Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/257798804

Fuzzy fault tree analysis: A review of concept and application

Article in International Journal of Systems Assurance Engineering and Management · March 2013
DOI: 10.1007/s13198-013-0145-x

CITATIONS READS

132 6,205

5 authors, including:

Yasser Mahmood Alireza Ahmadi


Luleå University of Technology Luleå University of Technology
12 PUBLICATIONS 141 CITATIONS 73 PUBLICATIONS 1,104 CITATIONS

SEE PROFILE SEE PROFILE

Uday Kumar
Luleå University of Technology
313 PUBLICATIONS 5,387 CITATIONS

SEE PROFILE

All content following this page was uploaded by Yasser Mahmood on 04 April 2014.

The user has requested enhancement of the downloaded file.


Int J Syst Assur Eng Manag (Jan-Mar 2013) 4(1):19–32
DOI 10.1007/s13198-013-0145-x

ORIGINAL ARTICLE

Fuzzy fault tree analysis: a review of concept and application


Y. A. Mahmood • A. Ahmadi • A. K. Verma •

A. Srividya • U. Kumar

Received: 30 October 2012 / Revised: 17 January 2013 / Published online: 16 February 2013
 The Society for Reliability Engineering, Quality and Operations Management (SREQOM), India and The Division of Operation and
Maintenance, Lulea University of Technology, Sweden 2013

Abstract Fault tree analysis (FTA) is a widely used 1 Introduction


method for analyzing a system’s failure logic and calculating
overall reliability. However, application of conventional FTA Reliability analysis of a system requires a proper under-
has some shortcomings, e.g. in handling the uncertainties, standing of logical functional relationships among sub-
allowing the use of linguistic variables, and integrating systems and their associated failure modes. Fault tree
human error in failure logic model. Hence, Fuzzy set theory analysis (FTA) is a widely used method for analyzing a
has been proposed to overcome the limitation of conven- system’s failure logic and calculating overall reliability. It
tional FTA. Fuzzy logic provides a framework whereby basic was first conceived in 1961 by H. A. Watson of Bell
notions such as similarity, uncertainty and preference can be Telephone Laboratories in connection with US Air Force
modeled effectively. The aim of this paper is to present a contract, to study the Minuteman missile launch control
review of the concept of fuzzy theory with fault tree analysis system (Lee et al. 1985). Since then, it is widely used for
and their applications since 1981, to reflect the current status system safety assessment, fault diagnosis and prognosis, as
of Fuzzy fault tree analysis (FFTA) methodologies, their well as reliability analysis of industrial systems, such as
strengths, weaknesses, and their applications. This paper nuclear reactor, aerospace, electric power, chemical,
explains the fundamentals of fuzzy theory and describes mechanical, civil engineering, petrochemical industry,
application of fuzzy importance for using FFTA. The con- pipelines, and electronics.
cept of the failure possibility and uncertainty analysis by FTA is basically a cause-and-effect analysis and traces a
using FFTA is discussed, and concludes with discussion on system failure ‘down’ to one or more failures at the lower
the application of FFTA in different fields. The review levels. The fundamental concept involved in the FTA is the
reveals the effectiveness of the FFTA in comparison with translation of a physical system into a structured logic
conventional FTA, when there is inadequate amount of diagram, in which certain sequence of basic events (causes)
accurate reliability oriented information. lead to one specified TOP event (i.e. system-level) of
interest (Gupta and Bhattacharya 2007). In fact, FTA is a
Keywords Fuzzy reliability  Fault tree analysis  deductive procedure for determining combinations of both
Fuzzy fault tree analysis component failures and human errors, which could result in
the occurrence of specified undesired events at the system-
level (Gupta and Bhattacharya 2007). The logic diagram is
Y. A. Mahmood (&)  A. Ahmadi  U. Kumar constructed using event and logic symbols which derive the
Luleå University of Technology, Luleå, Sweden
failure logic for a system and represent it in a form which is
e-mail: yasser.ahmed@ltu.se
ideal for communication to managers, designers, and
Y. A. Mahmood operators. If failure rate and other necessary data are
University of Mosul, Mosul, Iraq available for the basic events, the fault tree analysis will
provide estimates of the frequency of occurrence of the
A. K. Verma  A. Srividya
Stord/Haugesund University College, Haugesund, Norway undesired events. However, application of conventional

123
20 Int J Syst Assur Eng Manag (Jan-Mar 2013) 4(1):19–32

FTA has some shortcomings; vague, absence of accurate not allow treating different aspects of uncertainty, proba-
data, and uncertainty. bility and possibility at the same time. Even the most
In conventional FTA, the failure probabilities are con- sophisticated, precise, and well-constructed quantitative
sidered as an exact value, i.e. single estimated value or crisp model may give misleading results, if uncertainties are not
value. However, it is difficult to estimate a precise failure treated at some level.
rate or probability of the components due to lack of insuf- Therefore, it is necessary to develop a new methodology
ficient data, or vague characteristic of the events. This is to capture the subjectivity and the imprecision of failure
crucial specifically in the preliminary design stages, when data, to be used in the conventional FTA. Hence, using
the details of the components may have yet to be established, fuzzy set theory has been proposed by many researchers,
thus an exact failure rate could not possibly be known. such as Tanaka et al. (1983), Misra and Weber (1989) and
In the absence of accurate data, the experiences of field Kenarangui (1991) to overcome the limitation of conven-
experts, e.g. designers, operators or personnel, provide an tional FTA. Fuzzy logic provides a framework whereby
effective database supporting the rough estimation of basic notions such as similarity, uncertainty and preference
required data (failure rate and probability). However, inad- can be modeled effectively. A ‘‘fuzzy set’’ represents a set
equate data makes it difficult to determine the probability in with ill-defined boundaries, and processes vague terms
an objective manner. Hence, human judgments by linguistic with ‘‘grey boundaries’’. Fuzzy Logic is a branch of
variables become essential. Hence, they tend to apply natural mathematics that deals with fuzzy statements (linguistic
linguistic expression, such as ‘very low, low, high and fairly variables) and allows for partial membership in a set.
high’, to describe the probability of events. As an example, Fuzzy Logic provides a remarkably simple way to draw
the diagnostic procedure is based on the observed analytical, definite conclusions from vague, ambiguous or imprecise
heuristic symptoms and the heuristic knowledge of the pro- information (Zadeh 1978, 1988)
cess. Heuristic symptoms are the observations obtained The classification and the number of papers published in
through inspection by operating personnel in various forms, the last 30 years in the field according to Scopus biblio-
e.g. acoustic noise, oscillations or optical impressions like graphic database has been presented in Fig. 1. Tanaka et al.
colors or smoke. However, vague symptoms emerge fre- (1983) proposed the Fuzzy fault tree analysis (FFTA) and
quently which can usually be expressed only in the form of the possibility of failure instead of probability. They applied
qualitative measures or linguistic variables, such as ‘‘little,’’ fuzzy extension principle to calculate the probability of the
‘‘medium’’ or ‘‘much’’. In fact, conventional mathematical top event. Misra and Weber (1989) used the ‘‘max and min
ways cannot handle natural linguistic expression efficiently operators’’ to evaluate the top event ‘‘possibility’’ in terms of
because of its vagueness (Ding and Yu 2005). In fact, the the ‘‘possibilities’’ of the basic events, and illustrated a
crisp approach tends to have difficulty in conveying the procedure to analyze an emergency core cooling system of a
imprecision or vague nature in system modeling to represent nuclear reactor. Misra and Soman (1995) provided a simple
the failure rate. method for FFTA based on the a-cut method, also known as
In addition, in conventional FTA, the basic events con- resolution identity. This method was then extended to deal
sidered are normally associated with hardware failure. with multi-state FTA. Sawyer (1994) used the a-cut method
However, in a highly automated system, still people are the to calculate the failure probability of the top event in fuzzy
key component in the system. According to Lee et al. FTA of mechanical systems. An event tree analysis tech-
(Singer 1990), depending upon the degree of human nique, based on the use of verbal statements to characterize
involvement in the system, the human component is the probability of events, has been presented by Kenarangui
responsible for 20–90 % of the failures in many systems. As (1991). Fuzzy method might be the only way when little
a result of the unpredictability of human performance as quantitative information is available regarding fluctuations
well as the vagueness of system phenomena, the basic of the parameters, and the probabilities of basic events are
events in a man–machine system can be classified into two treated as fuzzy numbers (Ding and Yu 2005). Lin and
categories, i.e. human-related subjective events and proba- Wang (1997) combined fuzzy set theories with expert elic-
bilistic objective events. Thus, the integration of probabi- itation to the failure probability of basic events of a robot
listic estimation of hardware failures and subjective drilling system, based on triangular fuzzy numbers. Liang
evaluation of human performance is the major concern of and Wang (1993) discussing about human performance and
analyzing the hybrid system. In fact, the crisp approach used how to evaluate the exact failure rates of components in
in conventional FTA is unable to convey the imprecision or man–machine systems. Many other researchers (Geymayr
vague nature of human performance in system modeling. 1995; Bowles and Pelaez 1995; Cai 1996; Sharma 1993;
Moreover, the conventional FTA doesn’t allow alloca- Wang and Wu 2007; Wei et al. 2009; Dunyak and Wunsch
tion of a degree of uncertainty to the probability values in 1998; Hatoyama 1979; Ferdous et al. 2009; Zonouz and
the modeling of the basic event’s probability. Also, it does Miremadi 2006) contributed to support this approach and

