Professional Documents
Culture Documents
Safety Science
journal homepage: www.elsevier.com/locate/ssci
a r t i c l e i n f o a b s t r a c t
Article history: Over the past several decades, the number of disasters around the world has been amplified. Moreover,
Received 18 July 2015 the impacts of disasters on communities have attracted extra attentions, requiring efficient measures of
Received in revised form 14 October 2015 disaster management. Similarly, safety problems of bridges have become public topics since bridges rep-
Accepted 3 November 2015
resent essential parts of highway infrastructure that have been exposed to multiple hazards during their
life cycle. In this paper, a novel framework of disaster risk assessment is proposed, by combining Fuzzy
Analytical Hierarchy Process (FAHP) with fuzzy knowledge representation and fuzzy logic techniques
Keywords:
into a single integrated approach. A survey about collapsed bridges in the past few decades has been con-
Multiple hazards
Bridges
ducted. From this investigation the potential hazards have been identified. The FAHP approach is applied
Risk assessment to ranking risk factors since it is more systematic, accurate and effective than traditional AHP. Risk indi-
Bridge safety cators, the probability of occurrence, the impacts of disasters and disaster consequences are analyzed
Hierarchical risk breakdown structure from expert opinions. The bridge risk is computed by implementing fuzzy logic tools on the risk factors
Fuzzy AHP and their parameters. The proposed method is practical and efficient for a quick and reliable multi-
Vulnerability hazards risk analysis and assessment of bridges. Based on the level of disaster risks to which the bridge
is exposed, an appropriate bridge safety management plan is required to reduce or prevent the effects of
the disasters. A case study is examined for a bridge in Tianjin, People’s Republic of China. The results illus-
trate the applicability of the proposed risk assessment method of bridges under multiple hazards.
Ó 2015 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ssci.2015.11.001
0925-7535/Ó 2015 Elsevier Ltd. All rights reserved.
J.M. Andrić, D.-G. Lu / Safety Science 83 (2016) 80–92 81
Traffic hazards: ship collision with the bridge, vehicle collisions, ers to express their judgments in the form of the exact numeric
overloading. value of the 9-point scale. A decision maker feels more confident
Construction hazards: lack of design, construction quality, to express his or her opinion in the form of linguistic variables
deterioration. (Zheng et al., 2012).
Human-made hazards: fire, explosions, terrorist attacks. Therefore, fuzzy AHP is a more suitable method for evaluating
hazard prioritization for bridge risk assessment since it captures
Further, the collected risk factors are analyzed with the aim to imprecise judgements of decision makers (Das, 2011). This method
estimate the degree of their influences on the bridge structures. has been introduced by Van Laarhoven and Pedrycz (1983). Com-
pared to other MCDM methods, fuzzy AHP is more precise and effi-
cient in handling fuzziness of the data involved in measuring
2.2. Risk ranking
alternatives or judgements of different decision variables
(Sezhian et al., 2011). In this study, a method based on fuzzy
In different multi-criteria decision making problems, Analytical
AHP is preferred for risk ranking due to its advantage over the
Hierarchical Process (AHP), fuzzy AHP (FAHP), Analytical Network
other decision-making methods.
Process (ANP) and other similar methods have been employed for
solving these kinds of problems.
The AHP approach has been shown as a practical method for 2.3. Risk analysis
multi-criteria decision making. The decision problem in AHP is
decomposed into a break-down structure of interrelated compo- Risk analysis is the key process in risk assessment and safety
nents and it is organized hierarchically in order to determine the management and it embraces three points: (1) the likelihood of a
significance of each criteria (Vashishtha and Ramachandran, disaster occurrence (hazard analysis); (2) the impact of the disaster
2006). The AHP approach is developed by Saaty (1980) and it (vulnerability analysis); (3) the consequence of disasters (loss
outlines a break-down structure of goal, factors and sub-factors analysis).
in a hierarchical configuration. In addition, this method is shown Hazard analysis indicates the probability or likelihood of disas-
efficient for solving a variety of problems in IT project prioritiza- ter occurrence at a particular site of the bridge location. In general,
tion (da Silva Neves and Camanho, 2015); electric supply planning it can be assessed by qualitative or quantitative analysis. The qual-
(Rojas-Zerpa and Yusta, 2015); machinery safety (Caputo et al., itative analysis evaluates hazards based on subjective judgments of
2013); maintenance (Salmeron and Lopez, 2010). On the other experts. The quantitative analysis appraises the quantitative mea-
hand, the weakness of traditional AHP is that it doesn’t consider sures of hazards. Usually, the hazard analysis is carried out by
ambiguity which is entailed in human subjective judgement quantitative methods such as a probabilistic approach. Moreover,
(Mangla et al., 2015) since it appears inadequate for decision mak- the probabilistic theory is based on a series of experiments and
1951 – 1988
35
30
25
20
15
10
5
0
Fig. 2. Collapsed bridges in the USA during period 1951–1988 (Harik et al., 1990).
160
140
120
100
80
60
40
20
0
Fig. 3. Collapsed bridges in the USA during period 1989–2000 (Wardhana and Hadipriono, 2003).
J.M. Andrić, D.-G. Lu / Safety Science 83 (2016) 80–92 83
Fig. 4. Collapsed bridges in Columbia during period 1986–2001 (Diaz et al., 2009).
