Safety Science 101 (2018) 231–247

Contents lists available at ScienceDirect

Safety Science
journal homepage: www.elsevier.com/locate/safety

Fuzzy based risk prioritisation in an auto LPG dispensing station MARK

a b,⁎ c d
A. Maniram Kumar , S. Rajakarunakaran , P. Pitchipoo , R. Vimalesan
a
Department of Mechanical Engineering, Dr. Sivanthi Aditanar College of Engineering, Tiruchendur 628215, Tamil Nadu, India
b
Department of Mechanical Engineering, Ramco Institute of Technology, Rajapalayam 626117, Tamil Nadu, India
c
Department of Mechanical Engineering, P.S.R. College of Engineering, Sivakasi 626140, Tamil Nadu, India
d
General Manager, (Retired), HMT Machine Tools Limited, Pinjore, India

A R T I C L E I N F O A B S T R A C T

Keywords: A fuzzy rule based inference model assessing the failure modes for risk ranking in FMEA to manage risks and
Fuzzy logic make maintenance decisions is applied to a LPG refuelling station in this paper. Normally in FMEA, risk priority
FMEA number (RPN) is determined by multiplication of feature scores that are inferred from the degree or probability
Fuzzy FMEA of occurrence, severity and non detection of the problem, without taking into consideration the relative im-
Risk priority ranking
portance of factors. In fuzzy approach, linguistic assessment of factors is evaluated to obtain risk priority
Grey relational analysis
number. A rule based fuzzy inference engine generates the priority ranks of identified failure modes. Direct
evaluation is possible with the aid of grey theory by assigning weights to the features in the absence of expertise
to develop an inference rule base. The GRA approaches can solve the problem of risk prioritising, which needs
expertise to develop rule base while building the fuzzy inference system. By applying fuzzy FMEA and fuzzy logic
with grey relational approach (GRA), expert linguistic opinions are used to rank the identified failure modes and
the results are presented. The risk of the failure modes are ranked in an inclusive approach based on the fuzzy
domain projections. It is effective and feasible to handle various types of uncertainties, such as incompleteness,
fuzziness, imprecision, and so on, in the risk analysis process.

1. Introduction (Chen et al., 2014). Dynamic sequential accident models (DSAMs) uti-
lises precursor data to estimate the posterior risk profile quantitatively.
A significant number of studies have applied fuzzy logic to identify, DSAM also offers updates on the failure probabilities of accident bar-
quantify risks, enhance safety and also for maintenance functions. riers and the prediction of future end states (Al-shanini et al., 2014). A
Efforts were taken to assess the state of big diesel engine by calculating left–right (L-R) bell-shaped fuzzy number is applied for barrier and
the fuzzy distance between the fuzzy vector and standard fuzzy vector operational risk analysis (BORA-Release) method and an α-cut opera-
(Wei et al., 2009). Guo et al. (2009) developed a model, in which a tion is introduced to conduct the arithmetic operations of the fuzzy
three-layer BP neural network is combined with fuzzy comprehensive number, and a defuzzification method with total integral value is
evaluation and demonstrated with an application in an ethylene plant chosen to match the α -cut operations and acquire complete informa-
for allocating maintenance resources based on criticality evaluation. tion for the fuzzy numbers (Huang et al., 2015).
Fuzzy Probabilities of basic events and output events are derived In FMEA literatures based on fuzzy sets, the approaches have mostly
from fuzzy input data and fuzzy utility value (FUV) is introduced to concerned with the fuzzy inference approach by using a rule-base
perform risk assessment for natural gas pipelines (Shahriar et al., 2012). (Bowles and Peláez, 1995; Chin et al., 2008; Guimarães and Lapa,
Works on risk based maintenance (RBM) optimization suggest a fuzzy 2004a, 2004b, 2007; Pillay and Wang, 2003; Sharma et al., 2005; Tay
inference system (FIS) to minimize the suboptimal prioritizations of and Lim, 2006; Xu et al., 2002; Li et al., 2010). Linguistic terms were
functions in the FFR (functional failure risk) analysis performed using assigned to the factors and considering them as inputs, an if-then rule
an illustrative tailor-made risk matrix. (Ratnayake, 2014). A fuzzy logic base was created for risk evaluation. The outputs were termed as risk
system (FLS) was proposed for risk modelling to overcome the un- (Chin et al., 2008; Guimarães and Lapa, 2004a, 2004b), the critically
certainty of the RBM components in an oil refinery (Sa’idi et al., 2014). failure mode (Xu et al., 2002), priority for attention (Pillay and Wang,
The basic event occurrence probability was quantified and investigated 2003), and fuzzy RPN (Sharma et al., 2005; Xu et al., 2002) in the fuzzy
by a hybridised approach using fuzzy set theory and weight analysis FMEA studies which consider the fuzzy rule-base approach. A FMEA

⁎
Corresponding author.
E-mail address: [email protected] (S. Rajakarunakaran).

http://dx.doi.org/10.1016/j.ssci.2017.09.011
Received 20 July 2016; Received in revised form 3 May 2017; Accepted 18 September 2017
0925-7535/ © 2017 Elsevier Ltd. All rights reserved.

