Documente Academic
Documente Profesional
Documente Cultură
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/282505047
CITATIONS READS
9 129
2 authors, including:
Andreas Stavridis
University at Buffalo, The State University of New York
40 PUBLICATIONS 296 CITATIONS
SEE PROFILE
All content following this page was uploaded by Andreas Stavridis on 04 October 2015.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
SP-297—7
Analysis of Seismic Response of Masonry‐Infilled RC Frames through Collapse
P. Benson Shing and Andreas Stavridis
Synopsis:
The assessment of the seismic vulnerability and collapse potential of masonry‐infilled RC frame buildings
presents a significant challenge because of the complicated failure mechanisms they could exhibit and
the number of factors that could affect their behavior. In general, there are two types of analysis
methods that can be used to simulate the inelastic behavior of infilled frames. One is to use simplified
frame models in which infill walls are represented by equivalent diagonal struts, and the other is to use
refined finite element models that can capture the failure behavior of RC frames and infill walls in a
detailed manner. However, both types of models have limitations in simulating structural response
through collapse. While refined finite element models are not computationally efficient, simplified
models are less accurate because of their inability to represent some failure mechanisms that could
occur in an infilled frame. In this paper, possible failure mechanisms and causes of collapse of masonry‐
infilled RC structures are discussed, and both simplified and refined finite element analysis methods that
can be used to simulate the inelastic response of these structures and assess their vulnerability to
collapse are presented with numerical examples. Additional research and development work needed to
improve collapse simulations is discussed.
Keywords: masonry; infill walls; reinforced concrete frames; earthquakes; seismic; collapse; finite
element method; equivalent strut models.
7.1
P. B. Shing and A. Stavridis
P. Benson Shing is a Professor of Structural Engineering at the University of California, San Diego. He
received his BS, MS, and PhD from the University of California, Berkeley. He is a member of ACI and his
main research interests are in seismic design and analysis, and in the inelastic behavior and performance
of concrete and masonry structures, including large‐scale testing and analytical modeling.
Andreas Stavridis is an Assistant Professor of Civil Engineering at the University at Buffalo. He received
his BS degree from the National Technical University of Athens, Greece, and MS and PhD from the
University of California, San Diego. His main research interests include the experimental and analytical
evaluation of the seismic performance of structures, and the condition assessment and rehabilitation of
existing concrete and masonry buildings.
INTRODUCTION
Masonry‐infilled RC frames can be frequently found in seismically active regions around the world. They
represent a common form of construction in many countries. Many older buildings in the western
coastal regions of the United States have unreinforced masonry infill walls, and infill walls are frequently
used in newer buildings in the mid‐western and eastern parts of the country. Masonry infill walls are
normally neither reinforced nor tied to the RC frames. In spite of this, infill walls often enhance the
seismic resistance of non‐ductile RC frames, which would otherwise not be able to withstand a major
earthquake. However, the partial or total collapse of an infilled RC frame could be sudden as
demonstrated in past earthquakes. The assessment of their seismic vulnerability and collapse potential
presents a significant challenge because of the complicated failure mechanisms they can exhibit and the
number of factors that can affect their behavior. In general, there are two types of analysis methods
that can be used to simulate the inelastic behavior of infilled frames. One is to use simplified frame
models in which infill walls are represented by equivalent compression‐only diagonal struts, and the
other is to use refined finite element models that can capture the failure behavior of RC frames and infill
walls in a detailed manner. However, both types of models have limitations in simulating structural
response through collapse. While refined finite element models are not computationally efficient,
simplified models are less accurate because of their inability to represent some main failure mechanisms
that could occur in an infilled frame. In this paper, possible failure mechanisms and causes of collapse of
masonry‐infilled RC structures are first discussed, and both simplified and refined finite element analysis
methods that can be used to simulate the inelastic response of these structures and assess their
vulnerability to collapse are presented with numerical examples. Additional research and development
work needed to improve collapse simulations is discussed.
FAILURE MODES AND CAUSES OF COLLAPSE
Failure modes, load resistance, and ductility
When built with high quality masonry materials and workmanship, infill walls can significantly enhance
the earthquake resistance of RC frames. However, under severe seismic loads, their interaction with the
bounding frames can result in undesired failure mechanisms, which may lead to the collapse of the
structure. Hence, to evaluate the ability of this type of structures to survive an earthquake without
collapsing, the frame‐wall interaction and the resulting failure mechanism have to be accurately
simulated in the analytical model. Under earthquake loading, the interaction of an RC frame with
masonry infill walls can result in one of several possible failure mechanisms depending on the strength
and stiffness of the walls as compared to those of the frame. Figure 1 shows three failure mechanisms
that have been frequently observed in laboratories and in the field (Mehrabi et al. 1994, 1996).
7.2
Analysis of Seismic Response of Masonry-Infilled RC Frames through Collapse
Figure 1 – Failure mechanisms of infilled frames Figure 2 – Failure Mechanism (b) exhibited in a
(Mehrabi et al. 1996). quasi‐static cyclic loading test (Blackard et al.
2009).
Failure Mechanism (a) shown in Figure 1 may occur when infill walls develop profuse horizontal sliding
shear cracks because of weak mortar joints. This mechanism can result in flexural hinges at the top and
bottom of the columns. However, because of the relatively small horizontal forces exerted by the wall
against the columns in this case, the columns can avoid severe damage as long as the story drift is not
excessive. Mechanism (b) can occur in non‐ductile RC frames infilled with relatively strong masonry,
which tends to develop dominant horizontal and/or diagonal cracks as shown by a laboratory specimen
in Figure 2. In this case, shear failures can occur in the columns as a result of a large horizontal force
exerted by the cracked wall against the top or bottom of a column. Sometimes, shear failure can also
occur at the mid‐height of a column when the masonry bed‐joint at that level develops a through crack
leading to a short‐column effect. This can be the case when a weak horizontal sliding plane has been
introduced by a window opening. Mechanism (c) is associated with corner crushing in a wall as a result
of the racking force from the frame. This can be expected for walls constructed of weak masonry units,
such as hollow clay tile, which is vulnerable to compressive failure. This mechanism will result in flexural
hinges in the columns and may also introduce a short‐column effect.
