DOI: 10.1111/mice.

12496

ORIGINAL ARTICLE

Framework for city-scale building seismic resilience simulation

and repair scheduling with labor constraints driven
by time–history analysis

Chen Xiong1 Jin Huang1 Xinzheng Lu2

1 Guangdong Provincial Key Laboratory of

Abstract
Durability for Marine Civil Engineering,
Shenzhen University, Shenzhen, China A city-scale time–history analysis-driven framework is proposed for the quantitative
2 KeyLaboratory of Civil Engineering Safety evaluation of building seismic resilience and repair scheduling with repair resource
and Durability of China Education Ministry, constraints. First, a calculation method for the post-earthquake residual functionality
Department of Civil Engineering, Tsinghua
University, Beijing, China of buildings based on engineering demand parameters is proposed. Second, the repair-
scheduling unit (RSU) concept is proposed for city-scale repair scheduling. Moreover,
Correspondence
two repair priority indices are introduced to evaluate the repair priority of each RSU.
Xinzheng Lu, Key Laboratory of Civil Engi-
neering Safety and Durability of China Edu- Next, the concept of job block is proposed to compute the repair time of an RSU
cation Ministry, Department of Civil Engi- with insufficient repair resources. Subsequently, the workflow of repair simulation is
neering, Tsinghua University, Beijing 100084,
presented to calculate the community recovery curve and resilience index quantita-
China.
Email: [email protected] tively. Finally, 68,930 residential buildings of Beijing City’s 16 administrative dis-
tricts are simulated to demonstrate the proposed method. The outcomes of this work
Funding information
are expected to be a useful reference for building seismic resilience evaluation and
Institute of Engineering Mechanics, China
Earthquake Administration, Grant/Award repair scheduling of communities. Consequently, it could be an aid to pre-earthquake
Number: 2017D01; Intellectual Innovation disaster risk reduction planning and post-earthquake rapid recovery of building func-
Program of Shenzhen Science and Technol-
ogy Innovation Committee, Grant/Award
tionalities.
Number: JCYJ20180305123919731; SZU,
Grant/Award Number: 2017064

1 I N T RO D U C T I O N cept of urban building seismic resilience has emerged as one

of the most notable topics in order to cope with the aforemen-
With the development of seismic design methods, modern tioned type of problem (Burton, Deierlein, Lallemant, & Lin,
buildings that are compliant with seismic design codes typi- 2015; Franchin & Cavalieri, 2015; Kang, 2018; Mahmoud &
cally exhibit satisfactory seismic safety; accordingly, the prob- Chulahwat, 2018; Wang et al., 2017; Xu, Ouyang, Mao, & Xu,
ability of building collapse is minimal, and occupant safety 2019), and simulation methods can be useful tools when deal-
is ensured (Lu, Lu, Guan, & Ye, 2013). However, during the ing with community resilience (Batouli & Mostafavi, 2018;
2011 Christchurch earthquake in New Zealand, although the Mostafavi & Ganapati, 2019).
number of collapsed buildings in the central business district The seismic resilience represents the capability of infra-
of this city was limited, the functionalities of the area were structures to avoid severe functional loss when subjected to
severely disrupted because the buildings sustained irrepara- seismic effects and the ability to restore full functionality
ble damage (Kam, Pampanin, & Elwood, 2011; Parker & within a relatively short time (Bruneau et al., 2003). To quan-
Steenkamp, 2012; Wikipedia, 2012). For this reason, the con- tify the seismic resilience of a community, Bruneau et al.

© 2019 Computer-Aided Civil and Infrastructure Engineering

Comput Aided Civ Inf. 2019;1–20. wileyonlinelibrary.com/journal/mice 1

2 XIONG ET AL.

(2003) and Bruneau and Reinhorn (2007) proposed a concep- buildings after an earthquake. Bruneau and Reinhorn (2007)
tual framework. The seismic resilience of a community can adopted the fragility curve of buildings to determine their
be calculated according to Equation (1) as follows: structural and nonstructural losses. Lin and Wang (2016)
𝑡end
classified buildings into different groups and determined
𝑅= [𝑄0 − 𝑄(𝑡)]d𝑡 (1) buildings losses according to the building portfolio fragility
∫ 𝑡0 function of the corresponding building group. In addition to
fragility analysis, the nonlinear time–history analysis (THA)
where R, the seismic loss of community resilience, is defined
can fully consider the characteristics of ground motions
as the functional loss integral from the onset of a disaster (at
and the seismic performance of different buildings (Hori &
time t0 ) to the moment of full recovery (tend ); Q0 is the pre-
Ichimura, 2008; Mahin et al., 2018; Xiong, Lu, Hori, Guan,
event functionality of the community; and Q(t) is the post-
& Xu, 2015; Xiong, Lu, Huang, & Guan, 2019). Moreover,
event functionality function of the community that varies with
the THA can output multiple types of engineering demand
time.
parameters (EDPs) of each story of a structure, for example,
Cimellaro, Reinhorn, and Bruneau (2010a) further devel-
the interstory drift ratio, floor acceleration, and residual inter-
oped the analytical quantification of seismic resilience and
story drift ratio. The availability of these EDPs facilitates
proposed the resilience index as shown in Equation (2). The
the seismic performance assessment of city-scale buildings
resilience is defined graphically as the normalized shaded area
using the method proposed by the Federal Emergency Man-
underneath the functionality function Q(t), which is dimen-
agement Agency (FEMA) P58 report (FEMA, 2012; Zeng,
sionless and ranges from 0 to 1. The resilience index of Cimel-
Lu, Yang, & Xu, 2016). For example, Burton et al. (2015)
laro et al. (2010a) is easy to implement and has been applied in
obtained the EDPs of a typical building and subsequently
the seismic resilience assessment of water/gas supply systems,
evaluated the damage of different building components by
medical facilities, and bridges (Bocchini & Frangopol, 2012;
using the FEMA P58 method (FEMA, 2012), such that
Cimellaro, Reinhorn, & Bruneau, 2010b; Cimellaro, Solari,
the residual functionality of a building was determined.
& Bruneau, 2014; Decò, Bocchini, & Frangopol, 2013):
Nevertheless, the link between component-level damage
𝑡0 +𝑇LC to the overall residual functionality of a building remains
𝑅= 𝑄(𝑡)∕𝑇LC d𝑡 (2) unclear. Furthermore, the problem on how to evaluate the
∫ 𝑡0
residual functionality of an entire community based on
Here, t0 is the time of occurrence of the earthquake; TLC is the the THA and FEMA P58 method (FEMA, 2012) remains
control time of the system, which is defined as the period of unresolved.
interest for infrastructures in which the resilience (R) is con- As for the simulation of community functional recov-
sidered; and Q(t) is the functionality function of the system. ery process after an earthquake, most available methods
The seismic resilience consists of two resilience dimen- primarily focused on the functional restoration of lifeline
sions: (1) the community’s ability to maximize its residual infrastructures (Ballantyne et al., 1990; Chang, Shinozuka,
functionality after an earthquake (referred to as robustness) & Svekla, 1999; Chang & Shinozuka, 2004; Cimellaro et al.,
and (2) the capability of a community to achieve fast recov- 2014; Isumi, Nomura, & Shibuya, 1985; Xu et al., 2019);
ery (known as rapidity) (Miles & Chang, 2006). The residual nevertheless, few determination methods of the temporal
functionality of a community following an earthquake is nor- variation of building residual functionality Q(t) have been
mally measured in terms of certain indicators of community presented (Cimellaro et al., 2010a). Miles and Chang (2006)
utilities (e.g., available housing capacity; Burton et al., 2015). proposed a conceptual model to generate a community
To quantify the residual functionality, the damage states of recovery curve. The method can consider the influences of
each building together with the relationship between dam- various environmental and social factors on the recovery
age states and the loss of building functionalities have to process. Based on the conceptual model of Cimellaro et al.
be obtained. However, enormous building inventories and (2010a), Miles and Chang (2006) proposed linear, expo-
diverse building types in an urban area make it consider- nential, and trigonometric recovery curves. Consequently,
ably difficult to quantify the residual functionality of a com- a proper recovery curve can be selected according to social
munity. The second dimension of resilience (i.e., rapidity) preparedness and resourcefulness. By capitalizing on the
reflects the temporal evolution of functional recovery, which EDPs obtained through the THA, Burton et al. (2015)
is greatly influenced by public policies, available technical or adopted the FEMA P58 damage assessment method (FEMA,
labor resources, and social preparedness. A reasonable simu- 2012) to determine the necessary repair time of building
lation of such a recovery process is also a considerably chal- components and calculate the recovery curve of residential
lenging task. buildings.
In the existing literature, it is reported that fragility analysis The recovery curve is considerably affected by com-
is extensively used to calculate the residual functionality of munity resourcefulness. Bocchini and Frangopol (2012)
XIONG ET AL. 3

