Computers and Structures 87 (2009) 1534–1558

Contents lists available at ScienceDirect

Computers and Structures

journal homepage: www.elsevier.com/locate/compstruc

Upper bound limit analysis model for FRP–reinforced masonry curved

structures. Part II: Structural analyses
Gabriele Milani a,*, Enrico Milani b, Antonio Tralli b
a
Dipartimento di Ingegneria Strutturale (DIS), Politecnico di Milano, Piazza Leonardo da Vinci 32, 20133 Milano, Italy
b
Department of Engineering, University of Ferrara, via Saragat 1, 44100 Ferrara, Italy

a r t i c l e i n f o a b s t r a c t

Article history: A homogenized full 3D limit analysis model for the evaluation of collapse loads of FRP–reinforced
Received 16 February 2009 masonry vaults is presented. Six-noded rigid infinitely resistant wedges are used to model masonry.
Accepted 20 July 2009 Three-noded rigid infinitely resistant triangles are used to model FRP strips. Plastic dissipation is allowed
Available online 27 August 2009
only at the interfaces between adjoining elements. Unreinforced masonry homogenized failure surfaces
obtained in Part I of the present paper are used to evaluate plastic dissipation at the interfaces between
Keywords: masonry/masonry elements. A possible dissipation at the interfaces between FRP triangles and masonry
Masonry
wedges is also considered in order to take into account, in an approximate but effective way, the possible
FRP-reinforcement
Limit analysis
delamination of the strips from the supports. Italian code CNR DT 200 [CNR-DT 200, 2006. Guide for the
Vaults design and construction of externally bonded FRP systems for strengthening existing structures. C.N.R.,
Homogenization National Research Council, Italy; 2006.] formulas are used to evaluate peak interface tangential strength.
Kinematic approach While the delamination from the support can be modelled only in an approximate way within limit anal-
ysis, the aim of the paper is to accurately reproduce the change in the failure mechanism observed in
experimentations due to the introduction of strengthening elements.
A 3D approach to model masonry is used in order to take into account the presence of FRP strips either
at the extrados or at the intrados of the vaults.
Several numerical examples are analyzed, consisting of two different typologies of masonry arches (a
parabolic vault and an arch in a so-called ‘‘skew” disposition), a ribbed cross vault, a hemispherical dome
and a cloister vault. For all the examples presented, both the unreinforced and FRP-reinforced case are
discussed. Additional non-linear FE analyses are performed, modeling masonry through an equivalent
macroscopic material with orthotropic behavior at failure and possible softening, in order to assess limit
analysis results. Comparisons with experimental evidences, where available, are finally reported.
Reliable predictions of collapse loads and failure mechanisms are obtained with the model proposed for
all the cases analyzed, meaning that the approach proposed may be used by practitioners for a fast and
reliable evaluation of the effectiveness of a strengthening intervention.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction
The recent earthquakes occurred in Umbria and Marche (Italy
1997–1998) and Molise (Italy 2002) indicated that the historical
Italian buildings, essentially constituted by masonry structures,
are scarcely resistant to horizontal loads and highly vulnerable to
seismic actions. Such inadequate behavior of brickwork under
earthquakes is a common issue of masonry buildings in many
countries worldwide and is essentially due to mortar joints low
strength when loaded out-of-plane. Inadequate resistance under
seismic actions may be observed also for curved masonry
* Corresponding author. Tel.: +39 3495576064.
E-mail addresses: milani@stru.polimi.it (G. Milani), enrico.milani@unife.it
(E. Milani), atralli@ing.unife.it (A. Tralli).
structures, as for instance vaults, domes and arches, which typi-
cally are designed to withstand vertical loads under membranal
regimes only. Great impact on the scientific community and on
common people had the collapse of one vault of the S. Francesco
Basilica in Assisi during Umbria earthquake (26 September
1997), which caused both the death of four persons and an
unquantifiable artistic loss, due to the almost total destruction of
frescos realized by the great Italian medieval artist Cimabue.
The need of designing efficient and non-invasive strengthening
interventions to masonry structures in seismic area appeared
almost immediately clear to all technicians involved in the recon-
struction of collapsed vaults after Umbria and Marche earthquakes.
Conventional retrofitting techniques, such as external reinforce-
ment with steel plates, surface concrete coating and welded
mesh, have proven to be impractical, time expensive and add

0045-7949/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruc.2009.07.010
 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558 1535

F
zz
m only masonry uzzF

f FRP+masonry uyyF
F F
yy
FRP direction
y F
xx uxxF
FRP layer
Mesh
Thickness s utC
x (discretization at a structural level) F utB
(k)
t C u rC
(k) B B u
r r
Thickness
negligible
utA
Masonry/FRP uCs
I
z interface layer u sB
FRP triangles
(k)
u rA
s A
Masonry wedges
M
u sA
Thickness t
M
Masonry layer zz
uzzM

M uyyM
M
yy

M
y xx uxxM

Fig. 1. Multi-layer approach for a masonry shell reinforced with FRP. Unreinforced (Xm) and reinforced (Xf) domain discretization (left). Kinematic variables involved for Xf
discretization (right). FRP is modeled by means of infinitely resistant triangular elements, whereas masonry is modeled with infinitely resistant wedges. A possible jump of
displacement can occur at the FRP/masonry interface (delamination).

E
masonry wedge: E interface I ( 12
edge) masonry wedge: M masonry wedge: N M-N interface I

3 N
r 24
r 23
zz uzzN
M N
E zz uzzM yy 4 r14 3 r13
zz M
6 uzzE yy
1 P 2
z 2 uyyN r r
5 5 N s4 2 s3 2
z uyyM
E M sI 1 r11 r12
y
yy
E
uxxN r 2
I
2
uyy uxxM
3 E y 4 s1 s2
r1I N
r1I
uxxE 2 2 xx I
I
E
M
xx
2
t sI
x xx
P (xP,yP ,zP ) r 2I
4 4 x P 1
I

1 I

interface I stress action on a point P of interface I

1 1

Fig. 2. Masonry six-noded wedge element (left) and four-noded interface (right) between contiguous masonry elements (global and local frame of reference).

discretization (structural level) r(N )

z I-FRP
FRP triangles interface length L N (k)
E uzzN zz s axis
zz FRP direction P6
P3 (x 3, y3 , z3) u E Masonry wedges P7 - P5
zz Surface uyyN B
E N
N uxxN N
P6 - P8
P4 xx yy
N
uyyE A P4
E P2 (x 2, y2, z2) P2 M P2
yy I-FRP P8
zz
B P7 (k)
t axis
uzzM P3 P3 (k)
(k)
s axis
P1 r axis
E uxxE P1
xx M uyyM
P1 (x1, y1 , z1) M
M M
xx
uxxM yy
(k)
r axis

y r(M )
FRP direction r(N )
(M )
x P5
r

Fig. 3. FRP triangular element (left) and A–B interface between two contiguous M–N triangular FRP elements (right) with corresponding local frame of reference (a possible
jump of velocities along FRP direction may occur at A–B).
1536 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558

considerable mass to the structure (which may increase earth- λP 5 kN

quake-induced inertia forces). In this context, the utilization of 5 kN 5 kN
FRP strips as reinforcement instead of conventional methods
seems the most suitable solution for their limited invasiveness, 0.5 m
durability and good performance at failure.
Nevertheless, it is worth noting that, despite the great impor-
tance and the increasing diffusion of such innovative strengthening
technique, few numerical models devoted to the prediction of the 3m
ultimate load bearing capacity of out-of-plane loaded FRP–rein-
forced masonry [2,3] are nowadays at disposal.
Furthermore, when dealing with curved masonry structures,
the complex interaction between membrane and flexural actions
is very complex and brings additional complexity to the structural Line Load
analyses.
As well known, limit analysis (a valuable alternative to expen- 1.25 m
sive non-linear FE simulations) has been widely used for the anal-
ysis at failure of masonry structures (e.g. Heyman [4,5], Sinha [6],
Ferris and Tin-Loi [7], Orduna and Lourenço [8], Milani et al. [9],
etc.), because it requires only a reduced number of material param-
eters, providing limit multipliers of loads, failure mechanisms and,
3m
at least on critical sections, the stress distribution at collapse.
Reliable results have been obtained also in the specific case of Fig. 5. Unreinforced parabolic arch. Geometry, loading condition and FE discreti-
curved structures (see for instance Heyman [4], Livesley [10], zation adopted for the numerical analyses of the arch.
O’Dwyer [11], Oppenheim et al. [12], Milani et al. [13]). Very re-
cently, limit state approaches have been attempted for masonry ar-
ches also in presence of FRP reinforcement strips, see, e.g. Caporale The most important effect of a generic strengthening interven-
et al. [14] and Roca et al. [15]. tion executed with FRP strips is, indeed, to preclude the formation
As a matter of fact, non-linear complex damaging models (e.g. of the failure mechanism which causes the collapse of the unrein-
[16]) should be used for the analysis FRP–reinforced masonry. forced structure, with the subsequent formation of a new collapse
The FRP delamination from the support is, indeed, typically brittle, mechanism different from the un-strengthened case, with higher
as well as the tensile cracking of mortar joints. These aspects internal dissipation. Obviously, ‘‘hand” calculations may not be
preclude, in principle, the utilization of limit analysis, which is performed easily for complex structures, especially in the presence
based on the assumption of perfect plasticity for the constituent of curved shells with unsymmetrical loads. Therefore, the adoption
materials. of an upper bound approach combined with FEM seems particu-
Despite the aforementioned limitations connected to the larly suited for the prediction of FRP–masonry behavior prone to
hypotheses at the base of the approach proposed, following also collapse.
what suggested in the Italian Code CNR-DT 200 [1], limit analysis As discussed in the accompanying paper (Part I), the most suit-
may be useful for design purposes, to provide a fast and reliable able way for the analysis of FRP reinforced walls is the utilization of
estimation of collapse loads at a structural level. On the contrary, a two-steps approach based on homogenization concepts. First
no information is given by limit analysis concerning displacements step, relying in the simplified homogenization of unreinforced ma-
reached near collapse. However, such displacements are relatively sonry with a curved representative volume element has been
small and a rough estimation could be obtained by means of an widely illustrated in Part I and the reader is referred there for a
elastic analysis of the structure modeling the crack pattern previ- proper discussion of the limitations and the capabilities of the
ously evaluated by means of limit analysis. method.

