Collapse Fragility of Steel Structures Subjected To Earthquake Mainshock-Aftershock Sequences

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Collapse Fragility of Steel Structures Subjected to

Earthquake Mainshock-Aftershock Sequences


Yue Li, M.ASCE 1; Ruiqiang Song, S.M.ASCE 2; and John W. Van De Lindt, M.ASCE 3

Abstract: This paper investigates the collapse probability of mainshock-damaged steel buildings in aftershocks, as an essential part of
Downloaded from ascelibrary.org by UNIVERSITY OF HOUSTON on 12/08/14. Copyright ASCE. For personal use only; all rights reserved.

developing a framework to integrate aftershock seismic hazard into performance-based engineering (PBE). Analytical studies were conducted
utilizing structural degradation models derived from existing publicly available NEEShub data. During earthquake events, aftershocks have
the potential to cause severe damage to buildings and threaten life safety even when only minor damage is present from the mainshock. While
aftershocks are normally somewhat smaller in magnitude, their ground motion intensity is not always smaller. Aftershocks may have a higher
peak ground acceleration than the mainshock, even longer duration, and significantly different energy content as a result of the change in their
location relative to the site. To date, the description of seismic hazard in PBE has not included the probability of aftershocks. In this study, the
structural degradation model of a four-story code-compliant steel moment-resisting frame is calibrated using existing publicly available
NEEShub data. Three approaches to generate collapse fragility for the steel building that sustain a certain state of damage from a mainshock
are used to investigate the effect of damage states from mainshocks on the structural collapse capacity. It is found that structural collapse
capacity may reduce significantly when the building is subjected to a high intensity mainshock. As a result, the structure is likely to
collapse even if only a small aftershock follows the mainshock. In addition, the effects of mainshock records, fault types and spectral
shapes of aftershocks on the structural collapse capacity, are evaluated, respectively. DOI: 10.1061/(ASCE)ST.1943-541X.0001019.
© 2014 American Society of Civil Engineers.
Author keywords: Aftershock hazard; Collapse risk; Degraded system; Earthquake damage; Fragility; Performance-based engineering;
Reliability; Steel structures; Structural safety and reliability.

Introduction its aftershocks. In addition, the economic loss was estimated to


be around $150 billion (RMS 2008). There have also been occur-
During earthquake events, it is not uncommon to observe after- rences of several large earthquakes, seemingly related, but not
shocks following a mainshock. On April 11, 2012, a M8.6 earth- necessarily aftershocks. For example, consider the series of large
quake struck Indonesia, followed by several strong aftershocks earthquakes, known as the New Madrid Earthquakes of 1811–
with the largest measured at M8.2 just over two h later (USGS 1812, which included three earthquakes ranging from M8.1–M8.3
2012). After the 2011 Great Tohoku earthquake in Japan, 588 after- that caused extensive damage as a result of all three earthquakes.
shocks with M5.0 or greater were recorded with 60 aftershocks A mainshock may trigger aftershocks along the fault very far
being over M6.0 and three over M7.0 (USGS 2011). In the 24 h away from the mainshock center (Alliard and Léger 2008; Yeo
following the M8.8 earthquake in Chile on February 27, 2010, and Cornell 2009). The delay between a mainshock and the largest
approximately 90 aftershocks with magnitudes equal to or larger aftershock can range between several minutes to months. The delay
than 5.0 were recorded by the USGS (2010). The Wenchuan earth- is difficult to predict, while magnitudes of aftershock are relatively
quake occurred on May 12, 2008 with a magnitude of M7.9. easy to predict (Scholz 2002). In general, the occurrence rate de-
By September 8, 2008, there had been 42,719 total aftershocks, creases as time goes by after the mainshock. The magnitude of an
of which 34 were from M5.0 to M5.9, and eight were from M6.0 aftershock is usually less than the mainshock, but the aftershock
to M6.5(RMS 2008). These strong aftershocks contributed to
may have a higher peak ground acceleration (PGA) than the main-
the collapse of many of the buildings that sustained damage
shock, even longer duration, and different energy content (Li et al.
from the mainshock, causing even more loss of life. More than
2012). This combination of a mainshock and aftershocks would
70,000 people lost their lives in the Wenchuan earthquake and
require structures to dissipate more energy.
1 Aftershocks have the potential to cause severe damage to build-
Associate Professor, Dept. of Civil and Environmental Engineering,
Michigan Technological Univ., Houghton, MI 49931 (corresponding ings and threaten life safety even when only minor or no damage is
author). E-mail: yueli@mtu.edu present from a mainshock. Particularly, buildings with deteriorated
2
Graduate Research Assistant, Dept. of Civil and Environmental or degraded structural properties are more susceptible to damage.
Engineering, Michigan Technological Univ., Houghton, MI 49931. E-mail: In the 1999 Taiwan Chi-Chi earthquake, a gas station collapsed
rsong1@mtu.edu in an aftershock after it had sustained damage in the mainshock
3
George T. Abell Professor in Infrastructure, Dept. of Civil and Envir- (Lew et al. 2000). A M7.1 earthquake hit New Zealand’s second
onmental Engineering, Colorado State Univ., Fort Collins, CO 80523- largest city, Christchurch on September 4, 2010. After 5 months, a
1372. E-mail: jwv@engr.colostate.edu
M6.3 aftershock occurred on February 22, 2011. This sequence re-
Note. This manuscript was submitted on August 31, 2012; approved on
December 2, 2013; published online on June 4, 2014. Discussion period sulted in 185 deaths and approximately US$15 billion rebuild costs
open until November 4, 2014; separate discussions must be submitted (Parker and Steenkamp 2012). Fig. 1 compares the peak interstory
for individual papers. This paper is part of the Journal of Structural En- drifts of a steel building (described in the following section of this
gineering, © ASCE, ISSN 0733-9445/04014095(10)/$25.00. paper and used throughout as an example) from the mainshock and

© ASCE 04014095-1 J. Struct. Eng.

J. Struct. Eng. 2014.140.


4
system-oriented philosophy and thus the behavior of system should
Mainshock
be modeled accurately. Large-scale test specimens are very expen-
3 MS AS sequence sive to construct and test and therefore are often tested multiple
Story number

times even after they are damaged during the earlier tests. This type
of data is exactly what is needed for calibration of this type of de-
2 grading numerical model. Test data from one past Network for
Earthquake Engineering Simulation Research (NEESR) project
1 will be used to calibrate the system (global)-level model, which
includes numerous degrading hysteresis components. The array of
available test data from the NEESR program will allow this project
0 to eliminate some of the assumptions used in previous studies
0.0 1.0 2.0 3.0 4.0 5.0 6.0 (e.g., Li and Ellingwood 2007a; Yeo and Cornell 2005) and cali-
Interstory drift (%)
Downloaded from ascelibrary.org by UNIVERSITY OF HOUSTON on 12/08/14. Copyright ASCE. For personal use only; all rights reserved.

