Documente Academic
Documente Profesional
Documente Cultură
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/312666268
CITATION READS
1 116
2 authors:
Some of the authors of this publication are also working on these related projects:
Risk assessment of bridges under multiple hazards in operation period View project
All content following this page was uploaded by Jelena M. Andri on 21 February 2017.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
International Journal of Disaster Risk Reduction (xxxx) xxxxxxxx
A BS T RAC T
Communities are exposed to natural catastrophes, such as tsunamis, earthquakes, hurricanes, etc. Therefore,
building and making resilient communities could lead to the reduction of the disastrous negative impacts and
enable fast recovery. In this paper, novel seismic recovery functions and a metric for seismic resilience
assessment have been proposed employing the concepts from fuzzy sets theory. The basic resilience parameters
have been dened by fuzzy knowledge representation theory, the recovery process of bridges has been modeled
in terms of fuzzy functions and the concepts from fuzzy measure theory have been used to determine resilience
metrics. In addition, the presented model is integrated into a decision-making process for disaster preparedness
of communities. Moreover, the model has been simulated in Java. A bridge in Santa Barbara has been
investigated for this case study. The proposed method has revealed that fuzzy set theory has been a more
ecient tool for estimating the relationship between bridge damage and functionality since the collected data
has been established based on experts judgments. The conclusion has been drawn that improving the disaster
preparedness of communities would enhance the resilience of the bridge. The proposed model could be adapted
for seismic resilience assessment of other infrastructure components such as tunnels, highway segments and in
the case of bridges under multiple hazards.
Corresponding author.
E-mail addresses: jln.andric@gmail.com, jln.andric@hit.edu.cn (J.M. Andri), ludagang@hit.edu.cn (D.-G. Lu).
http://dx.doi.org/10.1016/j.ijdrr.2017.01.001
Received 26 May 2016; Received in revised form 7 January 2017; Accepted 9 January 2017
2212-4209/ 2017 Elsevier Ltd. All rights reserved.
Please cite this article as: Andric, J.M., International Journal of Disaster Risk Reduction (2017), http://dx.doi.org/10.1016/j.ijdrr.2017.01.001
J.M. Andri, D.-G. Lu International Journal of Disaster Risk Reduction (xxxx) xxxxxxxx
2
J.M. Andri, D.-G. Lu International Journal of Disaster Risk Reduction (xxxx) xxxxxxxx
such as spectral acceleration or peak ground acceleration based on For membership functions, the triangular fuzzy numbers are used since
seismological and geological data which is found in earthquakes they are very suitable to describing parameters of resilience due to their
catalogs [45]. An earthquake catalog is a list that provides the characteristics to refer to three possible case scenarios: pessimistic,
information of recorded earthquakes: time of occurrence (year, day, most-probable and optimistic. A triangular fuzzy number is determined
time), longitude and latitude of location, focal depth, and earthquake with three points A(a, b , c ). Likewise, for the residual functionality, the
size magnitudes [46]. other resilience parameters, i.e., target functionality, idle time interval,
and recovery time, are all represented using fuzzy set theory as
3.2. Bridge fragility analysis triangular fuzzy numbers. The subsequent denitions of the resilience
parameters are given and applied to four resilience parameters.
The level of bridge damage depends on the earthquake intensity In this research, the resilience parameter is dened as a convex,
and the bridges structural type. Moreover, the task of structural normalized fuzzy set a(i ) = <a1, a 2 , a3>, which is determined by the
vulnerability analysis is to estimate the expected level of damage which interval bounds by the value a1 and the value a3. The membership
is used to predict resilience. Commonly, this analysis is carried out by function is segmentally continuous, and has the functional va-
fragility curves, which represent the probability of exceeding certain lue (ai )=1, at exactly one value, for ai = a 2 . The membership function
damage states for a given peak ground motion intensity (PGA). Based for the resilience parameter is given as:
on the method which has been used in the determination of the fragility
ai a1 , a < a < a
curves, there are judgemental [47], empirical [48], and analytical a2 a1 1 i 2
fragility functions [49]. (ai ) = 3 i , a < a < a
a a
Besides fragility curves, the computational intelligence techniques a3 a2 2 i 3
for soft-computing can be applied to damage evaluation. After the 0, otherwise (1)
earthquake, the number of experts to perform the damage assessment
The residual functionality is a resilience parameter that is given by a
could be insucient and the nonexperts tend to overestimate or to
fuzzy set: Qr (i ) = <Qr1, Qr2 , Qr3> with a corresponding membership
underestimate damage due to the lack of experience [28]. In addition,
function:
the information obtained by all evaluation methods is highly subjective
since the damage levels are described by linguistic qualications, such Qri Qr1
as light, moderate, severe. In order to overcome these issues, the Qr Qr , Qr1 < Qri < Qr2
2 1
computational intelligence techniques have been applied which make it (Qri ) = Qr 3 Qri
possible to avoid or reduce the mistakes made by nonexpert building Qr Qr , Qr2 < Qri < Qr3
3 2
inspectors [28]. 0, otherwise (2)
However, analytical fragility functions are used in this research
since fragility curves are the most commonly used tools for vulner- The residual functionality is determined by the damage state of the
ability analysis of bridges. In the current literature, fragility curves are structure. Dierent damage states of a structure correspond to dierent
dened for ve dierent damage states of bridges [50,51]: levels of the residual functionality. Since, there are ve damage states,
ve linguistic terms corresponding to these damage states form the
no damage (N); linguistic scale.
