Article
An open-source framework
for regional earthquake
loss estimation using the
city-scale nonlinear time
history analysis
Earthquake Spectra
1–26
Ó The Author(s) 2020
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/8755293019891724
journals.sagepub.com/home/eqs
Xinzheng Lu, M.EERI1, Frank McKenna2, Qingle Cheng1,
Zhen Xu3, Xiang Zeng1 and Stephen A Mahin2
Abstract
Regional seismic damage simulation of buildings provides decision-makers with
important information for earthquake disaster prevention and mitigation. Utilizing
nonlinear time history analysis using multiple-degree-of-freedom (MDOF) models
for buildings, and the next-generation performance-based earthquake engineering, an
open-source general-purpose scientific workflow for seismic damage simulation and
loss prediction of urban buildings (referred to as SimCenter Workflow) is presented
in this study. To introduce the SimCenter Workflow process in detail and demonstrate its advantages, a seismic damage simulation and loss prediction for 1.8 million
buildings in the San Francisco Bay Area were performed using the SimCenter
Workflow. The open nature and modularization of the SimCenter Workflow facilitate its extensibility and make it practical for researchers to apply to seismic damage
simulations in other regions.
Keywords
city-scale nonlinear time history analysis; regional earthquake loss estimation; opensource framework; Bay Area
Date received: 13 November 2018; accepted: 23 September 2019
Introduction
Earthquakes are potential threats to modern cities with dense populations and infrastructures. Recently, powerful earthquakes such as the 2008 Wenchuan earthquake, 2010 Haiti
earthquake, and 2010 Chile earthquake have inflicted severe damage and led to extensive
1
Department of Civil Engineering, Tsinghua University, Beijing, China
Pacific Earthquake Engineering Research Center, University of California, Berkeley, CA
3
School of Civil and Environmental Engineering, University of Science and Technology Beijing, Beijing, China
2
Corresponding author:
Xinzheng Lu, M.EERI, Department of Civil Engineering, Tsinghua University, Beijing 100084, China.
Email: luxz@tsinghua.edu.cn
2
Earthquake Spectra 00(0)
casualties and significant losses (DesRoches et al., 2011; Lorito et al., 2011; Ye et al.,
2010). Seismic damage simulation of buildings on a regional scale provides valuable information that can be used to improve emergency preparedness and to mitigate possible
earthquake-induced losses in high seismic regions and populous modern cities by identifying current weaknesses.
Existing studies have applied seismic damage simulation to many regions using various
methods: Roca et al. (2006) conducted a seismic damage simulation for Catalonia, Spain,
based on the damage probability matrix (DPM) method (Applied Technology Council
(ATC), 1985); Lantada et al. (2009) performed a deterministic and probabilistic seismic
damage simulation for Barcelona with the DPM method; and Silva et al. (2015) performed
a seismic risk assessment for mainland Portugal using the DPM method. The capacity
spectrum method (Federal Emergency Management Agency (FEMA), 2012a) has been
used to simulate seismic damage for buildings in many regions in the United States,
including California, Los Angeles County, and Charleston (Chen et al., 2013; Kircher
et al., 2006; Schmidtlein et al., 2011). Using integrated earthquake simulation (Hori and
Ichimura, 2008), the seismic responses of buildings in the Zeytinburnu district of Istanbul
and center of Tokyo in Japan were simulated (Hori et al., 2018; Ichimura et al., 2015;
Sahin et al., 2016).
Overall, the seismic damage simulation methods used in these studies can be classified
as either empirical methods or analytical methods (Lang et al., 2012). The widely used
DPM method (ATC, 1985) is a typical empirical method. Using data for the damage to
buildings from historical earthquakes and expert judgment, the probability of various
buildings reaching various damage states under different seismic intensities can be
obtained. This method is easy to implement, but it is difficult to apply in areas where historical earthquake data are lacking or there are large differences between the inventory of
current and historical buildings. In addition, the characteristics of different buildings and
ground motions are difficult to account for using the DPM method. The analytical methods include static analysis and time history analysis (THA) methods, through which the
response of a building is calculated by establishing a structural mechanics model. The
capacity spectrum method used in the loss estimation software HAZUS (FEMA, 2012a) is
a typical static analysis method. The overall strength and ductility of different buildings
can be represented well using this method. However, the influence of higher-order vibrational modes of buildings and the time-domain properties of ground motions (e.g. velocity
pulses) cannot be easily represented. In contrast, the THA method can overcome the limitations of static analysis. Hori and Ichimura (2008) and Lu and Guan (2017) applied
THA to urban buildings in Japan and China, respectively. In particular, Lu and Guan
(2017) proposed a detailed procedure to determine the structural parameters of different
buildings according to building inventory data, which significantly improved the feasibility
of this technique for buildings in various regions. In addition, the next generation of
performance-based earthquake engineering (PBEE) methods, that is, the FEMA P-58
method, was used for loss estimation by applying the engineering demand parameters
(EDPs) obtained from the nonlinear THA of buildings (Lu and Guan, 2017; Zeng et al.,
2016). The city-scale nonlinear time history analysis method (Lu and Guan, 2017) includes
ground motion input, establishment of the building model, calculation of the seismic
response, and assessment of economic loss. Different modules are separated from each
other. These modules involve many technical and implementation details that cannot be
clearly described in published papers or books. The workflow also requires intense manual
labor to start the applications for the individual buildings, and to process the output data
Lu et al.
3
Figure 1. Flowchart and modules of the SimCenter Workflow.
from one program to create the input needed by the next application. These omissions
and amount of labor intensity limit the applicability of such simulations, essentially reducing them to academic exercises performed by a few researchers scattered around the world.
In addition, some methods for ground motion simulation and building inventory development have emerged recently, but there is currently no open-source platform to integrate
these methods together to perform seismic damage simulation on a regional scale.
