Engineering Structures 31 (2009) 125–137

Contents lists available at ScienceDirect

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

Seismic fragility analysis of low-rise unreinforced masonry structures

Joonam Park a,∗ , Peeranan Towashiraporn b , James I. Craig c , Barry J. Goodno d
a
Korea Railroad Research Institute, Ui-Wang City, Kyung-gi, 437-757, South Korea
b
AIR Worldwide Corporation, Boston, MA 02116, USA
c
School of Aerospace Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA
d
School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA

article info a b s t r a c t

Article history: Unreinforced masonry (URM) is one of the most common structural types for low-rise buildings in the
Received 4 March 2008 United States. Its dynamic behavior is highly nonlinear, and generally shows high vulnerability to seismic
Received in revised form loading. Despite the need for seismic risk assessment of this class of structures, the fragility curves for
31 July 2008
URM buildings based on analytical models are scarce in the field of earthquake engineering. This study
Accepted 31 July 2008
Available online 7 September 2008
performs seismic fragility analysis of a URM low-rise building. Fragility curves are developed for a two-
story URM building designed to represent a typical essential facility (i.e., a firehouse) in the central and
Keywords:
southern US (CSUS) region. A structural modeling method is proposed such that it can be effectively used
Unreinforced masonry for fragility analysis without significant increase in computational time, and maintains an acceptable
Seismic fragility level of accuracy in representing the nonlinear behavior of the structures. A set of fragility curves are
Out-of-plane wall stiffness developed and include different configurations of the out-of-plane walls and their associated stiffness.
Composite spring model The fragility analysis shows that the seismic performance of URM buildings is well below the desirable
building seismic performance level recommended by current seismic codes, indicating high vulnerability
of URM buildings within the CSUS region. It is also shown that the out-of-plane wall stiffness should not
be ignored in the risk assessment of URM buildings because the overall seismic performance of URM
buildings is rather sensitive to the out-of-plane wall stiffness. The analytical fragility curves developed
are compared with those of HAZUS. The comparison shows that the analytical fragility curves developed
have lower variation in the seismic response than those of HAZUS. Several reasons for the discrepancy are
discussed. The model-based analytical fragility curves developed in this study can increase the accuracy
and effectiveness of seismic risk assessment of essential facilities of the CSUS region. Moreover, the
structural modeling method introduced in this study can be effectively used for development of the
fragility curves of URM buildings.
© 2008 Elsevier Ltd. All rights reserved.

1. Introduction structures have high seismic vulnerability over the CSUS region.
This means that a moderate or major earthquake might result in a
Unreinforced masonry (URM) is one of the most common disastrous consequence associated with the URM structures in the
structural types for low-rise buildings in the United States. In region.
particular, French and Olshansky [1] have showed that nearly For rational estimation and reduction of the potential seismic
one third of the essential facilities (i.e., firehouses, police stations, losses associated with URM structures, the seismic performance
emergency management centers, etc.) distributed within the level should be quantitatively measured through risk assessment
central and southern US (CSUS) region are low-rise (i.e., two stories of such structures. Fragility analysis is an effective tool for risk
or less) URM structures. Due to the infrequent nature of major assessment of structural systems as it can be used for probabilistic
earthquakes in the CSUS region, however, most existing URM estimation of seismic losses and eventually enables decision-
building structures have been designed considering only gravity making activities for seismic risk reduction.
and/or wind loads. Moreover, past earthquake reconnaissance Fragility curves for URM structures are currently available from
reports have suggested that unreinforced masonry structures are HAZUS [2], a seismic loss estimation framework developed by the
highly susceptible to damage from earthquakes. Therefore, URM Federal Emergency Management Agency (FEMA). HAZUS uses a
systematic approach for probabilistic damage assessment of build-
ing structures, in which building response is characterized by
∗ Corresponding author. Tel.: +82 31 460 5293; fax: +82 31 460 5359. building capacity curves and seismic hazard is represented by
E-mail address: [email protected] (J. Park). demand spectra (i.e., capacity spectrum method). The building
0141-0296/$ – see front matter © 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2008.07.021
126 J. Park et al. / Engineering Structures 31 (2009) 125–137