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

based on fuzzy logic was implemented in diesel engine systems (Xu plant personnel and public. Various additives are added to detect its
et al., 2002). dispersion. Meticulous understanding of the failure of systems and its
Guimarães and Lapa (2004a) applied fuzzy concepts to analyze the components, the causes, the consequences and the actions needed to
failures in a chemical and volume control system of pressurised water avoid them is highly recommended for loss prevention in process in-
reactor. They also implemented a fuzzy logic system for risk ranking dustries. Fuzzy logic combined with expert elicitation was employed in
enhancement purpose in an auxiliary feed water system. (Guimarães a LPG refuelling station to quantify basic event failure probabilities by
and Lapa, 2004b). Guimarães and Lapa (2007) adopted fuzzy rule based qualitative information processing without reliance on historical
inference for risk analysis in nuclear power plant as a contemporary quantitative failure statistics. (Rajakarunakaran et al., 2015).
approach. Yeh and Hsieh (2007) conducted fuzzy FMEA assessment for In this work a Fuzzy FMEA model is proposed and applied to an auto
a sewage plant. Wang et al. (2009) made risk evaluation in FMEA using LPG refuelling station. Based on the literature review, the application of
fuzzy weighted geometric mean. Sharma et al. (2008) developed a Fuzzy based FMEA to an auto LPG refuelling station is a important task
fuzzy logic-based decision support system. required for defined safety decision making process. The proposed
Tay and Lim (2006) simplified the FMEA methodology based on framework of fuzzy FMEA exhibits two advantages: (i) an ability to
fuzzy logic by proposing a generic method to reduce the number of assess the failure modes of engineering systems using expert opinions
rules for inference system, provided by experts to model fuzzy RPN for articulated in linguistic terms; (ii) an ability to handle various types of
performance evaluation in a semiconductor manufacturing plant. uncertainties, such as imprecision, fuzziness, incompleteness of quan-
Pelaez and Bowles (1996) applied fuzzy cognitive map to FMEA. Pillay titative risk assessments. The rule base development in the case of fuzzy
and Wang (2003) proposed FMEA with fuzzy reasoning and grey rela- FMEA can be replaced with a multi-criteria decision making problem
tion for marine industry to conquer the traditional weaknesses in risk (MCDM) approach and the techniques may be applied to yield a ranking
evaluation. Braglia et al. (2003) used a fuzzy TOPSIS in FMEA. Keskin strategy. Grey relational analysis (GRA) is selected as an approach to
and Ozkan (2009) proposed FMEA supported by fuzzy ART algorithm. this problem to prioritise and rank the events in the absence of rule
A product design system based on fuzzy FMEA was developed by base.
Chin et al. (2008). Abdelgawad and Fayek (2011) implemented risk The uniqueness or novelty of this study are: (i) a use of linguistic
management in the construction industry through combined fuzzy values in terms of expert opinion, (ii) development of fuzzy interference
FMEA and fuzzy AHP. Tay and Lim (2010) enhanced FMEA metho- rules to apply for an auto LPG refuelling station (iii) an integration of
dology with fuzzy inference techniques and evaluated the performance the proposed fuzzy rules into the inference system (iv) determination of
of the fuzzy RPN models in the test handler process of a semiconductor the fuzzy RPN by defuzzification and thereby achieving the systems risk
manufacturing plant. Hao-Tien Liu and Yieh-lin Tsai (2012) utilised priority ranking (v) Application of linguistic expert opinions for multi
fuzzy analytic network process method to identify critical hazard causes criteria decision making (vi) Incorporation of grey theory for direct
and types and utilised fuzzy FMEA based inference approach to assess evaluation of failure modes by assigning weights to the features con-
the causes for a telecom engineering concern in South Taiwan. In a sidering the relative importance of factors (vii) Rule based inference is
chemical process industry, risks are prioritized using risk priority ma- replaced by the application of a grey relational framework. To reveal
trix and mitigation strategies are selected based on FMEA. Fuzzy esti- the feasibility of the proposed frameworks, the outputs generated are
mates obtained for the risk factors and bow-tie analysis are used to compared with the conventional FMEA risk analysis process.
calculate the aggregated risk probability and impact (Aqlan and
Mustafa, 2014). Mentes and Ozen (2015) integrated Ordered Weighted 2. Fuzzy FMEA
Geometric Averaging (OWGA) and Generalized Mixture Operators
(GMOs) to improve safety during design. The proposed approach is FMEA is a tool widely used to make out the probable failure modes
experimented with a motor yacht fuel system. Zhou and Thai (2016) of a process or a product. FMEA was first projected in 1963 by NASA
applied fuzzy set and grey relational analysis to evaluate the failure as a formal system analysis method for their reliability needs. Then in
modes in oil tankers transporting bulk crude oil. Relative ranking is 1977, it was put into practice by Ford Motors (Gilchrist, 1993). Since it
reflected by grey theory approach. The results show the both the emerged as a influential tool extensively used for risk analysis and re-
methods work similar. liability studies of systems in a broad scope of industries, together with
Researchers have conducted numerous investigations to enhance construction, automotive, aerospace, nuclear, and electro-technical.
the FMEA methodology by integrating artificial intelligence ap- Table 1 lists the FMEA scale to compute the three factors, occurrence
proaches. There are significant efforts made in FMEA literatures to (O), severity (S) and detection (D) to calculate RPN value which is the
overcome the shortcomings of the traditional RPN (Wang et al., 2009; product of the factors. The table show that the customary FMEA mea-
Liu, 2013). The risk parameters O, S, and D in fuzzy FMEA are ex- sures the probability of occurrence, severity and the probability of non
pressed as linguistic fuzzy terms in studies with expert opinions. The detection using five scales and scores of 1–10. The failures can be
traditional 10 point scale for O, S, and D risk factors are interpreted as eliminated or reduced by ranking the failures for corrective action by
linguistic variables. virtue of their risk implication.
Increased public awareness for environmental protection and The well accepted FMEA safety analysis method suffer from several
stricter pollution control norms propagating use of efficient fuels have impediments. In traditional approach, the risk factors are understood to
steadily pulled to the trend for LPG conversion for auto fuel and di- be similarly important (Braglia, 2000). Many operation and main-
version of domestic fuel to auto fuel at a steady growth @ 8% p.a. in tenance experts were found to give more weight to non detection factor.
LPG Consumption in India (Apurva Chandra, 2010). Automotive Li- The gathered failure data may often be unreliable or partial. Therefore
quefied Petroleum Gas (LPG) (BIS 14861) is a mixture of light hydro- based on knowledge and expertise of experts the risk factors are as-
carbons primarily Predominantly 100% Propane or combinations of sessed. The risk factors are hard to be calculated exactly due to the often
Propane: Butane ‘mixes’ derived from petroleum, which is gaseous at questionable and debatable mathemaical formula for calculating RPN
ambient temperature and atmospheric pressure, is liquefied at ambient mathematically. No rationale to multiply O, S, and D is found. From
temperature with application of moderate pressure. LPG due to its in- Table 1 it can be found that a non linear occurrence scale is considered
herent properties is susceptible to fire, explosion and other hazards. with linear severity and non detection scales. Multiplying the factor
Such hazards can have an impact on the process, property, equipment, scores after conversion of probability to a linear scoring system is