123
Int J Syst Assur Eng Manag (Jan-Mar 2013) 4(1):19–32 21

used the existing theory and methodology of the fuzzy represents a fuzzy number between zero and one, which
number based on FTA. can be assigned to the probability of an event (Misra and
The need to possess appropriate theoretical methods for Soman 1995). There are many forms of fuzzy numbers
handling above mentioned limitations with FTA is straight- used e.g. triangular fuzzy number (TFN) and trapezoidal
forward, and is of increasing importance with increasing fuzzy number (TZFN), which are commonly used in reli-
system complexity. Hence, a review of the concept of using ability analysis. Many researchers have used the triangular
fuzzy theory with fault tree analysis and their applications, to fuzzy number, although the pioneer worker in this field, i.e.
reflect the current status of FFTA methodologies, their Tanaka et al. used a trapezoidal membership function.
strengths, weaknesses, and their applications, has been pre- Let x, a, b, c € R, and fa: R ? [0,1] represents the
sented in this paper. membership function. Then, a triangular membership fuzzy
The fundamentals of fuzzy theory are presented in Sect. number A is clear in Eq. 1:
2. The fuzzy importance for using (FFTA) is described in 
 ðx  aÞ=ðb  aÞ; . . .. . .. . .. . .:. . .a  x  b
Sect. 3. The uncertainty analysis using FFTA is dealt with 

fA ð xÞ ¼  ðc  xÞ=ðc  bÞ;. . .. . .. . .. . .::. . .a  x  b ð1Þ
in Sect. 4 and is defined. In Sect. 5 the concept of the 
 0; . . .. . .. . .. . .. . .. . .. . .. . .. . .. . .. . .otherwise
failure possibility is defined. The most relevant applica-
tions are presented in Sect. 6 and finally, the conclusions of with a  b  c, this triangular fuzzy number can be denoted
the present research are presented in Sect. 8. by A˜ = (a,b,c). The parameter ‘b’ gives the maximal grade
of fA(x), ‘a’ and ‘c’ are the lower and upper bounds of the
2 Fundamentals of fuzzy theory available area for the evaluation data. Singer (1990) and
others (Kim 1996; Bian et al. 2009) used A˜ = (m,a,b) for
Zimmermann (1983) has explained the use of fuzzy meth- triangular to give the possible distribution of the basic
odology in operational research and given the following event probability. The value of ‘m = b’ but’a’ and’b’ the
observations: Vague phenomenon, vague relations in the left and right spreads ‘a = b - a’ and ‘b = c - b’ are
modeling of problems, that is, the problems themselves may shown in Fig. 2a.
be fuzzy in nature. Informational vagueness, that is, the The second type of the fuzzy number is a trapezoidal
input data available to the problems may be fuzzy. Heuristic fuzzy number (Eq. 2) if its membership function fA:R
algorithms, means although the problems may be accurate, it ? [0,1] is:
can be too complex or expensive to get exact solutions to 
 ðx  aÞ=ðb  aÞ; . . .. . .. . .. . .::. . .a  x  b
them. In these circumstances heuristic or fuzzy algorithms 
 1; . . .. . .. . .. . .. . .. . .. . .. . .. . .:. . .. . .b  x  c

can help. In this section some relevant definitions and f A ð xÞ ¼  ð2Þ
 ðc  xÞ=ðc  bÞ; . . .. . .. . .::. . .. . .c  x  d
principles of fuzzy mathematics are given to enable a better 
 0; . . .. . .. . .. . .. . .. . .. . .. . .. . .. . .. . .otherwise
understanding of the concepts on fuzzy theory.
witha  b  c  dthe trapezoidal fuzzy number are shown
2.1 Fuzzy number
in Fig. 2b and can be denoted by A˜ = (a,b,c,d) the interval
[b,c] are the most likely values of fA(x), ‘a’ and ‘d’ are the
Fuzzy number also called fuzzy probability (Tanaka et al.
lower and upper bounds of the available area for the
1983), describes ‘vagueness’ on equal terms with ‘crisp-
evaluation data (Lin and Wang 1997). Some researchers
ness’ (Weber 1994) such as ‘close to 5‘, ‘high reliability’
(Lee et al. 1985) treated probabilities of basic events as
and ‘low failure rate’. A fuzzy set in probability space
trapezoidal fuzzy numbers. There are different forms of
fuzzy numbers but the form of triangular and trapezoidal is
commonly used in the field of FFTA.
Using FFTA methodology requires converting the crisp
value into a fuzzy number. Fuzzy numbers can be generated
using different method such as expert knowledge elicitation,
3rr expression, or a percentage of lower and upper limits
(Cai 1996). There are two basic forms of expert elicitation in
form of a linguistic language: linguistic variables and
interval values (Cai 1996). Hence, if there is no statistical
data to find the failure rate as a crisp value, the 3rr
expression mode is used. The principle of the 3rr expres-
Fig. 1 Year’s distributions of published papers of FFTA, when sion method is by using three experts. Each expert will give
putting ‘‘Fuzzy fault tree’’ in Scopus the estimate value for the probability value of the fault

123
22 Int J Syst Assur Eng Manag (Jan-Mar 2013) 4(1):19–32

Fig. 2 (a) Fuzzy Triangle


Number (FTN) A˜ = (a, b, c).
(a) (b)
(b) Fuzzy Trapezoidal Number
(FTRN) A˜ = (a, b, c, d)