fragility curves are then generated assuming a lognormal distribu-

tion in which distribution parameters consist of the median capac-
ity and damage variability. Although it is desirable to obtain these
distribution parameters from either an analytical or experimental
approach, HAZUS defines these parameters in most cases based
on expert judgment [3]. More accurate fragility curves incorpo-
rating rigorous modeling of nonlinear behavior of URM structures
are scarce in the field of earthquake engineering at the present
time. This study performs seismic fragility analysis of low-rise URM
building structures using a simplified modeling method that can
be effectively used for fragility analysis without significant loss of
computation time, and which maintains an acceptable level of ac-
curacy in representing the nonlinear behavior of the structures.
Low-rise URM structures are usually characterized by their
relatively stiff yet quite brittle structural behavior. The behavior
of URM walls is typically characterized with respect to either in-
plane or out-of-plane loadings (referred to hereafter as ‘‘in-plane’’
or ‘‘out-of-plane’’). URM walls usually exhibit very high in-plane
elastic stiffness and very little out-of-plane stiffness. Out-of-plane
failures are generally quite brittle unless significant compressive
loads are present, but in-plane cracking can lead to relatively
Fig. 1. Typical low-rise URM building in the CSUS region.
ductile behavior of perforated (i.e., by windows, doors, etc.) walls
if significant pier rocking is developed. Many of the URM essential
facilities located in the CSUS region are constructed with wooden 2. Representative URM building structure
floor and roof diaphragms that are rather flexible and are often
poorly tied into the walls. As a result, the seismic response of Typical URM essential facilities in the CSUS region are relatively
such URM structures are often more complex than expected. small and low-rise (one or two-story in height). The structure
FEMA guidelines for seismic analysis and rehabilitation of building is composed of two main components: the URM bearing walls
structures [4] states that the out-of-plane stiffness of walls should and the floor and roof diaphragms. The walls are generally
be neglected while only masses should be considered in structural stiff and constructed with many openings or perforations for
modeling of URM buildings. This implies that the contribution of windows and doors. The diaphragms are usually constructed of
the out-of-plane wall stiffness to the dynamic response of URM timber and, as a result, are much more flexible than the walls. A
structures is minimal and should be ignored in the modeling and benchmark building is selected as representative of existing URM
analysis for conservatism. Some studies, however, show that the low-rise buildings in the CSUS region. This particular building has
dynamic response of an URM building structure could be greatly been designed and constructed for an experimental study [6] to
affected by the out-of-plane wall stiffness [4,5]. If out-of-plane determine the lateral load resistance of a URM building (Fig. 1).
walls are well-connected to the adjacent URM components such The full-scale 2-story test building has been designed to represent
as in-plane walls or floor diaphragms, the out-of-plane walls could the typical construction of an existing URM firehouse in the CSUS
contribute more to the overall response of the URM building. As region. Four unreinforced masonry walls, referred to as Wall A,
a consequence, we can say that no generally accepted approach Wall B, Wall 1, and Wall 2, define the test structure. The detailed
exists at the present time as to whether the out-of-plane wall dimensions of each wall and a plan view are shown in Fig. 2.
stiffness should be considered in modeling URM structures in the The building is 7.3 m (24 ft) by 7.3 m in plan and has story
absence of rigid diaphragms. heights of 3.7 m (12 ft) and 3.0 m (10 ft) for the first and second
This study has two objectives. One is to develop model-based floor, respectively. Wall A and, on the opposite side, Wall B are of
analytical fragility curves of low-rise URM buildings, of which the same configuration to maintain an axis of symmetry for the
fragility curves are only available based on expert opinion at building. Wall 1 only has a door on the first floor and two window
the present time. A low-rise URM building representative of the openings on the second floor. On the other hand, Wall 2 has a
essential facilities in the CSUS region is selected and structurally large door opening in the bottom floor designed to represent a fire
modeled with a simplified nonlinear spring model for effectiveness apparatus entrance.
in repetitive computation. The seismic responses of the URM Material properties are obtained from the experimental stud-
building subject to different levels of earthquake inputs are then ies [6], where masonry prism specimens and 4-brick specimens are
probabilistically estimated considering the randomness in its constructed for the tests. These material properties, however, are
material properties to yield the fragility curves. The model-based for a newly constructed URM building. The majority of the URM
analytical fragility curves developed in this study can increase firehouses in the CSUS region were constructed prior to 1970 [1].
the accuracy and effectiveness of seismic risk assessment of Strength and structural integrity of these buildings may experience
essential facilities of the CSUS region. Moreover, the structural some degree of degradation over such a long period of time. In ad-
modeling method introduced in this study can be effectively used dition, construction materials (i.e., brick and mortar) that are older
for development of the fragility curves of an individual URM than 50 years do not exhibit as high strength as similar materi-
building with a particular configuration. Another objective of this als today. In order to truly represent existing URM structures in
study is to investigate the effect of considering the out-of-plane the region, these factors should be taken into account in the struc-
wall stiffness on the seismic performance of the URM building tural models. One way to incorporate this age effect is to use lower
structures. The fragility curves are developed for three different values for the strength parameters of masonry to reflect the de-
configurations of out-of-plane wall stiffness and compared to one graded masonry condition. In this study, values offered by FEMA
another to investigate the effect of the out-of-plane wall stiffness guidelines [4] corresponding to URM structures in ‘‘fair condition’’
on the seismic performance. This effort demonstrates the needs are used for determination of the material properties. The mate-
of considering the effect of the out-of-plane wall stiffness in rial properties of the degraded URM buildings are listed in Table 1,
modeling and analysis of URM building structures for seismic risk along with those of the newly constructed URM building obtained
assessment. from the material test.
 J. Park et al. / Engineering Structures 31 (2009) 125–137 127

Fig. 2. Detailed dimensions of the test structure [6].

Table 1 other by the interface elements representing the mortar. However,

Basic material properties for the URM test structures for the purpose of seismic fragility analysis of the URM building,
Properties Newly constructed Degraded URM repeated dynamic analyses are required and this approach can
URM (from (from FEMA 356) — quickly become impractical. As a result, a more tractable structural
material test) used for this study
model is used in this study.
Masonry density (kN/m3 ) 18.84 18.84 Under the earthquake excitation, the particular direction of
Masonry compressive 10.00 5.37 seismic loading cannot be anticipated. It is unlikely that the
strength (MPa)
Masonry elastic modulus (MPa) 4134 2955
actual earthquake loading will be in a direction aligned with the
Masonry bed-joint shear 0.41 0.18 building axes. In addition, many buildings have irregular layouts
strength (MPa) that can result in a building structure that behaves in complex
three-dimensional ways under seismic loadings. Therefore, three-
dimensional analysis should produce a more accurate description
3. Modeling URM buildings of the behavior of the structure, and this might include torsional
responses due to asymmetry of the rigidity and mass distributions
Masonry is a nonhomogeneous material made up of two of each component with respect to the loading direction. However,
components: the masonry bricks and the mortar. Masonry in the interest of achieving a simpler model for dynamic analysis,
properties are dependent upon the properties of its constituents. the models developed in this study are based on two-dimensional
Structural behavior of the masonry can be very complex even behavior only and it is assumed that the earthquake excitation
under static loadings. The most refined approach for analyzing a is parallel to Wall A and Wall B of the benchmark URM building
URM wall is the so-called brick-by-brick approach which models (i.e., in the symmetry plane). The 2D model presented in this study
individual masonry brick as a solid element connecting to each is simple enough for repetitive dynamic analyses, yet it is able
128 J. Park et al. / Engineering Structures 31 (2009) 125–137

Fig. 3. Composite spring model of Wall A and Wall B.