In addition to the aforementioned failure modes, older multi‐story RC frames with strong infill walls
could develop lap‐splice failures at the base of the columns. In such a case, the infill walls and the
boundary columns act together to function like a monolithic shear wall with severe tensile and
compressive forces developing at the base of the columns.
Hence, the lateral load resistance contributed by the frame and the wall in an infilled frame is governed
by the frame‐wall interaction and the resulting failure modes. As a consequence, the resistance of an
infilled frame is not a simple sum of that of a bare frame and that of a masonry wall considered
independently.
The ability of an infilled frame to resist earthquake loads depends not only on its strength but also on its
ductility. As shown in Figure 3, a frame exhibiting Failure Mechanism (a) discussed above tends to be
more ductile than a frame developing Mechanism (b) because the latter has the shear failure of the
columns. Furthermore, as shown in Figure 4, increasing the axial load on an infilled frame governed by
Failure Mechanism (b) can increase its lateral load resistance but reduces its ductility. This can be
explained by the fact that the masonry infill is the main load resisting element of an infilled frame, and
7.3
P. B. Shing and A. Stavridis
the lateral resistance of a masonry wall is largely governed by the shear resistance of the bed joints,
which largely depends on the compressive stress in the joints. Nevertheless, masonry units will be more
susceptible to crushing failure when subjected to higher axial and shear stresses. Such crushing can
occur in the interior of a wall (which is different from Mechanism (c)), and will lead to a rapid strength
degradation. Such behavior can also be observed in frames exhibiting Mechanism (a).
Mechanism (a) Mechanism (b)
Figure 3 – Non‐ductile RC frames of the same design exhibiting Failure Mechanism (a) and Failure
Mechanism (b) due to different infill walls (Mehrabi et al. 1996).
Figure 4 – Influence of vertical load in non‐ductile RC frames exhibiting Failure Mechanism (b)
(Mehrabi et al. 1996).
Hence, to assess the shear strength of a masonry wall, it is important to know the amount of axial load
carried by the wall. The proportion of the gravity load initially carried by an infill wall depends on the
ratio of the elastic axial stiffness of the wall to that of the RC columns and the construction sequence,
i.e., whether the walls were constructed before the upper stories were built. If the infill walls were
constructed after the entire RC frame had been built, then these walls would experience almost no
gravity load right after construction. However, part of the sustained gravity load may be gradually
shifted to the walls as creep and shrinkage occurs in the RC columns and the clay units expand in time
due to moisture absorption. Therefore, one can expect that infill walls will eventually carry a portion of
the gravity loads unless gaps are introduced between the walls and the beams above. The vertical load
carried by a wall can be measured by in‐situ tests using apparatus like flat‐jacks. However, in the
7.4
Analysis of Seismic Response of Masonry-Infilled RC Frames through Collapse
absence of such data, it has to be estimated analytically. An analytical method to calculate the gravity
load distribution in an infilled frame considering the creep of the RC columns and masonry wall can be
found in Koutromanos (2011).
As compared to a bare frame, the ductility of an infilled frame is normally much lower. For a non‐ductile
frame with a strong infill wall, diagonal/sliding shear cracks usually occur in the infill at a story‐drift ratio
of 0.2 to 0.3%. This will be quickly followed by the shear failure of the RC columns, after which the in‐
plane lateral resistance starts to decrease rapidly with increasing drift. Normally, severe damage will be
inflicted in an infilled frame when the story‐drift ratio exceeds 1%.
In summary, the performance of an infilled frame and its ability to survive an earthquake depend on
many factors, including the strength and quality of the masonry infill, the reinforcing details of the
frame, and the resistance of the frame‐wall system as compared to the seismic load demand. In
addition, the strength and ductility of an infilled frame depend on the shapes, sizes, and locations of
openings in the walls (Stavridis 2009), and the locations of infill walls in the frame.
Causes of collapse
Even though an infilled frame may exhibit an undesired failure mode, such as Failure Mechanism (b)
shown in Figure 1, and the falling debris of damaged masonry can be a major life‐safety concern, both
field observations from past earthquakes and laboratory studies (e.g., see Stavridis et al. 2012) have
shown that masonry infill walls can significantly enhance the lateral resistance of a non‐ductile RC frame
and protect it from major damage or collapse in the event of a severe earthquake. The collapse of
infilled RC frames in past earthquakes was often associated with a weak story mechanism, which could
be attributed to the lack of infill walls in the bottom story of a building for certain practical reasons or to
the severe damage and subsequent loss of infill during strong shaking. Studies have shown that intact,
well‐built, multi‐wythe, infill walls can develop significant resistance to out‐of‐plane loads because of
the arching mechanism that can develop with the bounding frame (Dawe and Seah 1989, Angel et al.