events. Each average value of the probability values of the Very Fairly Fairly Very
low Low low Medium high High high
fault events is given by b, and the difference are the same and 1
shown as r. The guessed probability values belong to normal
distribution. According to the 3r law, the factual value is
0.5
located in the section of [b - 3r, b ? 3r]. And guess
‘aa = b = 3r’, each probability value of the fault event can
be expressed as A˜ = (b - 3r, b, b ? 3r). The second way is
to find the barycenter value of the subjection function b from 0 0.5 1
the statistical data as the probability of failure rate. Then the
Fig. 3 Fuzzy numbers represent linguistic value
lower and upper limit can be obtained as a range of uncer-
tainty of the basic events from the experts and technicians as cope with it by transforming these linguistic terms into
interval values (Wu et al. 2007). The other way is to use the graphical or mathematical representations. In conventional
crisp probability lower and upper limit failure rate b, then one FTA, it is not possible to reflect an exact picture of system
can use lower and upper limits (as an example ±15 %) safety as presented in the introduction. Hence, in order to
spreads (Kumar 2012), and the fuzzy number can be overcome this disadvantage, the combined expert elicita-
expressed as A˜ = (b - 0.15*b, b, b ? 0.15*b). tion with fuzzy set theories was proposed to evaluate
probability of the events by linguistic variables (Ding and
2.2 Linguistic variables and its advantages Yu 2005; Lin and Wang 1997; Pan and Wang 2007). Fuzzy
methods and linguistic variables might be the only resort
Linguistic variables as its name suggests, is a variable whose when little and no quantitative information is available.
values are words or sentences in a natural or synthetic lan- Linguistic variables can be used to present both human
guage. For example, ‘‘probability-of-failure’’ might be a error and hardware failure rates (Lin and Wang 1997).
linguistic variable whose values are from the term set: ‘‘very Fuzzy set theory provides a useful tool for linguistic
low’’, ‘‘low’’, ‘‘medium’’, ‘‘high’’ and ‘‘very high’’. These expression in reliability analyses. The analyst can use lin-
terms can be defined as a fuzzy set whose members are guistic variables to assess the events in a natural way, and
probabilities of failures as a linguistic variable, as shown in the failures possibility of the events can be approximated,
Fig. 3 (Bowles and Pelaez 1995). Each value of a linguistic by their membership functions. It is well suited for han-
variable represents a possibility distribution. These possi- dling ambiguous and imprecise information obtained in
bility distributions may be computed from the given possi- system safety engineering (Lin and Wang 1997).
bility distributions of the primary term (Zadeh 1988). But Weber (1994a, b), developed and applied an interval
the interval values may be expressed in terms of a pair of method for fault tree analysis in 1981. He used symbolic
single numbers [a,b], that is, 0 BB a B Bp B b B 1, processing or symbolic math. This simplified analytical
where p denotes the probist reliability (Cai 1996). approach used symbols at gates in place of probability
The main reason to treat probist reliability value by a values. The symbols in the middle matched the semantic
fuzzy number is the lack of observed data of system failure descriptions (probability terms) like extremely remote,
behavior. Having inadequate data makes it difficult to remote, and frequent. This was the first step to use the
determine the probability in an objective manner. Hence, linguistic variables in FTA. Shu et al. (2006) collected
human judgments by linguistic variables become essential. expert’s knowledge, to build the intuitionistic fuzzy num-
The concept of linguistic variable is very useful, in ber, in order to represent possibilities of failure basic
dealing with situations which are too complex or too ill- events. Pan and Wang (2007) and Ding and Yu (2005)
defined to be reasonably described in conventional quan- elicited the experts’ knowledge through several interviews
titative expressions. But conventional mathematical ways and questionnaires, to estimate the possibilities of failure.
cannot handle natural linguistic expression efficiently. Also Cheng et al. (2009) used expert’s knowledge on the
Therefore, fuzzy set theory was introduced by Zadeh to liquefied natural gas terminal emergency shutdown (ESD)

123
Int J Syst Assur Eng Manag (Jan-Mar 2013) 4(1):19–32 23

by using the intuitionistic fuzzy number for representing Fuzzy mean is a mean operation between AND and OR. The
possibilities of failure. Dokas et al. (2009) described a fuzzy sum operation produces an accumulating effect of
novel research investigation on how to combine expert input values. Kim et al. (1996) shows that fuzzy AND and OR
system (ES) technologies as a linguistic variables, together operators are used to evaluate the possibility of systems
with basic principles of the theory of fuzzy logic and a failure, but each value of m from (m,,ab) is determined to
widely used risk analysis fault tree analysis to develop an obtain the optimistic and pessimistic possibilities of the
EWS for landfill operations, and integrate possibility the- system accident. These are extended with major and minor
ory and risk analysis into one operational fuzzy expert TFNs to fuzzy FTA to get better results.
system providing early warning services. Ren and Kong In fault tree analysis, although AND, OR gates have been
(2011) proposed a new approach of fuzzy multi-state fault used as typical gates, but it is often difficult to model the
tree analysis based on the fuzzy expert system. They system structure with these two gates. The reason is that in
applied fuzzy theory to represent the degree of basic events early design stage, exact knowledge on system failure
then according to the process of fuzzy expert system they mechanism is not available in many cases (Pan and Yun
calculated the probability of the top event. 1997). Pan and Yun 1997 (1997) applied the fuzzy sets
The process diagnosis history, in the form of mainte- theory to modeling the fuzzy system structure, and assumed
nance, repairs, former faults, lifetime and load measures, that the gates are modeled as a fuzzy relation because it is
constitutes a further source of heuristic information. Sta- often impossible to obtain the exact relationship between
tistical data (e.g. MTBF, fault probabilities) achieved from components in the complex and large scaled system. While
experience with the same or similar processes can be some researchers (Yao and Zhang 2010; Song 2009) used a
added. Heuristic symptoms can be represented as linguistic new gate based on the T–S (Takagi and Sugeno) fuzzy model
variables (e.g. small, medium, large) or as vague numbers to overcome these difficulties.
(e.g. around a certain value) (Isermann 1997).

2.3 Fuzzy logic operator 3 Importance measures

Supposing that the fuzzy triangular number PI and P2 can In conventional FTA, the concept of importance is used to
be separately expressed by (a1, b1, c1) and (a2, b2, c2), the evaluate how far a basic event contributes to the top event.
algebraic algorithms about the fuzzy number P1 and P2 are Importance measure is determinable not only for the basic
shown as follows: events, but also for every higher-level event and contributor
as well. This provides a numerical significance of all the
1. ‘‘AND gate’’ fuzzy operators. In conventional FTA, fault tree elements and allows them to be prioritized
’’AND gate’’ operator is: PAND = G Pi in which (NASA). Importance measure also is used to optimize the
Pi(i = 1,2….n) represents the precise probability of system design and its reliability. It is used to improve system
the event i is Pi. Fuzzy operator is: reliability and cost because there is some tradeoff between
" #
Y
n Y
n Y
n Y
n them (He et al. 2007). Also, the fuzzy importance also
PAND ¼ Pi ¼ ai ; bi ; ci ; ð3Þ provides valuable information to improve fault diagnosis of
i1 i¼1 i¼1 i¼1
complicated system and provide a reasonable guidance for
If the events are dependent, then the algebraic algorithm maintenance (Mi et al. 2011). This can also be used to list the
is: PAND = min(P1, P2,……. Pn). examination table of fault diagnosis and to instruct the
operation and maintenance of the system (Peng et al. 2008).
2. ‘‘OR gate’’ fuzzy operators, In conventional FTA, ’’OR
The importance values are measured by the use of
gate’’ operator is: POR = G (1-Pi) in which Pi
probability information of both basic and top events (Gu-
represents the precise probability of event i is Pi, then
imarees and Ebecken 1999; Furuta and Shiraishi 1984).
’’OR gate’’ fuzzy operator can be denoted by:
However, these probability-based definitions may be
Y n
POR ¼ 1  ð1  Pi Þ insignificant for a case in which the occurrence of a basic
"i1 # event can be imprecisely specified. This means that if the
Yn Y
n Y
n state of a basic event is expressed by a fuzzy event, it may
¼1 ð1  ai Þ; ð1  bi Þ; ð1  ci Þ; ð4Þ not be possible to identify the importance of each event on
i¼1 i¼1 i¼1
the basis of only probability information.
if the events are dependent, then the algorithm is POR = Therefore, Furuta and Shiraishi (1984) proposed fuzzy
max(P1, P2,……. Pn) (Mao et al. 2010). importance (FI) to account for the ambiguities involved in
Fujino and Hadipriono (1994) developed two gates for the the certification process of the occurrence of basic events
purpose of FFTA, i.e. fuzzy mean and fuzzy sum operation. using max/min fuzzy operator and fuzzy integral. They

