You are on page 1of 12

Engineering Structures 40 (2012) 327338

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

A new discrete element model for the evaluation of the seismic behaviour
of unreinforced masonry buildings
Ivo Cali , Massimo Marletta, Bartolomeo Pant
Department of Civil and Environmental Engineering, University of Catania, Italy

a r t i c l e

i n f o

Article history:
Received 27 May 2011
Revised 20 October 2011
Accepted 5 February 2012
Available online 31 March 2012
Keywords:
Masonry building
Unreinforced Masonry structures
In-plane masonry response
Macro-element approach
Discrete element model
Seismic vulnerability
Nonlinear masonry analysis

a b s t r a c t
The evaluation of the nonlinear seismic response of masonry buildings represents a subject of considerable importance whose resolution is nowadays a main research topic in earthquake engineering. Rened
nonlinear nite element models require a huge computational cost that makes these methods unsuitable
for practical application. In this paper an innovative discrete-element model, conceived for the simulation
of the in-plane behaviour of masonry buildings, is presented. The basic idea of the proposed approach is
to approximate the in-plane nonlinear response of masonry walls by an equivalent discrete element. This
element is able to reproduce the typical in-plane collapse behaviour of a masonry wall subjected to earthquake loading. The reliability of the proposed approach has been evaluated by means of nonlinear incremental static analyses performed on masonry structures, for which theoretical and/or experimental
results are available in the literature. The proposed computational strategy provides a relatively simple
and practical tool which could be of signicant value for the design and the vulnerability assessment
of unreinforced masonry structures in seismic areas.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
The assessment of the nonlinear seismic behaviour of UnReinforced Masonry buildings (URM) represents a subject of great
importance, however it is rather difcult to solve. In order to estimate the seismic vulnerability of an existing building and to ascertain if the structure requires a seismic upgrade, a structural
engineer needs simple and efcient numerical tools whose complexity and computational demand must be appropriate for the
practical engineering purposes. However, the simulation of the
nonlinear dynamic behaviour of a masonry building represents a
challenging problem which rigorously requires the use of computationally expensive nonlinear nite element models and, above
all, expert judgment. For this reason, the seismic assessment and
subsequent rehabilitation of URM constitutes a contemporary issue in most seismic regions in which the historical masonry buildings represents a signicant number of the existing structures.
Even in the USA, the need of introducing practical approaches for
the evaluation of the seismic behaviour of URM has been recently
highlighted [13].
The different behaviour of masonry structures, compared to ordinary concrete and steel buildings, requires ad hoc algorithms
Corresponding author. Address: Department of Civil and Environmental Engineering, University of Catania, Viale Andrea Doria 6, 95125 Catania, Italy. Tel.: +39
(0)95 738 2255; fax: +39 (0)95 738 2249.
E-mail address: icalio@dica.unict.it (I. Cali).
0141-0296/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2012.02.039

capable of reproducing the nonlinear behaviour of masonry media


and providing reliable numerical simulations. Rened nite element numerical models, such as the smeared cracked and discrete
crack nite element models [4,5], able to predict the complex nonlinear dynamic mechanical behaviour and the degradation of the
masonry media, require sophisticated constitutive laws and a huge
computational cost. As a consequence these methods are nowadays not suitable for practical application and extremely difcult
to apply to large structures. An alternative approach to the nonlinear FEM is represented by the rigid-body spring models (RBSM),
specically formulated with the aim of approximating the macroscopic behaviour of masonry walls with reduced degrees of freedom. Some valuable application of this approach are relative to
historical masonry buildings [6,7]. However, the difculties in
the model implementation make the RBSM rarely used in the engineering practice. A comprehensive review in the current development on numerical issues on masonry mechanics is reported in
[8]. In the latter work it is also highlighted that a complex analysis
tool does not necessarily provide better results than a simple tool.
In the last three decades, many authors have developed simplied or alternative methodologies that, with a reduced computational effort, should be able to predict the nonlinear seismic
behaviour of masonry buildings and to provide reliable numerical
results for engineering practice purposes. The most commonly
used practical approach for the analysis of URM is the so called
equivalent frame model, in which the masonry building is represented by an equivalent nonlinear frame structure constituted by

328

I. Cali et al. / Engineering Structures 40 (2012) 327338

(a)

(b)

Fig. 1. The basic macro-element: (a) undeformed conguration; and (b) deformed conguration.

(a)
F

(b)
F

(c)

Fig. 2. Main in-plane failure mechanisms of a masonry portion: (a) exural failure; (b) shear-diagonal failure; and (c) shear-sliding failure.

nonlinear beam elements and rigid offsets [9,10]. According to the


equivalent frame model approach, each wall of the building is subdivided into macro-elements representing piers, spandrels and rigid zones. The nonlinear behaviour of piers and spandrels are
simulated by nonlinear frame element, while the rigid zones, in
which the damage cannot occur, are substituted by rigid offsets.
The substitution of a masonry portion with a frame element has
some restrictions due to the inaccurate simulation of the interaction between macro-elements, to the difculties that arise for complex geometry and to the weak modelling of the cracked condition
of panels. For these reasons some authors proposed the use of twodimensional macro-elements, DAsdia and Viskovic [11] model the
masonry by linear elastic macro-nite elements. Namely, the nonlinearity is simulated by updating the geometry of the element
excluding the volumes undergoing tensile stresses, or cracked, or
generally no more co-operating. A similar approach has been proposed by Braga et al. [12] that introduced the so-called multi-fan
panel element in which the stress eld of the panel is assumed
to follow a multi-fan pattern, the material behaviour is assumed
linear elastic in compression and non reacting in tension. More recently Vanin and Foraboschi [13] proposed a new approach based
on modelling masonry panels by means a strut and tie model.
Comparative evaluations of commonly used techniques for practical analysis of realistic URM structures subjected to seismic loadings are reported in the studies of Loureno [8], Seible and
Kingsley [5], Magenes and La Fontana [9] and Kappos et al. [10].
An overview on recent code developments and state-of-the-art
methods of earthquake resistant design of masonry buildings is
reported in [14] where the experimental results are also used in
order to justify the analysed numerical approaches.
In this paper a new modelling approach for the simulation of the
seismic behaviour of masonry buildings, suitable for current engineering practice applications, is presented. The proposed approach