slight damage (S); The target functionality is a resilience parameter presented by fuzzy
moderate damage (M); set Qt =<Qt1, Qt2 , Qt3>, with the following membership function:
extensive damage (E); and Qt Qt1
complete damage (C). Qt Qt , Qt1 < Qt < Qt2
2 1
(Qt ) = Qt Qt2
3.3. Fuzzy resilience parameters Qt Qt , Qt2 < Qt < Qt3
3 2
0, otherwise (3)
The disaster resilience parameters are residual functionality, target
functionality, idle time interval, and recovery time. The reduced In this research, it is suggested that the idle time interval varies for
functionality of a bridge after a disaster is directly associated with its dierent bridge damage states. More seriously damaged or collapsed
level of damage [52]. The relationship between bridge damage and bridges require a more complex rehabilitation project initiation,
residual functionality is critical to assess, since uncertainties arise in implementation, and commissioning prior to recovery which will take
the decision making processes of bridge rehabilitation after an extreme more time to prepare. The idle time interval is dened as a resilience
event. In order to determine the residual functionality, the information parameter and is represented by fuzzy set ti=<ti1, ti2, ti3> with the
of the relationship between bridge damage and its functionality could membership function:
be gathered by surveys, in which decision makers provide their
ti ti1
opinions of the expected level of trac capacity after the earthquake. ti2 ti1 , ti1 < ti < ti2
Experts have made decisions on which bridge lanes would have been
(ti ) = ti ti2 , t < t < t
opened or closed for trac. ti3 ti2 i2 i i3
3
J.M. Andri, D.-G. Lu International Journal of Disaster Risk Reduction (xxxx) xxxxxxxx
measure dierent levels for residual functionality, idle time interval recovery have been inuenced by a good inux of resources and good
and recovery time, and to analyze the possible case scenarios since organization; a trigonometric recovery is applied in the case when the
fuzzy resilience parameters are represented as fuzzy triangular num- recovery speed is slow due to limited resources and organization of the
bers indicated by (a, b, c). For example, in the case of residual society in the post-disaster period.
functionality, the minimum represents a system with low robustness, A novel recovery model is established via fuzzy mathematics
most likely value moderate robustness, and the maximum high regarding that the recovery process is characterized by vague uncer-
robustness. Then, in the case of idle time interval, the minimum tainty. The proposed model is called the fuzzy recovery model, since it
denotes a short idle time, most likely value moderate idle time, and is developed using concepts from the fuzzy set theory. In this model,
the maximum long idle time of system. In the case of recovery time, the fuzziness originates from lexical uncertainties considered. The
the minimum describes fast, most likely moderate, and the maximum source of lexical uncertainties is derived from resilience parameters.
slow recovery of the system. Furthermore, recovery functions are described by fuzzy functions. The
fuzzy recovery function could be dened as an uncertain mapping of
3.4. Fuzzy recovery model x(t ):F (T )F (X ) that assigns exactly one xF (X ) to each t F (T ).
Graphically, the fuzzy recovery process is described by trajectories that
Disaster preparedness represents an indicator that describes the are the real functions or elements x (t ) of x(t ) dened for all t T .
level of the systems preparedness for the disaster pertaining to The proposed types of fuzzy recovery functions are: a fuzzy positive-
emergency response, recovery plans and availability of resources exponential function in the case of unprepared communities; a fuzzy
during the recovery. Moreover, redundancy and resourcefulness have linear function for average prepared communities and a fuzzy negative-
been associated with the disaster preparedness. Redundancy is the exponential function for well-prepared communities. A bridge placed
quality of having alternative paths in the structure by which the in a well-prepared community restores its functionality faster in the
lateral forces can be transferred, and allows the structure to remain beginning of the rehabilitation process than at the end. Moreover, well-
stable following the failure of any single element [54]. prepared communities can gain back functionality faster since recovery
Resourcefulness is represented by the capacity to identify problems, plans, priorities and mobilized resources have already been established
establish priorities and mobilize resources when a condition exists for disasters. In addition, the behavior of the recovery process in such
that threaten to disrupt some element, system, or other unit of communities is characterized by a fuzzy negative-exponential function.
analysis [11]. Furthermore, resilience could be enhanced through For average prepared communities, the recovery speed is approxi-
resourcefulness and redundancy since they represent the means of mately constant during the entire recovery period, and it is the best
resilience. If these two properties change, it would aect the shape of tted with the fuzzy linear function. An unprepared community has the
recovery function. The community where the bridge is located could be lowest speed recovery in the beginning, and it increases over time, so at
well-prepared, average-prepared and unprepared for a disastrous the end of process it is the highest. To describe this behavior, a fuzzy
event. positive-exponential function is employed.
The recovery signies key aspects of resilience and it is labeled as a The fuzzy function for the functionality of a system for a well-
non-linear and dynamic process with aims to increase and return previous prepared system is a fuzzy negative-exponential function that is given
functionality of a system [55]. Furthermore, Smith and Wenger [56] have as follows:
denoted the recovery process as the dierential process of restoring,
rebuilding, and reshaping the physical, social, economic, and natural Qr , t < ti
environment through preventative planning and post-event actions. In the Q (t ) = (Q
ln(200) * t t ti
t Qr )*e r + Qt , ti < t < ti + tr
literature, four dierent approaches for modeling restoration process are
Qt , t > ti + tr
suggested: statistical curve tting, deterministic resource constraint, (6)
Markov process, and network models [56].
in which Qt , Qr , and tr are the fuzzy target functionality, fuzzy residual
The recovery process is complex to describe due to its stochastic
functionality and fuzzy recovery time.
nature. Besides, it is comprised of the rehabilitation and reconstruction
The fuzzy function for functionality in the case of an average-
processes. The duration of this phase depends on the activity planning,
prepared system is a fuzzy linear function:
construction technology and techniques, available mechanization,
building materials, and labor forces. Furthermore, various factors from
Qr , t< ti
dierent sources have inuenced the shape of the recovery process.
t t
Among these factors, the disaster preparedness of a community in Q (t ) = (Qt Qr )* t i + Qr , ti < t < ti + tr
r
which the bridge is located has a major impact on the recovery nature.
