EARTHQUAKE ENGINEERING & STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2015; 44:1783–1803

Published online 29 January 2015 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/eqe.2555

Seismic fragility of lightly reinforced concrete frames with masonry

infills

Jong-Su Jeon1, Ji-Hun Park2,*,† and Reginald DesRoches1

1
School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA
2
Division of Architecture and Urban Design, Incheon National University, Incheon 406-772, Korea

SUMMARY
Seismic fragility of lightly reinforced concrete frames with masonry infills is assessed through numerical
simulations considering uncertainty in ground motion and building materials. To achieve this aim, a numer-
ical model of the components is developed, a rational approach to proportion and locate individual struts in
the equivalent three-strut model is proposed, and an explicit nonlinear column shear response model
accounting for the infill–column interaction and soft-story mechanism is employed. The proposed numerical
model is used to (1) generate probabilistic seismic demand models accounting for a wide range of ground
motion intensities with different frequency content and (2) determine limit state models obtained from
nonlinear pushover analysis and incremental dynamic analysis. Using the demand and limit state model,
fragility curves for the masonry-infilled frames are developed to investigate the impact of various infill prop-
erties on the frame vulnerability. It is observed that the beneficial effect of the masonry infill diminishes at
more severe limit states because of the interaction with the boundary frame. In some cases, this effect almost
vanishes or switches to an adverse effect beyond a threshold of ground motion intensities. Copyright © 2015
John Wiley & Sons, Ltd.

Received 21 February 2014; Revised 19 December 2014; Accepted 30 December 2014

KEY WORDS: masonry-infilled reinforced concrete frame; equivalent strut model; infill–column
interaction; probabilistic seismic demand model; limit state model; seismic fragility

1. INTRODUCTION

Masonry-infilled reinforced concrete (RC) frames are composed of masonry infills and boundary
frames. In the past or current design practice for many countries, the infill has been regarded as a
non-structural component of which stiffness and strength are neglected. However, many studies have
pointed out significant influence of the infill on the seismic response of the structure. Traditionally, it
has been thought that the infill makes a beneficial contribution to the mitigation of seismic response
by enhancing the stiffness and strength of the frame [1–3]. Despite this advantage for the use of the
infill, the flexural or shear failure of the boundary frame because of the interaction between the infill
and boundary frame is a significant potential failure mechanism [4–6], which can be observed in
static and dynamic testing [7, 8]. Thus, overall seismic response of the infilled frames depends on
which one of the two contrary effects is dominant. Additionally, early failure of the infill at lower
stories causes vertical irregularity by forming a soft story after the collapse of the infill [9–11].
The complex behavior of the masonry-infilled RC frames necessitates more sophisticated models and
reliable analysis techniques that can incorporate uncertainty in ground motions and structural properties.
In spite of recent progress in computing power, continuum-based finite element models are still

*Correspondence to: Ji-Hun Park, Division of Architecture and Urban Design, Incheon National University, Incheon
406-772, Korea.
†
E-mail: [email protected]

Copyright © 2015 John Wiley & Sons, Ltd.

1784 J.-S. JEON, J.-H. PARK AND R. DESROCHES

computationally intensive, especially when conducting seismic risk assessment demanding numerous
simulations. Instead, an equivalent strut modeling approach that substitutes for the stiffness and
strength of infill panels has widely been employed in standards and research as an efficient alternative
to the continuum-based finite element model. Numerous equations are proposed for the stiffness and
strength of the equivalent strut. However, there is little consensus on which equation is the most
accurate and reliable [12]. This is attributed to the fact that the formulation of most equations as well
as comparison between them is based on limited testing results for specimens with local masonry
properties. Consequently, when evaluating the seismic performance and risk assessment of these
structures, the infill contains a high amount of uncertainty, which increases uncertainty in the system
response and failure mechanism through complex interaction with the uncertain boundary frame.
Therefore, a probabilistic framework for seismic fragility assessment accounting for uncertainty in
both seismic demand and capacity is required for the masonry-infilled RC frames.
Numerous studies on the fragility assessment of the masonry-infilled RC frames are available in the
literature [13–16]. Most of them have adopted a stripe method that utilizes incremental dynamic analysis
(IDA) [17], in which a ground motion is scaled up until the global system collapses without changing its
frequency content that varies with respect to ground motion intensity. However, this class of frames
experiences dramatic change in the stiffness and corresponding natural frequency so that their elastic
response at the serviceability limit state and inelastic response at the collapse limit state are relevant
to completely different frequency regions located far from each other in the response. Accordingly,
applying different levels of scaled ground motions with identical frequency content can distort
seismic demand on such frames. Moreover, most of earlier studies on the fragility assessment of the
masonry-infilled RC frames did not reflect column shear failure resulting from interaction between
the infill and boundary frame. Recently, Verderame et al. [11] used an equivalent strut model to
identify column shear failure only by post processing without explicitly simulating the column shear
failure. Celarec and Dolšek [16] not only adopted a strut model to simulate the behavior of masonry
infills for the seismic fragility assessment but also introduced a simple iterative pushover-based
procedure to approximate column shear failure with flexural hinges, which enables the application of
the method on more complicated structures. Also, nonlinear static pushover analysis (NSP) or
nonlinear time history analysis (NTHA) was performed combining multiple strut models with
nonlinear shear deformation models to identify the collapse capacity of the masonry-infilled RC
frames [18, 19].
The current study addresses the seismic fragility of lightly reinforced concrete frames with masonry
infills, in which the term lightly reinforced is used to refer to concrete frames that are designed against only
gravity load without considering masonry infills as structural elements, and, as a result, these frames have
relatively wide hoop spacing (larger than half of the section depth) and a strong possibility of column
shear failure when subjected to seismic load. The primary objectives of the current study are to (1)
propose explicit numerical models for infills and columns that can simulate the primary earthquake
response characteristics of masonry-infilled RC frames using OpenSees [20] and (2) employ these
component response models to assess the seismic vulnerability of lightly reinforced concrete frames
with masonry infills. A rational approach to proportioning and locating multiple struts as well as
accounting for column shear failure associated with infill–frame interaction is proposed and validated
using existing experimental results. For the fragility assessment of this class of frames, a gravity-load-
designed masonry-infilled school building in Korea is selected as a case study, and various frame
models are constructed with different infill characteristics. For each frame model, a demand model is
developed using an NTHA-based cloud method accounting for uncertainty in ground motions and
material properties and is then convolved with multiple limit states obtained from NSP and IDA to
develop fragility curves of the frame. Finally, median intensity measures resulting from the fragility
curves are estimated to examine the impact of infill characteristics on the frame vulnerability.

2. NUMERICAL MODEL OF MASONRY-INFILLED RC FRAMES

On the basis of typical potential failure mechanisms for masonry-infilled RC frames, numerical models
for infills and columns are proposed and validated using experimental results available in the literature.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
 SEISMIC FRAGILITY OF LIGHTLY REINFORCED CONCRETE FRAMES WITH INFILLS 1785

2.1. Potential failure mechanisms

The failure mechanism of RC frames with masonry infills comprises two categories of component
failure modes associated with infill panels and boundary frames, respectively. The combination of
different component failure modes can give rise to various (individual versus combined) potential
failure mechanisms. The failure of the infill panel includes three common component failure
modes such as sliding shear (SS), diagonal cracking (DK), and corner crushing (CC) [5–7, 21];
the SS mode is initiated by horizontal cracks in the mortar joints between masonry units; the
DK mode is caused by diagonal cracks across the panel; and the CC mode results from the
crushing in the corner of the panel. The failure of the boundary frame includes flexural failure,
shear failure, beam–column joint failure, and tension failure. Figure 1 shows some examples of
the failure mechanisms identified by the experiment of Meharabi et al. [7], where the SS mode
is termed ‘mid-height horizontal crack’ and ‘horizontal slip’. However, these potential failure
mechanisms are closely associated with the interaction between the infill and boundary frame,
and this combined action can make it more difficult to accurately predict the most likely failure
mechanism of masonry-infilled RC frames. Liauw and Kwan [22] and Mehrabi et al. [7]
suggested simple analytical methods to predict the lateral strength of their masonry-infilled RC
frame specimens using the lateral strength of both infill and boundary frame components and
obtained good agreement between the observed and predicted failure mechanisms as well as
lateral strengths. Such methods can be utilized as criteria to determine the component failure
mode and corresponding strength of each infill panel prior to performing dynamic analysis. The
component failure mode of the boundary frame can be determined during the analysis because
fluctuating axial forces affect the shear and moment capacities of columns.