for different failure modes, and this can be idealized as shown in

Fig. 4.
The stiffness calculation of a solid wall in FEMA, however, is only
applicable for walls with ideal end conditions (either fixed–fixed
or fixed-free). The masonry piers in a perforated wall are typically
connected to other wall components below and above them. As a
result, it is difficult to characterize the actual boundary conditions
for the masonry piers because they cannot be regarded as truly
Fig. 4. Idealization of hysteresis behavior of each failure mode. fixed–fixed. One approach to account for boundary conditions that
are less stiff than the assumed ideal fixed condition is to instead
to capture most of the important nonlinear behavior of the wall assume a finite rotational restraint at each end (Fig. 5(b)). Another
components and the diaphragms. approach is to simply increase the pier height while maintaining
the ideal fixed end conditions (Fig. 5(c)). For this study, the pier
height is altered to account for the less than ideal end fixity. In
3.1. Modeling of in-plane walls other words, the pier height is increased by a factor, r, to create
a new ‘‘effective height’’. Details about the effective height method
URM walls usually develop high initial elastic stiffness in the in- and the development of the composite spring model can be found
plane direction resulting in a high elastic stiffness and low elastic in Refs. [7,8]. Table 2 shows properties of each segment in a
structural period for URM buildings. While undamaged URM walls composite spring model for Wall A (or Wall B) and their governing
exhibit a very stiff linear elastic behavior for in-plane loading, the behavior calculated from FEMA 356 guidelines.
walls will ultimately fail through a fracture process that can involve
diagonal cracking or bed joint fracture and sliding. Perforated 3.2. Modeling of out-of-plane walls
walls (i.e., walls with openings) can exhibit much more complex
behavior that is associated with localized failures in the masonry An out-of-plane wall is a building wall that is perpendicular
piers, lintels and spandrels surrounding the openings. For these to the direction of the earthquake inputs, and its lateral stiffness
is significantly lower than that of the in-plane wall. Out-of-plane
cases, the behavior can range from almost pure ductile behavior
failures are generally quite brittle unless significant compressive
developed through pier rocking to highly nonlinear hysteretic
loads are present. Failure or even collapse of the out-of-plane walls
behavior developed through, for example, bed joint sliding.
under earthquakes is very likely if the connections between the
In order to more realistically model the nonlinear behavior of out-of-plane walls and the diaphragms and/or the in-plane walls
perforated in-plane walls, a simple composite nonlinear spring are not well constrained. Therefore, the dynamic behavior of out-
model is utilized for this study [7,8]. The basic approach to develop of-plane walls is quite complicated and difficult to characterize
a composite spring model for the in-plane behavior of a URM wall because it could have multiple simultaneous failure modes [5]. In
is to first subdivide the wall into distinct areas or segments, which this study, an out-of-plane wall is modeled with a single nonlinear
behave in a similar fashion as a solid URM wall. Each segment spring with bi-linear hysteresis behavior. Properties needed for
of the URM wall is then represented by a nonlinear spring, and describing the dynamic behavior of an out-of-plane wall are
the springs are assembled in series and parallel arrangements to then its stiffness and strength. As mentioned earlier, one of the
match the segment topology for the wall itself. Fig. 3 shows the objectives of this paper is to investigate the effect of considering
perforated URM wall (Wall A or B) of the benchmark structure the stiffness of the out-of-plane wall on the seismic risk assessment
on the left side and the schematic view of the corresponding of URM building structures. For this, three cases are examined for
composite spring model on the right. In this figure, springs number consideration of the out-of-plane wall stiffness as follows:
1 and number 8 are not representative of any wall component, but – Case 1 – The stiffness of the out-of-plane walls is neglected and
rather capture the story bending effects. In this composite spring assumed to be zero. Only the mass of the out-of-plane wall is
model, each component or segment is treated individually as a considered for the dynamic analysis of the URM building. This
solid masonry shear wall, and the stiffness and the strength are approach is consistent with the guideline provided by FEMA [4],
determined accordingly from the models in FEMA 356 [4]. The and could be considered as a conservative approach in terms of
post-elastic hysteresis behavior of each wall segment is modeled estimating displacements.
 J. Park et al. / Engineering Structures 31 (2009) 125–137 129

Fig. 5. Concept of effective height and effective stiffness of URM pier.

Table 2
Properties of the in-plane walls
Segment Wall A and Wall B
Governing behavior Elastic stiffness (kN/m) Strength (kN)

2 Sliding 1.79 × 106 256.59

3 Rocking 5.55 × 104 32.93
4 Rocking 9.46 × 104 40.01
5 Rocking 1.04 × 105 40.01
6 Rocking 1.22 × 105 57.67
7 Sliding 1.44 × 106 351.59
9 Sliding 2.30 × 106 222.95
10 Rocking 1.22 × 105 20.07
11 Rocking 1.04 × 105 13.97
12 Rocking 1.04 × 105 13.97
13 Rocking 1.22 × 105 20.07
14 Sliding 1.86 × 106 205.72

Table 3
Properties of the out-of-plane walls
Out-of-plane wall properties Case 1 Case 2 Case 3
Wall 1 Wall 2 Wall 1 Wall 2 Wall 1 Wall 2
1st fl. 2nd fl. 1st fl. 2nd fl. 1st fl. 2nd fl. 1st fl. 2nd fl. 1st fl. 2nd fl. 1st fl. 2nd fl.

Stiffness (kN/m) 0 0 0 0 3355 5252 3355 5252 13069 20148 8322 20148
Strength (kN) 0 0 0 0 2.36 2.71 2.36 2.71 9.12 10.41 5.79 10.41

– Case 2 – The out-of-plane walls are modeled in such a way wall stiffness and strength of the degraded URM buildings of the
that they are fixed at the top and bottom only, and the three cases.
connection to the in-plane walls along the side is ignored. Basic
bending theory is applied for calculation of the out-of-plane 3.3. Modeling of flexible diaphragms
wall stiffness. This configuration could be used for representing
the out-of-plane walls with degraded connection condition. Wooden floors and roofs in URM buildings are connected to
– Case 3 – The out-of-plane walls are assumed to have sound the bearing walls with simple connections that cannot transmit
connection to the in-plane walls and diaphragms along the bending. As a result, these components are modeled as flexible
perimeter. Therefore, the out-of-plane wall stiffness should be diaphragms with extensional and shear stiffnesses. Floors made
higher than that considered in Case 2. The stiffness of the from tongue-and-groove or plywood decking over wooden joists
out-of-plane walls of the newly constructed URM building are quite common in old URM construction in the region although
is obtained from the finite element analysis of the full-scale newer construction may involve poured concrete on light steel
building testing [6]. It is assumed that the out-of-plane wall decking. In either case, and especially for the wooden floors, the
stiffnesses of the degraded building can then be determined shear stiffness is much less than the extensional stiffness and
based on the ratio of the masonry elastic modulus of the newly usually is the defining characteristic of the floor. Laboratory tests
constructed building and the degraded building. of typical and modified wooden flooring systems [9] confirm this
behavior and provide representative stiffnesses. It should also be
For estimation of the out-of-plane wall strength, it is assumed noted that these stiffnesses are nonlinear and are often accurately
that a crack is initiated along the top or the bottom of the out-of- characterized by a bilinear stiffness with hysteretic behavior. Floor
plane wall when the maximum tensile stress exceeds the tensile and roof diaphragms for the newly constructed test structure are of
strength of the masonry, which is estimated as 89.6 kPa according identical construction, and their shear and extensional stiffnesses
to FEMA [4]. Table 3 shows the resulting values for the out-of-plane are found to be 700 kN/m and 152.4 × 103 kN/m, respectively,
130 J. Park et al. / Engineering Structures 31 (2009) 125–137

Fig. 6. Nonlinear spring model representing a URM test structure.