232

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

Table 1
FMEA scale for occurrence (O), severity (S), not detection (D).

Linguistic description Occurrence Severity Non detection Score

Very High Persistent failures Hazardous without warning Absolute certainty 10

Persistent failures Hazardous with warning Probably not detected 9
High Frequent failures Very high Poor chance of detection 8
Frequent failures High Poor chance of detection 7
Medium Occasional failures Moderate May detect 6
Occasional failures Low May detect 5
Low Occasional failures Very low Good chance to detect 4
Relatively few failures Minor Good chance to detect 3
Almost none Relatively few failures Very minor Almost certain to detect 2
Failure is unlikely None Certain to detect 1

supposed to have source inaccuracies and uncertainties in the deriva- • It combines expert experience and knowledge used in an FMEA
tion. The RPNs are found to be greatly distributed at the bottom with study in an organized way. This methodology presupposes that the
many holes and are not continuous in the 1–1000 scale. Interpretation association is non-linear among the RPN and the severity, occur-
problems coexist with the non continuous distribution. Again it is found rence, and not detection, instead of linear as in the customary case.
that the risk magnitude may be completely dissimilar with identical • The applicability of FMEA is improved by linguistic terms in fuzzy
value of RPNs, as different sets of O, S, D mathematically yield same methodology by facilitating the experts to allocate a more sig-
results. Carrying out a traditional FMEA has to overcome difficulties nificant value for the aspects considered. So the identified events are
such as knowledge shared within team associates, different risk depic- not ignored due of low RPN.
tion among uniform ratings, their relative risk significance, qualitative • The amalgamation of the input factors is considered during mod-
linguistic description and the subjectivity factor. This may lead to a elling. This solves the issue of RPN scores where the real risk is not
waste of time and resources. To surmount the above downsides and measured (Kumrua and Kumru, 2013).
recover the effectiveness of the customary FMEA methodology, a fuzzy • During modelling, human expertise is incorporated by fuzzy in-
based FMEA approach is developed. The approach then extends the ference techniques y vague and imprecise linguistic expressions. The
analysis using defuzzified linguistic terms. A risk assessment scheme understanding of inference techniques enables the customisation by
supported by fuzzy set theory is anticipated in this work to deal with experts and users.
these difficulties.
Fuzzy logic is a tool for transforming the vagueness of human sen- 3. Fuzzy FMEA with MCDM approach
sation and identification into a mathematical principle. The assessment
information is represented as linguistic terms by fuzzy approach. The Multi-criteria decision making techniques help us take decisions
fuzzy variables significantly have scope for steady changeover/transi- when there are a multitude of demands. In a FMEA analysis, failure
tion amid positions and in view of that, enjoy a inherent potential to modes are preferred to have least severity, minimum occurrence and
convey and handle uncertainties. Works increasingly use linguistic most detection capacity. This gives the risk prioritization enabling a
fuzzy terms for portraying the risk factors S, O, and D. Works like Yang multi-criteria approach decision making approach. The development of
et al. (2008), Gargama and Chaturvedi (2011), Keskin and Ozkan rule base in fuzzy FMEA can be complimented as a multi-criteria ap-
(2009) form good foundation for fuzzy-FMEA methodology. Most of the proach problem and MCDM techniques may be applied to yield a
existing studies in the fuzzy FMEA literature have concerned with the ranking strategy. In our approach we have selected grey relational
fuzzy rule base approach by using ‘If–Then’ rules. Fig. 1 shows an analysis (GRA) as an approach to this problem to prioritise and rank the
overall process diagram of fuzzy FMEA. events in the absence of rule base which is explained below.
Three major steps are to be carried to carry out Fuzzy FMEA as-
sessment using inference rules: 3.1. Grey relational analysis