F
zz
uzzF FRP direction Stresses acting on FRP/masonry
interface
Masonry/FRP interface 5
A 4 uyyF
ut B FRP direction
ut FRP F F r
yy
u rA A B
A
B u rB F
xx uxxF
r s
6
utC F-M
u sA (k) M s
(k)
I point P of the interface r
t zz fb
urC u sB (k) uzzM
r 1 C
C A
2
I
F-M
(k) (k)
B
u Cs s M uyyM t
M
Masonry 3 yy
(k) fb s
r
M
xx uxxM C fmt
(k)
s

Fig. 4. FRP/masonry interfaces. Left: delamination from the support (idealization of experimental test behavior). Right: discretization of interfaces with triangular elements
interacting with FRP triangles and masonry wedges.
 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558 1537

Table 1
Macroscopic homogeneous orthotropic model. Mechanical properties adopted for the simulations.

Example typea Parabolic Ribbed cross Hemi-spherical Cloister vault

arches vault dome
Young Modulus along 11 direction (E11) 5500 MPa 1700 MPa 1700 MPa 1700 MPa
Young Modulus along 22 direction (E22) 3600 MPa 1700 MPa 1700 MPa 1700 MPa
Poisson modulus (m12) 0.1762 0.15 0.15 0.15
Shear modulus (G12) 1400 MPa E/2 E/2 E/2
Mass density (cm) 1.8e9 N s2/ 2e9 N s2/mm 2e9 N s2/mm 2e9 N s2/mm
mm
Tensile strength along 11 direction (ft11) 0.62 MPa 0.05 MPa 0.1 MPa 0.02 MPa
Tensile strength along 22 direction (ft22) 0.32 MPa 0.1 MPa 0.17 MPa 0.08 MPa
Compressive strength along 11 direction (fc11) 2.4 MPa 2.3 MPa 1.8 MPa 2.2 MPa
Compressive strength along 22 direction (fc22) 3.0 MPa 2.7 MPa 2.0 MPa 2.6 MPa
Fracture energy in tension along 11 direction (Gf11) 2.0E2 N mm/ 0.0001 N mm/ 0.0001 N mm/ 0.0001 N mm/
mm2 mm2 mm2 mm2
Fracture energy in tension along 22 direction (Gf22) 2.0E2 N mm/ 0.00065 N mm/ 0.00065 N mm/ 0.00065 N mm/
mm2 mm2 mm2 mm2
Fracture energy in compression along 11 direction (Gfc11) 10 N mm/mm2 5 N mm/mm2 5 N mm/mm2 5 N mm/mm2
Fracture energy in compression along 22 direction (Gfc22) 5 N mm/mm2 2 N mm/mm2 2 N mm/mm2 2 N mm/mm2
Factor that determines the shear stress contribution to the tensile failure (a) 1.16 1.18 1.09 1.7
f f
(a ¼ txs2ty )
u

Factor which couples the normal compressive stressesb (b) 1.20 1.12 1.11 1.18
Factor which controls shear stress contribution to compressive failure (c) 2.4 3 3 3
Factor that specifies the equivalent plastic strain corresponding to the peak (kp) 8.0E4 6.0E4 6.0E4 6.0E4
compressive stress
a
11 direction coincides with x material axis (horizontal direction).
b
Parameter b was calculated fitting data obtained from the homogenization procedure with reference to formulas available in [32].

In order to validate the numerical model proposed, a number of

Table 2
FRP-strengthening mechanical parameters.
FRP reinforcement thickness (tFRP)
FRP reinforcement elastic modulus (EFRP)
Factor c1 (c1)
Factor cfd (cfd)
Factor cM (cM)
Fracture energy (CFK)
Design bond strength (ffdd)
In Part II, macroscopic anisotropic strength domains obtained in
Part I are implemented in a novel upper bound FE limit analysis
code for the analysis at collapse of entire FRP reinforced masonry
curved structures. Rigid infinitely resistant wedge-shaped 3D ele-
ments are used to model masonry at structural level. The utiliza-
tion of 3D elements is suitable to simulate the flexural strength
increase obtained by the introduction of FRP strips. On the other
hand, wedge-shaped elements are utilized with the aim of repro-
ducing possible diagonal out-of-plane failures, due to the develop-
ment of cracks (caused by bending and torsion) which zigzag
between contiguous bricks.
FRP strips are modeled by means of triangular rigid elements.
Masonry and FRP layers interact by means of interfacial tangential
actions between triangles (FRP) and wedges (masonry). Further-
more, a possible limited tensile strength for the FRP reinforcement
is considered at the interfaces between adjoining triangular ele-
ments. In this way, both delamination phenomenon at the FRP/
masonry interface and FRP tensile failure may be taken into
account. Despite the fact that delamination is a typical fragile
phenomenon, an equivalent ultimate shear strength for FRP/ma-
sonry interfaces is assumed in the framework of limit analysis,
following formulas provided by the recent Italian norm CNR-DT
200 [1] for the peak delamination strength. It has to be empha-
sized that the limit analysis approach here proposed is based on
the use a perfectly-plastic material response for masonry and for
the FRP/masonry interface, i.e. softening effect and limited ductil-
ity cannot be considered.
different structural examples are analyzed, consisting of two ar-
ches tested by Vermeltfoort [17] without reinforcement, a masonry
0.2 mm ribbed cross vault by Faccio et al. [18], a hemispherical dome and a
160,000 MPa
cloister vault, both tested by Foraboschi [19,20] in the presence
0.015
1.2
and absence of reinforcement.
3 Results obtained with the model proposed fit well both experi-
0.073 N/mm mental data (where available) and alternative expensive numerical
164 MPa load–displacement curves obtained using a macroscopic orthotro-
pic homogeneous model implemented in the commercial code
DIANA 9.3 [21]. From simulations results, it can be deduced that
the procedure proposed may represent a valuable tool for practi-
tioners involved in an inexpensive evaluation of ultimate loads
for masonry curved structures in the presence of FRP strips
reinforcement.
2. The structural level F.E. model: basic assumptions
The simplest FE discretization of a curved masonry structure
reinforced with FRP strips is represented by a piecewise linear
approximation of the middle surface by means of rigid flat six-
noded wedge elements. The utilization of wedges (i.e. 3D ele-
ments) instead of plate and shell triangles for masonry is useful
to reproduce the flexural behavior of the structures when a surface
reinforcement with FRP strips is introduced. This approach has, in-
deed, the further advantage of (1) automatically distinguishing be-
tween intrados and extrados reinforcement and (2) requiring only
in-plane and out-of-plane shear homogenized masonry failure sur-
faces, since flexural behavior is derived at a structural level by inte-
gration along the thickness.
In this framework, the less expensive limit analysis approach
which may be proposed is a model with rigid infinitely resistant
wedges. Thus, following a general approach widely diffused in
the technical literature for masonry plates out-of-plane loaded
(see e.g. Sinha [6]), in the model, plastic dissipation is allowed only
at the interfaces between adjoining elements. Such model has
shown to be sufficiently reliable even for masonry plates subjected
to out-of-plane loads and reinforced with FRP strips [3].
1538 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558

2.1. Six-noded wedge masonry element From Eq. (1), jump of displacements [U(P)] at a point P on the
interfaces I between two contiguous elements N and M can be eval-
The six-noded wedge element E utilized for bricks discretization uated as difference between velocities of P regarded belonging,
at a cell level (see Part I and Fig. 1) is used at a structural level to respectively, to N and M:
model homogenized masonry. As already discussed, kinematic
variables for each wedge element are represented by three ½UðPÞ ¼ UGM  UGN þ RM ðP  GM Þ  RN ðP  GN Þ ð2Þ
centroid velocities (uEx ; uEy ; uEz ) and three rotations around centroid
G (UEx ; UEy ; UEz ), Fig. 2.
Ce12 edge surface of an element E, connecting P1, P2, P4 and P5 2.2. Plastic flow relationships and power dissipation on masonry–
nodes is rectangular and jump of velocities on it is linear. In partic- masonry interfaces
ular, velocity field of a generic point P with global coordinates
(xP, yP, zP), on CE12 is expressed in the global frame of reference as: The main difference between the present model and the 2D
plate and shell approach presented by Milani et al. [13] relies
2 3 2 G3 2 32 3 essentially in the evaluation of the internal dissipation on
ux ux 0 UGy UGz x P  xG
6 7 masonry–masonry interfaces. In fact, while in the 2D model
UðPÞ ¼ 4 uy 5 ¼ 4 uGy 5 þ 4 UGy 0 UGx 54 yP  yG 5
uz uGz zP  zG dissipation occurs for mixed in and out-of-plane actions and a full
UG z UGx 0
characterization of masonry homogenized strength domain in
¼ UGE þ RE ðP  GÞ ð1Þ terms of membrane and flexural actions is needed, in the present
model flexural behavior is derived from integration of membrane
where U(P) is the point P velocity, UGE is the element E centroid actions along the thickness, once that the hypothesis of transverse
velocity and RE is element E rotation matrix. sections remain plane is assumed. It is worth noting that this latter

Fig. 6. Unreinforced parabolic arch. (a) Comparison among collapse loads provided by experimentation, limit analysis and non-linear FE code. (b) Deformed shapes at
collapse from limit analysis and non-linear FE code. (c) Normalized plastic dissipation patch obtained from limit analysis.
 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558 1539

