of 15.5mm height
4. Numerical Results
4.1. Finite Element Analysis of Butyl Strip

First, a material law for butyl needs to be derived from the butyl strip measurements
mentioned above. In this phase, a simple isotropic linear-elastic material law was
selected in order to get an overview of the mechanical behavior. For the calculation of a
representative stiffness value for butyl, a load curve gradient was calculated based on
collocation points at 5% and 10% engineering strain levels leading to a stiffness of

approximately 2 N/mm per mm strip. A finite element model of the butyl strip was set
up and the related material law was adjusted in such a way that the stiffness of the
model coincides with that of the test. For the material law Poisson’s ratio is required in
addition to the Young’s modulus which will be accordingly adjusted. The Poisson’s
ratio was selected close to 0.5 meaning almost incompressible behavior which is typical
for hyperelastic material. Regarding the boundary conditions of the finite element model
it was assumed that the butyl specimen perfectly sticks to the fittings of the testing
machine. This approach is clearly motivated by the sticky behavior of the strips when
manipulated.
It is obvious that the selection of a linear material law is only a rough approximation for
getting first insight. During unloading of the specimens at the end of the tests, the
flattened butyl was totally destroyed and did not significantly recover the original shape.
Thus, the material shows an inelastic behavior to a high level. From a numerical point
of view further improvement might be the application of an elasto-plastic material law.
Regarding testing and subsequent correlations, cyclic tests on the one hand and other
229
Challenging Glass 3
load schemes such as cyclic, tensile or shear tests of the butyl strips are of high interest
for future exploration.
Figure 15 : Finite element analysis of butyl strip presenting normal stress distributions.
4.2. Finite Element Analysis of Nominal D-shape Spacers

Both 3D and 2D linear finite element analysis were performed for the D-shape
aluminum and steel spacers including butyl strips of nominal thickness of 0.25 mm on
each side. The spacers were modeled by shell elements while the butyl is treated by
solids as done in the section before. The comparison between 2D and 3D analytical
models allows to judge edge effects. The comparison between aluminum and steel will
the drawing of conclusions with respect to the material. The following table presents the
stiffness values derived from the finite element analysis. Please note that high
uncertainty is related to geometry. The aluminum spacers are assumed to feature a
thickness of 0.4 mm while for steel spacers, the thickness seems to be lower. For
comparison, thickness values of 0.4 mm and 0.3 mm are presented here. The stiffness
refers to a strip of 1 mm width leading to a unit of N/mm2.
Table 2: Numerical stiffness values for D-shape spacers.
Spacer design Stiffness 3D model Stiffness 2D model Difference

Material – height - thickness N/mm2 N/mm2 %
Aluminum 13.5 mm, 0.4 mm 15.74 16.08 2.14
Aluminum 15.5 mm, 0.4 mm 15.63 15.97 2.13
Aluminum 17.5 mm, 0.4 mm 15.54 15.87 2.13
Steel 13.5 mm, 0.4 mm 17.66 18.07 2.32
Steel 15.5 mm, 0.4 mm 17.61 18.02 2.32
Steel 17.5 mm, 0.4 mm 17.58 17.98 2.32
Steel 13.5 mm, 0.3 mm 15.93 16.54 3.80
Steel 15.5 mm, 0.3 mm 15.84 16.44 3.79
Steel 17.5 mm, 0.3 mm 15.76 16.36 3.77
230
Figure 16 : Finite element analysis of D-shape spacer indicating large compressive strains in the butyl
strips.
The results of the tables allows for the following conclusions:
o Edge effects are negligible as the differences between 3D and 2D finite

element models are quite small.
o The spacer stiffness is not a key parameter as seen by the comparison between
aluminum and steel (Young’s modulus ratio approx. 3!) and by the comparison
of different thicknesses for steel.
o Nevertheless the trends in the differences are consistent with theory i.e. steel of
same thickness behaves stiffer than aluminum and thickness 0.4 mm behaves
stiffer than 0.3 mm.
o Obviously the butyl strips are the key in the mechanical behavior showing high
flexibility.
4.3. Comparison with Experimental Data

In a next step, the results were compared to the experimental values. It is obvious that
the numerical stiffness results presented in the section before are significantly higher
than the experimental ones. As the butyl strip was identified as key element for the
mechanical behavior a closer look was put on the butyl modeling. It was already
mentioned before that the butyl strips are applied for the test campaign based on 0.9 mm
thick strips. Thus, the tested spacer elements featured higher thickness for the butyl
strips than assumed for the finite element models leading to a lower stiffness of the
assembly. Therefore, calculations were performed with modified models featuring
0.9 mm thickness at each interface. Table 3 presents the results and comparisons with
experimental stiffness results based on displacement levels of 0.09 mm and 0.18 mm
(corresponding to 5% and 10% with respect to an assumed butyl thickness of 0.9 mm).
First, it should be mentioned that the experimental stiffness values are expected to
highly depend on pre-loading of the specimens. The number of test results for
evaluation is given in brackets. For a more quantitative approach it is evident that more
specimens are required in view of uncertain parameters. Nevertheless clear tendencies
231
Challenging Glass 3
are obvious. The experimental stiffness levels are typically bracketed by the numerical
results with larger butyl width on the lower side and with smaller butyl widths on the
upper side.
Table 3: Comparison of 2D stiffness values with experimental results for D-shape spacers.
Spacer design Test 0.25 mm butyl 0.9 mm butyl

Material – height - thickness N/mm2 N/mm2 N/mm2
Aluminum 13.5 mm 2.13 (1) 16.08 1.67
Aluminum 15.5 mm 2.65 (6) 15.97 1.67
Aluminum 17.5 mm 1.17 (1) 15.87 1.67
Steel 13.5 mm 2.86 (2) 16.54 1.68
Steel 15.5 mm 2.41 (2) 16.44 1.68
Steel 17.5 mm 2.73 (2) 16.36 1.68
5. Simplified Models
For global numerical models of insulating glass units e.g. in the context of
comprehensive SSG façade analysis, reduced order models are desirable for
representation of edge seal spacer characteristics. Thus, this section is dedicated to the
challenge to represent the edge seal spacer assembly by lumped stiffness parameters e.g.
by spring elements. Figure 17 sketches this approach in an illustrative manner.
Figure 17 : Simplification of mechanical edge seal spacer model.

For this simplified model, the axis where bending is decoupled from
tensile/compressive loading needs to be identified. The axis is obtained by relating the
constraint moments to the constraint forces. The following tables present the axis offsets
for the D-shape aluminum spacers with respect to the origin on left bottom corner. As it
can be seen the spacer height is of second order with respect to the offset value. This
finding underlines the dominance of the butyl strips as mentioned in the section before.
Table 4: Key figures of mechanical properties of D-shape edge seal spacers.
Spacer design Offset Tensile stiffness Bending stiffness

Material – height - thickness mm2 N/mm2 N/rad
Aluminum 13.5 mm 4.105 16.08 16.77
Aluminum 15.5 mm 4.107 15.97 16.76
Aluminum 17.5 mm 4.108 15.87 16.75
232
Figure 18 presents a comparison of the numerical results obtained with a detailed model
of the edge seal spacer and with a simplified model consisting of a spring element
representing the normal and bending stiffness characteristics. The deformation fields
presented in Figure 18 confirm the mechanical adequacy of the two models.
Figure 18 : Comparison of numerical models – left detailed model, right simplified model featuring a
representative spring element.
6. Conclusions and Outlook

Edge seal spacer designs were subjected to compressive load schemes by both
experiments and numerical studies. The butyl strips were identified to be dominant with
respect to the mechanical characteristics of the edge seal spacer assembly. Due to this
finding, the geometry of the butyl strips and the related deformations of the butyl during
the manufacturing process need more attention in the future for accurate mechanical
models.
Within this paper the butyl strips were exposed to monotonous compressive loading.
For future studies it is recommended to consider cyclic load histories as well allowing
to model the inelastic behavior of the butyl material in more detail. In addition the
behavior of the butyl strips shall be investigated with respect to different load schemes
such as shear or combined load schemes.
Furthermore, the mechanical characteristics of the edge seal spacer assembly were
reduced to a spring element defined by the location of the end grids, the normal and the
bending stiffness. This modeling approach showed adequate behavior and can thus be
applied to global modeling of insulating glass units of SSG designs for linear finite
element analysis.
233
Challenging Glass 3
7. Acknowledgements
For providing samples of the edge seal spacers, the author would like to thank the
contributing glass manufacturers – especially the companies Wagener, Roschmann and
Okalux. Furthermore Test-Ing Material GmbH [3] is acknowledged for performing the
experimental campaign of the edge seal spacers.
8. References
[1] NN; ETAG 002 Guideline for European Technical Approval for Structural Sealant Glazing System
(SSGS) – Part 1 Supported and unsupported systems, www.eota.be/pdf/ssgs-fin-am3.pdf
[2] Hagl, A., Dieterich, O., Glass Unit Corner Loading – Key Parameter in Durability, 4th Symposium on
Durability and Construction Sealants and Adhesives, June 2011, Anaheim, CA
[3] www.test-ing-material.de
234
doi:10.3233/978-1-61499-061-1-235
Tensile Loading of Silicone Point Supports

– Revisited
Anneliese Hagl
Test-Ing+ Material GmbH, Germany, www.test-ing-material.de
Oliver Dieterich
A. Hagl Ingenieurgesellschaft mbH, Germany, www.a-hagl-ingenieure.de
Andreas Wolf, Sigurd Sitte
Dow Corning GmbH, Germany, Andreas.Wolf@dowcorning.com,
s.sitte@dowcorning.com
An advanced high performance silicone film adhesive is ready to be exploited for

challenging glazing applications. This silicone adhesive differs in many aspects
from conventional silicones as it provides transparency and superior mechanical
characteristics. Furthermore, the transparent silicone material shows the tendency
of loosing transparency in highly loaded operating conditions (“stress whitening”).
Due to these specific characteristics, comparisons of the mechanical performance of
the film adhesive with commercially available structural sealants for glass façades
are of major interest. The paper compares the performance of conventional, gun-
grade silicone sealants with that of the novel silicone film adhesive for bonding
point supports on glass by evaluating various load schemes and bond thicknesses.
Keywords: Bonded Point Support, Point Support, Silicone Sealant Glazing (SSG),
Transparent Structural Silicone Adhesive (TSSA)
1. Introduction
During the last decades the performance, application and appearance of structural
silicone glazing (SSG) applications showed continuous but small improvements. Now –
with the launch of a new transparent structural silicone adhesive [1] – an advanced high
performance silicone material is ready to be exploited for challenging glazing
applications. This silicone film adhesive differs in many aspects from conventional
silicone sealants by combining excellent transparency and superior mechanical
characteristics. The adhesive material is based on a heat-curing one-component material
produced in thin sheets ready for application. Thus – compared to conventional silicone
adhesives – the bond thickness is much smaller.
Due to these specific characteristics, comparisons of the mechanical performance with

conventional structural sealants widely introduced into the market for glass façades are
of major interest. The transparent silicone film adhesive shows the tendency of loosing
transparency in highly loaded operating conditions. This unique feature allows tracing
the spread of the loading within the material on a local level up to the point of failure
with increasing loads In addition, this phenomenon of changing transparency is
reversible i.e. the transparency is regained by unloading which can be demonstrated by
adequate cyclic test procedures.
235
Challenging Glass 3
From previous test campaigns on circular point supports bonded with a widely available
two-component conventional silicone adhesive for SSG applications it is known that for
point supports of 50 mm diameter the cohesive failure of the silicone material initiates
in an annular zone that falls between approximately 30% and 60% of the point support
radius. The stress whitening observed with the transparent silicone film adhesive allows
monitoring of this behavior based on the change in transparency.
2. Next Generation of SSG Sealants

Transparent Structural Silicone Adhesive (TSSA)
The TSSA material is manufactured as a film adhesive of 1 mm nominal thickness. The
one-part material is based on a heat-activated addition-cure mechanism that requires
temperatures of 120°C-130°C for a period of 20-30 min. During the curing process a
pressure of typically 0.15-1.3 MPa is applied e.g. in an autoclave as used for the
production of laminated glass. The following characteristics of the cured silicone film
adhesive are presented in [1]:
Table 1: Typical properties of cured silicone film adhesive (extracted from [1]).
Property Test Method Typical Value Unit
Indentation Hardness JIS K 6253 [2] Durometer 70 JIS A
100% Modulus JIS K 6251 [3] (dumbbell #3) 4.0 MPa
Young’s Modulus ISO 527 Parts 1 and 2 [4, 5] 9.3 MPa
Max. Tensile Strength JIS K 6251 (dumbbell #3) 9.0 MPa
Elongation at break JIS K 6251 (dumbbell #3) 250 %
Tear Strength JIS K 6252 [6] (crescent specimen) 35 N/mm
The physical properties of the silicone film adhesive show a low dependency on
temperature. Furthermore, the structural silicone film adhesive combines high
transparency, strong adhesion performance, thermal stability and excellent
weatherability as reported in [1].
The TSSA material shows stress whitening when subjected to significant loads, e.g.,
during dog-bone tests, as can be seen in Figure 1. On the other hand, dog bone testing
does not show any peculiarities with respect to the stress-strain relationship as it
displays the typical elongation and failure behavior observed for elastomers. The stress
whitening phenomenon occurs for loads below the ultimate failure load and it is
typically not related to a special loading scheme. Stress-whitening is well understood
for thermoplastic materials as it is related to micro-cracking (crazing), however, for
elastomers it is less understood and interpreted as related to micro-void formation at the
polymer/filler interface [7]. The whitening phenomena was reported in reference [1] as
starting at loads of around 2.0-2.5 MPa regardless whether the specimen was subjected
to tensile or shear load schemes and independent of the test temperature. Furthermore,
the stress whitening phenomenon is reversible i.e. it disappears and re-appears under
cyclic loading histories.
236
Tensile Loading of Silicone Point Supports – Revisited
Figure 1: Transparent structural silicone adhesive (TSSA) dog-bone tests: Stress-whitening of specimen.
3. Point Support Results for Conventional Structural Silicone Adhesives

The behavior of point supports based on conventional two-component structural
silicone sealants was investigated and presented in detail in several papers, see for
instance references. [8, 9]. In contrast to dog bone tests the stress-strain curves show
three distinct phases in case of continuously increasing loads, as shown in Figure 2: The
initial phase features high stiffness and fully functional adhesive, the second phase
shows a significantly lower stiffness obviously related to some kind of micro-damage in
the adhesive material and the third and final – phase is related to total rupture of the
specimen. Furthermore, the experimental results lead to the hypothesis suggesting that
macroscopic breakage obviously starts in an annular range of 30%-60% of the total
radius for point supports of 50 mm diameter and 5 mm bonding thickness as indicated
in Figure 3.
Phase 1 Phase 3
Beginning Finally the
of macro core fails
cracks at
an inner
circle
Phase 2
Crack-progress
to inside and
outside
Figure 2: Tensile test characteristics of d=50 mm point support Figure 3: Point support failure hypothesis
bonded with conventional silicone sealant (thickness 5 mm). of conventional silicone sealant.
4. Experimental Results for TSSA Bonded Point Supports

Primary application fields identified for the market penetration of TSSA are point
supports in their multiple design varieties. Initial experimental and analytical activities
are related to circular point supports of 20 mm, 50 mm and 80 mm diameter in order to
cover a large range of potential applications. As TSSA is provided as a film of 1 mm
thickness, the baseline bond thickness of the point support specimens was selected as
one layer of TSSA. Nevertheless – in order to study the effect of bonding thickness
variations –samples were laminated also for 3 mm thickness corresponding to three
layers of TSSA.
237
Challenging Glass 3
The experimental campaign comprised tensile, shear and eccentric shear loading
schemes which introduces additional bending moments, see Figure 4. Special attempts
were made in the test set-up in order to keep the adhesive visible for monitoring its
transparency during the tests in order to allow identification of the whitening
phenomenon by synchronizing the video stream from a web cam to the load versus
deflection curve. The lever arms for the eccentric shear tests were varied between
29 mm and 49 mm length for the point supports of 50 mm diameter and between 26 mm
and 46 mm length for 20 mm diameter. Although additional tests were performed with
respect to durability and cyclic loading histories, this paper reports test results without
considering water immersion and cyclic loading, since the results were not significantly
be influenced by aging.
Figure 4: Experimental set-up for tension, shear and eccentric shear tests.
Nevertheless, the whitening point did not appear in all tests performed. Obviously, rapid
crack growth masked the appearance of the whitening point for two of the investigated
test configurations: shear applied to 50 mm and to 80 mm diameter point supports.
238
Point supports of large diameter exposed to shear only show a behavior where the
whitening point could not be observed. Nevertheless, the terminus “shear only” is not
totally appropriate as the load is introduced at the metal section of the point support
with an offset of at least half of the adhesive thickness with respect to the adhesive
center plane. Interestingly, the offset eccentric introduction of bending tensile stresses
allows a clear observation of the whitening phenomenon. Furthermore, it should be
noted that the appearance of the whitening phenomenon within the test campaign is not
linked to water immersion or cyclic load histories. The ratio of the load level related to
the occurrence of the whitening phenomenon as compared to the ultimate failure load is
compiled in Table 2 for the baseline tests (without water immersion and load cycles).
The number of specimens also is recorded so as to allow judgment of statistical
robustness.
Table 2: Comparison of whitening phenomena and ultimate load levels.

Whitening Number of Load
Test set-up: Tensile tests
point specimens ratio %
20 mm diameter, 1 mm thickness observable 10 38
Test set-up: Shear tests
50 mm diameter, 1 mm thickness not observable 10 n/a
80 mm diameter, 1 mm thickness not observable 4 n/a
Test set-up: Eccentric shear tests
26 mm offset, d=20 mm, t=1 mm observable 5 24

As next step, the load curves of the tensile tests are investigated. When compared to
conventional silicone bonding applications, the higher stiffness of the TSSA material in
combination with the lower bond thickness results in a quite stiff bond. The
compliances in the test set-up, e.g., in the attachments to the testing machine, are no
longer negligible. In order to account for this behavior, video-extensometry was applied
especially for the tensile tests. Nevertheless the resolution of the video-extensometer
also posed constraints for very stiff bond behavior and small displacements. In order to
allow comparison of the different bond geometries the loading is plotted in Figure 5 as
239
Challenging Glass 3
stresses versus strains for representative specimens of 20 mm, 50 mm and 80 mm

diameter and bond thickness values of 1 mm and 3 mm.
Ftensile ' tensile
V (1) H (2)
ATSSA t TSSA
Although the ratio of bond thickness to diameter covers a large range from 1:20 to 1:80,
the similarity of the curves is quite remarkable, which means that the impact of the free
(unconstrained) surface of the TSSA that allows lateral contraction appears to be small.
4
Stress - MPa
20-Z
3
50-Z-1
2 50-Z-3
80-Z-1
1
80-Z-3
Strain
0
0 0.2 0.4 0.6 0.8 1
Figure 5: Representative tensile test characteristics of 20, 50, and 80 mm diameter point support of TSSA
with thicknes of 1 and 3 mm. (Z: Tensile test)
Similarly, the shear test results were post-processed for specimens of 20 mm diameter,
50 mm diameter and 80 mm diameter – all specimens featuring 1 mm bond thickness.
Fshear ' shear
W (3) J (4)
ATSSA t TSSA
According to Figure 6, the shear behavior of point support configurations of 50 mm and
80 mm diameter show a high level of similarity, while deviations are visible for the
20 mm configuration as already explained before, the shear load is introduced at least
with an offset of the bond thickness, which is nominally the same for all specimens. On
the other hand, the resistance against bending due to this load introduction offset is
smallest for the 20 mm configuration. The extraordinary behavior of the 20 mm shear
test configuration was already highlighted in view of the appearance of the whitening
phenomenon, which occurred for 20 mm point supports but not for the other
configurations.
For the eccentric shear tests, both shear stresses and maximum bending stresses are
plotted versus the related strain measures in Figure 7. While the calculation of the shear
stresses and strains is quite straight forward assuming simple shear relationships,
maximum bending stresses obtainable using classical beam bending theory, i.e.,
assuming the interfaces to glass and steel remain planar but are rotated, lead to a linear
strain distribution which can be derived by kinematic analysis assuming rigid glass and
point support.
240
Stress MPa
4
3
20-S
2 50-S
1 80-S
Strain
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Figure 6: Shear test characteristics of 50 mm diameter point support bonded with TSSA.
Assuming linear stress-strain relationship which is an approximation for the

hyperelastic behavior, the maximum (engineering) bending stress can be related to the
bending moment by equation (5): As expected, bending stresses increase with
increasing eccentricity, while shear stresses decrease accordingly. Such kind of results
might allow assessing the applicability of mixed ultimate loads, governed by laws such
as shown in equations (6) and (7).
2 2 n n
M ecc F l ecc § V · § W · § V · § W ·
V (5) ¨¨ ¸¸ ¨¨ ¸¸ 1 (6) ¨¨ ¸¸ ¨¨ ¸¸ 1 (7)
WTSSA WTSSA © V max ¹ © W max ¹ © V max ¹ © W max ¹
15.0
Bending Stress max - MPa
12.5
20mm - 26mm offset
10.0 20mm - 36mm offset
20mm - 46mm offset
7.5
50mm - 29mm offset

50mm - 49mm offset
2.5
Bending Strain max
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
3.0
2.5
Shear Stress - MPa
20mm - 26mm offset

20mm - 46mm offset
1.5
50mm - 29mm offset
50mm - 49mm offset
0.5
Shear Strain
0.0
0 1 2 3 4
Figure 7: Eccentric shear test characteristics of 50 mm diameter point support bonded with TSSA.
In order to check the applicability of equation (6) and (7) for the point supports, the
ultimate load levels in tension and shear are plotted in Figure 8. As the eccentric shear
test results – with the test specimens experiencing both tensile and shear loads – do not
fall within the set of results obtained in shear and tensile tests, it is obvious that
241
Challenging Glass 3
equations (6) and (7) do not apply. The ultimate bending stresses plotted in Figure 7 for
eccentric shear lead to the assumption that the limits are dominated by the bending
moment stresses and thus by the bending moments of the point supports –
approximately independent from the eccentricity itself.
10.0
Shear Stress - MPa
shear
7.5 tension t=1
shear
tension t=3
5.0
eccentric eccentr. 1
shear eccentr. 2
2.5
eccentr. 3
tension Tensile Stress - MPa
0.0
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5
Figure 8: Ultimate load conditions of tensile, shear and eccentric shear tests
5. Numerical Analysis of TSSA Point Supports

As a first step, the hyperelastic material laws are derived from material tests of TSSA
such as dog-bone tests (uni-axial loading schemes), equi-biaxial loading schemes and
pure shear loading schemes. Hyperelastic material laws based on elastic energy
principles adequately describe elastic non-linearities regarding stress-strain relationships
but neglect plasticity and viscous flow phenomena which are related to load histories
and load rates. In case of the monotonous loading of the test specimens, softening
phenomena typically known as Mullins effect which are usually covered by a damage
parameter need not be taken into account. The following results demonstrate that the
behavior of TSSA can be adequately described by the Neo-Hooke material law for
tensile strains up to 250% elongation in dog bone tests: The Neo-Hooke law relates the
strain energy density W to the principal stretch ratios i via the shear modulus G. For
higher strains the extension of the Mooney-Rivlin material law by considering
additional strain energy terms will easily improve the correlation but is considered not
to be required within this research.
G 2
W O1 O22 O32 3
2 (8)
10
8
Eng. stress - MPa
L006 10mm/min
6
L007 10mm/min
L008 10mm/min
4
analytical
2
Eng. Strain
0
0 0.5 1 1.5 2 2.5 3
Figure 9: Comparison of dog bone experimental results with Neo-Hooke material law.
242
The various test configurations presented above were analyzed by means of FEA. Glass
and metallic sections of the point supports are assumed to be very stiff compared to the
silicone material and thus treated as rigid. For both the whitening load levels (if
identified) and the ultimate load levels, stress and strain distributions are of interest for
comparison purposes. Nevertheless it has to be noted that the behavior in region 2 of the
tensile test (and also region 3) is not covered by hyperelastic material laws. While the
analysis of the stresses and strains related to the ultimate loads might give indications of
the strength limits of the adhesive material, a comparison of the experimentally obtained
whitening patterns with the calculated stress and strain distributions shall provide
insight in the whitening phenomenon from a mechanical point of view.
For the FE analysis of tensile tests, axisymmetry can be assumed for the solution and
thus 2D axisymmetric FE models or – if not available – a small wedge of a 3D model
with adequate boundary conditions taking into account axisymmetry are adequate for
analysis. Figure 10 presents the tensile stress distributions in half cross sections of the
50 mm diameter point supports with thicknesses 1 mm and 3 mm for load levels
corresponding to the appearance of the whitening. Due to the axisymmetry the stress
states depend only on radial positions and the coordinate in thickness direction. The
following observations can be made: First, the smaller the bond thickness, the more
uniform is the stress distribution. This phenomenon can be expected, as then the free
surface is smaller compared to the diameter. Thus, the material has less possibility to
laterally contract. Second, the stress levels obtained are higher for the specimens with
3 mm than 1 mm bond thickness.
2.562 2.065 1.567 1.07 0.573 0.0754

3.389 2.771 2.153 1.535 0.917 0.299
t = 3 mm
3
sig zz - MPa
2 t = 1 mm
r - mm
0
0 5 10 15 20 25
Figure 10: Tensile stress distribution of 50 mm point support for tensile loading according to whitening point
(symmetry=center plane)
243
Challenging Glass 3
Figure 11 shows the beginning of the whitening phenomenon for both 1 mm and 3 mm
bond thickness. It is interesting to note that for the 1 mm specimens the whitening
appears at approximately 60% radius while for the 3 mm specimen the whitening seems
to start in the center region. This observation corresponds in some degree to the
numerical results as for the 3 mm configuration the center stresses are indeed higher.
Nevertheless, no direct indication from the stress analysis can be found why for 1 mm
bond thickness, the whitening starts at approximately 60% radius.
t = 1 mm t = 3 mm
Figure 11: Tensile Test Characteristics of 50 mm Diameter Point Support of TSSA: Whitening.
244
Regarding shear loading, the numerical model has to be modified to a half model as in
this case, only symmetry aligned to the load direction can be exploited. Please note that
the rotation of the point support is not constrained and that the load introduction point –
as already mentioned – is offset to the glass surface. Figure 12 presents the shear stress
distribution with peak stress amplitudes at the edge region close to the symmetry plane.
This concentration of high stress levels in small regions near the edge might explain
why for these samples whitening could not be observed easily during testing. Therefore,
the numerical model was loaded by the ultimate load level. Please note that different to
what was observed in the tensile tests, no softening of the specimen behavior is noted
for high loading according to Figure 6. Thus, it can be assumed that the hyperelastic
material law of the numerical model is valid up to the ultimate loads.
-1.713
-2.481
-3.248
-4.016
-4.783
-5.551
-6.318
-7.086
-7.853
-8.621
-9.388
-10.16
-3
Top Surface
Bottom Surface
tau xz - MPa
-5
-7
x - mm
-9
-25 -20 -15 -10 -5 0 5 10 15 20 25
Figure 12: Shear stress distribution, 50 mm point support for ultimate shear loading (top and bottom surface)
The eccentric shear tests are – from a numerical point of view – similarly treated as the
shear cases with the distinction that the load introduction point is offset. The stress
distributions in Figure 13 show that the loading related to the appearance of the
whitening phenomenon is mainly linked to the bending stresses and not to the shear
stresses according to the almost constant levels of bending stresses. Furthermore, the
shear stresses are significantly smaller compared to the maximum bending stresses.
245
Challenging Glass 3
6.1
5.182
4.264
3.345
2.427
1.509
0.591
-0.327
-1.245
-2.164
-3.082
Shear, Eccentricity 1 Shear, Eccentricity 2 Shear, Eccentricity 3
29 mm 39 mm 49 mm -4.0
5
29 mm offset
39 mm offset
2.5
49 mm offset
sig z - MPa
x - mm
0
-25 -20 -15 -10 -5 0 5 10 15 20 25
-2.5
-5
0.5
x - mm
29 mm bottom
0
29 mm top
tau xz - MPa
-25 -20 -15 -10 -5 0 5 10 15 20 25

-0.5 39 mm bottom
-1 39 mm top
49 mm bottom
-1.5
49 mm top
-2
Figure 13: Tensile and shear stress distributions of 50 mm point supports for eccentric shear loading (top and
bottom surface)
Figure 14 presents the advancement of the whitening during uploading with the
sequence of loading being from left to right. The figure on the left shows the first
occurrence of the whitening phenomenon for a still fully functional adhesive.
Interestingly, the phenomenon is obviously linked to tensile stresses but not to

compressive stresses. In the subsequent figure, the whitening has advanced substantially
across the adhesive. The two figures on the right show three distinct zones with the
ruptured adhesive on top, the whitening zone in the middle, and the still intact
transparent adhesive on the bottom. With increasing load levels, the borderlines
between the zones will move downwards as shown by the figure on the extreme right.
Whitening Rupture
Figure 14: Eccentric shear test (49mm offset) for 50 mm point support of TSSA. Whitening
246
6. Comparison of Conventional and Next Generation Silicone Adhesives

Due to differences in the bonding process, applications of the conventional and the
transparent silicones significantly differ in bond thickness. Representative thickness
values for point supports of conventional silicones amount to approximately 5 mm
while the transparent silicone is typically applied with 1 mm thickness. Nevertheless,
similarities are obtained especially in view of the behavior under tensile loading
schemes. Both silicone adhesives show three distinct zones: The first zone is linked to
high stiffness related to fully functional adhesives. Afterwards, the adhesive reacts
significantly softer, although no cracks are visible on a macroscopic level. Finally, the
adhesive suddenly fails. Another interesting detail is the comparison of the results
obtained for the TSSA film adhesive with the failure hypothesis developed for the
conventional silicone: An initiation of the cohesive failure was assumed at
approximately 60% of the total radius of the point support. This hypothesis is obviously
coherent with the appearance of the whitening phenomenon of the transparent silicone
at 1 mm thickness. Further studies are required in order to explain the dependency on
bond thickness.
7. Conclusions
For circular point supports of varying diameter, tests for a new transparent structural
silicone adhesive were performed for tensile, shear and eccentric shear loading schemes.
Special focus was directed at the whitening phenomenon indicating highly strained
regions. The test showed several similarities to conventional silicone adhesives.
Furthermore, bending moments were identified as critical load parameters for mixed
loading of point supports. The next steps will consist of a deeper analysis of the
appearance of the whitening phenomenon with respect to stress-strain curves. For
conventional silicone sealants, a decrease of the tangent stiffness is clearly visible for
tensile loading. This leads to the question whether this effect is also visible for the
transparent silicone and in case of a positive answer whether the stiffness decrease
coincides with the whitening phenomenon.
8. References
[1] Sitte, S., Brasseur, M.J., Carbary, L.D., and Wolf, A.T., Preliminary Evaluation of the Mechanical
Properties and Durability of Transparent Structural Silicone Adhesive (TSSA) for Point Fixing in
Glazing, Journal of ASTM International, Vol. 8, No. 10, 2011, Paper ID JAI104084.
[2] JIS Standard K 6253, 2006, Rubber Vulcanized or Thermoplastic – Determination of Hardness,
Japanese Standards Association, Akasaka Minato-ku, Tokyo, Japan.
[3] JIS Standard K 6251, 2004, Rubber Vulcanized or Thermoplastics – Determination of Tensile Stress-
Strain Properties, Japanese Standards Association, Aksasaka Minato-ku, Tokyo, Japan.
[4] ISO Standard 527-1, 1993, Plastics – Determination of Tensile Properties – Part 1: General Principles,
International Standardization Organization (ISO), Geneva.
[5] ISO Standard 527-2, 1993, Plastics – Determination of Tensile Properties – Part 2: Test Conditions for
Moulding and Extrusion Plastics, International Standardization Organization (ISO), Geneva.
[6] JIS Standard K 6252, 2007, Rubber Vulcanized or Thermoplastics – Determination of Tear Strength,
Japanese Standards Association, Aksasaka Minato-ku, Tokyo, Japan..
[7] Nathani, H., Dasari, A. and Misra, R.D.K., “On the Reduced Susceptibility to Stress Whitening Behavior
of Melt Intercalated Polybutene – Clay Nanocomposites during Tensile Straining”, Acta Mater., Vol.52,
2004, pp. 3217-3227.
[8] Hagl, A., “Beyond ETAG 002: U-type Bonding Geometries for Structural Glazing”, Steel Construction,
Vol.1, Issue 1/2009.
[9] Hagl, A., “Bonded Point-Supports: Understanding Today – Optimizing for the Future”, Challenging
Glass 2, Delft, 2010
247
doi:10.3233/978-1-61499-061-1-249
Investigation of Stress-Whitening in
Transparent Structural Silicone Adhesive
Anneliese Hagl
Test-Ing+ Material GmbH, Germany, www.test-ing-material.de
Andreas Wolf, Sigurd Sitte
Dow Corning GmbH, Germany, andreas.wolf@dowcorning.com,
s.sitte@dowcorning.com
The stress whitening behavior of a transparent structural silicone adhesive (TSSA)

under tensile and shear loading is investigated in this paper. The TSSA material has
been developed for point-fixing in glazing and combines high transparency, strong
adhesion performance, thermal stability and excellent weatherability. The
transparent film adhesive is a heat curing one-part material that shows strong
bonding to glass, metals, ceramics and even plastics typically without primer.
Keywords: Silicone, Adhesive, TSSA, Point Support, Bonding
1. Introduction
Bonded point fixed supports have recently received increased attention, as in contrast to
mechanical point supports they offer a number of advantages, such as lower visibility
from the exterior, a ‘smooth’ transfer of the load into the glass pane (avoiding stress
peaks), and the elimination of drilling holes from the glass [1-4]. Structural silicone
sealants have been used in linear adhesive fixing of glazing elements at a tertiary
structural level since the 1960s [5]. The long-term experience with silicone sealants in
this field has led to standardization of both the performance requirements on the
structural adhesive sealants as well as the glazing designs [6,7]. The room-temperature-
vulcanizing (RTV) structural silicone sealants used in linear fixing of glazing elements
(structural silicone glazing) display a low Young’s modulus, generally in the range of
about 1.0-2.5 MPa in tension, and a high elongation at break, generally in the range of
>100% when measured in tension on a tensile-adhesion joint with dimensions as
defined in ISO 8339 [8]. The resulting joint design with a bond thickness of minimum 6
mm allows compensation of thermally induced movements and dimensional tolerances
between the substrates, which is a necessity for linear structural bonded bearings.
However, for adhesively bonded point-fixed bearings, a higher Young’s modulus is
desirable to achieve higher stiffness with a smaller bonding area that still allows
carrying significant out-of-plane loads [4].
This paper presents experimental data obtained on a Transparent Structural Silicone

Adhesive (TSSA), a one-component, heat-curing silicone film adhesive showing high
transparency, superior mechanical performance, thermal stability and excellent
weatherability. TSSA is provided as pre-manufactured sheets with thicknesses of
typically 1 mm. Due to its appealing mechanical properties, point supports for glass
249
Challenging Glass 3
facades are identified as primary target application for the market introduction of this
next generation silicone adhesive material. The novel material has been described in
more detail in previous papers [9, 10].
The test campaign presented in this paper covers tensile and shear tests for point
supports of diameters 20 mm, 50 mm and 80 mm. In total, more than 100 test runs were
performed using representative circular point support specimens. Special attention was
given to the identification of the load level which leads to a significant transparency
change (stress whitening) of the adhesive material.
2. Experimental Procedure
2.1. Test Equipment

The tensile and shear tests of the point support specimens up to 10 kN maximum load
were performed on an INSTRON series 5566A dual column load frame. The load frame
was equipped with a 10 kN load cell and an advanced video extensometer (AVE) using
a lens of 55 mm for a field of view of 200 mm. Tests involving loads above 10 kN were
performed at the Institute for Lightweight Structures of the University of Munich.
2.2. Test Specimens

Test specimens, i.e., circular point support assemblies of diameters 20 mm, 50 mm and
80 mm were prepared by bonding stainless steel buttons with the TSSA (1 mm film
thickness prior to compression) to standard (uncoated) float glass coupons in a typical
autoclave process used for the production of laminated glass. The following procedure
was used for the manufacture of the point-fixing specimens:
1. Cleaning of glass and steel surface using Dow Corning® R40 Universal
Cleaner;
2. Application of Dow Corning® 92-023 Primer on both glass and steel surfaces;
3. Removal of polyester film cover on one side of the structural silicone film
adhesive and placing of the steel button on the film adhesive;
4. Removal of excess film adhesive around the button by cutting;
5. Removal of second polyester cover from the adhesive and placement of steel
button with the film adhesive face on the glass substrate;
6. Use of manual load equipment, placing pressure of approximately 0.7 MPa on
the button for a short period of time (10 s);
7. Placement of glass vertically with attached buttons in standard autoclave
process for laminated glass with the lamination process run at 13 bars and
140 °C for 4 hours;
8. Cutting of glass by water jet to generate individual test specimens.
The above process conditions were chosen primarily based on their suitability to fit into
a standard glass lamination process. The point supports made of stainless steel were
bonded to glass pieces typically based on a nominal adhesive thickness of 1 mm. In
addition, some specimens showed an adhesive thickness of 3 mm allowing the
evaluation of film thickness effects (in this case, three layers of TSSA film were
laminated together).
250
Investigation of Stress-Whitening in Transparent Structural Silicone Adhesive
2.3. Test Procedures

A key objective of this test campaign was the identification of the load levels at which a
change in the transparency of the adhesive material occurs (stress whitening). Due to the
peculiarities of the silicone film adhesive, test set-up and test procedures sometimes
differed significantly from conventional tests used for gun-grade RTV silicone sealants.
Thus, special fittings and attachments had to be developed and applied in order to allow
recording of the behavior of the silicone film adhesive through the glass of the specimen
with the help of a web cam. After fixing the specimen in the testing machine and
launching the test sequence the web cam was activated. Synchronization between the
test result stream and the web cam recording was done later via the identification of the
failure event recorded in both media. Figure 1 presents a sketch of the tensile test set-up
including a bracket which was designed to be applicable for all point support diameters
and the associated load levels and ensuring optical access to the glass side of the test
specimen from above.
Figure 1: Tensile test set-up for monitoring optical characteristics of silicone adhesive (schematic).
Figure 2 presents the bracket tailored to fit the tensile test specimens of various
diameters. Due to the expected high loads for the 80 mm point supports, the design was
substantiated with respect to strength issues in order to ensure the required load bearing
capabilities. The web cam (not shown here) was fixed to the bracket from its top side in
order to ensure a constant perspective during the test run.
Figure 3 displays a sketch of the shear test set-up. The glass was put in place vertically
by an L-shaped support structure allowing optical access.
Figure 2: Tensile test set-up for monitoring optical characteristics of silicone adhesive (photo).
251
Challenging Glass 3
Figure 3: Shear test set-up for monitoring optical characteristics of silicone adhesive (schematic).
A photograph of the shear test set-up is shown in Figure 4. The shear load was applied
by a fork-like device moved downwards during the test run. The curvature of this fork-
like pushing device was selected in order to be applicable to all investigated diameters
of point supports. In order to minimize the (unwanted) moment due to the out-of-plane
offset of the load introduction area, the test set-up is adjusted in such a way as to
minimize the distance between fork and glass.
Figure 4: Shear test set-up for monitoring optical characteristics of silicone adhesive (photo).
Tests were performed at room temperatures (approx. 20°C - 25°C) and under typical in-
door humidity conditions (approx. 40% r.h.). Use of the video extensometer required
the provision of two markers on the specimen. One marker was applied on the point
support itself, the other marker was applied on a screw attached to the glass pane as for
geometric reasons it was not possible to directly provide markers at the adhesive
boundaries.
252
2.4. Determination of Onset of Stress Whitening

The onset of the stress whitening was determined in the post-processing of the video
streams recorded by the web cam. Figure 5 shows typical whitening patterns that
occurred for the TSSA material and Figure 6 shows the earliest signs (onset) of the
whitening. This onset of whitening was used to determine the associated load levels.
Figure 5: Typical whitening patterns (shown here for point supports of 80 mm diameter, adhesive thickness
of 1 mm (left) and 3 mm (right).
Figure 6: Local onset of stress whitening.

Figure 7: Determination of whitening point WP and associated load level.
253
Challenging Glass 3
The procedure used to determine the load at the whitening point is shown in Figure 7.
Synchronization between video stream and load data stream was performed with the
help of the failure event. The recording speed for the web cam is usually set at 60
frames per second; only for cyclic load schemes this rate was significantly reduced due
to the long duration of the test run.
3. Test Results
3.1. Tensile Tests

The test series can be partitioned into baseline tests, i.e., tests with monotonous loading
and without water immersion; tests with specimens exposed to water immersion
(monotonous loading); and tests with specimens exposed to cyclic loading (without
water immersion). The load levels and nominal stresses (referring to the nominal
diameter of the point support specimens) at which stress whitening was observed are
listed below for the various tensile test series.
Table 1 presents the results of the tensile tests of point supports of 20 mm diameter and
1 mm adhesive thickness under baseline conditions, i.e., room temperature, monotonous
loading and no water immersion. The test series covers ten specimens labeled 20-Z-01
to 20-Z-10.
Table 1: Whitening point for baseline tensile tests with 20 mm diameter and adhesive thickness of 1 mm.
TSSA Thickness [mm] Test No. Load [N] Stress [MPa]
1 20-Z-01 657.6 2.09

1 20-Z-02 698.2 2.22
1 20-Z-03 618.0 1.97
1 20-Z-04 606.3 1.93
1 20-Z-05 644.3 2.05

1 20-Z-06 701.1 2.23
1 20-Z-07 489.7 1.56

1 20-Z-08 719.0 2.29
1 20-Z-09 717.9 2.29
1 20-Z-10 669.6 2.13

Max WP 719.0 2.29
Min WP 489.7 1.56

Mean Value WP 652.2 2.08
254
Table 2 displays the results of the tensile tests of point supports of 20 mm diameter
subjected to room temperature, monotonous loading, and water immersion. After water
immersion the specimens were conditioned at room temperature and humidity for 24
hours. The test series covers five specimens labeled 20-Z-W-01 to 20-Z-W-05.
Table 3 shows the results of the tensile tests of point supports of 50 mm diameter and 1
mm adhesive thickness under baseline conditions, i.e., room temperature, monotonous
loading and no water immersion. The test series covers eight specimens labeled 50-Z-01
to 50-Z-05 and 50-Z-10 to 50-Z-12. In addition, Table 4 presents the results of a test
series that was performed under baseline conditions for four specimens of 3 mm
adhesive thickness (labeled 50-Z-06 to 50-Z-09).
Table 2: Whitening point (in tensile) for 20 mm diameter after water immersion, adhesive thickness of 1 mm.
1 20-Z-W-01 756,2 2,41

1 20-Z-W-02 770,2 2,45
1 20-Z-W-03 796,2 2,54
1 20-Z-W-04 459,8 1,46
1 20-Z-W-05 622,5 1,98

Max WP 796,2 2,54
Min WP 459,8 1,46
Mean Value WP 681,0 2,17
Table 3: Whitening point (in tensile) for baseline tests of 50 mm diameter, adhesive thickness 1 mm.

1 50-Z-01 4,174.50 2.13
50-Z-02 3,089.97 1.57

1 50-Z-03 4,549.42 2.32

1 50-Z-04 4,422.90 2.25
1 50-Z-05 4,049.45 2.06
1 50-Z-10 3,934.89 2.01
1 50-Z-11 3,555.89 1.81
1 50-Z-12 3,837.50 1.96

Max WP 4,549.42 2.32
Min WP 3,089.97 1.57
Mean Value WP 3,951.82 2.07
255
Challenging Glass 3
Table 4: Whitening point in (tensile) for baseline tests of 50 mm diameter, adhesive thickness of 3 mm.
3 50-Z-06 3,387.35 1.73
3 50-Z-07 4,037.28 2.06

3 50-Z-08 4,158.00 2.12
3 50-Z-09 3,868.45 1.97

Max WP 4,158.00 2.12
Min WP 3,387.35 1.73

Mean Value WP 3,862.77 1.97
Table 5 presents the results of the tensile tests of point supports of 50 mm diameter
under room temperature, monotonous loading and with water immersion. After water
immersion the specimens were conditioned in a room climate for 24 hours. The test
series covers five specimens labeled 50-Z-W-01 to 50-Z-W-05.
Table 5: Whitening point (in tensile) for 50 mm diameter after water immersion, adhesive thickness of 1 mm.

1 50-Z-W-01 3165,7 1,61
1 50-Z-W-02 4740,7 2,42

1 50-Z-W-03 3800,1 1,94
1 50-Z-W-04 4852,8 2,47
1 50-Z-W-05 5011,5 2,55
Max WP 5.011,46 2,55

Min WP 3.165,74 1,61
Mean Value WP 4.314,15 2,20
Table 6 displays the results of the cyclic loading scheme defined by 200 N and 2000 N
as lower and upper boundary, respectively, based on three specimens.
Table 6: Whitening point (in tensile) for 50 mm diameter after cyclic loading, adhesive thickness of 1 mm.

1 20-Z-C2-01 4011,48 2,04
1 20-Z-C2-02 4525,42 2,31
1 20-Z-C2-03 4632,00 2,36
Max WP 4.632,00 2,36
Min WP 4.011,48 2,04

256
Table 7 shows the results of the tensile tests of point supports of 80 mm diameter and 1
mm adhesive thickness under baseline conditions, i.e., room temperature, monotonous
loading and no water immersion. The test series covers four specimens.
Table 7: Whitening point (in tensile) for baseline tests of 80 mm diameter, adhesive thickness of 1 mm.

1 80-Z-01 6.452,93 1,28
1 80-Z-02 5.773,08 1,15

1 80-Z-03 5.181,47 1,03
1 80-Z-04 3.507,44 0,70

Max WP 6.452,93 1,28
Min WP 3.507,44 0,70

3.2. Shear Tests

Table 8 shows the results of the shear tests of point supports of 20 mm diameter point
supports subjected to room temperature, monotonous loading and water immersion.
After water immersion the specimens were conditioned at room climate for 24 hours.
The test series covers five specimens labeled 20-S-W-01 to 20-S-W-05.
Table 8: Whitening point (in shear) for 20 mm diameter and adhesive thickness of 1 mm
(after water immersion).
TSSA Thickness Test No. Load [N] Shear Stress [MPa]

[mm]
1 20-S-W-01 541,3 1,72
1 20-S-W-02 507,0 1,62

1 20-S-W-03 252,6 0,81

1 20-S-W-04 407,6 1,30
1 20-S-W-05 651,2 2,07

Max WP 651,2 2,07
Min WP 252,6 0,81

Mean Value WP 471,9 1,50
In contrast to 20 mm point supports, no whitening point could be detected for point

supports of 50 mm and 80 mm diameters in shear. Nevertheless the ultimate shear
stresses observed for 50 mm and 80 mm point supports were of the same magnitude.
257
Challenging Glass 3
4. Conclusions
The onset of stress whitening in the transparent structural silicone adhesive (TSSA)
occurs relatively consistently at around 2 MPa when subjected to tensile loading. Under
shear loading, stress whitening was only observed for point supports with 20 mm
diameter. For bonded point supports with larger diameter no stress whitening was
observed in shear prior to failure. This raises the question whether this behavior depends
on the ratio of the diameter of the point support to the thickness of the TSSA layer.
Further tests in shear loading with increased adhesive thickness are planned for point
supports of 50 mm diameter and larger. Water immersion has a negligible on stress
whitening.
5. References
[1] Tasche, S., Strahlungshärtende Acrylate im Konstruktiven Glasbau, Ph.D. Thesis, Technische
Universität, Dresden, Germany, 2007.
[2] Weller, B. and Tasche, S., “Experimental Evaluation of Ultraviolet and Visible Light Curing Acrylates
for Use in Glass Structures”, Durability of Building and Construction Sealants and Adhesives, 3rd
Volume, A.T. Wolf, Ed., ASTM International , West Conshohocken, PA, USA, 2010, pp. 135-156.
[3] Hagl, A., “Bonded Point-Supports: Understanding Today – Optimizing for the Future”, Challenging
Glass 2, Conference on Architectural and Structural Applications of Glass, F. Bos and C. Louter, Eds.,
University of Technology, Faculty of Architecture, Delft, The Netherlands, 2010..
[4] Hagl, A., “Silicone Bonded Point Supports – Behaviour under Cyclic Loading”, Engineered
Transparency – International Conference at Glasstec, Düsseldorf, Germany, J. Schneider and B. Weller,
Eds., Technical University of Dresden, Dresden, Germany, 2010, pp. 139-148.
[5] Parise, C.J., Science and Technology of Glazing Systems, STP1054, ASTM International , West
Conshohocken, PA, USA, 1989.
[6] ASTM Standard C1401-09a, 2009, “Standard Guide for Structural Sealant Glazing”, Annual Book of
ASTM Standards, ASTM International, West Conshohocken, PA, USA.
[7] EOTA Recommendation, ETAG 002 Structural sealant glazing systems, Part 1: Supported and
unsupported systems, 1999; Part 2: Coated aluminum systems, 2002; Part 3: Thermal breaks, 2003,
European Organization for Technical Approvals (EOTA), Brussels, Belgium.
[8] ISO Standard 8339, 2005, Building construction - Sealants - Determination of tensile properties
(Extension to break), International Standardization Organization (ISO), Geneva.
[9] Sitte, S., Brasseur, M.J., Carbary, L.D., and Wolf, A.T., “Preliminary Evaluation of the Mechanical
Properties and Durability of Transparent Structural Silicone Adhesive (TSSA) for Point Fixing in
Glazing”, Journal of ASTM International, Vol. 8, No. 10, Paper ID JAI104084.
[10] Hagl, A., Dieterich, O., Wolf, A.T., and Sitte, S., “Tensile Loading of Silicone Point Supports –
Revisited”, Challenging Glass 3 – Conference on Architectural and Structural Applications of Glass,

Bos, Louter, Nijsse, Veer (Eds.), TU Delft, June 2012.
258
doi:10.3233/978-1-61499-061-1-259
Designing a Glass Bearing Connection with

a Probability to EN1990 CC2
Ron Kruijs
Glasimpex Schiedam, Holland, r.kruijs@glasimpex.nl
Member of TC129WG8, TC250WG3 and the Dutch construction glass workgroup
Deutscher Ausschuss für Stahlbau (DAst) [1] published a report about steel-glass
connections, Mascha Batinger [2] did her thesis about this subject. In both
publications an analytic model for bearing hole connection is presented. To verify
this analytic models experiments have been performed. The ultimate tensile
strength at the holes in those experiments are usually (almost)lower than the
residual stress from the tempering process. Analysing the design of the bearing hole
connection of both publications several problems were highlighted. In contrast to
the publication the load bearing capacity in this paper is designed from a stress
point of view. The acceptable level of stress is related to probability of EN1990.
Keywords: Glass, bearing hole, EN1990
1. General
There are no standards are available for the design of a glass connection. The design has
to be based on scientific publications and experiments. The subject of bearing hole
connection has been investigated by several authors. This connection has a number of
different parameters which have a function in the load capacity. The geometric
parameters as well as glass strength (residual part and annealed part) and condition of
the hole are important. In the design of the bearing hole connection in this paper the
geometric parameters and condition of the hole are as perfect as possible. The residual
stress part of the test samples are known. So when the experimental bearing strength is
known we can derive the tensile stress. And make a connection to CC probability.
2. Test results of DASt and Thesis

Table 1 and 2 present the results of the test from DASt and thesis. The failure load of
the test samples in table 1 and 2 can’t be compared. The failure load has a function in
sample size and hole position. In this paper the topic is the failure tensile stress.
2.1. Deutscher Ausschüss für stahlbau (DASt)

DASt 2/2007 table 4.17.1.4 shows results of the performed tests. The geometric
parameters of the hole where diameter 44mm, Hy50/70 ring 5mm and a metal insert
size unknown. The results of the test on axial loading are summarise at table 1. Only the
results with now residual stress from the tempering process are of interest.
259
Challenging Glass 3
Table 1: Results test DASt 2/2007
Toughened glass Failure Load Residual stress hole Failure stress Annealed part
10 mm 40,43 kN 96,00 N/mm2 90,48 N/mm2 5,52 N/mm2

10 mm 49,38 kN 96,00N /mm2 110,51 N/mm2 14,51 N/mm2
10 mm 42,73 kN 85,60 N/mm2 113,25 N/mm2 27,65 N/mm2
2 2
10 mm 50,13 kN 85,60 N/mm 132,84 N/mm 46.88 N/mm2
2.2. Zur bemessung van SL-belasteten anschüssen im konstruktiven glasbau Thesis von
Mascha Baitinger
[2] Table 8.7 shows results of performed tests. The geometric parameters of the hole
were diameter 44mm, Hy70 ring 5mm, 2mm aluminium ring and a bolt M30. The
average results of the test on axial loading are summarise at table 2.
Table 2: Mean results test thesis Mascha Baitinger.

Toughened glass Failure Load Residual stress hole Failure stress Annealed part
10 mm 44,34 kN 96,00 N/mm2 116,36 N/mm2 20,36 N/mm2

10 mm 50,86 kN 96,00 N/mm2 112,98 N/mm2 16,98 N/mm2
10 mm 41,43 kN 85,36 N/mm2 104,39 N/mm2 19,02 N/mm2
2x10 mm 88,59 kN 83,25 N/mm2 99,04 N/mm2 15,76 N/mm2
3x10 mm 141,54 kN 99,84 N/mm2 107,69 N/mm2 7,85 N/mm2
2x15 mm 115,77 kN 87,12 N/mm2 88,02 N/mm2 0,10 N/mm2
2x8 mm 86,02 kN 97,52 N/mm2 120,20 N/mm2 25,88 N/mm2
10 mm 49,03 kN 83,87 N/mm2 109,63 N/mm2 25,86 N/mm2
12 mm 52,81 kN 91,36 N/mm2 118,09 N/mm2 26,73 N/mm2
10 mm 57,33 kN 78,24 N/mm2 73,73 N/mm2 4,51 N/mm2
2.3. Remark
The tests are performed with new glass samples. The annealed part of the strength
should have an average about 60N/mm2. The average value here is much lower. So a
other parameter should be responsible for this phenomenon.
3. Design of the holes

It is known that the strength of annealed glass is covered by the condition of the cracks
at the high stress region. The maximum stress at a bearing hole connection is positioned
on the facet of the hole. For the test performed in this paper we used polish hole . On the
facet of the polish holes no cracks are visible (there are cracks only small ones), in
contrary to the holes used in [2] and assumable in [1].
260
Designing a Glass Bearing Connection with a Probability to EN1990 CC2
Figure 1 Polished hole Figure 2: Normal hole
4. Design of the insert.

Design figure 3 is used in [1] and [2]. The materials used from glass hole to middle are
mortal 5mm, aluminum 2mm and bolt. Design figure 4 is used in this paper. The
materials used are mortal 10mm, and stainless steel insert. Stainless steel insert size
diameter 70mm with tread M16. In this way there is no space left in the bearing
connection between bolt and hole in the pane.
Figure 3 Figure 4
5. Load introduction
The deformation of the mortal insert during load condition is about 0,5mm. So relative
small space between components lead tot asymmetric introduction of load and stress
concentration.
261
Challenging Glass 3
5.1. DASt and Thesis

The load introduction by figure 5 is combined with hole design Figure 3. This load
introduction will be asymmetric for the following reasons:
x Space between the aluminum insert and the bolt.

x Displacement in hole positions in steel plates.
x Small different in hole size.
x Possibility of stress concentration in mortal because of deformation of
aluminum insert.
Figure 5
5.2. This paper

The load introduction figure 6 is combined with figure 4 and 7.
This load introduction will be symmetric for the following reasons:
x The stainless steel disc has a threat M16. No space between bolt and disc.
x The stainless steel disc will not deform, no stress concentration will be
introduced in the mortal.
x Injection bolts are used, so displacements in hole positions or differences in
hole size will be avoided .
262
Figure 6 Figure 7
6. Experiments produced for this paper

Two sets of five samples are produced with a different hole design. The aim is to test
two bolts ad a row. The connection between glass and steel plates is made of four bolts
M16 10.9 and the connection between steel/steel is made of two bolds M16 10.9.
6.1. Bearing hole figure 1

All three plates of the laminated are produced with a hole of 90mm. The connection is
only bearing.
6.2. Combination of bearing and pretension bolt, hole figure 8.

The outer two ply of the laminated plate have a hole of 90mm and the inner plate has an
hole of 25mm. The advantage of this system is that an eccentric in the load introduction
at the hole is almost impossible. There is a low level of eccentricity because it is not
possible to use the injection bold on both site of the connection plates. The bearing
connection and the pretension bold load capacity can’t be activated together. The
pretension part needs too much displacement.
263
Challenging Glass 3
Figure 8
6.3. First tests

The test result of both types were limited by bolt capacity. Both connections failed
about 180KN.
6.4. Second test

For the second test two bolts M16 were added ad the steel/steel connection.
The test result of both types were again limited to bolt capacity. Both connections failed
at about 312KN. Meaning that 2 holes at a row using 12mm toughened glass had a
bearing capacity of more than 156KN.
6.5. Third test only figure 8 hole type

The aim was to test 2 bold at a row with a symmetric load introduction. We did the
following chances. We dismantled one bold and lost the pretension. In this way we had
the bearing connection at one hole and introducing an eccentricity. The level of
eccentricity is not known. The results are summarized in table 3.
Table 3: Results of test for this paper hole type par.6.2

Toughened glass Failure Load Residual stress at hole Failure stress Annealed part
2x12 mm 268 kN 136,80 N/mm2 184,48 N/mm2 47,68 N/mm2

2x12 mm 267 kN 136,80 N/mm2 183,79 N/mm2 46,99 N/mm2
2x12 mm 221 kN 136,80 N/mm2 153,13 N/mm2 16,33 N/mm2
2 2
2x12 mm 233 kN 136,80 N/mm 160,39 N/mm 23,59 N/mm2
2x12 mm 312 kN 136,80 N/mm2 214,76 N/mm2 77,97 N/mm2
264
The calculated failure stress did not take into account the eccentricity of load
introduction.
The first two results had a low eccentricity. The next two had a very high eccentricity.
The last one had no eccentricity, but did not break.
6.6. Conclusion
Compare table 3 with 1 and 2, and you can see that in table 3 the annealed part of the
glass strength is more like it would aspect to be. We know that the results are in no way
exact but it shows that when using a perfect hole and a symmetric load introduction, the
strength of this bearing hole connection is governed by glass strength.
When testing bearing hole connections it is important to know the failure stress and the
residual stress from the tempering process.
The problem of using bearing test results for calculation of bearing capacity is that there
is now correction for the parameters that together form glass strength. These parameters
are:
a) Time-depending of the annealed part

b) Aging of the annealed part
c) The influence of environment on the annealed part
d) The high variation of strength from new annealed glass.
e) The high difference in residual stress from the tempering process between
different manufactures.
So when only bearing results are used in statistics, this can lead to unsafe design.
7. Design to the probability of the EN1990 CC2

Using ultimate stress as a design aid we use the Dutch standard NEN2608 to derive the
permitted ultimate stress. The level of ultimate stress in this standard is related to CC2.
k e u k a u k mod u k sp u fg;k k e u k z u fb;k k sp u fg;k

fmt;u;d
J m;A J m;V
fmt;u;d ultimate stress;

ke factor for edge quality ke = 1;
ka factor for size effect ka = 1;
kmod factor for load duration kmod =0,29 ;
ksp factor for the glass surface profile ksp =1;
fg;k is the characteristic value of the bending strength fg;k = 45 N/mm2;
Jm;A; is the material partial factor Jm;A = 2,0;
kz is the correction factor for hole residual stress kz =0,65 ;
fb;k is the characteristic value of the residual stress of the tempering
proces fg;k = 120 N/mm2 (EN12150);
Jm;V is the material partial factor Jm;A =1,2;
265
Challenging Glass 3
Table 4: Ultimate strength hole.
Residual Residual stress + annealed part ad Ultimate

Standard/ product Kz Ym;v
stress hole stress
EN12150/NEN2608 75 N/mm2 0,65 1,2 40,62 + 6,66 N/mm2 47,28 N/mm2

106,66
Securipoint minimum 125 N/mm2 0,8 1,0 100 + 6,66 N/mm2
N/mm2
PaperCC2 152,25 128,86
0,8 1,0 121,80 + 6,66 N/mm2
testsamples N/mm2 N/mm2
Table 4 shows the different hole strength in relation with the level of residual stress.
When using only CE (EN12150) toughened glass the ultimate stress is 47,28N/mm2.
Only when higher demands are set for residual stress, in combination with a procedure
to control this minimum level of residual stress , higher levels of stress can be accepted.
Kz = 0,8 for other that NEN2608 ultimate stress design because of hole position. [3]
8. Conclusion
When table 4 is compared to table 3 to 1 you can see the enormous difference between
CE ultimate stress and the derivate experimental ultimate stress. Especially the level of
residual stress from the tempering process has a significant influence on bearing
strength. So when tests are done to derive a bearing strength of a connection. The level
of residual stress must be known. The level of residual stress of the toughened glass
used on site must, at least, be the same as the level of residual stress measured at the test
samples.
9. References
[1] Bailinger, Mascha, Zur bemessung von SL-belasteten Anschlüssen in konstructiven
[2] DASt,2/2007,Untersuchung von Stah-Glas-Verbindungen im Hinblick auf die
[3] Nielsen, Jens Henrik, Tempered Glass bolted connections and related problems.
266
doi:10.3233/978-1-61499-061-1-267
Influence of Various Factors on

Mechanical Properties of Adhesive Joint
in Glass Structures
Klára Machalická, Martina Eliášová
CTU in Prague, Faculty of Civil Engineering, Department of Steel and Timber
Structures, Czech Republic, klara.machalicka@fsv.cvut.cz, eliasova@fsv.cvut.cz
Ability of adhesive connection to distribute stress in uniform manner due to relative

large bonding area is a significant advantage in brittle and stress concentration
sensitive glass. For the glued joint design there are many aspects which have an
essential influence on shear load carrying capacity and behaviour of joint under
increasing load. The shear strength of adhesive connection is affected by correct
choice of glue for specific joint in the first place, but there are many other important
factors which have to be taken into the consideration during adhesive joint design.
Research is focused on effect of thickness of adhesive layer, different joining
materials including their surface preparation and surface roughness and exposure to
environment conditions for several chosen adhesives with different mechanical
properties.
Keywords: Adhesion, Artificial ageing, Glass structures, glued joint,

Environmental factors, Shear bonded connection, Thickness of adhesive layer
1. General
The contemporary architecture uses widely glass as a structural element due to great
advantage of transparency and glass structures are often combined with other materials,
too. Adhesive joint is more appropriate way how to make connection in glass structures
because of glued joint ability to distribute the loads arising from connection in more
uniform manner in comparison to bolted connection. There are other benefits like
possibility of joining different and also thinner materials, reduction in weight, easier
reaching of composite action of built – up sections, transparency and aesthetical quality.
Adhesive connections can be produced chemical resistant, depending on choice of the
adhesive. Adhesive layer can act, if it is necessary, also as a sealing. These above
mentioned advantages have a practical meaning for designing structures with a glass as
a load bearing elements.
Thanks to intensive research and development in improving material properties of

polymer adhesives in recent years there are many types of glue with different properties.
But missing standards or guidelines and lack of data about behaviour of adhesive in
particular joint are some of the difficulties, which have to be solved before correct and
safe design of adhesive connection.
267
Challenging Glass 3
1.1. Research experimental program

The research is focused on various types of adhesives and different joining materials to
examine influence on joint behaviour under increasing shear load and also specific part
of research is aimed at environmental and ageing effects. Investigation of adhesive
connections deals with elastic adhesives (one-component and two-component
polyurethanes), semi-rigid adhesives (two-component acrylic adhesive) and transparent
adhesives (UV-curing systems), which can be suitable in case of glass-to-glass bonding.
The research covers glass to glass and glass to metal (steel, stainless steel and
aluminium) joints with smooth or coarsened glass surface because of adhesion
improvement possibility. Adhesives have been applied in two different thicknesses to
determine influence on behaviour of the joint. Particular part of the research has been
performed for glued connection exposed to environmental conditions - changing
temperature (from -20 °C to 80°C), UV-radiation and increased relative humidity at the
same cycle of artificial ageing.
2. Choice of adhesives
Adhesives are polymer materials that are consisted of monomer units chained into
macromolecules. Chemical composition, molecular structure and cross-linking rate of
polymer determine the polymer properties. Adhesives used in glass structures can be
divided according to their modules of elasticity and shear modulus into flexible-elastic
or rigid systems.
2.1. Flexible adhesives

Flexible connection is created by using adhesive with low modulus of elasticity applied
in several millimetres thickness. Flexible adhesives (typically silicones or
polyurethanes) can be classified according to thermo-mechanical properties like
elastomers. Elastomers have low cross-linking rate and thus they can be stretched easily
several times their unstretched length and they rapidly return to their original
dimensions when the applied stress is removed. Elastomers have low tensile strength
(about 1 to 10 MPa) and high elongation at break even more than 350%. Due to low
modulus of elasticity they can easy distribute stress in uniform manner. The flexible
adhesives are very suitable for linear connections, accepting dynamic loads, damping
sound transmission between the components and functioning as a seal. Two flexible
adhesives have been selected to the research program – one-component and two-
component polyurethane. They have higher strength values then silicone and elongation
more than 200%, so they are suitable for flexible load-bearing connection of glass and
other material.
2.2. Rigid adhesives

Adhesives which can be classified according to their thermo-mechanical properties like
plastomers (thermosets or thermoplastics) can create rigid connection. Thermosets have,
due to their high cross-linked polymer chains, very high strength and low elongation at
break. Rigid adhesives (typically epoxy resins or acrylic adhesives) can be
differentiated to contact adhesives that require small adhesive thickness (often under 1
mm) and gap-filling adhesives, that are able to perform at thicknesses in excess several
millimetres [1], which can be useful for a hybrid structures due to imperfections in
flatness of joining materials. Some acrylics adhesives are transparent and cured by UV-
radiation. After curing they are UV-resistant, which can be benefit for glass structures.
268
Influence of Various Factors on Mechanical Properties of Adhesive Joint in Glass Structures
Three types of rigid adhesives were chosen to the research - gap filling semi-rigid two –
component acrylic adhesive and two types of transparent UV-curing adhesives, which
can be suitable in case glass to glass bonding.
3. Adhesion and cohesion

The strength of adhesive joint depends on two main factors – adhesion and cohesion.
Adhesion is a tendency of certain dissimilar molecules of adhesive and substrate to
cling together due to attractive forces. In contrast, cohesion takes place between similar
molecules of one material. Strong intermolecular bonds with high degree of cross-
linking cause high cohesion strength of adhesive. Strong atomic bridges between glue
and substrate make strong adhesive joint. Generally, failure of glued joint can happen
by loss of adhesion or by cohesive strength exceeding.
3.1. Types of adhesive joint failure

Shear adhesive connection in glass structure can fail by some of the three main modes
or their combination. Firstly, failure can occur as adhesive slip. This is the failure at the
adhesive – glass boundary; it is caused by poor adhesion of glue to glass surface, see
Fig. 1. Secondly, failure can develop within the adhesive layer by excess shear cohesive
strength of glue. This mode of failure can be caused by insufficient cohesion of
adhesive, see Fig. 2. Thirdly, some adhesives are stronger than the glass to which they
are attached and failure can occur within the glass by exceeding tensile strength of glass.
The behaviour and failure mode of the glued joint is dependent on the surface
preparation (cleaning, degreasing, primer coating). Adhesion is influenced by materials,
which are bonded together, as well as every surface treatment. Etching or sandblasting
of glass surface can improve the adhesion but it will reduce the strength of glass.
Furthermore, roughened surface can improve adhesion only with sufficiently liquid
adhesive. Glue with high viscosity can wet only surface protrusions and this can lead to
reduce of bonded area and subsequently to reduce of bond strength, see Fig. 3 [2].
Figure 1: Adhesive failure. Figure 2: Cohesive failure.
Figure 3: Surface wetting by a) adhesive with high viscosity, b) adhesive with low viscosity
269
Challenging Glass 3
4. Environmental condition effect to adhesive joint

Adhesive bond has to resist the service environmental conditions, which influence
strength and durability of joint. Chemical composition and cross-linking rate of polymer
affects the environmental resistance of adhesive and joint service time. Mechanical
properties of adhesive connection, which depends on adhesive layer itself as well as
interface between adhesive and substrate, may deteriorate upon exposure to moisture,
UV radiation or temperature range that the connection has to withstand during its
service life.
4.1. Glass transition temperature and thermal resistance

Thermal resistance of polymer adhesive depends on glass transition temperature Tg.
Temperature Tg is way to understanding molecular motion that occurs in polymeric
material. The degree of molecular motion affects adhesive and cohesive forces, polymer
chain and its structure, cross-linking, molecular weight, brittleness and other polymer
properties. At low temperatures, less than Tg, polymer behaves like a solids in which the
molecular segments has a moderate and independent motions. If the temperature of
polymer is increased, molecules become more flexible and mobile. Transition of
polymer from glassy to rubbery state signifies that temperature is close to Tg. If the
temperature is raised above Tg, distance between molecular segments is increased and it
is accompanied by increasing the specific volume of the polymeric material, [3].
Glass transition temperature should be above the upper service temperature for high
bond strength values. Elastomers have usually low Tg (below freezing) to have low
modulus of elasticity, low tensile and shear strength and high elongation at break. Rigid
adhesive have usually high Tg to ensure high strength values during common
temperatures, but they loss their stiffness and strength if temperature increases.
Furthermore, adhesive forces between substrate and glue become weaker during high
temperatures, which can cause peeling off adhesive layer from one of the bonded
surfaces (adhesive failure). On the contrary, with decreasing temperature contracted
bonded materials will cause increasing stiffness of the bonded connection but also
adhesive become brittle and the joint is prone to cohesive failure [4].
During repeated temperature changes adhesive layer has to be flexible enough to

equalize different elongations of different joining materials. This can be achieved by
using flexible and durable adhesive with optimal thickness of layer.
4.2. UV-radiation
UV-radiation is a main cause of organic materials damage and in glass structures
connection is important choose UV-resistant adhesives, because UV-radiation goes
through the glass and can break external layers of adhesive. It can lead to damage of
adhesive forces between glass and glue. In case of bonding glass to metal (or non-
transparent) structure, there is suitable to protect adhesive layer by glass coatings
(primer coating) if adhesive has not sufficient UV-resistance.
4.3. Moisture and water effect

Environmental moisture or water can be absorbed into polymer material and cause
swelling up of glue. But with decreasing relative humidity moisture can migrate out
from polymer material and cause other volume changing. Repeating of this process can
270
lead to worsening adhesive forces. Besides, absorbed moisture in polymer material can
migrate to interface between adhesive and substrate and can accumulate at micro-
cavities. This effect can lead to deterioration of adhesive forces and subsequently to
adhesive failure. In case of glass-metal bond, environmental humidity can also degrade
metal substrate by corrosive attack and consecutively adhesion of whole bonded
connection. Furthermore, water often in combination with heat can lead to hydrolysis
and cause changes in macromolecular structure of polymer which leads to changes in
material properties of polymer.
5. Performed experiments –joint thickness, material and surface treatment effects

The first part of experimental analysis for chosen adhesives is focused on aspects
influencing adhesion and effect of adhesive layer thickness on behaviour of joint.
5.1. Experimental setup

The small-scale tests, specialized in behaviour of adhesive joint under increasing shear
load, were performed according to the scheme in Fig. 4. Different types of adhesives
with different mechanical properties were chosen after consulting with SIKA CZ and
ProVetro Group affiliated cooperation's partner for Czech Republic. The sphere of
interest covers connecting glass with different materials (steel, aluminium, stainless
steel, glass) to examine adhesion to these materials. Required adhesion was reached by
using certified technologies of surface treatment and by primer coating for some of the
adhesives. Glass for specimens was utilized both with smooth glass surface (only
cleaned and degreased) and with roughened surface by sandblasting to obtain better
adhesion of glue to the glass surface. Elastic and semi-rigid adhesives (one-component,
two-component polyurethanes and two-component acrylic adhesive) were applied in 3
and 4 mm thick layer to join glass with metals. For transparent connection glass to glass
two types of rigid UV – curing adhesives were chosen, which were applied in 1mm
thick adhesive layer.
exchangeable middle part

from different materials
with glued glass specimens
adhesive layer
adhesive layer
float glass float glass
polyamide
setting blocks
Figure 4: Setup of the small-scale shear connection tests: on the left and middle – setup for metal to glass
connection, on the right – setup for glass to glass or timber to glass connection
271
Challenging Glass 3
5.2. Results
For every tested adhesive, summarized diagrams were prepared for comparisons of joint
types with various materials and glue thickness in one picture, see Fig. 5, 6 and 7. The
summarized shear stress-strain relationships show representative curves, which were
chosen according to their approaching to average values of results.
One-component polyurethane adhesive at all types of specimens reached strength value
approximately 4 MPa with elongation at break more then 350%. Two-component
polyurethane adhesive achieved strength values approximately 4.5 MPa with elongation
at break about 250%. The most of specimens made from both types of polyurethane
were broken in cohesive mode of failure. Sandblasting of glass surface has no
substantial effect on strength of glued joint by polyurethane adhesives because the
cohesive strength of adhesive was achieved also at specimens with smooth glass surface.
3 1 1
5 4 3
2 4
2 5
1...steel + glass (3mm)

3...steel + sandblasted glass (3mm)
4...stainless steel + glass (3mm)

5...aluminium + glass (3mm)
Figure 5: Shear stress-strain relationship – PU adhesives.
Two-component acrylic adhesive got to the strength value more than approximately 6
MPa with elongation at break 100 – 150%. Failure started predominantly in adhesion
manner but part of adhesive layer still behaved cohesively, so the joint was broken by
combined adhesive-cohesive failure. From this reason, specimens with sandblasted
glass reached up to 20% higher shear strengths than specimens with smooth glass
surface and specimens with sandblasted glass was broken only by cohesive failure.
Transparent glass to glass specimens achieved strength values approximately 10 – 15
MPa with elongation at break about 50 – 100 %. Difference between specimens without
surface roughening (surface was only cleaned and degreased before gluing) and with
surface sandblasting is showed at Fig. 6 right. Sandblasting has no essential effect on
272
strength of the glued joint, because the failure of specimens with smooth glass surface
was probably broken by glue cohesive failure together with glass rupture.
3 4
5
1 2...steel + glass (4mm)
6
3...steel + sand-blasted glass (3mm)
2
4...stainless steel + glass (3mm)
5... stainless steel + glass (4mm)
6... aluminium + glass (3mm)
Figure 6: Shear stress-strain relationship – acrylate adhesive.
b c
a...UV-adhesive 1 (1mm); smooth glass

d b...UV-adhesive 1 (1mm); sand-blasted glass
c...UV-adhesive 2 (1mm); smooth glass
d...UV-adhesive 2 (1mm); sand-blasted glass
Figure 7: Shear stress-strain relationship – UV-curing adhesives.
6. Performed experiments –environmental condition effect

The second part of experimental analysis deals with the environmental effects to bonded
joint. Moisture, UV-radiation, high and low (frost) temperature effect is in one typical
cycle of accelerated ageing. The second series of specimens were prepared for the same
adhesives and same bonding materials (substrates) like at the first part of research.
Specimens were subjected to laboratory ageing conditions and subsequent there were
performed shear tests according the scheme on Fig. 4. Artificial ageing effect was
273
Challenging Glass 3
assessed by naked eye to find out some modifications. Shear tests results (shear strength
values and failure mode) were compared for both parts of research.
6.1. Laboratory ageing

Typical cycle of artificial ageing, see Fig. 7, comprises eight-hour exposure to UV-
radiation alternating with demineralised water showers at 20 °C (i.e. weterometer
conditions). Afterwards, specimens were subjected to sixteen-hour exposure to high
temperature (+80 °C) or low temperature (-20 °C). This part is repeated four times and
than cycle is finished by conditioning at (20±2) °C and relative humidity (60±5) %.
Whole cycle is repeated nine-times. This whole procedure replaces 5 years in exterior
conditions in Czech Republic climate according to Research and Development Timber
Institute´s internal regulation for polymer coatings at metal materials.
temperature [°C ]
every 20 minutes UV-radiation or
water shower (weterometer)
increased / decreased temperature

without UV-radiation and showers
Conditioning for 64 hours

(20°C, 60% rel. humidity)
time [hour ]
Figure 7: Typical cycle of laboratory ageing.
6.2. Observable modifications caused by ageing

There were no significant changes at PU-adhesives observable with the naked eye. Both
of PU-adhesives have low UV-resistance, but this problem was solved by primer
coating to glass surface.
Specimens glued by 2-component acrylate adhesive showed adhesive faults at glass

surface and also there were small bubbles and small surface cracks at adhesive layer
edge, see Fig. 8. These faults were probably created due to temperature changes which
adhesive had to withstand. The adhesive should be thermally stable from -40°C to
+80°C according to technical data sheet. But glass transition temperature Tg is around
+50°C and it is within the temperature range of laboratory ageing cycle. There is
assumption that changes of specific volume when temperature went repeated over the Tg,
caused the small bubbles, cracks and adhesion faults in glue.
274
Figure 8: Faults caused by artificial ageing Figure 9: Unsealed specimen from

at 2C-acrylate adhesive. UV-curing adhesive 1.
One type of the UV-curing adhesives was found out like the least ageing resistant
adhesive from the selection of glues. All specimens glued by this UV-curing adhesive
with smooth glass surface were unsealed spontaneously during artificial ageing.
Specimens with this adhesive and roughened surface by sand-blasting had no
observable modifications. Because of UV-curing adhesives are UV-resistant after
hardening and they are thermal stable to 100 °C with glass transition temperature Tg
higher then temperature range of laboratory ageing, there is assumption that adhesion
weakening and subsequent pealing off was caused by high relative humidity at
weterometer.
The second type of UV-curing adhesive had also no noticeable changes, which can
affect the mechanical properties. But adhesive layer became light yellow, which can be
caused by raised relative humidity during accelerated ageing.
6.3. Results of artificial aged specimens

Shear test results in stress-strain relationship graphs including comparisons both
research parts are showed in Fig. 10, 13 and 14. Continuous line in the graphs marks
results of laboratory aged specimens and dashed line provides comparison with results
of specimens which were not exposed to ageing.
Specimens bonded by 1-component PU adhesive had similar shear strength and strain
values like same specimens without ageing effect. 2-component PU adhesive specimens
had shear strength values higher than same specimens without ageing. There is
assumption that this glue needs longer time for curing than is stated in technical data
sheet. Thermal and moisture resistance of these adhesives have been proved. UV-
resistance was sufficiently solved by black primer coating on glass surface. Cohesive
failure mode was observed at majority of specimens.
Specimens bonded by two-component acrylic adhesive showed similar shear strength

values like specimens without artificial ageing, no negative effect of small bubbles or
adhesive faults was observed. Average shear strain values were measured by 15%
higher and deformation had more plastic character than deformation at specimens
without ageing effect, see picture of broken specimens, Fig 11. Failure predominantly
started at adhesive manner like at specimens without ageing but cohesive failure was
decisive.
275
Challenging Glass 3
1 2*
3 3 1*
2
2
3*
2* 3*
1...steel + glass, lab. aged

2... stainless steel + glass, lab. aged
3... aluminium + glass, lab. aged
1*...steel + glass
2*...stainless steel + glass
3*...aluminium + glass
Figure 10: Shear stress-strain relationship – laboratory aged PU adhesives.

Figure 11: Specimen bonded by Figure 12: Specimen bonded by

2C-acrylate adhesive after shear test. UV-curing adhesive after shear test.
276
2*
1 1...steel + glass, lab. aged
1* 2... stainless steel + glass, lab. aged
3... aluminium + glass, lab. aged
3
2
* 1*...steel + glass
3
2*...stainless steel + glass
3*...aluminium + glass
Figure 13: Shear stress-strain relationship – laboratory aged 2C -acrylate adhesive.
Test specimens glued by the first type of UV-curing adhesive with sandblasted glass
surface had strength reduced to only about 40% of original value and shear strain was
measured by 50% higher. The glue was evaluated like unsuitable for structural bonding,
because of unsealing specimens with smooth glass surface and also for unsatisfactory
mechanical properties after ageing for roughened surface specimens.
Specimens bonded by the second type of UV-curing adhesive reached about 40% lower
strength values with approximately same deformation like specimens without laboratory
aging. The test specimens were broken by combination of cohesive failure in glue and
creating cracks in glass.
b...UV-adhesive 1; sand-blasted glass; lab. aged

b*
a*
c...UV-adhesive 2; smooth glass; lab. aged

c*
d...UV-adhesive 2; sand-blasted glass; lab. aged
d*
c
a*...UV-adhesive 1; smooth glass
d
b*...UV-adhesive1; sand-blasted glass
c*...UV-adhesive 2; smooth glass
b
d*...UV-adhesive 2; sand-blasted glass
Figure 14: Shear stress-strain relationship – laboratory aged acrylate adhesives.
277
Challenging Glass 3
7. Summary
There were performed numerous experiments focused on different joining material
(steel, stainless steel, aluminium + glass, glass + glass) including their surface treatment
(sandblasted glass) for 5 various types of glue applied at different thicknesses. Sphere of
interest covered also environmental influences.
Influence of different materials in bonded connection was assessed in the first part of
this research. There were proved that sandblasted glass surface can improve shear
strength values, if pure cohesive mode of failure is not reached at joint with smooth
glass surface. No decreasing strength values of sandblasted glass were observed because
of the cohesive strength of adhesive was crucial in the glued joints. Influence of
different metal materials in glued connection was not decisive, but it is remarkable that
specimens composed of glass with stainless steel glued by 2-component PU or acrylic
adhesive proved higher shear strength values then samples with common steel.
Effect of thickness was noticeable at semi-rigid 2-component acrylate adhesive -

strength of glued connection was decreased with increasing thickness of adhesive layer,
especially at glass – steel type of connection. At flexible PU-adhesives the influence of
thickness was not noticeable.
The second part of research was focused on environmental effects. Every adhesive from
selection showed different behaviour after laboratory ageing, which can simulated five
years at exterior conditions in middle Europe climate. PU-adhesives and 2-component
acrylate adhesive had no significant deterioration of mechanical properties, but some
changes, especially at acrylate adhesive specimens, were observed. One type of UV-
curing adhesive was showed as unsuitable for structural use because of weakened
adhesive forces after artificial ageing. The second type of UV-curing adhesive had
appropriate behaviour, but strength values were reduced by 40% from original (without
ageing) values.
But there is important to keep in mind that dependence between ageing affect and time
is not linear, so there is bad opportunity to predict behavior of joint at different time.
Mechanical properties can worsen less or more based on chemical composition and
macromolecular structure of particular adhesive. Except above mentioned, it is
important to say that long-term load effect during exposure to environmental conditions
can be significant, but one of the main goal of this investigation was evaluated
environmental resistance of selected adhesives.
8. Acknowledgements
This paper is carried out with a support of the project SGS10/237/OHK1/3T/11.
278
9. References
[1] Haldimann, M., Luible, A., Overend, M., Structural Use of Glass, International Association for Bridge
and Structural Engineering – ETH Zürich, Zürich, 2008
[2] Weller, B., Tasche, S., Vogt, I., Bonded Joints of Adhesives with Higher Strength, Proceedings of the
International Conference on Building Envelope Systems & Technology, pp 185-195, Bath, UK, 2007
[3] Petrie, E.M., Handbook of Adhesives and Sealants, McGraw-Hill, USA, 2007
[4] Huveners, EMP., van Herwijnen, F., Mechanical shear properties of adhesives, Proceedings of Glass
performance days, pp 367 – 370, Tampere, Finland, 2007
[5] technical data sheets [online], Sika CZ [vid. 13.4.2012], available on:
http://cz01.webdms.sika.com/fileshow.do?documentID=1790,
http://cz01.webdms.sika.com/fileshow.do?documentID=1143,
http://cz01.webdms.sika.com/fileshow.do?documentID=1111
[8] product data sheet [online], univarsc [vid. 13.4.2012], available on:
http://www.univarsc.com/DynamicContent/Documents/RITE-LOK%20Datasheets/RITE-
LOK%20UV50.pdf
[9] technical data sheets [online], MEVHA [vid. 13.4.2012], available on:
http://www.mevha.cz/editor/image/download1_soubory/conloc685.pdf
279
doi:10.3233/978-1-61499-061-1-281
Seismic Behaviour of
Point Supported Glass Panels
Luís Martins, Raimundo Delgado
Faculty of Engineering of the University of Porto, Portugal, www.fe.up.pt
Rui Camposinhos
Scholl of Engineering of Polytechnic of Porto, Portugal, www.isep.ipp.pt
Tiago Silva
FACAL - Façade Engineering, Portugal, www.facal.pt
Due to its transparency today’s architectures often rely on glazed façade solutions
to execute the building envelope. During a seismic event, glass breakage and fall
out can occur and threaten occupants and passers. So in earthquake prone regions
this must be accounted in the design project; however, there is little research on the
behaviour of glazed façades under seismic loads. This papers focus on the results of
a project under development to ascertain right well dimensioned and adequate
solutions to glass facades using point fixing systems under seismic or wind actions.
Numerical analyses were performed, using finite element commercial software, the
results are compared with simplified methods and conclusion are drawn.
Keywords: spider glass, glass façade seismic behaviour, curtain wall glass façade
1. Introduction
Recent developments in science and technology allowed glass to be used in several
structural demanding applications such as façades, roofs, girders and columns, etc.
Due to glass’ brittle behaviour this calls for more refined analysis methods and greater
design detailing to ensure structural stability and safety.
Currently modern facade buildings rely on glazed curtain wall systems. These systems
include either singular aluminium alloy frame glass curtain walls or frameless glass
curtain walls. This is the case of the so called spider fixing systems, which are pointed
supported.
Although there are some research on the behaviour of glass panels under out-of-plane
loads, e.g., wind loads, the combine effect of both in-plane and out-of plane loads that
are applied to the panels during an earthquake is a field of research still on its early
stages.
In fact, seismic action brings out specific problems to designers, due to the lack of, at
least well-known, international rules or regulations about this problem.
The scope of this paper is to present the relevant aspects of the seismic loading in point
fixed glass panels. Firstly a simplified method based upon the elastic response spectrum
is introduced, then the results of a time history dynamic analyses are presented.
281
Challenging Glass 3
2. Seismic design of glass façades

Earthquakes cause damage to the building main structure as well as to the non-structural
components. Falling façade fragments during an earthquake poses a serious hazard to
pedestrians and occupants as well, so in earthquake prone regions its must be accounted
in the design.
Sucuo lu and Vallabhan refer broken window glass as the second most serious non-
structural damage, for example in the 1985 Mexico City earthquake over 50% of the
263 office observed buildings had experienced some sort of glass breakage [1].
During an earthquake two types of lateral loads are considered acting in the façade
panels: the “in plane” loads and the “out of plane” loads. Both inertial loads are caused
by the horizontal displacements of the building’s floors, yet the in-plane actions causes,
mainly shear stresses, and the others give rise to extra inertial forces due to the panel’s
bending.
The frequency content of the dynamic loads transmitted to the panels is modulated by
the building natural frequency, so if it happens that it has a value very close to the
panel’s natural frequency, resonant effects occur with an agonisingly increase of the
dynamic response, a well-known phenomena that must be avoided, otherwise structural
safety may be compromised.
As there are no design regulations for determine seismic loads in glass façades, a
laboratory test procedure according the American Architectural Manufacturers
Association recommendations [2, 3] is followed in order to evaluate the maximum
seismic drift which may cause glass breakage and fall out of framed glass panels.
The dynamic test procedure considers a sinusoidal drift history for the procedure testing
with growing amplitudes to a maximum of 150 millimetres (Figure 1). This test method
has been applied in previous studies, like (Memari et al) [4], and is going to be used to
assess the seismic behaviour of point fixed glass panels in the Seismic Laboratory of the
University of Porto Engineering faculty.
Figure 1: Schematic of displacement time history for dynamic crescendo test [2].
282
Seismic Behaviour of Point Supported Glass Panels
3. Simplified method to assess the seismic forces

The simplified method to assess the seismic forces transmitted to the façade panels, was
adapted by Camposinhos [5] from the work of Singh [6, 7], and is based on the response
spectra of Eurocode 8 (EC8) [8]. Figure 2 presents the elastic response spectrum
adopted in this work.
Response spectrum (Seismic action type II; soil type D; seismic

region 2.1; =1.0)
10
8
Sa (m/s2)
6
4
2
0
0 1 2 3 4
T (s)
Sa
Figure 2: Response spectrum [8].
Thus the dynamic load transmitted to the façade panels (equation 1) depends on the
building’s natural vibration period and on the panel’s mass and natural vibration period.
0.40 u C fZ u S DS u J E u M E
FEk (1)
RE
Where,
FEk characteristic seismic force;
CfZ seismic coefficient of the panel (depends on the building dynamic
characteristics and the position on the panel in the building);

SDS ground acceleration value;
E importance coefficients of the panel (ranging between 1.0 and 1.5);
ME panel’s mass;
RE coefficient of performance of the panel (ranging between 1.5 and 3.5).
Prior to the application of the simplified method a parametric analysis was made to
evaluate the variation of seismic coefficient CfZ with the dynamic properties of the panel
and its position or height in the building. Two different cases were studied: in the first
case the panel is assumed to be in the last floor (m=N), and in the second, the panel was
assumed to be in placed at the penultimate floor (m=N-1). Figure 3 and Figure 4
presents the variation of Cfz with the natural period of vibration of the building for
different values of the natural period of vibration for the glass panel.
283
Challenging Glass 3
Variation of CfZ (m=N; =5%)

100
Tpanel=0.05s
80
60 Tpanel=0.2s
CfZ
40 Tpanel=0.4s
20
Tpanel=0.6s
0
0 1 2 3 4 Tpanel=0.8s
T(s) Tpanel=1s
Figure 3: Variation of Cfz (m=N).
Variation of CfZ (m=N-1; =5%)

100
Tpanel=0.05s
80
60 Tpanel=0.2s
CfZ
40 Tpanel=0.4s
20
Tpanel=0.6s
0
0 1 2 3 4 Tpanel=0.8s
T (s) Tpanel=1s
Figure 4: Variation of Cfz (m=N-1).
The results (Figure 3 and Figure 4) shows a very significant resonant effect for low
periods and a comparison between them allow to conclude that a panel in the last floor
corresponds to the most severe situation in what seismic excitation concern, with values
about 40% higher, once the coefficient, CfZ is directly correlated with the maximum
seismic force acting in the panel.
4. Case study
4.1. Field application

In order to analyze the behaviour of different solutions of point supported glass panels
systems under seismic loads a partnership study involving the Faculty of Engineering of
the University of Porto and FACAL – Façade Engineering, a Portuguese company
known for its work in the field of glazed façades, has been promoted. FACAL has a
great experience in technical innovative solutions, namely looking for the development
and easy installing thinner and lighter glazing, and is well known by the execution of
famous facade glazings such as the glass system in Casa da Música in Oporto, the
double skin façade Torre H in Lisbon, the glazing of the Spanish Pavilion in the
International Fair in Zaragoza or the glass roof of the Metro Station at Sá Carneiro
Airport in Porto.
On the basis of this work is the newly-built office building of Bouygues Imobiliária in
Lisbon. A particular attention was given to the safety requirements of a 20 meters high
284
point-supported glass façade over the main entrance of the building. The post-breakage
behaviour of the laminated safety glass panels is the primary concern. This is often a
neglected problem in glass façades yet of greater importance since Lisbon as many
other metropolis is prone to seismic activity.
A crucial concern on this matter is essential: The integrity of the glass façade must be
preserved and assured that there’s no risk of glass fragments fallings.
Among the various enhancement solutions the use of laminated safety glass with the
DuPont interlayer SentryGlas® provides an optimal behaviour due to its excellent post-
breakage performance with the glass fragments remaining adhered to the interlayer,
without falling down.
Figure 5: Point-supported glass façade - office building of Bouygues Imobiliária in Lisbon.
4.2. Geometrical and mechanical properties

This paper focus on a set of five point supported laminated glass panels. All the
specimens have a surface of 2350x2300 mm2 differing on the glass thicknesses, the
interlayer and distance from edges to holes (Figure 6).
Figure 6: Geometrical configuration of the glass panels: left) V1 and V2; right) V3, V4 and V5
(Drawings by: FACAL).
285
Challenging Glass 3
The panels identified as V1, V2, V3 and V4 are obtained from 10 mm strengthened
glass and a 1.52mm thick interlayer together with a 8mm also tempered glass.Panel V5
is made from two 12mm tempered glass sheets.
Panels V1 and V3 have a SentryGlass® film while the remaining panels have PVB
interlayer. Furthermore a panel with the same geometrical configuration of that of V5
but with SentryGlass® interlayer film was additional considered in the analyses.
The properties of the PVB’s Young modulus ranges from 3.2 MPa to 18 MPa [9, 10], so
in the panels with this type of interlayer three different values for this parameter were
considered: (i) 3.2MPa, (ii) 9.0 MPa and (iii) 18.0 MPa. In the case of SentryGlas® a
the value of 300 MPa was adopted for its Young modulus, as suggested by Delincé [10].
Table 1summarises the relevant data for the studied glass panels.
Table 1: Test specimen’s description.

Glass Dimensions Total Interlayer Interlayer’s Glass’ Glass’
panel [mm] thickness film Young Young Poisson
[mm] modulus modulus coefficient
[MPa] [GPa]
V1 2350x2300 10+1.52+8 SentryGlas® 300 70 0.20
V2(i) 2350x2300 10+1.52+8 PVB 3.2 70 0.20
V2(ii) 2350x2300 10+1.52+8 PVB 9.0 70 0.20
V2(iii) 2350x2300 10+1.52+8 PVB 18.0 70 0.20
V3 2350x2300 10+1.52+8 SentryGlas® 300 70 0.20
V4(i) 2350x2300 10+1.52+8 PVB 3.2 70 0.20
V4(ii) 2350x2300 10+1.52+8 PVB 9.0 70 0.20
V4(iii) 2350x2300 10+1.52+8 PVB 18.0 70 0.20
V5(i) 2350x2300 12+1.52+12 PVB 3.2 70 0.20
V5(ii) 2350x2300 12+1.52+12 PVB 9.0 70 0.20

V5(iii) 2350x2300 12+1.52+12 PVB 18.0 70 0.20
V5(iv) 2350x2300 12+1.52+12 SentryGlas® 300 70 0.20
4.3. FE model
To evaluate the structural response of the glass panels a set of numerical models were
made using commercial finite element (FE) software.
The glass panels and interlayer film were modelled with 8-node 3D finite elements. In
the interior of the panel the maximum size of the finite elements was limited to 2
centimetres, while near the supports the maximum size was reduced to half to attend the
stress concentrations near the holes to take in account the expected stress concentrations
in this regions.
286
Figure 7: FE mesh details - a) lateral view; b) interior of the panel; c) support region.
The degrees of Freedom (DOF) restrictions which enabled this stress concentration near
the support region and the brittle nature glass leads to a several numerical model
simulations stages until a solution that that correctly represent the real support condition
was achieved.
To assess the behaviour of the model two premises were advanced: (i) the allowance for
rotations of the panel according the real behaviour; (ii) Stress distribution around the
hole should be compatible with reality.
The best solution lead to the implementation of an external node where the support
constraints are included. The connection to the panel was provided throughout rigid
pseudo beam elements (Figure 8).
Figure 8: Lateral view of the support solution.
287
Challenging Glass 3
5. Results
5.1. Simplified method

The natural frequencies of the glass panels were determined using the FE model and are
depicted in Table 2.
The already mentioned simplified method was applied to assess the maximum seismic
effect on the façade panel and the results are presented in Figure 9.
As it can be observed, the peak values presented in the figure, nearly 10 times the
panel’s self weight, confirm the expected resonant effects in the panels.
Table 2: Glass panels natural frequency.

Glass panel Natural frequency (Hz)
V1 13.4
V2(i) 10.3
V2(ii) 11.5
V2(iii) 12.2
V3 8.1
V4(i) 6.4
V4(ii) 7.1
V4(iii) 7.5
V5(i) 8.0
V5(ii) 9.0
V5(iii) 9.5
V5(iv) 10.5
Seismic forces transmited to the façade panels

30
20
FEk (kN)
10
0
0 1 2
(s)
V1 V2(i) V2(ii) V2(iii) V3 V4(i)

V4(ii) V4(iii) V5(i) V5(ii) V5(iii) V5(iv)
Figure 9: Maximum seismic force transmitted to the glass panels.
288
5.2. Time history dynamic analyses

In order to qualitatively assess the results obtained by the simplified method a set of
time history dynamic analyses has been performed, using the El Centro ground motion
record (Figure 10) appropriately scaled so that the maximum spectral acceleration was
equal to the one calculated by EC8.
El Centro earthquake, May 18th 1940 (N-S)

4
2
üg (m/s2)
-2
-4
0 5 10 15 20 25 30 35
Time (s)
Ground acceleration
Figure 10: El Centro ground motion record [11].
To verify the existence of resonant effects in the panel a structure with a natural period
of 0.12 s (very close to the natural period of that of panel V5(i)) was analysed with the
selected ground motion record. The floor accelerations obtained, Figure 12, were then
applied to the façade panel and the corresponding response was gathered, Figure 13.
Floor accelerations
10
Acceleration (m/s2)
5
-5
-10
0 5 10 15 20 25 30 35
Time(s)
Floor accelerations
Figure 11: Floor accelerations.
289
Challenging Glass 3
Response of the panel V5(i)

100
Acceleration (m/s2)
50
-50
-100
0 5 10 15 20 25 30 35
Time (s)
Panel acceleration
Figure 12: Response of the panel V5(i).
The maximum response acceleration of the panel, about 81 m/s2, 10 times the peak floor
acceleration, confirms the existence of resonance in the façade panel.
As the simplified method does not consider any damping effects, new dynamic analyses
were made considering the panel’s damping ratio ( ) ranging from 0% to 5%.
Table 3: Differences in the panel V5(i) response for different damping ratio.
Damping ratio [%] Maximum deflection on the panel [mm]

0 16.7
2 11.8
5 8.58
Table 3 shows that there is a decay of almost 30% in the maximum deflection when a
2% damping ratio was considered.
Thus in buildings with natural vibration frequencies that might induce resonance in the
façade panels an energy dissipation device should be applied to reduce the dynamic
response of the panel.
5.3. Numerical simulation of the test procedure according to AAMA recommendations

Section 2 briefly described a test procedure to assess the maximum seismic drift causing
glass breakage and fallout in framed glass panels. This test procedure was numerically
simulated using the same FE models previously developed to determine the panel’s
dynamic properties.
Due to stress concentrations in the support region, material rupture was reached for a
drift less then10 mm (Figure 13).
290
Results for the AAMA dynamic test (Panel V1)

900
800
700
600
(MPa)
500
400
300
200
100
0
0 5 10 15 20 25
drift (mm)
Maximum Tension in the support Tension at 5mm from support

Tension at 10 mm from support f annealed glass
f heatstrenghtned glass f tempered glass
Figure 13: Example of results for the AAMA dynamic test.
The low drift level that causes material failure determined by the numerical test lead to
the formulation of the hypothesis that AAMA 501.6 dynamic test aims to determine the
maximum drift that causes panel’s detachment from the support rather than the material
rupture.
6. General conclusions
In a FEM analysis the importance of an adequate modelling of the support conditions in
the point supported glass panels is mandatory in order to control the stress concentration
phenomenon.
The simplified method to determine the seismic forces transmitted to façade panels
seems to be able to capture the relevant aspects of the whole issue regarding resonance
effects as well.
The numerical simulations using time history analysis confirmed the hypothesis of
resonance effects induced by the building’s natural frequency and are in agreement with
the peak values determined by the simplified method.
Tests showed up that the problem’s sensibility to damping is relevant. In fact a 2%

damping ratio for the panels lead to a 30% decrease in its maximum deflection.
It must be emphasized that in earthquake prone regions façade panels without energy
dissipation devices could be seriously and dangerously excited into non acceptable
limits.
291
Challenging Glass 3
7. References
[1] Sucuo lu, H. and C.V.G. Vallabhan, Behaviour of window glass panels during earthquakes.
Engineering Structures, 1997. 19(8): p. 685-694.
[2] AAMA, AAMA 501.6-09: Recommended dynamic test method for determining the seismic drift causing
glass fallout from a wall system. 2009, AAMA.
[3] AAMA, AAMA 501.4-09: Recommended static testing method for evaluating curtain wall and storefront
systems subjected to seismic and wind induced interstory drift. 2009, AAMA.
[4] Memari, A.M., R.A. Behr, and P.A. Kremer, Seismic behavior of curtain walls containing insulating
glass units. Journal of Architectural Engineering, 2003. 9(2): p. 70-85.
[5] Camposinhos, R.d.S., Revestimentos em pedra natural com fixação mecânica dimensionamento e
projecto. 2009, Lisboa: Edições Sílabo. 199 ISBN 978-972-618-561-1.
[6] Singh, M.P., et al., Seismic design forces. I: Rigid nonstructural components. Journal of Structural
Engineering, 2006. 132(10): p. 1524-1532.
[7] Singh, M.P., et al., Seismic design forces. II: Flexible nonstructural components. Journal of Structural
Engineering, 2006. 132(10): p. 1533-1542.
[8] CEN, Eurocode 8 - Design of structures for earthquake resistance Part1: General rules, seismic actions
and rules for buildings. 2010, IPQ: Caparica. p. 230.
[9] Chen, J., Q. Zhang, and B. Xie, Nonlinear finite element analysis on laminated glass panel based on
APDL. Computer Aided Engineering, 2010. 19(Copyright 2011, The Institution of Engineering and
Technology): p. 22-6.
[10] Delincé, D., et al. Post-breakage behaviour of laminated glass in structural applications. in Challenging
Glass. 2008.
[11] http://www.vibrationdata.com/elcentro.htm. Access Date: 18/05/2011
292
doi:10.3233/978-1-61499-061-1-293
The Mechanical Performance of Adhesives

for a Steel-Glass Composite Façade System
Shelton Nhamoinesu, Mauro Overend
University of Cambridge, UK, sn393@cam.ac.uk , mo318@cam.ac.uk
High strength adhesives provide potentially efficient load-bearing steel-glass linear

connections by enabling composite action. However, there is a lack of reliable
models that can accurately predict their mechanical behavior. This paper describes
experimental investigations undertaken to select suitable adhesives from short-
listed epoxies and acrylates. The selection was based on mechanical performance of
adhesive single-lap shear joints subjected to short-duration loads. The paper also
assesses the validity of an analytical and a viscoelastic-plastic numerical model
used for predicting the stress-state in adhesive joints. The investigation shows that
three of the tested adhesives may be suitable for use in a steel-glass composite
façade system. The analytical model provides good predictions at low strains but
the accuracy decreases with increasing adhesive strains. The non-linear numerical
model provides reasonable predictions but is sensitive to adhesive shear modulus
history.
Keywords: Structural Adhesives, Steel-Glass Composite Façade System, Single-

Lap Shear (SLS) Test
1. Introduction
Despite the ubiquity of bolted connections in structural glazing systems, adhesive
connections are gaining popularity. Unlike bolted connections that weaken the glass in
the vicinity of bolt holes, adhesive bonding ensures a more uniform load transfer
between glass and the supporting elements. As a result, efficient composite behaviour
between glass and the supporting elements can be achieved.
Studies aimed at understanding the mechanical behaviour of adhesive joints date back to
the mid-1940s when Goland and Reissner ‘[1]’ proposed an empirical moment
distribution approach for determining the stress-state in a lap joint. They derived
expressions for the distribution of shear stress across an adhesive in a lap-shear joint
with similar adherends. More recently, Bigwood and Crocombe ‘[2]’ proposed a general
elastic analysis where adherends act as cylindrically bent plates connected along
adjacent interfaces by an adhesive layer capable of transmitting both shear and tensile
loads. Their approach is an extension of Goland and Reissner’s empirical moment
distribution but has been extended to analyse dissimilar adherends.
There are many adhesives produced by different manufacturers which can be potentially
used for steel to glass connections. Perhaps the best known are the structural silicone
sealants which are increasingly being used to achieve flexible structural connections
between glass and aluminium or steel or between glass and glass. Compared to other
types of adhesives, silicones are better understood in terms of their mechanical
293
Challenging Glass 3
performance and durability. These are well documented in several standards and codes
such as BS 6262-6: 2005 ‘[3]’, EOTA 1988 ‘[4]’, AAMA CW-13-85 ‘[5]’ and ASTM C
1401-02 ‘[6]’. Structural silicone joints are relatively thick and flexible, thereby
allowing them to accommodate differential thermal strains between glass and metallic
sub-frames. However with tensile strengths of only 0.8 to 1.8MPa ‘[7]’ for dynamic
loading, structural silicones are unsuitable for transferring the higher longitudinal shear
required for composite action in a typical steel-glass composite façade system.
Several studies on high strength thermosetting adhesives ‘[8], [9], [10], [11], [12], [13],
[14]’ have shown that there is a possibility of using stiffer adhesives such as epoxies
and acrylates for bonding metal to glass. There is however a lack of confidence in the
use of such adhesives for structural applications partly because of a lack of reliable
models that can accurately predict their transient and long-term mechanical behaviour.
Significant research has been done to select the most suitable adhesives for metal to
glass connections ‘[13]’ and ‘[14]’; this paper caries forward the selection process by
adapting a selection criterion that is specific for a typical steel-glass composite façade
system. Six candidate adhesives were investigated:
x Two of the adhesives, the 3M two-part epoxy DP490 and the Huntsman two-
part acrylate Araldite 2047 were chosen on the basis of previous research by
‘[13]’.
x The other two adhesives, 3M two-part epoxy 2216B/A and the Holdtite two-
part acrylate 3295 were chosen on the basis of research done by ‘[14]’.
x The fifth adhesive, 3M two-part epoxy/acrylate hybrid 7271 B/A is a new
product on the market that was recommended by the manufacturer as a
potential metal to glass adhesive.
x The sixth adhesive, Dow Corning two-part silicone DC993 was chosen as a
control adhesive since its material properties are well documented in the Dow
Corning product data sheet ‘[15]’ and its mechanical behaviour has been
extensively investigated ‘[13], [14], [16]’.
x
This paper firstly outlines the determination of bulk material properties of the six
candidate adhesives by uniaxial tensile tests on dumbbell specimens. The material

properties were implemented into an analytical model ‘[2]’ as well as into a
commercially available finite element analysis (FEA) software, LUSAS v14.3 ‘[17]’.
The FEA analyses utilized a viscoelastic-plastic constitutive model as well as a simple
linear elastic perfectly plastic constitutive model. The models were used to predict the
mechanical performance of specially adapted steel-glass single-lap shear (SLS) joints.
Validation of the models was done experimentally by steel-glass SLS tests based on
ASTM D1002-99 ‘[18]’. Whilst different adhesives perform optimally at different bond
thicknesses, all the specimens in this investigation were prepared with a bond thickness
of 3mm which is the minimum gap-fill governed by the fitness tolerances in the end-
application. In addition to performing the tests at ambient temperature, specimens
previously exposed to 800C for 48hrs were also tested to investigate the effect of
extreme temperature on joint performance.
294
The Mechanical Performance of Adhesives for a Steel-Glass Composite Façade System
Finally, the results of the SLS tests and the validity of the analytical and numerical
models were discussed. The selection criterion for the adhesives most suitable for the
steel-glass composite façade system was as follows:
x cohesive or adherend failure preceded by substantial plastic strain in the

adhesive
x relatively high joint flexibility
x adhesive shear strength of at least 7.5MPa
x minimum loss of strength after exposure to temperatures of up to 800C
2. Analytical and Numerical Predictions
2.1. Experimental Determination of Material Properties

4mm thick dumbbells of the six candidate adhesives, sized to comply with ‘[19], [20]’
(Fig.1a) were prepared by casting the adhesives into an aluminium cut-out mould lined
by a PTFE release film (Fig.1b). Air bubbles caused by the chemical reaction of the
adhesive components were minimized by placing the cast mould into a vacuum
chamber; the aluminium mould also acted as a heat sink that reduced bubble formation.
Two different uniaxial tensile tests were performed on each of the adhesives in order to
determine the following bulk material properties: (i) visco-elastic shear modulus Gv, (ii)
visco-elastic decay constant Ⱦ, (iii) elasto-plastic stress-strain relationship and (iv)
poisson’s ratio .
Figure 1: (a) Dumbbell geometry ‘[18],[19]’ and (b) PTFE lined aluminium mould
The first test was performed to determine the visco-elastic properties Gv and . An
instantaneous tensile load was applied to the dumbbells at a high extension rate of
100mm/min up to an extension of 1mm followed by strain holding for approximately
500s while recording the decaying stress. The stress-time curve was converted into a
shear modulus-time curve (Fig.2a); was determined by curve fitting of Eq.1. Gv was
obtained by subtracting the residual shear modulus G from the initial shear modulus G0.
G (t ) Gv e E t G0 Gf e Et (1)
295
Challenging Glass 3
250 3
G0 2.5
200
Shear Modulus Gv(MPa)
True Stress (MPa)

2
150
Gv 1.5
100
1
50
0.5
G
0 0
0 100 200 300 400 500 0 0.02 0.04 0.06
Time (s) True Strain
Figure 2: (a) Shear modulus vs. time and (b) True stress/relaxation vs. true strain curves for 3M 2216 B/A
epoxy dumbbell. The dotted curve is the time-independent true stress vs. true strain
The second test was performed to determine the elasto-plastic properties of the
adhesives by adopting a discrete load-step strategy on the dumbbell specimens ‘[14]’.
The total loading period was divided into approximately ten intervals and at each
interval; the strain was increased by 0.005 followed by strain holding for a period equal
to the decay time td from Figure 2a above. A time-independent elasto-plastic
relationship was obtained by curve fitting of the discrete points on the true stress vs. true
strain graph (Fig.2b). The experimentally obtained material properties of the candidate
adhesives are summarised in Table 1 below.
Table 1: Material properties of the six candidate adhesives

Shear Elastic
Poisson’s Decay
Modulus Modulus Time independent elasto-plastic
Ratio Constant
Gv E0 stress-strain polynomial

(MPa) (MPa)
ߪ ൌ െʹ͹ͷ͸ͺߝ ଶ ൅ ͳ͵ͲͷǤͳߝ ൅ ͲǤͲͲͳ͵ߝ

൑ ͲǤͲͳͶ
ߪ ൌ െ͸ͷͷʹǤ͹ߝ ଶ ൅ ͸ͺͶǤͲ͵ߝ
3M DP490 Epoxy 0.38 239.0 659.6 0.001
൅ ͶǤͶͻͳͺߝ
൐ ͲǤͲͳͶ
ߪ ൌ െͳ͸ͺͺͳߝ ଶ ൅ ͹͸͸ǤͶ͵ߝ ൅ ͲǤͳͷͶͻߝ

Araldite 2047 ൑ ͲǤͲʹͷ
0.43 211.0 603.5 0.002 ߪ ൌ െͳ͵ͲͲǤͷߝ ଶ ൅ ͳͶͳǤ͵Ͷߝ ൅ ͸Ǥ͵͸͵ͳߝ
Acrylice
൐ ͲǤͲʹͷ
ߪ ൌ െ͵ͷ͵͸ͳߝ ଶ ൅ ͳʹͺ͵Ǥͷߝߝ
3M 7271 ൑ ͲǤͲͳ
0.29 559.0 1142.2 0.003 ߪ ൌ െͳͺ͸Ͷͷߝ ଶ ൅ ͻͶͷǤͻߝ ൅ ͳǤ͹Ͷ͸͵ߝ
Epxy/Acylic
൐ ͲǤͲͳ
ߪ ൌ െͳͲ͹ͷͷߝ ଶ ൅ ͳͻͻǤͳͶߝߝ
3M 2216 B/A ൑ ͲǤͲͳ
0.47 192.4 565.6 0.007 ߪ ൌ െͳ͵͵Ǥ͹͵ߝ ଶ ൅ ͵͵ǤͶͻͷߝ ൅ ͲǤ͸ͳ͹ʹߝ
Epoxy
൐ ͲǤͲͳ
Holdtite 3295 ߪ ൌ െͳʹͶ͵Ͳߝ ଶ ൅ ͷʹͷǤʹͷߝ െ ͲǤͲͲ͸ͳߝ

0.41 219.5 619.0 0.002 ൑ ͲǤͲͳͷ
Acrylate ߪ ൌ െͺͺ͸Ǥ͹ͻߝ ଶ ൅ ͳͶ͹Ǥ͹ߝ ൅ ͵ǤͲͲͻߝ ൐ ͲǤͲͳͷ
ߪ ൌ െͳ͹Ǥ͸ͳʹߝ ସ ൅ ʹʹǤͲͲʹߝ ଷ െ ͳͲǤͲ͸ͻߝ ଶ
DC 993 Silicone 0.48 3.9 11.5 0.004 ൅ʹǤ͵ͷͳߝ ൅ ͲǤͲͲͲ͵
296
2.2. Analytical Model

Using material data from Table 1 above, stresses across the adhesive layer in a lap-shear
joint (Fig.3) can be predicted by the Crocombe and Bigwood linear elastic analytical
model‘[2]’. The model is based on a seventh order differential equation (Eq.2) and a
sixth order differential equation (Eq.3) that describes the shear stress xy and the
transverse stress ɐy distributed across the adhesive layer of a lap shear joint.
d 7W xy d 5W xy d 3W xy dW xy
K1 K3 K5 0 (2)
dx 7 dx 5 dx 3 dx
d 6V y d 4V y d 2V y
K1 K3 K 5V y 0 (3)
dx 6 dx 4 dx 2
where K1 to K5 are constants dependent on the shear modulus of the adhesive and the
elastic moduli of the adherends.
Crocombe and Bigwood created a spreadsheet that solves equations 2 and 3 if the
applied loads (P1, P2, V1, V2, M1 and M2) shown in Figure 3 above are known. The
determination of the bending moments M1 and M2 is relatively complex, in this paper,
values of M1 and M2 were obtained using an analytical derivation presented in ‘[21]’.
The graphical presentation of the Crocombe and Bigwood elastic stress distribution
across the 30mm length of the SLS joint of each candidate adhesive is shown in Figure
4 below. The stress distributions in Figure 4 correspond to the six mean failure loads P
obtained from SLS experimental tests of each candidate adhesive.
Figure 3: Free-body diagram of element dx along single-lap adhesive joint ‘[2]’
297
Challenging Glass 3
25
DP490 Epoxy
P=28.2kN
20 H3295
Adhesive Shear Stress (MPa)
Acrylate
P=20kN
A2047
15 Acrylic
P=15.3kN
3M7271
10 Hybrid
P=7.8kN
3M2216
Epoxy
5 P=7.3kN
DC993
Silicone
0 P=1.6kN
0 5 10 15 20 25 30
Adhesive Overlap Length (mm)
Figure 4: Bigwood and Crocombe shear stress distribution plots across SLS adhesive joints
2.3. Numerical Model

The steel-glass SLS test was constructed as a 2-dimensional FEA model using LUSAS
v14.3 ‘[17]’. An eight node quadrilateral quadratic plane strain element type was
implemented throughout the model. Since the SLS joint is symmetrical about the
midpoint of the glass (line y-y in Fig.5), only half of the connection is modeled.
y ½ glass plate steel plate

x
y A B
‫ݔ‬ሶ
C D
Adhesive with dense mesh

of 12 elements in thickness
y
Figure 5: Two dimensional FEA model of the steel-glass adhesive SLS connection.
For the boundary conditions, the model was restrained in x and Mz along the symmetry
line y-y and also restrained in y and Mz along the lines AB and CD which represent the
contact of the steel plate to the testing machine. The glass and steel were modeled as
perfectly linear elastic materials with Eglass=70GPa, glass=0.23, Esteel=209GPa and
steel=0.3. The six candidate adhesives were modeled with visco-elastic and elasto-
plastic properties obtained from Table 1. A velocity ‫ݔ‬ሶ was applied at the end of the steel
plate, line BD and it corresponds to the experimental displacement rate of 0.2mm/min.
The analysis was run as a dynamic geometric and material non-linear analysis using the
implicit method and an updated langrangian approach. The results of the numerical
analysis were compared to the analytical and experimental test results in Section 4.
298
In addition to the non-linear analysis described above, a simple linear elastic analysis
was performed. The adhesives were modeled with linear elastic-perfectly plastic
material properties obtained from the time-independent elasto-plastic stress-strain plots
(Fig.2b).
3. Steel-Glass Single-Lap Shear Joint Tests
3.1. Specimen Preparation

Specially adapted SLS adhesive joints (Fig.6a,b) of the six candidate adhesives were
prepared based on ASTM D1002-99 ‘[18]’ guidelines. Each specimen was assembled
using two 120mm by 49mm by 6mm thick black mild steel plates and one 200mm by
200mm by 10mm thick fully toughened glass plate manufactured to BS EN12150-2
‘[22]’ standards. The bonding surface of the steel was sanded using a 220 grit sandpaper
to give a consistent finish for all specimens. All steel and glass surfaces were
thoroughly cleaned with acetone before adhesive application and in the case of the
silicone joints; a siloxane based primer was applied to the steel surface before bonding
the adherends. A specially machined aluminium jig lined with a PTFE release film was
used for assembling all specimens; the jig ensured alignment of the two steel plates as
well as maintaining a bond thickness of 3mm for all specimens. All joints had a bond
width of 49mm and a bond length of 30mm. All the six candidate adhesives were two-
part pot adhesives and mixing conformed to each of the manufacturer’s guidelines. The
prepared specimens were stored at ambient temperature and approximately 40% relative
humidity.
3.2. Test Procedure

The SLS tests were performed on an Instron 5500R testing machine with a 150kN load
cell. Specimens were attached to the testing machine by slotting steel pins into 12mm
holes which are 20mm from the end of each steel plate. A displacement gauge was
attached to each steel plate with the gauge probe resting on an aluminium plate glued to
the centre of the glass plate (Fig.6b). The displacement gauges separately measured the
vertical displacement in each adhesive joint. All tests were displacement controlled and
a displacement rate of 0.2mm/min was applied for all specimens except for the very
flexible silicone DC993 specimens which were tested at 1.0mm/min. Photographs were
taken before, during and after each test. The tests were divided into two phases.
Phase 1: Three specimens of each of the six candidate adhesives were tested at 210C.
Phase 2: Three specimens of the best-performing adhesives from Phase 1 were heat-
soaked in an oven at 800C for 48hrs and then tested at 210C.
299
Challenging Glass 3
End of steel plate

attached to Instron
49mm wide steel
49mm wide adhesive
Displacement gauge
attachment
Aluminium T-plate for
resting displacement gauge
200mm wide glass
Displacement gauge
Figure 6: (a) Schematic drawing and (b) photograph of the SLS test set up
3.3. Test Results

Phase 1: All DC993 silicone specimens failed cohesively with very low failure loads but
large extensions. Although the failure mode is desirable and the adhesive showed
remarkable flexibility, the calculated mean shear strength of only 1.1MPa makes the
silicone unsuitable for the steel-glass composite system being developed. All 3M
DP490 epoxy joints experienced glass failure preceded by brittle partial failure in the
adhesive. There was no plastic deformation observed (Fig.7) and the DP490 joints
experienced the highest failure loads of up to 28kN. The lack of flexibility and lack of
plastic deformation before failure makes DP490 unsuitable. The 3M 7271
epoxy/acrylate hybrid joints experienced very small strains before failure. Glass failure
at relatively low loads and very low extensions was observed in all specimens. Holdtite
3295 acrylate specimens carried significantly high loads and were relatively flexible.
The adhesive exhibited considerable plastic deformation (Fig.7) before local glass
failure on the glued glass edges. The 3M 2216 epoxy specimens showed good flexibility
but the load carrying capacity was relatively low with maximum loads of only 7.3kN.
Adhesion failure at the steel-adhesive interface was observed in all 3M 2216 specimens,
this seemed to suggest premature joint failure. This observation could be attributed to
potential inadequate surface preparation or to large bond thickness since previous
studies ‘[14]’ have shown that this epoxy predominantly fails cohesively. The Araldite
A2047 acrylate showed the best results. All specimens failed cohesively after
substantial plastic deformation. The joints were relatively more flexible yet they carried
significantly high loads of up to 15.3kN.
Phase 2: The 3M 2216 epoxy, the Araldite A2047 acrylate and the Holdtite 3295
acrylate adhesives were selected for SLS tests after heat soaking at 800C for 48hrs. At
the time of writing this report, test results for the Holdtite 3295 acrylate were not yet
available. Results for the 3M 2216 epoxy and Araldite A2047 showed that both
adhesives became less stiff (Fig.7). The 3M 2216 joints performed poorly; failing to
carry loads above 1kN. The Araldite A2047 joints still performed exceptionally well.
The mean load bearing capacity of the joints reduced by only 18% after exposure to
800C (Fig.7). Failure was still cohesive with substantial plastic deformation in the
adhesive occurring prior to failure. Table 2 below summarises the test results.
300
30
Araldite A2047
Acrylate
25
DP490 Epoxy
20 Holdtite H3295
acrylate
3M7271
15
Load (kN)
Epoxy/Acrylate
3M2216 Epoxy
10
A2047 Acrylate
after heat soaking
5
3M2216 Epoxy
after heat soaking
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Displacement (mm)
Figure 7: Typical SLS test load vs. extension curves for five candidate adhesives
(DC993 Silicone is excluded for clarity)
301
Challenging Glass 3
Table 2: Summary of experimental results

Extension Mean
Mean Failure at Shear
Load Mode of failure
failure Strength a
(kN)
(mm) (MPa)
Glass failure, no plastic strain in
3M DP490 Epoxy 28.2 0.56 19.2 b
the adhesive prior to failure
Cohesive failure preceded by
1.04
Araldite 2047 Acrylic 15.3 (12.5*) 10.4 (8.5*) substantial adhesive plastic
(1.28*)
strain
3M 7271 Glass failure, no plastic strain in

7.8 0.20 5.3 b
Epxy/Acylic the adhesive prior to failure
0.85 Adhesion failure at the steel-

3M 2216 B/A Epoxy 7.3 (0.7*) 5.0 (0.5*)
(1.71*) adhesive interface
Glass failure preceded by
Holdtite 3295 Acrylic 20.0 1.25 13.6 b significant adhesive plastic
strain
Cohesive failure preceded by
DC 993 Silicone 1.6 5.61 1.1
high adhesive plastic strain
*
specimens subjected to 800C for 48hrs before testing
a
based on equivalent constant shear stress along the lap joint and loading is short term
b
adhesive shear strength governed by glass failure
4. Discussion
Comparison of the linear elastic analytical Bigwood and Crocombe model to both linear
elastic and non-linear viscoelastic-plastic FEA models (Fig.8) generally reveal that there
is good agreement particularly for small loads. This is not surprising since the linear-
elastic analytical model by definition should predict the elastic deformation which is
predominant at small strains. As the failure load is approached and strains become
larger, the analytical model tends to underestimate the adhesive shear stress (Fig.8) by
magnitudes of up to 16.7%. This is uncharacteristic since elastic stresses are expected
to be larger than elasto-plastic stresses. Unlike the linear elastic FEA model, the non-
linear FEA model does not show prominent stress peaks near the joint edges at high
loads, this is due to prediction of plasticity in the non-linear model.
302
14
Linear FEA
P=15.3kN
12 Nonlinear FEA
P=15.3
Crocombe
10 P=15.3kN
Shear Stress (MPa)
Nonlinear FEA
8 P=10.2
Linear FEA
P=10.2kN
6
Crocombe
P=10.2kN
4 Nonlinear FEA
P=5.1kN
Linear FEA
2 P=5.1kN
Crocombe
P=5.2kN
0
0 5 10 15 20 25 30
Adhesive overlap length (mm)
Figure 8: Analytical and numerical shear stress distribution

across the Araldite A2047 acrylate SLS adhesive joint
Despite the limitations of the analytical model, experimental mean shear strengths of the
adhesives (Table.2) show very good agreement with the Bigwood and Crocombe’s
predictions of adhesive shear stress at failure load (Fig.4).
For most of the candidate adhesives, numerical results for load vs. extension generally
showed good agreement at low strains (Fig.9). As strains in the SLS joint increased,
numerical models showed varying degrees of accuracy for different adhesives. The
accuracy of the non-linear model seemed to heavily depend on the value of the decay
constant . Mathematical analysis of experimentally obtained shear modulus vs. time
plots suggested that the shear modulus history of the tested adhesives G(t) is best
described by a logarithmic function (Eq.4) below.
G (t ) D ln t J (4)
where and are constants. However the FEA constitutive model used in this study
describes G(t) by an exponential function (Eq.1). As a result, an estimation of was
required to allow Eq.1 to approximate Eq.4 and this invariably limited the accuracy of
the constitutive model.
The predictive capability of the models is also dependent on the adhesive failure mode;
lap shear joints that experience adhesion or glass failure tend to be predicted poorly
compared to those that fail cohesively. In the case of the 3M 2216 Epoxy adhesive
303
Challenging Glass 3
joints, for example, where failure was by adhesion at the steel-adhesive interface, the
numerical predictions were unsatisfactory.
16
14
12
10
Load (kN)
8 FEA: Linear elastic and perfectly

plastic constitutive model
6
FEA: Non-linear viscoelastic and
elasto-plastic constitutive model
4
Experimental
2
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Extension (mm)
Figure 9: Experimental and numerical load vs. extension plots for Araldite A2047
acrylate SLS adhesive joint
5. Conclusion
The main objective of this paper was to identify suitable structural adhesives for a steel-
glass linearly bonded system. SLS tests on the six candidate adhesives provided
significant information which indicated that at least three of the six adhesives may be
suitable for the proposed system.
Out of all the tested adhesives, it was concluded that the Araldite A2047 acrylate SLS
joints exhibit the most desirable mechanical characteristics under short-duration testing
conditions. The Araldite A2047 acrylate SLS joint:
x failed cohesively both in specimens tested before and after heat soaking at
800C for 48hrs
x showed good strength with mean shear strength of 10.4MPa
x exhibited relatively good flexibility with substantial plastic deformation
preceding failure
x was not significantly affected by exposure to extreme temperature, with a
maximum percentage drop in load bearing capacity of only 18%
x was relatively easy to prepare and handle.
Holdtite 3295 acrylate also performed well, the mean shear strength of 13.6MPa was
based on glass failure therefore it is likely that the adhesive is even stronger than this.
The Holdtite adhesive was relatively flexible and it experienced considerable plastic
strain before local glass failure. Although the 3M 2216 Epoxy showed poor adhesion to
304
the metal surface; its relatively good flexibility, significantly good strength, low cost
and relatively long handling time warrants its consideration for further investigation.
The other three adhesives were found to be unsuitable for different reasons ranging
from significantly low strength in the case of DC993 silicone to significant lack of
flexibility in the case of DP490 epoxy and 3M 7271 epoxy/acrylate hybrid. It must be
noted however that the elimination of some of these adhesives did not necessarily mean
they are not suitable for steel to glass connection; this study presented a specific bond
line thickness of 3mm, a constraint which limits the performance of some adhesives
which are otherwise suitable for bonding steel to glass.
The other objective of this paper was to validate the predictive capability of an
analytical model and a viscoelastic-plastic numerical model. It was shown that the
linear-elastic analytical model is useful in predicting adhesive joint behaviour at low
strains but the accuracy decreases as the adhesives start to experience plastic
deformation at large strains. It was also shown that the non-linear numerical model
provides reasonable predictions of stress distribution across adhesive joints but requires
good approximation of the adhesive shear modulus history function G(t). Research
aimed at improving the adhesive constitutive model by accounting for effects of
hysteresis and repeated cyclic loading is underway and results are due to be published in
the near future. It is hoped that these improved models can be used to predict the global
adhesive joint performance in full-scale steel-glass composite façade modules.
6. Acknowledgements
The study presented in this paper, which forms part of a broader ongoing research
aimed at developing a steel-glass composite façade system is funded by an Industrial
CASE studentship provided by the Engineering and Physical Sciences Research
Council (EPSRC) and a contribution from TATA Steel, the Industrial partner.
7. References
[1] Goland, M; Reissner, E, The stresses in cemented joints, J Appl Mech Trans ASME 66 (1944), Vol.11,
ppA17-A27.
[2] Bigwood, D.A; Crocombe, A.D, Elastic analysis and engineering design formulae for bonded joints, Int
J Adhes Adhes (1989), Vol.9, No.4, pp229-242.

[3] BS 6262-6: 2005, Code of practice for glazing for buildings – Part 6: Code of practice for special
applications, British Standards Institution BSI (October 2005).
[4] ATAG Nr 002, European Organisation for technical Approvals EOTA, Brussels (1998).
[5] AAMA CW-13-85, Structural sealant glazing systems, American Architectural Manufacturers
Association (AAMA), Schaumburg, USA (1985).
[6] ASTM C 1401-02, Standard guide for structural sealant glazing, ASTM Standards (2002).
[7] Haldimann, M; Luible, A; Overend, M, Structural use of glass, Structural Engineering Documents
SED10, International Association for Bridge and Structural Engineering IABSE, Zurich, May 2008.
[8] Moibob, D; Crisinel, M, Linear connection system for structural application of glass panels in fully-
transparent pavilions, Proceedings of Challenging Glass 1, Delft, Netherlands, 2008.
[9] Pye, A; Ledbetter, A, The selection of an adhesive for a glass-adhesive T-beam, Int J Adhes Adhes
(1998), Vol 18, pp159-165.
[10] Wellershoff, F; Sedlacek, G, Glued connections for new steel glass structures, Proceedings of Glass
Performance Days, Tampere, Finland, 2005.
[11] Louter, C; Veer, F; Hobbelman, G, reinforcing glass, effects of reinforcement geometry and bonding
technology, Proceedings of Glass Performance Days, Tampere, Finland, 2007.
[12] Weller, B; Schadow, T, Design of bonded joints in glass strucutres, Proceedings of Glass Performance
Days, Tampere, Finland, 2007.
[13] Belis, J; Van Hulle, A; Out, B; Bos, F; Callewaert, D; Poulis, H, Broad screening of adhesives for glass-
metal bonds, Proceedings of Glass Performance Days, Tampere, Finland, 2011.
305
Challenging Glass 3
[14] Overend, M; Jin, Q; Watson, J, The Selection and Performance of Adhesives for a Steel-Glass
Connection, Int J Adhes Adhes (2011), doi:10.1016/j.ijadhadh.2011.06.001.
[15] DC993 Product Data Sheet, http://www.geocel.co.uk/dynpdfs/416.pdf, http://www.dowcorning.com/,
Ref No. 62-0918H-01, Dow Corning Corporation, July 2001.
[16] Moibob, D, Glass panel under shear loading – use of glass envelopes for building stabilisation, PhD
Thesis, Swiss Federal Institute of Technology (EPFL), Lausanne, Switzerland, 2008.
[17] LUSAS Finite Element System, Lusas theory manual, FEA Ltd, UK, (2012).
[18] ASTM D 1002-99, Standard test method for apparent shear strength of single-lap-joint adhesively
bonded metal specimens by tension loading (Metal-to-metal), ASTM Standards (1999).
[19] BS EN ISO 527-1: 1996, Plastics – Determination of tensile properties – Part 1: General principles, pp.
1-16, British Standards Institution BSI (1996).
[20] BS EN ISO 527-2: 1996, Plastics – Determination of tensile properties – Part 2: Test conditions for
moulding and extrusion plastics, pp. 1-14, British Standards Institution BSI (1996).
[21] Cheng, S; Chen, D; Shi, Y, Analysis of adhesive-bonded joints with non-identical adherends, J
Engineering Mechanics (1991), Vol.117, No.3.
[22] BS EN ISO 12150-2: 2004, Glass in buildings: Thermally toughened soda-lime silicate safety glass,
evaluation of conformity/product standard, pp. 1-42, British Standards Institution BSI (2004).
306
doi:10.3233/978-1-61499-061-1-307
Load-Carrying Behaviour of Metal Inserts

Embedded in Laminated Glass
Kerstin Puller
Werner Sobek Stuttgart, Germany, kerstin.puller@wernersobek.com
IIT (Illinois Institute of Technology), Chicago, Illinois, USA
Werner Sobek
Werner Sobek Stuttgart, Germany, kerstin.puller@wernersobek.com
IIT (Illinois Institute of Technology), Chicago, Illinois, USA
ILEK (Institute of Lightweight Structures and Conceptual Design),
University of Stuttgart, Germany
The introduction of a new interlayer material for laminated glass – SentryGlas (SG)
– has enabled the development of an innovative glass connection technique in
which a metal element is embedded in the interlayer of the laminate and acts as a
load-carrying device. At the ILEK tensile tests on the interlayer material were
performed to determine the material properties of SG and to develop a numerical
material model. Pull-out tests of the proposed metal insert system were carried out
to study its load-bearing behaviour under tensile loading. A numerical simulation
was performed to model the pull-out tests and to verify the material models
developed. This paper focuses on the evaluation of the load transfer behaviour of
this connection under short-term tensile loading at various temperatures and the
simulated stress distribution within this multi-material system.
Keywords: Insert, SentryGlas, Connection technique, Laminated glass
1. Introduction
With the introduction of SentryGlas (SG), the company DuPont made available an
interlayer material with superior stiffness properties at room temperature compared to

previous interlayer products (PVB). This allowed the glass and also the laminate
thickness to be reduced. The main purpose of the interlayer material is to bond different
glass panes of the laminate in a planar manner, and, due to the ductility of the polymer
increase the residual load-bearing capacity of the composite compared to the individual
glass panes. But the increased stiffness and the ability of SG to adhere well to metal has
also presented the opportunity for a new and innovative glass connection technique.
This new connection technique uses a metal insert, which is partially embedded in the
interlayer material of the laminate, as load transferring element and as a connector to the
surrounding structure. First architectural applications of this connection technique are
the stair details of the Apple stores [1], the connection of the “Seele-staircase” presented
at the Glasstec 2006 [2] and the folding roof detail of the intermediate ceiling at the
Zürich swimming hall [3].
Research conducted at the ILEK (Institute of Lightweight Structures and Conceptual

Design, University of Stuttgart) studied the load transfer of inserts embedded in glass
307
Challenging Glass 3
laminates through experiments and numerical simulations, and analyzed the stress
distribution within the laminate in dependence of the insert geometry [4, 5]. Different
loading durations and various temperatures were considered. The present paper focuses
on the evaluation of the load transfer behaviour of this connection under short-term
tensile loading and the simulated stress distribution within the multi-material system.
2. Load-carrying behaviour of metal inserts embedded in laminated glass

In order to be able to describe the load carrying behaviour of metal inserts embedded in
laminated glass, the behaviour of the interlayer material itself (SG) was studied in
tensile tests using dumbbell-shaped specimens according to DIN ISO 527 [6]. Since the
mechanical behaviour of SG greatly varies with the surrounding temperature, the tests
were carried out at three temperatures (23 °C, 40 °C, 75 °C) which were found to occur
in glass laminates in central Europe [7]. Based on the experimental data, material
models for different temperatures and deformation rates were developed and used to
perform a numerical simulation with the finite element software Ansys 12.1. A detailed
description of the tensile tests performed, the material models and the numerical
simulation can be found in [4, 5, 8]. While the stress state in the dumbbell-shaped
specimens during the tensile tests was mainly uniaxial, a three-dimensional stress state
occurs in the laminate material of the proposed connection system (metal insert
embedded in the laminate). Thus, it was very important to verify that the material
models developed could also be used to simulate the load-carrying behaviour of the
proposed system. To do so, an experimental and numerical study was carried out.
2.1. Experimental investigations

A series of pull-out tests were performed on inserts embedded in glass laminates to
study the load-carrying behaviour of the connection system and to record the force-
displacement relationship. Figure1 and 2 show the test specimens.
Figure 1: Test specimen of pull-out test – front view. Figure 2: Test specimen detail of pull-out test – pers-
Metal insert (thickness = 1 mm) embedded in glass pective view. Metal insert (thickness = 1 mm)
laminate. embedded in glass laminate.
In order to minimize the influence of measurement errors the specimens were designed
so that the majority of the deformation occurred in the SG. Due to the fact that the
stiffness of SG varies significantly over the temperature range investigated, the test
specimens were dimensioned differently for the lower temperatures (23 °C, 40 °C) than
308
Load-Carrying Behaviour of Metal Inserts Embedded in Laminated Glass
for the higher temperature (75 °C). To avoid yielding of the insert under small
deformations, a high-strength steel (Domex 700 MC D) with a thickness of 4,56 mm
was used for the insert of the low temperature tests. The chosen thickness corresponds
to two times the largest foil thickness of SG. To prevent large eigenstresses in the
laminate, the large thickness SG-foils were cut out in the areas where the insert was
embedded. However, this procedure was not necessary for the high temperature test
specimens for which a stainless steel insert (product number according to [9]: 1.4301)
with a thickness of 1 mm could be used. The dimensions of the test specimens are listed
in Table 1 and a schematic drawing shows the test setup in Figure 3.
Table 1: Specimen dimensions in dependence of the test temperature.