(i) Fuzzification process converts the risk factors into the fuzzy do- The grey system theory was proposed by Professor J. Deng, and was
main using linguistic variables. In each input class, the linguistic born of the concept of a grey set. This theory is especially suitable for
input is then fuzzified to resolve the membership degree. data with uncertain, multi inputs and discrete properties. It is an ef-
(ii) Fuzzy inference consists of if–then rules representing the interac- fective method used to resolve uncertainty issues, under discontinuous
tions between various failure modes and effects formed by expert data and partial information (Deng, 1982). If the system information is
experience and knowledge. The output of the inference system is known totally, it is a white system and when it is unknown, it is black
RPN in fuzzy domain. system. Systems known partially are called as a grey system. The grey
(iii) Defuzzification process ranks the RPN in the crisp form in fuzzy system concept is shown in Fig. 2.
domain to give the priority for the failure event modes. Let X be the universal set. Then the following Eq. (1) defines a grey
set G (X).
This work is one of the foremost attempts to make the fuzzy FMEA
(X )
approach more appropriate for auto LPG refuelling station. In the ⎧ μu : X → {0,1}
G (X ) =
suggested approach, a rule base is used to position the probable failures ⎨ μ (X ) : X → {0,1}
⎩ 1 (1)
recognized by FMEA, which has similar RPN scores but dissimilar risk
inference. The advantages of the proposed fuzzy rule based approach where (x) ≥ (x); x ∈ X and (x) and μ(x) are the higher and lesser
can be summarized as follows: membership functions. If (x) = (x) then the grey set will become a

233

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

Fig. 1. Fuzzy FMEA process diagram.

fuzzy set. This gives an idea that the grey system theory reflects on Eq. (3).
fuzziness, and handled more flexibly with a fuzzy environment. GRA is (vi) Compute the grey relation coefficient using the Eq. (4).
a vital approach for grey theory in selecting from alternatives based on (vii) Introduce the weight vector of three risk factors. The weighting
decision criteria’s. The information of similar features is considered as a coefficient (βk) is decided based upon the analysis objective and
sequence. By degree of grey relational coefficient, the correlative de- reliability of the data for the parameters (Sf, Ss and Sd) to arrive at
grees involving two sequences are measured. The degree varies be- grey relation degree.
tween zero and one. The choices are made based on the degree. This (viii) Determine the degree of relation using for each failure mode in-
approach is useful for qualitative or quantitative analysis (see Fig. 3). corporating the weighted variables.
The process for carrying out a grey relation analysis in FMEA in- (ix) Rank the priority of risk: the stronger the degree of relation, the
volves several steps described by Liu et al., 2011 as follows: smaller is the effect of the cause.

(i) Express O, S, and D by linguistic terms (expert opinion) and the Initially the triangular membership functions of the three para-
membership functions as fuzzy variables. meters are defined. Five linguistic terms are defined for each parameter
(ii) Obtain crisp number for O, S, and D by defuzzification of fuzzy (O, S and D). The identified failure modes and its probability of oc-
sets. currence, severity and non-detectability are conveyed by linguistic ex-
(iii) Establish comparative series, which reflects the various linguistic pressions. The expert opinions are fuzzified by input triangular mem-
terms and decision factors of the study in a form of a matrix. bership functions.
(iv) Establish standard series, which reflects the ideal or desired level The next step is to provide a crisp number in place of each of the
of all the decision factors in a form of a matrix. assigned linguistic terms i.e. defuzzified form (Almannai et al., 2008).
(v) Calculate the difference between the comparative series and The defuzzified linguistic crisp values forms the comparative series as a
standard series. This is represented in a form of a matrix, D0 from matrix form. The optimal criterion of the parameters generates the

Fig. 2. Concept of grey system.

234

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

Fig. 3. Proposed grey analysis approach.

Fig. 4. LPG refuelling station process flow dia-

grams.

235

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

Table 2
Conventional FMEA results.