simulated the restoration process of bridge networks con- formed for Beijing City, and the effectiveness of different
sidering the repair funding allocated to each bridge; repair repair schemes is discussed.
scheduling is optimized by maximizing the network resilience
and minimizing the repair time and total cost. Luna, Bal-
akrishnan, and Dagli (2011) proposed a model for the 2 METHODOLOGY FRAMEWORK
post-earthquake recovery of water distribution systems.
The labor required for the restoration of pipeline dam- The proposed framework for the city-scale building seismic
age is estimated, and the resource allocation module is resilience simulation based on the THA mainly consists of
employed to prioritize the recovery process. Limited research five parts, as shown in Figure 1.
has been reported with respect to the repair scheme of
city-scale buildings. Zhang and Nicholson (2016) proposed a 1. Hazard analysis
multiobjective optimization model to support building retrofit
planning in a community. In the field of resilience evalua- In the first part, according to the predefined earthquake sce-
tion of individual buildings, the method of Resilience-based nario, the ground motion of a region can be obtained through
Earthquake Design Initiative (REDi) (Almufti & Willford, wave propagation simulation (Graves & Pitarka, 2010; Hori
2013) can predict the repair time and labor demand of each & Ichimura, 2008) or probabilistic hazard seismic analysis
building according to the widely accepted methods in con- (McGuire, 2008) together with artificial ground motion gener-
struction scheduling. This method is extensively used for the ation (Gasparini & Vanmarcke, 1976). The hazard analysis is
seismic resilience assessment of individual buildings (Dong scenario based; nevertheless, earthquake scenarios in differ-
& Frangopol, 2016; Hutt, Almufti, Willford, & Deierlein, ent return periods can be performed to consider the seismic
2015; Lu, Xie, Yu, & Lu, 2016; Tian, Lu, Lu, Li, & Guan, risk of the studied region.
2016). In the REDi method, the repair sequences of individ-
ual buildings and the labor demand of each repair sequence 2. Seismic response analysis
are explicitly presented. Accordingly, the REDi method can
be the basis upon which to develop a repair simulation method In this part, the nonlinear THA models of buildings are first
for city-scale buildings under resource constraints. generated according to the characteristics of different types of
In view of the foregoing discussion, the proposed frame- buildings in the region. For example, the multiple degree-of-
work incorporates nonlinear THA to calculate the story- freedom (MDOF) shear models can be used for regular multi-
level EDPs of each building. The story-level EDPs enable story buildings in urban areas (Hori & Ichimura, 2008; Xiong,
city-scale application of the FEMA P58 seismic damage Lu, Lin, Xu, & Ye, 2017). The MDOF flexural-shear model
assessment method (FEMA, 2012), which has primarily been can be applied (Miranda & Taghavi, 2005; Xiong, Lu, Hong,
used for individual buildings in previous studies. Based on & Xu, 2016) for regular high-rise buildings. For special or
the component-level damage results computed through the super-tall buildings with a complicated seismic performance,
FEMA P58 method, a probabilistic calculation method of a refined finite element model can be adopted (Lu et al., 2013;
building residual functionality is introduced. Capitalizing on Xiong, Lu, Hong, & Xu, 2019). Subsequently, a nonlinear
the component-level repair time obtained through the FEMA THA is performed for each building based on ground motion
P58 method and the building repair time calculation of the records obtained from hazard analysis. Thereafter, the EDPs
REDi method (Almufti & Willford, 2013), a city-scale build- of each building story can be calculated.
ing repair scheduling method is proposed. This work pro-
3. Damage assessment
vides a useful tool for the city-scale repair time and resource
demand estimation, which can inform the decision maker Following the seismic response analysis, damage assess-
about how many interim houses and emergency supplies are ments are performed for structural and nonstructural compo-
needed. The proposed method can assist in pre-event disaster nents according to the EDPs of each story of all buildings. For
reduction planning and post-event repair scheduling, thereby example, the probability of different damage states of struc-
improving the seismic resilience of cities. tural and nonstructural components at different stories of a
Specifically, in Section 2, the framework of the proposed building can be calculated through the seismic damage assess-
methodology is presented. In Section 3, a probabilistic calcu- ment method proposed by FEMA P58 (FEMA, 2012).
lation method of building residual functionality is introduced
based on component-level damage. Subsequently, a commu- 4. Calculation of residual functionality
nity repair scheduling method and a community repair sim-
ulation method that consider the labor resource constraints Based on the damage states of each component, the resid-
are introduced in Sections 4 and 5. Finally, in Section 6, ual functionality of each component is calculated. Thereafter,
a city-scale building seismic resilience simulation was per- by the integration of the residual functionalities of all building
4 XIONG ET AL.

Hazard Seismic Damage Evaluaon of Recovery

analysis response assessment funconality analysis
analysis
Define Construct Obtain fragility Calculate Obtain resource
earthquake building data of funconality of demand curve of
scenario performance performance buildings
models groups groups
Determine repair
Building
Damage priority
nonlinear Calculate
Ground moon assessment of
me–history performance funconality of Repair simulaon
generaon
analysis groups each building of the region

Outputs
Ground moon Engineering Damage states Building Recovery curve
demand of performance residual and resilience
of each building parameters groups funconality index

FIGURE 1 Framework of the methodology

components, the residual functionality of a building is calcu- ton et al., 2015). The socioeconomic functionality describes
lated (Burton et al., 2015). the relationship of a building with its social and economic
environment. For example, during the post-event recovery
5. Recovery analysis process, hospital buildings are more important than ordinary
office buildings; this indicates a higher socioeconomic func-
In this part, the number of workers required to repair each tionality. In order to consider both the physical and socio-
building is estimated based on the component-level struc- economic functionalities of buildings on a regional scale, the
tural and nonstructural damage of each story together with calculation method using Equation (3) is adopted to determine
the floor area and total floor area of a building (Almufti & the functionality of a building:
Willford, 2013). Subsequently, according to the residual func-
tionality and labor demand of each building, the repair scheme 𝑄0 = 𝛼𝑄Physical (3)
is determined. Finally, based on the repair scheme, the recov-
ery simulation of each building in the region is performed where Q0 is the weighted functionality of a building before
to obtain the recovery curve and resilience index, R, of the an earthquake; QPhysical is the index of physical functional-
region. ity (such as the total floor area of a residential building); and
The former three parts of the framework have been exten- 𝛼 is the building weight, which represents its socioeconomic
sively studied in previous research. They have been applied to functionality. The value of 𝛼 can be determined through expert
building seismic damage simulations of Tokyo, Beijing, and judgment.
San Francisco urban areas (Hori & Ichimura, 2008; Mahin
et al., 2018; Xiong et al., 2015; Xiong et al., 2019). More-
3.2 Residual functionality of individual
over, these methods have also been adopted as the framework
buildings
of the NSF-funded NHERI SimCenter (Mahin et al., 2018).
However, limited research has been conducted with respect to After an earthquake, the functionalities of a building may
the discussions in Parts 4 and 5 of the framework. Accord- be reduced or completely lost. According to the simulation
ingly, in the subsequent sections of this paper, the proposed framework presented in Figure 1, the residual functionalities
methodologies of these parts are introduced. of buildings are the prerequisites for the subsequent recovery
analysis. In this study, based on the FEMA P58 method (2012)
and the works of Burton et al. (2015), a probabilistic method
3 CA LCULAT ION OF RESIDUA L is proposed to quantitatively determine the post-earthquake
FUNCTIONAL I T Y residual functionality of buildings in a community. In particu-
lar, building damage is divided into four sequential limit states
3.1 Weighted functionality (Table 1); the residual functionality of each limit state is pre-
sented as follows:
The functionality of urban buildings can be divided into two
categories: physical and socioeconomic functionalities. The 1. LS0 : No damage
physical functionality denotes the utility and the amount of
services a building provides. For example, the housing capac- This state indicates that the structural and nonstruc-
ity is the physical functionality of residential buildings (Bur- tural components of a building remain intact; hence, the
XIONG ET AL. 5

TABLE 1 Description of different building limit states

Proportion of
Limit state Description functional loss
LS0 All structural/nonstructural components remain intact. There is no functional loss 0.0
LS1 Slight structural damage and the building is safe to occupy. Owing to 0.0–1.0
structural/nonstructural damages, the functional loss should be considered
LS2 Moderate or extensive structural damage and the building is unsafe to occupy. 1.0
Building experienced complete functional loss
LS3 The building is severely damaged or collapsed and the building needs to be 1.0
demolished and rebuilt. Building experienced complete functional loss

functionality of the building is not affected. In this state, the Limit state LS1 indicates that the structural PGs of a build-
proportion of building functional loss is 0.0, as shown in ing are not seriously damaged, and the building remains
Equation (4): safe for occupancy. However, the nonstructural PGs, such as
pipelines and elevators, may be damaged, and the functional
𝑃 (𝑄loss |𝐿𝑆0 ) = 0.0 (4) loss due to the damage of these nonstructural components will
be considered in this section.
where P(Qloss | LS0 ) represents the proportion of the func- In order to consider the functional loss caused by the dam-
tional loss of a building when limit state LS0 is attained. It is age of nonstructural PGs, a method similar to the system reli-
noteworthy that the value of P(Qloss | LS0 ) can be larger than ability approach is adopted (Burton et al., 2015). As shown
0.0 in real practice even if none of the structural and nonstruc- in Figure 2, different PGs of a building can be classified
tural components of the building are damaged. For example, into multiple functional groups (FGs) according to their util-
this may occur when buildings sustain the functional loss of ity; different FGs are connected in series. Once the FG fails,
external power or water supply. the overall functionality of the building is completely lost or
To calculate the probability of a building in the LS0 limit severely impaired. For example, in Figure 2, if the functional-
state, the concept of performance group (PG) is adopted ity of the floor access FG is completely lost, the building is no
(FEMA, 2012). The PG is a group of components with simi- longer useable; thus, the functionality of the building is com-
lar seismic fragility and is vulnerable to the same type of EDP. pletely lost. Nevertheless, if the functionality of water supply
Because the probabilities of no-damage state (DS0 ) in differ- FG is lost, and no alternative housing is available, the build-
ent PGs are independent events, the probability of a building ing can still be used with limited functionality. Therefore, the
reaching limit state LS0 is the product of the probabilities of proportion of functional loss, P(Qloss | LS1 ), of a building in
all PGs in the DS0 damage state, as shown in Equation (5): limit state LS1 can be calculated by Equation (6):
𝑙
∏ 𝑃 (𝑄loss |𝐿𝑆1 )
𝑃 (𝐿𝑆0 ) = 𝑃 (𝑃 𝐺𝑖 _𝐷𝑆0 |EDP) (5) ∏𝑚
(6)
𝑖=1 =1− [1 − 𝑃 (𝑄loss |𝐹 𝐿𝑗 )𝑃 (𝐹 𝐿𝑗 |𝐿𝑆1 )]
𝑗=1
where PGi _DS0 is the state of the ith PG with no damage
(DS0 ); P(PGi _DS0 | EDP) is the probability of the ith PG in where m is the number of FGs; P(Qloss | FLj ) is the proportion
the DS0 damage state given the corresponding EDP obtained of building functional loss with the jth FG completely lost.
through the THA in Part 2 of the proposed framework The value of P(Qloss | FLj ) for different FGs can be determined
(Figure 1); and l is the number of PGs. according to expert judgment. For example, if the floor access
FG fails, then the building is inaccessible and the value of
2. LS1 : Occupiable damage with loss of functionality P(Qloss | FLj ) can be set to 1.0; P(FLj | LS1 ) is the probability