Test temperature [°C]
23/40 75
Glass pane thickness [mm] 10 10
Glass pane width [mm] 400 400
Glass pane height [mm] 300 300
SG-thickness [mm] 6 x 2,28 2 x 1,52
Insert-thickness [mm] 4,56 1
Embedded length of the insert [mm] 50 50
Embedded width of the insert [mm] 25 25
Figure 3: Test setup for metal insert pull-out tests.
309
Challenging Glass 3
To avoid failure of the glass, supports were located close to the insert (support width
each: 70 mm, spacing between supports: 100 mm). The insert was directly clamped by
the lower jacks of the testing machine and a crosshead displacement rate of 1 mm/min
was applied for all tests. Two potentiometers on each side of the glass pane measured
the differential displacement between the insert and the glass pane. By averaging the
differential displacements of potentiometers on opposing glass pane sides rotational
displacements could be compensated. The recorded force-displacement relationships are
shown in Figure 4. Due to the fact that the stiffness of SG varies significantly over the
temperature range investigated, the data curves for the different temperatures can be
clearly distinguished: the stiffness of the system is reduced with a rise in temperature.
During the 23 °C pull-out tests, glass breakage occurred for two specimens around
26 kN. For another specimen, the insert started to slip out of the clamps due to an
insufficient clamping force. This reduced the applied load (at around 30 kN).
Subsequently the clamping force was increased and the test specimen was further
loaded.
Figure 4: Force-displacement relationship of metal inserts embedded in a SG-glass laminate, pull-out test,
specimen geometry according to Table 1, crosshead displacement rate: 1 mm/min,
test temperatures: 23 °C, 40 °C and 75 °C.
During the pull-out tests at 23 °C and 40 °C debonding of the metal insert and SG could
be observed at the insert end (Figure 5). The debonding typically started from the insert
face which was perpendicular to the loading direction. At 75 °C bubbles formed at the
insert end. The size of the bubbles increased with further applied deformation. After the
debonding/bubble formation occurred, the slope of the force-displacement relationship
310
starts to decrease. This effect is most evident for the 40 °C and 75 °C force-
displacement-curves and corresponds to a decrease of stiffness of the connection system
(consisting of the metal insert embedded in the glass laminate).
Figure 5: Photographs of the insert end with visible debonding or bubble formation, left: 23 °C (potentio-
meter displacement: 0,7 mm, force: 40 kN); middle: 40 °C (potentiometer displacement: 1,2 mm, force:
20 kN); right: 75 °C (potentiometer displacement: 0,8 mm, force: 2 kN).
2.2. Numerical simulation

To verify that the material models developed [4, 5, 8] can be used to simulate the load-
carrying behaviour of the proposed system (metal insert embedded in the laminate), a
numerical simulation of the pull-out tests was performed using the finite element
software Ansys 12.1. One main aspect of the numerical simulation was to incorporate
an iterative scheme adjusting the strain-rate distinct SG material model to the strain rate
which occurred during the experiments. The iterative scheme is explained in detail in [4,
5]. Figure 6 shows the simulated force-displacement curves together with the
experimental results.
Figure 6: Force-displacement relationship of metal inserts embedded in a SG-glass laminate: experimental

and numerical results, pull-out test, specimen geometry according to Table 1, crosshead displacement rate:
1 mm/min, test temperatures: 23 °C, 40 °C and 75 °C; Multi-linear SG-material model with von-Mises-yield
condition, enhanced-strain element formulation
311
Challenging Glass 3
In general, the results of the numerical simulation fit the experimental results quite well.
The largest discrepancy between the curves occurs for 40 °C. One reason for this
discrepancy may be the proximity to SG’s glass transition temperature at which its
stiffness significantly drops. However, due to the good agreement of the experimental
and numerical results the applied numerical procedure and the material models
developed can be used to simulate the load-carrying behaviour of metal inserts
embedded in glass-laminates. Since no failure models were integrated in the analysis,
the numerical model cannot estimate failure loads but can instead be used to evaluate
the stress-strain distribution within the multi-material system.
3. Stress distribution within an insert connection

After the numerical results showed a good coincidence with the experimental data, the
numerical model could be used to evaluate the stress-strain distribution within the
multi-material system. With a view to suspended glass panes in architecture, the
boundary conditions of the numerical model were adjusted: instead of modelling the
experimental setup with supports close to the insert, the insert itself was supported and a
line load applied to the lower glass pane edge (total line load: 1 kN; Figure 7). The
insert and the laminate dimensions were identical with the dimensions of the 75 °C-pull-
out test specimens (Table 1).
Figure 7: Support and loading conditions of the numerical model evaluating the stress distribution within an
insert connection (insert embedded in laminate); schematic drawing.
Figure 8 shows the stress distribution on the exterior and the interior glass pane face, in
the SG and within the stainless steel insert. Since tensile stresses lead to glass failure
and debonding of the SG, the first principle stresses are evaluated for glass and SG.
With metal yielding being dependent on the equivalent von-Mises-stress, this stress
distribution is shown for the stainless steel insert. All numerical investigations were
based on SG-material models for 23 °C and 75 °C with a strain rate of 0,0059 s-1 (This
strain rate is equivalent to the observed strain rate during the pull-out tests with the
identical specimen dimensions (75 °C test series)).
The numerical analysis shows that the load-bearing mechanism of the proposed
connection comprises two components: load is transferred over the embedded surface
area of the metal insert, and at the insert end. The force transferred by each component
depends on the relative stiffnesses of the different elements in the multi-material system.
Around 23 °C SG is relatively stiff, and most of the force at this temperature is
312
transferred over the embedded surface area. This results in high tensile stresses at the
upper glass pane edge. At 75 °C, the decreased stiffness of SG results in a relatively
high tensile force at the insert end which generates high local tensile stresses.
Figure 8: Stress distribution within an insert connection at 23 °C and 75 °C: first principle stresses at the ex-
terior and interior glass pane face, and within the SG, von-Mises equivalent stress in the stainless steel insert
(product number according to [9]: 1.4301), rectangular insert geometry (embedded length: 50 mm, embedded
width: 25 mm), total applied line load on the lower glass pane edge: 1 kN, multi-linear SG-material model
with von-Mises-yield condition, strain rate: 0,0059 s-1, enhanced-strain element formulation.
313
Challenging Glass 3
4. Conclusion and Outlook

The experimental tests with the proposed system show that a significant load can be
transferred through this type of connection. But the magnitude of load and the stiffness
of the system strongly depend on the polymer temperature and the loading rate.
Nevertheless, the results of the experiments are very promising. The force-displacement
relationships of the performed numerical simulations agree well with the experimental
data. Thus, the applied numerical procedure and the material models are suitable to
simulate the stress-strain distribution within an insert connection. Subsequently, the
stress distribution of a suspended glass pane with an insert connection was evaluated.
The evaluation showed that the load transfer of the insert consists of two components:
load is transferred over the embedded surface area of the metal insert, and at the insert
end. The amount of force transferred by each component depends on the relative
stiffnesses of the different elements in the multi-material system. Around 23 °C SG is
rather stiff, and most of the force at this temperature is transferred over the embedded
surface area. At 75 °C, the decreased stiffness of SG results in a relatively high tensile
force at the insert end. This load transfer generates for 23 °C high tensile stresses at the
upper glass pane edge, while for 75 °C high tensile stresses occur at the insert end.
An approach followed in [4, 5] uses the insight of the stiffness dependent load transfer
to optimize the load-carrying behaviour of the insert-laminate-system. In that approach
the insert material is systematically relocated to reduce stress concentrations in the glass
and to homogenize the stress distribution within the multi-material system. With an
adjusted insert geometry, the stress concentrations can be significantly reduced under
the same applied load and maintained embedded metal area.
5. Acknowledgements
The research was funded by the Federal Office for Building and Regional Planning,
Germany (Reference number: Z6-10.08.18.7-08.24/II2-F20-08-21).
The authors would also like to thank the members of the ILEK and the sponsoring
companies (seele sedak, DuPontTM, Hottinger Baldwin Messtechnik) for their support.
6. References
[1] OĆallaghan, James, A case study of the Apple computer stores - glass structures 2001 - 2005,
Proceedings of the Glass Processing Days, Tampere, Finland, 2005.
[2] Peters, Stefan; Fuchs, Andreas; Knippers, Jan; Behling, Stefan, Ganzglastreppe mit transparenten SGP-
Klebeverbindungen - Konstruktion und statische Berechnung, Stahlbau, 3/2007, pp. 151–156.
[3] Willareth, Philippe; Meyer, Daniel, A new folding glass roof for the historic city swimming hall Zürich,
Proceedings of the Glass Performance Days, Tampere, Finland, 2011.
[4] Puller, Kerstin, Untersuchung des Tragverhaltens von in die Zwischenschicht von Verbundglas
integrierten Lasteinleitungselementen, Phd-Thesis, ILEK, University of Stuttgart, expected: 2012.
[5] Puller, Kerstin; Denonville, Jürgen; Sobek, Werner, Hochleistungsfähige, materialminimale und
werkstoffgerechte Verbindungstechnik im Glasbau, Research Report, Forschungsinitiative: ZukunftBau,
Research project number: Z6- 10.08.18.7-08.24, ILEK, University of Stuttgart, expected: 2012.
[6] DIN EN ISO 527-2 – Kunststoffe: Bestimmung der Zugeigenschaften, Teil 2: Prüfbedingungen für
Form- und Extrusionsmassen, 1996:07, Beuth, Berlin, Germany, 1996.
[7] Sobek, Werner; Haase, Walter, Temperaturversuche an Verbundsicherheitsglasscheiben unter
Sonneneinstrahlung, Research Report, Zentrallabor des Konstruktiven Ingenieurbaus (now part of the
ILEK), Universität Stuttgart, 2001.
[8] Puller, Kerstin; Denonville, Jürgen; Sobek, Werner, An Innovative Glass Connection Technique Using
an Ionomer Interlayer, Proceedings of the Glass Performance Days, Tampere, Finland, 2011.
[9] DIN EN 10088-1 – Nichtrostende Stähle, Teil 1: Verzeichnis der nichtrostenden Stähle, 2005:09, Beuth,
Berlin, 2005.
314
doi:10.3233/978-1-61499-061-1-315
Direct Glass Fabrication –

New Applications of Glass
with Additive Processes
Lisa Rammig
HS-OWL, Detmold School of Architecture, Germany
Inhabit Europe Ltd., London, UK, lisa.rammig@inhabitgroup.com,
www.inhabitgroup.com
Direct fabrication technologies for different materials have been developed to great
detail, yet for glass, one of the most aesthetically pleasing materials known,
developments are still in the early stages. By examining material properties,
common technologies and analysing the parameters governing the direct fabrication
of glass the following conclusions can be made; Fabrication technologies are
available but must be adapted for the properties of glass. This case study which was
elaborated as part of a master thesis, tries to show these fabrication abilities by
moving away from computer-controlled additive processes to a manual process,
derived from conventional welding techniques used in the chemical industry.
Keywords: Glass-Glass connections, Case study, Welding
1. Vision
Digital technologies largely determine contemporary lives. In different sectors their
importance´s grow further and further. Starting in the 1970’s with the first personal
computers these technologies significantly influenced the functions of economy and
society. Computer technologies give the possibility of mass production processes, where
manual labour is not required and accuracy and therefore optimisation can be
guaranteed. The development of digital fabrication moves from mass production to a
more customized and individualized fabrication. It is possible to create the perfect
product without reverting to single piece - handcraft - manufacturing. The customer
becomes a part of the production again; the product grows to their taste.
Whilst the direct fabrication of materials such as plastics or metals are now a
sophisticated technology, the fabrication of glass, perhaps one of the most fascinating
building materials, is almost unexplored. Glass is strong but brittle, heavy yet looks
lightweight, and it is transparent. These properties have made glass an important
component of our built environment today. Several types of glass such as laminated and
insulated glazing, coated, curved, and free-formed glass panes have been developed
rapidly after the float glass process was developed. Besides this progress in glass
production, constructions have been further developed to increase a transparent
architectural appearance. Current architecture is heavily influenced by digital media and
modelling software, giving us the possibility to create almost anything. In turn, this has
developed the need for glass of an overall higher performance. Free formed glass panes,
315
Challenging Glass 3
each one different to the other, might fit perfectly together, but are singularly produced.
At this point the question arises; why not use additive fabrication methods for glass
production? It would give us the possibility to produce free-formed transparent building
parts, without the complexity of the classical steps of production.
In this case study, the opportunities and limitations of such a technology have been
made with respect to material properties, current technologies, design parameters, recent
developments, practical testing and possible areas of further research.
Figure 1: Glass processing: Scheme with integration of glass-glass connections and possible applications
2. Method
2.1. Glass the material

Glass can be described physically as a rigid inorganic silicate product of amorphous
nature. In contrast to crystalline materials it is characterised by its isotropy, which
means that all properties, or measured values, are the same in each direction of the
structure.
Conventional glass such as Lime-soda glass consists of silica sand, lime and soda, all of
which are natural raw materials. By the addition of other materials, properties such as
stiffness and colour can be changed.
Compared to other substances like metal or plastics, which can be defined by their
chemical affinity, glass is described by its structural composition, which must be seen
316
Direct Glass Fabrication – New Applications of Glass with Additive Processes
irrespectively of the chemical composition. The advantageous properties of glass such

as light transmittance, thermal behaviour and solidness result from its structure.
The composition of glass is described by its molecular structure, in which a freezing

process is required to change glass from a liquid to a hardened state. Silica glass has the
simplest structure; consisting of only silicon dioxide. Whilst the structure of a silica
crystal is sorted [Fig 1], the atomic composition in its glass state are unsorted, therefore
it is not a crystalline structure. The material seems to be ‘liquid’, but does not tend to
creep or flow.
The high proportion of silica sand [around 75 %] expresses the hardness, strength, and
brittleness of the material. A brittle material fails or breaks after minimal deformation.
The composition of the glass is a fundamental factor of the viscosity of the melt and the
properties of rigid glass. As described, pure Silica glass tends to crystallize but has a
very low coefficient of linear thermal expansion, which makes it resistant to
temperature differences and thermal shocks. To lower the melting point of 1700°C for
silica glass during the production process, Alkali is added to the melt. This enhances the
coefficient of linear thermal expansion of the rigid material such that it can be melted at
a lower temperature.
Table 1: Glass compositions

SiO2 Na2O CaO MgO Al2O3 B2O3
Silica Glass 100
Soda lime Glass 72 14 8 4 2
Borosilicate Glass 80 4 1 3 12
Table 2: Comparison of main properties of lime-soda and borosilicate glass

Soda lime Glass Borosilicate Glass
3
Density [kg/m ] 2490 2230
Scratch hardness on the Mohs hardness scale 6-7 4.5

-6 -
Coefficient of mean linear expansion 10 [K ] (20-300 °C) 8.4 3.3
2
Thermal conductivity [W/m K] 450 x 30 4500
Softening point [°C] 710-735 825
Processing temperature [°C] 1015-1045 1260
2
Modulus of elasticity E [N/mm ] 70000 63000
Poisson Ratio 0.2 0.2
2
Bending Strength [N/mm ] 30 30
2
Compressive Strength [N/mm ] 700-900 700-900
2
Tensile strength [N/mm ] 30-80 70
(at constant load) 7 7
Maximal thermal shock resistance Tmax [K] 68.02 192.4
317
Challenging Glass 3
2.2. Technical Parameters

This section gives a brief overview of the technical parameters governing the
development of Direct Glass Fabrication (DGF):
x The coefficient of linear thermal expansion is an important value for the

applications of glass; especially for the analysis of the material with respect to
rapid manufacturing processes.
x The theoretical strength of the glass expresses the stiffness as a result of the
bond strength of individual components and assumes a crack-free glass. This is
an unrealistic assumption as small cracks occur during the production.
x The practical strength takes into account such cracks and is therefore
significantly lower. This strength value is crucial for the dimensioning of
components.
Borosilicate glass is primarily used in chemical and pharmaceutical industries because it

offers better chemical resistance, and a lower coefficient of linear thermal expansion.
This significantly enhances its resistance to large temperature differences induced by
the additive fabrication process. As such, Borosilicate glass can be used as display
materials for windows, fireplaces, and fire protection glass due to its good thermal
properties.
Another important factor is the viscous performance of the material. When a crystalline
substance gets heated, it will melt at a certain temperature. Above this temperature, the
substance will be liquid. In contrast to that, glass does not have a melting point; it
softens continuously under the influence of temperature. The organized atomic structure
of a crystal leads to a hard solid material, whilst glass on the other hand, has an irregular
structure and therefore behaves like a cooled liquid. For a DGF addition on a usually
fabricated glass sheet, the whole pane of glass must be heated to a uniform temperature,
which is limited by the thermal shock resistance of the glass (Tmax =192.4 K) this is
typically 370°C for borosilicate glass.
Figure 1: Ratio of Volume and Temperature showing the viscous performance
318
3. Practical Application
Different types of connections were manufactured to compare and evaluate atomic
bonding. Firstly, the behaviour of glass-glass connections was tested and the degree of
precision as to which they could be made. The applications of layer-by-layer fabricating
were then completed to simulate a ‘real’ additive process.
The testing of the welded connections showed that the failure of the material in most
cases occurs within the region of the original material, confirming that the connections
are stiffer than the raw material. This verifies that it is possible to manufacture glass
with the additive processes to generate strong connections.
Figure 2a and b: First experiments of layer by layer application
This first step demonstrated that glass-glass connections can be produced manually but
furthermore it had to be proved that a layer-by-layer production is possible. This was
done by heating a thin glass rod, elongating it to a thin wire, and then welding it on to a
pane of glass. To verify that the connected layers have a constant high temperature, such
that bonding with the next layer is possible.
The question arises, how could such a production method be used in architecture or the
façade industry and what products could it possibly develop?
To develop applications from basic production principles, a glass point fixing was
designed (Fig.3+4) which can be attached to the glass pane by an additive process. This
was then fabricated and tested for numerous connections. The point fixings would be
mounted to the sub construction with a clamp, which gathers the loads via friction.
319
Challenging Glass 3
Figure 3: fracture mechanical testing of connections Figure 4: conchoidal fracture
When compared to conventional point fixations, one of the main advantages is that the
quantity and size of fixings can be determined individually for each pane. Attachments
are optimized based on the dead load, wind load, and other influences. The most
important advantage of the point fixing however, is the removal of the need to drill the
glass panes. Resulting in less stresses being induced and thermal losses are reduced, as
there is no thermal bridging. The penetration of the outer pane is avoided, which means
that the thermal separation of the facade is uninterrupted. This is of significant
importance for the northern European climate, especially with current energy savings
regulations and sustainability considerations when regarding energy resources.
Figure 5: Glass point fixing fabricated in a manual Figure 6: Scheme of point fixation via friction
additive process
320
4. Innovation
Direct fabrication technologies for different materials have been developed to great
detail, yet for glass, one of the most aesthetically pleasing materials known,
developments are still in the early stages. By examining material properties, common
technologies, and analysing the parameters governing the direct fabrication of glass, the
following conclusions can be made:
x Fabrication technologies are available but must be adapted for the properties of
glass. This case study research tries to show these fabrication abilities by
moving away from computer-controlled additive processes to a manual
process, derived from conventional welding techniques used in the chemical
industry.
x The major problem when dealing with larger panes of glass is that they have to
be heated carefully to avoid breakage of the material.
x Properties in which the behaviour of the glass can be modified are its viscosity,
the thermal linear expansion coefficient, and its resulting maximal thermal
shock resistance (Tmax)
x A proper annealing process has to be guaranteed, to avoid thermal stresses in
the connections.
For the fabrication of large products, which could not be built in the construction space
of rapid prototyping machines, special heating and cooling components are required to
provide the object with uniform heat. Partial heating of the glass would result in
material breakages caused by thermal stresses.
The developed point fixing technique highlights DGF´s applicability to facade design.
By producing a fixing which connects the glass panes to the primary facade without
penetration of the outer pane can lead to new opportunities in facade construction. It
would be possible to construct facades without thermal breaks and without the visual
disturbance of conventional point fixings. It has been shown that the
additive process of producing glass is successful, the results also encourage further
research to be undertaken within the field such that products can be developed which
are truly representative of such technological advances.
5. References
[1] Hopkinson, N; Hague, R.J.M.; Dickens, P.M; Rapid Manufacturing, John Wiley & Sons Ltd,
Chichester, 2006
[2] Neef, A.; Burmeister, K.; Krempl, S. Vom Personal Computer zum Personal Fabricator, Murmann,
Hamburg, 2005
[3] Lohmeyer, S. et al. Werkstoff Glas, Expert Verlag, Grafenau, 1979
[4] Wörner, J.D. ; Schneider, J.; Fink, A Glasbau, Springer, Berlin, 2001
[5] Petzold, A. ; Marusch, H. ; Schramm, B. Der Baustoff Glas, Verlag für Bauwesen Berlin, 1990
[6] Rammig,L. Direct Glass Fabrication-New applications of glass with additive Processes, HS-OWL, 2010
[7] Wiggington, M. Glas in der Architektur, Deutsche Verlags-Anstalt GmbH, München 2003
321
doi:10.3233/978-1-61499-061-1-323
Analytical Solutions for Detail Problems in

Structural Glazing
Matthias Seel, Geralt Siebert
Universität der Bundeswehr München, Germany, matthias.seel@unibw.de,
geralt.siebert@unibw.de, www.unibw.de/glasbau
Detail points play a crucial role in the calculation of stress in structural glazing.
Drill holes, point fixing by clamps or plane discs with drill holes and glued
connection are such detail points. The calculation of these details is usually done by
using the finite element analysis (FEA). But the scatter of the FEA results is very
big – which means that the quality can be evaluated as low. The FEA results
depend on the applied FEM software and the modeling by the user. Therefore, it is
always necessary to compare FEA-results with solutions of comparable problems
for excluding inadequate FEA models. In this paper, several analytical solutions for
detail problems in structural glazing are derived and presented. These analytical
solutions can be used for validation of point fixing systems (mechanical and glued).
Keywords: Detail points, Point fixing, Stress concentration, FEA validation, Plates
1. Introduction
The desire for more transparency in the architecture can be achieved by point fixings.
Due to this fact, punctual supports are popular among architects and builders. The profit
of transparency precludes the complex design.
The calculation of stress values (e.g by finite element analysis (FEA)) in the area of a
point fixing is sophisticated, if the results should be close to reality. Research activities
(e.g. ‘Beyer’ [1], ‘Siebert’ [2], ‘Seel’ [3]) on the mesh quality of different FEA-models
for point fixings show that stress and deformation values can be calculated more or less
accurately in such cases. It is possible, that the deviation from realistic values can be
more than 30 % without being identified as poor values.
The FEA software offers the engineer a good possibility to calculate complex
mechanical structures-like punctual supported structures-in a realistic way. Using
elaborate FEA software gives the user the feeling of having covered everything. The
FEA results depend on the applied FEA software and the modeling by the user.
Therefore, it is always necessary to exclude inadequate FEA models by validation with
solutions of comparable problems. A validation process can be done by testing or by
comparison with existing analytical or numerical solutions. In the field of structural
glazing, tests are used for validation of FEA results (e.g. with strain gauges and
displacement sensors). These tests are necessary to validate the results, but are at the
same time very time-consuming and require a lot of experience. Analytical solutions for
the validation of FEA results can be more suitable than complex tests.
323
Challenging Glass 3
The aim of this paper is to present analytic solutions for detail points in structural
glazing. These solutions can be used for a validation process in the field of FEA.
2. Analytic solutions for circular plates with and without a center hole
2.1. Theory
The bending surface of a thin, isotropic and elastic circular plate can be described by the
linear, partial inhomogeneous differential equation (ODE) of the fourth order:
§ w 2 w 1 ww 1 w 2 w · § w 2 w 1 ww 1 w 2 w · p(r , M )
''w ¨ 2 2 ¸¨ 2 ¸ (1)
¨ wr r wr r wM 2 ¸¨
w r 2
r Gr r wM 2¸
K
© ¹© ¹
with
w displacement in z-direction
r Radius
Angle
p(r, ) load function
K=Eh³/12/(1- ²) plate stiffness.
The detailed derivation of the Eq.(1) can be found in the relevant literature e.g.
‘Girkmann [4]’, ‘Timoshenko & Woinowsky-Krieger [5]’ and ‘Márkus [6], [7], [8]’.
Eq. (1) can be solved by using numerical or analytical approaches. In few cases it is
possible to get an exact solution. An exact solution must satisfy all boundary conditions
and Eq. (1). A circular plate with center hole and rotationally symmetric load p(r,) is
illustrated in figure 1. For such plates the equation can be solved. The presented
solutions here are based on analytic approaches and the classic plate theory of Kirchhoff.
In general, the solution for the unknown bending surface w consists of a homogeneous
wh and a particular function wp:
w (r , M ) wh (r ,M ) wp (r ,M ). (2)
If the displacement function w(r, ) of the plate is known, the internal forces (e.g.
moments) are determined by the following equations:
§ w2w § 1 ww 1 w 2 w · ·
mr K ¨ 2 P ¨ ¸¸ (3)
¨ wr ¨ r wr r 2 wM 2 ¸ ¸
© © ¹¹ ,
§ 1 ww 1 w 2 w w 2 w ·¸
mM K ¨ 2 P (4)
¨ r wr r wM 2 wr 2 ¸¹ ,
©
324
Analytical Solutions for Detail Problems in Structural Glazing
§ 1 w 2 w 1 ww ·
mr ,M mM , r (P 1) K ¨ ¸ (5)
¨ r wrwM r 2 wM ¸
© ¹.
Figure 1: Circular plate with center hole and rotationally symmetric load p(r,) (‘Seel [3]’)
The equations for the shear forces qr and q are not shown for clarity and limits of this
paper. The principal stresses H can be calculated by:
§ 2 ·
12 ¨ mr mM § mr mM · ¸
V H1/ H 2 z
3 ¨
r ¨¨ ¸¸ mr2,M ¸ (6)
h ¨ 2 © 2 ¹ ¸.
© ¹
The maximum/minimum stress in the cross-section is obtained by substituting z=h/2 in

Eq. (6) and the location of the maximum/minimum in the whole area is given by:
wV H 1 / H 2 wV H 1 / H 2
0, (7) 0. (8)
wr wM
325
Challenging Glass 3
Results of Eq. (2) can be determined by using adequate analytical approaches and by
consideration of boundary conditions (BC) as well as transition conditions (TC). In the
following sections analytical solutions (‘Seel’ [3]) are presented for some load cases.
2.2. Circular plates with center hole and rotationally symmetric load
For simple cases of circular plate there are some analytical solutions available in
literature e.g. ‘Márkus [6], [7], [8]’. ‘Beyer [1]’ presented a solution for partially
constant loaded circular plate with a center hole. In this section, an analytical solution
for partially loaded circular plate is presented ‘Seel [3]’. The plate with center hole and
linearly varying load p(r) is illustrated in figure 1. The displayed plate is divided into 4
areas (B). The origin is in the center of the circular plate. Area B-0 describes the center
hole, B-I the inner unloaded area next to the area B-0, B-II the loaded and B-III the
outer unload area. The plate is fixed by a simple support in z-direction at the outer circle
(r=a). In the following, the radius is replaced by a scaled radius:
r
U . (9)
a
The described problem can be solved by using the following analytic approaches for
each area:
wI ( U ) CI ,1 U 2 ln U CI , 2 U 2 CI ,3 ln U CI , 4 , (10)
a 4 U 4 ( p2 r1 p1r2 ) a 5 U 5 ( p1 p2 ) CII ,1 U
2
wII ( U )
64 K (r1 r2 ) 225 K (r1 r2 ) 2
§ 1· , (11)
C II ,3 U 2 ¨ ln U ¸
C II , 2 ln U © 2¹
C II , 4 ,
2
wIII ( U ) CIII ,1 U 2 ln U CIII , 2 U 2 C III ,3 ln U CIII , 4 . (12)
The approaches of the unload areas (B-I and B-III) have only a homogeneous solution
function wh,i presented by the integration constants Ci,j. Index i is the number of the area
and index j the number of the integration constants in the area i. The approaches for the
homogeneous solution wh,i are listed in ‘Girkmann [4]’ and ‘Márkus [6]’. The particular
solution wp,i is obtained by integrating the Eq. (1) by consideration of the load function
p(r). Each displacement function (Eq. 10 to 12) contains four integration constants Ci,j.
The integration constants are determined by the boundary and transition conditions. The
individual areas are linked via 4 transition conditions.
326
Between area B-I and B-II, for instance, the utilised transition conditions are:
wI wII , (13) mr , I mr , II , (14)
w w qr , I qr , II . (16)
wI wII , (15)
wr wr
For this case the boundary conditions at the hole edge are:
mr , I 0, (17) qr , I 0 , (18)
and two BC’s at the outer edge given by:
mr , III 0, (19) wIII 0. (20)
The 12 unknown integration constants Ci,j are calculated from linear equation system
with 12 equations. The 12 equation are based on the 4 boundary and 8 transition
conditions. Due to the size of the constants, they are not listed here.
With the known constants it is possible to calculate the displacement and stresses at
every point of the plate according to equations presented in section 2.1. The course of
principal stress and displacement are illustrated in figure 2 for example 1.
a: principal stress H b: displacement w()

Figure 2: Solution for a circular plate with center hole and rotationally symmetric load p(r) (see fig. 1)
327
Challenging Glass 3
2.3. Circular plates with center hole and rotationally antisymmetric load
A circular plate with center hole under a rotationally antisymmetric load is pictured in
figure 3. This load case presents a moment. A moment is transferred by mechanical and
glued point fixings. The plate is separated into four areas (see section 2.2.). In contrast
to a circular plate with a rotationally symmetric load the displacement and internal
forces are not constant over the angle for a rotationally antisymmetric load. The load
varies along an arc according to a cosine function. Because of the cosine load-function
the system behavior of the circular plate will change as cosinusoidal.
Figure 3: Circular plate with center hole and rotationally antisymmetric load p(r,) (‘Seel [3]’)
The following analytic approaches with a cosine function are used to solve Eq. (1) for
this problem:
§ C ·
wI ( U , M ) ¨¨ C I ,1 U ln U C I , 2 U 3 I ,3 C I ,4 U ¸¸ cos M , (21)
© U ¹
328
§ a4U 4 § r p 2 p1 · a 4 U 5 ( p 2 p1 ) ·
¨ ¨¨ p1 1 ¸¸ ¸
¨ 45 K © a r2 a r1 a ¹ 192 K r2 a r1 a ¸
wII ( U , M ) ¨ ¸ cos M (22)
¨ C II , 4 ¸
¨ C II ,1 U C II , 2 U C II ,3 U ln U U
3
¸
© ¹
§ C ·
wIII ( U , M ) ¨¨ C III ,1 U ln U C III , 2 U 3 III ,3 C III , 4 U ¸¸ cos M . (23)
© U ¹
The boundary and transition conditions for the problem pictured in figure 3 are the same
as in section 2.2 with one exception. At the edge there are three boundary conditions
(shear force, bending and twisting moment), but the plate theory of Kirchhoff requires
only two boundary conditions. The twisting moment mr, and the shear force qr have to
be combined to an equivalent shear force q*r at the free edges:
§ 1 GmrM ·
q*r r¨¨ q r ¸
¸ 0. (24)
© r GM ¹
The three BC’s are reduced to two BC’s with the help of Eq. (24). The integration
constants Ci,j are obtained by solving a set of 12 equations, which consider all boundary
and transition conditions. The integration constants are not listed here because of their
huge size. The internal forces, stresses and displacement can be calculated by using the
integrations constants, analytic approaches and equations in section 2.1.
a: principal stress H1 b: displacement w(,)

Figure 4: Solution for a circular plate with center hole and rotationally antisymmetric load p(r,) (see fig. 3)
329
Challenging Glass 3
2.4. Circular plates under rotationally symmetric load and antisymmetric load
Eq. (1) can be solved for circular plates without a center hole. The solution approaches
and transition conditions for rotationally symmetric/rotationally antisymmetric load are
the same as in section 2.2/2.3. Only the boundary conditions in area B-I are different.
The boundary conditions in the middle ( =0) of the circular plate are:
wwI wI z f . (26)
zf, (25)
wU
From these equations the integration constants can be obtained as:
CI ,1 CI ,3 0 . (27)
The remaining integration constants Ci,j can be determined in the same way as described
in section 2.2 and 2.3. The analytical solutions for these cases are described in detail in
‘Seel [3]’.
Figure 5: Circular plate without center hole and rotationally symmetric load p(r,) (‘Seel [3]’)
3. Comparison of Results
3.1. Introduction
In this chapter calculations are carried out to compare the results of the analytic
approaches with the Finite Element Analysis (FEA) results.
The material data from table 1 are the same for all examples. The geometry data for the
circular plate with center hole (example 1) are listed in table 2.
The used FEA-model for the comparison consists of thin 4-node shell elements with
128 approximately square elements around the center hole. This FEA-model is in
accordance with the requirements of ‘Siebert [2]’ for an adequate mesh around a center
hole. ‘Siebert [2]’ showed that the mesh quality around the center hole has an important
influence on stress value.
330
Table 1: Circular plate with and without center hole – material data
Parameter Symbol Unit Value
Young’s moduls E [N/mm²] 70000
Poisson’s ratio [-] 0.23
Material behavior - - linear elastic (Hook)
Table 2: Circular plate with center hole (example 1) – geometry data

Parameter Symbol Unit Value
thickness h [mm] 10
radius of center hole r0 [mm] 10
radius of p1 r1 [mm] 15.2
radius of p2 r2 [mm] 30
support radius a [mm] 60
3.2. Circular plates with center hole and rotationally symmetric load
The load values for example 1-sym. are given in table 3. Figure 1 shows the static
system of this example.
Table 3: Example 1-sym. - rotationally symmetric load

Load-parameter Symbol Unit Value
load at r1 p1 [N/mm] 2
load at r2 p2 [N/mm] 1
The analytic results (principal stress and displacement) for this example are shown in
figure 2. The results (maximum values) for the two calculation methods are listed in
table 4. The maximum deviation between the two methods is 0.05% at the hole. Figure
6 shows the course of the principal stress deviation over the radius r. The minimal
deviations can based on rounding and numerical errors. The results show that the two
approaches provide approximately the same results.
Table 4: results - rotationally symmetric load

Parameter Symbol Unit FEA Analytic FEA /
Solution Analytic Solution
Principal stress, max. H1 [N/mm²] 46.297 46.318 99.95 %
Displacement, max. w [mm] 0.0841 0.0841 100 %
331
Challenging Glass 3
Figure 6: Comparison FEA with analytic solution for a circular plate with center hole and symmetric load
3.3. Circular plates with center hole and rotationally antisymmetric load
The rotationally antisymmetric load data for the circular plate center hole (see figure 3)
are listed in table 5.
Table 5: Example 1-anti-sym. - rotationally antisymmetric load

Load-parameter Symbol Unit Value
load at r1 and =0° p1 [N/mm] 5
load at r2 and =0° p2 [N/mm] 10
The maximum displacements and principal stresses of each method are presented in
table 6 and figure 7. A stress difference of 0.25% at the center hole is the maximum
deviation between FEA and analytic solution according to section 2.3. Also here can be
shown, that both methods lead to the same results.
Table 6: results - rotationally anti-symmetric load

Parameter Symbol Unit FEA Analytic FEA /
Solution Analytic Solution
Principal stress, max. H1 [N/mm²] 43.375 43.385 99.98 %
Displacement, max. w [mm] 0.0334 0.0334 100 %
332
Figure 7: Comparison FEA with analytic solution for a circular plate with center hole and antisymmetric load
(=0°)
4. Conclusion
Due to the technical progress and claims in the field of architecture the structures
become more and more complex. The determination of the structural behavior is usually
done by FEA. It seems to be simple to produce results via the user-friendly FEA
software, but there can be a big difference in the quality of these results.
The reasons for the differences are usually based on inadequate FEA software and
actual modeling done by the user. It is always necessary to exclude inadequate FEA
results.
In this paper, analytical solutions for detail problems (e.g. punctual support glass panes
with bore holes) in the field of structural glazing are derived and presented. Due to the
extensive equations required for solving these problems a software tool for stress and
displacement calculations was developed. The tool can be downloaded from
www.unibw.de/glasbau/Download.
In the new German DIN standard DIN 18008-3 ‘Glass in Building-Design and
constructions rules-Part 3: Point fixed glazing [9]’ there is a simplified method - called
SLG-method ‘Beyer [6]’- for stress and displacement calculations of point fixed glass.
The analysis of two stress components of the SLG-method is done by the solutions
presented in sections 2.2 and 2.3. The results are listed in ‘AIF-Forschungsbericht
16320N [10]’.
The developed solutions ‘Seel [3]’for circular plates with and without center hole under
symmetrically and anti-symmetrically load can be used for a validation process of point
fixing systems (mechanical and glued).
333
Challenging Glass 3
5. References
[1] Beyer J., Ein Beitrag zum Bemessungskonzept für punktgestützte Glastafeln, Dissertation, Darmstadt,
2007
[2] Siebert B., Beitrag zur Berechnung punktgestützter Gläser, Dissertation, Technische Universität
München, 2004
[3] Seel, M.: Beitrag zur Bemessung von punktförmig gelagerter Verglasung, Dissertation
in Vorbereitung, Professur für Bauphysik und Baukonstruktion an der Universität der Bundeswehr
München, Neubiberg, 2012
[4] Girkmann, K., Flächentragwerke, Band 6, Springer Verlag, Wien, 1963.
[5] Timoshenko S.; Woinowsky-Krieger S, Theory of Plates and Shells, Second Edition, McGraw-Hill,
Tokyo, 1959.
[6] Márkus G., Theorie und Berechnung rotationssymmetrischer Bauwerke, 3. Auflage, Werner Verlag,
Düsseldorf, 1978.
[7] Márkus G., Kreis- und Kreisringplatten unter antimetrischer Belastung, 1. Auflage, Ernst und Sohn
Verlag, Berlin, 1973.
[8] Márkus G., Kreis- und Kreisringplatten unter periodischer Belastung, 1. Auflage, Werner Verlag,
Düsseldorf, 1985.
[9] E DIN 18008-3, Glass in Building –Design and construction rules Part3: Point fixed glazing, DIN
Deutsches Institut für Normung e. V., Berlin, 2011
[10] AIF-Forschungsbericht 16320N, Standardlösungen für punktförmig gelagerte Verglasungen -
Ermittlung der Standsicherheit und Gebrauchstauglichkeits, Deutscher Stahlbauverband, Düsseldorf,
2012
334
doi:10.3233/978-1-61499-061-1-335
Glazing with Countersunk Point Fittings

Geralt Siebert
Universität der Bundeswehr München, Germany, geralt.siebert@unibw.de,
www.unibw.de/glasbau
Tobias Herrmann
Ingenieurbüro Dr. Siebert, Germany, the@ing-siebert.de, www.ing-siebert.de
In contrast to raised head point fittings the countersunk alternatives enable a flat
and even surface. But there is an uncertainty amongst many engineers regarding the
correct structural design of those systems. Present paper collects and arrays design
influencing parameters. A lot of them are incorporated in an automated calculation
tool using the programming interface of a multipurpose FEA software. Extensive
calculations enable an estimation of sensitivity to respective parameters. Further
material behavior of synthetic intermediate layers is described with remarks on
hyperelasticity and viscoelasticity. Finally effects of the bolting torque are
presented.
Keywords: Glass, Countersunk, Point Fixing, FEA, Elastomer
1. Motivation
Plenty of authentic structural calculations treating point fixed glass elements showed a
high uncertainty (or unconcern) regarding correct solution of such a design task. It is
obvious that missing treatment of structural glass design during academic education in
the past is only one reason for this problem. Particularly the high complexity of these
comparatively small structural elements is often underestimated. E.g. it is not
uncommon that software developed for plates made of reinforced concrete is used, as
special glass software is unavailable or general FEA software is unaffordable.
Following design task (see Fig. 1) is solved by means of different FE – models of rising
complexity. That is to say, that the author intentionally started with a simple plate
model with nodal supports (Type 1) and subsequently developed various types (2 – 6)
of models implementing substitute springs, adapting the mesh around the bore,
introducing 3D-elements and finally considering contact mechanism. The goal is to
demonstrate the great bandwidth of results (see Fig. 2) that can be generated on the
basis of just one structural task, without adequate knowledge about the appropriate
translation into a computational model. The example further confirms that stress around
the glass bore is subjected to a larger uncertainty than at homogenous area of mid span.
Authors’ research findings shall serve to reduce uncertainties and hence improve the
quality of structural design of glazing with countersunk point fittings.
335
Challenging Glass 3
Figure 1: Definition of design task (extract) Figure 2: Stress results calculated by means of different
types of numerical models with raising complexity
2. Influencing Parameters
A large number of parameters effect final results of stress and deformation. They can be
assigned more or less clearly to the following classes [1]:
x Functional factors
x Structural characteristics
x Geometric parameters
x Material properties
x Load bearing mechanisms
x Boundary conditions
x FEA specifics
Functional factors comprise the purpose of the glazing, the associated type of load, the
type of load-bearing structure and special requirements (i.e. residual load bearing
capacity). Most of these parameters are obvious and often given by the architects draft
and common technical rules. Structural characteristics stand for the translation of above
requirements to a basic design that is to say choosing the type of glass and support
bearing in mind the substructure. Dimensions of the glass plate, its (countersunk)
drillings, the point fitting’s components and respective tolerances make up the class of
geometric parameters. While mechanic properties of glass and metal are well-known,
plastic interlayer as part of laminated glass or as part of the point fittings in particular
present the sophisticated section of the material properties (see Chap. 4).
A point fixed glazing consists of several detachable components. Loads are transferred
by means of contact and friction between these parts. Simple design software usually
does not offer these kinds of load bearing mechanisms. Shear transfer in laminated glass
is another part of this category. The class of boundary conditions comprises all loads
and supports that define the margin of the structural system. Eccentricities, flexibility of
bearings and the bolting torque (see Chap. 5) are important parameters in this context.
336
Finally all parameters concerning the translation of above mentioned into a

computational model and the computation itself are collected in the group of FEA
specifics.
Basically by means of the parameters of the first 6 classes an exact and complete
description of every structure is possible. But the following problems lead to a wide
scope of results:
x Lack of knowledge about the parameters

x Misinterpretation or neglect of the parameters
x Wrong or incomplete translation into computational model
3. Development of parameterized FEM model

To get an impression in what extend above parameters effect the resulting stress a
parameterized FEM model is developed [1]. It enables a quick examination of various
parameters by means of two systems:
x double symmetric glazing with four countersunk point fittings

x all sides line supported glazing with centric countersunk point fitting
The considered parameters of the former are given in Table 1.
Table 1: Parameters considered in automated calculation

Functional Structural Geometric Material Load Boundary FEA
bearing conditions specifics
Plate with Countersunk Glass plate: Linear Two-sided Loads Diameter of
four point fitting Height elastic contact: radial mesh
fittings Washer
Width
Thickness Bushing
Load type: Glass type: Bore: Young’s Friction in Support: Width of

Distributed Monolithic Location Modulus of contact Degrees of transition to

load all areas: freedom rectangular
Diameter components mesh
Center load Various Substitute
Countersunk coefficients springs
Fixing type: Clamping Poison ratio Preload of Mesh
Rigid disc, Washer: of all bolt density bore:
Diameter components Radial
Hinged
Thickness Tangential
Transversal
Bushing, Mesh
Countersunk density:
head: Point fitting
Thickness components
Eccentricities Geometric
of pin joint nonlinear
and computation
substructure
337
Challenging Glass 3
3.1. Programming
Two approaches are pursued:
x Parametric study
x Single Model analysis
The former shall enable an easy way to cover a large bandwidth of respective parameter
with tabular output for further manipulation to diagrams. The second one shall allow a
quick and user friendly calculation of one configuration.
The multipurpose FEA-software MSC.MarcMentat® provides a programming interface

(program language Python) to generate input files and handle the post-processing.
Besides it is possible to extend graphical user interface with own menus and input fields.
Figure 3 shows the procedures of both approaches. Examples are given in the next
section.
Figure 3: Procedures of parameterized calculation
338
3.2. Examples
The first example investigates the influence of the countersunk angle on glass stress. A
squared glass plate with one rigid point fitting is loaded by a transversal force V and
normal force N. The countersunk angle (and thereby the angle of the countersunk
head) varies between 60° and 120°. The diagram in Fig. 4 points out that maximum
glass stress at bore rises considerably under transversal loading with increasing angle ,
while under normal loading almost no change is recognized.
Figure 4: Example 1 - Influence of countersunk angle on glass stress
339
Challenging Glass 3
The second example (see Fig. 5) examines sensitivity of glass stress and deformation of
point fitting to the stiffness of washer and countersunk bushing. Results show that at
least for the configuration of this example lower Young’s moduli of the bushing lead to
lower stress and higher deformations. The influence of washer’s stiffness becomes
higher if a stiff bushing (i.e. aluminum) is chosen.
Figure 5: Example 2 - Influence of intermediate layer stiffness on stress
340
Final example is a squared plate with four point fittings under distributed load. The
eccentricity e of the pin joint is varied with influence on glass stress. Calculations are
performed for compression and tension load. Besides sensitivity to Young’s modulus of
washer Et is examined. The diagram (see Fig. 6) shows that a stiff washer increases
stress’ sensitivity to eccentricity of the pin joint.
Figure 6: Example 3 - Influence of pin joint’s eccentricity on stress
341
Challenging Glass 3
4. Polymer interlayer of point fixings
4.1. General
The washer and the countersunk bushing nearly represent the smallest components of
the whole glazing. Nevertheless their influence on the load bearing behavior is huge.
Their basic function is to prevent direct contact between fitting’s metal components and
glass. Therefore they must resist UV radiation and detergents. Machinability, stability
against changes in temperature and last but not least costs are further criteria.
Common materials are polymers like ethylene propylene diene monomere rubbers
(EPDM, not for bushings), polyoxymethylenes (POM), silicones and polyamides whose
compounds are chosen in a way that above requirements are satisfied.
EPDM is an elastomer. Elastomers are nearly incompressible and by comparison to

other polymers soft. Large imposed strains generate large transverse strains. If these are
inhibited a tremendous increase of stiffness results. This effect should not be neglected
during design. POM or other stiffer materials do not show this behavior because strains
and Poisson’s ratios are usually smaller.
4.2. Hyperelasticity of elastomers

As strains of soft materials like elastomers are large, Hooke’s law is not longer valid.
Special approaches for large strain material use the strain energy density W to define a
material law. W is calculated by means of the stretch ratios i (see Fig. 7).
Figure 7: Definition of stretch ratios i
342
For incompressible materials like elastomers Mooney [2] proposed
§ 1 1 1 ·
O O O C2 ¨ 2 2 2 3 ¸
2 2 2
W C1 (1)
¨ ¸
© O1 O 2 O 3 ¹ .
1 2 3
The stress tensor is derived by derivation of W with respect to . In the simple case of
uniaxial load principal stress becomes
ª § 1 ·¸ § 1 ·º
V 2 «C1 ¨ O 1 C2 ¨1 3 ¸» . (2)
«¬ ¨© O1 ¸¹ ¨© O1 ¸¹»¼
11
Above equation shows the strong nonlinear relation between stress and stretch. The
coefficients must be determined by testing. High performance FEA codes manage
hyperelastic material models, but computing time rises considerably. Therefore stiffness
of point fitting’s elastomer washer is approximated by Hooke’s law. Two cases (see
Fig. 8) are distinguished that take into account the limit states of friction:
x Free transverse movability at interfaces between elastomer / clamp disc and

elastomer / glass (free)
x Transversal strains set to zero at interfaces (blocked)
Figure 8: Section through elastomer washer – limit states ‘free’ and ‘blocked’
343
Challenging Glass 3
Beginning with the free boundary an equivalent Young’s modulus Efree can be
determined by comparison of a Hooke’s spring with a Mooney’s spring [1]:
§ 2
·
¨ 3 3 w §¨ w ·¸ ¸
¨ t © t ¹ ¸¹ § ·
2 ©
w
E 3 ¨ C1 C1 C2 ¸ . (3)
© ¹
free
§ w· t
¨1 ¸
© t ¹
The remarkable effect of blocking transversal strain at the interfaces is approximated by

equation (4) [1]. It is based on a parametric FEA study and only valid for hollow
cylinders with a ratio between inner radius ri and thickness t bigger than 2.5 and
compressive strains smaller than 10 % (see Fig. 9).
§¨ 0,345 2
1,44 ·¸ E free .
E blocked
©
f Form ¹
(4)
Figure 9: Effect of blocking transversal movement at interfaces of elastomer washer on compressing stiffness
4.3. Visco-elasticity
Beyond nonlinearity towards static strains the strain rate dependence of polymers causes
further efforts for the exact simulation of material behavior. Relaxation and creep are
the keywords regarding long-term loads. Fig. 10 shows corresponding courses of strain
and stress towards time.
344
Figure 10: Characteristic time-dependent behavior of visco-elastic material
In case of horizontal glazing creep of polymer interlayer (inside point fitting and
laminated glass structure) leads to additional deflections under dead load. Relaxation of
the point fitting generated by the bolting torque occurs as the compressive stress of the
elastomer washer is partially dissipated (see Chap. 5).
As mathematical description for a material model considering time-dependent behavior

following equations can be applied. For small strains (Hooke’s law) relaxation may be
expressed by means of a series function:
§ n ·
t
V (t ) H 0 © E f i 1 E i e i ¸¹ .
¨ ¦ W (5)
E, Ei and i can be interpreted as properties of a rheological model (Generalized

Maxwell model):
Figure 11: Generalized Maxwell model
345
Challenging Glass 3
Large strain material models need other formulations. The FE-code of MSC.Marc® [3]
applies
§ m § t
··
W 0 ¨© j 1G j © e j ¸¹ ¸¸¹

W (t ) ¨1 ¦ ¨1 O (6)
which again makes use of strain energy density function W apposite to (1) and a series
function similar to (5). Constants must be determined by curve fitting of relaxation test
results (example see [1]).
Consideration of further nonlinearities like temperature-dependent behavior, aging or

wear might be necessary to improve simulation of real structural behavior.
5. Effects of bolting torque

The bolt of the point fitting is fastened to clamp the glass pane between countersunk
head and inner clamping disk. The contraction causes a compression of the intermediate
layers (bushing and washer) and a tangential tensile stress around the bore hole, because
the countersunk head tends to expand the corresponding drilling. As shown in Chapter
4.3 and in [4] compression of polymer interlayers decreases over time and hence
preload force of the bolt decreases, too.
The example in Fig. 12 displays effects on glass stress and stiffness of the fitting by
means of a single point fitting under eccentric (bending!) transverse load V with
different levels of preload force F.
A higher preload causes a larger contact area for bending. Hence deflection w decreases
and (rotary-) stiffness rises. Example further confirms the incurrence of tensile stresses
around the bore due to bolting torque. Preload forces exceeding a certain value (here
1 kN) have negative influence on overall stress F + V.
346
Figure 12: Structure of parameterized calculation
6. Conclusions
The two explained aspects of material behavior and bolting torque show already the
strong sensitivity of glass panes with countersunk fitting to changes in small areas. Due
to numerous dependencies between the individual parameters general statements are
difficult to give. Further details on this will be published in [1], which also contains
entire information about presented examples.
The material behavior of the polymer components in particular will supply enough work
for future research. The aim has to be to define further limit cases that cover material
properties on the safe side and reduce design effort.
347
Challenging Glass 3
7. References
[1] Herrmann, T., Untersuchungen zu punktgestützten Verglasungen mit Senkkopfhaltern, PhD thesis,
Universität der Bundeswehr München, Germany, in progress
[2] Mooney, M., A theory of large elastic deformation, Journal of Applied Physics, 11(9), 1940, pp. 582-
592.
[3] MSC.Software Corporation, Experimental Elastomer Analysis, Training material course MAR 103,
Santa Ana, CA, U.S.A.,2010
[4] Herrmann, T., Aspects on Glass Panes with Countersunk Fixings, Proceedings of the 3rd International
Symposium on the Application of Architectural Glass (ISAAG 2010), Universität der Bundeswehr
München, Germany, 2010
348
doi:10.3233/978-1-61499-061-1-349
Reduction of Edge Effect in Adhesive

Joints of Glass Details
Olena Soroka
G.S. Pisarenko Institute for Problems of Strength, Ukraine, sorokahelen@gmail.com
Yurii Rodichev,
G.S. Pisarenko Institute for Problems of Strength, Ukraine, rym@ipp.kiev.ua
Alexander Shabetia
Institute of Applied System Analysis of National Technical University of Ukraine “Kiev
Politechnical Institute”, Ukraine, sabio111@gmail.com
Adhesive joints in load-bearing structures made of glass under tension are

characterized by a significant concentration of stresses at the edge. A lack of
sufficiently justified solutions on how to reduce the edge effect restrains the
development and use of adhesive joints. The edge effect in adhesive joints of glass
details is associated with the mechanism of transfer in a mechanical load from one
mating part to another through an adhesive layer. Operating conditions require
increase rigidity of the adhesive joints. As a result, the stress concentration
increases at the edges of the joint. A new technical approach to control the stresses
at the edge by directional change of the modulus of elasticity of adhesive layer in
the edge zone is justified. As an example of the calculation of model of adhesive
joint consisting of elements in the form of strips, joined by an adhesive layer, it is
shown that the reduction of stress concentration at the edge of the adhesive contact
with simultaneous overall high rigidity of joint can be achieved by reducing the
elastic modulus of adhesion layer next to the edge. The law of the elastic modulus
variation is proposed as a "mirror image" to the distribution of shear stresses at the
joint with a constant elastic modulus of adhesive.
Keywords: Architectural glass, Adhesive joints, Shear stress, Edge effect

1. Introduction
Nowadays, adhesive joints of load-bearing structures made of glass are widely used in
building and transport (Figure 1, 2). The edges of joints for different types of structural
elements in such composites are of significant difficulties in the design and production
of joints. Under tension these composites are characterized by a significant
concentration of stresses at the edge. The effects caused by large stress gradients on the
interface of adhesive joint are called the edge effects. High shear stresses at the interface
are responsible for delamination under applied loading which is much lower than the
ultimate loading for the materials of composite. The test results and analysis of failure
of such structures showed that the edge delamination reduces the efficient rigidity of
structure and its ultimate strength. However, the information on the investigation of
edge effect and, especially, methods to control this effect in adhesive joins is very
limited [1, 2]. At the same time operating conditions require increase in rigidity of the
adhesive joints. As a result, the stress concentration increases at the edges of the joint.
349
Challenging Glass 3
Figure 1 presents a scheme of composite load-bearing structural element under tension.

It consits of a laminated glass plate with an internal reinforcing part made of strong
metal. The stress concentration in the edges of glass-metal adhesive joints should be
taken into account in designing this building structure. Cyclic thermal stresses should be
regarded as the main cause of fracture for electric heated glazing of the cockpit
(Figure 2). This laminated structure of the aircraft glazing consists of the external
electric-heated glass layer - 1 joined with an internal load-bearing glass laminated part
through thick polymer adhesive interlayer - 2. Stress concentration near the counter of
electrically conducted coating and at the edge zone causes fracture of this glazing under
extreme operational conditions at low temperature.
1
h
1
3
1 – laminated glass plate; 1 – external

electric heated
electricglass
heated
layer;
glass layer;$\^
2- 2- thick ;
adhesive layer;
3, 44– `$
3, , {$$|`}$
– load-bearing layers \|^
with thin$$^
strong polimer layer
2 – metal reinforcing part
Figure 1: Scheme of composite load-bearing Figure 2: Scheme of electric heated glazing of the
structural element under tension cockpit
A lack of sufficiently justified solutions to the problem of reduction the edge effect
restrains the development and use of these composites.
The previous paper describes the possibility of reduction of the edge effect in adhesive
joints of glassceramics composite elements [3]. It was shown that the edge effect in
adhesive joints is associated with the mechanism of transfer in a mechanical load from
one mating part to another through an adhesive layer. The approach that ensures
reduction of the edge effect based on the choice of a law of adhesion interlayer elastic
modulus variation was proposed.
The paper is devoted to investigation of the edge effect and its control with regard to
architectural and transport laminated composite structures consist of carrying glass
plates and polymer adhesive layer. The influence of constructional parameters and
elastic modulus of composite elements on the shear stress concentration was studied.
The possibility of more efficient control of stress gradients at the edge without reduction
of general rigidity of composite is shown.
2. Problem Statement
To determine the edge effect in the adhesive joint a scheme where the uniform tensile
load P is applied only at the ends of central element of composite is adopted. The model
of adhesive joint is a symmetrical composition consists of elements in the form of strips,
joined by an adhesive interlayer (Figure 3). 2H, h and are the the thickness of glass
layers and interlayer respectivly. The length of the joint is 2l. The origin of coordinates
350
Reduction of Edge Effect in Adhesive Joints of Glass Details
is chosen at the middle point as it shown in the Figure 3. The axis z is a longitudial
symmetrical axis of the joint. Loading is transferred to the outer strip through an
adhesive layer. Shear stresses at the interface of “glass – adhesive layer” are
investigated. Due to symmetry the upper part of composition (above z) is considered.
2 h
P z
1 2H
0

2l
Figure 3: Scheme of loaded composite plate
Due to the fact that the elements of the system are in equilibrium and taking into
account that shear stresses in the cross section z = 0 are = 0, we obtain [2] a linear
differential equation of second order with constant coefficients for w(z) – difference of
the axial displacements of the centers of gravity of outer - 2 and inner - 1 glass strips.
d 2 w( z ) (1)
k 2 w( z ) 0
dz 2
Here k is the coefficient, which depends on elastic properties (lateral strain coefficient
μ; Young’s moduli of strips E1, E2 and adhesive interlayer Eai) and geometrical
parameters of composition elements
§ 1 1 ·
¨¨ ¸¸
2 © E1 H E2 h ¹
(2)
k
ª 2G h Hº
(1 P ) « »
E
¬ ai E 2 E1 ¼
The analitical solution w(z) of the equation and subsequent accounting of the
relationship between stear stresses and displacements resulted in the following equation
P E1 sh kz (3)
W
F1 § 2E § G h · ·¸ ch kl
(1 P ) Hk ¨¨ 1 ¨¨ ¸¸ 1
¸
© H E
© ai E 2 ¹ ¹
Where F1 is the cross sectional area of the inner element and
P
V (4)
F1
is the nominal stress under tensile conditions.
351
Challenging Glass 3
Using the equations (2) and (3) the correlation between the interface shear stress and the
nominal normal stress (hereinafter called relative shear stress / t) taking into account
the influence of Young’s modulus of glass and adhesive, as well as the effect of length
of plate 2l and thickness of layers (2H, h and ) was analysed.
3. Results and Discussion

The formula (3) makes it possible to examine distribution of shear stresses lengthwise
interface of adhesive joint. Basic physical and geometrical parameters of the structural
members for analyzed joints (Figure 3) of the glass composite details are given in
Table 1.
Table 1: Physical and geometrical parameters of analyzed joints of glass composite details.
Parameters of members Section Dimensions [mm]
Thickness of glass layer 1 (2H) mm 12, 6, 2
Thickness of glass layer 2 (h) mm 6, 3, 1
Thickness of adhesive layer 1 () mm 4, 1, 0.37, 0.1, 0.01
Length of the joint mm 1000, 400, 200
Young’s modulus of glass layers GPa 70
Young’s modulus of adhesive layers Eai GPa 70, 10, 5, 1, 0.1, 0.01,
The examples of calculated distribution for relative shear stress / when H=h=6mm;
=0,37mm; Eai =1GPa for different length of joint are presented in Figures 4 and 5. As
follows from the analisys of curves 1-3 (Figure 4), the maximum value of relative shear
stress for the same modulus of adhesive layer does not depend on the joint length, but
mesuare of shear stress localization depends on the length of composition: for short
joints zone of considerable shear stresses correlates with the length of the joint, and
stress concentration is more localized near the edge in the long joints while their middle
part is much less loaded with shear stresses (Figures 4 and 5). This decreases
performance of “short” compositions.
/t
/t
0,18
0,18
0,16
0,16
3 0,14
0,14
0,12
0,12 2
1 0,1
0,1
0,08
0,08
0,06
0,06
0,04
0,04 1 2 3
0,02
0,02
0 z,mm 0 z-l /l
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7
100 0 100 200 300 400 500
Figure 4: Distribution of relative shear stresses Figure 5: Measure of shear stress localisation for
lengthwise interface for composites with H=h=6mm; composites with different length 2l: 1- 1m, 2- 0,4m,
=0,37mm; Eai =1GPa and 2l: 1- 1m, 2-0,4m, 3-0,2m 3- 0,2m
352
An increasing in joint length for adhesive with higher Young’s modulus makes it
possible simultaneous composite structure rigidity and strength. But as a result of the
increased stress localization on the edges and high value of the maximum shear stress a
special technical approach should be developed to optimize the stress distribution at the
edges.
Dependence of the shear stress concentration on elastic modulus of adhesive interlayer

was studied for a plate of the length 1 m (Figures 6 and 7). The Young’s modulus of the
adhesive interlayer varies from 0.01GPa up to 70GPa. As shown in the Figure 6 an
increase of the value of Young’s modulus of adhesive layer Eai from 0.01GPa to 1GPa
causes higher shear stress up to 8.5 times. Shear stress concentration also increases
significantly. The length of the edge effect area decreases up to 100mm. About 80% of
adhesive layer is not loaded enough. The use of low rigid elastic adhesives causes more
uniform stress distribution. However low strength of these adhesives results in
insuffisient mechanical connection of load-bearing structural memebers as well as
decrease of strength and rigidity of composite glass detail.
An increase in adhesive layer Young’s modulus up to 10GPa causes to edge stress

rising in about 1.5 times. But, further increase in Young’s modulus of adhesive leads to
obtaining stable values of relative shear stress, which is about 0.27… 0.3 (See Figure 7).
Thus, an effecient control of edge stress value in the range of 0.05< / < 0.27 and
structural strength of this type of joint in the composite glass details can be realized by
the variation of Young’s modulus of adhesive materials in the range of 0,1…10 GPa.
The results of study of the influence of glass plates thickness on the edge shear stress
concentration are shown in Figure 8. The following parameters of joint were adopted in
the calculations: Eai = 1G}, =0.37mm, 1-500mm, H= 6mm. The thickness h of glass
layer 2 (Figure 3) was 1mm, 3mm and 6mm.The maximum values of shear stress at the
edge of the joint were decreased by about 1.5 times whith thickness h of 1 mm. The
length of edge stress zone decreased up to 40mm. The decrease of thickness h causes
the decrease of shear stress at the edge of the joint.
353
Challenging Glass 3
/
2 3
z,mm
Figure 6: Dependence of the edge shear stress concentration on elastic modulus of the adhesive interlayer for a
composite plate with the length 1 m. H = 6mm, h = 6mm, = 0.37mm. 1- Eai = 1GP, 2- Eai =0,1GP},
3- Eai =0,01GP}.
m/
Eai,GPa
Figure 7: Correlation between relative shear stress at the edge of composition and Young’s modulus of the
adhesive interlayer
The other way to decrease shear stress concentration in the edge of joints in glass
composite details by reducing the rigidity of adhesive layer is shown in Figure 9.
Simultaneous decrease of Young’s modulus Eai from 1GPa to 0.1GPa and an increase in
thickness of the adhesive layer from 0.37mm to 4mm causes a decrease in the stress
parameter / up to 0.02.
354
/
2
1
z, mm
Figure 8: Dependence of the edge stress distribution on the thickness of glass layer 2.
Eai = 1GP}, =0.37mm, 1-500mm, H= 6mm. 1-h=1mm, 2-h=3mm, 3-h=6mm
Correlation between the relative shear stress at the edge of composition and thickness of
the adhesive interlayer is shown in Figure 10. Stress distribution was calculated for
composite plates with a length 2l =1m and H= h=6mm. Young’s modulus of the
adhesive layer was Eai = 1 GPa (upper curve) and Eai = 0,1GPa (lower curve). The
thickness of the adhesive interlayer varied in the range of 0.01 mm…4 mm. Decrease
in the shear stress at the edge when thickness of adhesive interlayer increases up to
4 mm is significant. This effect can be used to optimize the shear stress distribution at
the edge zone of joints as well as to increase the bearing capacity of the plates under
mechanical and thermal loads.
In order to reduce the stress concentration at the edge of the adhesive joint with
simultinuous overall high rigidity of the composite, the method of decreasing the elastic
modulus in the rigion of edge effect was proposed. As an example a law of the
interlayer elastic modulus variation which is the mirror reflection of the law of shear
stress variation, when Eai varies from 1GPa to 0,01GP} was cosidered (Figure11).
355
Challenging Glass 3
/ /
0,3
0,25
1 0,2
0,15
0,1
2 0,05
3 0
z, mm , mm
0 0,5 1 1,5 2 2,5 3 3,5 4
Figure 9: Shear stress control by the change of Figure 10: Correlation between relative shear stress
adhesive layer rigidity: at the edge of composition (1=500mm, H=6mm,
1- Eai = 1GP}, =0.37mm, 2- Eai = 1GP} h=6mm) and thickness of adhesive interlayer.
=4mm or Eai =0, 1GP}, =0.37mm (Eai / ~const), Upper curve - Eai = 1 GPa ,
3- Eai =0, 1GP}, =4mm lower curve - Eai = 0,1GPa
Replacement of the constant Eai by functional dependence Eai=f (z) results in the
coefficient k (2) is variable. Therefore equation (1) transforms into a differential
equation of second order with variable coefficient:
d 2 w( z ) (6)
k 2 ( z ) w( z ) 0
dz 2
Figure 11: Shear stress control by replacement of the constant Eai modulus by variable Eai
The numerical solution of this equation allows one to obtain w(z) and shear stress
distribution in the case of variable elastic modulus of the interlayer. It is demonstrated
in Figure11 that the maximum shear stress at the edge of composite with variable elastic
modulus of the interlayer decreased by 1.4 times compared with one where elastic
modulus of interlayer is 1GPa, while the regidity of composite is retained.
356
Some other ways for reduction of the edge effect due to variation of elastic modulus of
the interlayer can be proposed, particulary it may be chosen as staircase characteristic:
from low value next to edge to higher ones in the middle part of the composition.
The obtained results provide a basis for design of architectural and transport composite
glass structures with an adhesive joints optimized on strength and rigidity parameters.
The following tendencies may be useful to ensure appropriate bearing capacity of these
structures taking into account shear stress concentration at the joint edge:
1) The minimum concentration of the edge stress is typical of joints with the lower
values of Young’s modulus and increased thickness of adhesive layer. But low
strength and rigidity of these joints leads to insufficient load-bearing behavior
under mechanical and thermal loading.
2) The compositions with high strength can be obtained using the strong and rigid
adhesive materials. However, the excess concentration and high level of the
maximum edge stress will cause fracture of these compositions. Inreasing of the
length of rigid joints is not efficient way for bearing capacity increasing on this
reason.
3) The combined technical approaches with optimisation of the edge stress
distribution on strength, rigidity and fracture resistance parameters have to be
developed to decrease the maximum value of the edge stress in composite glass
load-bearing structures.
4) Decrease in the elastic modulus of the adhesive interlayer at the edge region,
which enable to retain enough regidity to ensure bearing capacity of the joint,
causes decrease in the shear stress concetration and provides an adhesive strength
of composite.
Some suggestions how to reduce edge effect in a practical application are stated:
1) Technologically. It was suggested [3] to obtain an adhesion interlayer with
modulus, which depends on the distance to the joint edge when an adhesive
composition is used as a solution of a polymer in a volatile monomer with
additions of cross-linking plasticizer. During hardening of the adhesive in the
edge region the monomer volatilizes from the edge region, the dissolved
plasticized polymer is formed in the adhesive layer. As a result the elastic
modulus of adhesion interlayer decreases to the edge and varies over a wide range
due to variation of the amount of the added volatile monomer and plasticizer.
2) Design method.
- faired or stepped geometry of joint face leading to increase of the thickness of
the adhesive interlayer at the edge;
- stepped change of modulus due to application of different adhesive substances.
3) Decrease of elastic modulus of polymer that depends on strain value under
conditions of large deformation due to mechanism of viscous flow can be used.
357
Challenging Glass 3
4. Conclusion.
The results of investigation can be used as a basis of design-technological approach for
creation of laminated glass composite structures optimized in terms of adhesive strength
and rigidity properties. This approach will make possible to control performance of
glass joints at the stage of their design and production.
5. References
[1] Kiselev, A.G., Stress state at the edge points of diffusion joints of dissimilar materials, in: Collection of
Science Papers, Izd. MIFI, Moscow, 2000, pp 131-132.
[2] Bedly, M. R.; Ambrico, J.M., Delamination of thin films from two-dimensional interface flaws at
corners and edges, International Journal of Fracture, 3/2001, pp. 205-222.
[3] Maslov, V.M.; Soroka, O.B.; Lyashenko B.A.; Rodichev Yu.M., Reduction of edge effect in the adhesive
joint of pyroceramics, Strength of materials, 6/2005, pp.606-612.
358
Challenging Glass 3
Strength, Stability & Safety

359
doi:10.3233/978-1-61499-061-1-363
Improvement of Quality of Tempered Glass

with Numerical Modeling
Antti Aronen, Reijo Karvinen
Tampere University of Technology, Department of Energy and Process Engineering,
Finland,antti.aronen@tut.fi
Thin glass has increased problems in a tempering process. Thinner glass needs a
higher temperature and higher heat transfer coefficient to get tempered glass. At the
same time, bending stiffness of glass plate decreases and the possibility of faults in
visual quality increases. Numerical modeling is needed to understand how different
tempering parameters and support of glass affect tempering and visual quality. In
the paper, the reasons for higher temperature and higher heat transfer coefficients
for thin glass are shown. The traditional way of supporting glass with rollers is
studied and the effect of different parameters on deformations is shown.
Keywords: Tempering, Modeling, Quality
1. Introduction
The purpose of the glass tempering process is to improve glass strength. A tempered
glass should have the same visual quality as a float glass. The visual quality of tempered
glass can suffer during the tempering process due to high temperature or incorrect heat
transfer. The need to reduce the material used has created a demand for thin glass.
Thinner glass increases problems in the tempering process. Online measuring is difficult
and modeling is needed to understand the behavior of glass in the process.
Thin glass needs a higher temperature and a higher heat transfer coefficient to get
tempered glass. However, due to the high temperature and uneven support by rollers,
bending stiffness reduces and visual faults like roller waves and edge bending are
difficult to avoid [1,2]. The stress profile is not the only criterion for tempered glass.
Visual quality and the flatness of the tempered glass are criteria that are almost as
important.
Before conducting modeling, the theory of heat transfer and mechanical behavior of
glass has to be understood. The theory of heat transfer is based on energy equation and
boundary conditions [3] and mechanical behavior based on thermal stresses and
viscoelasticity [4,5].
The aim of this paper is to examine why problems increase with thin glass and how they
can be avoided. The paper also considers traditional roller support and studies how the
change of a support system affects the roller waves in tempered glass.
363
Challenging Glass 3
2. Theory
2.1. Heat transfer

Heat transfer and the temperature field form the basis of the heat treatment process.
Heat transfer and temperature range can affect the stress results in the tempering process.
Concerning residual stresses, the temperature field during the cooling process is the
most important factor. With thin glass, cooling is fast and radiation temperature is low.
So, radiation can be ignored and the calculation of temperature distribution is based on
the use of energy equation [3]
wT w § wT ·
Uc p ¨k ¸ (1)
wt wx i ¨ wx ¸
© i ¹
where U is the density, cp is the specific heat, T is the temperature, t is the time and k is
the heat conductivity.
In order to solve Eq. (1), boundary conditions have to be fixed, which can be easily
achieved using the heat transfer coefficient. For the one-dimensional case where the
temperature field is calculated over the thickness b, the boundary equation is
wT b 2 , t
q k h T b 2 , t Tf (2)
wz
for an upper surface and similarly for a lower surface. The temperature difference in
glass transition range has to be about 150 °C to get tempered glass.
In heat transfer calculations, it must be remembered that heat conductivity is linearly

dependent on temperature (Appendix A). In addition, the specific heat is time- and
temperature-dependent. The specific heat in the transition range is the combination of
heat capacity of equilibrium liquid cp,l and heat capacity in glassy state cp,g [6].
T Tf T
³ c p T ´ dT ´ ³ c p, l T ´ dT ´ ³ c p , g T ´ dT ´ (3)
T0 T0 Tf
The specific heat in a glassy state is linearly temperature-dependent and constant in

equilibrium liquid. The fictive temperature Tf is dependent on structural relaxation,
which will be presented in Chapter 2.3.
The temperature field can be calculated for the one-dimensional case using the finite
difference method [3].
2.2. Stresses and strains

At high temperature, above the transition temperature, the effect of viscoelasticity
occurs. Viscoelasticity can be presented using the Maxwell model. The bulk and shear
moduli, K and G, are time- and temperature-dependent and can be presented using
Prony’s series [5,7].
364
Improvement of Quality of Tempered Glass with Numerical Modeling
n § t ·
K t Kf K0 Kf ¦ w2i exp¨¨ W ¸
¸ (4)
i 1 © 2i ¹
n § t ·
Gt G 0 ¦ w1i exp ¨¨ ¸
¸ (5)
i 1 © W 1i ¹
When bulk and shear moduli are known the stress-strain relation is presented with a
hereditary integral [8]
t
w H 3H th t w eij
V ij t G ij ³ K t t´ dt´ 2 ³ G t t´ dt´ (6)
0 wt´ 0 wt´
where Vij is stress, H is the sum of Hxx, Hyy and Hzz, eij is deviatoric strain eij = Hij -1/3 H and
Hth is thermal strain.
In the heat treatment process, the temperature field is connected to thermal strain, which
governs the stress calculation. The numerical methods for the solution of stresses and
strains are shown in references [8,9].
2.3. Structural relaxation

Glass is a thermorheologically simple material. This means that the relaxation time at
different temperatures can be calculated using a shift function I. Because the structural
relaxation and the change of fictive temperature have to be taken into account, the shift
function is also time-dependent. The shift function is
W ref ªH § 1 x 1 x ·¸º
I t, T exp « ¨ » (7)
W «¬ R ¨ Tref T T f ¸¹»
© ¼
where W is relaxation time, H is activation energy, R is the universal gas constant and x
is a ratio of activation energy in the glassy state and a sum of activation energies in the
glassy state and for structure x=Hg/(Hg+Hs). [6]
When temperature changes during the process, the time should be replaced by a reduced
time. The reduced time [ takes the temperature-weighted relaxation time into
consideration.
t
[ t ³I T t´ dt´ (8)
0
The time in Eq. (6) can be replaced with the reduced time.
To calculate Eq. (7), the fictive temperature has to be known. The change of properties
is dependent on the fictive temperature and it can be described by the response function
365
Challenging Glass 3
p t p2 f T f t T2
Mp t (9)
p2 0 p2 f T1 T2
In the response function Mp above, p is a property, subscript 1 is the state before the
temperature change and subscript 2 is the state after the temperature change. The
response function of the material property Mp can be expressed by the analogy with the
bulk or shear relaxation function
n
Mp t ¦ Ci e t / O i
(10)
i 1
where Ci is a weight coefficient for a structural relaxation time Oi.
Using Eq. (9), the fictive temperature can be calculated.
t
wT t´
Tf t T t ³ M p t t´ dt´ (11)
0 wt´
The fictive temperature depends on the speed of temperature change. The numerical
solution of fictive temperature can be solved with the algorithm by Markovsky et al.
[10].
3. Effect of initial temperature and heat transfer coefficient on tempering
3.1. Effect of heat transfer coefficient on temperature field

The heat transfer coefficient needed for tempering depends on glass thickness. The
temperature difference between surface and mid-plane should be over 150 °C. By
keeping the initial temperature constant and high enough for tempering, the effect of
heat transfer coefficient on the temperature field during cooling and on maximum
temperature difference between surface and mid-plane can be studied. The results in Fig.
1 show the maximum temperature difference during cooling when the initial
temperature is 650 °C and the thicknesses used were 2, 4, 6 and 10 mm. The material
properties shown in Appendix A are used in following simulations.
With thinner glass, the heat transfer coefficient needed to get 150 °C temperature
difference is higher. The heat transfer coefficient needed is inversely proportional to the
glass thickness.
366
Figure 1: Maximum temperature difference during cooling with different heat transfer coefficient and
different thicknesses. Initial temperature is 650 °C.
3.2. Effect of cooling rate on glass transition temperature

Property change is dependent on the speed of the temperature change. In cooling with a
faster temperature change, the time for relaxation shortens. Thus, the change speed of
fictive temperature decelerates at a higher temperature. The constant value of the fictive
temperature after cooling shows the glass transition temperature.
For thinner glass, a higher heat transfer coefficient is needed to obtain a high enough
temperature difference between surface and mid-plane. Then the transition temperature
increases, because the maximum temperature difference is meant to form at the glass
transition range. So, temperature before cooling should be higher for thinner glass.
Fig. 2 shows the effect of the cooling rate on the behavior of fictive temperature and
glass transition temperature. Results in Fig. 2 are calculated using Eqs. (10) and (11).
Figure 2: Effect of cooling rate on fictive temperature and glass transition temperature.
367
Challenging Glass 3
For a 3 mm glass sheet, the average cooling rate in the glass transition temperature is
between 50 - 500 °C/s and with 6 mm glass the cooling rate is between 10 - 100 °C/s
depending on the temperature before the cooling and the heat transfer coefficient.
3.3. Effect of initial temperature and heat transfer coefficient on residual stresses
In the chapters above, reasons for higher temperature and higher heat transfer
coefficients for thinner glass have been presented. The effect of those two tempering
parameters can also be studied for one thickness. Residual stresses on the surface for 4
mm glass with modified initial temperature and heat transfer coefficient are shown in
Fig. 3. The results show that higher temperature and higher heat transfer coefficient
increase the residual stresses.
In the results, the reference value of heat transfer coefficient hr is the heat transfer
coefficient needed to reach 120 MPa surface compressive stress at temperatures above
the plateau level for 3 mm glass. The reference value of temperature Tr is the glass
transition temperature when the cooling rate is 100 °C/s. This can be calculated using
Eq. (11). Temperatures are in degrees Celsius.
Figure 3: Effect of the temperature and the heat transfer coefficient on surface residual stress.
Glass thickness is 4 mm.
The temperature should be over 650 °C to get more than 120 MPa compressive stress on
the surface. With higher temperatures, the heat transfer coefficient can be decreased.
368
4. Deformations
The deformations for a stationary glass plate in viscoelastic case wve can be calculated
using deformations in the elastic case we.
w ve x , t J t Ew e (12)
where J(t) is a creep compliance and E is Young’s modulus.
The elastic deformation for glass beam with uniform load q0 is
q0 L4
we x f x L (13)
EI
For a narrow rectangular beam, the moment of inertia I is
b 3W
I (14)
12
and the uniform load is
q0 UgbW (15)
In these equations, W is the width and the function f(x/L) depends on the geometry and
the support of the beam. Then, the effect of the dimensions of beam on deformations for
stationary glass can be found.
12 UgL4
w ve x, t Jt f x L (16)
b2
The effect of thickness and length for roller-supported glass in the motion on
deformations is presented in the sections below.
4.1. Effect of glass thickness on deformations

During the tempering process in glass, visual faults like roller waves and edge bending
are formed due to the uneven support of rollers and creeping at high temperatures.
Visual faults are more common in thin glass due to the high temperature and lower
bending stiffness. The result in Fig. 4 compares the displacement of 3 mm and 4 mm
thick glasses in the case of the deformation of the rear end of a glass plate that moves on
rollers.
The results show that ratio is proportional to b-2 ((3/4)-2 = 1.78). Most of the ratios are
between 1.7 and 1.8. Some errors are formed due to numerical error during the
calculations and rounding errors when displacement is small near the rollers.
369
Challenging Glass 3
Figure 4: Ratio of displacement of rear end of glass plate between two thicknesses (3 mm / 4 mm).
4.2. Effect of roller distance on deformations

The front and rear edges dominate the deformations and when the roller distance
changes, the maximum length from the front and the rear end to the nearest roller also
changes. The ratio of maximum deflections at two different roller distances (0.12 m and
0.08 m) is presented in Fig. 5.
Figure 5: Ratio of maximum deflections at two different roller distances (0.12 m and 0.08 m)
370
In the case of a stationary glass plate, the ratio is about 5.1. The ratio presented in Fig. 5
is about 6, which is higher than for results in a stationary plate. The change of support
should be taken into account, but proportion of length to the fourth power (L4) is a good
approximation.
5. Conclusions
In glass tempering, both tempering quality and visual quality are important. The quality
has to be increased especially for thin glass. The forming of stresses and deformations
can be studied with numerical modeling. The modeling gives good results and
approximations of glass behavior during the process.
Residual stresses can be affected by the temperature level and the heat transfer
coefficient. With thinner glass, the heat transfer coefficient should be raised to get
sufficiently high thermal strain difference between surface and mid-plane. Due to faster
cooling, the glass transition temperature rises, and the temperature level before cooling
should be increased.
The thickness of glass plate affects the deformations and visual quality of tempered
glass. Thinner glass has lower bending stiffness, which increases deformations. More
even support of rollers decreases deformations. For thin glass, the higher temperature
also increases deformations.
6. Appendix A
Material properties for soda-lime-silica glass
Table 1: Material properties [11-13]

Young’s modulus E = 70 GPa
Poisson ratio Q = 0.22
Thermal expansion coefficient for solid glass Dg = 9·10-6 1/K

Thermal expansion coefficient for liquid glass Dl = 32·10-6 1/K

Thermal conductivity k = 0.975 + 8.58·10-4 T W/mK ,where T in °C
Specific heat of solid glass cp,g = 893 + 0.4 T J/kgK ,where T in K
Specific heat of liquid glass cp,l = 1433 J/kgK
Ratio H/R = 76200 K
Constant x = 0.5
Density = 2530 kg/m3
Environment temperature Tf = 20 °C
371
Challenging Glass 3
Table 2: Characteristics of shear and bulk relaxation times and response function
for structural relaxation (Tref = 869 K) [12]
Bulk relaxation times
Shear relaxation times Structural relaxation
(K/K0 = 0.18)
w1i W1i w2i W2i Ci Oi

-5 -5
0.05523 6.658·10 0.0222 5.009·10 0.05523 5.965·10-5
0.08205 1.197·10-3 0.0224 9.945·10-4 0.08205 1.077·10-2
0.1215 1.514·10-2 0.0286 2.022·10-3 0.1215 0.1362
-2
0.2286 0.1672 0.2137 1.925·10 0.2286 1.505
0.2860 0.7497 0.394 0.1199 0.2860 6.747
0.2265 3.292 0.3191 2.033 0.2265 29.63
7. References
[1] Henriksen, Thomas; Leosson, Kristján, Anisotropy and Optical Distortion in Architectural Glass, Can It
Be Controlled, Proceedings of Glass Performance Days 2009, June 2009, Tampere, Finland, pp. 834-839,
2009.
[2] Abbott Mark; Madocks John, Roller Wave Distortion - Definition, Causes and a Novel Approach to
Accurate, On-line Measurement, Proceedings of Glass Processing Days 2001, 18-21 June 2001,
Tampere, Finland, pp. 226-230, 2001.
[3] Bejan, Adrian, Heat Transfer, John Wiley & Sons, Inc., USA, 1993.
[4] Boley, Bruno; Weiner, Jerome, Theory of Thermal Stresses, Dover, Mineola (NY), USA, 1997.
[5] Flügge, Wilhelm, Viscoelasticity, 2nd ed., Springer-Verlag, Berlin, Germany, 1975.
[6] Scherer, George, Relaxation in Glass and Composites, John Wiley & Sons, Inc., USA, 1986.
[7] Daudeville, Laurent; Carré, Hélène, Thermal Tempering Simulation of Glass Plates: Inner and Edge
Residual Stresses”, Journal of Thermal Stresses, 6/1998, pp. 667-689.
[8] Aronen, Antti; Karvinen, Reijo, Modeling of Deformations and Stresses during Glass Tempering,
Proceedings of the ASME 2011 International Mechanical Congress & Exposition IMECE 2011,
November 2011, Denver, Colorado, USA, 2011.
[9] Chambers, Robert, Numerical Integration of the Hereditary Integrals in a Viscoelastic Model for Glass,
Journal of American Ceramic Society, 8/1992, pp. 2213-2218.
[10] Markovsky, Alex; Soules, Thomas; Boyd, D.C., An Efficient and Stable Algorithm for Calculating
Fictive Temperature, Journal of American Ceramic Society, 4/1984, pp. C56-C57.
[11] Carré, Hélène; Daudeville, Laurent, Numerical Simulation of Soda-Lime Silicate Glass Tempering,
Journal de Physique IV,1/ 1996, pp. 175-185.
[12] Carré, Hélène; Daudeville, Laurent., “Load-Bearing Capacity of Tempered Structural Glass”, Journal of
Engineering Mechanics, 8/1999, pp. 914-921.
[13] Daudeville, Laurent; Bernard, Fabrice; Gy, René, Residual Stresses Near Holes in Tempered Glass
Plates, Materials Science Forum, vol. 404-407, 2002, pp. 43-48.
372
doi:10.3233/978-1-61499-061-1-373
Analytical Approaches for Buckling

Verification of In-Plane Loaded Laminated
Glass Columns and Panels
Claudio Amadio, Chiara Bedon
University of Trieste, Italy, amadio@units.it, chiara.bedon@phd.units.it
Modern and innovative architectural applications are frequently associated to the

structural use of glass beams and panels. Such brittle elements are in fact largely
adopted to sustain in-plane or out-of-plane loads (e.g. columns, stiffeners, stairs,
etc.). However, due to their typical slenderness, they could be affected by stability
problems. Because of these reasons, accurate investigations should be dedicated to
the analysis of their buckling response. In the paper, simple analytical formulations
are proposed to study the load bearing capacity of in-plane loaded 2-layer laminated
glass columns and panels. Comparisons with numerical simulations are proposed to
validate the analytical models. At last, useful formulations are suggested also for
the buckling verification of in-plane compressed 3-layer laminated glass elements.
Keywords: Buckling Verification, Laminated Glass, Equivalent Thickness,

Compressive Loads
1. Introduction
Laminated glass (LG) elements, originally used as purely architectural and decorative
accessories in buildings, constitute a fundamental structural component in the
realization of modern and innovative architectures. However, due to their typical high
slenderness, LG elements are frequently affected by stability problems. Because of this
reason, several authors investigated the buckled response of glass panels or beams in
different loading conditions, providing interesting experimental results and

sophisticated numerical validations. In the paper, an analytical approach based on the
concept of equivalent thickness is proposed to perform a rational buckling verification
of LG panels and columns under compression. The model well applies to 2-layer LG
elements, as well as to 3-layer LG elements, which are frequently used in practice. By
means of opportunely calibrated correction factors, the model accurately estimates the
critical buckling load of these elements and precisely describes the corresponding load
N-displacement w relationship. At the same time, it allows to simplify the numerical
modeling phase and to reduce the processing time required in simulations.
2. Analytical models for in-plane compressed laminated glass panels
2.1. Buckling verification of 2-layer laminated glass panels

In general, the buckling resistance of LG panels under in-plane compression is
estimated referring to the common linear elastic theory of sandwich elements [1]. As
proposed by Luible [2], the critical buckling load of a flat LG panel (length a, width b),
373
Challenging Glass 3
simply supported on the four edges (Figure 1), obtained by assembling two monolithic
glass sheets (thicknesses t1 and t2, Young’s modulus E, Figure 2) and a middle interlayer
(thickness tint, Young’s modulus Eint, shear modulus Gint) subjected to in-pane
compression, can be estimated as:
Figure 1: Simply supported flat LG panel subjected to in- Figure 2: 2-layer laminated glass element
plane compression. Geometry (a) and deformed (cross section).
configuration (b).
D1 D2 ª§ mb · º As b 2
2
2 «¨ ¸ 1» 2
D ¬«© a ¹ ¼» S Ds S D ,
2
§ mb a · 2
N y ,cr ,lam ¨ ¸ (1)
© a mb ¹ ª§ mb · 2 º As b2
«¨ ¸ 1» 2
¬«© a ¹ ¼» S Ds
with:
D D1 Ds D2 , Ds ( Et1 z12 Et 2 z 22 ) (1 Q 2 ) , (2) (3)

As Gint ( z1 z 2 ) 2 t int , Di Et i3 12(1 Q 2 ) , i= 1, 2, (4) (5)
z1, z2= distance between the centroidal axis of the interlayer and each glass sheet.
Depending on the value of Gint, the value for Ny,cr,lam calculated by means of Eq.(1) is
always comprised between the well known limit conditions, which are respectively
defined:
x in absence of any shear connection between the glass sheets (layered limit, that
is Gint 0)
a · S 2 ( D1 D2 )
2
§ mb
N y ,cr ,lam N y ,cr ,abs ¨ ¸ ; (6)
© a mb ¹ b2
374
Analytical Approaches for Buckling Verification of In-Plane Loaded Laminated Glass Beams and Panels
x in presence of a rigid shear connection between the glass sheets (monolithic

limit, that is Gint )
2
§ mb a · S 2D
N y ,cr ,lam N y ,cr , full ¨ ¸ . (7)
© a mb ¹ b 2
In this context, numerical and analytical comparisons performed by Luible [2]

demonstrated that Eq.(1) can be used to predict the bifurcation load Ny,cr,lam of simply
supported LG panels with a good level of accuracy. In particular, the mean ratio
between analytical and numerical critical loads of 200 LG panels characterized by
various geometrical and mechanical properties resulted equal to 1.05. However, the
estimation of the critical buckling load Ny,cr,lam does not constitute a useful criterion to
define the ultimate strength of a buckled LG panel, since the post-buckled regime is
typically characterized by membrane effects which allow sustaining greater loads.
An interesting analytical formulation for the buckling verification of 2-layer LG panels

subjected to in-plane compression can be derived from the simplified approach based on
the concept of equivalent thickness, originally formulated by Wölfel [3] and recently
applied by Bennison to LG elements in several boundary or loading conditions [4]. In
accordance with this theoretical model, the effective level of connection offered by the
interlayer can be expressed by means of a shear transfer coefficient *, (*= 0: layered
limit, *= 1: monolithic limit) defined as:
1
* , (8)
EJ s tint
1 9. 6 E
Gint t s2 O2
with:
O min( a, b) , ts 0.5(t1 t 2 ) tint , (9) (10)
t s ,1 t s t1 (t1 t 2 ) , t s,2 t s t 2 (t1 t 2 ) , (11) (12)
Js t1t s2, 2 t 2 t s2,1 (13)
and E a coefficient depending on the boundary and loading conditions [3]. For the
verification of simply supported LG beams in bending, for example, accordingly with
the original formulation, the value E= 1 should be taken into account [9]. In these
hypotheses, the deformation w of the LG panel can be evaluated referring to an
equivalent thickness defined as:
t eq ,w 3 t13 t 23 12*J s . (14)
375
Challenging Glass 3
Once teq,w is known, the critical buckling load of a given flat LG panel can be
reasonably calculated by means of the classical analytical formulation [5]:
2
§ mb a · S Etef3 ,w S2
2
N y ,cr ,lam ¨ ¸ 2 kV Def . (15)
© a mb ¹ b 12( 1 Q )
2
b2
In this context, a series of analytical calculations has been performed to detect if the
equivalent thickness approach is sufficiently accurate for the analysis of LG flat panels
simply supported along the edges and subjected to in-plane compression.
Specifically, a series of critical loads Ny,cr,lam has been evaluated by using the linear
elastic sandwich theory (Eq.(1)) and the equivalent thickness approach (Eq.(15), with
E= 1). Analytical calculations have been performed taking into account various
mechanical and geometrical parameters characterizing a typical LG panel, that is the
value of Gint (10-4 N/mm2 < Gint < 104 N/mm2), the aspect ratio D (1 D 5, with b=
1m), the thicknesses of glass sheets and interlayer (6/1.52/6mm, 8/1.52/8mm,
10/1.52/10mm). For the purpose of this work, only the first critical load has been
considered (m= 1). The main results are shown in Figure 3 for the 8/1.52/8mm LG panel,
as a function of the ratio
N crZenkert
RN , panel t
. (16)
N cref ,w
1.40 2.2
D= a/b 8/1.52/8mm m= 1
E coefficient for in-plane compressed
1 b= 1m
1.35 2-layer LG panels
2 Analytical calibration [6]
3 2.0
Fitting curve (Eq.(17))
1.30 4
5
1.25 1.8
1.20
RN, panel
1.6
1.15
1.10 1.4
1.05
1.2
1.00
Layered limit Monolithic limit
0.95 1.0
-4 -3 -2 -1 0 1 2 3 4
10 10 10 10 10 10 10 10 10 0 2 4 6 8 10
Gint [N/mm ] 2
D= a/b
Figure 3: Critical buckling load Ny,cr,lam for in-plane Figure 4: E coefficient for in-plane compressed LG
compressed LG panels (8/1.52/8mm) simply panels simply supported along the edges. Analytical
supported along the edges (m= 1). calibration [6] and fitting curve (Eq.(17)).
As shown in Figure 3, the examined formulations do not agree, and in general the
equivalent thickness approach overestimates the critical load Ny,cr,lam. Performed
calculations highlighted that RN,panel is independent on the thicknesses of the layers
constituting the LG panel. In contrary, the main differences depend on the values of Gint
and D. Similar differences between the analytical approaches can be avoided only if in
376
Eq.(8) an appropriate coefficient E, analytically calibrated as a function of D, is assumed

(Figure 4). The fitting curve for the so calibrated values of E is [6]:
1.09
E 1.09 . (17)
D2
In these hypotheses, E can be considered as the correction factor able to give RN,panel = 1.
The proposed formulation provides two contemporary advantages, since it allows a
suitable calculation of the critical buckling load Ny,cr,lam and at the same time it allows to
investigate, for the studied LG panel, the corresponding load N–transversal
displacement w relationship. As proposed in a previous effort [6], the load carrying
behavior of a generic LG panel, simply supported along the edges, subjected to a
uniform in-plane compression Ny, can in fact be investigated by means of Wolmir’s
formulation [7]:
§t·
2
° S 2 S2 ª§ w · 2 § w ·§ w0 · § w0 · º ½° w ,
2
Ny Eb¨ ¸ ® «¨ ¸ 3
¨ ¸¨ ¸ ¨ ¸ » ¾
2 (18)
°̄ 3(1 Q ) 8 © t ¹© t ¹ © t ¹ »¼ °¿ w w0
2
©a¹ «¬© t ¹
with w0 the maximum amplitude of the possible initial sinusoidal imperfection affecting
the LG panel. In [6], comparisons with experimental and numerical results have been
proposed to validate this analytical formulation.
In this context, according to the Limit State design approach, the buckling verification
of a flat LG panel under in-plane compression, simply supported along the four edges,
could be reasonably developed by simultaneously considering two different conditions,
referred to requisites of deformability and durability.
The maximum transversal displacement wmax (Eq.(18), for example, should satisfy the
condition:
a
wmax d , (19)
with a the length of the LG panel and k an appropriate coefficient, calibrated for each
type of glass. For this purpose, the check of deformability should take into account a
reasonable amplitude w0 of initial sinusoidal imperfection, representative of possible
geometrical imperfections, eccentricities of load or boundaries, residual stresses. At the
same time, the design compressive load Ny,Ed should be opportunely limited:
N y ,cr ,lam
N y , Ed d N y ,b, Rd , (20)
J M1
with Ny,cr,lam given by Eq.(15) and JM1 an appropriate safety coefficient depending on the
mechanical properties of glass.
377
Challenging Glass 3
2.2. Buckling verification of 3-layer laminated glass panels

Further analytical calculations have been performed to extend the validity of the
proposed approach to the analysis of 3-layer LG panels (Figure 5).
Figure 5: 3-layer laminated glass element (cross section).
In particular, the performed comparisons showed that the critical buckling load of a 3-
layer in-plane compressed LG panel, simply supported along the four edges, can be still
precisely estimated by means of Eq.(15), by assuming for the equivalent thickness the
expression:
teq ,w 3
2t13 t 23 12*J s , (21)
whit * and O respectively given by Eq.(8) and Eq.(9). In this specific circumstance, also
the following expression should be considered:
ts t1 t 2 2tint , t s ,1 0.5t1 0.5t 2 tint , (22) (23)
2.18
Js 2t1t s2,1 , E 2.18 . (24) (25)
D2
Eq.(25), specifically, represents the fitting curve for the corrective coefficient E able to
provide sufficiently accurate results. At the same time, the load N–transversal
displacement w relationship of the compressed LG panel can be accurately described by
means of Eq.(18). Consequently, the buckling verification can be rationally carried out
by contemporarily satisfying the conditions given by Eqs.(19) and Eq.(20).
378
2.3. Numerical validation

To validate the proposed analytical approach, several numerical analyses have been
performed on 3-layer in-plane compressed LG panels simply supported along the edges,
having different geometrical or mechanical characteristics. Three different FE-models
have been carried out with the commercial non linear code ABAQUS [8], to calibrate
the coefficient E (Eq.(25)) and to check the accuracy of each FE-model. In all the
performed simulations, glass was described as an isotropic linear-elastic material
(E = 70000N/mm2, Q= 0.23), whereas to characterize the PVB-interlayer, some
experimental data available in literature have been taken into account [9].
In the first and more accurate three-dimensional FE-model (3D+shell), the external
glass sheets (thickness t1) have been modeled by means of shell elements (S4R). At the
same time, the middle glass sheet (thickness t2) and the layers of PVB-film (tint) have
been described through 3D-8 node elements (C3D8H, hybrid formulation, incompatible
modes). 3D elements and shell elements were connected together using the same nodes.
Moreover, a section offset toffset= t1/2 from the centroidal axis of each glass sheet was
applied to the external shell elements. In-plane compression was introduced in the FE-
model in the form of uniformly distributed compressive pressures acting on the upper
and lower surface of 3D elements. To avoid possible eccentricities, boundaries have
been applied to the central nodes of the middle glass sheet.
In the second FE-model (Mshell), the 3-layer composite section was described by means
of multilayer composite shell elements (S4R), by taking into account the real thickness
of each layer. The third FE-model (TEQshell), finally, consists in a monolithic glass
shell (S4R) having an equivalent thickness estimated by means of Eq.(21).
379
Challenging Glass 3
2500 2000
a= 1m x b= 1m a= 1m x b= 1m
6/1.52/6/1.52/6mm 6/0.76/6/0.76/6mm
2000 FULL 1600
Analytical (Eq.(15))
ABAQUS (3D+shell)
ABAQUS (Mshell)
1500 ABAQUS (TEQshell) 1200
Ny,cr,lam [kN]
Ny,cr,lam [kN]
ABS
1000 800
FULL
ABAQUS (3D+shell)
500 400 ABAQUS (Mshell)
ABAQUS (TEQshell)
ABS
0 0
-2 -1 0 1 2 3 4 -2 -1 0 1 2 3 4
10 10 10 10 10 10 10 10 10 10 10 10 10 10
Gint [N/mm2] Gint [N/mm2]
Figure 6: Critical buckling load Ny,cr,lam for in-plane Figure 7: Critical buckling load Ny,cr,lam for in-plane
compressed LG panels simply supported along the compressed LG panels simply supported along the
edges (6/1.52/6/1.52/6mm). edges (6/0.76/6/0.76/6mm).
1800 1500
a= 1m x b= 1m
6/1.52/6/1.52/6mm
1600 w0= a/500
a= 1m x b= 1m Gint= 8.06N/mm2
6/0.38/6/0.38/6mm 1250
1400
Ny,cr,lam= 903kN
1200 1000
Ny,cr,lam [kN]
1000
N [kN]
750
800
600 500
FULL
400 ABAQUS (Mshell) Analytical (Eq.(18))
ABAQUS (TEQshell) 250 Analytical (Eq.(18)), w0= 0
ABS ABAQUS (3D+shell)
200
ABAQUS (Mshell)
ABAQUS (TEQshell)
0 0
-2 -1 0 1 2 3 4
10 10 10 10 10 10 10 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0
Gint [N/mm2] w [mm]
Figure 8: Critical buckling load Ny,cr,lam for in-plane Figure 9: Load N-transversal displacement w
compressed LG panels simply supported along the relationship for in-plane compressed LG panels simply
edges (6/0.38/6/0.38/6mm). supported along the edges (6/1.52/6/1.52/6mm).
At first, buckling analyses were performed with the three FE-models to predict the
critical buckling load Ny,cr,lam of 3-layer LG panels and to validate the analytical
procedure (Eq.(15), with E given by Eq.(25)). The main results are proposed for
6/1.52/6/1.52/6mm, 6/0.76/6/0.76/6mm and 6/0.38/6/0.38/6mm LG panels (1m x 1m).
Apparently, the 3D+shell FE-model is the more accurate, but the modeling of the LG
panel and the performance of the buckling analyses require rather long processing time.
Furthermore, the 3D+shell FE-model tends to lightly overestimate the real critical
buckling load Ny,cr,lam in presence of soft thermoplastic films (Figures 6-7) and generally
has convergence problems if used in presence of extremely thin layers (Figure 8). In
contrast, the Mshell FE-model can be quickly implemented and buckling analyses can
380
be performed in a very short time, but the accuracy of results is very poor and the model
does not converge if the material characterizing the interlayer is very soft (Figures 6-8).
This aspect should not be ignored, especially in the verification of LG panels assembled
with PVB-films. In this context, the TEQshell FE-model constitutes a major
simplification, both in the modeling of the LG panel and in the performance of the
buckling analyses. Moreover, the TEQshell has no convergence problems associated
with the presence of extremely thin layers or very soft films. In Figure 7, also a
comparison between some load N–transversal displacement w curves are proposed for a
6/1.52/6mm LG panel (a=1m x b=1m), having assumed Gint= 8.06N/mm2 and w0=
a/500 [6]. Numerical results obtained by performing static incremental analyses are
compared with the analytical curve given by Eq.(18), with teq,w= 15.48mm (Eq.(21)). In
this circumstance, the 3D+shell FE-model overestimates the effective buckling
resistance of the LG panel and the Mshell FE-model strongly underestimates it (Figure
9); in addition the analysis stops for convergence problems when N Ny,cr,lam. In
contrast, the N-w relationship obtained with the TEQshell FE-model and the analytical
approach (Eq.(18)) are in good agreement.
3. Analytical models for compressed laminated glass columns
3.1. Buckling verification of 2-layer laminated glass columns

Concerning laminated glass columns subjected to compressive loads, several analytical
formulations are available in literature. Luible [2], for example, proposed for
compressed LG beams the classical theory of sandwich columns. Blaauwendraad [10]
recently compared some of the existing analytical models for buckling of LG columns,
highlighting the similarities of these sandwich-based theoretical approaches and
deriving a new approximate formulation able to control the transition between the
layered and the monolithic limits. In a previous effort [11], the authors presented a new
analytical model developed on the basis of the elastic theory originally proposed by
Newmark et al. [12] for the analysis of the flexural behavior of composite elements with
deformable connections. According to this exact theoretical model, let us consider a LG
beam (width b, buckling length L0) obtained by assembling two external glass sheets
(thicknesses t1, t2, Young’s modulus E, Figure 2) and a middle interlayer (thickness tint,
Young’s modulus Eint, shear modulus Gint), pinned at the ends of its buckling length L0.
The transversal displacement w(z) due to a compressive load N can be evaluated as:
§ (D 2 EJ abs L20 EJ full S 2 )L20 N ·

wmax w0 ¨1 2 ¸, (26)
¨ D EJ L2 ( EJ S 2 NL2 ) EJ S 2 ( EJ S 2 NL2 ) ¸
© abs 0 full 0 full abs 0 ¹
with:
K EJ full , Gint b
D2 K , (27) (28)
EA* EJ abs t int
( EA1 )( EA2 ) Ebt1t 2 Eb 3 3

EA* , EJ abs (t1 t 2 ) , (29) (30)
EA1 EA2 t1 t 2 12
381
Challenging Glass 3
ª § t t ·2 §t t · º
2
EJ full EJ abs Eb «t1 ¨ 1 int ¸ t 2 ¨ 2 int ¸ » , (31)

¬« © 2 2 ¹ © 2 2 ¹ ¼»
A1 bt1 , A2 bt 2 , (32)
and w0 the maximum amplitude of the initial sinusoidal imperfection.

In these hypotheses, the critical buckling load Ncr of the compressed LG beam is:
S 2 EJ abs EJ full § D 2 L20 S 2 ·

N cr ¨ ¸. (33)
L20 ¨ D 2 EJ L2 EJ S 2 ¸
© abs 0 full ¹
A rational buckling verification should be performed by limiting the maximum

deflection wmax of the LG beam (Eq.(26)) and the maximum design compressive load
NEd, as suggested in Ed.(19) and (20) for the verification of LG panels. In [11], the
validity of this exact theoretical model has been largely checked. However, the
analytical model applies only to 2-layer LG beams and to simple structural systems.
Furthermore, according to the verification approach proposed in the previous sections, it
might be interesting to propose a univocal formulation for the analysis of LG columns
and panels in different loading and boundary conditions. For this purpose, let us
consider the expression for the equivalent thickness proposed in Eq.(14). The critical
buckling load Ncr could be estimated, once tef,w is known, as:
S 2 Ebt ef ,w S2
3
N cr EJ ef . (34)
L20 12 L20
In these hypotheses, analytical comparisons have been performed by evaluating Ncr with
Eq.(33) and Eq.(34), with E= 1. Calculations were carried out by taking into account for
LG columns a series of mechanical and geometrical parameters, that is the value of Gint
(10-4 N/mm2 < Gint < 104 N/mm2), the ratio J = b/L0 (0.01 J 0.5, with 100 b 500),
the thicknesses of glass sheets and interlayer (6/0.38/6mm, 6/0.76/6mm, 6/1.52/6mm).
The main results are shown in Figures 10-11 for the 6/0.38/6mm and the 6/1.52/6mm
LG columns (with b= 0.1m), as a function of the ratio
N crNewmark
RN ,beam t
. (35)
N cref ,w
382
6.0 6.0
J= b/L0 6/0.38/6mm J= b/L0 6/1.52/6mm
5.5 0.01 b= 0.1m 5.5 0.01 b= 0.1m
0.05 0.05
5.0 0.10 5.0 0.10
0.20 0.20
4.5 0.30 4.5 0.30
0.40 0.40
4.0 0.50 4.0 0.50
3.5 3.5
RN, beam
RN, beam
3.0 3.0
2.5 2.5
2.0 2.0
1.5 1.5
1.0 1.0
Layered limit Monolithic limit Layered limit Monolithic limit
0.5 0.5
-4 -3 -2 -1 0 1 2 3 4 -4 -3 -2 -1 0 1 2 3 4
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
Figure 10: Critical buckling load Ncr for compressed Figure 11: Critical buckling load Ncr for compressed
LG columns (6/0.38/6mm, b= 0.1m). LG columns (6/1.52/6mm, b= 0.1m).
As shown in the proposed figures, the equivalent thickness approach does not agree
with the theoretical exact model for compressed LG columns. Moreover, the ratio
RN,beam depends on the thicknesses of the 3 layers constituting the LG beam, as well as
on the aspect ratio J= b/L0. In general, for a fixed aspect ratio J, RN,beam increases as the
thickness tint of the interlayer increases. At the same time, for fixed thicknesses of the
glass sheets and the interlayer, the maximum value obtained for RN,beam increases as the
aspect ratio J increases However, analytical calculations allowed to notice that the
fitting curve representing the values E able to give RN,beam= 1 is:
2
§b ·
E 1.03¨¨ ¸¸ . (36)
© L0 ¹
As a result, Ncr can be accurately evaluated by means of Eq.(34) and the load-carrying
behavior of the compressed LG beam can be described by taking into account the
classical theory of bending for monolithic columns:
w0
wmax , (37)
1 N N cr
with w0 the maximum amplitude of the initial sinusoidal imperfection affecting the
beam. In these hypotheses, the verification can be still carried out by contemporarily
satisfying the conditions expressed by Eqs.(19) and (20).
383
Challenging Glass 3
3.2. Buckling verification of 3-layer laminated glass columns

The equivalent thickness approach proposed for the analysis of 2-layer LG beams under
compression can be also easily extended to 3-layer composite columns. In the specific
circumstance, tef,w can be estimated by means of Eq.(21), with:
2
§b ·
E 2.06¨¨ ¸¸ , (38)
© L0 ¹
and the buckling verification can be performed by taking into account the Eqs.(19), (20).
3.3. Numerical validation

Also in this circumstance, to validate the proposed approach, three different FE-models
have been realized with the code ABAQUS. 3D+shell, Mshell, TEQshell models were
used to describe the flexural behavior of 3-layer LG columns having different
geometrical and mechanical properties. Numerical results obtained by buckling analyses
were compared with analytical predictions (Eq.(34)). The curves proposed in the
Figures 12-14, for example, are referred to 0.1m x 1m LG columns (6/1.52/6/1.52/6mm,
6/0.76/6/0.76/6mm, 6/0.38/6/0.38/6mm).
In general, in presence of soft thermoplastic films, the 3D+shell FE-model
overestimates the exact critical load Ncr (Figures 12, 13), although very thin layers
compromise the accuracy of results (Figure 14). The Mshell FE-model provides
sufficiently accurate results only for stiff interlayers, whereas it tends to underestimate
the critical load Ncr for LG columns with soft interlayers (0.5 N/mm2 < Gint < 10N/mm2)
and it does not converge for Gint < 0.5N/mm2. In contrast, the TEQshell FE-model
always provides precise results. In the Figure 15, a comparison between some load N-
deflection w curves are shown to confirm the accuracy of the equivalent thickness
approach.
384
60 50
b= 0.1m x L0= 1m
6/0.76/6/0.76/6mm
b= 0.1m x L0= 1m
50
6/1.52/6/1.52/6mm 40
40
30
Ny,cr,lam [kN]
Ny,cr,lam [kN]
30
20
20 FULL FULL
Analytical (Eq.(34)) Analytical (Eq.(34))
ABAQUS (3D+shell) ABAQUS (3D+shell)
ABAQUS (Mshell) 10 ABAQUS (Mshell)
10 ABAQUS (TEQshell) ABAQUS (TEQshell)
ABS ABS
0 0
-2 -1 0 1 2 3 4 -2 -1 0 1 2 3 4
10 10 10 10 10 10 10 10 10 10 10 10 10 10
Figure 12: Critical buckling load Ncr for 3-layer LG Figure 13: Critical buckling load Ncr for 3-layer LG
columns (6/1.52/6/1.52/6mm, 0.1m x 1m). columns (6/0.76/6/0.76/6mm, 0.1m x 1m).
40 40
b= 0.1m x L0= 1m
b= 0.1m x L0= 1m 6/1.52/6/1.52/6mm
6/0.38/6/0.38/6mm 35 w0= L0/500
Gint= 8.06N/mm2
30 30 Ncr= 32kN
25
Ny,cr,lam [kN]
N [kN]
20 20
FULL 15
ABAQUS (3D+shell)
10 10
ABAQUS (Mshell)
ABAQUS (TEQshell) Analytical (Eq.(37))
ABS ABAQUS (3D+shell)
5
ABAQUS (Mshell)
ABAQUS (TEQshell)
0 0
-2 -1 0 1 2 3 4
10 10 10 10 10 10 10 0 5 10 15 20 25 30 35 40 45 50
Gint [N/mm2] w [mm]
Figure 14: Critical buckling load Ncr for 3-layer LG Figure 15: Load N-transversal displacement w
columns (6/0.38/6/0.38/6mm, 0.1m x 1m). relationship for 3-layer LG columns
(6/1.52/6/1.52/6mm, 0.1m x 1m).
385
Challenging Glass 3
4. Conclusions
Because of their high slenderness, LG elements can be frequently affected by stability
problems. In literature, several analytical models derived from the theory of sandwich
panels are available, but in general they are applied only to 2-layer LG elements, as well
as in specific loading or boundary conditions. In the paper, a simple and accurate
analytical model based on the concept of equivalent thickness is proposed to evaluate
the critical buckling load and the corresponding load N-deflection w relationship of LG
panels and columns under in-plane compression. By means of appropriate correction
factors, the analytical model can be easily applied to 2 or 3-layer LG panels under in-
plane compression as well as to 2 or 3-layer compressed LG columns. Numerical
comparisons are proposed to highlight the accuracy and the potentiality of the approach.
5. References
[1] Zenkert, D, The handbook of sandwich construction, UK: Eng. Mat. Advisory Service Ltd., 1997.
[2] Luible, A., Stabilität von Tragelementen aus Glas, Thése EPFL 3014, Ecole Polytechnique Fédérale de
Lausanne (free download: http://icom.epfl.ch/publications), Lausanne, 2004.
[3] Wölfel, E., Nachgiebiger Verbund eine Näherungslösung und deren anwendungsmöglichkeiten,
Stahlbau 6/1987, 1987, p. 173-180.
[4] Bennison, S.J., Quin, M.H.X., Davies, P.S., High-performance laminated glass for structurally efficient
glazing, Innovative light-weight structures and sustainable façades, Hong Kong, 2008.
[5] Timoshenko, S., Theory of Plates and Shells, McGraw-Hill Book Company, 1989.
[6] Bedon, C., Amadio, C., Buckling of flat laminated glass panels under in-plane compression or shear,
Engineering Structures, 36 (2012), p.185-197.
[7] Wolmir, A.S., Biegsame platen und schalen, VEB Verlag für Bauwesen, Berlin, 1962.
[8] ABAQUS® version 6.9, Simulia, Pawtucket, R.I. 02860 USA, 2009.
[9] Bennison, S.J., Jagota, A., Smith, C.A., Fracture of glassy/poly(vinyl butyral) (Butacite®) laminates in
biaxial flexure. J. Am. Ceram. Soc. 1999; 82(7); 1761-1770.
[10] Blaauwendraad, J., Buckling of laminated glass columns, Heron, 52(1-2), 2007.
[11] Amadio, C., Bedon, C., Buckling of Laminated Glass Elements in Compression, Journal of Structural
Engineering, Vol.137, No. 8, August 1, 2011.
[12] Newmark, N.M., Siess, C.P., and Viest, I.M., Tests and analysis of composite beams with incomplete
interaction, Proc. Soc. Exp. Stress Anal., 9(1), 75-92, 1951.
386
doi:10.3233/978-1-61499-061-1-387
Contact Damage Near the Supporting

Pillars in Vacuum Glazing Units
Minxi Bao, JianYang,
University of Birmingham, UK, mxb949@bham.ac.uk
Xiaogen Liu, Yiwang Bao
China Building Materials Academy, P.R.China, xtlxq88@163.com
The contact damage caused by the supporting pillar on the glass surface in vacuum
glazing units is investigated. XFEM numerical modeling is employed to simulate
the cone crack initiation and propagation. The critical indentation loads in localized
uniform pressure and that for a cylindrical indenter are calculated and compared.
The stress distributions due to the indenters of different geometries are investigated.
In light of the numerical results, an improved design of supporting pillar is
recommended, which is able to mitigate the severe concentration at the contact rim.
Keywords: Indentation, crack, contact stress, vacuum glazing units, XFEM
1. Introduction
Significant progress has been made on the development of vacuum glazing units after
this type of glass structure was first invented in 1913[1]. However, the high fracture
occurrence due to the construction features of vacuum glazing units is still a
predominant challenge in engineering application. As the test carried out in a previous
study, the strength of vacuum glazing unit is only 40% or 50% of that of normal
monolithic glass sheet [2]. This is because the barometric pressure induces severe stress
concentration at the contact area between the glass sheets and supporting pillars.
Consequently, the contact damage caused by the indentation of pillars would result in
the breakage under low wind or snow loads. The stress concentration at a pillar of the
vacuum glazing units was observed by polariscope, as shown in Figure 1. As is well
known, a cone cracks will occur due to the excessive indentation forces, as shown in
Figure 2a and b.
This contact damage depends on the indenter geometry, but it is still unclear of which
type of indenter is better, i.e. causing minimum contact damage. In the present paper, an
XFEM numerical method is introduced to simulate the crack initiation and propagation
due to the indentation load. The contact stress resulted from indenters are examined. In
order to optimize the pillar geometry, the stress fields at the contact area with uniform,
cylindrical and spherical indentation loading conditions are investigated, respectively.
387
Challenging Glass 3
(a) Cylindrical indenter (b) Spherical indenter

Figure 1: The stress concentration Figure 2: The cone cracks caused by the cylindrical and spherical
near the pillar indenters respectively [3, 4].
2. The Mean Strength Theory

Hertz[5] first published a research on indentation fracture in 1881. An elastic linear
analysis conducted by him indicated that the surface principle stress 1 at the edge of
contact circle was the maximum radial tensile stress. Hertz demonstrated that when the
applied indentation load was sufficient, the initiation of the cone crack beneath the
indenter would occur at wherever the principle stress 1 was greatest. Frank and Lawn
[6] pointed out that the crack path should start at the circumference of the contact circle
and follow the trajectory of the third principle stress outside the contact zone.
In indentation tests, a peculiar phenomenon is discovered in a great number of tests,

which contradicts the traditional strength theories mentioned above: a brittle material
does not always fracture in a high stress concentration areas even where the peak stress
is greater than the critical strength. Bao [7] proposed a Mean Strength Criterion (MSC)
in 1993 to address the problems. He stated that instead of a point stress on the surface of
the specimen, it is the maximum mean stress near the contact circle that determines the
crack initiation. The crack will not take place until the mean stress over a small process
zone in thickness direction reaches a critical value. The mean stress is calculated by
equation 1.
1 '
V ³ V dz
' 0 1 (1)
Where is the mean stress, 1 is the first principle stress near the contact circle; ' is the
integral thickness, which can be obtained by equation 2.
' 2 / S ( K IC / V 0 ) 2 (2)
where, 0 is the localized bending strength, KIc is known as the plain strain fracture
toughness, and can be readily measured.
An empirical equation deduced from the experiments is introduced [7] to calculate the
critical indentation load.
388
Contact Damage Near the Support Pillars in Vacuum Glazing
V 0Sa 2
Pc
[0.1373a 2 0.2862a 0.0236]
(3)
where Pc is the critical indentation load in N, a is the radius of contact area caused by a
spherical indenter.
In previous research [2-4], MSC has been used to determine the critical load of brittle
material, and good agreement has been discovered with experimental data. In this study,
it will be adopted to validate the numerical modeling.
3. Simulation of Crack Initiation and Growth with XFEM
3.1. Introduction of XFEM

In order to further study the indentation damage for vacuum glazing, numerical method
is employed. Many documented FEA simulations do not involve the crack initiation, so
it is difficult to predict the critical indentation load for crack initiation. In traditional
discrete cracking modeling methods, possible crack boundaries requires to be pre-
defined to avoid splitting elements, as cracks cannot appear within elements.
Nevertheless, it is difficult to precisely predefine the dimension and the location of the
cone crack unless sufficient experimental data are provided.
In this study, the XFEM module integrated in the ABAQUS is utilized to simulate the
crack initiation and propagation. XFEM is termed for “extended finite element method”,
which extends the piecewise polynomial function space of conventional finite element
methods with extra enrichment terms. Different from conventional crack simulations, it
allows arbitrary cracks independent of the mesh, and the discontinuous elements
separated by the crack. The incorporated Heaviside enrichment term enables the
displacement to jump cross crack. Therefore, two discontinuous elements can be
deemed as a superposition of two continuous elements with “phantom nodes”. Detailed
introduction of the methodology on XFEM simulation can be referred in the
ABAQUS/CAE user's manual [8].
3.2. The indentation modeling with uniform load

A flat glass disk with the radius 30mm and the thickness 4 mm is established as an
asymmetrical model. The indentation load is assumed to be uniform in this model. The
schematic diagram is depicted in Figure 3.
Figure 3: The schematic diagram of the axisymmetric model
389
Challenging Glass 3
The linear elastic material properties of tempered glass are assigned to the model; i.e.
Young’s modulus E= 70000MPa, Poison ratio = 0.24. In the XFEM simulation, the
initial crack will pass through an element in which the maximum principle stress
reaches a critical value. This critical maximum principle stress is defined as the
localized strength based on the MSC, and the element thickness is selected as the
integral thickness. The model is simply supported, and a uniform pressure is applied on
the central circular area, representing the indentation load. The indenter radiuses
modeled are 0.1mm, 0.15mm, 0.2mm, 0.25mm and 0.3mm respectively. The element
type CAX4R is a four-node bilinear quadrilateral with reduced integration.
Figure 4 present an image showing a cone crack caused by a spherical indenter. The
crack simulated in the XFEM analysis is shown in Figure 5 with a large scale factor.
The simulated crack is propagating in a cone-shape, which is in good accordance with
the experiment result.
Figure 4: The cone crack caused by spherical indenter[4] Figure 5: The cone crack section simulated by
the XFEM in axial symmetric mode.
It is noted in the simulation results that the crack always initiates at a circle larger than
the loading area, as presented in Figure 6.
surface stress
Mean stress


V p m
1.0 1.1 1.2 1.3 1.4 1.5

r0 / a

Figure 6: The crack circle larger than the loading area Figure 7: The surface stress and the mean stress
outside the contact area [4]
The large crack circle has also been observed and mentioned in the previous laboratory
work [9]. In theory, the crack is supposed to take place at the edge of the loading circle,
where 1 is calculated to be the maximum. The MSC is used to explain the contradiction.
As it is stated above, the crack initiation is determined by the maximum mean stress.
The distribution of the surface stress and the mean stress starting from the edge of
contact circle is depicted in Figure 7. The maximum surface stress is at 1.0a,, i.e., at the
390
edge of the loading area, while the maximum mean stress is found to be at about 1.1a,
where the crack initiation is likely to occur.
The history of the maximum local principle stress 1 versus the indentation pressure for
different loading radiuses are captured at the origin of crack, as shown in Figure 8. The
stress will drop immediately once the crack is formed, and experience a re-bounce
within a narrow range before completely drop to the zero value. The maximum value of
1 is always located at the tip of the crack to stimulate crack propagation, and the
previously formed crack does no longer carry any load. An interesting result is found
that the critical pressures for different loading radiuses remain almost consistent,
independent of loading radiuses.
Figure 8: The critical indentation pressure of different radiuses
The critical force can be obtained by integrating the critical pressure in the contact area.
The results of analytical approach and XFEM modeling are listed in Table 1, and the
curves are presented in Figure 9. It is found that the modeling results are in excellent
agreement with the MSC and the applicability of XFEM numerical approach is
validated. It is noted that the empirical equation developed by Bao [7] is based on the
spherical indentation tests. Therefore, the uniform indentation loading simulation is
proved to be very similar to the spherical indenter.
Table 1: The critical indentation loads derived from analytical method and XFEM simulation.
Calculated critical indentation load (N)
Radius(mm)
Mean strength equation XFEM modeling
0.1 185.64 186.14
0.15 419.73 418.56
0.2 747.45 743.66
0.25 1169.85 1161.27
0.3 1699.29 1671.21
391
Challenging Glass 3
Figure 9: The critical indentation load of different radii.
3.3. The crack simulation considering the contact stress

The support pillars adopted in vacuum glazing units are usually treated as cylindrical
indenter. Instead of the uniform loading applied in the previous modeling process, the
contact stress distribution induced by the cylindrical indenter is non-uniform. It is
difficult to derive analytical equations of real contact stress due to the complexity of
contact problems. So the numerical method is employed to explore the indentation
damage caused by real contact stress.
The cylindrical indenters are modeled with the radius 0.1mm, 0.15mm, 0.2mm, 0.25mm
and 0.3mm. The material properties of steel with E =175000 N/mm2, = 0.3 is assigned
to the pillars. A uniform pressure is applied on the top of the pillar. As for the contact
interface, the pillar surface is deemed as the master face and the contact glass is set as
the slave surface. To enhance the convergence of the simulation, the contact interface is
assumed to be frictionless. Figure 10 shows the critical indentation loads yielded by the
uniform forces and cylindrical indenters. The critical indentation loads of cylindrical
indenters are reduced significantly. When the contact stress is considered, the
indentation pressure is no longer a constant, but influenced by the size of indenters.
392
Figure 10:The critical loads comparison between uniform load and cylindrical indenter
Since the contact stress of cylindrical indenter is found to have an important impact on
the critical indentation load, it implies that the load bearing capacity of the vacuum
glazing is also affected by the geometry of indenters. Therefore, the stress distributions
in the glass resulted from three different loading conditions i.e. uniform pressure,
cylindrical indenter and spherical indenter are now further compared. Assuming the
same contact radius of three models, the maximum principle stress distributions at the
critical state are presented in Figure 12. Where, the horizantal axis represents the
distance from the centre of the indenter on the path indicated in Figure 11. Symbol “a”
denotes the radius of loading area, and the normalized stress for vertical axis is
expressed as V n V / V max .
Figure 11: The schematic Figure 12: The maximum principle stress distribution of uniform pressure,
diagram of the indentation cylindrical indenter and spherical indenter at critical state
393
Challenging Glass 3
As shown in Figure 12, the stress distribution of uniform pressure and spherical indenter
outside the contact area are very similar and both decreasing slowly. Underneath the
indenter, the spherical indenter exhibits a smoothly increasing curve, wherease severe
stress concentration takes place at the contact edge of the cylindrical indenter, and
results in a sudden peak stress outside contact circle. The integration of the maximum
principle stress beneath the cylindrical indenter reaches only one third of that under the
spherical indenter or uniform pressure. As the stress integration beneath the indenters at
critical state reflects the critical indentation pressure, it can be concluded that the
adoption of cylindrical pillars in vacuum glazing will lead to an earlier occurrence of
contact damage due to the intensive stress concentration near the edge of contact ring.
A novel pillar design is therefore introduced (see Figure 13). Instead of the conventional
flat surface, the contact section of the improved pillar is produced to be a curved surface
in order to reduce the stress concentration. The radius of the curve surface is determined
by the equivalent spherical indenter.
Figure 13: A novel pillar design with curved contact surface.
4. Conclusions
The contact damage triggered by the support pillar is studied. The XFEM numerical
method is employed and the applicability is validated by an analytical method using
MSC. It is demonstrated that the ring crack appears outside of the contact area, rather
than at the edge of the contact area where the surface stress is in the maximum.
The critical indentation loads with uniform pressure, the cylindrical indenter and
spherical indenter are compared. It is found that the uniform pressure model can be used
to represent the spherical indenter in determining the critical indentation load. The
critical indentation pressure applied on the pillar becomes dependent of the value of
contact radius when the real contact stress distribution is considered.
The stress distribution within and outside the contact zone under the uniform pressure, a
cylindrical indenter and a spherical indenter are presented. The results show that the
stress concentration due to a cylindrical indenter is more severe than that from the
sphere indenter or uniform pressure. So the critical load for indentation damage is much
lower than the latter. Based on the obtained conclusions, an improved pillar design is
proposed, which can be adopted in the engineering practice for the future manufacturing.
394
5. Acknowledgements
This work is supported by the international collaboration project (S2011ZR0397).
6. References
[1] Collins, R.E, Simko, T. M, Current status of the science and technology of vacuum glazing, Solar
Energy 62 (3), 1998, pp189-213.
[2] Liu, Xiaogen; Safety evaluation and failure detection of glass curtain wall, PhD Thesis, China Building
Material Academy, Beijing, China, 2005.
[3] Liu, Xiaogen; Bao, Yiwang; Xu, Haifeng; Wang, Xiufang; Propagation mechanism and control
criterion of fracture indentation in vacuum glazing, Materials Science and Technology, 6/2010,pp878-
882
[4] Bao, Y.W; Gao, S.J., Local strength evaluation and proof test of glass components via spherical
indentation, Journal of Non-Crystalline Solids, 2008, pp1278-1381
[5] Hertz, Heinrich, Hertz’s Miscellaneous papers, Nature 55, 11/1896, pp.6-9.
[6] Frank, C.K,; Lawn, B.R.; On the Theory of Hertzian fracture, Proceedings of the Royal Society,
Cambridge, Britain, 1967.
[7] Bao.Y; Jin, Z; Size effects and a mean-strength criterion for ceramics, Engineering Structure and
Materials, 8/1993, pp829-935.
[8] Hibbitt; Karlsson; Sorensen, Abaqus/CAE User's Manual, Pennsylvania State, USA, 2010
[9] Mouginot, R., Crack formation beneath sliding spherical punches, Journal of Materials Science,
22/1987, 989-1000
395
doi:10.3233/978-1-61499-061-1-397
Towards a European Structural Glass

network: COST Action TU0905
Jan Belis
Ghent University - LMO, Belgium, jan.belis@UGent.be, www.glassnetwork.org
Jürgen Neugebauer
FH Joanneum, Austria, juergen.neugebauer@fh-joanneum.at
Jens Schneider
TU Darmstadt, Germany, schneider@iwmb.tu-darmstadt.de
Mauro Overend
University of Cambridge, UK, mo318@cam.ac.uk
Danijel Mocibob
RI ISA (Permasteelisa Group), Croatia, d.mocibob@permasteelisagroup.com
COST Action TU0905 “Structural Glass: novel design methods and next generation
products” provides a platform for European scientists to cooperate and exchange
expertise in the research area of Structural Glass. The main objective of this Action
is to provide a strong contribution to the ongoing development of innovative high
performance structural glass products mainly in architectural and solar applications,
and to European standards and education in this field. This contribution intends to
give an overview of the context, organization, objectives and activities of this
successful network, which currently is about mid-term.
Keywords: COST, Action TU0905, Structural Glass, European Network,

Education Pack, Short Term Scientific Mission, Training School.
1. Introduction
Prior to zooming in to the European Structural Glass network provided by COST Action
TU0905, it is important to know the context of the COST framework, which is
consequently explained in this introduction. The information in the following
subsections is largely taken from the COST website [1].
1.1. About COST

COST is an intergovernmental framework for European Cooperation in Science and
Technology, allowing the coordination of nationally-funded research on a European
level. COST has a very specific mission and goal. It contributes to reducing the
fragmentation in European research investments and opening the European Research
Area to cooperation worldwide.
As a precursor of advanced multidisciplinary research, COST plays a very important

role in building a European Research Area (ERA). It anticipates and complements the
activities of the EU Framework Programmes, constituting a “bridge” towards the
scientific communities of emerging countries. It also increases the mobility of
397
Challenging Glass 3
researchers across Europe and fosters the establishment of scientific excellence in nine
key domains, among which “Transportation and Urban Development” (TUD).
1.2. About COST Actions

COST does not fund research itself but provides a platform for European scientists to
cooperate on a particular project and exchange expertise. These projects are called
"Actions". Each COST Action is a network centred around nationally-funded research
projects in fields that are of interest to at least five COST countries. COST provides the
COST Actions with financial support for joint activities such as conferences, short-term
scientific exchanges and publications. Each COST Action has an objective, defined
goals and clear deliverables. COST Actions have located their topic into one of COST’s
scientific Domains.
One of COST's main characteristics is its flexibility, allowing for an easy

implementation and light management of the research initiatives. Activities are
launched following a "bottom-up" approach, meaning that the initiative of launching a
COST Action comes from the European researchers themselves. After a competitive
selection procedure and peer review process, a number of successful proposals are
awarded with COST support.
A Memorandum of Understanding (MoU) provides the formal basis of an Action.

The member countries participate on an "à la carte" principle, in that only countries
interested in the Action participate. An Action is launched when at least five COST
member states have agreed the MoU and starts with the first Management Committee
meeting of the Action. It runs for an average of four years.
2. COST Action TU0905 “Structural glass: novel design methods and next
generation products”
This section summarises the objectives of the Action and explains its origin. In addition,
it gives an overview of its organisational structure and responsibilities at the time of
writing. Additional information on the COST Action on Structural Glass can be found
on the Action’s website [2].
2.1. Objectives of the Action

The main objective of this Action is to provide a strong contribution to the ongoing
development of innovative high performance structural glass products mainly in
architectural and solar applications, and to European standards in this field. The Action
will identify and share the outcomes of existing fragmented activities within the
European research community. In addition, the Action will establish a diverse multi-
disciplinary network that will encourage new research and collaborations. Finally, the
Action will strengthen the current and future generations of European glass designers by
developing a structural glass education pack for university curricula across Europe.
2.2. Origin of the Action

The initiative of this Action on Structural Glass was taken by the five authors of this
article, involving already right from the start international involvement of five European
countries (Austria, Belgium, Croatia, Germany, United Kingdom). Several of the
398
Towards a European Structural Glass network: COST Action TU0905
initiators had priory been involved in WG3 “Structural Glass” of COST Action C13
“Glass and interactive building envelopes”, which ended in 2005.
Action TU0905 “Structural glass: novel design methods and next generation products”
was launched on January 21, 2010 and started with the kick-off meeting in Brussels on
April 7, 2010. Consequently, the expected end date of the Action currently is April 6,
2014, although COST foresees the possibility to apply for an extension of the Action
duration.
2.3. Structure and members of the Action

During the kick-off meeting in Brussels, the Chair, Vice Chair and Grant Holder were
elected by the Management Committee (MC) members. Furthermore, the Domain
Committee Rapporteur, the COST Science Officer and the Administrative Officer play
an important role in the evaluation and/or support of the Action.
Chair, Vice Chair, Rapporteur and WG Chairs together form the so-called Core Group
(CG) of the Action, see Table 1.
Table 1: Overview of Core Group (CG) and COST Officers of COST Action TU0905.
NB: WG = Working Group
Name Function
Core Group Prof Jürgen Neugebauer (AT) MC Chair
Prof Jan Belis (BE) MC Vice Chair
Prof Jens Schneider (DE) Grant Holder
WG2 Chair
Prof Nina Penkova (BG) WG1 Chair
Dr Mauro Overend (UK) WG3 Chair
Dr Christian Louter (CH) WG4 Chair
Prof Kiril Gramatikov (MK) Domain Committee Rapporteur
COST Officers Dr Thierry Goger COST Science Officer
Ms Carmencita Malimban COST Administrative Officer
Four Working Groups (WG) are defined in the Memorandum of Understanding (MoU),
based on four priority areas detected to be critical for the further development of
European Structural Glass research. An overview of the WG’s and their respective
Chairs is given in Table 2.
Although not described in the original MoU, 13 Task Groups (TG) have been installed
as further subgroups of the WG’s during later MC meetings in Madrid and Sofia. The
main objective of installing the TG’s was to assess and further break down specific
tasks more easily in relatively small groups of experts. An overview of the TG’s and
their respective leaders is also listed in Table 2.
Finally, the 25 members states listed in Table 3 have joined the Action at the date of
writing, corresponding to 91 contributing members (WG and MC). More details are
available on the COST Action website [2].
399
Challenging Glass 3
Table 2: Overview of Working Groups (WG) and Task Groups (TG) of COST Action TU0905
(status on April, 2012).
WG TG Name TG Leader
1 Predicting complex loads on glass structures
1 Thermal Fracture Mr Marc Vandebroek (BE)
2 Insulating Glass Units Prof Jürgen Neugebauer (AT)
3 Dynamic Load Combinations Dr Olga Río (ES)
2 Material characterisation and material improvement
4 Thermal Tempering Prof Reijo Karvinen (FI)
5 Glass Strength Ms Maria Lindqvist (CH)
6 Interlayers Dr Gérard Savineau (FR)
3 Post-Failure Performance
7 Numerical Know –how and validation Mr Martin Larcher (DE)
8 Structural Design Philosophy Dr Mauro Overend (UK)
9 Learning from Failure Mr Daniel Honfi (HU)
4 Novel Glass Assemblies
10 Connections Prof Frank Wellershoff (DE)
11 Architectural Geometries Dr Ognen Marina (MK)
12 Stability Prof Jan Belis (BE)
13 Hybrid Components Dr Christian Louter (CH)
Table 3: Overview of member states which joined COST Action TU0905

(in alphabetical order; status on April, 2012).
Nr Member State Nr Member State
1 Austria 14 Lithuania
2 Belgium 15 Luxembourg
3 Bulgaria 16 Malta
4 Croatia 17 Netherlands
5 Czech Republic 18 Portugal
6 Denmark 19 Serbia
7 Finland 20 Slovenia
8 France 21 Spain
9 Germany 22 Switzerland
10 Greece 23 The Former Yugoslav Republic of Macedonia
11 Hungary 24 Turkey
12 Israel 25 United Kingdom
13 Italy
400
3. Activities
In this section, a short summary is presented of the activities of the Structural Glass
COST Action up till the submitting deadline of this article.
3.1. Meetings
Typically, minimum two Action meetings are organized every year, one of which is
(partly) devoted to a MC meeting, whereas the other usually is a WG or TG meeting.
Since the kick-off meeting in Brussels, six meetings have been organized up till the date
of writing this article, including two MC+WG meetings (in Düsseldorf and Sofia,
respectively), one WG meeting (in Madrid), one Core Group meeting (in Copenhagen)
and two TG meetings (in Cambridge and Graz, respectively).
3.2. Short Term Scientific Missions

Specifically to stimulate and help early stage Structural Glass researchers, the latter can
apply for so-called Short Term Scientific Missions (STSM) within the Action. The main
purpose of STSM’s is to provide financial support to young researchers for a stay at a
research institute in another member state, devoted to research which contributes to the
global objectives of the COST Action.
Up to date, six STSM’s have been approved within the Action, e.g. to the University of
Cambridge, Ghent University (UGent), Technical University of Denmark (DTU) and
Ecole polytechnique fédérale de Lausanne (EPFL). STSM topics included the
improvement of methods for numerical modeling of adhesively bonded joints,
reinforced glass beams, strength of monolithic glass beams in bending, hybrid glass
beams, and the Education Pack.
Several completed STSM’s have resulted in successful publications of research papers,

some of which have been listed at the end of this paper [3-6].
3.3. Education Pack

In contrast to more traditional building materials such as concrete or steel, structural
analysis courses related to glass usually are not (yet) part of structural, architectural or
civil engineering educational programs. However, high-level education obviously is

essential to further develop structural glass research and its further implementation in
practice by qualified engineers.
Consequently, one of the major objectives of this Action is to develop an Education

Pack based on the wide expertise and know-how available in a large variety of very
specialized subdomains in the area of Structural Glass. The final version of the
Education Pack is intended for academic education at universities throughout all
participating member states.
In general, the Education Pack is subdivided in three large parts, being A) Materials and
glass products, B) Components and connections, and C) Structural glass systems. In
addition, a fourth part is foreseen, called D) “Glossary of terms”. Each part is
coordinated by a responsible, but the content is basically produced and delivered at TG
level by small groups of specialists.
401
Challenging Glass 3
Given the availability of certain existing textbooks, e.g. [7] and the short-term need of
readily usable educational support for new courses triggered by and originating during
the COST Action, it was a MC decision to give the highest priority to a series of
elaborated lecture presentations. Currently a first draft of presentations on a large
variety of subtopics has already been produced, according to a uniform template created
especially for the Education Pack. The presentations are intended as an extensive
database (currently still subject to further development and changes) from which
teachers can select specific sections according to the specific target audience and aims
of their own custom made course.
However, even if this is already a huge and very useful step forward, it is a real
challenge not to lose essential information during transferring the presentations from the
authors towards other users. Consequently, in the longer run the Action has the ambition
to produce a new textbook which supports the final version of the presentations. As
such, an integrated Education Pack will hopefully be ready to be released by the end of
the Action.
3.4. Training School

Closely after the deadline of the current article, a first international Training School on
Structural Glass is scheduled which will be organized by and at Ghent University,
Belgium. The target audience of the Training School consists of beginning researchers
at early PhD or advanced Master level. Participants of over 20 countries are expected,
and during a full week (April 2-6, 2012) they will be submerged in the domain of
Structural Glass by lectures coming from all over Europe, but also by a student
colloquium, company visits, laboratory demonstrations, a workshop, and technical
excursions.
A more elaborated information report on the Training School is intended to be released

in a future mini-symposium on Structural Glass initiated by COST Action TU0905 in
cooperation with IABSE, which will be held in Portugal [8].
4. Conclusions and outlook

COST Action TU0905 “Structural glass: novel design methods and next generation
products” wants to help avoiding fragmentation in research efforts, to build a strong
professional network and to ensure the leading position of Europe in Structural Glass
research and education. The Action is currently ongoing and has an expected end date in
April, 2014.
The Action has a clear organizational structure now and currently counts over 90
individual members coming from 25 member states. Up till now the Action has
organized a number of successful activities, ranging from Action meetings over Short
Term Scientific Missions to a Training School (expected April 2012). One of the major
deliverables will be an Education Pack on Structural Glass, which will be made
available to universities of participating member states after the end of the Action and
which is expected to boost academic education in this young and challenging research
domain.
402
5. Acknowledgements
The authors would like to acknowledge COST Action TU0905 “Structural glass – novel
design methods and next generation products”, which provides excellent networking
and feedback opportunities in the field of Structural Glass and which is of great value
for the efficient further development of this challenging research domain. They would
also like to thank all members of COST Action TU0905 for their continuous efforts to
make this Action a success.
6. References
[1] http://www.cost.eu
[2] http://www.glassnetwork.org
[3] Louter, Christian; Belis, Jan; Veer, Fred; Lebet, Jean-Paul, Structural response of SG-laminated
reinforced glass beams; experimental investigations on the effects of glass type, reinforcement
percentage and beam size, Engineering Structures, Vol 36, March 2012, p. 292-301.
[4] Louter, Christian; Belis, Jan; Veer, Fred; Lebet, Jean-Paul, Durability of SG-laminated reinforced
glass beams: effects of temperature, thermal cycling, humidity and load-duration, Construction and
Building Materials, Vol 27, Issue 1, February 2012, p. 280–292.
[5] Vandebroek, Marc; Lindqvist, Maria; Belis, Jan; Louter, Christian, Edge strength of cut and polished
glass beams, Proceedings of Glass Performance Days, Finland, 2011, p. 476-479.
[6] Lindqvist, Maria; Vandebroek, Marc; Louter, Christian; Belis, Jan, Influence of edge flaws on failure
strength of glass, Proceedings of Glass Performance Days, Finland, 2011, p. 126-129.
[7] Haldimann, Matthias; Luible, Andreas; Overend, Mauro, Structural Use of Glass. Structural
Engineering Documents 10, IABSE, Zürich:2008.
[8] http://www.icsa2013.arquitectura.uminho.pt/
403
doi:10.3233/978-1-61499-061-1-405
How to Model Failure in Load-bearing

Glass Elements? A Discussion Based on
Analytical, Numerical and Experimental
Considerations
Fabrice Bernard
Université Européenne de Bretagne, Institut National des Sciences Appliquées de
Rennes, France, Fabrice.Bernard@insa-rennes.fr
Bouazza Fahsi, Baghdad Krour
Université de Sidi Bel Abbes, Algérie
Glass is a material that has been used for a long time in windows as a filling
material and has much to offer in this regard due to its possibility to carry high
compressive stresses. For several years, there has been a trend in architecture to use
glass not only as a part of the building envelope, but also as material for load-
bearing elements. This represents a special challenge because of the glass
brittleness. Knowing how to model the failure of such structures is then a very
important challenge and can contribute to a decrease of the partial safety
coefficients used in the design. Most of time, such a modeling needs to be based on
a statistical approach. This is the case for glass beams in current zones. The present
contribution focuses, in a first part, on the modeling of failure in annealed glass
single panes using Weibull model extended to take into account the subcritical
cracking. Then, in a second part, the modeling of laminated glass beams with a
SGP interlayer is presented. This modeling is performed with the FE software
Abaqus and takes into account the mechanical contribution of the interlayer thanks
to a Mooney-Rivlin model. The special challenge is here to reproduce the post-peak
behavior of the laminated glass beam and then the remaining load carrying capacity
of the structure. The third part of the paper deals with the failure in connection area.
Thanks to a combination of FE modeling, experimental campaigns and microscopy
observations (with optical and electronic microscopes) the deterministic aspect of
the failure in these special zones is put into evidence. This important result enables
to simplify the modeling of the mechanical behavior of such area
Keywords: Glass, load-bearing elements, connections, failure
1. Introduction
Glass is a material that has been used for a long time in windows as a filling material
and has much to offer in this regard due to its possibility to carry high compressive
stresses. For several years, there has been a trend in architecture to use glass not only as
a part of the building envelope but also as a material for load-bearing elements, i.e.
beams, columns or shear walls… This represents a special challenge because of the
glass brittleness. Indeed, new applications of glass in such structural parts need a good
knowledge of the load-bearing capacity, of its post-breakage behaviour and the lifetime
of the structural glass components.
405
Challenging Glass 3
Thus, knowing how to model the failure of such structures is a very important challenge
and can contribute to a decrease of the partial safety coefficients used in the structural
design process.
This communication investigates this point through the full study of three different glass
elements: a single-pane glass beam, a laminated glass beam and the connection area in
glass structures. Only the failure modes by overstresses are considered here.
2. Study of a single-pane glass beam
2.1. Statistical model

Glass does not possess an inherent ductility and disintegrates after failure. Moreover,
the sensitivity of the material to the damage of its surface (weak toughness) leads to
statistical and size-dependent failure behaviour. For example, the failure stress of a
macroscopic annealed glass element under tension is between 30 and 100 MPa whereas
the tensile stress of a glass fiber is more than 10 GPa. Griffith [1] explained this by the
presence of microcracks. Their origins are located on the machined edges. Failure is
then governed on one hand by their propagation and on the other hand by their random
distribution.
The Weibull model [2] is a statistical approach for the failure analysis of brittle
materials with random flaws. The failure probability Pf of a single glass plate is given
by:
ª 1 §V Vu ·
m
º
Pf 1 exp «
«¬ S 0
³S ¨¨© V 0 ¸¸ dS »
¹ »¼
(1)
where S0 is the unit area, S is the surface with flaws, V is the threshold stress (for Pf =0)
and m and V0 are the parameters of the Weibull law.
For a glass plate under bending, S is the polished surface under tension. It is well known,
so the integration of the law is explicit.
The Weibull model accounts for the size effect and the stress concentration effect.
Weibull parameters depend not only on the material but also on the loading rate. Indeed
glass is subjected to a subcritical crack growth phenomenon, especially when annealed
glass is considered.
2.2. Subcritical crack growth and how to take into account?

In classical Linear Elastic Fracture Mechanics (LEFM), failure due to the propagation
of cracks from the edge can be modeled by means of the stress intensity factor. As glass
is a brittle material, its fracture mode can be considered as a pure mode I. Instantaneous
failure occurs when KI reaches and exceeds the critical stress intensity factor, also called
the material toughness.
406
How to Model Failure in Load-bearing Glass Elements? A Discussion Based on Analytical, Numerical and
Experimental Considerations
This condition is called Irwin’s fracture criterion and is expressed as follows:
K I t K Ic (2)
A typical value for KIc for silica-soda -lime glass is 0.75 MPam.
However, in glass, a crack propagation occurs even if K I d K Ic . This subcritical crack
growth phenomenon under static loads, originally called “static fatigue” was discovered
by Grenet [3]. As revealed after by [4], this phenomenon is not observed in vacuum
conditions and is due to the effect of moisture at the crack tips.
This phenomenon explains also the dependence of glass on the rate and the duration of
the loading.
Figure 1 illustrates the subcritical crack growth and presents the cracking velocity
according to the stress intensity factor.
v (m/s)
Kcb
KIth KIc
10 3
III
1
II
Water
10 -3
Air 50 % R.H.
Air 25 % R.H.
10 -6 I
Vacuum
10 -9
10 -12
0 0,5 1,0 1,5 2,0 2,5 3,0
KI (MPa.m ) 1/2
Figure 1: Cracking velocity according stress intensity factor ([5], after [4])
407
Challenging Glass 3
The presence of the stress corrosion threshold KIth has not been clearly demonstrated or
put into evidence until now. A conservative (=safe) assumption would be to neglect it.
Evans [6] proposed a model including the dependence of the cracking velocity with the
stress intensity factor (region I, figure 1):
da
v AK In (3)
dt
with A and n the parameters of the law (A=0.124 and m=12.76) [5] for silica-soda-lime
glass, KI the stress intensity and a the flaw size.
The association of the Weibull and Evans models enables then to account for the
subcritical crack growth in the statistical analysis. The apparent Weibull parameters (m
and V0) depend on both the loading rate and the environment for a given edge finishing.
Intrinsic parameters m* and V0* can be defined for the strength in inert media. These
intrinsic parameters are independent of the loading conditions.
For a bending test, the stress intensity factor is equal to:
KI VY a (4)
where a is the flaw size, Y the shape factor and V the applied stress.
The Evans law leads to:
da n
A VY a (5)
dt
Assuming (1) a constant loading rate V =constant leading to V V t f (where tf denotes

the service life), (2) that the initial flaw length is small compared to the final crack
length ( a(t 0) a(t t f ) ), and (3) that KIth=0 as previously explained, and
integrating in time the previous equation, the failure probability can be expressed in
function of the Weibull intrinsic parameters and the Evans ones [5]:
ª § V ·m º
Pf 1 exp « ³ ¨¨ ¸¸ dS »
«¬ S © V 0 ¹ »¼
(6)
ª nm*
n2 § t · n2
m* m* º
« § V · § 1 · n2 »
1 exp ³ ¨¨ ¸¸ ¨¨ ¸¸ ¨ f
« ¸ dS »
© V * ¹ t
© ¹ © n 1 ¹
«¬ S 0 0
»¼
408
with m, V0 the Weibull apparent parameters, m* and V0* the intrinsic parameters, A and
n the Evans law parameters, tf the service life, V the applied stress, Y the shape factor
2 1 2
and t0 n2
V 0* .
n 2 AY K Ic
2
Consequently, the Weibull apparent parameters are:
1
n 1 m * and n2 ª 2 n1 ºn
m V0 V0 * n 1
«V n2 »
(7)
n2 ¬ n 2 AY K Ic ¼
2
It can be noted that only the Weibull apparent stress V0 depends on the stress rate.
Considering two loading rates, v1 and v2, V 0 v and V 0 v are linked through the
1 2
following relation:
1
§ v ·n
V0 v1
V 0 v2 .¨¨ 1 ¸¸ (8)
© v2 ¹
2.3. Application to a given problem
Carré and Daudeville [5] performed 4 points-bending tests under displacement control
on small annealed glass beams (L=230mm, h=37.5mm, e=19mm) with polished edges
and with different loading rates (0.5 and 0.05 m/s), see figure 2.
Figure 2: Four points bending test.

Specimen tested in the experimental campaign led by Carré and Daudeville [5].
409
Challenging Glass 3
Figure 3 presents the results of the large experimental database in terms of failure
probability vs. failure stress.
Figure 3: Experimental results obtained by [5]. Predictions with the Weibull model.
Apparent Weibull parameters, m and V0, for the loading rate of 0.5 m/s were firstly
identified with these previous tests results (continuous line in figure 3). For a loading
rate of 0.05 m/s, the failure prediction thanks to equations 6, 7 and 8 is given by a
dashed line in figure 3. The correct prediction validates the association of the subcritical
crack growth model with the Weibull theory.
Such developments can be easily included in a design code, as it was done in the
standard concerning glass tensile strength. Only the quality of the polishing should be
precisely evaluated since it governs directly the value of the Weibull parameters.
3. Modelling of the behaviour of laminated glass beams

Single conventional glass panes cannot be used as safety glass because of their
brittleness. The basic construction of laminated glass involves two pieces of float glass
together with an interlayer. To achieve that, Polyvinyl Butyral (PVB) interlayers have
been used for a long time. During the two last decades, some companies have put
considerable efforts into the development of new interlayer materials with increased
structural properties. One of these products is SentryGlassPlus (SGP) which is about
100 times siffer and 5 times stronger compared to conventional PVB.
With such laminated glass panes, brittle failure of an individual element may occur, but
the structure is able to redistribute loads to other elements, thereby providing
redundancy.
410
The following of this section will concern the modelling of the post-breakage
performance of a glass beam composed of 2 panes through a Finite Element Analysis
with the FE package ABAQUS.
3.1. Assumptions of the modelling: constitutive materials behaviour

Glass is assumed to be brittle in tension; A smeared crack model, Hillerborg-type [7] is
used to represent this tensile behaviour. Cracking is thus assumed to be the most
important aspect of the behaviour and it dominates its modelling. The model does not
track individual “macro” cracks. Rather constitutive calculations are performed
independently at each integration point of the FE model and the presence of cracks
enters into these calculations by the way that cracks affect the stress and material
stiffness associated with the integration point. A Rankine criterion is used to detect
crack initiation: this criterion states that “cracking” occurs when the maximum principal
stress reaches and exceeds the tensile strength of the material. Thus crack detection is
based only on mode I facture considerations. ABAQUS then assumes that “cracks” are
fixed and orthogonal to the direction of the maximum principal stress. The specification
of the post-failure behaviour needs to enter the post-failure stress as a tabular function
of the displacement across the crack (instead of strain which can introduce mesh
I I
sensitivity) or, in an equivalent way, the value of the fracture energy G F . G F
represents the energy required to form a unit area of crack surface. This material
property can be obtained thanks to the following LEFM formula:
I K Ic2
G F
(9)
E
where KIc is the glass toughness (0.75 MPam) and E is the Young’s modulus (70 GPa).
I
Consequently, G F 8 N/m for silica-soda-lime glass.
Moreover, the following assumptions are made:

x In a first simulation, two different values around the 5% fractile of figure 3

( f t 30 and 40 MPa) are used for the two considered glass panes;
x the compression behaviour of glass is assumed to be always linear elastic;
x in order to avoid any problems of convergence due to excessive distortion of
elements which can no longer carry stress, a brittle failure criterion called “kill
element method” is used. Thus, when the local cracking displacement in an
element reaches a critical value ( u ck 2G FI f t ), all the stress components
are set to zero and the corresponding element is removed from the mesh.
Besides, the obvious advantage for the computational convergence, this
technique allows also to visualize crack patterns in the beam.
The interlayer, whatever is its nature (PVB or SGP), exhibits a hyperelastic behaviour.
A material is called hyperelastic if the stress can be derived from an energy function W
that is uniquely related to the current state of deformation. The strain energy depends
solely on the deformation gradient.
411
Challenging Glass 3
There are several forms of strain energy potentials available in ABAQUS to model such
incompressible isotropic elastomers. The Mooney-Rivlin form has already shown its
capacity to reproduce accurately interlayer materials behaviours in laminated glass
structures [8]. This form is thus used in the present study; the strain energy potential is
given by:
W c01 I 1 3 c10 I 2 3 (10)
where:
- c01 and c10 are material parameters to be identified;
- I 1 and I 2 are the first and second deviatoric strain invariants defined as:
2 2 2 2 2 2
I1 O1 O2 O3 and I2 O1 O2 O3 with Oi the
deviatoric stretches.
ABAQUS enables an automatic identification of the material parameters c01 and c10 by
providing the results of an uniaxial test. In this study SGP is considered as the interlayer
materials and the schematic stress-strain curve of uniaxial tensile tests for a reference
loading rate of 5mm/min is given in figure below [9].
35
30
25
stress (MPa)
20
15
10
0
0 0.5 1 1.5 2 2.5 3 3.5 4
strain
Figure 4: Schematic stress-strain curve used in this modelling work for SGP.
3.2. Assumptions of the modelling: studied case

A beam composed of two annealed glass panes with a SGP interlayer subjected to a 3
points bending test is investigated in this numerical analysis.
Each individual glass component is 3000 mm length and 150 mm height. Their nominal
thickness is 6 mm. The interlayer thickness is 1.52 mm.
412
Solid brick linear Finite Elements are considered for all parts of the model. The mesh is
refined enough for the interlayer so that elements are not degenerated. Between all parts
of the model (glass 1-interlayer and interlayer-glass 2), the contact is supposed to be
perfect (“tie” contact). In a first approximation this assumption can be considered as
correct since [9] concludes that SGP beams do not show important local delaminations.
3.3. First results and future work

Figure 5 presents the load-deflection curve obtained by the modelling. Compared to a
single pane –glass beam, the laminated beam exhibits a small post-peak safety.
This first modelling aims only to show the feasibility of the numerical approach. In a
future work it is proposed to quantify this post-failure safety by evaluating an
“equivalent” fracture energy for the laminated glass beam. The methodology proposed
by the RILEM network (international union of laboratories and experts in construction
materials, systems and structures) is envisaged to achieve this purpose [10].
Such an equivalent fracture energy could be drawn in function of the failure stress or in
function of the associated failure probabilities of the two glass panes. This new result
might be of importance for structural engineers.
2.5
2
Load (kN)
1.5
0.5
0
0 1 2 3 4 5 6
Deflection at mid span (mm)
Figure 5: Load-deflection curve for laminated SGP beam. A first modelling result.
413
Challenging Glass 3
4. Study of the failure in connection area

In the past decade, an important and considerable effort has been performed on the
study of connection area in glass structures [11-14]. Indeed these particular zones
cannot be avoided, especially when large spans of high stiffness beams are considered.
Point bearings in holes are investigated in this study and more particularly countersunk
point fixtures with conical drillings.
In [14] an exhaustive experimental campaign has been presented. Push out tests adapted
for glass structures have been performed. Besides cylindrical holes, three geometries of
conical holes have been considered (namely holes b1, b2 and c1, see figure 6 and table
1). Both annealed and fully tempered 19 mm thick glass plates have been tested.
Figure 6: Cross section of conical holes.
Table 1: The 3 different conical geometries
Designation )int (mm) )ext (mm)
b1 24 40
b2 40 56
c1 30 40
More details concerning not only the connection (and particularly on the interlayer
washer enabling to avoid any contact between steel and glass) but also on the
experimental campaign itself, can be found in [14] or [15].
For conical holes, almost 60 samples have been tested until failure (40 on annealed
glass and 18 on tempered one more precisely). A special attention has been paid on the
initial torque applied to the bolt: indeed various values have been used. This initial
prestressing has been applied thanks to a torque wrench. Tables 2 a and b present all the
experimental results for respectively annealed and tempered glass. The number of
experiments as well as the standard deviation (if it is pertinent) are also given.
414
Table 2: Results of the experimental campaign on annealed glass.
Ultimate loads according to initial torque and standard deviations.
Exp. number
Hole geometry Initial torque (daN.m) Ultimate load (kN)
(stand. deviation kN)
Hole b1 0 12.7 1
1 23.3 5 (2.3)
2 22.7 4 (4.3)
>2.5 0 3
Hole b2 1 21.8 3 (2.1)
2 22 9 (2.5)
5 11.9 1
>5 0 2
Hole c1 2 24.4 2 (2.4)
2.5 16.8 3 (3.3)
3 20.5 5 (4.6)
5 26 2 (4.7)
One of the main conclusions is that the deviation is quite low although glass is sensitive
to surface flaws. Overall, the deviation is smaller than the deviation obtained during 4
points bending tests (see figure 3).
Table 3: Results of the experimental campaign on tempered glass.