2.2. Numerical response model

2.2.1. Masonry infill model. The most widely used modeling technique in the performance
evaluation and risk assessment of masonry-infilled RC frames is the equivalent strut modeling
approach, which has an ability to predict the contribution of infills to frame response in terms of
sufficient accuracy and computational efficiency when compared with continuum-based finite
element models. The equivalent strut models can be classified by the number of struts that
represents the diagonal strut action of the infill. A single strut model is the simplest one to
capture the contribution of the infill on the frame response, but it cannot simulate the interaction
between the infill and boundary frame without additional effort (e.g., iterative pushover-based
procedure [16]) because of the connectivity of the strut to the beam–column connections. This
interaction behavior can be simulated using multiple (two or three) strut models. In the current
study, the three-strut model developed by Chrysostomou et al. [23] has been adopted, as
shown in Figure 2, because it has more flexibility than the two-strut model in adjusting
the interaction forces between an infill panel and two boundary columns by proportioning the
three struts.
The strength proportions assigned to two off-diagonal struts and corresponding location of the
column and beam nodes, where the off-diagonal struts are connected, are determined on the basis of
bearing stress distribution along a contact length between the column and infill panel, as illustrated

Figure 1. Examples of potential failure mechanisms for masonry-infilled RC frames [7].

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
1786 J.-S. JEON, J.-H. PARK AND R. DESROCHES

Figure 2. Three-strut model of masonry infill panel.

in Figure 3. The contact length (z) is computed following the approach proposed by Stafford Smith
[24], which uses an analogy to the theory of an elastic beam on an elastic foundation:

π
where z¼ ≤ hm (1)
2λ
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4 E m t m sin 2θ
λ¼ (2)
4E c I c hm

and Em, tm, and hm are the elastic modulus, thickness, and height, respectively, of the infill panel, θ is
the angle between the diagonal and base of the infill panel, and Ec and Ic are the elastic modulus and
moment of inertia, respectively, of the column.
Upper and lower triangular parts of the actual infill panel located above and below the diagonal
are assumed to transfer an equal amount of strut action, as proposed by El-Dakhakhni et al. [21] on
the basis of observations by many researchers, so that 50% of the total strut action (F) in Figure 3
is transferred to the column and induces the bearing stress. However, this proportion relies on
various conditions including the aspect ratio of infill panels [25], and 64% and 50% of the infill
lateral force were found to be added to the column shear force through numerical analysis by
Combescure [25] and Celarec and Dolšek [16], respectively. The bearing stress distribution is
proposed on the same theoretical basis as the contact length. The portion of the column in
contact with the infill panel is regarded as a half of an elastic beam on an elastic foundation, of
which role is played by the infill panel. Deflection of the column along the contact length is
given by Hetenyi [26]:

Figure 3. Determination of off-diagonal strut node location.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
 SEISMIC FRAGILITY OF LIGHTLY REINFORCED CONCRETE FRAMES WITH INFILLS 1787

y cosðwÞ½coshðπ  wÞ þ coshðwÞ þ sinðwÞ½sinhðπ  wÞ  sinhðwÞ

¼ (3)
yc 1 þ coshðπÞ
where
 
π x
w¼ (4)
2 z

and y and yc are defined in Figure 3. The strain of the infill panel normal to the contact face is assumed
to be proportional to y as in the elastic foundation. A parabolic stress–strain relationship [27] is
assumed with a peak at the maximum strain of the infill panel. At the critical section of column
shear failure, assumed to be at the column effective depth d from the beam face, the contact length
is divided into two parts: one in contact with the lower portion of the central strut and the other in
contact with the lower off-diagonal strut. Proportions of the partial strut action FC2 and FL to the
half strut action 0.5F (Figure 3) is determined on the basis of the corresponding areas of the bearing
stress distribution. Proportions of FC1 and FU to 0.5F are determined in the same fashion, but the
bearing stress distribution on the opposite side of the panel is used. Consequently, the strength and
stiffness of the infill panel is distributed among three struts using the following three distribution
factors, as shown in Figure 2.

F C1 þ F C2 FU FL
γC ¼ ; γU ¼ ; γL ¼ (5)
F F F

where γC, γU, and γL are the distribution factors for the central, upper, and lower off-diagonal struts,
respectively, and summed up to be 1.0. The off-diagonal strut is connected to the boundary column
at the centroidal distance zc from the beam face for the bearing stress distribution and to the
boundary beam in parallel with the central strut. It is noted that zc needs to be checked to ensure the
basic configuration for equivalent struts in Figure 2, which is the basis of the column shear response
model described in Section 2.2.2. It is assumed that one-half of the total strut force is transferred to
each boundary column and allocated to the off-diagonal and central strut on the basis of the bearing
stress distribution. Both the central and off-diagonal strut responses are simulated using truss
elements with the OpenSees Hysteretic material model.
The equivalent strut model proposed in the current study employs the force–displacement envelope
model developed by Fardis [3], as illustrated in Figure 2. The elastic, second, and post-peak stiffnesses
(Ke, Ksec, and Kdeg) on the envelope are given by

Gm t m L m
Ke ¼ (6)
hm

E m bm t m
K sec ¼ cos2 θ (7)
dm

K deg ¼ αK e (8)

where
h i
bm ¼ 0:175ðλhÞ0:4 d m (9)

and Gm is the shear modulus of the infill panel and calculated as 0.4Em according to ASCE 41-06
[28], dm, h, and α are the diagonal length of the infill panel, height of the story, and post-peak
stiffness ratio, respectively, and bm is the width of the equivalent strut proposed by Mainstone
[1]. The relationship between the peak strength (Vmax) and cracking strength (Vcr) on the
response envelope can be expressed as follows [3]:

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
1788 J.-S. JEON, J.-H. PARK AND R. DESROCHES

V cr ¼ V max =1:3 (10)

where

V max ¼ 1:3τ d t m Lm (11)

and τ d is the diagonal shear strength of the infill panel. Equation (11) represents the DK mode of infill
failure, because the cracking and peak lateral forces are calculated using the cracking stress obtained
from diagonal tension tests. Equation (11) can be replaced with other existing lateral strength
models (Vmax), and other component failure modes of the infill can be considered in such a way.
The best strength model for each component failure mode will be selected in model validation
presented in Section 2.3.

2.2.2. Column shear response model. The flexural response of the boundary frame is modeled
using fiber-type displacement-based beam-column elements over the whole range of the members.
The confinement of concrete is considered using the model of Mander et al. [29] to account for
the increased concrete compressive strength and ductility of the columns. Concrete and
longitudinal reinforcement are modeled as the OpenSees Concrete02 and Hysteretic material
model with a hardening factor of 0.01, respectively. Two types of column shear failure in
masonry-infilled RC frames are considered in the numerical response model. One is shear failure
due to the strut action of an infill panel, which leads to the concentration of the shear force at
the interface of the column and infill panel. This shear failure is a force-controlled action,
occurring at a smaller story drift prior to infill failure. The other is flexure–shear failure
(deformation-controlled action) as a result of soft story mechanism, which is formed at lower
story levels after infill failure [10, 11, 15, 30]. As illustrated in Figure 4, the force-controlled
shear failure of the column is simulated using two zero-length springs at the face of the beams
and the column shear strength model developed by Sezen and Moehle [31], whereas the
deformation-controlled shear failure is modeled using a zero-length spring at the center of the
column and the limit state model proposed by Elwood [32]. All column shear models account
for the variation in the column axial force in the dynamic analyses to capture appropriate failure
mechanism. Because the location of the off-diagonal struts depends on the geometry of the infill
panel, applying the proposed column model (shown in Figure 4) to a column located between
two adjacent infill panels with different geometry requires two different off-diagonal strut
connection nodes for the left- and right-hand side infill panels between the middle and end shear
springs.