FG: Wall Paron FG: Water Supply FG: Floor Access FG: HVAC
PG: PG: Hot
PG: Stair PG: Fans
Paron_X Water Piping
…
PG: PG: Chilled PG: Steam
PG: Elevator
Paron_Y Water Piping Water Piping
…

…
…

FIGURE 2 Functional groups and performance groups of a typical building

6 XIONG ET AL.

of complete functional loss of the jth FG given the building The probability of a building reaching limit state LS1 can
is in the LS1 limit state. The value is related to the damage of be obtained by Equation (10):
each PG within the FG as presented below.
𝑃 (𝐿𝑆1 )
An FG consists of multiple PGs. The functional loss of
𝑛 ∏
∏ 𝑝
a PG reduces the functionality of the FG. The functional-
ity of the FG is completely lost only when all PGs within = 𝑃 (SafePlacarding|𝑃 _𝑃 𝐺𝑘 _𝐷𝑆𝑥 ) − 𝑃 (𝐿𝑆0 )
𝑘=1 𝑥=1
the FG fail. For example, in Figure 2, the functionality of (10)
floor access FG is completely lost when both stairs and ele-
vator PGs fail. Therefore, the probability of the complete where P_PGk _DSx denotes the proportion of the kth PG in the
functional loss of the jth FG P(FLj | LS1 ), given the build- damage state of DSx . P[SafePlacarding | P_PGk _DSx ] indi-
ing is in the LS1 limit state, can be calculated according to cates the probability of the building in safe placarding given
Equation (7): the proportion of the kth PG in the damage state of DSx , and
this probability can be determined according to the FEMA
𝑃 (𝐹 𝐿𝑗 |𝐿𝑆1 )
𝑛 P58 fragility database (FEMA, 2012).
∑ (7)
= 𝑃 (𝐹 𝐿𝑗 |𝑃 𝐿𝑘 )𝑃 (𝑃 𝐿𝑘 |𝐿𝑆1 , EDP)
3. LS2 : Unoccupiable damage
𝑘=1

where n is the number of PGs in the jth FG; P(FLj | PLk ) This limit state indicates that the structural PGs of the
is the proportion of functional loss of the jth FG when the building sustained moderate or extensive damage to the extent
functionality of the kth PG is completely lost. This value rep- that the building cannot be occupied and the functionality of
resents the functional contribution of a PG to the overall func- the building is completely lost. Therefore, the proportion of
tionality of the FG. Therefore, the summation of these values functional loss of the building in LS2 can be calculated accord-
of all PGs in an FG is 1.0. P(PLk | LS1 , EDP) is the proba- ing to Equation (11):
bility of the complete functional loss of the kth PG given the 𝑃 (𝑄loss |𝐿𝑆2 ) = 1.0 (11)
EDP, and the building is in the LS1 limit state, which can be
calculated using Equation (8): Note that the damage of a building is divided into four
sequential limit states LS0 –LS3 . Therefore, the sum of the
𝑃 (𝑃 𝐿𝑘 |𝐿𝑆1 ,EDP) probabilities of a building being in the four limit states is 1.0.
∑𝑝
(8) Consider that the probabilities of LS0 and LS1 are obtained
= 𝑃 (𝑃 𝐿𝑘 |𝑃 𝐺𝑘 _𝐷𝑆𝑥 )𝑃 (𝑃 𝐺𝑘 _𝐷𝑆𝑥 |𝐿𝑆1 , EDP)
𝑥=1 from former discussions. The probability of irreparable dam-
age, LS3 , is relatively easy to obtain and can be determined
where p is the number of damage states of the kth PG; according to the building residual interstory drift ratio in the
P(PLk | PGk _DSx ) is the probability of the complete func- next section. Therefore, the probability of a building reaching
tional loss of the kth PG with the assumption that the cur- limit state LS2 can be calculated using Equation (12):
rent PG reaches the damage state DSx . This value can be
determined through expert judgment. For example, if DS1 of 𝑃 (𝐿𝑆2 ) = 1 − 𝑃 (𝐿𝑆0 ) − 𝑃 (𝐿𝑆1 ) − 𝑃 (𝐿𝑆3 ) (12)
the PG means cosmetic damage, then P(PLk | PGk _DS1 ) can
be set to a relatively small value. On the other hand, if DS2 4. LS3 : Irreparable damage
of the PG indicates complete damage, the P(PLk | PGk _DS2 )
can be set to 1.0; P(PGk _DSx | LS1 , EDP) is the probability This limit state means that the building is severely dam-
of kth PG when it reaches DSx under the circumstance of LS1 aged or collapsed. If a building attains this limit state, then the
and the given EDP. building has to be demolished and rebuilt. Therefore, build-
Because the limit state LS1 of a building is primarily deter- ing functionality is also completely lost, and the proportion
mined according to the damage of structural PGs (FEMA, of functional loss in LS3 can be calculated by Equation (13):
2012), and the building functional loss discussed in this sec-
tion is mainly calculated based on the damage of nonstructural 𝑃 (𝑄loss |𝐿𝑆3 ) = 1.0 (13)
PGs, the occurrences of PGk _DSx and LS1 are independent
The probability, P(LS3 ), of a building reaching irreparable
events. Therefore, P(PGk _DSx | LS1 , EDP) can be calculated
limit state LS3 has been studied extensively (FEMA, 2012;
by Equation (9), and the value of P(PGk _DSx | EDP) can be
Ramirez & Miranda, 2012). It can be calculated based on the
computed according to the component fragility database of
fragility data of different types of buildings according to the
FEMA P58 (FEMA, 2012):
EDP of building residual interstory drift ratio.
𝑃 (𝑃 𝐺𝑘 _𝐷𝑆𝑥 |𝐿𝑆1 , EDP) = 𝑃 (𝑃 𝐺𝑘 _𝐷𝑆𝑥 |EDP) (9) According to the foregoing discussion, the probability
of a building reaching limit state LSy , together with the
XIONG ET AL. 7

functional losses of different limit states, P(Qloss | LSy ), can 3LS HV

+ 9$ &
3DUWLWLR QV
& HOOLQJ
' LVWULE XWLR Q 7LOHV
be obtained. Therefore, based on the total probability theorem,
the expected building residual functionality can be calculated ([WHULRU & ODGGLQJ
3DUWLWLRQV * OD]LQJ
by Equation (14):
0 HFKDQLFDO
[ ] (TXLSP HQ W
∑
4 6WUXFWXUH
𝐸(𝑄res ) = 𝑄0 ⋅ 1 − 𝑃 (𝑄loss |𝐿𝑆𝑦 ) ⋅ 𝑃 (𝐿𝑆𝑦 ) (14) (OHFWULFDO
6\VWHP V
𝑦=1
(OHYDWR UV

where Q0 is the weighted functionality of a building before an

earthquake (Equation (3)). 6WDLUV

FIGURE 3 Repair sequences of a typical building (Almufti &

Willford, 2013)
4 RECOVERY C U RV E A N D L A BOR
D E M A N D C U RV E
60
The recovery curve of a building is considerably affected by
the processes of post-event inspection, engineering mobiliza- 50
tion and review or redesign, financing, contractor mobiliza-

Worker demand
tion, permitting, and building repairs (Almufti & Willford, 40

2013; Burton et al., 2015). Because the former factors are 30

location dependent, and the time required for each process
can be determined according to local situations, this study is 20
focused on the time required for building repairs.
10

4.1 Repair-scheduling unit 0

0 100 200 300 400
It is extremely difficult to determine the repair priority for Days
each building in the city-scale simulation because of the enor-
FIGURE 4 Labor demand curve of a five-story building (10,000
mous building inventory involved. Accordingly, the concept
m2 of gross floor area) after the excitation of an earthquake with a
of repair-scheduling unit (RSU) is proposed for the repair sim- return period of 475 years
ulation. Each RSU contains buildings in the same neighbor-
hood with similar seismic performance. In repair simulation,
the repair scheme is determined as long as the repair priority used to calculate the maximum labor demand. Therefore, the
of each RSU is obtained. labor demand curve of irreparable buildings can be obtained
for the following repair scheduling. It is worth noting that the
4.2 Repair time and labor demand curve labor demand calculation method of REDi does not distin-
of an individual building guish between the structural and nonstructural repair labors.
Nevertheless, in real practice, owing to the technical expertise
According to REDi (Almufti & Willford, 2013), the repair of construction workers, not all construction workers can han-
sequences of a building are shown in Figure 3. Particularly, dle both structural and nonstructural repair work. This effect
the structural repair sequence is prior to nonstructural repair requires further investigation in future studies.
sequences. The nonstructural repair sequences A–F can be
performed simultaneously. The repair time of each PG within
4.3 Recovery curve and labor demand curve
a sequence is calculated according to the seismic damage of
of an RSU
each PG using the FEMA P58 fragility database (FEMA,
2012). Based on the repair sequence scheduling and labor According to the repair-scheduling method in the REDi report
assignment method in the REDi report (Almufti & Willford, (Almufti & Willford, 2013), the repair time of a building
2013), the repair time and labor demand curve, DBlgi (t), of depends on the available workers. In this section, a suffi-
a typical building can be computed as shown in Figure 4. cient number of workers are assumed for the RSU, and all
For buildings in the irreparable limit state (LS3 ), the study buildings in the RSU are simultaneously repaired with the
by Sun and Xu (2011) can be used to estimate the replace- required maximum number of workers. The repair time, TBlgi ,
ment time. Moreover, the maximum workers on site data pro- and the labor demand curve, DBlgi (t), of each building can
vided by the REDi method (Almufti & Willford, 2013) can be be obtained according to the method described in the REDi
8 XIONG ET AL.