is based on the concept of macro-element discretization [8] and has


been conceived with the aim of capturing the nonlinear behaviour
of an entire masonry wall and of the entire building, as an assemblage of several walls. The model is based on a plane nonlinear discrete element, able to simulate the behaviour of masonry wall in its
own plane. The basic macro-element consists of an articulated
quadrilateral with rigid edges in which two diagonal springs govern
the shear behaviour. The exural and sliding shear behaviour is
governed by discrete distributions of springs in the sides of the
quadrilateral that preside over the interaction with the adjacent
macro-elements. The calibration of the model require only a few
parameters to dene the masonry material based on results from
current experimental tests. The computational cost of the proposed
numerical approach is greatly reduced, compared to a traditional
nonlinear nite element modelling. Since the equivalence between
the masonry portion and the macro-element is based on very
simple physical considerations, the interpretation of the numerical
results is simple and straightforward. This novelty approach is intended as a tool, which requires low computational resources, for
investigating the nonlinear behaviour of URM buildings.
In the paper the efciency of the computational tool is shown by
means of nonlinear static analyses, carried out on case studies of
masonry walls and structures that have been object of theoretical
and experimental researches. The inuence of the discretization
of the model associated to the size of the mesh of macro-element
and the ne-tuning of the nonlinear links have been also
investigated.

2. The basic discrete-element


The basic element of the proposed approach has a simple
mechanical scheme, Fig. 1. It is represented by an articulated

329

I. Cali et al. / Engineering Structures 40 (2012) 327338

(a)

q
F

(b)

(c)

Fig. 3. Simulation of the main in-plane failure mechanisms of a masonry portion by means of the macro-element: (a) exural failure; (b) shear-diagonal failure; and (c) shearsliding failure.

quadrilateral constituted by four rigid edges connected by four


hinges and two diagonal nonlinear springs. Each side of the panel
can interact with other panels or elements or supports by means
of a discrete distribution of nonlinear springs, denoted as interface.
Each interface is constituted by n nonlinear springs, orthogonal to
the panel side, and an additional longitudinal spring which controls
the relative motion in the direction of the panel edge. In spite of its
great simplicity, such a basic mechanical scheme is able to simulate the main in-plane failures of a portion of masonry wall subjected to horizontal and vertical loads. These well-known
collapse mechanisms, namely the exural failure, the diagonal shear
failure and sliding shear failure, are approximately represented in
Fig. 2 where the typical crack patterns together with the qualitative kinematics of masonry portion are also sketched. Fig. 3 shows
how the proposed macro-element allows a simple and realistic
mechanical simulation of the corresponding failure mechanisms.
The exural failure mode, sketched in Fig. 2a, is associated to the
rocking of the masonry portion in its own plane. The loss of the
bearing capacity is related to the progressive rupture of the panel
in the tensile zone and/or to the crushing of the panel in the compressive zone. The latter mechanism is reproduced by the considered macro-element as shown in Fig. 3a. The axial and exural
deformability of the masonry panel is concentrated in the orthogonal nonlinear links according to a simple ber model discretization, as is usually adopted for reinforced concrete structures.
The diagonal-shear failure mode, drafted in Fig. 2b, is associated
to the loss of the bearing capacity of the masonry panel due to
excessive shear and to the consequent formation of diagonal cracks
along the directions of the principal compression stresses. Through
the proposed discrete-element this collapse mechanism can be
governed by means of the two diagonal nonlinear springs which
have the role to simulate and predict the nonlinear shear response
of the modelled masonry portion, Fig. 3b.
The sliding-shear failure mode is associated to the sliding of the
masonry panel in its own plane. In this case the loss of the bearing
capacity is associated to the formation of cracks parallel to the bedjoints (Fig. 2c). This failure mode, which generally occurs for low
levels of vertical loads and low values of friction coefcients, can
be controlled by the longitudinal nonlinear springs of the interfaces, Fig. 3c.
In the following paragraphs, the kinematics and the mechanical
characteristics of the basic element are outlined. Consequently, a
simple procedure, based on a ber modelling approach, for the
calibration of the mechanical parameters of the nonlinear links of
the discrete element, is presented.
2.1. The kinematics
According to the proposed discrete element approach, a masonry macro-portion is assimilated to an equivalent mechanical

scheme in which the physical role of each component is simple


and unambiguous. This feature is not common to all the simplied
models proposed in the literature and is noteworthy since it makes
the direct implementation of the model in many general-purpose
nite element platforms possible. As mentioned in the sequel,
the model can be also improved by using a rened mesh of elements in order to obtain a more detailed description of the kinematics of the structure and the collapse mechanism.
The degrees-of-freedom of the structural scheme are those
associated to the in-plane motion of the basic elements. Each element exhibits three degrees-of-freedom, associated to the in-plane
rigid-body motion, plus a further degree-of-freedom needed for
the description of the shear deformability. In Fig. 4 the chosen
Lagrangian parameters for the panel and the corresponding
deformed congurations are reported. The deformations of the
interfaces are associated to the relative motion between corresponding panels, therefore no further Lagrangian parameters must
be introduced in order to describe their kinematics. Each interface
can be identied by two nodes, i and j. Each node actually corresponds to the joints of the model belonging to the panels which
are connected by the interface.
Since the proposed modelling approach is suitable for describing the behaviour of a plane wall loaded in its own plane, a
three-dimensional masonry building can be modelled as an assemblage of plane walls. Therefore, in this simplied approach the
behaviour of the wall in the out-of-plane direction is not considered. The latter limitation is common to all the proposed simplied
approaches. The inclusion of the out-of-plane behaviour of the masonry wall in the proposed model constitutes a further development which, at present time, is a work in progress.
In Fig. 5 it is shown how a simple masonry wall can be modelled
by means of the proposed modelling approach. Namely, Fig. 5a
refers to a basic scheme composed by 12 panels and is characterised by 48 degrees of freedom, while in Fig. 5b the same wall is

(a)

u3

u4

(b)
u1

u3

u2

u4

u2

u1
Fig. 4. (a) Chosen Lagrangian parameters for the quadrilater; and (b) deformed
congurations corresponding to the activation of each degree-of-freedom.