123
24 Int J Syst Assur Eng Manag (Jan-Mar 2013) 4(1):19–32

gave two evaluation examples, structural importance (SI) fuzzy approach gives another advantage, by using FUIM
are calculated with fuzzy importance (FI) for comparison for ranking, which is based on propagated uncertainties.
between two approaches. The results for Example-1 FI and Liang and Wang (1993) proposed another importance
SI give the almost same order of importance, but for measure known as fuzzy importance index (FII). Calcula-
Example-2, as shown in Table 1, they give the different tion of FII is based on the ranking method of triangular
orders. For instance, the basic event No. 1 is evaluated by fuzzy numbers with maximizing and minimizing sets. To
FI as having the lowest importance. This difference can be establish the rank and the degree of sensitivity of compo-
explained by the fact that the SI is determined only by the nents Chanda and Bhattacharjee (1998) proposed the reli-
position of the underlying event in FTA. While using the ability improvement index (RII), and it is calculated by
fuzzy sets enables the calculation to account for the effect preventing the failure of each component and cut set sep-
of subjective uncertainties in the evaluation of the state of arately. According to this study, FIM, FIUM and FI are
basic events, and therefore yields a different order of widely used among other measures mentioned above.
importance for event No. 1 (Furuta and Shiraishi 1984) and
this is showing the advantages of using the second
approach (Table 2). 4 Uncertainty analysis
Suresh et al. (1996) introduced a new approach to rank
the system components or basic events depending on, their Using fuzzy concept allows allocation of a degree of
contribution to the top event failure probability and, their uncertainty to each probability value, and treats different
uncertainty contribution to the uncertainty of the top event, aspects of uncertainty, probability and possibility at the
based on fuzzy set theory. They introduced two different same time (Tanaka et al. 1983). Even the most sophisticated,
importance measures: fuzzy importance measure (FIM) precise, and well-constructed quantitative model may give
based on the Euclidean distance approach and fuzzy misleading results if uncertainties are not treated at an
uncertainty importance measure (FUIM), which are defines acceptable level. Uncertainty can range from modeling
as follow in equations 5 and 6: uncertainty, to incomplete and unreliable information. So,
  FFTA was designed to handle the uncertainty problem in
FIMi ¼ ED Qqi¼1 ; Qqi¼0 ð5Þ
incertitude in conventional FTA (Weber 1994). The uncer-
FUIMi ¼ ED½Q; Qi  ð6Þ tainties in the failure probability of system components or
where Q = top event failure probability, Qi = top event basic events can be propagated to find the uncertainty in
failure probability when error factor for component i is the overall system failure probability. The conventional
unity. FUIM plays an important role in the reduction of approach is Monte Carlo simulation by assuming a proba-
uncertainty, it is so important to identify the components bility distribution for the failure probability.
which have the maximum contribution of uncertainty to the The probabilistic approach to uncertainty analysis basi-
uncertainty of the top event. This helps in deciding the cally depends upon the assumption of a probability distri-
components for which more information should be col- bution of failure probability as explained earlier, which can
lected so that the uncertainty in the calculated system be obtained only when a sufficient amount of failure data is
failure probability can be lowered. The importance order available. To overcome some of the difficulties, the input
and ranking by using FIM and FUIM are shown in the parameter is treated as a fuzzy number and the variability is
Table 3 (Suresh et al. 1996). characterized by the membership function which can be
It is seen in Table 3 that there is a similarity between obtained based on available information or the expert’s
probability approach and fuzzy approach related to the opinion. The membership function of each fuzzy set is
ranking based of importance measure. As it is shown, the usually assumed to be a triangular or trapezoidal function
and is treated as a possibility distribution (Suresh et al. 1996;
Peng et al. 2008).
As we mentioned above in Sect. 3, FUIM was proposed
Table 1 Numerical results as a comparison between SI and FI in to identify those sources of uncertainties, which have the
Furuta and Shiraishi (1984)example
greatest impacts on the uncertainty of the top event. Suresh
Event No. SI FI et al. have made a comparison between FUIM and using
1 0.44 0
Mote-Carlo method for the system that was used in (Ras-
2 0.06 0
mussen 1400). The corresponding error factors after 1200
trials of the Monte Carlo simulation. As it is evident are
3 0.44 0.42
shown in Table 3, the computation time used for the
4 0.31 0.19
probabilistic approach is very high, compared to the fuzzy
5 0.31 0.09
set approach. Both FIM and FUIM have been calculated

123
Int J Syst Assur Eng Manag (Jan-Mar 2013) 4(1):19–32 25

Table 2 Fuzzy fault tree


Class Papers
classification
FFTA introduction, Uncertainty, Singer(1990), Tanaka et al. (1983), Misra and Weber (1989),
possibility, and importance Misra and Soman (1995), Sawyer (1994), Lin and Wang (1997),
Liang and Wang (1993), Cai (1996), Zonouz and Miremadi
(2006), Pan and Yun (1997), Yang et al. (1995), Page and Perry
(1994), Pan and Wang (2007), Kim (1996), Deshpande and
Khanna (1995), Chanda and Bhattacharjee (1998), Zhang et al.
(2008), Weber (1994a), Bian et al. (2009), Suresh et al. (1996),
Guimarees and Ebecken (1999), Furuta and Shiraishi (1984),
Fujino and Hadipriono (1994), and Mohan et al. (2003)
Linguistic variables and expert Ding and Yu (2005), Lin and Wang (1997), Cai (1996), Kim
knowledge with FFTA (1996), Fujino (1994), Cheng et al. (2009), Dokas et al. (2009),
Ulieru (2199), Sui (2011), Renjith et al. (2010), and La (2008)
FFTA Diagnosis Page and Perry (1994), Dokas et al. (2009), Sui (2011), Wu et al.
(2007), Peng et al. (2008), Yuliang and Tiejun (2010), Yao et al.
(2010), Ulieru (1994), Gmytrasiewicz et al. (1990), Lou et al.
(2010), Chang and Chang (2003), Hurdle et al. (2212), Chen
(2011), Xie (2010), Yao (2010), Khan 2000
FFTA application Ding and Yu (2005), Misra and Soman (1995), Sawyer (1994), Lin
and Wang (1997), Cai (1996), Page and Perry (1994), Kim
(1996), Chanda and Bhattacharjee (1998), Zhang et al. (2008),
Cheng et al. (2009), Dokas et al. (2009), lieru (2199), Peng et al.
(2008), Yuliang and Tiejun (2010), Yao et al. (2010), Ulieru
(1994), Lou et al. (2010), Chen (2011), Batzias and Batzias
(2003), Batzias and Siontorou (2004), Xiao-lin et al. (2010), Shu
et al. (2006), Kumar et al. (2008), Wu et al. (2007), Hu and Ge
(2010), Rong and Xin (2010), Szabó et al. (2010), and
Zimmerman (1983)

Table 3 Failure probability and ranking for different components can be utilized to provide insight on the design of data and
information gathering strategies that focus on the reduction
Event no. Failure Error FIM FUIM
of the total uncertainty (Suresh et al. 1996).
probability factor (rank) (rank)
Ferdous et al. (2009) also introduce a comparison of the
1 1.7e–5 10 4.69 (1) 3.0e–4 (2) conventional probability and fuzzy approaches for measur-
2 1.0e–3 3 2.9e–2 (3) 4.5e–5 (4) ing propagated error on the top-event. They have used the
3 3.6e–4 3 2.9e–2 (3) 1.6e–5 (5) conventional probability approach, and Monte Carlo simu-
4 1.0e–3 3 2.9e–2 (3) 4.5e–5 (4) lation for uncertainty analysis. Different levels of error (e.g.
5 3.6e–4 3 2.9e–2 (3) 1.6e–5 (5) 5, 10, 15 and 20 %) are considered in the basic event data of
6 6.1e–3 4 2.0e–2 (4) 2.8e–4 (3) FTA. Detailed comparison of uncertainty propagation for
7 6.1e–3 4 2.0e–2 (4) 2.8e–4 (3) both conventional probability and fuzzy approaches are
8 9.7e–4 10 8.5e–2 (2) 5.5e–4 (1) presented in Table 4. It also illustrates how much error is
9 9.7e–4 10 8.5e–2 (2) 5.5e–4 (1) propagated cumulatively, in the calculation of top-event
probability, if the basic event data are not accurately mea-
sured (Ferdous et al. 2009).
based on Eqs. (3) and (4) for all basic events. Even though According to Table 4, it is obvious that the fuzzy based
Monte Carlo simulation gives slightly different important approach gives a better result compared to the normal
measures for the events 2, 3, 4 and 5, the same rank was probability approach, which makes it more realistic and
given for all these events. In this approach, the fuzzy robust, to handle the vagueness and imprecision of the
importance measures are equal for the same rankings. basic event, in FTA.
However, the ranking for FUIM is 1st for events 8 and 9, Zonouz and Miremadi (2006) used another methodology
2nd for event 1, 3rd for events 6, 7, 4th for events 2, 4, and like fuzzy-Monte Carlo simulation (FMCS) approach to
5th for events 3, 5, which is different from FIM 1st for handle the uncertainty. The FMCS approach has two main
event 1, 2nd for events 8, 9, 3rd for events 2, 3, 4, 5 and advantages in comparison to the traditional Monte Carlo
4thfor event events 6, 7, as expected. FIM can be used to Simulation. It considers the cognitive uncertainty in addi-
find out the critical component which can be used for tion to the noncognitive vagueness. Also it has the capa-
design modifications of the system. The results of FUIM bility to model the situation in which components fail at a