70
60
50
40
30
20
10
0
Flood Construction Overload Collision Design Construction Debris
scheme quality
Fig. 5. Collapsed bridges in China during period 2000–2012 (Fu et al., 2013).
statistical data. However, disaster risk assessment in this research The direct social losses have been expressed in the number
is carried out based on subjective judgments of experts. Further, of dead and injured people who are on the bridge during a
risk analysis and assessment based on subjective judgments, which disaster. This number varies on the day time when the
includes individual experience and knowledge, is a more efficient disaster occurs and the age of people (Peek-Asa et al., 1998).
tool than the probabilistic approach according to Kuo and Lu During the light day, the number of victims is higher than at
(2013). night time.
The vulnerability analysis of bridges outlines the impact of a The indirect social losses have been associated with the number
disaster, expressed as bridge damage state. Additionally, the dam- of dead and injured people because of traffic disruption due to
age state is related to a damage ratio, which is defined as the bridge dysfunction.
reconstruction costs to the replacement cost. According to (DHS,
2009), five different damage states for bridges are defined: no 2.4. Risk assessment
damage, slight, moderate, extensive, and complete damage.
Disaster losses, or consequences of disasters, have been mea- Risk is the possibility of harmful consequences or expected
sured in terms of their influence on communities, environment losses resulting from interactions between natural and human-
and people. Disaster losses involve four aspects: direct economic, made hazards and vulnerable conditions. In general, disaster risk
direct social, indirect economic and indirect social losses represents a multiplication function of three parameters: hazard
(Cimellaro et al., 2010). (H) – the probability of disaster occurrence; vulnerability (V) –
Direct economic losses comprise bridge repair, rehabilitation the impact of disaster; and loss (L) – the consequences of the dis-
and reconstruction costs, and cost of bridge equipment replace- aster. Mathematically, disaster risk is expressed by following
ment, debris removal and temporary bypass construction. Such equation:
losses have been estimated for each structural bridge component
(foundations, bearings, piers, beams, girders). In addition, costs of R¼HV L ð1Þ
non-structural bridge elements, such as lightning, bridge drainage,
traffic and signaling installation, grounding, monitoring, and road-
way heating systems, have been included. 3. Hierarchical risk breakdown structure of bridges
Indirect economic losses have been caused by traffic disruption.
Bridge users have been required to select other routes since the In this study, 15 possible risk factors are listed, which could
functionality of a bridge is reduced or closed for traffic. Besides, a contribute to bridge collapse. Additionally, these identified poten-
new route will enhance travel time and distance which will result tial hazards have been arranged hierarchically in order to form
in higher travel expenses. To estimate this loss, information on Hierarchical Risk Breakdown Structure (HRBS). HRBS has been
traffic flow before and after the disaster, new total travel time introduced by Tah and Carr (2000) with the purpose to enable sys-
and new distance are required. tematic detailed risk analyses and ranking.
84 J.M. Andrić, D.-G. Lu / Safety Science 83 (2016) 80–92
In this paper, potential hazards have been organized and classi- e is constructed according to the pair-wise
judgment matrix X
fied in the HRBS referring to their origin. Further, the hierarchy comparison:
entails three levels. On the first level is the goal of the problem 2 3
which is the total bridge risk. The goal, bridge risk, has been
~x11 . . . ~x1n
impacted by two hazard categories. Therefore, the bridge risk has e ¼6
X 4 ... ...
7
... 5 ð2Þ
been decomposed into two factors: natural and man-made haz- ~xn1 . . . ~xnn
ards. They have been placed in the second level. The most relevant
sub-criteria for each of the criteria have been put in the third level. where ~xij is the importance of factor i over the factor j expressed as
Earthquakes, tsunamis, hurricanes, floods, debris, scour, ice, soil fuzzy judgment. Further, in matrix X e element ~xji represents the
and the age of the bridge belong to category of natural hazards reciprocal value of element ~xij .
(Fig. 6). Such risk factors are more unpredictable and uncontrol-
1
lable as their origin is in nature. The other category consists of col- ~xji ¼ ð3Þ
~xij
lision, overloading, deterioration, construction and design, fire and
terrorist attack. Human-made hazards are more controllable com- Many researchers proposed different linguistic scales for mea-
pared to the first group. suring the verbal judgment of importance. Usually, linguistic vari-
ables for expressing preference of alternatives of decision groups
have triangular or trapezoidal fuzzy numbers as membership func-
4. Fuzzy AHP-based method for disaster risk ranking of bridges tions (Sadi-Nezhad and Damghani, 2010). Further, definition and
operational laws of fuzzy numbers are provided. Triangular fuzzy
In order to estimate hazard weights and perform risk ranking, number represents m ~ ¼ ða; b; cÞ, in which the membership function
a method based on the fuzzy AHP concept is proposed. The pro- lm~ of m
~ is defined by:
posed procedure uses opinions of decision makers to estimate 8 xa
risk rankings. Every decision maker has been asked to express < ba ; a 6 x 6 b
>
their judgment on hazard importance compared to other hazards lm~ ðmÞ ¼ 1; x ¼ b ð4Þ
>
: cx
in the form of set of questionnaires. The importance of each cb
; b6x6c
hazard is denoted by fuzzy knowledge representation technique,
and fuzzy judgement matrices have been formed. Furthermore, The values a, b and c represent lower, medium and upper bound of
fuzzy number m, ~ respectively.
fuzzy weights have been computed. Afterwards, fuzzy weights
from different decision makers have been aggregated to form a In fuzzy set theory, the operational laws of two triangular fuzzy
single value. The aggregation process is carried out by aggrega- numbers a ~ ¼ ðb1 ; b2 ; b3 Þ are given as follows
~ ¼ ða1 ; a2 ; a3 Þ and b
tion operator. At the end, fuzzy weight values are converted into (Chen and Chen, 2003):
numerical values which will be used in process of disaster risk ~ ¼ ða1 ; a2 ; a3 ÞðþÞðb1 ; b2 ; b3 Þ ¼ ða1 þ b1 ; a2 þ b2 ; a3 þ b3 Þ
~ðþÞb
a ð5Þ
assessment.