Qt , t > ti + tr (7)
Due to this fact, the model has included uncertainties in possible
recovery paths. The fuzzy function for a poor-prepared/unprepared system is a
Generally, the recovery process is divided into two phases: the idle fuzzy positive-exponential function:
time interval and the recovery period. The rst phase represents the
Qr , t < ti
time between the earthquakes occurrence and the beginning of the
rehabilitation/reconstruction procedure. The functionality of a bridge t t
Q (t ) = (Q
ln(200) * i 1
Q )*( e tr * )+Qr , ti < t < ti + tr
during this period stays the same as it is after the earthquake. In t r 200
addition, activities that will not increase the functionality, such as
Qt , t > ti + tr
(8)
debris removal, are concluded in this phase. Therefore, during this
period, the functionality is constant and equal to the residual function-
ality. Hence, the functionality curve is straight. The second phase is a 3.5. Fuzzy resilience index
period between the time in which the rehabilitation process starts, and
the time in which the process ends. During this process, the function- In the existing research, the resilience is generally estimated by
ality increases from the residual to the target value. probabilistic approaches, and the value of resilience index is between 0
Cimellaro and Reinhorn [10] have suggested three dierent sim- and 1.
plied functions to describe the recovery process: a linear recovery However, to quantify the resilience index in this research, a novel
function in the case if there is no available information regarding the resilience metric is proposed. Since the resilience index is used for the
preparedness; an exponential recovery function when the response and fuzzy resilience model, it is implemented in fuzzy mathematics. The
4
J.M. Andri, D.-G. Lu International Journal of Disaster Risk Reduction (xxxx) xxxxxxxx
fuzzy resilience index represents a fuzzy measure on domain [0,1] as installation, grounding bridge, monitoring, and heating system of the
the fuzzy measure meets the set of requirements for the resilience roadway), debris removal, and temporary bypass construction.
denition and it is computed by a fuzzy integral. In this research, a novel formula for direct economic losses is
Mathematically, the fuzzy resilience index is expressed by a fuzzy proposed which evaluates the direct economic loss in terms of fuzzy
integral which represents the bridge functionality over the time period numbers. The direct economic loss of a bridge structure can be
noted as the time horizon: estimated by the following fuzzy integral:
th tr
Q (t ) u*L*W * tan
R = 0 th
dt
(9)
C = 0 (1+r )t
dt
(10)
where, Q (t ) is the fuzzy function representing the functionality of the in which, C the direct economic loss is expressed as a fuzzy triangular
bridge and th is a time horizon, and R is the fuzzy resilience index. number; u unit cost of the original bridge construction per square
Therefore, the concept of the fuzzy resilience index is introduced by meter; L bridge length; W bridge width; tan - recovery speed; r
a fuzzy triangular number, (R1, R2, R3), which are suitable for discount factor; and tr - recovery time.
representing resilience since each of the values relate to three possible Mathematically, the speed of recovery can be expressed as:
case scenarios: optimistic, most-possible and pessimistic. Moreover, to
Qt Qr
establish a crisp numerical value of resilience, the defuzzication tan =
tr (11)
process could be applied. In this case, the Incentre-of-Centroids
[57] method is employed for defuzzication. The validity of the above formula can be checked in the case of the
The -cut concept is applied to measure dierent levels of resilience complete damaged or total destruction of the bridge (residual function-
in order to analyze the possible case scenarios and possible recovery ality is equal to 0, Qr =0 ); herein the eect of the money discount rate is
paths. The value of is between 0 and 1, denoting that the degree of not considered (the discount rate is equal to 0, r = 0 ). In this case, a
uncertainty is the highest and the lowest, respectively. In the decision- new bridge with the same structural characteristics is needed to be
making processes, =0, =0.5 and =1 designates pessimistic, mod- built. The total restoration cost is equal to the value of construction of
erate and optimistic views, respectively. the new bridge. The target functionality is equal to 1 since the bridge
The functionality Q (t) is given as a fuzzy function, Fig. 2 depicts the returns functionality which it had before the disaster.
possible recovery paths that are referred as trajectories of the fuzzy
tr u*L*W *(Qt Qr ) tr u*L*W *(1 0)
function. The curve with =1 denotes the optimistic view for the C= 0 dt = 0 dt = u*L*W
particular damage state of the bridge. In addition, the optimistic view tr*(1+r )t tr (12)
indicates a highly resilient system that is characterized by a high
The indirect economic losses arise from transportation ow dis-
robustness and a rapid recovery. Moreover, the most-probable case
ruption of goods, services and people due to bridge damage or collapse.
scenario denotes a moderately resilient system with a moderate
It is very complex to compute these losses since a great deal of data of
recovery time and robustness. As a nal curve, the pessimistic view is
trac ow before and after disasters, alternative routes and others are
marked as =0 in Fig. 2, signifying the low resilience of the system that
required. The car/truck drivers must nd alternative routes that are
is labeled with low robustness, a long period of idleness and recovery
functional and didnt undergo damage. However, this route will have a
time. Using this curve (=0), the system functionality will be further
longer distance requiring time to reach the destination. Thus, the
described. The disaster occurred at time t0=0 (Point O). The idle time
reduction or loss of functionality will result in increased traveling
interval is the period between point O and point A. The recovery
distance and traveling time. In addition, the minimization of the
process begins at point A, in which the idle time interval is concluded.
recovery time and disaster preparedness is another salient aspect in
At point B, the recovery process is concluded and the system nally
the post-disaster period. The rapid recovery time will reduce the
reaches the target functionality.
indirect economic losses and enable the system to return to the initial
state as soon as practically possible, but it will increase the direct
3.6. Seismic loss model of bridges economic cost [58,59].
Social losses such as loss of life or injuries are dicult to compare to
The consequences of disasters are specied in terms that could economic losses since they are expressed in dierent units. Economic
aect and inuence communities, the environment and people. losses could be expressed as monetary units. However, diculties
Furthermore, disasters can result in highly uncertain losses. appear in estimating social losses such as loss of life, injury or social
Generally, these losses are classied into four dierent aspects: direct disruptions since these losses are non-monetary loss [60]. In some
economic, direct social, indirect economic and indirect social losses. literature, there have been a few attempts to quantify the economic
The direct economic losses depend on the damage ratio of the value of a human life. However, it is not common and not good practice
bridge. Such losses include repair fees of the bridges structural to discuss the value of human life in public.
components (foundations, bearings, piers, beams, and girders), non- The direct social losses are expressed in loss of lives or injuries
structural components (lightning, bridge drainage, trac and signaling which is estimated as a ratio of the number of the injured or dead
people Nid during the disaster and the total number of people Nt that are
present on the bridge:
Nid
LDC =
Nt (13)
5
J.M. Andri, D.-G. Lu International Journal of Disaster Risk Reduction (xxxx) xxxxxxxx
people. classied in one of the damage states. Therefore, the criteria used for
Indirect social loss arises due to the bridge dysfunction. After the assigning the damage state of the bridge for the specic IM is:
disaster has struck, injured people need medical treatment. If the
bridge collapses in disaster it is dicult to receive medical treatment "no damage" for F1(a) < RN;
since ambulances cant reach the location where the injured people are. "slight damage" for F2(a) < RN F1(a) ;
The indirect social loss is dened as a ratio of the people injured and "moderate damage" for F3(a) < RN F2(a) ;
dead number Nid due to bridge collapse or damage and the total "extensive damage" for F4(a) < RN F3(a) ;
number Nt of people who have been served in a hospital: "complete damage" for RN F4(a).