Table 4
Maximum displacements at the top of the in-plane wall (Wall A) and out-of-plane wall (Wall 1)
Ground motions Max. disp. at the top of Wall A (cm) Max. disp. at the top of Wall 1 (cm)
Case 1 Case 2 Case 3 Case 1 Case 2 Case 3

m02_01s 0.13 0.10 0.09 13.54 9.86 6.86

m02_02s 0.09 0.10 0.06 9.14 10.03 5.41
m02_03s 0.11 0.10 0.07 11.35 11.25 7.04
m02_04s 0.07 0.07 0.05 6.76 7.72 5.41
m02_05s 0.10 0.09 0.08 9.93 8.84 6.02
m02_06s 0.10 0.08 0.07 10.82 9.50 6.55
m02_07s 0.07 0.07 0.05 10.01 7.62 4.37
m02_08s 0.11 0.08 0.06 9.65 8.20 6.10
m02_09s 0.08 0.07 0.05 10.19 7.85 5.44
m02_10s 0.08 0.09 0.07 9.93 9.17 5.38
Mean 0.09 0.09 0.07 10.13 9.02 5.84

based on the experimental study [5]. The degraded properties are recent knowledge of the engineering modeling of URM building
determined based on the ratio of the masonry elastic modulus of structures, and therefore it is assumed that the peak displacement
the newly constructed building and the degraded building, which response obtained from the analysis using the developed model
is the same as the case of the walls. As a result, the shear and under a cyclic earthquake loading, which is of the main interest
extensional stiffnesses of the diaphragm are determined to be in the fragility assessment, is reasonably accurate with marginal
500 kN/m and 109.0 × 103 kN/m, respectively. errors.
Dynamic behavior of the URM building under earthquake
3.4. Modeling of URM buildings loadings is investigated by numerical computation of dynamic
responses using DRAIN-2DX. Ten uniform hazard synthetic ground
Computer program DRAIN-2DX [10] is selected in this study acceleration records with 2% probability of exceedance in 50 years
for modeling the behavior of the URM test structure. While developed for the Memphis area [11] are applied and the
DRAIN-2DX does not have a straightforward means to model URM displacement responses of the URM structure with three cases
structures, it does include a versatile, zero-length spring element of the out-of-plane walls configurations are recorded. Table 4
with a variety of possible nonlinear behaviors. As described in summarizes the maximum displacements of both the in-plane and
the previous sections, this simple nonlinear spring element can be out-of-plane walls computed during the application of the ground
used to develop basic nonlinear models for simple URM structures excitations. The table shows that the maximum displacements due
with flexible floor diaphragms. Each component in the model is to earthquake loadings are much higher in the out-of-plane wall
constructed using the DRAIN-2DX TYPE 04 zero-length nonlinear
than they are in the in-plane wall. This finding confirms that the
spring element. A simple 2D model can be constructed by orienting
out-of-plane walls are usually the most vulnerable component
the 2D axis system in a horizontal plane. The complete URM
in a URM building, and is consistent with the past research on
test structure is modeled by assembling the composite nonlinear
the nonlinear behavior of the URM out-of-plane walls, which also
springs for each wall with the lumped wall masses, as illustrated in
shows very low out-of-plane strengths [12,13]. The table also
Fig. 6. For illustration purposes, the in-plane walls are represented
shows that the wall responses (both in-plane and out-of-plane)
in Fig. 6 by a single spring in each floor. However, the actual
model incorporates the nonlinear behavior of in-plane walls at a decrease as the out-of-plane wall stiffness increases.
component-level as displayed in Fig. 3. In the following sections, a fragility analysis is performed for
It should be noted that the accuracy of the dynamic behavior evaluation of the seismic performance of the URM building under
of the model cannot be fully verified in this paper because of the consideration. The effect of the out-of-plane wall stiffness on the
lack of dynamic test data. Instead, the model incorporates the most seismic risk assessment of URM buildings is also investigated.
 J. Park et al. / Engineering Structures 31 (2009) 125–137 131

Fig. 7. Concept of fragility curve.

4. Fragility analysis of the URM building levels of damage to structural components. The inter-story drift
is computed as the relative lateral displacement between floors
As shown in Fig. 7, fragility is defined as the probability expressed as a percent of the story height at that floor (also called
of the case study structure reaching or exceeding a specified a drift angle). Building performance or damage levels (generally
limit state under a given earthquake intensity level [14]. In referred to as ‘‘limit states’’) are specified as a function of the
this figure, the shaded area represents a failure domain defined maximum drift the building sustains during an earthquake. In the
from specific criteria, where a damage measure, D, exceeds case of a URM building, three performance levels are defined in
a specific threshold, dsi . Under a given earthquake level San , FEMA 356. They include, Collapse Prevention (CP), Life Safety (LS),
P [D ≥ dsi |San ] is the probability of the structural response and Immediate Occupancy (IO) performance levels. On the other
exceeding the i-th limit state expressed as a threshold, dsi , and hand, HAZUS defines four limit states — Slight Damage, Moderate
generally increases as the earthquake intensity level increases. This Damage, Extensive Damage, and Complete Damage. Although the
probability could be calculated if the probability distribution of limit states are first described in a qualitative manner, they must
the structural damage under a given earthquake level is obtained be redefined in terms of a quantitative measure in order to perform
by accounting for stochastic variations of material properties and the fragility analysis. HAZUS uses the maximum drift ratio as the
the earthquakes mentioned above. Therefore, estimation of the damage measure following FEMA 356, and the threshold values
probability distribution of the structural damage for levels of of the maximum drift ratio corresponding to the limit states are
earthquake yields fragility curves. The fragility curves of the URM defined based on comprehensive review of past studies on building
building are constructed following this concept in the following seismic damage. Descriptions of and the maximum drift limits
sub-sections. for the building performance levels defined in FEMA 356 and
HAZUS are displayed in Tables 5 and 6, respectively. In this study,
4.1. Damage measure and performance levels the limit states defined in HAZUS are adopted and the fragility
curves are developed accordingly, because the fragility curves to be
Defining a measure for quantifying the buildings seismic developed in this study are to be compared with those of HAZUS.
damage is the first important step of a fragility analysis. A damage
measure should be determined such that it can represent the 4.2. Earthquake inputs
system level damage of a building caused by an earthquake.
A number of researchers have proposed damage measures In order to be relevant, the earthquake loadings that are applied
for buildings subject to earthquake loadings. Some utilize a to the test structure must resemble seismic characteristics in the
displacement-based measure such as a maximum roof drift ratio to same region. An ideal approach is to use actual ground motions
quantify damage [15]. Some use energy-based criteria that relate recorded from historical earthquake events. However, due to the
the amount of hysteretic energy to the levels of damage [16]. Some fact that the last major earthquake in this region was almost
researchers combine the two parts and derive hybrid measures [15, 200 years ago, the available recorded information for strong
17]. However, there has been little consistency on the most ground motions is scarce.
appropriate measure to quantify the seismic damage. In light of Wen and Wu [11] developed a suite of synthetic ground
these available damage measures, FEMA 356 proposes the use motions for three cities in Mid-America. These cities include
of the maximum drift ratio to assess building performance and Carbondale, IL, Memphis, TN, and St. Louis, MO. Earthquakes are
132 J. Park et al. / Engineering Structures 31 (2009) 125–137