Failure ID Failure mode Cause F C D RPN Rank

I.1 Pressure relief malfunction- LPG storage tank Mechanical failure 2 6 2 24 16

I.2 Over pressure and failure of storage tank- LPG storage tank Fill storage on cold day/High temperature environment 2 9 2 36 13
I.3 Pressure relief device failure- LPG storage tank Mechanical failure 2 9 2 36 13
I.4 Piping leak Mechanical failure/corrosion 4 8 2 64 8
I.5 Failure of Storage tank External fire near the storage tank by fuel spill 2 9 2 36 13
I.6 Piping failure- LPG Dispenser Vehicle impact to dispenser 2 8 3 48 11
II.1 Drive away with coupled to dispenser Human error 4 6 2 48 11
II.2 Hose failure- LPG Dispenser Mechanical failure 5 5 2 50 10
II.3 LPG Dispenser-Connection Leak Nozzle or O-ring damaged 5 5 2 50 10
II.4 Vehicle pressure relief device leaks- LPG Dispenser Mechanical failure 2 6 2 24 16
II.5 Nozzles leaks after disconnect- LPG Dispenser Mechanical failure 5 4 2 40 12
III.1 Vehicle tank isolation valve leaks- LPG storage tank Filling Mechanical failure 3 8 2 48 11
Process
III.2 Overfill of storage tank Human error or instrument failure 5 8 2 80 5
III.3 LPG transferring trailer bullet leak Vehicle impact to truck damages LPG piping 2 8 2 32 14
III.4 Unloading hose link leaks Inappropriate connection or mechanical failure 4 8 2 64 8
III.5 Release from connecting hose- LPG storage tank Filling Process Hoses in not vented condition earlier to disconnect- human error 5 4 2 40 12
IV.1 No delivery – LPG Filling Pump Incorrect direction of rotation. 3 5 2 30 15
IV.2 No delivery – LPG Filling Pump Suction filter/strainer blocked or leaking air. 5 8 3 120 1
IV.3 No delivery – LPG Filling Pump No power supply 3 8 2 48 11
IV.4 No delivery – LPG Filling Pump No liquid in storage tank 3 8 2 48 11
IV.5 No delivery – LPG Filling Pump Valves closed or air leaks in suction system. 3 4 2 24 16
IV.6 No delivery – LPG Filling Pump Pump vapour locked 3 5 2 30 15
IV.7 No delivery – LPG Filling Pump Excess internal clearances. 2 8 7 112 2
IV.8 Low output- LPG Filling Pump Pump speed too low 2 4 2 16 17
IV.9 Low output- LPG Filling Pump Cavitations or vaporization on suction side of pump 4 4 4 64 8
IV.10 Low output- LPG Filling Pump Obstruction in suction or discharge pipe 3 4 2 24 16
IV.11 Low output- LPG Filling Pump Air leakage in suction pipes or fittings. 3 4 2 24 16
IV.12 Excessive power consumption- LPG Filling Pump Obstruction in discharge line. 3 4 2 24 16
IV.13 Excessive power consumption- LPG Filling Pump Misalignment between pump and driver 3 4 3 36 13
IV.14 Excessive power consumption- LPG Filling Pump Bearings worn condition 4 5 3 60 9
IV.15 Excessive power consumption- LPG Filling Pump Air leakage in inlet piping 3 5 4 60 9
IV.16 Pump is Noisy- LPG Filling Pump Bearings worn condition 3 6 4 72 7
IV.17 Pump is Noisy- LPG Filling Pump Damaged 'O' ring seals or gaskets 3 8 4 96 4
IV.18 Pump is Noisy- LPG Filling Pump Misalignment between pump and driver causing excessive shaft 3 5 4 60 9
distortion
IV.19 Pump is Noisy – LPG Filling Pump Worn or damaged shaft in seal zone 3 8 4 96 4
V.1 No Vehicle Fill- LPG Dispensing Pump Dispenser malfunctioning or not authorizing 4 8 3 96 4
V.2 No Vehicle Fill- LPG Dispensing Pump Blocked filters 4 8 3 96 4
V.3 No Vehicle Fill- LPG Dispensing Pump Insufficient LPG in tank 4 8 3 96 4
V.4 Low Fuel Flow- LPG Dispensing Pump Restriction in nozzle 4 8 3 96 4
V.5 Low Fuel Flow- LPG Dispensing Pump Dirt/debris contamination in pump 2 5 4 40 12
V.6 Low Fuel Flow- LPG Dispensing Pump Faulty bypass valve 3 5 5 75 6
V.7 Pulsing Flow- LPG Dispensing Pump Insufficient LPG in storage tank or excessive inlet restriction 4 8 2 64 8
V.8 No Fuel Output- LPG Dispensing Pump No power supply to pump or pump contactor fault 4 8 2 64 8
V.9 No Fuel Output- LPG Dispensing Pump Drive Motor fail 3 8 2 48 11
VI.1 Valve Not Opening No power supply 3 5 2 30 15
VI.2 Valve Not Opening – ROV Solenoid failure 3 7 5 105 3
VI.3 Valve Not Opening No adequate pneumatic pressure 4 5 3 60 9
VI.4 Valve Closing Very Slowly-ROV Dirt in valve 2 8 4 64 8
VII.1 Control system not working No power supply 2 6 2 24 16
VII.2 Contactor/ Relay /detector failure Over loading, mechanical failures, improper current rating of 4 6 4 96 4
components
VII.3 Loose connection-Control system Vibration/ Arching 2 6 5 60 9
VII.4 Program fault Program crash 2 8 3 48 11
VII.5 Selector switch wrong selection Operator fault 4 6 3 72 7
VII.6 Program fault Wrong program loaded 2 6 5 60 9

standard series as a matrix. The deviation of the comparative series

from the standard series is obtained and used to determine the grey
relation coefficient. Weighting factors introduced for the three para-
meters are used along with grey relation coefficient to calculate the
degree of grey relation. Degree of grey relation embodies the order of
importance or rank for every identified failure modes.
Comparative series is derived with n decision factors or components
expressed as xi = (xi(1), xi(2), xi(3),…,xi(k)) ∊ X, Where xi(k) denotes
the kth factor of xi. The ‘n’ information series for the three linguistic
variables described in the matrix form as follows;
Fig. 5. Fuzzy triangular membership functions.

236

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

Fig. 6. Input membership functions plots (Occurrence, not detection and severity).

Fig. 7. Output membership functions plots.