123
26 Int J Syst Assur Eng Manag (Jan-Mar 2013) 4(1):19–32

Table 4 Error robustness of fuzzy approach and normal probability probability of failure. The failure possibility defined by
approach (Ferdous et al. 2009) Onisawa (1988) can effectively represent the variation of
Considered Approach Top-event Deviation in human performance, the vagueness of system phenomena,
error in data probability percentage a (%) and the degree of stability of system components. In this
sense, Failure possibility is a nonprobability subjective
No error (%) Exact probability 2.8e–4 –
approach b measure of reliability, using natural language expressions.
Fuzzy Approach – – The failure possibility can be defined by a fuzzy number
5 Conventional 2.97e–4 1 on interval [0,1], with any shape of membership function.
probability The triangular fuzzy numbers which is defined by an
approach interval i.e. [0,1] are used to denote the failure possibilities.
Fuzzy Approach 2.95e–4 2 The reason of using the triangular fuzzy numbers is because
10 Conventional 3.32e–4 13.3 it is intuitively easy for the decision-makers to perform
probability evaluation. It can convey the most possible failure rate of
approach
the system, and also reflect the dispersion of the evaluation
Fuzzy Approach 2.97e–4 3
data. It is very useful in conveying the variation of human
15 Conventional 3.48e–4 20.5
performance, the vagueness of system phenomena as well
probability
approach as the stability of machine components in man–machine
Fuzzy Approach 2.98e–4 13.5 systems (Liang and Wang 1993; Cai 1996; El-Iraki and
20 Conventional 4.63e–4 60 Odoom 1998; Huang et al. 2004).
probability Possibility theory becomes a specially powerful tool for
approach working with uncertainty when it is coupled with the
Fuzzy Approach 3.03e–4 5 concept of a linguistic variable (Bowles and Pelaez 1995;
a
Deviation = [TE probability (defined approach) - TE probability Zadeh 1973). Liang and Wang (1993) used failure possi-
(exact approach)]/[TE probability (exact approach)] bility defined by a triangular fuzzy number on the interval
b
Considering all basic event data are accurately measured [0,1] to characterize the possible deviation of the basic
events. Singer (1990) analyzed fuzzy reliability by using
L–R type (the Left and Right spreads) fuzzy numbers. He
single point of time as well as those which may fail considered the relative frequencies of the basic events as
gradually. fuzzy numbers and used possibility instead of probability
measures. He has shown the use of n-array possibilistic
AND, OR, and NEGATIVE operators to construct possible
5 Possibility fault tree allowing a straightforward evaluation for the
possibility distributions of the main events.
Possibility and probabilities are both defined on the interval In order to replace probabilistic considerations in the
[0, 1] and both provide a measure of uncertainty, but they FTA by the possibilistic ones and to reduce the difficulty
are quite different (Bowles and Pelaez 1995). Zadeh (1978) arising from the inexact and insufficient information of the
introduced the notion of a possibility in describing vaguely distribution functions. Kim (1996) proposed two cases of
defined events, failure and reliability problems. possibility distribution, i.e. a pessimistic possibility of
Using possibility instead of probability was proposed to system failure with major TFN and, an optimistic one with
characterize the failure occurrence of system events. minor TFN. The value of m from (m, a, b) is determined to
Possibility theory in the framework of fuzzy sets was obtain the optimistic and pessimistic possibilities of the
further developed by other researchers found its applica- system accident. We can extend major and minor TFNs to
tions in the area of reliability. Cai (1996a, b) and Cremona fuzzy FTA for better results.
and Gao (1997), have shown that the consideration of Lin and Wang (1997) and Ding and Yu (2005) used
fuzzy or possibilistic methodology in reliability and safety fuzzy possibility score (FPS) to reduce the difficulty arising
analysis is natural, and stems from the lack of observed from inexact and insufficient information of the distribu-
data regarding system failure in some cases; or the tion functions of basic event. In their approach, they used
inherent vagueness of human judgment in others (He et al. FPS to represent the failure occurrence of the fuzzy events.
2007). To integrate the hardware failure rate and the fuzzy pos-
The possibility of failure was first used as a substitute sibility score of human error, a transformation function is
for the probability of failure by Tanaka et al. (1983). In used to convert fuzzy possibility score into fuzzy failure
their definition, the possibility of failure includes the rate.

123
Int J Syst Assur Eng Manag (Jan-Mar 2013) 4(1):19–32 27

6 Diagnosis as application of FFTA failure vector to the basic events is finally solved in the
second step to determine estimates of the system failures in
System fault diagnosis is the process of identifying the terms of the basic events. The fault trees were used to con-
cause of a malfunction by observing its effect at various struct fuzzy relation equations that relate the failure modes to
test points (Hurdle et al. 2212; GazdikI 1985). The task of the basic events. Ulieru (1994) developed a dynamic
fault diagnosis consists of the determination of the type of approach for fuzzy fault trees processing. Based on similarity
fault with as many details as possible, such as the fault size, evaluation it determines. The possibilistic approach that
location and time of detection. The diagnostic procedure is allows the on-line processing of the dynamic fuzzy infor-
based on the observed analytical and heuristic symptoms mation gathered from the on-line detected dynamic and
and the heuristic knowledge of the process. subjective symptoms. Shu et al. (2006) From integrating
There are three inputs to the knowledge-based fault- expert’s knowledge and experience in terms of providing the
inferencing mechanism to represent symptoms in fault possibilities of failure of bottom events, to calculate fault
diagnosis: Analytical symptoms, Heuristic symptoms, and interval of system components. In order to find the most
process history and fault statistics. The latter depends on critical system component for the managerial decision-
the general status, based on the history (past life) of the making based on intuitionistic fuzzy fault-tree analysis to
process, that’s like FTA methodology its depending to the implement FTA on PCBA fault diagnosis.
failure rate in the past life. The information on the process
history in general is vague, and their facts have to be taken
as heuristic symptoms (Isermann 1997). 7 Industrial application
Heuristic symptoms share the observations of the oper-
ating personnel in the form of acoustic noise, oscillations FFTA is used in system safety assessment and reliability
or optical impressions like colors or smoke, obtained by analysis especially for complex and large scaled system,
inspection. These empirical facts can usually only be rep- such as nuclear reactor, aerospace, electronics, electric
resented in form of qualitative measures, e.g. as linguistic power, chemical, mechanical, civil engineering, petro-
expressions like ‘‘little’’, ‘‘medium’’ or ‘‘much’’, and the chemical industry, pipelines, and so on, as presented in
FFTA consider one of methods of fault diagnosis methods Table 5. Several papers have been presented that expound
can represent this qualitative measures (Zhu and Yu 2002). the virtues of fault-tree analysis and its applications.
The process of fault tree construction is a process of Numerous researchers use this benefit of FFTA to
searching the relationship between faults and its fault rea- employ linguistic expression to estimate the probability of
sons. The process of describing the logical relationship the occurrence of the event and assess the failure of the
between fault combination of actual system and transmis- system (Ding and Yu 2005; Pan and Wang 2007). Mohan
sion abstractly applying graphic symbol of the fault tree et al. (2003) and Pan and Wang (2007) used the FFTA to
(incident symbols and logic symbols). The fault is designed assess the failure rate of channel and bridge construction.
by deductive method, which manual analyze fault reasons Others used the linguistic variables to overcome the diffi-
and process, to find all possible reasons of each incidents culties in FTA for the machine-man basic events. Human
from top event to bottom event which have related in step performance is affected by many Performance Shaping
by step (Hongxia et al. 2008). Factors, e.g. stress level, task knowledge and experience,
Therefore, this is a good advantage in diagnosis in order and congnitionability. In fact, the unpredictability of
to connect all the component with symptoms as shown in human performance as well as the vagueness of system
Fig. 4. Fuzzy fault diagnosis is an approach that is based on phenomena lead to use the fuzzy approach to make it easy
fuzzy mathematics using fuzzy relation matrix and the (Zadeh 1988).
principle of maximal membership and threshold, according Some researchers (Mokhtari 2011; Ferdous 2011; Fer-
to the relationship between fault reasons and symptoms to dous 2011) used FFTA and FETA and integrated them by
diagnose the reasons on the base of considering all symp- using a bow-tie framework to analyze the risk assessment.
toms as shown in Fig. 4 (Ding and Yu 2005; Yao et al. Also FFTA can be used as early warning and emergency
2010; Hurdle et al. 2212; Hongxia et al. 2008; Isermann response where system parameters can be continuously
2006; Wang et al. 2009). monitored and it can evaluate in real time, Dokas et al.
Gmytrasiewicz et al. (1990) proposed a two-step (2009) and Szabo (2011) used this task in context of which
approach for fuzzy failure diagnostics. The first step intro- in turn relies on the ability to accurately risk estimation and
duces a causality matrix relating the symptoms to failure management. On the other hand, in maintenance activities
modes, which are constructed as combinations of basic Kumar et al. (2008) used FFTA as analytical tool for
events of the tree. The fuzzy matrix equation is then solved identifying the predominant failure modes to decide pre-
for the failure vector. A fuzzy matrix relation connecting the dictive maintenance tasks on process equipment’s.