Table 5
Structural performance levels for URM structures [4]
Damage levels Structural performance levels
Collapse prevention S-5 Life safety S-3 Immediate occupancy S-1
Severe Moderate Light

Overall damage descriptions Extensive cracking. Face course and Extensive cracking. Noticeable in-plane Minor cracking of veneers. Minor spalling
veneer may peel off. Noticeable in-plane offsets of masonry and minor out-of-plane in veneers at a few corner openings.
and out-of-plane offsets. offsets No observable out-of-plane offsets.
Drift ratio 1.0% 0.6% 0.3%

Table 6
Structural performance levels for URM structures [2]
Structural performance levels
Complete damage Extensive damage Moderate damage Slight damage

Overall damage Structure has collapsed or is in In buildings with relatively large Most wall surfaces exhibit diagonal Diagonal, stair-step
descriptions imminent danger of collapse due to area of wall openings most walls cracks; some of the walls exhibit hairline cracks on
in-plane or out-of-plane failure of have suffered extensive larger diagonal cracks; masonry masonry wall surfaces;
the walls. Approximately 15% of the cracking. Some parapets and walls may have visible separation larger cracks around door
total area of URM buildings with gable end walls have fallen. from diaphragms; significant and window openings in
Complete damage is expected to be Beams or trusses may have cracking of parapets; some masonry walls with large
collapsed. moved relative to their supports. may fall from walls or parapets. proportion of openings;
movements of lintels;
cracks at the base of
parapets.
Drift ratio (Low code) 3.5% 1.5% 0.6% 0.3%
Drift ratio (Pre code) 2.8% 1.2% 0.5% 0.2%

Table 7
Structural uncertainties for URM structures
Random parameters Distribution Mean COV

Masonry density (kN/m3 ) Lognormal 18.84 0.05

Masonry compressive strength (MPa) Lognormal 5.37 0.25
Masonry bed-joint shear strength (MPa) Lognormal 0.18 0.20
Damping (%) Uniform 5.0 0.115

are the major sources of structural uncertainty for the URM

structures in this study, and these structural parameters are
represented as random variables. The variations in the stiffness,
strength, and the masses of the URM building components can
be represented by incorporating these random variables into
the structural modeling. Statistical descriptions of these random
variables are presented in Table 7. The mean values of the random
parameters are obtained from the experimental study [5] and
Fig. 8. Acceleration spectra of the input ground motions. FEMA 356 while the probability distributions and dispersion
measures of the parameters are collected from several previous
generated based on two hazard levels: 10% and 2% probabilities of studies [18–20]. Uncertainties inherent in the earthquake loading
exceedance in 50 years. Earthquake events are simulated with their are also considered by using a suite of input ground motions. That
epicenters uniformly distributed around the New Madrid Seismic is, different earthquakes with the same intensity level might cause
Zone (NMSZ). The body wave magnitudes of these simulated different responses of the structure. The spectral acceleration at
earthquakes range between 5 and 8. Local site effects and soil the fundamental period of the building is considered as earthquake
amplification representative of the location of each city are also intensity measure in this study, and all ground motions in the suite
taken into account in the simulation process. At the end, a suite are scaled such that all have the same spectral acceleration value
of 20 ground acceleration records are generated for each city for each level of earthquake demand.
location (10 records for 10% probability level and 10 records for 2%
probability level). The 10 ground motion records for Memphis with 4.4. Construction of fragility curves
2% probability of exceedance in 50 years are used for this study.
The elastic acceleration spectrum of the input ground motions are The most direct means of statistical analysis to obtain proba-
shown in Fig. 8, and the time-history plots can be found in Ref. [11]. bilistic description of the response is running a series of nonlinear
time-history analysis through Monte Carlo simulation [21]. How-
4.3. Uncertain variables ever, it requires a relatively large number of simulations in order
to obtain a sufficiently reliable estimate for probability of dam-
Uncertainties associated with the URM building structure are age. Mann et al. [22] suggested that the number of simulations
represented by considering the variations of certain material might need to be of the order of 10,000 to 20,000 for approximately
properties. Past investigations suggest a number of the URM 95% confidence limit, depending on the function being evaluated.
material properties that exhibit uncertain properties [18–20]. It will be computationally impractical, if not impossible, to simu-
Referring to the literature, it is assumed that masonry density, late thousands or even hundreds of nonlinear time-history analy-
compressive strength, bed-joint shear strength, and damping ratio ses. Therefore, Latin Hypercube sampling (LHS) technique [23,24]
 J. Park et al. / Engineering Structures 31 (2009) 125–137 133

(a) Case 1. (b) Case 2.

(c) Case 3.

Fig. 9. Distributions of drift ratio for selected input ground motion intensities.