Table 3 ⎡ Δ01 (1) Δ01 (2) Δ01 (3) ⎤

Fuzzy numbers of conversion. Do = ⎢⋮ … ⋮ ⎥
⎢ Δ (2) … Δm (3) ⎥
Linguistic variables Triangular fuzzy numbers ⎣ 02 ⎦ (3)

a b c where Δ0j(k) = ||x0(k) − xi(k)|| and xi(k) is the comparative series and
x0(k) is the standard series. The grey relation coefficient, γ{x0(k), xi(k)}
Input Almost None 0 0 0.2 is calculated using following equation for the identified failure modes.
Low 0.1 0.25 0.4
The grey relation coefficient is calculated assuming ζ = 0.5 and as
Medium 0.3 0.5 0.7
High 0.6 0.75 0.9 shown in following equation
Very high 0.8 1 1
minmink |x 0 (k)−xi (k)| + ζ maximaxk |x 0 (k)−xi (k)|
Output None 0 0 0.2 i
x 0(k ),xi (k ) =
Very low 0.1 0.2 0.3 |x 0 (k)−xi (k)| + ζ maxmaxk |x 0 (k)−xi (k)|
i (4)
Low 0.2 0.3 0.4
High Low 0.3 0.4 0.5
Grey relational coefficient for O, S and D ((γf),(γs), (γd)) are calcu-
Low medium 0.4 0.5 0.6
Medium 0.5 0.6 0.7 lated. The weighting coefficient (βk) is decided based upon the analysis
High medium 0.6 0.7 0.8 objective and reliability of the data for the parameters (Sf, Ss and Sd) to
Low high 0.7 0.8 0.9 arrive at grey relation degree. Careful selection of the weighting coef-
High 0.8 0.9 1 ficient is necessary to get proper results and rankings. The weighting
Very high 0.9 1 1
coefficient will have a large influence on the final ranking of the failure
events and has to be selected carefully. The degree of grey relation for
each failure event incorporating the weighted variables can be calcu-
x x (1) x1 (2) x1 (3) ⎤
⎡ 1⎤ ⎡ 1 lated as in Eq. (5),
x = ⎢⋮ ⎥ = ⎢ ⋮ ⋮ ⋮ ⎥
⎣ x2 ⎦ ⎢ ⎥ n
⎣ x n (1) x n (2) x n (3) ⎦ (2)
Γ(x i,x j ) = ∑ βk γ {x i (k ),x j (k )}
k=1 (5)
The defuzzified crisp number is represented as xi(k) respectively for
each linguistic expression. The optimal level of parameters determines The degree of the grey relation coefficient stands for the extent of
the standard series. The safe operation of every system needs the least correlation with the reference sequence. If two sequences are identical,
of occurrence probability, severity and non-detection. The lowest level then the value of the degree of grey relation coefficient equals one. It
values are considered. Then D0 is calculated and reflected as a matrix, also indicates the degree of influence executed by the comparability
which is generated from the difference of comparative and standard sequence on the reference sequence. The degree of relation orders the
series. failure modes in ranks as per ascending order. This ranks the failure

237

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

Table 4
Fuzzy FMEA results with fuzzy RPN.