Nid
LIND = where RN random number, which is simulated by MCS; F1(a)
Nt (14)
fragility curve for slight damage; F2(a) fragility curve for moderate
damage; F3(a) fragility curve for extensive damage; F4(a)-fragility
curve for complete damage.
4. Case study: a bridge in Santa Barbara, California
4.1. Description of the bridge in Santa Barbara 4.4. Seismic resilience assessment
For the case study, a bridge located on the Route 225/Route 101 The residual functionality is assessed based on the collected data on
separation on the Las Positas Road in Santa Barbara, California, USA the relationship between bridge damage functionality. This data is
(structure id number: 510190 according to California Department of estimated by expert's opinions and gathered in order to develop a ve-
Transportation [62]) is investigated. The proposed procedure is applied member linguistic scale that is associated with ve dierent damage
to this multiple-span continuous girder bridge which was built in 1957. states. Padgett and DesRoches [70] have collected data based on
The total bridge length is 53.5 m and the width is 30.5 m with a left and experts subjective judgment of the expected level of trac capacity
right side walk of 1.5 m. The trac functions on the deck of the on a bridge after earthquakes. A ve-member linguistic scale consists of
structure. This bridge is selected since it is located in a seismic active linguistic terms dened by the membership function corresponding to
area and the seismic resilience of the bridge is used as a criterion for a fuzzy triangular number [47]. Furthermore, the linguistic scale for
disaster preparedness decision making. recovery duration is created by data taken from Shinozuka and Zhou
[71].
The linguistic scale for residual functionality is developed and the
4.2. Seismic hazard analysis of California
fuzzy linguistic terms with the corresponding fuzzy sets are illustrated
in Fig. 3. The estimated triangular fuzzy numbers for membership
California has high seismicity since a great number of active faults
functions of the linguistic terms in the case of the linguistic scale for
belong to this area. The seismicity of the California region has been
residual functionality are:
studied in numerous references. For example, Chang et al. [63] have
proposed a new method which will evaluate probabilistic seismic risk
1. Residual functionality in the case of "no damage": Qr1= < 1,1,1>
analysis of the region. This method detects a set of deterministic
2. Residual functionality in the case of "slight damage":
earthquake scenarios and for each scenario assigns hazard probability
Qr2= < 1,0.75,0.5>
in order to estimate the regional seismicity. In this case, from the
3. Residual functionality in the case of "moderate damage":
United States Geological Survey (USGS) 47 representative earthquake
Qr3= < 0.5,0.25,0>
scenarios have been identied. The state of bridge damage under these
4. Residual functionality in the case of "extensive damage":
conditions is evaluated in the next section.
Qr4= < 0.2,0.1,0>
5. Residual functionality in the case of "complete damage":
4.3. Bridge fragility analysis
Qr5= < 0,0,0>
The current literature reveals that the aim of many previous The values of the membership functions for the fuzzy set slight
research is the development of fragility curves for bridges in the damage (which is given by triangular shape) is < 1.0, 0.75, 0.5 > ,
Caltrans Network [6468]. Due to this fact, in this case study, the which means that 1.0 and 0.5 are the boundary values of this triangular
fragility curves which have been developed by Karamlou and Bocchini fuzzy number, that is why it is gone down to 0.5. In the case of the fuzzy
[69] are used for seismic resilience evaluation. The conditional set moderate damage, the value is < 0.5, 0.25, 0 > , which means that
probabilities of the fragility curves are given in Table 1. The rst 0.5 and 0 are the boundary values of this triangular fuzzy number, that
column in Table 1 represents Intensity Measure (IM) of ground is why it is gone down to 0.5.
motions of the earthquake. The membership functions for each linguistic term in the linguistic
In order to estimate resilience, 10,000 earthquake scenarios have scale for residual functionality are:
been simulated for each dierent level of PGA and the conditional The residual functionality in the case of "no damage": MF1 (x; 1.0;
probabilities of fragility curves for evaluating the damage state of the 0);.
bridge for a given level of PGA are listed in Table 1. The random The residual functionality in the case of "slight damage":
number RN is generated and according to its value, the bridge is
1 0 . 75 x , 0. 5x0. 75
0 . 25
Table 1 MF2(x;0. 75; 0. 25)= 1 x 0 . 75 , 0. 75x1
Fragility analysis of the bridge in Santa Barbara. 0 . 25
0, otherwise
IM (g) Slight damage Moderate Extensive Complete damage
damage damage The residual functionality in the case of "moderate damage":
6
J.M. Andri, D.-G. Lu International Journal of Disaster Risk Reduction (xxxx) xxxxxxxx
The residual functionality in the case of "extensive damage": holds for "moderate", "extensive" and "complete" damage with dier-
ent values for recovery time.