The process of risk ranking consists of seven steps which have
a ~ ¼ ða1 ; a2 ; a3 ÞðÞðb1 ; b2 ; b3 Þ ¼ ða1 b1 ; a2 b2 ; a3 b3 Þ
~ðÞb ð6Þ
been outlined below:
Step 1: Linguistic scale for evaluation of fuzzy judgment matrix:
a ~ ¼ ða1 ; a2 ; a3 ÞðÞðb1 ; b2 ; b3 Þ ¼ ða1 b1 ; a2 b2 ; a3 b3 Þ
~ðÞb ð7Þ
The relationships between goal and factors, and between factors
and sub-factors in traditional AHP approach, are established by
judgement matrix or pairwise comparison matrix X, constructed ~ ¼ ða1 ; a2 ; a3 Þð=Þðb1 ; b2 ; b3 Þ ¼ a1 a2 a3
~ð=Þb
a ; ; ð8Þ
for the factors on the same level. An element in row i and column b3 b2 b1
j represents the judged value of attribute i over the attribute j. The
values of verbal judgments in matrix X in traditional AHP are given ~ ¼ k ða1 ; a2 ; a3 Þ ¼ ðka1 ; ka2 ; ka3 Þ
ka ð9Þ
as crisp numerical values which are evaluated on 9-point scale
(Table 1). In fuzzy AHP, a linguistic scale with linguistic variables 1 1 1 1
¼ ða1 ; a2 ; a3 Þ1 ¼ ; ; ð10Þ
is used to compare the importance of sub-factors and fuzzy ~
a a3 a2 a1
Table 1 !1n
Y
n
Saaty’s scale of relative importance of attributes i over attribute j. v~ i ¼ ~xij ð14Þ
Verbal judgment of importance Value j¼1
according to their knowledge about the nature of the problem. here, k – is the total number of fuzzy sets in linguistic scale, Yi is the
Similarly, the linguistic scale is chosen according to the type of center of i-th fuzzy set, and lðyi Þ is the membership function of i-th
the fuzzy numbers which is suitable to describe and present partic- matching fuzzy set (Zeng et al., 2007).
ular linguistic variables for a certain problem. Step 8. Classification of bridge risk: The bridge risk can be classi-
Step 2. Linguistic data collection for hazard likelihood, bridge vul- fied in different risk categories and it is carried out based on crisp
nerability and disaster loss: Risk indicators have been evaluated risk ratings. The maximum risk range is determined based on the
for every risk factor by experts (Zeng et al., 2007). Expert judg- highest risk indicator rating in linguistic scale. Moreover, the like-
ments are provided in linguistic terms that have been defined in lihood of occurrence, the impact of disaster and the consequences
the previous step since linguistic terms are used to define logical of disaster are multiplied, and the fuzzy risk result in terms of crisp
judgments (Ölçer and Odabasßi, 2005). In this case, the risk param- score becomes risk rating. The crisp scores are calculated using
eters, the probability of disaster occurrence, bridge vulnerability ‘Center-of-gravity’ from previous step. Finally, the crisp risks have
and disaster consequences have been estimated by experts in been defined in different crisp risk ranges. A safety and manage-
bridge engineering. Their subjective judgment is provided through ment plan for the bridge can be recommended on the basis of risk
surveys and questionnaires about risk indicators for every hazard. level.
Step 3. Fuzzification: Fuzzification is a process which converts
values of linguistic variables into corresponding fuzzy sets using 6. Case study: A bridge in Tianjin, China
linguistic scale. Experts provide their opinions about risk parame-
ters using linguistic variables. In this step, the collected data about 6.1. Description of the bridge in Tianjin, China
the risk parameters for every hazard which have been expressed as
linguistic variables is converted into corresponding fuzzy sets. A single-tower cable stayed bridge in the downtown of Tianjin
Step 4. Fuzzy aggregation of expert opinions: Aggregation is a pro- in China, across over the Haihe River, is investigated for a case
cess in fuzzy logic that combines expert judgements and individual study (Fig. 7). The bridge was designed with three spans arrange-
opinions on risk indicators in order to form a single combined pref- ments: 145 + 48 + 42 m with the total length of 235 m; and the
erence fuzzy set. In this study, expert judgments are weighted width of the bridge is 15 m. The main steel box girder passes
equally. To aggregate their judgment on risk indicators, the follow- through single tower which is 78 m in height and 75° tilted. On
ing equations are applied (Chen, 2000): the deck, traffic is functioning in six lanes, three lanes in each
direction.
