This article appeared in a journal published by Elsevier.

The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elseviers archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
Engineering Structures 33 (2011) 19551965
Contents lists available at ScienceDirect
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
Nonlinear response of masonry wall structures subjected to cyclic and
dynamic loading
Jos Fernando Sima

, Pere Roca, Climent Molins

Department of Construction Engineering, Technical University of Catalonia, Barcelona, Spain
a r t i c l e i n f o
Article history:
Received 3 December 2009
Received in revised form
28 January 2011
Accepted 28 February 2011
Available online 9 April 2011
Keywords:
Masonry wall
Nonlinear analysis
Cyclic loading
Dynamic loading
Generalized Matrix Formulation
a b s t r a c t
The assessment of the dynamic or seismic performance of complex structures often requires the
integration in the time domain of the structural equation of motion in the frame of a nonlinear
analysis. Although sophisticated methods have been developed for the nonlinear analysis of masonry
wall structures, including the macro- and micro-modeling approaches, these require large computational
effort still limiting the extent andcomplexity of the structures analyzed. This paper presents analternative
method based on the Generalized Matrix Formulation for masonry skeletal structures and load bearing
wall systems, which has been proved as an efficient formulation for the analysis of the strength capacity
of these kinds of structures (Roca et al. (2005) [17]). The basic formulation has been complemented
with a uniaxial cyclic constitutive model for masonry and a time integration scheme. The ability of the
resulting approach to predict the nonlinear dynamic response of masonry structures is shown through
its application to the time domain analysis of an experimental scale masonry building with available
experimental results on its dynamic response.
2011 Elsevier Ltd. All rights reserved.
1. Introduction
During the past two decades, structural analysis of masonry
structures has experienced a significant progress thanks to the
development of sophisticate computer methods for both static
and dynamic analysis. These methods are based on two main
approaches, namely macro- and micro-modeling. The first does
not take into consideration any distinction between masonry units
and joints, masonry being regarded as an equivalent continuous
and homogeneous material. The average material properties are
usually obtained by means of homogenization techniques (Pegon
and Anthoine [1], Luciano and Sacco [2], Milani and Loureno [3],
among others). The micro-modeling approachconsists of modeling
individually the mortar joints and the masonry units [4]. In
some cases, simplifications on the micro-modeling have been
introduced by using zero-thickness interfaces for the joints [5,6].
Although the two approaches afford the simulation of many
aspects of the complex nonlinear behavior of masonry, they both
(an especially micro-modeling) require still large computational
effort preventing their application to the study of large and
complex masonry structures. This limitation is even more evident
when the assessment of complex masonry structures by means of
time history analysis is considered. Alternative efficient methods,
allowing the time history nonlinear analysis of masonry wall

Corresponding author. Tel.: +34 934017380; fax: +34 93 4054135.

E-mail address: jose.fernando.sima@upc.edu (J.F. Sima).
structures with a reasonable grade of accuracy and computational
effort, are still necessary. Some of these alternative methods are
based on modeling wall masonry buildings as equivalent frame
systems.
Due to its large advantage in terms of computational effort, the
possibility of using equivalent frames to model masonry wall sys-
tems has been explored since long time ago. Among the first at-
tempts, Karantoni and Fardis [7], utilized this approach to carry
pushover analysis using a nonlinear static analysis and compared
the results with more detailed FEM models. These authors identi-
fied the main limitations and problems of the method, which were
later overcome thanks to further sophistications in the description
of the material and connections. Through their pioneering work,
Magenes and Calvi [8] and Magenes and Della Fontana [9] provided
a powerful tool (the SAM method) for the seismic pushover anal-
ysis of masonry buildings using rigid links to improve the descrip-
tion between piers and lintels and appropriate inelastic models to
take into account the walls flexural and shear failure. The method
was validated by comparison with full scale experiments and FEM
numerical results on the seismic response of multi-story masonry
building faades. The method was later extended to the analysis of
3D systems [10]. Further applications were developed by Kappos
et al. [11] and Salonikios et al. [12], among others.
Seismic analysis of wall systems by frame equivalent modeling
is at present experiencing significant attention. Belmouden and
Lestuzzi [13,14] have formulated a linear finite element including
two multilayer connection hinges to model the deformations
0141-0296/$ see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2011.02.033
Author's personal copy
1956 J.F. Sima et al. / Engineering Structures 33 (2011) 19551965
Notation

f
,
m
Parameters of the generalized- integrationmethod
Parameter of the Newmark integration method
C Damping matrix

0
Strain at the elastic limit of masonry in compression

c
Strain at peak of the stressstrain curve of masonry
in compression

ct
Tensile strain corresponding to the tensile strength
of masonry

un
Unloading strain on envelope curve for masonry in
compression

re
Reloading strain on envelope curve after a complete
cycle of masonry in compression

pl
Residual plastic strain after the unloading curve of
masonry in compression

s
= (, )
T
=
_

x
,
y
,
z
,
x
,
y
,
z
_
T
Vector of sectional
strains and curvatures

0
s
Vector of initial strains of the section

XY
(s) = (, )
T
Strain vector of the section with curvilin-
ear coordinate s
d
0
Vector of displacements inB due to the initial strains
and stresses
d

Vector of displacements in B produced by the

deformation of the element in its basic isostatic
configuration under the effect of distributed loads