Ultimate loads according to initial torque and standard deviations.
Exp. number
Hole geometry Initial torque (daN.m) Ultimate load (kN)
(stand. deviation kN)
Hole b1 1 107 4 (7.1)
7.5 118.5 1
10 115.7 1
Hole b2 2 112 4 (6.1)
4 109 1
10 85 1
Hole c1 1 86.5 2 (9.2)
2.5 94.2 1
7.5 89 1
10 62 2 (2.9)
415
Challenging Glass 3
Fractography analyses, i.e. post mortem analyses of the failure origin, coupled with
SEM observations, have also shown that:
x for annealed glass:

- failure has been most of time initiated on the conical part of the hole chamfer
even if the cylindrical parts are less carefully machined;
- the interaction between aluminium interlayer and glass excites a flaw on the
glass surface: this contact is at the origin of the failure;
- mode accurately, the failure starts from a flaw located on the bottom of the
conical chamfer (see figure 6) near to the cylindrical part, where the tensile
stresses due to in-plane loading are the most important or, a few times, located
on the top of the chamfer where the contact is the most intensive.
x for tempered glass:
- failure is always located on the bottom of the chamfer;
- the FE modelling of the thermal tempering, presented in [16] has shown that
the decrease of the residual compressive stresses from the bottom to the top of
the chamfer is less important than the decrease of the applied tensile stress.
Thus during the push-out tests performed on tempered glass, the top of the
chamfer is never decompressed and failure cannot occur from this location
That explains why standard deviations are relatively less important for
tempered glass than for annealed one.
All these remarks tend to prove that a deterministic approach of the resistance of the
connection is sufficient and effective. Of course, such an approach is easier to perform
for structural engineers.
5. Summary and conclusions

In this paper a discussion concerning failure in glass structures is proposed. This
discussion compiles experimental, analytical and numerical considerations.
Single glass panes, laminated ones and connections are studied. Several conclusions
have been pointed out:
x the probabilistic model of Weibull is used in association with a subcritical

crack growth model. For a given surface finish, the model can take into
account the effects of the specimen size, of the stress distribution and of the
rate of loading;
x the use of a smeared crack model (for glass) and a hyperelastic one (for
interlayer) in a Finite Element Analysis can lead to a quantification of the post-
failure safety of a laminated beam;
x the failure in connections is quite different from the one observed in current
zones: indeed the cracks origins have always been found in a small area of the
chamfer leading to low deviations in experimental results. A deterministic
approach of the failure seems thus to be possible.
416
6. References
[1] Griffith, A. The phenomena of rupture and flow in solids, Philosophical Trans. Royal Soc., London,
A221/1920, pp. 163–198.
[2] Weibull, W. A. (1951). A statistical distribution function of wide applicability, J. Appl. Mech., 18/1951,
pp. 293–297.
[3] Grenet, L., Mechanical strength of glass, Bull. Soc. Enc. Nat., Paris, 5/1899, pp. 838–848.
[4] Michalske, T. A., and Frieman, S. W. (1983) A molecular mechanism for stress corrosion in vitreous
silica, J. Am. Ceramic Soc., 66/1983, pp. 284–288.
[5] Carré H.; Daudeville L., Load bearing capacity of tempered structural glass, ASCE Journal of
[6] Evans, A. G., Slow crack growth in brittle materials under dynamic loading conditions, International
Journal of Fracture, 10/1974, pp.251-259.
[7] Hillerborg, A.; Modeer, M.; Petersson, P.E., Analysis of crack formation and crack growth in concrete
by means of fracture mechanics and finite elements, Cem. and Concrete Research, 6/1976, pp. 773–782.
[8] Timmel, M.; Kolling, S.; Osterrieder, P.; Du Bois,P.., A finite element model for impact simultion with
laminated glass, Int. Journal of Impact Engineering, 34/2007, pp. 1465-1478.
[9] Belis, J.; Depauw, J.; Callewaert, D.; Delincé, D.; Van Impe, R.; Failure mechanisms and residual
capacity of annealed glass/SGP laminated beams at room temperature, Engineering Failure Analysis,
16/2009, pp. 1866-1875.
[10] RILEM Committee FM C50, Determination of the fracture energy of mortar and concrete by means of 3
points bending tests on notched beams, 1985.
[11] Laufs, W.; Sedlacek, G., Stress distribution in thermally tempered glass panes near the edges, corners
and hole, Glass Science and Technology, 72/1999, pp.1-14.
[12] Maniatis, I., Numerical and Experimental Investigation on the Stress Distribution of Bolted Connection
under In-Plane Load, Ph-D thesis, Technische Universität München, 2006.
[13] Schneider, J., Glass Strength in the Borehole Area of Annealed Float Glass and tempered Float Glass,
Research in Architectural Engineering Series, 1/2007, pp.157-167.
[14] Bernard, F., Daudeville, L., Point fixings in annealed and tempered glass structures: modeling and
optimization of bolted connections, Engineering Structures, 31/2009, pp. 946-955.
[15] Bernard, F., Sur le dimensionnement des structures en verre trempé: étude des zones de connexion, Ph-D
thesis, ENS Cachan, 2001.
[16] Daudeville, L.; Bernard, F.; Gy, R., Residual stresses near holes in tempered glass plates, Material
Science Forum, 404-407/2002, pp.43-48
417
doi:10.3233/978-1-61499-061-1-419
New Expressions for the Effective

Thickness of Laminated Glass
Laura Galuppi, Gianni Royer-Carfagni
Department of Civil-Environmental Engineering and Architecture, University of Parma,
Italy, gianni.royer@unipr.it
The deformability of the interlayer does not provide a perfect shear transfer
between the glass plies, so that flexural response of laminated glass is somehow
intermediate between that of a monolith and that of free-sliding plies. In the design
practice, this effect is usually taken into account through the definition of the
Effective Thickness (ET), i.e., the thickness of a monolith with equivalent bending
properties. Classical formulas for the ET have been proposed by Bennison et al. [1],
based on the original analysis by Wölfel [2]. Here, we propose new expressions for
the ET based upon strain energy minimization. Practical formulas are derived
which apply to the one-dimensional case of beams, as well as to the two-
dimensional case of laminated plates. The better efficiency of the proposed method
with respect to others is proved by the comparison with accurate three-dimensional
numerical simulations.
Keywords: Laminated glass, plate design, effective thickness, laminated plate,

composite structures, sandwich structure.
1. Mechanical behavior of laminated glass

An effective technique to enhance the post-glass-breakage performance of architectural
glazing consists in bonding glass plies together with polymeric interlayers via
lamination in autoclave at high temperature and pressure. In such a way a laminated
glass acquires safety properties because, after breakage, shards remain attached to the
polymer and the system maintains a small but significant load bearing capacity,
avoiding injuries due to catastrophic collapse. But the interlayer affects also the pre-
glass-breakage response because it allows the transfer of shear stresses among glass
plies, at the price of a relative sliding due to the deformation of the polymer.
Laminated glass is thus a sandwich structure, whose stiffness and strength may be
considerably less than those of a monolithic glass with the same total thickness, because,
due to the shear deformability of the polymer there is not a perfectly coupling between
any two consecutive glass plies. The behavior of the structure will be intermediate
between two limit cases:
x the monolithic limit for which the relative sliding is null (fig. 1a); this limit
corresponds to ‫ ܩ‬՜ λ;
x the layered limit (frictionless relative sliding of the plies), attempted for ‫ ܩ‬՜ Ͳ
(fig. 1b).
419
Challenging Glass 3
a) b) c)
Figure 1: Laminated glass composed of two plies and one interlayer under flexure. The two limit cases of a)
monolithic limit and b) layered limit; c) the intermediate real configuration.
In the design practice, the problem is simplified and reduced to a case in which all the
materials, including the viscoelastic interlayer, are considered linear elastic. In
numerical computations, the response of laminated glass could be modeled by a layered
shell element that takes into account the competing stiffness between glass and
interlayer, but most of the commercial numerical codes do not have such elements in
their library. On the other hand, a full 3D analysis is complicated and time consuming.
This is why, in the design practice and especially in the preliminary design, it is very
useful to consider approximate methods for the calculation of laminated glass.
1.1. The approach by Wölfel-Bennison (W-B)

Currently, the most used approach is probably the one proposed by Bennison [1] based
upon the theory for composed sandwich beams proposed by Wölfel [2]. To illustrate,
consider the laminated beam of length l and width b shown in Figure 2, composed of
two glass plies of Young’s modulus E, connected by a polymeric interlayer of shear
modulus G.
y E, A1,I 1
p(x)
h1
t G H
x
h2
l
b E, A2,I 2
Figure 2: Beam composed of two glass plies bonded by a polymeric interlayer. Longitudinal and cross
sectional view (not in the same scale).
420
New expressions for the effective thickness of laminated glass
Let
௛భ ା௛మ ௕௛భయ ௕௛మయ
‫ܣ‬ଵ ൌ ܾ݄ଵ , ‫ܣ‬ଶ ൌ ܾ݄ଶ , ‫ ܪ‬ൌ ‫ ݐ‬൅ , ‫ܫ‬ଵ ൌ , ‫ܫ‬ଶ ൌ Ǥ (1)
ଶ ଵଶ ଵଶ
When the layered limit is attained, the moment of inertia of the laminated beam equals
the sum ‫ܫ‬ଵ ൅ ‫ܫ‬ଶ . In the monolithic limit, the moment of inertia reads
‫ܣ‬ଵ ‫ܣ‬ଶ
‫ܫ‬௧௢௧ ൌ ‫ܫ‬ଵ ൅ ‫ܫ‬ଶ ൅ ‫ܪ‬Ǥ (2)
‫ܣ‬ଵ ൅ ‫ܣ‬ଶ
For intermediate cases, Wölfel [2] proposed a strong approximation according to which
the effective moment of inertia is of the form
‫ܣ‬ଵ ‫ܣ‬ଶ
‫ܫ‬௘௤ ൌ ‫ܫ‬ଵ ൅ ‫ܫ‬ଶ ൅ Ȟ ‫ܪ‬ଶ ǡ (3)
‫ܣ‬ଵ ൅ ‫ܣ‬ଶ
where the parameter accounts for the capability of the interlayer to transfer shear
stress between the glass plies. Hypothesis (3) is equivalent to assume that the bending
stiffness of the external layers has no influence on the coupling offered by the central
layer: the less the bending stiffness of the external layers, the more accurate is this
hypothesis. Wölfel proposed for the expression
ͳ
Ȟൌ ǡ
‫ܣ ܧݐ‬ଵ ‫ܣ‬ଶ (4)
ͳ൅ߚ ଶ
ܾ݈ ‫ܣ‬ଵ ൅ ‫ܣ‬ଶ
where depends upon the loading and boundary condition. Bennison [1] has adopted
Wölfel’s approach for the case of laminated glass, suggesting to using in (4) the value
ߚ ൌ ͻǤ͸ although in Wölfel’s theory this is a particular case associated with simply
supported beams under uniformly distributed load.
From (3), one can calculate the stress- and the deflection-effective thickness, i.e., the
(constant) thickness of the homogeneous plate that, under the same boundary and load
conditions of the considered problem, has the same maximal stress or maximal
deflection, respectively. Introducing, as per [1], the quantities
ு௛భ ு௛మ ଶ ଶ
݄௦Ǣଵ ൌ ǡ ݄௦Ǣଶ ൌ ǡ ‫ܫ‬௦ ൌ ݄ଵ ݄௦Ǣଶ ൅ ݄ଶ ݄௦Ǣଵ ǡ (5)
௛భ ା௛మ ௛భ ା௛మ
the deflection-effective thicknesses turns out to be:

య
݄௘௙Ǣ௪ ൌ ඥ݄ଵଷ ൅ ݄ଶଷ ൅ ͳʹȞ‫ܫ‬௦ , (6)
whereas the stress-effective thickness for glass plies number 1 and 2 is given by
య
௛೐೑Ǣೢ య
௛೐೑Ǣೢ
݄ଵǢ௘௙Ǣఙ ൌ ඨ ǡ ݄ଶǢ௘௙Ǣఙ ൌ ඨ Ǥ (7)
௛భ ାଶ୻௛ೞǢమ ௛మ ାଶ୻௛ೞǢభ
421
Challenging Glass 3
Although these expressions (referred to in the sequel as the Wölfel-Bennison approach)

refer to a very particular static scheme, they are commonly used in numerical
computations with models of monolithic plates with constant thickness.
1.2. The Enhanced Effective Thickness (EET) approach

An alternative formulation has been very recently proposed in [3]. This procedure,
called Enhanced Effective Thickness (EET) method, is based upon a variational
approach and consists in finding the best approximation for the response of laminated
glass among a restricted class of shape functions for the deflection surface through the
minimization of the strain energy functional. The main hypotheses for this model are:
x the interlayer has no axial or bending stiffness, but only shear stiffness;
x shear deformation of glass is neglected;
x all materials are linear elastic;
x geometric non-linearities are not considered.
Remarkably, not only the method applies to the one-dimensional case of beams under
bending, but it can be naturally extended to the two-dimensional case of plates [4] under
the most various load and boundary conditions. Tables for the calculation of the relevant
coefficients in the most common cases have been presented in [5] for ease of reference
and to facilitate the practical use.
The efficiency of the EET formulation is confirmed by comparison with the results of
precise numerical simulations on paradigmatic examples.
2. Enhanced Effective Thickness: the one-dimensional case
2.1. Theoretical model

The enhanced effective thickness (EET) method is manly based upon two assumptions:
x the equivalent moment of inertia IR is the weighted harmonic mean of the
I tot (associated with the monolithic limit) and I1 I 2
moments of inertia
(layered limit), that is:
ͳ ߟ ͳെߟ
ൌ ൅ ǡ (8)
‫ܫ‬ோ ‫ܫ‬௧௢௧ ‫ܫ‬ଵ ൅ ‫ܫ‬ଶ
where the non-dimensional weight parameter ߟ plays a role analogous to that

of in (3), because it tunes the response from the layered limit (ߟ ൌ Ͳ) to the
monolithic limit (ߟ ൌ ͳ);
x the deformed shape of the laminated glass beam has the form of the elastic
curve ݃ሺ‫ݔ‬ሻ of a monolithic beam with constant cross section under the same
loading and boundary conditions of the problem at hand.
422
The parameter K is of the form
ͳ
ߟൌ ǡ
‫ܫ ݐܧ‬ଵ ൅ ‫ܫ‬ଶ ‫ܣ‬ଵ ൅ ‫ܣ‬ଶ (9)
ͳ൅ Ȳ
‫ܫ ܾܩ‬௧௢௧ ‫ܣ‬ଵ ‫ܣ‬ଶ
where , as illustrated in [3], is a parameter that is calculated from minimization of the

strain energy. In particular, denoting with the one-dimensional domain representative
of the reference configuration of the beam, the quantity is defined as
‫׬‬ஐ ‫݌‬ሺ‫ݔ‬ሻ݃ሺ‫ݔ‬ሻ݀‫ݔ‬
Ȳൌ Ǥ (10)
‫׬‬ஐ ݃Ԣሺ‫ݔ‬ሻଶ ݀‫ݔ‬
Such a parameter depends upon the boundary and load conditions and its values are
recorded in [5] for the cases of most practical relevance. Notice as well that depends
upon the mechanical and geometrical properties of the laminated beam.
From (8), the deflection-effective thickness ĥw then turns out to be
ͳ
݄෠௪ ൌ Ǥ
ඩ ߟ ͳെߟ (11)
ଷ ଷ ൅ ଷ ଷ
݄ଵ ൅ ݄ଶ ൅ ͳʹ‫ܫ‬௦ ݄ଵ ൅ ݄ଶ
Recalling the definitions (5) of hs;1 and h s;2 , one also finds the following expressions
for the stress-effective thickness:
݄ଵଷ ൅ ݄ଶଷ ൅ ͳʹ‫ܫ‬௦ ݄෠௪

ଷ ݄ଵଷ ൅ ݄ଶଷ ൅ ͳʹ‫ܫ‬௦ ݄෠௪
ଷ
݄෠ଵǢɐ ൌ ඨ ൅ ǡ ݄෠ଶǢɐ ൌ ඨ ൅ Ǥ (12)
ʹߟ݄௦Ǣଶ ݄ଵ ʹߟ݄௦Ǣଵ ݄ଶ
The EET approach presents no additional difficulty with respect to the Wölfel-Bennison
formulations, giving the compact formulas (11) and (12) for laminated glass design.
Moreover, it can be readily extended to the two-dimensional case.
2.2. Examples
The results obtainable with the EET approach are now compared with those proposed
by Bennison [1] and with the numerical experiments performed by means of the finite
element software SJ-Mepla, specifically conceived of for laminated glass [6].
For the sake of comparison, in the present section, three paradigmatic cases are
analyzed in detail. With the same notation of Figure 2, assumed geometrical and
structural parameters are l = 3150 mm, b = 1000 mm, h1 = h2 = 10 mm, t = 0.76 mm,
E = 70 GPa, while the shear modulus G of the polymeric interlayer is varied to evaluate
its influence on the shear-coupling of the glass plies. The distributed load per unit length
becomes p = 0.75 N/m.
423
Challenging Glass 3
Three constraint conditions for a uniformly loaded beam are here considered:
x simply supported beam (Figure 3.a)

x beam with three supports (Figure 3.b);
x double clamped beam (Figure 3.c).
In the following graphs, the stress- and deflection-effective thicknesses, calculated

through (11) and (12), are plotted as function of G with a continuous line, whereas the
effective thicknesses calculated with the Wölfel-Bennison’s approach is represented
with a dashed curve. Results of numerical experiments are indicated with dots.
In the case of simply supported beams under uniform load the models give results that
in practice coincide, a finding that is not surprising because this is the simplest case
upon which the W-B approach is calibrated. Numerical results confirm the good
approximation that is achieved. In the cases of beams with three supports and of
clamped beams, there is a substantial deviation between the EET and W-B approaches
especially for the lowest values of G, but the numerical experiments are clearly in favor
of the EET approach. Observe that W-B is not on the side of safeness, because it
predicts effective thicknesses greater than in reality and, consequently, underestimates
deflection and stress.
424
a)
Deflection-effective thickness Stress-effective thickness
22 22
MONOLITHIC LIMIT MONOLITHIC LIMIT
20 20
18 18
[mm]
[mm]
16 16
LAYERED LIMIT
14 E.E.T. 14 E.E.T.
W-B W-B
LAYERED LIMIT
Numerical Numerical
12 12
-2 -1 0 1 -2 -1 0 1
10 10 10 10 10 10 10 10
G[MPa] G[MPa]

22
b) 22
20 20
18 18
[mm]
[mm]
16 16
14 E.E.T. 14 E.E.T.
LAYERED LIMIT
W-B W-B
Numerical Numerical
LAYERED LIMIT
12 12
-2 -1 0 1 -2 -1 0 1
10 10 10 10 10 10 10 10
G[MPa] G[MPa]

c)
22 22
20 20
18 18
[mm]
[mm]
16 16
14 E.E.T. 14 E.E.T.
W-B LAYERED LIMIT W-B
Numerical Numerical
LAYERED LIMIT
12 12
-2 -1 0 1 -2 -1 0 1
10 10 10 10 10 10 10 10
G[MPa] G[MPa]
Figure 3: Beam with different constraint condition under distributed load. Comparison of the effective
thicknesses obtained with: Wölfel-Bennison (WB) approach; the enhanced effective thickness (EET)
approach; the numerical simulations.
425
Challenging Glass 3
3. Enhanced Effective Thickness: the two-dimensional case
3.1. Theoretical model

Consider now a laminated glass plate identified by the xy domain under distributed
load p(x,y), composed of two glass plies of thickness h1 and h2 and Young’s modulus E
and Poisson’s ratio , connected by a polymeric interlayer of thickness t and shear
modulus G.
ா௛೔య
Defining the flexural rigidity of the i-th glass ply as ‫ܦ‬௜ ൌ , it can be
ଵଶሺଵିɋమ ሻ
demonstrated (see [4]) that the flexural rigidity for the monolithic limit reads
‫ܦ‬ଵ ‫ܦ‬ଶ
‫ܦ‬௧௢௧ ൌ ‫ܦ‬ଵ ൅ ‫ܦ‬ଶ ൅ ͳʹ ‫ܪ‬ଶ ǡ (13)
‫ܦ‬ଵ ݄ଶଶ ൅ ‫ܦ‬ଶ ݄ଵଶ
In analogy to the one-dimensional case, the deformed shape of the plate can be selected
as the elastic deformed surface ݃ሺ‫ݔ‬ǡ ‫ݕ‬ሻ of a monolithic plate with constant thickness
under the same loading and boundary conditions. Moreover, the equivalent rigidity DR,
in analogy with (8), can be assumed to be
ͳ ߟ 1-ߟ
ൌ ൅ . (14)
‫ܦ‬ோ ‫ܦ‬௧௢௧ ‫ܦ‬ଵ ൅ ‫ܦ‬ଶ
The minimization of the strain energy of the sandwich plate allows to determine the
counter part of (10) for the two dimensional case in the form
ͳ
ߟ= ,
‫ܦ ݐ‬ଵ ൅ ‫ܦ‬ଶ ͳʹ‫ܦ‬ଵ ‫ܦ‬ଶ (15)
ͳ൅ Ȳ
‫ܦ ܩ‬௧௢௧ ‫ܦ‬ଵ ݄ଶଶ ൅ ‫ܦ‬ଶ ݄ଵଶ
where
‫׬‬ஐ ‫݌‬ሺ‫ݔ‬ǡ ‫ݕ‬ሻ݃ሺ‫ݔ‬ǡ ‫ݕ‬ሻ݀‫ݕ݀ ݔ‬ (16)
Ȳൌ ǡ
ଶ ሺ‫ݔ‬ǡ ‫ݕ‬ሻ
‫׬‬ஐ ሾ݃ǡ௫ ൅ ଶ ሺ‫ݔ‬ǡ ‫ݕ‬ሻሿ݀‫ݕ݀ݔ‬
݃ǡ௬
depends upon the plate shape, the load distribution ‫݌‬ሺ‫ݔ‬ǡ ‫ݕ‬ሻ and the boundary conditions.
The stress- and deflection-effective thicknesses may be readily calculated and take
expressions analogous to (11) and (12), respectively.
It is important to note that the only “difficulty” of the proposed method consists in
calculating from (16), because all the other formulas are simple analytical expressions.
Values of that refer to the most common cases of the design practice (plate under
various boundary condition, under distributed and concentrated load) are reported in [5].
3.2. Examples
In the sequel, we compare the deflection- and stress-effective thickness for rectangular
plates under a uniformly distributed load with various constrain conditions, calculated
according to the proposed EET approach through equations (11) and (12), with the ones
calculated with the W-B formulas (6) and (7). Results are also validated by means of
426
numerical analysis performed with the finite element software SJ-Mepla. Assumed
structural parameters are the size of the plate 3000 mm x 2000 mm; the thicknesses of
the glass plies h1 = h2 = 10 mm; the thickness of the interlayer t = 0.76 mm; the elastic
parameters for glass E = 70 GPa and = 0.22. The shear elastic modulus G of the
polymeric interlayer is again varied between 0.01MPa and 10MPa. The distributed
pressure on the plate is taken equal to 0.75 10-3 N/mm2.
The considered case are:

x simply supported plate on four sides (Figure 4.a): in this case the two
formulations give different results at the qualitative level. Again W-B is not on
the side of safeness, because it underestimates deflection and stress;
x point-wise supported plate at the corners (applying to frameless glazing, Figure
4.b): in this case the EET and W-B give similar results, in agreement to
numerical outcomes;
x plate with two opposite edges simply supported and one edge built-in (for
example, glass banisters, Figure 4.c): it is evident that the EET model and W-B
approach give substantially different results and that numerical experiments are
in favor of EET.
427
Challenging Glass 3
22
Deflection-effec tive thick ness a) 22
Stress-effectiv e thic knes s
21 MO NO LITHIC LIMIT 21 MO NO LITHIC LIMIT
20 20
19 19
18 18
[mm]
[mm]
17 17
16 16
15 15
14 14
E.E.T. LAYERED LIMIT E.E.T.
13 W-B 13 W-B
Numerical Numerical
LAYERED LIMIT
12 - 2 -1 0 1
12 -2 -1 0 1
10 10 10 10 10 10 10 10
G [MPa] G [MPa]
22
Deflection-effectiv e thick nes s b) 22
Stress-effectiv e thic kness
20 20
19 19
18 18
[mm]
[mm]
17 17
16 16
15 15
14 14
13 W-B 13 W-B
LAYERED LIMIT Numerical Numerical
12 12
-2 -1 0 1 -2 -1 0 1
10 10 10 10 10 10 10 10
G [MPa] G [MPa]
22
Deflection-effectiv e thic knes s c) 22
Stress -effective thickness
20 20
19 19
18 18
[mm]
[mm]
17 17
16 16
15 15
14 14
13 W-B 13 W-B
LAYERED LIMIT Numerical Numerical
12 -2 -1 0 1
12 - 2 -1 0 1
10 10 10 10 10 10 10 10
G [MPa] G [MPa]
Figure 4: Rectangular plate with different constrained condition, under distributed load. Comparison of the
effective thicknesses obtained with: Wölfel-Bennison (WB) approach; the enhanced effective thickness (EET)
approach; the numerical simulations.
428
4. Conclusions
One of the currently most-used simplified approaches for the structural design of
laminated glass is that due to Bennison [1], which is based upon the original work by
Wölfel [2]. However, Wölfel’s model was primarily conceived of for a sandwich beam
with external plies with considerable axial stiffness but negligible bending stiffness and
an intermediate layer that can only bear shear stress, with zero axial and flexural
strength. Whenever the external layers present considerable bending stiffness, as in the
case of laminated glass, Wölfel proposed a very approximate solution that in any case,
as we have verified here, gives results in agreement with more accurate (numerical)
methods of analysis for the only case in which the load is uniformly distributed and the
deformed shape tends to be cylindrical.
In the other cases, the standard Wölfel-Bennison approach gives results that are not on
the side of safeness. Better approximations can be achieved with the Enhanced Effective
Thickness approach, which presents no additional difficulty with respect to the more
traditional formulation. Such an approach can be easily extended to the two-dimensional
(plate) case, for which it gives results that fit more closely the real situation both for the
deflection and the stress calculation.
The EET method furnishes compact formulas for both the beam case and the plate case
and, remarkably, the most relevant expression (11) and (12) are analogous to those
corresponding to the one dimensional case. The coupling offered by the interlayer can
be readily evaluated by using the values of that have been tabulated in [5] for all
those cases that are relevant for the design practice. However, using (16), the value of
can be calculated with no difficulty for any laminated plate under any load condition.
The enhanced effective-thickness approach thus seems to represent an accurate and
powerful tool for the practical calculation of laminated glass.
5. Acknowledgements
The autors acknowledge the Italian MURST for its partial support under the PRIN2008
program.
6. References
[1] Bennison, S. J., Stelzer, I., Structural properties of laminated glass, Short Course, Proceedings Glass
Processing Days, Tampere, Finland, 2009.
[2] Wölfel, E., Nachgiebiger Verbund Eine Näherungslösung und deren Anwendungsmöglichkeiten,
Stahlbau, 6/1987, pp. 173-180.
[3] Galuppi, L., Royer-Carfagni, G., Effective Thickness of Laminated Glass Beams. New Expression via a
Variational Approach, Engineering Structures, 38/2012, pp. 53-67.
[4] Galuppi, L., Royer-Carfagni, G., The effective thickness of laminated glass plates, Journal of Mechanics
of Materials and Structures, accepted for publication. Available on line at http://dspace-
unipr.cilea.it/handle/1889/1703.
[5] Galuppi, L., Manara, G., Royer-Carfagni, G., Practical expressions for the design of laminated glass,
submitted. Available on line at http://dspace-unipr.cilea.it/handle/1889/1720.
[6] SJ MEPLA, User’s manual, version 3.5, 2012.
429
doi:10.3233/978-1-61499-061-1-431
A Study of Polishing Performance of Glass

Using Fluid Jet Polishing
Anders Jönsson, Pia Lindahl, Johan Fredin
Blekinge Institute of Technology, Sweden, anders.jonsson@bth.se,www.bth.se
Christina Stålhandske
Glafo – the Glass Research Institute, Sweden, christina.stalhandske@glafo.se,
www.glafo.se
The strength of glass designs are relying on good surfaces. Complex shapes and
small holes calls for new polishing methods. Polishing of glass using fluid jet
polishing is well known as a suitable method for acquiring high quality surfaces. In
this study the combination of higher pressure and aluminum oxide as polishing
material is tested in terms of higher polishing performance. A design of experiment
study is done where important process parameters are varied. The results are
compared to fluid jet polishing using cerium oxide. The polishing effect on the float
glass is evaluated using visual grading and by using optical profilometry. In
addition an analysis of sustainability aspects are done comparing the two different
polishing materials. It is clearly shown that aluminum oxide is increasing the
polishing performance.
Keywords: Glass, Optical Profiler, Fluid Jet Polishing, Surface Analysis, Abrasive
1. Introduction
Polishing of glass is becoming more and more important for different glass products.
Traditionally, the polishing of glass in the art and design sector is a well established
process. Different polishing processes like acid polishing, mechanical and
chemical/mechanical polishing are well recognized. Due to complicated geometries and
ecological aspects, alternative polishing processes are developed and under evaluation.
Within the float glass industry more and more attention is given to the fact that the
making of holes is introducing defects in the glass. These defects are a disadvantage for
the strength of the glass product and therefore ways to reduce or remove these defects
are searched for. Fluid Jet Polishing (FJP) has a potential of being well suitable for
removing these defects.
The principle of FJP is conceptually described as the process of flowing a slurry that
contains a mix of a fluid and polishing abrasives. This process can be described as a
combination of the suspension jet technology [1] used for effective water jet cutting of
materials and bowl feed polishing used for reaching very fine surface qualities, down to
0.1 nm RMS, [2]. In FJP the pressure is typically quite low compared to traditional
suspension jets for cutting of materials, often the pressure is in the range of 0.5-2 MPa.
The FJP process is also unique in the sense that it both polishes and shapes the surface
in the same process step. This opens up for a wide range of potential areas of usage
besides the polishing of glass.
431
Challenging Glass 3
Compared to mechanical polishing, where the tool is in contact with the glass surface,
FJP has no mechanical contact between the “tool” and the surface to be polished. This
enables the jet to reach and polish arbitrary geometries. This also includes non-
symmetrical holes and small radii in corners for example which are extremely difficult
to polish with traditional mechanical polishing methods. This also implies that the
automation potential of FJP is high, since the accuracy needed of the tool path to follow
is decreased due to the flexible “tool”. Compared to acid polishing, there are distinct
environmental advantages both in terms of ecological and social aspects if the suitable
polishing abrasive is chosen.
Earlier work on FJP has shown that the process is capable of producing good surface
qualities. Several types of abrasives have been tested and evaluated. To the author’s
knowledge no earlier studies have been done, comparing cerium oxide and aluminum
oxide when using FJP with higher pressures. This work presents initial results on the
influence of the abrasives during FJP at higher pressures than traditionally used for
polishing. In addition an introductory study of environmental aspects is done in order to
compare the abrasives regarding other perspectives than pure technical.
2. Experimental
2.1. FJP setup

The experimental setup used in this work is a traditional FJP set-up shown in Figure 1.
It consists of a container, slurry buffer, where the slurry made of water and polishing
abrasives are buffered. From this container the pump is sucking the slurry and
increasing the pressure. The pressure is controlled by regulating the speed of the pump
and measured via an analogue pressure transducer. The pump speed is set manually to
provide the correct pressure for the test. After the pump the slurry is transported through
pipes to a nozzle where the potential energy due to the pressure of the slurry is
converted to kinetic energy in the form of jet velocity. The pressure before the nozzle,
the flow through the nozzle and the velocity of the jet after the nozzle are strongly
depending of the jet nozzle geometry. The relative position of the nozzle to the work
piece to be polished is controlled by a CNC-controller. The slurry in this work is a
mixture of tap water and commercial cerium oxide under the brand name Regipol with
an average particle diameter of 6.4 Pm or an aluminum oxide with average particle size
of 10.1 Pm. No measurements of pH-values of the slurry have been done in this test.
The temperature of the slurry is varying slightly between 20 and 30 degrees Celsius
during the test run.
432
A Study of Polishing Performance of Glass Using Fluid Jet Polishing
Figure 1: Schematic overview of the experimental setup for FJP.
An example of test-set-up is shown in Figure 2 where the polishing nozzle is adapted to

a 5-axis water jet cutting machine utilizing the 3D motion control possibilities.
Figure 2: 5-axis manipulator used in the experimental setup for FJP.
433
Challenging Glass 3
During the tests the following parameters are varied:
x velocity between nozzle and work piece in mm/s

x angle between work piece and jet in degrees
x stand-off distance between work piece and nozzle exit in mm
x pressure at pressure gauge just before the nozzle in bar
The parameters are varied between a high and a low value. The total test series is
consisting of 16 parameter sets. The nozzle is run forward and back once over the work
piece surface. This gives two passages of the jet for every set of parameters. In both test
series the same type of nozzle with a diameter of 1.3mm has been used. The work
pieces to be polished are made of 110 by 150 mm float glass and prepared by grinding
using 25μm abrasives.
2.2. Surface analysis

The polished surface is judged visually in order to have a subjective judgment of the
surface quality. The visual examination is performed in an enclosed area with black
background and light coming from above. Six categories are used where 0 is hardly any
impact and 5 is a bright surface corresponding to an acid polished surface. The
judgment is done by experienced glass professionals, but it is still difficult to provide an
objective judgment mainly due to the small surface analysed. In addition to the visual
inspection, an optical profiler, a Bruker NPFLEX is used for measurements of the
surfaces. The main parameters measured are the mean roughness of the surface, Sa, the
root mean square surface slope of the surface, Sdq, and the developed interfacial area
ratio Sdr. [3]. The measurements are made with a magnification of 2.7 times
magnification in VSI mode and the measured area is 1.7 x 2.3 mm. In addition a planar
fit is done.
3. Results and Discussion

The results reported in this work are based on two equal test series except for the
polishing abrasive used. In the first test cerium oxide is used and in the second test
aluminum oxide is used. In addition a brief analysis of sustainability aspects for the two
different abrasives are done.
3.1. Polishing
In general it is found that Sdq and Sdr parameters correlate better to the naked eye
evaluation than Sa values. It is the very fine cracks that should be removed and they
scatter light giving rise to a less bright appearance of the surface. Smooth structures as
polished grinding tracks and water cut structure remain even after small cracks are
removed and will influence the Sa value but not scatter light.
In Table 1 the results for the test series with 20 weight percent cerium oxide is shown.
434
Table 1: Test series using 20% cerium oxide, 1.3mm nozzle

Velocity Pressure Stand-off Angle Visual Sa Sdq Sdr
[mm/s] [MPa] distance [deg] [nm] [deg] [%]
[mm]
2 2 6 45 1 313 8.1 0.93
8 2 6 45 0 312 8.4 0.99
2 4.8 6 45 3 179 3.7 0.21
8 4.8 6 45 2 202 5.3 0.42
2 2 50 45 2 197 5.4 0.43
8 2 50 45 1 220 6.6 0.62
2 4.8 50 45 3 165 3.4 0.17
8 4.8 50 45 2 290 6.3 0.57
2 2 6 90 0 339 8.2 0.95
8 2 6 90 0 262 7.2 0.73
2 4.8 6 90 3 202 4.4 0.31
8 4.8 6 90 1 218 5.6 0.47
2 2 50 90 2 255 6.2 0.56
8 2 50 90 1- 342 8.9 1.14
2 4.8 50 90 2+ 281 5.0 0.36
8 4.8 50 90 2 305 6.9 0.68
It is clear from the table that the parameters that give the best results are low velocity
and high pressure. The other parameters seem to have less influence. The best values
reached are Sa 165 nm, Sdq 3.4 deg and Sdr 0.17when the velocity is low, the pressure
high, high stand-off distance and 45 degrees angle between the jet and the work piece
surface.
Figure 3: a) 20% cerium oxide 2mm/s, 4.8 MPa, distance 50mm and angle 45 degrees
b): 20% aluminum oxide 2mm/s, 4.4 MPa, distance 6mm and angle 90 degrees
435
Challenging Glass 3
The surface is shown in Figure 3a and the grooves from the grinding process are visible
but not clear. In Figure 3b, the corresponding aluminum oxide surface except that the
smaller distance is used.
In Table 2 the results for the test series with 20 weight percent aluminum oxide is
shown. It is here clear that a low velocity will increase the surface quality. It should be
noticed that the strong positive effect of a higher pressure is not evident in this test.
Instead the stand-off distance has remarkable higher influence on the surface quality. It
is also noticeable that the values for Sdq and Sdr are significantly lower than for the
cerium test series, (approximately 3 respectively 9 times better) even though the Sa
values are in the same order. The Sdq and Sdr are clearly better correlated to the visual
judgment then the Sa values [3].
Table 2: Test series using 20% aluminum oxide, 1.3mm nozzle

Velocity Pressure Stand-off Angle Visual Sa Sdq Sdr
[mm/s] [Mpa] distance [deg] [nm] [deg] [%]
[mm]
2 20 6 45 3 301 2.6 0.11
8 20 6 45 2 173 3.7 0.20
2 20 50 45 2 220 1.7 0.05
8 20 50 45 2 144 3.1 0.14
2 20 6 90 4 587 1.4 0.03
8 20 6 90 4- 189 2.8 0.12
2 20 50 90 2 267 2.0 0.06
8 20 50 90 2 155 3.5 0.18
2 44 6 45 3 564 2.1 0.08
8 44 6 45 3 278 2.0 0.06
2 44 50 45 2 1437 3.6 0.19
8 44 50 45 2 336 2.6 0.10

2 44 6 90 4 387 1.3 0.02
8 44 6 90 4- 314 1.7 0.04
2 44 50 90 2 700 3.3 0.16
8 44 50 90 2 249 3.4 0.18
In Figure 4a an area is measured that includes none polished and polished sub-areas,
where the polished are is clearly visible as a furrow. The groves origin from the
grinding preparation of the glass surface is still remaining and clearly visible. This is a
typical look from a cerium oxide polished surface. In this specific figure the
concentration of cerium oxide is 50%.
In figure 4b a similar 3D picture of an aluminum oxide polished surface with a similar

furrow. Here the grinding grooves are not easily seen.
436
Figure 4: a) Surface for cerium oxide, 0.5mm nozzle, 2mm/s, 3MPa, 6mm, 45 degrees and visual 4 [3]
b): Surface for aluminum oxide, 1.3mm nozzle, 2mm/s, 4.5MPa, 6mm, 90 degrees and visual 5
Figure 5 gives a complementary view of the surface by displaying the profiles at

specific X and Y coordinates of the surface through the furrow. The X profile is
showing the surface profile along the furrow bottom and the Y profile is representing a
cross section of the furrow.
Figure 5: Additional surface analysis performed using surface profiles.
437
Challenging Glass 3
3.2. Sustainability evaluation

In this work the cerium oxide and the aluminum oxide have been compared from a
sustainability perspective [4,5] while assuming that the material flows are linear, i.e. the
materials are not reused after the polishing process but will be emitted to water. The
evaluation is qualitative comprehensive and do not consider how assimilation to natural
systems will be affected by shape or size of the emitted substances.
However, even though neither cerium oxide nor aluminum oxide are listed in national
databases related to toxicity –since the work is dealing with small particles of the
materials, working environment precautions should be committed to avoid exposure –
specifically through inhalation. Lack of data regarding the mining of the materials
affects the ability to evaluate the extraction process. Information from suppliers has
indicated that both substances could be extracted in China, why similar working
conditions could be assumed but not confirmed.
In Table 3, criteria are shown that are used for comparing the substances. From the
evaluation of the criteria it is clear that aluminum oxide should be the preferable
alternative. This is based on the facts that aluminum is very common in the earth crust,
that the anthropogenic flows for cerium oxide is already larger than the natural flows
and that Cerium is a rare substance that should not be used in a dissipative way to
secure future generations availability of the material. Cerium is also listed on the EU list
of critical metals in the group of rare earths. In addition, and due to the reasons given
above, the price for cerium has been raised dramatically the last years, while the price
for aluminum is more stable.
438
Table 3: Comparing cerium oxide and aluminum oxide from a sustainability point of view.
Comparison of two materials Cerium Aluminum Comments
oxide oxide
A. Avoiding systematic - + The anthropogenic dominance of the
increasing concentrations in mobilization flows of the element are larger for
nature due to exchanges of Cerium than for Aluminum why there is a
basic elements with other greater risk that emissions of Cerium will
systems contribute to increasing concentrations in nature
[6]
B. Avoiding systematic - + Both Cerium oxide and Aluminum oxide
increasing concentrations in are naturally occurring substances; however, the
nature due to transformation natural availability of Aluminum oxide in water
of substances within the is greater than the availability of Cerium oxide
system [7].
C. Avoiding systematic Information is lacking regarding

physical degradation of implications on land-use change why further
ecological systems due to information is needed.
land-use change
D. Avoiding systematic Information is lacking regarding
physical degradation of implications on land-cover change why further
land-cover change
E. Avoiding systematic Information is lacking regarding
physical degradation of implications on renewable resources why further
depletion of renewable
resources
F. Avoiding systematic - + The natural availability in Earth’s Crust of
physical degradation of Aluminum is larger than the availability of
ecological systems due to Cerium [7]. Furthermore, even if Cerium is one
degradation of non-renewable of the most common Rare Earths metals, the
resources availability of the substance for future
generations are seen as critical [8].
G. Avoiding systematic - - The major extractions of both Aluminum
violations of economical, and Cerium are conducted in China, why we

social and cultural rights assume that possible contributions to
degradations of social systems are comparable
H. Avoiding systematic - - for Aluminum and Cerium.
violations of civil and
political rights
I. Avoiding systematic Information is lacking regarding
violations of any of the above implications regarding associated flows why
criteria due to associated further information is needed.
flows
Green color means that the positive advantages is clear, grey color means that there is a need for more
information, yellow color indicates an obvious negative effect.
439
Challenging Glass 3
4. Summary and Future Work

Two test series have been done in order to qualitatively compare the polishing
performance of fluid jet polishing using cerium oxide and aluminum oxide. The
polishing results are evaluated by visual examination and measured with an optical
profiler. The Sdq and Sdr values are correlated to the visual examination. Both visual
examination and Sdq and Sdr show clearly that aluminum oxide give significantly better
polishing results on the glass surfaces. An introductory study for sustainability aspects
of the materials are performed and indicates that most likely aluminum oxide is a better
choice from a strategic sustainable point of view. Further tests should be done in order
to statistically verify the results. The influence of temperature and pH value is also of
interest for investigation in future work. With more test results the correlation between
Sdq and Sdr values and visual examination can be consolidated. A more thorough
sustainability study is also suggested also including alternative abrasive materials and
including the recycling of the materials.
5. Acknowledgements
Funding’s have been provided by EU structural funds, Region Blekinge and Kalmar
Regionförbund. Orrefors Kosta Boda, Glasteknik i Emmaboda, Water Jet Sweden,
KMT Robotic Solutions have all contributed to the success of the project.
6. References
[1] Michele Monno, Massimiliano Annoni, Chiara Ravasio, Water jet, a flexible technology, Polipress -
Politecnico di Milano, ISBN 97888-7398-030-9, 2007.
[2] Booij, S.M., Fluid Jet Polishing - possiblites and limiations of a new fabrication technique, Technische
Universiteti Delft, 2003.
[3] Stålhandske, C. and M. Lang, Flexible glass polishing with a cerium oxide containin fluid jet, in
Proceedings of Glass Performance Days 2011: Tampere, Finland. p. 446-448, 2011.
[4] K-H Robèrt et al., Strategic Leadership Towards Sustainability, Psilanders grafiska, Karlskrona, Sweden,
2004.
[5] H. Ny et al., Sustainability Constraints as System Boundaries: An Approach to Making Life-cycle
Management Strategic, Journal of Industrial Ecology 10, no. 1–2 , 61–77, 2006.
[6] Klee, RJ, Graedel TE. 2004. Elemental cycles: A status report on human or natural dominance. Annual
Review Of Environment And Resources. 29:69-107.
[7] Emsley, J.,Web Site: Chemical Case Studies. Angewandte Chemie International Edition, 42: 4425.
doi: 10.1002/anie.200390584, 2003.
[8] Critical raw materials for the EU, European Commision, 30 July 2010.
440
doi:10.3233/978-1-61499-061-1-441
Design of Arches Strengthened with Cables

for Glass Roofs
Wim Kamerling
Delft University of Technology, Faculty of Architecture, The Netherlands,
M.W.Kamerling@tudelft.nl
Arches are structurally very efficient. Nevertheless the dimensions of an arch can
be reduced further by strengthening with cables. For arches of glass the efficiency
of strengthened arches will be showed. To prevent the cables to be subjected to
compression it is advisable to post-tension the cables. This paper focuses on the
design and structural analysis of arches composed of glass elements, strengthened
with post-tensioned cables to support transparent roofs.
Keywords: Glass, Arches, Cables, Post-Tensioning, Grid Shell,
1. Introduction
The transparency of glass roofs will be optimal in case the supporting structure is made
of glass and the dimensions of the glass elements are minimal. Generally form-active
structures [2] such as funicular arches are structurally very efficient. Arches are
constructed for at least two thousands years. However the dimensions of the Roman
arches of masonry were quite substantial. Later, thanks to Hook, Poleni and other
scientists and engineers arches could be designed less massive.
15,0 m
Figure 1: The roof of the GUM department store is supported by half circular arches, strengthened with
cables to reduce the bending moments acting on the arch [1].
Arches, subjected to normal forces only, can be dimensioned much smaller than section-
active structures, subjected to bending moments. However it is difficult to optimize the
dimensions of an arch, especially in case the structure is subjected to several load
combinations. To minimize the dimensions, the line of the system has to approach the
varying lines of thrust of the load combinations. For an arch, subjected to several load
combinations, the bending moments can be reduced considerately by strengthening the
arch with cables. At the beginning of the 20th century the Russian engineer V.G.
Shukhov designed half circular arches for the GUM department store in Moscow, see
figure 1. Post-tensioning the guy rods provided the arch with six additional bearings,
441
Challenging Glass 3
capable of working both in compression and tension [1]. The analysis of a cable
strengthened arch is complex. Nowadays an engineer can analyze these structures with a
finite element program [3], but probably Shukhov designed these arches with just a
slide rule. To understand these remarkable structures this paper focuses on the analysis
of the load transfer. Recommendations are given to design these arches with minimal
dimensions, to optimize the transparency of the glass roof.
2. Three hinged arch of glass

A structure of glass fails if the structure is exposed to a tensile stress exceeding the
ultimate tensile stress [5]. Due to their form, arches are mainly subjected to normal
compressive forces. Mostly these normal forces will compensate, entirely or partly, the
tensile bending stresses. So arches are very suitable for structures of glass, masonry and
concrete. To show the potentials at first an example of a three hinged arch of glass is
described. For form-active structures of glass subjected by bending and normal forces
the stresses are validated with equation (1) based on the Theory of Elasticity, provided
the effect of second order is small and can be neglected:
J g .N g J g .M g J e .N e J e .M e
d1 (1)
A. f t f W . ft f A. f t 0 W . ft 0
With:
Load factor permanent load: Jg = 1,2
Load factor live load: Je = 1,5
Area: A [mm2]
Modulus of the section: W [mm3]
Bending moment due to the permanent load: Mg [Nmm]
Normal force due to the permanent load: Ng [N]
Bending moment due to the live load: Me [Nmm]
Normal force due to the live load: Ne [N]

Ultimate tensile stress for permanent loads, t = f : f t =f 6,0 [MPa]
Ultimate tensile stress for instantaneous loads, t = 0: f t=0 14,4 [MPa]
2.1. Example 1: Three hinged arch of glass

Assume a roof is supported by arches of glass with a radius of 3,6 m and a span of 7,2
m. The roof is composed of glass panes, thickness 3*8 mm, spanning 1,2 m, supported
by glass purlins, section 30*180 mm2 , spanning 2,0 m. The purlins are supported by
beams, section 50* 360 mm2 with a span of 3,6 m. The beams are supported by the
arches and prevent the arches of buckling perpendicular to the main axis. The length of
sheets of glass is limited [4] so the arches, section 125 * 500 mm2, are composed of
stepwise glued plates with a thickness of t = 25 mm. At a joint three sheets are
continuously, consequently the area A and modulus of the section W are reduced with a
factor 0,6. The half circular arches are subjected to five concentrated loads, see figure 2
and 3. The wind and permanent load acting at the nodes of the arches are described in
table 1.
442
Design of Arches Strengthened with Cables for Glass Roofs
Table 1: Loads acting at a node of the arch:

Load Number Weight Length Fep
Roof glass panes 3* 0,024 * 1,2 * 25 * 2,0 = 4,32 kN
Purlins 3* 0,03 * 0,18 * 25 * 2,0 = 0,81 kN
Beam 0,05 * 0,36 * 25 * 3,6 = 1,62 kN
Arch 0,125 * 0,5 * 25 * 1,9 = 2,97 kN
Force due to the permanent load 9,72 kN
Force due to the wind load 1,0 * 3,6 * 2,0 = 7,2 kN
4
3 5 4
3 5
2 6
2 6
I
1 r 7 1 7
I r
Figure 2: Arch subjected to concentrated loads due to Figure 3: Arch subjected to an anti-metric radial
the permanent load. load.
Due to the permanent load the concentrated force is equal to Fg = 9,72 kN. The supports
are subjected to the thrust and the vertical reaction force equal to respectively Hrep =
1,134.Fg and Rrep = 2,5.Fg. The arch is subjected to bending, for node 5 and 6 the
maximum bending moment is equal to Mg rep = -0,232.Fg.r.
Due to the wind load, pe = 1,0 kN/m2 , the structure is subjected to radial nodal loads,
acting anti-metric on the arch, equal to Fe = 7,2 kN, see figure 3. Due to this load the
supports are subjected to a horizontal force equal to Hrep = 1,366.Fe. The vertical
reaction force is nihil, Rrep = 0. For node 5 and 6 the maximum bending moment is
equal to Me = Hrep.r.cos S/3.
For element S56 the stresses are validated with expression (1), with:
Reduced area: A = 0,6 * 0,125 * 0,5 = 37,5.103 mm2
Reduced modulus of the section: W = 0,6 * 0,125* 0,52/6 = 3,125.106 mm3
Reaction R = 2,5 * 9,72 = 24,25 kN
Thrust permanent load: Hrep = 1,134 * 9,72 = 11 kN
Bending moment due to the permanent load: Mg = 0,232*9,72* 3,6 = 8,1 kNm
Normal force due to the permanent load, S56: Ng = 18,1 kN
Thrust live load: Hrep = 1,366 * 7,2 = 9,9 kN
Bending moment due to the live load: Me = Hrep .r.cos S/3 = 17,7.kNm
Normal force due to the live load S56: Ne = 5,1 kN
443
Challenging Glass 3
1, 2 * 18 ,1 * 10 3 1, 2 * 8 ,1 * 10 6 1,5 * 5 ,1 * 10 3 1,5 * 17 , 7 * 10 6
|1
37 ,5 * 10 3 * 6 , 0 3,125 * 10 6 * 6 , 0 37 ,5 * 10 3 * 14 , 4 3,125 * 10 6 * 14 , 4
3. Strengthened arch
The dimensions can be reduced in case the arch is strengthened with cables running
from the supports to the nodes. The form-active structure is changed into a vector-active
structure, so the elements are mainly subjected to normal loads.
4 4
3 5 3 5
2 6 2 6
1 7 1 7
I r
I r
Figure 4: Arch composed of hinged bars and 4 cables Figure 5: Arch subjected to an anti-metrical radial
subjected to concentrated loads acting at the nodes. load.
Figure 4 and 5 show an arch composed of bars and cables connected with hinges. The
statically determinacy of the structure is defined with expression (2):
SD S 2 .K R (2)
With: SD = statically determinacy

S = number of bars
K = number of nodes
R = number of reactions
Substituting S = 10, K = 7 and R = 4 into (2) gives SD = 0, so the structure is statically

determinate. Successively the normal forces acting on the bars are defined from the top
to the supports. Table 2 shows, for the right half of the arch, the normal forces acting
parallel to the bars, due to concentrated nodal loads. To simplify the calculations the
loads acting at the nodes are solved into horizontal and vertical components. For the
members the sign is positive if the member is stretched and negative if the member is
compressed. A horizontal force or vertical force acting at a node is positive in case this
force is acting parallel to the positive X or Y axis.
444
Table 2: Solving horizontal loads and vertical loads acting on the nodes 4, 5 and 6 parallel to the members.
horizontal vertical
Node Member Normal force acting on node
component component
4 S45 = - ½ H4 /cos I½ V4 /sin I H5 = - S45 * cos I V5 = + S45 * sin I
5 S15 = + H5.sin 3Isin 5I V5.sin 3Isin 7I H5 = - S15 * cos 2I V5 = - S15 * sin 2I
S56 = - H5.sin 2Isin 5IV5.sin 4Isin 7I H6 = - S56 * cos 3I V6 = + S56 * sin 3I
6 S16 = + H6.cos I V6.sin I H6 = - S16 * cos I V6 = - S16 * sin I
S67 = - H6.sin IV6 cos I H7 = - S56 * sin I V7 = + S56 * cos I
Table 3 gives the normal forces acting on the members due to the horizontal and vertical
nodal loads including the forces from the supported members above. Table 4 gives the
normal forces, due to a vertical load acting on the nodes equal to Vi = -1p.
Table 3: Member forces due to the horizontal forces Hi and vertical forces Vi acting at the nodes including the
supported member forces.
Node Member Normal force Si due to a horizontal force Hi en vertical Force Vi
4 S45 = - ½ H4 /cos I ½ V4 /sin I
5 S15 = + (+ H5 – S45.cos I . sin 3Isin5I + (+V5 + S45.sin I sin 3Isin7I
S56 = - ( + H5 – S45.cos I sin 2Isin5I + (+V5 + S45.sin I sin 4Isin7I
6 S16 = + (+ H6 – S56.sin 3I.).cos I (+V6 + S56.cos 3I sin I
7 S67 = - ( + H6 – S56.sin 3I sin I (+V6 + S56.cos 3I cos I
Table 4: Normal forces due to a symmetrical vertical loading acting at the nodes equal to V4 = V5 = V6 = -1p
Normal force due to

Node Member Due to load V4 Due to load V3, V5 Due to load V2, V6
load 6Vi
4 S45 = -1,9318517 0 0 -1,9318517

5 S15 = +1,0 -0,7305080 0 +0,2679492

S56 = - 1,4142163 - 0,8965755 0 - 2,310789
6 S16 = + 0,7071068 + 0,4482877 - 0,2588191 + 0,8965755
S67 = - 1,2247449 - 0,7764571 - 0,9659258 - 2,9671278
Table 4 shows that for the permanent load the members of the arch are compressed and
the cables are tensioned, so this vector active structure transfers the vertical loads quite
well. Unfortunately some cables are compressed if the structure is subjected to
horizontal loads. Table 5 shows the forces due to anti-metrical radial loads. For the right
half of the arch the bars and cables are tensioned, but for the left part the bars and cables
are compressed. To avoid compressive loads acting on the cables the structure must be
post-tensioned or loaded by an extra weight at the top.
445
Challenging Glass 3
Table 5: Normal forces due to anti metrical radial loads acting at the nodes
Node Member Normal force due to a radial load Fi = 1 Fi = F
5 S15 = + (cos 4I).sin 3Isin 5I (sin 4I . sin 3Isin 7I +1.F
S56 = - (cos 4I .sin 2Isin 5I+ (sin 4I . sin 4Isin 7I +0,5176381.F
6 S16 = + [ - S56 .sin 3I + (+cos 2I)].cos I+ [S56.cos 3I + (+sin 2I)].sin I +0,7071068.F
S67 = - [ - S56.sin 3I + (+cos 2I)]..sin I>S56.cos 3I + (+sin 2I)].cos I +0,7071068.F
3.1. Example 2: strengthened arch

The arch described in example 1 is strengthened. Due to the permanent load the nodes
are subjected to a vertical force Fg = 9,72 kN and due to the live load the nodes are
subjected to a radial anti-metric load Fe = 7,2 kN. The cable S15 is subjected to a force:
S15 = + 0,268. Jg . Fg r 1,0.Jg . Fe
Substituting Jg = 0,9, Fg = 9,72 kN, Jg = 1,5 and Fe = 7,2 kN shows that this cable can be
subjected to a compressive load. To prevent the cable of subjected by a compressive
load we can decide to increase the vertical load acting at the top. Next the needed extra
force dFg acting at the top is calculated with Je = 0,9 and Je = 1,5:
S15 = 1,0 * dFg + 0,268 * 0,9 * 9,72 - 1,0 * 1,5 * 7,2 > 0 dFg > 9,4 kN p
4. Post-tensioned Arch
Post-tensioning can be helpful to avoid the cables loaded by compression. To post-
tension the arch two extra cables, S14 and S47, running from the supports to the top, are
added and tensioned. Table 6 shows the normal forces due to a vertical force P acting at
the top of the ach in case the two cables S14 and S47 are post-tensioned.
Table 6: Normal forces due to the post tensioning force P acting at the top, node 4
Node Member Normal forces due to the force P acting at the top Force
4 S45 = - ½ P /sin I -1,9318517.P
5 S15 = -S45.cosIsin3Isin5IS45.sinIsin3Isin7I +1,0.P
S56 = +S45.cosIsin2Isin5IS45.sinIsin4Isin7I - 1,4142163.P
6 S16 = + (-S56. sin 3I.cos I S56. cos 3I.sinI + 0,7071068.P
S67 = - (-S56. sin 3I. Sin I S56. cos 3IcosI - 1,2247449.P
446
4
3 5
2 6
1 7
I r
Figure 6: Ach loaded at the top due to the post tensioning of the cables S14 and S47.
4.1. Statically indeterminate arch

Adding the two cables S14 and S47 increases the statically determinacy of the structure,
so the transfer of the loads is effected by the stiffness of the members. To define the
load distribution the structure is partitioned into a statically determinate arch
strengthened with 4 cables and the triangular structure composed of the two diagonals
S14 and S47. The deformation u of both vector active structures is defined with the
following expression according to the Theory of Mohr.
6 in 1 N i . N 'i .l (4)
u
E i . Ai
With:
u= Deformation of the structure for a given node
Ni = normal force acting into element i due to the loading
N’i = normal force acting into element i due to the virtual load F = 1 acting at the given node parallel to
the deformation u
Ei = Young’s modulus of an element
Ai = Area of an element
1.1. Deformation of the triangular structure at the top

Due to the force F acting at the top both elements S14 en S47 are subjected to a force
½.F2. The length of the bars is equal to r2. The deformation of the triangle formed by
S14 en S47 with section Ad and Young’s Modulus Ed is equal to:
F .r 2 (5)
u
E d . Ad
4.2. Deformation of the statically determinate arch at the top

The deformation of the statically determinate arch due to a force F acting at the top is
calculated with expression 4 and table 7. The vertical deformation of the arch at the top
due to the force V4 = F acting at node 4 is equal to:
F .r .C p
u With: CP = 2 * (3,74359. EdAd/EsAs + 2,69798) (6)
E d . Ad
447
Challenging Glass 3
Table 7: Deformation of the ach due to a concentrated load acting at the top V4 = - Fp
Member N N' Length N.N'.length/EiAi
S45 -1,9318517.F -1,9318517 2.r.sin I EsAs 1,9318517.F.r/(Es.As)
S15 +1,0.F +1,0 2.r.cos 2I Ed.Ad 1,7320508.F.r/(Ed.Ad)
S56 - 1,4142163.F - 1,4142163 2.r.sin I Es.As 1,0352762.F.r/(Es.As)
S16 + 0,7071068.F + 0,7071068 2.r.cos I Ed.Ad 0,9659258.F.r/(Ed.Ad)
4.3. Load distribution for a concentrated load acting at the top

For the structure subjected to a force F acting at the top, the triangular structure,
composed of the cables S14 en S47, will transfer the force DF and the determinate arch
will transfer a force (1-D) F. The ratio D follows from the following equation:
(1 D 1 ) F .r .C p D 1 . F .r . 2 D Cp (7)
u 1
E d . Ad E d . Ad Cp 2
Probably the sections of the cables are smaller than the sections of the bars, so the
parameter Cp is much larger than 2 and the ratio D1 is within a range of 0,8 <D1 < 0,9,
see table 8.
Table 8: Range of the ratioD1

EdAd = EsAs Cp = 12,8831 D= 0,901
EdAd = ½ EsAs Cp = 9,13954 D= 0,866
EdAd = ¼ EsAs Cp = 7,26775 D= 0,837
EdAd = 0,1.EsAs Cp = 6,1447 D= 0,813
4.4. Deformation of the arch subjected to a symmetrical vertical loading

The deformation of the statically determinate arch subjected to vertical forces Vi = -F

acting at the nodes. The vertical deformation of the top is equal to:
F .r .C V with: CV = 2 * (5,50455.EdAd/EsAs + 1,69152) (8)

u
E d . Ad
Table 9: Deformation of the statically determinate arch subjected to loads V2 = V3 = V4 =V5 = V6 = - Fp

Member Load 6Vi = -F N' Length N.N'.length/EAi
S45 -1,9318517 .F -1,9318517 2.r.sin I EsAs +1,9318517.F.r/(Es.As)
S15 +0,2679492.F +1,0 2.r.cos 2I Ed.Ad 0,4667738.F.r/(Ed.Ad)
S16 + 0,8965755.F + 0,7071068 2.r.cos I Ed.Ad 1,2247048.F.r/(Ed.Ad)
448
4.5. Load distribution for a symmetrical vertical loading

For the structure subjected to a vertical symmetrical loading acting at the nodes the
triangular structure composed of the cables S14 en S47 will transfer the force DF and
the determinate arch will transfer a force (1-D) F. The ratio D is defined by the
stiffness of both parts and follows from the compatibility of the structure:
( C V D 2 .C P ) F .r D 2 . F .r . 2 D CV (9)
u 2
E d . Ad E d . Ad CP 2
Table 10: Forces acting at the members due to the vertical load 6Vi -D.F, with Vi = - Fp and D= 0,59
Member Normal force due to 6Vi V4 = -1 p V4 = D.Fn 6Vi -D.F
S45 -1,9318517.F -1,9318517 +0,67248.F -1,2594.F
S15 +0,2679492.F +1,0 -0,59.F -0,3221.F
S56 - 2,310789.F - 1,4142163 + 0,8344.F - 1,4764.F
S16 + 0,8965755.F + 0,7071068 - 0,4172.F + 0,4794.F
S67 - 2,9671278.F - 1,2247449 + 0,7226.F -2,24453.F
The area of the cables will be smaller than the area of the members of the arch. For
EdAd = 0,1.EsAs. The parameters CP and Cv are equal to Cp = 6,10 and Cv = 4,42 and the
ratio Dis equal to 0,59. So due to the vertical loads Vi = F the triangular structure is be
subjected to a vertical load equal to 0,59.F. Table 10 shows the forces acting on the
members for the structure subjected to the loading 6Vi -D.F.
4.6. Deformation of the arch subjected to a horizontal load acting at the top
The horizontal deformation of the arch due to a horizontal load H acting at the top
follows from:
H .r .C H with C = 2 * (0,26878.E A /E A + 0,22071) (10)
u H d d s s
E d . Ad
Table 11: Deformation of the arch subjected to a horizontal Force H4

Member H'4= H H'4= 1 Length N.N'.length/EAi
S45 - 0,5176381.H - 0,5176381 2.r.sin I EsAs 0.13870.H.r/(Es.As)
S15 + 0,2679492.H + 0,2679492 2.r.cos 2I Ed.Ad + 0,1243567.H.r/(Ed.Ad)
S56 - 0,3789374.H - 0,3789374 2.r.sin I Es.As - 0,0743305.H.r/(Es.As)
S16 + 0,1894687.H + 0,1894687 2.r.cos I Ed.Ad + 0,0963504.H.r/(Ed.Ad)
S67 - 0,3281694.H - 0,3281694 2.r.sin I Es.As - 0,0557471.H.r/(Es.As)
449
Challenging Glass 3
4.7. Deformation of the arch subjected to a radial anti-metrical load

The deformation of the arch due to the anti-metric radial load Fi acting at the nodes is
calculated with table 12. The horizontal deformation follows from:
F .r .C r with: Cr = 2 * (-0,22166.EdAd/EsAs + 0,73539) (11)

u
E d . Ad
Table 12: Deformation sue to anti metric radial load F2, F3, F5 en F6
Member Fr = F H'4 = 1o Length N.N'.length/EAi
S45 0 - 0,5176381 2.r.sin I EsAs 0
S15 + 1,0.F + 0,2679492 2.r.cos 2I Ed.Ad + 0,4641.F.r/(Ed.Ad)
S56 + 0,5176381.F - 0,3789374 2.r.sin I Es.As - 0,101537.F.r/(Es.As)
S16 + 0,7071068.F + 0,1894687 2.r.cos I Ed.Ad + 0,258819.F.r/(Ed.Ad)
S67 + 0,7071068.F - 0,3281694 2.r.sin I Es.As - 0,120118.F.r/(Es.As)
4.8. Load distribution for the horizontal load

For the structure subjected to a vertical symmetrical loading acting at the nodes the
triangular structure composed of the cables S14 en S47 will transfer the force DF and
the determinate arch will transfer a force (1-D).F. The ratio D is defined by the
stiffness of both parts and follows from the compatibility of the structure:
( C r D 3 .C H ). H .r D 3 . H .r . 2 D3 Cr (12)
u
E d . Ad E d . Ad CH 2
Mostly the area of the cables is smaller than the area of the members of the arch. For
EdAd = 0,1.EsAs the parameters Cr and CH are equal to Cr = 1,43 and CH = 0,47 then the
ratio Dis equal to 0,76. Due to the anti-metrical radial loading Fr the triangular
structure is be subjected to a horizontal load equal to 0,76.H. Table 13 shows the normal
forces acting on the members for the structure for D 0,76.
450
Table 13: Normal forces due to the anti-metric load F5=F6= Fr and H4 =D.Fr with D= 0,76.
Staaf Fr= Fr H'4 = 1.Fr o H4 = -D.Fr Fr-D.Fr
S45 0 - 0,5176381.Fr +0,3882.Fr r0,3882.Fr
S15 + 1,0.Fr + 0,2679492.Fr -0,2010.Fr r0,7990.Fr
S56 + 0,5176381.Fr - 0,3789374.Fr +0,2842.Fr r0,8018.Fr
S16 + 0,7071068.Fr + 0,1894687.Fr -0,1451.Fr r0,5650.Fr
S67 + 0,7071068.Fr - 0,3281694.Fr +0,2513.Fr r0,9532.Fr
4.9. The post-tensioning load

The post-tensioning load must be able to compensate the compressive loads acting on
the cables. The horizontal and vertical loads are for the better part resisted by the
triangular structure. To prevend compressive normal force acting on the diagonals S14
en S47 the required post-tensioning load P, acting at the top, follows from:
1,0.P J g .D 2 .Fv J e .D 3.Fr 0 (13)
4.10. Example 3, Post-tensioned arch

The roof described for example 1 is supported with a post-tensioned arch. The structure
is subjected to concentrated loads acting at the nodes due to the permanent load and the
anti-metrical radial load. The forces due to the permanent load and radial load are
respectively Fg = 9,7 kN and Fr = 7,2 kN. The section of the arch is 100 * 100 mm2,
Eglass = 70 GPa, the section of the cables is 314 mm2, Es=210 GPa, EdAd/EsAs = 0,1 D
0,59 and D 0,76. According to expression (13) the post-tensioning force P has to be:
P ! 1,2 * 0,59 * 9,7 1,5 * 0,76 * 7,2 15,1 kN
The maximum normal load is acting on the member S67. Due to the post-tensioning
force P, the permanent and the live load this member is according to table 7, 10 and 13
subjected to:
S 67 1,225 * 1,0 * 15,1 2,245 * 1,2 * 9,7 0,953 * 1,5 * 7,2 55 kN
The compressive stress due to this normal force is: V = Nd/A = 5,5 MPa.
4.11. Example 4, Grid shell

The arch is composed of slender elements, with a section of 100 * 100 mm2. This
section is much smaller than the section of the beams, 50 * 360 mm2, spanning 3,6 m
perpendicular to the arch. Reducing the centre to centre distance of the arches will
reduce the dimensions of the beams and improve the design.
The centre to centre distance is halved from 3,6 m until 1,8 m. The glass panes,
composed of three sheets of 8 mm, thickness 3* 8 mm, are supported at four sides by a
grid composed of arches and beams, section 80 * 125 mm2.
451
Challenging Glass 3
Table 14: Loads acting on the members of the grid:

Roof glass panes: 0,024 * 25 * 1,9 = 1,14 kN/m M = 1,14 * 1,92/12 = 0,343 kNm
2
Beam: 0,08 * 0,125* 25 = 0,25 kN/m M = 0,25 * 1,9 /8 = 0,113 kNm
2
Live load 1,0 * 1,9 = 1,9 kN/m M = 1,9 * 1,9 /12 = 0,572 kNm
2
Modulus of the section W: 80 * 125 /6 = 0,208* 10 mm3
6
The beams running perpendicular to the arches are subjected to bending, the bending
stresses are validated with expression (1):
1, 2 * ( 0 ,113 0 ,343 ) * 10 6 1,5 * 0 ,572 * 10 6

6
0 ,72 d 1
0 , 208 * 10 * 6 ,0 0 , 208 * 10 6 * 14 , 4
The arches are subjected to bending too but the bending stresses are partly compensated
by the compressive normal forces.
Table 15: Loads acting at the nodes of the arch:

Load Weight Length Fep
Roof glass panes: 0,024 * 1,8 * 25 * 1,9 = 2,05 kN
Beams: 0,080 * 0,125 * 25 * 1,8 = 0,45 kN
Arch; 0,08 * 0,125 * 25 * 1,9 = 0,48 kN
Node, plates, bolts: 0,40 kN
Force due to the permanent load 3,4 kN
Force due to the wind load 1,0 * 1,8 * 1,9 = 3,4 kN
The arch is subjected to a radial load Fr = 3,4 kN and a vertical load Fv = 3,4 kN, with
EdAd = 0,1.EsAs, D 0,76 and D 0,59. According to expression (13) the post-
tensioning force P has to be:
P ! 1, 2 * 0 ,59 * 3, 4 1,5 * 0 , 76 * 3, 4 1,85 * 3, 4 kN
The maximum normal load is acting on the member S67. Due to the post-tensioning
force P, the permanent and the live load this member is, according to table 7, 10 and 13,
subjected to:
S 67 1,225 * 1,0 * 1,85 * 3,4 2,245 * 1,2 * 3,4 0,953 * 1,5 * 3,4 21,7 kN
The compressive stress due to this normal force is: V = Nd/A = 2,2 MPa.
The beams and the members of the arches are connected with two bolts M12 to the
nodes. To avoid stress concentrations in the elements of glass the forces are transferred
to plates of steel, glued at both sides of the glass elements [6]. The nodes are composed
of two crossing plates to connect the beams and elements of the arches at every side, see
figure 7.
452
Figure 7: Vertical and horizontal cross section for the node at the top of the arch.
The shear stress acting on the glued surface between the plates and the elements of glass
has to be smaller than the ultimate shear stress, W = 5 MPa. The area of the glued surface
is equal to 80 * 40 mm2. Member S67 is subjected to a maximal force of F = 21,7 kN.
For the glued surface the shear stress is equal to:
21,7 * 10 3
W 3,4 5 MPa
2 * 40 * 80
5. Conclusions
For structures of glass, arches are very efficient, especially in case the system line of the
structure approaches the line of thrust and the normal forces compensate for the better
part the tensile bending stresses. The lines of the thrust will vary much in case the
permanent load is not much larger than the live load. For these structures it will be
efficient to strengthen the arch with cables. To prevent any compression into the cables
it is advisable to post-tension the cables. Post-tensioned arches are statically
indeterminate, the diagonals running from the top to the supports will transfer the better
part of the loads. Consequently the tensioning has to be pretty large to prevent the
cables to be subjected to compression. Nevertheless, the post-tensioned arches are
mainly subjected to normal compressive forces and can be dimensioned very slender.
The transparency of the described grid shell will be impressive. The analysis of the
post-tensioned arches is complex, but the given method and tables can be helpful to
design these structures. Further research is needed to test and construct these transparent
roof structures of glass.
6. References
[1] Belenya E., Prestressed Load-Bearing Metal Structures, MiR Publishers, Moscow, USSR, 1977;
[2] Engel, Heino, Structure Systems, Verlag Gerd Hatje, Germany, 1997;
[3] Haarhuis, Kars, Maximum Transparent Barrel-Vaulted Glass Roof, Proceedings Symposium IABSE-
IASS 2011: Taller Longer Lighter, London, Great Britain, 2011;
[4] Louter, Christian; Belis Jan; Bos, Freek; Veer, Fred; Hobbelman, Gerrie, Reinforced Glass Cantilever
Beams, Proceedings of the 9th International Conference on Architectural and Automotive Glass (Glass
Processings Days), Tampere, Finland, 2005;
[5] Louter, Christian; J.F van Heusden; F.A. Veer; J.N.J.A. Vambersky; H.R. de Boer; J. Versteegen; Post-
tensioned Glass Beams, Repository Library of Delft University of Technology, The Netherlands;
[6] Veer, Fred; Wurm Jan; Hobbelman, Gerrie, The Design, Construction and Validation of a structural
Glass Dome, Repository library of Delft University of Technology, The Netherlands.
453
doi:10.3233/978-1-61499-061-1-455
Design of Cable Strenghtened Beams for

Glass Structures
Wim Kamerling
Delft University of Technology, The Netherlands, M.W.Kamerling@tudelft.nl
This paper focuses on the design of beams of glass strengthened with post-
tensioned cables. Due to the strengthening the bending moments are decreased and
the structural efficiency is improved, consequently the dimensions of the beams can
be reduced. Post-tensioning the cables improves the bearing capacity and
transparency of the structure. The effect of the post-tensioning is analyzed and the
advantages as well as disadvantages are described. Recommendations and rules of
thumb are given for the design of beams of glass, strengthened with post-tensioned
cables and struts.
Keywords: Glass, Cable Strengthened Beams, Post-Tensioning
1. Introduction
Structurally it is efficient to strengthen beams with cables and struts in case the structure
is subjected to a downward load. Due to the strengthening the dimensions of these
beams can be reduced considerately. The slender structures are light and transparent,
especially in case the structure is composed of glass beams and glass struts strengthened
with steel cables, see figure 1.
q
D

Figure 1: Scheme of a beam strengthened with cables and a single strut
Post-tensioning the cables is effective [1]: the upward forces acting on the beam are
increased and the bending moments due to downward loads are decreased. For glass the
tensile strength is much smaller than the compressive strength. Post-tensioning the
cables increases the normal force acting on the beam and decreases the tensile stresses,
so the bearing capacity is increased. This paper focuses on the design of glass beams
strengthened with struts and post-tensioned cables. The advantages and disadvantages
are described, the load distribution is analyzed and recommendations for the design are
given.
455
Challenging Glass 3
2. Load transfer
A beam strengthened with struts and cables is not a truss. A truss is a vector active
structural system, composed of bars subjected to normal forces [2]. A strengthened
beam is subjected to normal forces as well as bending. The beam can be supported
with a single strut as well as several struts. Structurally it is efficient to increase the
number of struts [3]. The strengthened beam is statically indeterminate; the load transfer
is effected by the stiffness of the components.
3. Advantages and disadvantages

Strengthening a beam with cables and struts offers the following advantages:
x Dimensions: due to the reduction of the bending moments, the dimensions and
self weight of the beam can be minimized;
x Footprint: minimizing the need of materials reduces the embodied energy;
x Statically indeterminate: the strengthened beam is statically indeterminate,
consequently the cables can be post-tensioned to reduce the bending moments
and to increase the bearing capacity of the beam;
x Adaptability: the bearing capacity of strengthened beams can be adapted easily,
in case the spans or loads vary, by increasing the number or length of the
struts;
The strengthening has disadvantages too:

x Height: the cables and struts decrease the useful height of the inner space;
x Costs: the manufacturing and assemblage of struts, cables and joints increase
the cost of construction;
x Structural analysis: the analysis of the statically indeterminate post-tensioned
structure is rather complex;
x The cables can not resist compressive forces, the resistance for upward loads is
poor.
4. Load transfer, beam strengthened with one strut

A strengthened beam is statically indeterminate. The loads are transferred by two

systems: a section active structure system (the beam) and the vector active structure
system composed of the beam, cables and struts. To explain the load transfer, a
strengthened beam is considered with a span of 2*a, see figure 2.
q

EF

EF
D

Figure 2: Load distribution for a beam strengthened with a single strut and cables
456
Design of Cable Strengthened Beams for Glass Structures
The beam is subjected to an equally distributed load q and an upward force equal to E.F
acting halfway the span. The truss is subjected to a downward load E.F. The factor
Edepends on the stiffness of the beam and the truss. For E = 1 the upward force is
equal to the maximum force: F = c1.q.an with c1 = 5/4. In practice E is smaller than 1.
The factor E follows from the compatibility of the truss and beam. The deformation of
the beam and truss are respectively equal to:
5.q.a 4 E .F .a 3
Beam: u (1)
24.Eb I b 6.Eb .I b
Truss:
E .F .a.C1 Eb Ab 1 E A
u with: C1 tan D . b b (2)
Eb Ab E c .A c .2.sin D .cos D 2. tan D
2 2
E s As
With:
Young’s modulus: Area;
beam: Eb Ab
strut: Es As
cable: Ec Ac
The deformation of the beam is equal to the deformation of the truss. With equation (1)
and (2) the factor Eis defined with the following expression:
5.q.a 4 E .F .a 3 E .F .a.C1
= (3)
24.Eb I b 6.Eb .I b Eb Ab
With F = c1.q.a and c1 = 5/4 the factor E follows from:
1
E (4)
1 6.C1.I b /( Ab .a 2 )
This expression shows that the effect of the strengthening rises in case the factor C1
decreases. The bending moment halfway the span is equal to:
1 1
M q.a 2 E .F .a (5)
2 2
For E = 0,8 the bending moment halfway the span is zero, the upward force is equal to:
F = 0,8 * 5/4 q. a and the maximum bending moment is equal to: Mmax = 1/8 q a2. Due to
the strengthening the maximum bending moment is only 25% of the bending moment in
case this beam is not strengthened.
5. Load transfer, beam strengthened with two struts

It is profitable to increase the number of struts [3]. To show the efficiency of a beam
strengthened with two struts, see figure 3, the load transfer is analyzed. The structure is
statically indeterminate, the loads are transferred by two systems: the section active
457
Challenging Glass 3
structure system (the beam) and the vector active structure system composed of the
cables, struts and beam. The length of the beam l is equal to 2.a + b, see figure 3.
q

EFEF

EF E.F
D

a b a
Figure 3: Load distribution for a beam strengthened with two struts
The beam is subjected to an equally distributed load q and to two upward forces EF,
with F = c2.q.a. The truss is subjected to two downward forces equal to E.F. The factor
Edepends on the stiffness of the beam and the truss. If the truss is very stiff then E = 1,
in practice E is smaller than 1. If the truss is partitioned in three equal parts, so b = a,
then the factor c2 is equal to 1,1. The factor E follows from the compatibility of the truss
and beam. For x = a the deformation of the beam and truss is respectively equal to:
q.(2a b) 4 a4 2.a 3 a E .F .a 3 (2 3b / a)
u [ ] (6)
24.Eb I b (2a b) 4 (2a b)3 (2a b) 6.Eb .I b
E .F .a.C2
u with:
Eb Ab
Eb Ab / Ec Ac 1 Eb Ab b / a.(1 Eb Ab / Ec Ac )
C2 tan D . (7)
sin D .cos D
2
tan D2
E s As 2. tan 2 D
The deformation of the beam is equal to the deformation of the truss, the factor E is
defined with expression (6) and (7):
q.(2a b)4 a4 2.a3 a E.F.a3(2 3b / a) E .F .a.C2

[ ] (8)
24.EbIb (2a b)4 (2a b)3 (2a b) 6.Eb.Ib Eb Ab
For b = a equation (8) can be simplified:
22.q.a4 5.E.F.a3 E.F.a.C2

(9)
24.Eb Ib 6.Eb .Ib Eb .Ab
With F = c2 * q * a and c2 = 1,1 the factor E follows from:
458
1
E (10)
1 6 C2 .I b /( Ab .a 2 )
5
For b = a and F = 1,1.E.q.a the bending moment halfway the span is equal to:
M 9 q.a 2 1,1.E .q.a 2 (11)

8
For x < a the bending moment follows from:
M 3 q.a.x 1,1.E .q.a.x 1 q.x 2 (12)

2 2
The bending moment is maximal in case the shear force is minimal:
V 3 q.a 1,1E .q.a q * x 0 (13)

2
For E = 0,9 the shear force is zero for x = ½ a. The maximum bending moment is equal
to Mmax = 1/8 q * a2 . This moment is much smaller than the maximal bending moment
defined for the structure with one strut.
6. Post-tensioning
Especially for small values of E it is effective to post-tension the cables to increase the
loading capacity of the structure. For glass the tensile stress is decisive, the normal load
due to the post-tensioning will reduce the tensile bending stresses and increase the
loading capacity of the beam.
P.cos D
D
2.P.sinD
P
a a
Figure 4: Forces due to the post-tensioning
Figure 4 shows the forces in case the cables are subjected to a post-tensioning load P.
Due to the post-tensioning load the beam is subjected to a normal force P.cos D and an
upward force equal to 2*P.sin D. For the beam the tensile stresses are checked with the
following equation:
J p .P. cos D J p .M p J g .N g J g .M g J e .N e J e .M e
d1 (14)
A. f t f W . ft f A. f t f W . ft f A. f t 0 W . ft 0
459
Challenging Glass 3
With:
Load factor permanent load: Jg = 1,2
Load factor payload: Je = 1,5
Load factor post tensioning load: Jp = 1,0
Area: A [mm2]
Modulus of the section: W [mm3]
Bending moment due to the permanent load: Mg [Nmm]
Normal force due to the permanent load: Ng [N]
Bending moment due to the live load: Me [Nmm]
Normal force due to the live load: Ne [N]
Normal post-tensioning load acting on the beam: P.cos D[N]
Bending moment due to the pre-stressing load: Mp [Nmm]
Maximal tensile stress, time permanent loads, t = f : f t =f [Mpa]
Maximum tensile stress, instantaneous loads, t = 0: f t=0 [Mpa]
Figure 5: The bending moments due to the permanent load and the post-tensioning (red) causing
tensile stresses at the top and the bending moments due to the permanent load, post-tensioning
and live load (grey) causing tensile stresses at the bottom of the beam.
The minimal post-tensioning load can be defined with expression (14). The maximum
post-tensioning load follows from the limitation of the tensile stress for x = a above the
strut halfway the span:
J p .P. cos D J p .M p J g . fav .N g J g . fav .M g

d1 (15)
A. f t f W . ft f A. f t f W . ft f
With:
Load factor permanent load favourable: Jg fav = 0,9
Generally floors are not be subjected to an upward loading, nevertheless roofs can be
subjected to an upward wind load. For a roof subjected to an upward load the maximum
post-tensioning follows from the limitation of the tensile stress above the strut, halfway
the span:
460
J p .P. cos D J p .M p J g . fav .N g J g . fav .M g J e .N up J e .M up

d1 (16)
A. f t f W . ft f A. f t f W . ft f A. f t 0 W . ft 0
With:
Load factor permanent load favourable: Jg fav = 0,9
Bending moment due to upward load: Mup [Nmm]
Normal force due to upward load: Nup [N]
The cables cannot resist any compressive force so the normal force, due to the uplift,
has to be smaller than the normal force due to the permanent load and post-tensioning:
J e .N up J g . fav .N g
J p .P (17)
cos D cos D
7. Design
A not strengthened beam can be designed with the following rule of thumb:
J g .q g .l 2 J e .qe .l 2
W ! (18)
8. f gd 8. f ed
However a strengthened beam can be designed with the following rule of thumb:
J g .q g .a 2 J e .qe .a 2
W! (19)
8. f gd 8. f ed
Comparing these expressions show the effect of the strengthening.

8. Details
The details have to be designed according to the load transfer. For glass the joints are by
preference not tensioned but compressed.
Figure 6: Details, connecting the cables to the end plate and connecting the struts to the beam.
461
Challenging Glass 3
For the strengthened structure the force acting on the cables has to be transferred to the
beam and the struts. To avoid the struts to turn over, the struts and cables are split up in
two parts running besides the beam. The cable force Nc is solved into a horizontal force
Nc.cos D and a vertical component Nc.sin D. Generally the angle D is small, so the
vertical component is smaller than the vertical component. To connect the cable with
the beam, by preference the horizontal component is transferred by compression and the
vertical component is transferred by shear. According to the details designed by Louter
et al [5] the post-tensioning force is transferred with two wedge shaped plates glued to
the beam. The cables, running besides the beam, are bolted to the end plate.
Perpendicular to the end plate two wedge-shaped plates are made so the end plate can
be shifted over the wedge-formed plates glued to the beam. To transfer the upward load,
halfway the span, aluminum side plates are glued to both struts and bolted to the T-
shaped profiles, which are glued to both sides of the beam, see figure 6.
9. Example beams of glass

To show the effect of the strengthening for a transparent floor three alternative designs
are described and compared. The first beam is not strengthened, the second beam is
strengthened with cables and one single strut, and the third beam is strengthened with
two struts.
The beams support a floor composed of glass plates, thickness 3 * 10 mm. The plates
are supported by purlins, center-to-center 1,2 m, with a section of 45 * 250 mm2 and a
span of 2,4 m. The purlins are supported by the beams spanning 7,2 m, center-to-center
2,4 m. The maximal length of the glass is 6 m [4]. The beams are composed of sheets
with a length of respectively 1,2-6,0 m, 2,4-4,8 m, 3,6-3,6 m, 4,8-2,4 m and 6,0-1,2 m.
At a joint one sheet is cut and 4 sheets are continuous, so the section is weakened.
Consequently at the joints the stresses are calculated with a reduction factor equal to 4/5.
The maximal tensile bending stress is for the permanent loads equal to fgd = 6,0 MPa.
For the payload, acting instantaneously, the maximal tensile stress is equal to fed = 14,4
MPa. The payload acting on the floor is equal to 3,0 kN/m2. The dead load of the floor
and purlins is equal to 0,75 + 0,25 kN/m2.
Loads acting on the beams:

Permanent load, exclusive self weight: qg = (0,75 + 0,25) * 2,4 = 2,4 kN/m
Pay load: qe = 3,0 * 2,4 = 7,2 kN/m
9.1. Not strengthened beam

The beam is constructed of 5 sheets with a thickness 20 mm and a height of 900 mm;
the section is equal to100 * 900 mm2.
462
4
Area, reduced section A= /5 * (5*20) * 900 = 72,0 * 103 mm2
4
Moment of resistance W= /5 * 100* 9002/6 = 10,8 * 106 mm3
Permanent load qg = 0,1*0,9*25 + 1,0 * 2,4 = 4,65 kN/m
Pay load qe = 2,4 * 3,0 = 7,2 kN/m
2
Moment, permanent load Mgd = 4,65 * 7,2 /8 = 30,1 kNm
2
Moment, pay load Med = 7,2 * 7,2 /8 = 46,7 kNm
The tensile bending stresses are checked with equation (14) for x = a:
1,2 * 0 1,2 * 30,1 * 10 6 1,5 * 0 1,5 * 46,7 * 10 6

3
6
|1
72 * 10 * 6 10,8 * 10 * 6 72 * 10 * 14,4 10,8 * 10 6 * 14,4
3
9.2. Beam strengthened with one strut

The beam is strengthened with one strut, halfway the span, a = 3,6 m center and two
cables running on both sides with a diameter of 20 mm, S235, Ssteel = 210 GPa. To
prevent turning over the strut is split into two parts, both section 30 * 100 mm2, length
0,9 m, tan D = ¼, D = 14,040 , sin D = 0,2425, cos D = 0,97. The two parts are jointed at
both sides to the beam. The beam is constructed of 5 plates with a thickness 15 mm and
a height of 500 mm, section 75 * 500 mm2, Eglass = 70 GPa. The distribution ratio E is
defined with expression (2) and (4), with F = 5/4 q.a.
26,25 *108 1 0,25 * 26,25 *108

C1 = 184
1,32 * 108 * 2 * 0,24252 * 0,97 2 * 0,252 4,2 * 108
1
E 0,36
1 6 *184 * 7,8125 * 108 /(75 * 500 * 3600 2 )
The upward force is equal to F = 0,36 * 5/4 q.a.
Permanent load: qg = 0,075 * 0,5 * 25 + 2,4 = 3,34 kN/m

Bending moment Permanent load Mg = 3,34* 7,22/8 = 21,6 kNm
2
Bending moment Pay load Me = 7,2 * 7,2 /8 = 46,7 kNm
Forcen Permanent load EFg = 0,36 * 1,25 * 3,34 * 3,6 5,4 kN
Forcen Pay load E.Fe = 0,36 * 1,25 * 7,2 * 3,6 11,7 kN
Normal force Permanent load: Ng = ½ * 5,4/(tan D) = 10,8 kN
Normal force Pay load: Ne = ½ * 11,7 /(tan D) = 23,4 kN
4
Area, reduced section: A= /5 * 75 * 500 = 30,0 * 103 mm2
4
Moment of resistance W= /5 * 75* 5002/6 = 2,5 * 106 mm3
463
Challenging Glass 3
The tensile bending stresses are checked with the equation (14) for x = a:
1,2 *10,8 *10 3 1,2 * (21,6 5,4 * 7,2 / 4) *10 6 1,5 * 23,4 *10 3 1,5 * (46,7 11,7 * 7,2 / 4) *10 6 = 1,87 >1
3
6
3

30 *10 * 6,0 2,5 *10 * 6,0 30 *10 *14,4 2,5 *10 6 *14,4
The stresses do not fulfill the demands. To reduce the bending stresses the structure is
post-tensioned. The needed minimal post–tensioning is found for x = ½ a, with
expression (14):
1,0 * P * cos D * 10 3 1,0 * 1 * P * sin D * 3,6 * 10 6

2 >
30 *10 3 * 6,0 2,5 * 10 6 * 6,0
1,2 *10,8 *103 1,2 * ( 3 * 21,6 1 * 5,4 * 7,2 / 4) *106 1,5 * 23,4 *103 1,5 * ( 3 * 46,7 1 *11,7 * 7,2 / 4) *106
4 2 4 2 1
3
30 *10 * 6,0 2,5 *106 * 6,0 30 *103 *14,4 2,5 *106 *14,4
The minimum post tensioning force has to be P > 22,5 kN. Next the maximum post-
tensioning load is defined with expression (15) for x = a:
1,0 * P * 0,97 1,0 * P * 0,2425 * 3600 0,9 * 10,8 * 10 3 0,9 * (21,6 5,4 * 7,2 / 4) * 10 6
3
6
3
d1
30,0 * 10 .6,0 2,5 * 10 * 6,0 30,0 * 10 * 6,0 2,5 * 10 6 * 6,0
P < 33,3 kN.
For the minimum post-tensioning load P = 22,5 kN the strut is subjected to an upward
load 2.P.sin D = 10,9 kN, the beam is subjected to a normal force Np = P.cos D = 21,8
kN and a bending moment Mp = 2.P.sin D(2.a)/4 = 19,6 kNm.
9.3. Beam strengthened with two struts

Next the structure is designed with a strengthening composed of two struts center to
center a = 2,4 m and two cables running on both sides with a diameter of 20 mm. The
beam is constructed of 5 plates with a thickness 15 mm and a height of 300 mm, section
75 * 300 mm2, Eglass = 70 GPa. Both struts, length 0,6 m, are split in two parts which
are connected to the sides of the beam: 2 *30*100 mm2, tan D = 6/2,4 = ¼, D = 14,040 ,
sin D = 0,2425, cos D = 0,97, cables 20, area 2 * 314 m2. E steel = 210 GPa .
a tan D
a a a
Figure 7: Beam strengthened with two struts
464
Self weight G= 0,075 * 0,3 * 25 = 0,56 kN/m

3
Second moment of area Iby = 75 * 300 /12 = 168,75 *106 mm4
The factor C2 is defined with expression (7):

7 *104 * 22,5 *103 1 7 *104 * 22,5 *103 1 7 *104 * 22,5 *103 /(2,1*105 * 628)
C2 5 2
2
0,25 * 4

2,1*10 * 628* 0,2425 * 0,97 0,25 7 *10 * 2 * 30 *100 2 * 0,252
C2 = 330. Next the distribution ratio E is defined with expression (10):
1
E = 0,66
6
1 * 330 *168,75 *106 /( 22,5 *103 * 24002 )
5
Permanent load qg = 0,6 + 1,0 * 2,4 = 3,0 kN/m

Pay load qe = 2,4 * 3,0 = 7,2 kN/m
2
Permanent load Mg = 3,0* 7,2 /8 = 19,4 kN/m
2
Pay load Me = 7,2 * 7,2 /8 = 46,7 kNm
Permanent load EFg = 0,66 * 1,1 * 3,0 * 2,4 5,2 kN
Moment upward force Mg = EF.a = 5,2 * 2,4 = 12,5 kNm
1
Normal force Ng = EF/tan D 5,2/( /4) = 20,8 kN
Pay load, upward force E Fe = 0,66 * 1,1 * 7,2 * 2,4 12,5 kN
1
Normal force: Ne =EF/tan D 12,5/( /4) = 50,0 kN
Moment upward force Me = EF*a 12,5 * 2,4 = 30,0 kNm
4
Reduced section area Ab = /5*75 * 300 = 18,0 * 103 mm2
4
/5 *75* 3002/6 = 0,9 * 106 mm3
Moment of resistance W=
The tensile bending stresses are checked with the equation (15), halfway the span, for x
= 3/2 a:
1, 2 * 20 ,8 .10 3 1, 2 .(19 , 4 12 ,5 ). 10 6 1,5 * 50 ,0 .10 3 1,5 .( 46 ,7 30 ,0 ). 10 6

3
5
3
!1
18 * 10 * 6 9 * 10 * 6 18 * 10 * 14 , 4 9 * 10 5 * 14 , 4
The stresses do not fulfill the demands. To reduce the bending stresses the structure is
post-tensioned. According to expression (14) the minimal post-tensioning load follows
from:
1,0 * P * cos D *10 3 1,0 * P * sin D * 2,4 *10 6

!
18 *10 3 * 6,0 0,9 *10 6 * 6,0
1, 2 * 20 ,8 * 10 3 1, 2 * (19 , 4 12 ,5 ) * 10 6 1,5 * 50 ,0 * 10 3 1,5 * ( 46 ,7 30 ,0 ) * 10 6
1
18 * 10 3 * 6 9 * 10 5 * 6 18 * 10 3 * 14 , 4 9 * 10 5 * 14 , 4
465
Challenging Glass 3
P > 17 kN.
According to expression (15) the maximum post-tensioning load follows from:
1,0.P * 0,97 1,0 * P * 0,2425 * 2400 0,9 * 20,8 * 10 3 0,9 * (19,4 12,5) * 10 6
3
6
3
d1
18 * 10 .6,0 0,9 * 10 * 6,0 18 * 10 * 6,0 0,9 * 10 6 * 6,0
P < 23,5 kN.
For the minimum post-tensioning load P = 17 kN the strut is subjected to an upward

load P.sin D = 4,1 kN. The beam is subjected to a normal force Np = P.cos D = 16,5 kN
and a bending moment Mp = P.a.sin D= 9,9 kNm.
9.4. Evaluation
Due to the strengthening the area and self weight of the beam can be reduced
considerately. For the not-strengthened beam the required section is 100 * 900 mm2. For
the beam strengthened with one strut the required section is 75 * 500 mm2 and for the
beam strengthened with two struts the section is only 75 * 300 mm2.
10. Conclusions
Strengthening beams of glass with cables and struts is very effective to increase the
transparency, to reduce the dimensions, the self weight and the footprint of the structure.
Thanks to the statically indeterminacy the safety of the structure is increased too,
provided the struts are prevented to turn over perpendicular to the main axis of the beam.
Post-tensioning the cables is effective, especially if the stiffness of the truss is small, to
reduce the bending moments and tensile stresses. For the given example the height of
the beam, strengthened with one strut, was only 5/9 of the height of the not-strengthened
beam. Nevertheless it is effective to increase the number of struts. For the given
example the height of the beam, strengthened with two struts, was only 1/3 of the height
of the not- strengthened beam. Further research is needed to test these structures in
practice.
11. Literature
[1] Belenya E., Prestressed Load-Bearing Metal structures, MiR Publishers, Moscow, 1977;
[2] Engel, Heino, Structural systems, Verlag Gert Hatje, Germany, 2th edition 1999;
[3] Hollander J.P., Optihalisatie, Constructietypen en ontwerpaspecten voor eenbeukige stalen hallen,
Bouwen met Staal, The Netherlands, 2007;
[4] Louter, Christan; Belis, Jan; Bos, Freek; Veer, Fred; Hobbelman Gerrie, Reinforced Glass Cantilever
Beams, Proceedings of the 9th International Conference of Architectural and Automotive Glass (Glass
Processing Days) Tampere, Finland, 2005;
[5] Louter, Christian; J.F van Heusden, F.A. Veer, J.N.J.A. Vambersky, H.R. de Boer, J.Versteegen, Post-
tensioned Glass Beams, TuDelft Library, The Netherlands;
466
doi:10.3233/978-1-61499-061-1-467
The Design of an All Glass Roof

to the EN1990
Ron Kruijs
Glasimpex Schiedam, Holland, r.kruijs@glasimpex.nl
Member of TC129WG8, TC250WG3, and the Dutch construction glass workgroup
Architects van Mourik designed, in cooperation with Glasimpex Holland, an all

glass roof for a cultural centre at Utrecht. The roof is sized 13x17m1, the main
beams are 13m length. Because of the low budget the beams are split in four pieces
size 3,25m1. Bearing hole connection is used. The design have to meet the EN1990
and national annex of Holland. Meaning the probability of failure and the way the
structure fails must meet the requirements of the EN1990. Using the Dutch standard
NEN2608, the EN1990 is respected. NEN2608 annex D given method, based on
Fine and Kinney, is used to estimate the number of broken plies from the structural
member that can be expected during the servility of the structure. The damaged
structure must be able to carry a given level of load.
Keywords: Glass, EN1990, NEN2608
1. General
The reliability according to EN1990 can be split into the probability (chance of failure)
and the way the structure fails. For both parts demands are set. The layout for the way a
structure may fail is defined by the basic requirement (EN1990 section 2). For a
homogeneous plate of glass it is not possible to meet all requirements. Additions must
be made. In this paper the process of design for an all glass roof to the EN1990 is
described. This roof is built in March this year.
2. Probability of failure
Consequence Class (annex B EN1990)
The probability of failure has a function in the consequence of that failure. The glass
roof will be built above the lobby of a cultural Centre. Failure would have medium
consequence for loss of human life. In consultation with the building inspector the Class
of consequence is set at CC2. Meaning chance of failure ( =3,8) 1/10.000. Now we
have to investigate what values for glass strength can be used. The EN13474 given
values for glass strength but no connection is made with Consequence Class. The Dutch
standard NEN2608 given design value for the strength of glass are based on CC2. We
will use the NEN2608.
467
Challenging Glass 3
3. The way the structure fails

Demands for the way a structure may fail are given in the basic requirement of the
EN1990. A few of these requirements are difficult for glass and need special attention.
3.1. A structure shall be designed and executed in such a way that it will not be
damaged by events such as impact and human error to an extent disproportional to the
original cause (EN1990 2.1(1)P)
When using telescopic boom lift (or so) for cleaning aid there is always the risk of
breakage of glass by human error. Using homogeneous glass there is no residual
strength. So a small local damage can lead to the complete failure of a homogeneous
glass beam. The glass has to be laminated, but how many plies do we need in this
laminated glass? The number of plies is depending of the amount of layers that can be
broken by that event.
3.2. Potential damage shall be avoided or limited by appropriate choice of the

following. Selecting a structural form and design that can survive adequately the
accidental removal of an individual member or the occurrence of acceptable local
damage. Avoiding as far as possible structural systems that can collapse without a
warning. (EN1990 2.1.(5)P)
As we know glass has a brittle fracture behavior. So again we need laminated glass or,
and way be and, a second load path.
3.3. Conclusion
We need a procedure to estimate the amount of broken plies in the laminated glass
member ( the reduction in load capacity). This structure must be able to carry a certain
level of load when this level of broken glass is presented. The NEN2608 has a
procedure based on “Fine and Kinney”. With this procedure a level of broken glass can
be estimated. This procedure is informative, the goal is to oblige the designer to take
into account the possibility of breakage of glass. This breakage of glass can occur by
accident or vandalism.
3.4. Structural calculation

There is a need for at least two different structural calculations. According to the
EN1990 those will be.
x Fundamental combinations.
x Combinations of actions for accidental design situation, with a described level
of broken glass.
This procedure is set in the NEN2608.
468
The Design of an All Glass Roof to the EN1990
4. The all glass roof.

Now we have to implement the above mentioned demands and procedures.
Figure 1: Artist impression of the foyer by van Mourik Architects (Klaas van der Molen)
Figure 2: Glass roof plan and sections
469
Challenging Glass 3
4.1. Differed glass assembly of the roof

The main beam 1060x36mm consisted of a 12mm toughened/1,52mm pvb/12mm
toughened/1.52mm pvb/12mm toughened glass (Securipoint). The length of the beam
is 13m1 and it is spliced in 4 pieces. Lateral every 1625 mm a secondary beam is placed.
This beam is 520x16mm and consisted of 8mm toughened/1,52mm/8mm toughened
glass (Securipoint). The glass roof has a plates size 1730x1625mm and consist of a
laminated 8mm toughened/ 1,14mm pvb/6mm toughened glass.
4.2. Partial load factors

CC2 YG=1,2(National annex EN1990 Holland) en YQ=1,5
4.3. Fine and Kinney procedure according toNEN2608

Fine and Kinney allows to estimate the risk (RD) caused by an event. This is based on
the probability of damage, exposure and the effect of that damage. The probability of
this risk is than related to the level of damage of that structural member.
4.4. Method used in the NEN2608 based on Fine and Kinney

The standard describes three steps to be taken.
1) Determine the attack side of the member (The attack side is the side of
damage)
2) Determine the risk of damage pro attack side using formula E1
3) Determine pro attack side the level of damage according to table E2
Formula E1 Rd = Pd x Ed x Effd
Rd = risk of damage
Ed = exposure to the risk of damage
Effd = effect of the damage
Table E2:
Risk of damage Exposure to risk of damage Effect of damage
Impossible 0,1 Very rare 0,5 First aid 1,0
Practical impossible 0,2 Few time a year 1,0 Minor injury 3,0
Possible but highly unlikely 0,5 Monthly 2,0 Severe injury 7,0
Only possible on long term 1,0 Weekly 3,0 One dead 15
Possible 6,0 Daily 6,0 Several dead 40
Unavoidable 10 Always 10 Disaster many dead 100
One side lateral damage Rd<70

Two side lateral damage 70 Rd<400
Failure of one structural member (only when all components of that member are
accessible) Rd400
470
4.5. Below we will use this procedure to estimate the level of damage we have to take in
account for the different structural members of the glass roof.
4.5.1.Roofplates
The roof is walkable for cleaning and maintaining purpose
According to table E2 we have risk of damage possible 6
exposure of risk few times a year 1
effect of damage one dead 15
Risk of damage = 6 x 1 x 15 = 90
In this case only the upper sheet of the plate can be reached so the event of breakage of
this sheet must be considered.
4.5.2 Main beam

The inner side of the glass roof could be cleaned using a telescopic boom lift.
According to table E2 we have risk of damage possible 6
exposure of risk few times a year 1
effect of damage several dead 40
Risk of damage = 6 x 1 x 40 =240
The event of breakage of two lateral plies must be considered.
Engineer judgment; it is also possible that the telescopic boom lift breaks a complete
structural member of the main beam. This must also been considered.
4.5.3 Structural system of the roof

The glass roof plates are point fix at the corners. The corners of the glass plates are
connected to the main beams. The lateral beams are champed to the mean beams. The
upper edge of the main beam is connected with silicone to the roof plate. This system
foresees the main beam to resist lateral torsional buckling in both directions.
4.5.4 Safety requirements

Different structural members have to be designed with taking in to account a different
level of broken glass. So all different members will be designed separately.
1) Main beam, the risk of breakage of two lateral or a compete member must be
taken into account. So there is a need for;
a) Second load path thought the lateral secondary beams.
b) At least 3 plies of glass in a laminated because the breakage of two lateral
plies is possible. (Actually this is twofold because the second load path is
dealing with this problem).
2) We must also consider the risk of lateral torsional buckling of the secondary
beam when using this second load path.
3) Secondary beam, this beam has a function in minimizing the buckling length of
the mean beam en foresee in second load path. This beam has to be laminated.
When one ply of the laminated member is broken it must be able to foresee in
both functions. Also when both plies are broken it must be kept in place long
enough to evacuate the area underneath it.
4) Roof plate, when walking on the roof the upper ply of the laminated could
break. In that case the plate still have to be able to carry the load. So this
laminated must consist of at least two plies.
471
Challenging Glass 3
4.5.5. Structural calculations

A number of structural calculations are necessary for each component.
a) Main beam; fundamental combination, accidental situation with two broke
plies and second load path.
b) Secondary beam : fundamental combination, accidental situation with one
broke plies and second load path.
c) Roof plate; fundamental combination and accidental situation with broke upper
ply.
5. Basic requirements
How does this structural safety system fit in the basic requirements of section 3?
5.1.1 A structure shall be designed and executed in such a way that it will not be
damaged by events such as, impact and human error to an extent disproportional to the
original cause (EN1990 2.1(4)P)
When the upper ply of the roof plate is broken by human error it is still able to carry the
load. When two plies or a complete member of the main beam is broken it is still able to
carry the load. When one ply of the secondary beam is broken it is still able to fulfill its
function.
5.1.2 Potential damage shall be avoided or limited by appropriate choice of the

following. Selecting a structural form and design that can survive adequately the
accidental removal of an individual member or the occurrence of acceptable local
damage. Avoiding as far as possible structural systems that can collapse without a
warning. (EN1990 2.1.(5)P)
When a member of the mean beam or a secondary beam is removed the roof is still able
to carry the load.
There are more basic requirements and other rules written down in the EN1990 that
must be met. The two rules which we discussed are the most demanding for a glass
structure.
6. Conclusion
It is possible to build an all glass roof according EN1990. The problem, at this moment,
is that there are almost no standards that can be used for it. Investigating the EN13474
this standard seems to have no clear relation with EN1990. There is no connection
between CC Class and design strength of glass. There also is no procedure to meet all
the basic requirements. This last issue is written down in the scope of this standard. The
only standard that can be used is the Dutch NEN2608. This standard has a clear
connection to EN1990 and a procedure to meet the basic requirements. This procedure
to meet the requirements is informative. It is the goal of the authors to let the designer
think about the structural redundancy of the system.
472
7. Comment
When heat strengthened glass was used, a complete different safety system could be
used. The choice for toughened glass was done because the bearing hole connection
needed this glass strength (Securipoint is a brand name of Saint-Gobain and has a
negative residual stress of 125MPa or more).
8. References
[1] EN1990 and Holland Nation annex.
[2] NEN2608;2011
473
doi:10.3233/978-1-61499-061-1-475
Stability Analysis of
Structural Glass Systems
Peter Lenk, Franklin Lancaster
Eckersley O’Callaghan, London, N1 9AE, UK, info@eocsd.com
Glass, as a load bearing material, has been the focus of extensive research and
development over the past few decades. Stability of slender glass elements and
glass laminates has been studied by many researchers, initially studying behaviour
on small scale samples and comparing glass behaviour with the behaviour of other,
more established, structural materials. Similarities and comparisons have been
made between the behaviour of steel and glass. In this paper the behaviour of
structural glass frames will be discussed. A short introduction with case studies
showing the use of glass frames in recently completed landmark projects is
presented. Both global and local geometric stability is discussed with specific
attention given to the boundary conditions imposed by expected initial
imperfections and restraints. Practical modelling approaches addressing the
complex behaviour of multi-ply glass laminates with viscoelastic interlayer material
is also proposed and discussed. This leads into a geometrically non-linear analysis
of an example glass frame showing predicted P-- behaviour and comparison
between effective thickness (2D plate) simplifications and full build-up (3D brick)
numerical approximations. Throughout this paper numerical discretisation and
analysis is performed using the Strand 7 software package.
Keywords: Structural glass, Imperfections, Stability analysis, Jumbo glass panels
1. Introduction
In this paper two recently completed glass structures in which the glass performs as the
fundamental load resisting system for both gravity and stability loads is discussed. In
both structures a glass frame supports the outer glass shell which in turn acts as a large
shear wall to provide global stability. As this type of structural system is novel design is
undertaken from engineering fundamentals coupled with intense research and
development and rigorous analysis techniques. In the two presented projects the vertical
glass elements are designed as both columns, in a permanent axially-loaded condition,
and beams, in a short-term bending condition. Both global deflections of the structure in
its loaded state and initial installation and fabrication imperfections can govern the
design of these slender elements. In the design stages of these structures the individual
elements susceptibility to these geometric instabilities must be carefully considered. The
completed structures are shown in Figure 1.
475
Challenging Glass 3
Figure 1: Recently completed glass structures in Shanghai and New York.
2. Glass Frame Systems

In the design of glass structures one of the underlying aims is to compose them from
simple structural elements. To ensure that the expected load path is taken engineers
should aim, wherever possible, to make the structure statically determinate. In such
cases temperature loads and substructure movements have little effect on the
distribution of internal forces and the stability of the structure.
Secondary structure
Primary structure
Figure 2: (a) Glass frame, Apple Shanghai. (b) Glass frame, Apple New York
In Figure 2(a) and (b) the glass frames are schematically outlined. The primary structure
consisting of simply supported glass beams bearing on glass columns and the secondary
structure is shown. In Fig. 2 (a) a circular glass ring beam provides only lateral restraint
to the cantilevering primary structure. The glass columns are laterally restrained by
476
Stability Analysis of Structural Glass Systems
discrete steel connections to the perimeter glass shear wall which in turn also provide
out-of-plane support.
3. Analysis Inputs
3.1. Sway imperfections

Appropriate allowance should be incorporated in the structural analysis to cover the
effects of various imperfections. Interaction between the substructure and superstructure
needs to be carefully considered especially when both components are built from glass
(Fig. 1) as substructure movements will influence force distribution. A standard
representation of sway imperfection is shown in Fig. 3 where typical limits of rotation
for steel construction are set to h/200.
Figure 3: Sway imperfections EN 1993-1-1:2005 [1]
In the case of the glass structures presented in Fig. 1 the sway imperfection resulting
from wind load needs to be carefully considered as they do not fall into any codified
building category where general assumptions can be used. Whilst in the most general
case the vertical glass members behave as glass ‘fins’, resisting wind load through
major axis bending, in self-supporting glass frames global deflections must also be
considered in the stability checks. Schematic wind load coefficient diagrams for both
structures are shown in Figure 4. These wind loads are then applied to the structures to
establish their global behavior as shown in Figure 5
Figure 4: Wind load external pressure distribution
477
Challenging Glass 3
Glass structures stabilized with perimeter glass shear walls can vary in rigidity, with
storey rotation in the range of h/800 to h/2000. This lateral rigidity is inherently coupled
to the stiffness of the connecting components between the individual glass elements as
presented in [2].
Figure 5: Global analysis model for Shanghai and New York, Lateral deflections of frame under wind load
3.2. Bow imperfections from the fabrication process

In both of the projects presented above thermally toughened glass has been used and
many of the individual elements have large aspect ratios. The extract from EN 12150-
1:2000 [3] shown in Figure 6 defines typically acceptable tolerances for glass elements
however when considering panels in excess of 10m long the imperfections would be
unacceptable; 30mm of bow imperfection would not only be clearly visible but would
also lead to impractical connection details. For this reason the imperfections of each
element must be carefully investigated and fabrication tolerances reconsidered,
especially when dealing with very slender elements with aspect ratios of around 1:18.
With the latest top and bottom convection tempering ovens and good quality control,
much higher imperfection limits, in range of 0.0005mm to 0.001 mm in 1mm, can be
achieved. Belis et al. [4] presented research on initial bow imperfections where aspect
ratio, thickness, length, glass types and lamination techniques were considered. One of
the outcomes of this study was that different glass manufactures can deliver a wide
range of glass quality and therefore an emphasis on quality control is required.
478
Figure 6: Representation of glass imperfections EN 12150-1:2000 [3]
3.3. Boundary conditions

The boundary conditions imposed on the system govern the stability and therefore must
be carefully defined in the analysis of glass structures. Typical support conditions and
expected buckling modes are shown in Figure 7
Figure 7: (a),(b),(c) & (d) Glass fin boundary conditions and expected buckling shapes
Initial bow imperfections, discussed in 3.2, are shown in Fig. 7(a). When a glass
element of this shape is installed the expectation is that the edge in contact with the
façade panels is aligned and held in place using either an adhesive bond such as silicone
(Fig. 7b) or mechanical fittings (Fig. 7c) whilst the back edge remains unrestrained.
Whilst this installation technique reduces the overall magnitude of the imperfection it
also generates a torque in the element which will specifically motivate a lateral torsional
buckling mode. When performing a stability analysis on this type of system the
expected imperfection is explicitly chosen to be of the fundamental buckling mode. It
should be noted that the stiffness and frequency of the discrete fittings shown in 7 (c)
479
Challenging Glass 3
can have a significant effect on the buckling performance of the element. Figure 7 (d)
illustrates an element as part of a glass frame structure, in this case the fin has local
lateral restraint against the façade panels however deformations occur in the global
sense. For this type of structure it is clear that the magnitude of the global deflections
needs to be accurately established to ascertain the local stability of elements.
3.4. Effect of visco-elastic interlayer

The example structures detailed in this paper make use of glass elements laminated with
a structural interlayer. These thermoplastic polymers are viscoelastic, exhibiting
dependency on load duration, with mechanical properties such as shear modulus
spanning several decades. As the shear modulus of the laminating interlayer directly
affects the behavior of the glass elements it is essential to establish appropriate values
that correspond to the loading condition. The effect of interlayer stiffness can be simply
examined by relating the shear modulus to an effective bending thickness. The effective
thickness method proposed by Wölfel [5], and subsequently appearing in ASTM
E1300-09 [6], expresses a laminated element as an equivalent monolithic element with
reduced thickness. As a single effective thickness value cannot describe both the
expected stress and deflection in the laminated section different expressions are derived
for each. As this paper is concerned with geometric instability only the deflection based
effective thickness is duplicated here.
య
‫ݐ‬௘௙௙ ൌ ට݄ଵ ଷ ൅ ݄ଶ ଷ ൅ ͳʹ߁‫ܫ‬௦ (1)
ሺ௭భ ା௭మ ሻమ ሺ௛భ ௛మ మ ା௛మ ௛భ మ ሻ

‫ܫ‬௦ ൌ (2)
௛భ ା௛మ
Where h1, h2 are the thickness of the first and second plies with centroids z1, z2 and the
shear transfer coefficient, is defined as:
ଵ
߁ൌ ಶ಺ೞ ೓ೡ (3)
ଵାగమ
ಸ೔೙೟ ሺ೥భ శ೥మ ሻೌమ
Where hv is the thickness of the interlayer with shear modulus Gint, E is the elastic
modulus of glass and a is a length dimension defined as the shortest bending distance.
As presented in Luible [7] the critical axial buckling load can be calculated assuming an
effective thickness calculated using the critical buckling length Lcr as variable a.
గమ ாூ೥ ሺ௧೐೑೑ ሻ
ܰ௖௥ ൌ (4)
௅೎ೝ మ
The effective thickness expression for two plies is then extended to three plies and
subsequently five plies of glass by substituting the effective thickness of two plies back
into the equation as described by Zenkert [8]. This approach for five plies is shown
diagrammatically in Figure 8.
480
Figure 8: Five ply laminated composite
Using this model it is then possible to examine the effect of shear modulus on the
composite action of the laminated element. Figure 9 shows how varying the interlayer
shear modulus influences the effective thickness of the element between an upper and
lower bounds; the upper bound being defined as the homogeneous prismatic section
(66mm in this case for 5 plies of 12mm glass including interlayer) and a lower bound of
independent plies with no shear interaction.
Figure 9: Effective thickness of the glass composite
In the case of axial buckling a clear dependence on critical (effective) length can also be
seen: teff(G, Lcr). For members with simple restraints the critical length, and therefore the
effective thickness, is easy to establish however for complex boundary conditions and
481
Challenging Glass 3
member interactions this can become problematic. Improved results can be gained
through an iterative procedure by first assuming a trial buckling mode with an
associated critical length, establishing the effective thickness based on this length,
solving for the actual buckling mode and then re-establishing the effective thickness
based on this refined mode.
This type of modeling procedure is valuable as when dealing with large glass members
modeling the interlayer interaction using 3d-hexahedral finite elements can become
cumbersome, especially given the resolution that is required to accurately capture the
shear behavior of the interlayers. This type of simplification becomes even more critical
when several large glass members interact with one another in structures such as the
examples detailed in this document.
As with the effective thickness approach detailed for Euler buckling a similar technique
is sought for lateral torsional buckling. Challamel and Girhammar [9] explore a
variational technique to establish the critical buckling moment of a vertically layered
composite beam in pure bending.
The approach presented here uses results from several finite element studies where the
full build up of glass and visoelastic interlayers were modeled using hexahedral
elements. The models consisted of simply supported beams of varying length with an
applied uniform moment. Both glass and interlayer materials were assumed to be linear
isotropic. Initial imperfections of L/1000 were induced on each to establish the
equilibrium paths and an estimate of the buckling load. Once the buckling loads were
established for varying elastic moduli of interlayer the results were processed to
establish an effective thickness for each case. This was achieved by comparing the
analytical solution for a prismatic homogeneous beam with the calculated estimates.
The analytical solution for lateral torsional buckling is given as:
஠ గమ ாூೢ
௖௥ǡ௅் ൌ ඥ‫ܫܧ‬௭ ሺ‫ݐ‬௘௙௙ ሻ‫ܬܩ‬ሺ‫ݐ‬௘௙௙ ሻටͳ ൅ ௅ మ ீ௃ሺ௧ (5)
௅೎ೝ ೎ೝ ೐೑೑ ሻ
where J is the torsion constant dependent on effective thickness, G is the shear modulus
of glass and Iw is the warping constant of the composite section. For the case presented
here the effect of warping stiffness on the critical buckling load is assumed to be
negligible and so the closed form solution of equation (5) reduces to:
஠
௖௥ǡ௅் ൌ ඥ‫ܫܧ‬ሺ‫ݐ‬௘௙௙ ሻ‫ܬܩ‬ሺ‫ݐ‬௘௙௙ ሻ (6)
௅೎ೝ
The effective thickness can then be established from the estimated Mcr:
ల ଷ଺ ெ೎ೝ ௅೎ೝ
‫ݐ‬௘௙௙ ൌ ට మ ቀ ቁ (7)
ீா௅ ೎ೝ గమ
Figure 10 shows how varying the interlayer shear modulus influences the effective
thickness for lateral torsional buckling. It is noted that the dependence of the effective
482
thickness on the critical buckling length is not as pronounced as with Euler buckling
and instead also depend on the width of the element subject to buckling.
Figure 10: Effective thickness of the glass composite
4. Stability Analysis
Various methods and approaches can be taken to assess the stability of elements. While
simplified techniques such as those shown in Figure 11 for steel structures are
applicable in some cases complex boundary conditions require a more rigorous
numerical approach. We briefly discuss the application and considerations of these
finite element procedures to glass structures.
Figure 11: Replacement of initial imperfections by equivalent horizontal forces EN 1993-1-1:2005 [1]
483
Challenging Glass 3
4.1. Linear buckling analysis

A linear estimate for the critical buckling load factor can be obtained from the
Eigenvalues of the stiffness matrix. Linear buckling analysis is used to predict the load
factor and buckling mode for the element subject to imposed loads boundary conditions.
This type of analysis typically provides an upper bound solution which is generally
insensitive to member imperfections. Figure 12 (a) shows the fundamental buckling
mode from a global model of the structure. In this case the linear buckling result
highlights elements that may require more detailed analysis as shown in Figure 12 (b).
As a first approximation this procedure is very powerfull and easy to use as well as
generally providing suitable imperfections for use in nonlinear solutions. Imperfections
provided by linear buckling analysis will generally motivate the fundamental and most
critial buckling mode, providing a suitably conservative result.
Figure 12: (a) fundamental buckling mode in global model, (b) fundamental buckling mode for sub model
4.2. Nonlinear buckling analysis, P- - analysis

This type of analysis accounts for both global (P- ) and local imperfections (P-) by
establishing static equilibrium under deformed geometry. Non-linear analysis is
frequently used to verify linear buckling results and to more accurately predict the
buckling load factor and buckling mode. Figure 13 presents a nonlinear analysis of a
15m long glass fin subject to positive and negative wind loads. The fin behaves as a
simply supported beam spanning the full height of the façade with lateral restraint
provided at the locations of the fittings by the main façade panels. For the purpose of
the non-linear analysis an initial imperfection of 0.001mm in 1mm (L/1000) was
assumed which equates to a bow of 15mm on the unrestrained edge. The as-installed
assumed imperfection is shown in Fig. 13 (b). Lateral restraints with an assumed
stiffness model the connection of the glass fin to the façade panels with discrete patch
fittings. As previously discussed the fin was modeled using an effective thickness
approach to establish the equilibrium paths shown in Figure 14 and the estimated
buckling load.
484
For the typical fin with wind load in suction (negative pressure) the non-linear buckling
load factor is estimated to be around =4.8 which compares to the linear estimate of
=5.6. The fundamental buckling mode for this load factor is shown in Fig.13 (c)
(a) (b) (c) (d)

Figure 13: (a) Load applied to fittings [negative wind pressure on facade], (b) Post-installation imperfection
(c) Fundamental buckling mode [-ve wind pressure], (d) Fundamental buckling mode [+ve wind pressure]
For the typical fin subject to a positive wind pressure the non-linear buckling load factor
is estimated to be around =6.2 which compares to the linear estimate of =7.0. The
fundamental buckling mode for this load factor is shown in Fig.13 (d). As expected both
nonlinear estimates predict a lower (and therefore more critical) buckling load factor
than the linear estimate.
Figure 14: Buckling load factor vs. Lateral deflection for 50mm effective thickness
485
Challenging Glass 3
4.3. Material models

An important consideration when establishing geometric instabilities in glass elements
is material behaviour. The nature of glass is such that the material fails in a brittle
manner at relatively low tensile stresses compared to the stiffness of the material –
typically glass fails in strength before the onset of geometric instability.
(a) (b)
(c) (d)
Figure 15: Diagrams showing max tensile/compressive glass stress vs. load factor.
(a) & (b) 50mm glass effective thickness, (c) & (d) 35mm glass effective thickness
Feldmann and Langosh [10] developed a calculation procedure to establish the

permissible stress of glass based on stiffness. Given the relatively weak (compared to
compressive strength) tensile strength of fully tempered glass (~120MPa) the buckling
estimate should take account not only of the geometric stability but also the ultimate
strength. Elasto-plastic models or more appropriately elastic stress failure material
models can also be incorporated. Figure 15 shows how the glass stress develops against
load factor for global and local buckling of the described 15m tall glass fin. Two glass
effective thicknesses were assumed as an upper and lower bound. From Fig. 15 (a),(b)
and (c) it is clear that strength limits will govern design of the element over geometric
stability – the onset of buckling induces additional principal stresses which will lead to
failure before the fin can geometrically no longer support load. A special case (d) where
the onset of buckling occurs before glass failure requires low composite action.
486
5. Conclusion
In this paper some of the main contributing factors affecting the stability of glass
elements has been presented. Quantifying the imperfections, both from the
manufacturing processes, installation and serviceability is critical as it is these factors
that initially induce buckling. To reduce computational expense a simplification of the
problem into 2D shell elements by means of an effective thickness method has also been
discussed. This simplification is extremely useful when modeling large glass structures
such as those presented here however establishing the parameters and calculating an
appropriate effective thickness is non-trivial and is certainly an area for further study.
The dependence of the buckling effective thickness on the mode of instability is of
particular interest: different buckling modes and buckling types lead to different
effective thicknesses. For example in the case of lateral torsional buckling we see that
the length factor (effective length) is not simply based on a single length parameter but
instead also depends on the width of the element. Further work would usefully be
undertaken with the aim to relate the effective thickness in this scenario to an aspect
ratio of the element as well as investigating the effect of combined compression and
bending instabilities. The appropriateness of such 2D simplifications is also be of great
interest. With such a dramatic simplification of the problem the applicability will be
limited to problems on a more ‘global’ scale – to highlight the areas in a structure that
require more detailed investigation. As the limits of what can feasibly be fabricated
increase the emphasis on creating more dramatic and elegant structures at ever
increasing scales containing ever more slender elements will also increase. This will in
turn require ever more detailed and thorough analysis techniques.
6. Acknowledgements
Apple Inc., Bohlin Cywinski Jackson, Seele Sedak GmbH & Co. KG, Tripyramid Inc.,
Bejing North Glass Safety Glass Co.
7. References
[1] EN 1993-1-1:2005 Design of steel structures, General rules and rules for buildings, BSI, UK, 2006.
[2] Lenk, P., Lancaster, F., Behaviour of tall glass shear walls - Theory and Applications, GPD 2011
Tampere, Finland 2011, pp.583 – 588.

[3] EN 12150-1:2000 Glass in Building Thermally toughened soda lime silicate safety glass, BSI, UK, 2000.
[4] Belis, J., Mocibob, D., Luible, A.; Vandebroak, M., On The size and shape of initial out of plane
curvatures in structural glass components, Construction and Building Materials, 2011, pp. 2700-2712.
[5] Wölfel, E., Nachgiebiger Verbund Eine Näherungslösung und deren Anwendungsmöglichkeiten,
Stahlbau 6/1987, pp. 173-180.
[6] ASTM E 1300-09, Standard Practice for Determining Load Resistance of Glass in Buildings, ASTM
Int’ll, US, 2009.
[7] Luible ,A., Stabilitat von Tragelementen aus glas, PhD Thesis, EPFL, Lausanne, Switzerland, 2004.
[8] Zenkert, D., The Handbook of Sandwich Construction, EMAS, 1997.
[9] Challamel, N., Girhammar, U. A., Lateral-torsional buckling of vertically layered composite beams with
interlayer slip under uniform moment, Engineering Structures, Elsevier Ltd, 34/2012, pp. 505-513.
[10] Feldmann, M., Langosh, K., Buckling resistance and buckling curves of pane like glass columns with
monolithic sections of heat strengthened and tempered glass, Challenging Glass 2, Delft, Netherlands,
2010, pp. 319 – 330.
487
doi:10.3233/978-1-61499-061-1-489
Numerical Simulation of Façade / Window

Glazing Fracture under Impact Loading
Christoph Müller de Vries
Professor Pfeifer und Partner, Germany, muellerdevries@pfeifer-tragwerk.de
The sensitivity to fracture is a well-known characteristic of glass. This paper

therefore deals with the fracture of a laminated safety glass panel made of single-
pane safety glass during a dynamic impact event. Potential scenarios were
examined by experimental and numerical studies. For this purpose, different glass
panel geometries were subjected to a pendulum impact test using the test setup
commonly employed for fall-through-proof glazing (50 kg twin-tyre impactor) in
Germany. Depending on the glass panel size and support system, the actual impact
event is completed in less than 30 ms. However, the dynamic response of the
glazing can continue and fracture occur at a later stage.
Keywords: impact loading, fracture, pendulum test, laminated safety glass
1. General
In Germany, adequate impact and fall-through protection of façade and window glazing
is demonstrated by pendulum impact testing using a 50 kg twin-tyre impactor. The test
requirements are defined via the classification of the type of construction in accordance
with the Technical Rules on the Use of Fall-Through-Proof Glazing (TRAV). In future,
the test requirements will be regulated by the DIN 18008 standard series.
Normally, tests are required to demonstrate impact resistance. If comparable test results
are known testing can be dispensed with for glazing that is (roughly) identical in build-
up, geometry and support system. The decision rests with the commissioned test and
inspection body. Numerical modeling of the impact using the finite element method has
been examined in several research projects. Based on the basic parameters such as the
impact velocity and stiffness analyses of the impactor, these studies generally show
good agreement between experimental and modeling results.
As glass is a brittle material, its failure can be clearly described. Due to the construction
of laminated safety glass, the structure will retain a remaining stiffness even after failure
of one or both glass plies. These remaining stiffness can be taken into account in the
numerical simulation. By evaluating experimental results from impact events with and
without glazing failure, further parameters were established for the modeling studies.
Laminated safety glazing composed of thermally toughened glass supported on two
sides was used as a basis for the experimental and modeling studies.
489
Challenging Glass 3
2. Model
2.1. Design and FE model of pendulum impactor

The pendulum impactor was modeled on the basis of the requirements of DIN EN
12600 and TRAV. The impactor properties used as model input were a total mass of
50 kg ± 0.1kg and a tire inflation pressure of 4 bar.
Figures 1a and 1b: Pendulum impactor design as per DIN EN 12600 [1] (top) and FE model.
2.2. Parameters of pendulum impactor

The FE model was calibrated against experimental results available from own tests
(Fig. 2). For this purpose, the acceleration rates of the pendulum impactor on collision
with the base of a rigid steel-reinforced concrete column were measured for different
drop heights. The acceleration rates so determined partly vary from the published values
for drop heights of 900 mm, 700 mm and 450 mm (e.g. Schneider [2]). Additional tests
were carried out at lower drop heights to also account for lower acceleration rates
attained with the flexible glazing examined. In this case, the velocity of the pendulum
impactor is another parameter that must be considered in modeling. At drop heights
above 300 mm, the velocities can be determined using the physical equation for free fall.
For drop heights below 100 mm, tumbling movements and system resistances result in a
roughly 10 % reduction in the final velocities compared with the ideal free-fall equation
‫ ݒ‬ൌ ඥʹ݄݃Ǥ
490
Numerical Simulation of Façade / Window Glazing Fracture under Impact Loading
Figure 2: Experimental and modeling results for impact on a rigid “wall”
The modulus of elasticity of the tyre model was calibrated by comparing the modeling
and experimental results. Table 1 depicts the stiffnesses of the selected models as a
function of the impactor acceleration rate. At low acceleration rates, the pendulum
impactor must be modeled with a greater stiffness than at higher acceleration rates.
491
Challenging Glass 3
Table 1: Measured and calculated impactor acceleration rates

Column 1 Column 2 Column 3 Column 4 Column 5
test result “rigid modulus of elast.
drop height h FE-calculation velocity
wall” impactor
[mm] acceleration [m/s2] [N/mm2] acceleration [m/s2] v0 [m/s]
20 44 1,3 / 1,6* 46 / 46 * 0,63 / 0,59*
(50) (79) (1,15) (77) (0,99)
100 117 1,1 / 1,3* 122 / 120 * 1,40 / 1,33*
150 151 1.0 150 1.72
200 180 1.0 177 1.98
300 218 0.9 223 2.43
450 272 0.8 270 2.97
700 346 0.8 352 3.71
900 404 0.8 402 4.20
* In the second value, the deceleration due to tumbling movements of the impactor is considered.
The calculations of the acceleration rates (column 4) are all done with the given modulus in column 3.
() For the drop height 50 mm, no measurement of the velocity is available.
2.3. Glazing and interlayer

The model of the glazing must be capable of capturing the shear distortions in the
interlayer and the shortened chords. This makes it necessary to use volume elements
(Fig. 3).
Figures 3a and b: FE model and detail
492
The properties of the glazing under short-duration impact loading take into account the
stresses that can be accommodated and the modulus of elasticity. The values for the
dynamic modulus of elasticity indicated in the literature show great variability. While
limit values of between 69,000 N/mm² und 76,000 N/mm² are possible, this broad range
of variation has, however, little influence on the modeling results. For this reason, it is
recommended to use a modulus of elasticity of 73,000 N/mm² for modeling studies.
According to studies performed by Kerkhof [3], the potential failure stresses of the
glazing under short-duration loading can be determined as a function of the pre-stress
level and the depth of the initial crack. TRAV [4] specifies a short-term strength of
170 N/mm² for toughened glass.
With the type of glass construction examined, the time of glazing failure under dynamic
loading has an influence on the further modeling procedure. Assuming a failure of the
glass ply on the tensile side of the laminated safety glass panel, tensile stresses likely to
lead to failure will only be transferred to the other glass ply after fracture of the non-
impact-side ply. For this reason, variations in the time of fracture occurrence have to be
examined in the modeling studies.
Figure 4: Fracture of both plies of the laminated safety glass panel

The interlayer provides virtually the complete shear coupling between the two glass
plies. Only after glass fracture will the strain rates in the interlayer increase abruptly. A
study of the stress-strain relations at different strain rates was performed by Bennison et
al. [5] for PVB and SGP interlayers. According to this study, the modulus of elasticity
may rise dramatically shortly after glass fracture to values as high as 280 N/mm² for
PVB interlayers and even to up to 500 N/mm² for SGP interlayers.
3. Stiffnesses
The deformation and stress utilization occurring during the impact event result in
excessive loading of the glazing. Due to the composite effect, only one ply of the
laminated safety glass will be affected initially so that only this ply will be destroyed.
As a result of the reduced stiffness, the impact force of the pendulum impactor will
decrease. Depending on the dynamics of the pendulum impactor, further loads will be
subsequently imposed on the intact glass ply.
493
Challenging Glass 3
In order to determine the remaining stiffnesses of the destroyed glass ply, different tests
were performed:
a) Static loading tests:

Depending on the glass thickness, the tests yielded the following remaining
moduli of elasticity for the destroyed tensile-side glass ply:
4 mm glass ply: 5 N/mm²
10 mm glass ply: 32 N/mm²
Accordingly, the fractured glass ply does no longer make any significant
contribution to the stiffness of the system in the case of the small 1100 mm x
360 mm test specimens.
Figure 5: Repeated static loading tests on laminated safety glass

with one defective glass ply on the tensile side
b) Dynamic impact tests:

Comparison of deformations of a single glass pane, a laminated safety glass
panel with destroyed outer glass ply and an intact laminated safety glass panel
(Fig.6). The impact on a small 1200 mm x 200 mm test specimen consisting of
2x10mm single-layer safety glass likewise resulted in a virtually complete
stiffness loss of the broken glass ply. The measured difference between the
single pane and the single-side fractured laminated safety glass panel can be
explained by the partly defective glass panel, which causes a 3 mm
deformation without load.
494
Figure 6: Impact on test specimen 1200 x 200 mm
c) Vibration decay tests of the single side-fractured glazing:

The evaluation of the direction-dependent periods of oscillation for the uniform
vibration decay of the glass panel shows that a significantly higher modulus of
elasticity of 19,000 N/mm² is required for the numerical simulation in order to
bring the experimentally determined frequencies (Fig. 7) and the calculated
frequencies into better agreement. The stiffness increase of the large
2850 mm (*) x 1250 mm test specimens (* = span direction) with a glass
thickness of 2x10 mm is attributable to the influence of glass panel zones with
minor curvature and major interlocking of the fragments.
This high stiffness is of particular importance in the case of a renewed impact
on the single-side broken glazing and an early fracture of the non-impact-side
glass ply.
Figure 7: Vibration decay behaviour of glass panel sized 2850 x 1250 mm
495
Challenging Glass 3
4. Fracture process
As part of the studies conducted by Müller de Vries [6], various acceleration,
deformation and strain measurements were performed on the test specimens (Fig.8) in
order to provide information on the fracture event. On fracture of a single-layer
laminated safety glass panel, pre-stresses in the glass will result in the formation of
small fragments accompanied by a change in volume due to the voids formed by the
cracks and the locking-together of the fragments, resulting in permanent deformation. If
fracture occurs at an early stage of the impact event, the resulting curvature may result
in an independent increase in deformation due to the eccentric volume increase.
Figure 8: Graphical evaluation of measured quantities of two test specimens differing in the time of fracture
occurrence (2850 x1250 mm; 2 x 10 mm; drop height 900 mm)
For the numerical simulation and a comparison of the experimental and modeling
results, the curves of individual measured / calculated values must be superimposed to
be able to capture and assess model inaccuracies. The diagram below (Fig.9) depicts
such a superposition by the example of the measured and calculated glass panel
velocities during the impact event.
496
Figure 9: Glass panel velocities and deformations (test specimen no. 1)
As can be seen from the graph, the minor temporal variations between the measured and
calculated glass panel velocities add up to a significant variation at the time of the
fracture event. To ensure a good model fit for the glass ply velocities, the time of
fracture occurrence had to shifted by + 7 ms in the model calculation (Fig.10). Only in
this way was it possible to achieve good agreement between the superposed measured
glass panel velocities and the further post-fracture calculation results for the non-
impact-side glass ply.
Figure 10: Different modeling results compared to the test (test specimen no. 1)
497
Challenging Glass 3
Based on the available information on the remaining residual stiffness of the glass ply
after fracture, the stresses developing in the glass ply were calculated using different
stiffness assumptions (Fig.11). While the stresses in the local impact center can be
better represented by the lower theoretical modulus of elasticity of 32 N/mm² for the
broken glass ply, the higher modulus of elasticity of 19,000 N/mm² (depending on the
glass panel size and support system to take into account interlocking glass fragments) is
needed for a more accurate calculation of the overall stress profile of the impact-side
intact glass ply. The use of this high modulus of elasticity of the broken glass ply is
necessary because of the early time of fracture occurrence.
Figure 11: Comparison of stresses in the remaining intact glazing (test specimen no. 2)
5. Numerical simulation
For the numerical simulation, volume elements are used for both the glass plies and the
interlayer. The parameters of the materials used are described in chapter 2. The
interlayer itself is modeled with different moduli of elasticity for each direction so as to
simulate its function as spacer between the two glass plies without, however,
introducing its own bending stiffness into the calculation. The modulus of elasticity in
shear of the PVB interlayer is defined at 6 N/mm².
The first calculation is run without fracture of the glass in order to determine the time-
dependent distribution of the stresses in the glass plies and define the time of the
possible fracture. In a second calculation with the same parameters, the calculation is
stopped at this time step to change the material properties of the now fractured glass ply.
After that, the calculation continues.
This approach, i.e. the interruption of the calculation, also allows the simulation of an
impact on the single-side fractured laminated safety glass panel. In this case, the pre-
498
deformation has to be considered. For this purpose, the eccentric volume increase can
be simulated by a slow temperature rise and after the pre-deformation is reached, the
gravity in the model is activated to start the drop of the pendulum. An example of the
time-dependent simulation is shown in Fig. 12.
Figure 12: Example for the calculation history of the impact on a single side fractured laminated safety glass
6. Influencing factors
Because of the multitude of factors having a potential influence on the modeling result,
parameter studies are required in order to be able to make a sufficiently accurate
statement on the stress utilization of the remaining intact glazing after fracture of the
glass ply on the non-impact side. During the experiments, failure of both glass plies
normally resulted in the complete failure of the construction, so that this scenario was
excluded from the further study. The following criteria must be taken into account:
x Temporal variations may lead to different modeling results.

x The definition of the time of fracture occurrence must also consider higher
limit stresses of the glass.
x In the case of early fracture occurrence, the expansion of the broken glass ply
has to be taken into account.
x Varying moment loads during the impact event must be replicated (applies to
impact on partly defective glazing).
x The resulting interlocking of fragments leads to a permanent deformation and
loading of the glass panel.
x The resulting curvatures are a function of the load.
x These curvatures must be taken into account in the numerical simulation in the
case of a renewed impact.
499
Challenging Glass 3
x Remaining load carrying capacities in the case of fracture of both glass plies
must be taken into account.
x Deteriorations in the composite efficiency (delamination) must be separately
calculated.
7. Conclusion and outlook

The experimental and modeling studies showed that the use of a single model is rarely
valid to replicate the impact event. The fracture mechanisms in the glass and the
resulting changes in the system could be replicated by finite element simulations. If the
glass panel is in a state of large bending moments, only minor residual post-fracture
stiffnesses should be assigned in the case of fracture of the glass ply on the tensile side.
The geometry of the glass panel has a major influence on whether unloaded areas can
make a major contribution to the residual stiffness.
The impactor properties determined are well suited to modeling normal impact events
without glass fracture. In carrying out own modeling studies, each user should perform
a comprehensive calibration of his model against published test results in order to be
able to assess the multitude of influencing factors in any situation. With the limited
calibration with experimental results made so far, modeling with the aim of overloading
and hence, inducing glass fracture is likewise possible. Due to the high strength of the
single-layer laminated safety glass type examined, fracture in linear supported glass
panels is, however, rare. The transfer of this modeling approach to other glass types
such as float glass or other types of support systems such as point-supported glazing
units should therefore be the aim, but will require further studies in this specific area.
8. Acknowledgements
The dissertation on this subject was written during my activities at the Technical
University of Darmstadt, Structural Engineering Department, under the direction of
Prof. Dr.-Ing. J.-D. Wörner. I would like to take this opportunity to express my thanks
and appreciation for the generous support received.
9. References
[1] Schneider, J., Festigkeit und Bemessung punktgelagerter Gläser, Dissertation TU Darmstadt 2001.
[2] DIN EN 12600: Glas im Bauwesen – Pendelschlagversuch – Verfahren für die Stoßprüfung und
Klassifizierung von Flachglas, Ausgabe April 2003.
[3] Kerkhoff, F., Festigkeit von Glas – Zur Abhängigkeit von Belastungsdauer und -verlauf; Glastechnische
Berichte 54 Nr. 8.
[4] TRAV: Technische Regeln für die Verwendung von absturzsichernden Verglasungen, DIBt Berlin 2003.
[5] Bennison, S.J.; Sloan, J.G.; Kristunas, D.F.; Buehler, P.J.; Amos, T.; Smith, C.A., Laminated Glass for
Blast Mitigation: Role of Interlayer Properties, Glass Processing Days 2005, Tampere Finland.
[6] Müller de Vries, Chr., Ein Beitrag zur Berechnung von Pendelschlagversuchen unter Berücksichtigung
der Reststeifigkeit von gebrochenem VSG aus ESG, Dissertation TU Darmstadt 2011.
500
doi:10.3233/978-1-61499-061-1-501
Glazing Under Long Term Loads Induced

by Water
Arthur Pistora, Barbara Siebert
Ingenieurbüro Dr. Siebert, München, Germany, api@ing-siebert.de, bsi@ing-siebert.de,
www.ing-siebert.de
In modern architecture, swimming pools are often equipped with underwater

windows. The swimmer gets the possibility to have a look outside or inside the
house, while he’s enjoying his time in the water. Additionally the light can enter the
building through the water. For some projects our office, Ingenieurbüro Dr. Siebert,
made the engineering design and the static calculation. This paper shows by
presenting these projects, how to add value to a pool through clever use of glass
and what has to be considered from a static and an engineering point of view. For
the two projects in Germany a special building permit from the building authorities
was needed.
Keywords: Design of glass, Water load, TRLV, DIN 18008
1. General
In the last years glass has become one of the most popular building materials in modern
architecture. There is hardly a modern building, where glass is not used. On the façades
of many cities glass is the defining element, but the possibility of use is not limited to
façades, stairs or ceilings. An increasingly occurring area of application for glass is in
interaction with water, for example in fountains, built-in aquariums or in pools.
2. First Project, Office building in southwestern Germany

2.1. General conditions

Three accessible glass panes were built inside a water basin as a part of a new office
building in southwestern Germany. The basin is placed in front of the office building
close to the cafeteria and accessible to the public. The glazing is located above
underground parking lots and is designed as an insulating glass unit (IGU).
Because of the underground parking lot the glazing has to fulfill fire protection for 30
minutes. Above the glazing there is a water head of max. 150 mm. Resistance against
impact caused by a hard or soft body as well as the post breakage behavior had to be
proved.
2.2. Measurement and glazing details

The size of the each glazing unit is 1070 mm x 1870 mm. The glazing is linear
supported along its edges and the bearing has a width of 35 mm.
501
Challenging Glass 3
The glazing has a step of 45 mm, so the layers have individual bearings. Moreover, in
the fire protection layer there are no loads from the top layer induced. Because the fire
protection gel and the edge seal of the IGU would not be able to carry the load of the
upper bearing layer.
Glass set-up of the double glazing units (from top to bottom):
Load bearing layer:
x Laminated safety glass (LSG) consisting of

o 12 mm fully toughened glass (FTG) as wear layer
o 1.52 mm polyvinyl butyral – interlayer (PVB)
o 12 mm heat strengthened glass (HSG)
o 1.52 mm polyvinyl butyral interlayer (PVB)
o 12 mm heat strengthened glass (HSG)
x 16 mm spacing
Fire protection layer:
x 5 mm fully toughened glass (FTG)

x 15 mm fire protection gel
x Laminated safety glass (LSG) consisting of
o 4 mm float glass
o 0.76 mm polyvinyl butyral interlayer (PVB)
o 4 mm float glass
Figure 1: Section true the glazing
502
Glazing Under Long Term Loads Induced by Water
2.3. Design and static calculation

The following loads were considered for the structural design of the glazing.
Table 2: Loads for the structural design according to DIN 1055 [5]
Load description Load
Dead load of the top load bearing glazing 0.9 kN/m²
Dead load of the internal fire protection 0.51 kN/m²
glazing
Continuous load from the water layer 1.5 kN/m²
Traffic load 4.0 kN/m² respectively
2.0 kN on an area of 50 x 50 mm²
In the calculation the new recommendation for accessible glazing by the German
Institute for Building Technology from 2009 [4] was considered.
Figure 2: Installation of the glazing in 2011 (by Arnold Brandschutzglas)

2.4. Other verifications

The fire resistance (F30) of the glazing has been confirmed by an expert assessment
from the Institute for Fire Testing (MFPA Leipzig GmbH) in Leipzig. At this institute
the fire protection layer was tested in a fire test in the past. The impact resistance and
the residual resistance after breakage has been confirmed based on former experimental
testing.
3. Second Project, Swimming pool with glass wall
3.1. General conditions

At a project in the south of Germany a glass pane was planned as a side wall for a
swimming pool (figure 3 and 4).
503
Challenging Glass 3
Figure 3: Picture of the building and glazing of the swimming pool (by Design Associates)
3.2. Measurement and glazing details

The glazing has an inclination of 5° to the vertical toward the water and the size of 3470
mm x 915 mm. It is linear supported along the bottom and on the side edges. The
bearings at the bottom have a width of 47 mm and on the sides 50 mm. The top edge is
protected by a steel U-Profile.
Glass set-up of the laminated safety glass (LSG) (from inner to outside):

x 1.52 mm polyvinyl butyral – interlayer (PVB)
x 12 mm fully toughened glass (FTG) with step (see figure [5])
Figure 5: Glazing with dimension of the step pane
504
Figure 4: Swimming pool with glazing in the back - finished and on the building site (by Design Associates)
3.3. Design and static calculation

The following loads were considered for the structural design.
Table 2: Loads for the structural design according to DIN 1055 [5]
Load description Load
Dead load of the glazing 1.15 kN/m²
Continuous load from the water layer 0 to 9.15 kN/m²
Wind load on the outside -0.33 kN/m²
In the static calculation we tested if the glazing can bear the load in the case that a glass
panes breaks. The outer pane has no bearing function; it is only a wear layer. Because
the glass edges are protected, it can be assumed, that in the worst case a minimum of
two glass layers are intact.
For the case of glass breakage the allowable stresses (see table 5) had been increased by
the factor of 1.5.
4. Third Project, Glazing in swimming pool

At another project, for which we don’t have the permission to publish details, the
glazings were built into the floor and the surrounding walls of a swimming pool.
The glazing with the greatest dimension was built in the pool floor and has the
dimension of 1.5 m x 3.23 m and has a water head of max. 2.2 m.
5. Actual Building Administration – Regulation in Germany