1 0 . 1 x , 0x0. 1
0.1
The estimated triangular fuzzy numbers for membership functions
MF4 (x;0. 1; 0. 1)=1 x 0 . 1 , 0. 1x0. 2 of linguistic terms in the case of the linguistic scale for residual
0.1
functionality are:
0, otherwise
The residual functionality in the case of "complete damage": MF5 1. Recovery period in the case of "no damage": tr1= < 0,0,0>
(x; 0; 0). 2. Recovery period in the case of "slight damage":
In the case of no damage, it means that the bridge has stayed intact tr2= < 0.333,2.667,5.0>
after the earthquake since there is no any physical damage. Hence, 3. Recovery period in the case of "moderate damage":
there is no need to close any lane and the trac can ow without tr3= < 0.667,3.667,6.667>
interruption. Since all the lanes are functional, the residual function- 4. Recovery period in the case of "extensive damage":
ality is equal to full functionality 100% and it is represented with the tr4= < 2.0,5.167,8.333>
crisp number equal to 1.0. 5. Recovery period in the case of "complete damage":
Whereas in the case of complete damage, it means that it is tr5= < 2.5,6.25,10>
expected that the bridge will be collapsed according to HAZUS MR-4
Earthquake Technical Manual. Therefore, all the lanes must be closed The mathematical expressions for the membership functions of the
for trac in order not to endanger drivers and passengers. So, the recovery time taken from (DHS [51]) are:
residual functionality is null and it is represented with the crisp Recovery period in the case of "no damage": MF1 (x; 0; 0).
number 0. Recovery period in the case of "slight damage":
The target functionality is equal to 100% since it is assumed that the
bridge returns to its previous functionality before the earthquake. The 1 2 . 667 x , 0. 333x2. 667
2 . 333
target functionality can be presented as a single fuzzy number. The MF2 (x;2. 667; 2. 333)= x 2 . 667
1 2 . 333 , 2. 667x5
value of the target functionality is:
0, otherwise
7
J.M. Andri, D.-G. Lu International Journal of Disaster Risk Reduction (xxxx) xxxxxxxx
Recovery period in the case of "complete damage": Idle time interval in the case of "complete damage":
1 1 . 5 x , 1x1. 5
1 6 . 25 x , 2. 5x6. 25
0.5
3 . 75
MF5 (x;1. 5; 0. 5) = 1 x 1 . 5 , 1. 5x2
MF5 (x;6. 25; 3. 75)= 1 x 6 . 25 , 6. 25x10 0.5
3 . 75
0, otherwise
0, otherwise
Dec et al. [52] have assumed that the value of the idle time interval 4.5. Results and discussions
is between 1 and 2 months. In addition, they have used the same value
for every bridge damage level. In this case study, the values for the idle The seismic resilience index of the bridge in Santa Barbara is
time interval are assumed. The rst assumption is that the minimum obtained by performing a Monte Carlo Simulation (MCS). To compute
value of idle time interval is 40% of the minimum value of recovery the resilience and losses, the simulation is developed in the Java
time for the same damage state, while the maximum value represents programming language under NetBeans [72] and the graphs are
20% of the maximum value of recovery time, as shown in Fig. 5. plotted in Matlab [73]. As indicated previously, three levels of
The idle time interval values, which are assumed, for each damage disaster-preparedness of the community are considered for this bridge:
state of the bridge are: (1) well-prepared; (2) average-prepared; and (3) unprepared. The
simulation is run to generate a set of 10,000 samples for three dierent
1. Idle time interval in the case of "no damage": ti1= < 0,0,0> levels of community disaster-preparedness and dierent PGA (0.4g;
2. Idle time interval in the case of "slight damage": 0.5g; 0.6g; 0.7g).
ti2= < 0.133,0.566,1.0> The estimated fuzzy values of residual functionality, idle time
3. Idle time interval in the case of "moderate damage": period and recovery time for dierent ground motions are summarized
ti3= < 0.267,0.8,1.333> in Table 2. For each of these parameters, three dierent values are
4. Idle time interval in the case of "extensive damage": associated with three dierent views. For example, the expected
ti4= < 0.8,1.233,1.667> residual functionality of the bridge, for peak ground acceleration of
5. Idle time interval in the case of "complete damage": 0.7g is (0.73, 0.52, 0.31) where the value of 0.73 represents residual
ti5= < 1.0,1.5,2.0> functionality of the bridge for an optimistic view, the value of 0.52 is
the most-probable residual functionality and the value of 0.31 repre-
The mathematical expressions for membership functions for idle sents the pessimistic view of residual functionality.
time interval are: The predicted annual resilience values for dierent levels of
Idle time interval in the case of "no damage": MF1 (x; 0; 0). community preparedness and PGA are stated in Table 3. If the bridge
Idle time interval in the case of "slight damage": is located in a well-prepared community and an earthquake with a PGA
equal to 0.7g occurs, then the expected seismic resilience is equal to
1 0 . 566 x , 0. 133x0. 566
93.54%. However, the proposed method shows if the community is
0 . 433
MF2 (x;0. 566; 0. 433) = 1 x 0 . 566 , 0. 566x1 average-prepared for disasters, then the seismic resilience of the bridge
0 . 433 can be enhanced up to 5%, compared to the unprepared communities
0, otherwise
for ground motion intensity of 0.7g. In the case of the well-prepared
communities for disasters, the resilience of the bridge is enhanced up
Idle time interval in the case of "moderate damage": to 10% more than in case of the unprepared systems when the PGA is
0.7g.
1 0 . 8 x , 0. 267x0. 8
0 . 533
Fig. 6 displays the functionality of the bridge in Santa Barbara for a
MF3 (x;0. 8; 0. 533) = 1 x 0 . 8 , 0. 8x1. 333 period of one year after the earthquake struck. The estimated trian-
0 . 533
gular fuzzy numbers for membership functions of linguistic terms in
0, otherwise
the case of linguistic scale for residual functionality are represented
with the assumption that the bridge is located in the well-prepared
Idle time interval in the case of "extensive damage":
region which is exposed to the earthquake intensity measure of 0.7 g.
8
J.M. Andri, D.-G. Lu International Journal of Disaster Risk Reduction (xxxx) xxxxxxxx
Table 2 Section 4. The output result is a fuzzy number and it is obtained by the
Resilience parameters for three different views and different earthquake intensities. method of Incenter-of-centroids for the defuzzication process. The
results of the expected disaster losses are shown in Table 4. The unit
IM View Residual Idle time interval Recovery time
functionality (months/days) (months/days) cost of a bridge per square meter represents the cost of the construction
of a new bridge with the same structural characteristics. This data is
0.4g =1 0.94 0.09/3 0.22/7 taken from the ocial website of The State of Michigan [74] and for
=0.5 0.84 0.29/9 1.33/40
this case, the unit cost of construction of new bridge is $1620/m2. The
=0 0.73 0.49/15 2.43/73
discount factor is equal to 5%. Fig. 8 displays the change of direct
0.5g =1 0.86 0.16/5 0.40/12 economic losses towards the peak ground acceleration. Disaster losses
=0.5 0.70 0.47/14 2.15/65 and ground motion intensity change in the same order. If the ground
=0 0.55 0.78/24 3.89/117 motion intensity increases, then the disaster loss enhances.