brate system-level hysteresis models to represent the degraded and


Fig. 1. Comparison of interstory drifts damaged systems.
This paper focuses on utilizing the NEEShub existing data from
large-scale testing to calibrate degraded system-level models in
order to provide accurate estimates of the effect of multiple earth-
mainshock-aftershock (MS-AS) sequence, recorded at the REHS quakes (i.e., mainshock-aftershock) on collapse fragility (probabil-
station in New Zealand. It can be observed that the peak interstory ity of collapse conditioned on some intensities—most often the
drift dramatically increased in the MS-AS sequence. spectral acceleration at the fundamental periods a (T 1 ) of a struc-
Performance-based engineering (PBE) has been defined as a ture), a key component in the development of PBE considering the
seismic engineering approach based on specific performance ob- aftershock hazard.
jectives and safety goals of building occupants, owners, and the
public. It relies on probabilistic or deterministic evaluation of seis-
mic hazard and utilizes quantitative evaluation of design alterna- Prototype Steel Structures and Calibration
tives against the performance objectives. To date, the description
of seismic hazard in PBE has not included the probability of after- The steel structure used in this study is a four-story office building
shocks (ATC-58 2011). Luco et al. (2004) investigated the residual located in Los Angeles (34:110 N, 118.550 W), which is a typical
capacity of a mainshock-damaged building to withstand an after- urban California site with high levels of seismicity. A site class D
shock. Yeo and Cornell (2005) proposed a conceptual analytic soil is assumed. The design is in accordance to the 2003 Inter-
framework for the incorporation of aftershocks into performance- national Building Code (IBC 2003), the 2002 Minimum Design
based earthquake engineering. Li and Ellingwood (2007a) assessed Loads for Buildings and Other Structures (ASCE-7 2002), and
structural response of steel frame buildings under mainshock- the 2005 American Institute of Steel Construction seismic provi-
aftershock sequences. van de Lindt (2008) demonstrated experi- sions (AISC 2005). The structural lateral force resisting system
mentally that the effect of previous earthquakes on the performance consists of special moment-resisting frame (SMRF) perimeter
or integrity of a woodframe structure during the next earthquake steel frame with fully restrained reduced beam sections (RBS).
was significant if the structure sustains a certain state of damage The prototype four-story steel moment-resisting frame structure
in the first event. Yin and Li (2010) estimated the seismic in the East-West (EW) direction (shown in Fig. 2) is selected as
loss of wood-frame buildings subjected to mainshock-aftershock the focus of the study. The first three modal periods of the EW
sequences. frames are 1.32, 0.39, and 0.19 seconds, respectively. Two 1:8 scale
Past studies have calibrated component or subassembly-level model frames were built and tested until collapse at the NEES
hysteresis and then assumed the behavior of the components could facility at the State University of New York at Buffalo. Details of
be combined to predict system behavior. This is true for many mod- the model and tests can be found in Lignos and Krawinkler (2009).
The publicly available test data at the NEES data repository in
els at low deformation levels where nonlinearity is limited to
NEEShub provided an opportunity to calibrate damage models and
material behavior, but when geometric nonlinearity is present
evaluate building performance when subjected to multiple earth-
these models no longer provide accurate prediction. Ibarra and
quakes. It is very common to perform multiple tests on a specimen,
Krawinkler (2005) pointed out that some hysteretic models that
have been calibrated as damage models are component-based, but
calibrated system models are needed to accurately represent the 4
W21X93 W21X93
W24X76
W24X76

W24X76

system (global) performance. Furthermore, very few studies have


3@3.7m=11.1m

taken into account the deterioration of strength and stiffness in the 3


nonlinear range (Ibarra and Krawinkler 2005). When buildings are W21X93 W21X93
subjected to multiple earthquakes, the damaged building model fol-
2
lowing the first earthquake (e.g., mainshock) should be used in the W27X102 W27X102
W24X131
W24X131

W24X131

subsequent analysis and the cumulative damage from the main-


shock accounted for the throughout subsequent analyses. This is 1
W27X102 W27X102
typically done when comparing numerical predictions with test
4.6m

results if numerous tests are performed on the same unrepaired


specimen (Pei and van de Lindt 2009).
Clearly there is significant uncertainty in the capacity of main-
shock-damaged buildings as the characteristics of aftershocks 4@9.1m=36.6m
(e.g., frequency contents, durations and spectral shapes) are quite
Fig. 2. Four-story moment-resisting prototype building in EW
complex. Thus, a systematic methodology to integrate mainshock-
direction
aftershock seismic hazard into PBE would be needed. PBE is a

© ASCE 04014095-2 J. Struct. Eng.

J. Struct. Eng. 2014.140.


2
with one or more on an already damaged structure, making these
data sets perfect for inclusion in the proposed study. The data set is 1.8
CLE

Ground motion multiplier


used to calibrate damage models for analysis in the mainshock- 1.6
aftershock sequences. 1.4 MCE
The Open System for Earthquake Engineering Simulation 1.2
platform (OpenSees 2012) was used to simulate the EW moment
1
frame. The steel moment-resisting frame structure was modeled DLE
with elastic beam-column elements connected by zerolength ele- 0.8
ments, which served as plastic hinge rotational springs to represent 0.6
Experiment Data
the nonlinear structural behavior. The bilinear hysteretic response 0.4
SLE Analytical Prediction
of the spring is based on the modified Ibarra-Krawinkler deterio- 0.2
ration model owing to its ability to capture the structural strength 0
Downloaded from ascelibrary.org by UNIVERSITY OF HOUSTON on 12/08/14. Copyright ASCE. For personal use only; all rights reserved.