Fig. 7. Reinforced parabolic arch. FRP strips dimensions and disposition.

assumption is implicitly imposed at the interfaces when rigid ele-
ments are used to discretize masonry.
Several differences with respect to the cell level occur when
plastic flow and power dissipation are evaluated. We introduce
for each interface I between contiguous elements the vector field
T I
tI, defined as tI ¼ sI1 sI2 rIs and representing the stress acting
along local axis rI1 (sI1 ), rI2 (sI2 ) and sI (rIs ), as indicated in Fig. 2.
Power dissipated at the interface can be evaluated analytically
as:
Z X12
PI ¼ ðrIr1 Dr 1 þ rIr2 Dr2 þ rIs DsÞdX ð3Þ
0
where Dr 1 , Dr 2 and Ds are velocities jumps (two tangential and
mutually orthogonal and one perpendicular to the interface, see
Fig. 2) in the local coordinate system rI1 -rI2 -sI .
Velocities jumps in the local system may be easily evaluated
from (2) once that the rotation matrix RI for rI1  rI2  sI is evalu-
ated (see Part I of the paper for further details):
2 3
Dr 1
6 7
DUðPÞ ¼ 4 Dr 2 5 ¼ RI ½UðPÞ ð4Þ
Ds
where DU(P) is the jump of velocities vector in the local system.
For each interface I of area X12 connecting nodes 1–2–3–4
(Fig. 2), we suppose to have at disposal a homogenized (linearized)
strength domain in the local coordinate system rI1  rI2  sI and
constituted by mI planes. Such a linearization for each interface
(and, in principle, for each point of the interface) can be obtained
from Shom applying the procedure recommended by Krabbenhoft
et al. [22], and the reader is referred there for further details.
In particular, we assume that a generic linearization plane qI has
I I I I
equation Aqr1 sI1 þ Aqs rIs þ Aqr2 sI2 ¼ C qI 1 6 qI 6 mI . Introducing plastic
multipliers fields at the interface (one for each linearization plane)
from Eq. (3), power dissipated at the interface can be re-written as:
Z L12 X
I I I I
PI ¼ k_ qI ðr 1 ; r2 ÞðAqr1 sI1 þ Aqs rIs þ Aqr2 sI2 Þds ð5Þ
0 qI
I
Obviously, field k_ qI ðr 1 ; r 2 Þ assumes the same analytical expres-
sion found for the velocity field, i.e. is linear in r1–r2, see Eq. (4).
Therefore, k_ qI ðr1 ; r2 Þ field is fully determined introducing only three
plastic multipliers for each internal interface and for each lineari-
zation plane, corresponding e.g. to nodes 1, 4, and 2.
On the other hand, the numerical evaluation of integral (5) is
trivial and plastic dissipation at a generic interface can be obtained
as:
mI
X12 X I
PI ¼ ðk_ I;1I þ k_ I;4I þ k_ I;2I ÞC q
q q q I ð6Þ
3 qI
where all the symbols have been already introduced.
External power dissipated can be written as Pex ¼ ðPT0 þ kPT1 Þw,
where P0 is the vector of permanent loads, k is the load multiplier
for the structure examined, PT1 is the vector of variable loads and w
is the vector of assembled centroid elements velocities. As the
amplitude of the failure mechanism is arbitrary, a further normal-
ization condition PT1 w ¼ 1 is usually introduced. Hence, the exter-
nal power becomes linear in w and k.
3. Structural analysis
3.1. Three-noded flat FRP elements (triangles)
Three-noded triangular shell elements are utilized to model FRP
strips, as shown in Fig. 3, with nodes coordinates P i ¼ ðxi ; yi ; zi Þ;
i ¼ 1; . . . ; 3 and node numbers disposed in counter clockwise. Let
symbol X indicates the surface of a FRP element E, Fig. 3.
Let us introduce the two natural coordinates t and p for element
E, varying respectively from 0 to 1 and from 0 to 1  t [23]. The
global coordinate (x, y, z) of a point P within the triangular element
E can be expressed as:
1540 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558

T
X
3
T
Analogously to wedge masonry elements, FRP triangles are
P  ½x y z ¼ Ni ðt; pÞ½ xi yi zi  ð7Þ supposed infinitely resistant and rigid. Therefore, plastic dissipa-
i¼1
tion is allowed only at the interfaces between contiguous elements
where xi, yi and zi are global coordinates of node Pi (i = 1, ..., 3) and Ni due to stresses acting on the fibers direction. Continuity of the
are well-known shape functions. velocity field is imposed at each interface between contiguous

Fig. 8. Reinforced parabolic arch. (a) Comparison between collapse loads provided by limit analysis and non-linear FE code. (b) Deformed shapes at collapse from limit
analysis and non-linear FE code. (c) Normalized plastic dissipation patch in masonry elements obtained from limit analysis. (d) Normalized plastic dissipation patch on FRP/
masonry interfaces (with the simplifications assumed in the model, plastic dissipation indicates delamination of the strips).
 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558 1541

FRP triangular elements only along directions r(k) and t(k) (see Load [kN] Failure load increase
Fig. 3) whereas a possible jump of velocities is supposed to occur
along direction s(k).
With reference to Fig. 3, let two contiguous FRP elements M and 120.0
N be considered. Their centroid velocities and rotation rates are
   T h iT
M T
uM ¼ uM M
xx uyy uzz , uN ¼ uNxx uNyy uNzz , UM ¼ UM M M
xx Uyy Uzz Parabolich arch (FRP
h iT 100.0
extrados+intrados)
and UN ¼ UNxx UNyy UNzz . Jump of velocities on the common M
Skew parabolich arch
(FRP extrados+intrados)
and N interface (I-FRP) is linear: therefore, it is necessary to evalu- 80.0 Ribbed cross vault
ate jump of velocities only on the interface extremes A and B as
difference between velocities of nodes 1–3 and 2–4, respectively. Cloister Vault
Furthermore, a local frame of reference s(k)–t(k)–r(k) has to be 60.0
defined on the interface as follows (see also Fig. 3 for symbols Dome

meaning):
8 40.0 Values of bond-
ðP 7 P 5 ÞþðP 6 P 8 Þ
>
> sðkÞ ¼ kðP strength fb of
< 7 P 5 ÞþðP 6 P 8 Þk
technical interest
tðkÞ ¼ rðkÞ  sðkÞ ð8Þ
>
> 20.0
: rðkÞ ¼ rM þrN
krM þrN k 0.0 1.0 2.0 3.0
ðP 1 P 5 ÞþðP 2 P 5 Þ
where rðMÞ ¼ kðP ðP 6 P 4 ÞþðP 6 P3 Þ
and rðNÞ ¼ kðP . fb [MPa]
1 P 5 ÞþðP 2 P 5 Þk 6 P 4 ÞþðP 6 P3 Þk

If we denote with ½ xA yA zA  point A coordinates, node 1 Fig. 9. Synopsis of sensitivity analyses conducted for all the examples analyzed.
velocity is given by: Collapse load evaluation at different values of fb tangential bond strength (fb equal
to zero means unreinforced case).
2 3 2 M3 2 32 3
u1xx uxx 0 UM
yy UMzz x A  xM
4 u1yy 5 ¼ 4 uM 5þ6
4 UM 0 UM
7 4 yA  yM 5
yy yy xx 5
u1zz uM
zz UM UM 0 zA  zM
zz xx
λP 5 kN
M
¼ UM
G þ R M ðA  G Þ ð9Þ 5 kN 5 kN
M
where G ¼ ½ xM yM zM  is the centroid of element M.
0.5 m
Node 1 velocity can be easily re-written in the s(k)–t(k)–r(k) local
interface frame of reference by means of the rotation matrix
T(M, N) deduced from Eq. (9), i.e.:
 T M 3m
u1s u1r u1t ¼ TðM; NÞ½UM
G þ R M ðA  G Þ
ð10Þ

No difference occurs for node 2, provided that element N veloc-

ities and centroid are used instead of quantities related to M .
Consequently, A jump of velocities is evaluated (in the local
coordinate system) as: 1.25 m
M N
½uA  ¼ TðM; NÞ½UM
G  UM
G þ RM ðA  G Þ  RN ðA  G Þ ð11Þ
T T
where ½uA  ¼ ½ DuAs DuAr DuAt  ¼ ½ u1s  u2s u1r  u2r u1t  u2t  .
Analogous considerations can be repeated for node B, i.e.:
M N
½uB  ¼ TðM; NÞ½UM M
G  UG þ R M ðB  G Þ  R N ðB  G Þ ð12Þ
As already discussed, plastic dissipation is supposed to occur at
the interfaces only, due to stresses acting parallel to fibers direc-
Line Load
tion #k.
As a rule, low compressive stresses induce buckling of the
strips, due to the FRP negligible thickness. In order to take into
account this effect (at least in an approximate way), different limit
þ
stresses are assumed in tension and compression, namely fFRP (as- 3m
sumed equal to ffdd or ffdd,rid in agreement with CNR-DT 200 [1], see

the following section for details) for tensile failure and fFRP  0 for Fig. 10. Unreinforced skew arch. Geometry, loading condition and FE discretization
adopted for the numerical analyses of the arch.
compression buckling, respectively.
To be kinematically admissible, velocity jump at the interfaces
(Eqs. (11) and (12)) must comply to the following equality con-
straints (associated flow rule): On the other hand, from Eqs. (11)–(13), within each interface I-
2 3 2 3 FRP of length LI-FRP (thickness s), the power dissipated may be easily
Duis k_ iIFRPþ  k_ iIFRP evaluated as:
6 i7 6 7
½ui  ¼ 4 Dur 5 ¼ 4 0 5 ð13Þ
LIFRP
Duti
0 pF ¼ ðDuAs rA þ DuBs rB Þ
2
where i = A or B and k_ IFRPþ and k_ IFRP are plastic multiplier rates of LIFRP þ _ IFRPþ _ IFRPþ  _ IFRP
i i ¼ ðfFRP ðkA þ kB Þ þ fFRP ðk A þ k_ IFRP
B ÞÞ ð14Þ
point i (interface I-FRP) corresponding to fFRP þ 
and fFRP , respectively. 2
1542 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558

where rA and rB represent stress action along s(k) on nodes A and B,
respectively, and all the other symbols have been already
introduced.
3.2. FRP/masonry interfaces (delamination)
One of the most important aspects in the application of
composite materials for strengthening structural elements is the
adhesion between the reinforcing and reinforced materials. In
particular, when delamination from the support occurs, the
effectiveness of the reinforcement vanishes. This phenomenon is
very complex to model, especially in the framework of limit anal-
ysis, because it involves materials with different properties (ma-
sonry, FRP and glue layer) and depends on several parameters.
Experimental studies demonstrated that decohesion occurs for ma-
sonry failure [24]. In other words, the delaminated FRP presents a
significant layer of masonry material on the debonded surface.
A rigorous methodology to directly take into account in a
numerical model the behavior of the layer between masonry and
FRP is the use of the interface model concept. According to this