0.6g =1 0.78 0.22/7 0.57/17
=0.5 0.59 0.61/18 2.80/84
=0 0.40 1.00/30 5.02/151 5. Conclusions
0.7g =1 0.73 0.27/8 0.68/20 In this research, the recovery process has been considered as a fuzzy
=0.5 0.52 0.71/21 3.23/97
process and it has been modeled by fuzzy set theory. The proposed
=0 0.31 1.16/35 5.78/173
model has been followed by an innovative resilience assessment model
in order to develop an ecient system for the decision-making process
Curve (=1) depicts the optimistic view, wherein the residual function- for disaster preparedness. The suggested model has been based on
ality is equal to 0.73. Hence, this curve begins at this point, and it is fuzzy set theory and fuzzy knowledge representation approach, and the
horizontal during an 8 day idle time. Moreover, the recovery phase fuzzy methodology has been used for the resilience assessment
follows a negative-exponential function since the bridge is located in a procedure. Such an approach has been shown as an eective tool to
well-prepared community and during the recovery period of 20 days deal with uncertainties that arise from a lack of data. The input of the
the bridge succeeds to return to its full functionality of 100%. After this, model is an expected earthquake PGA where the bridge is placed, and
the curve is straight. The curve for the most-probable case scenario the output is the performance of the bridge after the earthquake
(=0.5) has the residual functionality equal to 0.52, and it starts from regarding seismic resilience and disaster loss. Finally, the prediction of
this point. During a 22 days idle time, the functionality is constant. The resilience has its applications for decision-making processes for
recovery phase then follows the negative-exponential function during disaster-preparedness of a community in which the bridge has been
the recovery period of 97 days when the bridge achieves 100% placed. The decision on the measures which will be applied in order to
functionality. Curve (=0.5) is smoother than curve (=1), since the prepare the community for disasters is made using the predicted
recovery time is longer and the bridge returns more slowly to full resilience results. The conclusions can be drawn that the resilience of
recovery. Curve (=0) represents the pessimistic view. The residual the bridge can be enhanced up to 10% if it is located in well-prepared
functionality in this case scenario is 0.31 for a 35 day of idle time communities compared to the unprepared, according to the simulation
period. Also, the recovery phase follows the negative-exponential results in the case of the peak ground acceleration of 0.7g.
function and during the recovery period of 174 days the bridge Compared to the traditional probabilistic approach, the advantage
functionality from 0.31 gradually proceeds to 1. The curve (=0) is of the proposed fuzzy resilience model is that: (1) it is more practical
the smoothest since it has the longest recovery time. for quick resilience assessment and decision-making about the disaster
Fig. 7 illustrates the results for dierent settings of disaster preparedness; (2) it deals with imprecise, vague and fuzzy data which
preparedness indicated in Table 2, and it shows the results of annual includes lexical uncertainties in the model; and (3) it can provide
resilience for dierent ground motion values. It can be observed that resilience assessment of a bridge under multiple hazards. The above
the resilience for well-prepared communities is decreasing when PGA suggested model can be easily applied to other highway infrastructure
increases and it is decreasing slower than in the cases of unprepared components such as tunnels and roadways after the data about the
and average-prepared communities. Also, the resilience in the case of relationship between damage-residual functionality and recovery time
an unprepared community decreases the fastest, compared to the other are collected.
two communities when the intensity of ground motion increases. The focus of further research will be to treat a recovery process as a
The direct economic losses are calculated by using formula given in fuzzy-random process as well as to develop a model according to this
consideration which is proposed in the paper [75]. Such a model would
Table 3
Annual resilience of the bridge in Santa Barbara.
Disaster Preparedness View (Case scenario) Fuzzy Resilience Value (%) Crisp Resilience Value (%)
9
J.M. Andri, D.-G. Lu International Journal of Disaster Risk Reduction (xxxx) xxxxxxxx
Fig. 7. Expected annual resilience for ground motion intensity between 0.4g and 0.7g in the cases of dierent disaster preparedness of communities.
Table 4
Direct economic losses of the bridge in Santa Barbara.
View Direct economic loss in ($1000) Crisp direct economic loss in ($1000)
Acknowledgements
Disaster Loss
1600
The nancial support received from the National Science
Disaster Loss in $ 1000
1400
Foundation of China (Grant Nos. 51678209, 51378162), the
1200
Research Fund from the Ministry of Science and Technology of China
1000
(2013BAJ08B01), the Open Research Fund of State Key Laboratory for
800
600
Disaster Reduction in Civil Engineering (SLDRCE12-MB-04) is grate-
400 Well-prepred community
fully appreciated.
Average-prepared community
200
Unprepared community
0
References
0.35 0.45 0.55 0.65 0.75
Ground motion (g) [1] E.A. Gencer, Natural disasters, urban vulnerability, and risk management: a
theoretical overviewThe Interplay between Urban Development, Vulnerability, and
Fig. 8. Economic loss estimation for ground motion intensity between 0.4g and 0.7g.
Risk Management, Springer, 2013, pp. 743.
[2] ISDR, U. Hyogo framework for action 20052015: building the resilience of
emphasize the characteristics of uncertainties randomness and nations and communities to disasters. in Extract from the nal report of the World
Conference on Disaster Reduction (A/CONF. 206/6). 2005.
fuzziness. In addition, the corresponding model for resilience assess-
[3] L.A.T. Cox, Community resilience and decision theory challenges for catastrophic
ment will be outlined based on a fuzzy-probabilistic approach and the events, Risk Anal. 32 (11) (2012) 19191934.
resilience will be modeled as a fuzzy-random variable. [4] D.R. Godschalk, Urban hazard mitigation: creating resilient cities, Nat. Hazards
Rev. 4 (3) (2003) 136143.
[5] W. Krger, E. Zio, Vulnerable Systems, Springer Science & Business Media, 2011.