Functionality Functionality
Q0, RSU TBlgi Q0
…

Qloss, RSU
Type 2

Qloss, Blgi
TBlg2
TBlg1

Qloss, Blg2
Type 1

QLoss, Blg1
t0 t1 t2 Time
Time (days)
FIGURE 6 Community recovery curves of two types of repair
FIGURE 5 Schematic of RSU recovery curve schemes

report (Almufti & Willford, 2013). It is noteworthy that in

practical applications, some RSUs may not be allocated with ciency of available resources. Different repair schemes result
sufficient numbers of workers; this means that the maximum in different types of community recovery curves. Two types
worker demand cannot be satisfied. Consequently, the recov- of repair schemes are presented in Figure 6. The objective
ery process of the RSU is delayed. Further details on the repair of type 1 repair scheme is to achieve the fastest full recov-
process with insufficient workers are discussed in Section 5.3 ery. For this scheme, more severely damaged RSUs, which
and a case study is presented in Section 6.4. require longer time to repair, are allocated with higher pri-
The recovery curve of a typical RSU is shown in Figure 5. ority. Therefore, the recovery rapidity at the early stage of
After the earthquake, the functional loss of each building in the repair process is slow and gradually accelerates as repair
the RSU can be calculated according to the method presented progresses. The objective of the type 2 repair scheme is to
in Section 3. The total functional loss of the RSU, QLoss, RSU , maximize the resilience index proposed by Cimellaro et al.
can be obtained by aggregating the functional losses of all the (2010a) (Equation (2)). Therefore, slightly damaged build-
buildings in the RSU, as shown in Figure 5. ings, which can easily be restored, are assigned with higher
Consider that the functionality of a single building is rela- priority. In the disaster recovery practice of San Francisco
tively small compared with that of an RSU; accordingly, the (SPUR, 2012), if a community cannot meet the shelter-in-
functional variation of a single building during the repair pro- place standard, the residents may leave because of the long
cess is ignored. Therefore, it is assumed that the function- recovery time. The type 2 recovery scheme enables fast recov-
ality of a building is abruptly restored to its pre-earthquake ery and reduces the demand for interim houses, which helps
functionality when the repair time, TBlgi , of this building is to alleviate the population loss of a community. Therefore,
attained. Based on this assumption, the recovery curve of an the type 2 repair scheme is adopted, and the resilience index,
RSU exhibits a stepped shape, as shown in Figure 5. R (Equation (2)), is used to evaluate the effectiveness of dif-
As for the labor demand curve of an RSU, DRSU (t), ferent repair-scheduling methods.
because all buildings in the RSU are simultaneously repaired
with abundant labor force, according to the aforementioned 5.2 Repair-scheduling methods
assumption, the labor demand curve of an RSU can be
obtained by aggregating the labor demand curve of all the Repair scheduling of regional buildings with limited labor
buildings in the RSU, as shown in Equation (15): forces is a nondeterministic polynomial-time-hard (NP-hard)
problem (Garey & Johnson, 1979). Moreover, the number of
𝑞
∑ buildings in a city is enormous, which makes it extremely
𝐷RSU (𝑡) = 𝐷b lg 𝑖 (𝑡) (15) difficult to find the global maximum of the resilience index,
𝑖=1
R (Equation (2)). Therefore, a simplified repair scheduling
where q is the number of buildings in the RSU. method is presented in this section.
According to the study by Cimellaro et al. (2010a), increas-
ing the recovery rapidity in the early stage of the recovery
5 REPAIR SC H E D U L I NG A N D process can effectively improve the seismic resilience (type 2
SIMULAT ION repair scheme in Figure 6). Therefore, two repair priority
indices, P1 and P2 , are proposed to achieve the type 2 repair
5.1 Community recovery curve scheme, as given by Equations (16) and (17):

In real practice, the repair resource is limited, and not all RSUs
can be assigned with sufficient workers. Hence, it is important 𝑄loss,RSU_𝑖 𝑇min
𝑃1 = ⋅ (16)
to determine the repair scheme so as to improve the repair effi- 𝑄loss,max 𝑇RSU_𝑖
XIONG ET AL. 9

Labor demand labor demand curve describes the maximum labor demand
of an RSU at different repair stages and the duration of each
repair stage. Therefore, the repair procedure of an RSU can be
…
regarded as multiple sequential job blocks, and the workload
Job block 1 Job block 2
Job block i
of each job block is the product of the maximum labor demand
Time
and corresponding duration. For example, if the number of
t1 t2 ti workers available is less than the maximum labor demand of
the first job block, the time required to finish the first job block
FIGURE 7 Labor demand curve and job blocks of a typical RSU can be computed by dividing the total workload of this job
block by the number of available workers. The repair proce-
𝑄loss,RSU_𝑖 𝑊min dures of the sequential job blocks follow the same rule, and
𝑃2 = ⋅ (17) the repair work of the RSU is completed only if all job blocks
𝑄loss,max 𝑊RSU_𝑖
in it are repaired. Based on the presented method, the repair
where Qloss, RSU_i is the post-earthquake functional loss of the time of an RSU with limited workers can be obtained.
ith RSU; TRSU_i is the repair time of the ith RSU given a suf- Based on the job block concept, the city-scale repair simu-
ficient labor force; WRSU_i is the worker day demand, which lation of large numbers of RSUs with limited labor resources
can be computed from the enclosed area of the labor demand can be performed according to the following steps (the
curve; Qloss, max is the maximum RSU functional loss among detailed explanation of the repair simulation workflow is pre-
all RSUs. Tmin and Wmin are the minimum RSU repair time sented in Appendix A).
and worker day demand among all RSUs.
As evident in Equations (16) and (17), the repair priority 1. Determine the time interval of each simulation time step
indices P1 and P2 are dimensionless and range from 0 to 1. (e.g., 1 day and 1 week).
An RSU with high repair priority indices (P1 and P2 ) indicates
2. Allocate available labor resources according to the
that the RSU can achieve large functional recovery within a
descending order of repair priority indices of RSUs in the
given time; this implies a more rapid recovery. Therefore, an
current time step.
RSU with larger repair priority index values is assigned with
a higher repair priority. That is, if the repair priority indices 3. Calculate the remaining workload of each RSU at the end
of all RSUs in the region are calculated, a rapid functional of the current time step.
recovery of regional buildings can be achieved according to 4. Record the restored functionality if an RSU is completely
the descending order of the repair priority index of each RSU. repaired by the end of the current time step and proceeds
Note that the repair priority index P2 , compared to P1 , con- to the second step until accomplishing the repairs of all
siders the labor demand. In cases where the labor force of a buildings.
region is limited, the repair scheme that uses repair priority
index P2 can better utilize the available labor force. For exam-
In addition to the constraint of available labor resources,
ple, in a constrained case, if the P1 value of RSU A is slightly
other recovery constraints, such as the constraints of limited
larger than that of RSU B, while the labor demand of RSU A
recovery funding or construction materials, also significantly
is significantly larger than that of RSU B, when using P1 as
affect the recovery process. However, the influence of other
the priority index, RSU A is prioritized. However, if priori-
recovery constraints can also be considered by following the
tizing RSU B, the spared labor force can be assigned to repair
above methodology.
RSU C, which may result in higher restored functionality. The
detailed comparison of the effectiveness between these two
repair priority indices is discussed in Sections 6.3 and 6.4.
6 CA S E STUDIES
5.3 Repair simulation method Beijing City is considered as an example to demonstrate the
In real practice, the repair resource in an urban area is always proposed methods in this study. First, the nonlinear THA is
limited, and not all RSUs can acquire sufficient resources implemented for residential buildings in the studied area. Sec-
for repairs. For an RSU with insufficient labor resources, the ond, the residual functionality and labor demand curve of all
repair time is longer than that given in Section 4.2. Accord- buildings are calculated through the method proposed in Sec-
ingly, the job block concept is introduced to calculate the tions 3 and 4. Third, the repair scheduling is performed, and
repair time of an RSU with limited labor resources. the effectiveness of the P1 and P2 repair schemes is compared.
The labor demand curve of an RSU according to the The proposed framework is implemented using the C++ pro-
method described in Section 4 is shown in Figure 7. The gramming language and compiled through Visual Studio 2010
10 XIONG ET AL.

RM > 30 F 1 – 3 F
URM 11.0% RC 1.5% 5.0% 1
frame 4–6F
1.4% 56.4%
29.3%

11 – 30 F
Shear wall 54.6% 7 – 10 F
27.3%
5.6%
(a) Distribution of building (b) Distribution of building
structural types stories

FIGURE 8 Structural types and number of stories of Beijing ದ 

City’s residential buildings (URM, RM, RC frame, and shear wall
ದ 
indicate unreinforced masonry, reinforced masonry, RC moment frame,
ದ 
and RC frame-shear wall/core tube stuctures, respectively)
ದ 
NP
ದ 
on a desktop platform with an i7-8700K CPU and 32 Giga-
FIGURE 9 Rental price distribution of Beijing City (CNY/m2
bytes of RAM.
per month)