and stiffness degradation associated with structural damages 0 0.05 0.1 0.15
(Lignos and Krawinkler 2009). The hysteretic model utilizes a Maximum roof drift
trilinear monotonic backbone curve to define the structural behav-
ior and a cyclic hysteretic model to take into account the basic Fig. 3. IDAs for the testing Frame 1 based on analytical prediction and
strength, postcapping strength, and unloading stiffness deteriora- experiment data
tion. An important parameter for this study is the cyclic hysteretic
energy dissipation capacity (Et ), defined as the product of the refer-
ence cumulative rotation capacity (Λ) and the effective yield different. The magnitude distribution of aftershocks is independent
strength (My ) of the structural component (Lignos et al. 2011). of elapsed time after the mainshock, which means large aftershocks
Model parameters for structural steel components can be deter- may occur several months later, as observed after the Wenchuan
mined from the empirical equations from the calibration of more and Christchurch Earthquakes. Meanwhile, the mean occurrence
than 300 steel specimens (Lignos and Krawinkler 2011). In order rate and the distribution of aftershocks have strong correlations
to simulate P-Delta effects, a leaning column carrying gravity loads with the mainshock magnitude (Reasenberg and Jones 1994; Yeo
is linked to the frame and modeled as beam-column elements and Cornell 2005).
jointed by zero length rotational spring elements with very small It is a great challenge to accurately predict aftershocks after a
stiffness values to avoid carrying significant moments. Since each mainshock. Both synthesized seismic sequences and as-recorded
flexural member is modeled as an elastic element with plastic hinge seismic sequences are employed in this study. In FEMA P695
rotational springs at the end, the structural properties of these com- (ATC-63 2009), there are a suite of 22 ground motions in the far-
ponents must be modified so that the equivalent structural proper- field record set and 28 ground motions in the near-field record set.
ties of the assembly is the same as the actual frame members (Ibarra The magnitude for each of the ground motions is between M6.5 and
and Krawinkler 2005). M7.9. Available strong motion databases, such as the Pacific Earth-
Based on the modeling parameters provided by Lignos and quake Engineering NGA Database (PEER 2011) and Center for
Krawinkler (2009), the analytical model was calibrated with the Engineering Strong Motion Data (CESMD 2012), provide the ac-
1:8 scale testing frames. The first and second mode periods of celeration time-histories of recorded mainshocks and their corre-
the analytical model are 0.46 and 0.15 seconds, respectively, which sponding aftershocks.
matches with the periods (0.47 and 0.13 seconds) of the test speci- The synthesized sequences can be generated by seeding the re-
men quite well. The intensities of earthquake ground motions in corded ground motions using the repeated approach or randomized
the test were increased incrementally until the structure collapses. approach. The synthesized sequences by the repeating approach,
The intensities include four levels: service-level earthquake (SLE), termed as repeated seismic sequence, are generated to repeat a
design-level earthquake (DLE), maximum considered earth- mainshock as an artificial aftershock and scale the intensity of
quake (MCE), and collapse-level earthquake (CLE) (Lignos and shocks based on ground motion models. This method can be con-
Krawinkler 2009). The maximum roof drift of the analytical sidered as a conservative way to estimate the seismic performance
model at the four intensity levels are shown by the dashed line of buildings on the assumption that characteristics of a main-
in Fig. 3. It can be seen that the peak roof drifts for SLE, DLE, shock and aftershocks, such as the frequency content and seismic
and MCE predicted by the analytical model match well with the duration, are the same. By comparison, the synthesized sequences
experiment data. For the CLE shaking, the predicted maximum roof using the randomized approach, termed as the randomized seismic
drift is slightly smaller than the experimental data due to the varia- sequence, are used to generate the artificial seismic sequences
tion of deterioration parameters. The modeling parameters of struc- taking into account the differences of the ground motion features.
tural components of the four-story EW prototype frame is listed in
Table 1 and more details can be found in Lignos and Krawinkler
(2009). The collapse fragility curves of the analytical model of the Table 1. Modeling Parameters of Components of the Four-Story EW
prototype steel frame are consistent with the results predicted in Prototype Frame
Lignos and Krawinkler (2009). Section Type M c =M y a kb θp c θpc d Λe
W24 × 76 Column 1.05 0.40 0.025 0.35 1.50
W24 × 131 Column 1.05 0.40 0.025 0.30 1.50
Earthquake Mainshock-Aftershock Ground Motions W21 × 93 Beam 1.05 0.40 0.025 0.19 1.90
W27 × 102 Beam 1.05 0.40 0.020 0.16 1.50
Structural performance during an earthquake is greatly impacted by a
Post-yield strength ratio.
variability in the seismic loading. The selection of ground motion b
Residual strength ratio.
records to represent realistic amplitude, frequency content, and du- c
Plastic deformation capacity.
ration for aftershocks is a challenge, in part because disaggregation d
Post-capping deformation capacity.
e
of seismic hazard due to mainshocks and aftershocks are typically Reference cumulative rotation capacity.

© ASCE 04014095-3 J. Struct. Eng.

J. Struct. Eng. 2014.140.


1.8
Recorded ground motions are selected randomly as seeds to gen-
erate seismic sequences by scaling the ground motion intensity. The 1.6
combination of a mainshock and only one aftershock is considered 1.4
as the first step to simplify the analysis in this study.
1.2
In order to examine the effects of a certain state of structural

Sa(T1) [g]
damage sustained from a mainshock followed by an aftershock 1
on the structural collapse capacity, the intensity level of the main- 0.8
shock is scaled by the factor γ m to obtain the specific structural Only Mainshock
0.6
damage state 0.7% Drift from MS + AS
0.4
Fm Sa;a 2.5% Drift from MS + AS
γm ¼ × ð1Þ 0.2
Sa;m Fa 5.0% Drift from MS + AS
0
Downloaded from ascelibrary.org by UNIVERSITY OF HOUSTON on 12/08/14. Copyright ASCE. For personal use only; all rights reserved.

0.00 0.05 0.10 0.15 0.20


where γ m = scaling factor for a mainshock in seismic sequences; Peak Drift
Fm = damage state factor of a building due to a mainshock; Sa;m ,
and Sa;a ¼ Sa of a mainshock and aftershock at the fundamental Fig. 5. IDA curves for different damage states from mainshocks
period of structure, respectively; Fa = scaling factor for intensity
level of an aftershock.
According to ASCE-41 (ASCE 2006), the transient drifts are larger than the 20% of the initial tangent slope of the IDA
used for different damage states. Then the spectral acceleration cor- (Vamvatsikos and Cornell 2002). The mechanism of failure of
responding to each damage state (i.e., Fm ) will be determined from the structure is sidesway collapse and multiple modes of failure
incremental dynamic analysis (IDA) of the buildings subject to a are observed. P-Delta effects, strength and stiffness degradation of
mainshock. The specified intensity of mainshock and aftershock components and loading history mainly contribute to the degrada-
can be combined together to carry out the IDA and determine tion of structural collapse capacity.
the residual collapse capacity of the building. According to ASCE-41 (ASCE 2006), the structural perfor-
mance for steel moment frames can be defined as three states.
The performance states: immediate occupany, life safety, and col-
Incremental Dynamics Analysis for Building lapse prevention, are defined by 0.7%, 2.5%, and 5% transient drift,
Collapse Capacity respectively. The three performance states can be viewed as minor,
moderate, and severe damage for the steel frame.
The seismic collapse capacity of a structural system can be deter- In order to investigate the effect of various damage states from
mined by IDA (Vamvatsikos and Cornell 2002). An IDA involves a mainshocks on the structural collapse capacity, the IDA analysis is
series of nonlinear dynamics time history analyses of the structure first carried out using earthquake No. 11 randomly selected from
subjected to a ground motion of increasing intensity. In addition, an the far-field record set, which is used as the source for scaling main-
ensemble of ground motion records, each record in the ensemble shocks and aftershocks. Fig. 5 presents the IDA curves for different
being scaled to multiple levels of intensity with respect to the damage states from mainshocks. It can be seen that the IDA curve
Sa ðT 1 Þ, is often used and is sometimes referred to as a multirecord for the building with minor damage state almost overlaps the
IDA. The scaling levels of seismic intensity are appropriately se- undamaged structural fragility curve. However, there are obvious
lected to force a building undergoing the entire range of behavior, differences between the IDA curve between the moderately or se-
from elastic to inelastic and finally to global dynamic instability in verely damaged building and the undamaged structure. The struc-
the form of large engineering demands (e.g., interstory drift), in- tural collapse capacity listed in Table 2 is decreased dramatically
dicating the collapse of the building. Fig. 4 shows the IDA results when the steel frame is in a severe damage state from a mainshock,
for the steel structure in which the 22 ground motions for the far- reduced from 1.5 g to 0.9 g. The collapse capacity has a loss of 13%
field record set in FEMA P695 (ATC-63 2009) is employed as when the building is subjected to a moderate state of damage. Addi-
mainshocks. The IDA curves present the maximum interstory drifts tionally, the peak interstory drift also has a slight increase for the
when the building is subjected to increasing levels of ground damaged building.
motion intensity. Based on the IDA curves, the structural collapse
capacity is determined as the last point on the IDA curve that is
Collapse Fragility for Undamaged and Damaged
4.0 Buildings
3.5
After the structural collapse capacity is determined by the IDA
3.0
curves, a lognormal distribution is fitted to generate the collapse
2.5 fragility curve using Sa as a random variable. The seismic fragilities
Sa(T1)[g]