is utilized in this study for efficiency of the simulation process. The walls is considered without consideration of the stiffness and the
region of the probability distribution of each random variable is relative displacements from the in-plane walls are only restrained
divided into ten intervals such that all the intervals have equal by the shear and extensional stiffness of the diaphragm. The out-
probability, and a value is randomly generated from each inter- of-plane walls, however, should sustain the seismic loadings to
val and stratified. The ten stratified values are then randomly com- some degree. In addition, the stiffness could be even higher with
bined without replacement along with ten ground motions to yield the contribution of its connectivity to the diaphragm and the in-
ten combinations of random variables. Taking each combination of plane walls, as mentioned in the previous section. Therefore, the
random variables as an input, a dynamic time history analysis is Case 1 modeling approach is considered to be infeasible for use
performed and the maximum drift ratio is calculated for each com- in generation of the fragility curves in this study, especially as the
bination. The values of the maximum drift ratio calculated for all damage is measured in terms of the drift ratio.
ten combinations of the random variables are then used for proba- According to Cornell et al. [25], a log-normal distribution for
bilistic description of the drift ratio for a given level of ground mo- the statistical description of the building response would be a
tion. This step of calculating the drift ratio is performed over the reasonable assumption. Parameters necessary to describe the log-
range of the input earthquake intensity from 0.1g to 3.0g with an normal distribution, i.e., the log-normal mean (µln D ) and the
standard deviation (σln D ), can be estimated from the log-normal
increment of 0.1g to obtain necessary data for generation of the
probability plot of the data points. Fig. 10 shows the data points of
fragility curves. The fragility curves are generated for all three cases
the maximum drift ratio for Case 2 and Case 3 plotted on the log-
of the out-of-plane configuration described in the previous section.
normal probability paper along with the fitted lines for selected
Fragility curves are developed for the limit states defined in
ground motion intensities. The log-normal mean and standard
HAZUS, as discussed earlier. Note that the limit states defined for
deviation can then be estimated from the y-intercept and the
‘‘pre code’’ URM low-rise structure in HAZUS is utilized, as it is
slope of the fitted line, respectively. The median value of the
assumed that the structure is designed without consideration of drift ratio for a given earthquake intensity is also calculated as
the seismic loadings. The distribution of the responses, i.e., the ^ µln D
m D = e . The fitted lines in Case 2 and Case 3 show that
maximum inter-story drift ratios, resulted from the analyses Case 2 apparently has larger median drift ratio response than that
with ten combinations of input random variables derived with of Case 3 for each level of earthquake intensity. The variance of
LHS mentioned above are obtained over the range of the input Case 3 is higher than that of Case 2, as the slopes of the fitted
earthquake intensity. Fig. 9 shows the response (i.e., drift ratio) lines of Case 3 are steeper that those of Case 2. The probability of
distribution of all three cases for selected earthquake intensity exceedance of the structures corresponding to different limit states
levels. The dotted horizontal lines indicate the thresholds for the (i.e., slight damage = 0.2%, moderate damage = 0.5%, extensive
HAZUS limit states, which are corresponding to the complete, damage = 1.2%, complete damage = 2.8%) can then be calculated
extensive, and moderate damage level from the top. In Case 1, most using the obtained log-normal parameters. For example, the log-
data points are above the complete damage level line even for low normal mean and standard deviation of the drift ratio response of
input earthquake levels. This shows that the structure is highly Case 3 corresponding to the earthquake intensity of Sa = 0.6g
vulnerable to even low earthquake intensity levels, and implies are −4.5114 and 0.41, respectively. The probability of exceedance
that the structure has almost no tolerance to the seismic loading. corresponding to complete damage is then calculated as,
This is not realistic, as many low-rise URM buildings sustained low 
ln(0.028) + 4.5114

to moderate level of earthquakes in the past. The reason why the P (D > 0.028|Sa = 0.6g ) = 1 − Φ
model yields extensive displacements is that the Case 1 structure 0.41
is modeled in such a way that only the mass of the out-of-plane = 0.0112 (1)
134 J. Park et al. / Engineering Structures 31 (2009) 125–137

(a) Case 2. (b) Case 3.

Fig. 10. Log-normal probability plot of max. Drift ratio for selected earthquake intensities.

(a) Case 2. (b) Case 3.

Fig. 11. Log-normal fitting of fragility.

Table 8 limit states. Assuming log-normal CDFs for the fragility curves, the
Log-normal parameters for description of fragility curves fragility corresponding to a given damage state can be expressed
Case 2 Case 3 as a log-normal cumulative distribution function:
µln Sa σln Sa µln Sa σln Sa   
ln Sa /Ŝa
Slight damage −2.4528 0.2074 −1.8887 0.2929
FR (Sa ) = Φ   (2)
Moderate damage −1.7071 0.2261 −1.2192 0.3496 σln Sa
Extensive damage −0.9269 0.2243 −0.5047 0.3319
Complete damage −0.1040 0.1869 0.2818 0.3269
where, Ŝa and σln Sa indicate the median capacity and the variability
of the system, respectively, both being expressed in terms of
where, Φ (•) denotes standard normal cumulative distribution spectral acceleration. The log-normal parameters can be estimated
function (CDF). The values of probability of exceedance corre- by plotting the calculated values of the probability of exceedance
sponding to extensive, moderate, and slight damage are also cal- on log-normal probability sheets as shown in Fig. 11 and Table 8.
culated in a same manner and they are estimated 0.4145, 0.9725, The fitted fragility curves for Case 2 and Case 3 are shown in
and 0.9999, respectively. Fig. 12. The standard deviation for each level of damage states
The fragility curves of the structures corresponding to different in Case 3 is also larger than in Case 2. This is a result of the
limit states can then be generated by plotting the input earthquake fact that the standard deviation of the drift response for a given
level represented in terms of Sa and the probability of exceeding earthquake intensity (i.e., σln D|Sa ) is larger in Case 3 than in
 J. Park et al. / Engineering Structures 31 (2009) 125–137 135

(a) Case 2. (b) Case 3.

Fig. 12. Fitted fragility curves.

Table 9
Probability of exceeding limit states for different earthquake levels
Natural period (s) Earthquake level (prob. of Spectral Probability of exceeding limit states
exceed.) acceleration (g)
Slight damage Moderate damage Extensive damage Complete damage