6.1 Buildings in Beijing City

apartment rental website (Fang, 2018); the rental price distri-
In this case study, 68,930 residential buildings of the 16 bution of each RSU is shown in Figure 9. It is evident in this
administrative districts of Beijing City are simulated. The dis- figure that the rental price is considerably higher in the core
tributions of structural types and number of stories of res- area of Beijing City. For example, the rental price can be more
idential buildings in the Beijing urban area are shown in than 200 CNY/m2 per month in the central area, whereas the
Figure 8. As evident in the figure, a large proportion of rental price in the suburban area of the city can be less than
the studied building inventory is composed of shear wall 50 CNY/m2 per month.
structures and high-rise residential buildings. This is mainly
because the housing data of Beijing City are obtained from
6.2 Earthquake data of Beijing City
the online apartment rental website (Fang, 2018). A large
proportion of the obtained building inventory is late 2000 The terrain of Beijing City is relatively flat, and the climate
high-rise residential buildings because of the accessibility of is dry. Therefore, the problems of landslides, topographical
more comprehensive data of these newly constructed build- effects, and liquefaction are not prominent in this area. More-
ings. Note that the primary purpose of the case study is not over, the microzoning of Beijing City is influenced by basin
to obtain the accurate seismic resilience of Beijing City, but effects (Fu, 2012). In the case study, the response spectra of
to demonstrate the effectiveness of the proposed method. If the entire Beijing City are acquired through probabilistic seis-
more comprehensive data become available in the future, a mic hazard analysis (McGuire, 2008). The response spectral
more accurate simulation can be performed using the meth- attenuation functions (Yu, 2002) and site amplification factors
ods presented in this work. (Lu & Zhao, 2007) of the area are adopted. Subsequently, the
As discussed in Section 3.1, the weighted functionality artificial ground motion generation program, SIMQKE (Gas-
reflects the physical and socioeconomic functionality of a parini & Vanmarcke, 1976), is used to generate the ground
building. In this case study, the number of buildings in the motion data of the region. The response spectral attenuation
area is enormous; accordingly, it is impractical to determine functions adopted in this study are shown in Figure 10a, where
the socioeconomic index, 𝛼, of each building manually. Note the long- and short-axis curves indicate the response spec-
that rental price is a good alternative to represent the socioeco- tral attenuation functions along the direction of the fault and
nomic attributes of residential buildings. A residential build- perpendicular to the direction of the fault; the typical ground
ing with a higher rental price in Beijing is usually located in motion acceleration and velocity time–history data are shown
areas with more economic activities and is closer to the work- in Figure 10b.
place. The rapid recovery of this type of residential buildings On September 2nd 1679 A.D., a major earthquake struck
can facilitate economic recovery and relieve traffic pressure. Beijing City, with the epicenter located approximately 50 km
Therefore, the rental price per square meter per month is used east from the Forbidden City. According to the description of
as the socioeconomic index, 𝛼, for each RSU. The rental price seismic damages in the literature, the magnitude of the earth-
of each RSU in Beijing City is obtained through the online quake was about M8.0. Moreover, according to the study by
XIONG ET AL. 11

30
1
25
Sa_0.3s (m/s2 )
20

15 Short axis
Long axis
10

0 (SLFHQWHU
0 50 100 150 200
Distance (km)
(a) The adopted response spectral attenuation functions ದ 
(M8.0) ದ 
4
3 ದ 
Acceleration (m/s2)

2 ದ 
1
0
NP
ದ 
-1
-2
-3
-4 FIGURE 11 Sa_0.3 s distribution and epicenter location of the
0 5 10 15 20 25 30 35 40 45
Time (s) Beijing M8.0 earthquake scenario (m/s2 )
0.6
0.4
Velocity (m/s)

0.2
0.0 of structural/nonstructural PGs, the four limit states of each
-0.2
-0.4
building were used to calculate the residual functionality of
-0.6 each RSU (Section 3). The unweighted residual functionality
0 5 10 15 20 25 30 35 40 45
Time (s) can reflect the damage of each RSU in the region as shown in
(b) Typical acceleration/velocity time–history Figure 12. In the M7.0 earthquake scenario (Figure 12a), the
FIGURE 10 Earthquake data of Beijing City. (a) The adopted residential buildings in the region experienced slight seismic
response spectral attenuation functions (M8.0) and (b) typical damage. The buildings in the eastern part of the city exhib-
acceleration/velocity time–history ited more severe damage than the buildings on the west. It is
noteworthy that in the M7.0 earthquake scenario, some build-
ings scattered in the core area of Beijing City sustain severe
Jiang et al. (2000), the accumulated probabilities of an M7.8 damage. These buildings are unreinforced masonry structures
and M6.9 earthquake occurring around the perimeter of the with a relatively weak seismic performance. The buildings in
city in the following 100 years are 0.42 and 0.73, respectively. the M8.0 earthquake scenario were more severely damaged
Therefore, two earthquake scenarios are considered: M8.0 compared with the buildings in the M7.0 earthquake scenario.
and M7.0 with the epicenter located at the eastern suburb of Nevertheless, buildings in the eastern part of the city expe-
Beijing; the fault strike is 50◦ (Fu, 2012). The epicenter rienced severe seismic damage in both earthquake scenarios
location and distribution of the 0.3-s spectral acceleration shown in Figure 12.
response (Sa_0.3 s) of the M8.0 earthquake scenario are pre-
sented in Figure 11.
6.3 Repair simulation of three RSUs
After acquiring the ground motion data for Beijing,
multistory and high-rise residential buildings were simu- Three RSUs in Beijing City were selected to demonstrate
lated with the MDOF shear models (Xiong et al., 2017) and validate the proposed method. The locations of the three
and MDOF flexural shear models (Xiong et al., 2016), RSUs are illustrated in Figure 13. The first RSU includes
respectively, through high-performance computing (Xu, Lu, 10 multistory residential buildings. The second and third
& Law, 2016). Because of the lack of a comprehensive RSUs include one high-rise residential building, respectively.
fragility database for structural and nonstructural PGs in The detailed information of these three RSUs is presented in
China and because the FEMA P58 fragility database is well Table 2. Moreover, the labor demand curves of the three RSUs
developed and world-renowned, the FEMA P58 fragility are calculated through the REDi method (Almufti & Willford,
database (FEMA, 2012) is adopted in this case study. There- 2013) and are illustrated in Figure 14.
fore, the seismic damage of structural/nonstructural PGs was As shown in Figure 13, RSU 1 is the closest to the epicenter
computed according to the FEMA P58 method (FEMA, among the three RSUs. Therefore, the proportion of residual
2012) and the details of the adopted PGs are listed in functionality for this RSU is the lowest (46.9%) in Table 2.
Appendix B. Subsequently, based on the obtained damages Moreover, RSU 1 consists of 10 multistory buildings, and the
12 XIONG ET AL.

0.310

0.535

0.504
P2
1.000

0.326

0.284
P1
(worker × days)
Worker days
(SLFHQWH
4,260

1,351

1,320
ದ 
ದ 
ದ 
Functional
loss (CNY/

ದ 
1,851,066

1,014,300

932,580
month)

NP ದ 

(a) M7.0 earthquake scenario

1
Repair time
(days)
22

37

39
functionality (%)
Proportion of

(SLFHQWHU
residual

ದ 
46.9

55.0

50.5

ದ 
ದ 
ದ 
Rental price

per month)
(CNY/ m2

NP ದ 

(b) M8.0 earthquake scenario

161

157
83

FIGURE 12 Distribution of unweighted residual functionality of

RSUs. (a) M7.0 earthquake scenario and (b) M8.0 earthquake scenario
Gross floor
area (m2 )
42,000

14,000

12,000

repairs can be conducted simultaneously for all 10 buildings if

given sufficient resources. Therefore, the repair time of RSU 1
is the smallest, while the worker day is the largest (Table 2).
–shear wall
Structural

–core tube

–core tube
RC frame

RC frame

RC frame

According to Equations (16) and (17), the priority indices P1

and P2 can be obtained as shown in Table 2.
type
Detailed information of the three RSUs

Based on the repair simulation method presented in Sec-

tion 5.3, repair simulations with a 50-worker constraint and
Year built

without constraints are performed. The detailed results and

2016

2004

2013

recovery curves are presented in Table 3 and Figure 15.

As shown in Table 3, the resilience index of the uncon-
strained case is larger than those of the constrained cases. For
Number of
buildings

the constrained cases, the repair scheme 2-3-1-Con yields the

largest resilience index (70.40%), which agrees with the order
10

of the P2 priority index values of RSU 1–3 in Table 2. Note

that the resilience index of the repair scheme using the P2
Stories

priority index (2-3-1-Con) is higher than the resilience index

28

20
7
TABLE 2

using the P1 priority index (1-2-3-Con). This is because the

RSU ID

required worker days of RSU 1 are significantly greater than

those of RSU 2–3, while the functional loss (i.e., the func-
1

tionality that can be restored after full recovery) is slightly

XIONG ET AL. 13

1 8,000,000

7,000,000

Functionality (CNY/month)
6,000,000

5,000,000

4,000,000
Uncon 1-2-3-Con
(SLFHQWHU
568 3,000,000 1-3-2-Con 2-1-3-Con
568 2-3-1-Con 3-1-2-Con
3-2-1-Con
568 2,000,000
0 20 40 60 80 100 120 140 160
Time (day)

FIGURE 15 Recovery curves of the three RSUs using different

NP
repair schemes

FIGURE 13 Locations of the three RSUs

6.4 Repair simulation of Beijing City
In this section based on the repair scheduling method pre-
TABLE 3 Summary of different repair schemes sented in Section 5, repair priority indices P1 and P2 of all
Repair scheme ID Repair scheme Resilience index (%) RSUs in Beijing City are calculated; the distribution of repair
1 1-2-3-Con 64.87 priorities is presented in Figure 16. The red dots in Figure 16
2 1-3-2-Con 64.70 represent RSUs with higher repair priorities. The distributions
3 2-1-3-Con 67.84 of repair priority using both P1 and P2 (Figures 16a and b)
4 2-3-1-Con 70.40 indicate that RSUs in the core area tend to have higher repair
5 3-1-2-Con 67.33 priority. This is because the rental price in the core area of
6 3-2-1-Con 70.23
Beijing is much higher. Note that all studied residential build-
ings are multistory to high-rise residential apartments, and
7 Uncon 90.29
there is no villa or high-end real estate included. Therefore,
Notes: 1-2-3-Con denotes the constrained repair scheme with the repair order of
the rental prices of residential apartments in Beijing City can
RSU 1, RSU 2, and RSU 3; Uncon means the unconstrained case; all resilience
indices are computed with the control time of 160 days. reflect their socioeconomic functionality. Moreover, a repair
simulation using equal weighting is presented in Figure 16c.
The result in Figure 16c shows that the prioritized RSUs are
250 scattered throughout the entire area, which is significantly dif-
RSU 1 ferent from the case for the weighted result in Figure 16b. The
200 RSU 2 difference between Figures 16b and c indicates that the use
of socioeconomic index can effectively influence the repair
Labor demand

RSU 3
150 priority. Note that the built environment consists of differ-
ent types of buildings, such as office, hospital, and school
100 buildings, and these buildings also contribute to the seismic
resilience of cities. Therefore, further works are required to
50 determine the socioeconomic indices of these buildings.
The repair simulation method proposed in Section 5.3 is
0 adopted for the M8.0 and M7.0 earthquake scenarios. The
0 10 20 30 40
Days number of construction workers for repair simulation is set to
724,000 according to Beijing statistical data (BMBS, 2018).
FIGURE 14 Labor demand curve of the three RSUs
For comparison, repair simulations with no labor constraints
are also performed for both earthquake scenarios. Consider-
larger than those of RSU 2–3 (Table 2). Therefore, when labor ing all the foregoing parameters, a total of eight repair simu-
resources are limited, prioritizing RSU 2–3 can make better lation scenarios are simulated, as summarized in Table 4. The
use of limited labor force and achieve a higher restored func- simulated community recovery curves are shown in Figure 17.
tionality. The above results indicate that the priority index P2 As evident in Figure 17, at 50 days, the IM8-P2-Con
is a very effective measure for use in repair scheduling with and IM8-P2-Uncon scenarios achieve 62.6% and 80.9% of
resource constraints. full functionality, respectively. The results indicate that the
14 XIONG ET AL.