Sl No. Failure ID Failure mode RPN Rank Fuzzy RPN Fuzzy rank

1 I.1 Pressure relief malfunction- LPG storage tank 24 16 0.3 5

2 I.2 Over pressure and failure of storage tank- LPG storage tank 36 13 0.47 3
3 I.3 Pressure relief device failure- LPG storage tank 36 13 0.47 3
4 I.4 Piping leak 64 8 0.5 2
5 I.5 Storage tank failure 36 13 0.47 3
6 I.6 Piping failure- LPG Dispenser 48 11 0.4 4
7 II.1 Drive away while connected to dispenser 48 11 0.4 4
8 II.2 Hose failure- LPG Dispenser 50 10 0.4 4
9 II.3 Leak in connection- LPG Dispenser 50 10 0.4 4
10 II.4 Vehicle pressure relief device leaks- LPG Dispenser 24 16 0.3 5
11 II.5 Nozzles leaks after disconnect- LPG Dispenser 40 12 0.4 4
12 III.1 Vehicle tank isolation valve leaks- LPG storage tank Filling Process 48 11 0.4 4
13 III.2 Overfill of storage tank 80 5 0.5 2
14 III.3 LPG transferring trailer bullet leak 32 14 0.4 4
15 III.4 Unloading hose connection leaks 64 8 0.5 2
16 III.5 Release from connecting hose- LPG storage tank Filling Process 40 12 0.4 4
17 IV.1 No delivery of LPG- LPG Filling Pump 30 15 0.3 5
18 IV.2 No delivery of LPG – LPG Filling Pump 120 1 0.5 2
19 IV.3 No delivery of LPG- LPG Filling Pump 48 11 0.4 4
20 IV.4 No delivery of LPG- LPG Filling Pump 48 11 0.4 4
21 IV.5 No delivery of LPG- LPG Filling Pump 24 16 0.3 5
22 IV.6 No delivery of LPG- LPG Filling Pump 30 15 0.3 5
23 IV.7 No delivery of LPG- LPG Filling Pump 112 2 0.6 1
24 IV.8 Low output- LPG Filling Pump 16 17 0.3 5
25 IV.9 Low output- LPG Filling Pump 64 8 0.5 2
26 IV.10 Low output- LPG Filling Pump 24 16 0.3 5
27 IV.11 Low output- LPG Filling Pump 24 16 0.3 5
28 IV.12 Excessive power consumption- LPG Filling Pump 24 16 0.3 5
29 IV.13 Excessive power consumption- LPG Filling Pump 36 13 0.3 5
30 IV.14 Excessive power consumption- LPG Filling Pump 60 9 0.4 4
31 IV.15 Excessive power consumption- LPG Filling Pump 60 9 0.4 4
32 IV.16 Pump is Noisy- LPG Filling Pump 72 7 0.4 4
33 IV.17 Pump is Noisy- LPG Filling Pump 96 4 0.5 2
34 IV.18 Pump is Noisy- LPG Filling Pump 60 9 0.4 4
35 IV.19 Pump is Noisy – LPG Filling Pump 96 4 0.5 2
36 V.1 No Vehicle Fill- LPG Dispensing Pump 96 4 0.5 2
37 V.2 No Vehicle Fill- LPG Dispensing Pump 96 4 0.5 2
38 V.3 No Vehicle Fill- LPG Dispensing Pump 96 4 0.5 2
39 V.4 Low Fuel Flow- LPG Dispensing Pump 96 4 0.5 2
40 V.5 Low Fuel Flow- LPG Dispensing Pump 40 12 0.4 4
41 V.6 Low Fuel Flow- LPG Dispensing Pump 75 6 0.4 4
42 V.7 Pulsing Flow- LPG Dispensing Pump 64 8 0.5 2
43 V.8 No Fuel Output- LPG Dispensing Pump 64 8 0.5 2
44 V.9 No Fuel Output- LPG Dispensing Pump 48 11 0.4 4
45 VI.1 Valve Not Opening 30 15 0.3 5
46 VI.2 Valve Not Opening – ROV 105 3 0.5 2
47 VI.3 Valve Not Opening 60 9 0.4 4
48 VI.4 Valve Closing Very Slowly-ROV 64 8 0.5 2
49 VII.1 Control system not working 24 16 0.3 5
50 VII.2 Contactor/Relay/detector failure 96 4 0.5 2
51 VII.3 Loose connection-Control system 60 9 0.4 4
52 VII.4 Program fault 48 11 0.4 4
53 VII.5 Selector switch wrong selection 72 7 0.4 4
54 VII.6 Program fault 60 9 0.4 4

modes with smallest degree of grey relation with highest ranking for 4.1. System description
priority.
A brief description of the system is found in one of the earlier works
of the authors (Maniram Kumar et al., 2015). The important systems, its
4. Case study components and the layout are briefly discussed in the above men-
tioned literature. Fig. 4 shows the process flow diagram.
The system selected for study is an auto LPG Dispensing Station The Liquefied petroleum gas (LPG) refuelling station under the
(ALDS) which a more widely available critical socio-technical system. discussion has the following key components: Filling Pump, Remote
The spread of these systems were widely noted in India favouring the operated valves, High-pressure storage, Dispensing pump, Dispenser,
conversion of auto fuel from gasoline to LPG. Control system.

238

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

Fig. 8. Structure of the Fuzzy Inference system.

4.2. Identification of the failure modes • LPG storage tank – I.1 to I.6 – (6 failure modes)
• LPG dispenser – II.1 to II.5 – (5 failure modes)
Broadly stated safety/operational concerns for an Auto LPG refuel- • LPG storage tank filling process – III.1 to III.5 – (5 failure modes)
ling station arise from LPG leaks and associated high-pressure systems, • LPG filling pump – IV.1 to IV.19 – (19 failure modes)
processes and the scope of study is preceded with this inclination by the • LPG dispensing pump – V.1 to V.9 – (Submersible) (9 failure modes)
FMEA experts. The failure modes are usually detected through visual • Remote operated valve – VI.1 to VI.4 – (4 failure modes)
inspection, condition monitoring techniques and time-based preventive • Control system – VII.1 to VII.6 – (6 failure modes)
maintenance actions. Typical causes of failures in systems are: improper
usage, over pressure and failure of tanks, pressure relief device failures, The failure modes are evaluated by allotting score for factors of
piping leak, leak in connection, loose connection, contactor/relay/de- occurrence, severity, and not detection. A 10 level score scale men-
tector failure, program fault, corrosion, valves failure, and etc. tioned in Table 1 is used for this purpose. While scoring the risk factors
The analysis has been limited to system components/equipment and an expert opinion is referred to. The failure modes which are more
processes for which the FMEA was performed. It has the following key possible to occur provided with higher ranking and other failure modes
areas: filling pump, remote operated valves, high-pressure storage tank, are ranked according to their percentage of occurrence.
dispensing pump, dispenser, control system, filling up of storage tank The failure modes which are related to safety and which has the
and control system. The analysis revealed 54 failure modes with its root potential to stop the process for more than two hours provided with
causes enlisted in Table 2. higher severity rating are considered. The severity ranking for the
failure modes which have the tendency to stop the process for less than
two hours are ranked according to their approximate repair time.
4.3. Analyzing the failure modes by classical FMEA Simple problems are quite easy to detect like rupture in hose, over
filling, storage, etc. so these failure modes categorized into low detec-
The FMEA analysis was carried for LPG automobile refuelling sta- tion rating. Similarly complex problems like pump clearance checking
tion and the allied RPN numbers were determined. Results were given loose connections in electrical panel are has higher detection rating and
in Table 2. The FMEA analysis categorized the failure modes into seven other failure modes are ranked as per the criticality of their detection.
major areas as follows: Based on the FMEA team’s recommendation the scores for