Therefore, we proposed an open-source general-purpose scientific workflow (referred to
as SimCenter Workflow) for the seismic damage simulation and loss prediction of buildings on a regional scale using the city-scale nonlinear time history analysis method. The
Workflow integrates different modules together and specifies the interface between the different parts, through which the simulation of the entire process—from the earthquake fault
to building loss—can be achieved. Consequently, users can quickly familiarize themselves
with the entire analysis process, apply to seismic damage simulations in other regions, and
modify the modules according to their needs. In addition, applications in the SimCenter
Workflow can be replaced with other alternative applications, which indicates its flexibility
and robustness. To introduce the SimCenter Workflow process in detail and demonstrate
its advantages, we present a scenario-based example run of this workflow application
which performs a seismic damage simulation and loss prediction for 1.8 million buildings
in the San Francisco Bay Area.
SimCenter scientific workflow application
A workflow is a repeatable set of activities that are performed in a sequence to complete a
task that often needs to be repeated. A scientific workflow is the automation of a workflow
such that each activity in the workflow is completed by separate applications and the flow
of outputs and inputs between the applications is managed. CyberShake (Graves et al.,
2011), for example, is such a workflow application used by the Southern California
Earthquake Center (SCEC) for creating seismic hazard maps. In this work, a scientific
workflow is presented that starts with an earthquake simulation for an urban region, and
results in the downtime and loss estimation for the buildings in that region. The applications involved in the workflow include the applications for the nonlinear THA analysis of
each building in the region and the subsequent repair time and loss calculation using the
FEMA P-58 method. The flowchart and modules of the SimCenter Workflow Application
are shown in Figure 1. It should be noted that any application in the SimCenter Workflow
can be replaced with other alternative applications. The SimCenter Workflow includes six
main applications. The function of each module is summarized in Table 1. Each module is
programmed in C ++ , and the relevant code has been open sourced to the GitHub code
4
Earthquake Spectra 00(0)
Table 1. Functions of the different SimCenter Workflow modules
Module name
Function
Building Model
Obtain the building information model file (BIM.json file) from the
building database, for example, the geographic information system
(GIS) database provided by UrbanSim (2018)
Obtain the input ground motion file (Event.json file) from the ground
motion database, for example, the M 7.0 Hayward Fault earthquake
scenario simulated by Rodgers et al. (2018)
Create a structural analysis model (SAM) based on the BIM.json
using the city-scale nonlinear time history analysis (Lu and Guan,
2017)
Identify the type of EDP for the building
Calculate the EDPs by performing a nonlinear THA based on the
structural analysis model and input ground motion
Determine the downtime and loss based on the obtained EDPs
Create Event
Create SAM
Create EDP
Perform Simulation
Create DL
EDP: engineering demand parameter; THA: time history analysis.
hosting platform (https://github.com/fmckenna/Workflowf1). This link includes key software libraries and technical details to run the whole workflow. The SimCenter Workflow
is a cross-platform application, which can be run on multiple operating systems (e.g.
Windows and Linux). Currently, the building inventory and ground motion input modules
are primarily targeted in the Bay Area case study. However, different building inventories,
ground motion simulation methods (e.g. OpenQuake engines (Pagani et al., 2014) and
SCEC Broadband Platform (Silva et al., 2018)), building modeling approaches (e.g. multiple-degree-of-freedom (MDOF) models, refined finite-element models (Xu et al., 2016), or
multiple level-of-detail (LOD) models (Xiong et al., 2019)), and loss modeling approaches
can be integrated into the framework. The SimCenter Workflow can serve as a convenient
open-source platform for seismic damage simulation of buildings in large urban areas.
Two of the main SimCenter Workflow modules, that is, Create SAM and Create DL
(shown in red in Figure 1), are introduced in detail in the following sections.
Create SAM
As mentioned previously, the THA method enables more rational consideration of the
characteristics of different buildings and ground motions than the empirical and static
analysis methods. However, for an urban area consisting of thousands of buildings, creating refined computational models is extremely time-consuming. To overcome this difficulty, the MDOF shear model (Figure 2) proposed by Lu and Guan (2017) is used in this
study. MDOF shear models are suitable for regional seismic response analysis and provide
good accuracy and high computational efficiency (Lu et al., 2014; Xu et al., 2016). The
MDOF shear model assumes that
1.
2.
3.
A multi-story building can be simplified to an MDOF model in which each story
has its nonlinearity;
The deformation mode of a building is dominated by the shear type; and
Depending on the structural system, the nonlinearity of the structure at each story
can be modeled using either a modified-Clough (Mahin and Lin, 1983), bilinear
(Miranda et al., 2002), or pinching model (Kinali and Ellingwood, 2007).
Lu et al.
5
Figure 2. MDOF shear model for buildings on a regional scale.
Based on the HAZUS database, Lu and Guan (2017) proposed a modeling approach
by which all the parameters of the MDOF shear model can be determined from the basic
building information (i.e. number of stories, height, year built, structural type, floor area,
and occupancy). The details are described below.
The first vibration period of the building is determined using Equation 1
T1 =
N
T0 ,
N0
ð1Þ
where T1 is the first vibration period of the specific building; N is the number of stories in
the building; and N0 and T0 are the number of stories and fundamental period, respectively, of the typical buildings presented in Tables 5.5 and 5.7 of FEMA (2013).
For regular buildings, the initial inter-story stiffness and mass of each story are assumed
to be uniformly distributed along the height of the building; the stiffness and mass matrices
of the MDOF model are then given by
2
2
6 1
6
6
½K = k0 6
6
4
2
6
6
½M = m6
6
4
1
2 1
.
1 . .
..
.
3
..
7
7
7
7 = k0 ½A,
7
1 5
.
2
1
ð2Þ
1
3
1
7
7
7 = m½I:
7
5
1
1
..
.
1
Therefore, the initial inter-story shear stiffness k0 is given as
ð3Þ
6
Earthquake Spectra 00(0)
Figure 3. Backbone curve of the MDOF shear model.
k0 = l
l=
4p2 m
,
T1 2
½F1 T ½I½F1 ½F1 T ½A½F1 ð4Þ
!
,
ð5Þ
where [F1] is the modal vector of the first vibration mode; and m is the mass of each story,
which is determined based on the floor area and the occupancy of the building (Lu and
Guan, 2017).