1 6.5E+10

6E+10

Functionality (CNY/month)
5.5E+10

5E+10

IM8-P1-Con
4.5E+10
IM8-P2-Con
(SLFHQWHU
4E+10 IM7-P1-Con
IM7-P2-Con
ದ  3.5E+10
ದ 
ದ  3E+10
ದ 
0 50 100 150 200 250 300
Time (day)
NP ದ 
(a) With labor constraint
(a) Using repair priority index P1
6.5E+10
1

6E+10

Functionality (CNY/month)
5.5E+10

5E+10

IM8-P1-Uncon
4.5E+10
IM8-P2-Uncon
(SLFHQWHU 4E+10 IM7-P1-Uncon
IM7-P2-Uncon
3.5E+10
ದ 
ದ  3E+10
ದ  0 50 100 150 200 250 300
ದ 
Time (day)
NP ದ  (b) Without labor constraint
(b) Using repair priority index P2
FIGURE 17 Community recovery curves of Beijing City with
1
or without labor constraint. (a) With labor constraint and (b) without
labor constraint

leaving the severely damaged buildings still under repair. At

this moment, the labor demand of the city has decreased to
become lower than the available labor force. Therefore, the
(SLFHQWHU rapidity results of functional recovery for IM8-P2-Con and
IM8-P2-Uncon scenarios are very close. Note that for severely
ದ 
damaged buildings, the rapidity of the recovery curve is low,
ದ 
ದ 
even though the labor force is sufficient. Therefore, from the
ದ  stakeholders’ point of view, it is important to retrofit buildings
NP ದ  with poor seismic performance before earthquakes to achieve
P2 with equal
(c) Using repair priority index fast recovery.
socioeconomic weighting Moreover, the repair times for attaining 90% of full func-
tionality are 46 and 138 days for the IM7-P2-Con and IM8-
FIGURE 16 Distribution of RSU repair priority. (a) Using
P2-Con earthquake scenarios, respectively. Therefore, from
repair priority index P1 , (b) using repair priority index P2 , and (c) using
the stakeholder point of view, the simulation results can pro-
repair priority index P2 with equal socioeconomic weighting
vide a reference for the contingency plan, for example, deter-
mining how many interim houses and emergency supplies
recovery rapidity of the unconstrained scenario is signifi- should be prepared. For the stakeholders of some impor-
cantly larger than that of the constrained scenario. This is tant public buildings, the necessity of structural or nonstruc-
because, at the early stage of the repair process, a large num- tural retrofitting should be assessed to ensure good seismic
ber of buildings need to be repaired, and the required labor resilience of the building.
demand is larger than the available labor force. After 100 days, In Figure 17, the community recovery curves that adopt
most of the slightly damaged buildings have been repaired, priority indices P1 and P2 approximate each other. To
XIONG ET AL. 15

TABLE 4 Summary of repair simulation scenarios

Scenario ID Earthquake scenario Repair priority index Labor constraint
IM8-P1-Con M8.0 earthquake scenario P1 724,000 workers
IM8-P2-Con M8.0 earthquake scenario P2 724,000 workers
IM8-P1-Uncon M8.0 earthquake scenario P1 Unlimited workers
IM8-P2-Uncon M8.0 earthquake scenario P2 Unlimited workers
IM7-P1-Con M7.0 earthquake scenario P1 724,000 workers
IM7-P2-Con M7.0 earthquake scenario P2 724,000 workers
IM7-P1-Uncon M7.0 earthquake scenario P1 Unlimited workers
IM7-P2-Uncon M7.0 earthquake scenario P2 Unlimited workers

100 98.51 3,000,000

98.37 98.43 98.51

2,500,000
95
Resilience index (%)

Number of workers
2,000,000
90
87.91 87.91
86.26 1,500,000
85.64
0.5X
85
1,000,000 1X
2X
80 500,000 3X
4X
0
75
0 50 100 150 200 250 300
Time (day)

FIGURE 20 Labor demand curves of IM8-P2-Con constrained

by different numbers of workers (0.5X–4X denote the cases using 0.5–4
FIGURE 18 Comparison of resilience indices, R, among times the original 724,000 workers)
different repair simulation scenarios (all resilience indices are
computed with the control time of 300 days)
resource, such that it yields a more rapid recovery in spite of
the labor constraint.
800,000 The community labor demand curves of different repair
720,000
700,000 700,000
simulation scenarios are shown in Figure 19. The results indi-
600,000
680,000 cate that the community labor demand curves that use pri-
ority indices P1 and P2 are considerably similar but are not
Number of workers

660,000

500,000 640,000

620,000
the same (shown as the enlarged curve in Figure 19). This is
400,000 600,000
50 55 60
because both P1 and P2 are very good priority indices, but P2
300,000 IM8-P1-Con is better for the case with resource constraints. In Figure 19,
200,000 IM8-P2-Con because the required labor demand of the city is significantly
IM7-P1-Con
100,000 larger than the available labor force, there exists a constant
IM7-P2-Con
0
labor demand segment at the beginning of the repair process.
0 50 100 150 200 250 300 The number of days during which the assigned workers were
Time (day)
maintained at a full capacity is 56 and 17 days for the M8.0
FIGURE 19 Labor demand curves of different repair simulation and M7.0 earthquake scenarios, respectively. Subsequently,
scenarios after the repair of some RSUs has been completed, the labor
demand significantly decreased.
The labor demand curves of simulation scenario IM8-P2-
further demonstrate the differences between repair scheduling Con under constraints of different numbers of workers are pre-
scenarios using P1 and P2 , the community resilience indices sented in Figure 20. For the cases using 0.5–3 times the origi-
of different simulation scenarios are presented in Figure 18. nal 724,000 workers, there also exists a constant labor demand
The results indicate that when the available labor force is lim- segment. However, when the number of workers equals or is
ited, the resilience index using P2 is larger than that when P1 larger than four times the original 724,000 workers, the max-
is employed. This is because P2 can fully utilize the limited imum labor demand of the scenario is attained.
16 XIONG ET AL.

     

(a) 10 days–repair work (b) 30 days–the region of repair (c) 50 days–repair work moves to
concentrated in the core area work expanded the suburban area

     

(d) 70 days–labor demand (e) 90 days–labor demand (f) 110 days–labor demand
decreasing decreasing decreasing

     

(g) 130 days–severely damaged (h) 150 days–severely damaged (i) 170 days–severely damaged
RSUs at the eastern part RSUs at the eastern part RSUs at the eastern part
continue to undergo repair continue to undergo repair continue to undergo repair

FIGURE 21 Temporal evolution of the worker distribution for the M8.0-P2-Con repair simulation scenario (workers per square kilometer)

To further demonstrate the repair process of regional are finished and the labor demand decreases to less than the
buildings, the temporal evolution of worker distribution of number of workers available for the following 70–110 days.
the M8.0-P2-Con repair simulation scenario is presented in For the subsequent 130–170 days, most of the slightly dam-
Figure 21. The figure shows the number of workers per square aged RSUs are repaired, and only a few of the severely dam-
kilometer at different moments and displays the progress of aged RSUs at the eastern part of the city continue to undergo
repair in the area. For example, the repair work is mainly con- repair.
centrated in the core area for 10 days because the rental price
in this area is considerably higher; accordingly, the repair pri-
orities of RSUs in this area are higher. As the repair work 7 CONC LU SI ON S
proceeds, the repair of some RSUs in the core area of the city
is completed. Thereafter, the repair work gradually moves to A city-scale THA-driven building seismic resilience simula-
the suburban area for the next 30–50 days. More repair works tion framework is introduced in this work. The framework
XIONG ET AL. 17