330

I. Cali et al. / Engineering Structures 40 (2012) 327338

(a)

(b)
void

void
void

void

Fig. 5. Masonry wall and corresponding macro-element discretizations with different mesh resolutions.

2.2.1. Calibration of the interface orthogonal springs


Since the masonry is considered as a homogeneous medium its
global behaviour should be ascribed to the exural and shearing
characteristics of a nite portion of an orthotropic inelastic continua. As mentioned before, the exural behaviour is simulated
by the interface orthogonal springs connecting the panel to adjacent elements. Each spring is here calibrated by adopting a specic
constitutive law for the masonry media, according to a simplied
ber model approach. In the description of the calibration procedure an elasto-plastic behaviour with limited deformability is assumed. The orthotropic nature of the masonry can be simply
considered by calibrating separately the horizontal and vertical
interfaces according to the mechanical properties of the corresponding directions.
The case regarding a horizontal interface connecting two generic panels, k and l, characterised by different mechanical and
geometrical properties, is represented in Fig. 6. The vertical orthogonal springs have the role of simulating the deformability of the
two panels in the vertical direction. The stiffness calibration of

modelled through a more rened mesh composed by 48 quadrilaterals which require 192 degrees of freedoms. The use of a more rened mesh is not mandatory however in some cases can provide
more accurate results and a better description of the collapse
mechanism.

2.2. Calibration of the nonlinear springs


It has been shown that the kinematics of the proposed model is
able to simulate the main failure mechanisms of a masonry element. However the effectiveness of the simulation of the nonlinear
behaviour relies on a suitable choice of the mechanical parameters
of the model inferred by an equivalence between the masonry wall
and a reference continuous model characterised by simple but reliable constitutive laws. This equivalence relies on a straightforward
ber calibration procedure, and is based only on the main mechanical parameters which characterise the masonry according to an
orthotropic homogeneous medium.
influence area

Panel k

spring 1

F
Fty1

u cy1

u cy1

L1

u ty1

L 1 /2
Fcy1

zero
thickness

L 2 /2

single spring
equivalent to springs
1 and 2 in series

F
Fty2

spring 2
u cy2

Panel k

u tu1

u cy2

u
u ty2

u tu2

Panel l

L2
Fcy2

Panel l

influence area

Fig. 6. Calibration of the interface springs.

I. Cali et al. / Engineering Structures 40 (2012) 327338

each panel is simply obtained by assigning to each link the axial


rigidity of the corresponding masonry strip. The masonry strip is
identied by the inuence area of the spring and the half-dimension of the panel in the direction perpendicular to the interface.
This procedure provides couples of springs in series. Each couple
of springs is then replaced by a single resulting nonlinear elastoplastic spring as described in what follows.
With reference to a single panel, according to the above-described procedure, the initial stiffness Kp, the compression and tensile yielding strengths Fcy, Fty and the corresponding ultimate
displacements ucu, utu are derived by the knowledge of the
mechanical material properties of the masonry as follows:

Eks
Kp 2
L
F cy skrc ;
ucu

L
ecu ;
2

1
F ty skrt

2a; b

L
etu
2

3a; b

utu

where E is the Youngs modulus in the direction orthogonal to the


interface; rc and rt are the compressive and tensile yielding stresses, ecu and etu are the ultimate compressive and tensile strains; s is
the thickness of the represented masonry and k is the distance between two nonlinear link.
Each couple of elasto-plastic springs in series, once the mechanical properties have been dened, can be replaced by a single
resulting nonlinear spring, Fig. 6, dened by appropriate stiffness,
yielding strengths and ultimate displacements. The substitution
of two nonlinear springs in series with a nonlinear single spring
leads to a reduction of the computational cost.
In the initial elastic range the resulting stiffness of the two
springs in series, of stiffness Kp1 and Kp2 is as follows:

K p1  K p2
;
K p1 K p2

while the yielding strengths of the resultant spring are given by the
minima between the yielding strengths of the two springs in series.
Considering two springs corresponding to panels of length L1 and L2
and characterised by ultimate deformation ecu1 < ecu2 in compression and etu1 < etu2 in tension, the ultimate displacements of the
combined resultant spring is simply obtained by adding the ultimate displacement of the yielded spring to the displacement of
the other spring that behaves elastically. These displacements in
compression and in tension can therefore be expressed as:

L1
F cy1
 ecu1
2
K c2
L1
F ty1
U tu  etu1
;
2
K t2
U cu

where Fcy1 and Fty1 are the compressive and tensile yielding forces
of the yielded spring and Kc2 and Kt2 are the corresponding stiffnesses of the elastic spring.
The rupture behaviour of the panel is dealt with by different criteria in compression and in tension. Precisely, once the compressive ultimate displacement is reached, the spring is removed
from the model and the relevant reaction is applied as an external
force loading the corresponding panel. On the other hand, when
the tensile limit displacement is attained, although the reaction
is again re-applied to the panel, the link is not removed from the
model, since it will be able to bear further compressive loads once
the contact with the corresponding panel will be restored.
The post-yielding behaviour of each orthogonal spring can be
characterised according to different constitutive plastic models
according to the experimental results obtained at the macro-scale,
as shown in the numerical applications.