At the moment glazings with water load must fulfill the regulation TRLV [1] and
additional a special building approval is needed.
The static design of the glazing must be done according to the “Technical rules for the
use of line supported glazing” [1]. Thereby no shear transmission of the PVB film can
be considered. Additional a reduced allowable tensile stress must be considered,
505
Challenging Glass 3
because the dead load of the water is a long term load. The allowable tensile stress for
the glass is given by the “Tips for the achievement of a special building approval” [6] of
the building administration of Bavaria (Germany).
Table 3: Allowable stress for glazing with water load according to [6]
Glass Allowable stress adm [N/mm2]
Fully toughened glass (FTG) 29
Heat strengthened glass (HSG) 13
Float glass (FG) 5
6. Static calculation according to the new building regulation DIN 18008
6.1. Overview
In the next time the new DIN 18008 [1] will most likely be established in Germany as
the design code for glazing. According to part 2 of the DIN 18008 additional regulation
must be considered for glazing with permanent load from water. These additional
requirements are not yet published so that a special building approval will be needed
6.2. Design criteria

In the new DIN 18008-1 [1] the design equation has the form:
‫ܧ‬ௗ ൑ ܴௗ
The design value of resistance Rd (max. tensile stress) of float glass (fk = 45 N/mm²) is
calculated according to this equation:
݇௠௢ௗ ή ݇௖ ή ݂௞
ܴௗ ൌ
ߛெ
The design value of resistance Rd of thermal treated glass (FTG: fk = 120 N/mm², HSG:
fk = 70 N/mm²) is calculated according to:
݇௖ ή ݂௞
ܴௗ ൌ
ߛெ
With:
fk = characteristic material strength
kmod = 0,25 (factor considering time dependence - long time load from water)
kc = 1,0
M = 1,8 (material safety factor for annealed glass)
M = 1,5 (material safety factor for thermal treated glass)
In case laminated safety glass (LSG) is used, Rd can be increased by 10 %.
506
For glazings consisting of float glass with a free edge and tensile stress along edges, the
characteristic material strength fk has to be reduced to 80%.
Table 4: Design value of resistance Rd for glazing with water load according to DIN 18008-1 [1]
Glass Design value of resistance Rd [N/mm2]
Monolithic Laminated safety glass (LSG)
Fully toughened glass (FTG) 80.00 88.00
Heat strengthened glass (HSG) 46.67 51.33
Float glass (FG), 4-side supported 6.25 6.88
Float glass (FG), 2 and 3-side supported with 5,0 5,5
free edges
7. Conclusion
The shown projects give an overview for the engineering design of glazings in pools or
basins. For both examples in Germany a special permit from the building authorities
was needed.
We think that in the future more glazings will be installed in swimming pools or
aquariums and that the dimension will increase. The new German Standard DIN 18008
gives the possibility to consider better the behavior of annealed and thermally treated
glass under long term loads induced by water. But in Germany further on a special
permit from the building authorities will be needed.
8. Acknowledgements
The authors would like to thank the following participants in these projects:
First project: Architect, hermannarchitekten, Porschestraße 2, D-74321 Bietigheim-
Bissingen; Supplier of the glazing: Arnold Brandschutzglas Vertriebs GmbH & Co. KG
Kastanienstraße 10, D-09350 Lichtenstein
Second project: Architect, Design Associates GmbH, Winterstraße 4, 81543 München
9. References
[1] TRLV - Technischen Regeln für die Verwendung von linienförmig gelagerten Verglasungen (Technical
rules for the use of line supported glazing), Deutsches Institut für Bautechnik e.V., Germany, Version
08/2006
[2] DIN 18008-1:2010-12 Glas im Bauwesen - Bemessungs- und Konstruktionsregeln - Teil 1: Begriffe und
allgemeine Grundlagen, (Glass in Building - Design and construction rules - Part 1: Terms and general
bases), Deutsches Institut für Normung e.V., Germany
[3] DIN 18008-2:2010-12 Glas im Bauwesen - Bemessungs- und Konstruktionsregeln - Teil 2:
Linienförmig gelagerte Verglasungen (Glass in Building - Design and construction rules - Part 2:
Linearly supported glazings), Deutsches Institut für Normung e.V., Germany
[4] Anforderungen an begehbare Verglasungen; Empfehlungen für das Zustimmungsverfahren, Deutsches
Institut für Bautechnik e.V., Germany, November 2009
[5] DIN 1055: Einwirkungen auf Tragwerke, German standard, (Actions on structures), Deutsches Institut
für Normung e.V., Germany
[6] Merkblatt ZiE–Nr. 3A, Hinweise zur Erlangung einer Zustimmung im Einzelfall (ZiE), (Tips for the
achievement of a special building approval) gemäß Art. 18 Abs. 1 und Art. 19 Abs. 1 Bayerische
Bauordnung (BayBO), Oberste Baubehörde im Bayerischen Staatsministerium des Innern,
www.innenministerium.bayern.de/bauen/baurecht/bautechnik/, Germany, Version January 2012
507
doi:10.3233/978-1-61499-061-1-509
Energetic Approach of Elastic Strain

Energy of Thermally Tempered Glass
Stefan Reich, Bernhard Weller
Technische Universität Dresden, Institute of Building Construction, Germany,
stefan.reich@tu-dresden.de
Nora Dietrich, Stephan Pfefferkorn
University of Applied Sciences Dresden, Germany
At the thermal tempering process elastic energy is stored in glass panes. At the
breakage of thermally tempered glass this energy is partly or fully released. The
broken glazing measurably expands its size with noise and kinetic movement. The
determination of elastic strain energy bases on the theory of glass fracture and its
further extension. The distinction into elastic and kinetic parts of the stored energy
was shown on the molecular dimension by Kerkhof in 1970. The authors of this
contribution developed an approach to transfer the molecular level into macro
dimensions. The contribution discusses the approach. A comprehensive test series
to check the developed approach is in progress.
Keywords: Glass, Residual Stresses, Energy, Energetic Approach
1. General
At the production of thermally tempered glass a certain energy amount is stored in the
fully tempered or heat strengthened glass panes. After failure of the glazing this stored
energy is released and lead to a volume expansion of a supported glass pane, free of
constraint forces. This volume expansion effect is generally known [1], [2] but not in
detail. In the event of failure several energies are released to make it difficult to
determine the elastic strain energy. To date no approach in the macroscopic dimension
exists which allows the determination of volume expansion of thermally tempered glass
fragments.
This contribution bases on [3] and discusses the principle of fracture mechanics to
determine the volume expansion or, under constraint, to determine the in-plane axial
forces at the supports. The approaches are based on the Griffith theory of brittle fracture.
This theory is extended by calculating the ratio between kinetic and surface energy on a
molecular level and transforming this ratio into a macroscopic level.
Experimental testing of a single pane size delivered results to be compared to the

theoretical approach in near future.
509
Challenging Glass 3
2. Energetic approach of brittle materials
2.1. General
Despite glass possesses a linear elastic behavior the fracture of glass is a complex
process. Different theories were developed to describe the interaction, the development
and the propagation of microscopic and macroscopic flaws of cracks, the mechanism of
dislocations and the geometry of the material. [4] One of such theories involved the
energy concepts of Griffith [5], later on developed by Irwin [6]. According to his
approach the potential energy Upot of a system is the sum of the inner elastic strain
energy U0 (including the outer strain applied by a force displacement) and the surface
energy UJ formed by the propagation of a crack. Dependent on the crack size the
potential energy is used less or more.
Later Mott [7] further developed the Griffith energy balance with an additional
constituent UK to integrate the influence of the kinetic energy in (1).
Upot = U0 - U - U -UK (1)
with
Upot the total potential energy of the system
U0 the elastic energy of the uncracked pane
UE the decrease of the elastic energy caused by introducing the crack in the
plate
UJ the increase of the elastic-surface energy caused by the formation of the
crack surfaces
UK the increase of the kinetic energy caused by a surplus of energy/fracture
instability
2.2. The elastic strain energy

A load applied on a linear elastic body proportionally deforms this body to the load. The
total elastic strain energy of a body UE is obtained by integrating the strain energy
density (x,y,z) over the body volume. This is stored in the body as an equal amount of
energy, the elastic strain energy (1st law of thermodynamics). The equations (2) and (3)
are valid for materials under Hooke’s law, see also Figure 1 and Figure 2 [1].
ͳ (2)
x,y,z =
ʹ
ͳ (3)
U = න ሺx,y,zሻdxdydz = ෍ ij V
ʹ ij
where
Fx,y,z strain energy density
510
Energetic Approach of Elastic Strain Energy of Thermally Tempered Glass
Figure 1: Work is the load-displacement function Figure 2: Strain energy density is the stress-strain
integrated over the displacement [1] function integrated over the strain [1]
2.3. The surface energy

Breaking a body into two or more parts requires the degrading of (ionic, atomic or
molecular) bonds between the material components. Therefore, the force which is
creating the bond has to perform a negative material specific work. This work is the
surface energy necessary to create new surfaces.
A crack is forming generally two surfaces, so that the surface energy UJ for one crack is
the product of the double specific surface energy Js, the thickness of the glass t and the
crack length x (4).
U = 2Js xt (4)
where
Js specific surface energy
t glass thickness
x crack length
The material property Js is the surface energy required per unit of crack area. Several
authors let the material property be constant, but suggest different values. Table 1
summarizes experimental results (Js = 1.7 to 11 Nm/m²) that significantly differ to the
theoretical values (Js = 0.3 Nm/m²).
511
Challenging Glass 3
Table 1: Specific surface energy critical fracture energy of soda-lime glass:
Reference Js [Nm/m²] Method and Environment
Acloque, 1975 [8] 0.3 not specified, theoretical obtained

quoted from Nielsen 2009 [9]
Petzold, 1990 [10] 0.3 not specified
Kerkhof, 1970 [6] 0.3 not specified, theoretical obtained
Shand, 1961 [11] 1.70 analytical technique
Gulati, 1997 [12] 1.75 not specified
Roesler, 1956 [13] 1.8 to 10 analytical technique
quoted from Wiederhorn, 1969 [14]
Bos, 2009 7 2 to 4 not specified
Shutov, et al., 1998 [15] 2.1 not specified
Nakayama, 1965 [16] 3.4 to 5.2 work of fracture method
Mecholsky et al. 1974 [17] 3.5 air, 22° C, 40% RH
Clif, 1957 [18] 3.7 to 4.3 cone crack, air (rel. hum. 20 %; temp. 20° C)
quoted from Linger, 1968 [19]
Wiederhorn 1969 [14] 3.91 r0.12 double-cantilever cleavage technique, dry

3.82 r0.10 gaseous nitrogen N2(gas), 27° C, <1%RH
Berdennikov, 1988 [20] 4.06 in vacuum, analytical technique

quoted from Wiederhorn, 1969 [14]
Proctor et al. 1967 [21] 5 not specified
Davidge, 1968 [22] 5 to 7 work of fracture method
6 to 8 analytical technique
8 to 11 compliance method
2.4. The kinetic energy

Until today the term of the kinetic energy is often ignored and a quasi-static failure
assumed. Kerkhof tried [6] to estimate the maximum velocity or kinetic energy
respectively of the molecular constituents by equating the kinetic energy with the
surface energy of a small, perfectly homogeneous, cubical volume element (5). The
material has no notches or dislocations (Figure 3). The constituents (ions or molecular
components) are situated in a distance wo from each other perpendicular to the
(imprinted) tensile stress to simplify the approach.
Vf = molecular strength
Figure 3: Schematic of the transformation of the elastic strain energy in kinetic energy and surface energy [6]
512
1 2
Uk,mic = v w ൌ ʹߛ௦ ൌ ܷఊǡ௠௜௖ (5)
2 f 0
where
U density = 2500 kg/m³ [23]
w0 atom equilibrium spacing = 0.2 nm [1], [6]
Js specific surface energy = 0.3 Nm/m² [6], [10]
With the use of the values mentioned above, Kerkhof calculated the theoretical
maximum velocity vf to 1550 m/s [6]. A comparison with experimental obtained
velocities listed in Table 2 indicates a good accordance.
The assumption of the specific surface energy Js = 0.3 Nm/m² strongly differs to the
experimental values listed in Table 1. Therefore all constitutes of (1) have be taken into
account that means elastic strain energy, the surface energy and the kinetic energy.
According to Kerkhof [6] the calculation of the kinetic energy is very complicated,
possess’ large uncertainties and received much critique. Factors like temperature, water
vapor in the atmosphere, a strict demarcation between static and dynamic stress
distribution, the probable changing amount of surface energy during failure and the
formation of secondary breaks play an important role [6].
Table 2: Velocity of crack propagation by high speed observations
Literature Velocity vf [m/s] Conditions

Takahasi [24] residual stresses on surface:
1500 172 N/mm²
1494 129 N/mm²
1486 123 N/mm²
Acloque, 1962 [25] quoted in Nielsen, 2009 [26] 1460 n/a
Nielsen 2009 [9] 1466 n/a

Acloque, 1963 [27] quoted in Nielsen, 2009 [9] 1500
Kerkhof, 1970 [6] 1520 median, testing method
based on ultrasound
Chaudri and Liangyi, 1986 [28] 1500 to 1800 tensile regions
from 900 decreasing to compressive regions
200 to 300
513
Challenging Glass 3
3. Transformation of molecular level to macroscopic dimensions

The characteristic size of small, more or less dice-shaped fragments of thermally
tempered glass is considered by the kinetic energy. Otherwise, with exclusion of this
energy form, the theoretically obtained patterns are much denser than in reality and
experiment [3].
Because of the uncertainties to value the kinetic energy, but in the need to include it into
the energy balance, the authors suggest transferring the molecular level to macroscopic
dimensions.
In this suggestion in (6) the surface energy Js is set in proportion to the kinetic energy
from (5) to determine a ratio rUk. According to experimental data ([11] to [21]) a median
surface energy Js = 3 Nm/m² (see Table 1) is chosen for calculation. This Js is about ten
times of the theoretically determined surface energy Js = 0.3 Nm/m² [6]. The terminal
velocity of glass vf,e= 1466 m/s was experimentally determined in [9]. The ratio
rUk = 0.0895 is the result of the defined values above.
1 2 (6)
୩ǡ୫୧ୡ 2 vf,e w0
୙୩ = ൌ
ஓǡ୫୧ୡ ʹɀୱǡୣ
୙୩ = ͲǤͲͺͻͷ
where
U density = 2500 kg/m³ [23]
w0 atom equilibrium spacing = 0.2 nm
vf,e terminal velocity = 1466 m/s (experimental value)
Js,e specific surface energy = 3.0 Nm/m² (median experimental value)
4. Energy equilibrium of thermally tempered glass fragments

4.1. Assumptions
Several authors [12], [29], [30] define energy assumption to predict the fragment size of
thermally strengthened glass that the authors adopt:
x only tensile strain energy is used for the process of thermal release
x no load applied on the glass
x fracture surfaces are smooth and normal to the surface of the plate
x all fragments are identical in shape and size (see Figure 4 for an idealized
fragment)
x ignoring of sonic effects, heat as the kinetic energy is used up in the kinetic
energy of moving fragments, noise and heat
x ignoring of stress waves, which reflect from the edges and may interact with
the approaching crack front [31]
x Hook’s Law is generally valid as well as the Bernoulli hypothesis
514
According to the assumptions of Warren [30], all of the tensile strain energy of the
fragmented samples is used up in the process. This assumption is realistic at fully
tempered glass but not a heat strengthened glass whose fragments obtain residual
stresses. Furthermore the energy approach from eq. (1) and the substitution of the
kinetic energy according to eq. (5) is extended by deformation work WE (see eq. (7)).
Therefore the release of tensile strain energy produces surface and kinetic energy, but
deformation work as well.
UE =U +UK +WE =U +U r+WE (7)
where
WE deformation work
In thermally tempered glass residual stresses are introduced by temperature differences

between the surface and the inner part of a glass pane due to rapid cooling. The inner
part of the ply cools down later than the surface. The shrinking of the inner parts is
prevented by the solidified surface. As a result elastic strain energy is stored in the pane.
If this energy is released through failure, the glass pane returns to its initial shape.
Linear elastic fragments deform proportionally to a specific ratio of elastic strain energy.
Therefore the material contracts in the tensile stress area. At the surface the reverse
effect exists, the fragment expands to stay in equilibrium with the center of the pane
(Figure 4).
failure
Figure 4: Fragment deformation and stress distribution before and after release of elastic strain energy [3]
The distortion of the fragments after failure of thermally tempered glass is observable
by the expansion of the whole glass pane. The thickness of the pane may be not
influenced according to this approach.
4.2. The elastic strain energy for one fragment

Usually a second order function describes the normal stress distribution through the
thickness of a thermally treated glass pane. The normal stresses Vxx and Vzz appear in
two perpendicular directions in planes parallel to the surface.
515
Challenging Glass 3
Figure 5: Schematic of an idealized remnant fragment shape used in the energy analysis [3]
Shear stresses may occur normal to the surface of small fragments. These may be of
relevant size near edges, corners and holes, but diminish at a distance t from the source
location [32]. In a plate of infinite size, they will vanish altogether. In the approach the
shear stresses will be neglected to keep it simple (not only because they are territorially
limited but also because their increasing influence at the edges and corners is partially
compensated by a decreasing influence of the normal stresses) [1]. Furthermore Vxx and
Vzz are equal. This leads to Eq. (8).
z2 (8)
xx ሺzሻ=yy ሺzሻ= t (1-12 )
t
where
t thickness of the glass plate
Vt tensile stress at pane center
z z = 0 pane center
Vxx, Vyy normal stresses parallel to the surface
-2Vt compressive stress at the surface of the plate
The stored elastic strain energy in the tensile strain area of the idealized fragment is
obtained in Eq. (10) by integrating the strain energy density per unit area (see eq. (3))
[1]. The factor z+ indicates the frontier between the pressure and tension zone in the
glass pane and is obtained by Eq. (9). Figure 5 shows the idealized fragment.
2 (9)
z 2 z+
0 = t · ൬1-12· ቀ ቁ ൰ 1 = 12· ቆ ቇ z+ = tൗ
d t 2ξ3
t
( )
2ξ3
x (10)
(1-Q) 2 24ሺ1-Qሻx2 t2t
U = න U = න න න xx dxdydz =
dV 0 0 E E·45ξ3
t
-( )
2ξ3
where
E Young’s Modulus
Q Poisson’s Ratio
x edge length of squared fragment
516
4.3. The surface energy for one fragment

One idealized fragment as cube is the result of four cracks (see Figure 6) and the
assumption that the fragment thickness is equal the pane thickness. The surface energy
UJ in Eq. (11) has divided by two as each crack always generates two surfaces. The
necessary surface energy to create one fragment is therefore:
1 (11)
U = 4· ൫2xt s ൯ = 4xt s
2
Figure 6: Schematic of a volume crack [1] and generation of one fragment [3]
4.4. The kinetic energy

The kinetic energy in macroscopic dimensions is:
UK = U rUk (12)
where
r = 0.0895 (see eq. (6))
517
Challenging Glass 3
5. Translation of the energy approach to a force approach
5.1. Ratio between elastic strain energy and deformation work

A validation this energy approach requires an experimental comparability. The transfer
of the energy approach translated in a force approach allows an easier experimental
measurement of existing constraint forces. The experiments were done at 19 mm FT
glazing supported under constraint. The failure and the release of elastic strain energy
leads to in-plane force to be measured with load cells.
The use of the force approach firstly requires the determination of the ratio rWE of UK to
UE in Eq. (13) and the values mentioned below.
ܷ௄ ൅ ܷఊ 4xt s ‫ݎ‬௎௞ ൅ 4xt (13)

‫ݎ‬ௐா ൌ ൌ s
ܷா 24ሺ1-Qሻx2 t2t
E·45ξ3
‫ݎ‬ௐா ൌ 0.182ξ͵ ൌ ͲǤ͵ͳ͸ ൌ ͵ͳǤ͸ Ψ
where
E Young’s Modulus = 70000 MPa
Q Poisson’s ratio = 0.23
rUK UK,mic/UJ,mic-ratio = 0.0895 (see eq. (6))
Js fracture surface energy = 3.0 Nm/m² [6], [10]
t thickness = 19 mm (by experiment )
Vt tensile stress at center = 40.12 N/mm² (median, by experiment)
x edge length of fragment = 7.6 mm (median, by experiment)
For the chosen 19 mm FT glazing the factor rWE = 0.316 shows that approximately 68 %
of the elastic strain energy is converted into deformation work. This value bases on an
average value of Js = 3.0 Nm/m². Changing assumption of Js (Table 1) cause significant
changes in rWE (Js = 2.0 Nm/m² rWE = 0.211, Js = 4.0 Nm/m² rWE = 0.397).
The fragment size and the tensile stress at the pane center have to be determined by
experimental data. Several authors described ways to predict the fragment size as
correlating to surface compressive stress. Often these theoretical approaches are used on
the Griffith theory neglecting any kinetic energy or deformation work. Therefore the
determined fragment size is too small or experimental determined factors have to be
added to the theory [30], [31]. Another authors [32], [12] are suggesting is that only a
part (35 to 43 %) of the stored energy is used up in generating new surfaces. A
comprehensive approach to link fragment size and tensile stress need to be developed
and proven by experiment.
518
5.2. Force approach

In the experimental study comparable in-plane forces should be measured whereas
several assumptions are necessary:
x A specific ratio of the elastic strain energy in the tensile stress zone is
transferred into deformation work.
x There is an equal amount of elastic strain energy, which is used up in the
compressive stress zone in deformation work to stay in equilibrium.
x If the fragment is prevented to deform or expand itself by constraint, it will
generate in-plane forces.
x The value of these in-plane forces are depending on the strain.
Figure 7: Schematic of force approach of one fragment [3]
A linear force ½Fx generated from one fragment with an edge length x is obtained by
means of the integration of eq. (8), over half of the tension zone respectively
compressive zone.
0,5t (14)
z 2 1
1/2Fx,l = න (z)= න 0 · ൬1-12· ቀ ቁ ൰ dz= ± t ·t·
t t 3ξ3
dz
2ξ3
The point force ½Fx,P is calculated by multiplication with edge length x
1 (15)
1/2Fx,p =(±t ·t· )·x
3ξ3
and multiplied with the ratio rWE to calculate the force ½Fx,P,WE, which is the result of
the deformation work.
1 (16)
1/2Fx,P,WE =(±t ·t· )·x·(1-rWa )=(±t ·t·x)·0.132
3ξ3
The force ½Fy,P,WE in y-direction is equal ½Fx,P,WE as only dependent on the tensile
stress in the pane center, the thickness, the edge length of the fragment and an energy-
ratio.
519
Challenging Glass 3
6. Testing, outlook and summary

The validation of the developed approach is in progress as many experiments to collect
data have been conducted [3]. The testing was conducted a 19 mm FT glazing of app.
1740 mm x 1210 mm. The glazing was point supported under constraint at the corners.
The testing has been finished and the results compared to the theoretical approach will
be published in near future.
The feasibility to determine the elastic strain energy and therefore in-plane forces of a
failed thermally tempered glass under constraint may be developed further to an
approach of increased post-breakage robustness.
The energetic approach may be a suggestion to integrate the kinetic energy at the energy
release of thermally tempered glass at the moment of failure. The authors invite
everybody to read the approach for critical response.
7. References
[1] Bos, F.: Safety Concepts in Structural Glass Engineering - Towards an Integrated Approach.
Dissertation. Zupthen: Wöhrmann Print Service, 2009.
[2] Kott, A.: Zum Trag- und Resttragverhalten von Verbundsicherheitsglas. Dissertation, ETH Zürich,
2007.
[3] Dietrich, N.: Nachbruchverhalten von vorgespannten Verglasungen unter zentrischem Zwang. Diploma
thesis. University of Applied Sciences Dresden, 2011.
[4] Jue, Jireh Y.: Energy Concepts for Fracture. Virginia Tech Materials Science and Engineering, 5. Mai
1997.
[5] Griffith, A. A.: The phenomena of rupture and flow in solids. Philosophical Transactions of the Royal
Society London Series A. (221) 1921, pp. 163-198.
[6] Kerkhof, F.: Bruchvorgänge in Gläsern. Frankfurt am Main : Verlag der deutschen glastechnischen
Gesellschaft, 1970.
[7] Mott, N. F.: Brittle fracture in mild-steel plates. In: Engineering (165), 1948, pp. 16-18.
[8] Acloque, P.: Déformation et rupture des verres. Ann Mines. 2, 1975, pp. 57-66.
[9] Nielsen, J. H.; Olesen, J. F.; Stang, H.: The Fracture of Tempered Soda-Lime-Silica Glass.
Experimental Mechanics. 2009, Bd. 49, 6, pp. 855-870.
[10] Petzold, A.; Marusch, H.; Schramm, B.: Der Baustof Glas. Berlin : Verlag für Bauwesen, 1990. pp. 35.
[11] Shand, E. B.: Correlation of Strength of Glass with Fracture flaws of Measured Size. J. Am. Ceram.
Soc. 1961, Bd. 44(9), pp. 451-55.
[12] Gulati, S. T.: Frangibility of Tempered Soda-Lime Glass Sheet, paper presented at Glass Processing
Days. Tampere: 1997. pp. 227-231, Conference proceedings.
[13] Roesler, F. C.: Brittle Fractures near Equilibrium. Proc. Phys. Soc. 1956, Bd. 69B, pp. 981-92.
[14] Wiederhorn, S. M. Fracture Surface Energy of Glass. Journal of the America. Ceram. Soc. Februar
1969, pp. 99-105.
[15] Shutov, A. I., Popov, P. B. und Bubeev, A. B. Prediction of the character of tempered glass fracture.
Glass and Ceramics. 55, 1998.
[16] Nakayama, J: Direct measurement of fracture energies of britlle heterogeneous materials. Journal of the
American Ceramic Society. 1965, Bd. 48.11, pp. 583-587.
[17] Mecholsky, J. J.; Rice, R. W.; Freiman, S. W.: Prediction of Fracture Energy and Flaw Size in Glasses
from Measurements of Mirror Size. Journal of the American Ceramic Society. 1974, Bd. 57, 10, pp.
440-443.
[18] Clif, C. J.: Fracture of glass under various liquids and gases. Journal of the Society of Glass
Technology. 1957, Bd. 41, pp. 157-167.
[19] Linger, K. R.; Holloway, D. G.: The Fracture Energy of Glass. Philosophical Magazine. 1968, Bd. 18,
156.
[20] Berdenikov, W. P.: Measurement of Surface Tension of Solids. Soviet Phys. Z.S. 1933, Bd. 4, pp. 397-
419.
[21] Proctor, B. A.; Whitney, I.; Johnson, J. W.: The strength of fused silica. Proc. R. Soc. Lond. 1967, Bd.
297, pp. 534-557.
520
[22] Davidge, R. W.; Tappin, G.: The Effective Surface Energy of Brittle Materials. Journal of Materials
Science. 1968, Bd. 3, pp. 165-173.
[23] EN 572-1. Glass in building. Basic soda lime silicate glass products. Part 1: Definitions and general
physical and mechanical properties. 2004.
[24] Takahashi, K.: Fast Fracture in Tempered Glass. Key Engineering Materials. 1999, pp. 9-18.
[25] Acloque, P.: High speed cinematographic study of the fracture process in toughened glass. Symposium
on Mechanical Strength of Glass and Ways of Improving it. 1962, pp. 581-886.
[26] Nielsen, J. H.: Tempered Glass - bolted connections and related problems. Dissertation. Kgs. Lyngby:
Technical University of Denmark, 2009.
[27] Acloque, P.: La fracture du verre propagation - influence des précontraintes. Verres Refract. 17, 1963, 3,
pp. 151-162.
[28] Chaudhri, M. M.; Liangyi, C.: The catastrophic failure of thermally tempered glass caused by small-
particle impact. Nature. 320, 1986, Bd. 6.
[29] Barsom, J. M.: Fracture of Tempered Glass. Journal of the American Society. 1968, 51, pp. 75-78.
[30] Warren, P. D.: Fragmentation of thermally strengthened glass. In: Varner, J. R. (Ed.): Fractography of
glasses and ceramics IV. American Ceramic Society, 2001.
[31] Tandon, R.; Glass, S. J: Controlling the fragmentation behavior of stressed glass. In: Bradt, R.C. (Ed.):
Fracture mechanics of ceramics. Houston: Springer Science+Business Media, 2005, pp. 77-91.
[32] Laufs, W.; Sedlacek, G.: Stress distribution in thermally tempered glass panes near the edges, corners
and holes, Part 1 and Part 2. Glastechnische Berichte Glass Science Technology. 1999, Bd. 72, 2, pp. 7-
14 and 42-48.
521
doi:10.3233/978-1-61499-061-1-523
Surface Cracked Layer and Damage

Resistance of Glass under Contact Loading
Yurii Rodichev, Yurii Yevplov, Helen Soroka
G.S. Pisarenko Institute for Problems of Strength, Ukraine, rym@ipp.kiev.ua,
sorokahelen@gmail
Frederic Veer,
Delft University of Technology, Netherlands, F.A.Veer@tudelft.nl
Strength and durability of glass structures depend directly on the state of the
cracked surface layer and the degree of surface damage [1, 2]. Lack of control of
surface micro-cracks and damage during the production of float glass and
processing of glass element as well as under operation conditions leads to
significant uncertainty in the estimates of the carrying capacity of glass structures
[3]. A technical approach based on precision control of critical surface micro-cracks
and damages was developed to guarantee a minimum level of strength of carrying
glass structures such as beams and plates under bending and tension. The results
of an experimental study of damage resistance of glass under contact loading were
studied by looking at the effects on the surface cracked layer. The glass surface was
damaged by a hard-alloy cutting roller using different cutting force. The
dependence of bending strength values and depth of critical surface micro-cracks
on the conditions of contact loading was found. Conditions of glass damage
resistance changing due to the changes in the surface cracked layer were defined.
The possibility of damage resistance and load bearing capacity of glass
constructions increasing as a result of protective coatings was investigated.
Keywords: Glass, Surface cracked layer, Strength, Damage resistance
1. Introduction
Glass is a surface defective structural material. The surface micro-cracks and damages
formed during float glass production and glass element processing as well as under
operation conditions create a specific surface cracked layer, which is an integral part of
glass structure. The level of defectiveness of the surface cracked layer is much higher
than that of the homogeneous internal structure of glass. Therefore, surface defects are
the fracture source of glass parts under thermo-mechanical loading. It was shown that
the strength and durability of glass parts depend directly on the state of micro-cracks
and damages located in this cracked surface layer [1-3]. The depth of these micro-cracks
does not exceed 100 μm usually. Contact interaction of glass part surface with hard
bodies leads to micro-crack growth, significantly increasing the damage in the surface
cracked layer and local increasing its depth. So controlling the surface damage as well
as critical micro-crack shape and size is an engineering challenge. The lack of control of
surface micro-cracks and damage during glass parts production and treatment leads to
significant uncertainty in estimating the load carrying capacity of structures [3, 4].
523
Challenging Glass 3
The thinness of the surface cracked layer [1, 3 and 4] and low scratch-resistance of glass
under contact loading [5, 6] cause the uncertainty of the actual state of surface micro-
cracks which are fracture sources in load carrying structures under operational
conditions. Sub-critical growth of micro-cracks occurs in the conditions of the complex
interaction with the environment, other cracks and surface defects. This feature of the
subcritical phase of glass fracture has not been studied extensively until now. Therefore,
the influence of the cracked surface layer on the damage resistance, fracture and
strength of glass carrying elements is a significant and actual scientific problem. The
literature on modification of scratch resistance of the surface does not take account of
the role of the cracked layer in the mechanical behavior of glass in the conditions of
contact loading [5, 6]. The influence of micro-cracks in the cracked layer was not
considered when investigating the different regimes of surface damaging processes
during the scratch-resistance test for annealed glass (Figure 1) and for the same
substrate with the protective nanoparticle coating. Influence of these scratches on the
structural strength of glass was also not studied.
Figure 1: Typical regimes of glass surface damaging by a Vickers penetrator: microductile regime,
microcracking regime and micro-abrasive regime [5].
The effect of the surface cracked layer on the damage resistance under contact loading
and structural strength of glass with damaged cracked layer is investigated in this paper.
A technical approach based on precision control of critical surface micro-cracks and

damages located in the cracked layer of glass was developed to guarantee a certain level
of strength of load bearing glass structures such as beams and plates under bending and
tension. Results of an experimental study of damage resistance of glass under contact
loading were obtained looking at the effects on the surface cracked layer. The glass
surface was damaged by a hard-alloy cutting roller with a different cutting force. The
micro-cracking regime was assigned under the different loads on the roller cutter. A
dependence of bending strength values and depth of the critical surface micro-cracks on
the conditions of contact loading was found. Conditions under which damage resistance
varies significantly due to the differences in the surface cracked layer were defined.
Using a polyamide safety film glued to the glass surface the possibility of damage
524
Surface Cracked Layer and Damage Resistance of Glass under Contact Loading
resistance and load bearing capacity increase of glass constructions as a result of

protective coatings application was investigated.
2. Technical approach and experimental procedures

The basis of the proposed technical approach to guarantee a certain minimum level of
strength of carrying glass structures such as beams and plates under bending and tension
was given in [1]. The core of this approach is that the parameters of crack growth
resistance of float glass are dependent on it physicochemical properties and production
technology as well as the data on surface defects in the specific cracked layer formed
under the glass production, glass processing and handling at the service conditions
compose the important aggregate basis for effective assessment of structural glass
strength. The concept of glass with its cracked surface layer is illustrated in Figure 2.
mean depth maximal depth
gas bubbles cracked inclusions

layer
Figure 2: The micro-cracks in surface cracked layer and internal defects in glass
The micro-cracks placed in the surface layer are more influential under thermal and
mechanical loading than internal technological defects like gas bubbles or inclusions as
well as than defects of the micro- and nano-structure of glass. External and internal
loads lead to micro-cracks growth and a local increase of cracked layer depth as well as
the formation of deep critical defects. The strength of glass parts decreases as the result
of surface damage continuously if the state of micro-cracks and micro-damage is not

controlled.
The core of the technical approach is connected with considering how the surface
cracked layer is influenced by the damage resistance under contact loading and the
structural strength of glass in service conditions. The formation of deeper micro-cracks
and rough damage under interaction of the roller cutter with the surface cracked layer of
the glass component under contact loading with an applied load Q is given in Figure 3.
The critical depth B of notches made by the cutter under different applied loads Q and
the resulting value of the bending strength of the damaged glass specimens tested in 4-
point bending were used as the parameters for damage resistance and structural strength
of glass after contact damage.
525
Challenging Glass 3
Q
Roller cutter Lateral chips
Initial
depth b
.
Cracked Deeper
Critical cracks
layer
depth B
Figure 3: Scheme of surface deeper cracks and damages formation under roller cutter action
The method of damaging by forming the cut-cracks using a hard-alloy cutting roller was
developed in this study. Flat specimens 130 mm×50 mm×6 mm were made from billets
with the dimensions 400 mm×50 mm×6 mm cut of from float glass on industrial
equipment in “Altis-Glass Ltd.” glasswork. Specimens were made with high quality and
low defectiveness sharp edges. They were packed with polymer film to protect the
specimen edge under laboratory handling. The state of the sharp edges was controlled
before inflicting controlled damage on the glass surface in the working area of
specimens with l = 50 mm (Distance between the lower supports was 100 mm). Nine
transverse notches-cracks were made in the working area on the surface of specimen
placed in tension under bending tests. The length of the notches was 40 mm (see
Figure 4).
l = 50mm
L = 100mm
9 notches
40mm
9 notches 50mm
Figure 4: Scheme of notched specimens of float glass at the 4-point bending test
and notches location in the working area in tension
526
The edges of the specimen were protected and did not damage. All notches were loaded
with the same stress. Therefore, only one of nine notches in the each specimen was the
fracture source. The probability of obtaining minimum values for the strength was
increased due to this method of specimens damaging.
Tests were made at a constant loading rate of 1 MPa/s. The scatter of tests results under
contact load Q in the range 20 … 50 N was very small. Therefore, only three
specimens were tested at each load level. The scatter of strength values for the tests with
load Q 10 N was larger. The number of tested specimens at these load levels was
increased to 5 items for this reason.
Test results are given in Tables 1-4 and in Figure 5. As it follows from the test results
and previous fractographic investigation of failed specimens, steady growth of the crack
continues under the cutting force (with given roller) equal 15÷20 N. If at Q = 10 N
division of the sample is still on the track roller, then under Q = 7 N, approximately
after 50% of crack length, it begins to branch, the mechanism of growth of the crack
changes. With a further reduction of cutting force, a shortening of straight portion and
an increase of crack branching are observed.
Table 1: Strength of float glass with damaged surface under 4-point bending tests
(L = 100 mm, l = 50 mm), load Q in the range 50 … 10 N
V, CV,
Load of cut Q, N
¡P} ¡P}
1 15.9
2 50 15.9 15.9
3 15.9
1 16.8
2 40 16.6 16.8
3 17.0
1 17.4
2 30 17.2 17.2
3 17.0
1 18.5
2 20 18.5 18.7
3 19.0
1 36.0
2 30.0
3 26.0
32.0
4 10 37.0
5 27.0
6 36.0
527
Challenging Glass 3
Table 2: Strength of float glass with damaged surface under 4-point bending tests
(L = 100 mm, l = 50 mm), load Q in the range 7 … 1 N
Load of cut V, meanCV,

Q, N ¡P} ¡}
1 41.0
2 43.0
3 41.0 44.6
4 7.0 48.0
5 50.0
1 58.0
2 48.0
3 61.0
54.5
4 52.0
3.5
5 56.0
6 52.0
1 56.0
2 66.0
3 53.0 56.0
4 3.0 53.0
5 52.5
1 58.0
2 58.5
3 69.0 60.5
4 2.0 61.5
5 56.0
1 80,0
2 79,0
3 83,0 84.3
4 1.5 94,5
5 85,0
1 115,0
2 109,0
3 98,0 109.0
4 1.0 111,0
5 113,0
528
1* – 146,0
*
2 – 92,0
*
3 – 133,0 134.5
*
4 – 134,5
*
5 – 167,0
*
Initial specimens without the notches
The results of three-point bending tests (L = 40 mm) of samples at Q = 50; 40; 30 N are
presented in the Table 3. Low spread in the measured strength and good agreement with
the results of pure bending tests, allow for the assertion that the use of the innovative
specimens designed for this work can significantly reduce the number of specimens
required to determine the lower limit of ultimate bending strength. It should be noted
that at Q = 2 N the crack growth deviates from the cut, which is explained by
considerable rigidity of the sample (with L = 40 mm) and decreasing depth of the crack.
¢}ble 3 : Strength of float glass with damaged surface under 3-point bending tests (L = 50 mm)
Load of cut V, CV,
Q, N ¡P} ¡P}
1 15.0
2 16.0
15.6
3 15.0
5.0 (15.9) *
4 16.0
5 16.0
1 16.5
2 16.5
16.6
3 16.5
4.0 (16.8) *
4 16.5
5 17.0
1 18.5
2 18.5
18.2
3 18.0
3.0 (17.2) *
4 18.0
5 18.0
*
Values of strength under 4-point bending tests
529
Challenging Glass 3
¢}ble 4 : Summary of tests results for the experimental curve of dependence of strength V
for damaged float glass 6 mm and depth of notch b on contact load Q
V, b o,
Q, dN Lg Q Lg V Lg bo
¡P} mm
1 5 15.9 0.70 1.201 0.425 -0.37

2 4 16.8 0.60 1.225 0.35 -0.46
3 3 17.2 0.48 1.236 0.25 -0.60
4 2 18.7 0.30 1.272 0.15 -0.82
5 1 32.0 0 1.505 0.06÷0.09 -1.12
6 0.7 42.6 -0.155 1.629
7 0.35 54.5 -0.46 1.736
8 0.3 56.0 -0.52 1.750
9 0.2 60.5 -0.70 1.782
10 0.15 84.3 -0.824 1.926
11 0.1 109 -1.0 2.037
5 , MPa
Lg
Lg V Lg Q
3,5
3
3
5
3
2,5
2
2
5
2
1,5
1
1
1
5
I II
0,5
0
-1 -0,5 0 0,5 1
0
-1 -0,5 0 0,5 1
Lg Q, dN
Figure 5: Dependence of bending strength of damaged float glass 6 mm

n load Q
Experimental dependence of the bending strength of float glass (thickness 6 mm) from
the force under cutting by roller cutter is shown in Figure 4. Two specific ranges of
contact loads are obvious. Range I is characterized by the load, which is insufficient to
form a continuous notch crack. The depth of the notch is commensurate with the initial
depth b of micro-cracks in the surface cracked layer of glass. Under the load equal to
1H bending strength (109 MPa) is comparable to that of the original glass. For this
range of loads impact of cracked layer is crucial.
530
In the range II loading is sufficient to "pierce" the cracked surface layer and to form a
crack-notch with a depth much greater than the thickness of the" fractured layer. In this
case, resistance to crack formation is more dependent on the crack growth resistance of
inner homogeneous and less defective structure of glass. This structure is not able to
resist the growth of the crack into depth as it grows inwards from the cracked surface
layer. Therefore, the increased sensitivity of glass to damage under contact loading and
the corresponding decrease of the material strength occur in area II.
Curve 3 in Figure 4, is the extrapolation of curve 2. The dependence of the bending

strength of float glass from the cutting force reflects a possibility to increase the strength
of glass parts up to 1000 MPa and more by optimizing the physical and mechanical
processing of glass, control of cracked state of the layer or its complete removal.
0 Lg Q, dN
0 0,2 0,4 0,6
0,2
0,4
0,6
0,8
1
1,2
Lg b0 ,mm
Figure 5: Dependence of the notch depth on contact load Q at damage

the float glass surface by roller cutter
Experimental dependence of the notch depth on contact load Q at damage of the float
glass surface by roller cutter is shown in the Figure 5. In logarithmic coordinates, the
dependence can be represented by a straight line, reflecting a power law increase

in damage of the glass as the cutting load increases. The dependency of the strength and
depth of the glass damage on the cutting force, shown in Figures 4 and 5 can be used to
optimize the technology of physical-mechanical processing of glass components on the
criteria of strength and permissible imperfection in view of anticipated operating
contact loads on carrying structures. The established pattern may also be useful in
selecting of cutting modes of glass on cutting machines as well as technical solutions to
protect the glass surface from damage during operation, for example, by applying
protective coatings.
The results of the evaluation of the effectiveness of safety building polyamide films
such as «Solar Guard» with the adhesive layer to improve performance of glass
structures under contact loading are given below.
Bending strength of damaged glass specimens with protective polymer coating.
531
Challenging Glass 3
The assessment of protective properties of polyamide safety film under contact loading
of glass parts was studied using flat specimens and bending strength tests. The
polyamide films with a thickness 40 μm and 150 μm were glued to the surface of flat
specimens of float glass 130 mm x 50 mm x 6 mm before the specimens were notched
with a tungsten carbide roller cutter. To get the transverse notch for specimens with the
thickness of film 40 μm and 150 μm specimens were loaded with Q = 40 N and
Q = 80 N, respectively. The safety film with damage was on the tension side of
specimens under the bending tests. The tests results are given in table 5. It was found
that the thickness of the protective coating must be optimized concerning with operation
contact loads and mode of indenter. Protective properties of the polymer film remain
until the indenter does not contact directly with the surface layer of glass under the load.
Increases in coating thickness increase the damage resistance of glass part under contact
loading. The thickness of the thin coating was insufficient at a contact load 40 N. The
bending strength of glass with a film 40 μm was almost the same as unprotected
damaged specimens. By increasing the coating thickness to 150 μm the strength of glass
has remained the same as that of the undamaged specimens while the contact load Q
was increased to 80 N.
Table 5: Bending strength of notched glass specimens with safety polymer film
Specimen Strength of specimens without film, Strength of specimens with film,
MPa MPa
Q = 40 N Thickness of the film 40 μm,
Q = 40 N
1 16.8 17.6
2 16.6 18.7
16.8 17.9
3 17.0 17.0
4 18.5
5 17.5
Q = 50 N Thickness of the film 150 μm,
Q = 80 N
6 15.9 122.0
7 15.9 141
15.9 119.6
8 15.9 91.0
(134.5 MPa
9 132.0 for undamaged
specimens)
10 112.0
532
3. Conclusion
A technical approach based on precision control of critical surface micro-cracks and
damages was developed, and results of an experimental study of damage resistance of
glass under contact loading were obtained by looking at the effects on the surface
cracked layer. The glass surface was damaged using a hard-alloy, tungsten carbide,
cutting roller with different cutting force being applied. The dependence of bending
strength values and depth of critical surface micro-cracks to the applied contact loading
was found. The conditions of increased glass damage resistance due to changes to the
surface cracked layer were defined. The possibility of improved glass construction
damage resistance and increased load bearing capacity by applying protective film
coatings is demonstrated.
4. References
[1] Rodichev Yu, Tregubov N., Veer F., Maslov V., The features of fracture and engineering strength of
glass materials with surface, internal and mixed modes of critical defects, GPD 2011 Proc..- Tampere,
Finland, 2011
[2] Veer F., Rodichev Yu., Corrosion effects on soda lime glass, Proc. of Challenge glass 2.- 2010.-TU
Delft, Netherlands.
[3] Rodichev Yu, Veer F., Fracture Resistance, Surface Defects and Structural Strength of Glass, Proc. of
Challenge glass 2.- 2010.-TU Delft, Netherlands.
[4] Rodichev Yu, Tregubov N., Veer F., Maslov V., Influence of surface and internal defects on the
fracture and engineering strength of glassmaterials, Proc. of Int. conf. “In-service damage of materials,
its diagnostics and prediction”.-2011, Ternopil, Ukraine.
[5] Tartivel, R., Reynaud, E., Grasset, F., Sangleboeuf, J. C. & Rouxel, T., Superscratch-resistant glass by
means of a transparent nanostructured inorganic coating. Journal Of Non-Crystalline Solids, January,
Volume 353, Number 1, Pages 108-110, 2007.
[6] Le Houerou V., Sangleboeuf J.-C., Deriano S., Rouxel T., Duisit G., Surface damage of soda–lime–
silica glasses: indentation scratch behavior, Journal of Non-Crystalline Solids Volume 316, Pages 54–
63, 2003.
533
doi:10.3233/978-1-61499-061-1-535
Surface Defects and Statistical

Characteristics of Glass Strength
Yurii Rodichev, Yurii Yevplov, Helen Soroka
G.S. Pisarenko Institute for Problems of Strength, Ukraine,
rym@ipp.kiev.ua, sorokahelen@gmail
Frederic Veer
Delft University of Technology, Netherlands, F.A.Veer@tudelft.nl
Nikolay Tregubov
Konstantinovsky glasswork “Kvarsit”, Ukraine
Vladimir Polivyany
Konstantinovsky glasswork “Spectechglass”, Ukraine, Polivvg@rambler.ru
Statistical distribution of the data on the bending strength of float glass

systematically deviates from the Weibull statistic distribution [1-3]. It was shown
that surface defects formed under glass processing and handing are the cause of the
multi linear nature of the Weibull plots. The results of an experimental study of
statistical distribution of bending strength values were obtained in view of the
effect of mechanical treatment and contact loads on glass surface damaging and
strength. The shapes and sizes of fracture sources - critical micro-cracks and
damages were investigated. It was shown that the statistical behavior of structural
glass strength may be controlled by monitoring critical surface defects, conditions
of their formation in production of load carrying glass structures basing on
experimental strength data.
Keywords: Glass, Surface damages, Strength, Statistical distribution
1. Introduction
Glass is used widely as a structural material for architectural constructions. The

challenge is that there is no reliable value for the strength of glass. The non-controlled
surface micro-cracks and damages formed during float glass production and glass
element processing as well as under operational conditions cause the fracture of glass
structures under mechanical loading. The inhomogeneity of shapes and sizes of micro-
cracks and defects in the cracked surface layer leads to a significant scatter in the
results of mechanical tests and systematic deviations of experimental plots from the
mono-linear Weibull statistic distribution [1-3]. Bending test results give a scattered
value for the bending strength, with a spread of some 30 to 50% of the mean strength in
certain cases. Therefore, the actual strength of glass and carrying capacity of glass
elements is difficult to predict up to now.
The experimental results of structural strength of float glass obtained under bending
tests of plate specimens 6 x 40 x 400 mm are shown in Figures 1 and 2 [4]. All four
point bending tests were conducted on a Zwick Z100 universal testing machine under
535
Challenging Glass 3
displacement control. Load span was 175 mm and support span – 350 mm. Test speed
was 5 mm/minute.
Figure 1 shows a Weibull plot of the lying test results with the damaged cut edge of
specimens in the compression zone.
Figure 1: Weibull plot of lying test result with the damaged cut edge upwards.
Two different parts of experimental curve are typical for these specimens. Upper part
with steep slope includes main group of specimens with probability of fracture more
than 5%. High spread of the results – 30% (variation coefficient -20%) is connected
with high inhomogeneity of fracture source (microcracks) parameters. Mean value is
61.5 MPa, maximum value - 100 MPa and minimum value - 40 MPa. The results
presented by lower part of experimental curve with the lowest strength value are typical
for the most rough defects and cracks in fracture focus. The slope of this curve is much
less. It shows that the spread of parameters of rough defects is less than in the upper part
of the experimental Weibull plot.
Figure 2 shows a Weibull plot of the lying test results with the damaged cut edge of
specimens in the tension zone. Three clearly eminent parts are shown by appropriate
lines with a different slope. The spread of these results is less than the previous. Mean
value is 51.5 MPa, maximum value is 59.0 MPa and minimum is 34.5 MPa.
Therefore, the larger the surface defects of the cut edge, the lower the scatter of results
as well as the deeper the depth of cracks in processed glass details.
It is important that the mean strength value is not an effective parameter of glass
strength as the result of statistic heterogeneity of strength values. More significant
uncertainty is related to the lower strength values assessment basing on mono-linear
mode of the Weibull distribution as the result of considerable deviation of test results in
lower part of experimental curve with lowest strength value (see Figure 1). The estimate
of the lower boundary of strength values and allowable stress in the design of carrying
536
Surface Defects and Statistical Characteristics of Glass Strength
glass structures is a complex task for this reason. Therefore, the influence of surface
defects on the fracture and statistical parameters of glass element strength is still an
actual scientific challenge.
Figure 2: Weibull plot of the lying test results with the damaged cut edge of specimens in the tension zone.
Assessment of the statistical behavior of glass plate specimens bending strength

considering the influence of surface defects made under edge grinding and contact
damaging was the focus of our experimental investigation.
The results of a study of the statistical distribution of bending strength values obtained
using the optical microscopy control of fracture sources in the mirror zone of fracture
surface of tested specimens. It was found that experimental data on glass strength
correlates directly with the parameters of critical crack size and damages obtained using
the micrograph analysis. Therefore, the influence of surface defects on the fracture and
statistical parameters of glass elements strength is proven. The results obtained in this
study showed that the statistical behavior of structural glass strength may be controlled
by monitoring of critical surface defects, conditions of their formation during the
production of load bearing glass structures basing on experimental strength data.
Accounting the multi-linearity of Weibull plots allows us to define more precisely the
allowable stress in the design of architectural glass structures and to improve the
technology of mechanical treatment and handling of glass parts using strength
parameters and data on defects in cracked layer.
2. Test method
Three groups of specimens were tested.
First group included the plate specimens cut from 8 mm float glass using a rotary
diamond wheel cutter into specimens 400 mm long and 60 mm width. All 80 specimens
were tested in four point bending (support roller distance 330 mm, loading roller
distance 133 mm). Tests were made on a Zwick z10 universal testing machine with a
constant traverse speed 10 mm/min. Half of the specimens were tested with the bur side
537
Challenging Glass 3
down. Thus, the surface with specific residual defects as a result of edge cutting was
under tension. Other specimens were tested with the bur side up, and the opposite
surface without these defects was under tension (see Figure 3).
The second group was made up of 6 mm float glass specimens manufactured using
other technology and tested in a different way.
Plate specimens 6 x 50 x 130 mm of float glass were cut and ground on diamond disc to
blunt the sharp edges with a chamfer 1 x 45º. End faces of specimens were not ground.
The specimens were tested in 4 point bending (support roller distance 100 mm, loading
roller distance 50 mm) on a ZD-4 universal hydraulic testing machine with a constant
loading speed 60 MPa/min. Conditions of specimens tests are given in the result
discussion.
Figure 3: Test setup with specimen of 8 mm float glass in position with bur side up
The third group of specimens 6 x 50 x 130 mm of float glass was damaged by a hard-
alloy roller cutter with different cutting force. Cutting force Q was changed in the range
50 … 10 N (high load) and 7… 1 N (low load) [5]. Nine transverse notches-cracks were
made in the working area on the surface of specimen placed in tension under bending
tests. The length of the notches was 40 mm. The edges of the specimen were protected
and did not damage. All notches were loaded with the same stress. Thus, only one of
nine notches in each specimen was the fracture source. The probability of obtaining
minimum values for the strength was increased due to this method of damaging
specimens. The modeling statistic distribution was made using the results of these
damaged specimen tests [5] to assess the mode of Weibull plot for glass elements with
roughly damaged surface in extreme operation conditions.
538
3. Results
The summary of test results for the first group is given in Table 1.
The mean bending strength values of specimens tested in position bur side up (less
damaged surface of the specimens is under tension) were noticeably higher than for
other part of specimens (51.3 MPa and 48.5 MPa correspondingly). The variation
coefficient as well as minimum and maximum strength values of specimens tested in
position bur side up was significantly higher as well. Thus, residual surface defects
formed under glass cutting were not eliminated during the grinding and polishing.
Table 1: Spread of strength for first group of specimens of float glass 8 mm.
Std
Test type. Test Mean Minimum Maximum
strength/
Quantity speed strength strength strength
mean
of specimens (mm/min) (MPa) (MPa) (MPa)
strength
Bur side
down. 10 48.5 15.9% 29.4 62.8
40 specimens
Bur side up
40 specimens 10 51.3 17.8% 37.2 85.4
Figure 4 shows Weibull plots for the 10 mm/minute data separated by side while Figure
5 shows the combined data for all specimens of the first group. The data in Figure 4 are
not mono-linear Weibull but certainly a closer approximation than the combined data in
Figure 5.
Figure 4: Weibull plot of data on bending strength for specimens of 8 mm float glass
539
Challenging Glass 3
Figure 5: Weibull plot of the combined data on bending strength for all specimens of 8 mm float glass
The results shown in Figure 4 (data of bur side down) and Figure 5 are characterized by
significant deviation of the lower values of strength at the probability of fracture less
than 10 % from mono-linear Weibull distribution. This feature is a natural consequence
of inhomogeneity of residual defects formed during uncontrolled glass cutting and
mechanical treatment.
The test results for the second group of the specimens with ground sharp edges are
given in Table 2 and Table 3. First part of these specimens in the amount of 20 pieces
made of 6 mm float glass was ground in several different ways. It was important to
evaluate the dependence of the glass strength relative to the direction of the sharp edge
grinding. The schemes indicating the direction chamfer 1u45 grinding are given in the
comments in Table 2. The position of the cracked surface layer with the initial micro-
cracks is shown in the schemes. The edge face formed during the roller cutting of glass
was low defected due to the high homogeneity of the internal structure of float glass.
The specimens 1 – 10 were with ground sharp edges. The specimens 11 – 20 were
ground in opposite bur side with uncontrolled surface damages formed during glass
cutting.
540
¢}ble 2: Strength of 6 mm float glass tested using the specimens with blunted the sharp edges by a chamfer
1 x 45º made on different method
Mean Vb
/ Vb, _____________________
Comments
¡P} Vmin / Vmax
1 52,5 Sharp edge ground from glass side with cracked layer
2 60,0 55,0 to edge face
3 53,3 _____________________
4 50,7* 50,7/ 60,0 Direction of chamfer 1u45
Cracked
grinding.
*
5 58,9** layer and **- central fracture.
6 57,6 Sharp edge ground from edge face to side with cracked
7 48,0 layer
8 48,0* 55,0
9 56,0 _____________________
Direction of chamfer
48,0/ 58,9
Cracked 1u45 grinding.
*
10 58,9 layer – central fracture
11 63,0* *
central fracture Bur side. Direction of chamfer 1u45
12 59,0 59,0 edge fracture grinding.
13 60,0* _________ *
central fracture
14 54,6 54,0…63,0 edge fracture
Cracked
15 58,0 edge fracture layer
16 51,6 edge fracture Bur side. Direction of chamfer 1u45

17 60,0 edge fracture grinding.
18 57,0 53,7 edge fracture
19 59,0 _________ edge fracture
41,0…60,0
20 41,0* *
central fracture Cracked
layer
The test results of four mentioned kinds of samples were comparable. So they were
combined into a single statistical sampling (see Figure 6). These results demonstrate
clearly the suitability of the mono linear Weibull distribution for statistical analysis of
the data on strength of the glass elements with homogeneous surface defects formed
during the mechanical treatment with the appropriate production technology.
541
Challenging Glass 3
F Y
0,999 2
0,995
0,99
0,95 1
0,90
0,75
0,63 0
0,50
y = ln [-ln(1-F)]
0,30 -1
0,20
-2
0,10
0,05 -3
0,025
-4
0,010
-5
0,005
0,0025 -6
0,001 -7
1,5 1,55 1,6 1,65 1,7 1,75 1,8 1,85
Lg V
Figure 6: Weibull plot of bending strength of 6 mm float glass

(specimens with a chamfer 1 x 45º made on different method)
Figure 7 shows the fracture source at the sharp edge of glass specimen blunted with the
ground chamfer 0.7 mm x 45º. The long surface cut with the depth about 30 μm was
detected in the result of micrograph analysis of the fracture source in the mirror zone of
the fracture surface of the specimen tested at the 4 – point bending.
Figure 7: The micrograph of the fracture source in the ground chamfer in tension under bending tests.
The critical value of stress intensity coefficient KCr = 0.45 MPam was calculated
basing on strength value Vb = 45 MPa, crack geometry factor Y =1.8 and equation of
linear fracture mechanics for surface long cut with the depth 30 μm. The result shows
the direct correlation of technological micro-crack parameters (see Figure 7) with the
542
strength level of processed glass element. Thus, scatter of tests data is the result of
heterogeneity of the surface defects formed during the grinding of chamfer.
As these tests utilized a small number of specimens it was necessary to repeat the tests
on a larger scale to obtain statistically significant results. Therefore the additional 30
specimen were made using the diamond disk to grind the bur side edge. These
specimens were tested in the same conditions. The test results were combined in a
single statistical sampling with the data on strength of the previous group of 20
specimens (see Table 3). Figure 7 shows the combined data for all specimens of the
second group.
¢}ble 3: Results of statistical analysis of the data on test of combined group

of 6 mm float glass specimens with a ground chamfer 1 x 45º.
Vb,
/ Lg Vb Probability, F F(y)
¡P}
1 2 3 4 5
1 36,0 1,55 0,01 -4,6
2 40,4 1,61 0,03 -3,49
3 41,0 1,613 0,05 -2,97
4 41,0 1,613 0,07 -2,62
5 43,0 1,63 0,09 -2,36
6 45,0 1,65 0,11 -2,15
7 45,6 1,66 0,13 -1,97
8 47,0 1,67 0,15 -1,82
9 48,0 1,68 0,17 -1,68
10 48,0 1,68 0,19 -1,56
11 48,0 1,68 0,21 -1,45
12 48,0 1,68 0,23 -1,34

*
13 48,0 1,68 0,25 -1,25
14 48,5 1,685 0,27 -1,16
15 49,0 1,69 0,29 -1,07
16 50,0 1,70 0,31 -0,99
17 50,0 1,70 0,33 -0,92
*
18 50,7 1,71 0,35 -0,84
19 51,0 1,71 0,37 -0,77
20 51,6 1,713 0,39 -0,70
21 51,6 1,713 0,41 -0,64
22 51,6 1,713 0,43 -0,58
23 52,5 1,72 0,45 -0,51
543
Challenging Glass 3
24 52,6 1,721 0,47 -0,45

25 53,0 1,724 0,49 -0,40
26 53,3 1,73 0,51 -0,34
27 54,0 1,732 0,53 -0,28
28 54,0 1,732 0,55 -0,23
29 54,6 1,74 0,57 -0,17
30 54,6 1,74 0,59 -0,115
31 56,0 1,748 0,61 -0,06
32 56,0 1,748 0,63 -0,006
33 56,0 1,748 0,65 0,05
34 56,0 1,748 0,67 0,103
35 57,0 1,756 0,69 0,16
36 57,0 1,756 0,71 0,213
37 57,0 1,756 0,73 0,27
38 57,6 1,76 0,75 0,33
39 57,6 1,76 0,77 0,39
40 58,0 1,763 0,79 0,445
41 58,9 1,77 0,81 0,51
**
42 58,9 1,77 0,83 0,57
43 59,0 1,77 0,85 0,64
44 59,0 1,77 0,87 0,713
45 59,0 1,77 0,89 0,80
46 60,0 1,78 0,91 0,88
47 60,0 1,78 0,93 0,98

48 60,0 1,78 0,95 1,10
49 60,6 1,78 0,97 1,25
50 63,0 1,80 0,99 1,53
MeanCVb = 52,6 ¡P}

St. Dev. S = 6,0MPA
Variation coefficient Q =11,4%
Maximum Vb = 63,0 ¡P}
Minimum Vb = 36,0 ¡P}
These results convincingly show the principal possibility to guarantee the homogeneity
of technological surface defects if the structural glass strength data and monitoring of
the defectiveness of the detail on the production line will be used for the development of
glass processing technology. Thus, the mono linear Weibull distribution may be
544
recommended in the glass structure design only for statistical analysis of the data on
strength of the glass elements with homogeneous surface defects formed during the
mechanical treatment. The production technology of load currying glass elements
should be optimized on statistical parameters in accordance with the concrete
technical requirements for modern architectural glazing.
The test results for the third group of the specimens damaged by roller cutter with a
different cutting force are given in Table 4 and Table 5. First part of these specimens in
the amount of 54 pieces made of 6 mm float glass was tested at the 4 – point bending
(see Table 4 and Figure 8). Three parties of the specimens were combined in this model
statistic sampling:
x specimens damaged by roller cutter with a high cutting force Q in the range
50…10 ;
x specimens damaged by roller cutter with a low cutting force Q in the range
7…1 ;
x undamaged specimens.
F Y
0,999
2
0,995
0,99
0,95
0,90 1
0,75
0,63 0
0,50
y = ln [-ln(1-F)]
0,30 -1
0,20
-2
0,10
0,05 -3
0,025
-4
0,010
-5
0,005
0,0025 -6
0,001 -7
1,5 1,55 1,6 1,65 1,7 1,75 1,8 1,85
Lg V
Figure 7: Mono-linear Weibull plot of data on strength of 6 mm flat glass specimens

with a ground chamfer 1 x 45º.
Such significant variation of surface damaging level of the specimens tested was
accepted to evaluate the Weibull statistics mode taking into account the influence of
different kinds of heterogeneous defects that can be applied to the surface of load
carrying glass structure during the careless processing and in the extreme operating
conditions.
545
Challenging Glass 3
¢}ble 4: Results of 4-point bending tests of the plate specimens damaged by roller cutter with a different
cutting force. Statistic analysis of the data made basing on the multi-linear Weibull distribution.
/ Vb,
Lg Vb Probability, F Y (F)
P
1 2 3 4 5
1 15,9 1,20 0,009 -4,7
2 15,9 1,20 0,027 -3,6
3 15,9 1,20 0,046 -3,06
4 16,6 1,22 0,065 -2,70
5 16,8 1,225 0,083 -2,45
6 17,0 1,23 0,10 -2,25
7 17,0 1,23 0,12 -2,06
8 17,2 1,235 0,14 -1,89
9 17,4 1,24 0,16 -1,75
10 18,5 1,27 0,176 -1,64
11 18,5 1,27 0,194 -1,53
12 19,0 1,28 0,21 -1,44
13 26,0 1,41 0,23 -1,34
14 27,0 1,43 0,25 -1,25
15 30,0 1,48 0,27 -1,16
16 36,0 1,56 0,29 -1,07
17 36,0 1,56 0,305 -1,01
18 37,0 1,57 0,324 -0,94
19 41,0 1,61 0,342 -0,87

20 41,0 1,61 0,361 -0,80
21 43,0 1,63 0,38 -0,74
22 48,0 1,68 0,40 -0,67
23 48,0 1,68 0,416 -0,62
24 50,0 1,70 0,435 -0,56
25 52,0 1,716 0,454 -0,50
26 52,0 1,716 0,472 -0,45
27 52,5 1,72 0,50 -0,37
28 53,0 1,724 0,51 -0,34
29 53,0 1,724 0,53 -0,28
30 56,0 1,75 0,546 -0,24
31 56,0 1,75 0,565 -0,18
546
32 56,0 1,75 0,58 -0,14

33 58,0 1,76 0,60 -0,09
34 58,0 1,76 0,62 -0,03
35 58,5 1,77 0,64 0,02
36 61,0 1,785 0,66 0,076
37 61,5 1,788 0,676 0,12
38 66,0 1,82 0,70 0,18
39 69,0 1,84 0,71 0,21
40 79,0 1,90 0,73 0,27
41 80,0 1,903 0,75 0,33
42 83,0 1,92 0,77 0,38
43 85,0 1,93 0,79 0,44
44 92,0 1,96 0,805 0,49
45 94,5 1,975 0,824 0,55
46 98,0 1,99 0,84 0,61
47 109,0 2,04 0,86 0,68
48 111,0 2,045 0,88 0,75
49 113,0 2,05 0,90 0,83
50 115,0 2,06 0,92 0,93
51 133,0 2,12 0,935 1,01
52 134,5 2,13 0,953 1,12
53 146,0 2,16 0,972 1,27
54 167,0 2,22 0,99 1,53
MeanCV = 58,72 ¡P}

St. Dev. S = 37,5 MPA
Variation coefficient Q = 64 %
Maximum V = 167,0 ¡P}
Minimum V = 15,9 ¡P}
The mean value of bending strength CVb = 58,72 ¡P} for roughly damaged glass
specimens in this model sampling is cloth to data on strength of glass with an industrial
processing (see Table 1) as well as for specimens with diamond ground chamfer (see
Tables 2 and 3). But very large spread of the strength data is negative difference of
these roughly damaged glass elements. The variation coefficient Q = 64 % as well as
difference between the minimum (CVb = 15,9 ¡P}) and maximum (CVb = 167,0 ¡P})
strength values were significantly higher also.
547
Challenging Glass 3
Figure 8 shows the important feature of lower part of experimental three-linear Weibull
plot associated with significant deflection of this part strength data at the probability of
fracture less than 20%. It is shown the actual stabilization of minimal values of strength
in the range 15…17 MPa. Basing on this feature of the Weibull plot for strength data of
intensively damaged glass elements it is possible to accept the strength value 15 MPa as
the lower boundary of bending strength at the fracture probability F = 0.1 % for
annealed glass parts in the hard operation conditions. It is obvious also that processing
technology non-controlled on parameters of glass defectiveness and strength as well as
on the data on statistical features of the lowest strength value can not be effective in the
architectural glass structures production.
F Y
0,999
0,995 2
0,99
0,95
1
0,90
0,75
0,63 0
0,50
y = ln [-ln(1-F)]
0,30 -1
0,20
-2
0,10
0,05 -3
0,02
-4
0,01
-5
0,005
0,0025 -6
0,001 -7
1 1,2 1,4 1,6 1,8 2 2,2 2,4
Lg V
Figure 8: Experimental Weibull plot of test results (4 - point bending) for specimens
damaged with cutting roller
– test results for high cutting force Q
(50,0…10,0 );
– test results for low cutting force Q
(7,0…1,0 );
– test results for undamaged specimens
The result of tests at the 4 – point and 3 – point scheme of the bending were similar. So
it was analyzes the joint results of 4-point and 3- point bending tests of the plate
specimens damaged by roller cutter with a different cutting force presented in Table 5
and Figure 9.
In spite of the reduction in the mean strength value up to CVb = 49,6 ¡P} and slow
increasing in the variation coefficient Q = 75,5 % in the general three-linear shape of
the Weibull curve has not changed.
548
¢}ble 5: Joint results of 4-point and 3- point bending tests of the plate specimens damaged by roller cutter
with a different cutting force. Statistic analysis of the data made based on the multi-linear Weibull distribution.
/ Vb,
Lg Vb Probability, F Y( F)
P
1 2 3 4 5
1 15,0 1,18 0,007 -4,96
2 15,0 1,18 0,022 -3,81
3 15,9 1,20 0,036 -3,31
4 15,9 1,20 0,050 -2,97
5 15,9 1,20 0,065 -2,70
6 16,0 1,204 0,080 -2,48
7 16,0 1,204 0,094 -2,32
8 16,0 1,204 0,108 -2,17
9 16,5 1,217 0,123 -2,03
10 16,5 1,217 0,138 -1,91
11 16,5 1,217 0,152 -1,80
12 16,5 1,217 0,166 -1,71
13 16,6 1,22 0,181 -1,61
14 16,8 1,225 0,195 -1,53
15 17,0 1,23 0,210 -1,445
16 17,0 1,23 0,225 -1,37
17 17,0 1,23 0,240 -1,29
18 17,2 1,235 0,254 -1,23
19 17,4 1,24 0,270 -1,16
20 18,0 1,255 0,282 -1,10

21 18,0 1,255 0,297 -1,04
22 18,0 1,255 0,311 -0,99
23 18,5 1,27 0,326 -0,93
24 18,5 1,27 0,340 -0,88
25 18,5 1,27 0,355 -0,82
26 18,5 1,27 0,370 -0,77
27 19,0 1,28 0,384 -0,72
28 26,0 1,41 0,398 -0,68
29 27,0 1,43 0,413 -0,63
30 30,0 1,48 0,427 -0,58
31 36,0 1,56 0,442 -0,54
549
Challenging Glass 3
32 36,0 1,56 0,456 -0,50

33 37,0 1,57 0,471 -0,45
34 41,0 1,61 0,485 -0,41
35 41,0 1,61 0,50 -0,37
36 43,0 1,63 0,514 -0,33
37 48,0 1,68 0,530 -0,28
38 48,0 1,68 0,543 -0,24
39 50,0 1,70 0,560 -0,20
40 52,0 1,716 0,570 -0,17
41 52,0 1,716 0,586 -0,13
42 52,5 1,72 0,60 -0,09
43 53,0 1,724 0,616 -0,04
44 53,0 1,724 0,63 -0,006
45 56,0 1,75 0,645 0,035
46 56,0 1,75 0,660 0,076
47 56,0 1,75 0,674 0,114
48 58,0 1,76 0,690 0,16
49 58,0 1,76 0,70 0,19
50 58,5 1,77 0,72 0,24
51 61,0 1,785 0,732 0,28
52 61,5 1,788 0,746 0,31
53 66,0 1,82 0,760 0,36
54 69,0 1,84 0,775 0,40
55 79,0 1,90 0,790 0,45

56 80,0 1,903 0,804 0,49
57 83,0 1,92 0,820 0,54
58 85,0 1,93 0,833 0,58
59 92,0 1,96 0,850 0,64
60 94,5 1,975 0,860 0,68
61 98,0 1,99 0,877 0,74
62 109,0 2,04 0,891 0,80
63 111,0 2,045 0,905 0,86
64 113,0 2,05 0,920 0,93
65 115,0 2,06 0,935 1,01
66 133,0 2,12 0,950 1,10
67 134,5 2,13 0,964 1,20
550
68 146,0 2,16 0,978 1,34

69 167,0 2,22 0,993 1,60
MeanCVb = 49,6 ¡P}

St. Dev. S = 37,5 MPa
Variation Coefficient Q = 75,5 %
Maximal Vb = 167,0 ¡P}
Minimal Vb =15,0 ¡P}
F Y
0,999 2
0,99
0,99
0,95 1
0,90
0,75
0,63 0
0,50
y = ln [-ln(1-F)]
0,30 -1
0,20
-2
0,10
0,05 -3
0,02
-4
0,010
-5
0,00
0,0025 -6
0,00 -7
1 1,2 1,4 1,6 1,8 2 2,2 2,4
lg V
Figure 9: Experimental Weibull plot of joint test results (4 – point and 3 – point bending)
for specimens damaged with cutting roller
(50,0…10,0 ; 4 – point bending);
– test results for low cutting force Q
(7,0…1,0 ; 4 – point bending);
(50,0…10,0 ; 3 – point bending);
– test results for undamaged specimens (4 – point bending)
551
Challenging Glass 3
4. Conclusion
The results of an experimental study of statistical distribution of bending strength values
obtained in the light of the effect of mechanical treatment and contact loads on glass
surface damaging and strength show that systematic deviation of the strength of float
glass elements from the Weibull statistic distribution is a regular phenomenon.
From the results it is concluded that:

x surface defects formed under glass processing and handing are the cause of the
multi- linear nature of the Weibull plots.
x shapes and sizes of fracture sources - critical micro-cracks and damages
influence directly on structural glass strength and its statistic parameters.
x statistical behavior of structural glass strength may be controlled by monitoring
of critical surface defects, conditions of their formation during production of
load carrying glass structural elements based on experimental strength data.
5. References
[1] Rodichev Yu., Tregubov N., Veer F., Maslov V., The features of fracture and engineering strength of
glass materials with surface, internal and mixed modes of critical defects, GPD 2011 Proc..- Tampere,
Finland, 2011
[2] Veer F., Rodichev Yu., Corrosion effects on soda lime glass, Proc. of Challenge glass 2.- 2010.-TU
Delft, Netherlands.
[3] Rodichev Yu, Tregubov N., Veer F., Maslov V., Influence of surface and internal defects on the
fracture and engineering strength of glassmaterials, Proc. of Int. conf. “In-service damage of materials,
its diagnostics and prediction”.-2011, Ternopil, Ukraine.
[4] Veer F.A., Rodichev Yu. M., The structural strength of glass; hidden damage, Strength of Materials
(Problems of Strength), Number 3, 2011, Pages 93 -109.
[5] Rodichev Yu., Yevplov Yu., Soroka H., Veer F., Tregubov N., Polivyany V., Surface Cracked Layer
and Damage Resistance of Glass under Contact Loading, Proc. of Challenge glass 3.- 2012.-TU Delft,
Netherlands.
552
doi:10.3233/978-1-61499-061-1-553
Load-Bearing Capacity of Thin Film

Photovoltaic Modules
Jens Schneider, Johannes Kuntsche, Jonas Kleuderlein
Technische Universität Darmstadt, Germany, www.iwmb.tu-darmstadt.de
The increasing use of thin film photovoltaic modules and the increasing occurrence
of damages raise the question about a sufficient safety level for the glass-glass-
modules used at present. This paper compares the experimental determination of
the load-bearing capacity used in the solar industry according to IEC 61646 and a
calculative method according to the German structural design standard for glass
structures (DIN 18008). Substantial differences in the safety level are pointed out.
Moreover, the load case „temperature“ can become relevant for the structural
design because of the high solar absorption of the modules. This influence is
currently not regulated in the standards. For typical installation situations, a
calculative consideration of this load case is discussed.
Keywords: Photovoltaic, Safety level, Temperature
1. Introduction
The use of photovoltaic (PV) modules is growing more and more due to falling prices of
the modules and governmental subventions. In 2010 Germany produced more than 11
billion kilowatt hours of electricity by photovoltaic systems according to the Federal
Ministry for Environment, Nature Conservation and Nuclear Safety (BMU) [1]. This
represents an increase by 77% compared to 2009.
Figure 1: Damage pattern in thin-film PV modules with glass-glass assembly
This is accompanied by the rising occurrence of damages at installed modules (Figure

1). Especially in thin-film modules with glass-glass assembly (Figure 2) increased cases
are noticed, where, due to glass breakage, the modules lose their function as a power
supplier and also become problematic due to a reduced residual load-bearing capacity of
553
Challenging Glass 3
the modules.
Upper Glass Upper Glass

Thin Film Lamination Foil
Lamination Foil Thin Film
Lower Glass Lower Glass
Figure 2: Systematic design of a thin-film module with glass-glass assembly
The solar industry is testing the performance of their products based on IEC 61646:
“Thin-film terrestrial photovoltaic (PV) modules - Design qualification and type
approval” [2]. Herewith, the achieved safety level for the load-bearing capacity of PV
modules that are currently in use is significantly below the usual safety level of the
German structural design standards in glass structures “Technische Regeln für
linienförmig gelagerte Verglasungen” (TRLV) or the new DIN 18008.
The draft standard to DIN VDE 0126-21: “Photovoltaik im Bauwesen – Photovoltaic in

building” [3] provides that PV modules must meet the requirements of DIN 18008,
depending on the application. In practice, however, the tests according to IEC 61646 are
considered to be sufficient so far.
Apart from the discussion of the standards, this article illustrates by examples, which
maximum load standard glass-glass PV modules can bear using computational design. It
is shown that a consistent application of the German structural design standards in glass
constructions would severe restrict the applicability of the already installed modules.
Framed and clamp-supported modules and modules with back rail systems are
compared.
2. Test according to IEC 61646

In the solar industry, thin film photovoltaic modules are tested according to IEC 61646,
which specifies the requirements for the testing of the modules. The term “type approval”
used in the title does not imply that by meeting the requirements of the standard, a
general technical approval for building constructions is achieved. The goal of the
standard is to determine the electrical and thermal characteristics of the tested modules.
However, the test series also include a mechanical load test, which is deemed to
determine the ability of the module to withstand wind, snow, ice or static loads.
A complete test series of thin-film PV modules according to IEC 61646 contains a total
of eight specimens. However, only one of these eight specimens is subjected to the
mechanical load test. In this test, the module is loaded cyclically. The load corresponds
to a constant surface load. The module is attached to a rigid support structure with the
manufacturer's prescribed method of attachment.
Firstly, the module is subjected to a uniform distributed compressive load of 2.4 kN/m2.
This load is hold for an hour. Then, without dismounting the module from the
substructure, the same load is applied as a suction load and also held over an hour. This
pressure-suction-cycle is repeated for a total of three times.
554
Load-Bearing Capacity of Thin Film Photovoltaic Modules
If big snow and ice accumulations are to be expected, the pressure load must be
increased to 5.4 kN/m2 in accordance to IEC 61646. But the installation situations
where this increased load are required are not defined. Moreover, it is pointed out that
even tougher test may be necessary if the tested loading does not cover the installation
situation. Therefore, theoretically a common design according to DIN 18008 or TRLV
would have to be performed. In practice, however, the common belief is that with
fulfilling the IEC 61646, the application range of any construction is covered.
In addition to the mechanical load test, IEC 61646 requires an thermic cycling test.
However, this test is only deemed to ensure the electric functionality and does not
consider sufficiently the glass-specific material behaviour under thermal loads.
3. Design method according to DIN 18008 and TRLV

The new German structural design standard in glass constructions DIN 18008 with the
already published parts [4] [5] describes the required analysis for the ultimate limit state
and the serviceability limit state of glass components. In contrast to the currently valid
German standard TRLV [6], where the concept of global safety factors is prescribed, for
the analysis according to DIN 18008, the concept of partial safety factors is specified.
The safety level aims at a failure probability of pf = 10-6 per year according to the
common construction standards, e.g. [9].
The computational load bearing capacity of glass constructions is proven by the fact that
the existing principal tensile stresses in the glass due to the loads according to DIN 1055
or EC 1 do not exceed the design strength of the glass. Generally, the design strength of
the glass Rd is determined by:
kmod kc fk (1)
Rd
JM
with:
x kmod: coefficient for consideration of the load duration of annealed float glass
(0.7 for wind, 0.4 for snow and 0.25 for dead weight),
x kc: coefficient for consideration the type of construction (1.0 or 1.8),
x fk: characteristic bending tensile strength (45 N/mm² for annealed glass,
70 N/mm² for heat-strengthened glass (HSG), 120 N/mm² for fully tempered
glass (FTG),
x M: partial safety factor for resistance (1.8 for annealed float glass, 1.5 for HSG
and FTG).
The strength of annealed float glass is highly dependent on the load duration due to the
subcritical crack growth. In DIN 18008, this is explicitly taken into account by the
coefficient kmod.
If tensile stresses occur at the free edge of the glass, for annealed float glass only 80%
of the glass strength can be applied due to the existing damage from the edge treatment.
If laminated glass (LG) or laminated safety glass (LSG) is used, the design value may
be increased by 10% (Table 1).
555
Challenging Glass 3
Table 1: Design values of glass strength according to DIN 18008 [4].

Rd [N/mm2]
Type of Glass Load duration kmod
monolithic + laminated + edge stress
long 0.25 6.3 6.9 5.5
Annealed float middle 0.4 10.0 11.0 8.8
short 0.7 17.5 19.3 15.4
HSG - - 46.6 51.3 51.3
FTG - - 80.0 88.0 88.0
DIN 18008 prescribes kc = 1.0 , unless otherwise specified. Only for vertical annealed
glazing with circumferential linear support kc = 1.8 may be applied. Below, the general
case kc = 1.0 is assumed, since PV modules are mainly used as horizontal glazing
(inclination to the vertical > 10°).
The determination of the design load values results from the respective valid standards
(DIN 1055 and DIN EN 1991).
Older regulations like the TRLV are based on the concept of global safety factors. Here
the load values are characteristic values. The resistance (glass strength) is reduced by a
global safety factor (Table 2).
Table 2: Allowable bending stress of according to TRLV [6].
Vzul [N/mm2]
Type of Glass
overhead vertical
Annealed float 12 18
Laminated (Annealed) 15 22.5
FTG 50 50
4. Comparison of the safety levels

according to IEC 61646 and TRLV or DIN 18 008
The IEC 61646 contains only the performance of one mechanical load test with a
pressure or suction load of 2.4 kN/m². The standard explains in the comments the origin
of this value with a wind pressure of 0.8 kN/m2, corresponding to a wind speed of
130 km/h and a safety factor of = 3.
This relationship between wind velocity and pressure can be confirmed by DIN 1055-4
[7]. However, a characteristic wind pressure of 0.8 kN/m2 is achieved already for a
building in wind zone 2 at a building height of 10 m to 18 m. This does not yet include
the cpe-factor, which takes into account the geometry of the building, the mounting
position in the building and the size of the loaded area.
556
Which actions (wind, dead weight, snow and ice loads, and combinations thereof) are
associated with the increased pressure load of 5.4 kN/m2 and which application should
be covered with this load is not described in IEC 61646.
However, the specified global safety factor of = 3 for the wind pressure of 0.8 kN/m2
is too low for a design based on only one test. A comparison with the safety level of the
TRLV demonstrates this. For horizontally mounted laminated glass made of annealed
glass, the global safety factor is determined by:
N
45
fk mm2 (2)
J global 3.0
V zul 15
N
mm2
This global safety factor is based on the characteristic value (5%-fractile) of the glass
strength. According to IEC 61646, the safety factor is, however, based on a mechanical
load test with only one test specimen. The (high) scattering of the glass strength is
therefore not considered. If, in addition to the 5%-fractile value, the mean value of
naturally aged annealed glass with fm = 80 N/mm² [8] and the 95%-fractile are estimated,
this results in the global safety factors according to Table 3.
Table 3: Safety factors according to TRLV for horizontal mounted LSG made of annealed float glass.
Strength of the tested 5%-fractile value mean value 95%-fractile value

glass [N/mm2] 45 80 115
global safety factor 3,0 5,3 7,7
Comparison to
100% 178% 256%
IEC 61646
This comparison can also be made with the new DIN 18008 to take into account the
concept of partial safety factors. For laminated safety glass made of annealed float glass,
the design value for the resistance under wind load results in:
N
0.7 1.0 45
kmod kc fk mm2 N (3)
Rd 1.1 1.1 19
JM 1.8 mm2
This, however, only considers the resistance side. The partial safety factor for the load
(wind) according to DIN 1055-100 [9] is Q = 1.5. This must be multiplied with the
partial safety factor for resistance to allow a comparison with the global safety factor.
Taking into account the scattering of the glass strength, the required safety factors are
shown in Table 4.
557
Challenging Glass 3
Table 4: Safety factors according to DIN 18008 for LSG made of annealed float glass under wind load.
Strength of the tested 5%-fractile value mean value 95%-fractile value

glass [N/mm2] 45 80 115
partial safety factor 2,4 4,2 6,1
global safety factor 3,6 6,3 9,1
Comparison to
118% 211% 303%
IEC 61646
The comparisons for global and partial safety concept show similar results. The
implementation of one mechanical load test according to IEC 61646 with the stated
global safety factor of = 3 does not provide a comparable level of safety to TRLV or
DIN 18008. If a glass with a high strength (e.g. 95%-fractile) would be tested, a global
factor of = 9.1 would be needed to meet the safety level of DIN 18008.
Moreover, the IEC 61646 does not consider sufficiently the time-dependent behaviour
of the strength of annealed float glass. The load duration for the mechanical load test is
longer than the short-term load (wind, 10 minutes) but well below the middle-term load
(snow, 30 days). Moreover, as the tests are carried out at room temperature, a certain
shear transfer between the upper and the lower glass is active due to the lamination foil.
This shear transfer does not exist in the real installation situation under solar radiation
with the usual viscoelastic lamination foils (PVB, EVA).
Basically, a proof of the bearing capacity with mechanical load tests is possible, but
should be based on the fundamentals of structural design. DIN EN 1990 [10] describes
in Appendix D how to perform a test-based design of construction elements. Thereafter,
with a test sequence which includes the real actions (storage conditions, load type and
duration, temperature, etc.) and a sufficient number of specimens, the design value of
the resistance can be determined. This can subsequently compared with the actions for
the particular installation situation.
5. Load-bearing capacity according to DIN 18008 for typical module assemblies

Photovoltaic modules are offered in different designs. Mostly, framed modules are used.
Here, the glass-glass module is held by a circumferential aluminium frame which is
fixed to the support structure. Unframed modules can be held either locally by clamping
brackets or linearly by using a back rail system which is glued to the rear glass of the
modules. The systems, each with typical dimensions, are shown in Figure 3.
558
Figure 3: Typical support structures of glass-glass modules, a: framed, b: clamped , c: back rail supported
A laminated glass with 2 x 3.2 mm float glass is used for the calculation below, which is
usual in the market. Thus, these types of modules represent assemblies, which have
passed in identical or similar design the test according to IEC 61646 and are currently
used in construction industry.
According to DIN 18008, the positive effect of shear resistance is not taken into account
in the calculation. The design value of resistance Rd is given by Table 1.
Figure 4: Finite Element Model for glass-glass modules,

a: clamped (SJ MEPLA), b: back rail supported (ANSYS 13)
The aluminium back rail profile, which is shown in Fig. 4, is glued to the rear glass of
the module by an adhesive tape. The adhesive tape has been simplified to a linear-
elastic material model (Young’s modulus E = 1.0 N/mm²; Poisson's ratio = 0.4). The
calculation takes into account geometrically nonlinear behaviour of the glass-glass
laminate (Figure 4).
The calculation gives the load capacity at which the glass strength Rd is reached. This
load corresponds to the design value of the action load Ed in the ultimate limit state:
Ed d Rd (4)
In order to obtain the characteristic load capacity Ek, the pre-determined design values
559
Challenging Glass 3
are divided by the partial safety factor according to the exposure time for permanent
load (e.g. dead weight) G = 1.35 and for short and medium loads (e.g. snow, wind)
Q = 1.5. In accordance with DIN 18008, the higher value of kmod is used when
considering load combinations. For example, in the combination of dead weight plus
wind kmod is set to 0.7. The results for the determined characteristic loads Ek are shown
in Figure .
Figure 5: Characteristic load capacity of typical PV modules
The comparison of the characteristic load capacity in buildings with usual wind and
snow loads shows, that the application of these modules is highly limited.
The framed module can bear about 0.6 kN/m2 in addition to its dead weight for a
medium load duration (snow) and about 1.3 kN/m² for a short load duration (wind). A
comparison of these values to the characteristic loads in accordance to DIN 1055
demonstrates the limited application possibilities of these modules. For example, a
building in snow load zone 2 (about 80% of Germany) with a roof slope of up to 30°
must withstand a characteristic minimum snow load on the roof of s1 = 0.7 kN/m²
according to DIN 1055-5 [11]. This already exceeds the characteristic load capacity of
the module.
The module which is clamped at four points reaches its full bearing capacity already
under its dead weight. With the approach of the lower value for kmod for combined loads
(dead weight plus snow or wind) after DIN 18008, the additional load capacity is so low
that it is hard to find a possible application.
The back rail supported system shows the best structural behaviour of the presented
systems. Despite the larger dimensions of the module, the bearing capacity is only
slightly below the framed system.
A neglected aspect in the examples is the position of the connection socket. This can be
located at the edge or on the surface area of the module. In an assembly on the surface
area, a hole in the rear glass is required. This hole reduces, depending on its location,
the load capacity of the modules again significantly. According to DIN 18008, holes in
float glass are not permitted, so the drilled rear glass would have to be thermally
toughened.
560
Furthermore, it has to be discussed, whether an increased coefficient for consideration

the type of construction kc may be applied for framed PV modules in conjunction with a
proof of sufficient residual load-bearing capacity. This would reflect the fact that the PV
modules are often mounted on existing roof structures or on a field and thus not directly
provide a risk on persons if they have sufficient residual load-bearing capacity after
glass berakage.
6. Load case temperature

The influence of solar radiation on photovoltaic modules compared to transparent
façade glazing is much higher due to the fact that the absorption coefficient increases.
Uncoated glasses have a solar absorption of approximately 8-20 % depending on the
glass thickness [12]. A thin film photovoltaic module (ASI), however, has an
absorbance between 60 % and 90 %.
The solar absorption basically leads to two superposed temperature profiles over the
module thickness. A constant temperature load can be superimposed with a variable
temperature profile over the cross section (Fig. 6).
Figure 6: Separation of the thermal stress in a constant and variable part
The constant temperature Tc only leads to compressive stress in the glass for framed
modules with an idealized system (constant linear horizontal support of the edges).
For locally clamped modules and temperatures up to 50 °C (compare [13]) with a 2 mm

elastic support of the glass edge and soft contact materials (Young's modulus = 5 –
200 N/mm², = 0.5, e.g. EPDM), no design-relevant tensile stresses occur, too. Only
for very stiff contact materials (Young's modulus = 1 000 – 3 000 N/mm², = 0.3 - 0.4,
e.g. POM), the principal tensile stresses would be in a design-relevant dimension.
For back rail supported modules, no significant tensile stresses are caused by constant
heating of the glass sheets due to the flexibility of the overall system.
With a variable temperature profile, the (suppressed) bending of the glass panes leads to
tensile stresses, which may be relevant to structural design.
A parametric study using the above framed module with two types of a variable
temperature profile shows this clearly. In the first variant (symmetric temperature
distribution), the upper and lower sides of the laminate are cooler than the thin film and
561
Challenging Glass 3
the lamination foil in the middle due to the ambient atmosphere. In the second variant
(asymmetric temperature distribution), a reduced cooling of the lower glass panel was
taken into account and therefore a constant temperature distribution assumed (Figure 7).
Figure 7: Investigated temperature distributions (symmetrical and asymmetrical)
For the frame, the limits "soft" and "rigid" were investigated. Here "soft" means free
glass edges (x, y-and z-direction), "rigid" means simply supported edges of both glass
panes in z-direction.
The interlayer is modeled with a Young's modulus of 1.0 N/mm² and a Poisson's ratio of
= 0.499. The temperature difference 'T was increased progressively up to 20 °C (e.g.
rain cooling). Figure 8 shows the stress distribution of the finite-element calculations
which takes into account geometrical nonlinearity.
Figure 8: Principle tensile stresses on the outer surface of the upper glass pane with symmetrical (left) and
asymmetrical (right) temperature distribution and soft frame
The maximum principal tensile stresses occur under symmetrical load T,1 at the outer
surfaces. With the asymmetrical load 'T,2 these are located on the outer surface of the
upper glass pane (Figure 9).
Figure 9: Thermal expansion (¤therm) and tensile stresses (max) for symmetrical (left) and asymmetrical (right)
temperature distribution
562
Figure 10 shows an almost linear relation between temperature difference and the glass
stresses. The small differences arise from the geometrically nonlinear analysis.
Furthermore, it indicates that the consideration of the load case temperature can be quite
relevant for structural design, especially if the capacity of the modules is already
reached with other loads.
Figure 10: Principal tensile stresses on the glass surface as a function of the temperature distribution and the
stiffness of the frame
7. Summary and Outlook

The mechanical load test according to IEC 61646 with only one specimen leads to a
much lower safety level than the application of the design standards for glass
constructions (TRLV, DIN 18008).
To consider the scattering of the glass strength, the number of specimens should be
increased significantly to reach a comparable safety level to DIN 18008 (see Table 3
and 4). Instead, according to IEC 61646, only a safety factor of = 3 is taken into
account referenced to a characteristic load of 0.8 kN/m². Thus, the allowable installation
and application situations that may be covered by the staggered test load according to
IEC 61646 with 2.4 kN/m² and 5.4 kN/m² have to be discussed.
Nevertheless, it is common practice to install PV-modules, which were tested according

to IEC 61646 without any additional structural calculations. The consequences are
already evident in numerous cases of damages. Only the positive post-breakage
behaviour of laminated glass ensures that the damages did not cause greater hazards to
people so far.
It was shown that temperature loads and especially variable temperature profiles over
the cross-section can lead to design relevant principal tensile stresses.
563
Challenging Glass 3
For the future, besides the need of a suitable structural design method for PV-modules,
the post-breakage behaviour of laminated glass has to be analysed systematically to
develop appropriate mechanical models. Moreover, specific loads on PV-modules, such
as thermal stresses due to the high solar absorbance of the thin film and the associated
thermal stresses in the glass, have to be investigated further.
8. References
[1] Bundesministerium für Umwelt, Naturschutz und Reaktorsicherheit: "Zeitreihen zur Entwicklung der
erneuerbaren Energien in Deutschland ", 12/13/2011.
[2] DIN EN IEC 61646: “Thin-film terrestrial photovoltaic (PV) modules - Design qualification and type
approval", 2009.
[3] DIN VDE 0126-21: “Photovoltaic in building (draft)", 2007.
[4] DIN 18008-1: Glass in Building - Design and construction rules - Part 1: Terms and construction rules",
2010.
[5] DIN 18008-2: “Glass in Building - Design and Construction Rules - Part 2: Linearly supported glazing",
2010.
[6] DIBt: "Technische Regeln für die Verwendung von linienförmig gelagerten Verglasungen (TRLV)",
2006.
[7] DIN 1055-4: “Action on structures – Part 4: Wind loads", 2005.
[8] Fink, Andreas: "Ein Beitrag zum Einsatz von Floatglas als dauerhaft tragender Konstruktionswerkstoff
im Bauwesen", Dissertation, TU Darmstadt, 2000.
[9] DIN 1055-100: “Actions on structures - Basis of design, safety concept and design rules", 2001.
[10] DIN EN 1990: Eurocode: Basis of structural design, German version EN 1990:2002 + A1:2005 +
A1:2005/AC: 2010", 2010.
[11] DIN 1055-5: “Actions on structures – Part 5: Snowloads and ice loads", 2005.
[12] Interpane Glas Industrie AG: „Gestalten mit Glas“, 8. überarbeitete Auflage, Lauenförde, 2011.
[13] Haase, W. et al.: “Temperaturversuche an Verbundsicherheitsglasscheiben unter Sonneneinstrahlung,
Versuchsbericht des Zentrallabors des konstruktiven Ingenieurbaus”, Stuttgart, 2001.
564
doi:10.3233/978-1-61499-061-1-565
Stress Distribution of Tension Structural

Glass Plates
Tomas Serafinaviius, Gintas Šauiuvnas
Vilnius Gediminas Technical University, Department of Steel and Timber
Structures, Lithuania, tomas.serafinavicius@vgtu.lt, gintas.sauciuvenas@vgtu.lt
www.vgtu.lt
Jörg Hildebrand
Bauhaus-Universität Weimar, Fakultät Bauingenieurwesen, Institut für
Konstruktiven Ingenieurbau, Weimar, Germany, joerg.hildebrand@uni-weimar.de,
www.uni-weimar.de
This paper deals with the results received by carried out tensile tests of the
structural glass plates supplied with three different types of stress raisors. The first
type of specimens is the plate with a circular hole located in the centre of the plate,
the second one is the plate with a slot hole located in the centre of the plate, and the
third one is the plate with the semicircular notches located at the opposite edges.
The results of test of these specimens were compared both among themselves and
both with the results of the specimens without stress raisors. The specimens were
made from tempered glass of 6 mm thick with the arrising edges. All tests were
carried out under equal conditions; the load level has been raised within each 1.0
kN. The experiment was performed with an optical device for measurement of the
stresses by birefringence. Detailed analysis of the obtained results is given in this
paper.
Keywords: Structural glass, Tempered glass, Stress raisors, Tensile stress, Stress
distribution, Stress birefringence, Fracture.
1. Introduction
Although in the glass structures are different types of joints, such as glued, 'Hagl' [1],
nevertheless the practice shows that the bolted connections are very often used in glass
structural assembly 'McDonnell & Thompson' [2], 'Kooymans' [3]. In Germany, the
technical rules for glass construction are available for point supported glazing 'TRPV'
[4] and for connection design 'Feldmann et al.' [5]. Baitinger and Feldmann [6] have
presented the calculation method of bolted glass connections, but these authors have not
carried out the analysis of stress area distribution. Beyer [7] has explored the specific
FZP-G-Z point fitting. Investigations of influence of experimental test type on the
determination of probabilistic stress distribution as well are done in 'Huerta et al.' [8].
There are also investigated residual stresses in tempered glass plates 'Anton et al.' [9]
and 'Nielsen et al.' [10], but these authors did not explore the influence of stress raisors
at different loads. The holes drilled in glass for bolted connectors lead to the stress
concentration; and the behaviour of bolted connection with two holes in-line is explored
in 'Eliasova' [11]. As it is seen from the above presented references influence of the
stress raisors on the load bearing capacity and stress distribution of structural glass
elements is not presented. Therefore, this paper is focused on the stress distribution in
565
Challenging Glass 3
tension glass plates based on the experiments with non-destructive measuring of surface
stresses, numerical simulation and as well the analytical analysis.
The tensile test was chosen due to the fact that the glass is considerably more sensitive
to tensile stresses 'Norville & Morse' [12] and such cases are less studied. Glass plate
with concentrator and without steel pin are less common, but it could be possible. For
example, the hole in glass beam for steel string. Stress concentrators types is selected
as close as possible to practical situations and also take into account of case in literature.
Tempered glass is selected because in construction practice it is applied more widely for
load bearing structures.
2. Test Methods
An universal testing machine Lukas for tensile tests was used. Stress measurements
were done with a machine Ilis StrainMatic M4/140 inline. Equipment operation was
performed with the software Ilis Strain Matic. Overall view of test equipment is given in
Figure 1. All experiments were carried out at Institute of Structural Engineering of
Faculty of Civil Engineering of Bauhaus-University Weimar. The analysis of
experimental data was carried out at Department of Steel and Timber Structures of
Faculty of Civil Engineering of Vilnius Gediminas Technical University.
Figure 1. The overall view of test machine.
The specimen placed in the sample compartment is illuminated with linearly polarized
light which is generated by a quasi monochromatic light source and a polarizer. This
linearly polarized light is converted into elliptically polarized light by birefringence in
the sample. A retarder converts the elliptically polarized light back into the linearly
polarized light. After a retarder the polarization plane is rotated compared with the
original polarization direction by an angle from which the different birefringence values
of the sample can be derived. The polarizer can be rotated during a measurement. The
intensity over the entire measuring range is recorded with a CCD-camera under
different polarizer settings (see Figure 2). The step width of the polarizer is given in
angle degrees and can be selected between 5° and 90° according to the orientation of the
expected stresses. If a polarizer step width of 15° is selected, for example, result images
are calculated for polarizer positions of 0°, 15°, 30°, 45°, 60° and 75° 'StrainMatic' [13].
In the interpolated mode the result image is not calculated by superposition (maximum
value) of the sub-images but by interpolation of three result images that have been taken
566
Stress Distribution of Tension Structural Glass Plates
at certain polarizer positions. For the relatively small measuring values these method
achieves the best results.
Figure 2. Stress intensity distribution at different polarizer positions.
The result images are evaluated automatically and the resulting stress distribution over
all measuring directions is displayed colour-coded in the operating software Ilis
StrainViewer. The legend shows the colour-coded percentage distribution of the
measured values. By default, red areas in the legend and in the image represent high
values, yellow / green areas average values and blue areas low values 'StrainMatic' [14].
3. Experimental Tests
All specimens are done from tempered glass of 6 mm thick. The force level has been
raised within each 1.0 kN. The results in pictures are given each 2.0 kN. The
comparison has been made between the three different types of stress raisors: plate with
a circular symmetric hole (4 pcs. of specimens), plate with slot hole (4 pcs. of
specimens), and plate with semicircular notches at opposite edges (4 pcs. of specimens).
As well results of above mentioned plates were compared with the specimens without
the stress raisors (4 pcs. of specimens). Stress measurements were performed in the
seven polarizer positions of 0o, 15o, 30o, 45o, 60o, 75o, 90o. Measurements were carried
out every 1.0 kN. From the value of load 9.0 kN the measurements were performed
every 0.5 kN. One step in the stress measurements on average took 3 minutes. In Tables
1, 2, 3, 4 only the maximum values of load at which the stress measurement was carried
out are given. The failure results are not significant for the topic, and in this paper the
tables and figures are not provided.
3.1. Plate with a Symmetrically Located Circular Hole

The circular type drilled holes are one of the most commonly used in fixing the glass
structures. For this reason, this type was chosen as a main stress raisor for tests, as well
as possibility to use an existing analytic calculation example for definition of the classic
stress concentration coefficient. Drawing of the specimen with a symmetrically located
circular hole including its dimensions and a real image of the specimen are given in
Figure 3. Cross-sectional areas of all specimens and the maximum values of the tensile
stresses as well as the values of the maximum load at which last it was possible to
perform the measurement are given in Table 1.
567
Challenging Glass 3
a) b)
Figure 3. Drawing of the specimen (a) and its real image (b).
Table 1. Resume of the measured values for the specimens T-1-(1–4).

Specimen Cross-sectional area [mm2] Thickness [mm] Max tensile stress [MPa] Max load F [kN]
T-1-1 415.18 5.90 47.6 9.5
T-1-2 415.00 5.89 47.1 9.5
T-1-3 412.05 5.88 47.3 10.0
T-1-4 412.90 5.87 45.9 9.5
a) b) c)
d) e) f)
Figure 4. Stress distribution at load values: 0.0 kN (a), 2.0 kN (b), 4.0 kN (c), 6.0 kN (d), 8.0 kN (e), 9.5 kN (f).
The measured results of the stress distributions are graphically presented in Figure 4 (a–
f). The hole centres are not located exactly in the centres of the specimens, and for this
reason the visible differences between the results of the left and right sides of specimens
568
are seen. The comparative examination of test results has been performed at the value
8.0 kN of tensile loads.
3.2. Plate with Symmetrically Located Slot Hole

In structural glass facades very often the slot hole are used for possible compensation of
the vertical and horizontal displacements. The influence of these slot holes are much
less investigated theoretically in comparison with a circular holes. Drawing of the
specimen with a symmetrically located slot hole including its dimensions and a real
image of the specimen are given in Figure 5. Cross-sectional areas of all specimens and
the maximum values of the tensile stresses as well as the values of the maximum load at
which last it was possible to perform the measurement are given in Table 2.
a) b)

Specimen Cross-section area Thickness Max tensile stress Max load F [kN]
[mm2] [mm] [MPa]
T-2-1 415.22 5.87 43.3 9.5
T-2-2 414.93 5.90 42.8 9.0

T-2-3 413.86 5.87 44.7 10.0
T-2-4 414.68 5.88 46.5 10.0
569
Challenging Glass 3
a) b) c)
d) e) f)
Figure 6. Stress distribution at load values: 0.0 kN (a), 2.0 kN (b), 4.0 kN (c),
6.0 kN (d), 8.0 kN (e), 9.0 kN (f).
The slot hole centres are not located exactly in the centres of the specimens, and this
also affects the distribution of the tensile stresses. The mean value of the maximum
tensile stress at load level of 8.0 kN is 39.6 MPa, and it is slightly lower than that of
specimen with a circular hole. The measured results of the stress distributions are
graphically presented in Figure 6 a–f.
3.3. Plate with the Semicircular Notches Located at the Opposite Edges
In some cases the glass plates may be fitted just through the edges of the plate. This has
led to the selection of stress raisors for exploration. Drawing of the specimen with a
semicircular notches located at the opposite edges including its dimensions and a real
image of the specimen are given in Figure 7 a, b. The measured results of stress
distributions are graphically presented in Figure 8 a–f. Resume of all specimens and the
maximum values of the tensile stresses and values of the load are given in Table 3.
Particular attention focuses residual stresses at the peaks and sharp edges. During the
load steps the residual stresses at the edges did not grow and was almost constant
around 39.5 MPa. Practical use of the options needs to avoid sharp and peak edges as
they may lead to collapse without any load.
570
a) b)
a) b) c)
d) e) f)
Figure 8. Stress distribution at load values: 0.0 kN (a), 2.0 kN (b), 4.0 kN (c), 6.0 kN (d),
8.0 kN (e), 10.0 kN (f).

Specimen Cross-sectional area Thickness Max tensile stress Max load F [kN]
[mm2] [mm] [MPa]
T-3-1 410.37 5.87 42.7 9.0
T-3-2 409.85 5.86 43.5 9.0
T-3-3 410.96 5.87 44.3 10.0
T-3-4 410.55 5.86 41.5 9.5
571
Challenging Glass 3
3.4. Plate without Stress Raisors

This type of specimens without the stress raisors was chosen for the comparative
analysis and as well as for the measurement of the mean stress value. Drawing of the
specimen without the stress raisors including its dimensions and a real image of the
specimen are given in Figure 9 a, b. Resume of all specimens and the maximum values
of the tensile stresses as well as the values of the maximum load at which last it was
possible to perform the measurement are given in Table 4.
The measured results of the stress distributions are graphically presented in Figure 10.
The mean value of the maximum tensile stress at load level 8.0 kN is 20.5 MPa.
a) b)

Specimen Cross-sectional area Thickness Max tensile stress Max load F [kN]
[mm2] [mm] [MPa]
T-4-1 555.07 5.89 26.7 10.0
T-4-2 586.35 5.87 28.8 10.5

T-4-3 589.05 5.87 26.3 10.0
T-4-4 586.44 5.86 29.1 10.5
572
a) b) c)
d) e) f)
Figure 10. Stress distribution at: 0.0 kN (a), 2.0 kN (b), 4.0 kN (c), 6.0 kN (d), 8.0 kN (e),
the last load 10.5 kN (f).
4. Analytical Stress Concentration Factors

It was investigated that stress concentrators significantly reduces the tensile elements
bearing capacity 'Pilkey' [15]. Extent of the reduction in bearing capacity depends on
various factors such as: stress raisors shape, the position of the element, the distance
between the stress raisors and type of the material – plastic or brittle one. The collapse
of plastic material is not as immediate as of a brittle material in the stress concentration
locations. The brittle materials collapse suddenly of the event of crack when the stress
reaches the material tensile strength. The resistance effect of the tempered glass is
visible not only in stress raisors areas, but also because the residual stresses after
tempering of the specimen and preparation of specimen edges.
There were examined three types of the stress raisors – plate with a symmetrically
located circular hole, plate with a symmetrically located slot hole and plate with a
semicircular notches located at the opposite edges. The aim of analysis is determining
whether the theoretical assumptions could be applied in order to determine the bearing
capacity of tempered glass plates with the stress raisors and finding out the impact of
residual stresses on the bearing capacity.
The values of the stress concentration factors Ktn and Ktg calculated by expressions and
relationships according to 'Pilkey' [15] and 'Faurie' [16] are listed in Table 5.
573
Challenging Glass 3
Table 5. Stress concentration factors.

Geometry of stress raisors Factor Ktn Factor Ktg
Plate with a symmetrically located circular hole 2.37 3.37
Plate with a symmetrically located slot hole 3.08 3.78
Plate with a semicircular notches located at the opposite edges 2.13 3.00
5. Numerical Simulation
Finite element modelling is widely applied for the analysis of various structures and
elements of the elastic-plastic materials. It allows obtaining the accurate results in
comparison with experimental results. Therefore, FEM was used in this analysis.
Numerical simulations were performed with the finite element SolidWorks software.
For modelling of structural glass plates the shell-type finite elements were used.
Because of computing resources and computational time saving just symmetric half of
the specimen was modelled.
The solution of problem using a linear analysis and non-linear analysis was objective to
determine whether the use of different methods will get the same results. The
consideration was made between the three types of the stress raisors described above.
The finite element mesh was generated such denser at that the stress raisors, see Figures
11 a, 12 a, 13 a. The stress distribution in plates was analyzed under a tensile load of 8.0
kN. The obtained character of stress distribution obtained by a linear analysis and non-
linear analysis for plates of the three types of the stress raisors is presented in Figures 11
b, c; 12 b, c and 13 b, c. The values of stress concentration factors were compared with
the experimental data and the results obtained by using the analytical expressions; the
comparison results are presented in Table 6. The results of the linear analysis showed
that the values of stress concentration factors are very large (for plates with circular
holes these values were 1.6 times higher than calculated analytically and 1.2 times
higher than for plates with a symmetrically located slot holes and plates with a
semicircular notches located at the opposite edges).
Due to the large difference between the values of stress concentration factors non-linear
analysis of the same plates was performed. The values of the stress concentration factors
by the non-linear analysis were significantly decreased and were even up to 1.6 times
smaller than the other way. If a linear analysis is conservative, it means that a non-linear
analysis at that case is in the opposite.
574
a) b) c)
Figure 11. FEA mesh (a), stress distribution due to linear model at load 8.0 kN (b), stress distribution
due to non-linear model 8.0 kN (c) for plate with a symmetrically located circular hole
a) b) c)
Figure 12. FEA mesh (a), stress distribution due to linear model at load 8.0 kN (b), stress distribution
due to non-linear model at load 8.0 kN (c) for glass plate with a symmetrically located slot hole.
a) b) c)
Figure 13. FEA mesh (a), stress distribution due to linear model at load 8.0 kN (b), stress distribution due to
non-linear model at load 8.0 kN (c) for glass plate with a semicircular notches located at the opposite edges.
575
Challenging Glass 3
6. Comparison
The comparison of obtained data is presented in Table 6 and the comparative
examination has been performed at 8.0 kN tensile load. Comparison was made between
three types of stress raisors that were mentioned earlier. The values of concentration
coefficients are compared between the values obtained by the analytical calculations,
the experiments and the numerical simulation.
Table 6. Mean values of the stress concentration factors at 8.0 kN tensile load.
Analytical Exp. I Exp. II Exp. III Num. L Num. NL
Geometry of
Ktn= Ktg = Ktn,II= Ktn,III =
stress raisors Ktn Ktg Kt,L Kt,LN
max / n max / g max / min max,r / min,r
Plate with a
symmetrically
2.37 3.37 2.15 3.08 2.40 2.34 3.67 2.09
located
circular hole
Plate with a
symmetrically
3.08 3.78 2.08 2.98 2.45 2.38 3.71 1.92
located
slot hole
Plate with a
semicircular
notches located 2.13 3.00 1.64 2.35 1.88 1.95 2.61 1.44
at the opposite
edges
Notes:
n = N/(t(b-d));
g = N/(t×b);
t, b, d – mean values of specimens cross-sectional dimensions;
max – max stress value of net section;
min – min stress value of net section;
max,r – max stress value of net section reduced by residual one;
min,r – min stress value of net section reduced by residual one;

L – linear analysis;
NL – non-linear analysis
As shown in the Table 6, the values of the analytical and experimental stress
concentration factors are not always close. The values of stress concentration factors for
plates with the circular hole obtained by the analytical calculation and testing are
practically identical. The differences form of stress concentration coefficient values of
the discrepancy is enough large. Mismatches might be explained as an uneven residual
stress distribution of tempering attenuated glass panels. It could be because the fact that
the slot hole is nearly circular form of hole and therefore does not have a significant
effect on the hole in which the length and width ratio is increased. Although the stress
concentrator and the values do not match, but the character of the stress distribution in
all cases is close to the theoretical distribution character. The linear and non-linear finite
element modelling and experimental results obtained by measurements of the
discrepancy may have an impact on the limited FEA software opportunities.
576
7. Conclusion
x Investigated the stress raisors, which usually occurs in glass structures, the
effect of stress distribution and the impact of the resistance.
x The experimentally determined stress distribution compared with the analytical
obtained expressions and numerical simulation results.
x It was found that a small length of the slot holes on the stress distribution
differs only marginally from the circular hole effect.
x Modeling with numerical methods and using the linear or non-linear
calculation method provides different results.
8. Acknowledgements
Acknowledgements to Research Council of Lithuania and Bauhaus-Universität Weimar
for funding of research plan. The specials acknowledgements are expressed to company
"GLASSBEL Baltic" www.glassbel.com for production of glass specimens.
Acknowledgements to COST Action TU0905 "Structural Glass – Novel design methods
and next generation product" for targeted information and valuable networking.
577
Challenging Glass 3
9. References
[1] Hagl, Anneliese, Bonded Point-Supports: Understanding Today – Optimizing for the Future,
Proceedings of the Conference on Architectural and Structural Applications of Glass (Challenging Glass
2), Delft, the Netherlands, 2010.
[2] McDonnell, Terry R; Thompson, David, Structural Glass Observation Boxes (Willis Tower Ledge),
Proceedings of the Conference on Architectural and Structural Applications of Glass (Challenging Glass
2), Delft, the Netherlands, 2010.
[3] Kooymans, John, Long span glass fin design, Proceedings of the Conference on Architectural and
Structural Applications of Glass (Challenging Glass), Delft, the Netherlands, 2008.
[4] TRPV, Technische Regeln für die Bemessung und die Ausführung punktförmig gelagerter Verglasungen
[Technical Rules for point supported glazing]. Mittlg. DIBt, Berlin, 2006.
[5] Feldmann, Markus; Kasper, Ruth; Pilsl, Mascha, Glass for Structural Applications- the development of
the connection design, Proceedings of the Conference on Architectural and Structural Applications of
Glass (Challenging Glass), Delft, the Netherlands, 2008.
[6] Baitinger, Mascha; Feldmann, Markus, Design concept for bolted Glass, Proceedings of the Conference
on Architectural and Structural Applications of Glass (Challenging Glass 2), Delft, the Netherlands,
2010.
[7] Beyer, Jorg, A new Concept for the Design of Structural Glass supported by Point Fittings, Proceedings
of the Conference on Architectural and Structural Applications of Glass (Challenging Glass), Delft, the
Netherlands, 2008.
[8] Huerta, Consuelo; Pacios-Alvarez, Antonia; Lamela-Rey, Maria-Jesus; Fernandez-Canteli, Alfonso,
Influence of experimental test type on the determination of probabilistic stress distribution, Proceedings
of the International Conference on Architectural and Automotive Glass (Glass Performance Days),
Tampere, Finland, 2011.
[9] Anton, Julian; Errapart, Andrei; Paemurru, Mart; Lochegnies, Dominique; Hodemann, Siim; Aben,
Hillar, On the Inhomogeneity of Residual Stresses in Tempered Glass Panels, Proceedings of the
International Conference on Architectural and Automotive Glass (Glass Performance Days), Tampere,
Finland, 2011.
[10] Nielsen, Jens Henrik; Olesen, John Forbes; Stang, Henrik, Experimental Investigation of Residual
Stresses in Toughened Glass, Proceedings of the Conference on Architectural and Structural
Applications of Glass (Challenging Glass), Delft, the Netherlands, 2008.
[11] Eliasova, Vencl, Bolted Connections of Glass Structures by means of Two Holes In-line, Proceedings of
the Conference on Architectural and Structural Applications of Glass (Challenging Glass), Delft, the
Netherlands, 2008.
[12] Norville, Scott; Morse, Stephen, Maximum Principal Stress and Probability of Breakage for Glass in
Buildings, Proceedings of the International Conference on Architectural and Automotive Glass (Glass
Performance Days), Tampere, Finland, 2011.
[13] StrainMatic, Basic Operation and Working Procedures. Ilis gmbh, Erlangen, Germany, 2007.
[14] StrainMatic, Reference of the Operating Software. Ilis gmbh, Erlangen, Germany, 2007.
[15] Pilkey, Walter D., Peterson's stress concentration factors. John Wiley & Sons, Inc, Hoboken, NJ, USA,
1997.
[16] Faurie, J. P., Guide du dessinateur: Les concentrations de contraintes, Cetim, France, 1997.
578
doi:10.3233/978-1-61499-061-1-579
Polishing Glass With Fluid Jet Technology

Christina Stålhandske, Maria Lang, Marita Jansson
Glafo – the Glass Research Institute, Sweden, Christina.Stalhandske@glafo.se,
www.glafo.se
Kent Persson
Lund University, Sweden, Kent.Persson@construction.lth.se
Anders Jönsson
Blekinge Institute of Technology, Sweden, anders.jonsson@bth.se, www.bth.se/waterjet
A fluid jet with 50% by weight of aluminum oxide is used to polish float glass.
Both water cut edges and holes as well as drilled holes are polished. Using a
pressure of 3.5 MPa, a 1.26 mm nozzle and 64 passes of the jet over the surface a
bright appearance is achieved. The polishing result is evaluated both with the naked
eye and an optical profiler. During mechanical testing of untreated holes it was
found that a drilled hole can withstand 20% higher forces than a water cut hole
before failure. The condition of the edge where the tensile stresses appear is very
important and future work includes polishing this edge.
Keywords: Glass, Optical profiler, Mechanical strength, Surface analyses,

Polishing.
1. Introduction
Glass is a brittle material but strong in compression. Durability and strength are
important properties and glass is found in more and more load bearing applications. The
very strong pristine glass degrades during handling and processing. The introduced
cracks have a negative effect on glass strength.[1] By polishing, surface and subsurface
micro-cracks are removed and some strength will be regained. [2-4] An unpolished hole
in a glass pane is the weakest spot and by polishing, the strength of the whole structure
can be increased. The edge of float glass and larger holes are polished today but cut outs
of complicated shapes or small holes are not so easily polished. An investigation by
Booij showed that glass could be polished by fluid jet polishing [5]. In our project
“Polishing glass by means of water jet technology – cross disciplinary collaboration for
sustainable development” we combining knowledge from the glass manufacturing
industry and the water jet technology. We have designed a polishing tool that is flexible
and has the potential of polishing odd shapes and small holes. The polishing tool
operates at pressures of around 5 MPa while cutting utilizes 100 times higher pressures.
It is thus a much gentler process and less material is removed. The advantaged
compared to laser polishing is that it is performed at room temperature and thus
minimizes the risk for thermal stresses.
Cerium oxide is a very common polishing agent for glass in the industry. The main
body of this investigation has addressed cerium oxide and aluminum oxide and found
that the latter is more efficient. Aluminum oxide is also cheaper and a better choice
579
Challenging Glass 3
from sustainability point of view. The work presented here only addresses aluminum
oxide polishing.
2. Experimental
2.1. Polishing technique

A pilot rig has been constructed for polishing flat surfaces with a fluid jet containing
slurry of polishing agent where the polishing agent is mixed with water in a tank and
recirculated, see figure 1a. The average grain size of the aluminum oxide is 6Pm. A
piston pump pumps the 50 % (by weight) aluminum oxide slurry to the computer
controlled head. The operating parameters used are the speed that the nozzle passes over
the glass surface with, the pressure of the slurry just before the nozzle, the distance from
the nozzle to the glass surface and the angle between the fluid jet and the glass surface
under treatment. While polishing a flat ground surface speed and pressure was found to
have the largest impact on the polishing effect, a stronger influence than angle and
distance.[6] The pressure should be as high as possible and a slow speed used.
The nozzle used is cylindrical with a diameter of 1.26 mm. With this setup the
maximum pressure of the slurry is 5 MPa but the pressure used in these experiments
was 3.5 MPa. The speed used is 2 mm/s. The glass surface is either a float glass with a
surface treated with a 25 m grinding wheel, a water cut edge or hole cut with 120 mesh
sand. To polish holes more flexibility is needed than achievable in the test rig and the
pumped and slurry equipment from the test rig was connected to a WJS NC 3015 E
water cut system using its 5-axis system, see figure 1b. The slurry is still recirculated.
The aim is too polish the inside of the hole and not the edge.
Figure 1 a) Pilot polishing rig b) The pump and recirculating system hooked up to a water cut system NC 3015 E.
2.2. Surface analysis

The polished surface is judged visually and with an optical profiler, a Bruker NPFLEX.
The main parameters measured are the mean roughness of the surface, Sa, the root mean
square surface slope of the surface, Sdq, and the developed interfacial area ratio Sdr. [6]
The measurements are made with a magnification of 5.1x in VSI mode and the
580
measured area is 0.9 x 1.2 mm. A planar fit is done for the edge measurements and
cylindrical correction when measuring inside holes but no filtering is applied. The
visual examination is performed in an enclosed area with black background and light
coming from above. Four categories are used where 0 is hardly no impact and 3 is a
smoothened bright surface but still with visible structure from the water cutting.
2.3. Mechanical strength

The mechanical strength of 6 mm float glass panes 46x46 cm with a centered Ø28 mm
hole water cut or drilled from both sides were determined in a MTS-testing machine, se
figure 2. Four types of specimens were tested which consisted of the combination of
water cut and drilled holes that were both polished and unpolished. Five specimens in
each test group were tested. The glass panes were subjected to plate bending so that
compressive stresses developed at the upper surface and tensile stresses at the lower
surface.
Figure 2: Test set-up for the testing of the glass-strength at the holes.
2.4. Non- linear acoustic crack determination

The industrial partner AcousticAgree used NAW, Nonlinear Acoustic Wave
modulation,[7, 8] to estimate the total amount of defects for two drilled holes and two
water cut holes before and after polishing. A transmitter and a receiver were mounted
on the glass panes in a fixture according to figure 3.
581
Challenging Glass 3
Figure 3: Set up for acoustic measurement of glass pane with hole.
3. Result and discussion
3.1. Polishing
The surfaces are improved with increasing numbers of passes of the jet over the surface
but only to a certain extent. The parameters were optimized for straight water cut edges
as these are easier to analyze. It is found that Sdq and Sdr parameters correlate better to
the naked eye evaluation than Sa values. It is the very fine cracks that should be
removed and they scatter light giving rise to a less bright appearance of the surface.
Smooth structures as polished grinding tracks and water cut structure remain even after
small cracks are removed and will influence the Sa value but not scatter light. The
results of polishing of a straight water cut edge are shown in table 1. The lowest Sdq and
Sdr values are found for 64 passes using the angle 20 degrees, see figure 4. The polished
holes were thus polished using these parameters. When 45 degrees is used the difference
between 32 and 64 passes is very small. The naked eye indicated that also here 64
passes would give the best result.
Table 1: Optimization of polishing parameters for polishing a flat water cut edge using a 1.26 mm cylindrical
nozzle and 3.5 MPa pressure of the slurry.
Angle No of passes Visual Sa [nm] Sdq [deg] Sdr [%]
5 64 2+ 5858 25.2 12.0
20 4 0 8272 53.6 53.3

20 16 2- 7245 31.4 10.3
20 32 2 6775 25.2 6.15
20 64 2+ 6116 13.5 1.77
45 32 2 5510 15.9 2.21

45 64 2+ 7348 16.9 2.68
582
Figure 4: Results from polishing a straight edge with 3.5 MPa, 1.26 mm nozzle and the angle 20 degrees. a)
Optical profiler picture of 64 passes, b) Optical profiler picture of 4 passes c) Photo of a water cut edge
polished with 64, 32, 16 and 4 passes.
The drilled and water cut surfaces are very different, see figure 5. The drilled surface
has stripes shaped as spirals while the water cut edge has line structures perpendicular to
the surfaces as seen in figure 4 and figure 5. The polishing degree of the water cut holes
varied. The holes denoted cut 4 and cut 5 have a lower degree of polishing than the rest
and their Sdq and Sdr values resembles the unpolished holes see table 2. The polishing
degree of water cut holes denoted cut 1-3 are similar to the straight edge polished with
64 passes seen in figure 4 both in Sdq, Sdr values and appearance. It is thus possible to
polish a hole as well as a straight edge. Most of the polished drilled holes have a slightly
lower Sdq and Sdr value than the polished water cut holes.
Figure 5: These images are data restored to facilitate the visualization of the shape of the surface. a) Water
cut and polished hole b) Drilled and polished hole
583
Challenging Glass 3
Table 2: Surface analysis of drilled and water cut holes untreated and polished with an angle of 20 degrees,
a pressure of 3.5 MPa and 64 passes over the surface.
Specimen Sa [nm] Sdq [deg] Sdr [%]
Cut Unpolished 5396 51.8 57.0

Cut Unpolished 5690 52.1 59.2
Cut 1 Polished 7411 9.2 1.06
Cut 2 Polished 6603 20.2 4.53
Cut 3 Polished 5728 16.7 3.17
Cut 4* Polished 8446 56.3 73.2
Cut 5* Polished 7563 43.2 28.7
Drilled Unpolished 3633 53.0 60.4

Drilled Unpolished 2974 53.1 65.5
Drilled 1 Polished 1983 29.3 11.5
*Not polished to the same extent as the other holes
3.2. Mechanical strength

Samples with untreated water cut holes had lower strength than corresponding drilled
holes, see Table 3 in accordance with the findings of Schneider.[9] The force at failure
was a20% lower for the water cut holes. A few blasting defects could be seen on the
edge of the water cut hole and these can act as crack initiation points thus having a
negative effect on the strength of the sample. Edge flaws can appear when drilling as
well but usually results in loss of larger pieces, chipping of the edge.
Veer compared untreated and polished manually cut and machine cut edges and found
an increase in strength by 24% for manually cut and 20% for machine cut polished
glass.[2] The surface from which the polishing is done has a small degree of polishing
at the very edge as well as inside the hole. Unfortunately this side was turned upwards
and did not have an effect on the mechanical strength as the edge with tensile stresses
was untreated. The mechanical strength of the polished and untreated holes were the
same for water cut holes and very similar for drilled holes. The inside of the holes are
clearly polished but has a limited effect on the strength of the hole. The edges of the
holes are thus more important than the interior sides. If the polishing removes cracks
and sharp indentations at the edges of the holes the strength should increase compared
to the untreated hole when tensile stress is applied.
584
Table 3: The mean forces at failure, N, and the standard deviation, std dev,
for plates with untreated drilled and water cut holes.
Drilled Water cut
Untreated Untreated
Mean 670 529
Std dev 95 70
3.3. Non-linear acoustic crack determination

Both water cut and drilled holes had a lower response after polishing. The normalized
defect responses where around 3 times higher before polishing than after, see table 4.
The total amount of defects is measured with the technique, including scratches of the
surface of the panes. In the measurements performed is it not possible to identify where
the decrease of defects occurs. The polishing is mostly effective in the interior of the
hole and not in the critical edge zone. Cracks are thus mainly removed in the interior of
the holes and the improvement indicated with these measurements will be higher than
the actual strength increase. If the slightly polished edge had been exposed for the
tensile stresses the strength would most likely increase. It will be interesting to see if
the strength difference between drilled and water cut holes evens out as indicated by the
acoustic measurements when one edge is polished as well.
Table 4: Normalized defect responses from NAW for panes

with untreated or polished drilled or water cut holes.
Specimen Untreated Polished
Drilled 2 64.6 19.9
Drilled 3 65.0 19.2
Cut 1 100 33.6
Cut 2 84.1 25.2
4. Conclusion and Future work

The water cut hole is somewhat weaker than the drilled hole which might be an effect of
blasting during the cutting process. Polishing of mainly the interior of the holes did not
improve the strength of the specimens and it is thus more important to improve the edge
than the interior.
The next step is to put more effort into polishing one edge as well as the interior of the
holes and measure the strength when tensile stresses are introduced in the polished edge.
These results will show if it is possible to achieve a similar strength with a polished
water cut hole as a polished drilled hole. The flexibility with water cutting is large than
drilling and none spherical holes can be produced.
585
Challenging Glass 3
5. Acknowledgements
Funding’s have been provided by EU structural funds, Region Blekinge and Kalmar
Regionförbund. Orrefors Kosta Boda, Glasteknik i Emmaboda, Water Jet Sweden,
KMT Robotic Solutions and AcousticAgree have all contributed to the success of the
project.
6. References
[1] Gulati, S.T. and T. Troe, Importance of Edge Finish on Thermal Tempering, in Glass Processing Days
2001: Tampere. pp. 72-78.
[2] Veer, F.A. and J. Zuidema, The strength of glass, effect of edge quality, in Glass Processing Days 2003:
Tampere. pp. 106-109.
[3] Bavelloni, R.C.Z., A. Kaonpää, and A.P. Nikkilä, Effects of different edge treatments on the 4-point
bending strength of normal and tempered glass. IMP Technologica, 2006. 02: pp. 12-16.
[4] Proctor, B., The effects of hydrofluoric acid etching on the strength of glasses. Physics and Chemistry
of Glasses, 1962. 3(1): pp. 7-27.
[5] Booij, S.M., Fluid Jet Polishing - possibilities and limitations of a new fabrication technique, 2003,
Technische Universiteit Delft.
[6] Stålhandske, C. and M. Lang, Flexible glass polishing with a cerium oxide containing fluid jet, in
Proceedings of Glass Performance Days 2011: Tampere, Finland. pp. 446-448.
[7] Zarembo, L.K., et al., On some phenomena accompanying forced non-linear vibrations of acoustic
resonators. Akust. Zh., 1966. 12: pp. 486-487.
[8] Zarembo, L.K. and V.A. Krasilnikov, Nonlinear interaction of elastic waves in solids. IEEE
Transactions on sonics and ultrasonics, 1967. SU-14: pp. 12-17.
[9] Schneider, J. and J.D. Wörner, Glass Strength of Annealed and Tempered Structural Glass in the Area
of Drilled Holes, in Glass Processing Days 2001: Tampere. pp. 193-198.
586
doi:10.3233/978-1-61499-061-1-587
Influence of the Load History on the Edge

Strength of Glass with Cut Edge Finishing
Marc Vandebroek
Faculty of Design Sciences, Artesis University College, Antwerp University Association,
Antwerp, Belgium, marc.vandebroek@artesis.be
Jan Belis
Laboratory for Research on Structural Models, Department of Structural Engineering,
Ghent University, Ghent, Belgium, Jan.Belis@UGent.be
Christian Louter
Steel Structures Laboratory (ICOM), École Polytechnique Fédérale de Lausanne
(EPFL) School of Architecture, Civil and Environmental Engineering (ENAC),
Lausanne, Switzerland, christian.louter@epfl.ch
The edge strength of glass is affected by the load history: i.e. the load duration and
the loading type, e.g. constant loading, cyclic loading or linearly increasing loading.
In standards, these influences are expressed by the factor kmod, which reduces the
strength depending on the load type, e.g. wind load, temperature load. In this study,
five series of glass specimens with a cut edge finishing were subjected to in-plane
four-point bending tests. First, the specimens were subjected to a linearly increasing
loading. Then identical specimens were subjected to either a constant loading or a
cyclic loading or both. These test results were compared to the theoretical
prediction, based on the theory of fracture mechanics, as well as to the guidelines of
the standards.
Keywords: Structural glass, Edge strength, Load history, Load duration, Load type,
Edge finishing, Fracture Mechanics, Stress corrosion
1. Introduction
Designers tend to use glass more and more as a structural element. Consequently, the
edges may be subjected to significant tensile stresses, as in structural glass beams or
façade mullions. In secondary construction elements such as windows, the edges may be
subjected to considerable tensile stresses due to e.g. thermal actions. However, the edge
strength, which is highly dependent on the edge finishing, is - in contrast to the surface
strength - insufficiently documented in literature and in the existing standards [1][2]. In
particular, experimental results of cyclic testing are scarcely documented in literature.
In this study, 5 series of specimens with a cut edge finishing and a thickness of either 4
or 8 mm, were tested in a four-point bending setup. First, two series were tested for
linearly increasing loading (constant stress rate, strength f). Then two series of
specimens, identical to the previous series, were tested for constant loading (constant
stress, strength fct). Finally, one series was tested for cyclic loading (cyclic constant
stress, strength fcycl).
587
Challenging Glass 3
The first objective of this investigation is to explore whether the test results correspond
to the theoretical prediction, based on the theory of fracture mechanics and stress
corrosion. The second objective is to explore whether the test results correspond to the
guidelines, mentioned in the current standards.
2. Test specimens and method

The nominal sizes of the soda lime silica glass specimens were 110 mm * 12.5 mm * 4
mm and 170 mm * 18.75 mm * 8 mm, see Figure 1, 2. The outer span (L) and inner
span (or load span l) lengths were 100 mm and 40 mm, respectively, for the 4 mm
specimens and 150 mm and 60 mm, respectively, for the 8 mm specimens. The edges
were simply cut, without any specific edge finishing. The cutting was performed by a
qualified glass processer using a standard automated cutting process. The scoring of the
specimens occurred at the air side.
The specimens under study were subjected to in-plane four-point bending tests in an
Instron 3369 testing machine at a test temperature of 20°C + 2°C and a relative
humidity of 65% + 4%, with the air side of the glass always in the same position (Figure
2). The specimens of thickness 4 mm were loaded either at a stress rate of 2 MPa/s or at
a constant stress (Figure 3). Next, the specimens of thickness 8 mm were either loaded
at a stress rate of 0.08 MPa/s, a constant stress or a cyclic constant stress (Figure 4). For
the constant cyclic stress, the holding phase amounted to 3 seconds and the time
between the cycles amounted to 20 seconds. The transition between the constant stress
and no stress was performed at a stress rate of 50 MPa/s.
An overview of this test series is given in Table 1. At least 90 days elapsed between
processing the edge and testing the specimens. During the 14 days before testing, the
specimens were kept at a temperature of 20°C + 2°C and a relative humidity of 65% +
4%, the same as the test conditions.
Figure 1: Schematic overview of the in-plane four-point bending test.
588
Influence of the Load History on the Edge Strength of Glass with Cut Edge Finishing
Figure 2: Detail of the in-plane four-point bending test setup.
Table 1: Overview of the test series.
Specimen Edge finishing Stress rate Number

Series
thickness [MPa/s] of specimens
4 mm Cut Stress rate: 2 MPa/s 100 A
Constant stress 40 B
8 mm Cut Stress rate: 0.08 MPa/s 40 C
Constant stress 20 D
Cyclic constant stress 20 E
[MPa] stress rate: 2 MPa/s

constant stress
70
failure
60
50
40
30
20
10
time [s]
0
0 10 20 30
Figure 3: Loading at constant stress rate and constant stress, specimens of 4 mm (series A, B).
589
Challenging Glass 3
ʍ stress rate: 0.08 MPa/s

constant stress
cyclic constant stress
70
loading
60
time: 3 s failure
loading and unloading at 50 MPa/s
50
40
30
20
10
time [s]
0
0 100 200 300 400 500 600 700
load of 20 N during 20 s
Figure 4: Loading at constant stress rate, constant stress and cyclic constant stress,
specimens of 8 mm (series C, D, E).
During testing at a constant stress rate (Table 1: series A and C), the load P in function
of the time was recorded, as well as the time to failure tf. The specimens which failed
outside the load span were excluded from the study.
During testing, the stress n was given in function of the load P by Eq. (1):
( P / 2).d 3.P.d
Vn (1)
(b.h 2 ) / 6 b.h 2
n [MPa]: maximum tensile stress, constant within the load span (Figure 1)
P [N]: total load (Figure 1)
d [mm]: distance between the load and the support (Figure 1)
b [mm]: width of the specimen (Figure 1)

h [mm]: height of the specimen (Figure 1)
After testing, the failure stress values or tensile strength values f were calculated based
on the failure loads Pf for series A and C (Table 1) with Eq. (2):
( P f / 2).d 3.P f .d
f (2)
2
(b.h ) / 6 b.h 2
f [MPa]: tensile strength corresponding to a linearly increasing loading (experimental

result: specimen loaded at a constant stress rate)
Pf [N]: experimental failure load
590
3. Fracture mechanics and stress corrosion

The theory of linear elastic fracture mechanics (LEFM), generally accepted in
describing glass failure strength, describes the relation between the tensile strength and
the flaw parameters, i.e. the flaw geometry and the flaw dimension.
According to the latter theory, at inert conditions, i.e. when no subcritical crack growth
occurs, the general relationship between the stress intensity factor KI (mode I), the
constant tensile stress normal to the flaws plane n , the geometry factor Y and the flaw
depth a is given by Eq. (3) [4][5][6][7]:
KI Y .V n .(S .a )1 / 2 (3)
1/2
KI [MPa.m ]: stress intensity factor in mode I
Y [-]: geometry factor
n [MPa]: tensile stress normal to the plane of the flaw (Figure 5)
a [m]: flaw depth (Figure 5)
Figure 5: Schematic view of the flaw, n denotes the tensile stress normal to the flaws plane.
Furthermore, according to LEFM, the critical stress intensity factor or fracture

toughness is the stress intensity factor which leads to instantaneous failure (i.e.
corresponding to the inert or short-term strength) [1][3][4]:
K Ic = Y . f .( .a ci )1 / 2 (4)
KIc = 0.75 MPa.m1/2: fracture toughness of modern soda-lime silica glass [1][8]
f [MPa]: inert strength
aci [m]: initial critical flaw depth
Considering stress corrosion and subcritical crack growth, the relation between the
crack velocity v and the stress intensity factor KI is given in literature [1][9][10] (see
Figure 6: region I):
v = v 0 .( K I / K Ic ) n (5)
v [m/s]: crack velocity
v0 [m/s]: crack velocity, when KI = KIc
n [-]: crack velocity parameter: constant value of 16 [11]
591
Challenging Glass 3
Since structural elements are generally expected to be in service for several years, only
region I in Figure 6 is of interest, i.e. the region with extremely slow subcritical crack
growth. The contribution of regions II and III to an element’s lifetime is negligible [12].
log v
III
v0
v = v0 .(KI / KIc)n II
I
log KI
Kth KIc
Figure 6: Relationship between crack velocity and stress intensity [1].
To find the relation between two different stress histories which result in the same crack
growth, the differential equation of crack growth is integrated [1][6].
Eq. (5) can be written as follows:
v = da / dt = v 0 .( K I / K Ic ) n (6)
t [s]: time
Using Eq. (3), integration of Eq. (6) yields [1][6]:
ac ( t ) t
䌿( 1 / a n/ 2
)da = 䌿v .( Y .
0 / K Ic ) n . n
n(t ).dt (7)
aci 0
aci [m]: initial critical flaw depth

ac(t) [m]: critical flaw depth at time t
n(t) [MPa]: stress normal to the plane of the flaw at time t
Thus, with n being constant, Eq. (7) yields:
t
2
䌿 ( t ).dt = ( n - 2 ).v
0
n
n n
0 .( Y . / K Ic ) .a ci
( n - 2)/2
( 1 - ( a ci / a c ( t )) ( n - 2 ) / 2 ) (8)
592
At failure time tf (or lifetime of the flaw under consideration), ac(t) equals ac(tf):
tf
2
䌿 ( t ).dt = ( n - 2 ).v
0
n
n n
0 .( Y . / K Ic ) .a ci
( n - 2)/2
( 1 - ( a ci / a c ( t f ))( n - 2 ) / 2 ) (9)
tf [s] : time period during which the flaw can resist the stress history
As n is large and assuming ac(tf) >> aci, Eq. (9) yields [1][6]:
tf
2
䌿 ( t ).dt = ( n - 2 ).v
n
n
0 .( Y .
( n - 2)/2
/ K Ic ) n .a ci
(10)
0
Eq. (10) means that two stress histories n1(t), t [0,tf1] and n2(t), t [0,tf2] cause the
same crack growth if:
tf1 tf 2
䌿
0
n
n1 ( t ).dt = 䌿
0
n
n2 ( t ).dt (11)
The value of these integrals increases from 0 at the beginning of the loading to the value
of Eq. (10) at failure [12]. The integration will be performed for values of KI > Kth, Kth
being the crack growth threshold (Figure 6)[1][12][13].
Comparing the strength values of two specimens tested at a different constant stress i.e.
assuming a constant value of n1(t) = fct,1 and n2(t) = fct,2 , Eq (11) yields:
f ct ,1 / f ct ,2 = ( t f ,ct ,2 / t f ,ct ,1 )1 / n (12)

4. Theoretical and experimental values

First, the series A and C were tested at a constant stress rate. These tests resulted in the
experimental strength f. Then the series B and D were tested at a constant stress.
Theoretically, the magnitude of this constant stress was determined by means of Eq.
(11). Therefore, experiments were first executed under the latter constant stress, aiming
at the same time to failure as the test with a constant stress rate, see Figure 3 and 4. If
the mean failure time of a certain number of specimens was higher than the mean value
of the corresponding series, tested at a constant stress rate, the level of the constant
stress was increased, and vice versa. In this way, the experimental strength fct was
determined. Finally, the same procedure was followed for series E, which determined
the experimental strength fcycl.
The magnitude of the constant strength value fct was calculated theoretically from Eq.
(11) ( n1(t) being the stress history, testing under constant stress rate, failing after a
period of time tf1; n2(t) being the constant stress history by which the specimens fail
after the same period of time tf,ct,2 = tf1). Thus, the theoretical rate fct / f was calculated.
593
Challenging Glass 3
This rate was compared to the experimental rate. In the same way, the theoretical rate
fcycl / f, by using Eq. (11), was compared to the corresponding experimental rate.
The results of this comparison are provided and discussed in section 6.
5. Results
For every series, the sample mean x and the sample standard deviation s for f, tf, fct, tf,ct,
fcycl and tf,cycl are listed in table 2.
Table 2: Sample mean x and sample standard deviation s for f, tf, fct, tf,ct, fcycl and tf,cycl.
f[MPa] tf [s] fct[MPa] tf ,ct[s] fcycl[MPa] tf ,cycl[s]
series x s x s x s x s x s x s
A 58.0 5.4 28.6 2.9 - - - - - - - -
B - - - - 48.3 1.5 34.0 49.2 - - - -
C 59.6 5.9 736 68 - - - - - - - -
D - - - - 50.0 0.5 757 487 - - - -
E - - - - - - - - 59.7 0.8 826 693
6. Discussion
Applying Eq. (11), the theoretical rate fct / f was calculated for the different series (see
Table 3). This rate was compared to the experimental rate, derived from Table 2. In the
same way, the theoretical rate fcycl / f (see Table 3) was compared using Eq. (11) to the
experimental value, derived from Table 2. For both rates, the deviation was calculated:
dev. = 100*(experimental rate – theoretical rate) / theoretical rate.
Table 3: Theoretical and experimental rate of fct / f and fcycl / f.

theoretical experimental
fct / f fcycl / f fct / f fcycl / f dev.
series
[%] [%] [%] [%] [%]

A-B 0.835 - 0.833 - - 0.2
C-D 0.833 - 0.839 - + 0.7
C-E - 0.951 - 1.00 +5.2
It can be concluded from Table 3 that:
x Eq. (11) yields rates fct / f which are only 0.2 % less to 0.7 % more conservative
than the experimental values (bold values of the dev.). Both the theoretical and
experimental values of fct / f are all very close to the theoretical value found by
Mencik [14], i.e. (1/(n+1))1/n = 0.838.
x However, Eq. (11) yields a rate fcycl / f, which is 5.2 % more conservative than
the experimental value.
594
According to prEN13474-3 [2], the design strength is calculated using kmod , which
considers the load duration of an action (e.g. for wind loads i.e. 600 s) and is given by
Eq. (13), which is basically derived from Eq. (12):
k mod (t test / t load )1 / n (13)

kmod [-]: factor for the load duration
ttest [s]: load duration of the test
tload [s]: load duration of the action
In case these guidelines are used for incorporating the effects of cyclic loading, a
discrepancy occurs.
Applying Eq. (12) or Eq. (13), the strength fct,1 cycle which corresponds to the time of one
cycle equals: fct,1 cycle = fct * (tf,ct / tf,ct,1cycle)1/n and delivers the values listed in Table 4 for
series D. Consequently, the deviation between fcycl and fct,1 cycle was calculated and listed
in Table 4: dev. = 100*( fcycle – fct,1 cycle) / fct,1 cycle
Table 4: Estimated mean experimental values for fct, fct,1 cycle and fcycl.
fct fct,1 cycle fcycl dev.
[MPa] [MPa] [MPa] [%]
D 50.0 69.1 - -
E - - 59.7 -13.6
Table 4 shows that, when the period between the cycles is 20 s and the number of cycles
is about 33 (series E), the guidelines in the standard overestimates the strength by
13.6 % compared to the experimental values.
Next, the experimental values of fcycl are compared to the values considering all cycles
fct,all cycles. Applying Eq. (12) or Eq. (13), the strength fct,all cycles which corresponds to the
total time of all cycles together is given by fct,all cycles = fct * (tf,ct / tf,ct,all cycles)1/n . Table 5
delivers the value for series D. Consequently, the deviation between fcycl and fct,all cycles
was calculated and listed in Table 5: dev. = 100*( fcycle – fct,all cycles) / fct,all cycles
and provides a value which is conservative (dev. of 7.2 %).
Table 5: Estimated mean experimental values for fct, fct,all cycles and fcycl.
fct fct,all cycles fcycl dev.