Masonry damage in compression

+
Masonry damage in tension

un
Damage at the unloading strain on envelope curve
for masonry in compression

re
Damage at the reloading strain on envelope curve
for masonry in compression
E
0
Initial elastic modulus of masonry
E
pl
Stiffness at zero stress after unloading of masonry
E
re
Reloading stiffness of masonry in compression
E
tre
Unloadingreloading stiffness of masonry intension
f
0
Stress at the elastic limit in compression
f

c
Compressive strength of masonry
f
ct
Tensile strength of masonry
F Flexibility matrix of the element
Parameter of the Newmark integration method
G
f
Fracture energy of the masonry in tension
= (X, Y, Z) Reference vector of points belonging to the
axial curve
K
s
Sectional stiffness matrix
K Stiffness matrix of the element
l

Characteristic length or crack band width

(s) Vector of displacements reference axial curve at the
point with curvilinear coordinate s
() = (
X
,
Y
,
Z
) Vector of sectional strains in the curvi-
linear coordinate point
M Mass matrix of the element
N(s, s
B
) Interpolation matrix that describes the exact equi-
librium forces between the transverse sections B
and s
(s) Vector of rotations of the reference axial curve at the
point with curvilinear coordinate s
() = (
X
,
Y
,
Z
) Vector of sectional curvatures in the
curvilinear coordinate point
p(t) Vector of applied loads
[p, m]
T
Vector of distributed forces and moments
P Vector of forces at the extreme of the element
P
0
Vector of reactions corresponding to the perfect
clamping at the ends of the element due to initial
strains and stresses.
P
A
Vector of forces at the end A
P
B
Vector of forces at the end B

s
= (R, M)
T
=
_
N, V
y
, V
z
, M
x
, M
y
, M
z
_
T
Vector of sectional
forces

0
s
Vector of forces due to initial stresses

XY
(s) Vector of sectional forces at a section of curvilinear
coordinate s

XY
(s) Vector of forces produced by the distributed loads
on a cantilever isostatic configuration
u Vector of displacements at the extremes of the
element
u(t) Vector of displacements at instant t
u
i
, u
i+1
Vectors of displacements at time t
i
and t
i+1
u
i
, u
i+1
Vectors of velocities at time t
i
and t
i+1
u
i
, u
i+1
Vectors of accelerations at time t
i
and t
i+1
u(s) Vector of accelerations of a point on the axis with
curvilinear coordinate s.
occurring in beam-to-column, column-to-footing or wall-to-
footing connections. A plastic fracturing model is used to describe
the cyclic response of concrete. The model has been validated
by comparison with experimental results on the cyclic response
of reinforced concrete structural walls modeled as a single beam
element. Pasticier et al. [16] have used equivalent frame modeling
for the study of the seismic performance for an existing two-story
masonry building faade. In their approach, the inelastic behavior
is lumped to plastic devices (flexural and shear plastic hinges and
nonlinear links) available in SAP 2000
r
v.10 package. Using these
arrangements, the authors have successfully carried out pushover
and time history analyses.
Recently, Milani et al. [15] have proposed a newstrategy for the
pushover analysis of in-plane loaded 2Dwall systems based on the
equivalent frame model. Their method involves two steps, the first
one utilized to determine the ultimate resistance of the masonry
walls using a heterogeneous limit analysis, and the second one for
assembling andanalyzing the equivalent frame. Flexural hinges are
introduced at both ends of the coupling beams when the capacity
is exceeded according to the failure loads determined in the first
step.
The aim of this paper is to present an alternative method for
nonlinear analysis of masonry wall structures in the time domain
based using equivalent frame systems. The basis of the model pro-
posed herein has been presented by Roca et al. [17]. In this work,
the authors proposed an efficient method on the framework of the
Generalized Matrix Formulation, which allows the modelization of
wall panels as a unique element with only two nodes with six de-
grees of freedom per node. Molins et al. [18] extended the basic
formulation to linear dynamic analysis. These authors proposed a
consistent element mass matrix which takes into account the exact
stiffness distribution throughout the element. The approach pre-
sented in this paper involves the extension of the formulation pro-
posed by Roca et al. [17] to nonlinear dynamic analysis of masonry
structures. The mainadvantage of the GeneralizedMatrix Formula-
tion, compared with traditional FEMbeammodels, is that provides
a virtually exact solution. It consists of a generalization of the
flexibility-based conventional matrix methods, in which the frame
deformation shape is a result of the exact integration of the equi-
librium and compatibility equations of an element. The absence of
interpolation errors allows using large elements without the ne-
cessity of intermediate nodes. However, it is necessary to use a
great number of control sections for the integration along the ele-
ment which may introduce precision problems and errors similar
to those obtained with the finite element method when the struc-
ture is discretized into a large number of elements.
Author's personal copy
J.F. Sima et al. / Engineering Structures 33 (2011) 19551965 1957
In order to extend the method to nonlinear dynamic analysis, a
phenomenological uniaxial cyclic constitutive model for masonry
is proposed. Additionally, an efficient time integration procedure
(the -generalized method) has been adapted and implemented
into the Generalized Matrix Formulation. In order to show the
capabilities of the proposed approach to predict the nonlinear
dynamic response of masonry wall structures, results obtained
with this method are compared with experimental results
obtained by Tomazevic and Weiss [19] on a three-story masonry
experimental model building subjected to a simulated ground
motion on a shaking table.
Compared to the previous developments based on frame
equivalent systems, the proposed method shows some advantages
because of possibility of carrying out pushover analysis or even
time history analysis on complex 2D and 3D systems involving
not only regular piers and lintels but also deformable floor slabs,
arches or vaults. With respect to the previous proposals, the
description of the walllintel connection is improved thanks to
the adoption of Kwans [20] approach for the modeling of the
rotational compatibility condition. As mentioned, the adoption of
GMF as frame formulation grants large accuracy in combination
with numerical efficiency even in the case of large and complex
systems.
2. Basic formulation
As in [17], the Generalized Matrix Formulation (GMF), a fle-
xibility-based formulation for 3D-framed structures composed of
curved elements with variable cross section, has been adopted
as a framework to model masonry linear or wall members. This
approach, firstly derived by Mar [21] and later extended by
Molins and Roca [22], consists of a hybrid formulation in which
the sectional internal forces across the element are expressed
as an interpolation of the external end forces acting on it.
This interpolation is not arbitrarily defined; actually, it is an
exact expression directly derived from the consideration of the
equilibrium between external and internal forces at any point
within an element, with no additional assumptions. Given that
the basic formulation has been presented in detail in the referred
works, only the fundamentals are outlined herein.
The equilibrium equation of the element is derived by consid-
ering a slice of element with infinitesimal length ds belonging to a
member subjected to a distributed load p and a moment m along
its axis, as is shown in Fig. 1. Acting on both ends of the slice, the
forces and moments R, M and R +dR, M +dM at point O and
O