123
28 Int J Syst Assur Eng Manag (Jan-Mar 2013) 4(1):19–32

Fig. 4 Fault–symptom relationship (Wang et al. 2009)

Table 5 Applications of FFTA


System analyzed by fault tree References
analysis
Traditional electrical system Misra and Soman (1995), Bowles and Pelaez (1995), Chanda and
Bhattacharjee (1998), Wu et al. (2007), Xiao-lin et al. (2010), Kumar
et al. (2008), Wu et al. (2007), Szabó et al. (2010), Hong (2008), Szabo
et al. (2011), and Chang (2009)
Traditional mechanical Sawyer (1994), Yuliang and Tiejun (2010), Yao et al. (2010), Yao (2010),
system Jinhua et al. (2011), Yao (2011), Li (2011), and Zhang (2010)
Ammonia storage tank Deshpande and Khanna (1995) and Deshpande (2011)
Robot motion Lin and Wang (1997)
Nitration unit Khan (2000)
Irrigation canal systems Mohan et al. (2003)
Electroplating system Batzias and Batzias (2003)
Biosensor Batzias and Siontorou (2004) and Siontorou (2008)
Civil construction Pan and Wang (2007), Zhao et al. (2010)
engineering
The fire water supply system Mao (2010)
PCBA process Shu et al. (2006)
Landfill operations Dokas et al. (2009)
Gas transmission pipelines Ding and Yu (2005)
Water tank system Hurdle et al. (2212)
Communication control Peng et al. (2008)
system
Emergency shutdown system Cheng et al. (2009)
GIS Hu and Ge (2010)
LPG tanks Rong and Xin (2010)
Cooler process plant Yang et al. (1995)
Ports and offshore terminals Mokhtari (2011)
risk analysis
Biomonitoring surveys Siontorou (2011)
Hoisting system Yao and Zhang (2010)
Automatic gun, weapon Chang (2006)
systems
Nuclear power plant Guimarees and Ebecken (1999), Gmytrasiewicz et al. (1990), Purba (2010),
and Guimarães and Lapa (2008)
Computer security system Kumar (2012)
Risk analysis Szabo et al. (2011), Wang (2011), Hu (2413), Wang (2011), Abdelgawad
(2011), Mentes (2011), Batzias (2010), Celik (2010), and Balog and Berta
(2001)
Construction projects delay Al Humaidi (2010)

123
Int J Syst Assur Eng Manag (Jan-Mar 2013) 4(1):19–32 29

8 Conclusions amount of accurate reliability oriented information. The pro-


posed method improves the analysis of the system and the
FTA is widely used to assess the operational performance, judgment ability of fault models, and greatly increases the
reliability prediction, lifetime, and system safety of various correction of fault component decision.
complex systems involved in, a nuclear reactor, aerospace,
the petrochemical industry, oil and gas transmission, and
other complex electrical mechanical systems.
The first issue as a comparison between FFTA and
References
conventional FTA is the dominance property of FFTA.
Conventional FTA uses the crisp value probabilistic con- Abdelgawad M (2011) Fuzzy reliability analyzer: quantitative
siderations, while linguistic variables and possibility is assessment of risk events in the construction industry using
considered in FFTA. Conventional FTA does not give fuzzy fault-tree analysis. J Constr Eng Manag 137(4):294–302
information concerning the tolerances and variation of the Balog E, Berta I (2001) Fuzzy solutions in electrostatics. J. Electrost
51–52(1–4):409–415
probability values, and the dependencies of the events, Batzias F (2010) Solving river pollution problems by means of fuzzy
while FFTA usually uses a triangular or trapezoidal pos- fault tree analysis., pp 228–233
sibility, and takes into account the uncertainty in calcula- Batzias A and Batzias F (2003) Fuzzy fault tree analysis as a
tions. In conventional FTA, the importance of the basic mechanism for technical support to small/medium electroplaters
on a quasi online/real-time basis. In: IEEE international confer-
event is measured based on the direct contribution to the ence on industrial technology, vol 1, p 36–41
top event, while in FFTA, the contribution of uncertainties Batzias F and Siontorou CG (2004) Investigating the causes of
also is considered. biosensor SNR decrease by means of fault tree analysis. In:
Further in FTA, it might be necessary to consider pos- Presented at instrumentation and measurement technology
conference, IMTC 04. Proceedings of the 21st IEEE, 2004
sible failure of components even if they have never failed Bian X, Mou C, Yan Z and Xu J (2009) Reliability analysis of AUV
before and there is no data available associated with the based on fuzzy fault tree. In: Presented at international
possible failure. The use of FFTA makes it possible to conference on mechatronics and automation, 2009
consider any possible failure events, using expert elicita- Bowles JB, Pelaez CE (1995) Application of fuzzy logic to reliability
engineering. Proc IEEE 83(3):435–449
tion and possibility approach. FFTA has been found Cai KY (1996a) Introduction to fuzzy reliability. Kluwer Academic,
capable to handle the linguistic variables and the impre- Boston
cision of the uncertainties associated with the modeling of Cai KY (1996b) System failure engineering and fuzzy methodology
failures and their dependency. an introductory overview. Fuzzy Sets Syst 83(2):113
Celik M (2010) A risk-based modelling approach to enhance shipping
In addition, FTA is not suitable where available data are accident investigation. Saf Sci 48(1):18–27
insufficient for statistical inferences, or the data show a Chanda R, Bhattacharjee PK (1998) A reliability approach to
large variation. Fuzzy methodology might be the only transmission expansion planning using fuzzy fault-tree model.
resort when little quantitative information is available Electr Power Syst Res 45(2):101–108
Chang J (2006) The reliability of general vague fault-tree analysis on
regarding fluctuations in the parameters. weapon systems fault diagnosis. Soft Comput 10(7):531–542
In fact, the experiences of field experts provide an Chang K (2009) A novel general approach to evaluating the PCBA
effective database supporting the estimation of required for components with different membership function. Appl Soft
data, although they have to face the numerous conflicting Comput 9(3):1044–1056
Chang S, Chang C (2003) A fuzzy-logic based fault diagnosis strategy
evaluations. Actually, the management of the large number for process control loops. Chem Eng Sci 58(15):3395–3411
of tangible and intangible attributes that must be taken into Chen L, Shinan C (2011) An approach of fault diagnosis for
account represents the main complexity of the problem. electronic system of aircraft based on trapezoid fuzzy fault tree.
Application of fuzzy set theory makes it possible to elicit Artificial Intelligence, Management Science and Electronic
Commerce (AIMSEC), 2011 2nd International Conference on.
the expert judgment which is often given in natural lan- Deng Leng, China, IEEE
guage as linguistic variables. Cheng S, Lin B, Hsu B, Shu M (2009) Fault-tree analysis for liquefied
Building a fuzzy relation matrix through FFTA makes it natural gas terminal emergency shutdown system. Expert Syst
possible to identify the relationship between failure modes, Appl 36(9):11918–11924
Cremona YGC (1997) The possibilistic reliability theory: theoretical
fuzzy symptoms and basic events, for diagnosis purposes. It aspects and applications. Struct Saf 19:173–201
takes into account the uncertainties and fuzziness of the Weber DP (1994) Fuzzy fault tree analysis. In: Fuzzy systems 1994b,
symptoms and facilitates the effective use of information IEEE World congress. Proceedings of the third IEEE conference
represented by FFTA. Combining fuzzy theory with FTA on computational intelligence, vol 3, p 1899–1904
Deshpande A (2011) Fuzzy fault tree analysis: revisited. Int J Syst
makes it possible to diagnose faults efficiently and can be Assur Eng Manag 2(1):3–13
easily designed with an aim of online prevention of the fault. Deshpande AW, Khanna P (1995) Fuzzy fault tree analysis: case
The review reveals the effectiveness of the FFTA in studies reliability and safety analysis under fuzziness. Physica,
comparison with conventional FTA, when there is inadequate Heidelberg