[6] G.P. Cimellaro, et al. The state of art of community resilience of physical
10
J.M. Andri, D.-G. Lu International Journal of Disaster Risk Reduction (xxxx) xxxxxxxx
infrastructures, in Structures Congress, 2011. [42] G. Chen, T.T. Pham, Introduction to Fuzzy Sets, Fuzzy Logic, and Fuzzy Control
[7] L.A. Dueas-Osorio, Interdependent response of networked systems to natural Systems, CRC press, 2000.
hazards and intentional disruptions, Citeseer, 2005. [43] J.-U. Klgel, Seismic hazard analysisQuo vadis?, Earth-Sci. Rev. 88 (1) (2008)
[8] M. Liu, D.M. Frangopol, Optimizing bridge network maintenance management 132.
under uncertainty with conicting criteria: Life-cycle maintenance, failure, and [44] L. Reiter, Earthquake Hazard Analysis: Issues and Insights, Columbia University
user costs, J. Struct. Eng. 132 (11) (2006) 18351845. Press, 1991.
[9] A. Rose, Dening and measuring economic resilience to disasters, Disaster Prev. [45] R.K. McGuire, Seismic Hazard and Risk Analysis, Earthquake engineering research
Manag.: Int. J. 13 (4) (2004) 307314. institute, 2004.
[10] G.P. Cimellaro, A.M. Reinhorn, M. Bruneau, Framework for analytical quantica- [46] A.F. Ares, A. Fatehi, Development of probabilistic seismic hazard analysis for
tion of disaster resilience, Eng. Struct. 32 (11) (2010) 36393649. international sites, challenges and guidelines, Nucl. Eng. Des. 259 (2013) 222229.
[11] M. Bruneau, et al., A framework to quantitatively assess and enhance the seismic [47] Applied Technology Council, Earthquake Damage Evaluation Data for California,
resilience of communities, Earthq. Spectra 19 (4) (2003) 733752. Applied Technology Council, 1985.
[12] M. Ouyang, et al., Vulnerability analysis of complementary transportation systems [48] A.A. Kiremidjian, N.I. Basz, Evaluation of bridge damage data from recent
with applications to railway and airline systems in China, Reliab. Eng. Syst. Saf. 142 earthquakesBulletin NCEER, US National Center for Earthquake Engineering
(2015) 248257. Research, 1997, pp. 17.
[13] M. Ouyang, L. Dueas-Osorio, X. Min, A three-stage resilience analysis framework [49] R.E. Melchers, Structural Reliability Analysis and Prediction, 2nd ed., JohnWiley
for urban infrastructure systems, Struct. Saf. 36 (2012) 2331. & Sons Ltd., West Sussex, England, 2001.
[14] M. Ouyang, L. Zhao, Do topological models contribute to decision making on post- [50] A. Karamlou, P. Bocchini, Computation of bridge seismic fragility by largescale
disaster electric power system restoration?, Chaos: Interdiscip. J. Nonlinear Sci. 24 simulation for probabilistic resilience analysis, Earthq. Eng. Struct. Dyn. 44 (12)
(4) (2014) 043131. (2015) 19591978.
[15] M. Ouyang, Z. Wang, Resilience assessment of interdependent infrastructure [51] HAZUS. MR4 earthquake model user manual. Whashington (DC): department of
systems: with a focus on joint restoration modeling and analysis, Reliab. Eng. Syst. Homeland Security 2009 Federal Emergency Management Agency, Mitigation
Saf. 141 (2015) 7482. Division.
[16] B.M. Ayyub, Systems resilience for multihazard environments: Denition, metrics, [52] A. Dec, P. Bocchini, D.M. Frangopol, A probabilistic approach for the prediction of
and valuation for decision making, Risk Anal. 34 (2) (2014) 340355. seismic resilience of bridges, Earthq. Eng. Struct. Dyn. 42 (10) (2013) 14691487.
[17] B.M. Ayyub, Practical resilience metrics for planning, design, and decision making, [53] B. Bede, Fuzzy Set-Theoretic OperationsMathematics of Fuzzy Sets and Fuzzy
ASCE-ASME J. Risk Uncertain. Eng. Syst., Part A: Civ. Eng. (2015) 04015008. Logic, Springer, 2013, pp. 1331.
[18] Y.-C. Kuo, S.-T. Lu, Using fuzzy multiple criteria decision making approach to [54] Building Seismic Safety Council. Prestandard and commentary for the seismic
enhance risk assessment for metropolitan construction projects, Int. J. Proj. rehabilitation of buildings. Report FEMA-356, Washington, DC, 2000.
Manag. 31 (4) (2013) 602614. [55] E. Jordan, A. Javernick-Will, Measuring community resilience and recovery: A
[19] L.A. Zadeh, Fuzzy logic, IEEE Computer 21 (4) (1988) 8393. content analysis of indicators. in Construction Research Congress, 2012.
[20] L.A. Zadeh, Fuzzy sets, Inf. Control 8 (3) (1965) 338353. [56] G.P. Smith, D. Wenger, Sustainable disaster recovery: operationalizing an existing
[21] J. Quelch, I.T. Cameron, Uncertainty representation and propagation in quantied agendaHandbook of disaster research, Springer, 2007, pp. 234257.
risk assessment using fuzzy sets, J. loss Prev. Process Ind. 7 (6) (1994) 463473. [57] C. Samantra, S. Datta, S.S. Mahapatra, Risk assessment in IT outsourcing using
[22] S. Bonvicini, P. Leonelli, G. Spadoni, Risk analysis of hazardous materials fuzzy decision-making approach: an Indian perspective, Expert Syst. Appl. 41 (8)
transportation: evaluating uncertainty by means of fuzzy logic, J. Hazard. Mater. 62 (2014) 40104022.
(1) (1998) 5974. [58] P. Bocchini, D.M. Frangopol, Restoration of bridge networks after an earthquake:
[23] V.J. Davidson, J. Ryks, A. Fazil, Fuzzy risk assessment tool for microbial hazards in multicriteria intervention optimization, Earthq. Spectra 28 (2) (2012) 426455.
food systems, Fuzzy Sets Syst. 157 (9) (2006) 12011210. [59] P. Bocchini, D.M. Frangopol, Optimal resilience-and cost-based postdisaster
[24] V. Kecman, Learning and soft computing: support vector machines, neural intervention prioritization for bridges along a highway segment, J. Bridge Eng. 17
networks, and fuzzy logic models, MIT press, 2001. (1) (2010) 117129.