Figure 4. Three-spring model for column shear failure.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
 SEISMIC FRAGILITY OF LIGHTLY REINFORCED CONCRETE FRAMES WITH INFILLS 1789

2.3. Model validation

The numerical models for masonry-infilled RC frames are validated through the comparison with
experimental results tested by Mehrabi et al. [7]. To account for different failure mechanisms
associated with the combination of different infill and boundary frame types (i.e., strong frame with
hollow or solid bricks and weak frame with hollow or solid bricks), five single-story and single-bay
specimens are selected in the model validation and presented in Table I. Only three parameters such
as Vmax, Em, and α, which are related to the force–displacement envelope of the equivalent strut, are
employed for model calibration to reduce the number of parameters and listed along with simulation
errors in Table I.
The calibration procedure used in the current study consists of the following steps: (1) search for
appropriate Vmax to fit the ultimate strength; (2) determine the shape of the force–displacement
envelope prior to capping (Ke and Ksec) by adjusting Em; and (3) calibrate α, which determines the
descending slope of the lateral load–drift envelope and the drift at abrupt strength loss due to
column shear failure. However, the above steps are performed repeatedly to find better results,
because these calibration parameters influence one another in a complicated manner. When cyclic
loading test data for the calibration of numerical models are not available, Vmax and Em should be
determined on the basis of design documents or material tests, and more reliable strength models for
equivalent struts are required. Against such a case, several strength models for equivalent struts are
evaluated for modeling recommendations later in this section. Regarding α, Table I can provide
illustrative values, and the following rules can be useful. First, relatively higher α is appropriate for
hollow masonry units when compared with solid ones because of rapid degradation. Secondly, the
SS mode has higher displacement capacity than other modes [6] for the same type of masonry units.
Finally, further adjustment of α is possible to prevent excessive drift capacity of infill panels
according to available guidelines such as FEMA 306 [6].
Figures 5(a) and 5(b) show the comparison of the simulated and observed responses for S5 (with
column shear failure) and S7 (without column shear failure), respectively. Figure 5 and Table I
reveal that the simulated and experimental results are in good agreement overall. Unlike S7,
which is dominated by infill failure, as shown in Figure 5(b), S5 is governed by column shear
failure and exhibits significant strength degradation after capping, as shown in Figure 5(a). This
deterioration is attributed to the column shear failure at the upper end of both columns. The drift
at the shear failure level for S5 is simulated well with only a 14% error, as presented in Table I
(12% error of the drift for S11). However, a number of limitations can be noted for the
calibrated model. For example, the strength degradation of S7, shown in Figure 5(b), is rapid
immediately after capping and then slows down, but it is hard to achieve this response
characteristic using the infill envelope with linear strength degradation in Figure 2, which is a
major source of modeling errors. Asymmetrical peak strengths for two opposite loading
directions are another source of errors in Figure 5(a). Therefore, response models of masonry
infills with more complicated envelope and hysteretic rules can provide better calibration results.
Adding concrete model parameters to the calibration variable might be another solution.

Table I. Comparison of observed and simulated responses.

Specimen S4 S5 S6 S7 S11

Experiment Masonry unit Hollow Solid Hollow Solid Solid

Boundary frame Weak Weak Strong Strong Weak
Aspect ratio (H/L) 0.67 0.67 0.67 0.67 0.48
Max. lateral load (kN) 158 247 197 467 287
Calibration of strut Strut strength, Vmax (kN) 103.1 235.2 91.63 349.3 262.4
Elastic modulus of masonry, 4139 8055 4199 9074 7684
Em (MPa)
Post-peak stiffness ratio (α) 0.010 0.006 0.006 0.006 0.003
Simulation error (%) Max. lateral load 4.42 7.26 5.90 5.78 2.87
Energy dissipation 11.75 16.26 2.53 1.94 19.15
Drift at major shear crack — 14.36 — — 11.70

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
1790 J.-S. JEON, J.-H. PARK AND R. DESROCHES

Figure 5. Comparison of observed and simulated responses [7].

Several existing lateral strength models for masonry infills are evaluated for the specimens with
relevant failure mechanisms in Table II, where the smallest error for each specimen is given in
bold. Parameters used in the evaluation are given in Table III. For the SS mode, the sliding shear
strength by Stafford Smith and Riddington [33] accounts for the friction force due to only
compressive strut action so that the friction force due to the external vertical load is added on the
basis of the vertical stiffness ratio between the infill panel and two columns. The same vertical load
is also used to compute the sliding shear strength equation of ASCE 41-06 [28]. The bond shear
strength, friction coefficient, and compressive prism strength of the masonry are computed by the
results of material tests [7]. For the DK mode, various strength values σ cr, f′me90, and τ d are
calculated using empirical equations given in Drysdale and Hamid [34] and FEMA 306 [6]. Here,
0.275 in the model of Drysdale and Hamid [34] is the mean value of the lower and upper bounds of
the coefficient.
For the seismic fragility assessment in the subsequent sections, the model of Stafford Smith and
Riddington [33] is adopted for the SS mode because of its small error and ability to account for
the friction force caused by the strut action. For the DK mode, the models of Fardis [3] and
Bertoldi et al. [35] show the smallest error for S5 and S11, respectively, but the former is
chosen because of its tendency to overestimate the equivalent strut strength and corresponding
likelihood of column shear failure. For the CC mode, the model of Bertoldi et al. [35] shows
good agreement for S7 but gives a significant error in S4 because of the combined action of the
CC and SS modes.
Simple analytical models based on the three-strut model are proposed in Figure 6 for guidance to
determine the component failure mode of infill panels. The first three failure mechanisms in
Figure 1 are adopted for those models. For the SS mode, ‘horizontal slip’ is adopted rather than
‘mid-height horizontal crack’, which was not observed in the five specimens [7]. In these
models, the lateral resistance of the masonry-infilled frame denoted by Vu equals the sum of

Table II. Comparison of existing strength models for infill panels to calibrated strengths.
Specimen S4 S5 S6 S7 S11
Observed failure mechanism SS + CC DK + shear SS CC DK + shear
Infill failure mode Model
Strength error (%) SS Stafford Smith and 4.16 — 9.36 — —
Riddington [33]
ASCE 41-06 [28] 41.86 — 49.33 — —
Bertoldi et al. (SS) [35] 40.81 — 42.26 — —
DK FEMA 306 [6] — 62.23 — — 67.24
Fardis [3] — 11.10 — — 25.00
Bertoldi et al. (DK) [35] — 26.03 — — 19.00
CC Bertoldi et al. (CC) [35] 67.28 — — 6.85 —

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
 SEISMIC FRAGILITY OF LIGHTLY REINFORCED CONCRETE FRAMES WITH INFILLS 1791

Table III. Parameters assumed for the verification of strut models.

Specimen S4 S5 S6 S7 S11
Infill failure mode Parameter
SS Bond shear strength (MPa) [7] 0.345 0.345 0.345 0.345 0.345
Friction coefficient [7] 0.9 0.9 0.9 0.9 0.9
DK Diagonal shear strength, τ d = 0.275(f′me)0.5 0.896 1.023 0.875 1.013 0.930
(MPa) [34]
Horizontal strength of masonry, f′me90 = 0.5f′me 5.309 6.929 5.068 6.792 5.723
(MPa) [6]
Cracking strength, σ cr = f′me90/20 (MPa) [6] 0.265 0.347 0.253 0.340 0.286
CC Compressive prism strength, f′me (MPa) [7] 10.62 13.86 10.14 13.58 11.45

Figure 6. Analytical models in horizontal equilibrium to determine component failure modes.

infill panel and column strengths. The column strength denoted by Vp is the minimum of the shear
strength Vn and two times the flexural strength Mn divided by the shear span, which is the clear
column height hm for the SS and CC modes or the distance zc between two points connected to
off-diagonal and central struts for the DK mode. For the DK mode, only the central strut is
taken into account using the distribution factor γC. A component failure mode that produces the
lowest maximum lateral resistance for the masonry-infilled frame can be adopted for detailed
numerical analysis. For verification, maximum lateral loads are estimated in Table IV with three
different infill strength models chosen from Table II and the corresponding analytical models
using parameters in Table III. In Table IV, the lowest Vu for each specimen is highlighted in
bold, and the corresponding component failure mode agrees with experimental observation given
in Table II except for S7, for which the smallest error (12.6%) is obtained with the observed
failure mode.