Thus, the inter-story backbone curve parameters of story i in Figure 3 are determined as
follows:
k0, i = l 4p2 m
,
T 12
Vy, i = SAy a1 m g N Gi ,
hi =
ð6Þ
ð7Þ
SAu SAy SDy
,
SDu SDy SAy
ð8Þ
SAu
,
SAy
ð9Þ
bi =
Dc, i = dCo h,
ð10Þ
where g is the acceleration of gravity; dCo is the inter-story drift ratio at the threshold of
the complete damage state, as suggested by HAZUS (FEMA, 2012a); h is the story height;
(SDy, SAy) and (SDu, SAu) are the yield capacity point and ultimate capacity point, respectively, of the capacity curve suggested by HAZUS (FEMA, 2012a), which are functions of
the design intensity and year built; a1 is the mode factor suggested by HAZUS (FEMA,
Lu et al.
7
Table 2. Parameters for Rayleigh damping
Structural type
z = z1 = z2
(FEMA, 2012a)
z = z1 = z2
(Newmark and Hall, 1982)
Steel structure (i.e. S1, S3)
Concrete structure (i.e. C1, C2, C3)
Masonry structure (i.e. RM2, URM)
Wood structure (i.e. W1)
0.05
0.07
0.10
0.10
0.02
0.02
0.07
0.05
FEMA: Federal Emergency Management Agency.
2012a); and Gi is the ratio between the inter-story shear strength of the ith story (Vy, i ) and
that of the ground story (Vy, 1 ), which is calculated as follows:
Gi =
Vy, i
:
Vy, 1
ð11Þ
The damage states for regular buildings are identical to the damage states defined by
HAZUS: slight, moderate, extensive, and complete. The inter-story drift ratio is used as
the threshold for each structural damage state, and the values for different structural types
are based on Table 5.9 of HAZUS (FEMA, 2013). An example of these values is presented
in Table A1 in Appendix A.
The empirical values of the damping ratio for Rayleigh damping (Chopra, 1995) are
estimated according to the structural type of the building. FEMA (2012a) and Newmark
and Hall (1982) have proposed the damping ratios for different types of buildings, as summarized in Table 2. It should be noted that because of the open-source nature of the
SimCenter Workflow, other advanced damping models (e.g. ATC-72 (ATC, 2010)) can
also be adopted.
To represent the pinching behavior subjected to cyclic loads, the commonly used pinching model proposed by Steelman and Hajjar (2009) is adopted in this study. In this model,
the unloading stiffness is constant and equal to the initial stiffness. All possible break
points in the reloading curve are located on the ‘‘break point line,’’ which is drawn from
the intersection of the unloading path with the horizontal axis to the yield point of the full
bilinear hysteretic loop in the loading, as shown in Figure 4. Thus, only one parameter, t,
is required to determine the hysteretic behavior of the model, as follows:
t=
Ap
,
Ab
ð12Þ
where Ap and Ab are the areas of the pinching envelope and full bilinear envelope, respectively (Figure 4); and t is a coefficient used to quantify the severity of degradation. This
model was chosen because the pinching parameter, t, can be easily determined from the
degradation factor, k, given in Table 5.18 of HAZUS (FEMA, 2013; Steelman and Hajjar,
2009). If other pinching models (e.g. that proposed by Ibarra et al. (2005)) are used, more
parameters (e.g., basic strength deterioration, post-capping strength deterioration, and
accelerated reloading stiffness deterioration (Ibarra et al., 2005)) should be calibrated.
However, the required calibration of the parameters for a large number of buildings has
not been found in literature to-date. Consequently, the single parameter-pinching model is
8
Earthquake Spectra 00(0)
Figure 4. Diagram of the pinching envelope of the pinching model.
adopted herein. It should be noted that this approach is suitable for conventional buildings with regular planar and vertical layouts. And the uncertainty of the parameters of the
structural models can be considered when the data are available (Lu et al., 2017).
Create DL
A regional scale seismic loss prediction method for buildings (Zeng et al., 2016) based on
the FEMA P-58 method (FEMA, 2012b, 2012c, 2012d) is adopted in the Create DL module of the SimCenter Workflow to calculate the loss of buildings.
FEMA P-58 is a component-level loss assessment method, in which the seismic loss is
calculated by adding up the repair cost of each building component under different damage states. The fundamental principle of the FEMA-P58 methodology is the PBEE framework developed by the Pacific Earthquake Engineering Research Center (Cornell and
Krawinkler, 2000). The procedure relies on the use of conditional probability, where the
probability of incurring damage at different damage states is conditioned based on the
occurrence of a given EDP. Once the damage state is identified, the repair action and
repair cost are calculated from a look-up table.
Figure 5 shows the flowchart of the repair cost calculation for a performance group
(PG), which consists of the components that are vulnerable to the same EDP (Zeng et al.,
2016). The EDP is first calculated using the THA. The damage state of the PG is then
determined using the fragility curves. Finally, the repair cost is determined from the total
quantity and unit repair cost of each PG. Details of this procedure are described in
Chapters 3 and 7 of FEMA (2012b).
The key challenges of using the FEMA-P58 method for a region are the assembly of the
performance models and the rapid calculation of the EDPs.
Assembly of the performance models. The performance model of a building contains the basic
information for both the structural and nonstructural PGs.
Lu et al.
9
Figure 5. Calculation of repair cost for a building component using the FEMA-P58 methodology (Zeng
et al., 2016).
(a)
(b)
The basic building data can be obtained from a GIS database, field investigation,
or design drawings. Generally, this information is easy to collect if a GIS database
of the target area is available.
The structural PGs generally obey a certain statistical law owing to the requirements of seismic design codes and economic considerations of structural design.
Through a statistical survey, the statistical law for the number of structural components per unit area can be obtained. Then, based on GIS data such as the structural types, floor area, and number of stories, the number of structural PGs can
be estimated. FEMA (2012b, 2012d) assumes that the number of nonstructural
PGs follows a lognormal distribution, and thus the number of structural PGs is
also assumed to obey a lognormal distribution in this study.