respectively, are in equilibriumwith the applied loads. The equi-

librium condition may be expressed as follows:
dR + pds = 0
R ds + dM + mds = 0.
(1)
The equations of equilibrium of an element can be obtained by
integrating the ordinary system of first-order equations (1). The
solution of the systemis formed by a solution of the homogeneous
system plus a particular solution of the complete system. As a
particular solution of the system, a cantilever configuration with
free end at point B has been selected. Under these assumptions,
integration of (1) leads to the following equation:
_
_
_
_
_
_
R
X
R
Y
R
Z
M
X
M
Y
M
Z
_

_
=
_
_
_
_
_
_
1 0 0 0 0 0
0 1 0 0 0 0
0 0 1 0 0 0
0 (Z Z
B
) (Y Y
B
) 1 0 0
(Z Z
B
) 0 (X X
B
) 0 1 0
(Y Y
B
) (X X
B
) 0 0 0 1
_

_
Fig. 1. Equilibrium over a slice of beam with a differential thickness.
Fig. 2. Cantilever basic static determined configuration.

_
_
_
_
_
_
R
XB
R
YB
R
ZB
M
XB
M
YB
M
ZB
_

_
+
_
_
_
_
_
_
R

X
R

Y
R

Z
M

X
M

Y
M

Z
_

_
. (2)
Eq. (2) may be rewritten in the following compact form:

XY
(s) = N(s, s
B
)P
B
+

XY
(s) (3)
where N(s, s
B
) is an interpolation matrix that describes the exact
equilibrium forces between the transverse sections B and s,
XY
(s)
is the vector of sectional forces at any section of curvilinear
coordinate s, and

XY
(s) is the vector of forces produced by
the distributed loads, p and m, on a cantilever basic statically
determined configuration (Fig. 2) configuration, calculated as

XY
(s) =

B
s
N(s, )[p, m]
T
d. (4)
The kinematic compatibility is introduced by means of the Navier
Bresse equations [21]:
(s) = (s
0
) +

s
s
0
()d (5)
(s) = (s
0
) +(s
0
) ((s
0
) (s))
+

s
s
0
(() +() (() (s))) d (6)
where (s) and (s) are the vector of displacements and rotations
of the reference axial curve at the point with curvilinear coordinate
s; () = (
X
,
Y
,
Z
) and () = (
X
,
Y
,
Z
) are the vectors
of sectional strains and sectional curvatures, in the curvilinear
coordinate point , and = (X, Y, Z) is the reference vector of
points belonging to the axial curve (Fig. 2).
The constitutive relationship between sectional forces and sec-
tional strains in local coordinates may be expressed as follows:
Author's personal copy
1958 J.F. Sima et al. / Engineering Structures 33 (2011) 19551965

s
= K
s
_

s

0
s
_
+
0
s
(7)
where
s
= (R, M)
T
=
_
N, V
y
, V
z
, M
x
, M
y
, M
z
_
T
are the sectional
forces, and
s
= (, )
T
=
_

x
,
y
,
z
,
x
,
y
,
z
_
T
are respec-
tively the sectional axial strain, the shear strains in the y and z axis,
the twist and the bending curvatures, K
s
is the sectional stiffness
matrix, which depends on the shape of the section and on the elas-
tic properties of the materials,
0
s
is the vector of forces due to ini-
tial stresses and
0
s
is the vector of initial strains of the section. The
following matrix expression is obtained by operating and combin-
ing expressions (1)(7):
_
P
A
P
B
_
=
_
N (s
A
, s
B
) F
1
N
T
(s
A
, s
B
) N (s
A
, s
B
) F
1
F
1
N
T
(s
A
, s
B
) F
1
_