1X n
FLi ¼ FLij ð18Þ
n j¼1 6.2. Procedure of risk prioritization for the bridge in Tianjin
1X n
FIi ¼ FIij ð19Þ Step 1: Linguistic scale for evaluation of fuzzy judgment matrix:
n j¼1
The linguistic scale of relative importance for pair-wise compar-
1X n
ison with a triangular fuzzy numbers adopted from Cheng
FC i ¼ FC ij ð20Þ (1997) is used to measure subjective judgements of experts.
n j¼1
This scale is given in Table 3, and it is shown in Fig. 8.
where i-th risk factor; j-th expert; and n – the number of experts; Step 2: Data collection for establishment of pair-wise comparison
FLi – is the fuzzy likelihood of occurrence of i-th risk factor; FIi – matrices: Three set of questionnaires have been established in
is the fuzzy disaster impact of i-th risk factor; FCi – is the fuzzy con- order to collect data from decision makers. The first set of ques-
sequences of disaster of i-th risk factor. tions is about the comparison of the importance of natural haz-
Step 5. Compute fuzzy risk factor: The multiplication operator is ards; the second set of questions is about the comparison of the
applied on the fuzzy likelihood of occurrence (FLi), fuzzy impact importance of man-made hazards, and the third is about the
of disaster (FIi) and fuzzy consequence of disaster (FCi) in order importance of natural hazards over the man-made hazards. In
to obtain fuzzy risk factor (FRFi). The proposed aggregated equation this survey, three members from the Department of Disaster
for calculating i-th fuzzy risk factor (FRFi) corresponding to i-th Reduction and Mitigation at Harbin Institute of Technology
hazard is: have been asked to give their opinion on the basis of their
knowledge and expertise for each sub-factor on the third level
FRF i ¼ FLi ðÞFIi ðÞFC i ð21Þ and factors on the second level in the hierarchy. The pair-wise
Step 6. Compute fuzzy risk index: The overall fuzzy risk index (FR) comparisons are made according to Table 3. Furthermore, judg-
is computed by ment matrixes for Level 2 and Level 3 are assembled for each
X
m
FR ¼ W i ðÞFRF i ð22Þ
i¼1
Table 3 Table 4
Saaty’s scale of relative importance used in the pair-wise comparison of fuzzy AHP. Computed weights of Level 2 appraisal factors by DM1.
Fig. 8. Linguistic scale of relative importance used in the pair-wise comparison of fuzzy AHP.
88 J.M. Andrić, D.-G. Lu / Safety Science 83 (2016) 80–92
Table 5
Computed weights of Level 3 appraisal factors by DM1.
b1 b2 b3 b4 b5 b6 b7 b8 b9 B1
b1 (1, 1, 3) (1, 1, 3) (7, 9, 9) (1, 3, 5) (1, 3, 5) (1, 3, 5) (1, 1, 3) (7, 9, 9) (1, 1, 3) (0.07, 0.21, 0.71)
1
b2 ; 1; 1 (1, 1, 3) (7, 9, 9) (1, 3, 5) (1, 3, 5) (1, 1, 3) (1, 3, 5) (7, 9, 9) (1, 1, 3) (0.06, 0.21, 0.63)
31 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
b3 ; ; ; ; (1, 1, 3) 7;5;3 7;5;3 9;7;5 5;3;1 5;3;1 ; ;1 (0.01, 0.02, 0.07)
91 91 7 91 91 7 51 31
b4 ; ;1 5;3;1
(3, 5, 7) (1, 1, 3) (1, 3, 5) (1, 1, 3) (1, 1, 3) (1, 1, 3) 5;3;1
(0.03, 0.08, 0.38)
51 31 1 1 1 1 1 1 1 1 1 1
5;3;1 ; ;1 ; ;1 5;3;1 5;3;1 5;3;1
b5 (3, 5, 7) (1, 1, 3) (1, 3, 5) (0.02, 0.06, 0.26)
1 1 51 3 51 3
b6 ; ;1 3 ; 1; 1
(5, 7, 9) 3 ; 1; 1
(1, 3, 5) (1, 1, 3) (1, 1, 3) (1, 1, 3) (1, 1, 3) (0.04, 0.11, 0.39)
51 3 1 1 1 1
3 ; 1; 1 ; ;1 ; 1; 1 ; 1; 1
b7 (1, 3, 5) (1, 3, 5) (1, 1, 3) (1, 3, 5) (1, 1, 3) (0.03, 0.11, 0.34)
1 1 1 51 31 1 31 1 1 31 1 1 1 1
b8 ; ; 9;9;7
(1, 3, 5) 3 ; 1; 1 5;3;1 ; 1; 1 ; ;1 (1,1, 3) 5;3;1
(0.01, 0.04, 0.14)
91 9 7 1 31 51 3
3 ; 1; 1 3 ; 1; 1 ; 3 ; 1; 1
b9 (1, 3, 5) (1, 3, 5) (1, 3, 5) 3 1; 1 (1, 3, 5) (1,1, 3) (0.03, 0.15, 0.36)
Consistency check: kmax = 9.89; CI = (9.89 9)/8 = 0.111; CR = CI/RI = 0.111/1.45 = 0.077 < 0.1.
Table 6
Computed weights of Level 3 appraisal factors by DM1.
Consistency check: kmax = 6.36; CI = (6.36 6)/5 = 0.072; CR = CI/RI = 0.072/1.24 = 0.058 < 0.1.
Table 7
Computed weights of Level 2 appraisal factors by DM2.
highest risk rating assigned to a risk in linguistic scale (Fig. 9
and Table 14). In Table 19, crisp risk ratings for linguistic risk
a1 a2 A parametric scale with reference to Table 14 are presented. Five
1
a1 (1, 1, 3) 5;3;1
1 (0.08, 0.25, 1.195) different crisp risk values have been computed by defuzzifica-
a2 (1, 3, 5) (1, 1, 3) (0.178, 0.75, 2.676) tion process using ‘Center-of-Gravity’ method and correspond-
ing to such crisp values, four possible risk categories (Risk
from linguistic scale and the membership degree of bridge Category 1–4) are defined with a range of (0.11–0.74). The high-
fuzzy risk belonging to these fuzzy sets are: est risk rating is 0.74 assigned to risk factors and the lowest risk
rating is 0.11. Further, risk categories describe four risk levels:
Very Low: l ¼ 0:77. Low Risk (LR), Moderate Risk (MR), High Risk (HR), and Very
Low: l ¼ 0:64. High Risk (VHR) as it is summarized in Table 20. In addition,
bridge safety management measures have been prescribed for
The procedure is illustrated in Fig. 10. every risk range.