By collecting relevant data from published documents and design drawings (i.e. structural layouts from tests, numerical models, or actual engineering projects), the median and
dispersion of the structural PGs of buildings with different structural types were determined, as listed in Table 3. It should be noted that these buildings are designed by researchers and cannot reflect all of the characteristics of a real building. Hence, the accuracy of
the estimation could be improved by enriching the statistical data and incorporating more
actual engineering projects in the future.
The nonstructural PGs can be identified using the FEMA P-58 database. FEMA
(2012b, 2012c, 2012d) investigated the number of nonstructural PGs in more than 3000
buildings with different occupancies (e.g. office, education, healthcare, hospitality, etc.),
and suggested typical values, which are listed in the normative quantity information provided in Appendix F of FEMA (2012b). Meanwhile, a tool to estimate the number of nonstructural PGs is provided in FEMA P-58. Therefore, based on those data and basic GIS
information such as floor area, number of stories, and occupancy, the nonstructural PGs
can be estimated. For example, the sanitary waste piping in an office building is assumed
to be 0.057 ft per 1 gross ft2 according to Appendix F of FEMA (2012b), so the length of
2.2 (0.3)b
3.4 (0.4)
1.5 (0.5)
Steel frame
Reinforced concrete
(RC) frame
RC shear wall
b
a
0.6 (0.3)
Number of
coupling
beams
per 100 m2
10.6% (0.4)
3% (0.7)
Shear wall
densitya
The shear wall density is defined as the ratio between the wall section area and the floor plan area.
The value in the parentheses is the dispersion.
Wood
Masonry
Number of
columns
per 100 m2
Structural type
Table 3. Median and dispersion of the structural PGs
50.1 (0.2)
Length of wall
per 100 m2
Del Carpio Ramos et al. (2015);
Hwang and Lignos (2017); Jones
and Zareian (2010); Khalil (2011);
Lignos et al. (2011); National
Institute of Standards and
Technology (NIST) (2018); Perrone
and Filiatrault (2017); Wang et al.
(2017); Yang et al. (2009)
Arroyo et al. (2017); Burton and
Deierlein (2014); CESMD (2018a);
CESMD (2018b); Haselton et al.
(2008); NIST (2018); Padgett and Li
(2014); Rahman et al. (2018);
Shokrabadi et al. (2015)
Aval and Asayesh (2017); Ge and
Zhou (2018); Hu et al. (2015); Lu et
al. (2015); Nazari and Saatcioglu
(2017)
Gülkan et al. (1990); Klingner et al.
(2012); Kollerathu and Menon
(2017); Park et al. (2009); Seible et
al. (1994)
Christovasilis et al. (2009);
Ellingwood et al. (2008); Kasal et al.
(1994); Kasal et al. (2004); Pang et
al. (2010); Pang et al. (2012)
References
10
Earthquake Spectra 00(0)
Lu et al.
11
sanitary waste piping in an office building with a floor area of 1000 m2 is 187 m in each
floor. For other specialized buildings, the nonstructural PGs can be estimated according
to Appendix F of FEMA (2012b).
Rapid calculation of the EDPs of a building group. Another difficulty in extending the FEMAP58 method from an individual building to a region is to obtain the EDP distribution in
an efficient manner. EDPs including the peak floor acceleration (PFA), peak floor velocity
(PFV), peak inter-story drift ratio (IDR), and residual drift at each story are determined
from a series of THA. The EDP distribution of a building group can be obtained efficiently using nonlinear THA of the MDOF shear models, making use of the results of the
SimCenter Workflow.
In addition to economic loss, the FEMA P-58 method can also be used to predict other
seismic consequences, such as the repair time and red-tag status of buildings. In the
Workflow, the rebuilding time is determined based on Table 15.9 of FEMA (2013) and
the parallel repair strategy is adopted.
Seismic damage simulation of buildings in the San Francisco bay area
In this study, the San Francisco Bay Area (hereafter referred to as the Bay Area) is considered for a case study simulating the seismic damage of buildings using the SimCenter
Workflow. And the case study area consists of nine counties (Alameda, Contra Costa,
Marin, Napa, San Francisco, San Mateo, Santa Clara, Solano, and Sonoma), as shown in
Figure 6.
The seismic risk of the Bay Area is significantly high. In 2003, earthquake scientists led
by USGS estimated that there is a 62% probability of a magnitude 6.7 or greater earthquake in the Bay Area in the next 30 years (USGS, 2019), as shown in Figure 6. Therefore,
it is necessary to perform the seismic damage simulation of the Bay Area under a significant earthquake.
Building inventory
An inventory of 1,843,351 buildings in the Bay Area was obtained from the GIS database
provided by UrbanSim (2018), including the number of stories, year built, floor area, occupancy, and so on. UrbanSim is a simulation system for supporting planning and analysis
of urban development, through which information for individual buildings and parcels
that represent location can be provided (UrbanSim, 2018). It should be noted that the
building inventory can be provided by other data source. Since UrbanSim has established
the building inventory for the Bay Area, these data were adopted in this work. The distribution of the year built and number of stories are shown in Figures 7 and 8, which show
that the majority of buildings in the Bay Area are low-rise buildings built before 2000.
Input ground motion
An accurate ground motion input not only reduces bias in the obtained structural
responses, but it also contributes to improving the reliability of the estimation of seismic
responses (Macedo and Castro, 2017). Rodgers et al. (2018) at Lawrence Livermore
National Laboratory (LLNL) and Lawrence Berkeley National Laboratory (LBNL) performed a simulation of a magnitude of 7.0 earthquake event on the Hayward fault (Figure
12
Earthquake Spectra 00(0)
Figure 6. Study area and earthquake probabilities for the Bay Area (U.S. Geological Survey (USGS),
2019).
Figure 7. Distribution of the year built for more than 1.8 million buildings in the Bay Area.
Lu et al.
13
Figure 8. Distribution of number of building stories of more than 1.8 million buildings in the Bay Area.
6) in the Bay Area using the National Energy Research Scientific Computing (NERSC)
Center Cori supercomputer. The propagation of seismic waves and the ground motions
at each grid point within the 120 3 80 3 30 km region were obtained in the simulation
(Figure 9). The input ground motion for each building can be determined from the
simulated ground motions based on the building location. The peak ground acceleration (PGA) distribution of the input ground motion for each building is shown in
Figure 10.