[MPa] [MPa] [MPa] [%]
D 50.0 55.7 - -
E - - 59.7 +7.2
595
Challenging Glass 3
7. Conclusions
x Testing the same specimens at a constant stress rate and at a constant stress
shows that the theoretical assessment of the rate fct / f, according to Eq. (11),
yields theoretical values which are only 0.2 % less to 0.7 % more conservative
than the experimental values of this study. Thus, the assessment seems
sufficiently accurate. Testing the same specimens at a constant stress rate and
at a constant cyclic stress shows that the theoretical assessment of the rate fcycl /
f, according to Eq. (11), yields theoretical values which are 5.2 % more
conservative compared to the experimental values. Eq. (11) is too conservative
in this case.
x In case the current guidelines in the aforementioned standards are also used for
incorporating the effect of cyclic loading, a discrepancy occurs. Applying these
guidelines to the cyclic loading test results with about 33 cycles, the strength
values calculated according to the standards under investigation were 13.6 %
less conservative than the experimental results. Yet, if the number of the
corresponding cycles was considered in the calculation, the assessed strength
value was 7.2 % more conservative compared to the experimental values. Thus,
considering the number of cycles will be a good basis for a strength calculation
method.
8. References
[1] Haldimann, M., Luible, A., Overend, M., Structural Engineering Document 10: Structural use of glass,
IABSE / ETH Zürich, Zürich, 2008.
[2] prEN 13474-3:2009, Glass in building – Determination of the strength of glass panes - Part 3: General
method of calculation and determination of strength of glass by testing, CEN, 2009.
[3] Griffith, A, The Phenomena of Rupture and Flow in Solids, Philosophical Transactions, Series A, 1920,
221: 163-198.
[4] Irwin, G., Analysis of Stresses and Strains near the End of a Crack Traversing a Plate, Journal of
Applied Mechanics, 1957, 24: 361-364.
[5] Anderson, T.L., Fracture Mechanics – Fundamentals and Applications, Second Edition, CRC Press,
Florida, 1995.
[6] Wörner, J.-D., Schneider, J., Fink, A., Glasbau: Grundlagen, Berechnung, Konstruktion, ISBN 3-540-
66881-0 Springer-Verlag Berlin Heidelberg New York, 2001.
[7] Weller, B., Nicklisch, F., Thieme, S., Weimar, T., Glasbau-Praxis: Konstruktion und Bemessung. 2 Aufl.,
Berlin: Bauwerk, 2010.
[8] Porter, M., Thesis: Aspects of Structural Design with Glass, Trinity, Oxford, 2001.
[9] Wiederhorn, S.M., Bolz, L.H., Stress corrosion and static fatigue of glass, Journal of the American
Ceramic Society, 1970, Vol. 53, p. 543 – 548.
[10] Lawn, B.R., Fracture of Brittle solids, 2nd edition, Cambridge University Press, Cambridge, 1993.
[11] Charles, R.J., Static Fatigue of Glass II, Journal of Applied Physics, 1958, 29, Number 11: 1554-1560.
[12] Haldimann, M., Thèse n° 3671: Fracture strength of structural glass elements – analytical and
numerical modelling, testing and design, EPFL, Lausanne, 2006.
[13] Fink, A., Dissertation D17: Ein Beitrag zum Einsatz von Floatglas als Dauerhaft tragender
Konstruktionswerkstoff im Bauwesen, Technische Universität Darmstadt, Institut für Statik, Bericht Nr.
21, 2000.
[14] Mencik, J., Strength and Fracture of Glass and Ceramics, Glass Science and Technology, 1992, 12.
596
doi:10.3233/978-1-61499-061-1-597
The Relation Between Fracture

Morphology and Failure Strength in
Annealed Glass
Frederic Veer
Delft University of Technology
Yurii Rodichev
Pisarenko Institute of Strength
The failure strength of glass as determined in bending tests shows a lot of scatter. In
addition there is no easy statistical description of the failure strength. Part of the
problem is that data sets are usually too small. To solve this a large research project
was conducted that involved the effects of stress corrosion, tin side and original
scoring side on the failure strength of cut, ground and polished annealed float glass.
The results show that there are different strength groups, with different fracture
morphologies. The change in fracture morphology is coincident with a change in
average strength and Weibull modulus. Allowing for this change in fracture
morphology allows a much better description of the failure strength of glass thus
improving the reliability of a strength value.
Keywords: Glass strength, fracture morphology, corrosion
1. Introduction
For several decades it has been clear that there is a relationship between the fracture
morphology of glass and it’s fracture behavior. In 1967 Wiederhorn, [1], showed that
there is a relation between humidity and glass failure mode. Michalske, [2], showed in
1984 that there is relationship between the crack velocity and the stress intensity. Cracks
blunted by corrosion having lower stress intensities and thus a lower stress intensity and
less crack branching. Other authors have shown similar results. The recent standard
work fractography of ceramics and glasses by Quin, [3], gives a good overview of
fracture morphologies. These works however are based on limited data sets. The
physical chemistry of the problem is well covered, the implications for statistical
strength determination not. Recent work by the authors, [4], using larger sets of test
results on cut glass suggest a more complicated relationship between Weibull modulus
and fracture mode. The statistical effects can however only be seen if a sufficiently
large number of test results is available.
To extend this work the authors conducted a series of 4p bending tests on cut, ground
and polished glass. These specimens were sawn from long beams where the short ends
were not ground. Thus the original side could be easily and accurately distinguished.
This allowed the authors to obtain a single series of data sets under controlled
conditions with a statistically valid number of results.
597
Challenging Glass 3
2. Experimental method
The beams were cut using a rotary diamond wheel cutter into specimens 450 mm long.
The resulting 190 specimens were tested in four point bending at different speeds. All
specimens were tested on a zwick z10 universal testing machine. The top surface of the
specimens was covered in a large piece of adhesive foil to keep the specimen together at
failure and to prevent fragments flying away. All specimens were photographed
following the test. Fracture mode and number of cracks were determined. A number of
specimens were examined with a microscope to look at fracture sources in more detail.
The test conditions are given in table 1.
Table 1: experimental conditions

parameter value
Specimen length 400mm
Specimen width 60 mm
Thickness 8 mm
Support roller distance 330 mm
Loading roller distance 133 mm
Temperature 20.2 to 20.6 °C
Humidity 50 to 52% relative humidity
Bur side of glass plate Air side (non tin side)
Figure 1: Test setup
598
The Relation Between Fracture Morphology and Failure Strength in Annealed Glass
3. Fracture typology
A number of failure types can be distinguished:
x the simple crack, designated S failure

x cracks that start as a simple crack and later branch, the simple branched crack,
designated SB failure
x the v crack, where branching starts at the beginning of the crack, designated V
failure
x the multiple v crack, where simultaneously multiples cracks initiate in a V
crack, designated MV failure
x the x-crack where failure does not start on an edge but on the specimen surface,
designated X failure
These are illustrated in figures 2 to 6.

Figure 2: Simple crack
599
Challenging Glass 3
Figure 3: Simple Branched crack

Figure 4: V crack
600
Figure 5: Multiple V crack

Figure 6: X-crack
601
Challenging Glass 3
4. Results
The for this paper critical parts of the results are given in tables 2 and 3. Full results are
given in [5].
Table 2: Distribution of failure source and failure type

Test Ratio of
Mean
Test type speed std to %L %R %X %V %SB %S
strength
(mm/min) mean
Bur side up 10 48.5 15.9% 32.5% 45% 25% 52.5% 22.5% 2.5%
Bur side
10 51.3 17.8% 30% 45% 27.5% 52.5% 17.5% 0%
down
Bur side up 1 40.9 17.6% 50% 25% 25% 35% 40% 0%
Bur side
1 43.4 15.6% 45% 35% 20% 35% 40% 5%
down
Bur side up 0.1 36.0 15.7% 40% 45% 15% 15% 60% 10%
Bur side
0.1 37.3 13.2% 55% 35% 10% 20% 65% 5%
down
Bur side up
tested in 0.1 31.6 15.6% 18.8%* 56.3% 0% 0% 68.8% 31.2%
water
Bur side
down tested 0.1 32.5 18.5% 33.3%* 41.7% 0% 0% 66.7% 33.3%
in water
* due to large number of simple crack failure identification of left or right failure is difficult
Table 3: Spread of strength
Mean Ratio
Test Minimum Maximum Minimum
strength of std Maximum
Test type speed for SB for SB for V
to for V (MPa
(mm/min) (MPa) (MPa) (MPa) (MPa)
mean
Bur up 10 48.5 15.9% 34.2 50.4 38.8 62.8

Bur down 10 51.3 17.8% 37.2 51.2 40.6 64.8*
Bur up 1 40.9 17.6% 29.8 42.4 36.0 55.6
Bur down 1 43.4 15.6% 29.8 45.4 38.0 58.6
Bur up 0.1 36.0 15.7% 28.0 42.6 40.2 45.8
Bur
0.1 37.3 13.2% 26.8 45.6 40.6 41.4
down
Bur up in
0.1 31.6 15.6% 29.9 39.6 n.a. n.a.
water
Bur down
0.1 32.5 18.5% 29.6 39.6 n.a. n.a.
in water
* there are three multiple V failures at higher stresses
602
5. Discussion
The results are comparable to those in the cited and non-cited literature. The analysis
used here shows however the importance of having a statistically valid set of result. The
first point is the danger of looking at mean strength values. Using all valid data points to
calculate a mean value gives a mean strength at 10 mm/minute of 50 MPa and 37 MPa
at 0.1 mm/minute. This is because the higher strength V failures count more heavily
than the lower strength SB failures. In large pieces of glass where longer edges are
tested the SB failures will be dominant because there will be more and more severe
flaws, even at high speeds. Looking at a Weibull plot of the SB failures, given in figure
7 suggests that the Weibull strength changes from 30 MPa at 10 mm/minute to 25 MPa
at 0.1 mm/minute. This value is similar to that for the SB failures for the tests conducted
under water. This is a much smaller change than the change in mean stress. These
values are also much more reliable for engineering purposes.
In addition there is a clear effect of crack morphology. Figure 8 compares the SB and V
failures for the bur side up tests at 10 mm/minute. The two data sets have well defined
single Weibull behavior, with significantly different Weibull constants. Figure 9 shows
the same results for the bur side down tests. This shows the same qualitative result but
suggest a small effect of the bur on the results.
Thus the effect of crack morphology has to be considered in evaluating results. Figure
10 shows how the probabilities of different crack morphologies changes with test speed
(stress corrosion attack). The change from a dominance of X and V failures at high
testing speed to a dominance of SB failures at low testing speeds is very important.
Typically tests are conducted at high speed while real life applications resemble very
low speed tests. Thus the high speed tests tend to be not representative and severely
overestimate the strength unless the change in crack morphology is allowed for.
Figure 11 shows a generalized strength diagram. At 20 MPa there is a change from

simple to simple branched crack morphology. Around 40 MPa there is a change from
simple branched to V failure morphology. At 60 MPa there is a change from V failure
to multiple V failure morphology. This failure diagram is in itself almost independent of
test speed. At lower test speeds we simply get the lower part of the diagram. If we
conduct the tests under water only SB and S failures are observed. The S failures are
now a much higher percentage of the total.
The X-failures are separate of the edge failures. As these do not seem to occur
significantly at lower test speeds their practical relevance is limited. Removing them
from the data set is also safe as a pure S or SB set is more conservative. As the X-
failures also seem to differ slightly between the Tin and the air side they should not be
looked at in a single data set with other failures. Including them will lead to non-
conservative results.
For lack of space the reasons for the observed crack morphology changes are not
discussed. A possible explanation is given in [5].
603
Challenging Glass 3
Figure 7: Weibull plot of SB failure data at all speeds for test results in air
Figure 8: Bur side up, SB and V failures separately plotted, 10 mm/min data
604
Figure 9: Bur side down, SB and V failures separately plotted, 10 mm/min data
Figure 10: Changing probability of crack morphology with test speed
605
Challenging Glass 3
Figure 11: Relationship between crack morphology, strength range and corrosion
6. Conclusion
From the results it is concluded:
x that using the mean strength of glass tested in four point bending is non-
conservative
x there is an important change in crack morphology with testing speed
x if the failure stresses are grouped to crack morphology, the individual groups
can be described by single Weibull functions.
x The Weibull constants are thus dependent on the failure morphology
x if the data is not separated it cannot be described by a single Weibull function
and no safe engineering strength can be calculated from this data
7. References
[1] S. M. Wiederhorn, Influence of Water Vapor on Crack Propagation in Soda-Lime Glass, J. Am. Ceram.
Soc. 50, 407-414 (1967).
[2] T.A. Michalske, fractography of slow fracture in glass, astm stp 827, 1984.
[3] G.D. Quin, Fractography of ceramics and glasses, NIST SP - 960-16, 2006
[4] F.A.Veer and Y.M.Rodichev, The structural strength of glass, hidden damage , Journal of strength of
materials, vol 43, no3, p 93, 2011
[5] F.A.Veer and Y.M.Rodichev, The structural strength of glass, hidden damage, part 2, Journal of
strength of materials, to be published1
606
doi:10.3233/978-1-61499-061-1-607
Strength of Flat Glass Subjected to Thermal

and Mechanical Load
Vladimir Zubkov, Nadezhda Kondratieva
Testing Centre «Samarastroyispytanya» of Samara State University of Architecture and
Civil Engineering, Russia, zubkov@sgasu.smr.ru
The results of the experimental and theoretical research of flat glass resting upon its
perimeter and subjected to distributed and point load are shown in the article.
Recommendations are given to avoid flat glass fracture because of thermal loads.
The stress distribution in glass subjected to distributed and point loads has been
determined. The method for calculating ultimate strength of glass at lateral bending
is offered. The method and program for calculating flat glass strength at designing
glass translucent structures, coverings and floorings have been worked out.
Keywords: Flat Glass, Covering, Thermal and Mechanical Loads, Glass Fracture.
1. Introduction
For estimating strength of glass used in modern translucent structures Testing Centre
«Samarastroyispytanya» of Samara State University of Architecture and Civil
Engineering made experimental and theoretical research of flat glass work at lateral
bending subjected to the distributed and point loads. The research showed that the
existing theories of plates' strength based on the application of elastic-plastic materials
can’t be used for calculating flat glass strength.
2. Special features of flat glass work in translucent structures

Considering flat glass as a structural material subjected to mechanical load we should

take into account special features of its work related to the strength characteristics of
glass and methods of their determining. Strength and deformation of the whole structure
depend on these characteristics. Strength characteristics of any construction material
are the following: ultimate tensile strength (Rp), ultimate compressive strength (Rc),
modulus of elasticity () the coefficient of lateral deformation ( ). Values of these
characteristics depend on the properties of material and the methods for their
determining. As glass is traditionally considered as the material for light transmission,
nowadays there is no common opinion about the method for getting strength
characteristics. There is the method of double coaxial ring (EN 1288-2). This method
has a number of advantages that exclude the effect of defects in sample’s edges on
measuring results. However the given method has some disadvantages for while testing
complex stress caused by tension, shear and cutting appears in glass So the result of
testing is the conventional ultimate strength when glass fractures but not the tensile
strength at bending. In this method the scale factor works depending on the ratio of
glass thickness to the diameter of the inside ring.
607
Challenging Glass 3
Our research showed that the most accurate method for getting ultimate tensile strength
(R p), modulus of elasticity (¥) and the coefficient of lateral deformation () of flat glass
at bending is the testing of samples of 650x120mm in size with the use of four-point
scheme on the installation shown in Figure 1. The size of samples is taken in accordance
with minimum influence of defects on the testing results, economical and technical
requirements. The edges of the samples are grinded.
The width of the sample should be five times less than its length so it will work
according to the beam scheme and the existing formula will be right for determining the
following characteristics:
R .. L b h 2 6,75 h (1)
Where: R - ultimate tensile glass strength at bending, [mPa];

– load at sample’s fracture, [¦];
L – distance between supports (L = 600), [mm];
b – sample’s width, [mm];
h – sample’s thickness, [mm].
E (0,213( P5 P3 ) L3 ) /( f 5 f 3 ) b h 3 ) (2)
Where: – modulus of glass elasticity, mPa;

P3 5 – Load on step 3 and step 5 (or step 2 and step 4), H;
f3 f5 – Flexure on step 3 and step 5 (or step 2 and step 4) .
Figure 1: Installation for Determining Ultimate Glass Strength.
In a number of countries where the outside temperature in winter is below zero Celsius
there appear mass deformations of glass in spring time because of dimming films and
608
Strength of Flat Glass Subjected to Thermal and Mechanical Loads
whole toned glass. In a number of countries where the outside temperature in winter is
below zero Celsius there appear mass deformations of glass in spring time because of
dimming films and whole toned glassThe value of such loads depends on the coefficient
of sun energy absorption (G) by glass and the coefficient of thermal linear expansion. If
the glass strength is not enough for such loads cracks appear in it. Figure 2 shows the
fracture of inside translucent glass because of thermal loads. The reason of the fracture
is a big value of thermal expansion coefficient of glass.
The researches show that glass won’t fracture because of thermal loads if the coefficient
of thermal linear expansion isn’t over 9¨10 – 6 ª – 1, and the coefficient of solar energy
absorption corresponds to the determined value of glass actual strength. These
requirements are given in Table 1. The requirements can be applied to external and
internal glasses.
Table 1: Dependence of glass strength on solar energy absorption coefficient
Coefficient of solar energy absorption G, % Ultimate glass strength,

not less than, MPa
1 2
Up to 25 50
Up to 35 75
Up to 45 100
Over 45 Over 110
Figure 2: Glass Fracture because of Thermal Loads.
3. Flat glass strength at lateral bending

Because of wind, snow and temporary point loads compressive and tensile stresses
appear in glass. When such stresses reach maximum values glass will break. For
avoiding the fracture glass in translucent and protective structures should be calculated
609
Challenging Glass 3
with the definite degree of reliability. Unfortunately in Russia and other countries there
are no regulating documents and methods for calculating flat glass strength at lateral
bending. While designing translucent structures glass thickness is taken instinctively, as
a rule without paying attention to its actual strength, reasonable theoretical or economic
grounds.
Taking into account this fact Testing Centre «Samarastroyispytanya» made

experimental and theoretical research of flat glass strength in glass translucent and
protective structures.
Flat glass as a construction may be shown as a plate resting upon its four sides and
subjected to distributed load. There are theories of such plates’ strength as the theories
of S.P. Timoshenko, A.S. Volmir, I.G. Galerkin, I.G.Bubnov. Some research sources
recommend the formula derived by S.P.Timoshenko for determining the thickness of
glass (h) subjected to distributed load:
Eqb 2
h (3)
Rp
Where:
– coefficient depending on of sides’ length ratio;
b – short side of glass plate;
Rp – calculated resistance of glass.
However our research showed that the theoretical strength calculated by the given
theories is not equal to the actual strength obtained at testing. The difference is up to
200% and more. The situation may be explained by the fact that the available theories
consider the plates with the ratio of short side to the thickness of not more than 100.
Maximum stress in such plates appears in the middle zone. In protective structures the
flat glass with the ratio of short side to its thickness from 100 to 300 is used. In such
plates maximum stress at lateral bending appears in corner zones.
3.1. Strength of flat glass subjected to uniformly distributed load

Stress in the flat glass plates with the ratio of short side to the thickness (b / h) from 100
to 300 and the ratio a/b<2 should be calculated by the formulas:
V max 0,5V 0,5 V 2 4W 2 (4)
2 E
b· §a· 2§
V 0 , 68 c1 q D 1 ¨ ¸ ¨ ¸ 10 3 , (5)
©h¹ ©b¹
610
b2
W 0,17 kc1qD12 , (6)
h2
Where: – maximum tensile stress in the corner zone of the plate, mPa;
max
– tensile stress in the corner zone perpendicular

to the plate’s diagonal, mPa;
q – load on the plate, kPa;
a – long side of the plate, mm;
b – short side of the plate, mm;
h – plate thickness, mm;
D1 – coefficient depending on b / h;
1 – coefficient considering the conditions of plate’s resting;
– coefficient depending on the ratio of long and short sides
of the plate (a / b).
The glass flexure in structures is recommended to be calculated by the following

formula:
f a 10 4 p J 12 (1 P 2 ) (7)
Where: – long side of the plate, mm;
– given inflexibility;
– coefficient depending on b/h relations;
– coefficient of lateral deformation.
Being subjected to prescribed load glass flexure shouldn’t be over 1/100 of long side of
the glass plate.
3.2. Strength of flat glass subjected to concentrated load

Tension along the axes x, in flat glass plates with the ratio of short side to thickness (b
/ h) from 50 up to 190 should be calculated by the formulas:
6O p1ac
V (8)
h2
Where: - tensile stress in glass along axis x or , mPa;
1 – support reaction, ¦:
1 ( /( 2 2b))1,5 (9)
– concentrated load acting in the centre of the plate, ¦;

– coefficient taking into account the position of the neutral axis;
, b – sides of the plate, mm;
h – thickness of the plate, mm;
611
Challenging Glass 3
– coefficient that takes into account the change of rigidity depending upon
the ratio value of length to the plate’s thickness ( = /h).
For rather thick plates when <120, coefficient () should be calculated by the formula:
pa
D1 ( 1 ] 2 )0.3 (10)
D
In flexible plates, when 200 > >120, great membrane tensile stresses appear and the
coefficient should be got by the formula:
pa
D 2 /( 1 ] 2 ) 0.3 (11)
D
Where: – coefficient depending on ;
D = Eh 3/12(1- 2
).
Flat glass with >200 is not recommended to be used for resisting concentrated load.
Maximum tensile stresses in the centre of the plate will be equal:
V max Vp
Flexure in the central zone of flat glass subjected to concentrated load can be calculated
by the formula:
f E ( Pa 2 / D ) J (12)
Where f – flexure in the central zone of the plate, mm;

40
E
(0,15] ) 0,5
P – concentrated load, ¦;

a, b – long and short sides of the plate, mm;
h – thickness of the plate, mm;
D = Eh 3/12(1- 2);
– coefficient that takes into account the change of plate’s rigidity when
changes.
612
The flat glass strength in translucent and protecting structures subjected to distributed
and point loads will be guaranteed if:
V max d R (13)
Where: max – maximum tensile stresses, mPa;

R – calculated tensile glass strength at bending, mPa;
Calculated tensile glass strength at bending R depends on ultimate glass strength at

bending and on the class of reliability of structures:
R R max c (14)
Where: R max – – ultimate tensile glass strength at bending equal to the maximum tensile
stresses obtained at testing of glass samples 600 ¨ 120 mm in size, mPa;
– coefficient of margin strength depending on the class of constructions’
reliability according to Table 2.
Table 2: Coefficient of extreme strength of glass

Class of Coefficient
Characteristics of buildings or constructions with the use of
reliability of of margin
translucent and protective structures
structures strength()
1 2 3
A1 Buildings of more than 50 floors or structures of more than 150

meters height 7
A2 Buildings of less than 50 floors or structures of less than 150 meters
height. Buildings and constructions erected in the areas with frequent 5
hurricanes, typhoons or in the seaside regions. Covering with
calculated load of more than 4000 ¦/m 2.
A3 Buildings up to 25 floors. Covering with the calculated load of less 4

than 4000 ¦/m 2.
A4 Buildings up to 12 floors 3
Class of reliability is estimated according to the degree of danger for people’s lives
while glass fracturing during construction and exploitation of a building and the
expenses for fractured glass replacement.
As the result of the research work there have been worked out special method and
program for calculating the strength of flat glass resting upon its contour and subjected
to distributed and point loads. The program takes into account actual strength and
modulus of glass elasticity as well as the class of reliability of buildings and structures.
The given method and the calculating program have already been used in designing
many buildings and constructions including high-rise buildings.
613
Challenging Glass 3
4. Experimental research
To prove the derived formulas there were made tests on flat glass tempered samples
1500x1500mm in size and the thickness of 6-19mm (Figure 3 and Figure5). Uniformly
distributed load was applied by vacuum. Concentrated load was applied by the
hydraulic jack. The perimeter of samples rested upon the elements of the installation
with the resin layer between the sample and the installation elements as in glass
translucent structures. At testing deformation of glass along axes x and , diagonals and
sample’s flexure were measured. Deformation was measured by strain gauge resistors
with accuracy of 1 ¨ 10 – 6 units related deformation. Flexure was measured with
accuracy of 0,01mm. The fracture of the samples happened immediately without any
elastic deformations. When the sample was subjected to the distributed load the fracture
started in the corner zone (Figure 4) while being subjected to the point load the sample
started fracturing in the central zone. (Figure 6).
Figure 3: Testing of flat glass samples subjected to distributed load.

Figure 4: Fracture of tempered glass samples subjected to distributed load.
614
Figure 5: Testing of samples subjected to point load.

Figure 6: Fracture of tempered flat glass subjected to point load
Deformation of glass subjected to distributed and concentrated loads is shown in

Figure 7 and Figure 8.
615
Challenging Glass 3
Figure 7: Glass deformation perpendicular to the diagonal of the sample subjected to distributed load.
a
Figure 8: Deformation of the sample along the axis (a) and the diagonal (b) in the central zone of
the sample subjected t point load.
616
5. Conclusion
The researches made at the testing centre show that glass won’t fracture because of
thermal loads if conditions given in table 1 are met.
The greatest values of stress in glass subjected to distributed load appear in the corner
zone perpendicular to the diagonal. In the samples subjected to point load the greatest
stress appear in the central zone along axes , and the diagonal. Stresses and flexures
of flat glass subjected to distributed and point loads calculated with the use of the given
formulas correspond to actual values. The error isn’t more than 15%.
The offered method and the calculating program for flat glass have been proved by
experimental research and can be used at designing translucent protecting structures,
coverings and floorings made of flat glass.
617
Challenging Glass 3
Laminates & Composite

Designs
619
doi:10.3233/978-1-61499-061-1-623
Laminated Glass Panels in Combination

with Timber Frame as a Shear Wall in
Earthquake Resistant Building Design
David Antolinc, Roko Žarni, Franci epon
University of Ljubljana, Faculty of Civil and Geodetic Engineering, Slovenia,
david.antolinc@fgg.uni-lj.si
Vlatka Raji, Mislav Stepinac
University of Zagreb, Faculty of Civil Engineering, Croatia, vrajcic@grad.hr
The idea of the present study is to determine the in-plane stiffness and load bearing
capacity of timber frame with laminated glass infill which acts as a shear wall and
finally to estimate the contribution of such wall to the overall behavior of the entire
building. For this reason we conducted an in-plane cycling load-displacement
experiment of above mentioned wall system. The shear wall is composed of cross
laminated timber frame and heat strengthened laminated glass panels, which are
connected just with friction without any adhesives. The consequence of this type of
connection is that we get larger displacements at a certain applied load which
means higher ductility of the whole wall. On the other hand we get the problem to
achieve the serviceability limit state. The comparison of the experimental results
obtained for different types of timber frame connection details is made. The
numerical simulation using ABAQUS software has also been done which gave us
inaccurate results due to the inadequate timber frame connection modeling. The
plan for the improvement of numerical modeling is made based on the additional
experimental analysis of the timber frame elements connections.
Keywords: Timber frame, laminated glass, dynamic load, seismic design, shear
wall, in-plane stability.
1. Introduction
There is no need for further discussion that glass is a unique material with its physical
and mechanical properties as it has been proofed in many applications as the most
competitive material. However, for the use of laminated glass in earthquake resistant
building design it is important to consider the dissipation of energy at the structural
level and the post fracture behaviour of the laminated glass itself. Considering
architectural design of modern ground plan of any type of residential building where is
the southern site of the building usually as transparent as possible and keeping all other
sites with low percentage of the openings, we get the non-regular building in plan. This
consequently means that such structures will show torsional behavior when exposed to
the earthquake excitations.
However, the current design philosophy considers the glass panels in glass facades as
non-structural elements (EC8, Design of structures for earthquake resistant building
design). This consequently means that we have to neglect these elements when
designing the building resistance for the lateral load imposed by the earthquake.
623
Challenging Glass 3
According to any modern design codes (e.g. EC8) it is necessary to design primary
structural elements within the limits of the prescribed story drifts ( d rQ d 0.005h ) in
order to protect non-structural elements. An important role in protecting of the non-
structural elements (glass façade panels) plays the connection detailing to the primary
structural elements and the structural system of the façade itself. Commercially
available curtain walls and façade panels can usually accommodate 10-15 mm of elastic
displacements, which is enough to sustain the displacements demand caused by the long
term deflection, wind, thermal expansion, shrinkage and creep together with other
movements that may arise during the service life of the structure [1]. During the event
of the earthquake the façade glass panels face additional inelastic story drifts which can
crucially impact the panels and consequently create life safety hazard.
The mission of the present study is to show an innovative timber – glass façade panels
which can sustain larger story drifts and dissipate a considerable amount of energy
during the event of the earthquake. Moreover, the idea is to show that such façade
system can be even used as a primary (secondary) structural element and that it can take
over some of the in-plane imposed load without creating life safety hazard.
2. Overview of the Specimens and the Racking Test Facility

Three different types of specimens have been tested with racking test facility in our
laboratory for three possible boundary conditions. The specimens are curtain walls
assembled as a laminated timber frame with laminated glass infill. All three types of
specimens are of the same geometry and dimensions (Figure 1) with different timber
frame element connection details (CD1, CD2, CD3) in corners shown in the Figures 2a,
b and c.
Figure 1: Dimensions of the test specimens with the positions of the steel connections to the foundation and
positions of the mesuring locations (SG1-12, LVDT0-10).
624
in Earthquake Resistant Building Design
Figure 2a, b and c: Connection detail of timber frame members with two bolts CD1 (a), one bolt CD2 (b) and
one bolt with nail plate reinforcement CD3 (c).
The glass panels with dimensions of 2900/2400 mm are made of two ply heat
strengthened glass with thickness of 10 mm for one ply and 1.6 mm for EVA SAFE
interlayer. As mentioned above is the assembly of each tested wall panel made of two
just mentioned laminated glass panels which are tightly inserted into the timber frame
and laterally closed with timber purlins attached to the main frame as shown in the
Figure 3. The connection between timber frame and glass panels is ensured with friction
only without any adhesives.
Figure 3: Detailed view of the timber frame and laminated glass panels assembly.
2.1. Racking test set up

In order to obtain the in-plane behaviour and the capacity of the considered shear wall
panels the in-plane quasi-static racking tests were conducted in collaboration with the
colleagues from the University of Zagreb. The racking test facility enables us to impose
vertical load, which was 80kN (|25kN/m) in our case and monotone horizontal as well
as cyclic load H at the bottom of the panel. In the Figure 4 the setup of the racking test
facility is shown where it is obvious that the specimen is fully restrained at the top into
the reinforced concrete foundation strip while at the bottom of the specimen the
boundary conditions can be changed (clamped, released rotation) which is shown in the
Figure 5a, b, c.
625
Challenging Glass 3
Figure 4: Quasi-static racking test facility.
Figure 5a, b and c: Boundary conditions B1 of cantilevered panel (a), panel with restrained rotation at the
bottom B2 (b) and with restrained vertical displacement and rotation at the bottom of the panel B3 (c).
In the Figures 6a and b below it is shown the monotone load for the specimen FR3 and
cycling story drift protocol for the specimen FR4 with boundary conditions B1 which is
defined based on the monotone load according to the prEN 12512.
70 20
60 15
Story drift amplitude [mm]
50 10
F [kN]
40 5
30 0
0 100 200 300 400 500 600 700 800 900 1000 1100
5
20
10
10
15
0
0 20 40 60 80 100 120 140 160 180 200 220 20
t[s]
t [s]
Figure 6a and b: Monotone load protocol (a) and cycling story drift protocol (b).
626
Ten quasi static racking tests have been conducted so far for the above mentioned wall
panel specimens for different boundary conditions (BC1, BC2 and BC3), loads
(monotone, cycling) and timber frame connections (two bolts, one bolt, steel nail plate
with one bolt). In the next table 1 is shown which boundary conditions, loading protocol
and timber frame connection details are attached to each of the wall panel specimens
with designations from FR1 to FR10.
Table 1: Boundary conditions, connection details and load protocol assignments to the specimens.
Specimen
FR1 FR2 FR3 FR4 FR5 FR6 FR7 FR8 FR9 FR10
BC1 9 9 9 9 9 9
Boundary
BC2 9 9
conditions
BC3 9 9
CD1 9 9 9 9 9 9
Connection
CD2 9 9
detail
CD3 9 9
Loading Monotone 9 9 9 9 9
protocol Cycling 9 9 9 9 9
2.2. Friction force test setup and results

For better understanding of the interaction between laminated glass panels and timber
frame we conducted an additional experiment where we measured the friction force
between glass and timber. The test setup is shown in the Figure 7a and as we can see it
is assembled of one sheet of laminated glass which is embedded into two timber
elements. For different levels (5 kN, 10 kN, 15 kN, 20 kN and 25 kN) of horizontal in-
plane compression applied loads we measured the force needed to push the glass ply in
the vertical (perpendicular to the horizontal compression load) direction which
represents the friction force. The result is friction coefficient (μ=0,15) between glass
and timber which is further used for numerical modelling. The results for each of the in-
plane load levels are shown in the Figure 7b.
3 5 kN
10 kN
F [kN]
2 15 kN
20 kN
1 25 kN
30 kN
0
0 10 20 30 40 50
u [mm]
Figure 7a, b: Friction force test setup (a) and relationship between friction force F and longitudinal
displacements (b).
627
Challenging Glass 3
2.3. Buckling force test setup

We also conducted the experiment to define the buckling force of the laminated glass
panels used in the shear wall panels. The investigated glass panel is 2400 mm high and
1250 mm wide and it is inserted into the timber embedment on the top and bottom
which is shown in the Figure 8. From the results of the test it is obvious that chosen type
of the laminated glass panels (quality of the glass and interlayer) have relatively
significant residual bearing capacity. This is also one of the reasons which encouraged
us to use this type of laminated glass in a shear wall panels.
Figure 8: Buckling force test setup.
3. Experimental results of the racking test

Regarding to the data acquisition plan which is shown in the Figure 1 we have measured
displacements (LVDT0-10) to capture the overall behaviour of the entire wall panel. We
also measured strains (SG1-12) of the laminated glass panel at the certain locations to
investigate the load transferring into the glass panel. In the next Figure 9 it is shown
comparison of the displacements responses of the specimens FR3, FR7 and FR9 to the
monotone load protocol at the story height (LVDT0) of the panel where each of the
panels has different timber frame element connection detail and the same boundary
conditions.
80
60
F [kN]
40
20
0
0 5 10 15 20 25 30 35 40 45 50 55
u [mm]
FR3_B1_CD1 F7_B1_CD2 FR9_B1_CD3
Figure 9: Comparison of the displacement responses of FR3, FR7 and FR9 at the story level to the horizontal
monotone applied load H.
628
From the comparison it is obvious that specimen FR3 which has frame elements
connection detail with two bolts achieved the highest load bearing capacity which is due
to the better stress distribution in the timber. The other two panels FR7 and FR9 both
have timber frame connection detail with one bolt only where FR9 has additional
reinforcement with the steel nail plate at the ends of the timber frame elements (in
connections) to prevent splitting of the timber. We can see that wall panel FR9 shows
more compliant behaviour comparing to the one without steel nail plate reinforcement
FR7 but on the other hand the specimen FR9 shows more ductile behaviour and can
achieve higher displacement level without total failure of the connection detail.
Meanwhile the initial stiffness of all three specimens is similar for all specimens.
Furthermore in the Figure 10 below it is shown the comparison of the wall panel
displacement response to the imposed cycling load at the story height. The comparison
is made for the specimens FR4, FR8 and FR10 which means that they have the same
boundary conditions and connection details as the panels compared in previous figure 9.
It is obvious that hysteresis of the specimen FR4 achieved the highest load bearing
capacity where meanwhile the specimen FR10 with steel nail plate reinforcement
achieved the highest displacements level. When comparing the dissipated energy the
specimen FR8 has the lowest one which is a consequence of low or no contribution of
timber frame connection after the early failure. The dissipated energies at the remaining
two specimens are very close but it is necessary to expose that the frame connection at
specimen FR10 stays relatively solid (because of the steel nail plate) at the end of the
test which it is not true for the specimen FR4.
65
45
25
F [kN]
5
-60 -40 -20 0 20 40 60

-15
-35
-55
-75
u [mm]
FR4 FR8 FR10
Figure 10: Comparison of displacement response of the wall panels FR4, FR8 and FR10 at the story height to the
horizontaly applied cycling load H.
629
Challenging Glass 3
4. Future work
We will continue with the experimental research of the wall panels with the CD2 and
CD3 timber frame elements connection details for the remaining two boundary
conditions (BC2, BC3) which have not been done to get the complete comparison of
responses for all three connection details (CD1, CD2, CD3). These experimentally
obtained responses will also be modelled with the numerical simulation using ABAQUS
software. We have tried to simulate the responses of already tested specimens with the
mentioned software and we obtained inaccurate results mainly because of inadequate
modelling of timber frame elements connections. In the Figure 11a it is shown the
global meshed FE model in ABAQUS where 3D solid (C3D8I) finite elements were
used together with general contacts to model the interactions between the wall panel
elements. In the Figure 11b the closer view of the modelling for timber frame elements
connection is shown.
Figure 11a, b: Global meshed FE model (a) and closer view of the timber frame elements connection (b) modeled
in ABAQUS.
In order to improve the FE model we will conduct the experimental analysis of the
timber frame elements connections separately to capture moment – rotation and force –
displacement (modulus of slippage Kser) behaviour and capacity which will enable us to
prepare detailed FE model of the connection and then finally use it in the global model.
However, there will be made the simulation of the whole building dynamic response
based on the validated wall panel model to estimate the contribution of the considered
wall panels to the safety of the entire building when exposed to the horizontal loads
(earthquake).
630
5. Conclusions
We can conclude that we get very ductile structural system with combining the timber
frame and laminated glass infill. The results show us that the failure of the considered
structural system appears at the timber frame elements connections. After the failure of
the timber frame the friction force between glass panels and timber frame elements take
over the imposed horizontal load. On the other hand we can conclude that timber frame
represents the protection of the glass panels and the connection between glass panels
and main structure, where even after the timber frame elements connections failure the
vertical load still can be sustained. Thus, we can fulfil one of the main targets
(philosophy) of any modern earthquake resistant building design codes which is to
design buildings with some damage allowance but without catastrophic failure to
protect the human lives.
6. Acknowledgements
I would like to thank to the Slovenian research agency for funding of young researcher
David Antolinc with the contract no. 261408-1/2008. I would also like to acknowledge
the Ministry of science, education and sport of Croatia for funding the scientific project
082-1491823-1463 Characteristics of the composite structures wood – structural glass
and wood – steel (leader Prof. Vlatka Raji).
7. References
[1] McBean, Peter, Drift Intolerant Façade Systems and Flexible Shear Walls. Do we have a Problem?,
Proceedings of the Annual Technical Conference of the Australian Earthquake Engineering Society
Albury, Albury, NSW Australia, 2005.
[2] Behr, Richard A, Architectural Glass for Earthquake-resistant Buildings, Proceedings of the 7th
international glass conference in Tampere (Glass Processings Days 2001), Tampere, Finland, 2001.
[3] Pantelides, C. P., Truman, K. Z., Behr, R. A., Belarbi, Development of a Loading History for Seismic
Testing of Architectural Glass in a Shop-front Wall System, Engineering Structures, 1996, Vol. 18, No
12, pp. 917-935.
[4] Huveners, E.M.P., Herwijnen, F., Soetens, F., Hofmeyer, H., In-plane loaded glass pane (shear wall),
Proceedings of the 10th International Conference in Tampere (Glass Performance Days), Tampere,
Finland, 2007.
[5] Mocibob, Danijel, Glass Panel under Shear Loading – Use of Glass Envelopes in Building Stabilization,
PhD thesis, Ecole Polytechnique Federale de Lausanne, Lausanne, Switzerland, 2008.

[6] Freitag, Claudia, Woerner, Johan-Dietrich, Glass as structural bracings – shear capacity of
mechanically pre-stressed windowpanes, Glass Performance Days 2009 Proceedings (Proceedings of the
11th international conference), Tampere, Finland, 2009.
[7] Sucuoglu, Haluk, Vallaghan, C. V. Girja, Behaviour of window glass panels during earthquakes,
Engineering Structures, 1997, Vol. 19, No 8, pp. 685-694.
[8] Sivanerupan, S., Wilson, JL., Gad, EF., Lam, NTK., Seismic Assessment of Glazed Façade Systems,
Proceedings of the Annual Technical Conference of the Australian Earthquake Engineering Society,
Newcastle, NSW Australia, 2009.
[9] Niedermaier, P., Shear-strength of Glass Panel Elements in Combination with Timber Frame
Constructions, Glass Processing Days 2003 Proceedings (Proceedings of the 8th international
conference), Tampere, Finland, 2003.
[10] Memari, Ali M., Shirazi, Ali, Kremer, Paul A., Behr, Richard A., Development of Finite-Element
Modeling Approach for Lateral Load Analysis of Dry-Glazed Curtain Walls, Journal of Architectural
Engineering, 2011, Vol. 17, No 1.
631
doi:10.3233/978-1-61499-061-1-633
Anchoring Triangular Glass Panels to

Prevent Collapse
Graham Dodd
Arup, United Kingdom, graham.dodd@arup.com
Leroy Reed
Kite Glass, United Kingdom, leroy.reed@kiteglass.co.uk
Where frameless glass supports people, the greatest design challenge is

to prevent the panel becoming detached from its supports if it
gets broken. The cladding below a foot bridge in Manchester has to be
crawled upon for cleaning, and the narrow triangular panels are only
supported at their corners. A novel method of anchoring the panels to
their supports was devised, using loops of wire rope laminated around
the edges. The preparation of the glass and the technique of vacuum bag
laminating with ionomer interlayer were keys to success in testing and
manufacture.
Keywords: triangular, laminated, robustness, wire
1. Introduction
Where frameless glass supports people, the greatest design challenge is to prevent the
panel becoming detached from its supports if it gets broken. Corporation Street Bridge
in Manchester, UK, was built in 1999 and has an external steel structure of tubes and
tension rods between rings, from which transverse steel beams support the bridge deck.
Stainless steel fittings from the external structure support frameless glazing, which
envelopes the bridge deck. The form of the bridge is an elliptic hyperboloid, so the glass
panels are all narrow isosceles triangles, supported only in a small area at each corner,
to make the glass envelope as transparent as possible. The original glazing suffered de-
lamination and by 2010 it required replacement. This paper describes the design, testing
and manufacture of replacement glazing.
1.1. Access strategy

The transparent glass envelope wraps around the bridge deck and encloses the
undercroft allowing a clear view of the underside of the deck. Therefore, the glass has to
be kept clean above and below the deck to maintain the transparency and visual quality
of the design. During the original design stages, it was concluded that the only practical
way to clean the undercroft glass would be to walk and crawl on it, gaining access
through hatches in the bridge deck. This access strategy was thoroughly reviewed when
designing replacement glazing and the same conclusion was reached.
1.2. Robustness
The need to work on the undercroft glass meant that it had to be designed to carry the
weight of cleaning operatives and that it had to be sufficiently robust to be a safe
633
Challenging Glass 3
working platform under a variety of conditions. Robustness criteria had been developed
and specified during the original design and these were reviewed in the light of recent
experience and published guidance such as CWCT Technical Notes 42, 66 and 67. [1]
Figure 1: Corporation Street Bridge, showing the transparent undercroft.
2. Performance Requirements
2.1. Loading cases

In addition to gravity, wind and snow, a distributed access load of 0.75 kN/m² and a
concentrated access load of 0.90 kN on an area of 125 x 125 mm area were applied.
The loaded area was placed towards the narrow corner of the panel where a combination
of bending moment and reduced section apply.
2.2. Condition cases

In order to ensure the construction was robust, several condition cases were specified,
under which the glass had to be capable of supporting load. The load factors were
varied between the cases to account for differences in probability:
x Both glass leaves intact JG = 1.2 JQ = 1.3 JW =1.5

x Upper leaf broken JG = 1.1 JQ = 1.1
x Lower leaf broken JG = 1.1 JQ = 1.1
The condition with all layers of glass broken was intended to represent the worst case,
possibly during maintenance or repair works, and so the silicone sealant joints had to be
removed from the analytical model because they could not be relied upon to provide
additional support to the panel.
634
Anchoring Triangular Glass Panels to Prevent Collapse
3. Panel design
3.1. Original panels

The original glazing for the glass below the deck, known as the undercroft, incorporated
two layers of flexible polyester (PET) film bonded with chemical cure resin. The PET
film extended as elongated tabs at each corner, to be rolled around aluminium plates and
anchored in slots in the support fittings at the corners. An alternative system was
required, which would provide a high level of confidence in achieving the robustness
requirements.
3.2. New panel design

The new panel construction was 12mm toughened and heat soaked upper layer, 1.52mm
SGP interlayer, 8mm toughened and heat soaked lower layer. The asymmetric
construction provided the redundancy to carry the design layer with either layer broken.
With the thin bottom layer broken, the thicker upper layer was strong enough to span on
its own. If the upper layer was broken, it would remain and act in compression,
transferring shear force via the interlayer to the bottom layer, which would act primarily
in tension. This enabled the glass thickness to be kept within the thickness and weight
originally used.
3.3. Choice of interlayer

Since the bridge was first glazed, DuPont Sentry Glass interlayer had been in use for
over 10years and a wide range of experience had been gained in a variety of projects.
One of the earliest of those projects was the Phoenix Special Proceedings Court ceiling,
designed with James Carpenter Design Associates, where we had extended tabs of SGP
from the corners of the quadrilateral panels to retain them into small patch fittings, to
achieve a similar objective to that on Corporation Street Bridge. It was known that SGP
had much higher strength than PVB, but experience had shown that the strength
advantage at higher temperatures was not so great.
3.4. Design of interlayer as a membrane and wire rope as catenary

The strength of the interlayer to support imposed load after the glass was broken, and
the strength of the wire ropes to support the interlayer membrane, were initially
determined by experiment, as described below. Later, it was confirmed by calculation
that the membrane stress in the interlayer and the tension in the cables when acting as
catenaries would be within acceptable limits. In this condition, large deflections reduce
the tension, so it was important to allow sufficient slack in the cables.
4. Prototype panels
Arup approached Kite Glass of Weybridge, Surrey, UK to discuss the feasibility of
making the triangular panels with SGP interlayer with extended tabs at the corners. Kite
Glass had amassed considerable experience in processing SGP and in toughening
shaped glass for architectural projects and volunteered to make a sample panel. The
extended tabs proved possible to produce with special supports during a vacuum bag
laminating process but they would require heating on site to form them to fit into the
anchorage slots. We were also concerned about the strength when tested at 400C, which
was a condition that had been established since the original design, from experience
with roof glazing and a number of tests.
635
Challenging Glass 3
4.1. Testing early prototypes

Kite Glass set up the first sample on a skeleton frame representing the corner fittings
and placed the whole assembly in their heat soak oven and raised the temperature to
400C. When the glass was broken with a centre punch, the panel collapsed and pulled
the tab off the narrow end of the panel. It clearly could not support additional weight
under those conditions and increasing the width of the tabs from the 50mm first tried
was not going to change the result.
Good results had been reported with embedding metal meshes into laminates, [2] so a
sample with perforated stainless steel mesh was made up and tested in the same way. In
this case the mesh proved to be the weak point and it was beginning to look like a
system that would be difficult to install anyway if a mesh had to be bent on site.
Figure 2a and b: The second prototype panel after hot testing and 18months exposure.
4.2. Wire rope anchor concept

A discussion in the Kite Glass factory, examining the broken samples, raised the
possibility of using wire rope rather than mesh or perforated metal. The rope would be
tougher and easier to connect to the fittings, but potentially difficult to bond to the
interlayer because the strands might pull out when loaded. A bevel on the edge of the
thicker pane, to create a pocket into which the wire rope could fit along the edges of the
panel offered a convenient location, and the length of bond could be extended without
encroaching on areas where it would be visible. There was concern about establishing
reliably the tensile capacity of a bond to wire rope, so the logical step was to continue
the rope around the panel in a continuous loop. Wire ropes are reliably spliced using
metal crimp sleeves for boat rigging, light aircraft control cables and stays, and many
other purposes.
636
4.3. Testing the wire rope concept

The first prototype with wire rope used old galvanized wire found in the maintenance
department at the factory and the ends were simply butted together near the middle of
the long edge of the panel. This prototype was broken and then loaded with four 25kg
bags of sand and left outside in the summer sun for two weeks. It gave confidence that
the concept would be workable.
Samples of 2mm and 3mm diameter stainless steel wire rope were obtained and
laminated along the edges of sample panels for testing. The cables were clamped to the
skeleton frame and the panels broken in the heat soak oven at over 400C. The sample
with 3mm diameter wire rope loop was able to support 100kg of distributed load for
over 2 hours. In fact the temperature rose to 540C and the panel supported the load until
it was removed 17 hours later at 300C.
Figure 3: Testing of production panels at 400C with both layers broken.
4.4. Durability
Later, the second prototype was removed to the factory yard and left for a number of
days under load. Eventually the load was removed but the prototype panel remained
exposed to the weather for over 18months and remained safely suspended. This allayed
an initial concern about whether the design relied on the bond between stainless steel
and SGP, which some researchers have found to be sensitive to moisture. Our
conclusion was that sufficient mechanical interlock existed to avoid reliance on
adhesion.
5. Development of the design
5.1. Wire rope loops

The technique to laminate the wire rope loops to the edges of the panels had to locate
the wire and ensure that it would be impregnated by and encapsulated in the SGP. The
rope type selected was 7x7 meaning seven strands each made of seven wires, which is
strong but reasonably flexible. It also has a comparatively open cross section, allowing
the interlayer to impregnate it and engage between the strands. The wire ‘gaskets’ were
procured from a specialist rigger and included a 1800 twist in the rope, so that three
crossing over loops could be formed, one for each corner. When made in this way, the
loops would lie naturally in a plane and could be temporarily fixed and then enveloped
637
Challenging Glass 3
in the vacuum bag for laminating. The crossing of the cables at each corner loop
ensured that peeling of the cable away from the edge of the glass was avoided when
loaded because there is a force component inward towards the glass edge. The finished
panels had a rounded loop of cable extending from each corner, ready to be located
within the stainless steel glazing fitting on site. The crimped sleeve splice was located in
one of the loops at a blunt corner of the panel, where the cable tension would lower than
at the narrow corner. By locating the splice consistently at one corner, it was also easy
to verify that the panels were installed the same way up.
Figure 4: Devising the arrangement of loops around a pulley with a rapid prototype fitting.
5.2. Anchor points for the panels

The existing stainless steel fittings on the bridge were to be re-used if at all possible, to
minimise cost, changes to the appearance and installation time. The fittings were held
closed by large cap head bolts, which would provide a convenient strong point for the
anchorage of the cables around the undercroft panels. When the panels were installed, it
was necessary to ensure that the wire loops from three adjacent panels would overlap
conveniently so that the fixing bolt would pass reliably through because it would not be
possible to inspect correct engagement once the fittings were closed. A nylon pulley
wheel, usually used as a piece of climbing equipment, was sourced as a locator for the
wire loops, which were sized to engage with it. The pulley also provided a large
diameter lead and some cushioning for the wire rope to maximise its strength under
dynamic load if ever called on to support someone on a broken panel. Clearance
between the pulley and the bolt and between the loops and the pulley provided some
slack in the system, which reduced the tension in the cables when acting to support the
glass by catenary action. Excess slack could allow dynamic loads to increase, which
would be damaging to the cables.
6. Testing
The following tests were carried out on three identical production panels to verify that
the robustness requirements would be met. The samples were maintained at 400C
throughout.
638
x Soft body impact test: 45kg, 3 times from 1.2m onto the top surface of the
centre of the test sample panel. No breakage occurred.
x Hard Body impact test: steel ball of 4.11kg dropped from 1.2 m at the centre
and at each of the corners. No breakage occurred, so the lower layer was
broken with a centre punch.
x Concentrated load test with one layer broken: 0.9kN on a 125mm square plate
at mid-span for 30 minutes. The panel supported the load with no further
breakage.
x Repeat Hard Body impact test: No further breakage occurred.
x Test for fragility: The upper layer of each panel was broken with a centre
punch within 50mm of the centre of the panel. A static load of 90kg, in the
form of two 45kg sand bags placed side by side on the centre line and either
side of mid span, was applied to the broken panel.
x The broken panels supported the imposed load beyond the required 30min.
The test sequence was successful and provided the confidence that the glazing for the
undercroft would be safe and robust for regular access for cleaning.
7. Manufacturing
A detailed survey by the glazing contractor had shown that each panel would need to be
unique, owing to construction tolerances of the original steelwork. Therefore, 360
different triangular shapes had to be made and correctly identified so that they could be
glazed in the correct position on the bridge. All panels were manufactured using a CNC
machine, which enabled the extra-large bevel on the edge of the glass to be controlled
within tight tolerances. If the bevel was too deep then the wire could look wavy and if
too small the wire would protrude along the edge and may not be attached correctly to
the interlayer.
7.1. Laminating
The panels above the deck did not have to support maintenance loads in the same way
because the top can be accessed from a mobile platform. These panels, known as the
Enclosure, were built to the same design but without the wire rope loops of the
Undercroft panels. These could be laminated in the conventional way with nip rollers to
assemble the panels then autoclaving to finish the laminating process.
For the Undercroft panels, the placing of the wire around the perimeter was a small
challenge in itself as each panel was a different size and had to match exactly the jointed
wire supplied by the rigger. This wire had to be placed very carefully into the edge of
the panel to ensure it overlapped the interlayer. The interlayer had to be cut exactly to
the shape of the triangular panel allowing just a small overlap on the size of the panel to
ensure it flowed around the wire. A novel method was devised to secure this wire and
allow the correct length of loop with a twist of 180 degrees. This process took several
attempts to get correct before proceeding with the actual panels for the bridge. Once the
two pieces of glass, interlayer and wire were assembled the whole package was sealed
into a vacuum bag made from aerospace film with aerospace breather material used all
around the perimeter, again to aid correct de-airing of the assembled panel and ensure
the bond with the wire was correctly assembled. The vacuum was maintained
throughout the autoclave process to ensure extraction of air as the interlayer softened.
639
Challenging Glass 3
Process checks included visual inspection with a magnifying lens to examine the flow
of interlayer around the wire, which was carried out on 100% of the first 20 panels and
subsequently on 20% at random. Small test samples were laminated along with each
batch and subjected to tensile tests to confirm that a consistent load was required to pull
the wire from the interlayer.
Figure 5: Wire loops secured around pulley.
8. Installation
The installation required Vitrine Systems to devise novel ways to support the gravity
load of the panels temporarily until each panel was correctly positioned and the joint
widths adjusted. The loops of wire were linked to each other with nylon ties during
installation, to ease assembly and ensure correct engagement of the fixing bolts through
the fittings.
The complex installation was completed on time and budget, and the access method into
the undercroft, supported by the new glass, was demonstrated to the client and the
building maintenance team.
9. Acknowledgements
The authors would like to acknowledge significant contributions to the success of the
project on the part of glazing contractor Vitrine Systems, main contractor Parkinsons,
and our client, Manchester City Council.
10. References
[1] Centre for Window and Cladding Technology, Safety and Fragility of Glazed Roofing; testing and
assessment, Technical Note 67, Bath, UK, 2010.
[2] To Increase the Residual Bearing Capacity of Glass with a Local Reinforcement, Neugebauer, j.,
Research in Architectural Engineering Series, Volume 1, 2007, EU COST C13 Glass and Interactive
Building Envelopes - Final Report, Crisinel et al Eds., ISBN 978-1-58603-709-3
640
doi:10.3233/978-1-61499-061-1-641
Through-Cracked Tensile Delamination

Tests with Photoelastic Measurements
Daniele Ferretti
Parma (I)
Marco Rossi
Glass Light and Special Structures Limited, London (UK)
Gianni Royer-Carfagni
Parma (I), gianni.royer@unipr.it
The critical mechanism in the post-glass-breakage response of laminated glass is

delamination of polymeric interlayer from glass shards. To investigate adhesion
properties between glass and PVB, “Through-Cracked-Tensile” tests have been
performed at room temperature, with different stain rate and polymer thickness.
Using photoelastic properties of PVB, the stress has been detected during the test,
while measuring force-displacement and force-delamination diagrams. Two
different failure modes have been recognized, influenced by the presence of friction
while relative sliding between the two materials occurs. A steady delamination
takes place only when friction is not present; on the other hand, friction can arrest
delamination, and the interlayer eventually breaks. Results have been interpreted
though a minimal model that predicts the delamination load.
Keywords: Laminated glass, interlayer, bond, debonding
1. The experimental procedure

1.1. Introduction
A critical mechanism in the post-glass-breakage response of laminated glass is the
delamination of the polymeric interlayer from the glass shards. In order to assess the
structural performance of laminated glass, including its capacity to adsorb impulsive
loading, a key issue is certainly the evaluation of the bonding performance of polymer
and glass. This effect has been studied by many authors. Suggested experimental
techniques are the most various and different in type, but two categories (this
classification is certainly not exhaustive) may be identified for those procedures referred
to as “Compression Shear Tests” (CST) or “Through-Cracked-Tensile” (TCT) tests.
In a classical CST apparatus [1], a small specimen of laminated glass is inserted in the
interface plane of two metallic units which is inclined at 45° with respect to the
compression loading axis, so that the compression and shear components acting in the
plane have approximately the same intensity. This test is simple and gives and average
value of the shear force W that provokes relative sliding of the plies for an interlayer
under a uniform state of strain. However, only partially can CST take into account the
641
Challenging Glass 3
influence of stress concentration at the front of delamination [2], that may provoke
progressive, “piecewise”, detachment of the polymer, at an average load level quite
lower than that producing the global detachment of the whole piece. Moreover, in CST
it is difficult to distinguish the contributions due to pure adhesion and Coulomb friction.
The purpose TCT tests is to measure the mechanical response of cracked glass bridged
by a polymeric interlayer ligament. In the experimental set up proposed in [3],
progressive delamination of the interlayer occurs. The modeling of this phenomenon is
based upon an energetic balance à la Griffith between the release of elastic strain energy
and the production of new delamination surface. This approach, based upon concepts of
fracture mechanics, naturally takes into account possible stress concentrations at the
delamination front and allows to define the key parameter to model the phenomenon,
i.e., the specific fracture energy * associated with delamination. Knowledge of * allows
the definition of the bonding strength under the most various boundary conditions,
possibly taking into account geometric and constitutive non-linearity in the polymer
response. In other words, the bonding strength is not an absolute value but depends
upon the specific stress configuration, possibly influenced by local effects.
In this paper, we present preliminary results for TCT tests performed on PVB-laminated
glass. Since the release of strain energy at delamination is strongly influenced by
interlayer thickness, four different heights for the PVB have been taken into account. To
account for the influence of manufacturing, specimens were prepared by two distinct
companies. At this stage, the influence of temperature was not considered (all tests were
performed at room temperature). To take into account viscoelastic effects [4], two
different strain rate were used. A simplified Linear Elastic Fracture Mechanics approach,
traditionally used to calculate the debonding strength of Fiber-Reinforced-Polymers
(FRP) attached to quasi brittle materials [5] like concrete has been adopted.
Viscoelasticity of the polymer has been considered, at least as a first order
approximation, by evaluating its secant elastic modulus at the different strain rates.
A peculiarity of the presented experimental campaign is the use of a polariscope.

Exploiting the high photoelastic properties of the PVB, it has been possible to recognize,
with extreme accuracy, the zone where delamination had occurred, and evaluate, at least
at the qualitative level, the state of stress in the polymer throughout the specimen.
Direction of principal stress and stress concentrations can be visually detected,
emphasizing their correspondence with the zones where delamination is triggered.
Test results have shown two different failure modes, which depend upon the presence of
friction while relative sliding between the two materials occurs. A steady delamination
takes place only when friction is not present; on the other hand, the presence of friction
can arrest delamination, and the interlayer eventually breaks.
The simplified model allows to readily calculate a reference value of the delamination
surface energy. Results appear to be sensible not only to strain rate but even more so on
the manufacturing, because sample prepared by different companies may give results
with noteworthy differences. In any case, the measured value are in agreement with the
order of magnitude recorded in the technical literature [3].
642
Through-Cracked Tensile Delamination Tests with Photoelastic Measurements
1.2. Specimens
Tests have been performed on 50×200 mm laminated specimens made with two glass
plies of thickness hg = 6 mm each and one PVB interlayer (Figure 1). Four samples,
labeled “A”, “B”, “C”, and “D” have been prepared with interlayer thickness t equal to
0.3, 0.76, 1.52, and 3.04 mm, respectively, with ten specimens for each class. The TCT
tests requires that both glass layers are pre-cracked in the middle, so that the two
resulting parts are connected by the interlayer only. Particular care had to be paid in pre-
cracking to avoid damage of the interlayer by contact with the sharp edges of glass.
To avoid slippage between the specimen and the wedge grips of the testing machine,
that represents a critical issue [3,6], specimens were provided with aluminum 50×50
mm tabs glued with an epoxy adhesive (Loctite 3425 A&B Hysol), allowing at least two
days for complete hardening of the adhesive. Sandpaper was used to improve friction
between tabs and grips of the testing machine.
Figure 1. Specimen description and relevant parameters for delamination.
1.3. Test setup

Tests were carried out in environmental laboratory conditions (temperature T = 19±1 °C
and relative humidity RH = 30±3%). Specimens were mounted on a universal testing
machine (INSTRON 8862) with wedge action grips. The testing machine was driven
controlling the displacement of the actuator at constant speed up to failure. Two
displacement speeds have been chosen, here referred to as “slow” (0.013 mm/s) and
“fast” (0.26 mm/s).
The evolution of delamination was recorded with a reflex camera (Canon EOS-1)
mounting a macro objective. The camera was connected to a computer to automatically
take pictures every 5 s and 20 s for the fast and slow tests respectively. Image analysis
permitted to measure crack width G and delamination fronts a1 and a2 (Figure 1). Some
643
Challenging Glass 3
specimens were contemporaneously observed with two distinct cameras, oriented at

right angle to analyze the front and the side of the specimen. This permitted to observe
the stretching and shrinking of the interlayer while detachment from glass occurred.
The photoelastic properties of PVB, already evidenced in some studies [7], was
exploited to observe the interlayer stress field during delamination in order to highlight
stress concentrations and asymmetric behavior.
A circular polariscope (Tiedemann & Betz) was employed. The polariscope is

composed of:
x White monochromatic light source (sodium bulbs);

x Milkglass diffusor placed in front of the light source;
x Rotating polarizer (calcite crystal plate) combined with a first quarter-wavelength
filter placed in front of the diffusor;
x Compensation analyzer combined with a second quarter-wavelength plate.
Both polarizer and analyzer are able to rotate and to pivot for Senarmont and Tardy
compensation [8].
Specimen was placed between polarizer and analyzer (Figure 2). With crossed axes of
polarized and analyzer, the polariscope produces images with dark background and dark
fields of unstressed zones, whereas stressed zones appear light. In particular, colored
fringes, called isochromatic, represent loci of equal principal stresses difference, i.e.
equal in-plane maximum shear stresses. Pivoting and rotating both polarizer and
analyzer lead to isoclinic fringes, i.e. loci of points of equal principal stress directions
[8]. Examples of isoclinic and isochromatic fringes are reported in Figures 3.
Figure 2a, b. Test setup and detail of the specimen within the polariscope.
644
Figure 3a, b. a) Isocliclinic fringes (axis of polarization vertical and horizontal; b) isochromatic fringes.
2. Test results
Forty specimens were tested. Most of them presented an anomalous behavior, with the
premature onset of a crack at one side (Figure 4.a) or in the middle of the interlayer
(Figure 4.b). The problem was more frequent in specimens “A” and “B”, which were
characterized by thin interlayers that seem more sensible to scratches produced during
pre-cracking. A few specimens only displayed a symmetric behavior (Figure 4.c). In the
following we limit to comment the most representative tests.
Figure 5 shows the Force vs. total elongation (FG) diagram for specimen CS1. Here,
the label stands for C series, Slow test, specimen #1; a similar reference will be used in
the following. Photos of delamination surround the curve. From the corresponding
pictures of ongoing delamination, one can notice that at a certain load level
delamination stopped, so that further elongation was due to interlayer straining only.
Failure occurred with fracture of the interlayer. A similar behavior was observed in
specimen CF2 (Figure 6), which was characterized by the same interlayer thickness but
fast test velocity. In this case the failure load was slightly higher, probably due to the
viscoelasticity of the polymer
However, an identical specimen, CF1, presented a different behavior because

delamination never stopped propagating. Correspondingly, the FG diagram presents a

peak followed by a horizontal plateau (Figure 7) associated to delamination progression,
a typical response already observed in the literature for TCT tests [6]. Delamination
stopped only when reaching the grips of the testing machine: at this point straining of
the interlayer occurs up to failure, like in the specimen CF2. This ascending branch is
not associated with delamination.
The behavior of specimens CF1 and CF2 is compared in Figure 8 through force-
displacement and force-delamination curves, using the same parameters defined in
Figure 1. It is possible to notice that in case of specimen CF2 (Figure 8a), the
delamination widths a1 and a2 propagated up to a certain load level, then they stopped
completely. In the case of specimen CF1 (Figure 8b) a delamination front stopped while
the other continued to propagate.
The different behavior of identical specimens might be caused by heterogeneities of the

interface related to lamination processes. Moreover, the considerable deformations of
645
Challenging Glass 3
the interlayer sometimes caused a small rotation of the specimen, so that interlayer
touched the sharp edges of one glass ply, engendering friction that can affect or even
stop delamination. In Figure 9 a sewing thread shows the gap between interlayer and
glass in the case of specimen CF1, which delaminated completely.
Specimen BF1, which is characterized by a thinner interlayer, presented an identical

response (Figure 10).
Also specimen DF1 showed a similar curve, with a peak followed by a sub-horizontal
branch (Figure 11). In this case, delamination was less pronounced and failure of the
interlayer occurred while in the plateau. Pictures evidence a different shape of the
delamination front. Indeed, series A-B-C and series D were prepared by different
companies; in particular, in sample D the polymer completely filled the gap at the
chamfered edges of the glass, thus increasing bond strength. Isoclinic fringes, obtained
with axes of polarization vertical and horizontal, permit to observe also the propagation
of the stressed zone ahead of the delamination front.
Specimen DF2 presented an intermediate behavior, with a plateau followed by a gradual

hardening (Figure 12).
Figure 4a, b and c. Responde of the interlayer: Crack at one side, in the middle, and symmetric propagation.
646
Figure 5. Force vs. displacement curve (specimen CS1), with pictures of the delamination.
Figure 6. Force vs. displacement curve (specimen CF2), with pictures of the delamination.
647
Challenging Glass 3
Figure 7. Force – displacement curve (specimen CF1).

Figure 8a,b. Crack width G compared with debonding propagation a1 and a2 for tests CF1 and CF2.
Figure 9a, b. Gap between interlayer and glass evidenced by the sewing thread.
648
Figure 10. Force – displacement curve (specimen BF1).

Figure 11: Force – displacement curve (specimen DS1).
649
Challenging Glass 3
Figure 12. Force – displacement curve (specimen DS2).
3. Discussion. Interface energy and evaluation of delamination load

The problem of debonding has been investigated in several fields. One of them is the
case of debonding of a carbon fiber plate (FRP) glued on a reinforced concrete substrate.
In this case, one of the simplest models is the one firstly proposed in [5] and adopted
also in some Code Standards. The concrete substrate is considered rigid and
delamination is described by means of a triangular bond-slip curve characterized by the
interface energy *.
The debonding load Fu may be computed as
‫ܨ‬௨ ൌ ܾξʹȞ‫ ݐܧ‬, (1)
where E = Young’s modulus of the FRP plate, t = thickness of the plate, b = base of the
plate. The only difference with respect to TCT tests is the fact that debonding occurs on
one surface instead of two, i.e. the model applies for the symmetric part of TCT test.
The value of interface energy * can be determined from Equation (1) as
ଵ ிೠమ
Ȟൌ , (2)
ଶ ଶ௕మ ா௧
where del coefficient ½ is introduced to consider two delamination surfaces. Equation

(2) requires the value of the effective secant modulus E (including viscous effects),
which depends on the strain rate. This can be determined from the experimental stress-
strain relationships depicted in Figure 13, where V = F/(b t) and H = G / (a1 + a2).
650
The results are summarized in Table 1. One must notice that the measured thickness t of
PVB is thinner then its nominal value. Table 1 shows the values of interface energy *
obtained through Equation (2). Results are rather scattered, confirming the dispersion of
the tests. Anyway, the order of magnitude of *is in agreement with data recorded in the
literature [3], which spans from 0.2 to 1.0 N/mm.
Figure 13. Experimental stress-strain relatioships.
Table 1. Test results.

Specimen Speed B t E Fu Vu *
[mm/s] [mm] [mm] [MPa] [kN] [N/mm2] [N/mm]
CF1 0.26 50 1.5 11.8 0.35 4.7 0.692
BF1 0.26 50 0.65 10.8 0.25 7.7 0.890
DS1 0.013 50 2.5 13.8 0.70 5.6 1.420
DS2 0.013 50 2.5 9.3 0.46 3.7 0.905
4. Acknowledgements
The authors acknowledge the Italian MURST for support received under the PRIN2008
program. We are grateful to Annalisa Franco who kindly assisted us in some of the tests.
651
Challenging Glass 3
5. References
[1] Schneider, K., Lauke, B., Beckert, W., Compression Shear Tests (CST) – A convenient apparatus for the
estimation of apparente shear strength of composite materials, Applied Comp. Materials, 8, pp. 43–62,
2001.
[2] Jagota, A., Bennison, S.J., and Smith, C.A., Analysis of a compressive shear test for adhesion between
elastomeric polymers and rigid substrates, International Journal of Fracture 104: 105–130, 2000.
[3] Muralidhar, S., Jagota, A., Bennison, S.J., and Saigal, S., Mechanical behaviour in tension of cracked
glass bridged by an elastomeric ligament, Acta Mater., 48, pp. 4577–4588, 2000.
[4] Hooper, P.A., Blackman, B.R.K., Dear, J.P., The mechanical behaviour of poly(vinyl butyral) at
different strain magnitudes and strain rates, J Mater Sci, 47, pp. 3564–3576, 2012.
[5] Bronsens, K., van Gemert, D., Anchorage design of externally bonded carbon fiber polymer plates,
Proceedings of the Fourth International Symposium on FRP Reinforcement for Concrete Structures
(FRPRCS-4), American Concrete Institute (ACI), Farmington Hills, MI, 635–645, 1999.
[6] Delincé, D., Sonck, D., Belis, J., Callewaert, D., Van Impe, R., Experimental investigation of the local
bridging behaviour of the interlayer in broken laminated glass, Proceedings of the Third International
Symposium on the Application of Architectural Glass (ISAAG), Munich, Germany, pp. 41-49, 2008.
[7] Hooper, P., Arora, H., Dear, J. P., Blast and Impact Resistance of Laminated Glass Structures,
Proceedings of the IMPLAST 2010 Conference October 12-14 2010 Providence, Rhode Island USA,
2010.
[8] Kobayashi, A.S., Handbook on Experimental Mechanics, 2nd, Revised Edition, Wiley-VCH Verlag
GmbH, 1993.
652
doi:10.3233/978-1-61499-061-1-653
Composite Structures Made of Glass and

Ultra High Performance Concrete
-Current Research-
Bernhard Freytag, Gerhard Santner, Lutz Sparowitz
Graz University of Technology, Austria, freytag@tugraz.at
Franz Götschl
Civil Engineer, Austria
Glass can be connected with UHPC in terms of direct bonding contact. Structures
built according to this technology are elegant and highly transparent but beyond
that they provide robustness similar to reinforced concrete. This paper focuses on
the fundamental mechanics of the bonding connection, which is the key element of
this building method. The design and assembling concepts are explained by means
of some prospective applications and an interesting pilot project. It is an 18.5 m
high noise barrier planned as standing folded plate. Finally, a detailed outlook
about an experimental programme concerning durability and fatigue of the bonding
glass-UHPC connection is given.
Keywords: glass, UHPC, composite, bond, folded plate,
1. Introduction
The brittle behaviour of glass led to the tradition that in the majority of cases glass is
flexibly supported by means of elastic sealing for instance. In this way enforcements
due to thermal deformations or other actions are minimised. And of course these
flexible connections between glass and its neighbour elements delimit also the
contribution of glass to load carrying tasks. Thus, the glass-friendly in-plane actions
were not activated for a long time.
The more and more growing desire for totally undisturbed transparency and advanced
elegance led to all-glass structures. These are mainly based on glued connections and
implicate better utilisation of the beneficial in-plane resistance of glass panes. But from
the point of view of safety all-glass structures are poor as the breakage of one pane can
induce a total collapse. In order to overcome that drawback, the safety was increased by
means of integral design concepts. These require the possibility for load redistribution
towards healthy carrying elements if one element fails. As a consequence, the economic
efficiency dropped down. Further technologies, also aiming at improved safety, are:
x Use of laminated safety glass (enhanced residual carrying capacity)

x Micro-reinforced laminated glass [1], [2]
x Steel reinforced glass beams [3], [4], [5]
x Glass-timber composite techniques [6]
653
Challenging Glass 3
All these concepts are based on glued connections between the composite materials.
The glues are always of organic nature and are highly sensitive to elevated temperature
and in particular to fire.
Linear Clamping is a well-proven method in the context of simple glass roofs. But the
flexibility of the interlayers used does not allow for the introduction of considerable in
plane loads. The glass-concrete composite technology is also based on the principles of
linear, clamped connections but beyond that it plays on the castability of concrete
resulting in a form-fit and rigid link [7] [8].
Ultra high performance Concrete (UHPC) is predestined for this type of composites. It
stands out in terms of compression strength, adhesive power and durability. These
properties allow for a filigree design of the concrete parts in order to keep the
transparency of the structure as far as possible. Of course, the appearance of glass-
UHPC-buildings may not be compared to all-glass-structures but it meets high demands
on transparency and initiates a new exciting architectural design where “glass grows out
of stone”.
2. Structural and assembling issues of glass-UHPC composites
2.1. Structural behaviour

The main concept is the linear continuous load transfer from glass panels into concrete
borders through a direct bonding connection. The edge of the glass is perfectly fitted in
a slot of the UHPC part by embedding it in fresh UHPC. The general functionality of
such structures can be explained by means of a simply supported 8m spanning beam
taken from [7] [8] (Figure 1). The glass web carries the shear forces, while the UHPC
borders act as compression flange, and in combination with the reinforcement
(prestressed or conventional) as tension flange respectively. Thus, the UHPC has to act
as structural element itself and has to provide the link to the reinforcement or other
structural elements. The glass surface in contact with the UHPC is pre-treated, specially
roughened. An extensive experimental study showed that an enamelled coating with
fine corundum offers the best capacity of the connection, but also sandblasted surfaces
are well applicable.
Figure 1: Glass-UHPC beam spanning 8m Figure 2: Experimental results of the 6 point

bending test
The experiments on glass-UHPC-composite beams show the impressive load carrying

behaviour (Figure 2). When cracks in the glass web occur, the load can still be increased.
This is an important result regarding the wish for a better safety level in structural glass
654
Composite Structures Made of Glass and Ultra High Performance Concrete -Current Research-
engineering. The load carrying capacity does not directly depend on the tensile strength
of glass.
2.2. Fabrication and Assembling

UHPC is a highly developed material and requires extra care in preparation and
handling [9]. UHPC differs from ordinary concrete in various points:
The packing density is optimised by means of precise grain size distribution in the
nanometre scale. Therefore quartz powders are used as fillers and silica fume works
both as filler and as binder in addition to cement. Of course these fine particles enlarge
the internal surface exorbitantly and therefore require much more water for getting a
workable mixture. Quite contrary to that, the amount of water related to the binder is
lowered significantly in order to avoid the formation of pores due to diffusing excess
water. The workability needed is ensured by adding high performance plasticiser based
on poly carboxylate. The quality and the workablility of UHPC also depend highly on
the mixing technology and the mixing procedure. If all these aspects are considered
carefully UHPC can become self-levelling and does not need further compaction. At the
end of the mixing procedure micro steel fibres are added for getting a sufficient post
cracking tensile strength on the one hand and compression toughness on the other hand.
It is obvious that the orientation of the fibres affect the tensile behaviour. Thus, it is
very important to cast UHPC according to strict specifications, which have to ensure
that the fibres align themselves in a desired kind.
All these UHPC-specific challenges clearly show that glass-UHPC members have to be
fabricated in precast plants. The formwork must always be of a two-part type and has to
be equipped with sealing at the contact line to the glass. The two parts are then pressed
together and clamp the glass in between. One has to note that both glass surfaces to be
wetted must be arranged vertically in order to achieve total wetting without bubbles.
The prefabricated glass-UHPC elements must be assembled on site and need therefore
practical connections. There are different possibilities of joint design and joint
arrangement with regard to the assembling.
The tasks of a longitudinal joint are comparable to those of an edge of a folded plate. It
has to transfer shear mainly, which can simply be attained by means of a high-strength
friction grip fastening. Lateral joints mainly have to transfer axial forces; tension or
compression. A butt joint combined with prestress applied by bondless inside tendons is
a well working solution for that but it needs either a high strength equalising layer (wet
joint) or extra low manufacturing tolerances (dry joint). Dry joints can be realised by
CNC-grinding indeed but it is very costly so far [10]. Last, it should be mentioned that
lateral joints can also be replaced by longitudinal overlapping joints in cases where
more than two glass-UHPC elements are involved (see also the example in 5.1).
Summing up, there is a wide range of mounting techniques but all of them except for
wet joints require high fabrication accuracy in order to minimise assembling
enforcements.
3. The mechanics of the force-fit bonding connection between glass and UHPC
The connection generally works in three ways: adhesion, interlocking and friction.
Before each mechanism as well as their interaction will be explained, a clear
655
Challenging Glass 3
terminology regarding adhesion and bonding connection shall avoid misunderstandings

(see also Figure 3): Adhesion means attraction directly in the interface and Connection
covers the entire mechanisms for the transfer of loads.
It is a well-known fact that friction only works if normal compression forces act on the
contact surface. In the case of glass-UHPC-connections, compression is induced by
shrinkage of the UHPC part. The width of the slot intends to become smaller, which
leads to clamping the embedded glass edge. Interlocking, on the contrary, means that
shear forces can be transmitted requiring neither adhesion nor clamping force.
The resistance of the connection to shear forces can be determined by simple “Push-
out” tests, but one does not get any information about the proportions of the three
mechanisms [7]. Focusing on the separation of the three mechanisms Plociennik carried
out numerical simulations [11] [12]. At least the separation of adhesion from friction +
interlocking was possible and is here presented in Figure 4.
Figure 3: Distinction between connection and adhesion Figure 4: Results from push-out test [7] with
indication of the three mechanisms.
3.1. Adhesion
In general, adhesion is based on intermolecular attraction. This topic with respect to
glass and UHPC has been explored in the project AdBond [13]. Glass and steel were
used as bond partners of UHPC. Both, adhesion on smooth surfaces and on rough
surfaces have been studied. Some of the most important findings are given here:
According to thermodynamic approaches, theoretical adhesion strength can be predicted

by means of calculating the work of adhesion, which is related to the surface energies of
the materials [14]. Depending on a theoretically based estimation of the intermolecular
distance (1,4·10-12m) the theoretical adhesive power is about 90 N/mm2 [15]. Of course,
this theoretical value is always far higher than real adhesion strengths, according to
Griffith [16].
From a practical point of view the adhesion strength of UHPC on the investigated
materials is still very high. The adhesion strength can reach or rather exceed the tensile
strength of UHPC. Of course this is tested on rough surfaces but also on polished steel
surfaces when the young UHPC is carefully cured. A factor of about 30 between
theoretical and real values has been determined from the comparison to adhesive tension
656
tests under nearly perfect conditions. The adhesive shear strength correlates with the
adhesive tensile strength by the friction coefficient P. This interesting phenomenon is
comprehensively explored by the author in [15]. The explanation is based on theoretical
studies concerning the molecular roughness of macroscopically smooth surfaces
verified by findings from atomic force microscopy taken from literature.
The shrinkage of UHPC, which always occurs during hardening, limits the strength of
the adhesive bond significantly. Cracks arise a few hours after the making of the contact
between the pasty UHPC and its partner surface. Until this moment UHPC can easily be
deformed and as a consequence no enforcement will come into being. But, in the further
course of the hardening process each crack becomes an origin for stresses in the
interface. They try to separate the UHPC from its partner surface. This phenomenon has
extra relevance in glass-UHPC-composite connections because of the large linear
dimension of the connection.
Relative slip of UHPC is reduced considerably by roughening the surface. Thus a

substantial part of the adhesion strength can be maintained. To get an idea of the
practical adhesion, results of direct tensile tests of glass-UHPC specimens are shown in
Figure 5. The treated glass surfaces in Glass-Concrete-Composite connections also
provide a significant grade of roughness, hence it is also adhesion that accounts for the
strength of the connection.
Dehydration protected
12.0 Roughness R0 -
smooth
11.0 Roughness R7 -
fine sandblasted
10.0
Adesive Tensile Strength [N/mm²]
9.30 Rougness R3 -
9.0 coarse sandblasted
8.66
8.0
UHPC Layer Not dehydration
7.0 protected
6.0
5.38
5.0
4.54
Glass 4.0
3.0 2.78
2.0 1.78
1.0
0.0
Tests
Figure 5: a.) Pull-off test, b.) Failure surface (Roughness R7), c.) Test results UHPC-glass
3.2. Friction
As already mentioned, the friction component is activated by a clamping force, which is
induced by the shrinkage of the concrete part. The magnitude of the clamping force
depends on the geometry and on the mechanical properties of UHPC and glass. The
embedding depth (tE) of a glass pane in a concrete border was optimised by means of a
numerical parametric study with regard to the clamping force in the scope of a diploma
thesis [17]. Whereas material parameters as well as the glass thickness are kept constant,
the embedding depth tE as well as the complementary depth (tB) are varied (Figure 6).
Both variants - keeping the height hB of the UHPC part constant or keeping the
complementary depth tB constant - have been investigated.
657
Challenging Glass 3
It turned out that it is not the embedding depth tE being the decisive geometric value, but
the complementary depth tB. The deformed figure of the numerical model clarifies that
fact (Figure 6). Due to shrinkage an open gap comes into being. The effective contact
zone tE,eff does not change considerably with the investigated variants.
The stiffness of the UHPC-part, characterised by tB, is essential for the magnitude of the
clamping force. High tensile stresses lead to concrete cracking near the front face of the
glass pane. If tB is too low, not enough stress is kept and the residual clamping force
will decrease. Thus, when keeping the height hB constant, a lager embedding depth
seems to worsen the effectiveness of the connection in terms of a lower clamping force
(Figure 6 right). In the case of the glass thickness tg=20mm the maximum clamping
force is observed at an embedding depth of 20mm. This can be reduced to the full
exploitation of the embedding depth. Further investigations show that the depth of the
effective contact zone generally agrees well with the thickness of the glass.
6
Figure 6: Finite element study on the dependency of the clamping force on the embedding depth
3.3. Interlocking
The interlocking mechanism works due to form closure. It needs neither intermolecular
attraction nor active compression force (clamping) in the contact surface. In [18] the
interlocking mechanism is considered by means of a global coefficient of friction global.
The mechanism is described using a simplified “tooth – model”, which approximates

the rough surface’s topography. The global coefficient of friction depends on the tooth
inclination as well as on the local coefficient of friction local, which acts in the
inclined surfaces of single teeth. global is calculated by transforming the clamping force,
which acts perpendicular to the global surface, to the inclined surface (Figure 7, eq.(1)).
(tan D Plocal )
P global (1)
(1 Plocal tan D )
If small clamping forces are of interest, this approach does not describe the mechanism
sufficiently, because the case without any clamping force will result in a prediction of
zero shear capacity.
In reality the interlocking mechanism works indeed without any active clamping force.
If the surfaces do not give way in normal direction, for geometric reasons sliding in
longitudinal direction is not possible. If shear loads are applied and the local maximum
658
force of friction is exceeded, relative slip between glass and UHPC intends to occur
(Figure 8). But slip in longitudinal direction (l) always requires a lateral movement (n)
of the UHPC flanks. These movements need a widening of the slot, which is restrained
by the lateral stiffness of the UHPC section as already seen in Figure 6. If it is strong
enough, an additional lateral force comes into being (FC,ind). It is denoted as passive or
induced clamping force. Assuming an infinitely stiff UHPC part, a maximum of passive
clamping force can be induced. In this case the strength of the connection is limited by
the failure (shearing off) of the UHPC teeth. Thus, the shear strength of the UHPC teeth
is the upper limit of the connection strength. In contrary, if the stiffness of the UHPC
part is rather low, the slot opens wide and the glass pane will slide through.
Figure 7: Global friction model [18] Figure 8: Induced clamping force
Summing up, the main parameters influencing the interlocking are the inclination of the
teeth (topography of the rough surface), the local coefficient of friction local and the
complementary depth tB. The contribution of the induced clamping force to the strength
of the connection can be subdivided into two main effects. The lateral compression
increases the shear strength of the UHPC teeth on the one hand and induces an
additional friction component on the other hand.
4. Prospective application
Glass-UHPC composites can be applied in all cases where thin-walled cross-sections

are useful from a structural point of view. That can be box girders, deep beams,
polyhedral shells and folded plates. Some prospective projects shall give an idea of
practical design work carried out during the last years.
Our first example is a self-supporting elevator tube (Figure 9a). In comparison to

classical glass-steel elevators the glass-UHPC composite alternative catches the eye
with its plain design without architectonic interference due to diagonal bracings. This
standing box girder consists of plane glass-UHPC elements, that are longitudinally
jointed by means of a bolted high strength friction grip connection (Figure 9b). Each
standard element is as high as two storeys. At the ground two elements of half the
height are arranged in order to join only two elements each storey. This overlapping
assembling method makes the exchange of a single element easier and lowers stress due
to shear forces.
659
Challenging Glass 3
Figure 9: Elevator Tube: a.) Rendering b.) Cross section of the edge Figure 10: Tower “Periscope”
The second project is the result of a student work [19] in the scope of the Austrian
Concrete Student Trophy 2010 (Figure 10). This building, called “Periscope”, is a look-
out, which can be used without going up. Standing on the ground one can appreciate the
nice view of the roofs of Graz by looking into mirrors. Thus, this tower can also be used
by handicapped people. This vertical inversion of the visitor’s location is further
transferred to the architectural design by decreasing the transparency and lightness
continuously from the bottom to the top. From the structural point of view it is again
conceived for the glass-UHPC segmental method of construction. The mounting is
based on pre-assembled tube-like segments, which are tied by subsequently threaded in
and tensioned strands.
A further impressive concept of a beam like structure is a footbridge between two

buildings at the University of Graz spanning 30 m (Figure 11). The bridge is again
planned as a transparent box girder with edges made of UHPC. The floor is completed
by folded perforated steel sheets which provide room for electrical installations. The
assembling follows the same principles as explained for the periscope. At the ends of
the bridge solid portals brace the cross section. The whole bridge including the portals
can be mounted at the ground and then lifted to its final position. The portals disappear
in the wall’s thickens of the existing buildings. A preliminary estimation of buckling

risk was done according to methods of steel construction.
Many other applications are imaginable. Glass-UHPC roofs can be constructed as

folded plates or conventional shed roofs could consist of glass-UHPC beams as vertical
components and traditional slanted roofs in between. The same light bringing
functionality can be utilised in office buildings where long bands for indirect light can
act as load carrying girders at the same time. Walls and facades made of glass-UHPC
elements can be designed as load-bearing elements covering additional bracing
functions at least.
660
Bolts
UHPC parts
Glass panes
Glass stiffener
Handrail
Figure 11: Footbridge University Graz; Visualisation and Cross Section
5. Pilot project noise barrier

The noise barrier presented here is part of a big housing project close to a motorway
located in Vienna. Due to the arrangement of the buildings the noise barrier is divided
into 4 sections. It is at maximum 18.5m high and measures about 100m in over all
length. For purposes of birds’ safety a print has to be applied onto the glass.
5.1. Structural Development

The wish for a high benefit from the structural advantages of glass-UHPC composite
technology led to the standing folded plate structure (Figure 12). In the first draft a
concrete plate was conceived at the top linked to the neighbour buildings at wall section
1 to 3. This plate should work as a lying beam supporting the folded plate.
Figure 12: Noise barrier in Vienna (source: ArchitekturConsultZT)
661
Challenging Glass 3
top plate
hor. joint with

prestressing rod
a.)
to ta
l to
lera
nce
ideal contact plane
f in
1
nt
al
me
en
UHPC ele
for
ele
ent
cross section me
ce
nt
me
cem
2
nt
for
l en
fina
b.)
Figure 13: Arrangement of joints Figure 14: a.) End fixing of glass, b.) diagram of
transposition according to Williot
The width of the composite elements is limited by the glass dimension to about 3 m.
The height of the wall is preliminarily divided by three, which results in about 6 m
height of the elements (Figure 13). Due to aesthetic reasons the concrete borders are
only arranged at the edges of the folded plate. Thus, upper and lower elements have
UHPC-borders at three edges, middle elements only at two. For purposes of easy
exchange of damaged elements the horizontal joints are planned with an alternating
offset of half the height. Thus, the axial forces in the UHPC-borders are not transferred
through the end face but via overlapping through the neighbour border. This technique
makes it possible that the composite elements can be longitudinally pretensioned.
Further, the more complicated threading in of monostrands throughout the full height of
the wall as well as the costly CNC-grinding of the end faces are avoided. Of course, the
prestress introduction brings about stress concentration at the ends of each single
element. The prestressing force must spread towards the glass pane, which results in
highly concentrated shear actions on the bonding connection. This local problem is
solved by means of “end fixing” of the glass pane. Figure 14a shows the slanted corner
combined with U-like reinforcement bars. The horizontal force component is carried by
a transversal tensioned thread bar running in the gap between two elements. This
transversal prestressing element is planned for any case. It works like a stirrup in the
context of structural concrete and enhances the bearing behaviour in the case of glass
breakage.
The longitudinal (vertical) assembling connection has to meet several demands. Beside
the shear bearing requirements (planned as high-strength friction grip fastening) it has to
provide possibilities for tolerance compensation; tolerances in width and in the folding
angel. Finite element analyses verify that especially width tolerances cause critical
membrane tension if tolerances cannot be compensated for. The diagram of
transposition according to Williot illustrates how the compensation plane has to be
oriented in order to avoid in-plane enforcements entirely and to portion the bending
662
enforcements equally to both elements. A kind of linear hinge, realised through a

rounded contact covers the angle tolerances. Figure 15 shows the detail marked in
Figur 13 for two variants following these principles. Both solutions locally need small
steely filler plates which have to be installed after mounting the elements but before
tensioning the lateral thread bars. They ensure the introduction of transversal prestress.
The big advantage of hexagonal-like cross section is the reduction of the number of
filler plates and the direct load transfer without flexure of the cross-section.
Filler plate
Compression
trajectories
Filler plate
a.) b.)
Figure 15: Two variants of cross-sectional formation: a.) triangle-like, b.) hexagonal-like
5.2. Some aspects of the structural analysis

x Proof of principle glass stress
The main action results from wind. Therefore the glass panes not only have to bear in-
plane loads from the folded plate mechanism but also bending action between the
folding edges. As a two panes laminated glass is used, the flexural stress is calculated
by hand according to [20], where the stiffness of the interlayer is taken into account.
Since the supported edges of both panes are fixed in the UHPC border, the panes cannot
slip with respect to each other and therefore the shear stiffness of the interlayer is not
relevant any longer at these edges. This shear rigidity at the edges reduces the flexural
stress by 17 %, calculated with FE. This positive effect is not considered in the design
but provides still safety reserves.
The membrane stresses from the global system (folded plate) are determined on a FE-
model considering a monolithic glass pane. For the final proof the stresses are
superposed with them of lateral bending in particular points and orientations. One has to
note, that the glass has reduced strength due to roughening at the edges as well as due to
the enamel based print (bird’s safety) all over the element. In order to compensate for
this strength reduction, the print is applied onto the inner glass surface. As a
consequence, the stress proof had to be processed at two points throughout the thickness.
As mentioned in the introduction crack formation in the glass can generally be allowed
for glass-UHPC composites. However, in this pilot project the glass web is not
continuous over the entire height of the wall, which is the reason for prohibiting glass
cracks in the serviceability as well as the ultimate limit state.
x Proof of the bonding glass-UHPC connection

The actions on the bonding connection are analysed by the use of a linear elastic finite
element model considering rigid connection behaviour. The resistance is composed of a
663
Challenging Glass 3
global coefficient of friction Pglobal=2.5 determined from former push-out tests (see also
Figure 4) and a clamping force fclamp calculated on a cross-sectional FE model of the
hexagonal type. With n= number of compressed contact surfaces and Jfric.=partial
safety factor the design value of the connection’s shear resistance results in
P global f clamp 2,5 0,6

tconn, Rd n 2 2,0 kN / cm (2)
J fric. 1,5
Because of the lack of experience and the cyclic nature of the wind load, the
contribution of adhesion to the capacity of the connection is not used. At the upper and
lower ends of the connection a certain amount of slip is allowed for redistribution of
shear peaks towards the u-shaped end fixing reinforcements.
x Proof of buckling
The principle compressive stress caused by wind as well as by prestressing intends to
buckle the glass pane. This phenomenon is studied on a single glass-UHPC element by
means of finite elements loaded through the worst section forces gained from the linear
global model. The most important resistance to that is provided by the flexural stiffness
of the pane. Therefore the rather soft bending behaviour of laminated glass has to be
considered again. To simplify matters, we determined an equivalent monolithic glass
thickness based on equal deflections for the case of the given span. The final calculation
is based on linear material but nonlinear geometry solved by the use of the Riks
algorithm (arc length method). As a result it can be stated that the buckling load is far
beyond the design load.
5.3. Final details and key data of the project

The structural solution presented in 5.1 was the bases of cost estimations calculated by
companies for the purpose of comparing it to conventional glass-steel variants. The
estimator was put off by the tricky shape of the UHPC cross section, which resulted in
very high formwork costs and consequently in the discarding of this solution. However,
far-reaching modifications in the structural concept, aiming at shapes which are more
acceptable for practitioners, brought about the final solution.

As shown in Figure 16 the edges of the folded plate are disintegrated by arranging a
separate column. These pretensioned columns run over the entire height and are turned
outwards which forms a very different but also pleasant design.
Figure 16: Part of wall 2, ground view
The concrete borders have now rectangular sections and are equipped with steely built-
in parts along the opposite side of the glass’ embedment (Figure 17). The outer edge of
these built-in parts have round steel bars which the element is clamped with onto the
664
respective column by means of clamping plates and bolts. This type of steel-steel
clamping connection offers all possibilities for tolerance compensation as discussed in
5.1 but beyond this the folding angel may vary in a wide range. This feature makes it
possible that only one size of glass-UHPC element suffices to build the wall. The
different spacings between the buildings (Figure 12) are considered by different folding
angels. It is obvious that the horizontal joint now runs through the entire length without
offset.
Figure 17: Final details
As a consequence of the disintegration of the folding edge the transversal prestress

cannot realised with one single bar per horizontal joint any longer. The modified detail
is illustrated in Figure 17 at the right. A short steely duct is welded onto the front face of
the built-in part for anchoring the prestressing thread bar. A broader introduction of the
prestress is ensured by a pair of short flanges. Now each composite element has its
“own” transversal prestressing bars (at the top and the bottom edge) which can be
tensioned comfortably in the plant before transportation. The concept of end fixing of
the glass remains but the reinforcement bars are welded onto the duct and run
backwards in order to avoid disturbance of the UHPC flanks beside the glass.
The final end bracing at the top of the folded plate is reduced to a plane truss consisting
of the horizontal UHPC-borders of the top elements and additional steel tube bars tying
the edges. At the bottom, one row of the columns is totally fixed whereas the other row
is supported in a way that tension and compression can be carried but horizontal
movements are possible. This is for minimising thermal enforcements.
The expected start of construction is early in 2013. Responsible for the planning are:
x Dipl.-Ing. Franz Götschl, Ingenieurkonsulent für Bauingenieurwesen

x ArchitekturConsultZT GmbH
6. Scientific future work

In a translational research project applied for at the Austrian Science Fund the following
major issues shall be studied:
x Separate analytical models for interlocking and friction
665
Challenging Glass 3
x Clamping force: development with time and influence of the used

interlayer of laminated safety glass
x Connection capacity with respect to permanent loads, cyclic loading
and freeze-thaw cycles
x Redistribution of shear peaks at the end of glass-UHPC connections
Figure 18: Test-setup for sensor calibrating Figure 19: Fundamental test set-ups (schemata)
Figure 20: Test set-ups for further load types (schemata)

Most of the work is based on experimental investigations on small scale specimens.

(Figures 19 and 20). The authors expect that friction is the most relevant mechanism
working against cyclic loads. Therefore it is of particular importance to determine the
real clamping force by means of experiments. It is intended to measure the clamping
behaviour within the specimen due to shrinkage, due to external forces and due to
induced forces. Sensors for direct measuring of load values are not suitable for the
planned experiments. Micro load cells are still too large and piezo force sensors are not
applicable for long time measurements. Thus, strain gauges applied on the front face of
the glass panels combined with Finite Element simulations, is the outlined method for
this investigation. Measuring the deformations will provide reliable data of the
clamping behaviour. In order to get knowledge of the magnitude of the clamping forces,
a special test-setup is planned for calibrating the sensors (Figure 18). The clamping
behaviour will be measured at several different experiments. Thus the separation of the
individual connection mechanisms should work.
666
7. References
[1] Neugebauer, J., Methods to increase the residual bearing capacity of glass, Proceedings of the 5th
International PhD Symposium in Civil Engineering, Delft University of Technology, the Netherlands,
2004
[2] Louter, P.C., High-strength fibre rods as embedded reinforcement in SentryGlas-laminated glass beams.
Glass Processing Days 2009, Tampere, Finland, 2009
[3] Louter, P.C., Adhesively bonded reinforced glass beams. Heron, Vol. 52(1/2), 2007, pp. 31-58
[4] Ølgaard, A. B.; Nielsen, J. H.; Olesen, J. F., Design of Mechanically Reinforced Glass Beams: Modeling
and Experiments, Structural Engineering International, Journal of IABSE, Volume 19, 2009, Number 2,
pp 130-136
[5] Wellershoff F.; Sedlacek G., Structural Use of Glass in Hybrid Elements: Steel-Glass-Beams, Glass-
GFRP-Plates, Glass Processing Days 2003, Tampere, Finland, 2003
[6] Kreher K.; Natterer Jul.; Natterer Joh., Timber-Glass Composite Girders for a Hotel in Switzerland,
Structural Engineering International, Journal of IABSE, Volume 14, 2004, Number 2, pp 159-161
[7] Freytag, B., Die Glas-Beton-Verbundbauweise, PhD Thesis, Graz University of Technology, 2002
[8] Freytag, B., Glass-Concrete Composite Technology, Structural Engineering International, Journal of
IABSE, Volume 14, 2004, Number 2, pp 111-117
[9] Juhart, J.; Freytag, B.; Linder, J.; Sparowitz, L., Preparation and Handling of UHPFRC for the
manufacture of thin-walled building elements, Central European Congress on Concrete Engineering,
Visegrád, Hungary, 2007
[10] Sparowitz, L.; Freytag, B.; Reichel, M.; Zimmermann, W., WILD Bridge - A Sustainable Arch Made of
UHPFRC, Sustainable Arch Bridges - 3rd Chinese-Croatian Joint Colloquium on Long Arch Bridges,
Zagreb, Croatia, 2011
[11] Plóciennik D., Continuous Bond between glass and steel by means of UHPC, ACEE – Journal 4/2008
[12] Plóciennik, D., Die Wirkungsweise von Glas-UHPC-Verbindungen – Theorie, Experiment, Bemessung,
PhD Thesis, Graz University of Technology, in preparation
[13] Freytag, B.: Adhesive Bond between UHPFRC and other Construction Materials, final report FWF
project L 412-N14, Graz, 2011
[14] Santner, G.; Freytag, B.; Juhart, J. et al., Adhesive Power of Ultra High Performance Concrete from a
thermodynamic point of view, ASCE Journal of Materials in Civil Engineering, 2012, in press
[15] Freytag, B.; Juhart, J.; Santner, G., Baumgartner, E., The adhesive Power of Ultra High Performance
Concrete, Bond in Concrete, June 17th – 20th 2012, Brescia, Italy, submitted
[16] Griffith, A. A., The phenomenon of rupture and flow in solids, Phil. Trans. Roy. Soc. London Vol. 7,
1920, pp 221-163
[17] Peintner, F., Detailuntersuchungen zur Umsetzbarkeit der Glas-UHPC-Verbundbauweise in vertikalen
Tragwerkstypen, Master Thesis, Graz University of Technology, 2011
[18] Plóciennik, D., Die Wirkungsweise von Glas-UHPC-Verbindungen – Presentation within the PhD-
presentation series, Institute of Structural Concrete, Graz University of Technology, 01.09.2010
[19] Biribauer, C.; Balak, K.; Peintner, F., Periskop, Highlight 2011 – Concrete Student Trophy 2010, Poster
presentation, Vienna, Austria, 2010
[20] Sobek, W.; Kutterer, M.; Messmer, R., Shear Stiffness of the Interlayer in Laminated Glass, Glass
Processing Days 1999, Tampere, Finland, 1999
667
doi:10.3233/978-1-61499-061-1-669
The Influence of Parameter Setting on the

Mechanical Properties of PVB in
Lamination Process
Maurizio Froli, Leonardo Lani
University of Pisa, Italy, m.froli@ing.unipi.it, l.lani@ing.unipi.it
In Compression Shear Tests (CST) specimens of Laminated Safety Glass (LSG) are
inserted in a special test device with an angle of 45° with respect of the loading
force so that the compression and the shear components acting in the plane of the
plastic interlayer have always the same magnitude. The CST procedure has been
adopted in the present research in order to assess the ultimate shear stresses of
PVB interlayers of different kinds. Critical values of shear and normal stresses have
been taken as measures of the adhesion properties of PVB to be implemented in
calculations to model interfacial adhesion. Different sets of specimens were
prepared under different bond process conditions and tested at the Laboratory for
Testing Materials and Structures of the University of Pisa. The influence of
lamination parameters on creep properties of PVB, have been investigated on large
specimens restrained like a cantilever beam. A numerical model of LSG was
adopted to compare the FEM predictions with experimental results.
Keywords: Laminated Glass, Adhesion, Creep, Relaxation
1. General
It is well known that adhesion between glass panes and plastic interlayers is a requisite
of fundamental importance for the safety and the structural integrity of laminated glass.
The level of adhesion depends on many factors: type of material, autoclave temperature,
pressure and time of bonding process, cleaning process etc.
Surprisingly, any national or international standard requires minimum adhesion

properties in spite of the importance to know and model bonding mechanisms between
glass and polymers in order to get high quality LSG and to avoid delamination
phenomena between glass and PVB.
Furthermore, the high adhesion of PVB to glass ensures, in the post-breakage phase,
that fragments remain attached to the plastic film. On the other hand, a Laminated
Safety Glass with low PVB adhesion guarantees a higher impact resistance, since more
energy is absorbed by elastic deformation of the plastic material. Therefore, the control
of the adhesion properties should be such to satisfy at the same time the capacity to
absorb impacts and the need for a sufficient bond strength [1].
Adhesion proprieties of PVB to glass are usually measured with the Compression Shear
Test (CST) [2] that allows to reach the ultimate shear stress of PVB before glass
collapses as it really often happens in single or double shear lap tests. As known, in a
669
Challenging Glass 3
CST test a small specimen of LSG is inserted in the interface plane of two metallic units
which is inclined of 45° with respect of the compression loading force so that the
compression and the shear components acting in the plane of the plastic interlayer have
at any instant, nominally the same intensity (see Fig. 1). The adhesion strength is given
by the minimum shear force that causes the collapse of PVB before the collapse of glass.
Besides the adhesion properties, the lamination parameters control also the creep
behavior of PVB, and the capacity of interlayer material to transfer the shear force
between glass plies. The same lamination parameters of CST specimens were adopted
to investigate the short and long time behaviour of large specimens restrained like a
cantilever beam subjected to a constant load.
2. Compression Shear Test
2.1. Introduction
Six rectangular LSG main panes each composed by two 500x150x6mm glass sheets
have been prepared under different laminating conditions of autoclave temperature and
pressure, as indicated in Table 1. The 0.76 mm thick PVB foils have been previously
stored under two different humidity conditions.
For each of the six panes, 30x50x50mm specimens have been cut and labeled as
indicated in Figure 1 in order to exactly specify their original position in the pane. The
first letter of the label is referred to humidity condition of PVB, the second is referred to
the conditions of the lamination process of autoclave named “R” or autoclave “S”.
Therefore the generated test population consists of 180 specimens subdivided in six
homogeneous families: N-R, O-R, N-S, O-S, N-S(R), O-S(R). The family N-S(R) and
O-S(R) are laminated with a autoclave “S” setted like autoclave “R”.
With reference to Fig.1, for example, label N-S/5B indicates a specimen cut from the
central part of pane N-S.
CST tests were all performed at room conditions of 18°C temperature and 55% relative
humidity with a test deflection velocity of 5 mm/min.
Table 1: Labeling and lamination process parameters.

Family Number of Thickness Dimensions Pressure Temperature Storage
name samples [mm] [mm] [bar] [°] Humidity
[%]
N-R 30 66.2 50X50 9.4 146 0
N-S 30 66.2 50X50 12 140 0
O-R 30 66.2 50X50 9.4 146 60
O-S 30 66.2 50X50 12 140 60
O-S(R) 30 66.2 50X50 9.4 146 60
N-S(R) 30 66.2 50X50 9.4 146 0
670
The Influence on Parameter Setting on the Mechanical Properties of PVB in Lamination Process
1A 2A 3A 4A 5A 6A 7A 8A 9A 10A
1B 2B 3B 4B 5B 6B 7B 8B 9B 10B
1C 2C 3C 4C 5C 6C 7C 8C 9C 10C
Figure 1: Labeling and location of the specimens over the original LSG pane.
Figure 2:CST test device.
2.2. Test results

For each of the tested specimen the applied load and the relative displacement of the
two steel units was recorded.
Figure 3 shows, for example, the results of test O-R/5A, where each curve is referred to
one of the two inductive transducers. The lack of coincidence between the curves
indicates a imperfect parallelism between the two loaded edges of the specimen. Table 2
collects the obtained test results.
From test results of Table 2 and 3 it can be deduced that the most important parameter
in the lamination process is the humidity storage of the PVB. Beside that these results
show until now that the specimen’s location on the main plate can be neglected.
The graph of Figure 3 indicates on the other hand that the stiffness in the ultimate limit
state is constant with a linear response of PVB and a shear modulus GPVB#10 MPa.
671
Challenging Glass 3
30
25
20
Load [kN]
15
10
0
0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 1,8 2
Displacement [mm]
Figure 3:Load vs Displacement for O-R/5A specimen.
Table 2: Test results - Adhesion value (Wmax).

Sample O-R O-S N-R N-S O-S(R) N-S(R)
[MPa] [MPa] [MPa] [MPa] [MPa] [MPa]
1A 6.88 9.16 12.36 5.53 10.51 10.09
2A 6.93 9.45 11.94 7.14 11.83 9.33
3A 7.13 9.64 14.06 7.01 11.22 9.68
4A 6.66 8.82 12.38 6.87 10.13 9.88
5A 7.60 8.64 14.17 6.62 10.17 10.97
6A 7.29 9.10 11.22 6.28 9.79 9.72
7A 6.92 9.84 12.73 6.12 10.99 9.45

8A 6.95 9.23 12.58 7.18 9.69 9.89
9A 6.43 9.64 11.00 6.02 10.76 9.75
10A 6.98 8.20 12.72 5.95 9.71 8.81
1B 7.36 7.70 11.53 6.03 12.17 9.44
2B 7.16 8.87 11.95 5.48 11.24 8.75
3B 7.00 9.59 13.36 7.14 10.35 8.96
4B 6.58 8.19 11.42 6.71 10.59 9.48
5B 7.06 8.46 14.74 6.84 10.55 8.48
6B 7.23 7.76 11.85 6.96 9.49 8.72
7B 7.14 8.76 11.92 6.82 10.70 8.58
8B 7.66 9.41 12.67 5.72 10.39 8.50
672
9B 7.15 9.56 11.66 5.72 10.30 8.78

10B 5.68 7.21 12.18 6.92 10.73 9.83
1C 7.32 6.82 10.64 4.94 9.85 11.95
2C 7.76 8.11 11.68 5.78 10.49 9.04
3C 6.87 8.69 7.18 6.65 10.74 10.09
4C 6.94 7.24 12.73 5.95 11.31 8.94
5C 7.21 7.64 12.75 5.65 8.62 8.23
6C 6.76 8.39 11.35 6.54 8.43 8.18
7C 7.14 7.78 11.90 6.99 9.92 9.28
8C 6.62 8.48 11.39 7.61 7.30 1080
9C 6.50 8.16 10.77 6.01 11.05 8.15
10C 7.17 7.64 8.52 6.43 9.45 8.43
Table 3 shows the collection of test results with average values of adhesion and
standard deviation. The high dispersion level of family N-R is due to the high number
of glass failures occurred in this case, therefore the real value of adhesion is greater.
Figure 4 shows the adhesion value as function of the normal probability of failure. The
limited number of samples does not allow to define completely each Gauss curve but it
is possible to compare the main values of adhesion and the dispersion of results for
different parameter of lamination process.
Table 3: CST test summery.

Family Wav Standard
name [Mpa] dev.
N-R 11.91 5.22

N-S 6.40 2.27

O-R 7.01 1.45
O-S 8.54 2.92
O-S(R) 9.31 3.03
N-S(R) 10.27 3.7
Results of table 3 show the importance on lamination process on adhesion properties

with the correlation between the humidity storage of PVB (samples N and O) and the
pressure and temperature of the autoclave (samples R and S). Humidity of PVB reveals
to be more influent in autoclave R with a temperature of 146°C and 9.4 bar of pressure.
The parameters of lower temperature and higher pressure of autoclave S (140°C and 12
bar) reduce the influence of humidity storage, and under these parameter setting
“humid” PVB shows a better adhesion than “dry” PVB.
673
Challenging Glass 3
30,00%
OR
OS
OR(9,4/146)
NR
25,00%
NS
NS(R)
OS(R)
20,00%
NS(12/140)
P(x) [%]
15,00%
NS(R)
10,00% OS(12/140)
NS(R)
5,00% NR(9,4/146)
0,00%
0,00 2,00 4,00 6,00 8,00 10,00 12,00 14,00 16,00
Adhesion stress at failure
Figure 4:Adhesion stress vs probability of failure.
3. Creep test on large plate
3.1. Introduction
The creep test have been conducted on four samples composed by two 1000x200x8mm
tempered glass sheets prepared under the same laminating conditions of autoclave “R”
or “S” of small samples adopted for Compression Shear Test, the parameters are shown
in table 4. Even in these tests the 0.76 mm thick PVB foils have been previously stored
under two different humidity conditions.
Creep tests were performed at room conditions of 19°C temperature and 60% relative
humidity with a constant load of 17 daN hanging with a metallic rope as shown in
figure 5. The displacement at the free edge of the plate was recorded during three days
with a inductive displacement transducer.
Table 4: Labeling and lamination process parameters.

Family Number of Tickness Dimensions Pressure Temperature Storage
name samples [mm] [mm] [bar] [°] Humidity
[%]
N-R 1 88.2 1000x200 9.4 146 0
N-S 1 88.2 1000x200 12 140 0
O-R 1 88.2 1000x200 9.4 146 60
O-S 1 88.2 1000x200 12 140 60
674
Figure 5:Creep test.
3.2. Test results

Figure 6 and 7 shows the short and long term behavior with displacement vs. time for
each sample laminated according to table 4. In the same graph it is possible to see the
behavior of monolithic and layered glass plates obtained from numerical analysis.
45
40
35
30
Displacement [mm]
25
NR*
20
OS*
15 NS*
OR*
10
Layered
5
Monolitic
0
0,00 1,00 2,00 3,00 4,00 5,00 6,00 7,00 8,00 9,00 10,00 11,00 12,00 13,00 14,00 15,00
Time [min.]
Figure 6: Displacement vs. time - Short time behavior.
675
Challenging Glass 3
45
40
35
NR*
30 OS*
NS*
Displacement [mm]
OR*
25
Layered
Monolitic
20
15
10
5
0
0 20 40 60 80 100 120 140
Time [h.]
Figure 7: Displacement vs. time - Long time behavior.
Table 5 shows the displacements measured at some representatives times and the factor
' between the minimum and maximum value at each time.
Table 5: Experimental creep results summary.

Displacements [mm]
Family T=1 min. T=2 min. T=5 min. T=10 min. T=120 h.
name
N-R 18.1 20.7 23.8 25.6 33.3
N-S 20.0 22.6 26.2 28.5 36.7
O-R 22.0 24.5 27.7 29.8 37.3

O-S 20.0 22.0 25.0 26.8 34.3
' 0.82 0.84 0.85 0.86 0.89
3.3. Numerical analysis

A numerical analysis was performed and compared with the results of the available
creep tests. The glass layer and PVB material were modeled with 8-node isoparametric
elements to evaluate the behavior of PVB and the capacity to transfer shear action
between glass layer. Soda–lime–silica float glass is modeled as a linear-elastic material
with a Young’s modulus of 70.000 MPa and Poisson ratio of 0.22.
676
Figure 8: Finite element model of glass plate.
The estimation of the shear modulus of the interlayer material can be performed by
comparing the numerical results of the modeled structure with the experimental data
obtained from laboratory experiences [4]. This kind of approach does not represent a
direct method to measure the shear modulus of PVB, but it allows a estimation of it
sufficiently accurate for use in current structural analysis and design.
By means of the numerical model we can calculate displacements that depends on the
characteristics assigned to the shear modulus G of the PVB; on the other hand, from
experiments we obtain corresponding quantities that depend on time and room
temperature. A process of structural identification can be performed by calibrating
conveniently the numerical model until the output data are optimally close to the
experimental results. When the model is sufficiently sound, it permits good estimations
of in situ values of G [5].
Table 6 collects, at each time of table 5, estimated values of G and the factor ' between
the minimum and maximum value.
Table 6: FEM creep results summary

G [Mpa]
Family T=1 min. T=2 min. T=5 min. T=10 min. T=120 h.
name
N-R 0.34 0.26 0.16 0.14 #0
N-S 0.28 0.20 0.10 0.06 #0
O-R 0.21 0.15 0.08 0.05 #0
O-S 0.28 0.21 0.14 0.10 #0
' 0.61 0.57 0.50 0.35 -
The sample N-R with PVB stored in absence of humidity and laminated under a
pressure of 9.4 bar and a temperature of 146°C is stiffer than all the others. At any
instant the ratio ' between minimum and maximum value of G appears not to be
negligible since it ranges within the interval 0.35y0.61.
677
Challenging Glass 3
4. Conclusions
A Compression Shear Test (CST) programme is presently running at the University of
Pisa over a population of 180 LSG specimens divided in 6 groups characterized by
different lamination conditions.
The variable parameters are: autoclave temperature and pressure, time of process,
storage humidity of the PVB.
The first tests confirmed the importance of the influence of storage humidity on the
adhesion property of PVB.
The viscoelastic properties of PVB of LSG were investigated with tests on large
specimen, the test confirm that the stiffness of PVB is sensible from lamination
parameters.
Numerical analyses were also performed to evaluate how sensitive numerical

simulations are with respect to the implemented shear modulus G(T,t) law of PVB.
5. Acknowledgements
The research was developed thanks the financial support of the Regione Toscana,
Programma Operativo Regionale FSE “Competitività Regionale e Occupazione 2007-
2013”.
6. References
[1] Keller U., Mortelmans, H., Adhesion in Laminated Safety Glass – What makes it work?, Glass
Processing Days , Tampere, Finland, 1999.
[2] Schneider K., Lauke B., Beckert W., Compression Shear Test (CST) – A Convenient Apparatus for the
Estimation of Apparent Shear Strength of Composite Materials, Applied Composite Materials, 2001,
pp.43-46.
[3] Ferry, J. D., Viscoelastic proprieties of polymers, 3rd Ed., Willey, New York.
[4] Schuler, C., Sackmann,V., Gräf, H., Bucak,Ö., Albrecht, G., (2004), Time and Temperature Dependent
Mechanical and Durability of Laminated Safety Glass, Structural Engineering International,Vol.14, n°2,
IABSE.
[5] Vallabhan, C.V.G., Das, Y.C., Magdi, M., Asik, M., Bailey, J.R., (1993), Analysis of Laminated Glass
Units, Journal of Structural Engineering, ASCE, Vol. 199, No. 5.
678
doi:10.3233/978-1-61499-061-1-679
Rectangular Plate Finite Element for

Triplex Laminated Glass
Ivelin Ivanov, Dimitar Velchev, Nikolay Georgiev, Ivo Ivanov
University of Ruse, Bulgaria, ivivanov@uni-ruse.bg
Laminated glass became a popular safety glass. The very soft PVB-interlayer
impedes the slippage between glass layers, bending in parallel under transverse
loading, by shear stress. Conventional Finite Elements (FE) are not capable to solve
the problem efficiently. Based on the assumption that the glass layers of a
laminated glass obey Kirchoff's plate theory and the PVB-interlayer transfer shear
stress only, a special rectangular triplex laminated glass plate FE is elaborated. The
element is validated by comparison with other computational FE models of
commercial software and experimental tests of a laminated glass strip in cylindrical
bending.
Keywords: Laminated glass, Plate finite element, Plate theory
1. Introduction
Laminated glass is widely spread as material for architectural glazing of contemporary
buildings. It is also used for windshields in automotive industry. Laminated glass, in
most of the cases, consists of only two glass plates bonded together by a polyvinyl
butyral (PVB) interlayer. The efficient application of laminated glass in the complex
structures is limited because of the difficulties in their strength calculations at the stage
of their design.
The PVB-material has elastic modulus which is thousands of times less than the elastic
modulus of glass. The great mismatch of the elastic moduli leads to layer interaction in
the bending of laminated glasses which does not obey the principle of straight normals
or pseudo normals in the theory of plates. Since the behaviour of laminated glass is not
very well modelled, its strength is considered as intermediate of two limiting cases: the
case of layered glass and the case of monolithic glass [1].
Vallabhan et al. [2] developed a mathematical model of laminated glass based on the
minimization of the total potential energy in which the bending and membrane strain
energy of the glass plates as well as the shear strain energy of the PVB-interlayer are
included. The assumptions for the glass plates correspond to von Karman's nonlinear
theory of plates. Sophisticated differential equations should be integrated, but it is not
easy to obtain a stable solution. A¬ik [3] developed an algorithm for implicit integration
of the equations and their unconditionally stable solution.
Although in [2], the theory of laminated glass was developed, it is not applied in finite
element formulation. Instead of numerical solutions simple analytical multilayer beam
models of laminated glass are developed [4]. An effective section modulus of the beam
679
Challenging Glass 3
could be calculated, which allows the maximum bending stress to be obtained. The
other opportunity is to utilise the Finite Element (FE) method to model laminated
glasses using 3-D solid elements [5]. This approach requires a lot of finite elements and
therefore expensive computations.
The aim of this work is to develop an efficient simple finite element, which incorporates
the basic theory of laminated glass. The glass layers are in bending while the PVB-
interlayer could transfer shear stress only. The element is proposed for static or quasi-
static transverse loading in a linear solution of the problem.
2. Stress and strain assumptions

The laminated glass considered here consists of two layers of glass and one interlayer of
PVB-material, which is popular under the name Triplex Laminated Glass (TLG). Only
the bending of TLG under transverse loading is considered. It is based on the
assumption that the glass layers of laminated glass obey the Kirchoff's plate theory. The
glass layers bend in parallel with the same curvature. The PVB-interlayer is in pure
shear and it resists with shear stress to the slippage between the glass layers.
Figure 1: Infinitesimal element of laminated glass.
An infinitesimal piece of laminated glass is considered here in order the main

relationships concerning the displacements, strain, and stress to be derived. The

infinitesimal element is shown in Figure 1. The displacements of point P from glass
layer number i, i = 1, 2, which is on distance i from the middle plane of the layer are
considered by using the projections of the infinitesimal glass layer of the element onto
xz- and yz-planes. The point moves to position P' and its displacements are denoted u, v,
and w as it is shown in Figure 2. They can be determined as:
ui u oi [ i w, x , vi voi [ i w, y , wi w( x , y ) (1)
where
ww ww
w, y T x , w, x Ty (2)
wy wx
are the rotations of the plate normal.
680
Rectangular Plate Finite Element for Triplex Laminated Glass
Figure 2: Displacements in glass layer i, i=1, 2.
The first derivatives of the displacements are the strains in the glass layers:
H xi H xoi [ i w, xx , H yi H yoi [ i w, yy , J xyi 2[ i w, xy (3)
Applying Hooke's law, the plane stress components are derived:
E E
V xi (H xi QH yi ) , V yi (QH xi H yi )
1 Q 2
1 Q 2 (4)
E
W xyi J xyi
2(1 Q )
The normal resultant forces for each glass layer can be found by integration of stresses
through the thickness of the layer:
hi / 2
E hi
n xi ³ V xi d [ i 1 Q 2
(H xoi QH yoi ) (5)
hi / 2
hi / 2
E hi
n yi ³ V yi d [ i 1 Q 2
(QH xoi H yoi ) (6)
hi / 2
The normal resultant forces of both glass layers should be equal and opposite because of
the equilibrium of the forces and the lack of membrane loading. They should create a
couple of forces which gives additional bending moment to the resistance in bending as
it is shown in Figure 3. Then the normal resultants can be denoted as:
681
Challenging Glass 3
nx { nx2 n x1 , n y { n y 2 n y1 (7)
Figure 3: Bending moments and layer interaction.
Applying Eqs. (5) and (6) for both glass layers, i = 1, 2, four equations are obtained,
which are solved for the membrane strains of the glass layers and by using the notation
(7), the result is written as:
n x Q n y n y Q n x
H xo1 , H yo1
E h1 E h1
(8)
n x Q n y n y Q n x
H xo 2 , H yo 2
E h2 E h2
The equations above give also the relationships
h1H xo1 h2H xo 2 , h1H yo1 h2H yo 2 (9)
The PVB-interlayer is assumed to have only pure shear strain and stress as it is shown
in Figure 4. The shear strain in both planes of projection can be derived from the
membrane strains, the rotations of both glass layers, and their relative displacements or
slippages:
h2 h
uo2 w, x u o1 1 w, x
J zx 0 M y T y 2 2 w, x
h0
h1 2h0 h1 h2
J x { J zx 0 (G u hC w, x ) , G u u o 2 u o1 , hC (10)
h0 2
where u is the slippage between the glass layers in x-direction and hC is the distance
between the glass layers midsurfaces as it is shown in Figure.3.
682
Figure 4: Strain and stress in PVB-interlayer.
In a similar way, the shear strain in the other plane is determined as
h1
J y { J yz 0 (G v hC w, y ) , G v vo 2 vo1 (11)
h0
by the slippage between the glass layers in y-direction. The slippage between the glass
layers is very important to explain the mechanism of glass layer interaction through the
PVB-interlayer. Also, the derivatives of the slippage can represent very important
relationship between the membrane straining of the glass layers:
h1 h2
G u,x H xo 2 H xo1 ( n x Q n y ) (12)
E h1 h2
h1 h2
G v, y H yo 2 H yo1 ( n y Q n x ) (13)
E h1 h2
The shear stress in the PVB-interlayer can be found easily applying Hooke's law for
shear:
W x { W zx 0 G0 J x , W y { W yz 0 G0 J y (14)
3. Strain energy
The derivation of the stiffness matrix is based on the stationary principle of the total
potential energy in equilibrium. The potential strain energy is necessary to be derived in
the volume of a plate finite element.
683
Challenging Glass 3
The strain energy of laminated glass element will be presented in vector and matrix
form in order to obtain then the stiffness matrix of the developed finite element. The
potential strain energy of the PVB-interlayer depends only on the pure shear stress and
strain.
1 T 1 T
2 V³ 2 V³
3J g t dV g CJ g d V ` (15)
0 0
where
ªG0 0º
g [J x J y ]T , t [W x W y ]T , CJ «0 G0 »¼
(16)
¬
The volume of integration is V0, V0 = h0 , where is the area of the plate.
The strain vector g can be expressed as a sum of two other vectors
ªJ x º 1 ªG u hC w, x º 1 h
g «J » «G h w » uG C uT (17)
¬ y¼ h0 ¬ v C ,y ¼ h0 h0
Then the potential strain energy can be constructed in another way:
1 T
2 :³
3J e J DJ e J d : (18)
where
ª Gu º ª1 0 hC 0º
«G » «0
« v », D « 1 0 hC »»
eJ J (19)
« w, x » «hC 0 hC2 0»
« » « »
¬« w, y ¼» ¬0 hC 0 hC2 ¼
The derivatives of displacement functions can be expressed by a differential operator,

L:
ª1 0 0º
«0 ªG º
« 1 0 »» « u »
eJ LJ u G (20)
«0 0 w x »« v »
« »« w »
¬«0 0 w y ¼» ¬ ¼
684
The potential strain energy of the PVB-interlayer is determined by three displacement

functions of two variables — the coordinates x and y, which are the slippage of the glass
layers u and v, as well as the deflection w.
The potential strain energy of each glass layer i, i = 1, 2, can be derived from the plane
stress field in bending as follows:
1 T 1 T
2 V³ 2 V³
3 gi ei s i d V ei Cei d V (21)
i i
where
ª H xi º ªV xi º ª º
« » « » E «1 Q 0 »
ei « H yi » , s i «V yi » , C «Q 1 0 » (22)
«J xyi » «W xyi » 1 Q 2 « 1 Q »
¬ ¼ ¬ ¼ «0 0 »
¬ 2 ¼
and the volume of the glass layer, Vi, is Vi = hi .
The strain in glass layers can be represented in vector form
ªH xoi [ i w, xx º ªH xoi º ª w, xx º
« » «H » [ « w »
ei «H yoi [ i w, yy » « yoi » i « , yy » e oi [ i eN (23)
« 0 2[ i w, xy » «¬ 0 »¼ « 2 w, xy »
¬ ¼ ¬ ¼
Substituting (26) into (24), one obtains

1
3 gi
2V³ (e oi [ i eN ) T C(e oi [ i eN ) dV 3 oi 3 ki (24)
i
where
1 1 hi3 T
hi ³ e Toi Ce oi d : , 3 Ni
2 12 :³
3 oi eN CeN d : (25)
2 :
Using the Eqs. (8), (12), and (13), the potential energy can be obtained as:
1 1 T
2 hi :³
3 oi eG CG eG d : (26)
685
Challenging Glass 3
where
ªG u , x º E ª1 Q º
h12 h22
eG «G » , CG « » (27)
¬ v, y ¼ 1 Q (h1 h2 ) ¬Q 1 ¼
2 2
Now, the potential strain energy of both glass layers can be summed:
1 T
2 :³
3g 3 o1 3 o 2 3 N 1 3 N 2 eg Dg eg d : (28)
where
ªCG 0 º
eg >G u,x G v , y w, xx w, yy 2 w, xy @T , D «
« 0
h13 3 »
h2
C»
(29)
¬ 12 ¼
The vector of pseudo strain eg can be represented as a differential operator, Lg applied

on the displacement functions u:
ªw x 0 0 º
«0 wy 0 »» ªG u º
«
eg Lgu «0 0 w x 2 » «G v » (30)
« »« »
«0 0 w x 2 » «¬ w »¼
«0 0 2w xy »¼
¬
The potential strain energy of the laminated glass plate is the sum of the potential shear
strain energy of the PVB-interlayer and the potential strain energy of the glass layers:
3 3J 3 g (31)
4. Stiffness matrix derivation

The derived strain energy expressions show that they depend on three functions of the
displacements forming the vector u, which totally describe the equilibrium state in the
triplex laminated glass. A rectangular four-node plate finite element with six degrees of
freedom per node as it is shown in Figure 5 is constructed in the global coordinate
system xyz.
686
Figure 5: TLG rectangular plate finite element.
The vector of the element nodal displacements consists of four nodal vectors with the
displacements of each node I, I = 1, 2, 3, 4. The nodal displacement vectors have six
displacements each: two relative displacements defining the slippage between the glass
layers, u, v, the deflection, w, and three rotations, !x , !y , ":
d [d1T d T2 d 3T d T4 ]T , d I [G uI G vI wI T xI T yI \ I ]T (32)
The displacement functions can be approximated by shape functions multiplied by the

nodal displacements, which are the unknown parameters:
u N d , N [ N1 N 2 N 3 N 4 ] (33)
The matrix of shape functions consists of four sub-matrices — one for each node I. The
nodal sub-matrices of shape functions have six shape functions depending on the nodal
coordinates — one shape function for each degree of freedom:
ª N 1I 0 0 0 0 0 º
NI « 0 N 2I 0 0 0 0 »» (34)
«
«¬ 0 0 N 3I N 4I N 5I N 6 I »¼
All shape functions are functions of two variables — the coordinates x and y. The first
two functions are equivalent, N2I { N1I. They describe the slippage between the glass
layers and they can be constructed as bi-linear shape functions from Lagrangian
polynomials [6]. The remaining shape functions describe the displacements in
Kirchoff's plate. They are C1 continuity shape function constructed by Hermitian
polynomials [6]. The basic plate element used here is Bogner-Fox-Schmit (BFS)
rectangular plate element which is well described in many sources like [6]. The basic
plate element could be also any plate element. Here, we added two degrees of freedom
and shape functions describing them.
687
Challenging Glass 3
Once the shape functions are chosen, the stiffness matrix, K, can be defined from the
potential strain energy, 3, by its partial derivatives with respect to the nodal
displacements:
w3 w (3 J 3 g )
Kd (K J K g )d , K KJ K g (35)
wd wd
First of all, the strain-displacement matrices should be defined. Starting with the shear
strain in the PVB-interlayer, the strain vector can be expressed by the shape function
matrix and defining the strain-displacement matrix the shear stiffness matrix is readily
determined:
³ B J DJ B J
T
eJ LJ u L J Nd BJ d , BJ LJ N , K J d: (36)
Z
Here Z is the area of the finite plate element. The other part of the stiffness matrix is
defined through the glass-layer strain vector:
³ Bg Dg Bg d :
T
eg L gu L g Nd Bg d , Bg Lg N , K g (37)
Z
5. Validation
Laminated glass is provided for experimental investigation. The laminate has total
thickness of 6.13 mm or the layers of glass/PVB/glass are 2.875/0.38/2.875 mm thick.
The elastic properties of glass are assumed to be as follows: the Young's modulus is
E = 72 GPa and the Poisson's ratio is Q = 0.23. The total mass density of the laminate is
measured as = 2.418 g/cm3. In order to characterise the PVB-material, compressive
shear tests are carried out. The approach used here to find the initial shear modulus is
the same as it is described in [7]. The results show that the shear modulus of PVB-
material is G0 = 0.5173 MPa.
Figure 6: Experimental setup for cylindrical bending of laminated glass strip.
688
An experimental setup with laminated glass strip has been prepared. The setup is given
in Figure 6. The simply supported strip is loaded by the weight of mass m and six strain
gauges measure the longitudinal strain on upper and lower surfaces.
a) b)
X Step: Step2
Y
c)
Increment 3: Step Time = 3.000
Deformed Var: U Deformation Scale Factor: +1.000e+01

Z
Figure 7: Computational models: a) TLG plate, b) continuum shell, c) 3-D brick element model.
Load m = 8 kg
250
3D brick
shell linear
shell nonlinear upper
200
shell nonlinear lower
TLG plate
experimental upper
150 experimental lower
strain (x106)
100
50
50
0 50 100 150 200 250 300 350 400
distance (mm)
Figure 8: Longitudinal strain magnitude along the strip.
The experimental results are compared with the results of three computational models of
the half of the laminated glass strip. The models are shown in Figure 7. The first model
is a FE model built by the TLG plate element developed here. The continuum shell
element model is very efficient model in the commercial software ABAQUS™ [8]. The
last model is using 3-D brick elements, again in ABAQUS™ software, which is
classical reference model, but computationally very inefficient.
689
Challenging Glass 3
Comparison of the results is given in Figure 8. The continuum shell element model has
the capability to solve the problem nonlinearly and the nonlinear solutions are given
together with the linear one. The nonlinearity is negligible here. The continuum shell
element model is softer than the others. The TLG plate model is close to the
experimental and to the reference 3-D brick element model.
6. Conclusions
The laminated glass is very difficult to be modelled by conventional finite elements,
since it does not obey the principle of straight normals remaining straight after the
deformation. The developed triplex laminated glass rectangular plate element is based
on Kirchoff's plate theory for the glass layers and on the assumption that the PVB-
interlayer undergoes pure shear. It is very efficient compared to the other computational
models based on the conventional finite elements and it gives very similar results for
linear solutions of the problem of laminated glass under transverse loading.
7. Acknowledgement
The support by National Science Fund of Bulgarian Ministry of Education, Youth, and
Science, through the grant agreement No DDVU 02/052- 20.12.2010, is greatly
acknowledged.
8. References
[1] Vallabhan, C.V.G.; Minor, J.E.; Nagalla, S.R., Stress in layered glass units and monolithic glass plates,
Journal of Structural Engineering, 113/1987, pp. 36-43.
[2] Vallabhan,C.V.G.; Das, Y.C.; Magdi, M.; A¬ik, M.; Bailey, J.R., Analysis of laminated glass units,
Journal of Structural Engineering, 119/1993, pp. 1572-1585.
[3] A¬ik, M.Z., Laminated glass plates: revealing of nonlinear behavior, Computers and Structures,
81/2003, pp. 2659-2671.
[4] Norville, H.S.; King, K.W.; Swofford, J.L., Behavior and strength of laminated glass, Journal of
[5] Duser, A.; Jagota, A.V.; Bennison, S.J., Analysis of glass/polyvinyl butyral laminates subjected to
uniform pressure, Journal of Engineering Mechanics, 125/1999, 435-442.
[6] Zienkiewicz, O.C.; Taylor, R.L., The finite element method, Vol. 2: Solid mechanics, fifth ed.,
Butterworth-Heinemann, Oxford, 2000.
[7] Jagota, A.; Bennison, S.J.; Smith, C.A., Analysis of a compressive shear test for adhesion between
elastomeric polymers and rigid substrates, International Journal of Fracture, 104/2000, pp. 105-130.
[8] ABAQUS Analysis User's Manual, ver. 6.11, Dassault Systèmes Simulia Corp., Providence, RI, USA.
690
doi:10.3233/978-1-61499-061-1-691
Reinforced Glass Beams Composed of

Annealed, Heat-Strengthened
and Fully Tempered Glass
Christian Louter
EPFL, Lausanne, Switzerland, christian.louter@epfl.ch
Jan Belis
Ghent University, Ghent, Belgium, jan.belis@UGent.be
Freek Bos
Witteveen+Bos, Deventer, The Netherlands, f.bos@witteveenbos.nl
Fred Veer
Delft University of Technology, The Netherlands, f.a.veer@tudelft.nl
Annealed, heat-strengthened and fully tempered SG-laminated reinforced glass

beam specimens were subjected to four-point bending tests to investigate the effects
of glass type on their structural response. During the test the beams showed linear
elastic response until initial glass failure, followed by a significant post-breakage
strength and ductility. From the results it was concluded that the application of
stronger glass types in reinforced glass beams significantly enhances their initial
glass breakage strength, but affects their post-breakage strength and ductility in a
negative way. The annealed and heat-strengthened beams reached similar post-
breakage strength levels, whereas this was reduced for the fully tempered beams.
Furthermore, compared to the annealed beams the post-breakage ductility was
reduced for the heat-strengthened and even further reduced for the fully tempered
beams. These differences originated from differences in fragmentation of the glass.
Keywords: Reinforced glass, Safety, Redundancy, Glass type, Hybrid, Composite

1. Introduction
This paper focuses on the effects of glass type on the structural response of SG-
laminated reinforced glass beams. In this beam concept, which has been progressively
developed and tested in preceding research [1, 2, 3], a stainless steel reinforcement is
bonded to the glass by means of SentryGlas® (SG) interlayer sheets, see Figure 1. In the
event of glass failure, the stainless steel reinforcement is intended to bridge the crack(s)
and to carry the tensile force over the crack(s). This way a post-breakage load-carrying
mechanism is created, which allows the beam to still carry load despite (extensive) glass
breakage.
Preceding research has demonstrated that this concept of laminating a stainless steel
reinforcement to a glass beam using SG interlayer sheets is a feasible concept [2]. The
shear strength and stiffness of the SG interlayer is sufficient to transfer the forces
between glass and reinforcement in order to activate the reinforcement action when the
glass is broken. Furthermore, preceding research has demonstrated that the SG-
691
Challenging Glass 3
laminated reinforced glass beams perform well at temperature levels ranging from -20
to +60°C, after thermal cycling and for long duration post-breakage loading [4].
In addition to the preceding studies, which have focused on SG-laminated reinforced
glass beams composed of annealed float glass, the current paper investigates the effects
of glass type on the structural response of the beams. This is done through three series
of four-point bending tests on SG-laminated glass beam specimens composed of either
annealed, heat-strengthened or fully tempered glass beams.
The following sections present the specimens, the test setup and the test results.
Subsequently, the results are discussed and conclusions from the research are provided.
2. Specimens
The specimens for this research consisted of three layers of glass, a hollow section
stainless steel reinforcement and SentryGlas® (SG) interlayer sheets, see Figure 1. The
typical material characteristics of the constituent materials are listed in Table 1.
Either annealed, heat-strengthened or fully tempered glass was applied. The cutting of
the glass, the edge-polishing and the heat-treatments (if applicable) were done by a
commercial glass processor. The stainless steel reinforcement was applied as a 10x10x1
mm hollow section. Apart from cleaning, no special surface pre-treatment was applied
to the stainless steel. The SG interlayer sheets were applied in a thickness of t = 1.52
mm.
The specimens were laminated in a standard lamination cycle using a vacuum bag
lamination technique. Prior to the lamination process the glass and stainless steel were
degreased. The lamination was performed by a commercial laminator.
125 mm
Figure 1: Schematic representation of the cross-section of the investigated SG-laminated reinforced glass
beams; (a) exploded view; (b) assembled view after the lamination process.
692
Reinforced Glass Beams Composed of Annealed, Heat-Strengthened and Fully Tempered Glass
Table 1: Typical material properties of glass, stainless steel and interlayer material, according to [5-13].
Property Units Glass (a, b, c) Stainless steel (d, e) Interlayer (g, h, i)
Heat- Fully
Specification - Annealed AISI 304 (f) SG
strengthened tempered
Tensile strength [N/mm2] 45(a) 70(b) 120(c) 520-750 34.5 (j)
Elastic modulus [N/mm2] 70 x 103 70 x 103 70 x 103 200 x 103 300 (j)
Transition
[°C] 530 (k) 530 (k) 530 (k) n/a ~55-60 (l)
temperature
Elongation at tear [%] - - - 45 400
3
Density [kg/m ] 2500 2500 2500 7900 950
Coefficient of 9 9 9 16.0 10-15
[K-1]
thermal expansion x 10-6 x 10-6 x 10-6 x 10-6 x 10-3
(a)
according to EN 572-1:2004 [5]; (b) according to EN 1863-1:2000 [6]; (c) according to EN 12150:2000 [7];
(d, e)
according to EN 10088-1:2005 [8] and EN 1008-2:2005 [9]; (f) EN name designation = X5CrNi18-10;
EN number designation = 1.4301; (g, h, i) according to Stelzer [10], Dupont [11] and Bucak & Meissner [12];
(j)
Strongly dependent on temperature and load duration [10].
(k)
as given for soda lime silica glass in [13], (l) as reported by Bucak & Meissner [12]
3. Test setup
The four-point bending tests were performed on a universal electromechanical test
machine, which was provided with a customized support frame to support the beam
specimens. The load span, support span and lateral support span corresponded to the
measures provided in Figure 2. The load was applied at a displacement rate of 2
mm/minute. During the test the applied load and the displacement at mid-span were
recorded.
mechanical load introduction

at preset displacement rate
cross head
lateral support lateral support
(a) test setup specimen
support frame
lat. sup. = 550 mm
load span
= 400 mm
(b) scheme
support span = 1400 mm
Figure 2: Schematic representation of the applied four-point bending test setup.
693
Challenging Glass 3
4. Results
The results of the four-point bending tests are presented in Table 2 and Figures 3, 4 and
5. The beams showed linear elastic response until initial glass failure. Upon initial glass
failure a significant drop in load was observed. For the annealed beams initial failure
typically occurred in one glass layer only. For the heat-strengthened and fully tempered
beams initial failure typically occurred instantaneously in all glass layers. As loading of
the beams was continued, all beams developed a ductile post-breakage response due to
plastic deformation of the stainless steel reinforcement. For the annealed and heat-
strengthened glass beams this was preceded and accompanied by additional glass
fracture, whereas for the fully tempered glass beams some crumbling of the glass
occurred. All beams finally collapsed due to local compressive failure of the glass.
Table 2: Results of the bending tests on the reinforced glass beams

composed of either annealed, heat-strengthened or fully tempered glass.
Heat- Fully
Annealed
strengthened tempered
Number of specimens [-] 5 5 5
Finitial = initial glass breakage load Mean [kN] 10.9 33.1 42.1
st.dev. [kN] 0.6 1.6 0.8
rel. st. dev. [%] 5.4 4.9 2.0
Fresidual = max. post-breakage load Mean [kN] 16.6 16.5 14.5
st.dev. [kN] 0.1 0.5 0.5
rel. st. dev. [%] 0.8 2.8 3.4
Fresidual / Finitial Mean [-] 1.5 0.5 0.4
st.dev. [%] 9.9 2.9 1.6
rel. st. dev. [%] 6.5 5.9 4.5
694
Figure 3: Results of the bending tests on the reinforced glass beams composed of annealed glass;
(a) load-displacement diagram; (b) cracking sequence; (c) photographs of a tested specimen.
695
Challenging Glass 3
Figure 4: Results of the bending tests on the reinforced glass beams composed of heat-strengthened glass;
696
Figure 5: Results of the bending tests on the reinforced glass beams composed of fully tempered glass;
697
Challenging Glass 3
5. Discussion
From the results of the four-point bending tests on the annealed, heat-strengthened and
fully tempered SG-laminated reinforced glass beams the following aspects are observed:
Firstly, the annealed, heat-strengthened and fully tempered glass beams showed
significant differences in initial glass breakage strength and fracture pattern. Obviously
this is directly related to the presence of compressive surface pre-stress in the heat-
strengthened and fully tempered glass. This pre-stress significantly enhances the initial
failure strength of these glass types. Consequently, the heat-strengthened and fully
tempered beams reached an average initial glass breakage load of 33.1 and 42.1 kN,
respectively, whereas this amounted to 10.9 kN for the annealed beams, see Table 2.
Furthermore, due to the presence of the compressive surface pre-stress the heat-
strengthened and fully tempered glass beams demonstrated more extensive
fragmentation than the annealed glass beams, see Figures 3, 4 and 5, as is commonly
known for these glass types.
Secondly, the annealed, heat-strengthened and fully tempered glass beams demonstrated
differences in residual load-carrying capacity. Whereas the annealed and heat-
strengthened glass beams reached an average post-breakage strength of 16.6 and 16.5
kN respectively, this was reduced for the fully tempered glass beams and amounted to
14.5 kN. Theoretically the maximum post-breakage strength is determined by
multiplying the maximum tensile capacity of the reinforcement by the maximum
internal leverarm between the compression force in the glass and the tensile force in the
reinforcement, see Figure 6. The maximum tensile capacity of the reinforcement
amounts to 30.7 kN (as determined by an uni-axial tensile test [2]) and the maximum
theoretical internal leverarm amounts to 120 mm, see d in Figure 6. When multiplied,
this results in a maximum moment capacity of 0.12 x 30.7 = 3.7 kNm, which
corresponds to a maximum load in the four-point bending test of 14.7 kN. Surprisingly,
the annealed and heat-strengthened glass beams exceed this theoretical limit.
Assumedly, an additional load-carrying capacity is generated by overlapping glass
fragments in the fractured annealed and heat-strengthened glass beams, which transfer
forces over the cracks in the glass by shear in the SG interlayer [2, 3]. For the fully
tempered beams, which demonstrated full fragmentation upon initial glass failure, this
shear transferring mechanism is absent. Contrary to the larger glass fragments in the
annealed and heat-strengthened glass beams, the very small glass fragments in the fully
tempered glass beams are not able to transfer significant forces through shear in the SG
interlayer. The absence of this mechanism reduces the post-breakage load-carrying
capacity of the fully tempered beams compared to the annealed and heat-strengthened
beams. However, it should be noted that even without this additional load-carrying
mechanism, the fully tempered beams still reached significant post-breakage loads. The
fully tempered beams could sufficiently exploit the tensile capacity of the reinforcement,
thereby reaching post-breakage strength levels close to the theoretical limit.
698
Figure 6: Schematic representation of the strain distribution in a fractured annealed reinforced glass beam.
Thirdly, the post-breakage ductility between the annealed, heat-strengthened and fully
tempered glass beams differed. The annealed glass beams demonstrated an extensive
ductile post-breakage trajectory, whereas this was reduced for the heat-strengthened
beams and even further reduced for the fully tempered beams, see Figures 3, 4 and 5.
This difference in post-breakage ductility originates from the difference in fracture
pattern between the annealed, heat-strengthened and fully tempered beams. Due to the
more extensive fragmentation upon glass failure – especially for the fully tempered
beams – the heat-strengthened and fully tempered beams finally collapsed much earlier
than the annealed beams, due to instability and failure of the glass compression zone.
The effects of glass type on the structural performance of hybrid/reinforced glass beams
has also been investigated by Kreher [14, 15]. Figure 7(a) shows an overview of the
results of four-point bending tests performed on hybrid timber-glass beams in which
timber flanges are adhesively bonded to either an annealed, heat-strengthened or fully
tempered glass web. From Figure 7(a) it can be seen that the post-breakage reserve
(Fresidual/Finitial, as plotted along the y-axis) reduces with increasing levels of pre-stress in
the glass web (as plotted along the x-axis). Figure 7(b) demonstrates a similar plot for
the current research results. When compared to the results of Kreher, the current
research results show a similar tendency of decreasing post-breakage reserve with an
increasing level of pre-stress in the glass. However, it should be noted that the decrease
in post-breakage reserve (Fresidual / Finitial) with increasing levels of pre-stress, mainly
originates from an increase in initial failure strength (Finitial) rather than from a
significant decrease in post-breakage residual strength (Fresidual), see Table 2.
699
Remaining load-carrying capacity Challenging Glass 3
after first crack [Fresidual / F initial]
2.5
Post-breakage reserve [Fresidual / F initial]
1.5
1
annealed
0.5 heat-strengthened
fully tempered
0
0 25 50 75 100 125 150 175 200 225 250
Calculated failure stress [N/mm2 ]
Figure 7: (a) Results of Kreher et al. for timber-glass beams (modified from [14, 15]);
(b) results of the current research into SG-laminated reinforced glass beams.
It should be noted that the x-axis in (b) depicts the calculated failure stress (using the methodology as
explained in [2]), whereas the x-axis in (a) depicts the measured pre-stress level in the glass using a
differential stress refractometer [15].
Furthermore, it should be noted that the y-axes of the plots can not directly be compared as they relate to
different beam geometries and different material use.
700
On the one side, the application of stronger glass types (i.e. heat-strengthened or fully
tempered glass) in reinforced glass beams seems advantageous, as the heat-strengthened
and fully tempered beams investigated in this research reach 3 or 4 times higher initial
glass breakage loads (Finitial, see Table 2) than the annealed beams. For practical
applications this may result in a more material efficient cross section of the beams.
Furthermore, contrary to annealed glass, the phenomena of stress corrosion [16] or
‘static fatigue’ is absent (or suppressed) for heat-strengthened and fully tempered glass,
which is beneficial for the long-term design strength of these beams.
On the other side, however, the application of stronger glass types introduces some
disadvantages for reinforced glass beams. As is demonstrated by the current research
results the application of stronger glass types – especially fully tempered glass – results
in reduced post-breakage strength and reduced post-breakage ductility. Furthermore, the
application of heat-treated glass types may introduce uncertainties about the actual pre-
stress level – which may vary between different producers and different production
batches – and thus about the expected failure pattern in the beams. A thorough quality
control, e.g. by means of a scattered light polariscope [17] by which the actual pre-stress
can be determined, is therefore necessary before applying heat-treated glass in a
reinforced glass beam. Additionally, nickel-sulphide inclusions may cause unexpected
failure in fully tempered glass. However, the probability of the latter can be
significantly limited by subjecting the glass to a standardized heat soak test [18].
The final choice of glass type for reinforced glass beams needs to be determined by the
engineer, taking the above mentioned considerations into account. Additionally, issues
such as lateral stability [19] at the post-breakage stage may need to be considered,
especially for the fully tempered glass beams.
6. Conclusions
From the four-point bending tests performed on SG-laminated reinforced glass beams
composed of either annealed, heat-strengthened or fully tempered glass, the following is
concluded.
The application of stronger glass types (i.e. heat-strengthened and fully tempered glass)
in reinforced glass beams significantly enhances the initial glass breakage strength, but
affects the post-breakage strength and ductility in a negative way. Whereas the annealed
and heat-strengthened beams reached similar post-breakage strength levels, the post-
breakage strength of the fully tempered beams was reduced. Due to their extensive
fragmentation, the fully tempered glass beams did not demonstrate the additional load-
carrying mechanism of shear transferring overlapping glass fragments as was present in
the annealed and the heat-strengthened beams. However, it should be noted that even
the fully tempered beams reached significant post-breakage strength levels. Furthermore,
compared to the annealed beams, the post-breakage ductility was reduced for the heat-
strengthened beams and even further reduced for the fully tempered beams. Again this
originates from more extensive fragmentation of the heat-strengthened and fully
tempered glass, which causes the beam to collapse earlier in the post-breakage
trajectory due to instability and subsequent failure of the glass compression zone
In addition to the current paper, the research results are integrally discussed in [20].
701
Challenging Glass 3
7. Acknowledgements
The material and technology support of Glas Trösch AG, Swisslamex is gratefully
acknowledged.
8. References
[1] Veer, FA, 10 years of zappi research, in J Vitkala editor, Glass Processing Days, Tampere Finland, page
424-428, 2005.
[2] Louter C. Fragile yet Ductile - Structural Aspects of Reinforced Glass Beams. Dissertation Delft
University of Technology, TU Delft, ISBN: 978-90-8570-743-1, Delft; 2011.
[3] Bos FP. Safety Concepts in Structural Glass Engineering – Towards an Integrated Approach.
Dissertation Delft University of Technology, TU Delft, ISBN 978-90-8570-428-7, Delft; 2009.
[4] Louter C, Belis J, Veer FA, Lebet JP, Durability of SG̺laminated reinforced glass beams: effects of
temperature, thermal cycling, humidity and load ̺ duration, Construction and Building Materials,
volume 27, issue 1, pages 280̺292, Febr.2012, http://dx.doi.org/10.1016/j.conbuildmat.2011.07.046,
[5] EN 572-1:2004. Glass in building - Basic soda lime silicate glass products - Part 1: Definitions and
general physical and mechanical properties. CEN; 2004.
[6] EN 1863-1: 2000. Glass in Building – Heat strengthened soda lime silicate glass – Part 1: Definition
and description. CEN; 2000
[7] EN 12150: 2000. Glass in Building – Thermally toughened soda lime silicate safety glass – Part 1:
Definition and description. CEN; 2000.
[8] EN 10088-1: 2005. Stainless steels - Part 1: List of stainless steels. CEN; 2005.
[9] EN 10088-2: 2005. Stainless steels - Part 2: Technical delivery conditions for sheet/plate and strip of
corrosion resisting steels for general purposes. CEN; 2005.
[10] Stelzer I. High Performance Laminated Glass. In: Bos, Louter, Veer, editors. Challenging Glass 2,
Delft: TU Delft; 2010, p. 467-74.
[11] DuPont. DuPontTM SentryGlas® architectural safety glass interlayer. E.I. du Pont de Nemours and
company; 2009.
[12] Bucak Ö, Meissner M. Trag- und Resttragfähigkeitsuntersuchungen an Verbundglas mit den
Zwisschenlage SentryGlas Plus –Abschlussbericht. München: AIF – Fachhochschule München; 2005.
[13] Haldimann M, Luible A, Overend M. Structural Use of Glass, Structural Engineering Documents 10,
IABSE-AIPC-IVBH, ISBN 978-3-85748-119-2, Zürich, Switzerland, 2012
[14] Kreher, K, Natterer Jul., Natterer, Joh. Timber-Glass-Composite Girders for a Hotel in Switzerland.
Structural Engineering International 2004: 2: 149-151
[15] Kreher K, Tragverhalten und Bemessung von Holz-Glass-Verbundträgern unter Berücksichtigung der
Eigenspannungen im Glas, Dissertation Nr. 2999, IBIOS/EPFL 2004.
[16] Gy R. Stress corrosion of silicate glass: a review. Journal of Non-Crystalline Solids, Volume 316, Issue
1, February 2003, Pages 1-11, ISSN 0022-3093, 10.1016/S0022-3093(02)01931-2.
[17] Nielsen JH, Olesen FO, Stang H. Characterization of the Residual Stress State in Commercially Fully
Toughened Glass. Journal of Materials in Civil Engineering 2010;22;2:179-85

[18] EN 14179. Glass in building: heat soaked thermally toughened soda lime silicate safety glass - Part 1:
Definition and description; June 2005.
[19] Belis J, Kipsterkte van monolithische en gelamineerde glazen liggers, Dissertation, Ghent University,
Belgium, 2005.
[20] Louter C, Belis J, Veer FA, Lebet J̺P, Structural response of SG̺laminated reinforced glass beams;
experimental investigations on the effects of glass type, reinforcement percentage and beam size, ,
Elsevier, Engineering Structures, Volume 36, Pages 292 ̺ 301, March 2012,
http://dx.doi.org/10.1016/j.engstruct.2011.12.016
702
doi:10.3233/978-1-61499-061-1-703
The Applicability Evaluation of Hybrid

Glass Structure Using Pre-stressing as
Structural Member
Naoya Miyasato, Akira Okada, Masao Saitoh
Nihon University, Tokyo, Japan, miyasato@arch.cst.nihon-u.ac.jp,
okada@arch.cst.nihon-u.ac.jp, m.saitoh@dome.arch.cst.nihon-u.ac.jp
Keisuke Nomoto
AXS Satow Inc, Tokyo, Japan, let_0572_be@yahoo.co.jp
Glass has high compression strength but low tensile strength. In this paper the
authors proposed Hybrid Glass Beam (HGB “model-a, b, c and d”) that can control
the stress of glass member to pre-stress into the tension member. The derivation
and characteristics of each model are shown through the results of experiments and
numerical analysis of the proposed models.
Keywords: Glass, Hybrid, pre-stress, stress control, failure examination
1. Introduction
Curtain walls using glass have been in widespread use from the late 20th century, and
flat glass has been used for architectural façade. Generally glass is designed not to bear
a long-term loading and a seismic force in Japan. However, recently glass has been used
as the member to bear long-term loading to make architecture more transparent. Glass
has high compression strength but low tensile strength. Glass mainly fails due to tensile
stress. Therefore, if tensile stress generated in the glass is reduced, glass can bear a
larger load. For example, tempered glass has been processed by controlled thermal
treatments adding compression stress increase its strength compared with normal glass.
2. Proposal of Hybrid Glass Beam

In this paper the authors proposed Hybrid Glass Beam (HGB “model-a, b, c and d”) that
can control the stress of glass member to pre-stress into the tension member. This
chapter describes the derivation and characteristics of the proposed models (HGB).
First of all, a stress control system using glass has been generated to replace the strut
and beam of Beam String Structure (BSS) with glass members (model-a), as shown in
Figure1. “Model-a” is almost equal to the crack control system of pre-stressed concrete.
The cable is expected to temporarily support its own weight after glass failure. The
other model of this system is generated to replace the strut of the cable girder or the
tension truss into the glass member (Model-b). “Model-b” can apply compression load
with tension force of the string and generate horizontal reaction of the cable. So “model-
b” is a little different from “model-a” in the string’s function.
Secondly, the “model-c” is generated by replacing the strut and the diagonal member of
703
Challenging Glass 3
Tension Truss with glass plates, as shown in Figure 1. The concept is shown below.
(1). The displacement and the stress control system for the beam member using pre-
stressing into the tension member are effective.
(2). This system doesn’t fall down and can bear load after some glass failure.
(3). Extend the span by using divided glass instead of a single glass in “model-a”.
Finally “mode-d” was derived from the Tension Truss, and from the pathway of
“model-c”, as shown in Figure 1. “Model-d”, which consists of tension members and
glass members, has maintained transparency which is required in its performance as the
supportive structure. This system is capable of ensuring the reduction of PS and the out-
of-plane stiffness. The concepts are listed below;
(1). Varying the resistance mechanisms can be expected to improve rigidity compared
to cable grid, and to reduce the tension difference of the strings compared to
Tension Truss.
(2). Glass beam with the cable doesn’t fall down immediately after glass failure.
(3). It is possible to extend the span using divided glass instead of a single glass plate
like “model-a”.
Figure 1: Proposal of Hybrid Glass Beam
704
The Applicability Evaluation of Hybrid Glass Structure Using Pre-stressing as Structural Member
3. Study and Results for HGB “model-a”[1]
3.1. Object of the experiment and numerical analysis

“Model-a” is consist of glass beam and tension member which arranged at bottom of
beam. The objective of this experiment was to confirm the effect of stress control and
strength improvement. The experiment with small scale model shows that the stress
control for the glass member using pre-stressing into the tension member is effective.
But it needs further consideration to apply to real scale structure. Therefore, the results
of failure test with a real scale model are described validity of this system, and the FEM
analysis is conducted to confirm the validity of experimental result and stress state. The
considering point of real scale model can be described as following.
(1). Connection between the top panel and the glass beam,
(2). Mechanism of Load transfer on this system,
(3). The feature of behavior after the glass failure,
(4). Validity of backup system with cable after the glass failure,
(5). Performance for additional Load after the glass failure.
The outline of the analysis is shown in figure 2. The outline of the real scale model is
shown in figure 3.
Table 1: Material data of the analysis

Glass E = 7.3×10͵N/mm²
Edge hardware (steel) E = 2.1×10ͶN/mm²
Cable 2 (SUS304) E = 1.2×10ͶN/mm² , A=1.86 mm²

Figure 2: The outline of the analysis
705
Challenging Glass 3
Figure 3: The outline of real scale model
Table2: Material data of the real scale model
Beam, Top board (float glass) Eg = 7 .3×10͵N/mm²
(edge) Long-term allowable stress = 7 N/mm²

(edge) Short-term allowable stress = 18 N/ mm²
Edge hardware, Point support hardware (SUS304) Es = 1.97×10ͶN/mm²
Cable4 [7×7strand] (SUS304) Ec = 1.20×10ͶN/mm²SGGAc = 7.45 mm²G

Figure 4: Comparison between experimental results and numerical analysis with a real scale model
706
3.2. Results of HGB “Model-a”

Comparison between experimental result and numerical analysis with real scale model
is shown in figure 4.
The real scale model has two glass beams that are arranged in parallel. The back side of
the glass beam failed firstly at the middle edge of the beam under a load of about 8.2KN.
Continuing to apply load, the remaining glass beam and two cables could bear the load.
The results are shown as follows:

(1). Stress control system to the glass using pre-stressing into the tension member is
effective.
(2). Glass beam doesn't fall down with the cable after the glass failure.
(3). Validity of the numerical analysis method is confirmed.
The availability and the validity of this system are verified with real scale experiments
and numerical analysis.
4. Study and Results for HGB “model-c” [2]

“Model-c” is consists of beam, struts of a glass plate and a tension member at bottom of
strut. This model can be extended the span by using divided glass plate. The results
obtained from the experiment with the small scale model indicated that this stress
control system using pre-stressing is effective. In this chapter, the author describes the
results of the experiment for a real scale model which has a span of 5m. Through the
experiment and numerical analysis the following factors and its practicality are
validated.
(1). Experimental value of the stress as all materials is under allowable stress value.
(2). Validity of the backup system after the glass failure.
(3). Bearing load after the glass failure.
(4). Influence of the gum plate as a buffer.
The outline of the analysis is shown in figure 5. The outline of the real scale model is
shown in figure 6.