The bridge fuzzy risk is converted into crisp risk value by Specific bridge safety management measures, in which infor-
‘Center-of-gravity’ method according to Eq. (23): mation technology is included, are proposed based on the assessed
bridge risk level:
0:1 0:77 þ 0:25 0:64
R¼ ¼ 0:17
0:77 þ 0:64 1. Bridge Management System (BMS) is used in case of moderate
Step 8. Classification of bridge risk: The bridge risk can be catego- bridge risk level. BMS integrates four basic modules: inventory;
rized in different risk categories depending on the bridge risk technical condition and safety; serviceability and operation;
level. Different risk categories are divided based on the risk and planning and budget (Bien, 1999; Bień, 2000). When bridge
ratings ranges. The maximum range has been decided on the exceptional conditions are noticed by multi-dimensional
Table 8
Computed weights of Level 3 appraisal factors by DM2.
b1 b2 b3 b4 b5 b6 b7 b8 b9 B1
b1 (1, 1, 3) (7, 7, 9) (7, 7, 9) (5, 7, 9) (7, 7, 9) (1, 3, 5) (3, 5, 7) (3, 5, 7) (1, 3, 5) (0.12, 0.32, 0.69)
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
b2 ; ; (1, 1, 3) (1, 1, 3) ; ;1 ; ;1 ; ; ; ; ; ; ; ; (0.01, 0.02, 0.05)
91 91 71 1 51 31 51 31 91 71 51 91 71 51 71 51 31 91 91 71
b3 ; ; 3 ; 1; 1
(1, 1, 3) 5;3;1 5;3;1 ; ; ; ; ; ; ; ; (0.01, 0.02, 0.05)
91 91 71 91 71 51 91 71 51 71 51 31 91 91 71
9;7;5 7;5;3 7;5;3 7;5;3 9;7;5
b4 (1, 3, 5) (1, 3, 5) (1, 1, 3) (1, 3, 5) (0.01, 0.04, 0.10)
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
b5 ; ; (1, 3, 5) (1, 3, 5) 5;3;1
(1, 1, 3) 9;7;5 7;5;3 7;5;3 ; ; (0.01, 0.03, 0.07)
91 91 7 91 91 7
5;3;1 5;3;1
b6 (5, 7, 9) (5, 7, 9) (3, 5, 7) (5, 7, 9) (1, 1, 3) (1, 3, 5) (1, 3, 5) (0.06, 0.15, 0.42)
1 1 1 1 1 1 1 1 1 1
b7 ; ; (5, 7, 9) (5, 7, 9) (3, 5, 7) (3, 5, 7) ; ;1 (1, 1, 3) 5;3;1 ; ; (0.03, 0.08, 0.23)
71 51 31 51 31 71 51 31
7;5;3 5;3;1 7;5;3
b8 (3, 5, 7) (3, 5, 7) (3, 5, 7) (3, 5, 7) (3, 5, 7) (1, 1, 3) (0.04, 0.10, 0.27)
1 1
5;3;1
b9 (7, 7, 9) (7, 7, 9) (5, 7, 9) (7, 7, 9) (1, 3, 5) (3, 5, 7) (3, 5, 7) (1, 1, 3) (0.10, 0.25, 0.58)
Consistency check: kmax = 10.03; CI = (10.03 9)/8 = 0.1287; CR = CI/RI = 0.1287/1.45 = 0.089 < 0.1.
J.M. Andrić, D.-G. Lu / Safety Science 83 (2016) 80–92 89
Table 9
Computed weights of Level 3 appraisal factors by DM2.
Consistency check: kmax = 6.51; CI = (6.51 6)/5 = 0.102; CR = CI/RI = 0.102/1.24 = 0.082 < 0.1.
Table 11
Computed weights of Level 3 appraisal factors by DM3.
b1 b2 b3 b4 b5 b6 b7 b8 b9 B1
b1 (1, 1, 3) (5, 7, 9) (5, 7, 9) (1, 3, 5) (5, 7, 9) (3, 5, 7) (5, 7, 9) (3, 5, 7) (1, 3, 5) (0.11, 0.33, 0.88)
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
b2 ; ; (1, 1, 3) (1, 1, 3) ; ; (1, 3, 5) ; ; ; ;1 ; ;1 ; ; (0.01, 0.03, 0.12)
91 71 51 1 71 51 3 71 51 3 51 31 51 31 71 51 31
b3 ; ; 3 ; 1; 1
(1, 1, 3) 5;3;1
(3, 5, 7) 5;3;1 5;3;1 5;3;1 ; ; (0.01, 0.04, 0.14)
91 71 5 71 51 3
5;3;1 5;3;1
b4 (3, 5, 7) (1, 3, 5) (1, 1, 3) (3, 5, 7) (1, 3, 5) (1, 3, 5) (1, 3, 5) (0.04, 0.14, 0.47)
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
b5 ; ; 5;3;1 7;5;3 ; ; (1, 1, 3) 5;3;1 7;5;3 5;3;1 ; ; (0.01, 0.02, 0.07)
91 71 51 71 51 3 91 71 51
7;5;3 5;3;1 7;5;3
b6 (3, 5, 7) (1, 3, 5) (1, 3, 5) (1, 1, 3) (1, 3, 5) (1, 1, 3) (0.03, 0.08, 0.28)
1 1 1 1 1 1 1 1 1 1 1 1
b7 ; ; (1, 3, 5) (1, 3, 5) ; ;1 (3, 5, 7) ; ;1 (1, 1, 3) 5;3;1 ; ; (0.02, 0.05, 0.19)
91 71 51 51 31 51 3 91 71 5
7;5;3 5;3;1 3 ; 1; 1 5;3;1
b8 (1, 3, 5) (1, 3, 5) (1, 3, 5) (1, 3, 5) (1, 1, 3) (0.02, 0.08, 0.27)
1 1
5;3;1
b9 (3, 5, 7) (3, 5, 7) (1, 3, 5) (5, 7, 9) (3, 5, 7) (5, 7, 9) (1, 3, 5) (1, 1, 3) (0.07, 0.23, 0.67)
Consistency check: kmax = 9.964; CI = (9.964 9)/8 = 0.121; CR = CI/RI = 0.121/1.45 = 0.083 < 0.1.