Case 2 0.450 10% in 50 years 0.330 1.000 0.996 0.209 0.000

2% in 50 years 1.057 1.000 1.000 0.999 0.803

Case 3 0.276 10% in 50 years 0.390 0.999 0.786 0.094 0.000

2% in 50 years 1.057 1.000 0.999 0.954 0.244

Case 2 as previously mentioned. The median capacity for each level observed that overall the Case 3 structure shows better seismic
of damage states in Case 3 is larger than in Case 2, which is as performance (i.e., less probability of exceeding a damage state)
expected because of the higher stiffness and strength of the out- over the Case 2 structure. For the earthquake level of 10/50 PE (10%
of-plane walls. It should be noted, however, that this cannot be probability of exceedance in 50 years), the probability of exceeding
used as a direct proof that the seismic performance of Case 3 is a damage state of the Case 2 structure is almost unity up to the
better than that of Case 2. This is because the earthquake intensity moderate damage state, but suddenly decreases for the extensive
measure used for generation of the fragility curves is a spectral and complete damage states. On the contrary, the probability of
acceleration, which is determined depending on the natural period exceeding a damage state of the Case 3 structure decreases more
of a particular structure. That is, single earthquake could result in gradually from the slight to the complete damage states. For the
different spectral accelerations for different structures. Therefore, earthquake level of 2/50 PE, both the Case 2 and Case 3 structures
for comparison of the seismic performance of more than two show high vulnerability. However, the Case 3 structure shows
structures, it is important to consider a spectral acceleration better performance in that it has the probability of exceeding the
corresponding to each structure and its probability of exceedance. complete damage state of 24.4%, whereas the Case 2 structure
The US Geological Survey (USGS) provides an interactive website has 80.3%.
(http://eqhazmaps.usgs.gov/) where the spectral accelerations of The difference of the seismic performance of the two cases
several probabilistic earthquake hazard levels (10%, 5%, and 2% is quite pronounced, which shows that the seismic performance
estimation of low-rise URM buildings is sensitive to the degree
probability of exceedance in 50 years) corresponding to several
of consideration of the out-of-plane wall stiffness and strength.
different period values (0.2, 0.3, and 1.0 s) can be obtained for any
This means that although the out-of-plane wall stiffness is much
particular location within the US by entering the zip code. Based
lower than the in-plane wall stiffness as previously mentioned, the
on the resulting spectral accelerations, the site class information
effect of the out-of-plane wall to the overall seismic performance
for the location is used to construct a response spectrum of
of an URM building structure should not be neglected. Therefore,
an earthquake with an arbitrary probability of exceedance.
the out-of-wall stiffness and strength should be appropriately
Using these response spectra, the spectral acceleration values
modeled considering the condition of its connection to the in-plane
corresponding to two different levels of probability of exceedance walls and diaphragms for effective seismic risk assessment of URM
– 2% and 10% in 50 years – for Case 2 and Case 3 are obtained. buildings.
The probability of exceeding various limit states for both cases The seismic performance of low-rise URM buildings observed
can then be read from the fragility curves. Table 9 shows the from Table 9 and Fig. 12 can be compared with the performance
probability of exceeding limit states for different earthquake levels. level recommended by the current seismic code to measure its rel-
The fundamental period of the structure is estimated as 0.450 s ative vulnerability. The seismic performance level corresponding
for Case 2 and 0.276 s for Case 3. These periods are consistent to the Basic Safety Objective (BSO) of building structures suggested
with values presented in results of prior experimental research [6] by FEMA 356 [4] is to ensure the ‘‘Life Safety’’ level and the ‘‘Col-
and also with values listed in HAZUS [2]. However, it should be lapse Prevention’’ level for the earthquake level of 10/50 PE and
noted that these periods are relatively higher than the fundamental 2/50 PE, respectively. From the damage descriptions and drift ra-
period calculated for the one-story URM building studied by tios presented in Tables 5 and 6, it can be stated that the ‘‘Life
Kim et al. [5]. The periods presented in this study are higher Safety’’ level and the ‘‘Collapse Prevention’’ level defined in FEMA
because the structure under consideration has two stories, and 356 are consistent with the ‘‘Moderate Damage’’ and the ‘‘Exten-
also because the current structure has flexible diaphragms. It sive Damage’’ of HAZUS, respectively. For both cases (i.e., Case 2
should also be noted that the out-of-plane walls are taken into and Case 3), the URM building analyzed in this study has more
consideration in the modeling of the structure in this study. It is than 70% probability of exceeding the moderate damage state for
136 J. Park et al. / Engineering Structures 31 (2009) 125–137

(a) Case 2 vs. HAZUS. (b) Case 3 vs. HAZUS.

Fig. 13. Comparison of fragility curves.

2. Construct the demand spectra (response spectra) correspond-

ing to the selected earthquake levels and the capacity spectrum
of the target structure (URM low-rise).
3. Obtain the spectrum displacements that are corresponding to
the selected earthquake levels from the intersections of the
demand spectra and the capacity spectrum.
4. From the HAZUS fragility curves for pre-code URM low-rise
building, read the values of the probability of exceedance
corresponding to the selected earthquake levels for various
limit states.
5. Construct the plot of the probability of exceedance versus the
spectral acceleration read corresponding to the natural period
of the URM structure from the response spectrum.
6. Obtain log-normal parameters (µln Sa and σln Sa ) from the curve-
Fig. 14. Comparison of median Sa of fragility curves for different limit states.
fitting for construction of the fragility curves.
The converted HAZUS fragility curves expressed in terms of
the earthquake level of 10/50 PE, and more than 90% probability spectral acceleration are compared with the fragility curves
of exceeding the extensive damage state for the earthquake level developed for the Case 2 and Case 3 structures in this study. Fig. 13
of 2/50 PE. Moreover, the probability of exceeding the complete shows the comparison of the two sets of fragility curves. Overall,
damage state for the 2/50 earthquake is more than 20% for Case the fragility curves developed in this study (Case 2 and Case 3)
3 and is more than 80% for Case 2. Overall, these results indicate are steeper than those of HAZUS. This implies that the variability
that the seismic performance of URM buildings is well below the of the seismic response of the Case 2 and Case 3 structures is
desirable building seismic performance level recommended by the smaller than that of HAZUS. In the comparison of Case 2 and
seismic code, indicating high vulnerability of URM buildings within HAZUS, the Case 2 structure is more vulnerable for lower limit
the CSUS region. states (i.e., slight, moderate and most of extensive damage states)
as the median capacity values of Case 2 are apparently smaller
5. Fragility comparison than those of HAZUS. For complete damage state, median capacity
values are almost the same but the variability of Case 2 is smaller
As previously mentioned, HAZUS fragility curves are the only
than that of HAZUS. In the comparison of Case 3 and HAZUS,
available ones for low-rise URM buildings at this time. The fragility
the Case 3 structure is more vulnerable for slight and moderate
curves of the URM structures developed in this study are compared
limit states. However, the median capacity of the Case 3 structure
with the HAZUS fragility curves. Note that the comparison is made
corresponding to complete damage state is greater than that of
for the HAZUS fragility curves developed for pre code low-rise URM
HAZUS. This tendency is better viewed with the plot of the median
building, as it is assumed that the building is designed without any
capacity values of the fragility curves for different limit states
consideration of seismic loading. For the comparison, the fragility
versus the corresponding drift ratios, which is presented in Fig. 14.
curves of HAZUS are converted such that their fragility parameters
Several explanations for the observed differences are as follows.
are consistent with those developed in this study. As previously
mentioned, the fragility curves in HAZUS are constructed by – The fragility curves presented in HAZUS are generated using
the nonlinear capacity spectrum method [2], where the capacity lognormal distribution parameters that are defined based on
is expressed in terms of spectral displacements. Accordingly, expert judgment, whereas the fragility curves developed in this
the HAZUS fragility curves are represented in terms of spectral study are based on analytical nonlinear modeling and time
displacement whereas the fragility curves developed in this study history analysis.
are described by spectral acceleration. Therefore, the x-axis of – The fragility curves developed above are of a particular
the HAZUS fragility curves should be converted from spectral structure and do not represent the URM building stock in
displacement to spectral acceleration. The conversion process is a region (i.e., a class of structure), although the structure
briefed as follows: represents a typical low-rise URM building in the CSUS region.
1. Select several input earthquake levels expressed in terms of On the contrary, the HAZUS fragility curves are developed
probability of exceedance (e.g., 2%, 5%, and 10% of probability aiming for seismic risk assessment of regional building stock.
of exceedance in 50 years). The level of uncertainty in assessing the damage distribution
 J. Park et al. / Engineering Structures 31 (2009) 125–137 137