123
30 Int J Syst Assur Eng Manag (Jan-Mar 2013) 4(1):19–32

Ding Y, Yu D (2005) Estimation of failure probability of oil and gas Huang HZ, Xin Tonga, Zuo Ming J (2004) Posbist fault tree analysis
transmission pipelines by fuzzy fault tree analysis. J Loss Prev of coherent systems. Reliab Eng Syst Saf 84(2):141
Process Ind 18:83–88 Hurdle EE, Bartlett LM, Andrews JD (2007) System fault diagnostics
Dokas I, Karras D, Panagiotakopoulos D (2009) Fault tree analysis using fault tree analysis. Proceedings of the Institution of
and fuzzy expert systems: early warning and emergency Mechanical Engineers, Part O: J Risk Reliab 221(1):43–55
response of landfill operations. Environ Model Softw 24:8–25 Isermann R (1997) Supervision, fault-detection and fault-diagnosis
Dong-an Z, Jun-jie Z and Ye-wei Z (2010) Risk analysis of shield methods—an introduction. Control Eng Pract 5(5):639–652
tunnel segment failure based on fuzzy fault tree method. In: Isermann R (2006) Fault-diagnosis systems. Springer, Berlin
Presented at (ICNC), 2010 sixth international conference on Jinhua M, Yanfeng L, Haiqing L, Weiwen P and Hong-Zhong H (2011)
natural computation 2010 Reliability analysis of CNC hydraulic system based on fuzzy fault
Dunyak JP and Wunsch D (1998) Fuzzy probability for system tree. In: Presented at 2011 international conference on quality,
reliability. In: Presented at proceedings of the 37th IEEE reliability, risk, maintenance, and safety engineering (ICQR2MSE)
conference on decision and control 1998 2011
El-Iraki A and Odoom ER (1998) Fuzzy probist reliability assessment Kenarangui R (1991) Event-tree analysis by fuzzy probability. Reliab
of repairable systems. In: Presented at fuzzy information IEEE Trans 40:120–124
processing society—NAFIPS, 1998 conference of the North Khan FI (2000) Analytical simulation and PROFAT II: a new
American methodology and a computer automated tool for fault tree
Ferdous R (2011a) Fault and event tree analyses for process systems risk analysis in chemical process industries. J Hazard Mater 75(1):1
analysis: uncertainty handling formulations. Risk Anal Kim C (1996) Multilevel fault tree analysis using fuzzy numbers.
31(1):86–107 Comput Oper Res 23(7):695–703
Ferdous R, Khan F, Sadiq R, Amyotte P, Veitch B (2011) Analyzing Kumar N (2012a) Reliability analysis of waste clean-up manipulator
system safety and risks under uncertainty using a bow-tie using genetic algorithms and fuzzy methodology. Comput Oper
diagram: an innovative approach. Process Saf Environ Prot Res 39(2):310–319
91(1–2):1–18 Kumar M (2012b) Reliability analysis of computer security system
Ferdous R, Khana F, Veitcha B, Amyotte PR (2009) Methodology for based on intuitionistic fuzzy fault tree. Adv Mater Res
computer aided fuzzy fault tree analysis. Process Saf Environ 403–408:3495–3502
Prot 87(4):217–226 Kumar EV, Chaturvedi SK, Deshpande AW (2008) Failure Proba-
Fujino T and Hadipriono FC (1994) New gate operations of fuzzy bility Estimation using Fuzzy Fault Tree Analysis (FFTA) with
fault tree analysis. In: Proceedings of the third IEEE conference PDM Data in Process Plants. Int J Perform Eng 4:271
on fuzzy systems, vol 2, p 1246–1251 La S, Jianping L and Min Q (2008) Study on applying fault tree
Furuta H, Shiraishi N (1984) Fuzzy importance in fault tree analysis. analysis based on fuzzy reasoning in risk analysis of construction
Fuzzy Sets Syst 12(3):205–213 quality. In: International conference on risk management &
Mao G, TU J and Du H (2010) Reliability evaluation based on fuzzy engineering management, ICRMEM ‘08, p 393–397
fault tree. In: 2010 IEEE 17th international conference on Lee WS, Grosh DL, Tillman FA, Lie CH (1985) Fault tree analysis,
industrial engineering and engineering management (IE&EM), methods, and applications—a review. IEEE Reliab Trans R-34:194–
p 963–966 203
GazdikI I (1985) Fault diagnosis and prevention by fuzzy sets. IEEE Li N (2011) ‘‘Research of the calculation method structure system
Trans Reliab 34(4):382 fuzzy reliability based on the fault tree’’. Adv Mater Res
Geymayr JAB (1995) Fault-tree analysis: a knowledge-engineering 201–203:968–973
approach. IEEE Trans Reliab 44(1):37–45 Liang G, Wang M (1993) Fuzzy fault-tree analysis using failure
Gmytrasiewicz P, Hassberger JA, Lee JC (1990) Fault tree based possibility. Microelectron Reliab 33(4):583–597
diagnostics using fuzzy logic. IEEE Trans Pattern Anal Mach Lin CT, Wang MJ (1997) Hybrid fault tree analysis using fuzzy sets.
Intell 12:1115–1119 Reliab Eng Syst Saf 58:205–213
Guimarães ACF, Lapa C (2008) Parametric fuzzy study for effects M Al Humaidi (2010) A fuzzy logic approach to model delays in
analysis of age on PWR containment cooling system. Appl Soft construction projects using rotational fuzzy fault tree models.
Comput 8(4):1562–1571 Civ Eng Environ Syst 27(4):329–351
Guimarees A, Ebecken N (1999) Fuzzy FTA: a fuzzy fault tree Ulieru M From fault trees to fuzzy relations in managing heuristics
system for uncertainty analysis. Ann Nucl Energy 26(6):523–532 for technical diagnosis. In: The IEEE International conference on
Guohuan L, Yuan Z and Zheng Y (2010) Study of hybrid intelligent systems, man and cybernetics. part 2 (of 5), Le Touquet, Fr,
fault diagnosis. In: Presented at 2nd International Asia confer- 10/17–20/93 ol. 21993
ence on, informatics in control, automation and robotics (CAR), Mentes A (2011) An application of fuzzy fault tree analysis for spread
2010 mooring systems. Ocean Eng 38(2–3):285–294
Gupta S, Bhattacharya J (2007) Reliability analysis of a conveyor Misra kB, Soman kP (1995) ‘‘Multistate fault tree analysis using
system using hybrid data. Qual Reliab Eng Int 23(7):867 fuzzy probability vectors and resolution identity’’ reliability and
Hatoyama Y (1979) Reliability analysis of 3-state systems. Reliab safety analysis under fuzziness. Physica, Heidelberg
IEEE Trans 28(R-5):386–393 Misra KB, Weber GG (1989) A new method for fuzzy fault tree
He L, Huang H and Zuo MJ (2007) Fault tree analysis based on fuzzy logic. analysis. Microelectron Reliab 29:195
In: Reliability and maintainability symposium, RAMS ‘07. p 7–82 Mohan S, Elango K and Sivakumar S (2003) Evaluation of risk in
Hong YY, Lee LH, Cheng HH (2008) Application of fuzzy fault-tree canal irrigation systems due to non-maintenance using
analysis to assess the reliability of a protection system for a fuzzy fault tree approach. In: Presented at INDIN 2003
switchyard. Int J Emerg Electr Power Syst 9(4). doi: proceedings IEEE international conference on industrial infor-
10.2202/1553-779X.1944 matics, 2003
Hu J (2011) Risk identification of sudden water pollution on fuzzy Mokhtari K (2011) Application of a generic bow-tie based risk
fault tree in beibu-gulf economic zone. Proc Environ Sci analysis framework on risk management of sea ports and
10:2413–2419 offshore terminals. J Hazard Mater 192(2):465–475