_
u
A
u
B
_

_
N (s
A
, s
B
) F
1
_
d

+ d
0
_
P

A
F
1
_
d

+ d
0
_
_
. (8)
Refer to Molins et al. [18] for a detailed explanation of the deriva-
tion of Eq. (8). This may be rewritten in a more compact form as
P = Ku + P
0
(9)
where P is the vector of forces at the extreme points, K is the
stiffness matrix of the element, u is the vector of displacements
at the extreme points and P
0
is the vector of reactions corre-
sponding to the perfect clamping at the ends of the element due to
initial strains and stresses. This last equation allows the construc-
tion of the global equation of the structural problem following the
conventional assemblage processes. The solution method for the
nonlinear problem in the framework of the Generalized Matrix
Formulationadoptedhereinwas proposedby Molins andRoca [22].
It consists of a combination of an incremental process, in which
some predefined load increments are gradually provided to the
structure, together with an iterative method to solve the problem
at each increment (NewtonRaphson). Aremarkable feature of this
solution strategy is that it combines a set of iterations at the global
level with a secondary set for each individual element. Molins and
Roca [22] observed that this procedure is significantly advanta-
geous in terms of total computer cost.
A consistent mass matrix, based on the present formulation,
has been derived by Molins et al. [18] by defining a sectional mass
matrix and providing a description of a displacement field inside
the element. A remarkable feature of the resulting formulation is
that the element mass matrix incorporates information related to
the geometry, stiffness distribution and mass density within the
element.
Roca et al. [17] proposed an equivalent frame method, where
both the walls and lintels beams are treated as discrete frame
members with the introduction of a set of special devices to
represent more realistically the shear deformation of the walls
based on the technique proposed by Kwan [20]. A detailed
explanation of this approach for the treatment of wall systems can
be found in the referred work. Integration along the axis of the
element is carriedout using a multiple Simpsons rule withvariable
number of points. In order to obtain the stiffness matrix of any
element with curved 3D axis and variable cross section is usually
sufficient to consider 11 integration points. On the other hand, the
computation of the mass matrix includes three integration levels
and needs more accurate rules, together with a set of additional
integration points [18].
3. Cyclic constitutive model for masonry
A significant number of papers have been presented on the
characterization of the behavior of masonry subjected to mono-
tonic compression or shearcompression loading (Priestley and
Elder [23] Magenes [24], Ewing and Kowalski [25] among others).
However, only a fewworks have been presented on the behavior of
masonry under cyclic compressive loadings. This is the case of the
works by Naraine and Sinha [26], AlShebani and Sinha [27,28] and
Oliveira et al. [29]. Naraine and Sinha [26] investigated the defor-
mationcharacteristics of firedclay brick masonry withlowlevels of
compressive strength under cyclic loading. This investigation was
later extended to the deformation characteristics of sand plast (a
form of calcium silicate) brick masonry with higher levels of com-
pressive strength subjected to uniaxial cyclic loading [27] and bi-
axial cyclic loadings AlShebani and Sinha [28]. Oliveira et al. [29]
researched on brittleness, energy dissipation and stiffness degra-
dation of masonry prisms under cyclic loading. Other authors, as
Chen et al. [30] or Macchi [31], have also reported on the cyclic
behavior of brick masonry with focus on seismic design of build-
ings. In all cases, the behavior shown by brick masonry subjected
to uniaxial cyclic loading presents significant similarity to that of
concrete. The latter has beeninvestigatedsince long time by Karsan
and Jirsa [32], among others.
Based on these results, the basic procedure proposed by Sima
et al. [33] for the derivation of a uniaxial cyclic constitutive loading
of concrete has been adopted herein. This procedure consists of the
modelization of experimental results available in the literature in
order to obtain, by means of linear regression, the necessary model
parameters for the material.
The present method considers the following equation, previ-
ously used for concrete, to model the envelope curve of masonry
in cyclic compression:
_
= E
0

0
=
_
1

_
E
0
>
0
(10)
where
0
is a strain value limiting the initial branch, E
0
is deforma-
tion modulus of the initial linear branch, and

=
_
1

0

(1 A) Ae
_

c
__
(11)
is the compression damage parameter which represents the mate-
rial degradation in compression, varying from 0 (material without
deterioration) to 1 (completely damaged material), and
A =
f

c

0
E
0
E
0
_

c
e

c
1

0
_ (12)
where f

c
is the compression strength and

c
is the strain at the peak
of the stressstrain curve.
The cyclic behavior is characterized by a set of equations that
reproduce a complete loop of unloadingreloading.
The proposed unloading curve is given by the equation:
= D
1
e
D
2
_
1

pl
un
pl
_
E
0
_

pl
_
(13)
where
D
1
=
r (1
un
)
(r 1)
D
2
= Ln
_
R (1
un
) (r 1)
r
_
(14)
with r =
pl

un
and R = E
pl
/E
0
.
pl
is the strain at zero stress, E
pl
is
the stiffness at the end of the unloading curve and (Fig. 3),
un
is the
compressive damage

at the unloading point which is the only

parameter used here to define the complete unloadingreloading
path.
The dependence of the other variables with this parameter has
been determined in a semi-empirical way. The cyclic compression
experimental tests performed by Naraine and Sinha [26] over
brick masonry panels have been reproduced with this model and
statistical regression has been performed in order to obtain the
proper dependency of the variables with the level of damage when
the unloading starts.
Author's personal copy
J.F. Sima et al. / Engineering Structures 33 (2011) 19551965 1959
Fig. 3. Complete unloadingreloading cycle in compression.
Fig. 4. Relationship between the reloading damage and the unloading damage
obtained by means of statistical regression on experimental results by Naraine and
Sinha [26].
Significant differences are found in the shape of the reload-
ing curves in compression obtained by different authors. Oliveira
et al. [29] obtained rather straight reloading branches, while
Naraine and Sinhas [26] experiments yielded strongly concave
curves. The character of the reloading curves may be very de-
pendent upon the specific properties of the masonry components
(brick and mortar). Unfortunately, a general description can be
hardly formulated based of the very limited available experi-
mental results. For the sake of simplicity, the constitutive equa-
tion adopted for the present research considers straight reloading
branches. Such a simple approach has obvious advantages on re-
quired input data and computer effort. It must be noted, however,
that it may produce, in some cases, an inaccurate estimation of the
overall energy dissipation throughout the hysteretic cycles.
It has been observed in test results [26] that the reloading curve
does not return to the envelope curve at the previous maximum
unloading strain and further straining in needed to taking up again
the envelope curve. It has beenfoundthat the relationshipbetween
the reloading compressive damage
re
(defined as the compressive
damage