of a building stock over a particular region would be higher of the authors and do not necessarily reflect those of the National
compared to an individual structure. This effect results in Science Foundation.
the relatively low variation of the analytical fragility curves
developed in this study. In order to develop analytical fragility References
curves for a regional building stock, the geometrical variation
should be considered in addition to the material uncertainty. [1] French S, Olshansky R. Inventory of essential facilities in Mid-America. Mid-
America earthquake center project SE-1 final report. 2000.
– Consideration of the out-of-plane wall stiffness in the analytical [2] NIBS — National Institute of Building Science. HAZUS 99 earthquake loss
fragility curves increases the seismic capacity. This tendency is estimation methodology, technical manual. Washington, DC. 1999.
more pronounced for higher limit states. [3] Wen YK, Ellingwood BE, Veneziano D, Bracci J. Uncertainty modeling in
earthquake engineering. Mid-America earthquake center project FD-2 report.
2003.
6. Conclusions [4] ASCE, FEMA 356. Prestandard and commentary for the seismic rehabilitation
of buildings. Publication No. 356, Washington (DC): Federal Emergency
Management Agency. 2000.
Model-based analytical fragility curves for a 2-story unrein- [5] Kim SC, White DW. Nonlinear analysis of a one-story low-rise masonry
forced masonry building typical of the CSUS region are devel- building with a flexible diaphragm subjected to seismic excitation. Eng Struct
oped in this study. These fragility curves are distinguished from 2004;26:2053–67.
[6] Yi T. Experimental investigation and numerical simulation of an unreinforced
the existing fragility curves in HAZUS, in that the HAZUS fragility masonry structure with flexible diaphragms. Ph.D. thesis. Georgia Institute of
curves are derived using selected parameters which are based on Technology. 2004.
the subjective judgments of experts. A simplified spring model is [7] Craig JI, Goodno BJ, Towashiraporn P, Park J. Response modification
applications for essential facilities. Mid-America earthquake center project ST-
developed to describe the highly nonlinear dynamic behavior of 4 final report. 2002.
URM structures in an effective manner, and a typical URM low-rise [8] Park J, Craig JI, Goodno BJ. Simple nonlinear in-plane response models for
building is modeled for the fragility analysis. The fragility curves assessing fragility of URM walls. In: Proceedings of the 7th US national
conference on earthquake engineering. Boston (MA): EERI. 2002.
are then generated for the structure considering the uncertainties [9] Peralta DF, Bracci JM, Hueste MBD. Seismic performance of rehabilitated
in both the seismic demand and the material properties. floor and roof diaphragms. Mid-America earthquake center project ST-8 final
The fragility analysis shows that the seismic performance of report. 2000.
[10] Prakash V, Powell GH, Campbell S. DRAIN-2DX base program description
URM buildings is well below the desirable building seismic per-
and user guide, Ver. 1.10, Berkeley (CA): Department of Civil Engineering,
formance level recommended by the seismic code, demonstrating University of California; 1993.
the high vulnerability of the URM buildings within the CSUS region. [11] Wen YK, Wu CL. Uniform hazard ground motions for Mid-America cities.
It is also shown that the out-of-plane wall stiffness should not be Earthq Spectra 2001;17(2):359–84.
[12] Simsir CC, Ashheim MA, Abrams DP. Response of Unreinforced masonry
ignored in the risk assessment of URM buildings because without bearing walls situated normal to the direction of seismic input motions. In:
consideration of the out-of-plane wall stiffness, the vulnerability of Proceedings of the 7th US national conference on earthquake engineering,
the URM building is estimated to be unrealistically high. The over- Boston (MA): EERI. 2002.
[13] Goodno BJ, Craig JI, Losiriluk T. Performance objectives for essential facilities.
all seismic performance of URM buildings is rather sensitive to the Mid-America earthquake center project ST-9 final report. 2003.
out-of-plane wall stiffness. This fact emphasizes the importance of [14] Ellingwood BR. Earthquake risk assessment of building structures. Reliab Eng
modeling the out-of-plane walls of URM buildings for more accu- Syst Saf 2001;74(3):251–62.
[15] Rodriguez ME, Aristizabal JC. Evaluation of a seismic damage parameter.
rate seismic risk analysis. Earthq Eng Struct Dyn 1999;28:463–77.
The comparison of the analytical fragility curves with the [16] Wong KKF, Wang Y. Energy-based damage assessment on structured during
HAZUS curves shows that the analytical fragility curves developed earthquakes. The Struct Des Tall Build 2001;10:135–54.
[17] Park YJ, Ang AHS. Mechanistic seismic damage models for reinforced concrete.
in this study have lower variability in the seismic response than
J Struct Eng 1985;111(4):722–39.
those of HAZUS. The first reason for the difference is that the [18] Abrams DP, Shinozuka M. Loss assessment of Memphis buildings. National
development approach of the two sets of fragility curves are center for earthquake engineering research project NCEER-97-0018 technical
different (i.e., model-based vs. expert judgment). Nonetheless, the report. 1997.
[19] Schueremans L, Gemert DV. Evaluating the reliability of structural masonry
HAZUS fragility curves are aimed for risk assessment of a building elements using the response surface technique. In: Proceeding of the 8th
stock in a region, whereas the analytical fragility curves presented international conference on durability of building materials and components.
here are developed for an individual building. For development 1999.
[20] JCSS — Joint Committee on Structural Safety. JCSS probabilistic model
of analytical fragility curves for a building stock in a region, the code, Internet Publication, http://www.jcss.ethz.ch/JCSSPublications/PMC/
geometrical variation as well as the material uncertainty should PMC.html; 2001.
be considered in the regional analysis. [21] Kleijnen JPC. Statistical techniques in simulation: Part 1. New York (NY):
Marcel Dekker Inc.; 1974.
[22] Mann NR, Schafer RE. Singpurwalla ND. Methods for statistical analysis of
Acknowledgements reliability and life data. New York (NY): John Wiley & Sons Inc.; 1974.
[23] Imam RL, Conover WJ. Small sample sensitivity analysis techniques for
computer models: With an application to risk assessment. Commun Stat 1980;
This work was supported primarily by the Mid-America 9(17):1749–842.
Earthquake Center (MAE) through the Earthquake Engineering [24] Song J, Ellingwood BR. Seismic reliability of special moment steel frames with
Research Centers Program of the National Science Foundation welded connections: II. J Struct Eng 1999;125(4):372–84.
[25] Cornell CA, Jalayer J, Hamburger OR, Foutch DA. Probabilistic basis for 2000
under NSF Award No. EEC-9701785. Any opinions, findings and SAC federal emergency management agency steel moment frame guidelines.
conclusions or recommendations expressed in this paper are those J Struct Eng 2002;128(4):526–33.