Fig. 11. Unreinforced skew arch. (a) Comparison among collapse loads provided by experimentation, limit analysis and non-linear FE code. (b) Deformed shapes at collapse
from limit analysis and non-linear FE code. (c) Normalized plastic dissipation patch obtained from limit analysis.
 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558
model, forces acting on the interface are related to the relative dis-
placement of the two sides (masonry and FRP), thus requiring the
utilization of interface elements.
It is worth noting that, at present, some national norms (see for
instance Japanese [25] or Italian [1] codes) are at disposal to eval-
uate with simple but effective formulas the delamination of a FRP
strip from the support, especially for RC elements. However, the
determination of the ultimate delamination strength when FRP is
applied to masonry elements remains an open issue, but a discus-
sion on the reliability of formulas proposed by codes of practice
regarding this topic is outside the scopes of the present paper.
For the sake of simplicity, in what follows, we refer to what stated
in the Italian technical norm [1]. Anyway, it is stressed that any for-
mula may be implemented in the code with no conceptual
differences.
In the Italian norm, a simplified approach is proposed to evalu-
ate the delamination phenomenon, suitably limiting force action
on the FRP strip. In particular, the ffdd design tensile strength of
FRP elements is:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 2  EFRP  CFk
ffdd ¼ pffiffiffiffiffiffiffi
Cfd CM t FRP
if the so-called bond length lb is greater than the optimal bond
length le or:
lb lb
ffdd;rid ¼ ffdd ð2  Þ
le le
if lb 6 le.
In Eqs. (15) and (16) the following symbols have been used:
– ffdd,rid, the reduced value of the design bond strength;
– ffdd, the design bond strength;
– EFRP, the FRP Young modulus;
– tFRP, the FRP thickness;
– cfd, safety factor (in what follows it is assumed equal to 1.20);
– cM, partial safety factor for masonry (see [26]) assumed in the
following equal to 1.0 in order to obtain characteristic values
of bond strength;
– lb, the bond length of FRP elements;
qffiffiffiffiffiffiffiffiffiffiffiffiffi
– le ¼ E2f FRP t FRP
mtm
, the optimal bond length of FRP corresponding to
the minimalmtm indicates masonry average tensile strength).
Finally, the term CFk in (15) represents the characteristic value
of the specific fracture energy of the FRP strengthened masonry
under a delamination test. In particular, when the debonding
Fig. 12. Reinforced skew arch. FRP strips dimensions and disposition.
1543
involves the first masonry layers, the term CFd (design value)
may be evaluated by means of the following formula:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
CFd ¼ c1 fmk  fmtm ½f in N=mm2  ð17Þ
where c1 is an experimentally determined coefficient, that typically
may range between 0.015 and 0.030 and fmk is the characteristic va-
lue of masonry compressive strength. Here, it is worth underlining
that the evaluation of the specific fracture energy (which is used
both in limit analysis and in any non-linear finite element code
for the calibration of mechanical properties of FRP/masonry bond)
remains an open issue, since the range of variability imposed by
(17) is very large, also because masonry mechanical properties in
tension are affected by a wide scatter.
The sb-slip constitutive law proposed by the Italian norm, see
Fig. 4, permits an indirect evaluation of shear limit stress (here de-
noted with the symbol fb) to use for masonry/FRP interface ele-
ments (and thus avoiding a discretization of FRP strips by means
of truss elements with limited strength ffdd), once that the ultimate
slip (usually fixed at 0.2 mm) is known (area under the sb-slip con-
ð15Þ stitutive law of Fig. 4 is CFd).
From the literature results, it appears clear that a damaging
model should be used for the analysis of masonry structures rein-
forced with FRP [16,27]. This precludes, in principle, the utilization
of limit analysis, which is based on the assumption of perfect plas-
ticity and infinite ductility for the constituent materials. Neverthe-
ð16Þ
less, following also what suggested in the CNR-DT 200 code [1],
limit analysis is particularly useful to have a fast and relatively reli-
able estimation of collapse loads at structural level. The most
important effect of a generic strengthening intervention executed
with FRP strips is, indeed, to preclude the formation of the failure
mechanism which causes the collapse of the unreinforced struc-
ture, with the subsequent formation of a new collapse mechanism
different from the previous, with higher internal dissipation. Obvi-
ously, ‘‘hand” calculations may not be performed easily for com-
plex structures, especially in presence of curved shells with
unsymmetrical loads. In this framework, an upper bound approach
combined with FEM (able to provide with a very limited computa-
tional effort field of velocities at collapse) seems particularly suited
for an inexpensive estimation of the load bearing capacity of FRP
reinforced masonry.
Let us consider a triangular FRP–masonry interface IF–M be-
tween elements F (FRP) and M (masonry), as depicted in Fig. 4.
uF ¼ ½ uFxx uFyy uFzz T and uM ¼ ½ uMxx uM
yy uM T
zz  indicate F and M
centroids velocities, respectively, whereas UF ¼ ½ UFxx UFyy UFzz T
T
and UM ¼ ½ UM xx UM
yy UMzz  indicate F and M rotation rates vectors.
1544 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558

Jump of velocities on the common IF–M interface is linear and where GM ¼ ½ xM yM zM  is the centroid of masonry element M.
may be evaluated on nodes A, B and C of the interface (Fig. 4) as We introduce a local s(k)–t(k)–r(k) frame of reference for IF–M de-
difference between velocities of nodes 1–4 and 2–5 and 3–6, fined as follows (see Fig. 4):
respectively. In particular, if ½ xA yA zA  represents point A coor- 8 ðkÞ
dinates, velocity of node 1 is given by: < s  FRP direction
>
ðBCÞ ðACÞ
tðkÞ ¼ kBCk  kACk ð19Þ
2 3 2 3 2 32 3 >
: ðkÞ
M M
u1xx uM
xx
0 Uyy U zz x A  xM r ¼ tðkÞ  sðkÞ
4 u1yy 5¼ 4 uM
yy
5þ6
4 UM 0
74
U 5 yA  yM 5
M
yy xx As already discussed in the previous section (Eq. (10)), node 1
u1zz uM UM UM 0 zA  zM
zz zz xx velocity in the local s(k)–t(k)–r(k) frame of reference is:
M  
¼ UM
G þ R M ðA  G Þ ð18Þ u1s u1r u1t ¼ TðI
FM
Þ½UM M
G þ R M ðA  G Þ
ð20Þ

Fig. 13. Reinforced skew arch. (a) Comparison between collapse loads provided by limit analysis and non-linear FE code. (b) Deformed shapes at collapse from limit analysis
and non-linear FE code. (c) Normalized plastic dissipation patch in masonry elements obtained from limit analysis. (d) Normalized plastic dissipation patch on FRP/masonry
interfaces (with the simplifications assumed in the model, plastic dissipation indicates delamination of the strips).
 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558 1545

λP

Position of
concentraded load
Point 1
1.15 m Point 2
Point 3

2.3 m

0.3 m
2.3 m

Fig. 14. Ribbed cross vault. Geometry, loading condition and FE discretization adopted for the numerical analyses of the arch.

!
Eq. (20) can be re-written for node 4, substituting M quantities AI X
C
F—M i i i
with F quantities. p ¼ ðDut ri þ Dus ssi þ Dur sri Þ
3 i¼A
Thus, jump of velocity on A (in the local coordinate system) may
MN
be evaluated as: I 3 NX
X
A PL