331

2.2.2. Calibration of the sliding springs of the interface


The longitudinal spring governs the sliding-shear failure by
considering the potential sliding between two adjacent elements.
The characteristics of this spring depend on the effective contact
surface between the adjacent elements. When the contact zone between the elements reduces to zero the sliding spring is no more
active. The sliding springs have been modelled by means of a rigid-plastic constitutive behaviour governed by a MohrCoulomb
yielding surface, in particular sliding occurs when the force in
the nonlinear link reaches its limit value, Flim, that is given by
the expression:

F lim c lrm Ao

in which c is a cohesion parameter and l is the friction coefcient


that can be attributed to the masonry, in the considered direction,
in order to control the sliding-shear failure mechanism; rm is the
current average value of the compressive stresses acting on the
interface and Ao is the effective contact area of the two adjacent
panels.
2.2.3. Calibration of the diagonal springs of the panel
In view of typical mechanical properties of masonry material
and the geometry of structural walls, shear failure is the most common type of failure of masonry wall subjected to seismic action in
its own plane. The diagonal springs have the fundamental role to
preside over the diagonal shear failure collapse. This is associated
to the loss of the bearing capacity of masonry panel due to excessive shear and the consequent formation of diagonal cracks along
the direction of the principal compressive stresses. Although the
use of two springs it is not compulsory, it allows a better interpretation of the physical response providing an improved model for
the shear-diagonal failure, since a different state of damage can
be associated to each spring depending on the direction and versus
of the shear force. The shear failure of URM piers associated with
diagonal cracking is difcult to describe with a simple expression
[15]. Many different yielding criteria can be adopted [16,17] all
accounting for a shear resistance strongly dependent on vertical
compression stresses in the wall. As highlighted in reference
[14], two basically different hypotheses, which lead virtually to
the same results, have been developed in order to model the shear
failure mechanism. In the rst case, which has been accepted by
Eurocode 6 (Design of masonry structures) [18] the shear strength
is dened according to a Mohr Coulomb law:

fv fv o lc rn

where fvo is the shear strength under zero compression strength, lc


is a friction coefcient dening the contribution of compressive
stresses, rn is the value of the compressive stress. Values of fvo
and lc should be determined by experimental test [16,19]. This criterion is the same suggested for the description of the sliding-shear
failure, although characterised by appropriate values of fvo and lc
that in general do not coincide with the corresponding values that
govern the sliding-shear failure. In the proposed macro-modelling
approach the two mechanisms are controlled separately.
Alternative theories associate the diagonal shear failure to the
principle tensile stresses which develop in the wall when subjected
to vertical loads and increasing horizontal forces. The most
adopted criterion based on this assumption is the well known
Turnsek and Cacovic criterion [16] in its modied form [17] that
takes into account the inuence of the geometry of the wall and
the distribution of action at maximum resistance. The latter criterion can be expressed as

fv

ft
b

r
ro
1
ft

332

I. Cali et al. / Engineering Structures 40 (2012) 327338

[20,21]. In the applications reported in the following displacement


controlled nonlinear static analyses have been performed. Further
details on the numerical procedures are reported in the subsequent
paragraph.
3. Simulation of experimental results

Fig. 7. Calibration of the diagonal springs in the linear elastic range.

where fv is the average shear stress in the wall attained at the maximum resistance, ft is the tensile strength of masonry, b is the shear
stress distribution factor (depending on the geometry of the wall
and on the value of the ratio between the vertical N and horizontal
H load), ro is the average compression stress due to vertical load N.
The calibration of the diagonal springs in the initial linear
elastic range is simply obtained by enforcing an elastic equivalence
between the panel and the corresponding masonry wall, considered as a pure shear deformable homogeneous plate with tangential modulus G, transversal area At and height h, Fig. 7. In the
homogeneous media the relationship between the shear force V,
and the top horizontal displacement d is given by

V
h
G  At

while in the macro-element the corresponding forcedisplacement


relationship is:

V
2 cos2 a  K diag

10

where Kdiag represents the initial elastic stiffness of each diagonal


link and a = arctan (b/h), being b the base length. By equating
expressions (9) and (10) the initial elastic stiffness of each diagonal
spring is derived:

K diag

G  At
2h  cos2 a

11

The ultimate shear load Vu and the corresponding displacement


uu of each diagonal spring are therefore derived by the knowledge
of the mechanical material properties of the masonry by means of
the adopted shear resistance criteria. Since the macro-element
must incorporate the mechanical property of the nite portion of
the masonry wall that represents, the ultimate shear displacement
of the element is directly associated to the ultimate shear generalised deformability of the masonry media, whose value can be evaluated experimentally or can be suggested by technical codes. If
experimental data are available, a better prediction of experimental results can be obtained by considering different values of ultimate shear loads and displacements for piers and spandrels.
According to the proposal of Magenes and Calvi [15], the ultimate
displacement of the wall can be associated to a specic value of the
ultimate angular deformation cu = du/h that is dependent on the
particular masonry media. The post-yielding behaviour of each
diagonal spring can be characterised according to different constitutive plastic models as shown in the numerical applications
reported in the following section.
It is worth to notice that in the proposed approach each macroelement inherits the plane geometrical properties of the modelled
masonry portion, as a consequence, contrary to the simplied
models based on equivalent frame element approach, there is no
need to dene an effective dimension of the structural element.
Once the constitutive laws have been dened both force and
displacement controlled load processes can be performed according to procedures currently used in nite element analyses

In this section the proposed discrete model is employed to simulate the nonlinear response of real masonry walls and structures
for which numerical and/or experimental results are available from
previous studies. In particular, the results of an extensive experimental and numerical research on the behaviour of unreinforced
brick masonry buildings [22,23] have been considered. The rst
applications are relative to numerical simulations of experimental
results of simple masonry panels characterised by various geometrical ratios [24]. Secondly, a case study relative to a large-scale
masonry model built and tested at the University of Pavia, in the
North of Italy, is considered [23].
3.1. Case study 1: masonry panels
The rst case study considers the results of cyclic shear tests on
brick masonry walls [24] that have been considered by several
authors aiming at validating different numerical approaches [25
28]. Namely the experimental behaviour of simple piers under
xedxed end conditions with the aim to reproduce exural
and diagonal shear cracking behaviours have been analysed. The
dimensions of the squat wall, that collapsed according to a diagonal shear behaviour failure mechanism, are 100  135 cm2. The
dimension of the slender wall, which exhibited a collapse response
dominated by the rocking, are 100  200 cm2. Both the walls are
characterised by a thickness of 25 cm and have been subjected to
an initial mean compressive stress of 0.6 MPa, associated to a
constant vertical load at the top, in addition to their own weights.
Further details on the specimens and test procedures are reported
on the referenced paper [24].
The experimental results of the considered panels are reported
in Fig. 8. In particular Fig. 8a reports the exural response characterised by a moderate hysteretic energy dissipation and an almost
nonlinear elastic behaviour with negligible strength degradation.
Fig. 8b reports the shear-diagonal typical forcedisplacement
curve dominated by shear cracking with high energy dissipation
and a signicant strength and stiffness degradation.
For the simulation of the cyclic degrading hysteretic behaviour
of a masonry wall, subjected to a combination of constant vertical
load and to a sequence of lateral load reversals, the use of an idealised bi- or tri-linear resistance envelope is recommended [14].
Here the cyclic shear-diagonal behaviour of the considered walls
has been modelled with the proposed approach by means of a
single macro-element. Furthermore, the shear-diagonal behaviour
has been modelled by means of the idealised bi-linear envelope
reported in Fig. 9. The relevant parameters of the bilinear envelope
have been identied according to the following simple rules:
The initial slope of the idealised envelope is evaluated according
to a secant stiffness, that can be associated to the formation of
the cracks, called effective stiffness of the wall Ke. The latter
has been obtained by assuming the resistance at the crack limit
as a percentage (60%) of the experimental maximum resistance,
as follows:

Ke

Hcr
dcr

12

The maximum resistance of the idealised curve is set equal to


the maximum resistance attained during the test Hmax;

I. Cali et al. / Engineering Structures 40 (2012) 327338

(a)

observe that a xed-ended masonry wall exhibits a deformation


that is partly associated to bending and partly to shear. Denoting
the shear and the exural stiffness as KV and KF, respectively, the
overall stiffness exhibited by the wall during the experiment, K,
can be considered as the result of two springs in series, from which
the shear stiffness can easily be obtained

80
60

force (KN)

40
20
0
-1,5

-1,0

-0,5

-20

0,0

0,5

1,0

1,5

KV

-40

-80

displacement (cm)

K F 12

100

40
20
-1,0

-0,5

0
-20 0,0

15

EI
h

16

where E is the Young modulus, I is the moment of inertia of the wall


cross-section and h is the height of the wall. The unloading stiffness
has been expressed according to the following simple expression

80
60

force (KN)

K  KF
KF  K

where the exural stiffness KF is provided, in simplied form, by the


following stiffness of an equivalent beam

-60

(b)

333

Ku Ko
0,5

1,0

-40
-60
-80
-100

displacement (cm)
Fig. 8. Experimental behaviour of simple piers, from Ref. [12]: (a) example of
exural response and (b) example of behaviour with diagonal shear cracking.

KI
Ko

b
17

where Ko is the initial stiffness, KI is the value of the stiffness that is


obtained by connecting the point corresponding to the current plastic deformation with the origin of the axes and b is a real number
that can assume value in the range 01. In the simulation the value
0.8 has been set. The re-loading stiffness Kr, in the cyclic behaviour,
has been set by pointing to the maximum reached plastic
deformation.
The exural behaviour is governed by the orthogonal interface
springs. These have been calibrated according to an elastic-perfectly-plastic constitutive law, with different tensile and compressive limits and an unloading stiffness expressed according to the
Eq. (17) with b = 0.8, the ductility of the orthogonal springs has
been set equal to 5. The other parameters, assumed in the simulation according to the above described criteria, are reported in
Table 1.
The results of the numerical simulation are reported in Fig. 10.
Even though the adopted constitutive models are very simple,
there is a satisfactory agreement between the experimental and
the numerical results.
3.2. Case study 2: masonry model (two-storey prototype brick
masonry building)

Fig. 9. Idealisation of experimental resistance envelope with bi-linear relationship.

The idealised ultimate states, determined by the maximum displacement attained during the test dmax and the corresponding
resistance Hdmax are set to be coincident to the experimental
ones.
The ratio between the resistance Hdmax at the ultimate displacement and the maximum resistance Hmax denes a strength
degradation factor Csd [14].

C ds

Hd max
Hmax

13

The slope of the envelope curve in the post-yielding behaviour is given by the kinematic hardening (or softening) parameter.

Hd max  Hmax
dmax  dH max

14

In order to calibrate the shear diagonal nonlinear link of the macroelement, that governs the pure shear behaviour, it is necessary to

This case study is relative to a two-storey prototype brick


masonry building tested by Magenes and Calvi [22,23] at the
University of Pavia and considered in [29] for comparing and validating different simplied numerical approaches. The rectangular
plan of the building is 6  4.4 m while the total height is equal to
6.4 m. The building consists of four two-wythe solid brick walls
with a total wall thickness of 250 mm. The prototype has been
tested by means of cyclic horizontal forces acting along the longitudinal direction. The building is composed by two masonry walls
with openings, in the longitudinal direction, and two masonry
walls with no openings in the transversal direction. These forces,
of equal value, have been applied directly to the two longitudinal
walls at the oor levels. As a result of the performed cyclic tests,
the cyclic forcedisplacement relationships for both the longitudinal walls have been obtained. The spatial conguration of the apertures in the walls with openings are different: the main entrance
wall of the building, denoted as door wall, possesses two windows
and two doors, while the wall opposite to this, denoted as window
wall, is characterised by six regularly located windows. The door
wall was disconnected from the adjacent transversal walls, while
the other (window wall) was connected to the adjacent walls with
an interlocking brick pattern around the corner. The oor consisted

334

I. Cali et al. / Engineering Structures 40 (2012) 327338

Table 1
Case study 1. Mechanical characteristics of the masonry.
Flexural

Table 2
Case study 2. Mechanical characteristics of the masonry.