uses the story-level EDPs obtained from the THA of each porated into the proposed model. Such processes should
building to achieve a reasonable seismic resilience evaluation be comprehensively considered in future studies.
and repair scheduling. Some of the conclusions drawn are as 3. The resilience of the water supply, power supply, and
follows: transportation system can significantly influence the
community resilience. This study primarily focused on
1. A probabilistic calculation method of the post-earthquake the seismic resilience of buildings and the repair process
residual functionalities of communities is proposed based as constrained by labor resource. The resilience of other
on the building seismic performance assessment method of infrastructures or the interaction effects of multiple
FEMA P58. Adopting the EDPs of each building obtained infrastructure systems with the resilience of buildings
from THA and the fragility database of FEMA P58, the should be further discussed in future studies.
residual functionality of each building can be reasonably 4. As this is a tentative study, only the labor resource con-
obtained by aggregating the damage of all structural and straint is considered in this work. Nevertheless, in real
nonstructural components. The proposed building resid- practice, the repair processes of city-scale buildings can
ual functionality calculation method cannot only be used also be influenced by the availability of construction
to estimate the residual functionality of buildings at city machinery and water/power supplies, financing, and traf-
scale, but can also be applied to individual buildings, fic conditions. Related research on the influences of other
thereby assisting in the performance-based design and resource constraints and the interdependency of various
seismic resilience design of buildings. resources should be further investigated.
2. Both physical and socioeconomic functionalities of each 5. The repair simulation assumes that the repair is managed in
building can be considered using the building function- a central-command way. However, real recovery scenarios
ality index. In the case study of Beijing City, the rental can be influenced by various factors, such as financing and
price was used to represent the socioeconomic function- the owner’s willingness to repair. It is suggested that the
ality index of residential buildings. Accordingly, this can influences of these factors should also be comprehensively
better guide decision making and repair scheduling. considered in future research.
3. A repair scheduling method for buildings at the city 6. The adopted fragility data of structural and nonstructural
scale is proposed based on the labor demand curve and components cannot consider the physical condition and
job block. Such a method can effectively encapsulate age of building, and such factors should be further investi-
the repair details of each RSU in the labor demand gated in future studies.
curve to significantly reduce the complexity of city-scale
7. The development of earthquake prediction and warning
repair scheduling. In future studies, this repair schedul-
system (Panakkat & Adeli, 2007, 2008, 2009; Rafiei &
ing method can also be applied to other types of infra-
Adeli, 2017) enables the prediction of possible locations
structures constrained by various resources as long as the
and magnitudes of future earthquakes. Such results can
corresponding demand curve is obtained.
be incorporated with the proposed method to estimate the
seismic resilience of cities.
This study provides useful tools for the city-scale seismic
resilience simulation of buildings under resource constraints.
Based on the city-scale building nonlinear THA, together
Since this is a tentative study, this work bears some limita-
with the component-level seismic damage assessment method
tions and there are still several extensions to be explored as
of FEMA P58 and the repair simulation method of REDi, the
follows:
proposed framework provides an insight into the community
repair simulation with labor constraint. Accordingly, these can
1. Residential buildings are selected in the case study to
assist in the seismic resilience evaluation of the built environ-
demonstrate the proposed method, and the rental price
ment, so as to ultimately achieve a resilient community.
is adopted as the socioeconomic index of the residen-
tial buildings. Note that the built environment consists of
different types of buildings, such as office, hospital, and ACKNOW LEDGMENTS
school buildings, and these buildings also contribute to the The authors are grateful for the financial support received
seismic resilience of cities. Therefore, further works are from the Scientific Research Fund of Institute of Engineering
required to determine the socioeconomic indices of these Mechanics, China Earthquake Administration (grant number:
buildings. 2017D01), Intellectual Innovation Program of Shenzhen Sci-
2. The lead time required for building inspection, evaluation, ence and Technology Innovation Committee (grant number:
finance planning, architectural/engineering consultations, JCYJ20180305123919731), and the Natural Science Founda-
contracting, and mobilizing for construction is not incor- tion of SZU (grant number: 2017064).
18 XIONG ET AL.

REFERENCES Fu, C. H. (2012). A study on long-period acceleration response spectrum

of ground motion affected by basin structure of Beijing (Ph.D. thesis).
Almufti, I., & Willford, M. (2013). REDiTM rating system: Resilience- Institute of Geophysics, China Earthquake Administration, Beijing,
based earthquake design initiative for the next generation of build- China (in Chinese).
ings. San Francisco, CA: Arup. Garey, M. R., & Johnson, D. S. (1979). Computers and intractability:
Ballantyne, D. B., Berg, E., Kennedy, J., Reneau, R., Wu, D., Taylor, C. A guide to the theory of NP-completeness. New York, NY: W. H.
E., … Tillman, C. (1990). Earthquake loss estimation modeling of the Freeman.
Seattle water system. Report No. 886005.00, Kennedy/Jenks/Chilton, Gasparini, D., & Vanmarcke, E. H. (1976). SIMQKE: A program for arti-
Federal Way, Washington, D.C. ficial motion generation. Cambridge, MA: Department of Civil Engi-
Batouli, M., & Mostafavi, A. (2018). Multiagent simulation for complex neering, Massachusetts Institute of Technology.
adaptive modeling of road infrastructure resilience to sea-level rise. Graves, R. W., & Pitarka, A. (2010). Broadband ground-motion simula-
Computer-Aided Civil and Infrastructure Engineering, 33(5), 393– tion using a hybrid approach. Bulletin of the Seismological Society of
410. America, 100(5A), 2095–2123.
BMBS (Beijing Municipal Bureau of Statistics). (2018). Retrieved from Hori, M., & Ichimura, T. (2008). Current state of integrated earthquake
http://www.bjstats.gov.cn/zxfb/201803/t20180316_394572.html simulation for earthquake hazard and disaster. Journal of Seismology,
Bocchini, P., & Frangopol, D. M. (2012). Restoration of bridge networks 12(2), 307–321.
after an earthquake: Multicriteria intervention optimization. Earth- Hutt, C. M., Almufti, I., Willford, M., & Deierlein, G. (2015). Seismic
quake Spectra, 28(2), 426–455. loss and downtime assessment of existing tall steel-framed buildings
Bruneau, M., Chang, S. E., Eguchi, R. T., Lee, G. C., O’Rourke, T. D., and strategies for increased resilience. Journal of Structural Engi-
Reinhorn, A. M., … von Winterfeldt, D. (2003). A framework to neering, 142(8), C4015005.
quantitatively assess and enhance the seismic resilience of commu- Isumi, M., Nomura, N., & Shibuya, T. (1985). Simulation of post-
nities. Earthquake Spectra, 19(4), 733–752. earthquake restoration of lifeline systems. International Journal of
Bruneau, M., & Reinhorn, A. (2007). Exploring the concept of seismic Mass Emergencies and Disasters, 3(1), 87–105.
resilience for acute care facilities. Earthquake Spectra, 23(1), 41–62. Jiang, W., Hou, Z., Su, Y., Zhang, Y., Xie, X., & Xiao, Z. (2000). Quan-
Burton, H. V., Deierlein, G., Lallemant, D., & Lin, T. (2015). Framework titative study of holocene activity of main active faults in the Beijing
for incorporating probabilistic building performance in the assess- plain and prediction of future seismic danger. Bulletin of the Institute
ment of community seismic resilience. Journal of Structural Engi- of Crustal Dynamics, 13, 1–15 (in Chinese).
neering, 142(8), C4015007. Kam, W. Y., Pampanin, S., & Elwood, K. (2011). Seismic performance of
Chang, S. E., & Shinozuka, M. (2004). Measuring improvements in the reinforced concrete buildings in the 22 February Christchurch (Lyt-
disaster resilience of communities. Earthquake Spectra, 20(3), 739– tleton) earthquake. Bulletin of the New Zealand Society for Earth-
755. quake Engineering, 44, 239–278.
Chang, S. E., Shinozuka, M., & Svekla, W. (1999). Modeling post- Kang, H. (2018). From rupture to recovery: Integrating probabilis-
disaster urban lifeline restoration. 5th U.S. Conference on Life- tic building performance assessment, decision-making and socioe-
line Earthquake Engineering, ASCE Technical Council on Lifeline conomic vulnerability to model post-earthquake housing recovery
Earthquake Engineering, New York, NY. (Ph.D. thesis). University of California at Los Angeles, CA.
Cimellaro, G. P., Reinhorn, A. M., & Bruneau, M. (2010a). Framework Lin, P., & Wang, N. (2016). Building portfolio fragility functions to
for analytical quantification of disaster resilience. Engineering Struc- support scalable community resilience assessment. Sustainable and
tures, 32, 3639–3649. Resilient Infrastructure, 1(3–4), 108–122.
Cimellaro, G. P., Reinhorn, A. M., & Bruneau, M. (2010b). Seismic Lu, X. Z., Lu, X., Guan, H., & Ye, L. P. (2013). Collapse simula-
resilience of a hospital system. Structure and Infrastructure Engi- tion of reinforced concrete high-rise building induced by extreme
neering, 6(1–2), 127–144. earthquakes. Earthquake Engineering & Structural Dynamics, 42(5),
Cimellaro, G. P., Solari, D., & Bruneau, M. (2014). Physical infrastruc- 705–723.
ture interdependency and regional resilience index after the 2011 Lu, X. Z., Xie, L. L., Yu, C., & Lu, X. (2016). Development and applica-
Tohoku Earthquake in Japan. Earthquake Engineering & Structural tion of a simplified model for the design of a super-tall mega-braced
Dynamics, 43, 1763–1784. frame-core tube building. Engineering Structures, 110, 116–126.
Decò, A., Bocchini, P., & Frangopol, D. M. (2013). A probabilistic Lu, H., & Zhao, F. (2007). Site coefficients suitable to China site cate-
approach for the prediction of seismic resilience of bridges. Earth- gory. Acta Seismologica Sinica, 29(1), 67–76 (in Chinese).
quake Engineering & Structural Dynamics, 42(10), 1469–1487. Luna, R., Balakrishnan, N., & Dagli, C. H. (2011). Postearthquake
Dong, Y., & Frangopol, D. M. (2016). Performance-based seismic recovery of a water distribution system: Discrete event simulation
assessment of conventional and base-isolated steel buildings includ- using colored petri nets. Journal of Infrastructure Systems, 17(1),
ing environmental impact and resilience. Earthquake Engineering & 25–34.
Structural Dynamics, 45(5), 739–756. Mahin, S. A., Deierlein, G. G., Govindjee, S., Lowes, L. N., Kareem,
Fang. (2018). Retrieved from http://www.fang.com/ A., Crittenden, C., … Schoettler, M. (2018). SimCenter computa-
FEMA. (2012). Seismic performance assessment of buildings (Volume 1- tional framework for simulating and mitigating the effects of earth-
Methodology). Washington, D.C.: Federal Emergency Management quakes. 11th U.S. National Conference on Earthquake Engineering,
Agency. Los Angeles, CA.
Franchin, P., & Cavalieri, F. (2015). Probabilistic assessment of civil Mahmoud, H., & Chulahwat, A. (2018). Spatial and temporal quantifica-
infrastructure resilience to earthquakes. Computer-Aided Civil and tion of community resilience: Gotham City under attack. Computer-
Infrastructure Engineering, 30(7), 583–600. Aided Civil and Infrastructure Engineering, 33(5), 353–372.
XIONG ET AL. 19