at the reloading strain

re
) and the unloading
compressive damage
un
shows a linear behavior (Fig. 4).
The difference between
un
and
re
represents the damage
accumulated in each cycle. The effect of cyclic stiffness degradation
and its dependence with the unloading compressive damage may
be observed in Fig. 4. Figs. 5 and 6 give the obtained relationship
between the parameter r (unloading strainplastic strain ratio)
and the unloading compressive damage, and the parameter R
(final unloading stiffnessinitial unloading stiffness ratio) and
the unloading compressive damage, respectively. Fig. 7 relates
the reloading stiffness degradation with the level of damage in
compression.
Fig. 5. Relationship between the reloading stiffness and the unloading damage
obtainedby statistical regressiononexperimental results by Naraine andSinha [26].
Fig. 6. Relationship between the unloading strainplastic strain ratio and the
unloading damage obtained by means of statistical regression on experimental
results by Naraine and Sinha [26].
A linear variation between the unloading damage
un
and the
reloading damage
re
has been considered for the compressive
damage

during the unloading path. For the reloading path, the

compressive damage has been maintained as a constant, which is
in agreement with experimental data (i.e. the test results suggest
that the reloading curve becomes highly nonlinear only beyond
the point of intersection with the unloading curves, often referred
to as the common point). These relationships can be expressed as
follows:

=
un
+

re

un

pl

un
(
un
) (15)
for the unloading path, and

=
re
(16)
for the reloading path.
Very good adjustments have been possible for the different
parameters of the model, with correlation coefficients varying
between 0.98 and 1 (Figs. 47). As a result, the proposed model
can be applied to simulate the experimental results presented by
Naraine and Sinha [26] and a good comparisonis obtainedbetween
the cyclic experimental and numerical results (Fig. 8).
Due to the lack of experimental information regarding the
general case of partial unloadingreloading, as well as for the cyclic
behavior in tension, the envelope curve in tension proposed for
concrete in [33] has been adopted herein. The basics features of
the cyclic tension model are outlined in the following paragraphs.
The tensile envelope curve adopted for the present formulation
consists of a linear elastic relationship until reaching the tensile
Author's personal copy
1960 J.F. Sima et al. / Engineering Structures 33 (2011) 19551965
Fig. 7. Relationship between the final unloading stiffnessinitial unloading stiffness ratio and the unloading damage obtained by means of statistical regression on results
by Naraine and Sinha [26].
Fig. 8. Application of the proposed model for masonry under cyclic compressive
loading to one of the tests presented by Naraine and Sinha [26].
strength, followed by an exponential curve to represent the
softening branch. This curve may be expressed as follows:
= E
0

ct
=
_
1
+
_
E
0
>
ct
(17)
where

+
= 1

ct

_
1

ct
_
(18)
is the tensile damage parameter, which measures the material
degradation in tension and varies from 0 (material without
deterioration) to 1 (completely damaged material), where
ct
is the
tensile strainthat corresponds with the tensile strength(Fig. 9) and
is defined by the following expression:
=
_
G
f
E
0
l

f
2
ct

1
2
_
1
0 (19)
where G
f
is the fracture energy (considered as a material prop-
erty), f
ct
is the tensile strength of masonry and l

is a characteristic
Fig. 9. Proposed cyclic envelope and unloading/reloading path for masonry in
tension.
length or crack band width introduced to guarantee the objec-
tivity of the results with respect to the size of the finite element
mesh [34].
The tensile cyclic behavior is modeled herein in a simplified
way. Astraight line is used for the unloading branch in tension. The
same curve is considered for the reloading branch when there is
no incursion in compression during a cycle. The criterion proposed
by Sima et al. [33] is considered, accounting for the stiffness
deterioration due to cyclic loading:
E
tre
E
0
=
_

ct
_
1.05
(20)
where E
tre
is defined in Fig. 9.
4. Integration of the dynamic equation of motion
A great number of integration schemes have been proposed
to date for the dynamic equation of motion. The selection of the
Author's personal copy
J.F. Sima et al. / Engineering Structures 33 (2011) 19551965 1961
proper integration algorithm becomes an important issue when
the assessment of complex masonry structures is considered.
The main characteristics of an appropriated method have been
addressed by Hilber and Hughes [35]. These include unconditional
stability, accuracy and the introduction of algorithmic damping.
Unconditional stability may be achieved by considering implicit
algorithms. The introduction of the algorithmic damping allows
controllable numerical dissipation over the higher-frequency
modes (to avoid spurious, non-physical oscillations caused by
the spatial discretization due to excitation of spatially unresolved
modes). Inaddition, it has beenobservedthat using high-frequency
dissipation schemes may improve the convergence of equation
solvers in the case of highly nonlinear problems [36].
A basic difficulty in designing this kind of algorithms lays in ad-
equately balancing high-frequency and low-frequency dissipation.
High-frequency dissipation must be attained without introducing
excessive algorithmic damping over the important low-frequency
modes. The so called generalized- algorithm proposed by Chung
and Hulbert [37] has been selected to be implemented in this work,
due to its optimal combination of high-frequency dissipation and
low-frequency dissipation. Moreover, its second-order accuracy
and stability properties have been proved even in the nonlinear
case [36].
The generalized- algorithm is well known and detailed expla-
nation of this scheme can be found in the referred works. However,
the implementation of the method for the nonlinear case deserves
to be commented herein. This has been carried out through a New-
tonRaphson procedure which for the iteration k can be expressed
as:
u
(k+1)
i+1
_
r
(k)
i+1
, u
(k)
i+1
, u
(k)
i+1
_
=
_
K
eff(K)
i+1
_
1
R
(k)
i+1
(21)
where u
(k+1)
i+1
is the vector of displacement increment, R
(k)
i+1
is the
vector of residual forces and the matrix K
eff(k)
i+1
is the effective stiff-
ness matrix, which takes into account inertial and viscous damping
effects and the tangent stiffness matrix of the structure K
(k)
i+1
as fol-
lows:
K
eff(K)
i+1
=
1
m
_
1
f
_
t
2
M +
1
m
_
1
f
_
t
C + K
(k)
i+1
(22)
where M and C are the mass and damping matrices of the element.
The vector of residual forces of the iteration k is calculated as:
R
(k)
i+1
= F
i+1
+