[25] J. Buckley James, E. Eslami, An introduction to fuzzy logic and fuzzy sets, Physica- [60] M.H. Faber, M.G. Stewart, Risk assessment for civil engineering facilities: critical
Verlag, A Springer-Verlag Company, Heidelberg, 2005. overview and discussion, Reliab. Eng. Syst. Saf. 80 (2) (2003) 173184.
[26] S. Tesfamariam, M. Saatcioglu, Seismic vulnerability assessment of reinforced [61] C. Peek-Asa, et al., Fatal and hospitalized injuries resulting from the 1994
concrete buildings using hierarchical fuzzy rule base modeling, Earthq. Spectra 26 northridge earthquake, Int. J. Epidemiol. 27 (3) (1998) 459465.
(1) (2010) 235256. [62] California Department of Transportation, S.M.I., California Log of Bridges on State
[27] S. Tesfamariam, M. Saatcioglu, Risk-based seismic evaluation of reinforced Highways District, 5, 2014
concrete buildings, Earthq. Spectra 24 (3) (2008) 795821. [63] S.E. Chang, M. Shinozuka, J.E. Moore, Probabilistic earthquake scenarios: ex-
[28] M.L. Carreo, O.D. Cardona, A.H. Barbat, Computational tool for post-earthquake tending risk analysis methodologies to spatially distributed systems, Earthq.
evaluation of damage in buildings, Earthq. Spectra 26 (1) (2010) 6386. Spectra 16 (3) (2000) 557572.
[29] F. Colangelo, Probabilistic characterisation of an analytical fuzzy-random model for [64] Y. Zhou, S. Banerjee, M. Shinozuka, Socio-economic eect of seismic retrot of
seismic fragility computation, Struct. Saf. 40 (2013) 6877. bridges for highway transportation networks: a pilot study, Struct. Infrastruct. Eng.
[30] Barreras;, F.E., et al., Seismic study of the Urban Bridges of the city of Ensenada, 6 (12) (2010) 145157.
Baja California, Mexico, in: Proceedings of the 14th World Conference on [65] Y. Dong, D.M. Frangopol, D. Saydam, Sustainability of highway bridge networks
Earthquake Engineering: Beijing, 2008. under seismic hazard, J. Earthq. Eng. 18 (1) (2014) 4166.
[31] M. Lamarre, W. Dong, Evaluation of seismic hazard with fuzzy algorithm, Eng. [66] M. Shinozuka, et al., Eect of seismic retrot of bridges on transportation networks,
Geol. 20 (1986) V1. Earthq. Eng. Eng. Vib. 2 (2) (2003) 169179.
[32] J.R.G. Crdenas, et al., Social Aggravation Estimation to Seismic Hazard Using [67] B.G. Nielson, R. DesRoches, Analytical seismic fragility curves for typical bridges in
Classical Fuzzy MethodsSimulation and Modeling Methodologies, Technologies the central and southeastern United States, Earthq. Spectra 23 (3) (2007) 615633.
and Applications, Springer, 2015, pp. 275293. [68] Y. Dong, D.M. Frangopol, Risk and resilience assessment of bridges under
[33] A. Ahumada, A. Altunkaynak, A. Ayoub, Fuzzy logic-based attenuation relation- mainshock and aftershocks incorporating uncertainties, Eng. Struct. 83 (2015)
ships of strong motion earthquake records, Expert Syst. Appl. 42 (3) (2015) 198208.
12871297. [69] A. Karamlou, P. Bocchini, Computation of bridge seismic fragility by largescale
[34] I. Karimi, E. Hllermeier, Risk assessment system of natural hazards: a new simulation for probabilistic resilience analysis, Earth. Eng. Struct. Eng. 44 (12)
approach based on fuzzy probability, Fuzzy Sets Syst. 158 (9) (2007) 987999. (2015) 19591978.
[35] M.L. Carreo, O.D. Cardona, A.H. Barbat, New methodology for urban seismic risk [70] J.E. Padgett, R. DesRoches, Bridge functionality relationships for improved seismic
assessment from a holistic perspective, Bull. Earthq. Eng. 10 (2) (2012) 547565. risk assessment of transportation networks, Earthq. Spectra 23 (1) (2007)
[36] I. Karimi, K. Meskouris, Risk management of natural disasters: a fuzzy-probabil- 115130.
istic methodology and its application to seismic hazard, Fakultt fr [71] M. Shinozuka, et al., Socio-economic eect of seismic retrot implemented on
Bauingenieurwesen, 2006. bridges in the Los Angeles highway network, Citeseer, 2008.
[37] M.L. Carreo, O.D. Cardona, A.H. Barbat, A disaster risk management performance [72] T. Boudreau, et al., NetBeans: the denitive guide, O'Reilly Media, Inc, 2002.
index, Nat. Hazards 41 (1) (2007) 120. [73] MATLAB, Users Guide, The MathWorks, Inc.: Natick MA, USA, 2010.
[38] J.M. Andri, D.-G. Lu, Risk assessment of bridges under multiple hazards in [74] Website, S.o.M.s.O. 2015 [cited 2015 September 6th]; Available from: http://
operation period, Saf. Sci. 83 (2016) 8092. www.michigan.gov/documents/mdot/Bridge_Repair_Cost_Estimate_Worksheet_
[39] H. Sucuolu, S. Akkar, Basic Earthquake Engineering, Springer, 2014. 2015_480468_7.xls.
[40] M. Carreo, O. Cardona, A. Barbat, of Building Damage and Safety After an [75] J.M. Andri, D.-G. Lu, Seismic resilience of a bridge based on fuzzy-probabilistic
EarthquakeIntelligent Computational Paradigms in Earthquake Engineering, IGI approach, in: M. Papadrakakis, V. Papadopoulos, V. Plevris (eds.), Proceedings of
Global, 2007, pp. 123157. COMPDYN 2015, Crete Island, Greece, 2015.
[41] G. Chen, T.T. Pham, Introduction to Fuzzy Systems, CRC Press, 2005.
11