¼ k_ iMF;k Dk ð23Þ
½uA  ¼ TðI FM
Þ½UM  UFG M F
þ RM ðA  G Þ  RF ðA  G Þ ð21Þ 3 i¼1 k¼1
G

where ½uA  ¼ ½ DuAs DuAr DuAt T ¼ ½ u1s  u4s u1r  u4r u1t  u4t T indi-
cates the jump of velocities on A in the local coordinate system
and all the other symbols have been already introduced.
No conceptual differences occur for nodes B and C, therefore Eq.
(21) can be utilized for all the vertices of the triangular interface.
In the framework of classic limit analysis theorems, to be kine-
matically admissible, jump of velocities field at the F–M interfaces
must obey an associated flow rule. A linearization of F–M failure
surface in the form Akssi + Bksri + Ckri = Dk, k ¼ 1; . . . ; N MF
PL
MF
(N PL
the number of planes used in the linearization of the failure
surface, ri, ssi and sri are defined in Fig. 4) is assumed. In the
framework of associated limit analysis, the following equality con-
straints must be imposed:
2P 3
2 3 MF
N PL
Duis Ak k_ iMF;k
6 k¼1 7 elements centroids velocities (w) and rotations (U) of both
6 7 6 P MF 7
½ui  ¼ 4 Duir 5 ¼ 6 NPL Bk k_ MF;k 7 ð22Þ
4 k¼1 i 5
Duit PNPL
MF
_ MF;k
k¼1 C k ki
where i = A, B or C, k_ MF;k
i is the kth plastic multiplier rate corre-
sponding to the kth plane. The Italian norm provides r–ss–sr failure
surfaces for masonry/FRP interfaces, see Fig. 4, where fmt represents
masonry tensile strength and fb is the interface shear strength.
As experimental evidences show [51], debonding of the strips
for curved structures occurs always in the presence of non-null
normal actions between the reinforcement and the support. Hence,
the determination of the actual strength domain of the interfaces is
particularly important. It is interesting to notice that interface
strength for normal stresses is assumed here equal in compression
and tension. A more realistic model should impose a different
resistance between compressive and tensile regimes (dashed line
in Fig. 4), since fmt is strictly related to bricks tensile and compres-
sive strength. Obviously, in the model, any non-linear bond failure
surface can be implemented without any conceptual difficulty.
However, the approach proposed by the Italian norm is adopted
hereafter for the sake of simplicity.
From Eqs. (20)-(22), within each interface F–M of area AI, the
power dissipated may be easily evaluated as:
3.3. The linear programming (LP) problem
A linear programming problem is obtained, after some elemen-
tary assemblage operations, where the objective function is the to-
tal internal power dissipated minus the power dissipated by
external loads not dependent on the load multiplier, i.e.:
8
in;ass _ I;assT
>
< minfPI (
> k  PT0 wg
is eq
> Aeq U ¼ b ð24Þ
: such that k_ I;ass P 0
>
where
– U is the vector of global unknowns and collects the vector of
FRP and masonry elements and the vector of assembled inter-
face plastic multiplier rates (k_ I;ass ). k_ I;ass collects plastic multiplier
rates of masonry–masonry interfaces, FRP–FRP interfaces and
masonry–FRP interfaces.
– Aeq is the overall constraints matrix and collects normalization
conditions, velocity boundary conditions and constraints for
plastic flow in velocity discontinuities (on FRP, masonry and
FRP–masonry interfaces).
– PIin;ass is a row vector that collects contributions to the internal
dissipation of masonry–masonry, FRP–FRP and masonry–FRP
interfaces (Eqs. (6), (14), and (23), respectively).
The reader is referred to [29] for a critical discussion of efficient
(classic) linear programming tools suited for solving (24).
On the other hand, it is worth noting that recent trends in limit
analysis have demonstrated that the linearization of the strength
domain can be circumvented using conic/semidefinite program-
ming (e.g. [30,31]). This tool is more powerful in terms of process-
ing time with respect to classic LP and could lead to a numerical
efficiency improvement for the structural analyses. Both free (e.g.
SeDuMi, http://sedumi.mcmaster.ca/) and commercial (e.g.
http://www.mosek.com) standalone tools are nowadays available;
1546 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558

nonetheless, since the aim of this paper is mainly concentrated on
the structural aspects related to the limit analysis of curved ma-
sonry elements, the classic interior point LP routine available in
Matlab is used for the sake of simplicity.
4. Structural examples
In this section, several structural examples consisting of a num-
ber of masonry curved shells loaded until failure are discussed in
detail. Limit analysis results are compared, where possible, with
experimental data available from the technical literature.
For all the examples, both the FRP-reinforced and the unrein-
forced case are considered, with the aim of evaluating both the dif-
ferences occurring in the failure mechanisms numerically obtained
and the increase of the ultimate load bearing capacity consequent
to the strengthening intervention.
In order to further assess the reliability of the procedure pro-
posed, results obtained through an alternative non-linear FE anal-
ysis conducted by means of the commercial code DIANA [21] are
finally reported. In this code, the orthotropic continuum macro-
scopic model originally presented in Lourenço et al. [32] is imple-
mented for masonry. A possible softening behavior both in tension
and compression may be modeled within DIANA. Furthermore,

Fig. 15. Ribbed cross vault. (a) Comparison among collapse loads provided by experimentation, limit analysis and non-linear FE code. (b) Deformed shapes at collapse from
limit analysis and non-linear FE code. (c) Normalized plastic dissipation patch obtained from limit analysis (left: top view. Right: bottom view).
 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558
Fig. 16. Reinforced ribbed cross vault. FRP strips disposition.
triangular interface elements with brittle behavior reproducing
delamination of the strips from the support are used for ma-
sonry–FRP bond. As a consequence, this approach represents a
valuable reference to compare limit analysis results, in absence
of experimental data available.
The first set of structural examples relies in a parabolic and a
parabolic skew arch, both tested by Vermeltfoort [17] in the
absence of reinforcement. No experimental data are available for
the example at hand in the presence of FRP, therefore only compar-
isons with [21] are provided. The second example is a ribbed cross
vault experimentally tested by Faccio et al. [18], the third is a
hemispherical dome (some numerical and experimental data are
available in Creazza et al. [33,34] and Foraboschi [20]), whereas
the last example is cloister vault with square plan (data from
Foraboschi [20]).
For all the cases analyzed, the homogenized limit analysis ap-
proach presented in the previous section has been used to predict
ultimate loads and failure mechanisms, assuming for the constitu-
ent materials (where available) experimentally determined
mechanical properties (see Vermeltfoort [17], Faccio et al. [18],
Creazza et al. [33,34], Foraboschi [20]).
Finally, a sensitivity analysis is conducted varying in a wide
range masonry–FRP interfaces ultimate tangential strength (fb),
with the aim of investigating the influence of bond resistance on
collapse loads increase and failure mechanism change.
4.1. Parabolic arch
A number of numerical and analytical models based on limit
analysis theorems are available in the literature for the evaluation
of collapse loads of unreinforced masonry arches. The original re-
search in this field goes back to 1960s and is due to Heyman
[4,5], and includes, among the others, associated, non-associated
and no-tension material approaches (see for instance recent contri-
butions by Gilbert and Melbourne [35], Lucchesi et al. [36], Olivito
and Stumpo [37], Baratta et al. [38], Drosopoulos et al. [39], Betti
et al. [40], Audenaert et al. [41], etc.). More recently, research fo-
cused on the FRP-reinforced case and includes mono-dimensional
models immediately applicable by practitioners, where the inter-
action between masonry and reinforcement is taken into account
in various ways (see, for instance [42–46], etc).
The common issue of almost the totality of the aforementioned
approaches is that they are specifically devoted to the analysis of
1D curved structures (beam elements), whereas only few papers
deal with 2D/3D structures (e.g. [47–51]). Since the aim of the
present paper is to analyzed 3D structures under complex loading
1547
conditions, the preliminary set of simulations hereafter reported
on a 1D arch in the presence and absence of reinforced must be
considered as a preliminary simple test for which reference solu-
tions exist and are obtained by consolidated models already avail-
able in the technical literature.
In particular, two parabolic unreinforced masonry arches exper-
imentally tested by Vermeltfoort [17] are considered. The first
arch, the only one for which a comparison with 1D existing models
is possible, is a barrel rectangular vault with clear span equal to
3 m, width equal to 1.25 m and a sagitta of 0.5 m, see Fig. 5. The
second structure has the same geometry, but presents a so-called
‘‘skew” disposition of the supports (i.e. supports are offset
perpendicularly to arch axis of 1.25 m). The offset makes the
behavior strongly tri-dimensional, thus precluding an analysis
with mono-dimensional models. The arches are one-head brick
structures with depth equal to 100 mm. Bricks are assumed of
dimensions 200  100  52 mm3 (Rijswaard soft mud bricks),
whereas joints thickness is equal to 12 mm. Following [17],
experimental bricks compressive strength is 27 N/mm2, whereas
mortar compressive strength is 2.5 N/mm2. Mechanical properties
assumed for joints and bricks are reported in Part I and the
reader is referred there for further details on masonry
macroscopic strength domain assumed for the simulations here
reported.
Geometry, loading condition and FE discretization adopted for
the analyses of the barrel rectangular vault are depicted in Fig. 5.
Only the second concentrated load from the left was increased
by Vermeltfoort [17] until failure, whereas the remaining loads
were maintained constantly equal to 5 kN.
Mechanical properties assumed in the DIANA [21] model for
masonry and FRP are summarized in Tables 1 and 2, respectively.
The reader is referred to [32] and [1] for a comprehensive discus-
sion of mechanical properties adopted, especially for what
concerns the inelastic range.
In Fig. 6a, a comparison among [21] load–maximum displace-
ment curves, experimental and limit analysis collapse loads in
the absence of reinforcement is reported. Results obtained using
a simple 1D approach (see [46]) are also represented. Furthermore,
in Fig. 6b, a comparison between deformed shapes near the col-
lapse provided by limit analysis and Lourenço et al. [32] macro-
scopic orthotropic model are depicted. As it is possible to note
from an overall analysis of the results, numerical models provide
basically the same collapse load, as well as similar failure mecha-
nisms. When dealing with the limit analysis model, internal plastic
dissipation (Fig. 6c) is concentrated under the line of application of
the external load, meaning that, in this case, failure occurs for the
1548 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558

formation of two well defined flexural ‘‘plastic” hinges. A total of
four hinges is present at collapse (two geometrical and two plas-
tic), a result clearly in agreement with simple mono-dimensional
predictions based on ‘‘hand” calculations (kinematical chains, see
for instance [46]).
In order to prevent the formation of the plastic hinges indicated
in Fig. 6c, we suppose to strengthen the arch by means of two sets
of FRP strips (width 100 mm), as illustrated in Fig. 7. The first set of
strips is disposed at the extrados of the arch, whereas the second is
applied at the intrados in order to prevent cracks opening due to
out-of-plane actions. It is worth remembering that the 3D model
here utilized allows to suitably distinguish between reinforcement
at the extrados and intrados in curved shells. For masonry–FRP
interface, a bond strength fb equal to 0.3 MPa is assumed.
External force–maximum displacements curves obtained for
the reinforced case with DIANA [21] and collapse load provided
by the homogenized limit analysis simulations are reported in
Fig. 8a. Collapse load obtained by means of a recently presented
1D approach [46] is also represented. As one can note, also in this
case limit analysis simulations are in good agreement both with