3. PROCEDURE OF FRAGILITY ASSESSMENT

This section describes the proposed approach for the fragility assessment of masonry-infilled RC
frames. The fragility function used in the current study can be derived convolving a probabilistic
seismic demand model (PSDM) and a limit state model. The demand model that relates the median
demand to the intensity measure (IM) is generated using the results of a set of NTHAs and is

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
1792 J.-S. JEON, J.-H. PARK AND R. DESROCHES

Table IV. Analytical estimation of the maximum lateral load with selected strength models.
Specimen S4 S5 S6 S7 S11
Infill failure mode Model
Estimation (kN) SS Stafford Smith and Riddington [33] 195 303 232 355 345
DK Fardis [3] 210 271 320 403 305
CC Bertoldi et al. [35] 260 421 344 526 420
Error (%) SS Stafford Smith and Riddington [33] 23.4 22.7 18.2 24.1 19.9
DK Fardis [3] 33.0 9.76 62.8 13.7 6.05
CC Bertoldi et al. [35] 64.6 70.1 75.1 12.6 46.3

expressed using a power form proposed by Cornell et al. [36]:

SD ¼ aIM b (12)

Equation (12) can be rewritten in the log-transformed space as follows (Figure 7):

lnðSD Þ ¼ lnðaÞ þ blnðIM Þ (13)

where SD is the median value of the demand as a function of an IM and regression coefficients a and b
can be computed by a linear regression analysis of ln(SD) on ln(IM) obtained from simulations. The
dispersion (βD|IM) accounting for the uncertainty in the relation is estimated:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
  
∑Ni¼1 lnðd i Þ  ln aIM b
βDjIM ¼ (14)
N2
where N is the number of simulations. There are special cases where the demand in the log-transformed
space follows a bilinear trend with respect to the chosen IM, specifically when the demand dramatically
increases beyond a specified earthquake intensity [37]. In this case, the demand model in Equation (13)
will be in the form of

lnða1 Þ þ b1 lnðIM Þ; IM ≤ IM o
lnðSD Þ ¼ (15)
lnða1 Þ þ b1 lnðIM o Þ þ b2 ½lnðIM Þ  lnðIM o Þ; IM > IM o

The parameters a1, b1, b2, and IMo are illustrated in Figure 7. In the bilinear model, the breaking
point (IMo) can be determined by minimizing the sum of the square of residuals between actual and
fitted values [38]. The dispersions, βD|IM1 and βD|IM2, are calculated by using Equation (14) for each
sublinear model.
The probabilistic limit state models also follow a two-parameter lognormal distribution [38–41] with
the median value (SC) and dispersion (βC). These parameters are estimated on the basis of experimental

Figure 7. Illustration of linear and bilinear PSDM.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
 SEISMIC FRAGILITY OF LIGHTLY REINFORCED CONCRETE FRAMES WITH INFILLS 1793

results, an expert-opinion survey, simulated responses, and/or a combination thereof. These limit state
models reflect multiple levels of structural functionality associated with observed damage such as
concrete spalling, rebar buckling, and infill collapse. Given the demand model and limit state
models, the fragility function can be expressed in a closed form:
2 3
6 lnðSD =SC Þ 7
P½D≥CjIM  ¼ Φ4qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi5 (16)
βDjIM þ βC þ βM
2 2 2

where βM is the modeling uncertainty and is assumed to be 0.2 per Celik and Ellingwood [40] and Φ[•]
is the cumulative normal distribution function.
Substituting Equation (13) into Equation (16) and rearranging the formulation, Equation (16) can be
expressed as follows:

lnðIMÞ  lnðIM m Þ
P½D ≥ CjIM  ¼ Φ (17)
where βIM


lnðSC Þ  lnðaÞ
IM m ¼ exp (18)
and b

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
β2DjIM þ β2C þ β2M
βIM ¼ (19)
b
where IMm and βIM are defined as the median value and dispersion, respectively, of the fragility curve
for the selected limit states. The fragility function for the bilinear PSDM can be obtained using
Equations (15) and (16) in a similar manner.

4. PROBABILISTIC NUMERICAL MODEL AND GROUND MOTION SUITE

This section explains a case study frame, deterministic and probabilistic frame models thereof, and a
suite of ground motions, selected in the current study.

4.1. Description of case study frame

For the fragility assessment of masonry-infilled RC frames, a typical existing school building in Korea
is selected as a case study, which was designed only for gravity loads. Figure 8 shows the plan,
elevation, cross-section details, and numerical frame model of the building frame. The frames in the
N–S direction comprises three frame types such as fully infilled (Frame A), partially infilled (Frame
B), and bare (Frame C) frames, of which elastic stiffnesses between the base shear force and roof
displacement are 87.9, 66.5, and 6.2 kN/mm, respectively, for a triangle load pattern and mean
properties of the masonry infill material designated as Infill ASCE presented later in Table VI.
Along the E–W direction of the plan, Frames B and C are alternately repeated to form multiple
classroom units, while Frame A is located only at the west and east bounds of the building. Thus,
the group of Frame B contributes more significantly to the overall stiffness than the group of Frame
A, as the number of the classroom unit increases, although Frame A is slightly stiffer than Frame B
because of additional small infill panels. As a result, Frame B is adopted for the case study to
reduce computational efforts. Additionally, the infill panels have no opening, and columns have
widely spaced transverse reinforcement with the column effective depth-to-spacing ratio of 1.1 to
1.44, which may result in brittle column shear failure.

4.2. Probabilistic numerical model

A basis model for the reference frame explained in Section 4.1 is created using the component response
models proposed in Section 2 and including additional modeling issues such as gravity load and mass,

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
1794 J.-S. JEON, J.-H. PARK AND R. DESROCHES

Figure 8. Plan, elevation, reinforcing details, and numerical model of case study frame.

foundation spring, and damping ratio that do not exist in the subassemblage model for the validation.
The basis model is utilized to build numerous uncertainty models accounting for material uncertainties
in masonry infills and boundary frames. These uncertainty models are incorporated to assess the
seismic vulnerability of various frame models with different failure mechanisms.

4.2.1. Basis model. Figure 8 shows the numerical model of the reference frame implemented in
OpenSees [20], employing the three-strut infill models. Considering tributary areas, gravity loads
from slabs and transverse beams are assigned to distributed loads on main beams and concentrated
loads on beam–column connection nodes with a load combination of 1.1(DL + 0.25LL), where DL
and LL are the floor dead and live loads, respectively. All masses are lumped at the beam–column
connections nodes. The geometric and material nonlinearities are taken into account. The columns
and beams are modeled using fiber-type displacement-based beam–column elements to capture the
flexure response of these components, and three zero-length shear springs are used to simulate
column shear failure associated with the strut action of the infill panel (two force-controlled shear
springs) and the inelastic response of column itself sustaining after infill failure (one deformation-
controlled shear spring). The slab thickness is 130 mm, and the corresponding effective width of the
flanged beam section is determined in accordance with ASCE 41-06 [28]: 2,430 and 1,430 mm for
the left and right spans, respectively, at every floor level.
Vertical and rotational springs are considered to model the foundation of the columns. The spring
constants are computed using the formulas in ASCE 41-06 [28] and the soil properties assuming site
class D in ATC-40 [42]; the computed vertical stiffnesses for the left, middle, and right column
footing (shown in Figure 8) are 295, 324, and 238 kN/mm, respectively, while the respective
rotational stiffnesses are 370 × 106, 497 × 106, and 219 × 106 kN-mm/rad. The vertical spring is
modeled using a compression-only spring to capture the uplift of the column foundations that can
reduce seismic force demand on superstructure [19]. Horizontal springs are neglected because of
their large stiffness associated with the integrated slab-on-grade and grade-beam. Five percent
Rayleigh damping is adopted for the damping model. Two frequencies of 0.25ω1 and ω3 are chosen
to define Rayleigh damping, where ω1 and ω3 are the natural frequencies of the first and third
modes, respectively. These frequencies are computed on the basis of the frame stiffness due to the
gravity loads. The frequency 0.25ω1 is included to capture the dramatic reduction of the stiffness
and natural frequencies after infill failure. This reduced first natural frequency for Rayleigh damping
is somewhat conservative compared with average reduction ratio 0.29 for the first natural frequency