f
1
f
F
i

1
m
1
f
M u
(k)
i+1


m
1
f
M u
i
C u
(k)
i+1


f
1
f
C u
i
r
(k)
i+1


f
1
f
r
i
(23)
where
f
and
m
are the parameters of the generalized- inte-
gration method, is the parameter of the Newmark integration
method, t is the time increment and r
i
is the vector of nonlinear
internal forces.
Once the incremental displacement vector is evaluated by
means of (21) and the total displacements are actualized through
u
(k+1)
i+1
= u
(k)
i+1
+ u
(k+1)
i+1
(24)
then, the velocities and accelerations can be actualized by means
of the well-known Newmark relationships.
The residual forces indicated in Eq. (23) are dissipated until the
desired level of accuracy is obtained, by means of convergence
criteria in both forces and displacements. The internal forces are
evaluated during a global iteration with a second set of iterations
at the element level as is shown in [17].
Fig. 10. Test setup (reproduced from [19]).
Fig. 11. Input ground accelerogram.
5. Application: three-story masonry building under seismic
loading
The example presented concerns the numerical simulation of
a three-story 1:5 scale plain masonry building subjected to a
simulated ground motion, experimentally tested at the Institute
for Testing and Research in Materials and Structures in Ljubljana,
Slovenia by Tomazevic and Weiss [19].
Fig. 10 describes the main characteristics of the experimental
model and test. In order to meet the requirements of similitude in
mass distribution and vertical stresses in the load bearing walls,
concrete blocks were fixed to the floor slabs (300 kg mass at each
floor level) and additional vertical stresses at the load bearing
walls were introduced by means of prestressed steel ropes at every
corner of the model, each providing a force of 12 kN, fixed to the
top slab and anchored into the foundation.
The structure was subjected to a series of ground shaking
simulations corresponding to the northsouth component of the
earthquake acceleration record of the Montenegro earthquake of
1979, with a peak ground acceleration of 0.43 g. The intensity of
the shaking was controlled by adjusting the maximum amplitude
of the shaking table displacement. The latter was obtained by
numerical integration of the earthquake accelerogram scaled
according to the laws of similitude.
The test identified as R47 has been modeled herein in order
to assess the capability of the proposed model. The shaking
table motion, in this case, presents a maximum acceleration of
1.10 g and the duration of the test was 5.5 s. The input ground
acceleration is shown in Fig. 11.
In order to apply the method presented, a structural model was
developed consisting of an equivalent system of beam elements to
model the entire system. To model the walls, a series of vertical
elements divided longitudinally into 13 integration sections were
Author's personal copy
1962 J.F. Sima et al. / Engineering Structures 33 (2011) 19551965
a
b
c
Fig. 12. Building model (a) overview (b) line model (c) cross section of the walls.
used. At the sectional level, the number of trapezoidal sub-
divisions used varied for each element, a higher number of sub-
divisions being included where high stress concentration was
expected (Fig. 12).
The mechanical properties of the masonry reported by Tomaze-
vic and Weiss [19] are compressive strength f
c
= 6.33 MPa, tensile
strength f
t
= 0.40 MPa and modulus of elasticity E = 6450 MPa.
The concrete slab presented a medium compressive strength of
f
c
= 25 MPa. Details of reinforcement for the slab were not
provided in the original paper. The following additional material
properties were assumed for the masonry walls: shear modulus
G = 1400 MPa and fracture energy G
f
= 120 J/m
2
. The ma-
sonry shear modulus considered herein is based on laboratory ob-
servations by Tomazevic [38] showing that the ratio G/E (shear to
Youngs moduli) obtained in experiments, is lower than that rec-
ommended in Eurocode 6 [39] (6%25% against 40%). The value as-
sumed herein for the fracture energy is in agreement with values
obtained experimentally by other authors [40]. The concrete slabs
were also modeled by means of a series of linear elements and the
material was assumed linear elastic with a modulus of elasticity
E = 31 500 MPa.
Fig. 13. Time history horizontal displacement results.
A linear elastic equation is adopted to describe the response
of the material in shear. Interaction between normal and shear
responses is considered through the use of a MohrCoulomb
criterion as axial strength envelope. In addition, the resulting shear
behavior is modified depending on the level of tensile damage
accumulated at each portion of the section. For tensile damage
equal to 0, the portion collaborates fully to the shear response,
while for tensile damage equal to 1, the portiondoes not contribute
at all. A linear interpolation is considered between these extreme
values.
As reported by Tomazevic and Weiss [19], the response of the
building is governed by the first mode of vibration. Therefore,
an eigenvalue analysis was performed before the test simulation
yielding a first natural frequency of 13.84 Hz, which is very similar
to the value obtained experimentally (13.81 Hz).
The time history analysis has been performed considering a
time step of t = 0.006 s and parameters for the generalized-
method
m
= 0.05,
f
= 0.1, = 0.33 and =
0.65. The selection of these values has been made by taking
into consideration the relationships between the values of the
parameters recommended by Erlicher et al. [36] to obtain an