Analysis of typical buildings
To describe the entire SimCenter Workflow analysis process in detail, seismic damage
simulations of two typical buildings (Table 4) in the Bay Area are performed. For each
building, the SimCenter Workflow executes the following steps:
1.
2.
Given the building tag, the Building Model module produces a building information model (BIM) file (BIM.json) from the UrbanSim (2018) GIS database. The
BIM.json files for the two typical buildings are shown in Figure A1.
Given the location of the building provided in the BIM.json file, the Create Event
module produces an Event.json file. This file contains the input ground motion for
that building obtained from the database of simulated ground motions. The time
histories of the input ground motions for the two typical buildings (grid points
S_28_28 and S_30_35) are shown in Figure 9 as examples.
14
Earthquake Spectra 00(0)
Figure 9. (a) Simulated ground motion for the M 7.0 Hayward Fault earthquake scenario; (b) time
histories of the input ground motion at grid S_28_28; (c) ground motion at grid S_30_35 (Rodgers et al.,
2018).
Lu et al.
15
Figure 10. Distribution of input ground motion PGA for each building.
Table 4. Description of typical buildings
Name
Structural
type
Occupancy
Floor
area (m2)
Number
of stories
Year built
Design
code
Replacement
value (106 $)
C1M
RC moment
frame
Wood
Office
1530
5
1925
Precode
20.368
Dwelling
170
1
1970
Moderate
code
0.272
W1
3.
4.
5.
Given the building information and ground motion (from the BIM.json and
Event.json files, respectively), a structural analysis model (SAM) file is then
created by the Create SAM module. The capacity curves for the two typical buildings predicted by the SimCenter Workflow and the HAZUS method are shown in
Figure 11, and are very similar.
Given the BIM.json, Event.json, and SAM.json files, the Create EDP module identifies the types of EDPs required for subsequent damage and loss calculations of the
building.
A nonlinear THA of the SAM.json and Event.json files is then performed to calculate the EDPs using pre- and post-processors and the OpenSees (McKenna, 2011)
application. The seismic responses of the two typical buildings are shown in Figures
16
Earthquake Spectra 00(0)
6.
12 and 13, respectively, from which the EDPs of the buildings can be obtained.
These outputs provide the foundation for the seismic loss assessment. The roof displacements of the two typical buildings predicted by the SimCenter Workflow and
the HAZUS method are compared in Figure 14. It should be noted that although
the capacity curves in Figure 11 are very similar to each other, there are some differences on the roof displacements predicted by the SimCenter Workflow and the
HAZUS method because they use different analysis methods (i.e. static analysis vs.
time history analysis). The nonlinear THA of the SimCenter Workflow can fully
represent the features of different ground motions and buildings (Lu et al., 2017).
Meanwhile, only the intersection point between the demand spectra and the capacity curve can be obtained using the HAZUS method, whereas the entire seismic
response (e.g. acceleration, velocity, and displacement) of each story can be
obtained using the proposed workflow, which is significant for loss estimation.
Finally, given the BIM.json and EDPs, the Create DL module provides damage
and loss estimates. The seismic losses of the two typical buildings predicted by the
SimCenter Workflow and the HAZUS method are shown in Figure 15. Similar to
the discussion in Figure 14, generally speaking, the losses predicted by the
SimCenter Workflow and the HAZUS method are similar to each other. There are
still some differences because they use different prediction methods. Especially for
the C1M structure, the loss ratios calculated by the SimCenter Workflow are
higher than those determined by the HAZUS method. The reason is that the
HAZUS method underestimates the seismic loss of the nonstructural accelerationsensitive (NSA) component for this structure. Specifically, the NSA loss ratio in
the HAZUS method is only 0.17%. However, according to the FEMA P-58 component database, the repair costs for NSA (e.g. elevators and chillers) will result in
a significantly larger economic loss. For instance, there are approximately three
chillers (D3031.012b) in the C1M building according to the Estimation Tool
(FEMA, 2012b). Chillers are fragile and expensive (Figure A2), resulting in a loss
ratio of up to 1.02%. However, such nonstructural component loss is obviously
underestimated in the HAZUS method. Note that FEMA P-58 method can provide a more accurate prediction if the information of the structural and nonstructural components in the buildings are clearly specified.
This section describes each step of the SimCenter Workflow in detail, through which
simulation of the entire process from earthquake fault to building loss can be achieved.
Note that the entire SimCenter Workflow procedure is open-source for the convenience of
future scientific research.
Distribution of building losses
Using the building information and input ground motion, the distribution of the median
building loss ratios and repair/rebuilding times were computed with the SimCenter
Workflow for all buildings. The results are as shown in Figures 16 and 17, respectively.
Such distributions can provide emergency responders information that can be used in
planning disaster scenarios and provide city planners an understanding of disaster risk,
and allow them to explore possible prevention and mitigation strategies that might be
effective in earthquake-prone urban areas. Overall, the distribution pattern of the loss
ratio and repair/rebuilding times are similar to the distribution pattern of the input ground
Lu et al.
17
Figure 11. Comparison of the capacity curves: (a) C1M; (b) W1.
motion PGA for each building. Buildings with a loss ratio greater than 60% accounted
for 7.5% of the total buildings, of which 62.6% were single-story wood structure built
before 1973. Therefore, single-story wood structures built before 1973 are potential weak
buildings in the Bay Area.
18
Earthquake Spectra 00(0)
Figure 12. Seismic responses of C1M: (a) top displacement versus time history; (b) peak inter-story
drift ratio
Lu et al.
19
Figure 13. Seismic responses of W1: (a) top displacement versus time history; (b) lateral hysteretic
behavior in the bottom story.
Detweiler and Wein (2018) performed a similar simulation based on the Hayward
Scenario using the HAZUS method. The distribution of the building damage ratio given
by Detweiler and Wein (2018) is similar to the results calculated by the SimCenter
20
Earthquake Spectra 00(0)
Figure 14. Comparison of roof displacements.
Figure 15. Comparison of seismic losses.