2.0 for most building types can be modeled by a lognormal distribution


1.5
1.0
Table 2. IDA for Different Damage States from Mainshocks

0.5
Damage state sustained from Collapse capacity Drift
a mainshock Sa ðT 1 Þ (g) (%)
0.0
0% 5% 10% 15% 20% N/A (only mainshock) 1.5 12
Maximum Drift Minor (I.O. + AS) 1.5 12
Moderate (L.S. + AS) 1.3 12.1
Fig. 4. IDA curves for steel structure subjected to mainshocks Severe (C.P. + AS) 0.9 13.9

© ASCE 04014095-4 J. Struct. Eng.

J. Struct. Eng. 2014.140.


(Li and Ellingwood 2007b; Li et al. 2010; Zareian and Krawinkler Table 3. Statistics of Collapse Capacity
2007). In order to investigate the effect of various states of damage Randomized
from a mainshock, the IDA curves for the three performance states: Repeated sequence sequence
immediate occupany, life safety, and collapse prevention are gen-
Damage state sustained Mean Mean
erated, and the corresponding collapse fragility analysis for the from a mainshock Sa ðT 1 Þ (g) COV Sa ðT 1 Þ (g) COV
three damage states are carried out.
Three types of mainshock-aftershock sequences are taken into N/A 1.35 0.56 1.35 0.56
Minor 1.33 0.55 1.33 0.61
account: repeated seismic sequence, randomized seismic sequence
Moderate 1.06 0.51 1.16 0.55
and as-recorded seismic sequence. The suite of 22 ground motions Severe 0.71 0.47 0.64 0.57
in the far-field record in FEMA P695(ATC-63 2009) is employed
to generate the synthesized seismic sequences. The earthquake
ground motion scaling methods in Eq. (1) is used to determine
are listed in Table 3. Inspection of this table reveals that the mean
Downloaded from ascelibrary.org by UNIVERSITY OF HOUSTON on 12/08/14. Copyright ASCE. For personal use only; all rights reserved.

the structural collapse capacity when the steel frame is subjected


value of structural collapse capacity of undamaged buildings is vir-
to three different damage states from mainshocks.
tually the same as that of the minor damaged building. The mean
capacity decreases from 1.35 g to 1.06 g, a loss of 22% collapse
Repeated Seismic Sequences capacity when the building is subjected to moderate damage from
the earthquake. In comparison, less than 50% of the original struc-
Due to the large variation of structural collapse capacity in terms of tural collapse capacity remains (i.e., from 1.35 g to 0.71 g) when
Sa ðT 1 Þ, different earthquake ground motion records may result in severe damage presents in the building. The analysis shows that the
different structural collapse capacities. In order to examine the ef- structural collapse capacity may be reduced significantly if the
fects of damage states sustained from a mainshock on the structural building is subjected to a high intensity mainshock. As a result,
collapse capacity, an IDA is carried out using the aftershocks only the probability of structural collapse increases even if aftershocks
to determine the collapse capacity of the undamaged building. with low seismic intensity level follow after the mainshock.
Then the mainshock-aftershock sequences with three damage states
already present in the structure as a result of the mainshocks are
Randomized Seismic Sequences
employed to determine the residual collapse capacity of the dam-
aged building. Comparing the collapse capacities determined by The randomized sequences seem more rational than repeated se-
aftershock only and mainshock-aftershock sequence with the speci- quences for generating seismic sequence because the differences
fied damage states, the effect of the damage state from the main- in ground motion features are included. Here, the component 1 of
shock on the structural collapse capacity can be quantified. earthquake ID No. 4 is randomly selected as the mainshock and the
Fig. 6 compares the collapse fragility curves for the undamaged remaining 21 ground motions are used as the artificial aftershocks.
and damaged building at three damage states. The collapse fragility Fig. 7 compares the collapse fragility curves for different ran-
curves of the undamaged building are shown by the “Only after- domized seismic sequences. It can be seen that the fragility curves
shock” curve. The fragility curves shown on the figure of 0.7%, are similar to those observed with the repeated seismic sequences.
2.5%, and 5.0% drift from MS + AS represent the probability of The fragility curve with the severe damage state in Fig. 7 is more
collapse of the building with minor, moderate, and severe damage vulnerable than the curve determined by the repeated seismic se-
due to the mainshock, respectively. The collapse fragility curve for quences. It indicates that the building with the severe damage state
the building with minor damage almost overlaps the fragility curves may have a higher collapse risk when the randomized sequences
of the undamaged building. However, there is an obvious difference instead of the repeated sequences is used to determine the collapse
between the collapse fragility curves between the moderately or capacity.
severely damaged building and the undamaged building. Given the The statistics of collapse capacity for different randomized seis-
same probability of collapse, the collapse capacity of a building mic sequences is listed in Table 3. It should be noted that the mean
with the moderate or severely damaged state is remarkably less than collapse capacity with severe damage state for these randomized
that of an undamaged building. sequences is reduced to just half of the mean collapse capacity
The mean and coefficient of variation (COV) of the collapse without damage state, from 1.35 g to 0.64 g. For the building with
damage state threshold for different structural performance states a moderate damage state, there is about 14% loss of the collapse

1.0 1.0
0.9 0.9
0.8 0.8
Probability of Collapse

Probability of Collapse

0.7 0.7
0.6 0.6 Only aftershock
0.5 0.5 0.7% Drift from MS + AS
Only aftershock
0.4 0.4 2.5% Drift from MS + AS
0.7% Drift from MS + AS
0.3 0.3
2.5% Drift from MS + AS 5.0% Drift from MS + AS
0.2 0.2
0.1 5.0% Drift from MS + AS 0.1
0.0 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Sa(T1) [g] Sa(T1) [g]

Fig. 6. Collapse fragility curves for repeated seismic sequences Fig. 7. Collapse fragility curve for randomized seismic sequences

© ASCE 04014095-5 J. Struct. Eng.