Glass [t=8] E=7.3×10͵N/mm²
Aluminum [C-80×50×t=3] E=0.7×10ͶN/mm²
Edge hardware (steel) E=2.1×10ͶN/mm²
Cable 2 (SUS304) E=1.2×10ͶN/mm² , A=1.86mm²
707
Challenging Glass 3
Figure 6: The outline of real scale model

Figure 7: Comparison between experimental results and numerical analysis with the real scale model
708
4.2. Results of HGB “Model-c”

Comparison between experimental result and numerical analysis with real scale model
is shown in figure 7. This experimental model consists of 5 divided glass struts, a
tension member and an aluminum beam. The experiment was performed three times.
For the 1st experiment model, all of the materials and the experimental values of the
stress were under the allowable stress limits. The ultimate state of the beam in the 2nd
experiment is the aluminum and the glass deformed out-of-plane (lateral buckling).
In the 3rd experiment, an impact fracture test was done for the full glass model as
follows:
(1). The load was applied until snow load level, and the middle glass section was
impacted with a hammer.
(2). After the load (about 600N) was added, the sections of glass next to the middle
section were impacted.
(3). This cycle was repeated three times until, finally, the three middle sections of
glass were at failure
The results of this model are shown as follows:
(1). The displacement and the stress control system of the glass beam using pre-
stressing were effective.
(2). Glass beam with cable doesn’t fall down immediately after glass failure.
(3). The failed glass beam can bear more than snow load level.
(4). The numerical analysis of this method was valid.
The reliability and validity of this system with real scale experiment and the numerical
analysis were confirmed.
5. Study and Results for HGB “Model-d”

An experiment using a small scale model which had a span of 1.6m was done to test the
following concepts and its practicality was validated.
(1). The impact of changes in the amount of PS and boundary conditions.

(2). Effect of a model with partially removed glass.
(3). Effect of sliding of the cable due to the tension difference between the strings.
(4). Confirming state of the connection of glass with tension member at both ends of
glass plate.
The parameters of the experiment were boundary conditions (presence of the vertical
support which is the outermost of point support hardware) and the amount of PS
(PS=0N, 500N, 1000N, 1500N). It should be noted that the boundary conditions were
provided with a pin, such as binding to nodal vertical displacement of the test piece at
both ends. The outline of the analysis is shown in figure 8. The outline of the real scale
model is shown in figure 9.
709
Challenging Glass 3

Glass(t=3mm) E=7.3×10͵N/mm² Hardend gum plate E=10.5N/mm²
Point support hardware E=2.1×10ͶN/mm² (90 degree of hardness)
Cable 4(SUS304) E=1.2×10ͶN/mm² , A=6.67mm² Rod 5 A=19.63mm²

Figure 9: The outline of small scale model
710
Figure 10: Comparison between experimental results and numerical analysis with the real scale model.
5.2. Results of HGB “Model-d”

Comparison between experimental results and numerical analysis with the real scale
model is shown in figure10. It should be noted the influence of the amount of PS was
confirmed in only non-vertical support model. For vertical support model, in all graphs,
which shows generally linear properties, a decrease in stiffness was observed due to loss
of tension of measurement point T3 for P=1.4kN. On the other hand, for non-vertical
support model, in all graphs, although showing a slight non-linear property,
phenomenon of loss of tension was not identified. With the increase in amount of PS,
the stiffness was observed to increase substantially.
The reliability and validity of this system is shown by the experiment and the numerical
analysis confirmed the following;
(1). The effects of changes in the amount of PS and boundary conditions

(2). The failed glass beam can bear sufficient load.
(3). The numerical analysis method was valid.
(4). The effects of the sliding of the cable due to the tension difference between
the strings.
711
Challenging Glass 3
6. Summary
In this paper, Hybrid Glass Beam (HGB “model-a, b, c and d”) is proposed and the
derivation and characteristics of each model is described. The results of each model are
shown below and in figure 11.
Figure 11: Summary of Hybrid Glass Beam (HGB “Model-a,c,d”)
712
Results of “model-a” demonstrate (1) stress control system to the glass using pre-
stressing; (2) glass beam don’t fall down with the cable after glass failure; and (3)
validity of the numerical analysis method.
Results of “model-c” demonstrate (1) the displacement and the stress control system
using pre-stressing are effective; (2) HGB doesn’t collapse immediately after the glass
failure because the tension member was installed on the bottom of the glass beam; and
(3) the failed glass-beam could bear a higher load than the dead load of the glass beam.
Results of “model-d” demonstrate (1) the impact of changes in the amount of PS and
boundary conditions; (2) the failed glass beam can bear sufficient load; and (3) the
effects on the sliding of the cable due to the tension difference between the strings.
7. References
[1] Naoya MIYASATO, Akira OKADA, Keigo FUJIWARA, Yusuke MIYAKE and Masao SITOH., The
Applicability of Glass Structure Using Pre-Stressing, IASS, 2009, Shanghai
[2] Naoya MIYASATO, Akira OKADA, Keisuke NOMOTO, Yusuke MIYAKE, Masao SITOH., An
Applicability Evaluation of Hybrid Glass Beam as Structural Member, IASS, 2011, London.
[3] Asahi Glass Company Limited, General Catalogue of Glass Construction Materials (Japanese). 2010.
[4] P.C. (Christian) Louter. Adhesively bonded reinforced glass beams, Heron Vol.52 2007.
[5] IABSE, Structural Engineering Documents 10 Structural use of glass, 2008.
[6] Tony Robbin, Engineering a new architecture, 1996, p.109 –p.111.
[7] IASS 50ht Anniversary Jubilee book: Fifty Years of Progress for Shell and Spatial Structures
713
doi:10.3233/978-1-61499-061-1-715
Structural Design of
Composite Steel-Glass Elements
Michal Netusil, Martina Eliasova
Czech Technical University in Prague, Czech Republic, michal.netusil@fsv.cvut.cz
Composite glass beams, stiffened by adhesively bonded high-strength element,

show high load bearing capacity and robustness (post breakage behavior) in
comparison with pure glass beams or fins. Wide range of polymer adhesives can be
used in composite structure with real load bearing role to provide horizontal shear
interaction between both materials and on the other hand to keep all other
requirements on compliancy. Numerical and analytical approach was developed
and calibrated by the results of full scale tests of composite steel-glass beams.
Reached load carrying capacity of the beam was very high and influence of a lot of
parameters was defined and included to analytical model. Generalized results of
performed experiments, numerical and analytic studies can say, how to predict an
ultimate load of such a hybrid structure, where the key element is the semi-rigid
shear connection, provided by modern and progressive polymer adhesives used to
be known mainly from industry until now, although their potential is very high.
Keywords: Composite steel-glass element, polymer adhesive, analytic solution
1. Fundamental Experiments
After the choice of adhesives in respect to many initial requirements, material tests and
steel-glass connection tests were performed to find out the real material properties,
especially their shear strength and progress of shear modulus during the range of load.
Procedure of finding of these parameters as well as the calibration of different
numerical models for each of the adhesive type was described in detail in [1].
1.1. Experiments of full-scaled hybrid beams

Composite steel-glass beams consist of steel flanges and glass web bonded together, see
Figure 1 on the left.
Figure 1: Composite Steel-Glass Beam, left – view, right – details of the steel-glass connection.
715
Challenging Glass 3
The shear force is carried by the glass web, whereas the bending capacity of the hybrid
beam is significantly increased by slender steel flanges compared to the pure glass pane.
The shear forces between steel and glass are only sustained by the adhesive between
them. Different details of connection between steel and glass designed to be used in
hybrid beams are drawn in the Figure 1 on the right. To maximize the exploitation of
steel and glass, chosen adhesive has to ensure an adequate stiffness but on the other
hand must be soft enough allowing a compensation of different temperature elongation
of steel and glass and reduction resp. redistribution of stress peaks or other constraints.
The best aesthetic appearance was achieved with direct connection between glass web
and steel flanges, see Figure 2.
Figure 2: Composite Steel-Glass Beam with direct connection between web and flanges.
This kind of connection offers the best way to investigate and model the behaviour of
hybrid beam with neglected effect of additional profiles welded on flanges, side-
bonding or influence of interlayer on numerical and analytic studies. Therefore, this
paper stay focused only on this representative connection type with acrylic adhesive,
although the other details of connection shown in Figure 1 can provide higher load
bearing capacity, post breakage behavior or even better stress distribution in cross-
section, which was proved experimentally.
Figure 3: Set-up of the full-scale tests of the hybrid beams (with lateral support in midspan)
716
Structural Design of Composite Steel-Glass Elements
Beams were subjected to 4-point bending tests until the glass failure, see Fig. 3, where
the lateral support in the middle of the span was arranged to avoid lateral-torsion
buckling effects. Measured values were deflection in the middle of the span and under
the load introduction points, normal stresses along the length of the beam and horizontal
sliding between steel and glass, see Figure 4 on the left.
Tested beams were 4,25 m long (span of 4 m) and consisted of steel grade S235 flanges
with dimensions of 100x10 mm. Final choice of adhesives resulted to acrylic adhesive
from producer Sika, which is in the scope of this paper and which shows advantageous
properties in connection subjected to shear [1]. Designed thickness of adhesive was
3mm. Single-layered 19 mm toughened glass with the height of 290 mm was used as a
web of the beam. In this case, adhesive joint is performed only along the front edge of
the glass web, though the area of adhesion is restricted by the thickness of the glass web.
As it was described above, this setup is beneficial for modeling and prediction of
behaviour in term of minimizing other effects of different parameters (PVB foil, side
connection, additional profiles), but single layered toughened glass would never be used
in real structure because of level of residual load-bearing capacity equal to zero, see
Figure 4 on the right.
Figure 4: Experiment of the beam with single layered toughened web

1.2. Results
3 beams were tested with acrylic adhesive and described type of direct connection
between web and flanges. Force F was applied in different velocities, see Figure 5. In
case of Beam 2, stabilized values were checked and also unloading was performed.
Summarized results are in the Table 1, where can be seen, that Beam 1 failed by the loss
of adhesion at the interface of adhesive layer and steel flange.
Table 1: Summary of the results

Specimen Ultimate Force Reached Max. vertical Speed of load Firstly
at midspan F Glass stress deflection introduction failed
[kN] [Mpa] [mm] [N/s]
Beam 1 65,10 116,97 19,5 50 adhesion
Beam 2 89,79 139,75 23,6 150 with unload. glass
Beam 3 72,34 104,36 17,9 150 glass
Average 75,74 120,36 20,3
717
Challenging Glass 3
Figure 5: Different velocities of load introduction on tested beams
Nevertheless, reached stress in glass web was close to 120 MPa, therefore it can be
assumed, that this adhesion failure didn’t affect the results significantly. Effect of
different speed of load introduction on growing stress in the glass pane can be seen in
the Figure 6, where can be also seen, that sustained (but still short-time) measured
stresses in the beam 2 under constant value of applied force F=40 kN reached almost
the same value like in case of beam 1, which was loaded very slowly to provide a time
for adhesive layer to relax (in the range of minutes, not long-term relaxing).
Figure 6: Influence of the velocity of load introduction on reached stress in glass pane
718
Stress distribution, measured during the experiment on mostly relaxed beam 2, is drawn
in the Figure 7 for different bending moments acting on the cross-section of the beam.
Horizontal shear interaction between components is higher when the bending moment is
smaller, because the shear stiffness of adhesive connection decreases by increasing
stress within the adhesive layer [1].
Figure 7: Sustained value of stress distribution in the cross-section of the hybrid beam with acrylic adhesive
2. Analytical approach
2.1. General analytic model

Adhesively bonded steel glass beam is a flexible composite section. Load bearing
characteristics significantly depend on the stiffness of the connection between flanges
and web. First limit case presents a fully rigidly connected section. There is no
displacement between steel and glass. Steel flanges are loaded with moments and high
normal forces. This leads to a high load carrying capacity. Second limit case is
presented by totally unconnected section members. There is a high displacement

between steel and glass. All section members are loaded with bending moments only.
Because of the low bending stiffness and resistance of flanges, such a beam shows low
carrying capacity and high deflections compared to the beam in first limit case. The
behaviour of real bonded hybrid steel-glass-beam is located in between these two cases.
There is a displacement between steel and glass due to adhesive flexibility. The load
carrying capacity generally increases with the increase of the connection stiffness,
because of the higher normal forces in the flanges. Shear resistance and shear modulus
of adhesive are important characteristics and have high influence on the load carrying
capacity and structural behaviour of the composite steel-glass element. The stiffness of
the bonded connection depends mainly on the stiffness of the adhesive itself and
geometry of the bonded joint.
Different analytic models for composite beams are available and can be used for the
design of hybrid steel glass beams. Since most adhesives behave nonlinear, a linear
stiffness of best fit should be used for analytical calculations.
Flexible composed hybrid-beams can be pre-designed according to Möhler [2], who
developed his method (which is today known as the -method) in 1956 for the design of
719
Challenging Glass 3
timber structures. It can be applied to the hybrid steel glass beam considering the
different material properties of steel and glass as well as the flexibility of the joint with
an effective moment of inertia with reduced contribution of the flanges. Therefore, it is
qualified very well for the pre-design of hybrid steel glass beams.
2.2. Application on composite steel-glass beam

Assume the simplest bonded connection between steel flange and glass web - direct
connection on the face of glass-sheet with one adhesive layer, see Figure 8.
Figure 8: Detail of the direct glued joint
Average stiffness of the connection KK (along its length) can be determined by the
formula (1), where GK is experimental found shear modulus of used adhesive in [MPa],
b is the width of adhesive in [mm] and tk is the thickness of the adhesive layer in [mm].
b
KK GK (1)
tk
Efficiency factor and effective moment of inertia Iy,eff of double symmetric cross
section are given by the following formulas:
1
J , (2)
1 k
I y,eff 2 I a n I G 2 J Aa za2 , (3)
where
Ea Aa
k S2 , (4)
L2 KK
EG
n , (5)
Ea
720
where Ea is the Young’s modulus of steel [MPa], EG is the Young’s modulus of glass
[MPa], Aa is the area of one steel flange [mm2], L is the span of the beam [mm], Ia is the
moment of inertia of one steel flange [mm4], IG is the moment of inertia of the glass web
[mm4], za is the vertical distance between the centre of gravity of the steel flange and
centre of gravity of whole cross section [mm].
Figure 9: Stress distribution in composite cross-section
Values of the normal stress distribution along the cross section drawn in Figure 9 can be
obtained from following formulas:
My § tf ·
V r 1a r ¨¨ J z a ¸¸ (6)
I y ,eff © 2 ¹
My § tf ·
V r 1i r ¨¨ J z a ¸¸ (7)
I y ,eff © 2 ¹
My
V s1 r J za (8)
I y ,eff
My hw
V r2 r n (9)
I y ,eff 2
721
Challenging Glass 3
where, except the values described above, My is the maximal bending moment caused
by external load [Nmm], tf is the thickness of the steel flange [mm] and hw is the height
of the glass web in [mm]. Shear stress within the adhesive layer can be calculated
(simplified) by (10), where Q is the maximum shear force caused by external load.
Q J za Aa
W (10)
I y,eff b
Finally, vertical deflection in the middle of the span of the hybrid beam subjected to
uniformly distributed load q [kN/m] can be determined by formula (11).
5 q L4
G . (11)
384 E a I y ,eff
As the most of the adhesives behave non-linear and their shear modulus doesn’t have
constant value by increasing load, specific secant shear modulus has to be used as an
input into the formula (1). This modulus has to be adequately found in the shear stress –
shear strain diagram of specific bonded connection according to the current stress state
of the glued joint caused by external load, acting on the beam at the same moment. Or
Möhler’s method can be modified by using software included the analytic description of
the shear stress – shear strain behaviour of the specific glued joint. Than, after a few
iterations, an equilibrium state of specific input shear modulus corresponding to the
current shear stress caused by external load can be found [3].
Figure 10: Stress prediction by Möhler’s method assuming the non-linar change of adhesive stiffness
722
Very accurate prediction of stress distribution in all components of the cross-section

determined by described analytic approach, with assuming the non-linearity of the
adhesive behaviour and changeability of the adhesive shear modulus, can be seen in
Figure 10.
3. Numerical approach
Finite element analysis supported by software package Ansys was applied earlier to
create properly working non-linear material models of different adhesives in overlapped
connection, subjected to shear [1]. The best fitted material model for acrylic adhesive
was multi-linear isotropic with appropriate input material curve. This models was later
on fine-tuned according to the real thickness of the joint and included into a complex FE
model of the hybrid beam, see Figure 11.
Figure 11: Left – FE model of the beam, right – resulting stress distribution in the glass web
4. Comparison and conclusions

In both cases, numerical simulation of hybrid beam under increasing load and bending
moment led to accurate prediction of the behaviour of the hybrid beam as well as
determining of important values of stresses and deflections, which were slightly higher
than experimentally obtained, therefore can be assumed on the safe side, see Table 2.
Numerical approach confirmed, that FE analysis is a strong tool for design of new
composite structure or prove the existing one.
Table 2: Comparison of experimental, analytic and numerical results

Value Analytic FEA Experiment
Max. normal stress in the glass pane r2 [MPa] 38,4 40,3 36,7
Max. normal stress in steel r1a [MPa] 31,5 33,0 32,8
Shear stress in adhesive [MPa] 1,45 1,65 -
Vertical deflection of the beam [mm] 6,3 8,6 6,3
Based on the large scale tests and the finite element modeling of the hybrid beams,
modified analytical model is proposed to be used. This model has been specifically
calibrated using the material laws developed for chosen adhesive and enable the load
bearing capacity of the hybrid beam to be determined for varying spans and loading
magnitudes.
723
Challenging Glass 3
Design guidance relating to hybrid beams was presented and enables designer to follow
a procedure, which includes selecting the most appropriate adhesive, obtaining the
material laws of the adhesive and the determination of the load bearing capacity of the
hybrid beam using simplified analytical model and appropriate finite element models
[4].
Future research in the field of hybrid beams is needed for complex description of their
behaviour under different factors and for defining safe and reliable design rules. Main
topics to be investigated are the behaviour of the beam under long-term load (creeping
of adhesive), determining the ageing effects of adhesives, protection from corrosion and
fire, robustness (post breakage behaviour of multi-layered glass web), cyclic load or
lateral torsion stability.
5. Acknowledgements
This research, experimental, numerical and analytic research was supported by the grant
of the Ministry of Education of the Czech Republic nr. LD11037.
6. References
[1] Netusil, Michal; Eliasova, Martina, Experimental and numerical analysis of glued steel-glass joints,
Challenging Glass 2 – Conference on Architectural and Structural Applications of Glass, Delft,
Netherlands, 20-21.5.2010, ISBN 978-90-8570-524-6
[2] Möhler K.: Über das Tragverhalten von Biegeträgern und Druckstäben mit zusammengesetzten
Querschnitten und nachgiebigen Verbindungsmitteln. TH Karlsruhe, 1956, Technischen Hochschule
Fridericiana zu Karlsruhe.
[3] Netusil, Michal; Eliasova, Martina, Hybrid steel-glass beams with polymer adhesive, Glass Performance
Days Conference Proceedings, Tampere, Finland, 2011, ISBN 978-952-5836-02-8
[4] Abeln B., Preckwinkel E., Yandzio E., Heywood M., Eliasova M., Netusil M., Grenier C.: Final Report
of RFCS-CT-2007-00036 INNOGLAST ”Development of innovative steel-glass-structures in respect to
structural and architectural design.“ Research Programme of the Research Fund for Coal and Steel,
TGS8 “Steel products and applications for buildings, construction and industry.
724
doi:10.3233/978-1-61499-061-1-725
In-Plane Loaded Glass-Steel Elements

Geralt Siebert
Universität der Bundeswehr München, Institute for Structural Design, www.unibw.de/ki
Andreas Haese
Ingenieurbüro Dr. Siebert, München, Germany, aha@ing-siebert.de
For this project one element is considered as a glass pane with a (small) edge beam
on the vertical edges. In a first step different load-paths and boundary conditions
were investigated. Even for very slender edge beams the load carrying capacity is
far higher when introducing the loads into the steel instead of the glass.
Accompanying to the numerical analysis experimental testing was done to verify
the calculation model and outcome. In the second step multiple systems with
different geometries, measures, stiffness and loads were examined determining
deformations, stresses and stability-parameters. The analysis of the results showed
some communality for all models that led to a surprisingly easy design and safety
concept.
Keywords: glass-steel elements, in-plane load, combined loading, façade

application, load carrying capacity, element failure scenarios
1. General Idea
The basic idea behind this project was to find a solution for a transparent building
envelope, e.g. for an existing monument under preservation order, that cannot take the
vertical loads of the façade structure. The typical approach would be a mullions and
transoms sub-structure with glass elements. But in order to minimize the visible steel or
non-transparent façade elements all load-bearing elements should be integrated in the
façade plane. For the given example of ancient buildings, a massive top or lateral
structure, as it would be necessary for a cable system, is not acceptable. As a result, the
vertical loads have to be brought down in the façade-plane. Figure 1 illustrates the basic
idea.
Figure 1: Basic idea: Standard (left) and integrated (right) façade structure
725
Challenging Glass 3
2. System definition
As the investigation is not based on a single project with defined boundary conditions
and dimensions, but is intended to give more general results on in-plane loaded façade
elements, the façade structure is analyzed on the basis of single elements as a cut-out of
the entire structure. Figure 2 shows the definition of the element parameter in respect to
the façade structure.
Figure 2: Main stiffness parameters as cut-out of the façade structure
Figure 3 shows the most important element parameters (B, H, EIRB, t or D) and the
possible loads (F, r, q) in Detail.
Figure 3: Detailed element parameters
726
3. Model
Of course state of the art finite element programs are capable of representing the
detailed connection of glass and edge beams. But due to the complex material behavior
of the interlayer (e.g. EPDM) on the one hand and to the small bearing width (to
minimize the non-transparent parts) on the other hand an effective clamping is usually
not given for this connection. Except for a glued connection friction cannot be assumed
either. As a result, the only relevant – mechanical – connection between the edge beam
and the glass pane is direct contact perpendicular to the glass surface. Therefore the
edge beam and the glass pane are modeled separately and connected only by springs,
representing the interlayer stiffness for direct contact.
a) b)
Figure 4: Connection edge beam and glass pane in reality (a) and in the FE model (b)
4. Load path comparison

In a first step, possible load paths for in-plane loads were investigated.
4.1. Load paths

The vertical loads can be transferred from element to element from edge beam to edge
beam (Fig. 5a), partly between the edge beams and between the glass edges (Fig. 5b) or
between the glass edges, either concentrated at the corners (Fig. 5c) along the glass edge
(Fig. 5d).
The FEM implementation of the different loading scenarios is shown in Fig. 6.
a) b) c) d)
Figure 5: Load paths for in-plane loads
727
Challenging Glass 3
a) b) c) d)
Figure 6: FEM implementation of the loading situations
4.2. Results
Table 1 shows the buckle loads of some investigated systems as an excerpt of the results
of the load path comparison. The results are given for “small” edge beams
(EIedge beam EIGlass) and rather big/stiff edge beams (EIedge beam 10 · EIGlass). As a
reference to load case d) the buckling loads of a four side linearly supported glass pane
are given in the right column. (for the values in brackets, the allowable stress in the edge
beam would be exceeded).
The data in table 1 shows that for all cases (glass and edge beam dimensions) the load
transfer between the edge beams leads to the highest load bearing capacity.
Table 1: Results of the load path comparison (excerpt)

B·H Load type Pcrit („small“) Pcrit („big“) Pcrit, glass*)
[mm · mm] [kN] [kN] [kN]
a) (112,5) (1038) -
d) 34,9 36,9 37,1
1500 · 1000
b) 67,9 73,7 -
c) 80,9 100,3 -
a) (111,6) (1036) -
d) 40,3 46,6 47,4
1000 · 1000
b) 74,4 93,1 -
c) 68,6 75,9 -
a) 50,4 (464,3) -
d) 31,9 51,1 51,5
1000 · 1500
b) 45,8 102,2 -
c) 34,9 46,3 -
)
* reference value from [3]
728
5. Load-deformation behavior
Based on the results of the load path study primarily the load transfer via the edge
beams was analyzed. As a consequence, the glass pane itself is not charged directly with
vertical loads but acts as a stiffening element to the loaded edge beams. By that
additionally the critical or complicated detail with concentrated load application on
glass corners disappears.
As the considered façade elements are exposed to in-plane and out-of-plane loads
(typically wind), the buckling load usually is not decisive. For façade elements,
especially for isolating glass units strict deformation limits have to be observed.
Therefore the following investigation concentrated on the load-deformation behavior of
the described elements for various dimensions and boundary conditions (parametric
study). The parameters were defined on the assumption of realistic sizes and dimensions
for a façade application.
5.1. Parameter definition

The main decisive and independent parameters are given in table 2:
Table 2: Main parameters

a aspect ratio B/H
s stiffness ratio EIedge beam / EIGlass
q surface load
t glass thickness
Based on these parameters and some assumptions (e.g. glass thickness), the further
parameters (EIGlass, EIedge beam,…) can be determined.
5.2. Parametric study

The first results of this parametric study were load-deformation curves for any
parameter combination within the analyzed range.
The decisive points to evaluate the element behavior are
x center of the glass pane (SMP),

x center of the edge beam (RBM) and
x center of the horizontal glass edge (QRM).
Fig. 7 shows an example of a load deformation curve for a certain parameter

combination. Additionally to the deformation behavior, the resulting stresses on the
glass surface have been determined depending on the parametric setting and the applied
vertical load.
729
Challenging Glass 3
Figure 7: Example for a load-deformation curve
6. Analytical approach
As the vertical load is applied onto the edge beam, the structure can be interpreted as
initially curved or loaded column (see fig. 8) with its determined behavior (ref. [1]).
Figure 8: Initially curved column according to [1]
The element stiffness results from the interaction of the edge beam and the glass pane.
The investigation showed that with an increasing aspect ratio (B/H) the relative
effective stiffness (EIeff / (EI edge beam + EIglass) decreases, meaning that – expressed as
effective glass width – less of the glass pane acts as stiffening to the edge beam.
Therefore in the next step the investigation concentrated on the determination of the
effective stiffness of the glass-steel element postulating an analogy of the element and
the interpretation as an initially curved column.
730
The effective stiffness was determined by two ways:
x Determination of EIeff based on the element buckling load and

x definig EIeff in that way, that the load-deformation-curve of the FE-model is
identical to that of the analogy-system
This analysis was done for all parameter combinations and both ways led to the same
result: The effective stiffness can be assumed as depending only on the aspect ratio. All
other parameters (e.g. q, s, t) do have an influence, but a small one compared to the
element geometry.
The effective stiffness EIeff can be expressed as a composed stiffness:
EIeff = EIedge beam + bm/B · EIglass (1)
with bm as effective stiffening part or effective width of the glass pane.
Figure 9 shows the relationship of the aspect ratio and the effective width of the glass
pane.
Figure 9: Relationship between aspect ratio and effective width of the glass pane
Figure 9 contains the data of 54 parametric models. It shows that the effective width
mainly depends on the aspect ratio.
Based on these results, the model parameters that led to the smallest effective width
within the same aspect ratio were identified and with these parameters further aspect
ratios were analyzed. The results are shown in Figure 10.
731
Challenging Glass 3
Figure 10: Effective width of the glass pane as a function of the aspect ratio
7. Experimental study
To verify the numerical results a test set-up was installed (Figure 11a). With that mock-
up two different aspect ratios and different load introductions (edge beam: Fig. 11b or
glass edge: Fig. 11c) were tested.
a) b) c)
Figure 11: Test set-up (a) with loading of the edge beam (b) or the glass edge (c)
Figure 12 shows an example of the numerical and the experimental curves. A major
problem conducting the tests was the poor quality of the edge beams. The differing
cross sections of the U-shaped edge beams made it necessary to determine the cross
section characteristics for every individual sample to be able to compare the tests with
the numeric results.
“LERB” in Figure 12 refers to a load introduction into the edge beam and “LEGK”
refers to a load introduction into the glass edge concentrated near the corner.
732
Figure 12: Numerical (FEM…) and experimental (RS1H…) results
8. Results
Based on the given results, the load deformation curves were determined for all
investigated models and compared to the analytical approach. As the effective width
had partly been determined by comparison of the deformation curves for the point at the
middle of the edge beam (RBM), these curves naturally are very close. It showed, that
using the same amplification factor V = wRBM(F) / wRBM (F=0) – with wRBM(F)
determined by using the analogy-system with bm – for the other decisive points (plate
center: SMP and middle of the horizontal glass edge: QRM) lead to a good estimation
of the resulting deformations. With an increasing aspect ratio, the deformations at the
points QRM and SMP are over estimated. In any case the assumption of V is on the safe
side.
Figure 13: Load deformation behavior according to numerical results and analytical approach
733
Challenging Glass 3
Using the same amplification factor for the determination of the glass surface stresses
leads to good or “safe” results for the points SMP and QRM, but the stresses at the
vertical corner of the glass plate (RBM) would be underestimated.
Further investigation showed that the stresses at RBM can be approximated by setting
Vstress,SMP = V1,2. Figure 13 gives an example of the load stress behavior according to the
numerical results and the analytical approach.
Figure 14: Load stress behavior according to numerical results and analytical approach
9. Summary
In the presented project with laterally reinforced glass elements were investigated
regarding its load carrying capacity and its load deformation behavior for in-plan and
out of plane loads. In several parametric studies the behavior was determined for typical
façade parameters and boundary conditions. Based on the results an analytical model
was found, that enables the engineer to easily estimate the resulting deformations and
stresses for a certain application. More detailed results can soon be found in [2].
The authors want to thank the “DASt – Deutscher Ausschuss für Stahlbau” for its
support of this project and the “AIF – Allianz Industrie Forschung” for funding main
parts of this project.
11. References
[1] Petersen Ch., Statik und Stabilität der Baukonstruktionen, 2. Auflage, Vieweg Verlag 1982
[2] Haese A., Beitrag zur Bemessung scheibenbeanspruchter Stahl-Glas Elemente, Universität der
Bundeswehr München, Institut für Konstruktiven Ingenieurbau, in progress
[3] Englhardt O., Flächentragwerke aus Glas – Tragverhalten und Stabilität, Dissertation, Schriftenreihe
des Departments Nr. 12 – Dezember 2007, Universität für Bodenkultur Wien, 2007
734
doi:10.3233/978-1-61499-061-1-735
Load-Bearing Behaviour of Non-monolithic

Glass Laminate Beams for Wide Spans
Erich Trösch
Glas Trösch Holding AG, Bützberg, Switzerland
Ernst Wälchli
Bauingenieur SIA/USIC, Langenthal, Switzerland
Thomas Baumgärtner,
Glas Trösch GmbH, Nördlingen, Germany
In modern architecture, glass is increasingly being used as a structural material. For

example, glass fins, roof beams, facade supports and even entire glass bridges are
being made from laminated safety glass. In the past, however, this was much
restricted by the achievable dimensions. Glas Trösch took up the challenge,
presenting for the first time a 21 meter long laminated glass beam at the glastec
2010 in Düsseldorf. (Image 1) This world-first met with considerable interest from
architects and builders. That prompted the company to conduct a scientific
investigation on the load-bearing behaviour of large beams of this type.
Keywords: Glass beam, wide span, laminate, Segment
1. Introduction
Glass is one of the oldest artificially manufactured materials used for building. Its
fascination is today greater than ever. No other material so opens up the building shell
to both the inside and the outside, permitting an incomparable transparency. Its many
uses and its excellent resistance to the effects of weather make glass an ideal material
for building. So it is no surprise that a trend has been developing for some years now in
building towards open architecture, to spatial concepts with increasing transparency and
to light-flooded rooms.
Thanks to ongoing research work, new findings are coming to light on the behaviour of
basic glass and its refinement products. More efficient manufacturing processes, new
joining techniques and improved verification procedures make possible applications that
couldn't be predicted in the past.
Glass beams up to a length of more than 20 meters are today technically feasible. They
allow the visions of architects, such as facade structures having only vertical glass fins
or beams without any horizontally load-bearing elements, to become reality without any
problem.
735
Challenging Glass 3
Figure 1: Laminated glass beam, 21 meters long, at the Glastec 2009 in Düsseldorf
Figure 2 a and b: Detailed view of glass beam
Today, the structural strength of glass elements made from multiple laminates can be
easily determined with appropriate stability verifications and supplementary FEM
analyses. However, investigations to date have been based only on multiple-laminate
glass beams in which the individual glass sheets have the same geometry as the overall
beam. At present, individual glass sheets can only be tempered up to a maximum length
of 9 meters. To cater for present-day architectonic requirements, however, the glass
beams must have a span of more than 20 meters. For glass laminate beams
manufactured in this way and abutting along their longitudinal sides, some of the
scientific basics are not yet known.
This work thus deals with the determination of the load-bearing capacity of glass beams
of this type. Laboratory tests are planned for investigating the load-bearing behaviour of
736
Load-Bearing Behaviour of Non-monolithic Glass Laminate Beams for Wide Spans
these glass laminate beams. In a first phase, tests are conducted on so-called small
specimens. To that extent, it must be ensured that the failure states and sensitive
influencing factors occurring in long beams are encountered in the same form as in the
small specimens too. The maximum load-bearing capacity of the glass beams is
ascertained by increasing the load accordingly until the element fractures. The
subsequent evaluation of the measured data indicates the extent to which the values
measured at the test element match those in the supporting structure concept prepared in
the finite-element model.
2. Geometry and setups of test elements

The so-called short glass beams substantially consist of triple laminated safety glass.
The test specimens vary in the glass type (float, toughened safety glass, partially
tempered glass, chemically tempered glass), in the glass dimension and in the number of
glass joints.
Figure 3. Plan view of glass beam in a laminate
The geometric ratios of the short glass beams to their length and height were selected to
be consistent in comparison with the long beams. The spans between the supports are
for the short beams 3000 mm, and for the long ones 20000 mm. The entire beam length
is 3200 mm with a 100 mm projection on both sides at the supports, and 20200 mm for
the long beams. The height of the long beams is 1000 mm, that of the short beams is
accordingly 150 mm.
The glass beams are each designed with three glass planes with a laminate fim between
them. In the version with abutting sheets, the two outer ones are have a triple point and
the inner one is split centrally (see Picture 3). As a reference, glass beams with
continuous glass sheets are tested.
737
Challenging Glass 3
Figure 4: Schematic view of test arrangement
To prevent the beams from tipping over in the test stand, the beams are guided laterally.
Path control is achieved by vertical movement of the beams. No horizontal movement
takes place due to the lateral mounting. Failure of the beams thus occurs as a result of
glass fracture or failure of the film.
The beam is supported at its ends only in the axis directions x, y and z. Clamping for the
beam in the y and z axes is assured. The glass edges in the supports rest on a sliding
layer of reinforced teflon, thus permitting turning about the y axis with the least possible
friction.
Figure 5: Test setup similar to [1], lateral mounting is not shown.
The load is introduced at the one-third points. Since the test series also include glass
lengths split into thirds over the beam length, the load is introduced next to the vertical
butt joints, thus preventing any movement of the individual glass pieces relative to one
738
another. The load (deformation-based) is introduced only in the vertical direction. If

there is any lateral evasion by the glass beams, the pressure cylinder compensates.
Image 5 shows the test arrangement as for the dissertation by Luible [1]. Diverging
from the latter, the individual loads are introduced at the one-third points and not in the
centre of the field. In addition, beams with a thicker structure are to be tested here. With
the "short beams" up to three sheets are adjacent to one another. The "long beams"
made of laminated safety glass have a five-sheet structure.
3. Presentation of the computation model

Both the glass sheets and the film (SGP 5000) are modelled with volume elements. The
computations [2] were conducted with the FEM program ANSYS (Version 11.0). For
the volume elements of the glass and of the SGP 5000 film, 20-node SOLID95 elements
were used in computing the glass beams. Linear material behaviour was taken into
account for the buckling analysis, and non-linear material behaviour of the films for the
geometrically non-linear computation (with pre-deformation).
Image 6: Load introduction element in one-third point

Image 7: Support mounting with glass Figure 8: Detail of load introduction element
projection
739
Challenging Glass 3
4. Test method
The following presents the testing of a test specimen in a photo-documentation
Figure 9: Schematic view of the test setup

Figure 10: Test setup
740
Figure 11: Load introduction element Figure 12: Measuring pickup for bending
glass beam centre
Figure 13: Beam clamping, fixed point Figure 14: Beam clamping, non-fixed
mounting
Figure 15: Test specimen under load (with clearly visible bending)
741
Challenging Glass 3
Figure 16: Test specimen after glass fracture
At the time of writing the tests had not yet been concluded. A detailed evaluation of the
test results was therefore not available.
Here are the first results:

Image 17: Comparison of fractile - calculated value for laminate beams from TVG
742
Figure 18: Comparison of fractile - calculated value for laminate beams from Float
5. Summary
From the test results obtained, and taking into account the computed safety concept, a
practically-minded and economical dimensioning concept is devised, i.e. the study is
intended to show that abutting glass laminate beams can in future be used as load-
bearing glass laminates.
This new laminate technology enables a bendable beam of more than 20 meters in
length to be manufactured, thus ending the current limitation of the beam length to the
maximum glass dimensions with a maximum length of 9 meters permitted by
manufacture. Thanks to the results obtained from the research project, a completely new
method for building can now be introduced. This will give a marked boost to the
increased use of glass for future building projects. The use of non-monolithic glass
laminate beams represents an innovative and material-efficient connection technology
that will above all strengthen the competitive situation of companies engaged in glass
and facade construction.
6. References
[1] Luible, A.; Stabilität von Tragelementen aus Glas, Technische Universität München; Dissertation No.
3014, Lausanne, EPFL; 2004
[2] Dr. Burmeister, Delta-X, Stuttgart, Statische Bemessung von gestoßenen Glasverbundträgern, 2010
743
doi:10.3233/978-1-61499-061-1-745
Experimental Investigations on Continuous

Glass-GFRP Beams; Preliminary Non-
Linear Numerical Modelling
Luís Valarinho, João R. Correia, Fernando Branco
Instituto Superior Técnico/ICIST, Portugal, luis.valarinho@civil.ist.utl.pt
José Sena-Cruz
School of Engineering, University of Minho, Portugal, jsena@civil.uminho.pt
This paper describes results of experimental and numerical investigations about the
structural behaviour of composite beams made of annealed glass panes and GFRP
pultruded profiles. A brief description of flexural tests previously carried out on
simply supported glass and glass-GFRP composite beams is first presented. Then,
results of flexural tests on two-span glass-GFRP composite beams, bonded with
three different structural adhesives, are described in detail. Finally, a preliminary
numerical study of the glass-GFRP composite simply supported beams is presented.
In this study, two-dimensional finite element models were developed in order to
simulate and analyse the serviceability and post-cracking behaviour of those beams.
Experimental and numerical results presented in this paper prove the advantages
and technical viability of glass-GFRP composite beams.
Keywords: Glass-GFRP, Continuous beams, Numerical simulation, ductility
1. Introduction
Glass has played a central role on modern architecture since the 19th century, namely
due to its many aesthetical possibilities combined with its main feature: transparency.
Since then glass has had an important use on building façades. A few decades ago, glass
has also started to be used as a structural material and there are already several
examples of civil engineering applications in roofs, floors, beams and columns.
Structural elements made of float glass present several limitations, including relatively
low tensile strength and brittle behaviour, which contrasts with the current design
philosophies associated with more conventional materials, such as steel and reinforced
concrete, for which ductility of structural members must be guaranteed.
The traditional alternatives to overcome the above mentioned limitations of float glass
consist of using either toughened glass or laminated glass [1]. Toughened glass presents
higher tensile strength compared with float glass, however it still exhibits a fully brittle
behaviour at failure. On the contrary, laminated glass is capable of displaying a pseudo-
ductile and redundant behaviour – if one of its glass panes cracks or breaks, polyvinyl
butyral (PVB) films not only keep them in place but also transfer the tensile stresses to
the other panes.
745
Challenging Glass 3
More recently, a different approach has been pursued by several authors (e.g., [2-5]),
which consists of joining glass panes to other structural materials, namely stainless steel,
carbon fibre reinforced polymer (CFRP) laminates, glass fibre reinforced polymer
(GFRP) rods, concrete, wood and steel. The underlying principle of those composite
members is similar to that of reinforced concrete and relies on the stress transfer
between the glass pane and the strengthening material used when the tensile strength of
glass is attained.
This paper first describes the main results of an experimental programme about the
structural behaviour of composite beams made of annealed glass panes and glass fibre
reinforced polymer (GFRP) pultruded laminates. In a first stage of the experimental
campaign, flexural tests on simply supported glass and glass-GFRP composite beams
were carried out, in which the effects of the geometry of the GFRP strengthening
elements and the type of adhesive used to bond them to glass panes were investigated.
The main findings of these tests, already described in detail in [6], are briefly
summarized here. The second stage of the experimental campaign, whose results are
described in detail in this paper, included flexural tests on continuous two-span glass-
GFRP composite beams with an I-section made of a glass web and GFRP flanges. In
these tests, the serviceability (stiffness, cracking loads) and ultimate behaviour (failure
loads, crack pattern, failure modes, force redistribution and ductility) of the beams was
analysed and compared, allowing the evaluation of the potential advantages of the
proposed glass-GFRP structural system and structural adhesives in hyperstatic members.
The final part of this paper describes the numerical simulation of the simply supported
beams tested. In particular, two-dimensional finite element models were developed
using FEMIX software [7], in order to simulate and analyse the serviceability behaviour
of glass-GFRP composite beams (prior to glass breakage), as well as their post-cracking
behaviour. A multi-fixed smeared crack model, available in FEMIX computer program,
was used. For now, the numerical investigations focused only on the beams in which the
strengthening material was bonded to the glass beam with an epoxy adhesive. For these
beams, test results showed that the epoxy adhesive provides a high level of shear
interaction at the bonded interfaces – therefore, complete shear interaction was assumed
in the numerical models. Experimental and numerical results are compared in terms of
initial stiffness, cracking load and crack pattern.
2. Experimental programme
2.1. Test programme

The experimental programme included material characterization tests (to more
information about these tests see Valarinho [8]) and flexural tests on (i) simply
supported beams and (ii) continuous two-span beams.
2.2. Beam geometry, flexural test setup and procedure

The simply supported (SS) beams comprised the following three types of geometries:
(i) rectangular reference glass beams, with a cross section of 12 × 100 mm2, without
GFRP reinforcement (SS-S series); (ii) rectangular composite beams (SS-R series),
similar to the former but strengthened in the bottom edge with a GFRP pultruded
laminate with a cross section of 12 × 10 mm2; and (iii) beams with I geometry (SS-I
series), composed of the same glass panes strengthened in the top and bottom edges
746
Experimental Investigations on Continuous Glass-GFRP Beams; Preliminary Non-linear Numerical Modelling
with GFRP flanges (cross section of 76 × 10 mm2) and angles (cross section of
30 × 20 mm2 with a thickness of 4.8 mm) - Fig. 1. In both rectangular and I-section
simply supported beams, the GFRP profiles were adhesively bonded to the glass panes
with a 2 mm thick layer of two different types of adhesives: an epoxy structural
adhesive (EPa, Sikadur 330) and a high performance elastic gap-filling polyurethane
adhesive (PUa, Sikaflex 265). All simply supported beams, with a span of 1.50 m, were
tested in a symmetrically 4-point bending configuration with a load span of 0.50 m..
Unlike the I-beams (SS-I-EPa and SS-I-PUa), in both rectangular beams (SS-R-EPa and
SS-R-PUa), in order to prevent lateral deformation, four pairs of vertical metal guides
were symmetrically positioned throughout the span - the outer pairs were placed at the
support sections while the inner pairs were 0.725 m apart themselves - Fig. 2. All beams
were monotonically loaded until failure under load control, at approximate speeds of
27 N/s and 10 N/s for the glass beams and the composite beams, respectively.
Figure 1: Geometry and cross section of beams from Figure 2: Experimental setup of the flexural tests
series S, R, I and I2 (dimensions in mm). on simply supported beams [6] (beam I-PUa).
For the continuously supported (CS) two-span beams only an I-section was tested, with
a slightly different geometry than that used in the simply supported beams: the flange
width was reduced to 50 mm – Fig. 1. A total of six beams were produced with the
following three adhesives (2 beams of each type): (i) the polyurethane adhesive used in
the simply supported beams (CS-I2-PUa); (ii) an alternative structural epoxy adhesive
(Sikadur-31 cf, beams CS-I2-EPb); and (iii) an alternative polyurethane adhesive
(Sikaforce 7710_L100, beams CS-I2-PUb). All interfaces were bonded with a 2 mm

thickness layer of adhesive, except when PUb adhesive was applied – here, a 1 mm
thickness was used given its low viscosity. The continuously supported beams, with two
spans of L = 1.4 m, were tested in a 5-point bending configuration - Fig. 3. The load
was applied using a 200 kN hydraulic jack reacting against a steel loading frame. A
steel load distribution beam, placed between the jack and the tested beams, allowed
applying a symmetrical point load distanced from the central support of 0.56 m (0.4 L,
the configuration that ensures the maximum moment at the central support section and a
ratio of 1.53 between the maximum negative and positive bending moments). In order
to guarantee a symmetrical force distribution in both spans (in the linear stage), a steel
roller was placed between the distribution beam and the hydraulic jack. In addition, to
avoid any transverse loading, metal plates and spheres were placed between the
distribution beam and the top surface of the tested beams. The supports consisted of
cylindrical rollers, placed in-between metal plates. The central support was fully fixed,
while the lateral supports allowed longitudinal sliding. In order to correct possible
altimetry differences between supports, a thin layer of plaster was applied underneath
747
Challenging Glass 3
the supports, wherever needed. Support reactions and applied load were measured with
load cells placed respectively below the supports (capacity of 50 kN in the outer
supports and 100 kN in the central one; precision of 0.01 kN) and between the hydraulic
jack and the distribution beam (capacity of 100 kN; precision of 0.01 kN).
Displacements at the centre of each span were measured with displacement transducers
(25 mm stroke; precision of 0.01 mm). Axial strains were measured throughout the
depth of two cross-sections under negative and positive bending. All beams were
monotonically loaded until failure under load control at an approximate rate of 130 N/s.
Figure 3: Experimental setup of the flexural tests on continuously supported beams (beam CS-PUb-2).
2.3. Materials
The beams tested comprised the following three different types of materials, whose
properties are listed in Table 1: (i) 12 mm thick annealed glass panes, with edge
treatment; (ii) GFRP laminates, made of an isophtalic polyester matrix reinforced with
alternating layers of E-glass rovings and mats; and (iii) four different adhesives.
Table 1 – Mechanical properties, in tension, of the materials used on glass_GFRP beams (N.A. not available).
Material u [MPa] E [GPa] Source
Testing (NP EN 1288-1:2007
Glass 58.9 ± 12.6 80.6
and NP EN 1288-3:2007)
475.5 ±
GFRP 32.8 ± 0.9 Testing (ISO 527-1,4)
25.5
Epoxy Sikadur 330 (EPa) 22.5 ± 3.9 5.13 ± 0.11 Testing (ISO 527-1,4)
Polyurethane Sikaflex 265 (PUa) 3.4 (1.49 ± 0.22) × 10-3 Testing (ISO 527-1,4)
Epoxy Sikadur-31 cf (EPb) 18 to 24 5 Manufacturer
Polyurethane Sikaforce
13 N.A. Manufacturer
7710_L100 (PUb)
3. Results of flexural tests
3.1. Flexural tests on simply supported beams

Results of flexural tests on simply supported beams (described in detail in [6]), are
summarized in Fig. 4 (in Table 2 are presented the main results of I geometry simply
supported beams), in which the load-deflection behaviour of the beams tested is
illustrated. Results of this stage of the experimental programme, which included also
748
material characterization tests and tests on double lap joints between glass and GFRP
adherends, allowed drawing the following main conclusions:
x The flexural tests proved the advantages and technical viability of glass-GFRP
composite beams. In particular, it has been shown that it is possible to obtain
relatively safe and ductile failure mechanisms in glass panes, provided that these
are bonded to GFRP strengthening elements. In fact, after the development of the
first crack in the glass pane, all strengthened beams kept their integrity,
exhibiting a residual strength that varied with the type of adhesive and, as
expected, with the geometry of the strengthening element (Fig. 4). In general, the
load-deflection behaviour could be divided into two stages separated by the
appearance of the first visible crack: on the first stage the behaviour was linear,
given the mechanical characteristics of the main materials involved, while the
second stage comprised a progressive loss of stiffness due to the damage
progression on the glass pane, which ultimately led to the beam failure.
x In terms of post-cracking residual strength and ultimate load capacity, epoxy
bonded composite beams presented much better performance than their
polyurethane counterparts. For beams from series SS-R, even with a small
strengthening cross-section, after glass cracking beam SS-R-EP was still able
to fully recover the maximum load; in opposition, beam SS-R-PU did not
present any post-cracking residual strength. In beams with I geometry, the
strengthening cross-section increase augmented the post-cracking residual
strength and both adhesives were able to mobilize a considerable residual
strength (153% and 199% for beams SS-I-PU and SS-I-EP, respectively),
providing significant safety levels. The ultimate strength of composite beams
with epoxy adhesive was 1.37 and 3.95 times higher than that of beams with
polyurethane adhesive in series SS-R and SS-I, respectively.
x In what concerns ductility, the fragile behaviour observed in the annealed glass
beams was not repeated in none of the composite beams – these exhibited a
pseudo-ductile behaviour after initial cracking, which, similarly to strength,
varied with the strengthening geometry and, especially, with the type of
adhesive. As expected, beams with polyurethane adhesive presented much

higher ductility than beams with epoxy adhesive - the higher ductility of the
former beams stemmed not only from the distribution of stresses between the
two materials (also observed in the latter beams) but, essentially, from the high
deformation capacity and low stiffness of the polyurethane adhesive, which
caused significant slipping between the two materials. It should be mentioned,
however, that the achievement of higher ductility levels had a counterpart,
namely the lower values of initial stiffness, post-cracking strength and ultimate
load capacity.
x The different types of adhesives led to different cracking patterns: beams with
polyurethane adhesive had a cracking pattern characterized by few cracks, with
a considerable spacing between them; beams with epoxy adhesive had a much
more regular crack pattern (roughly similar to that exhibited by reinforced
concrete beams), with vertical bending cracks in the central part of the beam
and increasingly inclined shear cracks towards the supports along the shear
span (Fig. 5).
749
Challenging Glass 3
Figure 4: Experimental load vs. deflection curves Figure 5: Beam SS-R-EP: crack pattern in the brink of
from the simply supported beams [6]. collapse [6].
3.2. Flexural tests on continuous two-span beams

Figures 6 to 9 summarize the experimental results obtained for the continuous beams in
terms of the following parameters: (i) load vs. midspan deflection curves and strength
(Fig. 6); (ii) distribution and variation of reactions and bending moments, both as a
function of the applied load, in which the theoretical curves plotted were obtained from
elastic force analysis; (iii) moment redistribution; and (iv) composite action. The results
are presented separately for the different types of adhesives. The main results of the
experimental tests on continuous beams are summarized in Table 2.
Figure 6: Load vs. midspan displacement of CS beams (LD – left span; RD – right span), from left to right,
beams with EPb, PUb and PUa adhesives.
Figure 7: Load vs. distribution of reactions, load vs. variation of reactions and load vs. variation of bending
moments of CS beams (only beam I-EPb is plotted).
750
Figure 8: Load vs. redistribution of bending moments of beams, from left to right, CS-I-PUa-1 beam, CS-I-
Pub-2 beam and CS-I-EPb-1.
Figure 9: Axial strains vs. section depth for increasing total load [kN] of, from left to right, positive moment
section on CS-I-PUa-1, negative moment section on CS-I-PUb-1 beam and positive moment section on CS-I-
EPb-1CS beam (prior to glass cracking).
Figure 6 shows that the general load-deflection behaviour of the continuous beams was
similar to that observed in the simply supported beams. Accordingly, there are two
behavioural stages separated by the occurrence of the first visible crack. In the first
stage all beams exhibited a linear behaviour with a similar stiffness in both spans. As
expected, beams CS-I2-EPb exhibited the highest stiffness (14.5 kN/mm), followed by
beams CS-I2-PUb (13.0 kN/mm), with the lowest stiffness being registered in beams
CS-I2-PUa (5.05 kN/mm). The second stage was characterized by the propagation of
cracks and by the corresponding progressive loss of stiffness, resulting in a pseudo-
ductile behaviour.
With respect to the cracking load, as for stiffness, beams CS-I2-EPb and CS-I2-PUb
presented the best performance, with cracking loads being more than two times higher
than those of beams CS-I2-PUa. Although the average cracking load of CS-I2-EPb
beams was slightly higher than that of the CS-I2-PUb beams, one of the beams of the
latter series presented a higher cracking load than the average one registered on CS-I2-
EPb beams. When the first visible crack developed, the midspan deflection of all beams
was about 2.5 mm (L/585 of the span).
The crack pattern development was of two types: beams CS-I2-PUa exhibited few
cracks that had a continuous development during the test and were particularly
concentrated over the central support and on the loaded sections; on the remaining
beams, the glass pane displayed a more distributed crack pattern. Those distinct
behaviours can be attributed to the level of interaction at the bonded interfaces which, as
discussed in [6], is low for the PUa adhesive and high for adhesives with higher
stiffness, such as epoxy adhesives and Pub polyurethane. It is worth mentioning that all
beams first cracked above the central support with the exception of one of the beams of
CS-I2-EP2 series. In this beam the first crack appeared at the right midspan (most likely
751
Challenging Glass 3
due to material heterogeneity, as the bending moment in the support was higher); this
occurrence can be noticed in the load vs. deflection curve at the right span, in which a
premature loss of stiffness can be identified.
Figure 10: Cracking pattern of beams CS-I-PUa and CS-I-PUb-1
In the beams bonded with the PUb and EPb adhesives the failure modes were very
similar and were caused by the sudden and explosive disintegration of the glass web (in
most specimens, this only occurred in one of the spans – Fig. 11) after attaining a high
level of damage with extensive cracking in the glass web. One of the beams with the
PUa adhesive (CS-PUa-1) was unloaded without having collapsed (i.e., without web
disintegration) after a considerable lateral (out of plane) deformation became visible,
particularly in one of the loaded sections (Fig. 12). In the other beam of that series (CS-
PUa-2), the test was not interrupted when such out of plane deformation began and the
beam eventually failed due to a mechanism that involved lateral bending and crushing
of the glass web below one of the loaded sections.
Figure 11: Failure mode of beam CS-I-PUb-1. Fig. 12 – Deformation of CS-I-PUa beam prior to unloading.
Regarding the maximum load attained, it can be seen that the three types of beams
behaved differently. Again, beams CS-I2-EPb presented the best performance attaining
an ultimate load of 58.3 kN. The beams bonded with polyurethane adhesives presented
much lower strength, especially beams CS-I2-PUa, with a failure load that was almost
three times lower than that of beams bonded with epoxy. Beams CS-I2-PUb presented
an intermediate strength of 34.8 kN. Despite the marked difference in terms of ultimate
load between beams CS-I2-EPb and CS-I2-PUa, both types of beams presented a very
similar post-cracking strength (ratio between ultimate load and cracking load) of 183%,
indicating that a similar design philosophy can be used in those beams.
752
Unlike deflections at cracking, deflections at failure were very dissimilar for the
different types of adhesives. Before unloading, beams CS-I2-PUa exhibited a deflection
of 18.9 mm at the left midspan, a much higher deflection than that exhibited by beams
CS-I2-EPb and CS-I2-PUb (6.4 mm and 8.6 mm, respectively). Consequently, the
ductility index (defined as the ratio between the deflection at the first visible crack and
the deflection at failure), was much higher in beams CS-I2-PUa (almost 1000%) than in
the other beams (for the right span it was around 320% and 304%, respectively for
beams CS-I2-PUb and CS-I2-EPb).
Table 2 - Summary of results of flexural tests on simply supported beams with I geometry and on continuous
two-span composite glass-GFRP beams (average results are presented for series CS-I2-PUa and CS-I2-EPb).
SS-I- SS-I-
Beam series CS-I2-PUa CS-I2-PUb CS-I2-EPb
PUa EPa
Span - - Left Right Left Right Left Right
Initial stiffness (kN/mm) 1.74 4.55 5.05 5.01 13.5 12.7 14.7 14.1
Cracking load (kN) 5.09 15.50 11.4 30.6 32.1
Maximum load (kN) 7.80a 30.81 20.9 34.8 58.3
Post-cracking strength (%) 153 199 183 110 180
Deflection at first visible
3.00 3.53 2.27 2.34 2.31 2.42 2.21 2.30
crack (mm)
Deflection at failure (or
26.5b 14.9 18.89 22.97 6.42 7.75 8.56 7.01
before unloading) (mm)
Deflection in terms of span
500 425 616 599 605 579 632 608
at first visible crack
Deflection in terms of span
57 101 74 61 218 181 164 200
at failure
Ductility index (%) 883 426 831 984 275 302 380 303
a b
Did not correspond to beam failure Deflection at 80% of maximum load
The flexural tests on continuous composite beams also allowed analyzing the capacity
of force redistribution between the central support and the loaded sections. The
maximum bending moments at those sections and the corresponding maximum
redistribution capacities are summarized in Table 3.
All beams were able to redistribute internal forces, following the damage propagation in
their cross-sections. Yet, such capacity was different amongst the beams tested. It can
be seen that beams CS-I2-PUa presented by far the highest redistribution capacity in
line with their highest ductility index, compared with beams bonded with adhesives
PUb and EPb. This result is consistent with the differences in the mechanical properties
of PUa adhesive and the two other adhesives (PUb and EPb), and the influence of such
properties on the ultimate strength of the beams (and also on the maximum moment and
the redistribution capacity). Beams CS-I2-PUb and CS-I2-EPb, despite having similar
values of ductility index, showed considerably different redistribution capacities, with
beams CS-I2-PUb exhibiting higher capacity than beams CS-I2-EPb. Further studies
753
Challenging Glass 3
will be developed within this project (namely tests on adhesively bonded glass-GFRP
joints) in order to understand better the reasons for such differences.
Since the redistribution of moments is a consequence of the loss of stiffness on several

sections (due to the damage increase) and, in this case, is not due to the mechanical
behaviour of the materials (as in steel or reinforced concrete structures), the moment
redistribution from the central support to the spans was only momentarily observed,
most of it occurring after the appearance of the first crack. With the development of the
crack pattern and with the appearance of cracks in the spans, the beams had the
tendency to re-equilibrate the force distribution, approaching the original elastic one –
Fig. 8.
Table 3 - Results for failure behaviour of continuous two-span composite glass-GFRP beams.
Ultimate Maximum moment (kN.m) Ductility Maximum redistribution (%)

Beam
load (kN) Left Support Right indexa (-) Support Span
CS-I2-PUa-1 23.5 2.57 2.49 2.46 988% -75% 52%
CS-I2-PUa-2 18.3 2.07 2.18 2.34 810% -59% 15%
CS-I2-PUb-1 26.7 3.05 3.49 3.16 221% -45% 39%
CS-I2-PUb-2 42.9 5.00 4.27 4.91 356% -27% 28%
CS-I2-EPb-1 53.8 5.46 5.90 5.56 269% -7% 7%
CS-I2-EPb-2 62.7 6.30 7.68 6.31 414% -5% 9%
a
average from both spans
4. Numerical simulation
4.1. Initial considerations

Smeared crack models have been used for the simulation of concrete in tension since the
1970s. In these models, the fracture process is initiated when the maximum principal
stress in a material point exceeds its tensile strength. The propagation of the cracks is
mainly controlled by the shape of the tension-softening constitutive law and fracture
energy of the material. Normally, the mesh objectivity is guaranteed by associating the
dissipated energy in crack propagation process with a characteristic length of the finite
element. In order to avoid snap-back instability, the mode I fracture energy must be
greater than a threshold value which depends on the tension-softening constitutive law.
Typically, the fracture propagation in mode II is based on the concept of shear retention
factor [9].
The numerical investigations described in this section comprised a parametric study

carried out with the aim of evaluating the applicability of smeared crack models for the
simulation of annealed glass structural elements strengthened with GFRP using an
epoxy adhesive. For that purpose a multi-fixed smeared crack model [9] was selected
from the FEMIX computer code, which is a general tool for the analysis of structures by
the Finite Element Method [7]. The main analysed parameters were the fracture energy
and the shear retention factor.
754
4.2. Description of the FE model

The strengthened beam SS-R-EPa was modelled as a plane stress problem. Fig. 13
shows the geometry, mesh, support conditions and load configuration used to develop
the parametric study. To simulate the glass and GFRP, 4-node Serendipity plane stress
elements were used with 2×2 Gauss-Legendre integration scheme. Linear elastic
behaviour under compression was adopted. Perfect bond was assumed between both
materials. This assumption is corroborated by the experimental observations (c.f.
section 3). The shape of the tension-softening law was assumed as linear. The crack
band width was assumed equal to the square root of the area of the finite element in
order to assure that the results are not dependent on the mesh refinement. In the multi-
fixed smeared crack model used, for a specific integration point, a new crack is initiated
when the maximum principal stress exceeds the uniaxial tensile strength, and the angle
between the direction of the existing cracks and the direction of the maximum principal
stress exceeds the value of a predefined threshold angle. In the present study the
threshold angle was assumed constant and equal to 30q. A maximum of 2 cracks per
integration point was allowed to arise.
As referred before, the parametric study analysed the influence of the fracture energy
and the shear retention factor on the load vs. deflection at midspan relationship. The
numerical responses were compared with the experimental one. Additionally, in some
cases the crack patterns were also compared.
Figure 13: Mesh, support conditions and load configuration.

4.3. Results and discussion

For studying the effect of the mode I fracture energy (Gf) on the structural response of
the annealed glass beam strengthened with GFRP, the following values were
considered: Gf,min, 1.5Gf,min, 2.0Gf,min and 4.0Gf,min, where Gf,min is the minimum fracture
energy required to avoid the snap-back instability [9]. According to the literature, the
value of the glass fracture energy is about Gf,min/100 [1], although to the authors’ best
knowledge there is no experimental work reporting the determination of such value
(3×10-3 J/m2). It is also worth mentioning the considerable scatter of Gf reported in
other more conventional materials, namely concrete, for which differences of the same
order of magnitude have been reported by several authors [10]. In the simulations of the
present section the parameter p defining the shear retention factor was assumed to be
equal to 2.0.
Fig. 14 depicts the relationships between the load and midspan deflection responses,
both numerical and experimental. In this figure it can be seen that the simulation of the
elastic branch matches the experimental response. With the exception of model
“4.0Gf,min” all the numerical models predicted the crack load initiation. After this point a
755
Challenging Glass 3
sudden load decay is observed for model “Gf,min”. This load decay is similar to the one
observed in the experimental test. However, when the corresponding deflection is
compared a large difference can be observed. This difference can be attributed to the
fact that the data acquisition speed (1 Hz) was not fast enough to capture such drop in
the experimental test. After this phase several cracks arose and then grew in terms of
width and depth. At this stage, a similar response is observed for all the models (with
the exception of model “4.0Gf,min”), which predicted quite well the experimental
response including the failure load.
Figure 14: Effect of fracture energy on the load vs. mid-span deflection.
Fig. 15 presents the crack patterns obtained for different deflection levels of the models
“Gf,min” and “2.0Gf,min”. For all the stages analysed, the existing cracks are mainly “fully
opened” (in purple), i.e. cracks where the mode I fracture energy is fully exhausted. In
spite of model “2.0Gf,min” predicted a greater number of flexural cracks with higher
depth, the model “Gf,min” showed a better similarity with the experimental observations
in terms of crack pattern at the upper part of the strengthened beam. In addition, for both
models, the horizontal cracks developing on the shear span at the GFRP vicinity can be
perfectly identified in the experimental prototype.
Figure 15: Effect of the fracture energy on the crack pattern, for the models with Gf,min and 2.0Gf,min.
The nonlinear material model used allows the evaluation of the shear retention factor, E,
in two distinct ways [9]: (i) a constant value; (ii) a non-constant value defined by E= (1
– Hcr/Hcr,ult)p, where Hcr and Hcr,ult are the crack normal strain and the ultimate crack
normal strain, respectively, and p is a parameter that can assume the values of 1, 2 or 3.
756
Figs. 16 and 17 show the influence of the shear retention factor on the structural
response when the strategies (i) and (ii) are followed, respectively.
Figure 16: Effect of shear retention factor on the Figure 17: Effect of the parameter p on the load vs.
load vs. midspan deflection. midspan deflection.
In these simulations a linear tension-softening constitutive law was used and the fracture
energy was assumed equal to Gf,min. When a fixed value for E is assumed (see Fig. 16),
after crack initiation, the numerical models overestimated the experimental result. This
behaviour was expected since during the crack propagation the numerical shear
resistance degradation does not exist. When a non-constant value for the shear retention
factor is adopted (see Fig. 17), the numerical model predicts quite well the overall
response. Minimum differences were found for the cases of p=1, 2 and 3.
5. Conclusions
This paper presented results of experimental and numerical investigations on composite
structural beams that combine annealed glass panes and GFRP pultruded profiles, the
latter being used as strengthening elements and bonded to the former with different
types of adhesives. The following main conclusions are drawn:
x The main advantage of the composite beams proposed in this study is their
post-cracking residual strength and pseudo-ductility - the experimental tests on
simply supported and continuously supported beams attested such better

performance.
x The results obtained for the continuously supported beams were in line with
the ones reported earlier for the simply supported beams - as expected, for
similar adhesives, hyperstatic beams exhibited an increase of ultimate strength
and a reduction of deflections.
x Amongst the continuous beams tested, the ones bonded with the PUa adhesive
presented the highest values of ductility, much higher than those obtained for
the other two types of adhesives, which were very similar to each other. Beams
with PUa and EPb adhesives presented the highest post-cracking strengths,
considerably higher than those exhibited by beams PUb. Although presenting
similar values of post-cracking strength, as in the simply supported beams, the
higher levels of ductility in beams PUa were obtained at the expense of lower
values of initial stiffness, cracking and ultimate load.
x The redistribution capacities presented by the continuously supported beams
were associated to the loss of stiffness of the cracked sections and were
757
Challenging Glass 3
strongly dependent on the type of adhesive – as expected, the highest force

redistribution was obtained with the most deformable PUa adhesive. Due to the
symmetry of the structural system and especially due to glass brittleness, the
highest values of moment redistribution were only achieved momentarily, with
the beams rapidly tending to re-equilibrate the distribution of internal forces.
x A numerical parametric study was performed with a multi-fixed smeared crack
model that includes a linear tensile-softening law. The fracture energy and the
shear retention factor were the main parameters analysed.
x The model with the minimum fracture energy required to avoid the snap-back
instability, although being considerably higher than that referred in the
literature for glass, predicted with high accuracy the main aspects observed
experimentally, such as the crack initiation, stiffness degradation, load carrying
capacity and crack patterns.
x According to the studies performed, the shear retention factor cannot be
constant during the numerical test in order to include the shear degradation.
6. Acknowledgements
The authors wish to acknowledge FCT, ICIST and ADI (project n.º 3456/2009) for
funding the research and companies SIKA, Guardian, STEP and ALTO for having
supplied the adhesives, the glass panes and the GFRP pultruded profiles used in the
experiments, respectively. The first author also thanks FCT for scholarship nº
SFRH/BD/80234/2011.
7. References
[1] Haldimann M.; Luible A.; Overend M., Structural Use of Glass. Structural Engineering Documents 10,
IABSE, Zurich, 2008.
[2] Louter P.C., Adhesively bonded reinforced glass beams, Heron 2007; 52: 31-58.
[3] Louter C.; van de Graaf A.; Rots J., Modeling the Structural Response of Reinforced Glass Beams using
an SLA Scheme, Proceedings of Challenging Glass 2, Conference on Architectural and Structural
Applications of Glass (eds. Bos, Louter, Veer), Delft, The Netherlands, 2010.
[4] Ølgard A.B.; Nielsen J.H.; Olesen J.F.; Design of mechanically reinforced glass beams: modelling and
experiments, Structural Engineering International 2009; 19(2): 130-136.
[5] Louter C.; Leung C.; Kolstein H.; Vambersky J, Structural Glass Beams with Embedded Glass Fibre
Reinforcement, Proceedings of Challenging Glass 2, Conference on Architectural and Structural

Applications of Glass (eds. Bos, Louter, Veer), Delft, The Netherlands, 2010.
[6] Correia J.R.; Valarinho L.; Branco F.A., Ductility and post-cracking strength of glass beams
strengthened with GFRP pultruded composites, Composite Structures 2011; 93(9): 2299-2309.
[7] Sena-Cruz, J.M.; Barros, J.A.O.; Azevedo, A.F.M.; Ventura-Gouveia, A., Numerical simulation of the
nonlinear behavior of RC beams strengthened with NSM CFRP strips, Proceedings of the CMNE 2007 -
Congress on Numerical Methods in Engineering and XXVIII CILAMCE - Iberian Latin American
Congress on Computational Methods in Engineering, Abstract pp. 289, Paper nº 485 published in CD –
FEUP, 20 pp., Porto, 13-15 June 2007.
[8] Valarinho, L., Construction in structural glass: behaviour of glass–GFRP hybrid beams, MSc
Dissertation in Civil Engineering, Instituto Superior Técnico, Technical University of Lisbon, 2010.
[9] Sena-Cruz, J.M, Strengthening of concrete structures with near-surface mounted CFRP laminate strips,
PhD Thesis, Department of Civil Engineering, University of Minho, 2004, 198 pp. URI:
http://hdl.handle.net/1822/11781.
[10] Net, P.; Alfaiate, J.; Almeida, J.R.; Pires, E.B, The influence of mode II fracture on concrete
strengthened with CFRP, Computers and Structures 2004; 82(17-19): 1495-1502.
758
doi:10.3233/978-1-61499-061-1-759
Design of Glass-Polycarbonate
Composite Panels
Thorsten Weimar
University of Siegen, Institute of Building Structures, Germany,
weimar@architektur.uni-siegen.de
Glass-polycarbonate composite panels are mainly used for security glazing

applications. Glass possesses a high material stiffness whereas polycarbonate
provides high impact strength in combination with low weight. The composite
panels combine high resistance against attack with slender cross-sections and
reduced dead load in comparison to common laminated safety glass. In
consequence, the structural design of glass-polycarbonate composite panels
depends mainly on static loads. This requires a profound knowledge of the material
properties and the interaction of multiple layers in composite structures. In addition
to the structural behaviour of the glazing, the post-breakage performance is also
assessed. The results provide the basis for a suitable design concept of glass-
polycarbonate composite panels.
Keywords: Glass, Polycarbonate, Polyurethane, Composite Panels
1. Introduction
The systematic combination of different materials forming a composite element enables
the engineer to adjust and optimise the desired structural properties for a specific
application, which a single building material might not be able to cope with. When
using glass in building construction, its combination with more ductile materials
effectively contributes to broaden the range of applications beyond the limitations of the
base materials involved. The material properties of security glazing units and especially
their resistance against external attack can be significantly improved by using hybrid
elements made of glass and polycarbonate, ‘Weller & Weimar [8]’.
Due to the fact that this build-up is about 33 % more slender and 50 % lighter compared
to common laminated safety glass, the composite panel enables the use of more slender
frame and support structures. The supplementary installation in existing buildings and
the subsequent processing to insulation glass is simple. Security glazing made of glass-
polycarbonate composite panels consists of two glass panes and one or more
polycarbonate (PC) panels. The single layers are laminated in a semi-automated method
using specially developed transparent polyurethane (PUR). Costs depend on the
materials used and the manufacturing process and range approximately 45 % higher
than for conventional special security glazing.
759
Challenging Glass 3
Figure 1: Glass-polycarbonate composite panel Figure 2: Security glazing with resistance against
manual attack with the highest standard according to
‘EN 356 [3]’. Laminated glass (left) and glass-
polycarbonate composite panel (right).
Security glazing with resistance against manual attack is proved with test-methods in
accordance with ‘EN 356 [3]’ to define minimum cross-sections of the required
resistance class. These cross-sections do not depend on size or intended application as
vertical or overhead security glazing. Table 1 shows typical minimum cross-sections of
glass-polycarbonate composite panels used as security glazing with resistance against
manual attack for the three highest resistance classes.
Table 1: Minimum cross-sections of glass-polycarbonate composite panels

used as security glazing with resistance against manual attack
Resistance
Minimum cross-section
class
Annealed Glass / PUR / PC / PUR / Annealed Glass

P6B
4 mm / 2 mm / 4 mm / 2 mm / 4 mm
Laminated Safety Glass / PUR / PC / PUR / Laminated Safety Glass
P7B
2 x 3 mm / 2 mm / 6 mm / 2 mm / 2 x 3 mm
Annealed Glass / PUR / PC / PUR / PC / PUR / Annealed Glass
P8B
4 mm / 2 mm / 5 mm / 2 mm / 5 mm / 2 mm / 4 mm
Number and thickness of the single layers can be modified to adapt the experimental
proven minimum cross-section to the static loads. Security glazing made of laminated
glass generally does not require a structural design due to their relatively large thickness.
The slender glass-polycarbonate composite panels, however, can require additional
dimensioning. Considering the resulting loads in dynamic and static part separately
leads to an economic and efficient design of the materials used.
760
Design of Glass-Polycarbonate Composite Panels
2. Material Properties
The structural design of glass-polycarbonate composite panels requires a profound
knowledge of the material properties and their interaction in the multiple layers
structure. Therefore, the physical and mechanical properties of the materials used –
glass and polycarbonate as well as the polyurethane interlayer – were defined. The two
polymeric materials were analysed in detail, particularly with regard to the temperature
dependence and load condition. Physical and mechanical properties of glass,
polycarbonate and polyurethane are given in table 2.
Table 2: Material properties of glass, polycarbonate and polyurethane

Property Glass Polycarbonate Polyurethane
Density 2,500 kg/m3 1,200 kg/m3 1,035 kg/m3
Thermal expansion
9 · 10-6 K-1 70 · 10-6 K-1 224 · 10-6 K-1
coefficient
Elasticity modulus 70,000 N/mm2 2,300 N/mm2 0.76 N/mm2
Shear modulus 28,500 N/mm2 833 N/mm2 0.27 N/mm2
Poisson’s ratio 0.23 0.38 0.40
Ultimate stress 45 N/mm2 70 N/mm2 0.39 N/mm2
Ultimate strain ~ 0.1 % 120 % 110 %
The composite panels comprise soda lime silicate glass, which is the commonly used
glass in building. Material properties confirm steady behaviour with regard to the
relevant conditions in building and are specified in ‘EN 572-1 [4]’.
Physical and mechanical properties of both polymeric materials are not sufficiently
available. The datasheet of the used polycarbonate provides basic material properties
given in ‘Lexan [5]’. The influence of the parameters temperature, environmental
conditions, load duration and load rate were derived from additional tensile tests.
Subsequently, the analysed data resulted in specific factors to consider the influences
examined. These influence factors can be combined to a modification coefficient, which

reflects the application conditions on site.
Polyurethane was classified as a weakly cross-linked, thermoplastic elastomer. In

context of this research, the physical and mechanical properties of the polyurethane
interlayer were fully determined by means of tensile and shear tests. The characteristic
values examined consider the influence of temperature, environmental conditions, load
duration and load rate. The resulting influence factors permit the adaption of
characteristic values using a modification coefficient with regard to the intended
application.
The modification coefficient Amod is calculated in equation 1 multiplying four influence

factors - A1 for temperature, A2 for load rate, A3 for load duration and A4 for
environmental conditions.
Amod A1 A2 A3 A4 [ ] (1)
761
Challenging Glass 3
In accordance with ‘BÜV Kunststoffe [1]’ the design values for stress and deflection
calculations can be determined reducing the characteristic value by the modification
coefficient. The design value of resistance to the actions as a function of life time cycle
is expressed by equation 2.
Rk
Rd (2)
J M Amod
The results are described in detail in ‘Weimar [7]’ and provide the basis for a suitable
design concept of composite panels.
3. Structural Behaviour and Post-breakage Behaviour

The structural behaviour of glass-polycarbonate composite panels was investigated in
detail by means of four-point bending tests comprising short and long-term load
durations. The analysis also covered the interaction of the composite’s multiple layers.
The specimens examined consisted of one inner polycarbonate sheet and two outer
panes of annealed glass and laminated by 2 mm thick polyurethane interlayers. The
thickness of the single panels, given in table 3, varied, but the total thickness remained
22 mm. All build-ups possessed a length of 1,100 mm and a width of 360 mm. Each 15
specimens were tested for the three design studies.
Table 3: Build-up of different specimens tested in four-point bending test

Specimen
Cross-section
type
1
8 mm / 2 mm / 2 mm / 2 mm / 8 mm
2
6 mm / 2 mm / 6 mm / 2 mm / 6 mm
3
4 mm / 2 mm / 10 mm / 2 mm / 4 mm
The glass panes and their coherence influence the structural behaviour of the composite
panels. The results provide elementary knowledge about the load bearing capacity and
composite action of the polyurethane interlayer. Table 4 shows the shear modulus of the
different types of specimens determined after load durations of 10 s, 24 h and 1,000 h.
Table 4: Shear modulus after different load durations measured by four-point bending tests
Shear Shear Ratio of Shear Ratio of
Specimen
Modulus Modulus Shear Moduli Modulus Shear Moduli
type
Gt=10s Gt=24h Gt=24h / Gt=10s Gt=1,000h Gt=1,000h / Gt=10s
1 0.34 N/mm2 0.28 N/mm2 0.82 0.25 N/mm2 0.74
2 0.33 N/mm2 0.28 N/mm2 0.85 0.25 N/mm2 0.76
3 0.31 N/mm2 0.28 N/mm2 0.90 0.22 N/mm2 0.71
762
The shear modulus of the polyurethane, measured in the tensile tests, also depends on
load duration, which is described in equation 3 with influence factor A3(t).
A3 (t ) 1.104 0.034 ln (t) (3)
where
t Load duration [h]
Equation 4 expresses the decreasing of the shear modulus as a function of time.
G0
G (t ) (4)
A3 (t )
For example, load duration of 24 h reduces the initial value of shear modulus to about
83 % and after 1,000 h under load to about 75 %. The values measured in tensile tests
agree well with the results of the shear modulus determined by four-point bending tests.
The tests show a partial transfer of shear forces in the interlayer and the results match
well with the measured shear modulus of the polyurethane. The interlayer is able to
transfer shear forces for short and medium load durations.
In addition to the structural behaviour of the glazing, the post-breakage performance

was also assessed. The tests were repeated to analyse for the post-breakage behaviour of
the composite panels with broken glass panes. Figure 3 shows the deflection of a
specimen with broken glass panes and a 2 mm thick polycarbonate sheet. Figure 4
illustrates the residual strength of a specimen with a 10 mm thick polycarbonate sheet.
Figure 3: Deflection of a specimen type 1with Figure 4: Deflection of a specimen type 3 with
broken glass panes and a 2 mm thick broken glass panes and a 10 mm thick
polycarbonate sheet. polycarbonate sheet.
The post-breakage behaviour of the glass-polycarbonate composite panels demonstrates

sufficient and inherent redundancy. The interaction between the polycarbonate sheet
763
Challenging Glass 3
and the glass fragments in the compression zone influence essentially the post-breakage
performance. The behaviour does not depend on the fracture pattern of the panes used.
Figure 5 shows the ratio of bending stiffness in the unbroken state to the state with
upper or both glass panes broken.
Figure 5: Ratio n1 of bending stiffness in the unbroken state to the state with upper glass pane broken and
ratio n2 of bending stiffness in the unbroken state to the state with both glass panes broken in according
to the thickness of polycarbonate sheet.
The ratio of bending stiffness in the unbroken state to the state with upper glass pane
broken depends on the thickness of polycarbonate sheet. Values range from 0.82 with
2 mm thick polycarbonate sheets to 0.94 with 10 mm thickness. This behaviour can also
be found for the ratio of bending stiffness in the unbroken state to the state with both
glass panes broken. The specimens with broken glass panes reach values up to 0.47 with
10 mm thick polycarbonate sheets. For comparison, common laminated glass with
annealed or toughened glass loses its post-breakage performance in a static system of
two points.
4. Calculation Method
As part of the investigation on hybrid elements, experimental testing was carried out to
examine the composite panel’s load capacity and residual strength. The test results
provided the basis for the development of an analytical design approach of composite
panels made of glass and polycarbonate. By using the sandwich theory expanded to
three layers the composite panels were calculated analytically in ‘Weimar [7]’.
Generally, the total bending stiffness Bt of composite panels comprise the bending
stiffness Bi of the single layers and Bs of the sandwich themselves. Equation 5 applies to
symmetrical cross-sections, which are determined in the four-point bending tests. Ba and
Bb mean the bending stiffness of the two outer glass panes and Bm of the inner
polycarbonate sheet.
764
Bt Ba Bm Bb Bs
E b Em b (5)
2 d3 d m3 2 E b d a 2
12 (1 Q )
2
12 (1 Q m2 )
where
E Elasticity Modulus [N/mm2]
Q Poisson’s Ratio [-]
a Distance of centre lines [mm]
b Width [mm]
d Thickness of layers [mm]
The stiffness of the polyurethane interlayer can be calculated using equation 6.
b a2
S G (6)
t
where
t Thickness of interlayers [mm]
In accordance with ‘Stamm [6]’, the maximum deflection of the composite panel as well
as the maximum stress of the single layers were calculated for the specimens tested in
the four-point bending test. Subsequently, the calculations were repeated numerically
using two different FEA-programs Ansys 11.0 and SJ Mepla 3.5. Table 5 compares the
analytically and numerically calculated values with the measured results to determine
the quality of the calculation methods.
Table 5: Deviation ratio of calculated values from the measured results

Stress Stress Stress
Method Deflection
Glass pane above PC sheet Glass pane below
Sandwich Theory 3.8 % 6.0 % +33.3 % +6.0 %
Ansys 11.0 5.3 % 6.7 % +18.2 % +11.0 %

SJ Mepla 3.5 5.7 % 7.0 % +18.2 % +10.5 %
The results of the approach match well with the measured values of deflection and stress
in the glass panes. Due to the small stress values measured in the polycarbonate in a
range of 0.1 N/mm2 to 0.9 N/mm2, the deviation ratio is significantly higher. However,
the prepared numerical models realistically reproduce the composite panels and can be
used to calculate any cross-section. Geometrically nonlinear analysis should be used
under large deformations.
5. Design References
Finally, design references for a semi-probabilistic safety concept are defined applying
with the current state of the art. ‘DIN 18008-1 [2]’ covers a design-method for glazing
whereas in ‘BÜV Kunststoffe [1]’ a design concept for constructions with plastics is
discussed. Both design-methods presented base on the semi-probabilistic safety concept.
765
Challenging Glass 3
The design loads for the ultimate and the serviceability limit state are determined by the
same combination of actions. Thus, the resulting stresses and deformations of the glass
panes and polycarbonate sheets can be calculated with the sandwich theory.
The changing mechanical properties of polycarbonate and polyurethane as a function of

time also influence the total mechanical behaviour of the composite panel. Therefore,
the stresses and deformations of the sandwich are calculated with modified material
values of the polycarbonate and the polyurethane.
Subsequently, the ultimate limit state is proved by comparing the maximum stress
calculated for the design load and the allowable stress separately for the glass panes and
the polycarbonate sheets. Deformation can be limited by the relevant serviceability
criterion for the composite panel, which is defined to 1/100 of the decisive span.
The different thermal expansion coefficients of glass and polycarbonate cause to

additional strains in the polyurethane interlayers under the effect of temperature. These
strains are to be limited by meeting a design value for the serviceability limit state.
Equation 7 specifies the relevant serviceability criterion based on the ultimate strain of
the polyurethane interlayer.
H B, k
Cd (7)
J M Amod

where
HB,k Characteristic value of ultimate strain [-]
JM Material partial factor [-]

Amod Modification coefficient [-]
The partial shear transfer of the polyurethane interlayer shown in the four-point bending
tests may be considered for a short to medium load duration. The accessible shear
modulus can be calculated with equation 4. For long or constant load durations no shear
resistance are taken in account.
The use of glass in building generally requires the proof of a sufficient post-breakage
behaviour for applications with load transfer or to prevent people from injury. The post-
breakage performance of the glass-polycarbonate composite panels demonstrates an
inherent redundancy and can be ensured with the following constructive specifications.
x Glass-polycarbonate composite panels should be supported on four sides with a

bearing width of at least 10 mm.
x The post-breakage behaviour does not depend on the fracture pattern of the
glass type used. Therefore, all glass types – annealed glass, heat-strengthened
glass and toughened safety glass – can be used.
x Due to the composite between the polyurethane interlayer and the glass panes,
a sufficient protection from fractured glass pieces is provided.
x Four-sided supported glass-polycarbonate composite panels do not require an
experimental analysis of the post-breakage performance.
766
6. Conclusion
The proof of glass-polycarbonate composite panels used as security glazing with
resistance against manual attack can be considered separately for dynamic and static
loads. Minimum cross-sections are defined for the resistance against manual attack
based on experimental tests. Number and thickness of the single layers can be modified
to adapt the experimental proven minimum cross-section to the static loads. The
structural design of glass-polycarbonate composite panels requires a profound
knowledge of the material properties and the interaction of multiple layers in the
composite construction. Therefore, the mechanical properties of the polycarbonate
sheets and the polyurethane interlayers derived from testing comprise the influence of
several parameters. The resulting influence factors are combined to a modification
coefficient, which reflects the application conditions on site.
The structural behaviour of glass-polycarbonate composite panels was investigated in

detail by means of four-point bending tests. As a result of the tests, the polyurethane
interlayer is able to transfer shear forces for short and medium load durations. The tests
were repeated to account for the post-breakage behaviour of the composite panels with
broken panes. Glass-polycarbonate composite panels demonstrate sufficient post-
breakage behaviour and inherent redundancy. By using the sandwich theory expanded
to three layers as well as two different FEA-programs the composite panels were
calculated and compared with the measured values to determine the quality of the
calculation methods. However, the methods realistically reproduce the composite panels
and can be used to calculate any cross-section. Based on the semi-probabilistic safety
concept references are defined for the design of glass-polycarbonate composite panels.
The results presented offer a material-efficient design of glass-polycarbonate composite
panels applying with the current state of the art.
7. Acknowledgements
The author would like to thank the Federal Ministry of Economics and Technology, the
Arbeitsgemeinschaft industrieller Forschungsvereinigungen “Otto von Guericke” e.V.
for supporting the research project. Special thanks go to SILATEC Sicherheits- und
Laminatglastechnik in Gelting, Germany for sponsoring the test specimens. The tests
were carried out by the author as scientific researcher at the Institute of Building
Construction, Technische Universität Dresden.
8. References
[1] BÜV-Empfehlung. Tragende Kunststoffbauteile im Bauwesen. Entwurf, Bemessung und Konstruktion.
Berlin, 2010.
[2] DIN 18008-1: Glass in Building - Design and construction rules - Part 1: Terms and general bases.
Berlin: Beuth, 2010.
[3] EN 356: Glass in building – Security glazing - Testing and classification of resistance against manual
attack. Berlin: Beuth, 2000.
[4] EN 572-1: Glass in building. Basic soda lime silicate glass products. Part 1: Definitions and general
physical and mechanical properties. Berlin: Beuth, 2004.
[5] Lexan 9030: Datasheet. Sabic Innovative Plastics, 2011.
[6] Stamm, Klaus; Witte, Horst: Sandwichkonstruktionen. Berechnung, Fertigung, Ausführung. Wien, New
York: Springer, 1974.
[7] Weimar, Thorsten: Research on Glass-Polycarbonate Composite Panels. PhD. Technische Universität
Dresden, 2011.
[8] Weller, Bernhard; Weimar, Thorsten: Sicherheitsverglasungen als Hybridbauteil. Stahlbau 77 (2008).
Issue 11. Pages 802-808.
767
doi:10.3233/978-1-61499-061-1-769
Further Research About the

Short and Long-Term Breakage Behaviour
of Hybrid Glass-Steel Elements
Bernhard Weller, Philipp Krampe
Technische Universität Dresden, Institute of Building Construction, Germany,
www.bauko.bau.tu-dresden.de
Stefanie Retsch
Formerly Technische Universität Dresden, Institute of Building Construction, Germany
Laminated glass-steel-beams bonded together with a transparent acrylate adhesive

can improve the post-breakage performance. One result of the short-term tests is the
higher capacity of the hybrid beams compared to conventional laminated glass.
Another conclusion is the post-breakage behaviour of the beams, which highly
depends on the thermal pre-stressing used glazing. Annealed glass beams are able
to bear higher loads after failure while fully tempered glass shows almost no post-
breakage behaviour. Much more surprising are the results of the long-term research
investigation. Several hours after the breaking of the glass the deformation of the
bonding joint was observed. This behaviour led to independent deformations of the
single materials glass and steel of the hybrid beam. Their deformation curve is
directly influenced by the load bearing capacity of the component materials. The
compensating function of the acrylate adhesion will be given in the paper.
Keywords: Glass beams, hybrid, breakage behaviour, long-term study
1. Introduction
Load bearing elements of common mullion-transom structures as façade system are
usually made of opaque materials such as steel or aluminium. Typically, glazing

elements are supported and mechanically fixed horizontally by means of pressure caps.
The trend in modern architecture towards higher transparency in facades cannot be met
satisfactory by conventional systems in all cases. Therefore, glass will not be used
merely as an enclosing element or infill panel but increasingly rather with load bearing
and bracing function. Examples are beams, fins, columns or even entire frames. [1] - [4]
The development of hybrid elements and its concept is described in former articles [5]
including the presentation of results of small sample tests and short-time loading tests.
The general technical feasibility and the acceptance of such elements by planning
architectures are shown as well.
2. Motivation and Idea

The motivation is based on the conception of hybrid beams and fins for load bearing
purposes and their integration into a building structure [5] - [7]. The fact of the lower
tensile strength of glass compared to steel will be faced by joining steel elements at the
769
Challenging Glass 3
edges of the glass beam. The load transferring connection will be achieved by an
acrylate adhesive [8].
The linear bonding enables the continuous load transfer between the steel elements and
the glass and avoids local stress concentrations. Therefore, the governing requirements
for the adhesive are a high strength for load transfer along with a sufficient elasticity for
the compensation of temperature strain. Such temperature influence was mitigated and
excluded by the constriction of the hybrid beams to interior conditions. Further
mechanical or durability problems caused by the temperature dependent properties of
the adhesive are avoided as well.
The behaviour of the joint was tested with small samples and short term loading tests on
entire beams previously [9]. The load bearing capacity and the post breakage behaviour
were improved significantly. But also the creeping of the adhesive requires also long
term loading tests in order to determine the time dependent stress distribution over the
cross section of the element. Due to the flexibility of the joint the stresses in the steel
element will be redistributed onto the glass. An overstressing of the glass panes caused
by such redistributions must be avoided. Therefore, knowledge about the redistributing
effects is indispensible to design the section for the situation of finished creeping (t = )
[10].
Apart from the technical problems to be solved and the unknown mechanical interaction
of glass, steel and adhesive the usage of such hybrid beams is still a special construction
which requires special approvals according to German building law. In order to increase
the acceptance of such elements comprehensive testing and research is necessary.
Especially the utilisation of structural adhesives – apart from silicones for SSG-
applications – causes some scepticism.
Therefore, the focus of this research was put on the mechanical behaviour of the hybrid
beams under short term and long term loading. Varying parameters were the type of
glass (annealed or fully tempered glass) in order to determine the bearing capacity and
the post breakage behaviour respectively and the shape and application of the steel
element (figure 1). Several scientific investigations show that the post breakage
behaviour of glass beams consisting of conventional non-reinforced laminated safety
glass does not meet the required safety standards, independently of the type of glass.
Usually the beam is overdesigned in a manner that the bearing capacity can still be
guaranteed by remaining intact case, even if several panes are broken.
3. Specimens for Load Bearing Tests

The hybrid beams are made of laminated safety glass with annealed and fully tempered
glass as per codes and standards [1], [12]. The specimens for the short term test
consisted of three panes, the ones for the long term tests of two. The thickness of one
pane was in each case 6 mm. Additionally, stainless steel elements of different shapes
were applied on both edges [9]. The optimal thickness of the adhesive joint of 2 mm
was taken from other publications dealing with the same subject, though mostly
investigated with small sample tests ([9], [13]).
770
Further Research About the Short and Long-Term Breakage Behaviour of Hybrid Glass-Steel Elements
Figure 1: investigated cross sections – short term test. Same build-up for long term test but only two panes
The dimensions of the laminated safety glass beams are 1,100 mm in length and 100
mm in height. The hybrid beams are higher due to the application of the steel elements,
while the height of the glass remains the same. The hybrid beam with section S1 shows
the easiest way of applying a flat steel element, while S3 requires higher diligence in
producing the laminated glass. S0 is the reference section to compare the behaviour of
the hybrid beams to conventional glass beams.
4. Testing Procedure
4.1. Test Set-up

The hybrid beams are loaded about the strong axis due to their specific arrangement as
part of the façade. For experimental testing of the load bearing capacity and the post
breakage behaviour a modified four-point bending test appears purposefully [13]. The
original set-up allows a span of 1,000 mm. The hinged vertical support conditions are
the same as for single span beams. Fork bearings at the vertical support and close to the
loading points prevent rotating about the centre line and lateral movement of the
specimens in order to exclude stability effects (figure 2).
The load is applied at the two third points of the beams. This allows for a constant
tensile stress along the lower edge over a quite reasonable length.
The load was applied either by means of a hydraulic jack in case of annealed glass or by
a manually driven cylinder in case of fully tempered glass.
771
Challenging Glass 3
Figure 2: Experimental set-up of modified four point bending test
4.2. Short Term Tests

The tests with the specimens made of annealed glass were already performed in
2009/2010. The results were published in [15].
Altogether 80 specimens were tested: 10 beams of each section as per figure 1 made of
both annealed and fully tempered glass. The load was applied force controlled with a
speed of 10 N/s. This corresponds to a stress increase of 0.55 N/(mm2·s) related to
section S0 without any steel reinforcement. The deflections were measured at the
middle of the beam and at the supports until the first crack occur. Due to the zero-
setting of the measurement equipment influences caused by the dead load of the beams
could be avoided. Additionally, 5 specimens had strain gauges at the lower.
The force was applied until the first crack appeared. After a short stop the broken beam
was loaded again until the destruction of all plies of the laminated safety glass. For the
evaluation of the post breakage behaviour the load was applied under the fully damaged
condition until a stress or force drop respectively of 10 %.
Additionally to the already presented results the series of short term tests were
completed with tests on fully tempered glass. The following figures show the results of
the four-point-bending test (short term) for both annealed and fully tempered glass.
772
Figure 3: Breakage load of the several sections for annealed and fully tempered glass
As assumed, the results of the breakage load show that the reference section S0
possesses a lower capacity than the hybrid beams with steel reinforcement. The
maximal forces are approximately three times higher for the beams made of fully
tempered glass compared to the ones consisting of annealed glass. This is obvious due
to the higher strength of fully tempered glass, but it is visible that the factor between the
capacities is the highest for section S0; the relative improvement is lower for the hybrid
beams. However, the ratio between the stress in the steel element and the glass is
independent of the glass type used. Therefore, a linear relationship between the
maximum bearing capacities should be assumed. Consequently, the comparison of the
breakage loads between section S0 and S1 shall result in same ratio for both annealed
and fully tempered glass. This is not the case, so that other mechanical effects must be
active as well. However, a significant slip or creeping of the joint was not detectable.
The loading time was too short to generate a time dependent behaviour.
The relative improvement of the capacity is at maximum doubled for annealed glass and
approximately 35 % higher for fully tempered glass. Therefore, such an application of
steel elements to create hybrid beams possesses a higher potential the lower the glass
strength is. While the bearing capacity is almost doubled for annealed glass comparing
section S0 and S1/S3 the deflections are not differing with the same factor. But the
results of figure 4 are not representable because the higher breakage load generates also
higher deflections. Therefore, no conclusion regarding the stiffness can be made only
considering the deflection values. Otherwise, section S3 would be the weakest which is
apparently not the case.
773
Challenging Glass 3
Contrary, the hybrid beams consisting of fully tempered glass becomes stiffer from
section S0 to S3 (figure 1 and figure 4). Considering only the steel parts, this behaviour
is obvious because the inertia moment increases from a single plate with bending about
the weak axis (S1), over two plates with bending about the strong axis (S2) to a T-
section (S3). Generally, the bearing capacity should be increased the same way, but as
figure 3 shows, section S1 possesses the highest capacity. This leads to the conclusion
that section S2 and S3 collapse due to other breakage mechanisms than S1.
Delamination or failure of the adhesive joint are quite likely scenarios.
Figure 4: Deflection of the hybrid beams at the point of first crack

Table 1 shows the different measured values of the deflections. The variation coefficient
is relatively high. This is owing to the fact that the beams are hand-made and the
combined materials glass and adhesive possess inherently strongly deviating parameters
in terms of the strength and elasticity (adhesive).
Table 1: Comparison of the deflection

Annealed Glass Fully Tempered Glass
Standard Variation Standard Variation
Section Deflection Deflection
Deviation Coefficient Deviation Coefficient
[mm] [-] [%] [mm] [-] [%]
S0 2,36 0,31 13 7,74 1,13 15
S1 2,99 0,83 28 5,55 1,09 20
S2 2,78 0,46 17 4,69 0,94 20
S3 3,38 0,36 11 4,30 0,30 7
774
The following conclusions can be drawn from the short term tests:
x fully tempered glass possesses a three times higher bearing capacity until the
first crack which is reasonable due to the higher glass strength;
x at the point of failure all fully tempered glass plies are breaking almost
simultaneously. Using annealed glass, the load can be increased after the first
pane is broken. There is no abrupt complete failure of the glass;
For further investigations pertaining to the long term behaviour the tests are limited to
section S1. The reason is the easiest application of the steel element using the lowest
amount of adhesive.
4.3. Long Term Test

The interaction between adhesive, steel elements and glass under permanent loading is
not entirely known. The evaluation of the short term tests shows that the mechanical
combination of steel and glass leads to a partly notable increase of the bearing capacity
and stiffness [10]. While the mechanical behaviour under short terms loading is
calculable using an approach similar to flexible combined elements the influence of
creeping and other time dependent effects on the mechanical behaviour and stress
distribution is unknown. A mere material test does not allow a statement about creeping
properties because the influencing boundary layers are omitted. Therefore, this
behaviour was investigated in long term tests using the same principle set-up as for the
short term tests.
The load was applied by concrete blocks with a known weight (figure 5). Deviating to
the short term tests additional displacement transducers were horizontally fixed at the
end of the upper steel plate and the glass close to the upper edge so that the slip in the
joint could be recorded. Technically it was not possible to measure the horizontal
differential movement of steel and glass directly. It is assumed that due to the time
dependent and elastic mechanical behaviour of the adhesive joint a horizontal slip will
occur between steel and glass. Hence, possible time dependent shear deflections are
investigated.
775
Challenging Glass 3
Figure 5: Testing set-up for long term tests
Due to the time consuming characteristic of long term tests a reduced testing program
was developed leaving three beams: two made of annealed glass and one of fully
tempered glass. Therefore, resilient, statistically based results cannot be obtained.
Table 2: Testing program long term tests

Annealed F2_01 1 Annealed F2_02 1 Fully Tempered E2_01 1
Laminated safety glass with 2x 6 mm glass and 0.76 mm PVB,
Build-up height 100 mm, length 1,100 mm
2 steel plates, 6 mm width, thickness 2 mm, acrylate joint thickness 1.8 mm
Load 123 kg 102 kg 143 kg
Allowable
12 N/mm2 12 N/mm2 50 N/mm2
Stress
Measured 16 N/mm2 (Glass) 10 - 12 N/mm2 (Glass) 20 N/mm2 (Glass)
Stress 53 N/mm2 (Steel) 43 N/mm2 (Steel) 54 N/mm2 (Steel)
After 110 h breakage Breakage as scheduled
of both plies Breakage as scheduled after 1,000 h
Test Progress
after 1,000 h
Æexceed of allowable stress Æcomplete failure
1
F2 indicates annealed (Float) glass with two panes. E2 indicates fully tempered glass with two panes.
The loading was chosen in such way so that reasonable stresses will occur in the joint
but a breakage of the glass is avoided. The stress distribution cannot be predicted and
calculated due to the unknown creeping properties.
Though the load was calculated a premature failure of the beam F2_01 occurred. The
load was determined close to the strength of the annealed glass. At this point subcritical
crack growth happened which was additionally amplified by the time dependent stress
redistribution. Due to the viscoelasticity of the joint the stress in the glass increased
776
along with negative impact on the crack growth. Nonetheless, the test was continued in
order to gain knowledge about the post-breakage behaviour. Consequently, the load was
reduced for the second test.
Figure 6: Long term test F2_01: Breakage after 109 und 113 hours
It is obvious that the steel element initially carries a quite significant part of the load.
For example, the increase of the stiffness between section S0 and S1 with the build-up
according to table 2 amounts to 15% (considering the acrylate joint as rigid and non-
flexible). This amount is transferred to the glass over the time due to the time dependent
behaviour of the adhesive. The stress increased not only in the steel element after
breakage of the first glass ply but also in the remaining one. After breakage of the
second ply the stress increased highly in the steel plate because the broken glass was not
able to carry any tensile but only compression stresses. The behaviour is similar to a
cracked reinforced concrete beam.
In the time range from 50 to 109 hours there is no significant change in both the stresses
and deflection. The system came to an equilibrium which means that creeping has
stopped or at least the effects were not detectable anymore. After breakage of all glass
panes the deflections were still recorded. Then the steel plate will take the tensile
stresses while the glass is only able to take the compression. This behaviour is also
comparable to a composite beam.
The second test with reduced loads shows an analogue behaviour as the first test.
Initially the steel takes a significant part of the stress which will be continually reduced
mostly during the first 40 hours. The stress in the glass is increased respectively.
777
Challenging Glass 3
Figure 7: Long term test F2_02 for the first 40 hours
After 1,000 hours no breakage occurred. Figure 8 shows that the centre deflection of the
beam continually increased. That means that creeping has not come to an end. The
horizontal deflection difference between steel and glass is quite small. But nonetheless,
this would allow an evaluation of the creeping properties.
Figure 8: Long term test F2_02: vertical and horizontal difference deflection
The first glass pane was destroyed as scheduled after 1000 hours and 24 hours later the
second one. The first breakage did not result in substantial changes in terms of
deflections. After the destruction of the second ply the deflection increased abruptly
778
along with an increased horizontal deflection of both steel and glass. This effect might
be substantially caused by a larger support rotation due to the higher middle deflection.
After a certain time the horizontal movement equalises, which means that creeping
effects are occurring.
Figure 9: slip of the steel element due to the adhesive’s viscoelasticity
For the third sample fully tempered glass was used. The load amounts to 140 kg on each
loading position. Theoretically a higher load would be possible but the set-up was
geometrically limited as to apply more concrete blocks.
Figure 10: Long term test E2_02: vertical and horizontal deflections
Figure 10 shows the change of the deflection during 1,000 hours. After the initial elastic
deflection the vertical movement increases relatively slow but continually. To gain the
real middle deflection the value of the support deflection must be deducted from the
middle one. Then the change is even lower. But it is visible as well that the vertical
deflection is accompanied by the horizontal deflection. That means that the load
redistribution leads to increased vertical deflections.
779
Challenging Glass 3
After more than 1,000 hours both plies were destroyed as scheduled. Close after the
destruction of the second ply the entire beam collapses due to shear failure at the left
loading point. This effect might be intensified due to the fact that both plies were
punched at an approximately same location. Nonetheless, while the hybrid beams
consisting of annealed glass have a certain post-breakage bearing capacity those made
of fully tempered glass fail completely without any post-breakage bearing capacity.
Consequently, hybrid beams made of fully tempered glass possess a high bearing
capacity at intact condition but no residual capacity. Therefore, such build-ups are not
purposeful and sufficient to meet all safety requirements.
A reference test without an adhesive joint shows the effect of the acrylate bonding. The
reference beam was tested in the same manner as the test F2_02 with 120 kg load at
both third points. The steel plate was only fixed with short tape strips as to prevent a
load transferring connection between steel and glass. The steel plate is assumed to slip
independently. A PVC-bar modelled the thickness of the adhesive layer.
The strain measurement in figure 11 show that the missing composite action leads to
stresses of the glass far above the allowable stresses. Consequently the glass panes
broke after approximately 48 hours (subcritical crack growth).
Figure 11: Long term test: comparison of F2_02 (solid line) and the reference beam (dashed line) – stresses
The effect of the stress redistribution by means of the acrylate interlayer is clearly
visible. The adhesive joint transfers the load from the glass to the steel. The single
components of the beam are not working independently but in combination almost as
one rigid section (except for some deduction such as creeping of the adhesive).
The positive effect of the adhesive joint is also shown in the deflections. The deflection
of the reference beam without any connection between steel and glass is more than two
times higher than the one of the hybrid beam. Additionally the tome dependent
behaviour of the hybrid beam is clearly visible.
780
Figure 12: Long term test: comparison of F2_02 (solid line) and the reference beam (dashed line) – deflections
5. Summary and Conclusions

Hybrid beams made of both annealed and fully tempered glass show no significant
differences in terms of their mechanical behaviour in short term loading tests. The
bearing capacity is able to be increased up to a factor of 2, dependent on the glass type
and steel element used. A sufficient load transferring capacity of the adhesive joint was
proved. But this applies only for intact conditions.
In case of a scheduled or unscheduled destruction the beams made of fully tempered

glass lose all their bearing capacity. The beam collapses totally without any residual
strength. Additionally, the collapse occurs abruptly.
Therefore, annealed glass is more appropriate to meet the safety requirements though it
has quite limited bearing capacity. A feasible solution to meet the safety requirements
along with an increased bearing capacity at intact conditions might be the use of heat
strengthened instead of annealed glass. But this is still matter of further research.
6. Acknowledgements
The research project was financially supported by Bundesministerium für Wirtschaft
und Technologie and was supervised by VDI/VDE Innovation + Technik GmbH.
Special thanks go to the parties involved DELO Industrie Klebstoffe GmbH & Co. KG,
Glasid AG, Hunsrücker Glasveredelung Wagener GmbH & Co. KG, Jansen AG and
ViscoTec Pumpen- und Dosiertechnik GmbH.
781
Challenging Glass 3
7. References
[1] Hess R.: Glasträger. HBT Bericht Nr. 20. Institut für Hochbautechnik ETH. Zürich 2000.
[2] Louter, P.C.: Adhesively bonded reinforced glass beams. HERON Volume 52 (2007) issue 1/2 special
issue: Structural Glass.
[3] Menkenhagen, J., Koschecknick, K.: Geklebte Fassadensysteme. In: Tagungsband; glasbau2009;
Institut für Baukonstruktion der Technischen Universität Dresden; Dresden 2009.
[4] Weller, B.; Härth, K.; Werner, F.; Hildebrand, J.: Hybridbauteile im Konstruktiven Glasbau. In:
Stahlbau Spezial (2009) – Konstruktiver Glasbau.
[5] Weller, B.; Meier, A.; Weimar, T.: Glass-Steel Beams as Structural Members of Facades. Challenging
Glass 2; Conference on Architectural and Structural Applications of Glass; Delft 2010.
[6] Weller, B.; Weimar, T.; Krampe, P.; Walther, A.: Glass Steel Beams - Development of Hybrid
Structures. In: Tagungsband; Glass Performance Days 2009; Tampere 2009.
[7] Weller, B.; Meier, A.; Retsch, S.; Weimar, T Produktentwicklung hybrider Stahl-Glas-Verbundträger –
HybridGlas. Research Project 16IN00659. Final Report, Dresden 2012, unpublished.
[8] Weller, B.; Tasche, S.: Experimental Evaluation of Ultraviolet and Visible Light Curing Acrylates for
Use in Glass Structures. Journal of ASTM International (JAI), Vol. 6, No. 9. 2009.
[9] Tasche, S.: Strahlungshärtende Acrylate im Konstruktiven Glasbau. Dissertation. TU Dresden Institut
für Baukonstruktion. Dresden, 2007.
[10] Härth, K.: Beitrag zum Tragverhalten hybrider Träger aus Glas und Kunststoff. Dissertation. TU
Dresden Institut für Baukonstruktion. Dresden, manuscript.
[11] DIN EN 572-2: Glass in building. Basic soda lime silicate glass products. Part 2: Float glass. Berlin:
Beuth, 2004.
[12] DIN EN 12150-1: Glass in building – Thermally toughened soda lime silicate safety glass - Part 1:
Definition and description.. Berlin: Beuth, 2000.
[13] Prautzsch, V.: Prautzsch, Volker: Klebungen in Abhängigkeit von Klebfläche und Klebschichtdicke.
Diploma Thesis. Dresden: Technische Universität, 2005, unpublished.
[14] DIN EN 1288-3: Glass in building. Determination of the bending strength of glass. Part 3: Test with
specimen supported at two points (four point bending). Berlin: Beuth, 2000.
[15] Weller, B.; Meier, A.; Weimar, T.; Menkenhagen, J.; Koschecknick, K.: Hybride Glasträger als
Fassadenelemente. In: Stahlbau Spezial (2010) – Konstruktiver Glasbau.
782
Challenging Glass 3
Curved & Bended Glass

783
Challenging Glass 3 – Conference on Architectural and Structural Applications of Glass,
Bos, Louter, Nijsse, Veer (Eds.), TU Delft, June 2012.
Reprinted with permission from ASTM International.
Copyright © 2012 by ASTM International,
100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.
doi:10.3233/978-1-61499-061-1-785
Durability of Cold-Bent
Insulating Glass Units
Keith Besserud, Arkadiusz Mazurek
Skidmore, Owings & Merrill, LLP, Chicago, IL, USA
Lawrence Donald Carbary, Kenneth Rubis
Dow Corning Corporation, Midland, MN, USA
Andrew J Black, Donald Misson
Permasteelisa NA, Mendota Heights, MN, USA
Matt Bergers
Viracon Inc., Owatonna, MN, USA
Architectural influences on the commercial building industry have resulted in the

design and development of high performing complex curtainwalls that may include
bent or warped glass. Glass on modern day curtainwall projects is mostly insulating
glass. When an insulating glass unit is bent out of plane, a shearing action occurs to
both the primary and secondary seals. The testing described in this paper was done
to determine the effect on the durability of the insulating glass unit as measured by
argon retention, frost point change and visual changes after aging in a displaced
condition. Full size IG units 5 ft by 10 ft by 1 in (1.52 by 3.05 m by 25.4 mm) were
fabricated, bent out of plane at 2-12 in. (50-300 mm) in 2 in. (50 mm) increments
and placed under a 100 psf (4788 Pa) wind load. One unit was displaced 15 in. (380
mm) where breakage occurred. An assessment was made based on the data as to
what would be a reasonable amount of bending to avoid excessive glass stress and
avoid breakage. This full size unit was then modeled and measured to predict and
validate the stresses and strains on the primary and secondary seals. Upon
completion of the bending tests and the review of modeling, small IG units 14 by
20 in. (350 by 500 mm) were tested according to the protocol specified in ASTM
E2188-10 and E2190-10 with an equivalent amount of edge seal displacement in X,
Y and Z directions. Argon retention and frost point measurements were taken
before and after the durability testing and results reported. During the durability
testing the glass of several of the units broke in thermal cycling due to high local
stresses in the glass. However, the units that did survive the aging protocol passed
both frost point and argon retention tests, while maintaining their appearance.
Additionally, the methodology developed in this research provides a strong
foundation for future testing in the area of cold-bent glass durability.
Keywords: Insulating Glass Durability, ASTM E2190, ASTM E2188, Bent Glass,
Cold-bent Glass, Structural Silicone, Architectural Design Feature, PIB Durability,
Argon Retention, Warped Glass, Doubly Curved Glass
1. Introduction
1.1. History and Reasons for Research

Over the past decade the practice of architecture has witnessed the widespread
introduction and adoption of new modeling software programs that facilitate the rapid
785
Challenging Glass 3
conceptualization of highly complex, curvilinear building geometries. Based on Non-

Uniform Rational B-Spline (NURBS) geometry that was developed in the automotive,
nautical, and aerospace industries, these types of programs have become very popular
amongst designers. Often resulting in surfaces that are doubly-curved, these building
designs inevitably proceed through some process of "post-rationalization" in which the
architect's aesthetic design intent must be reconciled with the realities of construction
limitations (budgets), material behavioral properties, and laws of physics.
One approach often considered in this process of post-rationalization is known as "cold-

bending". Cold-bending refers to a practice of fabricating unitized curtainwall panels in
a standard process (without any curvature induced) and then bending the panels into a
cold-bent condition as they are installed on the face of the building. The theoretical
appeal of this approach is primarily two-fold: 1) Aesthetically, it has the potential to
allow designers to realize a continuously smooth, industrial design quality reflective
surface (as opposed to a faceted surface); and 2) It can prove to be a cost effective
strategy for cladding a building surface of double curvature (relative to the other options
available).
Currently, however, architects are often hesitant to pursue this approach. Naturally,
there are limits to the amount of bending that can be induced in the panels, but
unfortunately there is currently little information available that clarifies these limitations.
Manufacturers, fabricators, installers, and designers (and their insurers) are all left to
establish their own comfort levels with respect to cold-bending practices in an ad-hoc
and extremely conservative manner because so little is known about the structural and
other performance related qualities of the glass, the sealants, and the other components
of the panels when exposed to long and short term bending forces, in combination with
the complications of weathering.
Therefore, the goal of this research was to begin to establish some empirical data with
respect to cold-bending. Because this study was necessarily constrained to a single
specific set of conditions (particular panel size & aspect ratio, particular panel
components by particular manufacturers, etc.) and because the study was primarily
focused on the behavior of the polyisobutylene (PIB) primary seal, much more research
will be required to develop a more comprehensive understanding of the behavior of
cold-bent curtainwall panels. Nevertheless, this research has yielded some valuable
insights.
1.2. Past Experience

The behavior of the glass and frame system under cold-bending has been previously
studied on several projects. One of the first appearances of cold-bent glass in the
architectural façade industry was in insulating glass strip windows at the City Hall
building of Alphen aan den Rijn in the Netherlands in 2002 [1]. Another Netherlands
project also constructed in 2002, the Floriade Pavilion in Haarlemmermeer, used point
supported laminated glass panels cold formed onsite to glaze three large openings in the
Pavilion. The cold-bent panels created a curved glass surface matching the free-form
exterior of the Pavilion [1]. In addition to building façades, the smooth curved
transitions between cold-bent laminated panels gave designers the freedom to create an
undulating point supported glass roof for the Tramstation at Zuidpoort in Delft,
Netherlands [2]. More recently this innovation was employed in the renovation of the
786
Durability of Cold-Bent Insulating Glass Units
Victoria and Albert Museum in London where the glass panels were cold worked on
site to create a 4000 ft2 (370 m2) twisting roof over a previously unused courtyard space
[2].
However, despite its growing presence in the international construction scene this
technology has been used sparingly in the United States. Concerns regarding long term
stresses and deformations induced on the glass and silicone by the cold-bending process
as noted in prior research [1, 3] may be one of the reasons for its sparse use.
Nevertheless, the engineers and designers of these successful cold-bent projects and
others were able to use various methods of finite element modeling and physical
experimentations to determine the structural resiliency, code compliance, and
warrantability of the curved glass [3, 4, 5]. From these analyses and testing, glass
manufacturers developed enough confidence to warrant their glass products for use in
cold-bending applications. However, the amount of cold-bending utilized in previous
projects was relatively small compared to the bending performed during this research.
This research seeks to gain a more complete understanding of the limits of the durability
of a cold-bent insulating glass unit (IGU) under cold deformation by exceeding the
amount of deflection previously used, and to assess the durability of such a deflected
IGU.
1.3. Scope of Research

The scope of this research was to determine the durability of the primary weather seal of
an insulating glazing unit under cold-bending deformation. The research team
hypothesized that the primary cause of IGU failure (moisture infiltration) would be
caused by a strain in the edge seal. The standard test for determining durability is
specified in American Society of Testing and Materials (ASTM) E2188-10 [6] and
E2190-10 [7] and these standards specify an IGU of a certain size and make-up.
Because of scaling problems, these standard IGUs cannot be deformed in the same way
as a full scale curtainwall panel to obtain the same edge-seal strains. Therefore, the team
devised a test procedure which had the following steps: 1) model a full-sized
curtainwall panel with Finite Element software, 2) perform a full scale test on a
curtainwall panel for model verification, 3) deform the edge seal of a small IGU
specimen equal to the deformation in the finite element model and full scale test, 4)
perform ASTM durability testing on the small specimen to determine the relative
performance of a cold-bent IGU as compared to a “flat” IGU.
This testing and evaluation was all done within the scope of inch pound (IP) units. SI
units are provided in parenthesis throughout this document .
2. Full Scale Testing
2.1. Overview
The full scale test units consisted of a 10 ft high by 5 ft wide (3.05 by 1.52 m) insulating
glass unit framed by vertical and horizontal extruded aluminum profiles. Three identical
test units were fabricated and each one was anchored at all four corners to its own wood
test frame. Each individual wood frame was rigidly connected and sealed to a strong
wall integrated with an air compressor capable of producing both negative and positive
787
Challenging Glass 3
pressures. During testing the anchor at the same top corner of each test frame was
removed to apply the out of plane displacement.
A preliminary finite element model of a full scale test unit was created in order to
determine the limit of out of plane deflections that can be applied in actual construction
practice. In the model beam elements represented the framing members and plate
elements simulated a single glass layer. Two adjacent edges of the glass were restrained
in the out of plane direction while the other two edges were allowed to freely translate.
An incremental displacement was applied to the free corner of the model to determine at
what displacement the maximum long term stress in the glass (factored to account for
the stiffness two layers of glass in the actual unit) would exceed limits specified in
ASTM E1300-07 [8] and Glass Association of North America (GANA) 50th
Anniversary Edition [9]. The finite element analysis indicated that 12 in. (300 mm) was
the maximum amount one corner of the full size unit could be pulled out of plane before
exceeding the long term allowable stresses in the glass. A more refined model was
developed later in the project to accurately predict edge seal strains.
The three test assemblies (a test assembly is the full scale test unit, wood frame, and
associated measurement devices) were each subjected to unique test procedures and
data acquisition methods in order to evaluate the various behaviors of the insulating
glass unit under applied cold-bending, and also to validate computer finite element
models. The first test assembly was set up as a baseline test to evaluate the structural
capacity of the test unit in cold-bending. Measurements were taken at incremental out
of plane displacement up to the maximum of 12 in. (300 mm). The second test
assembly provided information about the edge seal deformations between the outer and
inner glass layers at the same increments used in the first test assembly. The third test
assembly was designed to validate the results of the first and second test assembly
through comparisons between their strain data, and also to determine if the unit under
maximum bending was capable of withstanding repeated applications of static pressure.
All three test assemblies were also designed with measurement devices to correlate their
cold-bending and compare strains at specific locations on the glass. However, only test
assemblies one and three were subjected to a baseline test performed in accordance with
ASTM E330-02 [10] Procedure A. Test assembly two required access to the displaced
corner for measurements and thus could not be sealed to the test wall for pressurization.
During the pressure test, the air compressor applied a pressure of 100 lbf/ft2 (4788 Pa)
to the glass surface through the sealed pressure chamber. Engineering judgment and
prior job experience were used to identify 100 lbf/ft2 as a typical maximum wind
pressure that a high rise building might experience in a 50 year return period. The
structural silicone used as the secondary seal for the IG unit and also to attach the
insulating glass to the metal frames was sized and designed around the above mentioned
windload so that the structural silicone would maintain its industry standard 20 psi (138
kPa) design stress. Due to the limited sample size and inherent imperfections in glass,
this test provided a necessary baseline performance criterion that units had to pass in
order to be accepted as fit for cold-bending. This test also provided additional
information about the deformation states of the cold-bent surfaces under pressure.
788
2.2. Curtainwall Frame

The curtainwall panel frame was pin-anchored in three corners by steel angles to
restrain in-plane and out of plane displacement and allow for the rotation at the corners.
The fourth corner was pulled out of plane using a hook-and-slide mechanism that is
typical to a curtainwall anchoring system. The anchoring and pulling mechanism was
designed to allow for a full 12 in. (300 mm) of deflection in one direction.
The framing profiles used were designed for a previous flat-glazed project and no prior
considerations for cold-bending the frame were developed into the frame design. The
framing members were open channel vertical shapes and closed tubular horizontal
shapes. The horizontal members were attached to the vertical members with three to
four fasteners, which were attached through clear holes in the vertical framing members
and threaded into screw races in the horizontal member. See Figure 1 for framing
member profiles and section properties.
The IG units were structurally glazed in-shop into the frame using structural silicone
sealant with a bite of 0.75 in. (19 mm) and a depth of 0.25 in. (6.4 mm) so that the 20
psi (138 kPa) industry standard structural silicone design strength would be maintained
at the 100 lbf/ft2 (4788 Pa) windload.
789
Challenging Glass 3
Figure 1: Aluminum framing profiles and section properties; a) Male Vertical Mullion, b) Lower Sill
Transom, c) Upper Head Transom, d) Female Vertical Mullion
2.3. Glass Make-Up

The insulating glass units for the full scale test were fabricated using two 0.25 in. (6.4
mm) pieces of fully tempered clear glass. A 13.2 mm mill finish aluminum spacer
established a 0.520 in. (13.2 mm) air space. On both sides of the spacer, an average of
0.015 in. (0.4 mm) of pressed PIB provided a primary seal between the spacer and the
glass. The IGU was glazed with 3/8 in. (9.5 mm) of structural silicone insulating glass
sealant.
790
2.4. Silicone
The structural silicone that was used as the secondary seal for this study was also the
silicone used to attach the IGU to the aluminum frame. This two part structural silicone
conforms to both ASTM C1184 [11] and ASTM C1369 [12]. This particular silicone is
described as Sealant “D” previously by Wolf [13] as shown in Figure 2. Data gathered
on the tensile strength of the IGU edge seal assembly per the ASTM C1265 [14]
configuration is presented in Table 1. This data includes ultimate tensile strength as well
as stress data at 2.5%, 5% and 10% strain.
Figure 2: Stress-strain relationship of silicone at 22°C
2.5. Full Scale Test Procedure

Test Assembly One consisted of a curtain wall frame anchored to a wood test frame at
each corner. See Figure 3. The wood frame was rigidly anchored and sealed to the test
wall with the exterior lite of glazing facing the test wall. All gaps between the curtain
wall unit and wood test frame were also sealed. It was subjected to initial pressurization
of four positive and four negative cycles of ten seconds each at 100 lbf/ft2 (4788 Pa) in
general accordance with ASTM E330-02 [10] to ensure that the glass could attain
design pressures prior to cold-bending. The upper right anchor was removed and the top
corner of the unit was pulled two inches (50 mm) out of plane away from the test wall.
The anchor was reinstalled, and one ten second cycle was run for the positive and the
negative pressures. This procedure of displacement followed by pressurization was
repeated in 2 in. (50 mm) increments until a total displacement of 12 in. (300 mm) was
reached. After pressurization at a displacement of 12 in (300 mm), the unit was
displaced till breakage at 15 in. (380 mm). Linear and rosette strain gages provided
strain information at key locations, and linear variable displacement transducers
(LVDTs) were attached to the glass in the corners to provide accurate displacement data.
Data from these gages was recorded for each displacement increment and during all
load cycles. Figure 4 shows the locations of these gages on Assembly One.
791
Challenging Glass 3
Figure 3: 120 by 60 by 1 in. (1.5 by 3.0 m by 25 mm) Full scale test assembly #1.
Figure 4: Full scale test assembly #1 gage location schematic
792
Assembly Two was installed in a partial wood test frame and anchored at three locations,
leaving the top right corner free (Fig. 5). The partial frame allowed access to the side of
the glazing where six dial indicators were mounted to the outer pane of glass (Fig. 6) to
monitor the edge seal strains as the unit was displaced. Consequently, the frame was
unable to be sealed and no pressure was applied to Assembly Two. The free corner was
displaced in two inch (50 mm) increments until a total displacement of 12 in. (300 mm)
was achieved. Dial indicator and linear strain (Fig. 7) readings were recorded at each 2
in. (50 mm) increment and LVDT’s were again used to measure glass bending.
Assembly Three followed the same installation as assembly one. For this assembly, the
top right corner was displaced to ten inches (250 mm) and secured. Eight ten-second,
100 lbf/ft2 (4788 Pa) test loads were applied in accordance to ASTM E330-02 [10],
alternating between positive and negative directions. Linear and rosette strain gages and
LVDT’s were applied to this assembly as shown in Fig. 8. During the time of testing,
the authors decided that design cold bend would be 8 in. (200 mm) and the 10 in. (250
mm) displacement would correspond to 125% of the design displacement of Assembly
1. This was done to determine if repeated loading beyond the 100% cold bend design
would result in glass breakage of this single assembly.
Figure 5: 120 by 60 by 1 in. (1.5 by 3.0 m by 25 Figure 6: Dial gage used to measure relative
mm) Full scale test assembly #2 in partial test frame displacement between glass panes in full scale test
assembly #2
793
Challenging Glass 3
Figure 7: Full scale test assembly #2 gage location Figure 8: Full scale test assembly #3 gage location
schematic schematic
3. Small Unit Durability Testing

3.1. Overview
The purpose of the small unit test was to determine the durability of a cold-bent IGU
using the industry-standard testing protocol as specified in ASTM E2188-10 [6]. The
edge-seal conditions of a full-scale, cold-bent curtainwall panel cannot be simply
converted into an equivalent degree of bending in a small test specimen. Because of the
specific requirements of this testing protocol, the glass stiffness, the IGU spacer size,
and PIB size could not be scaled and still maintain comparable results to a “standard”
ASTM E2188-10 test. Therefore, the edge-seal was strained an amount equal to that
observed in the full scale testing and validated by the finite element model. Following
this method a similar edge seal condition was re-created between the full-scale and
small-scale tests.
The small scale test was broken into four sets of identically constructed specimens with
varying amounts of edge-seal strain. The first set was a control and was tested without
any additional edge-seal deformation. The second set had edge-seal deformations which
794
corresponded to a “100%” allowable amount of bend. The “100%” allowable amount

of bend was determined through a combination of engineering judgment and the results
of full scale testing. The third and fourth set contained edge seal deformations that
corresponded to “50%” and “150%” of the allowable amount of bend. It is important to
note that “50%” and “150%” are not indicative of the amount of edge-seal strain, but
relate to the amount of displacement in the full scale test. The amount of edge-seal
strain in these cases was determined through a combination of the full-scale test and
finite element model.
3.2. Design of Test Apparatus

The testing procedure for the small scale durability test (as specified in ASTM E2190-
10 [7]) required a compact and portable displacement mechanism to impart movements
on the two lites of glass that would replicate the edge seal strains in the full scale test
unit. The final design consisted of a four sided frame made from a bent and welded steel
1/4 in. (6.4 mm) plate. The 14 by 20 in. (350 by 500 mm) IGU was placed into the
frame and the lower lite was secured to the frame on all four sides using epoxy resin.
The frame was designed to surround the four edges of the IGU in order to adequately
apply load to displace the pane of glass, however, there was a nominal clearance of 3/8
in. (9.5 mm) between the steel and IGU edge seal. This allowed for adequate clearance
around the edge seal as specified in ASTM E2188-10 [6]. Fasteners and bearing plates
were used to displace the upper pane of glass in-plane relative to the lower. In plane
displacements are defined as X and Y directions. Another set of fasteners attached to
plates adhered by epoxy to the surface of the glass was used to pull the upper pane of
glass away from the lower. This out of plane displacement is defined as the Z direction.
See Fig. 9 for a photograph of the small test frame. In this photograph, there are steel
plates and fasteners bonded to the glass at the mid span of the long 20 in. (500 mm)
dimension. The screws shown at these locations were used to lift the outer lite away
from the inner lite in the Z direction.
Figure 9: 14 by 20 in (350 by 500 mm) small test frame will IGU installed
795
Challenging Glass 3
To apply the desired displacement between the two panes of glass, two #6-40 socket
head machine screws were used in each in-plane direction. The data in Table 1 was used
to estimate the forces required to strain the outboard lite of the small test units and to
determine the size of the screws used to produce the strains. Steel shims and silicone
bearing plates were used to distribute load and protect the glass edge (Fig. 10) when
moving the plates in the X and Y directions.. The applied displacement was measured
between the steel frame and displaced glass pane using a dial caliper. The out of plane
displacement was applied using two #6-40 flat head machine screws that were adhered
to the upper unfixed pane of glass with an epoxy resin. This displacement was measured
between the steel frame and glass pane using an outside micrometer. Additionally,
measurements were taken between the lower fixed pane of glass and the steel frame to
ensure that the entire IGU did not move relative to the steel frame. All of these
measurements were recorded at the time of application and subsequently measured
throughout testing. See Table 2 for recorded measurements and Figure 11 for the
locations of displacements.
Figure 10: Application of small specimen edge seal displacements
796
Figure 11: Edge seal displacement table locations
A slight modification was made to the original test frame design due to the high loads
required to displace the glass out of plane in the Z direction for the 150% design test
specimens. The outer pane of glass broke in two glass specimens during the application
of displacements and these were discarded. Also, two of the tension mechanisms that
were secured to the glass with epoxy broke free and had to be reinstalled. To avoid
stress concentrations for the 150% tests, a revised tensioning mechanism was designed
and installed for these specimens. This mechanism consisted of a longer steel plate
adhered to the glass to more effectively distribute load and thereby reduce stress. The
new tension mechanism allowed the glass to be deflected in tension to the required
degree without additional glass breakage.
3.3. Insulating Glass Make-Up

Twenty four units of IGU for the durability testing were manufactured using standard
14 by 20 in. (350 by 500 mm) configurations as prescribed by ASTM E2190-10 [7].