123
Int J Syst Assur Eng Manag (Jan-Mar 2013) 4(1):19–32 31

NASA Office of Safety and Mission Assurance (2002) Fault tree Tong W, Guangyu T, Bo ZQ and Klimek A (2007) Fuzzy set theory
handbook with aerospace applications. NASA Headquarters, and fault tree analysis based method suitable for fault diagnosis
Washington DC of power transformer. In: International conference on intelligent
Onisawa T (1988) An approach to human reliability in man-machine systems applications to power systems, ISAP 2007; p 1–5
systems using error possibility. Fuzzy Sets Syst 27(2):87–103 Tong W, Guangyu T, Yi L and Jun W (2007) Study on power
Page LB, Perry JE (1994) Standard deviation as an alternative to transformer fault diagnosis method based on fuzzy tree. In:
fuzziness in fault tree models. IEEE Trans Reliab 43(3):402 Presented at international power engineering conference, IPEC
Pan H, Huang J and Liu G (2008) Fault diagnosis of circuit board 2007
based on fault tree. In: Presented at 10th international conference Ulieru M (1994) Diagnosis by approximate reasoning on dynamic
on control, automation, robotics and vision, ICARCV 2008 fuzzy fault trees. In: Presented at IEEE world congress on
Pan N and Wang H (2007) Assessing failure of bridge construction computational intelligence proceedings of the third IEEE
using fuzzy fault tree analysis. In: Fourth international confer- conference on fuzzy systems, 1994
ence on fuzzy systems and knowledge discovery, p 96–100 Wang YF (2011) Quantitative risk analysis model of integrating fuzzy
Pan H, Yun W (1997) Fault tree analysis with fuzzy gates. Comput fault tree with bayesian network. In: IEEE international confer-
Ind Eng 33(3–4):569–572 ence on intelligence and security informatics (ISI), 2011;
Peng Z, Xiaodong M, Zongrun Y and Zhaoxiang Y (2008) An p 267–271
approach of fault diagnosis for system based on fuzzy fault tree. Wang Y (2011) Quantitative risk assessment through hybrid causal
In: International conference on multimedia and information logic approach. Proc Insti Mech Eng. 2011; 225(3): 323–332.
technology, MMIT ‘08, 2008; p 697–700 Wang F and Wu D (2007) Design and implementation of a missile
Purba JH (2010) A hybrid approach for fault tree analysis combining fault diagnosis system based on fault-tree analysis. In: Presented
probabilistic method with fuzzy numbers’’ lecture notes in at machine international conference on learning and cybernetics,
computer science, vol 6113 LNAI(PART 1)., pp 194–201 2007
Rasmussen NC (1975) Reactor safety study. WASH-1400, US Wang Y, Li L, Chang M, Chen H, Dong X, Ren Y, Li Q and Liu D
Nuclear Regulatory Commission (2009) Fault diagnosis expert system based on integration of
Renjith V, Madhua G, Nayagam V, Bhasi AB (2010) Two-dimen- fault-tree and neural network. In: Presented at international
sional fuzzy fault tree analysis for chlorine release from a chlor- conference on computational intelligence and software engi-
alkali industry using expert elicitation. J Hazard Mater neering, CiSE 2009
183(1–3):103–110 Weber DP (1994a) Fuzzy weibull for risk analysis’’ reliability and
Rong W and Xin D (2010) Application of fuzzy fault tree analysis on maintainability symposium, proceedings, annual, 1994; p 456–461
burning and blasting of LPG tank. In: International conference Wei L, Yanjiao J and Yumin S (2009) Assessment of grid
on logistics systems and intelligent management, p 1093–1096 construction project risk of process based on fuzzy-event tree-
Sawyer JP (1994) Fault tree analysis of fuzzy mechanical systems. fault tree.In: Presented at international symposium on informa-
Microelectron Reliab 34(4):653–667 tion engineering and electronic commerce, IEEC ‘09, 2009
Sharma U (1993) Use of recursive methods in fuzzy fault tree Xiao-lin L, Yan-xia Z and Zeng-hui Z, (2010) Research on application of
analysis: an aid to quantitative risk analysis. Reliab Eng Syst Saf fuzzy fault tree analysis in the electronic equipment fault diagnosis.
41(3):231–237 In: The 2nd international conference on computer and automation
Shengwu H and Xiaosan G (2010) GIS reliability analysis based engineering (ICCAE), 2010; p 65–67
trapezoid fuzzy fault tree, In: 18th international conference on Xie G (2010) Fault diagnosis platform for radar circuit based on virtual
geoinformatics, 2010; p 1–5 instrument. In: International conference on measuring technology
Shu M, Cheng C, Chang J (2006) Using intuitionistic fuzzy sets for and mechatronics automation (ICMTMA), p 245–248
fault-tree analysis on printed circuit board assembly. Microelec- Yao Z (2010) ‘‘Design of coal-mechanical online fault diagnosis
tron Reliab 46(12):2139–2148 based embedded system’’ Informatics in Control. Autom Robot
Singer D (1990) A fuzzy set approach to fault tree and reliability (CAR) 1:178–181
analysis. Fuzzy Sets Syst 34(2):145 Yao Z (2011) Fault diagnosis analysis of gear based on fuzzy fault
Siontorou CG (2008) Carbohydrate detection failure analysis via tree. Adv Mater Res 204–210:1994–1997
biosensoring. IEEE Trans Instrum Meas 57(12):2856–2867 Yao C and Zhang Y (2010) T–S model based fault tree analysis on the
Siontorou CG (2011) Error identification/propagation/remediation in hoisting system of rubber-tyred girder hoister. In: Presented at
biomonitoring surveys—a knowledge-based approach towards WASE international conference on information engineering
standardization via fault tree analysis. Ecol Ind 11(2):564–581 (ICIE), 2010
Song H (2009) Fuzzy fault tree analysis based on T-S model with Yao Z, Lou G, Song X and Zhou Y (2010) On-line fault diagnosis
application to INS/GPS navigation system. Soft Comput study for roller bearing based on fuzzy fault tree. In: Presented at
13(1):31–40 2nd international Asia conference on informatics in control,
Sui Y (2011) Reliability assessment of urban anti-disasters system automation and robotics (CAR), 2010
based on fuzzy fault tree analysis In: 2nd IEEE international Yi R and Leixing K (2011) Fuzzy multi-state fault tree analysis based
conference on emergency management and management sci- on fuzzy expert system. In: Presented at 9th international
ences (ICEMMS), p 159–162 conference on reliability, maintainability and safety (ICRMS),
Suresh P, Babar A, Raj V (1996) Uncertainty in fault tree analysis: a 2011
fuzzy approach. Fuzzy Sets Syst 83(2):135–141 Yuliang C and Tiejun Z (2010) Research on the application of fuzzy
Szabó S, Németh B and Kiss I (2010) Eliminating bias in fuzzy fault trees fault tree analysis method in the machinery equipment fault
for electrostatic risk assessment. In: Presented at 4th international diagnosis. In: 2nd international Asia conference on informatics
workshop on soft computing applications (SOFA), 2010 in control, automation and robotics (CAR), 2010; p 84–87
Szabo SV, Kiss I, Németh B, Berta I (2011) Complex system for risk Zadeh LA (1973) Outline of a new approach to the analysis of
assessment in ESD hazardous processes with unusually high complex systems and decision processes. IEEE Trans Syst Man
risks. J Phys Conf Ser 301(1):012037 Cybern 1:28
Tanaka H, Fan LT, Lai FS, Toguchi K (1983) Fault tree analysis by Zadeh LA (1978) Fuzzy sets as a basis for a theory of possibility.
fuzzy probability. IEEE Reliab Trans 32:453–457 Fuzzy Sets Syst 1(1):3

123
32 Int J Syst Assur Eng Manag (Jan-Mar 2013) 4(1):19–32

Zadeh LA (1988) Fuzzy logic. Computer 21(4):83–93 Zimmerman H (1983) Using fuzzy sets in operational research. Eur J
Zhang X (2010) ‘‘Engineering machinery engine system reliability Oper Res 13(3):201–216
analysis’’. Adv Mater Res 139–141:2587–2590 Zong-Xiao Y, Suzuki K, Shimada Y and Sayama H (1995) Fuzzy
Zhang Z, Wang Z and Zhang B (2008) Studies on median value of fault diagnostic system based on fault tree analysis. Proceedings
fuzzy numbers based on confidence level. In: Presented at of 1995 IEEE international conference on international fuzzy
international conference on machine learning and cybernetics, engineering symposium, 1995; vol 1: p 165–170
2008 Zonouz SA and Miremadi SG (2006) A fuzzy-monte carlo simulation
Zhu DQ, Yu SL (2002) Survey of knowledge—based fault diagnosis approach for fault tree analysis. In: Reliability and maintain-
methods. J Anhui Univ Technol 19(3):197 ability symposium, RAMS 06, 2006

123

View publication stats

You might also like