J. Struct. Eng. 2014.140.


1.0
capacity, decreased from 1.35 g to 1.16 g, compared to the intact
structure. 0.9
0.8

Probability of Collapse
0.7
As-Recorded Seismic Sequences 0.6
The as-recorded seismic sequences are employed to investigate 0.5
the effect of the damage states from mainshocks on the structural 0.4 As-recorded aftershock
collapse capacity. The following criteria are applied to select 0.3 As-recorded MS-AS sequence
mainshock-aftershock sequences from the PEER (2011) and
0.2
CESMD (2012): the magnitude of aftershock is greater than 5.0;
0.1
the acceleration sensor is located on free field or low-height build-
ings to neglect the soil-structure interaction effects; and the peak 0.0
0.0 1.0 2.0 3.0 4.0
Downloaded from ascelibrary.org by UNIVERSITY OF HOUSTON on 12/08/14. Copyright ASCE. For personal use only; all rights reserved.

ground acceleration (PGA) of the horizontal component of a main- Sa(T1) [g]


shock is greater than 0.4 g. Under these criteria, eight seismic se-
quences are first selected from the 1994 Northridge earthquake, the Fig. 8. Collapse fragility curves for as-recorded seismic sequences
1999 Chi-Chi earthquake, and the 2011 Tohoku earthquake. Table 4
lists relevant information about the selected sequences.
The statistics of collapse capacity for as-recorded sequences the high potential to cause severe damage to buildings and threaten
shows the mean collapse capacity is only 0.67 g with a COV 0.72 life safety due to the reduction of structural collapse capacity from a
for the as-recorded sequence. Compared with a mean collapse mainshock.
capacity of 0.92 g with a COV of 0.66 for the as-recorded after-
shock, indicating the structure lost 27% collapse capacity when
the structural capacity is compromised by the mainshock. Sensitivity Analysis for Earthquake Characteristics
Fig. 8 shows the collapse fragility curves for the as-recorded to Collapse Fragility
aftershock and mainshock-aftershock sequences. The fragility
curve for the as-recorded MS-AS sequence “shifts to the left” of
Effects of Different Mainshocks
the curve for the as-recorded aftershock, meaning with the higher
probability of collapse. It is evident that the damage state sustained Even if both the peak drift of a building from a mainshock and an
from the mainshock has a great influence on the structural collapse aftershock ground motion are the same, the building may have
capacity, where there is a noticeable reduction in the structural a very different response to the mainshock-aftershock sequence.
collapse capacity when the steel building is damaged from the In this study, the influence of different mainshocks in seismic se-
mainshock. quences on the structural collapse fragility is investigated. Each
An interesting observation is that the structural collapse capacity mainshock may cause a specific damage state to the building due
is sensitive to the damage state sustained from a mainshock, no to the differences in earthquake characteristics, such as the fre-
matter how the seismic sequences are generated. All of the repeated quency content and earthquake duration. Even if the maximum in-
sequences, randomized sequences and as-recorded sequences dem- terstory drifts from mainshocks are the same, the structural collapse
onstrate that the collapse capacity may reduce dramatically when capacities determined by the following aftershocks could be dif-
the structural performance is affected by a mainshock. Therefore, ferent, which may be due to the difference of structural damage
the accumulative damage from a mainshock should be consid- accumulation for the post-mainshock building.
ered when determining the structural collapse capacity in post- The randomized seismic sequences based on the far-field record
earthquake investigation. In addition, the aftershocks may have set are applied. From the database, one ground motion is selected as

Table 4. Seismic Sequences Selected from PEER and CESMD


Earthquake Station name Date Magnitude Component
Northridge Castaic—Old Ridge Route 1/17/1994 6.69 NGA_no_963_ORR090.AT2
1/17/1994 5.93 NGA_no_1676_CASTA090.AT2
Rinaldi Receiving Sta 1/17/1994 6.69 NGA_no_1063_RRS318.AT2
3/20/1994 5.28 NGA_no_1728_E-RRS318.AT2
Sylmar—Converter Sta East 1/17/1994 6.69 NGA_no_1085_SCE288.AT2
3/20/1994 5.28 NGA_no_1737_E-SCE288.AT2
Chi-Chi CHY035 9/20/1999 7.62 NGA_no_1202_CHY035-E.AT2
9/20/1999 6.20 NGA_no_2709_CHY035-E.AT2
CHY024 9/20/1999 7.62 NGA_no_1193_CHY024-E.AT2
9/20/1999 6.20 NGA_no_2457_CHY024-E.AT2
Japan HARAMACHI 3/11/2011 9.00 FKS0051103111446.EW
4/7/2011 7.10 FKS00520110407233247.EW
HARAMACHI 3/11/2011 9.00 FKS0051103111446.NS
4/7/2011 7.10 FKS00520110407233247.NS
KITAKAMI 3/11/2011 9.00 IWT0121103111446.EW
4/7/2011 7.10 IWT00720110407233249.EW
Shiogama—MYG012 3/11/2011 9.00 MYG0121103111446.EW
4/7/2011 7.10 MYG01220110407233244.EW
Sendai—MYG013 3/11/2011 9.00 MYG0131103111446.NS
4/7/2011 7.10 MYG01320110407233245.NS

© ASCE 04014095-6 J. Struct. Eng.

J. Struct. Eng. 2014.140.


1.0 1.0
0.9 0.9
0.8 0.8

Probability of Collapse
Probability of Collapse

0.7 0.7
Only Aftershock
0.6 0.6 Only Aftershock(FF)
0.7% Drift from MS-1 + AS
2.5% Drift from MS-1 + AS 0.7% Drift from MS + AS(FF)
0.5 0.5
5.0% Drift from MS-1 + AS 2.5% Drift from MS + AS(FF)
0.4 0.4
0.7% Drift from MS-4 + AS 5.0% Drift from MS + AS(FF)
0.3 2.5% Drift from MS-4 + AS 0.3 Only Aftershock(NF)
5.0% Drift from MS-4 + AS 0.7% Drift from MS + AS(NF)
0.2 0.2
0.7% Drift from MS-10 + AS
2.5% Drift from MS + AS(NF)
2.5% Drift from MS-10 + AS 0.1
0.1 5.0% Drift from MS + AS(NF)
5.0% Drift from MS-10 + AS
Downloaded from ascelibrary.org by UNIVERSITY OF HOUSTON on 12/08/14. Copyright ASCE. For personal use only; all rights reserved.

0.0
0.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Sa (T1) [g] Sa (T1) [g]

Fig. 9. Collapse fragility curves for seismic sequences with different Fig. 10. Collapse fragility curves for seismic sequences with different
mainshocks fault types of aftershocks