Table 12
Computed weights of Level 3 appraisal factors by DM3.
Consistency check: kmax = 6.409; CI = (6.409 6)/5 = 0.082; CR = CI/RI = 0.082/1.24 = 0.066 < 0.1.
90 J.M. Andrić, D.-G. Lu / Safety Science 83 (2016) 80–92
Table 13 Table 16
Fuzzy, crisp and normalized risk rank. Impacts of disasters assigned by experts in linguistic terms.
Table 14
Table 17
Linguistic classification of grades of risk factors.
Consequences of disasters assigned by experts in linguistic terms.
Hazards (H) Vulnerability (V) Losses (L) Triangular fuzzy numbers
Hazards E1 E2 E3 E4 E5
Very rare No damage Very small (0, 0, 0.3)
Earthquake B VS VB VB S
Rare Slight damage Small (0, 0.25, 0.5)
Tsunami VS VS M VB M
Moderate Moderate damage Middle (0.3, 0.5, 0.7)
Hurricane M VS M S S
Frequent Extreme damage Big (0.5, 0.75, 1)
Flood M VS M M S
Very frequent Complete damage Very big (0.75, 1, 1)
Debris VS VS B VS VS
Scour B VS S S S
Ice S S S VS VS
Table 15 Soil M S S S S
Likelihoods of hazards assigned by experts in linguistic variables. The age of the bridge M M M S M
Collision B VS B S S
Hazards E1 E2 E3 E4 E5 Overloading VB M M B S
Earthquake F VR R R R Deterioration B M M M M
Tsunami VR VR R R VR C&D M M M VS B
Hurricane R R VR M VR Fire M VS B B S
Flood M R VR R M TA VS VS VB B M
Debris VR VR R R R
Scour M R M M M
Ice M VR R M R
AHP, fuzzy knowledge representation and fuzzy logic into an inte-
Soil R VR M M R
The age of the bridge M M F R VR grated approach for disaster risk assessment of a bridge. Firstly, 15
Collision M VR R M VR potential hazards have been identified from the literature review
Overloading F VF F F VF involving collapsed bridges worldwide during different time
Deterioration F M F M F
periods (the USA, China, and Colombia). Further, the fuzzy AHP
C&D M VR M R F
Fire R VR R R VR
procedure is applied to evaluate risk weights in order to rank risk
TA VR VR VR R VR factors. The fuzzy set theory is suitable to express expert’s logic
opinion about risk indicators. In addition, risk factors are identified
7. Conclusions as linguistic values with corresponding membership functions.
Moreover, risk factors are presented as triangular fuzzy
In this research, a new method is developed in order to assess numbers. According to the crisp risk rating, the optimal strategies
bridge risk level and provide corresponding measures for bridge and plans for bridge safety and disaster management may be
safety management. The proposed technique combines fuzzy selected.
Table 18
Aggregated fuzzy risk’s parameters and fuzzy degree of risk factors for every hazard.
Hazards Hazard probability Bridge vulnerability Disaster consequences Fuzzy degree of risk factors
Earthquake (0.1, 0.3, 0.5) (0.36, 0.6, 0.78) (0.38, 0.6, 0.76) (0.014, 0.108, 0.296)
Tsunami (0, 0.1, 0.38) (0.36, 0.55, 0.74) (0.26, 0.4, 0.6) (0, 0.022, 0.169)
Hurricane (0.06, 0.2, 0.46) (0.12, 0.35, 0.58) (0.12, 0.3, 0.54) (0.001, 0.021, 0.144)
Flood (0.12, 0.3, 0.54) (0.28, 0.5, 0.72) (0.18, 0.3, 0.54) (0.006, 0.045, 0.210)
Debris (0, 0.15, 0.42) (0, 0.15, 0.42) (0.1, 0.15, 0.44) (0, 0.003, 0.078)
Scour (0.24, 0.45, 0.66) (0.2, 0.45, 0.7) (0.1, 0.3, 0.56) (0.004, 0.061, 0.259)
Ice (0.12, 0.3, 0.54) (0.06, 0.2, 0.46) (0, 0.15, 0.42) (0, 0.009, 0.104)
Soil (0.12, 0.3, 0.54) (0.06, 0.3, 0.54) (0.06, 0.3, 0.54) (0, 0.027, 0.157)
The age of the bridge (0.22, 0.4, 0.64) (0.24, 0.45, 0.66) (0.24, 0.45, 0.66) (0.013, 0.081, 0.279)
Collision (0.12, 0.25, 0.5) (0.36, 0.6, 0.84) (0.2, 0.4, 0.66) (0.008, 0.06, 0.277)
Overloading (0.58, 0.85, 1) (0.32, 0.55, 0.78) (0.36, 0.6, 0.78) (0.067, 0.281, 0.608)
Deterioration (0.42, 0.65, 0.88) (0.22, 0.45, 0.68) (0.34, 0.55, 0.76) (0.064, 0.161, 0.455)
C&D (0.22, 0.4, 0.64) (0.28, 0.45, 0.68) (0.28, 0.45, 0.68) (0.017, 0.081, 0.296)
Fire (0, 0.15, 0.42) (0.36, 0.6, 0.84) (0.26, 0.45, 0.7) (0, 0.041, 0.247)
TA (0, 0.05, 0.34) (0.34, 0.6, 0.8) (0.3, 0.45, 0.66) (0, 0.014, 0.179)
Table 19
Crisp risk levels.