Shear-diagonal

Flexural

E (MPa)

rc (MPa)

rt (MPa)

G (MPa)

ft (MPa)

2100

6.20

0.1

500

0.275

1.5

0.8
Piers
Spandrels

(a)

60

Base Shear (KN)

force (KN)

E
(MPa)

rc

rt

ft
(MPa)

fo
(MPa)

(MPa)

G
(MPa)

(MPa)

2100
2100

6.2
6.2

0.05
0.05

500
500

0.18
0.225

1.5
1.5

0.2
0.2

0.5
0.5

(a)

20
0
-1,0

-0,5

-20

0,0

0,5

1,0

1,5

-40
-60
-80

(b)

150
100
50
0
-3,0 -2,5 -2,0 -1,5 -1,0 -0,5 0,0
-50

1,5

2,0

2,5

3,0

1,5

2,0

2,5

3,0

-100

(b)

80

150

60

100

Base Shear (KN)

force (KN)

1,0

Top displacement (cm)

100

40
20
0
-0,5

0,5

-150

displacement (cm)

-1,0

Sliding

80

40

-1,5

Shear-diagonal

-20 0,0

0,5

1,0

-40
-60

50
0
-3,0 -2,5 -2,0 -1,5 -1,0 -0,5 0,0
-50

0,5

1,0

-100

-80
-150

-100

Top displacement (cm)

displacement (cm)
Fig. 10. Numerical simulation of experimental tests on piers: (a) example exural
response and (b) example of behaviour with diagonal shear cracking.

Fig. 12. Door prototype wall. Comparison between (a) experimental and (b)
numerical cyclic tests.

of a series of isolated steel beams designed to simulate a very


exible diaphragm. The constant vertical forces acting on the walls
are due to the self-weight of the masonry and to the gravity and
accidental loads of the oor slabs (248.4 kN for the rst storey
and 263.8 kN for the second storey). The seismic forces were simulated by the application of equal concentrated horizontal forces
directly applied at the longitudinal walls at the oor levels by
means of displacement controlled screw jacks.
The results of the numerical simulations reported in the following aim to predict the response of the door wall. In Fig. 11 the geometric layout of the wall and its basic representation by means of
the proposed macro-element approach is reported.

In Table 2 the mechanical masonry characteristic that have


been used for the numerical simulations are summarised. The results of the numerical simulation are relative to a displacementcontrolled push-over analysis. Since the experimental test has
been performed by applying equal concentrated forces at the oor
levels, the nonlinear numerical analyses have been conducted by
imposing, at each step, different oor displacement values leading
to equal resisting forces. These latter values have been obtained by
means of a NewtonRaphson iterative procedure imposing a global
equilibrium condition.
The diagonal springs of the panels have been modelled by considering the well known Turnsek Cacovic criterion [16,17] in which

(a)

(b)

void

void

void

void

Fig. 11. The door wall (a) geometric layout; and (b) equivalent mechanical scheme.

335

I. Cali et al. / Engineering Structures 40 (2012) 327338

Base Shear (KN)

150
100
50

-2,5

-2,0

-1,5

-1,0

0
-0,5 0,0
-50

0,5

1,0

1,5

2,0

2,5

2,0

2,5

-100
-150

First interstory displacement (cm)


150

Base Shear (KN)

the damage has been controlled by a strength degradation parameter a = 0.08. The hysteretic behaviour has been dened according to the bi-linear idealisation described in the previous
subsection.
In Fig. 12 a comparison between the experimental and numerical results in terms of base shear versus the absolute value of the
top displacement is reported. By observing the maximum reached
forces and displacements as well as the exhibited hysteretic behaviour and keeping in mind that the shear-diagonal behaviour has
been calibrated according to a simple bi-linear representation,
the agreement between the experimental and numerical results
can be considered satisfactory. Fig. 13 reports a comparison between the experimental and numerical results in terms of base
shear of the wall versus the rst inter-storey drift; also in this case
the numerical prediction appears to be acceptable.
In Fig. 14 a simplied representation of the damage scenarios
predicted by the model, corresponding to the value of drift of 3%
and to the ultimate value of 4.3%, are compared with the corresponding crack patterns obtained experimentally. It is worth to notice how the proposed approach is able to grasp the progressive
distribution of damage in the wall. The used representation in
the interface allows to distinguish the reactive compressive zone
from the cracked one due to tensile forces; diagonal bars inside a
panel indicate the yielding of the diagonal springs.
The further representation of the ultimate state, reported in
Fig. 15, clearly shows the composite shear-exural failure of the
piers and the diagonal shear failures of spandrels as well as the
crack patterns due to the partialisation of the interfaces.

100
50

-2,5

-2,0

-1,5

-1,0

0
-0,5 0,0
-50

0,5

1,0

1,5

-100
-150

First interstory displacement (cm)


Fig. 13. Door prototype wall. Comparison between (a) experimental and (b)
numerical cyclic tests.

Fig. 14. Simplied drafts of damage scenarios corresponding to different values of the applied drifts: (a) 0.3% and (b) 0.43% (maximum drift).

336

I. Cali et al. / Engineering Structures 40 (2012) 327338

Fig. 15. Damage scenarios corresponding to a maximum applied drift 0.43%.

In Fig. 16, the resultants of the compressive and shear forces in


the piers are reported as functions of the top displacement. It is
worth to notice how the proposed approach is able to model the
variation of the coupled normal and shear forces during the nonlinear response of the structure.
3.3. Inuence of the model discretization
In the considered approach the macro-element inherits the
geometry of the masonry portion that is modelled, this aspect

(b)

200
160
120

rigth pier

80
40
0
0,0

40

20

Shear (KN)

Panel Compression (KN)

(a)

constitutes a great advantage that is not common to all the simplied approaches based on a macro-element discretization. Furthermore the consistent geometry of the element makes possible the
implementation of models with an irregular distribution of the
openings and allows the implementation of models characterised
by different level of discretization associated to the mesh resolution and to the ne-tuning of nonlinear links in the interfaces. In
Fig. 17 the results of push-over analyses, performed on the door
wall of Pavia prototype, associated to different mesh resolutions
are reported. Three different mesh resolutions have been

0
0,0

0,5

1,0

1,5

2,0

2,5

-40

2,5

Top displacement (cm)


100

200

60

central pier

160

Shear (KN)

Panel Compression (KN)

2,0

rigth pier
0,5

240

120
80
40

20
-20 0,0

0,5

1,0

1,5

0,5

1,0

1,5

2,0

2,0

2,5

central pier

-60
-100

2,5

Top displacement (cm)

Top displacement (cm)

200

80

160

Shear (KN)

Panel Compression (KN)

1,5

-20

Top displacement (cm)

0
0,0

1,0

120
80

left pier

40
0
0,0

40

0
0,0

0,5

1,0

1,5

2,0

2,5

left pier
0,5

1,0

1,5

Top displacement (cm)

2,0

2,5

-40

Top displacement (cm)

Fig. 16. Door prototype wall: the resultants of compressive (a) and shear forces (b) in the piers of the door wall as a function of the absolute top displacement.