Str: structural components, NSD: nonstructural drift-sensitive components, NSA: nonstructural acceleration-sensitive
components.
Workflow, namely, the region near the HayWired fault is the area with high building damage ratio. It is important to point out that the results of the SimCenter Workflow are more
fine-grained as they are analyzed on the building level, whereas the results of the HAZUS
analysis in the Hayward Scenario are aggregated over zip code areas. In addition, the seismic loss of the component in the building can be given by the SimCenter Workflow
because it incorporates the FEMA-P58 method.
It should be noted that the computational expense of the SimCenter Workflow is
acceptable. And the simulation of the Bay Area case study can be finished within three
days using three desktop computers (with Inter(R) Xeon(R) CPU E5-2630 v3
@2.40 GHz).
Lu et al.
21
Figure 16. Distribution of median building loss ratios in the Bay Area.
To further demonstrate the advantages of the proposed technique, the 3D dynamic
visualization method proposed by Xiong et al. (2015) is used in this study. The buildings
in downtown San Francisco are selected as a demonstration, and the results are shown in
Figure 18, its animation is available in the electronic supplement (Animation 1). Different
color contours in Figure 18 represent different displacement magnitudes, and the seismic
response of each story in every building is visualized realistically. This type of information
can provide an understanding of disaster risk to planners and policymakers to explore
possible building mitigation and resilience strategies that might be effective in earthquakeprone urban areas.
Conclusion
An open-source general-purpose scientific workflow application for seismic damage simulation and loss prediction, SimCenter Workflow, is presented in this study using the cityscale nonlinear time history analysis. To introduce the SimCenter Workflow process in
detail and illustrate its advantages, a seismic damage simulation and loss prediction for
1.8 million buildings in the San Francisco Bay Area were performed using this application.
The main conclusions are as follows:
1.
2.
The SimCenter Workflow can serve as a convenient open-source platform for seismic damage simulation of buildings in large urban areas.
Compared with the widely used HAZUS method, the Workflow can provide more
detailed information for seismic loss assessment owing to the usage of actual
22
Earthquake Spectra 00(0)
Figure 17. Distribution of the median building repair/rebuilding time in the Bay Area
Figure 18. Visualization of the seismic response of buildings in downtown San Francisco (t = 13.2 s).
3.
4.
building inventories, the MDOF models, the nonlinear THA of buildings, and the
FEMA P-58 method.
The case study shows the applicability of the SimCenter Workflow in large-scale
seismic damage simulation and loss prediction.
The open-source nature and the modularization of the SimCenter Workflow facilitate its extensibility, and thus make it practical for researchers to apply the
SimCenter Workflow to seismic damage simulations in other regions.
Lu et al.
23
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/
or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/
or publication of this article: The authors are grateful for the financial support received from the
National Natural Science Foundation of China (No. U1709212). The SimCenter is supported
through a grant provided by the National Science Foundation (#1612843).
Supplemental material
Supplemental material for this article is available online.
References
Applied Technology Council (ATC) (1985) Earthquake Damage Evaluation Data for California
(ATC-13). Redwood City, CA: Applied Technology Council.
Applied Technology Council (ATC) (2010) Modeling and Acceptance Criteria for Seismic Design and
Analysis of Tall Buildings (PEER/ATC-72-1). Redwood City, CA: Applied Technology Council.
Arroyo O, Liel A and Gutiérrez S (2017) A performance-based evaluation of a seismic design
method for reinforced concrete frames. Journal of Earthquake Engineering 22(10): 1900–1917.
Aval SBB and Asayesh MJ (2017) Seismic performance evaluation of asymmetric reinforced concrete
tunnel form buildings. Structures 10: 157–169.
Burton H and Deierlein G (2014) Simulation of seismic collapse in nonductile reinforced concrete
frame buildings with masonry infills. Journal of Structural Engineering 140(8): A4014016.
Center for Engineering Strong Motion Data (CESMD) (2018a) Information for strong-motion
station. Available at: https://strongmotioncenter.org/cgi-bin/CESMD/stationhtml.pl?staID=
CE12284&network=CGS (accessed 10 November 2018).
Center for Engineering Strong Motion Data (CESMD) (2018b) Information for strong-motion
station. Available at: http://www.strongmotioncenter.org/cgi-bin/CESMD/stationhtml.pl?sta
ID=CE12493&network=CGS (accessed 10 November 2018).
Chen R, Branum DM and Wills CJ (2013) Annualized and scenario earthquake loss estimations for
California. Earthquake Spectra 29(4): 1183–1207.
Chopra AK (1995) Dynamics of Structures. Upper Saddle River, NJ: Prentice Hall.
Christovasilis IP, Filiatrault A and Wanitkorkul A (2009) Seismic Testing of a Full-Scale Two-Story
Light-Frame Wood Building: NEESwood Benchmark Test. New York: Multidisciplinary Center
for Earthquake Engineering Research.
Cornell CA and Krawinkler H (2000) Progress and challenges in seismic performance assessment.
PEER Center News 3(2): 1–3.
Del Carpio Ramos M, Mosqueda G and Hashemi MJ (2015) Large-scale hybrid simulation of a
steel moment frame building structure through collapse. Journal of Structural Engineering 142(1):
04015086.
DesRoches R, Comerio M, Eberhard M, et al. (2011) Overview of the 2010 Haiti earthquake.
Earthquake Spectra 27(S1): S1–S21.
Detweiler ST and Wein AM (2018) The HayWired earthquake scenario—Engineering implications.
Scientific Investigations Report 2017-5013-I-Q, p. 429. Available at: https://pubs.er.usgs.gov/
publication/sir20175013v2 (accessed 3 January 2020).
Ellingwood BR, Rosowsky DV and Pang W (2008) Performance of light-frame wood residential
construction subjected to earthquakes in regions of moderate seismicity. Journal of Structural
Engineering 134(8): 1353–1363.
Federal Emergency Management Agency (FEMA) (2012a) Multi-hazard loss estimation
methodology HAZUS–MH 2.1 advanced engineering building module (AEBM) technical and
24
Earthquake Spectra 00(0)
user’s manual. Technical Report. FEMA-P58, Federal Emergency Management Agency,
Washington, DC.