the mainshock (MS) from component 1 of each earthquake: with different characteristics may have an apparent effect on the
Records No. 1, No. 4, and No. 10 due to the significant difference structural collapse fragility of the moderate or severe damaged
in the spectral shape, i.e., the epsilons are 1.63, 0.00 and −2.30, buildings. The mainshock with a higher value of epsilon tends
respectively. The rest of ground motions in the set are considered to result in a lower collapse capacity of damaged buildings.
as the aftershocks (AS). The earthquake scaling method in Eq. (1)
is applied to conduct the IDA of mainshock-aftershock sequences Effects of Far Fault and Near Fault Aftershocks
and the structural collapse capacities are determined by taking into
It is common that the locations of a mainshock and aftershocks
account the three damage states.
in an earthquake event are not the same. For example, the
Fig. 9 presents the collapse fragility curves for seismic sequen-
September 2011 Canterbury Earthquake epicenter was at Darfield
ces with different mainshocks. For example, the fragility curve of
in New Zealand, which is about 40 km from the location of the
0.7% Drift from MS-1 + AS shows the probability of collapse for
February 2011 Christchurch aftershock. For some cases, the build-
the building with minor damage from the mainshock No. 1. It can
ings are subjected to the far-fault mainshock followed by the near-
be observed that the fragility curves of the building with minor
fault aftershock. In order to investigate the effect of fault type of
damage from all three mainshocks almost overlap, which means
aftershocks, two cases of seismic sequences: far-fault mainshock
the effect of different mainshocks on the structural collapse fragility followed by a far-fault aftershock and far-fault mainshock followed
can be neglected when the structure only sustain minor damage by near-fault aftershock, are considered in the collapse fragility
from a mainshock. However, the probability of collapse of the analysis. Randomly selecting the component 1 of earthquake ID
building are quite different with moderate and severe damage from No. 4 as the mainshock, the suite of 22 ground motions for the
the three mainshocks. far-field record set and 28 ground motions for the near-field record
Table 5 summaries the statistics of collapse capacity for seismic set are employed to generate the seismic sequences.
sequences with different mainshocks. The three mean collapse Fig. 10 shows the collapse fragility curves for seismic sequences
capacities with minor damage are almost the same, 1.34 g, 1.33 g, with different fault types of aftershocks. The high damage states
and 1.34 g for mainshocks No. 1, 4, and 10, respectively. However, from the mainshock have an obvious influence on the structural
the mean collapse capacities for the moderate or severe damage are collapse fragility curves, regardless of the kinds of fault type of
dissimilar. For instance, the mean collapse capacities are 1.31 g and aftershocks.
0.82 g for moderate and severe damage from the mainshock ID 1, The statistics of collapse capacity for seismic sequences with
respectively, compared with 1.16 g and 0.64 g from the mainshock different fault types of aftershocks are listed in Table 6. The mean
ID 4, respectively. It can be noted that the different mainshocks structural collapse capacity determined by only near-fault after-
shocks is smaller than the one by only far-fault aftershocks. Near-
fault pulse effect may contribute to the reduction of mean structural
Table 5. Statistics of Collapse Capacity for Seismic Sequences with
different Mainshocks
Table 6. Statistics of Collapse Capacity for Different Fault Types of
Mean
Aftershocks
MS Seismic sequence Sa ðT 1 Þ (g) COV
mean
N/A N/A 1.35 0.56
Damage states Fault type of AS Sa ðT 1 Þ (g) COV
1 Minor damage + AS 1.34 0.57
Moderate damage + AS 1.31 0.56 None Far-fault 1.35 0.56
Severe damage + AS 0.82 0.49 Minor (I.O.) 1.33 0.61
4 Minor damage + AS 1.33 0.61 Moderate (L.S.) 1.16 0.55
Moderate damage + AS 1.16 0.55 Severe (C.P.) 0.64 0.57
Severe damage + AS 0.64 0.57 None Near-fault 1.26 0.43
10 Minor damage + AS 1.34 0.56 Minor (I.O.) 1.23 0.45
Moderate damage + AS 1.08 0.45 Moderate (L.S.) 1.16 0.44
Severe damage + AS 0.67 0.44 Severe (C.P.) 0.82 0.45

© ASCE 04014095-7 J. Struct. Eng.

J. Struct. Eng. 2014.140.


collapse capacity. The structural collapse capacity reduces dramati- Table 7. Statistics of Collapse Capacity for Positive and Negative Epsilons
cally after sustaining damage from a mainshock for both the after- Mean
shock fault types. It seems that the structural collapse capacity is Damage states Epsilon of aftershock Sa ðT 1 Þ (g) COV
more sensitive to the far-fault aftershock than near-fault aftershock
None Positive 1.66 0.46
when the severe damage occurs in the building, as the reduction of Minor (I.O.) 1.63 0.46
the mean structural collapse capacity is up to 50% for far-fault Moderate (L.S.) 1.30 0.42
aftershock compared with 35% for near-fault aftershock. The rea- Severe (C.P.) 0.75 0.43
son may be that the natural periods of the building are remarkably None Negative 0.91 0.54
elongated when it suffers a severe damage, resulting in the weak- Minor (I.O.) 0.91 0.54
ened effect of near-fault pulse on the structural collapse. A further Moderate (L.S.) 0.73 0.43
look at the influence of earthquake fault types can be done by ap- Severe (C.P.) 0.64 0.56
plying more ground motions and structure models.
Downloaded from ascelibrary.org by UNIVERSITY OF HOUSTON on 12/08/14. Copyright ASCE. For personal use only; all rights reserved.

the collapse capacity, considering 0.75 g and 0.64 g mean collapse


Effects of Epsilon capacity for earthquake with positive and negative epsilon, respec-
One key characteristic of ground motions affecting structural tively. Figs. 12 and 13 show the collapse fragility curve for seismic
collapse capacity is the spectral shape measured by epsilon. sequences with positive and negative epsilon, respectively. It can be
Epsilon can be computed by subtracting the mean predicted seen that the positive epsilon leads to a low probability of collapse
ln Sa ðT1Þ from the record’s ln Sa ðT1Þ, and then dividing by the while the negative epsilon leads to a high probability of collapse.
logarithmic standard deviation estimated by ground motion models It can be concluded that epsilon of the aftershock has a significant
(Baker and Cornell 2005). The ground motion model proposed effect on the structural collapse fragility for both the intact building
by (Abrahamson and Silva 2008) is adopted in this study. Based and the damaged building.
on the repeated approach, the suite of 22 ground motions in the Because epsilon has a significant effect on the structural col-
far-field record set are employed to generate the seismic sequence. lapse fragility, it is necessary to properly account for the effect
The average epsilon value for the far-field record set at T 1 ¼ 1.32 s of epsilon when evaluating structural collapse risk. The epsilon ad-
is 0.014. justment approach proposed by Haselton et al. (2011) is used to
Fig. 11 shows the relationship between the structural collapse
capacity determined by a specific ground motion and the corre- 1.0
sponding epsilon. It can be observed that larger epsilon tends to 0.9
have a larger collapse capacity prediction, while smaller epsilon
0.8
leads to smaller collapse capacity prediction. This is consistent with
Probability of Collapse

0.7
what has been found in other studies (Baker and Cornell 2005;
Goulet et al. 2007; Yin and Li 2010). In addition, there is quite 0.6
a dispersion of epsilon among the 22 ground motions because 0.5
the ground motions have various spectra shapes. 0.4
In order to investigate the influence of epsilon on the structural Only mainshock
0.3 Aftershock (Positive epsilon)
collapse fragility, the 22 ground motions in the far-field record set
0.2 0.7% Drift from MS + AS (Positive epsilon)
are divided into two groups according to the epsilon of the after- 2.5% Drift from MS + AS (Positive epsilon)
shock, i.e., the earthquake group with positive epsilon and the 0.1
5.0% Drift from MS + AS (Positive epsilon)
group with negative epsilon. The collapse fragility analysis of 0.0
the intact building and building with three damage states are carried 0.0 1.0 2.0 3.0 4.0
Sa(T1) [g]
out for both earthquake groups. Table 7 lists the statistics of col-
lapse capacity for positive and negative epsilons. It is observed that Fig. 12. Collapse fragility curves for seismic sequences with positive
the mean collapse capacities with positive epsilon are larger than epsilon
those with negative epsilon. The minor damage building has a mean
collapse capacity of 1.63 g with positive epsilon, 82% larger than
the mean collapse capacity of 0.91 g with negative epsilon. For 1.0
severely damaged buildings, the epsilon seems a slight effect on 0.9
0.8
4
Probability of Collapse