Likelihood (H) Impact of disaster (V) Consequences (L) Fuzzy risk rating (H V L) Crisp risk (rating)
Very rare No damage Very small (0, 0, 0.027) 0.11
Rare Slight damage Small (0, 0.016, 0.125) 0.14
Moderate Moderate damage Middle (0.027, 0.125, 0.343) 0.20
Frequent Extensive damage Big (0.125, 0.422, 1) 0.50
Very frequent Complete damage Very big (0.343, 1, 1) 0.74
Table 20
Crisp risk ranges and categories.
The following conclusions can be drawn from this research: (3) Fuzzy logic has presented as more practical compared to the
traditional probabilistic approach as it has been verified in a
(1) The case study has demonstrated the applicability of the case study. Also, lexical uncertainties that come from impre-
proposed method for disaster risk assessment in bridge engi- cision, vague and fuzzy data have been included in risk
neering. Further, it has been an effective tool for quick disas- assessment process.
ter risk analysis and decision-making for optimal bridge (4) The proposed fuzzy logic-based method could be applied for
safety management. risk assessment in other industry or engineering areas. For
(2) Fuzzy AHP approach has shown as a more effective method example, it could be applied in construction project risk
for risk ranking compared to traditional AHP since uncer- management. Firstly, the HRBS should be formed from the
tainties from vague and fuzzy data have been considered possible risk factors which could occur during the construc-
in risk ordering process. tion period. Further, ranking procedure should be applied to
92 J.M. Andrić, D.-G. Lu / Safety Science 83 (2016) 80–92
risk factors. Finally, the total risk of construction project Giangarra, P.P. et al., 2011. Smarter bridges through advanced structural health
monitoring. IBM J. Res. Develop. 55 (12), 1–9, 9: 10.
could be evaluated using the fuzzy logic-based method, in
Godschalk, D.R., 2003. Urban hazard mitigation: creating resilient cities. Nat.
which the data about the risk indicators, the probability Hazards Rev. 4 (3), 136–143.
and consequences should be collected from construction Goerlandt, F. et al., 2015. A risk-informed ship collision alert system: framework
project managers. and application. Safety Sci. 77, 182–204.
Harik, I. et al., 1990. United States bridge failures 1951–1988. J. Perform. Constr.
Facil.
Harms, T. et al., 2010. Structural health monitoring of bridges using wireless sensor
Acknowledgements networks. Instrum. Measur. Mag., IEEE 13 (6), 14–18.
Ivezić, D. et al., 2008. Fuzzy approach to dependability performance evaluation.
Quality Reliab. Eng. Int. 24 (7), 779–792.
The financial support received from the National Science Foun- Kuo, Y.-C., Lu, S.-T., 2013. Using fuzzy multiple criteria decision making approach to
dation of China (Grant Nos. 51378162, 51178150, 50978080), the enhance risk assessment for metropolitan construction projects. Int. J. Project
Manage. 31 (4), 602–614.
Research fund from Ministry of Science and Technology of China Leung, L.C., Cao, D., 2000. On consistency and ranking of alternatives in fuzzy AHP.
(2013BAJ08B01), the Open Research Fund of State Key Laboratory Eur. J. Operat. Res. 124 (1), 102–113.
for Disaster Reduction in Civil Engineering (SLDRCE12-MB-04), Liu, H.-T., Tsai, Y.-L., 2012. A fuzzy risk assessment approach for occupational
hazards in the construction industry. Safety Sci. 50 (4), 1067–1078.
and the Specialized Research fund for the doctoral program of Liu, M., Frangopol, D.M., 2006. Optimizing bridge network maintenance
higher education (20112302110005) is gratefully appreciated. management under uncertainty with conflicting criteria: life-cycle
maintenance, failure, and user costs. J. Struct. Eng. 132 (11), 1835–1845.
Malinowska, A., 2011. A fuzzy inference-based approach for building damage risk
References
assessment on mining terrains. Eng. Struct. 33 (1), 163–170.
Mangla, S.K. et al., 2015. Risk analysis in green supply chain using fuzzy AHP
Andrić, J., Lu, D.-G., 2014. Risk assessment of bridges using fuzzy logic controller. approach: a case study. Resour. Conserv. Recycl.
EURODYN, Porto, Portugal. Möller, B., Beer, M., 2013. Fuzzy Randomness Uncertainty in Civil Engineering and
AS/NZS4360, 1999. Risk Management Standards Australia. Sydney. Computational Mechanics. Springer Science & Business Media.