1994, Mander et al. 1993, Bashandy et al. 1995, and Flanagan and Bennett 1999). However, the
effectiveness of the arching mechanism and the stability of a wall depend on its height‐to‐thickness
(slenderness) ratio, and such mechanism can be jeopardized when the masonry infill has been damaged
by the frame‐panel interaction due to in‐plane shaking. Single‐wythe infill walls are especially vulnerable
to out‐of‐plane collapse but their contribution to the structural resistance is small to begin with. Figure
5(a) shows the collapse of a single‐wythe clay tile wall during an earthquake while the building remained
standing. Figure 5(b) shows the collapse of an infilled‐frame building in the same earthquake, which can
be attributed to the lack of sufficient walls in the first story. One can see from the picture that significant
drift occurred in the first story of the building leading to the separation of the RC frame from the
exterior wall, which together with the damage inflicted on the columns might have contributed to the
loss of the vertical load carrying capacity of the structure. Walls having window openings are more
susceptible to collapse, which was observed in a shaking‐table test as shown in Figure 6.
Computational models
To simulate the earthquake response of a masonry infilled RC frame through collapse, it is important
that the model can directly or indirectly account for the aforementioned failure mechanisms including
the loss of stability of infill walls due to in‐plane and out‐of‐plane loads. In particular, it is important that
the model be able to account for the following possible local and global behavior of a structure.
7.5
P. B. Shing and A. Stavridis
1. The inelastic behavior of RC members, including flexural hinging, diagonal shear failure, axial
load failure of RC columns, lap‐slip failures, and the failure of beam‐to‐column joints.
2. The inelastic behavior of masonry walls, including the cracking and shear sliding of mortar joints
and the crushing failure of masonry units.
3. The loss of stability and out‐of‐plane collapse of infill walls, considering the influence of the
arching mechanism, wall damage induced by in‐plane loads, and wall openings.
4. The frame‐wall interaction and the resulting failure mechanism.
5. The P‐ effect on the RC frame.
6. The three‐dimensional response of a building to multi‐axial ground motions.
Figure 5 – Damage of buildings with masonry infill walls in 2008 Wenchuan Earthquake in China.
It is also important that the computational models should be general enough to account for the
construction quality. However, a model that has all the aforementioned simulation capabilities can be
computationally prohibitive, and simplifications will be normally needed. In the following sections, two
modeling methods are discussed. One is the use of detailed finite element models using continuum
elements and the other is the use of simplified models with frame elements.
7.6
Analysis of Seismic Response of Masonry-Infilled RC Frames through Collapse
Figure 6 – Collapse of masonry infill with a window opening in a shaking‐table test (Stavridis et al. 2012).
NONLINEAR FINITE ELEMENT MODELS
A detailed nonlinear finite element model can capture the global as well as local failure behavior of an
infilled frame. However, these models require a significant computational effort. There are two general
finite element modeling approaches to simulate the fracture behavior of quasi‐brittle materials, such as
concrete and masonry. They are the smeared and discrete crack approaches. The smeared‐crack
approach has been often used to model diffused tensile cracks as well as the compressive failure of
concrete in reinforced concrete structures. However, this approach suffers from several inherent
limitations, including stress locking (Rots 1991, Lotfi and Shing 1991) that limits its ability to simulate the
brittle behavior of an RC member failing in diagonal shear and the shear sliding behavior of masonry
joints. This limitation can be overcome by representing cracks in a discrete manner using zero‐thickness
interface elements. Hence, combining the smeared and discrete approaches is necessary to model the
failure behavior of infilled frames in a realistic way (Stavridis and Shing 2010, Koutromanos et al. 2011).
An example of this is shown in Figure 7, in which zero‐thickness interface elements were used to model
flexural and diagonal shear cracks in RC columns and cracks in masonry joints, while smeared crack
elements were used to model diffused damage, including compressive failure in the concrete members
and brick units. In this model, each reinforcing bar in the columns was divided into multiple truss
elements (Stavridis and Shing 2010) so that each discrete crack was resisted by the right quantity of
reinforcement. This also improved the robustness of the numerical solution by having as many smeared‐
crack elements connected to a truss bar as possible.
Constitutive laws that can be used for the smeared‐crack and zero‐thickness interface elements are
described in Koutromanos (2011) and Koutromanos et al. (2011). A smeared‐crack model represents
cracks in a distributed fashion and can also simulate the compressive failure behavior of concrete and
masonry. The model used by Koutromanos et al. (2011) has an uncracked material represented by a J2‐
plasticity constitutive law. As shown in Figure 8(a), it has a von Mises failure surface with a tension cut‐
off. When the maximum principal stress reaches the tensile strength of the material, cracks initiate in a
direction normal to the direction of the maximum principal stress. For a cracked material, it adopts a
nonlinear orthotropic law with the axes of orthotropy normal and parallel to the crack, whose direction
is assumed fixed. At each point, two orthogonal cracks are allowed to occur. The uniaxial stress‐strain
relation for each direction of the orthotropic law is shown in Figure 8(b).
7.7
P. B. Shing and A. Stavridis
Figure 7 – Finite element modeling of a masonry infilled RC frame.
INITIAL YIELD SURFACE σ2
-f ’m -fo f ’t ft
Initial stiffness σ Exponential
unloading/reloading softening
σ1 ft
f ’ t ε2 ε1
LINEAR ε
ELASTIC -fo
Secant stiffness
unloading/reloading
ELASTOPLASTIC Exponential
softening
-f ’m f΄c
fc parabola
VON MISES
FAILURE SURFACE
Figure 8 – Smeared‐crack constitutive model.
The shear failure of RC columns and the sliding shear behavior of mortar joints are best captured by
zero‐thickness interface elements with a cohesive crack constitutive law. A cohesive crack elastic‐plastic
constitutive model has been proposed by Koutromanos and Shing (2012) to capture the mix‐mode
fracture as well as the cyclic crack opening‐closing and sliding shear behavior of concrete and masonry.