The units were fabricated using tempered glass with a thickness of 0.185 in (4.7 mm)
for each lite. The outer lite has a Low-E coating on the #2 surface that is deleted to the
PIB interface. The air space was established using a 13.2 mm, mill-finish aluminum
spacer. The PIB thicknesses ranged between 0.010 in. (0.25 mm) and 0.030 in. (0.76
mm) and an average PIB thickness of 0.015 in. (0.38 mm) was attained. Two legs of
the spacer were filled with a 3A molecular sieve desiccant. The units were filled with
argon to greater than 90% fill level and then sealed with 3/8 in. (9.5 mm) of silicone
insulating glass sealant.
3.4. Weathering Test Procedure

The durability of the test specimens was evaluated by industry established and accepted
methods of exposing the units to high humidity environment, alternating temperatures
and periods of UV exposure and water spray. The method chosen is described in detail
in ASTM E2188-10 [6]. The procedure and the cyclical phase are shown in Figure 12.
Evaluation of argon retention was conducted per ASTM E2649 [15].
797
Challenging Glass 3
Figure 12: ASTM E2188-10 [6] Test Cycle
The test began by establishing an initial frost point per ASTM E546 [16]. The ASTM
E546 test method placed a cold plate on the surface of the glass. Condensation in the
form of frost or, at higher temperatures, dew would be observed if there was sufficient
moisture in the IG unit. Factory-made units are expected to demonstrate low
temperature frost point at the start of the test, because the desiccant is new and not
subjected to any extreme environmental influences. Typically the initial frost point
results are expected to be lower than -90 °F (-68 °C).
Additionally, testing for Argon in the units was conducted per ASTM E2649 [15]. All
units began the test with greater than 90% argon concentration.
After these initial tests, the units were placed in a high humidity chamber for a period of
two weeks. The chamber was maintained at 140 °F (60 °C) and 95% relative humidity
for the entire two weeks.
Once this phase was completed, the units were removed and allowed to equilibrate to
standard temperatures and conditions for a period of 24 hours. The units were then
measured again for frost point and argon retention and the values were recorded.
At the completion of the data collection, the units were placed in a weather cycling
phase. During this phase, the units experienced cold temperatures to -20 °F (-29 °C).
After a hold period, they were allowed to return to standard temperatures at which point
they are exposed to ultraviolet light and mist spray while their temperatures were raised
to 140 °F (60 °C). The spray was turned off and after a hold at high temperature and
while the ultraviolet light is still on, the units were allowed to return to standard
temperatures where the cycle begins again. This six hour cycle was repeated for nine
weeks.
798
At the completion of the weather cycling phase, the units were once again removed,
allowed to equilibrate for 24 hours and the frost point and argon values were measured
and recorded.
Finally, the units were placed back in the humidity chamber (as described above) for an
additional four weeks. At the completion of the four weeks, they were removed,
allowed to equilibrate for 24 hours and the final frost point and argon values were taken
and recorded.
3.5. Argon Testing Procedure

Testing for argon gas content was performed on all of the small testing units at four
specific time frames, (initial, after two weeks in high humidity exposure, after nine
weeks of weather cycling, and then again after four more weeks of high humidity
exposure). Units were tested by non-destructive spark emission spectrography (SES) in
accordance with ASTM E2649 Determining Argon Concentration In Sealed Insulating
Glass Units Using Spark Emission Spectrography [15]. In order for an insulating glass
unit to be listed as complying with gas content certification, the unit must demonstrate
an average argon containment of greater than 90% before weathering, and the average
argon levels of the entire sample group must be greater than 80% after weathering.
3.6. ASTM E2188-10 [6] Test Procedure: Expectations

Test unit failure due to workmanship issues related to the initial fabrication of the 14 by
20 in. (350 mm by 500 mm) units is normally expected to be seen very early in the
testing process typically in the first couple of weeks of high humidity. Once past this
point, test units of the configuration in this study can be expected to complete the entire
cycle without evidence of increased frost temperatures or argon loss due to
workmanship of the construction. If there was a sign which indicated that a unit was
either demonstrating higher frost temperatures (therefore gaining moisture from the
outside atmosphere) or losing argon, that could be interpreted as the induced effects of
edge seal strains at the PIB to glass or PIB to spacer interface.
4. Finite Element Modeling and Validation

4.1. Full Scale Unit Modeling

The glass has been modeled using 0.5 in. (12.7 mm) square shell elements. The
structural effect of the PIB layer is negligible and was not considered in the model. The
silicone has been modeled using 0.5 by 0.25 in. (12.7 by 6.35 mm) rectangular shell
elements. Due to the one- and two-dimensional nature of the beam and shell elements,
rigid links were used to hold the actual three-dimensional locations of the elements in
space. Rigid links are linear elements that connect two nodes rigidly in space. The
silicone was modeled as nonlinear material with shell elements configured
perpendicular to the surface of the glass. The stress strain curve of the material has been
obtained from physical tests at room temperature (22 °C) and is presented in Figure 2
[13]. The corner portion of the finite element (FE) model is presented in Figure 13.
The aluminum framing members are typical of a four sided, structurally glazed, unitized
curtainwall system. Each framing member profile was unique and the properties of the
sections have been calculated and are presented in the Figure 1. The volume of the air
799
Challenging Glass 3
cavity has been maintained with fluid elements with bulk stiffness of 100 kPa. The four
corners of the frame have been modeled with pinned anchors where one corner had non-
zero out of plane restraint. The end of the horizontal beam elements were moment
released at the shared node of the vertical beams to represent a pinned connection
between the horizontal and vertical framing members.
There are certain elements of the physical specimen that have not been modeled
numerically. These parameters include negligible effects such as: the stiffness of PIB,
the effect of gravity, and local distortions of framing profiles. Other parameters that
were not modeled and may have significant effects include the flexibility/plasticity of
the frame corner connections and thin-walled beam behavior of the frame members. The
torsional behavior of the thin-walled frame members differs from the as-modeled solid
members. The influence of warping of the thin-walled section changes the torsional
shape of deformation of the frame members and may have significant impact on the
overall results. A difference in torsional deformation to the one observed in the physical
test has been obtained in the numerical model.
Figure 13: Corner of FE model of full size test
4.2. Analysis of the Full Scale Unit Model

There are several possible failure scenarios during the cold-bending of a curtain wall
unit. Some of the major ones are: structural failure of the frame, breakage of glass,
structural silicone failure or PIB seal separation in the IGU. Structural failure of the
frame can be safely predicted through the typical structural analysis involved in
curtainwall design. Glass can be designed to perform with the sustained stresses induced
through cold-bending. Structural silicone is considered to be a strong component in the
system and not a critical link. On the other hand, it is suspected that the PIB seal is the
800
governing element during cold-bending of IGU’s. Long term performance of such a seal
under sustained strains is unknown. However, the seal is known to fail under a regime
of ASTM testing procedure even without applied strains.
Based on the results of the FE model, the locations of maximum edge seal strains in the
PIB were predicted. Preliminary FE analysis of the full scale unit revealed that the
maximum edge seal deformations were located on the long side of the panel about 40 in.
(1016 mm) away from the loaded corner. Intuition and engineering judgment, however,
led to the conclusion that strains at the corner of the unit may also be significant.
Therefore, these two locations were selected for the measurements of the edge seal
strains on Assembly Two. The strain state at each of the locations was measured using
three displacement dial gages attached to the external ply of the glass to measure the
relative in plane edge displacements of the internal ply of the glass. Figure 7 shows
locations of all six of these gauges. From these measurements the edge seal strains in
the PIB were calculated and presented in Table 3.
In addition to measuring the in-plane movements of the glass, several strain gages were
attached to the glass in order to understand its true bending behavior. These gages were
used to validate the results obtained between the three full scale test assemblies and FE
model. The locations of strain gages in the full size test specimen of test one and three
are shown in Figures 4 and 8 respectively. Two rosette strain gages (numbered 1
through 6) were placed at the center of the glass on the inner and outer pane. Rosette
gages 7 through 9 are placed at the location of maximum stress in the glass as predicted
by FE model. These gages helped in understanding the shape of deformation of the
glass during its bending. Unidirectional strain gages were placed on all three test
assemblies on the outer glass layer at the midpoints of the free edges. Readings of gages
10 and 11 are affected mainly by the stiffness of the frame members.
The amount of applied out of plane deformation of the IGU in the physical testing was
measured using four LVDT’s attached near the corners of the glass (Fig. 14). The out of
plane displacement in the numerical model however, was applied using forced
displacement values of the node at the loaded corner with the other corners of the
frames restrained. Since these measurements were taken at different locations, a
correlation between the out of plane deformation in the physical and numerical models
needed to be developed. Therefore, the displacements obtained directly from LVDT
readings in the physical test were compared to the corresponding nodal displacement
results of the glass in the FE model. This comparison between physical and FE model
out of plane displacements (bending) is presented in Fig. 15. Data presented in this
paper corresponds to displacements as measured in the physical test.
801
Challenging Glass 3
Figure 14: One of four LVDTs measuring bending in the full scale test specimen
Figure 15: Relation between physical and numerical out of plane displacements
802
4.3. Validation of Models

In order to correctly create a structural model we need to understand the behavior of
glass bending, its expected displacements and other factors that have an effect on the
glass deformation. To understand the cold glass bending, it is useful to study simplified
models of deformation. The shape of the cold-bent glass is more complex than the
following two theoretical models, but they are the major contributors to the overall state
of deformation.
For a cold-bent plate, the first idealized deformation shape is one where straight lines
parallel to the edges remain straight after the plane is deformed. The deformed shape
that follows the straight lines rule is presented in Figure 16.
Such deformation will create a state of stress in the glass such that:
V xx z V yy z W xz z W yz z 0
(1)
and
W xy t
2 W xy t
2 z0
(2)
Where:
t = thickness of the plate
This state of stress represents two directional bending along x`-y` directions where x`-y`
are axes rotated 45 ° away from x-y (Fig 16).
The second idealized state of deformation is unidirectional bending. There are two
statically equivalent states of deformation with unidirectional bending where the
bending could occur about either of the two diagonals. Both of these states are presented
in Figure 17. Depending on the initial deformation, a structure can arrive at either state
of equilibrium. The diagonal about which the glass bends can be selected by forcing the
bend during the initial deformation. This effect is very difficult to obtain numerically.
Depending on the initial deformation state, nonlinear FE analysis will return various
outcomes. However, simple tests such as bending a credit card by hand will reveal that
unidirectional bending requires the least amount of energy to force four corners of a
rectangular plate out of plane. Applying external pressure to the surface of such bent
glass can cause an effect known in the literature as “snap through buckling” [17]
803
Challenging Glass 3
The major differences between the two cold-bent shapes described above are:
x In the bidirectional bending example, the edges of the rectangular glass

remained straight (Figure 16).
x In the pure unidirectional bending example, the edges of the glass deform
freely (Figure 17). One of the diagonal lines remains undeformed. The
direction of bending is perpendicular to the undeformed diagonal.
x In both pure unidirectional and bidirectional conditions, if the glass edge is
framed the framing members are subject to torsion
x The energy required to obtain unidirectional deformation is much smaller from
the energy required in the bidirectional bending condition. Therefore the
deformation state of bidirectional bending is possible but unstable.
Figure 16: Idealization of purely cold-bent rectangular plate

Figure 17: Two states of unidirectional bending
804
The above states of deformation were observed with experimental testing [4] where the
two directional bending is observed with small deformations and the unidirectional
bending is a post buckling form.
The duality of the large deformation state was observed during the full scale testing
presented earlier in the paper. The application of pressure to the IGU in the cold-bent
condition forced the deformed glass from one state of minimum energy to another. This
is referred to as “snap through buckling” in the literature. This was quite a visual
surprise during the full scale testing applying the windload to deformed unit.
An intuitive understanding of the principles laid out above would lead us to the
following conclusions: a cold-bent plane with infinitely stiff edges would deform purely
in a bidirectional manner and a cold-bent plane with no frame at all would result in a
simple unidirectional bend (about one of the diagonals). In our test, there are frame
members that stiffen the sides of the rectangular IGU and the outcome was somewhere
in between these two idealized cases.
Considering the above concepts, the deformation of the glass during the cold-bending
process depends on the proportions between the flexural stiffness of the stiffening frame
members and the glass panel itself. It should be noted here that the torsional
deformation of the frame is a result in both of the idealized cases. Therefore, to allow
for this deformation without high torsional forces in the frame, some members should
be torsionally weak (i.e. thin-walled, open section frames). The two vertical frame
members in the full size specimen are significantly weaker in torsion than the two
horizontal members. A combination of framing members with different bending and
torsional stiffness creates a complex system where the state of bending deformation may
not be intuitive.
The physical tests of cold-bending of glass were intended to proceed to failure and large
deformations were a part of the testing protocol. While the behavior of many of the
materials (such as glass or aluminum) had a linear physical behavior, the silicone
connecting these parts had nonlinear physical behavior. Therefore, a model considering
material nonlinearity and large deformation needed to be built.
4.4. Sources of Error and Modeling Inaccuracies

As previously discussed, the modeling of the full scale test unit required consideration
of the many different variables and how they interacted with one another. Given these
possible variables and interactions, sources of error and modeling inaccuracies are
inevitable. The research team sought to mitigate as many of these factors as possible
through the testing regime and through identifying potential areas of inaccuracy. The
warped shape of the glass and modeling of the framing connections are two such areas
of possible error.
The dual deformation modes of the glass proved to be one main area of uncertainty in
the results. Because of the complex deformation of the glass, the readings of the rosette
strain gages did not show a close correlation to the strains in the numerical model.
Buckling of the glass can cause a dramatic change in stress values which could not be
obtained in the finite element model. However, the linear strain gages showed a much
805
Challenging Glass 3
closer correlation to the tested unit (see Fig. 16 and Fig. 17), so it was reasonable to
consider the model a good representation of the behavior of the full scale test unit.
As previously stated, the connections between the framing members in the full scale
testing consisted of several machine screws fastening the horizontal members to the
vertical members. In the finite element model these connections are modeled as pins
(not restraining any moment) or fixed (restraining relative rotation). However, the actual
connections are able to transfer some amount of moment before there is enough rotation
to consider the connection pinned. This difference between the actual connection and
the modeled connection can impact the correlation between recorded and modeled
strains. The stiffness of the connections is unknown and it is difficult to predict without
additional tests. It has been decided that the assumption of a pinned connection is the
closest prediction of a real behavior because it ultimately led to a closer correlation in
data.
To illustrate the effect of the stiffness of the connections on the behavior of the model,
two graphs are presented (Figs. 18 and 19). These graphs show unidirectional strain
gages #10 and #11 (see Figures 4 and 8) readings for pinned and fixed conditions of the
connections respectively. Graphs noted as FE are numerical results, and Test #1 and #3
are physical results of the two tests. Note that the fixed model captures an event at 6 in.
(150 mm) of applied bending. This is possibly a location where the system deforms in
one of two possible states. Also, the outcome of two very similar tests being different
past this point reveals a potential instability of the system. Note, that non-zero initial
strains in test #1 are residual strains after the specimen had been loaded to 4 in. (100
mm) displacement for the first time. A reset in the test procedure was required to correct
the loading mechanism. The graph shows strains after this process.
Figure 18: Comparison of strains at unidirectional strain gauges locations for pinned frame connections. TOP
XX are strains in the location and direction of strain gauge #11 and SIDE YY of strain gauge #10
806
Figure 19: Comparison of strains at unidirectional strain gauges locations for fixed frame connections. TOP
XX are strains in the location and direction of strain gauge #11 and SIDE YY of strain gauge #10
4.5. Determination of Small Unit Discrepancies

The four specimen sets that were tested in the weatherization chamber have been
subjected to edge seal deformations that correlate to maximum strains in the full size
specimen at four values of out of plane displacement. These groups represented 0%,
50%, 100% and 150% of the maximum design displacement. The maximum design
displacement was determined with an engineering judgment. In practical applications,
this would depend on the wind pressures, the size or shape of the panels, the makeup of
the IGU and many more parameters. The engineering judgment has been made
considering the maximum glass stress from the FE model. Numerical tests performed on
the model revealed that 8 in.(200 mm) of displacement produces maximum stresses of
4.7 ksi (32 MPa). This is still well below the long term limit stress of the fully tempered
glass but considering additional possible stresses from positive and negative wind
pressures, climatic loads and other safety factors, 8 in. (200 mm) of bending has been
determined to be a reasonable limit of engineering design. Therefore the maximum
applied edge seal deformations during cold-bending of the full size unit have been
recorded for 4, 8 and 12 in. (100, 200 and 300 mm) of corner displacement. The
location of where the PIB is strained the most varies depending on the stiffness of the
framing elements, connections between them and many other factors. From preliminary
numerical modeling it has been determined that the location is about 40 in. (1.016 m)
below the loaded corner. After the physical tests were performed, the numerical model
needed to be revised. These revisions included applying the proper glass offsets,
modeling the air pressure in the cavity, considering the pin behavior for the framing
members and modeling an accurate location of the applied load. The outcome of the
analysis was very sensitive to these minor model modifications but after the above
modifications have been applied, the location of the maximum shear displacements of
the PIB did not change significantly from the initial model and the modeled system
correlated much more accurately to the test data.
To obtain edge seal strain between the two panes of glass, three dial gages per location
of concern have been used in phase two of the full scale test. The three dial gages have
been configured to obtain the differential movement along the short edge, the
differential movement along the long edge and in plane rotation of the panes with
respect to each other at each of the edge locations (Figure 7). The recorded differential
807
Challenging Glass 3
displacements between panes have been converted to the displacement at single points
of interest and they have been compared with the numerical results. Because of the
sensitivity of the model and some modeling inaccuracies (described above) the results
of the model match very well only for selected displacement values. Readings of all six
dial gages compared with numerical results are presented in Figure 20. Refer to Figure 7
for dial gages numbering. Values of edge seal strains from various locations of the
numerical model have been put side by side with the physical results in Table 3. It
should be noted here that the edge seal deformations are applied to the small test
specimens on all four edges simultaneously, therefore a displacement that is
perpendicular to the long edge of the small specimen will be at the same time parallel to
the shorter edge. The strains applied to the small specimens are shown in the right-most
column of Table 3 and are summarized in Figures 21, 22, 23 and 24 which graphically
depict the deformations for the control, 50% design, 100% design and 150% design
specimens respectively.
Figure 20: Overlay of PIB in plane displacements as measured in test #2 and obtained from numerical model
808
Figure 21: Dimensions of primary (PIB) and secondary (silicone) at control no bending
Figure 22: Deformations of PIB and silicone at 50% or 4 in. (100 mm)of bending
809
Challenging Glass 3
Figure 23: Deformations of PIB and silicone at 100% or 8 in. (200 mm) of bending
Figure 24: Deformations of PIB and silicone at 150% or 12 in.(300 mm) of bending
810
5. Results
5.1. Review
Table 4 summarizes the test results of the ASTM E2188-10 [6] weathering protocol of
the small IG units. During the test protocol the small IG units have the frost point and
argon percentage measured initially, after two weeks of high temperature and humidity,
after nine weeks of accelerated weathering (hot, cold, UV and water spray) and after
four weeks of high temperature and humidity. Each argon-percentage test point for a
specific specimen is an average of five readings according to ASTM E2649 [15]. These
averages are further averaged and plotted in Figure 25.
%ArgonasmeasuredduringASTME2188testing
94
92
90
88
Initial
86
2WHH
84 9WAW
4WHH
82
80
78
76
0% 50% 100% 150%
Figure 25: %Argon within the test units as measured during the ASTM E2188-10 [6] testing protocol
All six of the control specimens qualified per ASTM E2190-10 [7], meaning that the
frost points of the specimens remained below -90 °F (-68 °C) and the average argon
retention of all the specimens was greater than 80%. On the other hand, the 50%, 100%,
and 150% samples did not meet the testing criteria. Six specimens must meet the
requirements described in section 4 of ASTM E2190-10 but, because there were no
available replacement specimens when a specimen broke during testing, the entire
sample group could not be qualified. The reason for the limited number of specimens
was because the research team did not strictly follow the requirements of the ASTM
E2188 standard [6] and procure twelve total samples as put forth in section 5.7. Had
this been done, the broken specimens could have been replaced with a new one from the
remaining specimens. Any lites that broke are noted in the table as “breakage” or
“thermal break”. The “breakage” label refers to breaks that occurred during the initial
811
Challenging Glass 3
displacement of the specimens before the testing began. “Thermal Break” refers to lite
failure that occurred during testing.
All of the 50%, 100%, and 150% units that did not experience breakage showed frost
points below -90 °F (-68 °C). Additionally, the average argon retention of these
specimens was greater than 80%.
An unforeseen result from the durability testing was the failure of the epoxy adhesive in
tension. The test units used an epoxy adhesive in several locations. The epoxy was used
to keep the small IG units attached into the steel frame. This epoxy held fast in all of the
tests, however, the epoxy adhesive used to induce displacement in the Z direction
adhesively released from the glass during the humidity and accelerated weathering
cycles. Since the epoxy failed in all of the tests where tension was applied, this testing
represented stressing the PIB in the Z direction for only a portion and not the entire test.
The measurements that were taken to document the edge seal displacement were
monitored after the first high-humidity phase to ensure that the edge seal strain was
maintained. The measurements between the steel frame and the pane of glass that was
adhered to it were maintained and no relative movement occurred between these two
elements. The measured distance between the steel frame and the displaced piece of
glass, however, increased by approximately 0.005 in. (0.13 mm) from its initial value.
All other measurements, including the “z” displacements were maintained after the first
high-humidity phase. It was during the weathering phase of testing that epoxy adhesive
failure caused the tensioning mechanism to fail and release strain on the edge seal in the
“z” direction.
6. Conclusions
6.1. Review
The full scale units that were subjected to wind load of ± 100 lb/ft2 (4.8 kPa) while
under a cold-bend did not break. The initial modeling suggesting at 8 in. (200 mm) of
bending as the limitation of our design was a good engineering judgment. The
successful completion of testing on Assembly 3 shows that the curtainwall system was
able to survive a pressure which was greater than what the profiles were designed for on
a previous project. In fact, the immediate failure of the glass during the over-deflection
of 15 in. (38.1 cm.) suggests that the initial calculation of long term glass stress at 12 in.
(30.48 cm.) was indeed a good estimation of allowable bending from a glass stress
standpoint.
There were no thermal failures in the control set of small insulating glass units tested to
the ASTM E2188-10 [6] protocol. Additionally, all of the control specimens qualified
under ASTM E2190-10 [7] by maintaining a low frost point and high argon retention.
This is evidence of the quality of workmanship in the specimens as they were all
procured at the same time using the same methodology. Therefore, the workmanship of
the other specimens is not in question and the weathering data collected bears this
conclusion out.
812
The argon was retained in each of the small units that represented deflected large scale
units to the same degree or better than the control group. This is a very positive signal.
In no case were the frost points reduced below -90 °F (-68 °C). Therefore the surviving
units must have had the insulating glass primary and secondary seals remaining intact.
Further testing would require an appropriate number of specimens for each specimen set.
The epoxy that was used in this testing was not adequate to be placed under load in the
accelerated weathering environments. The adhesive loss of the epoxy was a significant
disappointment because the epoxy was thought to be a very high performing product.
When the epoxy was performing keeping the glass strained in the Z direction, the glass
failed.
The deformations in the X and Y direction of the insulating glass units did not affect the
frost point and argon retention of the surviving units. In fact, due to the creep of the
structural silicone secondary seal and the relative stiffness of the silicone protection pad,
the X and Y direction displacements increased throughout the high humidity testing.
This means that the displacements induced were conservative because they increased
throughout the high humidity phase of testing. This also is a very positive signal. It is
very likely that a revision in the method to deflect small insulating glass units subjected
to the ASTM E2188-10 [6] protocol can be done and a full compliance with the ASTM
E2190-10 [7] specification can be obtained.
This study was not intended to test the strength of the glass during the weathering cycles,
but to test the effect of strains on the primary and secondary seals. The thermal
breakage that occurred during the weathering cycling does not constitute a failure in the
spirit of this testing, but it is a result of an underestimation of the physical strength of
the tempered glass. The measurements that were taken while the glass was intact
suggest that a strained edge seal is quite resilient to moisture infiltration. Further testing
may provide evidence in favor of these preliminary conclusions.
6.2. Limitations
This work was done on clear, tempered 60 by 120 by 1 in. (1.52 by 3.05 m by 25.4 mm)
insulating glazing units to determine the durability of the secondary and primary seals
through the ASTM E2190-10 [7] and ASTM E2188-10 [6] testing protocol. The data
generated targeted a specific sized unit, a specific spacer system design and a specific
aspect ratio. It is unknown if this data applies to every case of IG size, glass make up
and spacer system, and it is unlikely that it does. Before a project is undertaken
mimicking this type of cold-bending, this testing should be repeated using actual sizes
and anticipated bending dimensions. This is the first data generated on this topic of
durability with cold-bending, and is likely just the beginning of many research projects
on this topic. Cold-bending of IGU’s is easiest when the units are attached to the
glazing frame using structural silicone. It may be specified that cold-bent IGU’s are to
be used in a mechanically attached curtainwall system, but the structural silicone used to
attach an IGU to a frame is a combination adhesive and sealant retarding air and water
infiltration. Mechanically held cold-bent IGU’s may put undue stresses on gasket
systems that result in unwanted air and water infiltration. Due to the likeliness of
structural silicone attachment of cold-bent glass the durability testing was done to the
ASTM E2190-10 protocol which uses a UV cycle in the test. This protocol is not the
same as the EN1279 Glass In Building – Insulating Glass Units [18] protocol used in
Europe. It is unknown if the two protocols produce similar results.
813
Challenging Glass 3
6.3. Future Work

The breakage of the small insulating glass units in the thermal cycling is most likely due
to the way the assemblies containing the test specimens were fabricated. This was the
first attempt and the epoxy anchorage method appeared to have the best chance for
success. Additional testing of small IG units to the protocol in a deflected position is
the most challenging, yet the protocol is the basis for specifications of pass or fail. In
hindsight, the glass in the small test units should have been increased to a 3/8 in. (9 mm)
thickness in lieu of the thinner glass suggested by the specification. There was also a
consideration regarding the thickness of the assembly to be able to fit into the test
chamber. Finally the epoxy that was used to displace the glass in the Z direction should
have been replaced with an extremely high strength silicone adhesive such as the one
also presented in this symposium. [19]
Additionally, future work needs to be done to more strongly correlate the amount of
bending in a full size panel to the exact amount of edge seal strain in an IGU. The
sample size of the full scale test was too small (one panel; assembly #2) to properly
conclude that the induced edge seal strains are exactly equal to those used in the small
scale durability test. Also, this research has not correlated other factors to edge seal
strain such as: glass aspect ratio, framing member section properties, glass thickness or
makeup, spacer design, etc.
Future research would refine the testing methods as laid out in this project by
reexamining the design of the small unit displacement apparatus to lower the induced
stress on the glass and reduce the risk of thermal breakage, to procure glass with a
higher resistance to thermal fracture (thicker glass or higher edge strength), to improve
the method of application of “z” displacement so that it survives the weathering process,
and to anticipate thermal breakage and procure more specimens to complete the testing.
Future projects that consider cold-bend IG unit should indeed have a level of
understanding whether or not the bent glass will indeed hold up to the durability
standards. This knowledge must come from a study of the particular variables present in
such a project. As mentioned above, the affects of glass makeup, frame behavior,
spacer design, silicone size, or unit typology and geometry could drastically affect
durability as well as other factors that were not examined in this study such as glass
stress, silicone stress or aesthetic appeal.
814
7. Acknowledgements
The authors would like to specifically thank the personnel at Architectural Testing Inc.
in St Paul Minnesota for the willingness to customize the testing to support this project.
Specifically Dan Johnson, Eric Schoenthaler, Brian Goetzke, Mark Lewke, Michael
Resech and Zane Wybest at ATI made this work possible.
The authors have collaborated with each other in the spirit of advancing the
understanding the physical attributes and science behind cold-bent glass. The
collaboration was necessary as each party brought their own industry experience to the
project. It is not likely that a single industry entity could produce this type of work. For
this the authors wish to thank their employers for the time and expense that this work
has required all for the spirit of industry advancement.
8. Disclaimer
This publication is for informational purposes only. None of the parties involved in this
research make any representations that this data is appropriate for use in other
applications. Any party electing to use any of the data contained herein assumes all
risks related to such use. To the extent permitted by applicable law, no liability
whatsoever is accepted by any of the parties participating in this research for any direct
or consequential loss, damage, costs or prejudices whatsoever arising from the use of
this publication or its contents.
815
Challenging Glass 3
9. References
[1] Eekhout, M., Lockefeer, W., and Staaks, D. , 2007, “Application of Cold Twisted Tempered Glass
Panels in Double Curved Architectural Designs”, Glass Performance Days 2007, pp. 213-220.
[2] Eekhout, M. and Niderehe, S., 2009, “The New, Cold Bent Glass Roof of the Victoria & Albert Museum,
London”, Glass Performance Days 2009, pp. 408-412.
[3] Dodd, G. and Thieme, S. 2007, “Comparison of Curved Glass and Cold Bent Panels”, Glass
Performance Days 2007, pp. 83-86.
[4] van Herwijnen, F., Staaks, D., and Eekhout, M., 2004, “Cold Bent Glass Sheets in Façade Structures”,
SEI Volume 14, Number 2/2004, pp. 98-101.
[5] Belis, J., Inghelbrecht, B., Van Impe, R., and Callewaert, D., 2007, “Experimental Assessment of Cold-
Bent Glass Panels”, Glass Performance Days 2007, pp. 115-117.
[6] ASTM Standard E2188-10: Standard Test Method for Insulating Glass Unit Performance, Annual Book
of ASTM Standards, ASTM International, West Conshohocken, PA, 2010.
[7] ASTM Standard E2190-10: Standard Specification for Insulating Glass Unit Performance and
Evaluation, Annual Book of ASTM Standards, ASTM International, West Conshohocken, PA, 2010.
[8] ASTM Standard E1300-09a: Standard Practice for Determining Load Resistance of Glass in Buildings,
Annual Book of ASTM Standards, ASTM International, West Conshohocken, PA, 2009.
[9] Glass Association of North America, 2010, “GANA Glazing Manual - 50th Anniversary Edition”, pp.
55-60.
[10] ASTM Standard E330-02: Standard Test Method for Structural Performance of Exterior Windows,
Doors, Skylights and Curtain Walls by Uniform Static Air Pressure Difference, Annual Book of ASTM
Standards, ASTM International, West Conshohocken, PA, 2002.
[11] ASTM Standard C1184-05: Standard Specification for Structural Silicone Sealants, Annual Book of
ASTM Standards, ASTM International, West Conshohocken, PA, 2005.
[12] ASTM Standard C1369-07: Standard Specification for Secondary Edge Sealants for Structurally Glazed
Insulating Glass Units, Annual Book of ASTM Standards, ASTM International, West Conshohocken, PA,
2007.
[13] Wolf, A. T., and Cleland-Host, H. L., “Material Properties for Use in FEA Modeling: Sealant Behavior
with Ambient Laboratory Climate Aging,” Durability of Building and Construction Sealants and
Adhesives, ASTM STP 1453, A. T. Wolf, Ed., ASTM International, West Conshohocken, PA, 2004.
[14] ASTM Standard C1265-94(2005)e1: Standard Test Method for Determining the Tensile Properties of an
Insulating Glass Edge Seal for Structural Glazing Applications, Annual Book of ASTM Standards,
ASTM International, West Conshohocken, PA, 2005.
[15] ASTM Standard E2649-09: Standard Test Method for Determining Argon Concentration in Sealed
Insulating Glass Units Using Spark Emission Spectroscopy, Annual Book of ASTM Standards, ASTM
International, West Conshohocken, PA, 2009.
[16] ASTM Standard E546-08: Standard Test Method for Frost/Dew Point of Sealed Insulating Glass Units,
Annual Book of ASTM Standards, ASTM International, West Conshohocken, PA, 2008.
[17] Wempner, G., and D. Talaslidis, D., Mechanics of Solids and Shells: Theories and Approximations,
CRC Press 2003
[18] EN 1279 Glass in Building – Insulating Glass Units: Part 2 Long term test method and requirements for
moisture penetration. Part 3: Long term test method and requirements for gas leakage rate and for gas
concentration tolerances European Committee for Standardization: rue de Stassart, 36 B-1050 Brussels,
2002
[19] Wolf, A.T, Sitte, S., Brasseur, M., J., and Carbary L. D, “Preliminary Evaluation of the Mechanical
Properties and Durability of Transparent Structural Silicone Adhesive (TSSA) for Point-Fixing in
Glazing” Fourth International Symposium on Durability of Building and Construction Sealants and
Adhesives, Journal of ASTM International, published online August 2011, Volume 8, Issue 10
(November 2011), JAI 104084, available
www.astm.org/DIGITAL_LIBRARY/JOURNALS/JAI/PAGES/JAI104084.htm.
816
10. Tables
817
Challenging Glass 3
818
doi:10.3233/978-1-61499-061-1-819
Application of Thermally Curved Glass in

the Building Industry
Michael Elstner
Interpane Glasgesellschaft mbH, Germany, michael.elstner@interpane.com
Markus Kramer
IB Kramer Tragwerksplanung, Germany, info@ib-kramer.de
Curved, or bent, glass units are used more and more often in modern architecture.
The curved elements can create flowing contours, organic bodies, or bold arch and
vault constructions. Technical developments in the manufacture and refinement of
glass are constantly attaining new heights. Necessary, for this reason, is not only a
scientific examination of this product but also a detailed description of its
application. Since there presently exist no – or very few –standards, regulations, or
guidelines bearing on this product, the Bundesverband Flachglas has drawn up a
manual for thermally-curved glass which may be considered as a brief work dealing
with the fundaments of this topic. There are described here both the boundary
conditions in terms of German building legislation and the particular features of the
dimensioning of this type of glass. In addition to this, advice and pointers are given
regarding possible tolerances and assembly/installation. Curved glass (made from
float glass, fully TG, heat-soaked fully TG or LSG) may, at present, only be applied
and installed in Germany on the basis of an Exceptional Approval (Zustimmung im
Einzelfall, or ZiE) or of an already-acquired National Technical Approval
(Allgemeine bauaufsichtliche Zulassung, or AbZ). Also explained are the structural
dimensioning of curved glass, which differ fundamentally from those applying to
flat sheets of glass. The manual provides here, among other information, general
strength- and resistance-values for guidance in pre-dimensioning.
Keywords: Application, curved glass, dimensioning of glass, product standards,

tolerances, climatic pressures, blocking
1. Introduction
The modern urban landscape is dominated by buildings with large glass facades. The
spirit of contemporary architecture is not, however, one which demands only cubic
building-geometries involving flat, straight surfaces. Rather, architecture today
comprises curved elements as well, which serve to create flowing contours, organic
bodies, or bold arch and vault constructions. It is in the creation of such structures that
curved or bent glass units find their application. Technical developments in the
manufacture and refinement of glass are constantly attaining new heights, thus making
possible the manufacture of this technically demanding building product. Increasing use
of curved-glass glazing in building envelopes means that this type of glass must also
fulfil requirements in respect of such properties as heat-insulation, solar control,
acoustic insulation, and burglar resistance. Available as curved-glass glazing units are
both monolithic float glass and thermally or chemically toughened elements in the form
of single glazing, laminated glass, laminated safety glass, or insulating glass units. The
necessity to deal in greater detail and precision with the special properties of this
819
Challenging Glass 3
building product also becomes evident, however, in the fact of its displaying a different
load-bearing behaviour, and various other different material characteristics, than those
displayed by flat glass. It was in view of these considerations, as well as of the fact that
there presently exist no (or very few) standards, regulations, or guidelines bearing on
this building product, that the Bundesverband Flachglas initiated, in 2009, the working
group “Curved Glass”. The goal of this working group was to draw up a manual [1]
which would provide answers to the most important questions likely to be put by the
user – for example, a planner or window and façade manufacturer– of this product. This
manual was then indeed – in June 2011 and in collaboration with, among other firms,
the firms Finiglas, Flintermann, Döring Glas, Glas Freericks and Glas Trösch –
successfully drawn up. The manual is divided into various sections, some of the main
themes and topics of which we introduce here. Most especially, we deal in some detail
with the topics: the area of the manual’s applicability; the significance of construction
legislation for curved glazing units; manufacturing tolerances, including visual quality,
along with correct blocking and dimensioning of curved-glass glazing units. It is
necessary above all for planners and architects to concern themselves in detail with the
specific characteristics of building products and of their application. Thus it is stipulated
in the official German Regulations for Civil Construction Contracts
(Verdingungsordnung für Bauleistungen, or VOB) that a service or performance must
be clearly and thoroughly described. This means that it is necessary not only to explain
and justify the feasibility of the product in question but also to clarify the way in which
it fits into the framework of existing construction regulations. In addition to this,
indication must also be given of the relevant technical values.
Figure 1: © archimation for Studio Daniel Libeskind – KoeBogen, Duesseldorf
820
Application of Thermally Curved Glass in the Building Industry
2. Area of Application
The manual’s area of application comprises all possible uses of the product both in and
around the building. This means that there is described in it the product’s use both in the
façade and in the interior fittings. Such specialized applications as that in cabinetry and
furniture construction, or yacht and ship glazing, cannot be described within the
framework of a general manual, due to the special demands made, in such cases, on the
technical and optical qualities of the product. Likewise left undescribed are “cold-
formed” or “laminate-curved” types of glazing. These glass-types represent a discipline
in the manufacture of curved and bent glazing that is still very new. It remains at present
the preserve of just a few specialized companies. Due to the small amount of experience
with these procedures hitherto acquired, and their significantly greater complexity, a
general description, in a manual, of said procedures and said glass-types is likewise not
possible.
3. Building Physics
The Federal German Energy Saving Ordinance (EnEV) which represents the
implementation within Germany of the EU’s “Energy Performance of Buildings”
(EPBD) Directive, sets requirements in respect both of thermal insulation and of
thermal protection in summer for the window and façade parts of buildings. In cases
where, besides flat glazing, curved glass is also used, this latter must also fulfil the
corresponding requirements. Very often necessary, in order to do this, are functional
coatings designed to reduce Ug values but also to ensure thermal protection in summer.
Important particularly in the case of solar-control coatings are, besides the functional
requirements, also the aesthetic ones. A clear building-physics description of a glazing
unit – including one made of curved glass – must therefore comprise, at the very least,
the following pieces of information:
x Thermal transmittance coefficient - Ug value

x Total solar energy transmittance coefficient - g value
x Light transmittance - TL
x External light reflectance - RLa
x Colour rendering index - Ra,D

x Selectivity Coefficient - S
x Weighted sound-reduction index – RW
Which values are possible in the case of a curved-glass unit depends, among other
factors, which coatings can be used in curved glass. This in turn depends on the
geometry, the glass configuration, the size etc. There must in every case here be
consultation and coordination with the manufacturer. It should be taken into account
that the optical qualities and technical values that can be satisfied by flat glazing units
cannot always be satisfied where such glazing units are used within a building together
with curved-glass units. This means that it may well prove necessary to execute the flat
glazing in a building with the products of one manufacturer and the curved glazing with
the products of another. The curved glazing will, then, for example, display other values
in terms of for example light transmittance and g value. With regard to the visual quality
of a piece of glazing, the ideal way of forming an impression is through inspecting
samples of the product one intends to use, including its coating. These should always be
821
Challenging Glass 3
displayed, together with the products intended to be used for the flat glazing, at the
premises/offices of the building contractor and architect.
4. Tolerances
The factor “tolerances” is one of greater importance in the case of curved-glass glazing
units than in that of almost any other building product, since curved glass is mostly
custom-manufactured and not mass-produced. This means that, even if the basis of a
curved-glass manufacture is a batch of glass sheets with identical glass configuration,
these sheets can display different behaviors during the curving process that creates an
individual processing of the product. Thus, the range of possible quality between these
sheets can be a wide one. The manual for curved glass works out and establishes, for the
first time, with various manufacturers tolerances for this form of glass by constant
reference to the product standards for flat glass. Denominations here were chosen in a
manner accordant with ISO 11485 Part 1. One tolerance that is extremely important for
this product is “shape accuracy” (PC) and “twist deviation” (V), since these describe the
geometry of the curved glazing unit as a whole and provide the user with the indications
that are necessary for the planning of the construction. This is intended to ensure that no
restrains are introduced into the glass by the design and construction process itself.
Curved glass, due to its specific form / geometry, tends to react more sensible to
restrains, and any additional mechanical stress that is placed on it can lead to breakage.
Figure 4a: Shape Accuracy (PC) Figure 4b: Twist Deviation (V)
Besides “shape accuracy” and “twist deviation”, “arc”, “edge straingthness deviation”
and “edge offset” are all important parameters when attempting to acquire an
impression of the accuracy of curved glass and thereby of its likely effects on the whole
of the construction in which it is to be installed. Regarding tolerances for spherically
curved glass, the user will need to consult with the manufacturers in each individual
case. Particularly in cases where curvature of this latter sort is in question, it is
recommended that samples be prepared during the planning phase and as part of a
feasibility study so as to be able to acquire an impression of possible tolerances and of
the visual quality. In view of what has just been said, it should be borne in mind that, in
the case of curved glass, the window- and façade-constructions used must always be
those executed with the wet-sealing method, so as to ensure that the glass is not
subjected to stress and strain arising from the substructure.
822
This means that quality consists in the sum of the following factors
Geometry/ shape (tolerances)

Type of glass

¶

Dimensions
Complexity of glass configuration
Sequence of coatings/printings
Manufacturing process
Process parameters
Figure 5: Parameters for the Quality of Curved Glass
The manual itself [1] can be consulted for the numerical values for these tolerances and
their graphic presentation. In the case of very complex façade geometries and glazing
designs, it may be necessary to execute feasibility studies before proceeding to any call
for tenders. Such studies may be very important in securing a degree of certainty for the
further planning process and thereby a certain fixed latitude for planning of time and
costs. At this early stage it may still be possible to carry out, if necessary, optimizations
of the basic design.
5. Dimensioning of Curved Glass

Due to its “shell-bearing action”, or “vault effect”, curved glass displays an entirely
different load-bearing behaviour from flat glass panels. This must be taken into account
not only in the blocking but also in the dimensioning of curved glass panels. To this end,
it is absolutely necessary that simulations be performed using appropriate finite-element
software which are capable of accurately reflecting the curvature of the glass panels in
question while also taking into account the specific bearing conditions. The precise
reflection, within the model, of these determining conditions resulting from the
positioning (rigidity of substructure) is extremely important, since even slight
deformations tend to have significantly greater effects on curved panes than they do on
flat ones.
5.1. Shell load-bearing capacity of curved glass

When establishing the necessary glass thickness, the stiffening effect of the shell-form
can bring about a significant reduction of said thickness, especially in the case of single-
glazed glass assemblies and correspondent positioning. This is made clear in the
example demonstrated below (see Fig. 7). This comparison involves exposing a single-
glazing pane, supported at two of its edges – but with different curve radii – to a
constant wind-load, in the form of a pressure load, on its convex side. The flexural
tensions and deflections which result from this are represented as specific relative
magnitudes on a flat pane of the same dimensions. The shell load-bearing capacity can
be clearly observed here. In the case of a curve radius of two metres, for example, the
deflection amounts to only 5 per cent in comparison to a flat pane of the same size,
while the principal tensile stress amounts to only 16 per cent. Fig. 6 shows the stress
distribution. In this case, maximum stress is located at the exposed edges of the pane.
823
Challenging Glass 3
Figure 6: Distribution of Stress on a Curved Pane – Radius: 2 m
100%
90%
80% related deformation
70%
60% related stress
50%
40%
30%
20%
10%
0%
R=10 m
R=15 m
flat pane
R=1 m
R=2 m
R=5 m
Figure 7: Comparison of Deformation / Stress in the Cases of Different Curve Radii
5.2. Climatic Pressures

As positive as the stiffening effect exerted in curved single glazing may be, it should
still not remain unmentioned that, where the glazing installed takes the form of
insulating glass units (IGU), relatively powerful inner forces arise as a result of climatic
changes. Due to the climatically-induced greater internal stresses within the curved
insulating glass unit the advantage accruing from the stiffening effect of the curved
single-glazing pane is partially depleted. Generally speaking, the advantages arising
824
from the shell load-bearing capacity of the curved single-glazing sheets tend not to be as
great in cases where these sheets are applied as component parts of “IGU” as in cases
where they are employed directly as single glazing. Greater climatically-induced inner
stresses tend to have negative effects on the likely service life / period of freedom from
condensation of insulating glass units. This is generally the case regardless of whether
the insulating-glass panels in question are supported at two, at three, or even at all four
of their edges.
Climatic pressures, in particular, must in no case be established using the procedures

described in [4] since these derive from plate theory as this applies to flat panes of glass.
Fig. 8 shows the changes in climatic pressure that occur where the glass configuration
and the pane surface remain the same but the curve radii differ. The glass panel with the
curve radius of 1 m is taken here as the 100 per cent “climatic dummy load”. The flat
glass panel – of identical size and with identical glass configuration – displays only 2
per cent of the climatic loads displayed by the curved glazing.
120%
100%
related equivalent
80% area loads
60%
40%
20%
0%
R=10 m
R=20 m
R=1 m
R=2 m
R=5 m
flat pane
Figure 8: Comparison of “Climatic Dummy Loads” Applied to Different Curve Radii
5.3. Bases For Calculation

Series of research projects have shown that curved glass planes display different
characteristic bending tensile strengths from flat planes. This fact must be taken into
account during dimensioning. As explained above there are, at present, no product
standards extant for curved glass units, which means that there are also no characteristic
bending tensile strengths. Characteristic bending tensile strengths are, at present, fixed
only for curved-glass units which are subject to a particular National Technical
Approval of an individual manufacturer. In the manual [1] bending tensile strengths are
given which have been derived from the research studies [2] and which are to be
understood as benchmark figures for the pre-design and preliminary dimensioning
process.
Table 1 shows the comparison of the characteristic bending tensile strengths of flat and
curved glass.
825
Challenging Glass 3
Table 1: Comparison of Bending Tensile Strengths

Characteristic Bending Tensile Strength in Glass Area [MPa]
Type of Glass
flat curved differential
Float glass anneleaed 45 40 0%
Heat-strengthened float glass 75 55 73,4 %
Toughened float glass 120 105 87,5 %
In order for a piece of glass to be considered “fit for use” the following must be
demonstrated: that, despite deformations and the shortening of the chords associated
with it, there is no possibility of the pane’s slipping out of the sill boards. The values
mentioned in table 1 are reduced on the edges. So the final characteristic bending
strength will only be about 80 % of that for the glass area.
6. Blocking
The basic principles governing the blocking apply to curved glass just as they do to flat
glass and are described [3]. The blocking must conduct the dead load of the glazing unit
safely into the substructure. The glazing units do not, as a rule, take on the task of
bearing loads originating within the structure. In the case where it is intended that the
glazing units should indeed take on the function of bearing loads or weights originating
in the structure, this must be taken into account already at the stage of planning out the
basic structural-engineering principles that will apply in the given project. The glass
manufacturer or system provider should also be contacted and consulted here. Blocking
is a task which belongs rightly to the planning stage and should be carried out before the
execution of the assembly work. Curved single glazing or IGU must, where they are
installed in a vertical position, have their glazing blocks placed in the same way as flat
panes. For example, in the case of spherical curving, or profiles inserted into the edge
seal, or of application in glass construction engineering consultation with the
manufacturer will always be necessary. For curved glazing units the following
recommendations in respect of glazing block placement are also made: the placing of
setting blocks must be carried out in such a manner that the glazing unit is in a state of
equilibrium and cannot tip over. This means that the setting blocks must be arranged in
such a way that a line connecting the central points of the two glazing blocks cuts across
the line of the gravity center of the glazing. At this point of maximum load the dead
load of the glazing unit is conducted away into the surrounding structure. Position is
dependent on geometry, size, and glass configuration (see Fig. 9). The position of the
setting blocks must be taken into account in the dimensioning of the substructure. Their
distance from the corner of the glass must correspond to the regular distance of
minimum100 mm. Recommendations for the execution of blocking for curved glass are
to be found in [1].
826
Figure 9: Establishing the Gravity Center in a Curved Glazing Unit
7. Summary
It is advisable, already at the stage of establishing the proposed architecture of a
building and its building envelope, to make enquiries among manufacturers regarding
the possibilities inherent in the building product “curved glass”. This would involve,
among other things, finding out the different types of glass that might play a role in the
execution of the building design: cylindrically or spherically curved float glass; fully
toughened glass; laminated and laminated safety glass; or insulating glass units
manufactured from one of the above glass-types. Possibilities – with regard, for
example, to tolerances, visual quality, and building physic properties – known from
experience with flat glass cannot be carried over “1 to 1” onto curved glass. The manual
on thermally curved glass, developed in collaboration with several different
manufacturers, offers an initial orientation here with regard to the use of curved glazing.
At the same time, however, it cannot provide answers for all of the questions which
arise in relation to the use of this complex building product. For this reason, where any
doubt exists, the manufacturers, or the relevant specialized planners, should be

contacted and consulted.
827
Challenging Glass 3
8. References
[1] Manual 009/2011, Bundesverband Flachglas, Guidelines for Thermally-Curved Glass in the Building
Industry,
[2] Bucak, O., Feldmann, M., Kasper, R., Bues, M.Illguth, M.: Das Bauprodukt „warm gebogenes Glas“ –
Prüfverfahren, Festigkeiten und Qualitätssicherung. (“The Building Product ‘Thermally Curved Glass’
– Testing Procedures, Strengths and Resistances, and Quality Assurance”). special edition (2009) of
Stahlbau on “Structural Building With Glass”, pps. 23 – 28
[3] Technical Guidelines of the Glaziers’ Guild, No. 3 “Blocking of Glazing Units”, Verlagsanstalt
Handwerk GmbH , Düsseldorf 7th edition, 2009
[4] TRLV 2006-08, DIBt, Technical Regulations for the Use of Linear-Supported Glazing
[5] TRPV 2006-08, DIBt, Technical Regulations for the Measurement and Execution of Point-Supported
Glazing
[6] TRAV 2003-01, DIBt, Technical Regulations for the Use of Barrier Glazing
[7] Elstner, Schäfer, Gebogene Gläser in der Fassade (“Curved Glass Units in the Façade”) Glas +
Rahmen 10-2010
[8] Technical Guidelines of the Glaziers’ Guild, No. 8: “Safety of Passage In Public Buildings and
Facilities Using Glass” Verlagsanstalt Handwerk GmbH , Düsseldorf 2nd edition, 2006
[9] GLASBAU 2012, “The use of thermally curves glass in buildings” Scheideler, Broich, Elstner,
828
doi:10.3233/978-1-61499-061-1-829
Cold-Bent Single Curved Glass;

Opportunities and Challenges
in Freeform Facades
Mark Feijen
i-Saac, the Netherlands, mark@i-saac.nl
Ivo Vrouwe
Sint-Lucas School of Architecture, Belgium, info@ivovrouwe.net
Peter Thun
BRS, the Netherlands, p.thun@brs.nl
The use of freeform geometry in architectural design necessitates a progressive

building industry to be materialized. Great progress has been made in shaping
materials into double curved surfaces by moulds and milling techniques. However,
using these techniques in a material and energy efficient way proves to be a great
challenge. Current solutions often result in less affordable techniques. In this paper
the opportunities and challenges of cold-bent single curved glass are discussed. In
comparison to most design industries, the architectural industry often constructs
large surfaces with relatively little curvature. By subdividing these double-curved
surfaces into single-curved parts, the initial design is translated into a specific
material efficient geometry. After translation, the geometry can be materialized in
Cold-Bent Glass. In most cases these techniques can compete well with preformed
glass solutions.
Keywords: Cold-Bent, Glass, Geometry, Freeform
1. Introduction
In the contemporary building industry, an interesting dialogue is conducted. On the one

hand, designing architects push the industry forward to develop techniques and
materials to be able to materialize their progressive digitally driven designs. On the
other hand, interesting materials and technologies are generated to assist forthcoming
demand with sufficient techniques.
When, in design, material reflection takes place at the end of the process, one can speak
of a top-down or non-deterministic [1] approach. In this case the design forms the start
point of the engineering process, materialization comes second. In conventional
architecture, this routine is quite common and successful. In more advanced geometries,
the translation from the digital model to the physical one, often behaves less fluently.
The top-down approach takes a broad-minded building and engineering partner as well
as an experienced architect to provide an appropriate result. When one of these parties
don’t line up, the end result can differ greatly from the initial form or idea.
829
Challenging Glass 3
The bottom-up or deterministic [1] development, requires an industry that is skilled in

communicating design orientated information and an architect that is capable of
working within material and geometrical parameters. In most cases, the success rate of a
proper materialization is higher than in the first case.
In this paper technical information is presented to assist architects and designers in

understanding the use of curved surfaces, materialized in glass. In the first part of the
paper, the basic concept of freeform glass is discussed. In the middle part a parameter
study of a specific case is described. In the last part of the paper a well managed top-
down design approach for a double-curved glass façade is presented.
2. Cold-Bent Single Curved Glass

The concept of “cold” bending of flat panels is a well know concept in industry in
general. Materialization can range from plastics to wood and metals as shown in fig.1.
Figure 1a,b and c: “cold” bent Plexiglas in Bio Solar Haus in Germany; “cold” bent laminated wooden panels
for stage setting of Barbara van Loon; “cold” bent corrugated steel panels (partly) in sheep shaped artwork in
New Zealand
In cold bent materialization, glass does not seem to be the most obvious choice. The
general perception of glass is still a brittle material with hardly any strength capabilities.
The perception of glass as a brittle material is correct. The perception of glass having
little strength is generally based on a misunderstanding. Over the last 20 years, an
increasing amount of project and experiments were conducted with glass as main
carrying component. To support this development, codes and standards [2] have been
developed on maximum allowable tensile bending strength of glass.
Supported by FEM-Software and these well defined codes and standards at hand,
investigation of material’s boundaries is more accessible. A study of cold bending of
glass is one of the possibilities to support freeform architecture with a transparent
materialization.
2.1. Basic concept

The concept of Cold bent glass is based on the pre-tensioned state of tempered glass.
Simplified to a 2D problem, the bending stresses are superposed on the pretension state
of the glass as presented in figure 2.
830
Cold-Bent Single Curved Glass - Opportunities and Challenges in Freeform Facades
Figure 2: Superposing pretensions stresses and bending stresses
Theoretically the cold bending stresses can compensate the pretension stresses of the
tempered glass entirely. However, bended to this limit won’t allow for additional
loading as required by the codes and standards.
Due to cold bending, the stresses can be calculated by the formula:
E t
V (1)
2R
Formula description: ; bending stress, E; modulus of elasticity of glass, t; panel thickness, R; bending radius
Geometrically a perfect radius can be created by applying a constant bending moment

on two opposite sides of the cold bent plate. The static scheme of this approach is
shown in figure 3a.
Figure 3a and b: left the static scheme for a perfect radius, right the approximation of a radius
In practice, it is difficult to apply a constant bending moment on a plate. By extending

the plate on both sides of the supports first, creating a cantilever, and applying a force
on both ends next, distance is introduced to apply the load (figure 3b). Doing so, the
middle part of the plate will describe a perfect radius. Both ends will have deformed like
a cantilevered beam with an end load.
The approximated approach does not differ vastly from a perfect radius as presented in
figure 4. The blue line represents a perfect radius. The orange line represents the
deformation of both ends of the plate; cantilever deformation.
831
Challenging Glass 3
Figure 4: deviation of a circle (2D approximation of a 3.0x1.5 m panel with thickness of 6 mm)
2.2. Parameter Study

The simplified structural approach, as described above, provides a sufficient description
of the bending process. However, the study ignores the influence of the width of the
cold bent glass panel. To obtain a better understanding and a solid exploration of the
influence of the different parameters and limitations of cold bending, a parameter study
has been performed. The study involves parameters like panel width, thickness and
radius.
The study is based on a finite element analysis performed with the non linear FEM
software ABAQUS. Cold bending of the glass panel is simulated by means of a contact
simulation. This approach results in a better understanding of stress distribution in the
panel during bending along a curved edges first, and the stress state after bending
second. Sideways, the contact simulation produces contact stresses between the glass
and the backing structure.
In figure 5a and b, the typical geometry of a model is presented. The model consists of a
quarter of a glass panel, plotted in light blue, with symmetric boundary conditions. The
curved backing structure, guiding the plate deformation, is a rigid part, plotted in green.
The red curve element is a rigid element that assists deformation of the panel with the
applied angle (figure 5b; hoek ).
Figure 5a and b: typical geometry of model for contact simulation and applied deformation.
832
The start dimension of the model is a panel of 3500 x 1500 x 6 mm. The model is based
on the panel dimension of a realized case study. This case is not presented in this paper.
In the parameter study the radius varies from 3.5 to 6, 9 and 12 meters, the thickness
varies from 4, 6, 8 to 10 mm and the width is varies from 500 to 3500 mm in steps of
500 mm. For this case, the length of the plate was a fixed value.
A characteristic result of bending stresses in the plate model is presented in figure 6. A

colour contour of the stresses in tangential direction is plotted (top fiber stresses). A
stress peak in the plate occurs near the described structural contact egde.
Figure 6: Typical bending stress (stresses in tangential direction) of a cold bent panel in end situation.
Panel size: 3000x1500x6 mm, bending radius 3.5 meter
The variation of the radius and thickness results in relatively easily scalable stress
results. The results follow the standard stress and strain relations between bending
radius and thickness as expected acoording to formula (1).
Less predictable are the results of the variation of width. The panel is only kept in shape
along the two longitudinal edges. In between these two edges the panel is
unconstrainted and will tend to flatten out. Over a certain length to width ratio this
effect will be visible in the reflection of the glass. This flattening effect is clearly visible
when one compares the stress results in different longitudinal panel sections of two
separate panels with a different width (graphs in figure 7).
In the graph in figure 7 the end situation of a cold bent panel is shown for a 3500x 500 x
6 mm panel (A) and a 3500 x 3500 x 6 mm panel (B). The bending radius is 3.5 meter.
The smooth lines represent the section along the symmetry line of the panel (the exact
middle) and the lines with the“peak” stresses are a along the edge of the panels.
For panel (A) the stress in both section lines behave much alike (except for the peak
stress), whereas the behaviour of the lines of panel (B) differs significantly. This large
difference in the stresses is clearly visible in the deformation plots.
833
Challenging Glass 3
90
80
70
Topfiber stress N/mm²
60
50
40
30
20
10
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
True distance along path
Panel (A); panel edge Panel (A); Symm side Panel (B); Panel Edge Panel (B); symm side
Figure 7: End situation of cold bent panels. Panel (A) is 3000x500x 6 mm and Panel (B) is 3000x3500x6 mm.
The bending stress at the top fiber of the panels are shown along the model at the contact edge (with the peak
stress) and along symmetry axis side (smooth curve).
The exact boundary between acceptable and inacceptable deformation difference is

difficult to determin. A practical guideline is width/length ratio of 1/2.
The peak stresses in the longitudinal edge sections occurs nearby the contact area
between the glass and the backing structure. The contact location indicates a similar
“cantilever” effect as described in section 2.1 (figure 2 and 3).
The expected cold bending stress based on formula (1) for these panels is 60 N/mm2. If
the peak stresses found in the finite element results are related to the expected cold
bending stress in a graph as presented in figure 9, the peak stress values can be
predicted for a given panel width. This assumption creates a procedure to determine the
maximum stress in the panel due to cold bending.
Figure 9: Peak stress factor in relation to the width of the panel (Length 3.5 m, radius 3.5 m)
834
3. New Headquarters of the Council of the European Union

The study for the new headquarters of the Council of the European Union in Brussels, a
design by Samyn and Partners, is an example of a non-deterministic colt bent glass
materialization.
The initial design was made by double curved glass panels. By describing alternatives
in cold-bent glass, different options were compared.
Figure 10a and b: Design Geometry and Surface Curve Analysis
The geometry of the initial design is described by connecting thirteen elliptic forms. The
ratio of length to width varied per segment. Every ellipse was elevated by 1.18 meter
per segment. Next, every elevation was filled in with three horizontal lines of curved
glass segments.
By using custom scripts, every panel within the geometry is measured in radius. Panels
with a bending radius of 12 meters and more, displayed in figure 10b in purple, were
assumed to be realized in cold-bent glass. Panels curved in a radius of 12 meters or less,
plotted in red, were assumed to be critical for cold bent insulated glass and had to be
realized in an alternative way.
Figure 11a, b and c: Design Geometry and Surface Curve Analysis
Due to the limited curve differences and a small horizontal shift from one ellipse to the
next, every horizontal section strip was conceived as a straight tubular cut. Unrolling the
section provided a flat single curved panel in projected view. By subdividing the
unrolled section, the individual panels were generated.
835
Challenging Glass 3
To be able to cut cutting costs, several panels within the geometry were compared to
elements with straight instead of curved outlines. Deviations of 5 millimeter or less
could be assimilated within the connection detail. A small selection of parts behaved
within the parameters of this group.
Figure 12a, b and c: The initial design, a quarter of the surface geometry and the section described by single
curved panels. (3a Computer rendering polygon graphics 2008 by Philippe Samyn and Partners architects &
engineers, Lead and Design Partners)
4. Conclusion and Further Research

Using single curved cold bent glass for double curved architectural envelopes shows
great potential. Evolving double curved surfaces to material efficient, single curved
geometries, proves to be a less expensive alternative. Due to large surfaces used in
architecture, visual distortion due to faceted construction is reduced to a minimum.
Architecturally, further research is needed in material connection in various angles. In
further studies, mockups will be used to explore different connection types and
principles.
Structurally, determining the limitations of cold bent glass is a work in progress. The
scope of the study for the near future will be a further investigating of the behavior of
layer panels of glass (laminated and insulate). The aim of these upcoming research steps
is to develop practical and time efficient calculation methods for cold bent laminated
and insulated glass. Mockups will be use to explore boundary conditions of cold bent
glass curvature. Hereby, different techniques of curving the glass will be tested.
5. References
[1] Hanafin, Stuart; Pitts, Greg, Non-Deterministic Exploration through Parametric Design, international
journal of architectural computing, issue 4, volume 7, 2009.
[2] NEN 2608, 2nd design code januari 2011; Glass in building – Requirements and determination method.
[3] Belis,J. , B.Inghelbrecht, R. Van Impe, D.Callewaert ‘Cold Bending of Laminated Glass panels;
HERON Vol. 52 2007 No,1/2
836
doi:10.3233/978-1-61499-061-1-837
Quality Control of Bent Heat Strengthened

and Fully Tempered Glass by
the Application of Photoelasticity
Markus Feldmann, Pietro Di Biase, Ruth Kasper
RWTH Aachen University, Institute of Steel Construction, Germany,
www.stb.rwth-aachen.de
Irregularities of the prestress distribution have been detected in some products of

bent heat strengthened resp. fully tempered glass [1]. However, until now, a basis
for the interpretation of these irregularities has not been found. A reason for this
may be the lack of knowledge on the correlation of the visualized colours from
polarization filters and the state of stress. Therefore, the product quality of bent
glass was investigated by visualizing the distribution of the differences of
orthogonal stress integrals. This combined with the visualized colours gives a
deeper insight in the resulting fracture patterns of glass. The results shall also serve
for a future online quality monitoring system.
Keywords: Glass, heat strengthened glass, fully tempered glass, bent glass,
photoelasticity, fracture pattern, thermal prestress, orthogonal stress integral
1. General
Irregularities of the prestress have been detected in some products of bent heat
strengthened or fully tempered glasses by the use of polarizing filters [1]. Thereby in
some cases, correlations between inhomogeneous distributions of prestress and non-
conforming fracture patterns were found. However, an interpretation basis for the
evaluation of the polarizing filter images in view of the fracture pattern is not existing
so far. A reason for this is the missing knowledge on the relation of the visualized
colours and the stress state in prestressed glass panels. Hence the product quality in
terms of the isotropy of the prestress was to be investigated by the use of polarizing
filter technologies [2].
2. Photoelastic investigations and fracture pattern analysis
2.1. Principles of photoelasticity

Stress-free glass is optically isotropic, i.e. in the glass, the light spreads in all directions
at the same speed. If we have a thermal prestress, both the propagation velocity and the
refractive indices change: the glass then becomes birefringent. The birefringent effect
can be made visible through polarized light. If we now consider a body under a plane
stress state (V3=0 and V1¸V2), the stress difference (V1–V2) of this body can be
determined by polarizing filter images.
837
Challenging Glass 3
Anyhow, for thermally tempered panels, the stress magnitude cannot directly be
determined by this method because the stress changes over the thickness. In addition, in
bent glass perfect prestress parabolas do not exist: in a bent glass pane slight differences
in the shape and position of the prestress parabola may occur. As a consequence,
equilibrium of the prestress then needs to be considered over a larger area. The usual
consideration of a stress section of a single point location in view of equilibrium is no
more sufficient (Fig. 1). Since the varying shapes and positions of the prestress
parabolas yield into a resulting stress integral ¸0, this influences also the polarizing
filter images.
Figure 1: Examples of prestress parabolas measured with SCALP 04 over the thickness.
The colour scale by Michel-Lévy can be used as a reference to the polarizing filter
image [3]. It shows an interference colour sequence beginning with black and going
over into a fixed sequence of colours (redÆblueÆgreenÆyellowÆorangeÆred), being
repeated continuously (Fig. 2).The interference of the colours can be ordered after the
“red-to-blue transfer”. Here the scale can be used to get information on the orthotropie
stress state of prestressed glass panels.
However with polarizing filter images the quantitative determination of the stress
distribution across the glass is not possible. To measure the distribution of the thermal
stress over the thickness (locally at a single point in the glass panel), the optical
measuring system SCALP 04 can additionally be applied [4].
2.2. Application and optimization

Since linearly polarized light produces distracting black lines (so called isoclines) in the
images, circularly polarized light was used which eliminates the isoclines. In the
experimental setup an additional retarder with a defined optical path difference of
560nm is positioned, so that interference colours in the range of the first order can be
created. Fig. 2 illustrates this on the colour scale by Michel-Lévy. If there would be no
retarder and the prestress is isotropic (V1 V2), only black and white areas were
838
Quality Control of Bent Heat Strengthened and Fully Tempered Glass by the Application of Photoelasticity
visible (Fig. 2 left). With the use of a retarder, the origin of the scale (i.e. an isotropic
stress state) moves into the coloured area.
Figure 2: Colour scale by Michel-Lévy, effect of the retarder for V1 V2.
In order to gain more precise information between the polarizing filter images of
thermally tempered bent panels and the condition of prestress, a large number of
polarizing filter images were taken (uniformly and irregularly prestressed). To minimize
measurement errors, measuring from both sides of the panel was necessary [4]. So
detected colour-striking areas were marked. At these locations the prestress in
x direction and perpendicular to it (y direction) could be determined with the SCALP 04
system.
Figure 3: Polarizing filter image with marked points and measurement direction
Also the measurements have been performed from both the convex and the concave
surface. For each direction at each point a single prestress parabola has been constructed,
averaging the stress ordinate at the centre plane. From the resulting parabola finally the
integrals of the prestress in the x and y direction and their difference can be calculated.
The results of stress integral differences that are obtained by the SCALP 04 method
now can also improve the evaluation of the polarization filter images. In this respect, the
main equation of photoelasticity for varying stress distribution across the thickness-
direction z can be used (1).
(1)
Here C describes the material constant of the glass and the characteristic wavelength.
The stress integrals in the x- and y direction are shown with V1 and V2. It shows the
839
Challenging Glass 3
linear dependency between the "difference of the principal stress integral" and the
"phase shift".
We define the phase difference s = · , which is used in the colour table by Michel-
Lévy. From this any colour of the polarizing filter image for the corresponding optical
path difference can be found. Considering the retarder delay (sretarder), with the
"difference of the principal stress integrals" (see eq. (2)) a higher colour resolution in
the considered stress range then can be obtained.
(2)
For each of the about 600 measuring points, the colour of the polarizing filter image was
accordingly defined in the RGB colour space. Sorting the "differences of the principal
stress integrals" and their corresponding colours in ascending order by the magnitude of
the differences (Fig. 4) results into an improved colour scale band of the colour scale by
Michel-Lévy.
Figure 4: Product related spectrum band in the colour spectrum of Michel-Lévy.
The border values for the differences lay in the derivated spectrum band at
-84,42 Mpa·mm as minimum and at +97 Mpa·mm as maximum.
If we now measure the differences V1–V2 on the surface (with SCALP 04), as already
mentioned before [4], then the question is, whether we can also correlate these surface
differences with the differences of the stress integrals:
(3)
840
This is shown in Figure 5 and Figure 6.
Figure 5: Spectrum from the differences of surface stress in x- and y direction (convex side).
Colours of surface stress are placed at positions of respective stress integral differences.
Figure 6: Spectrum from the differences of surface stress in x- and y direction (concave side).
Colours of surface stress are placed at positions of respective stress integral differences.
The comparison to Fig. 4, shows that often small colour strips mismatch with the stress
integral differences. The reason for this is that a large part of the stress gradient, which
has an influence on the colour, is neglected in case of only measuring at the surface.
Considering only the spectrum for the convex side, as usual according to the colour
scale by Michel-Lévy, the colour gradient starts again from orange/red through blue and
goes into yellow. The “difference of the surface compressive stresses” comprises a
margin for the convex side from -9,9 Mpa (left) to +18,5 MPa (right).
In the project with the polarizing filter images and the measurement of the prestress
three prestress profiles could be established, which characterize the isotropy of the
prestress. In profile type 1, (Fig. 10) recurring and homogeneous colour structures can
be seen which correspond to a slight inequality of the prestress integrals.
Panes with the prestressing profile type 2 (Fig. 11) and type 3 (Fig. 12) were
inhomogeneously prestressed on intent and differ primarily in the intensity of the
colours.
Especially for prestressing profile type 3 white areas are identified which have a stress
difference of integrals of about -122 Mpa·mm and thus clearly lie below the newly
created scale.The lowest minimum of measured stress integral differences of all
specimen was about -390 Mpa·mm.
As stated, the polarization filter images give no information about the size of the
prestress, but only on the isotropy of the stress integrals. Therefore no distinction
between heat strengthened or fully tempered glass panels of the same batch is possible
by analyzing the colours of the filter images. Nevertheless, panels with a various
thickness and similar prestress conditions have deviations in the polarizing filter images.
These are due to the fact, that the integral of the prestress parabola increases linearly
with the thickness of the glass and the stress ordinates remain constant. Therefore
greater colour differences occur more frequently in a thicker panel than in a thinner one.
841
Challenging Glass 3
3. Fracture pattern analysis and correlation to polarizing filter images
3.1. Fracture pattern tests

Additionally, fracture pattern tests were performed according to the current flat glass
standard EN 12150 [6] resp. EN 1863-1 [7]. To ensure a stress-free bearing, the glass
panels were placed vertically on the curved edge of the glass and struck with a spring-
operated firing pin. The flat-glass quality requirements have been fulfilled by the bent
tempered specimen by providing the required minimum number of crumbs in a counting
mask of 50mm x 50mm. The homogeneous polarizing image and a locally measured
average surface compressive prestress of -90 Mpa confirm this.
The fracture pattern tests for heat strengthened glass only partly fulfill the requirements
for the fracture pattern. Also here the well-known fact could be observed that the
thickness of the panels has a big influence on the fineness of the fracture pattern: panels
of different thickness but of same thermal surface compressive prestress have a different
fracture pattern. The effect is shown in Figure 9: glass panel of 6mm fulfill the
requirements a glass panel of 10mm shows to much fracture lines.
Figure 9: Fracture pattern heat strengthened glass: left (d = 6mm, VP = -50 N/mm²);
right: (d = 10mm, VP = -50 N/mm²).
3.2. Prestress profiles vis-á-vis the fracture patterns

Now the question will be discussed whether the prestress images with regard to
homogeneity of colours and structure correlate with the fracture pattern.
Prestress profile type 1 (Fig. 10):

The tests show that a slight non-isotropy of the prestress integrals has no effect on the
fracture behavior, as here a uniform fracture pattern was observed.
Figure 10: Type 1, polarizing filter image and fracture pattern (lines are redrawn) e.g. heat strengthened glass.
The strong anisotropy of the prestress in type 2 leads to an irregular fracture pattern. In
panes of this type, the fracture pattern requirements, especially for heat strengthened
842
glass could not be fulfilled. Along the upper edge an area with partly fine broken
crumbs could be detected. There the polarizing filter images show a clear, orange-
coloured stripe. In this area a correlation between the polarizing filter images and the
fracture pattern can be observed.
Figure 11: Type 2, polarizing filter image before and after the fracture.
The fracture patterns of panels of type 3 neither fulfill the fracture pattern requirements
for flat heat strengthened nor for flat fully tempered glass. Also for this type fine broken
crumbs at the upper edge can be observed. At these points the polarizing filter image
shows an array of red, orange and yellow tones, which are concentrated at the edge.
Figure 12: Type 3, polarizing filter image before and after the fracture.
In all cases a clear correlation between the isotropy of the prestress integrals and the
homogeneity of the fracture pattern of a bent glass panel can be observed. Larger colour
differences in the polarizing filter image are indicators for an irregular fracture pattern.
However, a conclusion about the fineness of the fragmentary cannot be made by use of
polarizing filters. Additional measurements to evaluate the stress distributions across the
thickness at some points of the glass panel are necessary, to finally give information
about the fracture pattern.
4. Summary
It is shown that by polarizing filter images the isotropy of the cross-section integrals of
thermal stress can be evaluated used as a quality criterion next to the fracture pattern.
Since it can be included into the production line of semi-finishing glass companies, the
method is very advantageous for the production e.g. of bent glass panels.
To the visualized interference colour a so-called “integral of the stress difference” can
be assigned, which is a measure for the imbalance of perpendicular prestress (e.g. an
irregular stress distribution in the x and y direction, or an asymmetric parable through
the thickness). On the basis of local measurements else than with polarizing filter
images the derivation of the prestress parabola over the thickness and an evaluation of
the stress distribution is also possible.
843
Challenging Glass 3
It should always be noted, that through polarizing filter recordings alone, quantitative
values of stress so far cannot be determined. That means for example, that tempered
heat strengthened and fully tempered glass panels of the same thickness often have
similar polarizing filter images in case they are uniformly prestressed.
Opto-mathematical algorithms allow for an evaluation of the colour differences in the

RGB space and together with the determination of assessment criteria for an online
monitoring system in the industrial production.
5. Acknowledgements
The aforementioned interdisciplinary research work was conducted in collaboration
with four partners. The glass samples were provided by Flintermann GmbH. ISRA
Surface Vision GmbH provided further photoelastic knowledge and translated the
results into an online quality control system. Scientifically, the project was accompanied
by RWTH University Aachen and University of applied sciences of Munich. The
authors and project partners are grateful for the financial support of the Federal Ministry
of Economics via the AiF-ZIM-programme.
6. References
[1] Bucak Ö., Feldmann M., Kasper R., Bues M., Illguth M.: Das Bauprodukt „warm gebogenes Glas“ –
Prüfverfahren Festigkeiten und Qualitätssicherung. Stahlbau Spezial (2009), Konstruktiver Glasbau, S.
23–28.
[2] Bucak Ö., Feldmann M. et al.: Reproduzierbare und prüfbare Erhöhung der Qualität und damit der
Leistungsfähigkeit thermisch vorgespannter gebogener Gläser. ZIM Endbericht 2010 (unveröffentlicht)
[3] Magnus M.: Michel-Lévy Farbtafel – Mineralbestimmung im polarisierten Licht. Carl Zeiss
MicroImaging GmbH
[4] Feldmann M, Kasper R, Langosch K: Glas für tragende Bauteile. 1. Auflage 2012 Werner Verlag
[5] Blank K., Grüters H., Hackl K.: Contribution to the size effect on the strength of flat glass.
Glastechnische Berichte 63 (1990) Nr. 5
[6] EN 12150: Glass in building – Thermally toughened soda lime silicate safety glass
[7] EN 1863: Glass in building – Heat strengthened soda lime silicate glass
844
doi:10.3233/978-1-61499-061-1-845
Layout Strategies and Optimisation of Joint

Patterns in Full Glass Shells
Thiemo Fildhuth, Sebastian Lippert, Jan Knippers
ITKE - Institute of Building Structures and Structural Design, University of Stuttgart /
Germany, t.fildhuth@itke.uni-stuttgart.de, www.itke.uni-stuttgart.de
Continuously curved, full glass shells have a large potential for application in free-
form façades or roofs. Glass elements needed for such a construction can be
fabricated either by heat bending or cold lamination bending. But as the curvature
and size of glass laminates are limited by manufacturing restrictions, structural
joining is necessary for assembling large shells. Unfortunately, such joint
discontinuities conflict with the desired membrane state of the shell. The joint
pattern layout has a large influence on the load bearing behaviour, stability and
appearance of glass shells. Hence, computational procedures have been used to
optimise the joint geometry layout by targeting either at minimising particular joint
forces or by following the principal membrane force trajectories of the shell.
Keywords: Heat and cold bent glass laminates, full glass shells, joint pattern layout,
optimisation, trajectories
1. Introduction
Approaching the design and construction of continuously curved glass shells for use in
free-form façades or roofs implies many complex, inter-dependent subjects: Material
and manufacturing restrictions limit the size, the load bearing capacity and the possible
curvature of heat or cold bent glass elements, which can be used to compose large shells.
The necessary structural joining of the glass panels results in surface and stiffness
discontinuities which conflict with the desired membrane state of the shell. As the
joining also largely influences the montage, load bearing behaviour, stability and not
least the appearance of the glass shells, an optimisation of the joint pattern layout is a
promising approach on the feasibility of such full glass constructions. Furthermore,
optimisation can contribute to reduce particular joint forces.
An application of double curved, heat bent glass elements allows for a higher degree of
curvature of the shell, but panel sizes are limited and tempering may not be possible. By
contrast, fully tempered glass can be used for elements produced by cold lamination
bending, but their large curvature radii only allow for assembling low-curvature shells.
This fact e.g. impedes optimising of the load bearing behaviour by modifying the main
curvatures.
In the following, heat and cold bent glass is principally compared as an important
precondition for an optimisation procedure presented thereafter. Examples of joint
geometry manipulation targeting on minimising the joint forces produced either via an
evolutionary computational approach or by adapting to the principal membrane force
trajectories conclude the article.
845
Challenging Glass 3
2. Curved Glass Elements as a Prerequisite for Shells
2.1. Single Curvature

Heat bent, single curved glass is widely available for building purposes. Tempered,
single curved glass even up to 2.85x12.00m with radii down to 4.00m can now be
manufactured. But the residual stress quality due to tempering may vary significantly
depending on the manufacturing [1] in such glass elements; in some cases even tensile
surface stress up to 40MPa has been measured [2]. Lamination of heat bent glass panes
may result in additional unknown stress enclosed in the glass [1]. Furthermore, heat
bent glass having a low curvature often shows disturbing optical surface blurs.
Generation of single curvature via cold lamination bending [3, 4] allows for using heat
strengthened (HTG) or fully tempered (FTG) glass. Therefore, stacks of tempered glass
panes and interlayer material are deformed together by means of a bending (mould)-
“device”, which keeps deformation constant during the following autoclave heat and
pressure lamination [3, 5]. Once removed from the autoclave / bending device, the now
coupled glass laminate suffers initial springback and long-term creep, which reduces the
applied deformation by approximately 10-15% for single curvature. Springback and
creep are largely depending on the shear stiffness properties of the interlayer, thus e.g.
SentryGlas® Plus 5000 (SGP) [6, 7, 8] is typically used due to its high short and long
term shear modulus. The maximum deformation imposed prior to lamination is always
limited by the acceptable tensile surface stress in the glass (here: ~80MPa for FTG)
caused by cold, elastic bending. Cold deformation into single curvature mainly depends
on bending and thus on the glass thickness t; e.g. curvature radii of down to 4.50m (no
springback included) could theoretically be reached for t=6mm for a tensile stress limit
of <50MPa.
2.2. Double Curvature

Spherically heat bent glass up to 4.60x2.60m is generally available in float glass quality.
Anticlastic curvature (only float glass) as an expensive, particular custom product is
rarely fabricated. Usually, double curved, heat bent glass is not available in fully
tempered standard glass (FTG) quality. Therefore, only chemical tempering could be an
option, which has e.g. been used for the spherically bent glass elements (R=8.00m) of
the 8.50m span, full glass calotte shown in [9].
Cold lamination bending as described in chapter 2.1 allows producing glass elements
having a limited, low synclastic or anticlastic curvature.
Figure 1: Small, double curved panel achieved through cold bending and lamination
(photo: T. Fildhuth, panel: seele sedak GmbH & Co. KG)
846
Layout Strategies and Optimisation of Joint Patterns in Full Glass Shells
In contrast to bending into single curvature, membrane effects prevail in elastically

double curved glass panes, where the dimensions and the deformation geometry are the
main factors for the inner stress states. This is shown (fig. 2) for the minimum principal
curvatures R1=R2 at ~80MPa of 3x3m FTG-panes deformed into a synclastic paraboloid
(Ri=15m), an anticlastic hyperbolic paraboloid (HP) (Ri=20m) and a saddle shape
(Ri=22m). Such geometries, as per fig. 2 (left), are typical for many shells. Analytical
description of cold deformation into double curved geometries is unfortunately
hampered by the unknown exact deformation of the glass – even minor mistakes in
horizontal displacement assumptions cause large calculation discrepancies [10].
Figure 2: Max. principle tensile stress, radii and nodal displacement during cold deformation
into paraboloid-, saddle- and hp-shape (no springback included)
Figure 3: springback simulation by time steps in dynamic, non-linear FE analysis –

max. nodal displacement for paraboloid shapes (short term SGP shear module)
847
Challenging Glass 3
Typically, the short-term springback mentioned above and long-term creep reduce the
internal stress state in the laminates to values below 40MPa. These stresses have to be
taken into account for the dimensioning of the global structure. In multilayer laminates
(typically min. three panes), the stress included in the inner and the outer glass panes
after springback may significantly vary in between 20 to 40MPa [3]. FE-simulation of
springback (2x10mm ESG, 0.9mm SGP) by time steps is shown for paraboloid
geometries of different sizes in fig. 3. Springback and creep normally reduce the initial
deformation by about 50%. Generally, rectangular laminate shapes tend to develop
single curvature zones around the corners during springback. Thus, for achieving a
desired target deformation geometry after springback and creep, iterative deformation-
and-springback loop runs have to be performed in FE-simulation.
3. Modelling and Abstraction of Joints

Layered, four-sided shell elements (Sofistik 2010) having an elastic material law are
used for FE-simulation of the glass laminated from 3x10mm FTG (E=70,000MPa,
Q=0.23) with 2x0.9mm SGP (E=10MPA, G=3.3MPa). Dead load (gk=0,77kN/m2) and
snow load (sk=0,75kN/m2) are applied evenly distributed to the shells in characteristic
(1.00g+1.00s) and design combinations (1.35g+1.50s). Asymmetric load cases are not
presented here.
For simplified modelling of joints, either shell / volume elements usually having an
isotropic or orthotropic material law or springs of a corresponding stiffness are often
used. For abstraction reasons and to simplify the joint force post-processing, spring
quadruples geometrically arranged per Frenet Frame are used here in analogy to the
components of Kirchhoff-Love shell theory (fig. 5). Rotational springs m12 are
neglected; translational springs n11, n12, q1 and rotational m11 are modelled continuously
along the glass element edges having a joint or support. Basic stiffness values provided
by an industry partner are shown in table 1. Conversion of material properties of a given
joint geometry of width b and thickness t into spring stiffness cn,i,j / cm,i,j and vice-versa
is similar to the description in [11]. Previous studies [5] have revealed a non-linear
relationship of the joint stiffness and the load bearing behaviour of shells, where already
a low joint stiffness allows invoking a certain membrane performance (fig. 4).
Table 1: Spring stiffness and a selection of comparative material parameters.
Joint material spring parameters spring stiffness used in the

(for comparison, t=15mm) study (equivalent thickness teq)
spring type monolithic silicone epoxy resin joints: equival. supports

glass E=1MPa E=1,000MPa E= 102MPa,
E=70,000MPa teq=24.5 mm
cn11 [kN/mm2] 184.8000 0.0033 3.1348 0.2500 0.2500
cn12 [kN/mm2] 71.1000 0.0008 0.8621 0.0500 0.0500

2
cq1 [kN/mm ] 71.1000 0.0008 0.8621 0.0500 0.0010
[kNm/(rad
cm11 9,623.7000 0.1713 163.2706 12.5000 0
m)]
848
ଵଵ ɐଵଵ
‫ ۍ‬ଶଶ ‫ې‬ ‫ ۍ‬ଶଶ ‫ې‬
ɐ
‫ ێ‬ଵଶ ‫ۑ‬ ‫ ێ‬ଵଶ ‫ۑ‬
୲ ‫ ێ‬ɐ
‫ ێ‬ଵଵ ‫ۑ‬ ‫ۑ‬
ଶ Ʌଷ ɐଵଵ
‫ ێ‬ଶଶ ‫ ۑ‬ൌ න ‫ێ‬ ‫ ۑ‬Ʌଷ
‫ێ‬ଵଶ ‫ۑ‬
ଷ ଶଶ
ି ‫ێ‬Ʌ ɐ ‫ۑ‬
୲
ଶ
‫ۑ ێ‬ ‫ێ‬Ʌଷ ɐଵଶ ‫ۑ‬
ଵ
‫ۑ ێ‬ ‫ ێ‬ɐଵଷ ‫ۑ‬
‫ ۏ‬ଶ ‫ے‬ ‫ ۏ‬ɐଶଷ ‫ے‬
Figure 4: Non-linear relationship of stress and joint Figure 5: Spring Model - Frenet Frame orientation and
stiffness (10x10m HP shell). Kirchhoff-Love shell theory force components.
4. Optimisation
As in-plane tensile joint forces +n11 are particularly difficult to deal with, a main
objective of the experimental optimisation is to reduce peak +n11-forces while
complying with geometrical boundary conditions like glass elements size etc.
Stimulated by methods and results from the research project [12], a diploma thesis [13]
concerning a code-based optimisation scheme for joint patterns applicable to user-
defined NURBS shell geometries using evolutionary algorithms [e.g. 14] has been
accomplished at the ITKE. Two distinct functions of the program are to be
distinguished as per chapters 4.1 and 4.2:
4.1. Geometrical and Modelling Features

Modelling and joint pattern layout function, involving finite element modelling and
analysis: The program first creates a user-controlled joint grid, which constitutes the
base for a following FE-modelling and analysis. Geometrically, the joints are laid out as
interpolation “streamline” curves (using inverse distance weighting) through a
parametric array of vectors situated on the shell surface. The curves can be modified by
angular variation (Re,2, see below) of the vectors with respect to principal directions
defined in the optimisation steps. This geometry is then converted into a finite element
model in Sofistik. Thus, geometrical and FE-analysis results are available for the target
functions of the evolutionary optimisation process. The streamline function for creating
joints allows adapting joints to particular force trajectories of the shell, which could
significantly reduce target joint force components. Such an orientation along the
maximum principle membrane force vectors (nI) is demonstrated in chapter 5.1 for the
example of an elliptic paraboloid shell.
Plane stress state assumed, the resulting principal membrane forces and their angle can
be calculated by means of the classical equations (1) and (2) for known in-plane forces
nx, ny und nxy related to global or local coordinates.
୬౮ ା୬౯ ୬౮ ି୬౯ ଶ
୍ǡ୍୍ ൌ േ ටቀ ቁ ൅ ଶ୶୷ (1)
ଶ ଶ
849
Challenging Glass 3
ଵ ଶή୬౮౯ ¹
୍ ൌ ή ൬ ൰ Ǣ୍୍ ൌ ୍ ൅ (2)
ଶ ୬౮ ି୬౯ ଶ
4.2. Evolutionary Optimisation Feature

Evolutionary optimisation loop procedure implying fitness ranking of individual
solutions, selection via stochastic universal sampling, recombination, mutation and
reinsertion [e.g. 15]: The correlating automated geometrical adaption of the joining
layout is conducted evaluating the following geometrical / manufacturing restrictions
(Re,i) of the glass elements and the targets (Ta,i) (figure 6):
x Re,1: Ratio of rise fi to length li for curved

element boundaries (projected on mean plane)
x Re,2: Angular variation of joints to principal
directions (also defining the corner angle of
the glass elements); this angular variation is
related to the setting of fi/li from Re,1
x Re,3: Number of element boundary curves
(triangular, quadrangular, …)
x Ta,1: “Bounding box” - glass element
dimension limits (a, b) prior to the cutting of
the final shape
x Ta,2: Max. / mean spring force results n11, n12,
Figure 6: Glass element parameters.
q1, m11 of joints and supports
Chapter 5.2 gives the evolutionary optimisation results for a hyperbolic paraboloid (HP)
shell. All joints and supports are configured as per table 1.
5. Examples
5.1. Trajectory Joint Scheme Examples

A section from an elliptic paraboloid shell of 12.00m span, which could e.g. serve as a
subway entrance, was taken as an example for joining patterns derived from principal
membrane force trajectories (nI, nII). Either the nI- or the nII-vectors were used for
creating the streamline for the joining as per fig. 7. Subsequently, the shear forces
nearly vanished in the joints (fig. 8). Initially, the glass element dimensions were
restricted to a<4.60m and b<2.60m. Thus, the shell could be either designed with low
curvature to theoretically permit assembly from cold bent glass laminates (min.
principal radii R1=18m, R2=60m; 03a_traj_cold_bent) or with an increased curvature by
applying synclastic, heat bent glass (R1=8m, R2=24m; 06_traj_heat_bent), see figure 9.
As cold bending theoretically permits bigger element sizes, a variant with a<12.50m
was added (01_traj_cold_bent). In another version, the trajectory scheme was rotated by
45° to get a joint orientation along the principal membrane shear force vectors
(08_traj_45°_cold_bent). Joint forces and principal stress were calculated for design
loads as per chapter 3. Buckling analysis was performed using load increments (related
to an evenly distributed surface load of 1kN/m2) of the imperfect shell. To establish
these imperfections, the most unfavourable eigenforms have been scaled to L/300
(~40mm max. vertical imperfection).
850
Figure 7: Streamlining of the joints along principal Figure 8: Shear forces n12 in joints vanish due to
membrane force vectors principal membrane force orientation
Figure 9: Schematic overview of elliptical shell joint patterns along trajectories
Results clearly show the changing joint loads depending on the joint orientation (fig.
10). Avoiding any joint which crosses the principal load bearing direction (nI) results in
the low n11-joint forces in version 01_traj_cold_bent. Figure 11 shows the relatively
poor buckling stability of shell 03a_traj_cold_bent caused by the joint layout. Rotating
the joints by 45° (08_traj_45°_cold_bent) or applying fewer joints (01_traj_cold_bent)
significantly enhances the stability behaviour. The best performance is reached in the
shell with high curvature (06_traj_heat_bent), although it has the same joint pattern as
the worst shell example with low curvature (03a_traj_cold_bent). Principal design
stress results in the layered glass are shown in figure 10.
851
Challenging Glass 3
Figure 10: Comparative results of n11-spring forces and principal bottom stress (design loads)
852
Figure 11: Stability analysis of the imperfect shell, load-displacement diagram.
5.2. Evolutionary Joint Force Optimisation Example: HP-shell

The performance of the optimisation routine minimising the peak tensile n11-joint forces
(Ta,2) is verified using the example of a 10x10m HP. The target functions include Re,3:
4-sided elements; Ta,1: 1.5<a<14.0m and 1.5<b<2.5m; Re,2: angular variation <±30°;
Re,1: fi/li<0.1. Optimisation leads to slightly curved joints rotated towards the principal
tensile membrane force direction (diagonal). Fig. 12 shows the development of the
geometrical conditions and of the n12- and +n11-joint forces during the procedure.
Compared to a regular, straight joint pattern created manually, the peak / mean +n11-
joint forces have been reduced by 70% / 45% and joint shear n12 is reduced by 16% as a
side effect. Compressive joint forces and support reaction forces rise (fig. 13). Figure 14
shows a design stress plot.
853
Challenging Glass 3
Figure 12: Evolution of design parameters during optimisation from [13].

Figure 13: Design n11-spring forces in joints and supports (blue: tensile);
left: regular joint pattern, right: optimised version
854
Figure 14: Principal bottom stress vectors in the glass prior to (left) and after optimisation (right)
6. Conclusion
The contrivable sizes, geometries, curvatures and quality of glass elements to be used in
full glass shells are the basic limiting factors for the aspired design of such structures.
Heat bending of glass allows achieving sufficient curvature necessary for a shell, but the
fabricable element dimensions especially for double curvature are limited. Usually, such
double curved glass is only available in synclastic shapes. Tempering of large, double
curved, heat bent glass elements is still difficult to achieve or even not possible. Thus,
either float glass quality or chemically tempered glass are options for double curvature,
whereas single curved elements are available as HSG or FTG. Cold lamination bending
of glass, which is possible for every glass quality (ANG, HSG, FTG), permits less
curvature, but results in a very high surface quality. The achievable curved element size
(<15m length) is only limited by the available autoclave dimensions and by physical
(tensile stress) properties of the glass. As for double curvature, the attainable radii
unfortunately grow with increasing glass dimensions.
Based on these findings, FE-simulation procedures for full glass shells were deployed.
As the feasible shell geometries are limited, stiffness modification of the joints and the
optimisation of their geometrical layout with respect to the shell surface turned out to be
useful for enhancing the load bearing behaviour. Therefore, two different,
computational approaches are presented here for a synclastic and an anticlastic shell
design: A direct joint adaptation to the principle membrane force trajectories was
performed for an elliptic paraboloid shell, whereas an evolutionary optimisation
procedure targeting minimisation of the particularly problematic, tensile (+n11) joint
forces was used for a HP-shell. In both cases certain boundary conditions for the panel
sizes / angles had to be fulfilled.
855
Challenging Glass 3
Depending on the shell geometry and the glass element size, both optimisation
approaches resulted in significant changes or reductions of the target joint forces
compared to standard, regular joint patterns. Slight curving and diagonal orientation of
joints in the anticlastic HP reduced tensile joint forces by more than 60% (fig. 13). On
the other hand, compressive joint forces and tensile support reactions rise, but this may
be preferable compared to high tensile forces in the joints.
Trajectory joint layout is very efficient for exposing joints to only one, chosen
component of the principal membrane force (the other, bigger one being parallel to the
joint); in addition shear (n12) nearly vanishes (compare Mohr´s Circle). E.g. if a
principal tensile membrane force direction exists in a shell, this component could be
avoided in the joints by orientating the joints parallel to the tensile force vectors.
Nevertheless, layout difficulties may occur near disturbances like supports, curvature
changes etc. If small glass panels are used, two principal joint directions have to be
managed, which may expose one of the joint directions to an undesired force component.
Stability studies (fig. 11) show the important influence of the joint pattern layout and
the overall curvature on the buckling behaviour of synclastic shells. Hence, the
reduction of the number of joints by using large glass elements and an orientation of the
joint to counteract typical buckling (eigen-) forms are recommended.
7. References
[1] Bucak, Ö., Feldmann, M., Kasper, R., Bues, M., Illguth, M.: Das Bauprodukt „warm gebogenes Glas“ –
Prüfverfahren, Festigkeiten und Qualitätssicherung. Stahlbau Spezial 2009, pp. 23-28.
[2] Colvin, J. B.: The Analysis of Breakages of Curved Heat Treated Glass Breakages. GPD 2011
Proceedings, Tampere 2011, p. 130
[3] Kassnel-Henneberg, B.: Purely structural glass building envelopes. GPD 2011 Proceedings, Tampere
2011, p. 304
[4] de Vericourt, R.: Inextensional Theory Applied to Cold Bent Glass Built-in Stress Evaluation. GPD 2011
[5] Fildhuth, T. and Knippers, J.: Double Curved Glass Shells from Cold Bent Glass Laminates. GPD 2011
[6] Bennison, S. J., Qin, M. HX., Davies, P.S.: High Performance Laminated Glass for Structurally
Efficient Glazing. Innovative Light-Weight Structures and Sustainable Facades Conference Paper, Hong
Kong 2008
[7] Feirabend, S.: Steigerung der Resttragfähigkeit von Verbundsicherheitsglas mittels Bewehrung in der
Zwischenschicht. PhD Thesis, Universität Stuttgart 2010, p. 20
[8] Belis, J., Vander Beken, J., Van Impe, R., Callewaert, D.: Performance of Ionoplast Laminates above
Room Temperature. Conference Proceedings Glass Processing Days 2007, Tampere (SF): 2007
[9] Blandini, L.: Structural Use of Adhesives in Glass Shells. PhD Thesis D93, ILEK - Universität Stuttgart:
Verlag Grauer 2005
[10] Keller, C. M.: Modellierung der Kaltverformung von Glas in doppelt gekrümmte Formen mit Platten-
und Schalentheorien. Diploma thesis at the IBB and the ITKE, Universität Stuttgart 2012
[11] Bagger, A.: Plate Shell Structures of Glass. PhD Thesis, DTU Civ. Eng., Lyngby 2010
[12] ITKE, seele GmbH & Co. KG: Kalt gekrümmte Glaslaminate – Herstellung und Verwendungschancen.
AiF-research project KF2386401SU9, unpublished.
[13] Lippert, S.: Optimierung des Fügungsverlaufs bei Ganzglasschalen aus gekrümmten Glasbauteilen.
Diploma thesis at the ITKE, Universität Stuttgart 2011
[14] Harzheim, L.: Strukturoptimierung: Grundlagen und Anwendungen. Verlag Harri Deutsch, Frankfurt a.
M., 2008
[15] Rechenberg, I.: Evolutionsstrategie ´94. Verlag Frommann-Holzboog, Stuttgart, 1994
856
doi:10.3233/978-1-61499-061-1-857
Special Aspects in Planning of Thermally

Curved Glass: Structural Design,
Constructional Hints
Markus Kramer
IB KRAMER, Germany, info@ib-kramer.de, www.ib-kramer.de
Michael Elstner
INTERPANE, Germany, Michael.elstner@interpane.com, www.interpane.com
The Federal Association of Flat Glass in Germany (Bundesverband Flachglas),

published a new guideline „Guidelines for Thermally Curved Glass“ in 2011. The
guideline contains information about the processing of curved glass, tolerances
which may occur as well as for the assembling of this type of glazing. The current
paper deals with special aspects in planning and calculating structures including
thermally curved glass, given in the guideline. Bases for calculating curved glass
are explained, as well as problems due to the higher stiffness of bent glass panes
used for insulating glass units. The need to use appropriate computational
procedures, implicating shell action is shown in examples. Further more, some
essential constructional aspects are discussed. Especially the influence of tolerances
between glass and substructure, and deformation of the substructure is shown.
Keywords: Curved Glass, Static Calculation, FEA, Structural Design
1. Introduction
Building projects with the use of free-form façade shapes became more and more over
the last years. Such façades often require the use of curved glass elements. From
practice, there is a lack of information about curved glass, and thermally curved glass in
particular. Due to this, the Federal Association of Flat Glass in Germany built up a
workshop over two years, attending glass bending companies, glass refining companies,
and also attendees from research and engineering offices. The result was a guideline,
including the most important aspects, to be considered with the use of curved glass for
buildings [5]. Looking to the characteristics of calculating and designing structures with
curved glass, this is a very important aspect, often causing problems and failures. In the
following the material characteristics, due to the production process, bases for
calculation of curved glass, and approvals for the design of the supports will be
discussed.
2. The product “curved glass”
2.1. Bending process

With regard to the most important material properties of the building material glass, it is
ideal elastic, but has no ductility, so that failure occurs spontaneously. Due to this, the
preferred way to produce curved glass elements should be bending them thermally.
857
Challenging Glass 3
Thermally bending avoids unwanted preloading of the glass, and reduces loads to the
substructure, as it happens in the use of cold bending glass elements.
To bend a flat glass pane, first a mold, defining the final shape, has to be built. For the
thermally bending process, the flat glass pane is put on the mold, and heated up to about
620°C, the transformation temperature. During the heating process, the glass becomes
soft, and sinks into the mold. After this, the bent glass is cooled down. This process
occurs very fast, to get tempered, curved glass, or very slowly, getting curved float glass,
avoiding any residual stress. The cooling process for float glass can take lots of hours (s.
figure 1).
Figure 1: Results for deformation.

Curved glass can be finished to all common glass products, like laminated safety glass,
or insulated glass units.
2.2. Gravity based process

To get curved float glass, normally a gravity based process is used. Following the
principle described in 2.1, the glass sinks into the mold due to gravity only, when it
becomes soft after heating up. This process offers a lot of very complicated, even
spherical, biaxial curved shapes, and very small radii. The mold can be built like a ‘u’,
or like an ‘n’, depending on the wanted shape. So the flat pane sinks into the mold, or
sets on the mold. The gravity based process can also be used for tempered curved glass.
In this case, the curved and hot glass element will be cooled down rapidly, when leaving
the oven. This procedure is a little bit old fashioned, and causes problems with a
continuous pre stress of the panes. But it gives the opportunity, to realize quite small
radii for tempered curved glass.
A further advantage of the gravity based processes is the ability, to bend more than one
pane, laying over each other, separated by fleece layer, at the same time. This
guarantees minimal tolerances between the single curved elements, when used for
laminated safety glass units. This ability is not given for tempered curved glass.
2.3. Combined process, gravity and flexible molds

For the production of tempered curved glass, nowadays, a combined process using
gravity and flexible molds is used. These modern types of machine include the heating,
bending, and tempering process in one single unit. The principle is the same as
described in 2.1, but the mold is not a rigid block or element, but made of flexible
elements, moving from both sides to the heated glass. The jet nozzles for the tempering
process are often included in the mold elements. The production process with these
machines, takes only a few minutes. But it is a very complicated, and high engineered
process, to find the right parameters for the heating and cooling phase, to get a proper
858
Special Aspects in Planning of Thermally Curved Glass: Structural Design, Constructional Hints
tempered curved glass. All the needed parameters differ from shape to shape, so the
bending company needs a lot of experience to get good results.
The two main disadvantages of this process are that only cylindrical shapes can be
produced, and that the single panes for laminated safety glass have to be produced in
separate stages. So tolerances between the single elements of a laminated safety glass
unit have to be taken by the use of thicker interlayers.
3. Bases for calculation: Curved glass vs. flat glass
3.1. Stiffness
The main difference between the calculation of flat and curved glass elements is caused
by a trivial mechanical reason: Higher stiffness due to the shell action. Depending on
the type of supporting, and the amount of curvature, the stiffness of the glass element
growth up substantially. An effect, what can easily be shown by a sheet of paper: You
will fail, trying it holding vertically, fixed at the bottom edge, when it is flat. But giving
only a small curvature at the bottom edge, it will stand properly in a vertical direction.
To activate the shell action of the system, the type of supporting is the most important
point. Following the most frequently used types for cylindrical shapes are described.
Type a), where only the straight edges are supported, is a statically defined system, what
is equal to a simple beam. There will be not shell action, and not difference to a flat
system with regard to the bending moment, effecting the bending tensile stress (figure
2a). Supporting of type b) effects a shell action, but in the same way an arch does.
Looking for the fact, that glass is able to take 10 times more pressure than tensile stress,
what effects a better resistance. But the radial supporting at the straight edges has to be
rigid enough, to guarantee, that the arch can act. Because of this, the substructure need
to be proper designed to take the horizontal loads from the glass, e.g. by the use of
tensile elements between the supports (figure 2b). The usual types of support, activating
the shell action of a curved glass, are to support both curved edges in radial direction, or
also additionally bending stiff, for supporting at one edge only, or to support all edges
of the glass element.
Figure 2a and b: Simple beam and arch action.
The shell action and the higher stiffness of the curved glass element, have some positive,
and some negative effects. Due to the higher stiffness of curved glass elements, the
deformation under external loads becomes less, and the mechanical resistance increases.
859
Challenging Glass 3
Otherwise, exactly these effects are negative for curved insulating glass units, because
the lower deformation under surface loads in radial direction, causes higher climatic
loads, as stiffer than the elements are [1]. Even if the resistance of a curved glass is
higher, the increasing inner load can effect higher tensile bending stress.
Generally you can say, as smaller the bending radius, as stiffer the element, as better the
resistance to tensile bending stress, as higher the climatic load in curved insulating glass
units. The increase of stiffness, and its effect to the needed thickness of glass is shown
in a very simple example below. A cylindrical curved glass is supported at the curved
edges, in radial direction, and is loaded by a surface load, acting radially, and
alternatively by a line load. The bending radius varies from 10000mm to 5000mm.
figure 3 shows the results for the tensile bending stress and figure 4 for the deformation
in comparison with the results from a flat glass plane with the same spanning and
loading. The results show a decrease to about 60% of the tensile bending stress of the
flat glass, and to about 35% of the deformation of the flat glass, by a radius of
10000mm only. The decrease is much higher by a radius of 5000mm (to about 30% of
the tensile bending stress, and about 13% of the deformation). So even a week curvature
has a high effect to the stiffness of the glass elements. More information about the effect
of curvature for curved insulating glass can be found in [1] and [2].
Figure 3: Results for tensile bending stress.
Figure 4: Results for deformation.
860
3.2. Resistance to tensile bending stress

The main criteria to proof and dimension glass elements on building structures, is the
permissible tensile bending stress. As a result of many research projects concerning this
point, in the last few years, curved glass, especially curved tempered and heat
strengthened glass, is a complete different product from flat glass. Regarding to the
production process, described in chapter 2, properties of flat glass products cannot be
transferred 1:1 to curved glass products. For the design of curved glass elements it can
be good practice, to use the characteristic values of tensile bending stress, shown in the
table below. Due to the different production procedures in the different glass bending
companies it is needed, to test and approve the tensile bending stress for the current
shape and glass configuration, using small test pieces, produced under the same
conditions, and with the same parameters as used for the originals. As special hint
should be given to the reduced tensile bending stress at the edges of float glass. This
will be also an important point in the discussion of proper modeling for the calculation.
Table 1: Permissible tensile bending stress fk (characteristic values, [N/mm2]) according to [4] and [5].
Type of glass Value for inner area Value at the edge
Curved floatglass (gb-Float) 40 32
Curved heat strengthened glass (gb-TVG) 55 55
Curved tempered glass (gb-ESG) 105 105
4. Methods of calculation
4.1. Depending on type of supporting

The linear elastic material behavior of glass offers the opportunity to use any well
known method to get the tensile bending stress, even for curved glass. But there is one
serious restriction: With the exception of single curved glass elements (no insulated
glass), supported at the straight edges only, without using the arch effect, every curved
glass element has to be proofed considering the shell action. Even if the curvature is
quite small, to simplify the curved shape as a flat one, leads to complete wrong results.
This effect is shown in figure 5. The maximum tensile bending stress, caused by a
surface load, occurs at the middle of the non supports edges, on the opposite side of the
loading, but also at the curved edges, on side of the loading. These effect would be
never found, idealizing the curved glass as a flat one.
Figure 5a, b and c: FEA-model, tensile bending stress on surface of load and on opposite surface.
861
Challenging Glass 3
It is not important how to consider the shell action of the shape. By using an analytic
way, table books or a mechanical software tool, that is able to generate curved shell
elements. But it is important to do it. Nowadays, to use a useful calculation software,
based on the method of the finite elements (FEA) will be normally.
As mentioned above, the stiffness of the glass increases very fast with the increase of
curvature. This has the effect, that the glass element is more sensible for deformation, or
different elasticity of the supporting and the substructure. Curved glass should not be
designed without any knowledge of the supporting structure, and the substructure. Most
problems with the use of curved glass are cause by the fact, that glass and substructure
are designed from different engineers, without any communication. In best case, the
complete structure is designed and calculated from one hand, but minimum the glass
structure including its support. The stiffness of the support should be considered in the
calculation, or designed to be uncritical for the dimensioning of the glass.
4.2. Depending on type of glass

As described in chapter 3, the amount of inner loads of an insulating glass unit is highly
depending on the stiffness of the glass panes, covering the included gas volume.
Regarding to the results of the example, comparing the results from flat and curved
glass, the need to consider the curvature in calculating the inner forces is evident, even
in the case of a very small bending radius. The theoretical bases to get the inner forces
in insulating glass units are comprehensively discussed in [1], [2] and [3]. At this point,
it is only mentioned, that generally, there are two methods to get the influence of the
enclosed gas. The first one is, to get the gas pressure “by hand”, using the general gas
equation, by considering the change of the enclosed volume, and the stiffness of the
single glass elements. The second one is, to build up a complete FEA-model of the
insulating glass unit, including the spacer and the edge sealant. Additionally it is needed,
to consider the influence of the type of support, and how to generate it. This method is
quite complex, because every material need to be generated with a proper material data.
But it offers the opportunity, to consider all effects, even the positive ones, like the
increase of the volume, due to expansion of the edge sealant, as described in [3].
5. Constructional aspects
5.1. Interaction substructure / glass

In general, curved glass elements need curved substructural elements. The substructure
will usually be made of aluminum, steel or timber. To bend aluminum or steel profiles
is also a complicated process, and the results will also have tolerances, as more, as
complicate the shape of the curvature is. So the first point in the planning process has to
be, to consider realistic tolerances of both structures, glass and substructure, to find a
proper solution for the design of the supports. If the substructure is already finished, it is
useful, to make a detailed 3D-laser-optical measurement, to get the actual shape. The
mostly used method, to fix glass elements to the substructure is, to hold it by screwed
bands. In case of unavoidable tolerances between glass and the substructural profile, and
when distance sleeves are missing, the uncontrolled pulling of the screws will effect an
unwanted loading to the edge of the curved glass elements. In the best case, this leads to
a prompt failure, so that the mistake is found during the installation. But in general,
such an insufficiency will not be found, before the glass fails, due to the occurrence of a
862
regular loading, what has been considered in the calculation, but not in addition to the
unwanted loading.
What is also a problem in the use of flat glass, becomes a bigger problem in the use of
curved glass, because the higher stiffness of the curved elements disables them to follow
the deformation of the substructure like flat glass would do. This means, that the
deformation of the substructure, or even the superstructure, has to be considered in the
design process for the glass.
5.2. Approved support constructions

Regarding to the aspects, discussed above, it is approved to use wet glazing systems for
the support of curved glass elements, mainly. To lay down the elements into a bed of
silicone gives the ability, to cover unavoidable tolerances, and guarantee a supporting,
that is elastic enough, to absorb smaller deformations from the substructure.
A further ability to get a proper support, is to use pre-fabricated solutions, where the
connection between glass and the direct substructural element has been done in factory,
under better conditions, where tolerances can be found and corrected easier. An
example is shown in figure 6. To support a vertical standing, curved glass element, a u-
section profile was fixed to the edge of the curved glass, using the silicone DC 993 in
factory. At the bottom of the Profile, a flat tin was fixed, to enable a bolted connection
to the base structure. The profile could be installed and adjusted in factory. The
thickness of the silicone was chosen thick enough, to take the tolerances between
aluminum and glass. The result was a very simple installation on site, where the bolts
had to be fixed only, without any risk, to overscrew them and cause any problems to the
glass.
Figure 6a and b: An example for an approved supporting.
863
Challenging Glass 3
6. Conclusion
As discussed above, the product curved glass has to be seen as separate glass product.
Characteristics from flat glass cannot be assigned to curved elements, without serious
investigations. To proof the characteristics, used in a calculation, normally tests with
specimen are needed. For the calculation of curved glass elements proper methods,
considering the shell action are needed. Simplifications of the mechanical model, e.g.
the assumption of a curved element to be flat, cannot be accepted and lead to wrong
results. The higher stiffness of curved glass elements causes higher climatic loads in
curved insulated glass units. Well known methods to calculate the climatic loads of flat
insulated glass cannot be assigned to curved ones. Support structures and substructures
have to be designed regarding the special requirements of curved glass. Especially the
tolerances between glass and substructure, and deformations of the substructure have to
be considered.
7. References
[1] Neugebauer, Jürgen, Climatic Loads in Curved Insulated Glasses, Proceedings Challenging Glass
Conference, Delft, Netherlands, 2008.
[2] Feldmeier, F, Klimabelastung und Lastverteilung bei Mehrscheiben-Isolierglas, Stahlbau 75 (2006),
Heft 6, Verlag Ernst & Sohn, Berlin, Germany, 2006.
[3] Neugebauer, Jürgen, The Influence of the Edge Sealing in Curved Insulated Glass, Proceedings
Challenging Glass Conference, Delft, Netherlands, 2010.
[4] Bucak, Ö., Feldmann, M., Kasper, R., Bues, M., Illguth, M.: Das Bauprodukt “warm gebogenes Glas” –
Prüfverfahren, Festigkeiten und Qualitätssicherung, Stahlbau Spezial (2009), Verlag Ernst & Sohn,
Berlin, Germany, 2009.
[5] Bundesverband Flachglas.: Leitfaden für thermisch gebogenes Glas im Bauwesen, BF-Merkblatt 009 /
2011, Troisdorf, Germany, 2011.
864
doi:10.3233/978-1-61499-061-1-865
Curved Glass in Four-Point Bending

Bernhard Weller, Michael Engelmann, Philipp Krampe, Stefan Reich
Institut für Baukonstruktion, Technische Universität Dresden, Germany,
michael.engelmann@tu-dresden.de, www.bauko.bau.tu-dresden.de
The European Standard EN 1288-3 regulates the four-point bending test for flat
glass in order to receive strength data. No testing code for determining material
properties of curved glass exists today. Therefore, the knowledge from flat glass
testing might be transferred to curved glass in order to reduce the testing
complexity. An adaption of the experimental set-up is possible and creates
reproducible results. Those can be used to create a finite element model for further
analysis. In conclusion it is possible to determine a simplified connection between
load and maximum tensile stress using newly defined correlation factors in the style
of EN 1288-3. By means of these factors strength testing of bent glass can be
performed according to the standard without applying strain gauges.
Keywords: curved glass, four-point bending, strength, effective stress
1. General
1.1. Motivation
Today there is no code available which describes design properties of curved glass.
Hence further tests and investigations have to be performed in order to use this product
in modern transparent architecture. This will increase the cost and the effort to prove
that the construction is safe.
On the other hand testing of flat glass to determine the glass strength is a common
procedure and can be performed by analyzing the maximum force applied rather than
measuring the strain directly. In order to reason from strain to stress in curved glass
bending directly strain gauges have to be set in place. This extra effort is to be
minimized by this research [1].
Figure 1: Curved glass in four-point bending at failure.
865
Challenging Glass 3
1.2. Intention of Research

The European code EN 1288-3 [2] describes the four-point bending test for flat glass
based on results by Kurt Blank [3]. In his research analyses of strain measurement and
breakage behavior as well as numerical calculations were performed to gain a stress-
load relationship. Therefore using the EN 1288-3 [2] testing method no strain gauges
have to be applied during flat glass testing. A method is presented to determine the
maximum tensile stress on the surface and at the edge during failure of flat glass by
analyzing the force applied and the origin of failure. Using this method bending strength
data is produced including the load bearing behavior and the failure properties of glass.
However, the EN 1288 testing procedures were developed for flat glass and cannot be
transferred directly to curved glass. Even the draft of ISO/CD 11485 [7] does not
contain a “standardized method of evaluation/measurement of the mechanical strength
of curved tempered glass”. The aim of the presented research is to adapt knowledge
from flat glass testing to curved glass in order to reduce the effort and to describe a
common basis for testing.
1.3. Course of Action

The experimental set-up from flat glass testing has to be altered for curved glass first.
Furthermore, the distribution of mayor tensile stress has to be determined using a
numerical model which represents the experiment. Therefore, data from testing has to
be compared to those from modeling. Once the model shows sufficient accuracy, once it
is verified, this data is used to analyze the load bearing behavior of curved glass during
four-point bending. Therefore, testing and calculations including a wide parameter
spectrum were carried out [1].
2. Failure of Glass
2.1. Strength of Glass

The term strength of glass is defined by the European code EN 1288-3 [2] as the
bending stress or effective bending stress leading to failure. In general, strength of glass
is not a material constant but a description of the quality of the surface. So the strength
depends on the geometry used, loading history, and speed of loading as well as further
factors. Moreover, the edge of the glass usually contains more critical surface flaws due
to mechanical processing. Therefore, there are different values for the strength at the
edge and on the surface.
The manufacturing of curved glass might lead to further damage. This is potentially one
reason why today’s strength values for curved glass are about 15-20 % lower than those
for flat glass products [4], [5], [9].
866
2.2. Effective Stress

While the quality of the surface determines the strength of glass this property cannot be
described in a detailed way because of the microscopic characteristic of the mechanical
flaws on the surface. They are not distributed homogenously but randomly on the
surface. Thus the strength of glass is distributed randomly as well and leads to a
strength design value that limits the probability of failure only rather than limiting the
stress that is applied. For mathematical analysis a Weibull distribution is commonly
used and is used for this research accordingly. Therefore, a strength design value called
effective stress Veff (Equations (1), (2); [3]) unites two characteristics:
x load bearing behavior in terms of principal tensile stress VI and Vx on the

surface AK and at the edge la,
x random breakage in terms of Weibull shape parameter E
1
ª 1 ºE
V eff .surface « ³ V IE dA » (1)
«¬ K AK
A »¼
1
ª1 ºE
V eff .edge « ³ V xE dx » (2)
«¬ la la »¼
AK … tested surface area

la … edge between rollers
VI ; V x principal tensile stress (plane; linear)
E… shape parameter (Weibull distribution)
Hence the effective stress acting homogenously on the surface leads to the same
probability of failure as the real load [3]. The design limits the effective stress rather
than the local limit of strength.
2.3. Bending of Flat Glass

The four-point bending of flat glass and the determination of an effective stress is
analyzed in [3]. The stress distribution on the glass surface between the load-applying
rollers results from nonlinear finite element calculations. They are verified by
conducting experiments and measuring the load bearing behavior. An effective stress
value is created this way separately for the edge and the surface. These values are
compared with the result from approximation using linear beam theory (Vb):
V eff .surface
kS (3)
Vb
V eff .edge
ke (4)
Vb
867
Challenging Glass 3
In case of the center plate stress the kS factor (Equation (3)) is sufficiently small to
allow for an approximation of the stress at this location using linear beam theory with
satisfactory accuracy although the Bernoulli hypothesis is not valid for glass beams
sized for four-point bending. A mayor difference (ke > 1,05) was just found in case of
the edge stress (Equation (4)). In order to give an approximation the result has to be
altered using the ke-factor which can finally be found in Figure 3 of EN 1288-3 [2].
2.4. Bending of Curved Glass

Bending of flat glass is just one special case within the possible geometry parameter set
of curved glass. The stress distribution during this test depends on the curvature. Thus
two different set-ups for curved glass testing need to be considered leading to different
stress distributions (Figure 2).
Figure 2a, b and c: convex (R > 0), concave (R < 0) and flat position (R ).
In comparison to the flat glass analysis the stress within a curved sample is determined
by Equation (5) [10]. Looking at the possible parameter range for curved glass during
four-point bending according to EN 1288-3 [2] the problem can be handled according to
a curved beam with a low curvature [1] by Equation (6).
N M § 1 t 2 ·
V b.curved ¨1 ¸ (5)
A R A ¨© N R t 2 ¸¹
N M
V b.low curvature (6)
A W
N normal force
M bending moment
A cross section area
W… statical moment of cross section area
R… radius
t… glass thickness
N… stress factor for curved beam
Furthermore, there is a horizontal component AH from the support force in a curved

glass test (Figure 2). This component increases the bending moment in the center of the
beam due to an extra cantilever arm caused by the rise of the arc and induces a normal
force. The deflection of the glass happens in the same direction making the use of
second order theory necessary to cover effects from deformation.
868
3. Extension to Curved Glass
3.1. Course of Action

For an approximation of stress and deflection the linear beam theory equation is altered
to include the curvature of the specimen. During a second step an effective stress
Vcurved.eff for curved glass of different thicknesses and radiuses is determined. Therefore,
a numerical model is created on the basis of [4]. This parameterized model is used for
the special case of flat glass first in order to reproduce the calculation in [3].
The results from curved glass modeling are to be verified by experiments where the
origin of failure is observed as well. The experimental set-up from EN 1288-3 [2] is
analyzed and adjusted [1], [4]. Finally a load-stress relationship is proposed using
k-factors for curved glass [1] in the style of EN 1288-3 [2].
3.2. Beam Theory

Basis for the linear beam theory is the hypothesis of Bernoulli which states that the
cross section remains even during bending. Thus there is only stress in the direction of
the span. The strain perpendicular to the longitudinal axis of the beam might happen
without disturbance because the width and the thickness of the cross section is much
smaller than the span and thus not resulting in high stress. The sections of the glass
specimens in EN 1288-3 [2] contrariwise are 360 mm wide and spanning 1000 mm. In
this case plane stress occurs that is not covered by Bernoulli beam theory. Using the
method the bending stress result Vcurved.b might need to be adapted to achieve sufficient
accuracy.
Equation (1) from EN 1288-3 [2] is extended for the use of curved glass during this
research in order to give an approximation of the load bearing behavior and the
deflection. Furthermore, this value will be a reference for a k-factor for curved glass in
the style of EN 1288-3 [2] and [3] as presented in Equation (7) and Equation (8).
V curved .eff .surface

kcurved .S (7)
V curved .b
V curved .eff .edge

kcurved .e (8)
V curved .b
4. Experimental Results
4.1. Experimental Set-Up

According to EN 1288-3 [2] the glass spans 1000 mm between two bearings (Figure 5)
which are fixed in translation but can rotate without disturbance (Figure 3, Figure 4).
This way a frictionless connection between the glass and the bearing is assumed for
later calculation. Resulting support forces are oriented perpendicular to the glass only.
This set-up is widely used and can be taken as commonly accepted (e.g. [4], [5]).
The load is applied by two rollers at a distance of 200 mm in a way that the stress
increases with a speed of approximately 2 N/mm2·s.
869
Challenging Glass 3
Contact between the glass and the metal rollers is avoided by a striped piece of
elastomer (EPDM, t = 5 mm, not reinforced, Shore-A 70) which is fixed all around the
roller in order to accommodate large rotation angles of the bearing.
Figure 3: experimental set-up (concave). Figure 4: roller bearing.
4.2. Samples
Various samples were chosen to verify a wide range of parameters. Consequently there
were 28 samples from 4 mm to 12 mm of thickness and three types of radiuses
according to Table 1. All samples were made of fully tempered glass with polished
edges.
Table 1: Number of Curved Specimens.
Radius Specimen Thickness [mm]
[mm]
4 6 8 10 12
840 3x 3x - - -
1016 3x 3x 2x 3x 2x
2000 3x 3x - 3x -
4.3. Measurement
The strain is measured by up to eight uni-directional strain gauges which were applied
according to Figure 5. First of all the difference between center and edge strain was
recorded. Subsequently the difference between strains of both surfaces was noted with
the purpose of reasoning to membrane stress.
The deflection was measured in the center of each plate and at the edge between the
rollers using linear displacement transducers. By using this method a deflection of up to
100 mm can be recorded which equals a relative displacement of 100 mm / 4 mm = 25
in case of the thinnest plate and accommodates the highest relative displacement in
Figure (3) of EN 1288-3 [2].
870
Figure 5: position and numbering of strain gauges (top & bottom side).
Using a Scattered Light Polariscope (SCALP-04) residual stress at the surface of all
fully tempered samples was measured. Furthermore, the dimensions were measured in
order to check geometry accuracy.
At first non-destructive tests were performed up to a load of 80 % of the predicted
failure load (Fn/u.max). This test was repeated three times in both positions (convex and
concave) to verify the experimental set-up and make sure that the results are
reproducible. Later on the load was increased up to failure (Fn/u.fail). A thin plastic foil on
the compression side of the sample allowed the definition of the origin of failure.
4.4. Results
The residual stress on the surface was analyzed according to [6] and showed a mean
value of Vv = -126,7 N/mm2 and a standard deviation of s = 30,18 N/mm2. The
minimum residual stress level according to EN 12150 was achieved but the results show
that its distribution within the surface is not homogenous. In [8] there is a more detailed
investigation of measuring stress level and distribution in tempered curved glass.
The correlation between load and strain at the edge and the surface is almost
linear (Figure 6). There is just a slight curvature so all results were approximated
without major difference using a quadratic polynomial function. In convex
configuration the rise of the arch is reduced with increasing load resulting in smaller
leverage arms. Therefore a stiffening effect is recorded. On the contrary during concave
arrangement the samples react less stiff with increasing load. This is the expected result
according to [4] and [5]. The conclusion of different strength values for convex and
concave configuration [5] is a matter of the mechanical system first and not necessarily
a matter of the material itself.
Looking at the deflection results the same general behavior with increasing stiffness at
convex configuration and decreasing stiffness at concave arrangement was
871
Challenging Glass 3
recorded (Figure 7). An approximation using second order polynomials was done as
well (black continuous line). Most of the 4 mm and 6 mm samples did not fail within
the maximum measureable deflection of 100 mm.
Figure 6: experimental load-strain relation. Figure 7: experimental load-deflection relation.
In summary the tests are repeatable and the defined set-up proved its usability. Samples
in convex arrangement are not reacting stabile. Consequently this configuration is not to
be used. The horizontal fixing in the middle of the beam which is used for calculations
is not feasible constructionwise during the real test (see Figure 2). The experimental
set-up is free to move but theoretical stabile during the specific load case. So
imperfections in the real life sample are leading to a sudden kinetic behavior.
Furthermore, the origin of the failure was just observed in a very small number of
samples. Due to the large deformation up to 100 mm of the thinner samples (Figure 8)
and the high level of destruction of the thicker samples (Figure 9) no general results
concerning the origin of failure were obtained.
Figure 8: large displacement. Figure 9: high level of destruction.
To verify the correlation of load and strain for the whole parameter spectrum further
tests are necessary.
872
5. Numerical Results
5.1. General
The modeling of the support conditions has a mayor effect on the results [1], [4]. During
the experiment the bearing rollers remain in place while the specimen rolls over them
during loading. Therefore, a geometrical nonlinear contact approach has to be
implemented to find realistic results and represents the correct load bearing effect. The
FEM software SOFiSTiK 23 with 4-node shell elements (QUAD) and nonlinear spring
elements (FEDE) was used. A quarter of the whole glass specimen was modeled
assuming double symmetry (Figure 10).
Figure 10a, b and c: contact model; convex(R > 0), concave (R < 0) and flat position (R ).
5.2. Results for flat glass

In the first step a flat glass model was produced in order to comprehend the results
from [2] and [3]. Firstly a linear support condition was chosen according to Figure 11 c.
Figure 11a, b and c: deformed beam assuming linear support conditions.
The support remains fixed at the smaller edge of the model and displaces horizontally
once the plate is loaded vertically. While a calculation using second order theory is
performed the leverage arm of the support force shortens with increasing deflection at
every iterative step. The support force is in line with the load and does not allow any
lifting of the glass. This does not meet the experimental requirements exactly but meets
the experimental results for flat glass panes sufficiently well [3].
The principal tensile stress at the surface was calculated and an effective stress was
assessed according to the previously defined scheme. As a result k-factors for flat glass
are printed in Figure 12 and Figure 13 (continuous lines) in comparison with results
from EN 1288-3 [2] and [3] (dashed lines). The difference between both results is
sufficiently small (< 5 %) to conclude that this approach is suitable for a complex
curved glass study.
873
Challenging Glass 3
Figure 12: recalculation ke factor. Figure 13: recalculation kS factor.
5.3. Results for curved glass

In a second step a model for curved glass was created using the very same linear bearing
conditions. The support force is directed towards the center of the circle defined by the
curvature of the glass arch. This situation does not change while the model deflects
during a second order calculation. Therefore, the leverage arm of the support forces
decreases in case of a concav curvature (see Figure 11) and increases in case of a
convex one.
Figure 14: deformation at contact support.
The support force is not in line with the load. Its horizontal component causes an extra
bending moment in the center of the glass. Using linear conditions the load bearing
mechanism from the experiment is not represented sufficiently well (Figure 15).
Hence, in a third step, a curved parametric model was created using recommendations
from [3] and [4] which include a contact approach at the bearing. The leverage arm of
the support force changes as well during the loading but it remains within a physically
possible area which is limited by the size and the radius of the bearing rolls.
Because of discretization and the use of the contact approach there is no continuous
load-stress coherence. The results are smoothened by describing them in the same
manner as the experimental results using second order polynomials. Additionally in this
way an analytical comparison between experimental and numerical results is possible.
The load bearing situation is simulated and compared with the experimental results with
sufficiently small difference (Figure 15 right). Occasionally a significant deviation is
observed. The reason for that is probably the relatively small numbers of specimens that
874
lead to a large influence from runaway values. Therefore further tests have to be
performed to verify the model in the whole parameter spectrum.
Figure 15: experimental and numerical results, using linear support and contact approach
In summary a model including a contact approach represents the real experiment

sufficiently exact and is therefore used for further investigations.
In all available studies the stress was measured at the axes of symmetry only and
compared to the finite element results. For the determination of the two-dimensional
tensile stress situation between the rollers a larger area is under
investigation (Equation (1), (2)). So, especially the area of load application is not
verified. The load is modeled by a linear load and distributed by the software to the
nodes.
During the experiment the loading machine remains in place and thus does not obstruct
the horizontal movement of the glass pane. So the horizontal distance between the
rollers and the center of the glass remains constant during the experiment. This effect is
neglected for modeling because the respective distance is much smaller than in case of
the support rollers.
Furthermore, a constant load (deflection or linear load) is assumed for numerical
calculations. Due to the shell effect and plane strain the glass may be curved
perpendicular to the span as well. This will lead to a non-constant loading. This effect is
not investigated in detail today (e.g. by using a contact approach). Its impact on the
result is assumed minor during this investigation.
875
Challenging Glass 3
In Figure 16 and Figure 17 k-factors for flat 6 mm thick shell elements on concave
position are presented in three different aspects:
x Results from flat glass testing (EN 1288-3 [2] and [3]),
x Results for flat glass using linear support conditions (recalculation),
x Results for flat glass using a contact approach (a radius of ± 10000 mm defines
an upper and a lower tolerance for flat glass modeling).
Figure 16: kcurved.e factor, concave (t = 6 mm). Figure 17: kcurved.S factor, concave (t = 6 mm).
Up to a relative deflection of about w/h = 5 all curves are proceeding pretty close so no
mayor difference is observed. It can be concluded that the type of modeling (linear or
contact bearings) have no mayor effect on the result. Up to this state a tensile stress of
about 73 N/mm2 is approximated using Equation (5). This stress value is considerably
lower than the expected stress at failure of fully tempered glass. Hence the specimen is
expected to carry this load in the experiment. For relative deflections of w/h > 5 the
k-factor from [2] and [3] drops while the newly defined kcurved-factor remains
considerably high and even increases further once the load is increased. Therefore, the
approximation using the beam theory result in lower stress values than those that would
be measured during the experiment. The strength of the glass is assumed lower. The
results for linear supports leave the expected range at higher loading levels.
Consequently this way of modeling will result in more conservative stress values at
higher loading levels. Once a complete set of k-factors is produced for curved glass the
results for flat glass have to be reviewed in order to root in the same calculation concept.
In fact both, the kcurved.e and kcurved.S factor increase because of the effect from second
order theory and a conduct of decreasing stiffness with increasing deflection as well as
further factors.
In Figure 17 the strain along the y-axis from the center of the plate (y = 0) to the
edge (y = 90 mm) from the experiment (+ ×) and from numerical calculations
(continuous line) is depicted for increasing load levels. During the whole process the
stress level at the edge in concave configuration is usually higher than on the
surface (Figure 19). Thus the weakest part of the glass pane is stressed more leading to a
higher probability of failure at the edge.
On the other hand there are samples with a higher stress level in the center than at the
edge (Figure 18). While the relation of the stress level at both locations (center and
edge) reaches the relation of strength of the edge and the surface the probability of
876
breakage is constant within the whole glass pane. Furthermore there are samples where
the location of the higher probability of breakage switches from center to the edge or
vice versa during loading. In Figure 18 a sample is depicted where the stress level in the
center is higher than at the edge in the case of a relative small loading level of about
20 % Fmax. At a load of 60 % Fmax the values are almost constant and with further
increasing load the edge is stressed more. Thus a failure at the edge is more probable
assuming that the strength of the edge is lower than the strength of the surface.
Figure 18: strain y-axis convex configuration Figure 19: strain y-axis concave configuration
(R=2000 mm t=6mm) [1]. (R=2000 mm t=6mm) [1].
Those charts are made for fully tempered glass samples. Annealed glass is expected to
fail at a load of about 40 % Fmax. In this case the probability of failure is almost the
same for the center and the edge. Consequently while annealed glass fails at a lower
load level the probability of the location of failure might differ from a specimen of same
geometry but with a different residual stress situation. The tempering of the glass allows
a higher level of load and determines as well as the geometry the result from four-point
bending of curved glass. Consequently it is not possible to tell by the curvature and
thickness only whether a failure is more likely to fail at the edge or on the surface.
Further factors such as glass type and residual stress have to be taken into account.
5.4. Summary and Further Prospects

The results for flat glass that are published in EN 1288-3 [2] and [3] were reproduced
with adequate accuracy. Afterwards a model including a contact approach was created
and it is able to calculate the experimental results with good agreement. However, the
use of this flat glass model results in differing k-factors for high loads. Usually the edge
of curved glass is stressed more during four-point bending which leads to a higher
probability of failure at this location. But there are specific combinations of radius,
thickness and test-configuration where the location for the highest probability of failure
switches during loading from the surface to the edge or visa versa. Thus further factors
have to be taken into account for strength analysis.
In summary the load bearing of curved glass in four-point bending can be summarized
in a set of kcurved-factors in the style of EN 1288-3 [2].
877
Challenging Glass 3
6. Conclusions
The experimental set-up is serviceable in concave configuration. Using a geometrically
nonlinear contact model the strain during the test can be calculated with sufficient
accuracy. In conclusion a set of k-factors for curved glass in the style of EN 1288-3 [2]
can be presented. But for evaluation of the whole parameter set further investigations
have to be made in order to verify the modeling completely.
The location of the higher probability of failures of curved glass might change for
certain geometries and during loading. Therefore, no general information can be
provided whether a sample is more likely to fail at the edge or on the surface.
7. Acknowledgements
Support by Bundesverband Flachglas and Prof. Dr.-Ing. Bernd W. Zastrau (Institute of
Mechanics and Shell Structures, Technische Universität Dresden) is gratefully
acknowledged.
8. References
[1] Engelmann, Michael, Gebogenes Glas im Vierpunkt-Biegeversuch, Diploma Thesis, Institut für
Baukonstruktion, Technische Universität Dresden, Germany, 2011. (not published)
[2] EN 1288-3, Glass in Building – Determination of the bending strength of glass – Part 3: Test with
specimen supported at two points (four point bending), 2000.
[3] Blank, Kurt; Dürkop, Detlev; Durchholz, Michael; Grüters, Hugo; Helmich, Gerd; Senger, Wolfgang,
Strength tests of flat glass by means of four-point bending, Glastech. Ber. Glass Sci. Technol. 67 No. 1,
1994, pp. 9-15.
[4] Baatz, Andreas, Experimentelle und numerische Untersuchungen der Biegezugfestigkeit von gebogenen
Glasplatten, Diploma Thesis, Lehrstuhl für Stahlbau, Rheinisch-Westfälische Technische Hochschule
Aachen, Germany, 2002. (not published)
[5] Bucak, Ömer, AiF Abschlussbericht – Trag- und Resttragfähigkeitsverhalten von unterschiedlich
gebogenen Glasscheiben im Bauwesen, Entwicklung von Berechnungsmethoden, Prüf- und
Qualitätssichernde Kriterien, FH München, 2009.
[6] DIN 53804-1, Statistische Auswertung – Teil 1 Kontinuierliche Merkmale, 2001.
[7] ISO/CD 11485-3, Glass in building – Curved Glass – Part 3: Requirements for curved tempered and
curved laminated safety glass (Draft), 2011.
[8] Weller, Bernhard; Krampe, Philipp; Engelmann, Michael, Prüfung vorgespannter, gebogener Gläser, in:
Weller, Bernhard; Tasche, Silke (editors), Glasbau 2012, 2012.
[9] Bundesverband Flachglas (editor), Leitfaden für thermisch gebogenes Glas im Bauwesen, 2011.
[10] Dankert, Jürgen; Dankert, Helga, Technische Mechanik. Statik, Festigkeitslehre, Kinematik/Kinetik,
Teubner, Wiesbaden, Germany, 4. korrigierte und ergänzte Auflage, 2006.
878
Challenging Glass 3
Architectural Design &

Lighting
879
doi:10.3233/978-1-61499-061-1-881
Glassolutions LED-in-Glass Outdoor at

Oskomera Group HQ
Rens Demarteau, Esther Hebly
Oskomera Group B.V., Netherlands, info@oskomera.com, www.oskomera.com
Paul Roman, Rino Messere
Glassolutions, Netherlands, www.glassolutions.nl
The Oskomera Group moved into her new headquarters around Christmas last year.
One of the distinguishing features of this new building is the LED-in-Glass
Outdoor (LIGO) glazing supplied by Glassolutions. This special glazing is a part of
the ‘glass box’ inserted into the mass of the main building. It shows the corporate
logo, very clearly at night, subtly by day.
Keywords: Glass, LED, façade integration, media facade.
1. Introduction
When one approaches the Oskomera Group headquarters during daylight hours, LED-
in-Glass may elude the less observant visitor. One might notice the subtle letters of the
company logo, not suspecting the innovative lighting system that they represent. During
low-light hours and night, that is when LED-in-Glass really shines, as the illuminated
window panes show their stuff and become the eye catcher they were designed to be.
1.1. Oskomera Group

The Oskomera Group is a technical specialist that develops, engineers, manufactures
and installs facades and load bearing constructions and aluminum and steel. Oskomera
is also a major supplier of solar installations.
When the Oskomera Group, which has its home in Deurne, the Netherlands, decided to
have her new company headquarters built, she wanted to show off some of the
possibilities of their product range. As said above, she builds facades, which also
includes integrating all kinds of architectural features.
That is where Glassolutions came in.
1.2. Glassolutions
Glassolutions is one of the leading manufacturers and distributors of glazing panes in
Europe. But they are also experts in creating all sorts of decorative glazing, with a high-
tech specialization. When they approached Oskomera with a special lit glazing system
which is called LED-in-Glass Outdoor (LIGO), the connection between Oskomera’s
project and the Glassolutions quickly became evident.
881
Challenging Glass 3
Figure 1: The LED-in-Glass façade can be clearly seen at night
2. The Project
2.1. Description
The LIGO façade within the Oskomera Headquarters consists of a relatively small
number of panes, 16 in total, 2 per character of the company name. This configuration
was chosen for two reasons. The first was the rhythm of the façade, which has a narrow,
high push-out panel every two windows. The other reason was the height of the
characters. As the Oskomera font is very wide, a narrow window quickly restricts
character height.
2.2. Design requirements

Integrating a product like LIGO in a façade requires very meticulous designing. One of
the primary functions of a façade is to keep water out. But the functionality of LIGO
contradicts this, as, to light the LED bars, power is required. This is supplied by a cable
which must come from the inside and thus must go through the protection layer. But
just watertight isn’t enough, as an air leak will, because of the pressure difference,
invariably lead to water leakage.
Also, aluminum facades are usually chosen because they are very slim, which leads to a
very limited amount of space to put these cables and their entrance openings. Therefore,
it was chosen to lead the cables to the outside and lead them over the façade into the
building. Of course this only works with low facades of maximum three stories, so with
higher facades a more complex solution is required.
882
Glassolutions LED-in-Glass Outdoor at Oskomera Group HQ
Figure 2: The Cables from the LED bars are transferred first to the outside of the façade,
then over it to the inside of the building.
3. The Product LIGO
3.1. Glazing panes

The product LIGO consists of two parts working together; the input side, provided by
the electronic controls, and the output side, which consists of the window panes with
their attached LED bars.
The window panes for LIGO consist of a double glazed unit, of which one pane is fitted
with LED bars along its edge. These LEDS emit their light into the side of the window
pane. A special print on one side of the window pane then lights up when it is hit by the
LED light. The number of LED bars on one window pane is determined by the height of
the window and the size of the lit area. Each bar is 45 cm long and that determines the
number of bars required on each side of the window. To get an even lighting, LED bars
on both sides of the window pane are required. Each bar has its own cable.
Each LED bar can light up prints up to a distance of 90 cm at the moment. No LED bars
can be placed under the window pane, due to weight and water drainage restrictions to
the cables. This means the maximum width of a fully lit pane is 180 cm.
883
Challenging Glass 3
3.2. Control
Each LED bar is controlled through 4 wires integrated in the cable: red, green, blue, and
white. These cables lead to a DMX box, which can be directed by the computer
software required for managing the LED installation. The DMX boxes come in sizes of
eight and twelve, and can be joined to create larger LIGO installations.
This software is specially designed for this product, and can be installed on a computer,
but can also be used from a tablet computer or Smartphone. It is also possible to
combine this software with existing Building Management Systems to provide
additional flexibility.
3.3. Lighting Possibilities

The flexibility of the LIGO control program makes it possibel for a wide range of
lighting schemes to be used. Colors can be changed on the fly, with every LED bar
emitting a different color. Also the lighting intensity can be altered per LED bar,
making it possibel to highlight certain sections.
The color pattern can be changed fluently, making it possible to “animate” the lighting
pattern. This can be programmed to happen very slowly up to very quickly.
Figure 3a, b and c: DMX box, single color lighting, and multicolor lighting.
This wide range of possible lighting is achieved with a minimum of power input
required, as all lighting comes from LEDs. The 24 meter long Oskomera façade can be
fully animated by a power supply of less than 1000 Watts, with power requirements
depending on animation intensity and color setting.
4. Conclusion
After installation in November, Oskomera has had the dark days of winter to experience
and experiment with her new façade, and the product itself has been an unqualified
success.
Apart from some minor cable and installation optimizations, installation and usage of
the building feature has been smooth sailing, and the Oskomera building will be
showing her logo for a long time to come.
Glassolutions has had the invaluable opportunity to experience the issues and
procedures that come with installing her new LIGO product.
All in all, LIGO adds another dimension to interactive façade lighting.
884
doi:10.3233/978-1-61499-061-1-885
Special Glazing
Christian Eckhardt
Evonik Industries AG, Germany, christian.eckhardt@evonik.com
Transparency is something like nothing. Transparency gives the impression of

lightness and liberty. Reflections and lense effects play with the transparency. In
different examples we want to give ideas what can be done with special effects. In
some cases you want to make aware the transparency, in some cases the glazing
should not catch someone’s eyes. Often the mix of both represents the tension
architects want to show.
Keywords: Plexiglas, Mega-glazing, formed transparency
1. Introduction
The architecture digs for new transparent materials for special applications. In different
examples this paper gives ideas what can be done with special effects. One example is
mega glazing with the dimension of 3.8 m by 17 m in Raiding, Austria. This glazing
opens a concert hall to the birth place of the composer Franz Liszt. Another example is
a glazing with convex and concave finished surfaces. These surfaces give the effect that
the viewer walking along these blocks has the impression to look into water. It is
installed at the flagship store of the flagging producer Raab Karcher in Berlin. Another
flagship store in London has a facade with milled Plexiglas. The milling surfaces have
different designs, for examples matt and transparent surfaces. The whole façade looks
like a huge product code which can be illuminated at night with a very special mix of
transparency and translucence. All those projects have very special specifications for a
vey special glazing.
One of the plastic materials that can be used in those outdoor applications is PMMA,
also known as acrylic or under the brand Plexiglas. We want to give an overview about
the material with the basic characteristics and detail designing with some examples.
2. Material PMMA
Polymethyl methacrylate, PMMA for short, is the chemical name of a highly
transparent plastic. Brand names for it are Plexiglas® or Acrylite® for example. PMMA
belongs to the group of transparent thermoplastics and stands out due to good
weathering resistance and a relative hard surface as well as its transparency. For sheet
material there are mainly two manufacturing processes, extrusion and cast.
The material characteristics are dependent on the type of manufacturing and the
environmental influences such as temperature and exposure time. The most important
material characteristics under standard climate conditions according to DIN 7823 are
summarized in Table 1.
885
Challenging Glass 3
Table 1: Material characteristics according to DIN 7823

Extruded PMMA Cast PMMA Unit
Young´s Modulus 2900 3000 MPa
Tensile Strength 60 70 MPa
Tensile Strain 2 4 %
Flexural Strength 100-115 100-115 MPa
Density 1,19 1,19 g/cm³
-5 -5
Thermal Expansion 7 x 10 7 x 10 1/K
Coefficient
In Figure 1 the dependence that Young’s Modulus has with the temperature is shown.
At 23°C you can see a young´s modulus of 3300 MPa. At 60°C it drops down to 2500
MPa.
Figure 1: Temperature dependency of young´s Figure 2: Relaxation. [3]

modulus. [3]
The strain can be divided into three different parts. An energy elastic strain occurs
immediately when a load is applied. The second part is an entropic elastic strain, with
which the position of the molecule chains relative to one another is changed. This part
of the strain is slowly reduced after the load is relieved. The third part is a viscous flow
that represents an irreversible deformation. In Figure 2, the stress relaxation over the
duration of the load is plotted. With a steady deformation, the stress is relaxed
significantly over time. The relaxation is dependent on the level of stress that is applied.
PMMA has a very high transmission, which results in the material heating up only a
very small amount in the sun.
3. Examples
3.1. Konzertsaal Liszt, Raiding

This concert hall in Austria is homage to the artist Franz Liszt, one of the concise artists
of the 19th century. To improve the link to the artist, the hall has a huge window to the
house of birth. This megawindow was made of Plexiglas® and has the dimensions of
17 m by 3.7 m with a thickness of 50 mm. With a transparent glue line the view is
undisturbed. With this gluing technique it is possible to produce very huge windows.
Actually it is not really a gluing line but cold welding because the material has a real
connection, not only interplay adhesion.
886
Special Glazing
Figure 3: Concert hall Raiding, Austria.
3.2. Raab Karcher Store, Berlin

The company Raab Karcher opened their first Flagshipstore “home couture” in the
middle of Berlin 2004. In this showroom the newest flagging is displayed. The basic
design element for the store is the shop window with a deforming effect. A 50mm
Plexiglas sheet was milled, formed and polished to get convex and concave surfaces.
This form has the effect of a moving room while passing the window and adjusts the
view into a swimming pool. In PMMA you always have the possibility to polish the
surfaces. A disturbed surface – by a milling machine or scratches – can be repaired.
Figure 4: Raab Karcher Store, Berlin.
887
Challenging Glass 3
3.3. Reiss Façade, London

The fashion label Reiss built his new flagship store in the middle of London, GB. The
facade of the store had to illustrate the brand philosophy into architecture.
The idea for the light transmitting facade with the machined Plexiglas® was born by the
London architects Squire & Partner. The whole façade is routed with different structures
in the surfaces. Those different structures give a linear impression. The refraction of
light causes by different routing patterns creates a constantly shifting play of light for
the viewer, both outside and inside, depending on the light situation.
Figure 5: Reiss Façade, London.
3.4. Lichtkirche, Germany

The first movable church in Germany wants to illustrate that religion comes to the
people. A very important implementation was not only the moving idea of a church but
also the lightness of the building. This lightness was realized with a translucent wall
made out of Plexiglas®. The church was built by the Protestant church in Hesse and
Nassau. In 2010 she was displayed at the Landesgartenschau (Regional Garden Festival)
in Bad Nauheim, Germany.
Figure 6: Lichtkirche, Germany.
888
Special Glazing
4. Summary and Outlook

When you want to make something special you often need a special material. But if you
use those special materials it is absolutely important to know all the details of the
material. Otherwise you can not design in accordance with the material. This
contribution wants to give you a feeling about the material Plexiglas, first of all with the
material properties but also with some samples of realized buildings.
The possibilities with the material Plexiglas are manifold. One unique application is the
windows for submarines. Those windows are very thick and have a form of a
hemisphere or a whole bowl. In the bowls you sit inside the ocean and have a really
perfect view around. Those very special applications show the possibilities and can be
adapted in new fields of use in the building industry.
5. References
[1] Wörner, Jan; Stahl. Jochen; Eckhardt, Christian, Structural Transparency – A new Wood Plastic
Composite Girder, Proceedings of Challenging Glass 2, Delft, Netherlands, 2010.
[2] Eckhardt, Christian, Ein Beitrag zur Konstruktion von Trangsystemen unter Verwendung von
transparenten Thermoplasten, Dissertation, Darmstadt, 2009
[3] Evonik Industries, Results of internal tests
889
doi:10.3233/978-1-61499-061-1-891
Architectural Aspect of Structural Design of

Glass Façades / Glass Skin Applications
Aleksandra Krstic-Furundzic, Tatjana Kosic, Jefto Terzovic
Faculty of Architecture, University of Belgrade, Serbia,
akrstic@arh.bg.ac.rs, tkosic@arh.bg.ac.rs, jefto@arh.bg.ac.rs
The objective of the work is to study the different solutions of the design and
construction of structural glass facades. In the paper different forms, assembling
and connection types of structural glass facades are discussed. On the basis of
analyses of the geometry of different shapes, position of facade surface, structural
facade functions and supporting system, the general classification of form types is
given. According to number of layers, a ventilation and connection option, the clear
idea of the assembling types is achieved. Based on form types, structural system
concept and stability is analyzed. The diversity of combinations of structural glass
components is the potential for different options of facade transparency and
building appearance.
Keywords: Structural glass, Facades, Form types, Assembly types, Structural

system concept and stability.
1. Introduction
While glass has been used as a building material for centuries, the greatest impact on
building envelope design was achieved during 20th century. The new approach to the
use of glass facades has been made at the end of last century (1980s) when their
structural properties became a matter of serious researches. Since that time the
functional aspects of glass facades, such as achievement of optimal indoor daylighting,
visual and thermal comfort condition, reduction of solar gains, saving energy etc, have
been expanded by improving the structural characteristics that includes load-bearing and
wind loading behavior of glass components and their structural safety.
The objective of the work is to study the different solutions of the design and
construction of structural glass facades. The paper is intended to serve as state-of-the-
practice summary regarding different forms, assembling and connection types of
structural glass facades and in that sense the paper is structured. On the basis of analysis
of the geometry of different shapes, position of facade surface, structural facade
functions and supporting system, the general classification of form types is given.
According to number of layers, a ventilation and connection option, the clear idea of the
assembling types is achieved. The typology of structural glass component connections is
based on the fact that the interdependence of structural glass component types,
connection place and technology type is crucial. Based on form types, structural system
concept and stability is analyzed. The diversity of combinations of structural glass
components is the potential for different options of facade transparency and building
appearance.
891
Challenging Glass 3
Methodological approach entails three steps:
x creation of design typology which includes form and assembly types,

x design of hypothetical models of structural glass panels and
x examination of structural stability of hypothetical models.
The design of hypothetical models is carried out according to defined form types.
Stability of three types of glass panels (flat, curved and sawtooth), is examined through
stress and deformation analyses.
The paper is indicating various approaches to design of structural glass facades and
might be helpful to architects and engineers.
2. Design Typology
According to generally accepted definition, the structural glass facades are long-span
applications categorized by the various structural systems employed as support, new
glazing and connection systems. Based on this, the design typology of structural glass
facades can be created and described through form and assembly types.
2.1. Form types

Form types of structural glass facades are defined according to position, geometry,
function and supporting system of glazing surface strongly influencing structural glass
facade appearance.
In relation to horizontal line, position of glazing surface can be: vertical,

sloping/angle90q and sloping /angle!90q (Figures 1 and 2).
According to geometry following types of glazing surfaces are present: flat,

sawtooth/accodion/stepped, curved/flexible and combination of previous, as shown in
Figure 2.
Figure 1a, b and c: Examples of structural glass facade form types.
892
Architectural Aspect of Structural Design of Glass Façades / Glass Skin Applications
Figure 2: Form types of structural glass facades according to position and geometry of glazing surface.
Following form types of structural glass facades according to functional characteristics

can be noticed: conventional facade, energy providing facade (solar facade solar
thermal and photovoltaic) and multimedia facade.
Classification of structural glass facades according to supporting system can be based

on:
x supporting system type: strongback system, truss system, cable trusses, grid
shells, cable nets, glass fin systems, glass structures,
x supporting system position: inside – behind the glass layer and outside – in
front of the glass layer.
893
Challenging Glass 3
Glass layer is emphasized if applied in front of supporting system, while the supporting
system is dominant building appearance feature if created in front of glass layer.
Generally, glass layer is expressive design feature if supporting system has large
structural grid. Especially the property of transparency is a valued feature of glass which
might be pointed out by large structural grid of supporting system. The combination of
being protected from rain, wind and cold and still be able to see what is going on in the
outside world is in fact an ideal combination [1].
2.2. Assembly types

Various assembly types of structural glass facades can be created due to the different
combinations of glass layers according to number of layers, ventilation and connection
options.
Figure 3: Assembly types of structural glass facades according to number of layers and ventilation type
According to number of layers the following structural glass facades are present: single-
skin and multiskin (only glass skins or glass + massive skins), as shown in Figure 3.
894
According to ventilation options the not ventilated and ventilated glass facades can be
recognized (Figure 3). Ventilated glass facades can be created as multistory ventilated
and partitioned by story (box-window facade, corridor facade and shaft-box facade) [2].
Discussing connection types in the case of structural glass facades, the following
occasions can be recognized:
x glass components connections and

x glass and other structural components connections,
which determines specific conditions in sense of component types and materials,

connection place and technology.
In case of glass facade components mutual connection, classification can be specified

according to: a) glass facade component types (connections between linear components,
linear and plane components and plane components), b) connection place (at the corner,
along the edges and within the surface) and c) connection technology (dotted – bolted
and linear - adhesive, with metal sections), as shown in Figure 4.
Figure 4: Structural glass facade components connection types
895
Challenging Glass 3
Similar typology can be adopted in case of the other supporting system types listed in
the subtitle 2.1. Form typology.
Also, construction of structural glass facades includes design of glass and other
structural components connections which can be classified according to structural
components type (connection with the roof, connection with the foundation and
connection with the gables) and according to component material (glass and concrete,
glass and wood and glass and metal).
3. Structural System and Stability
3.1. Structural components and system

Assigning the supporting function to facade glass panels means the minimum of the
quantity of the bearing substructure of structural glass facade. As for the structural glass
facade there is a tendency to reduce a number of contact points with ceiling construction,
the extreme option is only two contact zones, i.e. at the level of the highest and the
lowest line of the facade surface. By special geometry of facade glass surface, i.e. by its
design in the sense of vertical and horizontal partition on facade surface (dimensions of
glass panels), it is possible to impact on parameters that determine position, disposition
and dimensions of the cross-section of elements of the bearing substructure of the glass
surface or even to completely avoid those elements.
3.2. Influence parameters

First of all glass, surface should be dimensioned having in mind the horizontal wind
effect. The pressure of the wind on the facade is concerned in this study through the
equivalent static load that creates some certain impacts in the cross-section of the glass
element, i.e. stresses that must be within permitted limits for the certain type of glass [3].
It is very important that deformation of the glass surface caused by wind pressure is of
acceptable values.
The subject of numerical analysis are the glass panel models of 18 mm thick toughened
glass. In this study different model types are defined according to geometry
characteristics shown in the Figure 2. First three analyzed glass panels are of the same
characteristics: 1200 mm width, 4000 mm height simulating the glass panel that is built
in a level of a floor. The glass panel is supported by substructure. The equally divided
surface load of intensity of 1,00 kNm2, simulating wind effect, is applied on glass
panels. The analysis was done by software package Tower 6.0 [4], based on the Finite
elements method. In computer model glass elements are designed as follows: in data
base the glass as a material is created according to its characteristics, which are
described by adequate Module of elasticity, volumetric weight, Poisson's coefficient,
thermal coefficient.
3.3. Structural system stability

Analysis of the flat glass shows high values of normal stress in cross-section, in the
inside-surface of glass (Figure 5) and deformations (Figure 6).
Deformation of the flat panel of 18 mm thick toughened glass, caused by standard wind
effect, is unacceptable because its value on the half of panel height is 98.66 mm.
896
Figure 5: Normal flexural stress of the flat glass Figure 6: Deformation of the flat glass panel caused
panel caused by wind effect pressure by wind effect pressure
Figure 7: Normal flexural stress of the slightly Figure 8: Deformation of the slightly curved panel
curved panel caused by wind effect pressure caused by wind effect pressure
897
Challenging Glass 3
If the glass panel of the slightly curved surface is observed, with the same initial
parameters in the sense of intensity of load, conditions of leaning and dimensions of
panel, considerably more convenient results appear (Figures 7 and 8).
It is obvious that normal flexural stress of panel comparing the previous type, is reduced
10 times, and that the reduction of deformation is more intensive; comparing the type of
the flat glass – flexure is less 340 times. Relatively slight curves of the glass surface
very much increase its stiffness making it applicable on significantly bigger height span,
what will be shown in the analysis of the third type – densely curved or sawtooth glass
surface, shown in Figures 9 and 10.
Figure 9: Normal flexural stress of densely curved Figure 10: Deformation of densely curved glass
glass surface surface
Shown results are even more convenient then of previous case, so that the max. normal
flexural stress is 2,54 MPa while flexure of glass panel is only 0,14 mm.
By creating the glass panel in this way some kind of corrugated structure of glass is
obtained, with the big stiffness in the case of bending. This characteristic enables
designer to reduce the thickness of glass or to increase vertical span of the panel, shown
in Figures 11 and 12.
898
Figure 11: Normal flexural stress of densely curved Figure 12: Deformation of densely curved glass
glass panel whose verical span is double-floor height panel whose verical span is double-floor height
(H=8m) (H=8m)
The values of normal flexural stress of 6,95 MPa and flexure of the 1,47 mm on the half
span, show that the glass panel can be supported by substructure of the bigger structural
grid. In that sense the structural glass facade with large panels contains fewer contact
points with bearing substructure or structure of the building, increasing transparency of
the structural glass facade.
899
Challenging Glass 3
4. Conclusion
In the paper different forms, assembling and connection types of structural glass facades
are discussed. Design of hypothetical models of structural glass panels and their
examination through stress and deformation analyses are carried out.
The generally conclusion is that geometry of facade panel directly impacts on global
process of design and can define:
x visual identity of facade;

x grid of structural glass facade envelope;
x bigger transparency of facade;
x flexibility of the interior space.
In that sense the curved and sawtooth forms are the most appropriate, which is
demonstrated in the paper.
The paper is indicating various approaches to design of structural glass facades pointing
out the challenge of its construction.
5. Acknowledgements
This work has resulted from research within the COST Action TU0905 – Structural
Glass – Novel design methods and next generation products and scientific projects
“Spatial, Environmental, Energy and Social Aspects of Developing Settlements and
Climate Change – Mutual Impacts” and “Development and implementation of scientific
methods in design and construction of economic structural systems by application of
new technologies“, financed within the program Technological Development by the
Ministry of Education and Science of the Republic of Serbia from 2011 to 2014.
6. References
[1] Nijsse, Rob, Glass in Structures. Elements, Concepts, Designs, Birkhäuser, Basel, Switzerland, 2003,
pp.8.
[2] Loncour, X., Deneyer, A., Blasco, M., Flamant, G., Wouters, P., Ventilated Double Facades.
Classification & illustration of facade concepts, Belgian Building Research Institute, Department of
Building Physics, Indoor Climate & Building Services, Belgium, pp. 9-10.
[3] Porter, Mark, Aspects of Structural Design with Glass, A Thesis for the degree of Doctor of Philosophy,
Balliol College and the Department of Engineering Science, The University of Oxford, 2001.
[4] http://www.radimpex.rs/about.php?lang=en&id=1
900
doi:10.3233/978-1-61499-061-1-901
Cones Made of Glass

Juergen Neugebauer
University of Applied Sciences FH-Joanneum Institute for architecture and civil
engineering, Austria, juergen.neugebauer@fh-joanneum.at
SFL Technologies, Austria, neugebauer.juergen@fibag.at
A recently finished and very interesting project is the new entrance of the museum
quarter ´Joanneumsviertel´ in the center of Graz in Austria. The complex of the
museum’s buildings consists of two wings of the existing structure. For the
connection of those old parts of the museum the architects designed the new
entrance with two basement levels between them. The architectural challenge of
this project was to bring daylight into these two lower floors. The concept of the
architects was to let natural daylight flow in via vertical funnels in the form of
small round courtyards with a diameter of up to approx. 16 m into the basement.
Laminated and insulated glass was used for the cladding of these conically-shaped
funnels.
Keywords: Conically curved Laminated Glass, Conically curved Insulated Glass
1. Introduction
A very interesting project, finished at the end of 2011, is the new entrance situation of
the museum quarter ´Joanneumsviertel´ in the center of Graz in Austria. The complex
the museum’s buildings consists of two wings (museum of natural science and the
museum of modern art) of the existing structure. For the connection of those old parts of
the museum the architects - eep architects, Graz/A; Nieto Sobejano Arquitectos,
Madrid/E - designed the new entrance between them. The visitors of the museum can
reach the biggest cone designed as the museum entrance via a specially designed public
place. The picture below shows this cone with an escalator (see in figure 1 bottom left).
Figure 1: New Entrance - ´Joanneumsviertel´. [1]
901
Challenging Glass 3
The basement with a depth of approx. 10 m was excavated and the two levels were
covered with wide spanned reinforced concrete slabs. An architectural challenge of this
project was to bring daylight into these two lower floors. The concept of the architects
was to let natural daylight flow in via vertical funnels into the basement. These funnels
have the form of small round courtyards with different diameters of up to approx. 16 m.
Laminated safety glass and insulated glass were used for the cladding of these
conically-shaped funnels. The cones have a central axis which are inclined up to 15°
from the vertical and for this reason the inclination of the glass panes vary from the
vertical position to a position of up to 30° from the vertical. Two of the six cones
interpenetrate and another one is posed on its top and situated in the center of a larger
one.
2. Cone 1&2
Cones 1&2 are the cones with an interpenetration located on the northern part of the
public place. Cone 1, with a diameter of approx. 9 m, extends into the first basement
level and Cone 2, with a diameter of approx. 6 m, extends into the second basement
level (see figure 2 below). For the balustrade, laminated safety glass with a thickness of
24 mm which consists of 2 x 12 mm conically-curved glass panes was used. The
cladding in the basement levels consists of insulated glass with conically-curved 12 mm
glass on the outer side, a 16 mm space and laminated safety glass which consists of 2 x
8 mm conically-curved glass on the inner side. The guaranty of the tightness against the
rain for the parabolic curve (see in figure 3 below) of the interpenetration was a big
challenge, as well as the geometrical challenge which had to be solved.
Cone 2
Cone 1
Figure 2: Principle of Cone 1&2.
Figure 3: Interpenetration of cone 1&2.
902
Cones Made of Glass
3. Cone 3&4
Cone 3 has its larger diameter on the upper side in comparison with cone 4 which was
posed on the top and has the larger radius on its bottom edge. Smaller cone number 4 is
situated in the center of cone 3 which is larger. For the balustrade, laminated glass with
a thickness of 24 mm which consists of two 12 mm conically-curved glass panes was
used. The cladding in the first basement levels was made for both cones of insulated
glass with conically-curved 12 mm glass on the outer side, a 16 mm space and a
laminated safety glass which consists of 2 x 8 mm conically-curved glass on the inner
side. The glazing in the second basement level, which is used as a depot for the exhibits,
was designed to be laminated glass. Cone 6 is equal to Cone 3.
Cone 3
Cone 4
Figure 4: Principle of Cone 3&4

Figure 5: Cone 3&4 – View from the inner.
903
Challenging Glass 3
4. Cone 5 – The new Entrance of the Museum

The biggest cone 5 with a diameter approx. 16 m was designed as the new entrance for
the visitors of the museum. Via an escalator the people reach the first basement level
and can enter the museum through a sliding door. For the balustrade laminated glass
with a thickness of 24 mm which consists of 2 x 12 mm conically-curved glass panes
was used. The cladding in the basement level was made of insulated glass with
conically-curved laminated glass consists of 2 x 8 mm glass on the inner and the outer
side and a 16 mm space between.
Cone 5
Figure 6: Principle of Cone 5.