Bede, B., 2013. Fuzzy set-theoretic operations. Mathematics of Fuzzy Sets and Fuzzy Ölçer, A., Odabasßi, A., 2005. A new fuzzy multiple attributive group decision making
Logic. Springer, pp. 13–31. methodology and its application to propulsion/manoeuvring system selection
Bien, J., 1999. Expert Functions in Bridge Management Systems. Wroclaw University problem. Eur. J. Operat. Res. 166 (1), 93–114.
of Technology, Poland. Peek-Asa, C. et al., 1998. Fatal and hospitalized injuries resulting from the 1994
Bień, J., 2000. Neural networks in bridge management systems. Structures Congress Northridge earthquake. Int. J. Epidemiol. 27 (3), 459–465.
of the American Society of Civil Engineers, Advanced Technology in Structural Petrović, D.V. et al., 2014. Risk assessment model of mining equipment failure based
Engineering Philadelphia. on fuzzy logic. Expert Syst. Appl. 41 (18), 8157–8164.
Bruneau, M. et al., 2003. A framework to quantitatively assess and enhance the Rojas-Zerpa, J.C., Yusta, J.M., 2015. Application of multicriteria decision methods for
seismic resilience of communities. Earthquake Spectra 19 (4), 733–752. electric supply planning in rural and remote areas. Renew. Sustain. Energy Rev.
Caputo, A.C. et al., 2013. AHP-based methodology for selecting safety devices of 52, 557–571.
industrial machinery. Safety Sci. 53, 202–218. Ross, T.J., 2009. Fuzzy Logic with Engineering Applications. John Wiley & Sons.
Chen, C.-T., 2000. Extensions of the TOPSIS for group decision-making under fuzzy Saaty, T.L., 1980. The Analytical Hierarchical Process. J. Wiley, New York.
environment. Fuzzy Sets Syst. 114 (1), 1–9. Sadi-Nezhad, S., Damghani, K.K., 2010. Application of a fuzzy TOPSIS method base
Chen, S.-J., Chen, S.-M., 2003. A new method for handling multicriteria fuzzy on modified preference ratio and fuzzy distance measurement in assessment of
decision-making problems using FN-IOWA operators. Cybernet. Syst. 34 (2), traffic police centers performance. Appl. Soft Comput. 10 (4), 1028–1039.
109–137. Salmeron, J.L., Lopez, C., 2010. A multicriteria approach for risks assessment in ERP
Cheng, C.-H., 1997. Evaluating naval tactical missile systems by fuzzy AHP based on maintenance. J. Syst. Softw. 83 (10), 1941–1953.
the grade value of membership function. Eur. J. Operat. Res. 96 (2), 343–350. Sezhian, M.V. et al., 2011. Performance measurement in a public sector passenger
Cimellaro, G.P. et al., 2010. Framework for analytical quantification of disaster bus transport company using fuzzy TOPSIS, fuzzy AHP and ANOVA – a case
resilience. Eng. Struct. 32 (11), 3639–3649. study. Int. J. Eng. Sci. Technol. (IJEST) 3 (2), 1046–1059.
da Silva Neves, A.J., Camanho, R., 2015. The use of AHP for IT project priorization – a Tah, J., Carr, V., 2000. A proposal for construction project risk assessment using
case study for oil & gas company. Proc. Comput. Sci. 55, 1097–1105. fuzzy logic. Constr. Manage. Econ. 18 (4), 491–500.
Das, P., 2011. Selection of Business Strategies for Quality Improvement using Fuzzy Van Laarhoven, P., Pedrycz, W., 1983. A fuzzy extension of Saaty’s priority theory.
Analytical Hierarchy Process. Center for Quality. Fuzzy Sets Syst. 11 (1), 199–227.
Dawen, P., Wenda, D., 2009. Notice of retraction research about the bridge risk Vashishtha, S., Ramachandran, M., 2006. Multicriteria evaluation of demand side
assessment. In: 1st International Conference on Information Science and management (DSM) implementation strategies in the Indian power sector.
Engineering (ICISE), 2009. IEEE. Energy 31 (12), 2210–2225.
DHS, H.-M., 2009. MR4 Earthquake Model User Manual. Department of Homeland Wardhana, K., Hadipriono, F.C., 2003. Analysis of recent bridge failures in the United
Security, Federal Emergency Management Agency, Mitigation Division, States. J. Perform. Constr. Facil. 17 (3), 144–150.
Washington (DC). Yucel, G. et al., 2012. A fuzzy risk assessment model for hospital information system
Diaz, E.E.M. et al., 2009. Investigation of common causes of bridge collapse in implementation. Expert Syst. Appl. 39 (1), 1211–1218.
Colombia. Pract. Period. Struct. Des. Constr. 14 (4), 194–200. Zadeh, L.A., 1965. Fuzzy sets. Inform. Control 8 (3), 338–353.
Du, X., Lin, X., 2012. Conceptual model on regional natural disaster risk assessment. Zeng, J. et al., 2007. Application of a fuzzy based decision making methodology to
Proc. Eng. 45, 96–100. construction project risk assessment. Int. J. Project Manage. 25 (6), 589–600.
Engel, A., Last, M., 2007. Modeling software testing costs and risks using fuzzy logic Zheng, G. et al., 2012. Application of a trapezoidal fuzzy AHP method for work safety
paradigm. J. Syst. Softw. 80 (6), 817–835. evaluation and early warning rating of hot and humid environments. Safety Sci.
Fu, Z., et al., 2013. Statistical Analysis of the Causes of Bridge Collapse in China. 50 (2), 228–239.
Forensic Engineering 2012: Gateway to a Safer Tomorrow, pp. 75–83.