337

I. Cali et al. / Engineering Structures 40 (2012) 327338

(a) 175

(a)

150

Base Shear (KN)

150

Base Shear (KN)

175

125
100
75

mesh A

50

mesh B

25

mesh C

0
0.00

0.50

125
100
75
50
25

1.00

0
0.00

1.50

0.50

(b) 175

(b)
Base Shear (KN)

Base Shear (KN)

0 .10 MPa

t=

0 .20 MPa

1.50

2.00

175
150

150
125
100
75
mesh A
50

mesh B

25
0.50

1.00

1.50

displacement (cm)

Table 3
mesh resolutions.
Label

Number of macroelements

Number of degrees of
freedom

Maximum Nlink
distance (cm)

A
B
C

16
64
144

64
256
576

10
10
10

160
140
120
100
80
60

10 cm

40

20 cm

20

40 cm

1.00

1.50

100
75
50

f t= 0 .12 MPa
f t= 0 .075 MPa
f t= 0 .05 MPa

0
0.00

0.50

1.00

1.50

2.00

Top displacement (cm)

Fig. 17. Door prototype wall: Push-over analyses associated to different meshresolutions: (a) strain-softening behaviour; and (b) strain-hardening behaviour.

0.50

125

25

mesh C

0
0.00

Base Shear (KN)

0 .05 MPa

t=

Top displacement (cm)

Top displacement (cm)

0
0.00

1.00

t =

2.00

Fig. 19. Door prototype wall: Push-over analyses associated to different values of
tensile strength associated to the exural (a) and to the shear-diagonal (b)
behaviours.

when strain softening models are used. For this reason with the
aim to investigate the inuence of the mesh in the numerical
simulations, in Fig. 17, for the Pavia prototype, two different comparisons have been considered according to a strain-softening and
a strain hardening shear-behaviour characterised by a = 0.08 and
a = 0.08, respectively. All the other parameters are reported in Table 2. It can be observed as, in presence of strain softening plasticity (Fig. 17a), the localisation of plastic deformation causes a mesh
dependence in the softening branch of the push-over curve, while
in presence of strain-hardening behaviour (Fig. 17b) there is a good
agreement between the results obtained by different mesh
resolutions.
With reference to the mesh A and an elasto-perfectly plastic
behaviour, in Fig. 18 the inuence of the ne-tuning of the Nonlinear Links in investigated. Namely, the results of push-over analyses
associated to different numbers of Nlinks for unit length are compared. The comparison shows that, in the analysed case, all the considered Nlinks distributions provide results in very good agreement.
3.4. Inuence of the tensile and shear strength

2.50

Top displacement (cm)


Fig. 18. Door prototype wall: Push-over analyses associated to different Nlinks
distributions in the interfaces.

considered whose increasing computational cost is summarised in


Table 3. Mesh A correspond to the basic mesh size in which piers
and spandrels are modelled by a single macro-element. Mesh B
and C represent more rened mesh resolutions in which piers
and spandrels are represented by several macro-elements. It is well
known that, in the case of the nite element approach, conventionally conducted within the framework of the local continuum
theory [20,21], the numerical solution is sensitive to the mesh size

In the following a sensitivity analysis on the tensile strength,


associated to the exural and shear-diagonal behaviours, has been
performed in terms of monotonic push-over curves on the door
wall of the Pavia prototype. The discretization of the model is consistent with the mesh A and an elasto-perfectly plastic behaviour,
according to the parameters reported in Table 2. Namely, in
Fig. 19a the push-over curves for three different value of the tensile
strength rt, that governs the exural behaviour of the panel, are
represented. In Fig. 19b the sensitivity analysis has been conducted
by varying the tensile strength ft associated to the Turnsek and
Cacovic criterion. It is worth to notice, that the present modelling
approach allows to calibrate each fundamental collapse behaviour
(exural failure, diagonal shear failure and sliding shear failure)

338

I. Cali et al. / Engineering Structures 40 (2012) 327338

according to ad hoc adopted collapse criteria, this constitutes a


novelty with respect to the other approaches presented in the
literature. The parameters associated to each failure mechanism
are based on results from current experimental tests.
4. Conclusions
The problem of the evaluation of the seismic vulnerability of
unreinforced masonry buildings represents a subject of high practical relevance but, at the same time, a very difcult task. For the
assessment of URM building seismic vulnerability and its possible
seismic retrotting, the structural analyst needs simple and efcient
numerical tools whose complexity and computational demand
must be appropriate. In the present work an advanced model, based
on the denition of a plane macro-element, for the simulation of the
seismic behaviour and the evaluation of the seismic vulnerability of
unreinforced masonry buildings has been introduced. The proposed
model has been conceived to provide a simulation method capable
of predicting the seismic behaviour of a masonry structure at a
low computational cost, compared to a nonlinear nite element
simulation. A considerable reduction of the computational cost is
achieved since the nonlinear behaviour of a masonry portion is described by very few degrees of freedom and by means of uniaxial
nonlinear springs. A suitable denition of the parameters of the
model that leads to a reasonable equivalence between the masonry
wall and its simplied nonlinear mechanical representation has
been proposed. This method possesses many advantages with respect to other existing simplied approaches. The ability of the
macro-element to model a portion a masonry which inherits the
same plane conguration, and without distinguishing between
piers, spandrels and immune zones, makes the modelling of the global in-plane response of masonry faades with irregular opening
pattern possible. The presence of the interfaces between macroelements allows also the modelling of the interaction with nonlinear
frame elements. As a result inlled frame structures could also be
modelled. Furthermore the introduction of ad hoc interface elements could lead to a simple evaluation of the global behaviour of
three dimensional masonry building in presence of concrete ring
beams at the oor levels. This latter aspects constitute a research
in progress that will be the subject of a paper to come.
Although simple to comprehend, the proposed approach for
wide spread practical application requires the implementation of
ad hoc user interfaces able to simplify the application by third
party or practicing engineers. The effectiveness of the proposed
modelling has been evaluated by means of nonlinear incremental
static analyses performed on masonry walls and structures which
have been the object of theoretical and experimental research in
the past. The numerical results describe fairly well the experimental results, in terms of the cyclic response, stiffness, strength and
dissipative behaviour, as well as damage distribution.
Acknowledgement
This research has been supported by the Italian Network of
Seismic Engineering University Laboratories (ReLUIS).
References
[1] Associated Press. Quake panel urges law on retrotting buildings. AP breaking
news, May 8 2004.