Structural collapse capacity 0.7


3 Epsilon at 1.32s 0.6

2 0.5
Only aftershock
0.4
Sa(T1) [g]

1 Aftershock (Negative epsilon)


0.3
0.7% Drift from MS + AS (Negative epsilon)
0 0.2
0 5 10 15 20 25 2.5% Drift from MS + AS (Negative epsilon)
-1 0.1
5.0% Drift from MS + AS (Negative epsilon)
0.0
-2 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Sa (T1) [g]
-3
Earthquake ID number
Fig. 13. Collapse fragility curves for seismic sequences with negative
Fig. 11. Structural collapse capacity and the corresponding epsilon epsilon

© ASCE 04014095-8 J. Struct. Eng.

J. Struct. Eng. 2014.140.


Table 8. Statistics of Collapse Capacity for Three Target Epsilons
Coefficients Mean Sa ðT 1 Þ (adjusted epsilon)
Mean Sa ðT 1 Þ
Seismic sequence c0 c1 (w=o epsilon) −2 0 2
No damage + AS 0.15 0.24 1.35 0.71 1.16 1.89
Minor damage + AS 0.13 0.24 1.33 0.71 1.14 1.85
Moderate damage + AS −0.07 0.27 1.06 0.55 0.93 1.59
Severe damage + AS −0.49 0.14 0.71 0.46 0.61 0.82

take into account the effect of epsilon in the collapse assessment. building, the lower mean collapse capacities can beexpected by us-
A general set of ground motions without considering epsilon is ing the sequences instead of repeated sequences. The influence of
Downloaded from ascelibrary.org by UNIVERSITY OF HOUSTON on 12/08/14. Copyright ASCE. For personal use only; all rights reserved.

used first to determine the structural collapse capacity, and then characteristics of mainshock-aftershock sequence, including differ-
the collapse capacities are adjusted to account for epsilon in the ence mainshocks, aftershock types, and epsilon of ground motions
collapse risk analysis. The relationship between collapse capacity to structural collapse fragility are also examined. Different main-
Sa ðT1Þ for each ground motion record and epsilon (ε) can be de- shocks have a significant effect on the structural collapse fragility,
termined by the regression analysis according to when subtaintial damage occured from the mainshock. Epsilon has
a significant effect on the structural collapse fragility. Therefore, the
ln Sa ðT1Þ ¼ c0 þ c1 ε ð2Þ collapse fragility needs to properly account for effect of epsilon
where c1 and c0 are parameters determined by regression analysis. when evaluating structural collapse risk of undamaged or damaged
After determining the values of c1 and c0 , the adjusted collapse building.
capacity corresponding to the target epsilon of interest can be Undamaged fragilities and conditional (damaged) fragilities can
calculated by Eq. (2). be combined with aftershock hazard models to quantify the effect
In this study, three values, −2, 0 and 2, are selected as the target of including aftershock seismic hazard in PBE. A better under-
epsilons of interest to take into account the low, moderate, and large standing of building performance under mainshock-aftershock se-
epsilon of ground motions for structural fundamental period 1.32 s. quences will facilitate achieving the objective of PBE, i.e., making
The collapse capacities are determined by the IDA for the undam- design strategies and risk levels consistent with occupant expect-
aged and damaged building at three (i.e., minor, moderately and ations and social objectives.
severe) damage states by a mainshock, and then they are adjusted
based on the three target epsilons. Table 8 lists the statistics of col-
lapse capacity at three target epsilons. It can be observed that for Acknowledgments
using only aftershock case, the expected mean collapse capacity is
1.89 g for ε ¼ 2, 40% larger than the mean collapse capacity 1.35 g The authors thank Dr. Nicolas Luco at United States Geological
without considering ε. However, for ε ¼ −2 and 0, the mean col- Survey for the discussion during the preparation of this paper. The
lapse capacities using only aftershock are 0.71 g and 1.16 g, 47%, research described in this paper was supported, in part, by the
and 14% smaller than the value without considering ε. There are National Science Foundation (NSF) CMMI Division of Civil,
similar trends for the mean collapse capacity of the damaged build- Mechanical, and Manufacturing Innovation under Grant No.
ing. In order to appropriately estimate the collapse risk for both CMMI-1100423. The support is gratefully acknowledged. How-
undamaged and damaged structures, the collapse fragility obtained ever, the writers take sole responsibility for the views expressed
from the selected ground motion should be adjusted to account for in this paper, which may not represent the position of the NSF
epsilon effects. or their respective institutions.

Conclusions References
Abrahamson, N., and Silva, W. (2008). “Summary of the Abrahamson &
To promote disaster-resilient communities, building design and
Silva NGA Ground-Motion Relations.” Earthquake Spectra, 24(1),
construction practices should address the overall risk from earth- 67–97.
quakes, including aftershock hazard. In order to examine the effects Alliard, P. M., and Léger, P. (2008). “Earthquake safety evaluation of grav-
of different structural damage states from a mainshock and the ity dams considering aftershocks and reduced drainage efficiency.”
following aftershocks on the structural collapse fragility, a scaling J. Eng. Mech., 10.1061/(ASCE)0733-9399(2008)134:1(12), 12–22.
method is used to combine the mainshock and the aftershock. American Institute of Steel Construction (AISC). (2005). “Seismic Provi-
Three approaches to generate mainshock-aftershock sequences sions for Structural Steel Buildings.” Including Supplement No. 1,
were applied to investigate the effect of different damage states AISC, Chicago, IL.
on structural collapse capacity, which includedthe repeated seismic ASCE-41. (2006). Seismic rehabilitation of existing buildings, American
sequence, the randomized seismic sequence and the as-recorded Society of Civil Engineers, Reston, VA.
seismic sequence. The structural collapse capacity may reduce ASCE-7. (2002). Minimum design loads for buildings and other structures,
significantly when the building is subjected to a high intensity Structural Engineering Institute (SEI), Reston, VA.
ATC-58. (2011). Development of next generation performance-based
mainshock and the structure is likely to collapse even if a small
seismic design procedures for new and existing building, 75% draft
aftershock follows the mainshock. It can be noted that the dispe- ATC-58-1, Applied Technology Council, Redwood City, CA.
rsion of the collapse capacity are low, moderate and large for re- ATC-63. (2009). Quantification of building seismic performance factors,
peated sequence, randomized sequence, and as-recorded sequence, FEMA P695, Redwood City, CA.
respectively. It indicates that using synthesizing mainshock- Baker, J. W., and Cornell, A. C. (2005). “A vector-valued ground motion
aftershock sequences in the analysis may underestimate the intensity measure consisting of spectral acceleration and epsilon.”
variation of collapse capacity. For the severe mainshock-damaged Earthquake Eng. Struct. Dynam., 34(10), 1193–1217.