7.8
Analysis of Seismic Response of Masonry-Infilled RC Frames through Collapse
It is implemented in a 4‐node, zero‐thickness, isoparametric, line element. The model can simulate the
initiation and propagation of cracks under combined normal and shear stresses. It also accounts for
reversible shear dilatation induced by the roughness of the crack surface, which can have a significant
effect on the response of a confined crack. As shown in Figure 9(a), the yield surface for the model is
hyperbolic and its evolution is controlled with internal variables that are functions of the mode‐I and
mode‐II fracture energies as well as the frictional work. Figure 9(b) shows an example on the modeling
of the shear behavior of a mortar joint. As shown, the model accounts for the shear dilatation and joint
compaction due to damage.
Shear Displacement (in)
-0.4 -0.2 0.0 0.2 0.4
n 1.5 218
1.0 146
-10 -5 0 5 10
Shear Displacement (mm)
0.5 0.02
0.0 0.00
-0.5 -0.02
-1.0 -0.04
-1.5 -0.06
-2.0 experiment -0.08
-2.5 analysis -0.10
-3.0 -0.12
-10 -5 0 5 10
Shear Displacement (mm)
Figure 9 – Cohesive crack interface model.
It is well known that the interaction between the mortar joints and the masonry units due to the
different moduli of elasticity and Poisson effects of the two materials has a significant influence on the
strength and failure mechanism of a masonry assembly. However, when a zero‐thickness interface is
used to represent the behavior of a mortar joint, the 3‐D behavior of a mortar layer and its interaction
with the adjacent masonry units cannot be simulated. In this case, the influence of the brick‐mortar
interaction on the masonry strength has to be accounted for in an indirect way by assuming that the
compressive strength of the masonry units is equal to the compressive strength of the masonry prisms.
For brick masonry, prism strengths are normally lower than the brick strengths due to the tensile
splitting stress introduced by the softer mortar to the stiffer brick. In addition, the stiffness of the
interface elements representing mortar joints has also to be so determined that the stiffness of the
masonry assembly is properly represented. A procedure to calibrate these constitutive models is given
in Stavridis and Shing (2010), and Koutromanos et al. (2011).
7.9
P. B. Shing and A. Stavridis
This modeling approach has been proven to be able to accurately capture the failure mode and load‐
displacement response of an infilled frame up to a severe damage state, at which the lateral load
resistance decreases to as much as 50% of the peak resistance and the story‐drift ratio exceeds 1%.
Figure 10 shows the simulation of a quasi‐static test conducted by Blackard et al. (2009) on a masonry‐
infilled non‐ductile RC frame. A picture of the damaged specimen is shown in Figure 2. It can be seen
that the model was able to capture the crack pattern as well as the load‐displacement response of the
test specimen up to a state with severe load degradation.
800
400
force (kN)
0
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5
-400 Experiment
Analysis
-800
drift ratio (%)
Figure 10 – Analysis of an infilled frame subjected to quasi‐static cyclic loading. (Koutromanos et al.
2011).
This modeling method has also been successfully applied to simulate the response of a three‐story,
masonry‐infilled, non‐ductile, RC frame tested on a shaking table by Stavridis et al. (2012). As shown in
Figure 11, the model is able to accurately capture the failure mechanism, response time histories, and
base shear‐vs.‐bottom story drift relations of the structure through the second last test on the shaking
table that had the Gilroy motion (from the 1989 Loma Prieta Earthquake) scaled to 120%. The specimen
collapsed out‐of‐plane in the subsequent test that had the El Centro record from the 1940 Imperial
Valley Earthquake scaled to 250%. This occurred after severe diagonal shear failures had developed in
the bottom‐story columns and the bottom‐story infill wall that had a window had partially collapsed (as
shown in Figure 6). However, neither the out‐of‐plane collapse of the frame nor the collapse of the wall
could be simulated by the model. This is because of the plane‐stress and small displacement
assumptions adopted in the model formulation.
7.10
Analysis of Seismic Response of Masonry-Infilled RC Frames through Collapse
0.4
0.3
0.2
1.2 2.0
1.5
0.8
1.0
drift ratio (%)
0.4 0.5
Vb/W
0.0 0.0
123.0 123.5 124.0 124.5 125.0 125.5 126.0 -1.2 -0.8 -0.4
-0.5 0.0 0.4 0.8 1.2
-0.4
-1.0
-0.8 Experiment Experiment
-1.5
Analysis Analysis
-1.2 -2.0
drift ratio (%)
time (sec)
(c) Bottom‐story drift time history for Gilroy (d) Base shear vs. story drift for Gilroy motion
motion scaled to 120% scaled to 120%
Figure 11 – Analysis of a three‐story infilled frame tested on a shaking table (Koutromanos et al. 2011).
It should be noted that even though all the validation studies were based on laboratory specimens that
had good quality materials and workmanship, the material parameters, such as the mortar joint
properties, used in the models can be calibrated to account for the deterioration due to aging and for
the less‐than‐ideal construction quality that may be encountered in real structures. To model the failure
of lap splices in RC columns, a bond‐slip model can be used to connect reinforcing bars to smeared‐crack
elements. Such a model has been developed by Murcia‐Delso (2013) to simulate the bond‐deterioration
between the reinforcing steel and concrete, and can be used to simulate the failure of bar anchorages
and lap‐splices.
The constitutive models presented here can be extended to 3 dimensions to simulate the out‐of‐plane
collapse of an RC frame or that of a masonry infill wall. While this extension is conceptually
straightforward, it can significantly increase the computational effort. Furthermore, to simulate the
collapse of a masonry wall due to brick dislocation and the collapse of an RC column with dominant
shear cracks under an axial load, the cohesive crack interface model presented here needs to be
extended to account for the loss of contact due to shear sliding. This can be achieved with a contact law.
7.11
P. B. Shing and A. Stavridis
However, combining a cohesive crack model with a contact law will introduce additional computational
overheads.