adequate combination of accuracy, stability and dissipation over
higher modes.
Fig. 13 shows the results obtained for the time history analysis
in terms of horizontal displacements for each one of the floors. A
good overall agreement between the model predictions and the
test results was obtained, with the amplitudes and the frequencies
exhibiting acceptable deviations. The difference between the
maximum computed displacements on the 3rd floor and those
obtained experimentally was 8%, while at the 2nd floor was 3%
and at the first floor was a 25%. Figs. 14 and 15 show the damage
distribution at the masonry walls after the test. Even that the
Author's personal copy
J.F. Sima et al. / Engineering Structures 33 (2011) 19551965 1963
a
b
Fig. 14. Tensile damage distribution in the direction of the strong motion after the
test (a) center walls (b) lateral walls.
crack pattern for this test was not reported in the original work of
Tomazevic and Weiss [19] (i.e. the authors only provided the crack
pattern for the test that produced the collapse of the structure)
the damage distribution follows the trend observed at the crack
pattern at the ultimate state (Fig. 16).
Fig. 17 shows the displacement-response Fourier spectra for
the top floor. It can be observed that the higher contribution
to the response is produced at a lower value in the computer
model than the experimental result (3.45 Hz against 4.99 Hz).
Compared with the frequency measured for the intact structure
(13.81 Hz in the experiment vs. 13.84 in the simulation), it turns
out that the measured decrease in the value of the frequency
due to cumulated damage is of 64% in the experiment and 75%
in the numerical simulation. Given the large complexity of the
phenomena analyzed, this numerical value can be regarded as an
acceptable estimation of the experimental result. The difference
between both values can be attributed to a slight overestimation
of damage degree in the numerical simulation.
Interstory drift is an important parameter which is closely
related to damage sustained by the building during the seismic
excitation. Fig. 18 shows the interstory drift profile obtained by
means of the time history analysis. The maximum drift angle
numerically obtained for the first level was 0.81% corresponding to
a
b
Fig. 15. Compressive damage distribution in the direction of the strong motion
after the test (a) center walls (b) lateral walls.
a maximum base shear coefficient of 23.5%, given as a percentage
of the total gravity load, while the values obtained experimentally
by Tomazevic and Weiss [19] were 0.86% and 19% for the first floor
drift and base shear coefficient respectively.
A remarkable feature of the proposed formulation is that the
computational cost as well as the time required for the analysis
is very low due to the reduced number of elements necessary for
the analysis and the simplicity of the nonlinear material models. As
a matter of example, in this application the time required for the
complete analysis was less than two hours in a standard personal
computer (dual core 2 GHz, 4 GB RAM).
6. Conclusions
A novel method for the nonlinear dynamic analysis of masonry
structures in the time domain has been presented. The method is
applicable to skeletal structures or to load bearing wall systems
using, in the later case, equivalent frame configurations to model
the walls. The basic formulation for the description of linear
spatial systems, namely the Generalized Matrix Formulation, has
been complemented with a uniaxial cyclic constitutive model for
masonry adjusted with available experimental data. A significant
feature of the proposed model is that the required input data can
Author's personal copy
1964 J.F. Sima et al. / Engineering Structures 33 (2011) 19551965
a
b
Fig. 16. Crack pattern at the ultimate state (a) lateral walls (b) center walls [19].
Fig. 17. Displacement-response Fourier spectra for the top floor.
Fig. 18. Analytical drift profile obtained through the time history analysis.
be obtained through conventional monotonic compression and
tension tests. Additionally, a time integration scheme has been
adapted and implemented into the frame formulation.
The ability of the method to accurately and efficiently sim-
ulate the response in the time domain of a complex masonry
construction is shown through the application to a three-story ex-
perimental scaled building subjected to ground motion. The results
obtained show the capacity of the model to adequately predict the
dynamic and cyclic response of masonry structures at very reason-
able computer cost.
Acknowledgements
The studies presented here were developed within the research
project BIA2006-04127 funded by DGE of the Spanish Ministry of
Science and Technology, whose assistance is gratefully acknowl-
edged. The first author expresses his gratitude to the Educationand
Science Ministry of the Spanish Government for the financial sup-
port by means of a Grant for the Formation of University Teachers.
References
[1] Pegon P, Anthoine A. Numerical strategies for solving continuum damage
problems with softening: application to the homogenization of masonry.
Comput Struct 1997;64(14):62342.
[2] Luciano R, Sacco E. Homogenization technique and damage model for old
masonry material. Internat J Solids Structures 1997;34(24):3191208.
[3] Milani G, Loureno PB, Tralli A. Homogenized limit analysis of masonry walls.
Part I: failure surfaces. Comput Struct 2006;84:16680.
[4] Loureno PB. Computational strategies for masonry structures. Ph.D. disserta-