Fig. 17. Reinforced ribbed cross vault, first FRP disposition. (a) Comparison between collapse loads provided by limit analysis and non-linear FE code. (b) Deformed shapes at
collapse from limit analysis and non-linear FE code. (c) Normalized plastic dissipation patch in masonry elements obtained from limit analysis. (d) Normalized plastic
dissipation patch on FRP/masonry interfaces (with the simplifications assumed in the model, plastic dissipation indicates delamination of the strips).
 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558 1549

results obtained through expensive simulations conducted with
commercial non-linear FE codes and with simple 1D approaches.
Arch deformed shape at collapse and power dissipation patch
(in masonry elements and at the interface between FRP and ma-
sonry elements) are reported in Fig. 8 from b to d. The effect of
the reinforcement, which tends to preclude the formation of the
plastic cylindrical hinges which cause the collapse in the unrein-
forced case is particularly evident, as well as the delamination of
the strips, which is obviously critical near the cylindrical hinges
(in agreement with [1]).
In Fig. 9, the increase of ultimate loads in presence of FRP and
varying fb masonry/FRP interface strength in a wide range (from
0 to 3 MPa) is represented. As it is possible to notice, failure load
reaches asymptotically a maximum for high fb values, but a consid-
erable increase in the collapse multiplier is obtained even for rela-
tively low fb (namely between 0.3 and 0.8 MPa). From simulations
results, it is interesting to notice that the correct evaluation of fb
remains a very important issue: therefore, a sensitivity analysis
should be always performed to collect information on strengthen-
ing efficiency.

Fig. 18. Reinforced ribbed cross vault, second FRP disposition. (a) Comparison between collapse loads provided by limit analysis and non-linear FE code. (b) Deformed shapes
at collapse from limit analysis and non-linear FE code. (c) Normalized plastic dissipation patch in masonry elements obtained from limit analysis. (d) Normalized plastic
dissipation patch on FRP/masonry interfaces (with the simplifications assumed in the model, plastic dissipation indicates delamination of the strips).
1550 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558
As can be deduced comparing deformed shape in the presence
or absence of FRP (Figs. 8 and 6, respectively), FRP strips introduc-
tion results in a clear change of the failure mechanism (an out-of-
plane sliding can be observed in the reinforced case under the line
of application of the load), confirming that reinforcement act as a
tie, tending to preclude the formation of cylindrical hinges on
masonry.
The second numerical simulations sub-set relies on the same
parabolic arch analyzed previously, but with a so-called ‘‘skew”
disposition, as illustrated in Fig. 10, obtained offsetting external
supports of 1.25 m. Supports offset allows the arch to fail asym-
metrically under a complex combination of bending and torsion
along yield lines. It is therefore an interesting example to test the
capabilities of the 3D model here presented in the presence of
combined internal actions.
Obviously, in this case, a mono-dimensional analysis would be
not useful and a 3D shell model is necessary to reproduce the
torsional behavior of the arch.
No experimental force–displacement curves are at disposal for
the example at hand. From [17], only the experimental collapse
load (around 26 kN) is available. When mechanical properties
reported in Part I are adopted, a numerical collapse load of
33 kN is obtained, in acceptable agreement (when an associated
limit analysis approach is adopted) with experimental evidences.
A comparison among experimental collapse load, limit analysis
failure multiplier and load–maximum displacement curve
obtained by means of the DIANA [21] orthotropic continuum
model is summarized in Fig. 11a. Furthermore, in Fig. 11b, a com-
parison between deformed shapes near the collapse provided by
limit analysis and the macroscopic orthotropic model are
represented. As in the previous case, numerical models essen-
tially provide the same collapse load, as well as very similar fail-
ure mechanisms. Finally, normalized plastic dissipation patch
provided in the post processing phase by limit analysis is
depicted in Fig. 11c. As it is possible to note, internal dissipation
Fig. 19. Hemispherical dome. Geometry, loading condition and FE discretization adopted for the numerical analyses of the arch.
is concentrated not only under the line of application of the
external load, but also along two inclined lines, thus demonstrat-
ing that failure occurs as a consequence of the limited strength of
the arch in torsion.
With the aim of investigating the effect of a possible FRP rein-
forcement disposed in such a way to preclude the formation of
plastic hinges, the disposition illustrated in Fig. 12 is numerically
tested, with both numerical models proposed. Mechanical proper-
ties assumed for the reinforcement and the FRP–masonry bond are
summarized in Table 2.
Also in this case, strips are disposed both at the extrados and
intrados, in order to properly preclude cracks opening only where
required. For masonry–FRP interface, a bond strength fb equal to
0.3 MPa is assumed.
External force–maximum displacements curves obtained using
[21] and collapse load provided by the homogenized limit analysis
simulations are compared in Fig. 13a. As one can note, also in this
case limit analysis simulations are in good agreement with results
obtained through expensive simulations conducted with commer-
cial non-linear FE codes (see Fig. 13b).
Arch deformed shape at collapse and power dissipation patch
(in masonry elements and at the interface between FRP and ma-
sonry elements) are reported in Fig. 13 from b to d. Similarly to
the previous case, the effect of the reinforcement is somewhat
evident.
Finally, a sensitivity analysis varying fb masonry/FRP interface
strength from 0 to 3 MPa is conducted. In particular, in Fig. 9, the
ultimate loads so obtained using limit analysis are represented
increasing fb. Also in this case, collapse load reaches asymptotically
a maximum increasing fb, but a satisfactory benefit is obtained
even for relatively low fb values.
Similarly to the barrel rectangular vault, failure mechanism
changes introducing reinforcement (compare Figs. 11 and 13).
From the results, it is possible to argue that the numerical proce-
dure proposed may be of interest for all technicians involved in
 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558 1551

Fig. 20. Hemispherical dome. (a) Comparison among collapse loads provided by experimentation, limit analysis and non-linear FE code. (b) Deformed shapes at collapse from
limit analysis and non-linear FE code. (c) Normalized plastic dissipation patch obtained from limit analysis (left: with near top cupola elements. Right: near top cupola
elements hidden).
1552 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558

strengthening design of masonry structures, because it is able to 4.2. Cross vault

provide a predictive information on the most suitable disposition
to use in building practice to obtain the maximum strength A ribbed cross vault, experimentally tested by Faccio et al. [18],
increase of the structure. formed by the intersection of two barrels vaults with an external