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
 SEISMIC FRAGILITY OF LIGHTLY REINFORCED CONCRETE FRAMES WITH INFILLS 1795

near collapse state observed in Figure 13 later. However, Rayleigh damping applied to the numerical
model does not assign damping ratios lower than 0.04 for frequencies between 0.25ω1 and ω3.
Therefore, conservatism due to the overestimation of frequency reduction is insignificant.
To account for the effect of potential failure mechanisms on seismic demands, the current study
builds two frame models for the case study frame in OpenSees [20]: one is the bare frame model
simulating only deformation-controlled flexure–shear failure in all columns; and the other is the
frame model accounting for the force-controlled shear failure of the boundary columns associated
with infill–frame interaction along with the deformation-controlled flexure–shear failure in all columns.
For the frame models with masonry infills, the current study considers RC frames with three
different types of the masonry infill given in Table V: Infill ASCE is the weak and stiff infill based
on the default masonry properties in good condition recommended by ASCE 41-06 [28]; Infill
KOR100 is the strong and soft infill based on the test of Korean masonry by Kim et al. [43]; and
Infill KOR550 is the strong and stiff infill obtained by increasing the ratio of elastic modulus to
prism compressive strength of Infill KOR100 from 100 to 550, of which value is the same as that in
ASCE 41-06. Also, Korean masonry infills have a similar ratio of the elastic modulus to shear
modulus [43], and thus, this ratio is set to be 0.4, as indicated in Table V. Infill KOR550 is added
because Korean masonry tested by Kim et al. [43] has an unusually low elastic modulus although it
is supported by other laboratory tests of masonry prisms [44] and masonry-infilled RC frames [45].
Experimental results by Shin et al. [45] indicated that the frame reached ultimate strength at 2%
drift and degraded to the strength of a reference bare frame at around 4% drift similar to Figure 9.
The failure mode of each infill type is determined to contrast the weak (ASCE) and strong
(KOR100/500) infill types. In the model validation, the SS mode tends to produce the smallest
equivalent strut strength among the three infill failure modes, and thus, it is adopted for the weak
infill type (ASCE). On the other hand, the DK mode is applied to the two strong infill types
(KOR100/550) because the specimens exhibiting column shear failure in the model validation failed
in the DK mode rather than the CC mode, which is observed in specimen S7 with a strong frame
contrary to the case study model. For general seismic fragility assessment, analytical models
proposed in Section 2.3 are applicable to determine the failure mode of infills.
The post-peak stiffness ratios (α) for Infill ASCE, Infill KOR100, and Infill KOR550 are selected as
0.002, 0.05, and 0.01, respectively. The value of 0.002 is smaller than the lowest α used in the model
calibration presented in Section 2.3, considering that the solid masonry unit of the case study frame can

Table V. Response envelope parameters for different masonry infills.

Type Source Em/fm Gm/Em Assumed failure mode α

Infill ASCE ASCE 41-06 [28] 550 0.4 SS 0.002
Infill KOR100 Kim et al. [43] 100 0.4 DK 0.05
Infill KOR550 Kim et al. [43] 550 0.4 DK 0.01

Figure 9. Normalized load–drift envelope of an infill panel for three infill types.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
1796 J.-S. JEON, J.-H. PARK AND R. DESROCHES

dissipate energy in a more stable manner under the assumed component failure mode (SS mode) than
the specimens S4 and S6 with hollow masonry unit subjected to the SS mode in the validation. The
value of 0.01 is close to the upper bound, used in the model calibration presented in Section 2.3,
whereas the value of 0.05 is applied to Infill KOR100 because the value of 0.01 leads to an
excessively large collapse drift. Figure 9 shows the associated force–displacement envelopes for the
mean material properties listed in Table V.

4.2.2. Uncertainty model. For the fragility assessment, uncertainties in material properties (modeling
parameters) such as concrete strength (fc), steel yield strength (fy) for boundary frames, prism
compressive strength (fm), diagonal shear strength (τ d), and bed-joint sliding shear strength (τ o) for
masonry infills are incorporated in the frame model, described in the previous section, using the
Latin Hypercube sampling (LHS) technique [46], as carried out by Celik and Ellingwood [40] and
Jeon et al. [41]. The modeling parameters (mean and coefficient of variation) and associated
probability distributions are presented in Table VI. The current study assumes that the uncertainty of
the elastic and shear modulus (Em and Gm) of the infill depends on the prism compressive strength
(fm) of the infill because these moduli are functions of fm given in Table V. These material
parameters, sampled using the LHS, are used to determine response model parameters for each
material model, namely, each frame model has different properties of material models implemented
in OpenSees [20]. Thus, the technique provides different governing component response
mechanisms associated with material uncertainties that can affect the system collapse.
The two off-diagonal strut connection nodes in the column (Figure 4) are located on the basis of
Equations (1) through (4) and associated with material uncertainty in Table VI. However, for some
sampled material parameters, the upper one of the two nodes is moved to the lower half of the
column and vice versa. This erroneous configuration should be avoided because the column
response model in Figure 4 is intended to have only one shear spring between the off-diagonal strut
and its adjacent central strut. Thus, the location of the off-diagonal strut connection node is assumed
to be constant and determined using the mean value of material parameters.

4.3. Ground motion suite

A suite of 80 recorded ground motion pairs (total of 160) are used for generating the fragility curves of
the masonry-infilled RC frames. Each pair has two orthogonal components in the horizontal directions.
The recorded ground motions in California identified by Medina and Krawinkler [48] are extracted
from the Pacific Earthquake Engineering Research Center’s Strong Motion Database. The ground
motions have an even selection of recorded time histories from four bins that include combinations
of low and high moment magnitudes (Mw) and large and small epicentral distances (R), as presented
in Table VII. The magnitude varies between 5.8 and 6.9, whereas the epicentral distance varies
between 10 and 60 km.

5. FRAGILITY ASSESSMENT OF MASONRY-INFILLED RC FRAMES

This section presents the generation of PSDMs, characterization of limit state models, and development
of fragility curves for various frame models reflecting different failure mechanisms. The PSDMs and

Table VI. Material uncertainties in numerical model.

Random variables Mean (MPa) Coefficient of variation Distribution
Boundary frame Concrete compressive strength (fc) 18 [47] 0.150 [13] Normal
Rebar yield strength (fy) 300 [47] 0.110 [13] Lognormal
Infill ASCE Prism compressive strength (fm) 8.07 [28] 0.200 [28] Lognormal
Bed-joint sliding shear strength (τ o) 0.24 [28] 0.267 [28] Lognormal
Infill KOR100/550 Prism compressive strength (fm) 9.14 [43] 0.214 [43] Lognormal
Cracking stress in diagonal tension (τ d) 0.61 [43] 0.193 [43] Lognormal

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
 SEISMIC FRAGILITY OF LIGHTLY REINFORCED CONCRETE FRAMES WITH INFILLS 1797

Table VII. Classification of the ground motion suite [48].

Bin Combination Number of pairs Magnitude (Mw) Distance (R)
1 Large Mw–short R 20 6.5 < Mw < 7.0 13 km < R < 30 km
2 Large Mw–long R 20 6.5 < Mw < 7.0 30 km ≤ R ≤ 60 km
3 Small Mw–short R 20 5.8 < Mw < 6.5 13 km < R < 30 km
4 Small Mw–long R 20 5.8 < Mw < 6.5 30 km ≤ R ≤ 60 km

fragility curves developed in this section are used to examine the impact of masonry infill properties on
the seismic demand and to estimate the relative vulnerability of simulated frame models.