Federal Emergency Management Agency (FEMA) (2012b) Seismic performance assessment of
buildings volume 1: methodology. Technical Report. FEMA-P58, Federal Emergency
Management Agency, Washington, DC.
Federal Emergency Management Agency (FEMA) (2012c) Seismic performance assessment of
buildings volume 2: implementation guide. Technical Report. FEMA-P58, Federal Emergency
Management Agency, Washington, DC.
Federal Emergency Management Agency (FEMA) (2012d) Seismic performance assessment of
buildings, volume 3: Supporting electronic materials and background documentation. Technical
Report. FEMA-P58, Federal Emergency Management Agency, Washington, DC.
Federal Emergency Management Agency (FEMA) (2013) Multi-hazard loss estimation
methodology: Earthquake model, HAZUS—MH 2.1 technical manual. Technical Report.
FEMA-P58, Federal Emergency Management Agency, Washington, DC.
Ge PL and Zhou Y (2018) Investigation of efficiency of vector-valued intensity measures for
displacement-sensitive tall buildings. Soil Dynamics and Earthquake Engineering 107: 417–424.
Graves R, Jordan TH, Callaghan S, et al. (2011) CyberShake: A physics-based seismic hazard model
for southern California. Pure and Applied Geophysics 168(3–4): 367–381.
Gülkan P, Clough RW, Mayes RL, et al. (1990) Seismic testing of single-story masonry houses: Part
1. Journal of Structural Engineering 116(1): 235–256.
Haselton CB, Goulet CA, Mitrani-Reiser J, et al. (2008) An assessment to benchmark the seismic
performance of a code-conforming reinforced-concrete moment-frame building. Technical
Report, Pacific Earthquake Engineering Research Center, Berkeley, CA.
Hori M and Ichimura T (2008) Current state of integrated earthquake simulation for earthquake
hazard and disaster. Journal of Seismology 12(2): 307–321.
Hori M, Ichimura T, Wijerathne L, et al. (2018) Application of high performance computing to
earthquake hazard and disaster estimation in urban area. Frontiers in Built Environment 4: 1.
Hu Y, Zhao ZZ and Qian JR (2015) Comparison of Chinese and American seismic design methods
for frame-core wall structures in high seismic zone. Journal of Building Structures 36(2): 1–9.
Hwang SH and Lignos DG (2017) Earthquake-induced loss assessment of steel frame buildings with
special moment frames designed in highly seismic regions. Earthquake Engineering & Structural
Dynamics 46(13): 2141–2162.
Ibarra LF, Medina RA and Krawinkler H (2005) Hysteretic models that incorporate strength and
stiffness deterioration. Earthquake Engineering & Structural Dynamics 34: 1489–1511.
Ichimura T, Fujita K, Quinay PEB, et al. (2015) Implicit nonlinear wave simulation with 1.08T DOF
and 0.270T unstructured finite elements to enhance comprehensive earthquake simulation. In:
Proceedings of the international conference for high performance computing, networking, storage
and analysis, SC ’15, Austin, TX, 15–20 November. New York: ACM.
Jones P and Zareian F (2010) Relative safety of high-rise and low-rise steel moment-resisting frames
in Los Angeles. The Structural Design of Tall and Special Buildings 19(1–2): 183–196.
Kasal B, Collins MS, Paevere P, et al. (2004) Design models of light frame wood buildings under
lateral loads. Journal of Structural Engineering 130(8): 1263–1271.
Kasal B, Leichti RJ and Itani RY (1994) Nonlinear finite-element model of complete light-frame
wood structures. Journal of Structural Engineering 120(1): 100–119.
Khalil AA (2011) Enhanced modeling of steel structures for progressive collapse analysis using the
applied element method. Journal of Performance of Constructed Facilities 26(6): 766–779.
Kinali K and Ellingwood BR (2007) Seismic fragility assessment of steel frames for consequencebased engineering: A case study for Memphis, TN. Engineering Structures 29(6): 1115–1127.
Kircher CA, Whitman RV and Holmes WT (2006) HAZUS earthquake loss estimation methods.
Natural Hazards Review 7(2): 45–59.
Klingner RE, McGinley WM, Shing PB, et al. (2012) Seismic performance of low-rise wood-framed
and reinforced masonry buildings with clay masonry veneer. Journal of Structural Engineering
139(8): 1326–1339.
Lu et al.
25
Kollerathu JA and Menon A (2017) Role of diaphragm flexibility modelling in seismic analysis of
existing masonry structures. Structures 11: 22–39.
Lang DH, Singh Y and Prasad JSR (2012) Comparing empirical and analytical estimates of
earthquake loss assessment studies for the city of Dehradun, India. Earthquake Spectra 28(2):
595–619.
Lantada N, Pujades LG and Barbat AH (2009) Vulnerability index and capacity spectrum based
methods for urban seismic risk evaluation: A comparison. Natural Hazards 51(3): 501.
Lignos DG, Krawinkler H and Whittaker AS (2011) Prediction and validation of sidesway collapse
of two scale models of a 4-story steel moment frame. Earthquake Engineering & Structural
Dynamics 40(7): 807–825.
Lorito S, Romano F, Atzori S, et al. (2011) Limited overlap between the seismic gap and coseismic
slip of the great 2010 Chile earthquake. Nature Geoscience 4(3): 173–177.
Lu XZ and Guan H (2017) Earthquake Disaster Simulation of Civil Infrastructures: From Tall
Buildings to Urban Areas. New York: Springer.
Lu XZ, Han B, Hori M, et al. (2014) A coarse-grained parallel approach for seismic damage
simulations of urban areas based on refined models and GPU/CPU cooperative computing.
Advances in Engineering Software 70: 90–103.
Lu XZ, Li MK, Guan H, et al. (2015) A comparative case study on seismic design of tall RC framecore-tube structures in China and USA. The Structural Design of Tall and Special Buildings 24(9):
687–702.
Lu XZ, Tian Y, Guan H, et al. (2017) Parametric sensitivity study on regional seismic damage
prediction of reinforced masonry buildings based on time-history analysis. Bulletin of Earthquake
Engineering 15(11): 4791–4820.