Fig. 21. Reinforced hemispherical dome. (a) Comparison between collapse loads provided by limit analysis and non-linear FE code. (b) Deformed shapes at collapse from limit
analysis and non-linear FE code. (c) Normalized plastic dissipation patch in masonry elements obtained from limit analysis. (d) Normalized plastic dissipation patch on FRP/
masonry interfaces (with the simplifications assumed in the model, plastic dissipation indicates delamination of the strips).
 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558
radius of 2.3 m, is consider as second example, see Fig. 14. Com-
mon Italian bricks of dimensions 120  250  55 mm3 were used,
with joints thickness equal to 10 mm. Masonry homogenized fail-
ure surface obtained by means of the compatible model proposed
in the present work has been already presented in the accompany-
ing paper (Part I) and the reader is referred there for further details.
Numerical results obtained with a macroscopic continuum non-
linear model (similar to that implemented in [21] are also available
– only for the unreinforced case – from [33,34].
In Fig. 15a, load–maximum displacement curves provided by
the continuum macroscopic models (both [34] literature results
and new simulations performed with DIANA [21] are represented),
experimental data and limit analysis collapse load are depicted. In
Fig. 15b, a comparison between deformed shapes near the collapse
provided by limit analysis and DIANA [21] approach are repre-
sented. Similarly to previously analyzed arches, limit analysis re-
sults are in very good agreement with both experimental
evidences and alternative expensive non-linear simulations. Inter-
nal plastic dissipation patch is finally represented in Fig. 15c.
From an analysis of the dissipation patch, it is particularly evi-
dent an out-of-plane sliding of the elements under the zone of
the application of the external load, with the formation of five
bending hinges in the principal arch of the cross vault.
In order to prevent the formation of such plastic hinges, two
strengthening interventions depicted in Fig. 16 are proposed and
their efficiency is numerically analyzed. While in the first rein-
forcement scheme only extrados of the arch near the point of
application of the load is strengthened, in the second one both
the extrados and the intrados are reinforced. In both cases, a ma-
sonry–FRP bond strength fb equal to 0.3 MPa is assumed.
External force–maximum displacements curves obtained for
the reinforced case with DIANA [21] and collapse load provided
by the homogenized limit analysis simulations are reported in Figs.
17 and 18a, respectively for the first and second disposition. The
good agreement between limit analysis and non-linear FE simula-
tions confirms the capabilities of the model proposed.
Vault deformed shape at collapse and power dissipation patch
(in masonry elements and at the interface between FRP and ma-
sonry elements) are finally represented in Fig. 17 (Fig. 18) from b
to d for the first (second) disposition of the strips analyzed. Simi-
larly to previous examples, FRP precludes the formation of inclined
yield lines, with a consequent increase of the failure load. Delami-
Unreinforced case
0 0
N rr [daN/cm]
-5
-10
-15
-20
6.0
N ss eccentricity [cm]
dome thickness
4.0
2.0
0.0
-2.0
-4.0
-6.0
Fig. 22. Hemispherical dome. Internal hooping actions at collapse and eccentricity of meridian actions in the unreinforced (left) and reinforced case (right).
1553
nation of the strips occurs, in agreement to what stated in the CNR
DT 200 [1] code, i.e. starting from the extremes.
In Fig. 9, the ultimate loads obtained with the model proposed
in the presence of FRP and varying fb masonry/FRP interface
strength in a wide range (from 0 to 3 MPa) are represented (sensi-
tivity analysis for FRP first disposition). From the simulations, it
can be stated that optimal reinforcement effect is obtained ranging
fb approximately between 0.3 and 0.6 MPa.
The differences in the failure mechanisms between the rein-
forced and unreinforced case (compare, for instance, Figs. 18 and
15) are worth noting. Finally, it is stressed that, in general, high val-
ues of bond strength do not correspond to a proportional increase
of the failure load (if compared to moderately low values of fb, see
Fig. 9).
4.3. Hemispherical dome
The third example here analyzed relies on a hemispherical
dome with inner diameter equal to 2.2 m and thickness of 12 cm
and experimentally tested by Foraboschi [20].
Common Italian bricks of dimensions 120  250  55 mm3
were used, with joints thickness approximately equal to 10 mm.
Mechanical properties assumed for joints and bricks are reported
in Part I and the reader is referred there for further details. In
Fig. 19, the geometry and the loading condition are represented.
The dome is loaded until failure by means of a concentrated verti-
cal increasing load applied at the top of the structure. A steel plate
with dimensions reported in Fig. 19 is placed between the load and
the external loaded surface in order to diffuse vertical stresses.
In Fig. 20a, load–maximum displacement curves (both a heter-
ogeneous and a macroscopic approach have been used) provided
by the non-linear model [21], experimental data and limit analysis
collapse load are depicted. In Fig. 20b, a comparison between de-
formed shapes near the collapse obtained by means of limit anal-
ysis and DIANA [21] are represented. Considerations similar to
the previous examples can be repeated here. Also in this case, in
fact, limit analysis deformed shape is very similar to that obtained
with alternative expensive non-linear models, but optimization re-
quires only a small fraction of processing time needed by commer-
cial FEM in the inelastic range.
Internal plastic dissipation patch is finally represented in
Fig. 20c. As can be noted, internal dissipation is concentrated along
FRP reinforced case
N rr [daN/cm]
-5
-10
-15
-20
6.0
N ss eccentricity [cm]
dome thickness
4.0
2.0
0.0
-2.0
-4.0
-6.0
1554 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558
a circular crown, with the formation of one annular bending hinge.
Moreover, an adding amount appears along the meridians of the
dome, essentially due to non-null membrane annular actions. In
fact, evident openings along meridians can be observed in the de-
formed shape.
In order to prevent the formation of meridian cracks, the
strengthening intervention depicted in Fig. 19 is proposed and its
efficiency is numerically evaluated. A masonry–FRP bond strength
fb equal to 0.3 MPa is assumed.
Fig. 23. Cloister vault. Geometry, loading condition and FE discretization adopted
for the numerical analyses of the arch.
External force–maximum displacement curve obtained for the
reinforced case with a continuous macroscopic model [21] and col-
lapse load provided by the homogenized limit analysis simulations
are reported in Fig. 21a. As can be noted, limit analysis simulations
are in good agreement with non-linear FE results.
Dome deformed shape at collapse and power dissipation patch
(in masonry elements and at the interface between FRP and ma-
sonry elements) are finally depicted in Fig. 21 from b to d. Similarly
to previous example, FRP effect is rather evident. The formation of
meridian cracks is, in this case, partially precluded by the hooping
strips, with a consequent significant increase of the failure load in
presence of fibers. Furthermore, it is interesting to notice from the
deformed shape at collapse that, in the presence of reinforcement,
failure is more constrained to localize under the zone of applica-
tion of the external load, with an evident out-of-plane sliding of
the loaded area.
In order to evaluate also internal actions at collapse for the
problem at hand, associated to the failure mechanism found with
the upper bound approach here presented, the dual problem of
(24) has to be solved:
8
>
> minfPT0 ½w UT þ Pin;ass
I k_ I;ass g
>
> 8 8
>
> T
>
>
< > > maxf^kg
> HT ½w UT þ Ain k_ I;ass ¼ 0 >
< < (
>
such that RT ½w UT ¼ 1 HR þ R^k  P0 ¼ 0
>
> >
> >
> such that
>
> : : T
>
> k_I;ass
P0 Ain R 6 Pin;ass
I
>
:
primal dual
ð25Þ
where R collects elements membrane actions and moments of each
element and ^ k is the collapse multiplier.
In Fig. 22, the radial membrane actions distributions (Nrr) and
the meridian actions eccentricity obtained from the dual problem
are represented in absence (left) and presence (right) of FRP. As
it is possible to notice, in both cases collapse occurs for the forma-
tion of a hinge along the parallels. However, when FRP is intro-
duced, radial membrane action absolute value tends to increase,
especially near the top of the dome, thus increasing also meridian
strength. Furthermore, while in the unreinforced case, annular
hinge is located between fifth and sixth row of elements, in the
reinforced case it is observed between third and fifth row. An
out-of-plane sliding between the second and the third row can
be also distinguished from the deformed shape.
In Fig. 9, results in terms of collapse load from a sensitivity anal-
ysis conducted varying fb masonry/FRP interface strength in a wide
range (from 0 to 3 MPa) are finally represented. The same consid-
erations reported in the previous sub-sections can be repeated
here, i.e. optimal reinforcement effect is obtained assuming fb rang-
ing between 0.3 and 0.6 MPa.
4.4. Cloister vault
The last example relies in the limit analysis of a square cloister
vault of side 2 m.
The cloister vault is represented in Fig. 23 and was tested until
collapse by Foraboschi [19,20] both in absence and presence of
reinforcement. The FRP strips disposition arranged in the experi-
mental test is slightly different to that sketched in Fig. 23, but
unfortunately no information is available from [19,20] on both
geometrical details of the reinforcement and the failure load in-
crease obtained in the reinforced case. For these reasons, here we
compare limit analysis model only with results provided by the
DIANA [21] non-linear FE code. The vault is loaded until failure
with a concentrated vertical force applied at the top, as illustrated
in Fig. 23. A large steel beam with suitable dimensions is placed
between the concentrated force and the external loaded surface
 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558 1555

in order to prevent premature local failures due to stress
concentration.
In Fig. 24a, load–maximum displacement curves provided by
[21] models (both a heterogeneous and a continuum orthotropic
macroscopic approach were used), experimental data and limit
analysis collapse load are depicted for the unreinforced vault. In
Fig. 24b, a comparison among deformed shapes near the collapse
provided by limit analysis and DIANA [21] models are represented.
Also in this case, limit analysis failure mechanism is very close to
that obtained by means of an alternative expensive numerical
approach.
Internal plastic dissipation patch is finally represented in
Fig. 24c. As in the previous example, out-of-plane masonry
strength is crucial for the overall behavior of the vault at failure.
In particular, the formation of four horizontal cylindrical hinges
approximately at 2/3 of the height of the structure and parallel
to the edges is rather evident.
In order to prevent the formation of such hinges near the top of
the vault, a crossed strip disposition with annular ties, as illus-
trated in Fig. 23, is numerically analyzed. In the simulations, a ma-
sonry–FRP bond strength fb equal to 0.3 MPa is assumed.
External force–maximum displacement curve obtained for the
reinforced case with DIANA [21] – only the macroscopic model is
tested in this case – and collapse load provided by the homoge-
nized limit analysis are reported in Fig. 25a.
Vault deformed shape at collapse and power dissipation patch
(in masonry elements and at the interface between FRP and ma-
sonry elements) are finally depicted in Fig. 25 from b to d. Similarly
to previous examples, the disposition of strips analyzed tends to
preclude both the formation of cylindrical hinges and the local

Fig. 24. Cloister vault. (a) Comparison among collapse loads provided by experimentation, limit analysis and non-linear FE code. (b) Deformed shapes at collapse from limit
analysis and non-linear FE code. (c) Normalized plastic dissipation patch obtained from limit analysis.
1556 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558

out-of-plane sliding near the zone of application of the external Finally, the increase of ultimate loads in presence of FRP
load. Dissipation at FRP/masonry interfaces is, in this case, rather and varying fb masonry/FRP interface strength in a wide range
diffused, meaning that the strengthening disposition tested is very (sensitivity analysis) is represented in Fig. 9 for the sake of
effective. completeness.