SIMPLIFIED ANALYSIS METHODS
A simple and efficient approach to model infill walls in frame structures is to use the equivalent diagonal
strut concept as proposed by Holmes (1961), Stafford Smith (1967), and Mainstone and Weeks (1970).
This concept has been investigated and extended by others in numerous studies. However, the original
equivalent strut theory was based on the observed behavior of small‐scale steel frames infilled with
either brickwork or concrete. For masonry infilled RC frames, the equivalent strut representation
appears to be an over‐simplification of the actual behavior and fails to capture some main failure
mechanisms, such as Mechanisms (a) and (b) depicted in Figure 1. A study by Stavridis (2009) using
detailed nonlinear finite element models has demonstrated that the compressive stress field in a
masonry infill wall may not be accurately represented by a single diagonal strut. As shown in Figure 12,
the distribution and orientation of compressive forces developed in an infill wall can be quite different
from that introduced by a single diagonal strut. Hence, replacing a wall by a diagonal strut will not lead
to a realistic representation of the frame‐wall interaction. Furthermore, it will not simulate the possible
shear failure of a column that could be induced by the frame‐wall interaction. Models to simulate the
shear and axial load failures of columns using zero‐length springs have been proposed (e.g., Elwood
2004). The zero‐length spring concepts can be combined with a strut model in the way shown in Figure
13 to account for the axial and shear failures of a column as a result of the frame‐wall interaction.
However, three issues are associated with such a model. One is that it ignores the shear transfer that
develops between the beam and the wall, as shown in Figure 12, and it will, thereby, induce
unrealistically large shear demands on the columns causing their premature shear failure. Second, such
a model ignores the variation of the axial force along the height of each column due to the interface
stresses introduced by the infill. Finally, in such a model, the shear failure of a column may result in the
immediate collapse of the column under the axial load, which will lead to the collapse of the frame
without being able to account for the vertical load carrying capacity of the wall. Moreover, the
orientation of the compressive forces in a wall can change as damage evolves, while the angle of a
diagonal strut is fixed. Multiple‐strut models as proposed in some studies might partially overcome
these problems. Nevertheless, this will increase the complexity of a model and its benefits remain to be
demonstrated.
Vtot = 591 kN
Figure 12 – Strut model vs. forces on an infill from Figure 13 – Strut model with zero zero‐thickness
finite element analysis. springs.
7.12
Analysis of Seismic Response of Masonry-Infilled RC Frames through Collapse
Lateral Load
Qmax
Qy K2
1
Qres
K
1
y Q
max
Q Story Drift
res
(a) Load‐vs.‐displacement envelop for an infilled (b) Comparison of the simplified curve with the
fame experimental result for the structure shown in
Figure 10
Figure 14 ‐ Comparison of a simplified curve with experimental result.
In spite of the aforementioned issues, the use of strut models is probably the most efficient way for
Monte‐Carlo‐type simulations. For this purpose, one can treat diagonal struts as purely
phenomenological models. However, the struts should be calibrated in such a way that they not only
reflect the behavior of the infill walls but also the response of the frame members associated with
mechanisms that are not represented in the model, such as the shear failure of the RC columns. Two
steps are needed for such calibration. The first is to divide a multi‐bay, multi‐story, structure into
multiple single‐bay, single‐story frames, and the second is to derive the lateral load‐vs.‐displacement
curve for each of these frames. The second step can be accomplished by creating a detailed finite
element model for each bay of the structure. However, this can be impractical for structures with many
bays and stories. To circumvent this difficulty, Stavridis (2009) has used experimental data and finite
element analysis results to derive a set of simple rules to define ASCE 41‐type pushover curves for
infilled frames. His study focused on non‐ductile RC frames with strong infill walls consisting of multi‐
wythe solid clay brick units, which typically lead to Failure Mechanism (b) shown in Figure 1. The
idealized lateral‐vs.‐displacement curve adopted is shown in Figure 14 and it is calibrated with the
following steps.
1. The initial stiffness of an uncracked infilled frame is calculated with a shear beam model as proposed
by Fiorato et al. (1970).
1
K (1)
1 1
K fl K shl
in which K fl and K sh represent the flexural and shear stiffness of an uncracked cantilever wall. With this
approach, the structure is assumed to be a composite beam with the RC columns being the flanges and
the masonry wall being the web of the beam. Hence, for the flexural stiffness, K fl , the equivalent
properties of the composite beam should be used, but for the shear stiffness, only the contribution of
7.13
P. B. Shing and A. Stavridis
empirical equations have been proposed by Stavridis (2009) based on the parametric study.
1
Q 0.86 AR w for AR w 2.15 (4a)
max
3
Q max
0.15 for AR w 2.15 (4b)
7.14
Analysis of Seismic Response of Masonry-Infilled RC Frames through Collapse
corresponds to a brittle behavior with a high post‐peak negative stiffness.
The aforementioned procedure has been calibrated only for non‐ductile RC frames infilled with solid
masonry panels. All these frames exhibit Failure Mechanism (b) in Figure 1, with shear cracks developed
in the infill and the columns. If a different failure mechanism develops due to different infill and frame
properties, the pushover curve has to be recalibrated. It is also important to point out that this study has
considered in‐plane seismic loads only with the out‐of‐plane loads ignored.
The load‐vs.‐displacement relation of an infilled frame is also affected by the existence of window and
door openings in the infill walls. The openings reduce the strength and stiffness of an infilled frame.
Stavridis (2009) has examined the behavior of 30 infilled frames with various opening shapes, sizes, and
locations in the infill using nonlinear finite element models. In his study, the ratio of the opening area to
the gross surface area of the wall is between 7 and 19%. Based on this study, he has proposed the
following recommendations to derive a conservative backbone curve for infilled frames with openings.