tion. Delft (The Netherlands): Delft University of Technology; 1996.
[5] Loureno PB, Rots JG. A multi-surface model for the analysis of masonry
structures. J Eng Mech, ASCE 1997;123(7):6608.
[6] Lofti HR, Shing PB. Interface model applied to fracture of masonry structures.
J Struct Eng, ASCE 1994;120(1):6380.
[7] Karantoni FV, Fardis MN. Effectiveness of seismic strengthening techniques for
masonry buildings. J Struct Eng 1992;118(7):1884902.
[8] Magenes G, Calvi GM. Prospettive per la calibrazione di metodi semplificati
per lanalisi sismica di pareti murarie. In: Proc. of the Italian national congress
la meccanica delle murature tra teoria e progetto. 1996.
[9] Magenes G, Della Fontana A. Simplified non-linear seismic analysis of masonry
buildings. Proc Br Masonry Soc 1998;8:1905.
[10] Magenes G. A method for pushover analysis in seismic assessment of masonry
buildings. In: Proc. of 12th world conference on earthquake engineering. 2000.
[11] Kappos AJ, Penelis GG, Drakopoulos CG. Evaluation of simplified models for
lateral load analysis of unreinforced masonry buildings. J Struct Eng, ASCE
2002;128(7):8907.
[12] Salonikios T, Karakostas C, Lekidis V, Anthoine A. Comparative inelastic
pushover analysis of masonry frames. Eng Struct 2003;25:151523.
[13] Belmouden Y, Lestuzzi P. An equivalent frame model for seismic analysis
of masonry and reinforced concrete buildings. Constr Build Mater 2009;23:
4053.
[14] Belmouden Y, Lestuzzi P. Analytical model for prediction nonlinear reversed
cyclic behavior of reinforced concrete structural walls. Eng Struct 2007;29:
126376.
[15] Pasticier L, Amadio C, Fragiacomo M. Non-linear seismic analysis of a masonry
building by means of the Sap2000 v.10 code. Earthq Eng Struct Dyn 2008;
37(3):46785.
[16] Milani G, Beyer K, Dazio A. Upper bound limit analysis of meso-mechanical
spandrel models for the pushover analysis of 2D masonry frames. Eng Struct
2009;31(11):2696710.
[17] Roca P, Molins C, Mar AR. Strength capacity of masonry wall structures by the
equivalent frame method. J Struct Eng, ASCE 2005;131(10):160110.
[18] Molins C, Roca P, Barbat AH. Flexibility-based linear dynamic analysis of
complex structures with curved-3D members. Earthq Eng Struct Dyn 1998;
27:73147.
[19] Tomazevic M, Weiss P. Seismic behavior of plain and reinforced masonry
buildings. J Struct Eng, ASCE 1994;120(2):32338.
[20] Kwan AKH. Analysis of coupled wall/frame structures by frame method with
shear deformation allowed. Proc Inst Civ Eng Struct Build 1991;91:27397.
Author's personal copy
J.F. Sima et al. / Engineering Structures 33 (2011) 19551965 1965
[21] Mar AR. A general method for the analysis of curved beams and space frames.
Barcelona: Department of Construction Engineering, Technical University of
Catalonia; 1985.
[22] Molins C, Roca P. Capacity of masonry arches and spatial frames. J Struct Eng,
ASCE 1998;124(6):65363.
[23] Priestley MJN, Elder DM. Stressstrain curves for unconfined concrete
masonry. ACI J 1983;80(3):192201.
[24] Magenes G. Comportamento sismico di murature di mattoni: resistenza e
meccanismi di maschi murari. Ph.D. thesis. Structural Mechanics Department.
University of Pavia; 1992 [in Italian].
[25] Ewing BD, Kowalsky MJ. Compressive behaviour of unconfined and confined
clay brick masonry. J Struct Eng, ASCE 2004;130(4):65061.
[26] Naraine K, Sinha SN. Behavior of brick masonry under cyclic compressive
loading. J Struct Eng, ASCE 1989;115(6):143245.
[27] AlShebani MM, Sinha SN. Stressstrain characteristics of brick masonry under
uniaxial cyclic loading. J Struct Eng, ASCE 1999;125(6):6004.
[28] AlShebani MM, Sinha SN. Stressstrain characteristics of brick masonry under
cyclic biaxial compression. J Struct Eng, ASCE 2000;126(9):10047.
[29] Oliveira D, Loureno P, Roca P. Cyclic behavior of stone and brick masonry
under biaxial compressive loading. Mater Struct 2006;39(2):24757.
[30] Chen SWJ, Hidalgo PA, Mayes RL, Clough RW, McNiven HD. Cyclic loading tests
of masonry single piers, volume 2height to width ratio of 1. EERC rep. no.
78/28. Berkeley: Environmental Engineering Research Council. University of
California; 1978.
[31] Macchi G. Behavior of masonry under cyclic actions and seismic design. In:
Proc. 6th international brick masonry conference. LI-XXIV. 1985.
[32] Karsan ID, Jirsa JO. Behavior of concrete under compressive loadings. J Struct
Div, ASCE 1969;95(ST12):254363.
[33] Sima JF, Roca P, Molins C. Cyclic constitutive model for concrete. Eng Struct
2008;30(3):695706.
[34] Oliver J. A consistent characteristic length for smeared cracking models. Int J
Numer Methods Eng 1989;28:46174.
[35] Hilber HM, Hughes TJR. Collocation, dissipation and overshoot for time
integration schemes in structural dynamics. Earthq Eng Struct Dyn 1978;6:
99117.
[36] Erlicher S, Bonaventura L, Bursi OS. The analysis of the generalized- method
for non-linear dynamic problems. Comput Mech 2002;28:83104.
[37] Chung J, Hulbert GM. A time integration algorithm for structural dynamics
with improved numerical dissipation: the generalized- method. ASME J Appl
Mech 1993;60:3715.
[38] Tomazevic M. In: Elnashai , Dowling , editors. Earthquake resistant design of
masonry buildings/Miha Tomazevic. Series on innovation in structures and
construction, vol. I. London: Imperial College Press; 1999.
[39] European Committee for Standardization (CEN EN-1996-1-1). Eurocode 6:
design of masonry structures. Part 1-1 general rules for reinforced and
unreinforced masonry structures; 2005.
[40] Rots JG. Structural masonry. Centre for civil engineering research and codes,
Rotterdam: Balkema; 1997.