5.1. Probabilistic seismic demand model

The NTHAs are conducted for each pair of ground motion and probabilistic bare or masonry-infilled
frame model (total of 160 pairs) to monitor key responses incorporated into a PSDM that is a linear
or bilinear regression of demand (D)–intensity measure (IM) pairs in the log-transformed space. This
regression determines the slope, intercept, and dispersion for linear or sublinear models. In the
current study, the maximum story drift (θmax) widely used in the seismic risk assessment of building
frames is chosen for the seismic demand. To determine an optimal IM that meets the research
objective of the current study, two IMs typically used in the risk assessment are considered here:
peak ground acceleration (PGA) and spectral acceleration at the first mode (Sa-T1). Figure 10 shows
the linear and bilinear PSDMs for Infill ASCE model with respect to each IM. The figure indicates
that both IMs have a similar value of dispersion for both the linear and bilinear models. However, a
structure-independent IM is needed for evaluating the relative vulnerability of different frames [38],
which is a primary objective of this research. Thus, the current study selects PGA as the IM,
although Sa-T1 provides slightly better estimate (slightly smaller dispersion), and is a better IM for
the evaluation of individual frames.
Using the linear or bilinear regression model expressed in Equations (13) or (15), the regression
coefficients for various frame models are computed in Table VIII. A primary assumption in the
regression is that the residual follows a normal distribution with zero mean. Thus, a better PSDM is
required to possess the residual following a normal distribution in a logarithmic space. For example,
Figure 11 shows the plot of PGA and residuals between the predicted and actual demands for the
linear and bilinear PSDMs of the Infill ASCE model corresponding to Figure 10(a). Because the
residuals are not normally distributed in Figure 11(a), the best fit for the linear PSDM deviates from
the simulation data at the lower and higher levels of PGA (Figure 10(a)). In particular, the deviation
at the higher PGA levels results in the underestimation of the maximum story drift demand, which
leads to the underestimation of the probability of reaching more severe damage states. In contrast,

Figure 10. PSDMs of Infill ASCE model for PGA and Sa-T1.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
1798 J.-S. JEON, J.-H. PARK AND R. DESROCHES

Table VIII. Linear and bilinear PSDMs for various frame models.
Linear PSDM Bilinear PSDM
Frame a b βD|PGA a1 b1 βD|PGA1 PGAo b2 βD|PGA2

Bare frame 5.55 0.98 0.67 1.14 0.35 0.30 0.14 1.77 0.84
Infill ASCE 3.05 1.67 0.58 0.80 1.13 0.34 0.15 2.46 0.69
Infill KOR550 0.58 1.06 0.40 0.39 0.89 0.32 0.24 2.96 0.60
Infill KOR100 1.65 1.30 0.53 0.54 0.85 0.37 0.18 2.28 0.61

Figure 11. Plot of PGA and residuals of PSDMs for Infill ASCE model.

the residual plot of the bilinear PSDM shown in Figure 11(b) indicates that the residuals of the bilinear
model are normally distributed (almost symmetrically with respect to zero residual) when compared
with those of the linear model. Thus, to capture the scatter-plot in the whole range of PGA, the
current study adopts the bilinear PSDM. The second sublinear model beyond a certain IM (PGAo)
accounts for the dramatic increase in the maximum story drift demand associated with the
significant strength and stiffness loss due to infill or column failure.

5.2. Limit state model

The closed form of fragility function indicated in Equations (16) or (17) necessitates reliable limit state
models. As mentioned in Section 3, limit state models are determined on the basis of expert judgment,
experimental results, analytical models, or a combination thereof. The expert opinion and experiment-
based methods can provide reasonable estimates of the limit state models, only when seismic behavior
of structures is well identified with sufficient data collected from laboratory tests or historical seismic
events. However, response mechanism observed during experiment or post-earthquake reconnaissance
efforts for masonry-infilled RC frames depends on regional masonry infill properties that may have
discrepancy with those adopted in the current study. Thus, for consistency with the adopted infill
properties, the current study employs the analysis-based method for the frame models with and
without masonry infills, which characterizes multiple limit state models expressed in terms of the
maximum story drift, using the NSP and IDA. The NSP is performed for 160 realizations of each
frame model, whereas the IDA is conducted for 40 realizations of each frame model and the same
number of the strongest ground motions identified from the NTHAs for the PSDM to reduce the
computational time. For the NSP, the inverted triangle load pattern along the frame height following
the recommendation of ATC-40 [42] might cause numerical instability in OpenSees as a result of
the complex interaction between two shear springs and six struts assembled around every beam–
column connection located between two adjacent infill panels. Thus, to avert this issue, the pushover
loading is applied to only the second floor level in order to simulate the soft-story mechanism of the

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
 SEISMIC FRAGILITY OF LIGHTLY REINFORCED CONCRETE FRAMES WITH INFILLS 1799

Figure 12. Determination of limit state models.

frame selected. Figure 12(a) compares the pushover curves for a realization of all frame models
sampled from Table VI (fc = 18.89 MPa and fy = 325 MPa for all the frames, fm = 9.26 MPa and
τ o = 0.21 MPa for Infill ASCE, and fm = 9.25 MPa and τ d = 0.54 MPa for Infill KOR100/550).
Figure 12(b) depicts the IDA curves for the 40 realizations of Infill ASCE model sampled from
Table VI.
The NSP is used to identify key points indicating the transition of simulated response of the
components in the first story such as immediate occupancy (IO) and life safety (LS). Here, IO and
LS are defined as the drifts at the yield and maximum lateral force demands at the story level,
respectively. For simplicity, the yield drift is calculated using the equal area rule for bilinearization
in ATC-40 [42]. The current study adopts only two limit state models from the NSP because, as
shown in Figure 12(a), it is difficult to identify the common type of the pushover curves beyond the
peak load that can be applied to all the frame models. On the other hand, the IDA is employed to
define the drift corresponding to the global dynamic instability of the frame models termed collapse
prevention (CP). Vamvatsikos and Cornell [49] suggested that, for bare frames, CP is defined as the
story drift at which the ratio of the local tangent to the elastic slope (γ) reaches 0.2, which was
adopted in FEMA 350 [50]. However, because the IDA curves for masonry-infilled RC frames have
exceptionally large elastic slopes compared with bare frames, the value of 0.2 might underestimate
the collapse drift. To alleviate this problem, the current study defines CP through the comparison of
the NSP and IDA results. Table IX lists parameter values used to define the limit state models on
the basis of the maximum story drift (%). As indicated in Table IX, the incipient collapse point on
the pushover curve denoted by CPo is assumed to be the story drift at 20% loss of maximum story
shear force, whereas that on the IDA curve is computed using three different values of γ. Generally,
larger γ results in smaller SC for CP. Final CP points marked with bold in Table IX are selected
under the principle that the SC for CP should be larger than that for LS and close to that of CPo so
far as possible. For the masonry-infilled frame models, the CP point on the IDA curve is defined
with γ = 0.2 in spite of a large elastic slope, because the drastic loss of story shear results in collapse

Table IX. Medians and dispersions of maximum story drift (%) for limit state models.

Pushover results IDA results

IO LS CPo CP (γ = 0.2) CP (γ = 0.1) CP (γ = 0.05)

Frame SC βC SC βC SC βC SC βC SC βC SC βC
Bare frame 0.44 0.07 0.85 0.07 2.43 0.17 3.04 0.50 — — — —
Infill ASCE 0.16 0.31 0.49 0.26 0.84 0.54 0.91 1.05 1.24 0.82 1.74 0.70
Infill KOR550 0.05 0.22 0.17 0.19 0.56 0.36 0.40 0.74 0.74 0.89 1.17 0.87
Infill KOR100 0.09 0.14 0.28 0.17 0.53 0.17 1.93 1.20 2.36 1.05 3.00 0.95
Note that the bold values refer to the limit state models chosen in the current study.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
1800 J.-S. JEON, J.-H. PARK AND R. DESROCHES

at relatively small drift levels. Consequently, the current study employs three limit state models for
each frame model; IO, LS, and CP are the story drifts at yield shear force, maximum shear force,
and incipient collapse, respectively.
Strictly speaking, there exists a certain extent of inconsistency between the limit state and demand
models because of the type or condition of analysis, the NSP with a prescribed lateral loading
distribution does not reflect the dynamic characteristics of ground motions, and only a portion of the
ground motion suite is used and scaled up in the IDA unlike the demand model (unscaled ground
motions). Despite such inconsistency, the NSP or IDA has been combined with the NTHA-based
cloud method for the demand model, as used in the previous work [40, 51, 52]. The uncertainty
resulting from this inconsistency is implicitly reflected as a source of modeling uncertainties in the
fragility modeling used in the current study. Specifically, this uncertainty is represented by βM of
Equation (16).
Using the IDA results in Figure 12(b), the fundamental periods at the elastic and near collapse
conditions (γ = 0.2) for the Infill ASCE model are plotted in Figure 13 This figure reveals that the
mean value of the fundamental periods at the near collapse state (1.03 s) is more than three times as
high as that at the elastic state (0.30 s). As mentioned in Section 1, this significant change of the
fundamental period supports the appropriateness of the cloud method (using unscaled ground
motions) adopted in the development of fragility curves for this class of frames. Note that different
elastic fundamental periods in Figure 13 are associated with different uncertainty models of the frame.