1. The initial stiffness of a frame with an opening is to be estimated with the following formula.
K
1 RA (6)
K solid
in which K solid is the initial stiffness of the same frame with a solid infill calculated with Equation 1,
RA Aop AWtot , which is the ratio of the opening area, Aop , to the area of the solid infill wall, AWtot , and
α is a factor depending on the shape of the opening. Based on the results of the parametric study, it has
been suggested that 2 for infills with a window and 1.6 for infills with a door.
2. The yield force, Q y , corresponds to the load at which cracks initiate at the corners of the openings.
Based on the parametric study, the ratio, Qy / Qmax , is between 65 and 80%, similar to the case with
solid infill walls. Hence, this ratio can be assumed to be 2/3 for infilled frames with openings as well.
3. For RA between 7 and 19%, Qmax can be conservatively estimated as follows.
Qmax 0.8 Qmax
solid
(7)
Solid
in which Qmax is the peak strength of the same frame with a solid infill.
7.15
P. B. Shing and A. Stavridis
4. The drift at the peak load, Qmax , can be conservatively assumed to be equal to that for the frame with
a solid masonry panel, which can be estimated with Equation 4. Alternatively, the drift, Qmax , can be
determined with the post‐yield hardening stiffness, K 2 , as shown in Figure 14(a). This stiffness can be
estimated as a fraction of the initial stiffness.
K2 K (8)
with recommended to be 0.10 for infills with a window and 0.08 for infills with a door.
5. The residual strength Qres can be determined as a fraction of the peak strength.
Qres 0.5 Qmax (9)
6. The drift Qres , can be assumed to be equal to 1.40 Q , which is the same as that for frames with
max
solid infills.
Once the backbone curve for an infilled frame has been defined with the procedure described above or
with a finite element analysis, the curve can be used to determine the load‐displacement relation for an
equivalent diagonal strut. It should be emphasized that this curve accounts for the load degradation
exhibited by an infilled frame due to the shear and/or flexural failure of the RC columns as well as
damage in the infill. For a multi‐bay, multi‐story, structure, the calibration can follow these steps:
Step 1: Develop the backbone curve for the lateral load‐vs.‐drift relation for each bay and each story of
the structure using the aforementioned procedure or a finite element model.
Step 2: Construct a bare RC frame model for each bay and each story of the structure, and determine
the lateral load‐vs.‐displacement backbone curve for the frame considering the appropriate gravity load.
Step 3: Calibrate the strut model so that when it is added to the bare frame considered in Step 2, one
can obtain the behavior of the infilled frame estimated in Step 1.
Step 4: Develop a model for the entire structure consisting of the frame elements determined in Step 2
and the diagonal struts calibrated in Step 3.
The above method was applied to model the response of the three‐story infilled frame tested on a
shaking table by Stavridis et al. (2012), with the structural configuration shown in Figure 11. The frame
model with equivalent diagonal struts is shown in Figure 15(a). The analysis was carried out with the
software platform OpenSees. The axial behavior of the diagonal struts was modeled with a constitutive
law for concrete, which was calibrated with simplified pushover curves derived with the procedure
described above. The strut for each bay was individually calibrated. The load‐displacement curves for
the infilled frame models representing the two bottom‐story bays are shown in Figures 15(b) and 15(c).
One was for the solid infill while the other was for the infill with a window opening. In both cases, the
numerical model in OpenSees deviated slightly from the idealized (simplified) pushover curve because of
the specific constitutive law used. In Figure 15(d), the base shear‐vs.‐bottom story drift hysteresis curves
7.16
Analysis of Seismic Response of Masonry-Infilled RC Frames through Collapse
obtained from the dynamic analysis of the three‐story frame with the strut‐based model, for a sequence
of ground motion records up to the 120% Gilroy motion, are compared to the test results. The model
shows more rapid load degradation than the test results due to the conservatism introduced in the
simplified pushover curves. The quality of the simulation could probably be improved if a detailed finite
element model was used to calibrate the model in place of the simple rules.
This type of phenomenological model can be extended to simulate the response of an infilled frame to
out‐of‐plane loads by using two bi‐axial fiber‐section beam‐column elements to represent a diagonal
strut. A fiber‐section beam‐column model can simulate both the arching mechanism and the out‐of‐
plane failure of an infill wall. This approach has been considered by Kadysiewski and Mosalam (2008),
who have proposed an interaction curve to estimate the capacity of an infill wall subjected to
simultaneous in‐plane and out‐of‐plane loads. The proposed interaction curve is calibrated with finite
element models. This approach deserves to be further explored and evaluated with experimental data
(e.g., the data of Dawe and Seah 1989, Angel et al. 1994, Mander et al. 1993, Bashandy et al. 1995, and
Flanagan and Bennett 1999).
700
400
80
300
200
40
100
0 0
0 0.2 0.4 0.6 0.8 1 1.2
Drift ratio, %
(a) Diagonal strut model (b) Pushover analysis of a bay with solid infill
1800 400
700
1350 Shake-Table Tests 300
600 OpenSEES model
Simplified curve 900 Strut model 200
120
Base shear, kips
Base shear, kN
450 100
Lateral force, kN
400 0 0
80
300 -450 -100
-900 -200
200
40
-1350 -300
100
-1800 -400
0 0 -1.5 -1 -0.5 0 0.5 1
0 0.2 0.4 0.6 0.8 1 1.2 st
1 Story drift, %
Drift ratio, %
(c) Pushover analysis of an infilled bay
(d) Hysteretic curves from dynamic analysis
with a window
to up 120% Gilroy
Figure 15 – Frame analysis using equivalent diagonal struts.