Figure 7: Cone 5 – View from the interior.
904
Cones Made of Glass
5. Principle concept of the Glazing

The special boundary condition of the great deformation of the concrete slab of more
than 30 mm (for the long term deformation) causes the structural system of the whole
cladding. The glass panes of the balustrade had to be staked on the insulated glass of the
level below. The lower glass has to carry dead load of the glass above. The glass panes
were glued to stiff stainless steel ring sections and those are discrete supported on their
ends. These hinged support systems transfer the horizontal loads to the concrete slabs
and guarantee the freedom for the vertical movements of the concrete slabs (see in
figure 8).
Figure 8: Cross Section – Principle of the Structural System of the Cones.
6. Design of the Cones

All the different cones were design with a big finite element model which covers all
glass panes. The loads were defined with dead loads, wind loads and horizontal loads
due to human impact. For the design of the balustrade a horizontal load of h=3.0 kN/m
was used. This level of the load is based on the possibility of a big gathering. Beside the
structural design an experimental test for the dynamic loads the so called pendulum
impact test was successfully done.
7.2
Z
X
Y
Figure 9: Cone 5 Deflection Results of the Design with a Finite Element Program.
905
Challenging Glass 3
7. Acknowledgements
All these conically-curved laminated or insulated glass as well as the steel substructure
was produced by the company SFL Technologies in Austria. The structural
development and the structural design were made by the Research Center for Integral
Construction Engineering “fibag”.
8. References
[1] Homepage oft he Museum „Joanneumsviertel“ , http://www.joanneumsviertel.at
906
doi:10.3233/978-1-61499-061-1-907
Green Houses – Why Can´t They Be Built

as Our Ancestors Did?
Holger Techen
University of Applied Sciences Frankfurt/M., Germany, techen@fb1.fh-frankfurt.de
Matthias Michel
imagine structure GmbH, Frankfurt/M., Germany michel@imagine-structure.eu
Codes of practice and regulations are the basis for today´s glass design. They
ensure the same safety levels for this brittle material as it is common for many other
construction materials. In combination with research results as they are enclosed in
this proceeding, delicate glass constructions that resemble the slenderness of the
glass houses of the 19th century may become feasible again. During the design
process, the detailing of adhesive connections for fixation and stabilization within
the glass structure itself had been developed.
Keywords: Constructability of historical glass houses, stabilizing with glass
1. Walsgrove Garden
For the architect it was a demand to design a new Walled Garden near London in the
typical way of the green house architecture of the 19th century. The main characteristic
of those green houses is a minimized steel structure, a scaly glass cladding and
simplified mechanical openings within the glass cladding.
As the design’s archetype the Claydon Garden green house had been chosen, which is in
continuous use since the beginning of the 19th century. It’s steel structure is made of
slender rolled steel T-sections which are stabilized by tension bars (see fig. 1). The
cladding´s glass panels are clamped alongside tensile members interconnecting the
major members. Due to limited technical capabilities only small glass sizes had been
produced in the prime time of the green house architecture. The water tightness of those
buildings had been realized by a scaly cladding configuration or by agglutinating the
glass panels and the steel structure.
In 2007 the British architect Tony Hunt was looking for design solutions to combine the
historic construction principles with nowadays regulations and codes for glass structures.
The very first design didn´t respect the historical archetypes. Edge supported insulating
glass units, T-sections with 140 mm flange width, screwed L-profiles used for glass
fixation had been the construction principles. Those principles of course didn´t match
the primal ideas. During the intense planing process between architect and structural
engineer the basic demands had been isolated again in order to design a primary
structure made out of CorTen steel sheets cut and welded to bended T-profiles. To
minimize the steel structure´s flange width it was initially necessary to redesign the
glass fitting principles.
907
Challenging Glass 3
Figure 1a + b: Claydon Garden, detail oft he scaly glazing
Neither the fixations of the historic prototype building at Claydon Gardens with its
numerous tension rods and glass clips nor nowadays recommendations and design codes
did show any valid advice to realize a delicate steel and glass structure. However in
reference to the German code DIN V 11531 (green houses) the demand of insulating
glass or laminated glass panels could be substituted by heat soaked toughened glass.
Correspondingly to the historical green houses the scaly cladding got at minimum a roof
pitch of 25° in order to avoid the drip off any condensation into the green house.
Figure 2: Walsgrove Garden, main structure supporting the imbricated glazing
For the fixation of the glass panels a clamp made out of a helical hinge had been
developed. These clamps are fixed by two bolts along the T-profiles web. By pre
cambering the 1,5 mm thick steel clamp both overlapping glass panels are elastically
held. In specific the glass panels are clamped by a two point fixation within the steel
908
Green Houses – Why Can´t They Be Built as Our Ancestors Did?
clamp. In order to avoid any direct contact between glass and steel the relevant glass
edges are sealed by small, more or less invisible silicone strips (see fig.3)
Figure 3 a+b: Archetype of the elastical glass clamp
With the development of this specific detail it was possible to design the dimensions of
the main steel structure independently from the glass panel supporting conditions.
With a flange width of 50 mm and a web height of 75 mm the T-profile dimensions are
similar to the historical archetype. Because this slender profile is not suitable for a total
span of 5,0 m, a prestressed tension rod was introduced to fix and stabilize all T-profiles
in a way it was accomplished at Claydon Garden green house. The tensile member was
preloaded by a pack of encapsulated disc springs at adjacent walls to comepensate for
temperature change. Each T-profile was attached to the prestressed rod by dogs on each
side of the web in order to achieve the stability against lateral buckling (fig 4).
Figure 4 a+b: Girder stabilzation by a prestressed tension bar, clamped with two brackets
Prestressing device in the adjacent side walls
909
Challenging Glass 3
2. Glass house at Abbey Beilstein

The abbey in Beilstein is highly frequented place at the river Mosel. To expand the
abbey´s cafe the architect designed a glass winter garden within the patio as an
extension to the historical, listed buildings. As shown in the first project it was the
intention to follow the construction´s design to historical archetypes by using the
potential of glass as a structural material. The extension (groundfloor 10,15 x 4,95 m)
consists of rampant frames in a grid of 1,15 m which carry the overhead and facade
glazing units. The horizontal frame members do not touch the existing wall to keep the
historic facades untouched by any solid structure. They end 2,30 m before the wall and
are stabilized by the overhead glazing only (fig 5).
Figure 5: View from the courtyard
Laminated, 2 x12 mm heat strengthened glass is used for the overhead glazing. The
glass panels are point fixed with the horizontal frame members and edge supported
along the wall. The frames are hinged at their foot support and would incline towards
the building. They are only held in their position by the horizontal glazing. Forces are
carried from the steel frames by the punctual fixations into the glass and through the
glass via membrane and bending stresses into the edge supports along the wall.
For this it was necessary to fill up the gap between steel bolt and the glass bore-hole
with a compressive, fine-grained mortar in order to transmit the bearing pressure
uniformly along the bore-hole edge.
The edge support is differentiates between vertical and horizontal loads. While each
glass panel is fixed at two points in horizontal direction only, the vertical support is
continuous. This specification was realized by using silicone and EPDM-materials with
different stiffness characteristics (fig. 6).
910
Figure 6: Detailing of the edge support along the wall for vertical and horizontal actions
As mentioned, the Frames (high 4,05 m) have a hinged supported base plate which is
highlighted by a recess above the flooring and the profile itself. Figure 7 shows the
detail of the uplifted base point.
Figure 7: Section of the uplifted base point
The steel frames are made out of a hollow welded section of 60 x 210 mm. The frames
are connected together by a strut (Ø 30 mm) which is fixed to the adjacent walls for
lateral stability and in order to achieve a redundancy in case of glass failure.
911
Challenging Glass 3
The steel structure and the glass panels had been analysed with its different material
properties and mechanical characterisitcs in one FEM-model only. Especially the
coupling of the glass and the steel frame within the punctual fixations did need a
modeling close to reality. Beside the stress analysis, the FEM-model was used to
determine critical buckling forces which are essential to proof the stability of the entire
system. In account of initial sway imperfections (glass and steel) and calculation by the
theory of 2rd order the stability of the total system had been verified even for raised
snow and wind pressure actions on the overhead glazing.
Figure 8: The primary structure stands isolated from the historic buildings
and only held and stabilized by the overhead glazing.
The glass structure of the adjacent alleyway consists of laminated heat strenghtened
glass as well. The overhead glazing is edge supported along the adjacent wall and
towards the patio punctual supported on several free standing glass panels. These 3,0 m
high panels are made of 2 x 12 mm heat strengthened glass and are rigidly fixed in U-
profiles along the pavement. The overhead glazing holds the tip of the vertical panels in
place. The edge support along the adjacent wall is designed as a structural sealant
glazing (SSG) fixation. In general all adhesive connections are prefabricated in stainless
steel profiles (U- or L-profiles) and mounted on site only.
The most critical part is the connection between vertical and horizontal glass panels.
Due to the slenderness of the 3 m high vertical glass elements, it is important to center
the vertical loads by specific formed steel profiles onto the vertical elements. This
connection was realized by prefabricated SSG-adhesive connections (fig.9,10). All
connections and glass elements had been designed and analyzed according to EOTA-
regulations. The stability and buckling behavior of the slender glass elements had been
analyzed according to respective technical literature [1-5] and results of the latest
research.
912
Figure 9: Centric connection of the glass roof on top of the glazed wall
Figure 10: Centric connection of the glass roof on top of the glazed wall
By the intense collaboration of architect, engineer and manufacturer slender and delicate
glazing can be realized today that will be in accordance with nowadays sophisticated
regluations, but presenting itself with the airiness of historic masterpieces. To achieve
this, especially the glass detailing and handling should be given attention and the design
of details shall appropriate to the material characteristics.
3. References
[1] M. Feldmann, K. Langosch, R. Kasper, M. Baitinger: Einsatz von Glas in tragenden Bauteilen –
Aktuelle Entwicklungen des konstruktiven Glasbaus, VDI-Berichte 2084-2009
[2] M. Feldmann: DAST-Forschungsprojekt Nr. 15060/N: Vereinfachte und einheitliche
Stabilitätsnachweise für Bauteile aus ESG und VSG für Druck und Biegung, 2006-2008
[3] J. Liess: Bemessung druckbelasteter Bauteile aus Glas, Diss. Universität Kassel, 2005
[4] F. Wellershoff: Nutzung der Verglasung zur Aussteifung von Gebäudehüllen, Diss. RTWTH Aachen,
2006
[5] H. Techen: Fügetechnik für den konstruktiven Glasbau, Diss. TH Darmstadt, 1997
913
Challenging Glass 3
Glass in Façades
915
doi:10.3233/978-1-61499-061-1-917
Novel Testing Techniques for Building

Glass in Service Conditions
Yiwang Bao, Xiaogen Liu, Yan Qiu, D.T.Wan
State Key Laboratory of Green Building Materials , China Building Materials
Academy, BeijingˈCHINA, byw@ctc.ac.cn
Minxi Bao, Jian Yang
University of Birmingham, UK, mxb949@bham.ac.uk
Developing the detection techniques and safety evaluation methods for the glass
curtain walls is a challenging task. In this work, some new testing techniques were
developed to determine the probability of failure by using dynamic tests and photo-
elastic scanning method. A small device of polariscope scanning was designed and
machined to detect the defects and stress concentration in tempered glass. In order
to predict the risk of falling of the glass curtain wall, a testing device was developed
on the basis of dynamic tests, i.e., determining the risk through evaluating the
degradation of the nature frequency of the glass panel. Furthermore, the bend-
stiffness test was used to estimate the air tightness of insulated glass and a simple
tester was developed. The testing techniques have been applied in high buildings
and the tested results show that the testing methods described above are feasible
and effective.
Keywords: building glass, safety evaluation, dynamic tests, photo-elastic scanning,

bend-stiffness test
1. Introduction
Since glass has been widely used in building as a transparent curtain wall and
windows material, its safety and reliability in service condition have attracted
increasing research attentions. During past decades, various accidents due to

building glass falling occur at a worryingly high frequency, which have caused
great losses in economy and even human beingÿs life. The failure and the
resulting damage of the building glass usually exhibits following aspects: (1)
spontaneous breakage of tempered glass used in buildings; (2) falling off of a glass
plate from height because of loose edge support or deterioration of the sealant; (3)
edge seal failure in the insulating glass due to air leak and accumulated damage.
Recently, a great deal of research on testing methods of the safety of the building
glass have been carried out by scientists and engineers in many countries [1-5], but
most of the research outcomes are still at an initial stage, and lack of effective tools
and unanimous standards. In this work, new testing techniques were developed to
determine the failure due to the abovementioned three reasons. The aim of this work
is to provide a nondestructive approach to predict the risk or failure of building
glass in service.
917
Challenging Glass 3
2. Detecting defects and stress concentration of building glass

Spontaneous breakage of tempered glass is mostly due to the stress concentration in
the tensile stress zone of the strengthened glass. And the stress concentration may
be caused by the defects, impurity or inhomogeneous fixing [6]. In order to detect
the location where the breakage is originated, a photoelastic scanning device was
designed to detect the stress concentration caused by the defects in building glass,
which includes a scanner, image transportation and stress analysis systems.
The photoelastic scanner has two working forms, i.e., transmission mode and
reflection mode, both of which are made up of two polarization plates perpendicular
each other. Images are taken by a high resolution digital camera that is connected to
a computer. By this way, the stress concentration in the glass can be observed easily.
As an example, the contact stress at the pillars in vacuum glass was tested and the
stress concentration points are displayed as in Fig.1.
Figure 1: Photoelastic image of the vacuum glass, showing the distribution

of the pillars in the vacuum glass. (by self-made photoelastic device.)
The photoelastic picture of the vacuum glass indicates that the stress concentration
near the pillars is clear and the stress is weak at the pillars near the edge of the glass.
In many cases, a small defect or an impurity in the glass also causes stress
concentration like a pillar in the vacuum glass. For example, a chromic oxide
particle in the glass is observed by a common microscope as shown in Figure 2(a),
and the particle under polariscope is shown as figure 2(b). The comparison between
(a) and (b) demonstrated that the photoelastic method could observe the stress
concentration in glass and to find the defects that may cause the spontaneous
breakage.
918
Novel Testing Techniques for Building Glass in Service Conditions
Figure2: (a) micrograph of chromic oxide particle in glass, (b) observation of the particle under polariscope,
showing the stress concentration near the particle. (with KEYENCE digital microscope, VEX-600E)
As for the glass panel fixed in building, we could find out the defects and the risk of
spontaneous breakage by using photoelastic scanning on the glass. Furthermore, the
size and shape as well as the location of the defect also are determined
quantitatively.
3. Test method for predicting loose support and debonding of building glass
$nother failure ofbuilding glass in service is that the glass plate falls off from a
high building because of the loosen edge support or deterioration of the structural
sealant. Any change of support condition will result in the variation of the natural
frequency of the glass window or glass curtain wall. In fact, the frequency of the
glass will fall with a reduced level of edge clamp. Therefore, if the relationship
between the support damage and the frequency degradation are established, it is
possible to evaluate the risk of falling of the building glass by measuring and
comparing the variation of the natural frequency.
A dynamic frequency tester is used to measure the building glass nondestructively.

In order validate this method, the tests are carried out with small glass sample
(300x300x4mm) under elastic clamp support with different fixing degrees by bolts
at edges. The experimental data are recorded in Fig.3, and the results indicate that
the frequency declined with a slacked clamp.
919
Challenging Glass 3
250 Bolts is not slacked

Bolts is partially slacked
200
Accleration FRF(mm2/s)
150
100
50
200 600 1000 1400 1800

Frequency(Hz)
Figure3: Vibration figures changs with the clamp condition for a square glass (300x300x5)
The clamp force can be changed by loosening up bolts that are used for fixing the
edges of the glass plate. The relationship between the frequency and the clamp
force is investigated, which shows that the nature frequency of the glass plate
decreases with a reduced fixing force (slackness in the screw thread) and a nearly
linear relation is observed as shown in Figure 4.
340
330
320
Natural frequency/Hz
310
300
290
280
270
260
250
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Loose laps of the bolts
Figure4: Measured frequency of the square glass plate vs. the clamp force (loose laps of the bolts)
For mechanical analysis, the curtain wall glass in a stick system is usually treated as
a rectangle plate fixed by an elastic clamp at four edges[7], and the structural
stiffness of the elastic clamped rectangle glass should be within that of the simply
supported plate (lower bound) and the clamped rectangle plate (upper bound).
The 1st order natural frequency of the rectangle plate simply supported at four
edges can be determined by [8]:
920
1 1 D
Z S 2( 2
2 ) (1)
a b m
The 1st order natural frequency of the clamped rectangle plate can be determined
by:
504 D ª 4 4 º
Z2 « a b 4 a 2b 2 »
a b m ¬
4 4
7 ¼ (2)
Where Z is the minimum (first orders) natural frequency (rad/s), m is the mass of
unit area, a is the length, b is the width, D is the bend stiffness and defined as:
D Eh 3 / 12(1 X 2 ) (3)
where E is the elastic modulus, h is the thickness of the plate, Q is the Poisson
ratio.
In fact, the natural frequency for a building glass depends on many factors, and
mostly the reliability is in proportional to the frequency. Determining the natural
frequency of the curtain wall glass and their variation with time are important for
estimating the reliability of glass curtains or window glasses. Relative comparison
between the measured and expected frequency might yield the risk of loosening or
falling off of the glass.
For a given building glass panel supported by frame (such as exposed frame and
concealed frame glass curtain wall), the natural frequency should be higher than
that when it was simply supported, in other words.
As an example, the mechanical parameters of the glass specimen are: E =72 GPa,
Q
=0.22, m =10kg/m2, the size is 300x300x4mm, then the lower limits natural
frequency of the glass specimen can be determined by Eq.(1):
1 1 D 1 1 7.2 u 1010 u 0.0043

Z S 2( 2) 3.14 2 u ( )
a 2
b m 0.32 0.32 10 u 12 u (1 0.22 2 )
1375.˄
3 rad / s )
f Z / 2S 1375.3 / 6.28 219( Hz )
The value of 219Hz is the lower limit of the frequency of the glass panel, and the upper
limit was determined as 266 Hz by Eq(2).
921
Challenging Glass 3
4. Test for tightness in insulating glass unit in service

Failure in the air tightness of the insulating glass in building may result in two
serious problems, i) reduction in the thermal insulation and thermal performance,
ii) reduction in the interface strength between the glass and sealant. Therefore,
nondestructively test and predicting the air leakage of the insulating glass unit is
significant for safety service of the glass.
A new testing technique was developed based on the following principle. The load
and the deformation would be carried by both panels in the insulating glass because
the air in the glass cavity is compressed, when the air leakage occurs, only the front
glass panel carries the load, and the deflection is greater than that of a sealed
insulating glass. In other word, giving a applied load at the center of the insulating
glass unit, the deflection of the two panels is similar and the distance between both
glasses at the loaded point has little change for the tightly sealed hollow glass unit,
but if it is leaked, there will be greater deflection in the front panel and smaller
distance between both panels. The deformation is shown schematically as in Fig.5.
According to this observation, we designed a simple approach and testing device to
predict the tightness of insulating glass unit in service, by using loading device and
a distance gauges. Usually, for series of insulating glass units with the same shapes
and same sizes, applying the fixed load at the center of the glass units and measure
the change of the distance between both panels, the unit in which the deformation
occurs in only the front panel would be the leaked unit.
outer panel
outer panel
P P
'P
inner panel inner panel
(a) sealed (b) leaked

Figure 5: Schematic deformation in comparison between sealed and leaked
hollow glass unit under a load at the center
The test device could have different forms. In either form, it should have the
loading and distance measurement. In this work, a simple device was used to
measure the air leakage of the isolating glass units in building.
922
Figure 6: Schematic of leakage test for isolating glass units in service
As shown in Fig.6, a normal load P is applied on the glass through the weight W,
and the load value is determined by the moment equivalence to the stiff bar.
PuH W u ( d / L) u S (4)
P W u ( d / L) u ( S / H ) (5)
The load P would result a deflection on the outer glass for a leaked unit. Measuring
the change of the distance between two panels by using the displacement gauge or
comparing the change in the deflections among series glass units with the same
sizes, we can evaluate if an isolating glass unit is leaked or perfectly sealed.
923
Challenging Glass 3
5. Conclusions
Defects that cause stress concentration in building glass could be detected by using
photoelastic tests. A polarized scanner was designed to test nondestructively the
reliability of the building glass or glass curtain wall in service.
Risk of loosening and falling down of glass panel in building can be predicted
through the degradation of the natural frequency of the panel, because the frequency
depends on clamp condition.
The stiffness of the outer panel of a leaked insulating glass unit will become smaller
(nearly a half of well-sealed unit). Therefore, the air leakage is determined by
measuring and comparing the deflections of the outer panel of the hollow glass units in
service.
6. Acknowledgements
This work was supported by the “973” project (2009CB623103), Natural Sciences
Foundation of China under Grant(No.572), Beijing Natural Sciences Foundation
under Grant (No.2082026), and international cooperation project (S2011ZR0397).
7. References
[1] Foss R V, Safety glass test developments, Proceeding 5th Glass Processing Days, Finland, 1997, pp. 96–
100.
[2] Xin Dafan; Lu Jinlong; Xu Qin, Research on estimation of curtain wall glass in use , Building
Construction, 1/26,pp. 64-65.
[3] Weggle David; Zapata Brian; Kiefer Michael, Properties and Dynamic Behavior of Glass Curtain Walls
with Split Screw Spline Mullions, Journal of Structural Engineering,10/ 133 ,pp.1415-1425.
[4] William J; Nugent M; Mark K,. Preventing Failures in Metal and Glass Facades, Forensic Engineering,
25,2008,pp. 540-549.
[5] Efstathiades Ch; Baniotopoulos C C ; Nazrko P; Ziemianski L, Application of Neural Networks for the
Structural Health Monitoring in Curtain-Wall Systems, Engineering Structures, 29,2007,pp.3475–3484.
[6] Bao Yiwang; Liu Lizhong; Han Song, A New Mechanism of Spontaneous Breakage of Tempered
Glass——Monolithic Silicon particle, Journal of The Chinese Ceramic Society,9/35,pp. 1273-1276.
[7] ASTM E1300. Standard Practice for Determining Load Resistance of Glass in Buildings, 2003.
[8] Timoshenko S; Krieger S.W, Theory of plates and shells, New York: McGraw – Hill Book Co(1977).
924
doi:10.3233/978-1-61499-061-1-925
Fire Resistance
Without Fire Resistant Glazing
Csilla Csoke, Johan Koudijs
DGMR Bouw B.V., The Netherlands
Fire resistant glazing is a high tech product but it has its price. In some cases, for
instance for buildings near transport routes of petrochemical products, large areas
of façade have to be able to withstand heat from large pool fires in order to allow
for the safe evacuation of these buildings. This paper will explain the loads of fire
resistant glazing tested according to European standards and will compare these
loads to facades exposed to pool fires. It will show an example of building
developments near rail and waterways that led to research on a low budget solution
to fireproof facades without using fire resistant glazing and will also show the
results of this research.
Keywords: Glass, Low-E coating, Fire Resistance, External Fire Exposure
1. Introduction
In a densely populated country like The Netherlands, the transportation of dangerous
goods on roads, railways, and by container ships is a serious threat to Dutch citizens.
Transportation is only allowed on designated routes, as assigned by the government.
This does not mean that all risk is eliminated.
In the legislation in The Netherlands, Qualitative Risk Analysis (QRA) is compulsory in

Environmental Planning. In this analysis, the risk of dying of one individual or a group
of individuals is determined. Measures have to be taken if the risk exceeds certain
values.
Keeping a safe distance between the transportation route and building development is a
simple solution. As the following case shows, however, this is not feasible in many
situations.
2. Case: The Rotterdam Tower
2.1. Project description

There is a new development of several high-rise buildings (140-160 m) on the
waterfront of the river ‘New Maas’. It is located on a peninsula in the centre of
Rotterdam, along the quay that used to be the ‘home’ of the Holland-America line. The
river is still used for the transportation of a large amount of goods from the port of
Rotterdam to the German industry in the Ruhr area.
According to Environmental Planning [1], the distance from the edge of the water to the
nearest building must be at least 25 m. Furthermore, in all buildings between 25 and 40
925
Challenging Glass 3
m from the quay, measures have to be taken to ensure that the occupants of the
buildings have sufficient time to evacuate. Only buildings at a distance of at least 40 m
are safe without taking measures.
The peninsula (called ‘Kop van Zuid’), on which the high-rise tower ‘The Rotterdam’
will also arise, is only 175 m wide. Developing buildings 40 m from the quay means a
huge waste of potential area for building development. Building closer to the river
requires a fire resistant façade at each waterfront-facing elevation.
Our company was asked to find a cost effective way to protect ‘The Rotterdam’
building without fire resistance façade construction, especially without fire resistant
glazing.
Figure 1: Peninsula ‘Kop van Zuid’, Rotterdam

926
Fire Resistance Without Fire Resistant Glazing
Figure 2: High-Rise Development ‘The Rotterdam’
2.2. Scenario
The scenario that, according to the local regulations [2], had to be taken into account
was a collision between inland shipping vessels. In that case, the assumption is a spill of
75 tonnes of highly flammable liquid floating on the water. This amount of liquid can
cause a pool fire with a diameter of 25 m and a height of almost 40 m. This serious fire
can last for 30 minutes (given the amount of fuel). At the lower parts, the heat radiation
emitted by the flames will be approx. 130 kW/m² (compare to the 0.7–0.8 kW/m² solar
radiation). At higher levels in this fire, the radiation will drop to 50 kW/m² or even
lower [2].
Besides for the effect of the high radiation, the influence of the wind has to be taken into
account. According to the research of TNO [3], wind speeds up to 9 m/s can blow the
flames towards the building. If the pool fire is close to the quay, the flames will come
into contact with the façade. In these parts of the facade, fire resistant glazing will be
installed. The other parts of the façade, for instance on ground floor level, will only
suffer from the high radiation levels emitted by the fire.
The amount of radiation received by the façade can be calculated using a numerical
model of the software WinTherm 8.1.1. This is a 3-D thermal software that DGMR used
to calculate thermal radiation exchange between three-dimensional solid surfaces. By
modeling a large flame from a pool fire as a solid body with a prescribed surface
temperature distribution and constant emissivity, the pattern of radiant flux incident on
the complex building facade was determined. The calculations show that large areas of
the water-facing elevation shall receive heat flux up to 41 kW/m².
Radiation below 15 kW/m² will not propagate fire according to the European standard
EN 13501-2. In the areas receiving more than 15 kW/m², measures should be taken to
prevent the propagation of fire by the so-called auto-ignition: if the heat radiation
reaches a certain level, materials can burn without actually coming into contact with the
flames.
927
Challenging Glass 3
Figure 4: Pool fire exposure of the facade
3. Target
As was mentioned in the introduction, the traditional solution in these situations is to
use fire resistant façade constructions with fire resistant glazing. In this case, it should
be sufficient to apply a façade system – and glazing – that has 30 minutes of fire
resistance (Integrity and Radiation) based on the external fire conditions according to
EN 1363-2; classified as EW30-ef.
There are actually two problems with this requirement: Testing according to the external
fire conditions means not only heat radiation but also exposure to flames and hot gases.
In our case, only the heat radiation has to be taken into account. The other problem is
that there is no fire resistant glazing with classification EW30-ef available on the market.
Glass suppliers only provide glass tested according to the standard fire curve. Basic
knowledge on the conditions in furnaces that is used to determine fire resistance
according to standard ISO or EN fire curves gave us the insight that, during these tests,
the constructions are exposed to over 100 kW/m² heat radiation. In our case, the façade
is exposed to 41 kW/m2.
Our goal was to find an economic solution without using fire resistant glazing. The
target was set by the local authorities: keep the amount of radiation coming through the
glass below 10 kW/m², preferably even below 3 kW/m². This performance should last
for the duration of the fire (30 minutes).
4. Research and development
4.1. First fire test

‘Normal’ glass has no fire resistance if exposed to the standard fire curve. The
behaviour of fire resistant glazing is very well documented, but our company has found
little evidence that non-fire resistant glass was exposed to an external fire curve. DuPont
and Pilkington showed, in developing bushfire resistant glass, that a laminated Low-E
coated glass could withstand radiation levels of 40 kW/m² for a short period of time (2-5
928
minutes) (2004) [4]. Out of curiosity concerning the behaviour of ‘normal’ insulation
glass units exposed to external fire conditions, we organised fire tests at the Efectis Fire
Laboratory in Rijswijk (The Netherlands) at the end of 2007.
We tested a simple solar-control IG unit with a laminated inner pane without Low-E
coating according to the EN 1363-2 fire curve for external fires. The glass configuration
withstood exposure for 9 minutes.
4.2. Modification of furnace

As was mentioned hereinabove, during a fire test according to the European standards,
the test specimen is not only exposed to heat radiation but also to the hot gases of the
furnace. The hot gases will transfer heat to the test specimen causing an additional
temperature rise in the specimen. Therefore, together with Efectis Test Laboratory, we
decided to modify the furnace in such way that it better meets the test requirements for
this building.
We designed a special wall in front of the test furnace in order to limit the exposure to
radiation only. The bottom half of the sidewalls was left open to help the hot gases
escape the furnace. Practical issues on how to control the temperature of the furnace and
thereby the heat radiation on the test specimen was solved by the testing laboratory. The
by the furnace emitted radiation is measured in front of the test specimen, in the
furnace-side.
Figure 3: Test specimen in front of the modified furnace
929
Challenging Glass 3
4.3. Further testing

After the modification of the furnace, we tested several different glass configurations,
all of them without fire resistant glass pane. IG units composed by a single outer pane
and PVB laminated inner pane with Low-E coating are tested in aluminium and steel
frames. Both pyrolytic and high performance Low-E coatings on different positions
were tested. In some cases, some of the glass panes were composed of heat strengthened
glass (Figure 4).
The results of these tests are remarkable. Basically, the IG units with Low-E coating on
the right position were able to withstand heat flux of 41 kW/m² for 27 minutes and
reduced the radiation level at 1 metre behind the glass to 2 kW/m². The configuration
with heat strengthened glass panes did not fail for 60 minutes. These results are only
useful for the project ‘The Rotterdam’. The furnace was modified for the testing
according to the special requirements. The results are not applicable for commercial
targets (CE marking).
Figure 4: Compositions of IG Units. The arrows shows the possible positions of the Low-E coating.
The proper compositions is not illustrated.
4.4. Testing again to external fire exposure

Because of the results of the IG unit, where heat strengthened glass is used for some of
the panes, we decided to test this composition according to the European standards for
external fire exposure. This glass configuration withstood exposure for 40 minutes. The
IGU unit has received the classification EW30-ef according to EN 13501-2 and,
therefore, it is exploitable for CE marking. This is the first insulation glass unit
worldwide composed of non-fire resistant glazing. The exact composition of the IG unit
is available on request.
5. Conclusion
Because of the curiosity to find a cost effective way to protect the project “The
Rotterdam” without fire resistant façade construction, especially without fire resistant
glazing, our engineers exposed ‘normal’ insulation glass units to external fire conditions.
The results show that a smart composition of a laminated inner pane, Low-E coating,
and a heat strengthened glass pane at the right positions can achieve the classification of
EW30-ef. This insulated glass unit is an economical solution for the façade of projects
where external fire exposure poses a risk.
930
6. References
[1] Molag, .M; Leur, P. van der; Vredeveldt, L., Stax, L., Aspects and backgrounds for a development-free
zone along the New Waterway and the New Maas related to transport of dangerous goods, TNO Milieu,
Energie en Procesinnovatie, TNO-MEP-R 2002/266, July 2002.
[2] Committee for the Prevention of Disasters Voorburg, Methods for the Calculation of Physical Effects:
Part 2 CPR 14E, 1997/2005, ISBN 9012084970.
[3] Molag, M; Reinders, J. E. A.; Wiersma, T., Safety Evaluation for the High-rise project ‘The Rotterdam’
on the Wilhelminapier in Rotterdam, TNO, 2 September 2005, no. 35443.
[4] Bushfire Resistant Windows Introduced by Pilkington and DuPont, News Item Building Product News,
02/02/2004.
931
doi:10.3233/978-1-61499-061-1-933
Building Design Might Soon Be Driven by

Energy Requirements Only
Christof Erban
Fraunhofer ISE Freiburg,Germany, Christof.Erban@Ise.Fraunhoder.de
TU Delft, The Netherlands, c.w.erban@tudelft.nl
This paper deals with the near future constraints on energy consumption of
buildings, their effect on the required accuracy of glass g values and the amount of
photovoltaic required on buildings.
Keywords: photovoltaic, BIPV, total solar transmittance
1. Introduction
According to the European Parliament Directive 2010/31/EU, Article 9.1, all new
buildings will have to be “nearly net zero energy” by 2020 (2018 for public buildings).
Thus any energy required for the operation of the building will have to be compensated
by the equivalent amount of energy supplied back by the building over the course of one
year. In order to be creditable, the energy provided back necessarily has to be obtained
from renewable sources, since otherwise again primary energy would be used. This
directive immediately raises three implications for the future usage of energy within
buildings:
x The consumption of energy has to be reduced significantly.

x The energy usage within the building has to be as efficient as possible.
x The usage of renewable energy sources is imperative to compensate for any
energy acquired from sources external to the building.
Due to the fact that the provision and the usage of various renewable energy resources –
such as hydroelectricity, wind energy or biomass – within any building is very limited,
it is much more likely that the energy used for possibly paying back will have to be
converted from solar energy. Solar energy not only possesses the advantage of being
comparatively evenly available to everybody but furthermore that it is easily accessible
and harvestable in systems that range from very small – such as those used in individual
households - to very large - for office or industrial buildings. Solar-thermal and
photovoltaic products that ensure reliable and sustainable kWh generation have been
developed and are available worldwide at reasonable prices.
The energy consumption of buildings is typically in the form of thermal or electrical

energy. The thermal energy is required to either heat or cool the building to a desired
operation temperature. Thus the energy required primarily is related to the type of usage
of the building and the thermal losses or gains through the building envelope. This
933
Challenging Glass 3
correlation is mainly described by the U value and g value of the product or the
combination of products. As both properties are available for the products used in the
building skin, almost any wall composition and construction can be calculated; the
energy requirements can be predicted.
100 %
90 %
80 %
Irradiation 70 %
60 %
G - value [%]
50 %
40 %
Absorption 30 %
Transmission 20 %
Solar heat gain
Reflection Secondary heat 10 %
transfer
0%
Christof Erban, Schüco International KG
0 % 10 % 20 % 30 % 40 % 50 % 60 % 70 % 80 % 90 % 100 %
Cell coverage [%]
Figure 1a, b: (a) Interaction of solar radiation and PV modules. (b) Total solar energy transmittance (g value)
of semi-transparent PV elements vs. cell coverage for monolithic panels and insulating glazing.
Unfortunately, constant boundary conditions are assumed when the U or g value is

calculated or determined in experiments as described in the relevant standards.
Especially the g value – describing the net solar energy flow – through a component is
strongly affected by its reflection, absorption and transmission, which as can be seen in
Figures 2 and 3 are not constant for varying incident angles.

1.0 1.0
0.8 0.8
0.6 0.6
6 mm Float fg023066
Tdd [-]
Tdd [-]
T 0° s-pol. (=p-pol.) calc.

T 0° p-pol. (=s-pol.) calc.
0.4 0.4
T 20° s-pol. calc.
T 20° p-pol. calc. T 30° s-pol. calc.
0.2 T 60° p-pol. calc. 0.2
T 75° s-pol. calc.
T 75° p-pol. calc. T s-pol. mess: 0°, 30°, 45°, 60°, 75°
T p-pol. mess: 0°, 30°, 45°, 60°, 75°
0.0 0.0
300 400 500 600 700 800 300 400 500 600 700 800
Wellenlänge [nm] Wellenlänge [nm]
Figure 2 Transmission spectra of p and s-polarized light through 6 mm float glass for different angles of
incidence [1]
934
Building Design Might Soon Be Driven by Energy Requirements Only

0.50 1.0
6 mm Float fg023066 6 mm Float R 0° s-pol. (= p-pol.) calc.
0.45 fg023066 R 20° s-pol. calc.
R 0° p-pol. (=s-pol.) calc. R 30° s-pol. calc.
0.40 0.8 R 45° s-pol. calc.
R 20° p-pol. calc.
R 30° p-pol. calc. R 60° s-pol. calc.
0.35
R 45° p-pol. calc. R 75° s-pol. calc.
0.30 R 60° p-pol. calc. 0.6 R s-pol. mess: 20°, 30°, 45°, 60°, 75°
Rdd [-]
Rdd [-]
R 75° p-pol. calc.
0.25
R p-pol. mess: 20°, 30°, 45°, 60°, 75°
0.20 0.4
0.15
0.10 0.2
0.05
0.00 0.0
300 400 500 600 700 800 300 400 500 600 700 800
Wellenlänge [nm] Wellenlänge [nm]
Figure 3 Reflection spectra of p and s-polarized light through 6 mm float glass vs. for different angles of
incidence [1]
As both reflectance and transmission significantly depend on the glass composition,

such as DGU, coatings, surface structures, etc. different glass specimen do not only
have a varying U values and g values, but furthermore show a different behavior when
changing the incident angle. Figure 4 shows that especially coated glass types show
large deviations from the g value being determined at 0° incident angle.

Figure 4: Calculated angular functions of total solar transmittance g for three DGU [2]
The non-adequate consideration of the incident angle dependency would not be of

interest when calculating the thermal impact onto buildings if the irradiation energy
sums of these angles would not contribute significantly, since the heat gain is the
product of g value and energy impact. But as figure 5 shows the opposite is the case.
The incident angle of 0° for which the solar heat gain coefficient conventionally is
determined contributes insignificantly to the annual sum, whereas the significant heat
impact occurs at angles deviating from 0°.
935
Challenging Glass 3

E[kWh/m²/a/°
30 30
E[kWh/m²/a/°
25 25
20 20
15 15
10 10
5 5
0 0
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
J [°] J [°]
Figure 5: Annual irradiation energy sums for different angles of incidence in Freiburg.
South vertical (blue) South vertical (blue),
south 30° tilted (red) west vertical (red),
south 60° tilted (green) west-south-west vertical (green) [3]
The incident angle of the solar irradiation onto a glass surface depends on the position
of the sun, the orientation and the tilt angle of the surface itself, and the location of the
building. Consequently areas that are oriented differently will have an entirely different
heat impact onto a building.
2. Conclusion
The “flat-rate” solar heat gain coefficient G - as determined in the relevant standards -
provide a simple means to compare one product with another one, but they are, without
further considerations, practically useless when engineering an energy consumption
wise optimized building. Without considering the incident angular dependency even
simulations that account for the orientation of the surface as well as the sun, are not
exact enough to really optimize the thermal behavior of a building.
A low value for a solar heat gain coefficient is not desirable per-se. It does not
necessarily lead to the optimum when optimizing the annual energy performance of
buildings. In winter a rather high but controlled solar heat coefficient is desired for
passive heating whereas in summer a rather low solar heat gain coefficient is desired to
reduce the risk of overheating, resp. omit the necessity of active cooling.
Figure 6 shows the incident angle dependency of monthly irradiation energy sums for
south vertical oriented areas in Freiburg. Both images clearly show that:
x The weather has a very significant effect on the absolute number of irradiation
sum per angle, as indicated by the shown difference for January and December
(left), respectively June and July (right)
x (Meteonorm test reference year data have been used)
x The distribution of incident angle dependency of the monthly irradiation
energy sums is very little effected by the weather.
x The maximum energy impact in winter is not significantly smaller than in
summer for south oriented vertical areas.
936
E[kWh/m²/a/°
5 5
E[kWh/m²/a/°
4,5 4,5
4 4
3,5 3,5
3 3
2,5 2,5
2 2
1,5 1,5
1 1
0,5 0,5
0 0
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
J [°] J [°]
Figure 6: Monthly irradiation energy sums for south vertical oriented areas for different angles of incidence in
Freiburg.
January (blue), June (blue),
December (red) July (red), [3]
3. Considerations concerning the compensation of energy

The European Parliament Directive 2010/31/EU, Article 9.1 requires all new buildings
to be “nearly net zero energy” by 2020. As stated above solar energy most likely will
most likely provide the means to possibly compensate the use of energy in order to
achieve an annual energy-saldo of zero.
In order to fulfill this requirement, one - in addition the knowledge concerning the
impact onto the building - needs to know
x how much electricity a photovoltaic system is able to produce per given m² at

the location the building is located
x how much electricity a photovoltaic system is able to produce per given m² of
ground floor considering the orientation of the building and the tilt angle of its
roof
x how much electricity a photovoltaic system is able to produce per given m² of
ground floor considering the topology of its roof.
Figure 7 shows the global horizontal irradiation (GHI) map as published by PV-GIS [4].
It shows that the irradiation covered by the scope of European Parliament Directive
2010/31/EU varies roughly by the factor of two - from approximately 800 kWh/m²/a to
approximately 1600 kWh/m²/a.
937
Challenging Glass 3
Figure 7: Annual total global horizontal irradiation [4]
Figure 8 shows the irradiation map for differently oriented and tilted surfaces.
annual irradiation Tilt angle

+170
North
-170
+16
-16 0
+1
0
0
10°
- 15
50
40
+1
-1
40
+1 30%
30 0
-13 20°
+1 40%
0
20 - 12
50% 30°
+ 110 -110
60%
+100 -1 00 40°
70%
West 0°10° 20°30°40°50°60°70°80°90° East
80% 50°
+80 -80°
90%
-70° 60°
+70 95%
0 -60
+6 ° 100% 70°
0 -5
0°
+5
0
-4
80°
+4
0°
+3 0
-3
0
+ 20
-20 °
°
+10
-10°
South 90°
© Christof Erban
Figure 8: Annual total irradiation for different orientations and tilt angles
Thus on a ground area of 100 m² the GHI ranges from 80.000 kWh in Scandinavia to
160.000 kWh in southern Spain. Considering an average system efficiency of 13% for
crystalline pv-cells (which account for 81,1% of the pv-market in Europe [5]) one can
estimate 10.400 kWh of electricity in Scandinavia and 20.800 kWh in southern Spain.
938
This theoretical value needs to be reconsidered in respect to suboptimal orientation of

the building’s roof and the relative size of the roof when compared to the 100m² ground
area. Respective values are shown in figure 9 and range from 30% to 115% depending
on the arrangement of the pv modules onto the roof, their tilt angle and their orientation.
Thus the range of electricity generation on a 100 m² building ranges from 3120 kWh
(type 2 building in Scandinavia) to 23920 kWh (type 4 building in Southern Spain).
active roof area

inactive roof area
inactive wall area
E
tilt angle 0° 30° 30° 30° 15° 30° N
S
rel. area covered1 100% 30% 58% 115% 104% 115% W
energy yield2 85% 100% 100% 100% 85% 78%

energy yield3 85% 30% 58% 115% 83% 98%
Figure 9: Energy yield for different roofs topologies and roof orientation
1
active roof area/ ground area, 2in comparison to optimum orientation, 3in comparison to optimum orientation
and ground area
In Europe photovoltaic arrays on residential roofs typically range from 3-5 kWp and
provide throughout the entire year enough electricity to compensate for all electricity
demands in the building, if the building is not heated or cooled by electricity. The
surplus in summertime is fed into the grid and sold. Photovoltaic in Middle Europe can
on an annual basis even provide enough electricity for heating using a heat pump and
thus fulfill the European directive demands for an annual net zero energy balance, but as
figure 10 shows there is a miscorrelation from when the electricity is provided by the
photovoltaic – in summer – to which the electricity is required by the heat pump – in
winter.
500
450
400
350
300
pv
250
Electricitydemand
200 Electricitydemadheatpump
150
100
50
0
Jan Feb Mrz Apr Mai June July Aug Sept. Oct Nov Dec
Figure 10: Monthly electricity balance for a residential house with a 3 kWp photovoltaic array vs.
a residential house with a 3 kWp photovoltaic array and heat pump (passive house standard) [6]
Today there exist no financially viable means or products that could provide the
required seasonal storage of electricity for single households. In addition there are
limitations to transfer large amounts of solar electricity to e.g. centralized seasonal
939
Challenging Glass 3
storage means by the capacity of utility grid. In countries with high amount of
photovoltaic systems installed – such as e.g. in Germany – photovoltaic contributes
significant amounts of the noon peak electricity (figure 11). On weekends, the public
grid reaches its capacity limits and cannot accept more photovoltaic generated
electricity, since lignite and uranium powered power plants cannot be reduced in their
within hours.
70000
60000
50000 photovoltaic
wind
40000
gas
30000 coal
run_of_the_river
20000 lignite
uranium
10000
0
1.Juni2012 8.Juni2012 15.Juni2012 22.Juni2012 29.Juni2012
Figure 11: Simulated electricity production for June 2012 in Germany [7]
To summarize: Depending on the location of the building and the building topology,
photovoltaic systems can principally generate enough electricity to fulfill the EPD
directive. But with the fast increasing number of photovoltaic systems, crucial timing,
storage and transfer problems will develop.
Photovoltaic systems can provide a shift of the daily noon peak as well as of the
seasonal summer winter peak just by being mounted at different orientations and tilting
angles.
4. Summary
The European Parliament Directive 2010/31/EU, Article 9.1, requires that all new
buildings will have to be “nearly net zero energy” by 2020 (2018 for public buildings).
The thermal impact onto buildings can be significantly be improved by accounting the
incident angle dependency when determining the solar heat gain factor g.
Photovoltaic mounted on buildings can on an annual basis provide principally enough

electricity to enable buildings to fulfill this requirement, but there is a large
miscorrelation of when the electricity is required to when the electricity is being
provided. In countries with a high amount of photovoltaic already today the limited
capacity of utility grid limits the amount of photovoltaic. A more demand driven
provision of photovoltaic is possible by adapting orientation and tilting angles to
achieve a peak shift of the electricity conventionally generated around noon time.
940
5. Acknowledgements
I would like to thank Helen Rose Wilson, Wendelin Sprenger and Tilmann Kuhn from
the Fraunhofer ISE Freiburg for their questions, proposals and comments and our
fruitful discussions.
6. References
[1] Polarisationseffekte bei winkelabhängigen Messungen von Reflexion und Transmission, Wilson, H.R.,
Colloquium Optische Spektrometrie COSP 2007 Berlin
[2] Angular-dependent Light and Total Solar Energy Transmittance for Complex Glaszings, Final Report,
Dr. Platzer, W., Fraunhofer ISE, Freiburg 2000.
[3] Irradiation data provided by W: Sprenger, Fraunhofer ISE, Freiburg 2011.
[4] Photovoltaic Geographical Information System, PV-GIS, JRC 2004, http://re.jrc.ec.europa.eu/pvgis
[5] EPIA, Global Market Outlook for Photovoltaics until 2015, Brüssel 2011
[6] Einfluss gebäudenaher Photovoltaik-Anlagen auf den Primärenergiebedarf von Gebäuden nach EnEV,
Roos, M., Henze, N., Boyanov, N., Maas, A., 4. Anwenderforum Bauwerkintegrierte Photovoltaik,
Kloster Banz, Bad Staffelstein, Bad Staffelstein 2012
[7] 30° geneigt, nach Süden orientiert – Sackgasse oder Königsweg, Erban, C., 4. Anwenderforum
Bauwerkintegrierte Photovoltaik, Kloster Banz, Bad Staffelstein, Bad Staffelstein 2012
941
doi:10.3233/978-1-61499-061-1-943
Ceramic Digital Printing -

Customizing Glass Façade Design
Bernd Hoffmann
Hoffmann GTD Glas-Technik-Design, Germany, GTD@gmx.com, CERDIP.com
Modern digital printing technologies using ceramic inks are providing new
interesting opportunities for longlasting decorations and optical effects on glass.
This way even multicoloured very large format images and decors can be printed
and fired on glass providing reflective high resolution prints, turning the functional
glass skin of a building into a more or less translucent artistic screen. After a short
introduction into the basic principles of the Ceramic Digital Printing technologie
and the world of Glass Façade Design the presentation will give an overview about
the advantages and limitations of the Ceramic Digital Printing for architectural
applications. Different design intentions, challenges and achievements will be
presented and discussed. TOPAZ - a helpful concept of preceding considerations
for the design process of glass facades will be introduced.
Keywords: Ceramic Digital Printing, Glass Facade Design, Facade Glass Design,
Glass Design Engineering, XXL-imaging
1. Ceramic Digital Printing
1.1. Ceramic Printing

The demand for a longlasting quality for printings on architectural glass is strongly
restricting the use of organic materials in combination with in-organic, mineralic glass.
In the last century Silkscreen Printing has become the major technique for the graphical
application of ceramic inks on glass. There have been many developments in the
Silkscreen Printing technology trying to enhance the quality and to reduce the efforts.
Even here the (organic) Digital Printing has entered the process, helping to avoid films
and the photographic stencil development on screens using CtS (computer-to-screen).
The Silkscreen Printing is an analog process still, even though many professionals may
often wrongly talk about a "Digital Silkscreen Fritting" when they are actually looking
for a "Ceramic Digital Printing".
943
Challenging Glass 3
Figure 1 : Ceramic Digital Printing : Harlem Hospital, NYC [3]
1.2. Digital Printing

Since the introduction of the applicable computer to the world in the 60´s of the last
century, most social and industrial processes have undergone revolutionary changes.
The quality, quantity and resolution of data processing has at least doubled in every
decade. So nowadays the digital processing has passed the challenges of it´s analog
ancestors in almost every working field.
Figure 2: HiRes-Ceramic Digital Printing-6-color-print [1]
944
Ceramic Digital Printing - Customizing Glass Façade Design
Even the limits of photografic image reproduction have yielded to the quality of modern
digital printing.
Table 1 : Decades of Resolution [1]
decade 60´s 70´s 80´s 90´s 2000 2010
resolution [dpi] 10 90 180 360 720 1440 2880
black & white , grey colour
1.3. XXL-Printing
According to the use of modern graphic programs, data processing and storage the
world of large format design has opened not only for highly sophisticated professionals.
Every image can now be processed and reproduced in high qualities in gigantic formats.
Figure 3 : White CDP with colour-illumination, Rockheim , Norway [3]
Table 2 is showing the amount of graphical data that can be processed by modern data
storage devices, to give a feeling for the amount of graphical information needed.
Table 2: Data sizes for graphical reproductions using BMP-files with a 360 dpi resolution [1]
1 m² = 14.400 * 14.400 pix = 200 Mpix data storage device

Black & White 25 MB
Greyscale (8-bit) 200 MB
RGB (8-bit) 600 MB CD
7 m² (RGB) 4.700 MB DVD
40 m² (RGB) 25 GB Blue Ray
Here we can see, that the amount of information for a high resolution image print might
be important for a smooth data-processing. One of the greatest challenges of a
professional Facade Glass Design preparation is it, to determinate the best composition
of vector- and bitmap-information according to the most important distances and effects.
945
Challenging Glass 3
Figure 4 a,b,c: CDP-detail seen from 30 m , 3 m , 30 cm distance , "One Reading Central", UK
1.4. Ceramic Digital Printing

The Ceramic Digital Printing (CDP) can be seen as a synergy of the advantages of the
Ceramic Silkscreen Printing (CSSP) and the Digital Printing with organic inks (solvent,
watersoluble, UV-cured, etc.).
For a successful development of the Ceramic Digital Printing on glass two main issues
had to be mastered:
A new system of colour-pigments with particle diameters less then some 1/100th of the
conventional sizes, performing with reliable results in the temperature range of flat glass
tempering (550-750°C) and sufficient mixing properties had to be found and established.
Figure 5: REM-images of CSSP and CDP pigments [3]
And even more challenging, the erosive character of ceramic pigments that are usually
grinding on the delicate architecture of multi-nozzled printheads, eating them up step by
step within an unpredictable period of time, had to be tamed.
946
Table 3: Advantages of the direct Ceramic Digital Printing on Glass

Organic Ceramic Analog Digital
water resistant + no films needed +
solvent resistant + no screens needed +
UV resistant + no colour change +
chemical resistant + individual colouring +
mechanical + variable thicknesses +
resistant within one layer
scratch resistant + multi colour registration +
+ colour spectrum fine positioning and +
reproduction
+ high resolution high resolution +
+ very thick layers
2. GlassJet
This issues of particle size and erosion have been solved, leading to the production of a
Ceramic Digital Printing system named "GlassJet" which has been introduced to public
during the Glasstec 2004 in Düsseldorf. This standard industrial printing solution is
capabel of printing 6 colours of ceramic pigments on flat glass directly, allowing a
resolution of more then 360 dpi, using a drop-on-demand technology on areas of 2800 x
3700 mm and even 3300 x 6000 mm in a single pass.
Figure 6: Ceramic Digital Printing system GlassJet PRO 24 [3]
2.1. Colourful Hi-Resolution

A high quality colourful fotorealistic reproduction can be achieved. Due to the low
viscosity of the inks on glass a resolution of 360 dpi in most cases will be the best
setting for most spectacular results. Every single point of an artwork can be coloured
and controlled individually, granting the highest quality in positioning and registration
(very important for multi-coloured prints).
947
Challenging Glass 3
Within one print the density of each part can be defined from highly translucent to solid,
controlling the amount of ink preferred.
Figure 7: Worldmap, opaque colours with transparent blue
2.2. Optical properties of GlassJet inks

The GlassJet inks are mostly reflective to garantee a bright colour reflection. Only the
blue ink is of transparent nature to allow a nice reproduction of water and air qualities.
It can be made reflective by adding other more reflective inks (white).
2.3. Certifications
Performance reliability and qualities are comparabel to the common industrial standards
of the professional Digital Printing. Thus digitally printed safety glass products have
aquired most of the certifications needed for their architectural outdoor application.
Their applicability monolithic or inside of laminated glass and/or in combination with

solar-control coatings is proofen.
2.4. Results
The range of chances for creating new artwork or reproducing design solutions on
architectural glass is tremendous. Here you can find a selection of some designs that
have been developed:
948
Figures 8 a, b and c: Imitations of Streakmetal ; Glassbricks ; PV-Elements [1]
Material Imitations
Optical imitations of fabrics, streakmetal, metal surfaces, carbon fibre, stone, marble,
plants, clouds, streaks, drops, glassbricks and other natural or technical objects and
effects have been produced with Ceramic Digital Printing on glass successfully and
built into facades and rooftops :
Figure 9 : Marble imitations with CDP , Ave. Friedland, Paris [4]
Figures 10 a, b, c : Brisbane towers micro-blinds; Stones; Etch-imitation, French consulate London
949
Challenging Glass 3
Figures 11 a, b, c: Wood-imitations and PV-cell-imitations
3. Conclusions
The Ceramic Digital Printing on glass can substitute most of the known techniques of
glass decoration up to a very high extend.
Most of the effects that can be attained can be combined with one another
simultaneously in just one step of production.
Technical and decorative effects for new individual indoor and outdoor design solutions
can be generated that have been impossible before.
For architectural applications biggest advantage of the CDP will probably be the
multicoloured XXL-format production, where one image can be tiled and distributed to
various tiles so that every panel can own its own design without pushing the costs and
efforts of production skyhigh.
A further advantage of the CDP is the opportunity of shortening the time necessary for a
glass design development process by providing different variations of CDP-samples
within very short periods of time. The production time (the print, drying and firing) of
one small sample can be reduced to be less then one hour if everything has been
prepared nicely.
4. Glass Design Engineering

The Ceramic Digital Printing is not just a technique of reproduction but can be used as a
totally new technique providing effects and artwork solutions which have formerly been
impossible.
With CDP the faculty of Glass Facade Design is no longer restricted to standard sheet
materials and standard design patterns like monochrome silkscreen rasters, for example.
Sophisticated colourful graphical compositions can now be calibrated exactly according
to architectural intents.
950
Figure 12 a, b : 6-color silkscreen imitation by CDP, Bundesministerium BMF, Berlin

( the top-down-gradation, staircases and the green tree are part of the printed artwork )
This advanced processes of creating most suitable skins for a building are no longer just
about picking an artwork and ordering some print-versions on glass, but they are
bearing the chance and challenge to optimize the design individually.
This optimizations are depending on combinations of design parameters that often have
not been relevant to be taken into consideration before. Sometimes those parameters
even seem to possess uncertain definitions.
Therefore we are presenting a checklist for glass design projects that will help to
consider those parameters to round-up a project definition and may provide useful
conventions for a plain design language:
5. T O P A Z a checklist for facade glass design projects
5.1. TILING
The first step will be the determination of the maximum panel format to provide the
chance for a smooth workflow without exceptions.
5.2. OPACITY
The lightflow is the major difference between "Glass Design" and "Paper Design".
Paper Design is based on light-reflection mainly, whereas Glass Design is handling 3
different important main characteristics of lightflow :
Reflection (REF) Transmission (TRM) Absorption (ABS)
951
Challenging Glass 3
Reflection:
Shadow box applications will be printed opaque or even solid to avoid the background
to shine through. Those are working almost completely on reflection.
Most of the inks of the GlassJet are of reflective nature to be compatible for outer
facade design, avoiding the "church-window-effect", the fact, that transparent colours
are great for projecting transmitting light but tend to "fall into grey" being seen in
reflection.
Figures 13 a, b : The Church-Window-Effect: Cathedral-windows in transmission & in reflection
Transmission:
Most facade applications will use the wide range of light-transmission.
Functional and aesthetical aspects will determine the transmission wanted.
These effects can be determined qualitatively by defining the scattering of the light
and quantitatively by calculating the lightflow within the visual spectrum.
Absorption:
The absorption will be important for the energy flow during the production process and
for the energetic ecology of the mounted panel (IGU) and its environment.
Figures 14 a, b : CDP in transmission (daylight) & in reflection (black background)
952
5.3. POSITIONING
A design printed on transparent substrates can often be seen from both sides. Even if
inks are printed simultaneously there will always be more or less significant differences
in the performance of surfaces printed with different sequences. Therefore it might be
important to determine, which sides will be most important.
For translucent or transparent panels it will be essential to consider the positionings of

the light and of the viewer to get an idea about the resulting design effects.
5.4. ACTIVE WHITE

Opaque prints on glass are using the color White to print light-reflections. On paper
prints those will be left unprinted. On glass, White can be functioning as a substitute for
the usual paper background.
Translucent prints might need the colour White to show light in reflection, but they
usually will not be specified in the artworks automatically.
In most cases it will be necessary to discuss the use of the White on glass in advance.
5.5. ZOOM
For large architectural artworks it often is very important to chose an interesting graphic
strategy for the transfer of the relatively small basic image bitmaps into large facade
cladding prints, especially if people will have to live with it closely.
Figure 15 a, b : Photoshop-file and printed glass panel with recomposed design
A more practical version of the TOPAZ checklist can be found on CERDIP.COM
953
Challenging Glass 3
Figure 16 : CDP-printed facade , "One Reading Central" [4]
6. Summary
The technology of the Ceramic Digital Printing is providing an instrumentation that has
proofen itself to be a real advantage in the field of designing and producing architectural
flat glass, inspiring professionals to create new visions and solutions for future projects.
The Ceramic Digital Printing is not just a machine of standard image reproduction, but
an instrument of versatile approaches and effects, that might help to create really
progressive individual design solutions, helping to add more character to the functional
glass skins of modern architecture.
7. Facade Glass Design Service

Many developments have been made according to the new advantages and challenges of
the CDP and the design and production process surrounding it.
New designs and technologies are demanding proper tools, experience and sometimes
even new hard- and software solutions to be developed and applied successfully.
A combined Facade Design Consulting, Facade Design Glass Production, Prepress and
Production Service will be helpful to optimize a creative and productive workflow.
8. Design Software Solutions

Design software solutions like Bitmap-Analyzers, Bitmap-Modulators and Bitmap-
Synthesizers are helping to optimize the composition of colours. (B2B-software).
954
Even more interesting for the XXL-format imaging will be Bitmap-to-Vector

modulators and synthesizers that are able to modulate or generate graphical structures
by the use of vector objects according to mathematical algorithms (B2V-software).
This tools will help to provide a wider spectrum of alternatives for the production of
new designs.
Figure 17: Analysing printed contrasts with CERDIP-software [1]

Figure 18: Analysing colourspace-distributions with CERDIP-software [1]
9. Acknowledgements
x Hoffmann GTD, Germany
x DipTech, Israel
x Interpane-Sicherheitsglas, Germany
10. References
All texts, pictures and tables by Hoffmann GTD unless stated differently.
[1] Hoffmann GTD
[2] Cerdip.com
[3] Dip-Tech.com
[4] Interpane.com (Ipadecor K)
955
doi:10.3233/978-1-61499-061-1-957
Behaviour of Electrochromic Glass in the

Mediterranean Area
Gianraffaele Loddo, Daniela Ludoni, Marco Pittaluga, Gian Piero Cossu
Civil and Environment Engineering and Architecture Department, Cagliari University,
Italy, grloddo@unica.it
For several years, our university research group is interested in innovative materials
in the field of energy saving, with particular attention in transparent building
components, regarding, specifically, the use of electrochromic glass (EC). The high
sunshine that characterizes the climate of the Mediterranean regions makes this
technology particularly interesting in relation to indoor comfort, energy saving and
consequently the environmental respect. Currently no specific studies, full scale,
have been set for testing and evaluation of the EC in this very challenging. Our
research group built, for this purpose, two identical test rooms. In the first room we
implemented low-emission glass and in the second one EC glass. Indoor and
external comparative measurements have performed and recorded. In this paper we
present in detail the results of this experiment and the conclusions for the tested
January period.
Keywords: Technological Innovation, Energy saving, Electrochromic glass
1. Preface
In recent decades the need to make sustainable interventions in buildings has led to the
development of new materials and building systems that allow having high levels of
comfort, a significant saving non-renewable energy and reducing CO2 emissions.
Among the materials studied, but known for some time, are of special interest those
defined chromogenic, capable of adapting to different environmental conditions: the use
of these materials over the past decade has led to the production and commercialization
of electrochromic windows. At present the production, and therefore most of the
experiments related to it, takes place in the USA. Particular climatic conditions in
Europe and especially the Mediterranean ones, with very important seasonal variations
of temperature and sunshine, can help find effective use of these special glass, so it is
necessary to study these technologies to verify performance and the ability to achieve
with them, smart and technological envelopes and sustainable too.
2. Electrochromic glazing
The dynamical glass exploit the characteristics of the chromogenic materials: they are
able to modify their transparency and/or their colour and then return to the initial state.
The change can take place actively, on command, or passively by the action of light or
temperature. Electrochromic glass (EC) are able to modify, on user demand, the
intensity of colouring of the glass. The electrochromic base panel is made of depositing
vacuum, on a slab Low Emissive, 5 layers of ceramic material.
957
Challenging Glass 3
The panel is then assembled and integrated in a unit which consists of:
x External Slab of tempered Low-E glass (called LE), with a thickness of 6 mm,
on whose inner surface is applied a film of metal oxides;
x Internal Slab of laminated or tempered glass, thickness of 6 mm;
x Air gap filled with argon gas thickness of 12.7 mm;
x Stainless Steel Spacer.
The activation of a low electric field produces a reaction of oxidation-reduction, in

which lithium ions, migrate of layers. By reversing the polarity is obtained reverse
switching. In the fully tinted state ions prevent the passage of part of the visible light, of
UV and by more than 91% of incoming solar radiation. The increase in colouring
intensity keeps the visibility and the outdoor relationship and, in the same time, prevents
the vision from outdoor to indoor. The variation occurs in a period of time ranging from
3' up to 5', depending on the size of the glass and air temperature, and can be activated
manually or by automated system. The variation of the coloration is able to significantly
change the g value (Solar Heat Gain Coefficient). In a Low-Emissivity glass (LE),
values of SHGC and the Transmission of the visible field are respectively 0.42 and 71,
while in an EC these values are in the fully tinted state 0.09 and 3. The transmittance
(U) is 1.30 W/m² K for the LE and 1.49 W/m² K for the EC. The electrochromic
glazings have exceeded 100,000 hours of durability testing and are certified for 50,000
hours (NREL Test E-2141). Currently the EC is produced in various colours including
green, blue and gray. The consumption of electricity necessary to activation is 2.8 W/m²
during the passage of colouring, and 1.0 W/ m² to system.
3. Experimental set
The studies were conducted in two phases: a simulation and testing. The test is
performed on two identical rooms for geometric features, construction and orientation,
built on the roof of the Department Architecture, University of Cagliari. The two phases
are important both to verify the reliability of the software to accurately evaluate the
behaviour of windows EC and LE in different climatic conditions.
3.1. Simulation
The simulation was performed on two pairs of rooms: the first one in a cellular brick
bearing wall and the other one with balloon frame wood: no evident difference was
noted. In the two versions has been simulated and compared the use of LE and EC glass.
The results showed very similar values in the two construction types. For the real test,
therefore, has been chosen the second system, which offers the advantage of the rapid
implementation and easy disassembly and recycling of the material used at the end of
experiment. [1] The team did several studies regarding the envelope, to found the best
solution conforming to the Italian rules. So that one chosen is simply not a box to
experimentation; it has all energetic behaviour performances requested to the actually
buildings.
3.2. The construction of test rooms

Construction work started in late June 2011. Two rooms are identical; the only
difference is the type of glazing. The external dimension is 400x400 cm in plan with an
internal height of 275 cm. The orientation is north-south axis along a diagonal so that
958
Behaviour of Electrochromic Glass in the Mediterranean Area
the two walls (one with the door the other one with the window) are respectively
oriented to the South East and South West while the two blank walls are oriented to the
North and East North West. On the existing roof, for each room, were resting 8 wooden
beams (8x8x400 cm). The floor (8 cm thick) is set on these ones, and also pillars
(16x5x275 cm), that are dual at the corners and openings. The roof is made of beams
(16x5x400 cm) resting on the uprights. The floor, walls and roof are insulated and its
sealing is slated adhesive. There are two openings: one door (80x220 cm) with opaque
panel (80x90 cm) and transparent panel (80x115 cm), a window (160x115 cm). The EC
glass are in the blue version. The distance between two test rooms is 20 m.
Figure 1a, b, c,: Some pictures of construction phases: structures, external insulating wall, and the 2 test
rooms.
4. Characteristics relating to the thermal exchange of the elements of the envelope

To define the heat exchange is necessary to describe the opaque elements of test rooms.
The thermal transmittance is a function of the stratigraphy of the elements that
constitute test rooms: vertical walls, flat roof and floor. (Figure 2a, b and c)
Figure 2a, b and c: Details of the foundation, wall and roof.
Test rooms components are composed by layers adjacent to each other materials. It is
shown that the total thermal resistance is given by the sum of the thermal resistances of
each layer. For a homogeneous material, the thermal resistance is the ratio between the
thickness of the material and its thermal conductivity:
s
R [m².K/W]
O (1)
Summarized in the table below the main constituent material of the test rooms envelope
and some physical properties such as density and thermal conductivity (Table 1):
959
Challenging Glass 3
Table 1: Characteristics of the materials

Rif. Material Thickness Density Thermal
Fig.2 [cm] [kg/m³] Conductivity
[W/mK]
0 Existing slab = = =
1 Oriented Strand Board (OSB) 1 650 0.13
2 Thermal insulation (EPS) 5 30 0.036
3 Screed (lean concrete) 4 2200 0.93
4 Planking (pine) 2 450 0,12
5 Thermal insulation (rockwool) 16 35 0.036
6 Skim coat cementitious 0.2 1500 0.90
7 Coating to coat 0.3 1800 0.70
8 Vapour barrier 0.05 = =
9 Ventilated air gap 4 = =
10 Panelling (pine) 1 450 0.12
11 Waterproofing membrane 0.2 = =
The conductance is defined, in case of a not homogeneous material. This one is the
parameter which characterizes the thermal performance of not homogeneous materials
with certain features. The thermal conductance is the inverse of the thermal resistance:
1
C
RT (2)
This last parameter is useful because the total thermal conductance of different adjacent
materials along the direction of heat flux is given by the sum of the thermal conductance.
The thermal conductance is adopted also in the definition of the thermal properties of
the air gaps which also vary depending on the thickness. It should be noted that, in
addition, heat transfer by conduction through the wall, we have energy exchange
(convection and radiation) between the room air and wall surfaces, these exchanges are
taken into account in surface thermal resistances Rsi and Rse which simplified
calculation is described in the UNI EN ISO 6946 (Tab. 2).
Table 2: Determination of superficial thermal resistance

Direction of heat flow
Upward Horizontal Down
Internal superficial thermal 0.10 0.13 0.17
resistance [m² K/W]
External superficial thermal 0.04 0.04 0.04
resistance [m² K/W]
960
The phenomenon of heat transmission between two rooms makes reference to the
temperatures of the air contained in the two rooms, in accordance with the following
equation:
x
q tot U A (Ti Te )
(3)
Where U is the thermal transmittance of the wall, A the surface of the wall; Ti and Te,
respectively, the indoor and outdoor air temperature.
U, thermal transmittance of the wall or global heat transfer coefficient [W/(m².K)] is:
1
U n l m
s 1 1
Rsi ¦ k ¦ ¦ Rse
k 1 Ok j 1 Cj y 1 Cy
(4)
Where:
x Rsi = Internal superficial thermal resistance [(m².K)/W];

x k = Thermal conductivity of the k-th homogeneous layer [W/(m.K)];
x sk = Thickness of the k-th homogeneous layer[m];
x n = Number of homogeneous layers;
x Cj = Thermal conductance of the j-th layer not homogeneous [W/( m².K)];
x m = Number of layers for which the thermal conductance is known;
x Cy = Thermal conductance of the y-th air gap [W/( m².K)];
x l = Number of air gaps;
x Rse = External superficial thermal resistance [(m².K)/W];
The average thermal transmittance is equal to the weighted average of the different
elements transmittances (along the direction of heat flow) in which the weight is
represented by the surface (in the perpendicular direction to the heat flow). Following
values of average transmission are obtained applying the general equation (4) for the
calculation of the transmittance to elements of test rooms’ envelope (Figures 2a, b, c)
and to other special elements; and using the properties of Table 1 and 2.
Table 3: Average Transmittance values and legal limit transmittance value for the envelope elements
Element Average transmittance Legal limit Transmittance
[W/m² K] [W/m² K]
Vertical walls 0.17 0.40
Flat roof 0.26 0.38
Floor 0.36 0.42
961
Challenging Glass 3
The averages calculated transmittances are compared with the limit values prescribed by
the Legislative Ordinance 311/2006 for the municipalities such as Cagliari, sited at
climate zone C (Daily Degree = 990), the transmittance values are all lower than limit
values, vertical walls have a thermal transmittance value of less than 58% of limit
values.
5. Collecting Data
5.1. Survey instruments

In each of two rooms is a microclimatic station consists of sensors, connected to a data
logger capable to measure: dry and wet bulb temperature, relative humidity, solar
radiation and transmittance. The combined sensors of temperature and humidity work in
a range between -20°C and +80°C and between 5% and 98% RH. The accuracy is ±
0.30°C for temperature and ± 2% for humidity. The pyranometer measures the solar
radiation between 305 and 2000 nm, with an accuracy of 10 microvolt/W/m².
Figure 3: Instruments in the EC room.
5.2. Method of survey and working of test rooms

The detection covers 24 hours with the recording of the internal data. The outside
meteorological data are quoted daily. In January 2012 in the room with EC glass, the
activation of electrochromic panel is from 7.15 to 15.20 for door (South-East), and from
8.40 to 16.55 for window (South-West). During the first week of February 2012 the EC
glazing were kept in the Off position to audit the behaviour of windows in these
conditions. In continuation of the trial period, activation (ON) or not (OFF) of the EC
windows will be modified in relation to the solar duration, intensity and inclination.
Figure 4a, b and c: Room with EC glazing: in state Off (indoor) and state On (indoor and outdoor).
962
6. Data for January 2012
6.1. Outside temperatures

In January 2012, the meteorological data of Cagliari (Lat. 39°13'N - Long. 09°7'E) were
varied in a limited range: the temperature had a minimum of 0°C (18th Jan.) and a
maximum of 18°C (2th Jan.) with an minimum average daily of 6°C (18th Jan.) and a
maximum of 14°C (22th Jan.), relative humidity between 51% (6th Jan.) 84% (29th Jan.).
20,00
15,00 T min
10,00
T max
5,00
0,00 T average
-5,00
1-gen
3-gen
5-gen
7-gen
9-gen
11-gen
13-gen
15-gen
17-gen
19-gen
21-gen
23-gen
25-gen
27-gen
29-gen
31-gen
Graph 1: January 2012, Temperatures of Cagliari (minimum, average, maximum).
The analysis data shows that, in January, the days of 2th (hottest day), 6th (minimum
humidity), 18th (coldest and minimum average temp.), 22th (maximum average temp.)
and 29th (maximum humidity) are considered most significant in the period.
6.2. Inside collected data (temperatures)

The data collection occurred on 01th, every 5' from 0.00 hours until 12.00 of 19th , and
was followed every 15', from 12.00 of 19th until 24.00 of 31th. The new range was
considered sufficient to evaluations while the previous one was overly, and
unnecessarily closes. The collected data in the rooms LE and EC were detected by
identifying, in each day, the time to maximum and minimum temperature and maximum
and minimum internal solar radiation. These data are summarized as follows:
Table 4: Detecting outdoor temperature range in significant days of January 2012

Date 02 06 18 22 29
Ext. Temp. 12 13 6 14 8
av.ge [°C]
Ext. Temp. 6 8 0 13 6
min [°C]
Ext. Temp. 18 16 14 17 10
max [°C]
963
Challenging Glass 3
Table 5: Detecting indoor temperature range in EC room in significant days of January 2012
Date 02 06 18 22 29
T.min [°C] 13.30 12.76 8.28 13.43 10.92
T.max [°C] 17.51 13.99 12.27 15.78 13.13
T [°C] 4.21 1.23 3.99 2.35 2.21
Table 6: Detecting indoor temperature range in LE room in significant days of January 2012.
Date 02 06 18 22 29
T.min [°C] 13.89 13.39 9.32 14.24 11.62
T.max [°C] 23.20 18.58 18.94 20.96 14.41
T [°C] 9.31 5.19 9.62 6.72 2.79
20 ,00
15 ,00
EC min.
10 ,00
LE min.
5 ,00
0 ,00
01-gen
03-gen
05-gen
07-gen
13-gen
15-gen
17-gen
19-gen
2 1-gen
2 3-gen
25-gen
27-gen
29-gen
31-gen
09-gen
11-gen
Graph 2: January 2012, Comparison of minimum indoor temperatures in EC and LE room
25,00
20,00
15,00 EC max.
LE max.
10,00
5,00
01-gen
03-gen
05-gen
07-gen
09-gen
11-gen
13-gen
15-gen
17-gen
19-gen
21-gen
23-gen
25-gen
27-gen
29-gen
31-gen
Graph 3: January 2012, Comparison of maximum indoor temperatures in EC and LE room
964
The comparison between the Graphs 2 and 3 shows, in an evident way, as in the rooms
of test, while it is negligible the difference between the minimum temperatures (it has
been found, in the month, a daily average T of 0.782°C favourable for the room LE)
the difference between the maximum temperature is considerable (measured in months,
a daily average T of 4.184°C for the LE room, even with daily peaks of 9°C). It's very
interesting to note how minimum and maximum temperature peaks occur with the same
hourly frequency. The most critical hours were 8.00 (95%) for the minimum
temperature and 16.00 (80%) for the maximum one.
6.3. Inside data collected (incoming solar radiation)

Similarly incoming solar radiation data were detected. We show those ones for 2th (max
external temp. = 18°C and clear sky) and January 18th (external temp min = 0°C and
external average min temp. = 6°C and clear sky).
During on 2th the global solar radiation in the room EC (expressed as [W/cm²= 10-
6
W/104m²]) had a low value close to 0 with two maximum around 10.30 with 201
W/cm² and at 15.35 with 251 W/cm² (see graph 4)
300
250
200
150
100
50
0
0.00
1.30
3.00
4.30
6.00
7.30
9.00
10.30
12.00
13.30
15.00
16.30
18.00
19.30
21.00
22.30
0.00
Graph 4: 02 th January 2012 (hottest day), incoming solar radiation in the EC room
The same day 2th global solar radiation in the room LE had a value close to 0 at night
and evening and grew up between 8.00 and 8.30 going from 95 W/cm² to 571 W/cm²
and continuing to grow until at 11:00 when it reached the maximum of 16,520 W/cm²
(see graph 5).
18000
15000
12000
9000
6000
3000
0
0.00
1.30
3.00
4.30
6.00
7.30
9.00
10.30
12.00
13.30
15.00
16.30
18.00
19.30
21.00
22.30
0.00
Graph 5: 02th January 2012 (hottest day), incoming solar radiation in the LE room
965
Challenging Glass 3
During the day 18th global solar radiation measured in room EC has shown a very low
value, almost always close to 0 W/cm² with two maximum around 10.30 am with 231
W/cm², and 17.00 with 635 W/cm² (see graph 6).
The same day 18th global solar radiation measured in the room LE still showed a
reduced value, very close to 0 at night and evening with a value that has begun to grow
between 8.00 and 8.30 going from a value of 88 W/cm² to 519 W/cm² continuing to
grow until 11.00 when it reached its peak of 18,640 W/cm² (see graph 7).
The incoming solar radiation during the 2th. in EC room was 1.50% of that one in the LE
room. On the 18th.in EC room solar radiation was 3.50% of that one in the LE room.
700
600
500
400
300
200
100
0
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
11.00
12.00
13.00
14.00
15.00
16.00
17.00
18.00
19.00
20.00
21.00
22.00
23.00
0.00
Graph 6: 18th January 2012 (coldest day), incoming solar radiation in the EC room
21000
18000
15000
12000
9000
6000
3000
0
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
11.00
12.00
13.00
14.00
15.00
16.00
17.00
18.00
19.00
20.00
21.00
22.00
23.00
0.00
Graph 7: 18th January 2012 (coldest day), incoming solar radiation in the LE room
7. The data of February 2012

In order to verify the correct operation of measuring instruments and homogeneity in
the energy behaviour of two test rooms, throughout the first week of February 2012, the
EC glazing were always kept in its OFF state, and it proceeded to measure external and
internal parameters, in the same way of January 2012.
7.1. Outside temperatures

The days from 1th to 7th February 2012 Cagliari had a minimum temperature of -3°C
(07th.) and a maximum of 12°C (01th.) with an average daily minimum of 3°C (06th.) and
maximum one of 9°C (01th.), relative humidity ranged between 59% (01th.) 83% (03th.).
966
14,00
12,00
10,00
8,00 T min
6,00 T max
4,00
2,00 T average
0,00
-2,00
-4,00
2.01
2.02
2.03
2.04
2.05
2.06
2.07
Graph 8: 01th - 07th February 2012, Temperatures of Cagliari (minimum, average, maximum).
Among seven days the only one, which represents the average, is on 03th., when there
were 4°C minimum, 7°C maximum, and 6°C average temperature. From the
meteorological point of view it was a very special week with temperatures much lower
than the annually averages that are during the same period.
7.2. Inside data collected (temperatures)

The data collection was continuous throughout the period and took place every 15'. Data
collected in the EC and LE rooms identify, in each day, the time to maximum and
minimum temperature and maximum and minimum internal solar radiation. These data
are summarized as follows:
Table 7: Record EC room temperature in the first week of February 2012

Date 01 02 03 04 05 06 07
T.min [°C] 9.80 9.50 6.93 5.23 5.32 5.40 3.99
T.max [°C] 16.37 11.12 9.07 10.33 14.94 12.60 6.25

T [°C] 6.57 1.62 2.14 5.10 9.62 7.20 2.26
Table 8: Record LE room temperature in the first week of February 2012

Date 01 02 03 04 05 06 07
T.min [°C] 9.45 9.50 6.96 5.18 5.51 5.87 4.42
T.max [°C] 17.51 11.80 9.06 10.92 12.82 13.31 6.93
T [°C] 8.06 2.30 2.10 5.74 7.31 7.44 2.51
967
Challenging Glass 3
T intern. EC
8 T intern. LE
6
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
11.00
12.00
13.00
14.00
15.00
16.00
17.00
18.00
19.00
20.00
21.00
22.00
23.00
0.00
Graph 9: 03th February 2012, Comparison of indoor temperatures of the EC and LE rooms
The analysis of the Graph 9 shows how, in two rooms, when the EC are in state OFF, is
negligible the difference between the minimum temperatures (T average daily in week
of 0.11°C in favour for LE). The difference between the maximum temperatures is of
the same order (average T daily in week of 0.24°C in favour for LE). The EC (in state
OFF) glass have a similar behaviour than the LE glass, about the outside temperature,
and they provide the same shielding too.
7.3. Inside data collected (incoming solar radiation)

In order to the solar radiation we show the example relating to 3th, for its outside average
temperature = 5.50°C (very cloudy sky). During the day, the global solar radiation in the
room EC (OFF) showed a very low value close to 0 that has begun to grow at 8.00, to
reach a maximum of 385 W/cm² to 16.30, and, from 18.00, get values close to 0.
The same day there were similar measured values in the EC (in OFF state) and LE
rooms. Values, almost always close to 0, have gradually started to grow around at 8.00
to reach a maximum of 455 W/cm² at 16.30, and then to descend fast to values close to
0 (see graph 10). During the day on 3th, the incoming solar radiation had, in two rooms,
a very similar trend. The maximum value, reached in EC room (in state OFF) of 385
W/cm², is comparable (84%) with that one of 455 W/cm² of the LE room. The
difference is due to the light colour which the glass EC keeps in state OFF. The data
obtained are very different from those ones in January (compare graph 4, 5, 6, 7, 10).
500
400 Solar Rad. EC
300 Solar Rad. LE
200
100
0
0.00
1.30
3.00
4.30
6.00
7.30
9.00
10.30
12.00
13.30
15.00
16.30
18.00
19.30
21.00
22.30
0.00
Graph 10: 03th February 2012 (average temperature), Comparison incoming solar radiation EC and LE rooms
968
8. Conclusions
This first phase of experimentation, with this configuration of test rooms, will last
through 2012 and it allows making interesting observations: as expected, but now we
can quantify with precision, the EC glass, even in OFF, are slightly disadvantageous in
the short cold period that characterizes the Mediterranean climate. The advantage, in the
warm months, compared to LE of EC windows is already announced by the great power
that the shielding EC have been shown to have. The goals for the search are not simply
a quantification of the performance of the EC but also their relationship with the
building envelope both in terms of typology and material of the relationship between
quantity of glass surface and blank wall. The special conditions in the Mediterranean (in
Cagliari there are over 2,600 hours of sunshine per year) and the redevelopment of the
rich historical buildings are the most important variables in research.
9. Acknowledgements
The research team wishes to thank the partners: eng. Ubaldo Carlini, Laborvetro Srl
(Cagliari - Italy); Casa Extra Clima sas (Cagliari - Italy).
10. References
[1] Loddo, Gianraffaele; Ludoni, Daniela; Cossu, Gian Piero; Pittaluga, Marco; Pinna, Maria Laura,
Experimenting and testing the use of electrochromic windows in the Mediterranean climate, Proceedings
of International Conference on Glass and Solar in Sustainable Development (Glass Performance Days),
Tampere, Finland, 2011.
969
doi:10.3233/978-1-61499-061-1-971
Building Integrated Photovoltaics

Barbara Siebert
Ingenieurbüro Dr. Siebert, Germany, mail@ing-siebert.de, ing-siebert.de
In terms of rising energy costs and limited fossil fuels the solar energy becomes
more and more important. In the last years a change of applications was visible:
From energy plants on fields and grasslands only for producing energy to the so
called BIPV (building integrated Photovoltaic). Here the PV acts not only for
producing energy: Other purposes are e.g. acting as building envelope and/or acting
as architectural element. In this paper the basics of PV are shortly presented. Other
important points are the aspects of safety e.g. for overhead glazing and large façade
systems with PV and aspects of static analysis. Finally some examples of BIPV are
presented.
Keywords: solar, glass structure, linear fixing, point fixture, solar clamp, BIPV,
Façade, Overhead glazing, railing
1. Introduction
Until now most applications with photovoltaic are standard applications with framed
panels on roofs or installations on fields. More and more attractive applications from the
architectural point of view are built or are under construction. The possible glass sizes
are increasing, so there are many new ways of application. The electric demands of the
electric engineers and the safety aspects of the structural engineers are worlds apart. To
design a BIPV application knowledge and background is required about many points
like the electric part, possible position of cells on the panes, properties of glass, finite
element analysis, remaining load carrying capacity, influence of temperatures, approval
of the building authorities and many others things. Focus in this paper will be mainly
the part of the structural engineer.
2. Basics of photovoltaic
Photovoltaic panels and solar cells respectively can be classified in many ways like e.g.
thickness, material or production process. A common feature of most solar panels is the
fact that the cells are placed on a surface of one glass to which a second layer (plastic,
foils, etc.) is laminated or the cells are placed between two layers of glass.
Three principle techniques are the thin-film technology (see Fig. 1) - a special coating
on the glass, monocrystalline cells with the standard dimensions of 156mm x 156mm
which are laminated between two glass panes (Fig. 2) or polycrystalline cells (Fig. 3).
The size of the cells is founded in the size of the silicium blocks, where the cells are cut
off. The efficiency of monocrystalline cells is a little bit higher than the efficiency of
polycrystalline cells. It is possible to produce the thin-film panels translucent for
applications like windows or overhead glazings.
971
Challenging Glass 3
Figure. 3: Polycrystalline
Figure 1: Thin-film coating Figure 2: Monocrystalline Cells
Cells
3. Definition Building integrated Photovoltaic

Until now photovoltaics are used mainly for energy plants on grasslands or covering
existing roof constructions like shown in Fig. 1, 2. There the inclination and direction is
optimized of course to the sun. Looking on Fig. 3, also for other orientations there is a
significant efficiency. So with the decreasing costs of the cells other applications like
vertical panels are getting more and more interesting, not only from the architectural
point of view but also from the financial point of view.
Figure 4: “Old “ Aplication Covering Figure 5: Energy plant on grass- Figure 6: Orientation and
of roof land (Picture Schott solar) efficiency
Building integrated Photovoltaics (BIPV) are photovoltaics that are used to replace
conventional building materials in parts of the building envelope such as the roof,
overhead glazing, windows, facades or railings. BIPV is realized more and more for
new projects but also for rehabilitations of buildings. Advantage is to reduce the cost by
reducing the amount of cost and building materials. Also the architectural appearance is
an important point and advantage. Disadvantage can be – at present - also the costs of
the panels if custom-made panels are necessary.
972
4. Support of glass
4.1. Linear Fixation

There are different kinds of support of the glass possible; beginning from standard
panels with linear support with aluminium profiles (Fig. 7) up to glued applications. In
all cases it is important not to shade the active area of the solar cells. There must be also
space for the cables, what is more a problem in case of the punctually fixation and not
of the linear fixation (the cables can run along profiles). In case of glued panes (often
with so called backrails) the performance and durability of the adhesive and sometimes
mechanical locking if the connection are important points. The optic appearance of the
standard framed panels like shown in Fig. 7 is considered as poor and sometimes not
acceptable to architects. Opposite to this exists some special application like linear
supported canopies (Fig.8).
Figure 7: Linear support Figure 8: linear supported canopy
4.2. Point-fixtures
Here can be made a division in point fixings in holes and point fixings situated on the
edges of the solar panels. The disadvantage of the point fittings in holes is the large
distance from the glass edge to the hole (according German regulations: minimum
80mm). So there is a loss of area for the PV-cells. Point fixings used for facades are a
challenge for calculation, production and installation. For solar panels this is even a
bigger challenge because of the solar cells in the interlayer. Tolerances – which are
always in the substructure and the glass panes - are a very important point for the design.
Here are often solutions with clamps on the edges of the solar panels better and cheaper
than solutions with point fixings in holes (Fig.9, 10). Easy installation guarantees low
cost for installation. This is valid for first installation as well as for change of eventually
broken or not working panels.
973
Challenging Glass 3
Figure 9:: solar clamp Figure10: Clamp for facades
5. Demands
For the previous applications like energy plants main aspects were the electric demands.
Because of the situation of the panels on the ground no additional safety aspects has
been considered. For the “new” applications in the building envelope aspects of safety
like static calculation of the glass and remaining load carrying capacity have to be
considered strictly.
5.1. Applications
First of all it is important to divide into the different possible applications for BIPV. Fig.
11 shows applications like overhead glazing (also accessible for work), vertical glazing
and Anti drop device like railings and simple applications like the covering of roofs.
Depending on the application different points like static calculation, remaining load
carrying capacity and the need of special permits of the building authority are important.
Overhead glazing Vertical glazing Anti drop device
Accessible glazing for work Covering of roof
Figure 11: Applications
5.2. Static Analysis

Of course a static calculation of the glass is necessary. Using minimum two layers of
glass every solar panel can be seen as laminated glass; depending on the material of the
interlayer in some cases it can be classified as laminated safety glass. So the basic
principles for calculation and design of laminated glass can be applied also for solar
panels. Depending on the kind of support, this can be very easy or very difficult. So the
static analysis of point-fixed glass is very complex, there is many experience necessary
974
to do this calculation with close to reality results. Because of the brittle behaviour of
glass this is very important, in case of too high punctually stress the glass breaks
without any yielding like in case of steel. Some impressions are shown in Fig. 12-14.
V1 -120.
L2 -480.
C1
-80.
-40.
-520.
0.
40.
-560.
80.
120.
-600.
123
123123
123
123
123123123
123
123
-640.
Z -680.
X
Figure 13: Mesh of a point fixed Figure 14: Stress peaks in

Figure 12: Punctually fixation
glass analysis
5.3. Remaining load carrying capacity

It must be sure that a glass construction cannot collapse in case of breakage of glass, so
that the safety of people, e.g. falling against a glass façade or standing under a glass roof
at the moment of breakage, is guaranteed. Depending on the kind of application the
testing of remaining load carrying capacity is done with different testing methods. In
case of glass-panes of facades a pendulum impact test like shown in Figure 15 is done.
The behaviour after breakage of a glass pane depends on many factors. The kind of
glass (thermally toughened or heat strengthened glass) or the geometry of the point
fixing have influence. The interlayer between the glass panes - especially for PV-
applications - are not only PVB-foils but also EVA foils or cast in resins. The remaining
load carrying capacity is not the same using “standard” laminated glass ore laminated
glass with photovoltaic cells. Also the temperature has an influence. So often tests in
laboratory are necessary. Figure 16 shows a Remaining load carrying test and figure 17
an Impact test at high temperatures. In addition specific topics like e.g. adhesion of
photovoltaic cell and interlayer have to be considered.
Figure 16: Remaining load

Figure 15: Pendulum impact test carrying capacity
Figure 17: Impact test at T= 70°C
975
Challenging Glass 3
6. Built Examples
6.1. Solar Facade in Würzburg

At present a photovoltaic facade in front of dormitory is under design. One façade has a
width of about 56m and a height of about 14m. The size of the glass panes is 1100 mm
x 2450mm and 1100 mm x 2800mm, the glass panes are fixed linear. To match the
German regulations, some changes were necessary; of course a static calculation was
done. (Fig. 18, 19).
6.2. Roof construction in Haar

Second example is a roof in front of an outdoor swimming area in Haar near Munich.
The fixation of the solar panels is linear but not in the classical way with aluminium
frames: clamping profiles made of stainless steel are used for fixing solar panels with
translucent thin layer technology. The size of the panels matches to the substructure. To
avoid time consuming tests and permits of the building authority, a 3-layerd glass was
chosen: So the glass pane with PV acts only as additional element and not as load
bearing element. (Fig. 20, 21).
Figure 18, 19: Solar façade in Würzburg
976
Figure 20: Solar roof in Haar Figure 21: Solar roof in Haar
6.3. Solar Decathlon, Spain

For the competition solar decathlon in Spain 2010 punctually fixed solar panels were
used for the building envelope [18]. The glass acts as structural element for façade and
roof constructions (Fig. 22, 23). A static calculation for the glass was necessary to
transfer the wind- and snow loads to the substructure.
Figure 22: Detail of facade Figure 23: Solar decathlon, Spain
6.4. Solar roof private House in Munich

For the rehabilitation of a private House in Munich a new system for the roof was
developed. Instead of roof ties and additional solar panels, the solar panels were fixed
direct to the wooden substructure with special clamps. A leak-proof covering behind the
PV-elements is not necessary. The solar panels are situated overlapped like the ties of a
classical roof. Advantage is the ventilation of the system with a better efficiency, the
reducing of costs because of only one layer of covering and a very fast installation
because of large scale elements.
977
Challenging Glass 3
Figure 24, 25, 26: mounting of

Figure 27: building site solar roof
panels, clamp and static analysis
Fig. 27 shows an impression of the building site. It is possible to change single elements
without demounting the complete construction. This is reasoned by a special
mechanism inside the clamp.
6.5. PV-Railing
In case of balustrades the wish of many architects is often a more or less translucent
appearance. A balustrade with PV-cells can fulfil these wishes and produce additional
energy (Fig. 28, 29). Important is here the demand that the balustrade has to stay in
position if a person is falling against the railing. To proof this, pendulum impact tests
are necessary (Fig. 30).
978
Figure 28: solar balustrade Figure 30 Pendulum impact test

Figure 29: solar balustrade, real project
animation of solar balustrade
7. Discussion, Conclusions and Acknowledgements

The architectural use of solar panels helps to save energy and reducing Co2 output. The
so called BIPV (building integrated Photovoltaic) becomes more and more popular,
because the production of energy in the building envelope together with an attractive
architectural appearance is possible. The applications are not limited to roof
constructions; there are applications possible also for façade, balustrades and overhead
glazing’s. For the design and construction of these applications several points have to be
considered: Not only the electric aspects and the aspects of the efficiency are important.
Because of special fixations like point fixings and special applications like overhead
glazing additional points like static calculation or remaining load carrying capacity are
very important. By these new applications the portion of renewable energy hopefully
will increase in the future.
8. References
[1] DIN 18516-4 (02-1990): Cladding for external walls, ventilated at rear; tempered safety glass;
requirements, design, testing.
[2] prEN 13474-1 2005-07 Glass in building - Determination of the strength of glass panes - Part 1: Glass
and glass products for fenestration,
[3] prEN 13474-3 2005-04 Glass in building – Design of glass panes Part 3 - Basis of Design - Design of
glass by calculation - Design of glass by testing

[4] DIN 18008: Gas im Bauwesen – Bemessungs- und Konstruktionsregeln. (Glass for building design and
construction rules). Deutsches Institut für Normung e.V. Germany 2006, Teil 1: Begriffe und
allgemeine Grundlagen (part 1: Basics), Teil 2: Linienförmig gelagerte Verglasungen (part 2: linear
supported glazing), revised draft expected for 2008, part 3-6 only internal working draft exist
[5] ISAAG 2004, 2006, 2008: International Symposium on the application of architectural glass
(www.isaag.com), Munich
[6] Siebert, G.: Entwurf und Bemessung von tragenden Bauteilen aus Glas, Ernst und Sohn Verlag, Berlin
2001, new version 2012
[7] Siebert, G., Siebert B. Design of safe glass structures: interaction between glass product, application,
calculation and design, GPD India Glass processing days, Delhi 2008
[8] Siebert, G., Systems for fixing of solar panels, IABSE Bangkok 2009
[9] Maniatis, I., Innovatives Bauen mit PV-Elementen, OTTI 2009
[10] Maniatis, I., Siebert, G, Systems for fixing of solar panels, GPD Glass Performance days, 2009
Tampere
[11] Siebert, B, Siebert, G, Modern façade and roof constructions with photovoltaic panels, Challenging
glass 2, 2010 Delft
[12] www.pauli.de
[13] www.schottsolar.com
[14] www.mgt-esys.at
[15] Hochschule Stuttgart (solar decathlon)
979
doi:10.3233/978-1-61499-061-1-981
Design Methods and Structural

Components of Blast Enhanced Façades
Frank Wellershoff
HafenCity Universität Hamburg, Façade Systems and Building Envelopes,
frank.wellershoff@hcu-hamburg.de
A “blast resistant” façade requires different numerical design methods and different
structural components than a “normal” façade. The impulsive load requires a
dynamic analysis with consideration of mass and inertia. To withstand an extreme
blast load a façade should be flexible and able to absorb energy in properly
designed and analyzed crash zones. These zones are the laminated glass and
additional elasto-plastic connections along the load path to the sub structure.
Keywords: Blast, load, façade, design, element, connection, cable, curtain wall
1. Introduction
In the last decade the topic “civil safety” was becoming a key objective in our
multicultural world. All efforts to decrease the risks of assassinations by higher control
of passengers in the public transportation are limited to a certain extent and are
counteracting our request for fast und unrestricted mobility. The protection of buildings
and especially facades against bomb blast attacks became therefore a more relevant
topic in all design stages of potential targets itself and the surrounding buildings.
The dynamic behavior of facades under shock waves with short time durations but high
impulse loads requires different design strategies, specialized analysis tools, and
specialized connections between the façade elements itself as well as the connections to
the primary building structure. The design philosophy is primarily to save lives and
prevent injuries, and secondarily to protect buildings, functions, and assets. The design
criteria take a balanced approach to safety, considering cost effectiveness and
acknowledging acceptance to some risks.
For cable net facades as lightweight and transparent structures particularly at the
podium or entry levels of buildings the connections from the glass panels to their
fittings, from the fittings to the cables, and the cable end connections must be well
designed and in most cases, in order to minimize their size, are designed to the limits.
For curtain walls facades the load path from the glazing to the mullions and transoms
into the brackets has to be blast enhanced designed and calculated. For both façade
types – cable facades and curtain walls – new energy dissipative connectors with
residual strength capacity were recently developed as well as design rules and new
appropriated software packages. The intensive research was initiated, financed, and
operated by Gartner Steel and Glass and the Permasteelisa R&D Group.
981
Challenging Glass 3
2. Explosion Loads
2.1. Analytical Determination of explosion Loads

As a result of the detonation of an explosive charge a pressure wave spreads initially
spherical in all directions, until it is reflected by surfaces (building, floor). In the
explosion, a very large amount of energy is released within a few nano seconds. The
pressure increase is in a time range of nanoseconds and the length of the overpressure
phase in the rage of mille seconds. The short period of overpressure is characterized by
the peak overpressure p̂10 and by the time t d . The integration of pressure over time
results in the specific impulse I. The negative pressure phase is longer than the
overpressure phase, and the magnitude of the negative pressure is usually much lower
than the magnitude of the overpressure. The barometric pressure is under normal
conditions 101,3 kPa (1,013 bar) at sea level.
The peak reflected overpressure pˆ r 0 is formed by the reflection of the incident plane
shock wave which encounters a structural system in a certain angle. The ratio of the
peak reflected overpressure and incident peak overpressure is called the reflection factor.
The reflection factor depends on the incident peak overpressure, of the angle between
the shock front and the surface and the density and the rigidity of the material of the
reflection surface. In detail, the main effects and influencing factors for the
determination of reflection coefficients are shown in [1]. With known reflection factor,
the reflected pressure time history can be derived, which has a similar time history as
the incoming pressure if interaction effects are neglected. Figure 1 shows a typical
reflected pressure-time history of an explosion in air.
Figure 1. Schematic diagram of pressure time history of an explosion in air
The determination of the complete reflected pressure time history is essential for the
structural analysis, because this affects the structure and its components. Using only the
reflected peak pressure is insufficient. The specification of quasi static loads is not
possible without the consideration of the system-dependent reaction. For explosion
loads, a (usually nonlinear) dynamic calculation is thus necessary in any case. Above all
the effective mass of explosive material, its height above the ground and the distance to
982
Design Methods and Structural Components of Blast Enhanced Façades
the building affects the reflected pressure time history of an explosion. The mass of the
explosive material is usually defined as the TNT equivalent mass (TNT= trinitrotoluene,
commonly used military explosive). Other parameters are possible obstacles, such as
protective walls or upstream buildings, as well as the type and geometry of the building
itself. For light and flexible membrane structures and large-area cable net facades, the
reflected pressure time history is affected additionally by the fluid-structure interaction
and the aerodynamic damping [2].
2.2. Determination of explosion Loads with Hydro Codes

The analytical approaches discussed in chapter 2.1 are applicable only for simple cases.
Loads that differ significantly from the analytically determined pressure time curves can
result through reflections on neighboring buildings. This is the case, for example, if
explosions in adjacent streets or the influence of upstream buildings or walls are taken
into account. In these cases, the pressure time history can be derived from either with
the help of (possibly scaled) experiments or financially cheaper by numerical studies
with so-called hydro codes. Hydro codes such as Ansys Autodyn or LS-Dyna, are based
on the physical principles of mass, momentum, and energy conservation, and calculate
the explosion and the subsequent propagation and reflection of the shock wave.
2.3. Normalized explosion Load Assumptions

Because it is very complex to determine the complete pressure time history due to a
detonation and all reflection effects, standardized explosion load assumptions were set
out first in the United States and then internationally (ISO) [3, 4, 5]. These explosion
load assumptions provide a linear sloping triangular history for the reflected pressure.
The reflected pressure is characterized by the reflected peak overpressure p ˆ r 0 and by
the positive pulse i r . The duration of the overpressure phase in this linear approach is
defined by:
2ir
t d ,lin (1)
pˆ r 0
The influence of the negative phase is neglected in these standardized approaches. This
is justified in the dynamic calculation of rigid or heavy structures (e.g. reinforced
concrete structures), because the negative phase hardly affects the structural response in
these cases. On the other hand the negative phase can affect significantly the structural
response of lighter and more flexible systems with lower natural frequencies [2, 6].
Despite this influence, which is present in cable net facades, only standardized
explosion load scenarios in accordance with US or ISO standard are specified in most
cases. It is assumed that the failure of the façade to the internal side is the critical design
intent. Therefore the impact on people in the Interior of the building should be
minimized. A failure of the system to the outside due to the negative phase is accepted.
In table 1 and 2 the essential design loads are grouped according to the US GSA/ISC
standard and according to the international ISO standard. The specified quantities of
explosives (TNT equivalent mass) and stand-off specify which explosives would create
these loads in a ground detonation in front of a large façade. In an experimental attempt
to detect the façade performance not the whole facade, but a mockup facade of 3.05 m x
983
Challenging Glass 3
3.05 m (GSA/ISC) and 3.15 m x 3.15 m (ISO) is usually tested. Due to clearing effects
at relatively small mockups the mass of the explosive material and the stand-off must be
adapted to create the same pressure and impulse values as for a large façade. The
denotations of the ISO scenarios (e.g. EXV 45) consider this phenomenon. EXV xx
means that the specified peak overpressure and impulse values are achieved with a
typical test mockup of 3.15 m x 3.15 m with an explosion of 100 kg TNT at a distance
of xx m (see table 2). This corresponds to an equivalent explosion scenario in front of a
large facade (see table 2 and Annex C. 1 ISO 16933) [4]).
Table 1. Explosion scenarios of the US General Services Administration (GSA/ISC) [3]
scenario
pˆ r 0 ir td ,lin mass TNT stand-off
[kPa ms] [kg] [m]

[kPa] [ms]
GSA C 27,58 193,06 14,0 47,5 30
GSA D 68,95 675,71 19,6 340 34
Table 2. Explosion scenarios (vehicles bombs) of the ISO 16933, Annex C1 [4])
Reflected Length of Stand-off 100 kg Equivalent

Peak
Impulse overpressur TNT in front of explosion scenario
reflected
e phase small mockup in front of large
overpressure
(linear) (3,15m x 3,15m) facade
Class
pˆ r 0 ir td ,lin Stand-off TNT
[Pa s] [m] [kg]

[kPa] [ms]
EXV 45 30 180 12 45 30
EXV 33 50 250 10 33 30
EXV 25 80 380 9,5 25 40
EXV 19 140 600 8,6 19 64
EXV 15 250 850 6,8 15 80
EXV 12 450 1200 5,3 12 100

EXV 10 800 1600 5,0 10 125
3. Safety Requirements for explosive Loads

Even slight pressure wave generated by small bombs or large bombs at a great distance
(e.g. in a neighboring street), can lead to major damage if the facade is not properly
constructed. To protect persons behind the facade from major injuries, an explosion-
resistant function of the facade is frequently specified. Most specifications refer to a
classification of the performance condition according to the US GSA standard [7]. The
GSA method classifies facades into six protection and risk classes (protection and
hazard levels, Figure 2). For the highest protection class 1 the glass must not break. In
the other protection classes it is defined how far glass splinters are allowed to fly into a
standardized test box. Most commonly the protection class 3B is specified in which the
splitters may fly in maximum 10 ft. (3.05 m) into the test box.
984
The GSA protection classes are developed for window glazing spanning in maximum
from floor to ceiling. The direct use of the GSA protection classes would therefore be
inappropriate for multi-story high cable net facades. The key factors for the protection
of individuals are: firstly, that the pressure wave is considerably damped to protect ears
and lungs and, secondly, that glass splinters do not act as projectiles that cause heavy
injuries. Therefore, a definition of the permitted flight distance of glass splinters in
relation to the height of the glazing would be more logical and could be used for all
types of facades. For numerical evidence, the speed of the glass splinters at the moment
when the breaking strength of the glass is reached could be defined as a protection class.
For this method it needs to be considered that the fracture stress significantly depends
on the glass product and the load duration due to the effects of the surface pre stress and
the crack growth of flaws in the glass surface. For very short impact loads, such as
under explosions, much higher breakage strength is known compared to the breakage
strength for wind loads.
Figure 2.GSA/ISC performance conditions for window glazing according [7]
Table 3: GSA/ICE performance conditions for window system response

Performance Protection Hazard Description of Window Glazing Response
Level Level
Condition
1 Safe None Glazing does not break. No visible damage to glazing or frame.
2 Very High None Glazing cracks but is retained by the frame. Dusting or very
small fragments near sill or on floor acceptable.
3a High Very Low Glazing cracks. Fragments enter space and land on floor no
further than 3.3 ft. from the window.
3b High Low Glazing cracks. Fragments enter space and land on floor no
further than 10 ft. from the window.
4 Medium Medium Glazing cracks. Fragments enter space and land on floor and
impact a vertical witness panel at a distance of no more than 10
ft. from the window at a height no greater than 2 ft. above the
floor.
5 Low High Glazing cracks and window system fails catastrophically.
Fragments enter space impacting a vertical witness panel at a
distance of no more than 10 ft. from the window at a height
greater than 2 ft. above the floor.
985
Challenging Glass 3
4. Cable Net Facades without Blast Load Enhancement
4.1. Cable Net Façade Types

Cable nets façades can generally be divided into:
x Type 1: Systems with straight directions of the cables - as an example, see

Figure 3.
x Type 2: Systems with polygonal directions of the cables - as an example, see
Figure 4.
Figure 3.Cable net facade with straight cable Figure 4.Cable net facade with polygonal cable direction
direction (type 1): Hamad Medical City, Doha, (type 2): Sony Center, Berlin (Gartner)
Qatar (Gartner)
For type 1 is to bear in mind that a straight cable needs to be deformed into a significant
curvature before it can carry loads orthogonal to the cable line. The required curvature
can be reduced by a pre stress in the cable. Additional cables (length l2) with an angle to
the primary cables (length l1) could be used to stabilize the primary cables as well as to
carry unevenly distributed loads. Unless the lengths in both directions do not
significantly differ (approx.l2/l1 < 1.5), both directions can be considered for evenly
distributed loads. However, the deformations of straight cables are significantly larger
than deformations of elements with bending resistance.
986
The limit of the allowable deformation of a facade perpendicular to its surface depends
on different factors:
x Obstacles in front of and behind the facade (e.g. building columns behind the
facade, see Figure 3)
x Maximum permissible tension and compression of the glass joints
x Required glass support
x Permissible warping of the glazing in the transition areas to rigid connectors
(edge of the facade, door frame). Forced twists may constitute a crucial
boundary condition, in particular for insulating glass, which allow only a
limited deformation of the spacer.
With suitably careful planning, deformations are possible up to 1/40 of the facade span.
For type 2 the positive and the negative wind loads are carried by two separated cables
which are already shaped into a corresponding polygonal line. The pre stress in the
cables can be much lower than in type 1 facades. To force a pre stressed cable into a
polygonal line additional orthogonal forces, acting at the nodes of the polygon, are
required. Compression struts between the two cables are usually used for this purpose,
causing the typical fish-shaped cable networks.
4.2. Cable-End-Connectors
In addition to the allowable deformation of the facade, also the sustainability and the
stiffness of the sub-structure can be a determining factor for the type of cable network.
Often, the necessary high pre stressing forces for straight cables cannot be anchored.
Balance springs at the cable end connections can be used to control the cable pre stress
into a low range even under temperature changes and deformations of the sub-structure.
4.3. Glass Fitting

Spiders with drilled holes in the glazing (figure 4, 5) are the so far most frequently used
form of glass fittings in cable facades. In the last years there was a clear trend to
clamped connections (figure 3, 6, 8). Holes in the glazing are not required.
987
Challenging Glass 3
Figure 5. Spider connection, Court Square Project, Figure 6.Clamped connection,

Long Iceland, United States (Gartner) Chicago Airport (Gartner)
5. Cable Net Facades with Blast Load Enhancement
5.1. Facades with maximum Flexibility

Since the attacks on the World Trade Center Towers in New York 2001, more
explosion-resistant cable net facades are built. In particular cable net facades with
straight cable directions can be designed with blast enhancement. By the softness of this
type of façade a pressure wave can be better absorbed and a higher aerodynamic
damping is possible. For smaller explosion loads a sufficient protection can be achieved
already, if the high deformations are considered in all details of the facade. Figure 7
shows an example of a facade that is designed in this type.
988
Figure 7.Explosion resistant cable net façade, Court square project, long Iceland, United States (Gartner)
5.2. Facades with Connectors for Blast Enhancement

To reduce the stresses and forces in the components or to increase the residual strength
of the system, special connectors are required [8]. Figure 8 shows an example of this
type of facade.
Figure 8. Explosion resistant entrance facade of World Trade Center Tower 7, New York (Gartner)
While a pressure wave acts on a facade, two main phases can be defined, in which
various connectors are acting with their function of blast enhancement. A definition of
phases is shown in Figure 9 [9].
In phase 1, the pressure wave hits the glazing of the façade and the load is transferred by
the glass clamp-cable connectors (see chapter 5.3). Already, through the use of
plasticizers in these connectors, explosive energy can be absorbed at this stage. This
results into lower surface tensile stresses in the glass and lower cable forces as with
rigid couplings between the glass and the cables.
989
Challenging Glass 3
In phase 2, the stresses in the glass, as well as the cable forces rise. In case the glass
breaks, the glass clamp connector described in chapter 5.4 provides residual strength
and positioning. The cable end connectors described in chapter 5.5 can contribute to
targeted energy dissipation at this stage and limit the maximum anchor forces at the
main bearing structure.
Activation of Glass Clamp Activation of Cable End Connector

– Cable Connector and if necessary of
Glass Clamp Connector
Figure 9.Load and response phases in explosion protected cable net facades. Gartner / Permasteelisa
5.3. Glass Clamp - Cable Connector

The connection between the glass clamp and the cable shown in Figure 10, allows a
significant reduction of glass stress as well as the cable force by converting kinetic
energy into deformation energy [9]. So it prevents already in the phase 1 either, that the
glass break, or that in the case of glass breakage the splinters detach with to high speed
from the laminated foil. A crash absorber, e.g. with aluminum foam, which is positioned
between the glass clamp and the cable clamp will be deformed. To transfer the dead
weight of the glass also after triggering the central bolt is decoupled from the crash
absorber.
990
Figure 10.Patented glass clamp-cable connector before and after plastic deformation [10]
1.1 Glass Clamp Connector

The glass clamp connector shown in the Figure 11 allows a significant larger residual
strength of broken laminated glass panes [9]. Much higher loads can be carried before
whole laminated glass panes break out, however, so even higher loads affect the cables.
In this connection a perforated plate of carbon is embedded into the layers of the
laminated glass. The carbon plate is connected by Kevlar twines with a cone on the
outside of the pane. The cone is free of any forces under regular loads and attaches the
clamp only in the case of glass breakage and a larger deformation of the glass pane.
Figure 11.Patented glass clamp [11]
5.4. Cable End Connector

Cables with a straight direction require a high axial pre stress to limit the deformation
under wind loads to an acceptable level. There is therefore the risk that the cables
exceed their breakage strength and tear under explosion loads. A newly developed cable
end connector is schematically represented in figure 12. This connector is very stiff until
a defined activation force FA to ensure a minimal deformation of the cable under the
regular load combinations with pre stress, self-weight, temperature, and wind. Under a
blast load a triggering device allows a controlled elongation of the cable end connector
with simultaneous energy dissipation. In the version shown in the figure 12 the
activation force is determined by a breaking point in the primary load path. After
triggering the axial tensile force is redirected to a secondary load path, in which one or
more crash absorber, e.g. with aluminum foam, are integrated.
991
Challenging Glass 3
Figure 12. Patented cable-end-connector before and after fuse breakage [12]
In the crash absorber a yield force Fc ("crash force") is activated that is smaller than the
activation force FA. Crash materials show a hardening under increasing crash
deformations. The challenge is to find the best force displacement function for the
explicit facade and to realize this function with the belonging crash absorber (defined by
the crash material, cross section and initial crash absorber length). A too slow hardening
of the crash material requires a too long deformation length until the required amount of
energy is dissipated (defined by the integral of the crash force over the crash
deformation). A too fast hardening limits the possible dissipation energy before the
cable strength is reached.
Two significant effects are achieved by the new cable end connector:
x Significant energy dissipation due to high plastic deformations.
x Reduction and control of cable forces and thus reduction and control of the
forces which are forwarded to the frame construction.
In the development of appropriate cable end connectors the following aspects were
considered:
x a low increase of the trigger load under high strain rates as they occur under
blast loads
x a small scattering of the activation force FA,

x a small scattering of crash force Fc, and
x a choice of crash materials with different hardening behavior
The newly developed cable end connectors were tested quasi-static, as well as with
realistic deformation speeds as they occur under explosion loads (1 m/s up to 5 m/s).
5.5. Development of an Assessment Tool

The structural behavior of explosion-resistant cable net facades is extremely complex.
This is particularly true if the non-linear behavior of dissipative cable end connectors
should be considered. In a joint project of the Department of structural analysis of the
University of Armed Forces Munich and Gartner Steel and Glass GmbH in Würzburg,
the dependencies of the different parameters were examined among themselves with a
standardized load of explosion [13]. The results were implemented in a design tool,
developed with Microsoft Excel, which is used for preliminary design and parametric
studies [14].
992
As the load scenario the GSA-D level (table 1) was chosen. In the United States this is a
common level for buildings with higher security requirements.The examined parameter
combinations are summarized in table 3. Five idealized force displacement curves for
the cable-end-connectors (SEV types) were defined (figure 13). The ratio of activation
force to crash force is 60% for all cable-end-connectors (FA/ F(c) = 0.6). This ensures
that under consideration of the scatter of both forces the activation force is always
higher than the crash force.
The five SEV types differ in the plastic hardening area (Figure 13). For SEV 1 it is
twice as large as for SEV 2 and SEV 3 has an ideal plastic plateau without hardening.
SEV 1, 2a, and 3a are unlimited in their plastic deformation. For SEV 2b and 3b the
plastic deformation is limited to 200 mm. After this deformation they harden again to a
linear elastic behavior with high rigidity. This behavior considers that the length of the
crash absorber and therefore the energy that could be absorbed by plastic deformation is
limited. The variations of SEV 1, 2a and 3a are suitable to determine the required plastic
deformation length and to design related crash absorbers. The assessment tool can be
used to design cable-end-connectors with "tailor-made" capacity of energy absorption.
Figure 13. Idealized force-deflection curves of five cable end connectors (SEV) [13]
Figure 14. Screenshot of the developed design tool with Excel [14]
993
Challenging Glass 3
6. Curtain Walls with Blast Load Enhancement

In the last years Permasteelisa has developed its own software for the design of linear
supported glazing under blast wave loads. This software simulates blast waves and the
dynamic reaction of the façade [8].
The software makes it possible to perform two different types of analyses: single
analysis and isodamage curve analysis. With the single analysis type, verification is
performed on a façade panel governed by a blast law that simulates the wave derived
from the explosion. Analysis of the isodamage curves, which can only be performed on
façade panel with rigid supports, makes it possible to launch iterative (and automatic)
single analyses. The final purpose of such analyses is to find pairs of points on the
diagram (maximum pressure-impulse) that determine an equivalent condition ¾ in
terms of displacement and failure ¾ over the dynamic behaviour of the panel.
Figure 15 shows the key steps involved in writing and solving a dynamic equilibrium
equation for a system with a single degree of freedom. The concentrated parameters
must be provided according to the characteristics of the system (in this case the façade
panel) and, together with definition of the forces, they determine the unambiguity of the
solution to the dynamic problem. This solution is evaluated through the choice of a
specific stepped integration method for which the user can modify the parameters in
order to counteract problems related to divergence of the solution. Once the outputs
have been obtained, they are studied to ensure their conformity with the set objectives.
•a) System inputs •b) Load inputs

•c) Integration
method
m u c u k u F (t )
•d) Outputs
Figure 15: Structure of a dynamic equilibrium equation for a system with one dof [15]
994
One mayor aspect of the software is the realistic consideration of the resistant function
of laminated glass, as shown in figure 16. Here four different components can be
identified:
x 1) During the first phase, the resistance function shows a nearly linear trend,
guided by the fragile performance of the glass
x 2) Then, with the failure of the first slab of glass, the resistance collapses
suddenly and the second phase takes over in which the second slab of glass
effectively determines the overall resistance
x 3-e) When the second slab of glass breaks, the “package” deformation of the
layers of PVB ensues. First of all, the resistance offered acts within the elastic
field with a nearly linear branch
x 3-p) Finally, once the yield strain has been reached, we enter the field of large-
scale membrane deformation, i.e. the area where most of the blast wave energy
must be dissipated.
Figure 16: Progressive phases of the behaviour of laminated glass [15]
The further development of the software allows now the analysis of complex façade
systems with double or triple glazing (including laminated glass), plastic hinges in
mullions and transoms, and elasto-plastic deformations of the brackets. Figure 17 shows
exemplary brackets that can be trigger to the appropriate force-deflection function for a
specific blast load scenario.
Figure 17: Patented elasto-plastic bracket plates for curtain wall systems [16]
995
Challenging Glass 3
7. Acknowledgements
This paper gives a short overview of design methods and structural components that
were developed by Gartner Steel and Glass and the Permasteelisa R&D Group in the
last years – special thanks Marc Zobec and Guido Lori. Financed by Gartner /
Permasteelisa several tests and calculations were done at external institutes - special
thanks to BAM in Berlin and the Universities of Applied Science and Armed Forces in
Munich.
8. References
[1] Gebbeken, N. und T. Döge: Der Reflexionsfaktor bei der senkrechten Reflexion von Luftstoßwellen an
starren und an nachgiebigen Materialien. Bauingenieur, 81(11):496–503, 2006
[2] Teich, M. und Gebbeken, N., The Influence of the Underpressure Phase on the Dynamic Response of
Structures Subjected to Blast Loads, International Journal of Protective Structures, 1(2):219–233, 2010
[3] US General Services Administration: Security Design Criteria. General Services Administration (GSA),
Interagency Security Committee (ISC) (for official use only). 2001
[4] International Organization for Standardization: ISO 16933:2007: Glass in building – Explosion-resistant
security glazing – Test and classification for arena air-blast loading, 2007
[5] Johnson, N. F.: International Standards for Blast Resistant Glazing, Journal of ASTM International,
3(4):1–16, April 2006
[6] Krauthammer, T. und A. Altenberg: Negative phase blast effects on glass panels. Intenational Journal of
Impact Engineering, 24(1):1–17, Januar 2000
[7] US General Services Administration: GSA-TS01-2003: Standard Test Method for Glazing and Window
Systems, Subject to Dynamic Overpressure Loadings, 2003
[8] Wellershoff, F.: Blast enhanced facades for the new World Trade Centre Towers NY, Challenging Glass
Conference Proceedings, page 643 – 653, 2008
[9] Wellershoff, F.: Blast enhanced cable facades, Proceedings of Glass Processing Days 2011, page 525 –
529
[10] Gartner Steel and Glass GmbH: Haltevorrichtung für Flächenelemente, Patenting documents
[11] Gartner Steel and Glass GmbH: Composite pane and retaining structure with at least one composite
pane of this type, Patenting documents
[12] Gartner Steel and Glass GmbH: Cable End Anchorage with Overload Protection, Patenting documents
[13] Wellershoff, F., Teich, M.; Nehring, G.: Gebbeken, N.: Konstruktion und Berechnung von
explosionshemmenden Seilnetzfassaden, Stahlbau 81 (2012), Heft 1, Seite 13-25
[14] Nehring, G.: Zur Bemessung explosionssicherer Seilnetzfassaden, Diplomarbeit, Universität der
Bundeswehr München, Lehrstuhl für Baustatik, 2010
[15] Lori, G., Zobec, M., Franceschet, A., Manara; G.: The Behaviour of Facades due to Blast Loads ̽ A

Single Degree of Freedom Performance Evaluation Approach, Presentation at Glass Processing Days
2009
[16] Josef Gartner GmbH, Befestigungsanker und Verfahren zum Befestigen einer Fassade, Patenting
documents
996
Challenging Glass 3
Indices
997
Author Index
Surname Name Affiliation Page
A
Amadio Claudio University of Trieste, Italy 373
University of Ljubljana, Faculty of Civil and Geodetic
Antolinc David 623
Engineering, Slovenia
Tampere University of Technology, Department of Energy and
Aronen Antti 363
Process Engineering, Finland
B
387
Bao Minxi University of Birmingham, UK
917
State Key Laboratory of Green Building Materials, China 387
Bao Yiwang
Building Materials Academy, Beijing, China 917
Baldassini Niccolo RFR, Paris, France 93
Baumgärtner Thomas Glas Trösch, Switzerland 735
Bedon Chiara University of Trieste, Italy 373
177
397
Belis Jan Ghent University, LMO, Belgium
587
691
Université Européenne de Bretagne, Institut National des
Bernard Fabrice 405
Sciences Appliquées de Rennes, France
Bergers Matt Viracon Inc., USA 785
Besserud Keith SOM, USA 785
Black Andrew J. Permasteelisa, USA 785
Blandini Lucio Werner Sobek Stuttgart, Germany 41
Delft University of Technology, Faculty of Architecture, the
Bokel Regina 139
Netherlands
47
Bos Freek Witteveen + Bos consulting engineers, the Netherlands

691
Bostick Charles Seele Sedak GmbH&Co.KG., USA 57
Branco Fernando Instituto Superior Técnico/ICIST, Portugal 745
C
Callewaert Dieter Ghent University, LMO, Belgium 177
Camposinho Rui School of Engineering of Polytechnic of Porto, Portugal 281
Lawrence
Carbary Dow Corning, USA 785
Donald
Carpenter James F. James Carpenter Design Associates, USA 3
Carvalho Paulo School of Architecture, University of Minho, Portugal 187
University of Ljubljana, Faculty of Civil and Geodetic
epon Franci 623
Engineering, Slovenia
Correia João R. Instituto Superior Técnico/ICIST, Portugal 745
Civil and Environment Engineering and Architecture
Cossu Gian Piero 957
Department, Cagliari University, Italy
Cruz Paulo J.S. School of Architecture, University of Minho, Portugal 187
Csoke Csilla DGMR Bouw BV, the Netherlands 925
999
Challenging Glass 3
D
Delgado Raimundo Faculty of Engineering of the University of Porto, Portugal 281
Demarteau Rens Oskomera Group BV, the Netherlands 881
RWTH Aachen University, Institute of Steel Construction,
Di Biase Pietro 837
Germany
ArcelorMittal Chair of Steel and Façade Engineering, University
Dias Vincent 195
of Luxembourg
Dieterich Oliver A. Hagl Ingenieurgesellschaft GmbH, Germany 235
Dietrich Nora University of Applied Sciences Dresden, Germany 509
Dispersyn Jonas Ghent University, LMO, Belgium 177
Dodd Graham Arup, UK 633
E
Eckhardt Christian Evonik Industries AG, Germany 885
Egeraat, van Erick designed by Erick van Egeraat, the Netherlands 7
CTU in Prague, Faculty of Civil Engineering, Department of 267
Eliášova Martina
Steel and Timber Structures, Czech Republic 715
819
Elstner Michael Interpane, Germany
857
Institut für Baukonstruktion, Technische Universität Dresden,
Engelmann Michael 865
Germany
Fraunhofer ISE, Freiburg, Germany / Delft University of
Erban Christof 933
Technology, Faculty of Architecture, the Netherlands
F
Fahsi Bouazza Université de Sidi Bel Abbes, Algeria 405
Feijen Mark i-Saac, the Netherlands 829
Feldmann Markus 837
Germant
Department of Civil-Environmental Engineering and
Ferretti Daniele 641
Architecture, University of Parma, Italy
ITKE - Institute of Building Structures and Structural Design,
Fildhuth Thiemo 845

Fredin Johan Blekinge Institute of Technology, Sweden 431
Freytag Bernhard Graz University of Technology, Austria 653
Froli Maurizio University of Pisa, Italy 669
Fröling Maria Lund University, Sweden 209
G
Department of Civil-Environmental Engineering and
Galuppi Laura 419
Architecture, University of Parma, Italy
Georgiev Nikolay University of Ruse, Bulgaria 679
Götschl Franz Independent, Austria 653
H
Haese Andreas Ingenieurbüro Dr. Siebert, Germany 725
221
A. Hagl Ingenieurgesellschaft GmbH / Test-Ing+ Material
Hagl Anneliese 235
GmbH, Germany
249
Hebly Esther Oskomera Group BV, the Netherlands 881
1000
Author Index

ArcelorMittal Chair of Steel and Façade Engineering, University
Hechler Oliver 195
of Luxembourg
Heide, van der Rogier Philips Lighting, the Netherlands 9
Heijden, van
Tom Scheuten Absoluut Glastechniek, the Netherlands 47
der
Henriksen Thomas Waagner Biro, Austria 67
Herrmann Tobias Ingenieurbüro Dr. Siebert, Germany 335
Heyder Frank hpl-Ingenieure, Germany 75
Bauhaus-Universität Weimar, Fakultät Bauingenieurwesen,
Hildebrand Jörg 565
Institut für Konstruktiven Ingenieurbau, Germany
Hoffmann Bernd Hoffmann GTD Glas-Technik-Design, Germany 943
I
Ivanov Ivelin University of Ruse, Bulgaria 679
Ivanov Ivo University of Ruse, Bulgaria 679
J
Jansson Marita Glafo – the Glass Research Institute, Sweden 579
431
Jönsson Anders Blekinge Institute of Technology, Sweden
579
K
Delft University of Technology, Faculty of Architecture, the 441
Kamerling Wim
Netherlands 455
Tampere University of Technology, Department of Energy and
Karvinen Reijo 363
Process Engineering, Finland
Kasper Ruth 837
Germany
Kassnel-
Bruno Glas Trösch, Switzerland 151
Henneberg
Kleuderlein Jonas Technische Universität Darmstadt, Germany 553
83
Knippers Jan University of Stuttgart / Knippers Helbig Advanced Engineering,
845
Germany
Testing Centre «Samarastroyispytanya» of Samara State
Kondratieva Nadezhda 607
University of Architecture and Civil Engineering, Russia
Kosic Tatjana Faculty of Architecture, University of Belgrade, Serbia 891
Koudijs Johan DGMR Bouw BV, the Netherlands 925
Institut für Baukonstruktion, Technische Universität Dresden, 769
Krampe Philipp
Germany 865
819
Kramer Markus IB Kramer Tragwerksplanung, Germany
857
Krour Baghdad Université de Sidi Bel Abbes, Algeria 405
Krstic-
Aleksandra Faculty of Architecture, University of Belgrade, Serbia 891
Furundzic
259
Kruijs Ron Glasimpex Schiedam, the Netherlands
467
Kuntsche Johannes Technische Universität Darmstadt, Germany 553
L
Lambert Harriet Eckersley O'Callaghan, London, UK 101
Lancaster Franklin Eckersley O'Callaghan, London, UK 475
1001
Challenging Glass 3

Lang Maria Glafo – the Glass Research Institute, Sweden 579
Lani Leonardo University of Pisa, Italy 669
Larsson Oskar Lund University, Sweden 209
Leduc Nicolas RFR, Paris, France 93
101
Lenk Peter Eckersley O'Callaghan, London, UK
475
Lindahl Pia Blekinge Institute of Technology, Sweden 431
Lippert Sebastian 845
State Key Laboratory of Green Building Materials, China Building 387
Liu Xiaogen
Materials Academy, Beijing, China 917
Gianraffa Civil and Environment Engineering and Architecture Department,
Loddo 957
ele Cagliari University, Italy
587
Louter Christian EPFL, Lausanne, Switzerland
691
Civil and Environment Engineering and Architecture Department,
Ludoni Daniela 957
Cagliari University, Italy
M
Macfarlane Tim Glass Light and Special Structures Ltd., UK 11
CTU in Prague, Faculty of Civil Engineering, Department of Steel
Machalická Klára 267
and Timber Structures, Czech Republic
Martins Luís Faculty of Engineering of the University of Porto, Portugal 281
Mazurek Arkadiusz SOM, USA 785
Messere Rino Glassolutions, the Netherlands 881
Michel Matthias imagine structure GmbH, Frankfurt am Main, Germany 907
Misson Donald Permasteelisa, USA 785
Miyasato Naoya Nihon University, Tokyo, Japan 703
Mocibob Danijel RI ISA (Permasteelisa Group), Croatia 397
Müller de Vries Christoph Professor Pfeifer und Partner, Germany 489
N
CTU in Prague, Faculty of Civil Engineering, Department of Steel
Netusil Michal 715
and Timber Structures, Czech Republic
University of Applied Sciences FH-Joanneum Institute for 397

Neugebauer Jürgen
architecture and civil engineering / SFL Technologies, Austria 901
Nhamoinesu Shelton University of Cambridge, UK 293
Nijsse Rob Technical University of Delft, the Netherlands 111
Nomoto Keisuke AXS Satow Inc, Tokyo, Japan 703
O
O'Callaghan James Eckersley O'Callaghan, London, UK 57
ArcelorMittal Chair of Steel and Façade Engineering, University of
Odenbreit Christoph 195
Luxembourg
Oikonomopoul Delft University of Technology, Faculty of Architecture, the
Phaedra 139
ou Nehterlands
Okada Akira Nihon University, Tokyo, Japan 703
Oppe Matthias Knippers Helbig Advanced Engineering, Germany 83
Out Bas Delft University of Technology, Adhesion Institute, the Netherlands 177
293
Overend Mauro University of Cambridge, UK
397
1002
Author Index
P
Paulu Franziska hpl-Ingenieure, Germany 75
209
Persson Kent Lund University, Sweden
579
Pfefferkorn Stephan University of Applied Sciences Dresden, Germany 509
Pistora Arthur Ingenieurbüro Dr. Siebert, Germany 501
Civil and Environment Engineering and Architecture Department,
Pittaluga Marco 957
Cagliari University, Italy
Polivyany Vladimir Konstantinovsky glasswork “Spectechglass”, Ukraine 535
Werner Sobek Stuttgart, Germany / IIT (Illinois Institute of
Puller Kerstin 307
Technology), Chicago, USA
Prandelli Silvia Buro Happold Ltd, UK 121
Q
State Key Laboratory of Green Building Materials, China Building
Qiu Yan 917
Materials Academy, Beijing, China
R
Raji Vlatka University of Zagreb, Faculty of Civil Engineering, Croatia 623
HS-OWL, Detmold School of Architecture, Germany / Inhabit
Rammig Lisa 315
Europe Ltd., London, UK
Raynaud Jacques RFR, Paris, France 93
Reed Leroy Kite Glass, UK 633
Institut für Baukonstruktion, Technische Universität Dresden, 509
Reich Stefan
Germany 865
Retsch Stefanie Independent, Germany 769
Riederer Jochen Knippers Helbig Advanced Engineering, Germany 83
349
523
Rodichev Yurii G.S. Pisarenko Institute for Problems of Strength, Ukraine
535
597
Rogan Damian Buro Happold Ltd, UK 121
Roman Paul Glassolutions, the Netherlands 881

Rossi Marco Glass Light and Special Structures Ltd., UK 641
Department of Civil-Environmental Engineering and Architecture, 419
Royer-Carfagni Gianni
University of Parma, Italy 641
Rubis Kenneth Dow Corning, USA 785
S
Santner Gerhard Graz University of Technology, Austria 653
Vilnius Gediminas Technical University, Department of Steel and
Šauiuvnas Gintas 565
Timber Structures, Lithuania
397
Schneider Jens TU Darmstadt, Germany
553
Schreurs Pieter Scheuten Absoluut Glastechniek, the Netherlands 47
Vilnius Gediminas Technical University, Department of Steel and
Serafinaviius Tomas 565
Timber Structures, Lithuania
Universität der Bundeswehr München, Institute for Structural
Seel Mathias 323
Design, Germany
Sena-Cruz José School of Engineering, University of Minho, Portugal 745
1003
Challenging Glass 3

Institute of Applied System Analysis National Technical University
Shabetia Alexander 349
of Ukraine“Kiev Politechnical Institute”, Ukraine
501
Siebert Barbara Ingenieurbüro Dr. Siebert, Germany
971
323
Universität der Bundeswehr München, Institute for Structural
Siebert Geralt 335
Design, Germany
725
Silva Tiago FACAL – Façade Engineering, Portugal 281
Sitoh Masao Nihon University, Tokyo, Japan 703
235
Sitte Sigurd Dow Corning, Germany
249
Werner Sobek Stuttgart, Germany / IIT (Illinois Institute of
Technology), Chicago, USA / ILEK (Institute of Lightweight 41
Sobek Werner
Structures and Conceptual Design), University of Stuttgart, 307
Germany
349
Soroka Helen G.S. Pisarenko Institute for Problems of Strength, Ukraine 523
535
Sparowitz Lutz Graz University of Technology, Austria 653
431
Stålhandske Christina Glafo – the Glass Research Institute, Sweden
579
Stepinac Mislav University of Zagreb, Faculty of Civil Engineering, Croatia 623
Sugizaki Kenichi Tokyo Institute of Technology, Sat-Consultant, Yasuda artier, Japan 129
T
Takeuchi Tori Tokyo Institute of Technology, Sat-Consultant, Yasuda artier, Japan 129
Techen Holger University of Applied Sciences Frankfurt am Main, Germany 907
Terzovic Jefto Faculty of Architecture, University of Belgrade, Serbia 891
Thun Peter BRS, the Netherlands 829
Timm Christoph SOM, USA 17
Tregubov Nikolay Konstantinovsky glasswork “Kvarsit”, Ukraine 535
Trösch Ernst Glas Trösch, Switzerland 735
V
Valarinho Luís Instituto Superior Técnico/ICIST, Portugal 745

Van Hulle Arno Ghent University, LMO, Belgium 177
Vandebroek Marc Artesis Hogeschool, Antwerpen, Belgium 587
139
187
Delft University of Technology, Faculty of Architecture, the 523
Veer Frederic
Netherlands 535
597
691
Velchev Dimitar University of Ruse, Bulgaria 679
Vrouwe Ivo Sint-Lucas School of Architecture, Belgium 829
W
151
Wälchli Ernst Wälchli + Partner AG, Switzerland
735
State Key Laboratory of Green Building Materials, China Building
Wan D.T. 917
Materials Academy, Beijing, China
Weber Felix Arup, UK 157
Weimar Thorsten University of Siegen, Institute of Building Structures, Germany 759
1004
Author Index

509
Institut für Baukonstruktion, Technische Universität Dresden,
Weller Bernhard 769
Germany
865
HafenCity Universität Hamburg, Façade Systems and Building
Wellershoff Frank 981
Envelopes, Germany
235
Wolf Andreas Dow Corning, Germany
249
Y
387
Yang Jian University of Birmingham, UK
917
535
Yevplov Yurii G.S. Pisarenko Institute for Problems of Strength, Ukraine
523
Yasuda Koichi Tokyo Institute of Technology, Sat-Consultant, Yasuda artier, Japan 129
Z
Zemborain Axel Pelli Clarke Pelli Architects, USA 167
University of Ljubljana, Faculty of Civil and Geodetic Engineering,
Žarni Roko 623
Slovenia
Testing Centre «Samarastroyispytanya» of Samara State University
Zubkov Vladimir 607
of Architecture and Civil Engineering, Russia
1005

Challenging Glass 3 Conference On Architectural and Structural Applications of Glass 1nbsped 9781614990611 9781614990604 Compress

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Challenging Glass 3 Conference On Architectural and Structural Applications of Glass 1nbsped 9781614990611 9781614990604 Compress

Uploaded by

Copyright:

Available Formats

Challenging Glass 3

Copyright © 2012. IOS Press, Incorporated. All rights reserved.

Faculty of Civil Engineering and Geosciences

Delft University of Technology

Bos, Louter, Nijsse, Veer (Eds.)

ISBN 978-1-61499-060-4 (print)

Organizing Committee Scientific Committee

Christian Louter Fred Veer, secretary (NL)

Welcome to Challenging Glass!

Wishing you a stimulating and enjoyable conference,

Freek Bos, Christian Louter

Rob Nijsse, Fred Veer

Platinum Main Sponsor

Host & Partners

Light in the Public Realm 3

Fusing Design, Innovation and Light 9

Engineering Invention in Glass Architecture 11

Case Study 1 World Trade Center – Podium Wall Design Development 17

Projects & Case studies 39

The Glass Screens of the Japan Post Tower 41

The Glass Sphinx: A Massive Stacked Glass Sculpture 47

The Apple Glass Cube: Version 2.0 57

Future Application of Structural Use of Glass 67

A Laminated Glass Wall Will Protect Warnemünde From High Water 75

Lincoln Center Canopies – Performance in Glass 83

Project for the Eiffel Tower: Constructive Geometry 93

Challenges in the Design, Fabrication and Installation of Glass Structures

Designing a Glass Pavillion to Protect an Ancient Greek Temple 139

A True All-Glass Staircase 151

Two Lines – Arup with David Chipperfield Architects 157

Torre Iberdrola, Bilbao, Spain 167

Joints, Fixings & Adhesives 175

Experimental Investigation of Unconventional Canopy Prototypes, Suspended

by Adhesive Bonds 177

Connecting Through Reinforcement – Experimental Analysis of a Glass

Determination of Adhesives Properties for Non-linear Numerical Simulation

Shear Capacity in Adhesive Glass Joints 209

Experimental and Numerical Analysis of Edge Seal Spacers of Insulated

Tensile Loading of Silicone Point Supports – Revisited 235

Investigation of Stress-Whitening in Transparent Structural Silicone Adhesive 249

Designing a Glass Bearing Connection with a Probability to EN1990 CC2 259

Influence of Various Factors on Mechanical Properties of Adhesive Joint in

Seismic Behaviour of Point Supported Glass Panels 281

The Mechanical Performance of Adhesives for a Steel-Glass Composite

Load Carrying Behaviour of Metal Inserts Embedded in Laminated Glass 307

Direct Glass Fabrication – New Applications of Glass with Additive

Analytical Solutions for Detail Problems in Structural Glazing 323

Glazing with Countersunk Point Fittings 335

Reduction of Edge Effect in Adhesive Joints of Glass Details 349

Strength, Stability & Safety 359

Improvement of Quality of Tempered Glass with Numerical Modeling 363

Analytical Approaches for Buckling Verification of In-plane Loaded

Towards a European Structural Glass Network: COST Action TU0905 397

How to Model Failure in Load-Bearing Glass Elements? A Discussion Based

New Expressions for the Effective Thickness of Laminated Glass 419

A Study of Polishing Performance of Glass Using Fluid Jet Polishing 431

Design of Arches Strengthened with Cables for Glass Roofs 441

Design of Cable Strenghtened Beams for Glass Structures 455

The Design of an All Glass Roof to EN1990 467

Stability Analysis of Structural Glass Systems 475

Numerical Simulation of Façade / Window Glazing Fracture Under Impact

Glazing Under Long Term Loads Induced by Water 501