5.3. Fragility curves

With the bilinear PSDMs and limit state models determined in the previous sections, the closed form of
the fragility function in Equation (17) is estimated and developed for the bare frame and infilled frames
with different masonry properties, as shown Figure 14. The different frame models are compared by
evaluating the relative change in the median value of the fragility curves (PGAm, the PGA
associated with a 50% probability of reaching a certain limit state). As presented in Table X, all of

2
Fundamental period (sec)

Elastic
Collapse
1.5

0.5

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
Ground motion number

Figure 13. Fundamental periods at elastic and near collapse condition for Infill ASCE model.

1 1 1
0.9 0.9 0.9
0.8 0.8 0.8
P[D≥C|PGA]

P[D≥C|PGA]

P[D≥C|PGA]

0.7 0.7 0.7

0.6 0.6 0.6
0.5 0.5 0.5
0.4 0.4 0.4
Bare frame Bare frame Bare frame
0.3 0.3 0.3
Infill ASCE Infill ASCE Infill ASCE
0.2 Infill KOR550 0.2 Infill KOR550 0.2 Infill KOR550
0.1 Infill KOR100 0.1 Infill KOR100 0.1 Infill KOR100
0 0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
PGA (g) PGA (g) PGA (g)
(a) IO (b) LS (c) CP

Figure 14. Comparison of fragility curves for all frame models.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
 SEISMIC FRAGILITY OF LIGHTLY REINFORCED CONCRETE FRAMES WITH INFILLS 1801

Table X. Median values of fragility curves (PGAm) for all frame models.
PGAm PGAm,infill/PGAm,bare
Frame IO LS CP IO LS CP

Bare frame 0.067 0.180 0.369 — — —

Infill ASCE 0.189 0.297 0.382 2.82 1.65 1.04
Infill KOR550 0.101 0.329 0.440 1.51 1.83 1.19
Infill KOR100 0.120 0.253 0.590 1.79 1.41 1.60

the masonry-infilled frame models have higher PGAm at all three limit states when compared with the
bare frame model. For simplicity, the relative vulnerability of the masonry-infilled frame models is
examined with respect to the bare frame model: PGAm of IO is 1.5 to 2.8 times as high as the bare
frame model; PGAm of LS is 1.4 to 1.8 times as high as the bare frame model; and PGAm of CP is
1.0 to 1.6 times as high as the bare frame model. Thus, for a more severe limit state, the infill has
less impact on the PGA associated with a 50% probability of exceeding the limit state. In particular,
Infill ASCE has little effect on PGAm of CP, and the fragility curve for the Infill ASCE model
almost overlaps that of the bare frame in Figure 14(c). However, this observation does not refer to
the similarity of the actual responses between the bare frame model and Infill ASCE model, because
these two models have apparently different PSDMs (shown in Table VIII) and limit states (shown in
Table IX). Rather, the reduction in the seismic demand due to the presence of the infill is canceled
out with the decrease of the drift threshold for the CP limit state. For the Infill KOR550 model, the
exceedance probabilities for PGAs beyond 0.12 g at IO limit state (Figure 14(a)) and for PGAs
beyond 0.55 g at CP limit state (Figure 14(c)) are greater than those for the bare frame model. In
addition, the Infill KOR100 model significantly mitigates the probability of reaching IO limit state
for PGAs below 0.16 g but shows negligible effect for PGAs higher than 0.16 g. Therefore, the
impact of different masonry infill properties on the seismic vulnerability of the lightly reinforced
concrete frames depends on ground motion intensities as well as infill properties.

6. SUMMARY AND CONCLUSIONS

This paper proposes an analytical procedure for fragility assessment of lightly reinforced concrete
frames with masonry-infills reflecting uncertainty in ground motions and building material
properties. Response models for infills and columns are proposed employing existing OpenSees
material models and element formulations, a rational approach to proportion and locate individual
struts in the equivalent three-strut model is suggested, and a nonlinear column shear model
reflecting the infill–frame interaction prior to infill failure as well as soft-story mechanism after infill
failure is proposed. The component models are validated using existing experimental results for
specimens exhibiting infill failure and column shear failure. The simulated responses are well
correlated with observed test data with regard to stiffness, strength, and energy dissipation, and thus,
the models enable reliable seismic risk assessment by reducing modeling uncertainties. However, it
is desirable to improve the proposed model to extend its application to the three-dimensional frames
demanding computational efficiency.
To examine the influence of infill properties on the seismic vulnerability of masonry-infilled RC
frames, a gravity-load-designed school building typical of Korean construction is selected as a case
study. Numerical models of a representative plane frame in the building are developed for three
different types of the masonry infill considering material uncertainty. For the seismic fragility
assessment of these frame models, probabilistic seismic demand models are generated using
nonlinear time history analyses for a suite of ground motions, and limit state models are proposed
combining analytical results from nonlinear static pushover analyses and incremental dynamic
analyses. Inconsistency between the demand and limit state model caused by different numerical
analysis technique is implicitly reflected using a modeling uncertainty parameter in the seismic
fragility function used in the current study. A bilinear model is used to capture the nonlinearity of

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
1802 J.-S. JEON, J.-H. PARK AND R. DESROCHES

scatter-plots for demand-intensity measure pairs instead of commonly used linear models and
combined with respective simulation-based limit state models to derive fragility curves. Overall, the
presence of masonry infills in the case study frames mitigates the seismic demand and further
decreases the probability of exceeding a specific limit state when compared with bare frames.
However, the beneficial effect of the infill diminishes at more severe limit states and almost vanish
especially for the weak and stiff infill (Infill ASCE) subjected to the CP limit state. Moreover, the
strong and stiff infill (Infill KOR550) aggravates the seismic vulnerability against the CP and IO
limit states for the ground motion intensity higher than certain thresholds when compared with the
bare frame. Therefore, the impact of the masonry infill on the seismic vulnerability of lightly
reinforced concrete frames relies on ground motion intensities as well as infill properties.

ACKNOWLEDGEMENTS
This research was supported by a grant (13AUDP-B066083-01) from Architecture and Urban Development
Research Program funded by Ministry of Land, Infrastructure and Transport of Korean government.