McGuire, R. K. (2008). Probabilistic seismic hazard analysis: Early his- seismic damage prediction of multi-story buildings. Journal of
tory. Earthquake Engineering & Structural Dynamics, 37(3), 329– Earthquake Engineering, 21(3), 461–485.
338. Xu, M., Ouyang, M., Mao, Z., & Xu, X. (2019). Improving repair
Miles, S. B., & Chang, S. E. (2006). Modeling community recovery from sequence scheduling methods for postdisaster critical infrastructure
earthquakes. Earthquake Spectra, 22(2), 439–458. systems. Computer-Aided Civil and Infrastructure Engineering, 34,
Miranda, E., & Taghavi, S. (2005). Approximate floor acceleration 506–522.
demands in multistory buildings. I: Formulation. Journal of Struc- Xu, Z., Lu, X. Z., & Law, K. H. (2016). A computational framework for
tural Engineering, 131(2), 203–211. regional seismic simulation of buildings with multiple fidelity mod-
Mostafavi, A., & Ganapati, N. E. (2019). Toward convergence disaster els. Advances in Engineering Software, 99, 100–110.
research: Building integrative theories using simulation. Risk Analy- Yu, Y. X. (2002). Study on attenuation relationships of long period
sis, https://doi.org/10.1111/risa.13303 ground motions (Ph.D. thesis). Institute of Geophysics, China Earth-
Panakkat, A., & Adeli, H. (2007). Neural network models for earthquake quake Administration, Beijing, China (in Chinese).
magnitude prediction using multiple seismicity indicators. Interna- Zeng, X., Lu, X. Z., Yang, T. Y., & Xu, Z. (2016). Application of the
tional Journal of Neural Systems, 17(1), 13–33. FEMA-P58 methodology for regional earthquake loss prediction.
Panakkat, A., & Adeli, H. (2008). Recent efforts in earthquake prediction Natural Hazards, 83(1), 177–192.
(1990–2007). Natural Hazards Review, 9(2), 70–80. Zhang, W., & Nicholson, C. (2016). A multi-objective optimization
Panakkat, A., & Adeli, H. (2009). Recurrent neural network for approxi- model for retrofit strategies to mitigate direct economic loss and pop-
mate earthquake time and location prediction using multiple seismic- ulation dislocation. Sustainable and Resilient Infrastructure, 1(3–4),
ity indicators. Computer-Aided Civil and Infrastructure Engineering, 123–136.
24(4), 280–292.
Parker, M., & Steenkamp, D. (2012). The economic impact of the Can-
terbury earthquakes. Reserve Bank of New Zealand Bulletin, 75(3),
13–25. How to cite this article: Xiong C, Huang J, Lu X.
Rafiei, M. H., & Adeli, H. (2017). NEEWS: A novel earthquake Framework for city-scale building seismic resilience
early warning system using neural dynamic classification and neural
simulation and repair scheduling with labor constraints
dynamic optimization model. Soil Dynamics and Earthquake Engi-
driven by time–history analysis. Comput Aided Civ Inf.
neering, 100, 417–427.
Ramirez, C. M., & Miranda, E. (2012). Significance of residual drifts 2019;1–20. https://doi.org/10.1111/mice.12496
in building earthquake loss estimation. Earthquake Engineering &
Structural Dynamics, 41(11), 1477–1493.
SPUR. (2012). Safe enough to stay. San Francisco, CA: San Francisco
Bay Area Planning and Urban Research Association. APPENDIX A : R EPAIR SIMULAT ION
Sun, C., & Xu, J. (2011). Estimation of time for Wenchuan earthquake WORKFLOW
reconstruction in China. Journal of Construction Engineering and
The repair simulation workflow is shown in Pseudo-code 1;
Management, 137(3), 179–187.
Tian, Y., Lu, X., Lu, X. Z., Li, M. K., & Guan, H. (2016). Quantifying
it is explained as follows:
the seismic resilience of two tall buildings designed using Chinese
1. Initialize the current time and aggregate the workloads of
and US codes. Earthquakes and Structures, 11(6), 925–942.
Wang, Z., Wang, Q., Zukerman, M., Guo, J., Wang, Y., Wang, G., all RSUs in RemainWork. Calculate the postearthquake
… Moran, B. (2017). Multiobjective path optimization for crit- residual functionality of all RSUs and aggregate them
ical infrastructure links with consideration to seismic resilience. in ResidualFunctionality[0]. Thereafter, start the WHILE
Computer-Aided Civil and Infrastructure Engineering, 32(10), 836– loop until the workload of the entire urban area in Remain-
855. Work is less than zero; this means the completion of all
Wikipedia. (2012). List of tallest buildings in Christchurch. repair works.
Retrieved from http://en.wikipedia.org/wiki/List_of_tallest_build-
ings_in_Christchurch 2. Each WHILE loop represents the repair process within
Xiong, C., Lu, X. Z., Hong, G., & Xu, Z. (2016). A nonlinear computa- a specific interval of time TimeStep (e.g., 1 day and 1
tional model for regional seismic simulation of tall buildings. Bulletin week). At the beginning of each WHILE loop, the num-
of Earthquake Engineering, 14(4), 1047–1069. ber of available workers, AvailWorkers, is initialized to a
Xiong, C., Lu, X. Z., Hori, M., Guan, H., & Xu, Z. (2015). Building seis- total number of workers, TotalWorkers, for the subsequent
mic response and visualization using 3D urban polygonal modeling. worker assignment. Next, enter the FOR loop to simulate
Automation in Construction, 55, 25–34.
the repair process of each RSU. It is noteworthy that RSUs
Xiong, C., Lu, X. Z., Huang, J., & Guan, H. (2019). Multi-LOD seismic-
damage simulation of urban buildings and case study in Beijing CBD.
in the RSU container, RSUList, have been sorted accord-
Bulletin of Earthquake Engineering, 17, 2037–2057. ing to the repair priority indices (Equations (15) and (16))
Xiong, C., Lu, X. Z., Lin, X. C., Xu, Z., & Ye, L. P. (2017). Parame- in descending order; the repair process will be conducted
ter determination and damage assessment for THA-based regional according to the RSU order in RSUList.
20 XIONG ET AL.

Pseudo-code 1 5. At the end of the FOR loop of each RSU, assess whether
Repair simulation the workload of the current RSU has been completed. If
Time = 0 so, then update the community recovery curve in Residu-
RemainWork = GetTotalRemainWork() alFunctionality, and erase the current RSU from RSUList.
ResidualFunctionality[0] = GetInitialFunctionality() At the end of each WHILE loop, update the current time,
WHILE RemainWork > 0
AvailWorkers = TotalWorkers obtain the remaining workload of the urban area; there-
FOR I = 1 to Sizeof(RSUList) after, move on to the WHILE loop of the next time
IF AvailWorkers == 0 interval.
BREAK
END IF
IF RSUList[I].RemainWork >0
A P P E N D I X B: A D O P T E D P E R FO R M A NC E
IF AvailWorkers >RSUList[I].WorkerLimit
RSUList[I].RemainWork –= G RO U P S
TimeStep*RSUList[I].WorkerLimit
AvailWorkers –=
RSUList[I].WorkerLimit TABLE B1 List of adopted performance groups
ELSE NISTIR
RSUList[I].RemainWork –= classification NISTIR name
TimeStep*AvailWorkers
AvailWorkers = 0 D2021.011a Cold or hot potable water piping––small diameter
END IF threaded steel (2.5″ in diameter or less)
END IF D2021.021a Cold or hot potable water piping (greater than 2.5″
IF RSUList[I].RemainWork <= 0 in diameter)
ResidualFunctionality[Time] += D2031.021a Sanitary waste piping––cast iron w/bell and spigot
RSUList[I].FuncLoss coupling
RSUList.erase(I) D2022.011a Heating hot water piping––small diameter threaded
END IF steel––(2.5″ in diameter or less)
END FOR
Time += TimeStep D2022.021a Heating hot water piping––large diameter welded
steel––(greater than 2.5″ in diameter)
ResidualFunctionality[Time]
= ResidualFunctionality[Time – TimeStep] D3041.031a HVAC drops/diffusers in suspended ceilings––no
RemainWork = GetTotalRemainWork() independent safety wires
END WHILE C3011.001c Wall partition, type: gypsum + wallpaper, full
3. At the beginning of each FOR loop of the RSU simu- height, fixed below, slip track above w/returns
(friction connection)
lation, assess whether AvailWorkers is zero. If so, then
B1071.041 Exterior wall––type: gypsum with wood studs, full
it means that all workers have been allocated and there
height, fixed below, fixed above
is no available worker for the current RSU; accordingly,
D1014.011 Traction elevator––applies to most California
BREAK and exit the FOR loop. If AvailWorkers is greater installations 1976 or later, most western states
than zero, then assess whether the remaining workload of installations 1982 or later and most other U.S.
the current RSU in RSUList[I].RemainWork is greater than installations 1998 or later
zero. If so, then the repair work will be conducted for the C2011.021b Monolithic cast-in-place and precast concrete stairs
with no seismic joints––replace in kind if
current RSU.
replacement is required
4. Because the RSUs in RSUList have been sorted in B1044.101 Slender concrete wall, 18″ thick, 12′ high, 15′ long
descending order according to the repair priority index B1044.102 Slender concrete wall, 18″ thick, 12′ high, 20′ long
of each RSU, the repair priority of the RSU in the
B1044.103 Slender concrete wall, 18″ thick, 12′ high, 30′ long
current FOR loop is always higher than the latter
B1041.061a ACI 318 OMF with weak columns, Conc Col &
one. Therefore, the maximum labor demand of the cur- Bm = 24″ × 24″ , beam one side
rent RSU is satisfied as long as there are sufficient B1041.061b ACI 318 OMF with weak columns, Conc Col &
workers in AvailWorkers. Specifically, assess whether Bm = 24″ × 24″ , beam both sides
the number of remaining workers in AvailWorkers is B1041.063a ACI 318 OMF with weak columns, Conc Col &
greater than the maximum labor demand of current RSU Bm = 36″ × 36″ , beam one side
RSUList[I].WorkerLimit. If so, then the number of required B1041.063b ACI 318 OMF with weak columns, Conc Col &
maximum workers is allocated to the current RSU, and Bm = 36″ × 36″ , beam both sides
the remaining workload of the current RSU is reduced B1051.011 Ordinary reinforced masonry walls with partially
grouted cells, shear dominated, 8″ –12″ thick, up
by TimeStep * RSUList[I].WorkerLimit; otherwise, allo-
to 12′ high
cate all available workers in AvailWorkers to the current
RSU for repair.