Fig. 25. Reinforced cloister vault. (a) Comparison between collapse loads provided by limit analysis and non-linear FE code. (b) Deformed shapes at collapse from limit
analysis and non-linear FE code. (c) Normalized plastic dissipation patch in masonry elements obtained from limit analysis. (d) Normalized plastic dissipation patch on FRP/
masonry interfaces (with the simplifications assumed in the model, plastic dissipation indicates delamination of the strips).
 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558
5. Conclusions
A homogenized FE limit analysis approach for the numerical
evaluation of collapse loads and failure mechanisms of FRP–rein-
forced masonry curved structures has been presented.
The approach consists of two steps. In step I, described exten-
sively in the companying paper (Part I), unreinforced masonry
strength domains have been obtained with a FE limit analysis pro-
cedure applied to a representative element of volume constituted
by a central brick interacting with its six neighbors by means of
rigid plastic interfaces (mortar joint). In step II, described in the
present paper, FRP strips have been applied on the already homog-
enized material. Six-noded rigid infinitely strong elements with
possible plastic dissipation at the rectangular interfaces between
adjoining elements have been utilized to model masonry, whereas
three-noded rigid infinitely resistant elements have been used for
FRP strips. Unreinforced masonry homogenized failure surfaces ob-
tained in Part I of the present paper have been used to evaluate
plastic dissipation at the interfaces between adjoining wedge-
shaped elements. A possible dissipation at the triangular interfaces
between FRP and masonry elements has been also considered in
order to model, in an approximate but effective way, the possible
delamination of the strips from the supports. Italian code CNR DT
200 [1] formulas have been used to evaluate peak interface tangen-
tial strength.
Several numerical examples have been analyzed, consisting of
two different typologies of masonry arches (a barrel vault and an
arch in a so-called ‘‘skew” disposition), a ribbed cross vault, a
hemispherical dome and a cloister vault. For all the cases, both
the unreinforced and FRP reinforced case have been discussed.
Additional non-linear FE analyses have been conduced for all the
examples presented, modeling masonry through both a heteroge-
neous and an equivalent macroscopic material with orthotropic
behavior, in order to assess limit analysis results. Comparisons
with experimental evidences, where available, have been finally
reported.
Acknowledgements
A. Tralli and E. Milani gratefully acknowledge the financial sup-
port of the research project MIUR COFIN 2007 (‘‘Multiscale prob-
lems with complex interaction in structural engineering”,
coordinator Prof. A. Corigliano).
References
[1] CNR-DT 200. Guide for the design and construction of externally bonded FRP
systems for strengthening existing structures. C.N.R., National Research
Council, Italy; 2006.
[2] Korany Y, Drysdale R. Load–displacement of masonry panels with unbonded
and intermittently bonded FRP. I: analytical model. J Compos Constr
2007;11(1):15–23.
[3] Milani G. Homogenized limit analysis of FRP–reinforced masonry walls out-of-
plane loaded. Comput Mech 2009;43(5):617–39.
[4] Heyman J. The safety of masonry arches. Int J Mech Sci 1969;43:209–24.
[5] Heyman J. Equilibrium of shell structures. Oxford: Oxford University Press;
1977.
[6] Sinha BP. A simplified ultimate load analysis of laterally loaded model
orthotropic brickwork panels of low tensile strength. J Struct Eng ASCE
1978;56B(4):81–4.
[7] Ferris M, Tin-Loi F. Limit analysis of frictional block assemblies as a
mathematical program with complementarity constraints. Int J Mech Sci
2001;43:209–24.
[8] Orduña A, Lourenço PB. Three-dimensional limit analysis of rigid blocks
assemblages. Part I: torsion failure on frictional joints and limit analysis
formulation. Int J Solids Struct 2005;42(18–19):5140–60.
[9] Milani G, Lourenço PB, Tralli A. Homogenised limit analysis of masonry walls.
Part I: failure surfaces. Comp Struct 2006;84:166–80.
[10] Livesley RK. Limit analysis of structures formed from rigid blocks. Int J Numer
Meth Eng 1978;12:1853–71.
[11] O’Dwyer D. Funicular analysis of masonry vaults. Comp Struct 1999;73(1–
5):187–97.
1557
[12] Oppenheim IJ, Gunaratnam DJ, Allen RH. Limit state analysis of masonry
domes. J Struct Eng ASCE 1989;115:868–82.
[13] Milani E, Milani G, Tralli A. Limit analysis of masonry vaults by means of
curved shell finite elements and homogenization. Int J Solids Struct
2008;45:5258–88.
[14] Caporale A, Luciano R, Rosati L. Limit analysis of masonry arches with
externally bonded FRP reinforcements. Comp Meth Appl Mech Eng
2006;196:247–60.
[15] Roca P, Lopez-Almansa F, Miquel J, Hanganu A. Limit analysis of reinforced
masonry vaults. Eng Struct 2007;29:431–9.
[16] Luciano R, Sacco E. Damage of masonry panels reinforced by FRP sheets. Int J
Solids Struct 1998;35(15):1723–41.
[17] Vermeltfoort AV. Analysis and experiments of masonry arches. In: Lourenço
PB, Roca P, editors, Proc. historical constructions. Guimarães PT; 2001.
[18] Faccio P, Foraboschi P, Siviero E. Masonry vaults reinforced with FPR strips [In
Italian: Volte in muratura con rinforzi in FRP]. L’Edilizia 1999;7/8:44–50.
[19] Foraboschi P. Strengthening of masonry arches with fiber-reinforced polymer
strips. J Compos Constr 2004;8:191–202.
[20] Foraboschi P. Masonry structures externally reinforced with FRP strips: tests at
the collapse. In: Proc I Convegno Nazionale ‘‘Sperimentazioni su Materiali e
Strutture”, Venice; 2006 [in Italian].
[21] DIANA 9.3 version user’s manual. TNO Building and Construction Research,
Department of Computational Mechanics, Delft, The Netherlands; 2008.
[22] Krabbenhoft K, Lyamin AV, Hjiaj M, Sloan SW. A new discontinuous upper
bound limit analysis formulation. Int J Numer Meth Eng 2005;63:
1069–88.
[23] Zienkiewicz OC, Taylor RL. The finite element method. Vol. 1. Basic
formulations and linear problems. London: McGraw-Hill; 1989.
[24] Ceroni F, Pecce MR, Manfredi G, Marcari G. Experimental bond behaviour in
masonry elements externally reinforced with FRP laminates. In: Proc of
International Conference Composites in Constructions, Cosenza, Italy; 2003.
[25] JSCE, Japan Society of Civil Engineers. Recommendations for upgrading of
concrete structures with use of continuous fiber sheets. Concrete Engineering
Series 41; 2001.
[26] D.M.LL.PP. Norme tecniche per la progettazione, esecuzione e collaudo degli
edifici in muratura e per il loro consolidamento [Technical norms for the
design, execution and test of masonry buildings and for their rehabilitation],
Italy; 1987.
[27] Marfia S, Sacco E. Modelling of reinforced masonry elements. Int J Solids Struct
2001;38:4177–98.
[29] Sloan SW, Kleeman PW. Upper bound limit analysis using discontinuous
velocity fields. Comp Meth Appl Mech Eng 1995;127(1–4):293–314.
[30] Makrodimopoulos A, Martin CM. Lower bound limit analysis of cohesive-
frictional materials using second-order cone programming. Int J Numer Meth
Eng 2006;66(4):604–34.
[31] Krabbenhoft K, Lyamin AV, Sloan SW. Formulation and solution of some
plasticity problems as conic programs. Int J Solids Struct 2007;44:1533–49.
[32] Lourenço PB, de Borst R, Rots J. A plane stress softening plasticity model for
orthotropic materials. Int J Numer Meth Eng 1997;40:4033–57.
[33] Creazza G, Saetta A, Matteazzi R, Vitaliani R. Analyses of masonry vaults: a
macro approach based on three-dimensional damage model. J Struct Eng
2002;128(5):646–54.
[34] Creazza G, Saetta A, Matteazzi R, Vitaliani R. Analyses of masonry vaulted
structures by using a 3-D damage model. In: European congress on
computational methods in applied sciences and engineering, ECCOMAS
2000, Barcelona, SP; 2000.
[35] Gilbert M, Melbourne C. Rigid-block analysis to masonry arches. Struct Eng
1994;72:356–61.
[36] Lucchesi M, Padovani C, Pasquinelli G, Zani N. On the collapse of masonry
arches. Meccanica 1997;32:327–46.
[37] Olivito RS, Stumpo P. Analisi numerico-sperimentale di strutture murarie
voltate soggette a carichi mobili. In: Proc XXXI AIAS congress, Parma, Italy;
2002.
[38] Baratta A, Zuccaro G, Binetti A. Strength capacity of a No-Tension portal arch-
frame under combined seismic and ash loads. J Volcanol Geothermal Res
2004;133(1–4):369–76.
[39] Drosopoulos GA, Stavroulakis GE, Massalas CV. Limit analysis of a single span
masonry bridge with unilateral frictional contact interfaces. Eng Struct
2006;28(13):1864–73.
[40] Betti M, Drosopoulos GA, Stavroulakis GE. Two non-linear finite element
models developed for the assessment of failure of masonry arches. C.R.
Mecanique 2008;336:42–53.
[41] Audenaert A, Fanning P, Sobczak L, Peremans H. 2-D analysis of arch
bridges using an elasto-plastic material model. Eng Struct 2008;30:
845–55.
[42] Como M, Ianniruberto U, Imbimbo M. Strength of masonry arches reinforced
with FRP sheets [La resistenza degli archi murari rinforzati con fogli in FRP]. In:
Proc Conv Naz Meccanica delle Strutture in Muratura Rinforzate con FRP-
materials, Venice; 2000. p. 155–64.
[43] Orduña A, Lourenço PB. Cap model for limit analysis and strengthening of
masonry structures. J Struct Eng 2003;129(10):1367–75.
[44] Drosopoulos GA, Stavroulakis GE, Massalas CV. FRP reinforcement of stone
arch bridges: unilateral contact models and limit analysis. Compos Part B: Eng
2007;38(2):144–51.
[45] Baratta A, Corbi O. Stress analysis of masonry vaults and static efficacy of FRP
repairs. Int J Solids Struct 2007;44(24):8028–56.
1558 G. Milani et al. / Computers and Structures 87 (2009) 1534–1558

[46] Focacci F. Rinforzo delle murature con materiali compositi [Masonry
strenghtening with composite materials]. Flaccovio; 2008.
[47] Lucchesi M, Padovani C, Pasquinelli G, Zani N. The maximum modulus
eccentricity surface for masonry vaults and limit analysis. Math Mech Solids
1999;4:71–87.
[48] D’Ayala D, Casapulla C. Limit state analysis of hemispherical domes with finite
friction. In: Lourenço PB, Roca P, editors. Proc historical constructions,
Guimarães PT; 2001.
[49] Huerta S. Mechanics of masonry vaults: the equilibrium approach. In:
Lourenço PB, Roca P, editors, Proc historical constructions, Guimarães PT;
2001.
[50] Block P, Ciblac T, Ochsendorf J. Real-time limit analysis of vaulted masonry
buildings. Comp Struct 2006;84(29–30):1841–52.
[51] De Lorenzis L, Dimitri R, La Tegola A. Reduction of the lateral thrust of masonry
arches and vaults with FRP composites. Constr Build Mater
2007;21(7):1415–30.