7.17
P. B. Shing and A. Stavridis
COLLAPSE CRITERION
The detailed finite element modeling and simplified analysis methods presented above can be used to
simulate the response of an infilled frame up to a severe damage stage close to collapse. Nevertheless,
neither method can simulate the actual collapse of an infilled frame in a satisfactory manner without
further enhancements. For Monte‐Carlo‐type simulations, the use of detailed finite element models is
not practical and simplified strut‐based models should be used.
Regardless of the type of model being used, one critical issue in a collapse simulation of any structure is
the definition and identification of collapse. While collapse can be defined as the loss of the vertical load
carrying capacity of a structure, the distinction between the partial and total collapse needs to be made.
The simulation of either the partial or the total collapse of an infilled frame is a big challenge because of
the interaction and possible changes of the load transfer mechanisms among the columns, beams, and
walls. Once the RC columns in an infilled frame have suffered severe damage such as shear failure, the
gravity load can be shifted to the infill walls as long as these walls remain standing. A severely cracked
masonry wall could still carry significant gravity load even though it might have lost a large portion of its
in‐plane and out‐of‐plane lateral load resisting capacity. This can be perceived as an unstable
equilibrium state, which has been often observed in in‐plane quasi‐static tests (see Figure 4) as well as
nonlinear finite element analyses (Koutromanos et al. 2011). Hence, the collapse of an infilled frame has
to be preceded by the collapse of the infill walls. The collapse of infill walls could occur before or after
the RC frame has suffered severe damage. The walls could collapse prior to the development of severe
damage in the frame if the masonry is weak or large openings exist in the walls, and this will
immediately result in a weak‐story mechanism. For this situation, simple models like those proposed by
Elwood (2004) can be used to simulate the subsequent collapse of the RC columns.
In view of the aforementioned modeling challenges and the large uncertainties in the collapse
prediction for an infilled frame, it is prudent to treat the first attainment of an unstable equilibrium state
of the structure as the total collapse condition. This can be considered as a state in which the infilled
frame has lost a significant portion of its lateral in‐plane load carrying capacity due to the failure of the
walls at one or more stories. Until more experimental data are available, this state can be assumed to be
reached when the post‐peak in‐plane resistance of the infilled frame drops to 60% of the peak. The
story‐drift level at which this load degradation is reached can be taken as the collapse criterion for
Monte‐Carlo‐type simulations. This critical drift limit can be identified with a pushover analysis of a
planar structure using a detailed finite element model or a simplified model. Experimental results from
different studies have shown that under in‐plane loads only, this can occur at a story‐drift ratio between
1.0 and 1.5%. Quasi‐static tests and finite element analyses have shown that frames with weak infill
walls or walls with a large opening tend to have lower strengths but higher ductile. Hence, under static
loads, these frames may not reach 40% strength loss till a relatively large drift level has been reached.
Nevertheless, walls in these frames will be less stable under dynamic loads as mentioned in a previous
section. Therefore, to be prudent, the critical drift limit should not be greater than 1.5%. If a wall has
been damaged by the out‐of‐plane load, it can be expected that the in‐plane resistance will drop more
rapidly and a strength loss of 40% can occur at a smaller story‐drift level. Further studies are needed to
establish a drift limit for such cases.
CONCLUSIONS
Simplified models using equivalent compression‐only diagonal struts are most efficient for Monte‐Carlo‐
type collapse simulations for masonry‐infilled RC frames. However, because of the inability of such
models to represent some main failure mechanisms of an infilled frame, the strut model should be so
calibrated that the load degradation induced by the unaccounted mechanisms, such as the shear failure
7.18
Analysis of Seismic Response of Masonry-Infilled RC Frames through Collapse
of the RC columns, can be represented in the load‐displacement response of the entire frame. To this
end, the axial behavior of a strut can be modeled with a phenomenological law that reflects the damage
evolution of an entire infilled frame. Such a strut model can be calibrated with a detailed nonlinear finite
element analysis or the simple procedure described in the paper. The simple procedure presented here
was derived for structures that have non‐ductile RC frames and strong infill walls. Further work is
needed to extend the procedure to structures with different frame and wall properties.
In a simplified model, RC beams and columns can be modeled with beam‐column elements with flexural
hinging capabilities. To simulate the damage and possible collapse of an infill wall caused by out‐of‐
plane loads, a diagonal strut can be modeled with two bi‐axial fiber‐section beam‐column elements,
which account for the out‐of‐plane bending as well as the arching mechanism developed in an infill wall.
Nevertheless, further work is needed to calibrate these models with available experimental data.
For Monte‐Carlo‐type collapse simulations, a convenient criterion for collapse must be determined.
Until more data are available, it can be assumed that an infilled frame will be on the verge of collapse
when its in‐plane resistance drops to 60% of the peak resistance. This is based on test data currently
available. The story‐drift level at which this load degradation will be reached under monotonically
increasing loads can be taken as a collapse criterion. This drift limit can be identified with a pushover
analysis of a planar structure using a detailed finite element model or a simplified model, but the critical
drift ratio for collapse should not be greater than 1.5%.
For future numerical studies and the calibration of simplified models, finite element models that can
simulate collapse in a detailed fashion are desired. To simulate out‐of‐plane failures, 3‐D finite element
models are needed. Other simplified collapse simulation strategies such as element removal methods
should also be investigated. Different modeling methods should be compared and evaluated with
benchmark examples.
ACKNOWLEDGMENTS
The modeling work presented in this paper is based on a prior research project supported by the
National Science Foundation Grant No. 0530709 awarded under the George E. Brown, Jr. Network for
Earthquake Engineering Simulation Research (NEESR) program. However, opinions expressed in this
paper are those of the authors and do not necessarily represent those of the sponsor.