Macedo L and Castro JM (2017) SelEQ: An advanced ground motion record selection and scaling
framework. Advances in Engineering Software 114: 32–47.
McKenna F (2011) OpenSees: A framework for earthquake engineering simulation. Computing in
Science & Engineering 13(4): 58–66.
Mahin SA and Lin J (1983) Construction of inelastic response spectra for single-degree-of-freedom
systems, computer program and applications. UCB/EERC-83/17. University of California,
Berkeley, CA.
Miranda E and Ruiz-Garcia J (2002) Influence of stiffness degradation on strength demands of
structures built on soft soil sites. Engineering Structures 24(10): 1271–1281.
National Institute of Standards and Technology (NIST) (2018) Structural robustness prototype
buildings. Available at: https://www.nist.gov/el/materials-and-structural-systems-division-73100/
structural-robustness-prototype-buildings (accessed 11 January 2020).
Nazari YR and Saatcioglu M (2017) Seismic vulnerability assessment of concrete shear wall
buildings through fragility analysis. Journal of Building Engineering 12: 202–209.
Newmark NM and Hall WJ (1982) Earthquake Spectra and Design. Oakland, CA: Earthquake
Engineering Research Institute (EERI).
Padgett JE and Li Y (2014) Risk-based assessment of sustainability and hazard resistance of
structural design. Journal of Performance of Constructed Facilities 30(2): 04014208.
Pagani M, Monelli D, Weatherill G, et al. (2014) OpenQuake engine: An open hazard (and risk)
software for the global earthquake model. Seismological Research Letters 85(3): 692–702.
Pang W, Rosowsky DV, Pei S, et al. (2010) Simplified direct displacement design of six-story
woodframe building and pretest seismic performance assessment. Journal of Structural
Engineering 136(7): 813–825.
Pang W, Ziaei E and Filiatrault A (2012) A 3D model for collapse analysis of soft-story light-frame
wood buildings. In: Proceedings of the world conference on timber engineering, Auckland, 15–19
July.
Park J, Towashiraporn P, Craig JI, et al. (2009) Seismic fragility analysis of low-rise unreinforced
masonry structures. Engineering Structures 31(1): 125–137.
Perrone D and Filiatrault A (2017) Automated seismic design of non-structural elements with
building information modelling. Automation in Construction 84: 166–175.
26
Earthquake Spectra 00(0)
Rahman MM, Jadhav SM and Shahrooz BM (2018) Seismic performance of reinforce concrete
buildings designed according to codes in Bangladesh, India and US. Engineering Structures 160:
111–120.
Roca A, Goula X, Susagna T, et al. (2006) A simplified method for vulnerability assessment of
dwelling buildings and estimation of damage scenarios in Catalonia, Spain. Bulletin of Earthquake
Engineering 4(2): 141–158.
Rodgers AJ, Pitarka A, Petersson NA, et al. (2018) Broadband (0-4 Hz) ground motions for a
magnitude 7.0 Hayward fault earthquake with three-dimensional structure and topography.
Geophysical Research Letters 45(2): 739–747.
Sahin A, Sisman R, Askan A, et al. (2016) Development of integrated earthquake simulation system
for Istanbul. Earth, Planets and Space 68(1): 115.
Schmidtlein MC, Shafer JM, Berry M, et al. (2011) Modeled earthquake losses and social
vulnerability in Charleston, South Carolina. Applied Geography 31(1): 269–281.
Seible F, Priestley MJN, Kingsley GR, et al. (1994) Seismic response of full-scale five-story
reinforced-masonry building. Journal of Structural Engineering 120(3): 925–946.
Shokrabadi M, Banazadeh M, Shokrabadi M, et al. (2015) Assessment of seismic risks in code
conforming reinforced concrete frames. Engineering Structures 98: 14–28.
Silva F, Maechling PJ, Goulet CA, et al. (2018) The SCEC broadband platform: Open-source
software for strong ground motion simulation and validation. In: Poster Presentation at 2018
SCEC annual meeting, San Francisco, CA, 3–7 December.
Silva V, Crowley H, Varum H, et al. (2015) Seismic risk assessment for mainland Portugal. Bulletin
of Earthquake Engineering 13(2): 429–457.
Steelman JS and Hajjar JF (2009) Influence of inelastic seismic response modeling on regional loss
estimation. Engineering Structures 31(12): 2976–2987.
UrbanSim (2018) Home-UrbanSim. Available at: http://www.urbansim.com/home/ (accessed 10
November 2018).
U.S. Geological Survey (USGS) (2019) Earthquake probabilities for the bay area. Available at:
https://earthquake.usgs.gov/earthquakes/events/1906calif/virtualtour/modern.php (accessed 11
May 2019).
Wang S, Lai JW, Schoettler MJ, et al. (2017) Seismic assessment of existing tall buildings: A case
study of a 35-story steel building with pre-Northridge connection. Engineering Structures 141:
624–633.
Xiong C, Lu XZ, Hori M, et al. (2015) Building seismic response and visualization using 3D urban
polygonal modeling. Automation in Construction 55: 25–34.
Xiong C, Lu XZ, Huang J, et al. (2019) Multi-LOD seismic-damage simulation of urban buildings
and case study in Beijing CBD. Bulletin of Earthquake Engineering 17(4): 2037–2057.
Xu Z, Lu XZ and Law KH (2016) A computational framework for regional seismic simulation of
buildings with multiple fidelity models. Advances in Engineering Software 99: 100–110.
Yang TY, Moehle J, Stojadinovic B, et al. (2009) Seismic performance evaluation of facilities:
Methodology and implementation. Journal of Structural Engineering 135(10): 1146–1154.
Ye LP, Lu XZ and Li Y (2010) Design objectives and collapse prevention for building structures in
mega-earthquake. Earthquake Engineering and Engineering Vibration 9(2): 189–199.
Zeng X, Lu XZ, Yang TY, et al. (2016) Application of the FEMA-P58 methodology for regional
earthquake loss prediction. Natural Hazards 83(1): 177–192.