© ASCE 04014095-9 J. Struct. Eng.

J. Struct. Eng. 2014.140.


CESMD. (2012). “Center for engineering strong motion data.” 〈http://www OpenSees. (2012). “The open system for earthquake engineering simula-
.strongmotioncenter.org/〉 (Aug. 28, 2012). tion.” 〈http://opensees.berkeley.edu/〉 (Sep. 5, 2012).
Goulet, C. A., et al. (2007). “Evaluation of the seismic performance of a Parker, M., and Steenkamp, D. (2012). “The economic impact of the
code-conforming reinforced-concrete frame building—from seismic Canterbury earthquakes.” Reserve Bank of New Zealand Bulletin,
hazard to collapse safety and economic losses.” Earthquake Eng. Struct. 75(3), 13–25.
Dynam., 36(13), 1973–1997. PEER. (2011). “Pacific earthquake engineering research center PEER
Haselton, C. B., Baker, J. W., Liel, A. B., and Deierlein, G. G. (2011). strong motion database.” 〈http://peer.berkeley.edu/peer_ground_motion
“Accounting for ground-motion spectral shape characteristics in struc- _database/spectras/16318/unscaled_searches/new〉 (Oct. 1, 2011).
tural collapse assessment through an adjustment for epsilon.” J. Struct. Pei, S., and van de Lindt, J. W. (2009). “Methodology for earthquake-
Eng., 10.1061/(ASCE)ST.1943-541X.0000103, 332–344. induced loss estimation: An application to woodframe buildings.”
Ibarra, L. F., and Krawinkler, H. (2005). “Global collapse of frame struc- Struct. Saf., 31(1), 31–42.
tures under seismic excitations.” Rep. No. TB 152, The John A. Blume Reasenberg, P. A., and Jones, L. M. (1994). “Earthquake aftershocks: up-
Earthquake Engineering Center, Stanford Univ., Stanford, CA. date.” Science, 265(5176), 1251–1252.
IBC. (2003). International building code, International Code Council RMS. (2008). “Reconnaissance report: The 2008 wenchuan earthquake:
Downloaded from ascelibrary.org by UNIVERSITY OF HOUSTON on 12/08/14. Copyright ASCE. For personal use only; all rights reserved.

(ICC), Falls Church, VA. Risk management lessons and implications.” Risk Management
Lew, M., Naeim, F., Huang, S. C., Lam, H. K., and Carpenter, L. D. (2000). Solutions, Newark, CA.
“Geotechnical and geological effects of the 21 September 1999 Chi-Chi Scholz, C. H. (2002). The mechanics of earthquakes and faulting,
earthquake, Taiwan.” Struct. Des. Tall. Spec. Build., 9(2), 89–106. Cambridge University Press, Cambridge, U.K.
Li, Q., and Ellingwood, B. R. (2007a). “Performance evaluation and
U.S. Geological Survey (USGS). (2010). “Magnitude 8.8—offshore
damage assessment of steel frame buildings under main shock–
Bio-bio, Chile.” 〈http://earthquake.usgs.gov/earthquakes/eqinthenews/
aftershock earthquake sequences.” Earthquake Eng. Struct. Dynam.,
2010/us2010tfan/#summary〉 (Apr. 5, 2012).
36(3), 405–427.
U.S. Geological Survey (USGS). (2011). “Magnitude 9.0—near the east
Li, Y., and Ellingwood, B. R. (2007b). “Reliability of woodframe residen-
coast of Honshu, Japan.” 〈http://earthquake.usgs.gov/earthquakes/
tial construction subjected to earthquakes.” Struct. Saf., 29(4), 294–307.
eqinthenews/2011/usc0001xgp/#summary〉 (Apr. 5, 2012).
Li, Y., Song, R., van de Lindt, J. V., Nazari, N., and Luco, N. (2012).
“Integration of aftershock seismic hazard into performance-based engi- U.S. Geological Survey (USGS). (2012). “Magnitude 8.6—off the west
neering.” 15th World Conf. Earthquake Engineering, Lisbon, Portugal. coast of Northern Sumatra.” 〈http://earthquake.usgs.gov/earthquakes/
Li, Y., Yin, Y., Ellingwood, B. R., and Bulleit, W. M. (2010). “Uniform recenteqsww/Quakes/usc000905e.php#summary〉 (Jun. 6, 2012).
hazard versus uniform risk bases for performance-based earthquake Vamvatsikos, D., and Cornell, C. A. (2002). “Incremental dynamic
engineering of light-frame wood construction.” Earthquake Eng. Struct. analysis.” Earthquake Eng. Struct. Dynam., 31(3), 491–514.
Dynam., 39(11), 1199–1217. van de Lindt, J. W. (2008). “Experimental investigation of the effect of
Lignos, D. G., and Krawinkler, H. (2009). “Sidesway collapse of deterio- multiple earthquakes on woodframe structural integrity.” Pract. Period.
rating structural systems under seismic excitations.” Rep. No. TB 172, Struct. Des. Construct., 10.1061/(ASCE)1084-0680(2008)13:3(111),
The John A. Blume Earthquake Engineering Center, Stanford Univ., 111–117.
Stanford, CA. Yeo, G. L., and Cornell, C. A. (2005). “Stochastic characterization and de-
Lignos, D. G., and Krawinkler, H. (2011). “Deterioration modeling of steel cision bases under time-dependent aftershock risk in performance-
components in support of collapse prediction of steel moment frames based earthquake engineering.” Rep. No. TB 149, The John A. Blume
under earthquake loading.” J. Struct. Eng., 10.1061/(ASCE)ST.1943- Earthquake Engineering Center, Stanford Univ., Stanford, CA.
541X.0000376, 1291–1302. Yeo, G. L., and Cornell, C. A. (2009). “Building life-cycle cost analysis due
Lignos, D. G., Krawinkler, H., and Whittaker, A. S. (2011). “Prediction and to mainshock and aftershock occurrences.” Struct. Saf., 31(5), 396–408.
validation of sidesway collapse of two scale models of a 4-story steel Yin, Y. J., and Li, Y. (2010). “Seismic collapse risk of light-frame wood
moment frame.” Earthquake Eng. Struct. Dynam., 40(7), 807–825. construction considering aleatoric and epistemic uncertainties.” Struct.
Luco, N., Bazzurro, P., and Cornell, C. A. (2004). “Dynamic versus static Saf., 32(4), 250–261.
computation of the residual capacity of a mainshock-damaged building Zareian, F., and Krawinkler, H. (2007). “Assessment of probability of col-
to withstand an aftershock.” Proc., 13th World Conf. Earthquake lapse and design for collapse safety.” Earthquake Eng. Struct. Dynam.,
Engineering, Vancouver, Canada. 36(13), 1901–1914.

© ASCE 04014095-10 J. Struct. Eng.

J. Struct. Eng. 2014.140.

You might also like