REFERENCES
Angel, R, Abrams, D, Shapiro, D, Uzarski, J, and Webster, M, 1994, “Behavior of Reinforced Concrete
Frames with Masonry Infills,” Report No. UILU‐ENG‐94‐2005, Dept. of Civil Engineering, University of
Illinois, Urbana‐Champaign, IL.
Bashandy, T, Rubiano, NR, and Klingner, RE, 1995, “Evaluation and Analytical Verification of Infilled
Frame Test Data,” Report No. 95‐1, Dept. of Civil Engineering, University of Texas, Austin, TX.
Blackard, B, Willam, K, and Mettupalayam, S, 2009, “Experimental Observations of Masonry Infilled RC
Frames with Openings,” ACI SP 265‐9, American Concrete Institute, pp. 199‐222.
Dawe, JL and Seah, CK, 1989, “Out‐of‐Plane Resistance of Concrete Masonry Infilled Panels,” Journal of
the Canadian Society for Civil Engineering, V. 16, pp. 854‐864.
Elwood, KJ, 2004, “Modelling Failures in Existing Reinforced Concrete Columns,” Canadian Journal of
Civil Engineering, V. 31, pp. 846‐859.
Fiorato, AE, Sozen, MA, and Gamble, WL, 1970, “An Investigation of the Interaction of Reinforced
Concrete Frames with Masonry Filler Walls,” Report No. UILU‐ENG 70‐100, Dept. of Civ. Eng., University
of Illinois at Urbana‐Champaign, Urbana, IL.
7.19
P. B. Shing and A. Stavridis
Flanagan, RD and Bennett, RM, 1999, “Bidirectional Behavior of Structural Clay Tile Infilled Frames,”
Journal of Structural Engineering, V. 125, No. 3, pp. 236‐244.
Holmes, M, 1961, “Steel Frames with Brickwork and Concrete Filling,” Proceedings of the Institute of
Civil Engineers, V. 19, No. 6501, pp. 473‐478.
Kadysiewski, S and Mosalam, KM, 2009, “Modeling of Unreinforced Masonry Infill Walls Considering
In‐Plane and Out‐of‐Plane Interaction,” PEER Report 2008/102, Pacific Earthquake Engineering Research
Center, University of California at Berkeley, Berkeley, CA.
Koutromanos, I, 2011, “Numerical Analysis of Masonry‐Infilled Reinforced Concrete Frames Subjected
to Seismic Loads and Experimental Evaluation of Retrofit Techniques,” Ph.D. Dissertation, University of
California at San Diego, La Jolla, CA.
Koutromanos, I, Stavridis, A, Shing, PB, and Willam, K, 2011, “Numerical Modeling of Masonry‐Infilled
RC Frames subjected to Seismic Loads,” Computers and Structures, V. 89, pp. 1026‐1037.
Koutromanos, I and Shing, PB, 2012, “A Cohesive Crack Model to Simulate Cyclic Response of Concrete
and Masonry Structures,” ACI Structural Journal, V. 109, No. 3, pp. 349‐358.
Lotfi, HR and Shing, PB, 1991,” An Appraisal of Smeared Crack Models for Masonry Shear Wall
Analysis,” Computers & Structures, V. 41, No. 3, pp. 413–425.
Mainstone, RJ and Weeks, GA, 1970, “The Influence of Bounding Frame on the Racking Stiffness and
Strength of Brick Walls,” Proceedings of the 2nd International Conference on Brick Masonry, Stoke‐on‐
Trent, pp. 165‐171.
Mander, JB, Nair, B, Wojtkowski, K and Ma, J, 1993, “An Experimental Study on the Seismic
Performance of Brick‐Infilled Steel Frames with and without Retrofit,” Report No. NCEER‐93‐0001,
National Center for Earthquake Engineering Research, State University of New York, Buffalo, NY.
Mehrabi, AB, Shing, PB, Schuller, MP, and Noland, JL, 1994, “Performance of Masonry‐Infilled R/C
Frames under In‐Plane Lateral Loads,” Report No. CU/SR‐94‐6, Dept. of Civil, Environmental, and
Architectural Engineering, University of Colorado, Boulder, CO.
Mehrabi, AB, Shing, PB, Schuller, MP, and Noland, JL, 1996, “Experimental Evaluation of Masonry‐
Infilled RC Frames,” Journal of Structural Engineering, V. 122, No. 3, pp. 228‐237.
Murcia‐Delso, J, 2013, “Bond‐slip Behavior and Development of Bridge Column Longitudinal
Reinforcing Bars in Enlarged Pile Shafts,” PhD Dissertation, University of California, San Diego, La Jolla,
CA.
Rots, JG, 1991, “Numerical Simulation of Cracking in Structural Masonry,” HERON, Netherlands School
for Advanced Studies in Construction, The Netherlands, V. 36, No. 2, pp. 49‐63.
Stafford Smith, B, 1967, “Methods for Predicting the Lateral Stiffness and Strength of Multi‐Story
Infilled Frames,” Building Science, V. 2, pp. 247‐257.
Stavridis, A, 2009, “Analytical and Experimental Study of Seismic Performance of Reinforced Concrete
Frames Infilled with Masonry Walls,” Ph.D. Dissertation, University of California at San Diego, La Jolla,
CA.
Stavridis, A and Shing, PB, 2010, “Finite Element Modeling of Nonlinear Behavior of Masonry‐Infilled
RC Frames,” Journal of Structural Engineering, V. 136, No. 3, pp. 285‐296.
Stavridis, A, Koutromanos, I, and Shing, PB, 2012, “Shake‐Table Tests of a Three‐Story Reinforced
Concrete Frame with Masonry Infill Walls,” Journal of Earthquake Engineering and Structural Dynamics,
V. 41, pp. 1089‐1108.
7.20