[2] Chen SY, Moon FL, Yi T. A macroelement for the nonlinear analysis of in-plane
unreinforced masonry piers. Eng Struct 2008;30:224252.
[3] Moon FL, Yi T, Leon RT, Kahn LF. Recommendations for the seismic evaluation
and retrot of low-rise URM structures. ASCE J Struct Eng
2006;132(5):66372.
[4] Penelis Gr G. An efcient approach for pushover analysis of unreinforced
masonry (URM) structures. J Earthq Eng 2006;10(3):35979.
[5] Seible F, Kingsley GR. Modeling of concrete and masonry structures subjected
to seismic loading. Experimen Numer Meth Earthq Eng; 1991.
[6] Casolo S, Pea F. Rigid element model for in-plane dynamics of masonry walls
considering hysteretic behaviour and damage. Earthq Eng Struct Dynam
2007;36(8):102948.
[7] Casolo S, Sanjust CA. Seismic analysis and strengthening design of a masonry
monument by a rigid body spring model: The Maniace Castle of Syracuse.
Eng Struct 2009;31(7):144759.
[8] Loureno PB. Computations on historic masonry structures. Progr Struct Eng
Mater 2002;4(3):30119.
[9] Magenes G, La Fontana A. Simplied nonlinear seismic analysis of masonry
buildings. In: Proceedings of British masonry society, vol. 8; 1998. p. 1905.
[10] Kappos AJ, Penelis GG, Drakopoulos CG. Evaluation of simplied models for
lateral load analysis of unreinforced masonry buildings. J Struct Eng
2002;128(7):8907.
[11] DAsdia P, Viskovic A. Analyses of a masonry wall subjected to horizontal
actions on its plane, employing a non-linear procedure using changing shape
nite elements. Trans Modell Simul 1995;10:51926 [WIT Press].
[12] Braga F, Liberatore D, Spera G. A computer program for the seismic analysis of
complex masonry buildings. In: Pande GN, Middleton J, Kralj B, editors.
Computer methods in structural masonry, vol. 4. London: E & FN Spon; 1998.
p. 30916.
[13] Vanin A, Foraboschi P. Modelling of masonry panels by truss analogy part 1.
Masonry Int 2009;22(1):110.
[14] Tomazevic M. Earthquake-resistant design of masonry building. In: Elnashai
AS, Dowling PJ, editors. Series on innovation in structures and construction,
vol. 1. London: Imperial College Press; 2006.
[15] Magenes G, Calvi GM. In plane seismic response of brick masonry walls. Earthq
Eng Struct Dynam 1997;26:1091112.
[16] Turnsek V, Cacovic F. Some experimental result on the strength of brick
masonry walls. In: Proceedings of 2nd international brick masonry conference.
Stoke-on-Trent; 1971. p. 14956.
[17] Turnsek V, Sheppard P. The shear and exural resistance of masonry walls. In:
Proceedings of international research conference on earthquake engineering.
IZIIS, Skopje; 1981. p. 51773.
[18] Eurocode 6: Design of masonry structures, part 1-1: general rules for
buildings. Rules for reinforced and unreinforced masonry. ENV 1996-1-1:
1995. CEN, Brussels, 1995.
[19] Corradi M, Borri A, Vignoli A. Experimental study on the determination of
strength of masonry walls. Construct Build Mater 2003;17:32537.
[20] Jirasek M, Bazant Z. Inelastic analysis of structures. Wiley; 2001.
[21] Zienkiewicz OC, Taylor RL. The nite element method. Solid mechanics, fth
ed., vol. 2. Butterworth Heinemann; 2000.
[22] Calvi GM, Magenes G. Experimental research on response of URM building
systems. In: Proceedings of the USItaly workshop on guidelines for seismic
evaluation and rehabilitation of unreinforced masonry buildings. Pavia,
Technical Report NCEER-94-0021. National Centre for Earthquake
Engineering, Buffalo, 20 July 1994.
[23] Magenes G, Calvi GM, Kingsley GR. Seismic testing of a full-scale, two-story
masonry building: test procedure and measured experimental response. In:
Experimental and numerical investigation on a brick masonry building
prototype numerical prediction of the experiment, Report 3.0 G.N.D.T.
Pavia, January 1995.
[24] Anthoine A, Magonette G, Magenes G. Shear compression testing and analysis
of brick masonry walls. In: Proceedings of 10th European conference on
earthquake engineering, vol. 3. Vienna, A. A. Balkema: Rotterdam, 1995. p.
165762.
[25] Gambarotta L, Lagomarsino S. Damage models for the seismic response of
brick masonry shear walls. Part II: the continuum model and its application.
Earthq Eng Struct Dynam 1997;26:44062.
[26] Calvi MG, Kingsley GR, Magenes G. Testing of masonry structures for seismic
assessment. Earthq Spectra 1996;12(1):14562.
[27] Calderini C, Lagomarsino S. Continuum model for in-plane anisotropic inelastic
behaviour of masonry. J Struct Eng 2008;134(2).
[28] Brenchich G, Gambarotta L, Lagomarsino S. A macroelement approach to the
three-dimensional seismic analysis of masonry buildings. In: Proceedings of
11th European conference on earthquake engineering. Paris, A. A. Balkema:
Rotterdam 602; 1998.
[29] Marques R, Loureno PB. Possibilities and comparison of structural component
models for the seismic assessment of modern unreinforced masonry buildings.
Comput Struct 2011;89:207991.

You might also like