REFERENCES
1. Mainstone RJ. On the stiffnesses and strengths of infilled frames. Proceedings of the Institution of Civil Engineering,
Supplement IV, 57–90, 1971.
2. Saneinejad A, Hobbs B. Inelastic design of infilled frames. Journal of Structural Engineering, ASCE 1995;
121(4):634–650.
3. Fardis MN. Experimental and numerical investigations on the seismic response of RC infilled frames and recommen-
dations for code provisions. Report ECOEST-PREC8 No 6. Prenormative research in support of Eurocode 8, 1997.
4. Paulay T, Priestley MJN. Seismic Design of Reinforced Concrete and Masonry Buildings. John Wiley & Sons, Inc.:
New York, 1992.
5. Crisafulli FJ. Seismic behaviour of reinforced concrete structures with masonry infills. Ph.D. thesis, Department of
Civil Engineering, University of Canterbury, Christchurch, New Zealand, 1997.
6. FEMA 306. Evaluation of Earthquake Damaged Concrete and Masonry Wall Buildings: Basic Procedures Manual.
Federal Emergency Management Agency: Redwood City, CA, 1998.
7. Mehrabi AB, Shing B, Schuller MP, Noland JL. Experimental evaluation of masonry-infilled RC frames. Journal of
Structural Engineering, ASCE 1996; 122(3):228–237.
8. Stavridis A, Koutromanos I, Shing PB. Shake-table tests of a three-story reinforced concrete frame with masonry
infill walls. Earthquake Engineering and Structural Dynamics 2012; 41(6):1089–1108.
9. Negro P, Colombo A. Irregularities induced by nonstructural masonry panels in framed buildings. Engineering Struc-
tures 1997; 19(7):576–585.
10. Dolšek M, Fajfar P. Soft storey effects in uniformly infilled reinforced concrete frames. Journal of Earthquake En-
gineering 2001; 5(1):1–12.
11. Verderame GM, De Luca F, Ricci P, Manfredi G. Preliminary analysis of a soft-storey mechanism after the 2009
L’Aquila earthquake. Earthquake Engineering and Structural Dynamics 2011; 40(8):925–944.
12. Chrysostomou CZ, Asteris PG. On the in-plane properties and capacities of infilled frames. Engineering Structures
2012; 41:385–402.
13. Mosalam KM, Ayala G, White RN, Roth C. Seismic fragility of LRC frames with and without masonry infill walls.
Journal of Earthquake Engineering 1997; 1(4):693–720.
14. Dymiotis C, Kappos AJ, Chryssanthopoulos MK. Seismic reliability of masonry-infilled RC frames. Journal of
Structural Engineering, ASCE 2001; 127(3):296–305.
15. Ricci P, De Risi MT, Verderame GM, Manfredi G. Influence of infill distribution and design typology on seismic
performance of low- and mid-rise RC buildings. Bulletin of Earthquake Engineering 2013; 11(5):1585–1616.
16. Celarec D, Dolšek M. Practice-oriented probabilistic seismic performance assessment of infilled frames with consid-
eration of shear failure of columns. Earthquake Engineering and Structural Dynamics 2013; 42(9):1339–1360.
17. Vamvatsikos D, Cornell CA. Incremental dynamic analysis. Earthquake Engineering and Structural Dynamics 2001;
31(3):491–514.
18. Uva G, Raffaele D, Porco F, Fiore A. On the role of equivalent strut models in the seismic assessment of infilled RC
buildings. Engineering Structures 2012; 42:83–94.
19. Burton H, Deierlein G. Simulation of seismic collapse in non-ductile reinforced concrete frame buildings with ma-
sonry infills. Journal of Structural Engineering, ASCE 2013. doi:10.1061/(ASCE)ST.1943-541X.0000921
20. McKenna F, Scott MH, Fenves GL. Nonlinear finite-element analysis software architecture using object composition.
Journal of Computing in Civil Engineering ASCE 2010; 24(1):95–107.
21. El-Dakhakhni W, Elgaaly M, Hamid A. Three-strut model for concrete masonry-infilled steel frames. Journal of
Structural Engineering 2003; 129(2):177–185.
22. Liauw T, Kwan K. Unified plastic analysis for infilled frames. Journal of Structural Engineering, ASCE 1985;
111(7):1427–1448.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe
 SEISMIC FRAGILITY OF LIGHTLY REINFORCED CONCRETE FRAMES WITH INFILLS 1803

23. Chrysostomou CZ, Gergely P, Abel JF. A six-strut model for nonlinear dynamic analysis of steel infilled frames. In-
ternational Journal of Structural Stability and Dynamics 2002; 2(3):335–353.
24. Stafford Smith B. Methods for predicting the lateral stiffness and strength of multi-storey infilled frames. Building
Science 1967; 2(3):247–257.
25. Combescure D. Some contributions of physical and numerical modelling to the assessment of existing masonry
infilled RC frames under extreme loading. Proceedings of First European Conference on Earthquake Engineering
and Seismology, Geneva, Switzerland, 2006.
26. Hetenyi MI. Beams on Elastic Foundations: Theory with Applications in the Fields of Civil and Mechanical Engi-
neering. University of Michigan Press: Ann Arbor, Michigan, 1946.
27. Stavridis A, Shing PB. Finite-element modeling of nonlinear behavior of masonry-infilled RC frames. Journal of
Structural Engineering, ASCE 2010; 136(3):285–296.
28. ASCE/SEI Seismic Rehabilitation Standards Committee. Seismic Rehabilitation of Existing Buildings (ASCE/SEI 41-
06). American Society of Civil Engineers: Reston, VA, 2007.
29. Mander JB, Priestley MJN, Park R. Theoretical stress–strain model for confined concrete. ASCE Journal of Struc-
tural Engineering 1988; 114(8):1804–1826.
30. Celarec D, Ricci P, Dolšek M. The sensitivity of seismic response parameters to the uncertain modelling variables of
masonry-infilled reinforced concrete frames. Engineering Structures 2012; 35:165–177.
31. Sezen H, Moehle JP. Shear strength model for lightly reinforced concrete columns. Journal of Structural Engineer-
ing, ASCE 2004; 130(11):1692–1703.
32. Elwood KJ. Modelling failures in existing reinforced concrete columns. Canadian Journal of Civil Engineering
2004; 31(5):846–859.
33. Stafford Smith B, Riddington JR. The design of masonry-infilled steel frames for bracing structures. The Structural
Engineer 1978; 56(1):1–7.
34. Drysdale RG, Hamid AA. Tensile strength of brick masonry. International Journal of Masonry Construction 1982; 2
(4):172–177.
35. Bertoldi SH, Decanini LD, Gavarini C. Telai tamponati soggetti ad azione sismica, un modello semplificato:
confronto sperimentale e numerico (in Italian). Atti del 6 Convegno Nazionale ANIDIS: Perugia, Italy, 1993.
36. Cornell AC, Jalayer F, Hamburger RO. Probabilistic basis for 2000 SAC Federal Emergency Management Agency
steel moment frame guidelines. Journal of Structural Engineering, ASCE 2002; 128(4):526–532.
37. Ramamoorthy SK, Gardoni P, Bracci JM. Probabilistic demand models and fragility curves for reinforced concrete
frames. Journal of Structural Engineering, ASCE 2006; 132(10):1563–1572.
38. Jeon J-S. Aftershock vulnerability assessment of damaged reinforced concrete buildings in California. Ph.D. Thesis,
School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA, 2013.
39. FEMA. HAZUS-MH MR4 Technical Manual, Earthquake Model. Federal Emergency Management Agency: Wash-
ington, DC, 2003.
40. Celik OC, Ellingwood BR. Seismic fragilities for non-ductile reinforced concrete frames—role of aleatoric and ep-
istemic uncertainties. Structural Safety 2010; 32(1):1–12.
41. Jeon J-S, Lowes LN, DesRoches R, Brilakis I. Fragility curves for non-ductile reinforced concrete frames that exhibit
different component response mechanisms. Engineering Structures 2015; 85:127–143.
42. Applied Technology Council (ATC-40). Seismic evaluation and retrofit of concrete buildings. Report No. SSC96-01.
Seismic Safety Commission, 1996.
43. Kim HC, Kim KJ, Park JH. Experimental study on the material properties of unreinforced masonry considering earth-
quake load (in Korean). Journal of the Earthquake Engineering Society of Korea 2001; 5(2):93–101.
44. Yi W-H, Lee J-H, Kang D-E, Yang W-J. An experimental study of material characteristics of brick masonry (in
Korean). Journal of the Architectural Institute of Korea 2004; 20(12):45–52.
45. Shin JH, Ha GJ, Kim KY, Kwon JB, Jun JH. Improvement and evaluation of seismic-resistant capacity of reinforced
concrete infilled masonry wall with lintel types of opening. Proceeding of the Architectural Institute of Korea 2001;
21(1):163–166.
46. Mckay MD, Conover WJ, Beckman RJ. A comparison of three methods for selecting values of input variables in the
analysis of output from a computer code. Technometrics 1979; 21(2):239–245.
47. KISTEC. Guideline for the Seismic Evaluation and Retrofit of Existing Buildings (in Korean). Korean Infrastructure
Safety & Technology Corporation: Goyang, Korea, 2011.
48. Medina RA, Krawinkler H. Seismic Demands for Nondeteriorating Frame Structures and Their Dependence on
Ground Motions, John A (ed.). Blume Earthquake Engineering Center, Stanford University: CA, 2003.
49. Vamvatsikos D, Cornell CA. Applied incremental dynamic analysis. Earthquake Spectra 2004; 20(2):523–553.
50. FEMA. Recommended Seismic Design Criteria for New Steel Moment-Frame Buildings (FEMA-350). Federal Emer-
gency Management Agency: Washington, DC, 2000.
51. Wen YK, Ellingwood BR. The role of fragility assessment in consequence-based engineering. Earthquake Spectra
2005; 21(3):861–877.
52. Ramamoorthy SK, Gardoni P, Bracci JM. Seismic fragility and confidence bounds for gravity load designed rein-
forced concrete frames of varying height. Journal of Structural Engineering, ASCE 2008; 134(4):639–650.

Copyright © 2015 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2015; 44:1783–1803
DOI: 10.1002/eqe