239

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

Fig. 9. The rule viewer in fuzzy logic toolbox.

frequency, occurrence and detection were applied. The RPN scores high-medium, low-high, high and very high. The fuzzy numbers of
allow for the prioritization of efforts for the refuelling station’s unin- conversion are given in Table 3.
terrupted continuous operation and safety improvement. Due to un- The expert opinion as linguistic terms are used for occurrence, se-
available and uncertain data, expert opinions were used to scale the risk verity and not detection values of the failures. A rule base of 125 rules
parameters. maps the inputs to output in the inference system to establish risk
priority (See Appendix A). Some rules are described below.
4.4. Analysis of failure modes by fuzzy FMEA with rule base
“IF occurrence IS almost none AND severity IS almost none AND not
detection IS almost none (found) then RPN IS none (no risk),
The fuzzy inference system is modelled by using the fuzzy logic
IF occurrence IS almost none AND severity IS medium AND not
toolbox of Matlab software to calculate the RPN values. The model has
detection IS very high then RPN IS Low Medium.
3 inputs and 1 output variable as given in Fig. 8. Combining the 3 input
IF occurrence IS very high AND severity IS medium AND not de-
factors the inference system calculates the RPN values. Triangular
tection IS very high then RPN IS Very high,
membership functions are found easier to generate and the most fre-
IF occurrence IS very high AND severity IS very high AND IF not
quently used functions of applications (Wang et al., 2011). The real
detection IS almost none then RPN IS Low high.”
numbers are translated to fuzzy sets through triangular (TFN) function
which is described below in Eq. (6). Fig. 5 shows the fuzzy triangular
The inference was carried on by mamdani min/max method (input
membership functions. A 5-level and 10-level triangular membership
method: min; aggregate method: max) and the defuzzification was
functions were used for the input variables and for the output variable
carried by center of gravity (COG) method. This center of gravity
RPN respectively (Figs. 6 and 7).
method can be articulated as a centroid defuzzification approach
finding the center of gravity point of the fuzzy set, on the fuzzy interval.
⎧ 0, x⩽a
⎪ x − a, a ⩽ x ⩾ b This technique was developed by Sugeno in 1985 (Sugeno, 1999). This
b−a
TFN = μA (x ) = is most commonly used technique and is shown in Eq. (7).
⎨c−x, b ⩽ x ⩾ c
⎪c−b μ
⎩ 0, x⩾c (6) ∫i (x ) xdx
X∗ = μ
For input variables 5 levels are being represented by triangular ∫i (x ) dx (7)
membership functions as almost none, low, medium, high and very
high. The output RPN is represented by the 10-level scale are re- The above formula can be shown as follows for defuzzification of a
presented as none, very-low, low, high-low, low-medium, medium, triangular fuzzy number as Eq. (8):

240

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

Table 5
GRA based Fuzzy FMEA results.

Failure ID Failure mode Equal weights 0.3333 Weighing factors (0.21, 0.26, 0.53) for (S, O, D)

Grey relation Rank Grey relation Rank

I.1 Pressure relief malfunction- LPG storage tank 0.90383693 7 0.9249999 9

I.2 Over pressure and failure of storage tank- LPG storage tank 0.824778258 5 0.8633335 6
I.3 Pressure relief device failure- LPG storage tank 0.824778258 5 0.8633335 6
I.4 Piping leak 0.754584469 3 0.8230049 5
I.5 Storage tank failure 0.824778258 5 0.8633335 6
I.6 Piping failure- LPG Dispenser 0.850737538 6 0.8835819 8
II.1 Drive away while connected to dispenser 0.807683861 4 0.8644228 7
II.2 Hose failure- LPG Dispenser 0.807683861 4 0.8644228 7
II.3 Leak in connection- LPG Dispenser 0.807683861 4 0.8644228 7
II. 4 Vehicle pressure relief device leaks- LPG Dispenser 0.90383693 7 0.9249999 9
II.5 Nozzles leaks after disconnect- LPG Dispenser 0.807683861 4 0.8644228 7
III.1 Vehicle tank isolation valve leaks- LPG storage tank Filling Process 0.850737538 6 0.8835819 8
III.2 Overfill of storage tank 0.754584469 3 0.8230049 5
III.3 LPG transferring trailer bullet leak 0.850737538 6 0.8835819 8
III.4 Unloading hose connection leaks 0.754584469 3 0.8230049 5
III.5 Release from connecting hose- LPG storage tank Filling Process 0.807683861 4 0.8644228 7
IV.1 No delivery of LPG- LPG Filling Pump 0.90383693 7 0.9249999 9
IV.2 No delivery of LPG – LPG Filling Pump 0.754584469 3 0.8230049 5
IV.3 No delivery of LPG- LPG Filling Pump 0.850737538 6 0.8835819 8
IV.4 No delivery of LPG- LPG Filling Pump 0.850737538 6 0.8835819 8
IV.5 No delivery of LPG- LPG Filling Pump 0.90383693 7 0.9249999 9
IV.6 No delivery of LPG- LPG Filling Pump 0.90383693 7 0.9249999 9
IV.7 No delivery of LPG- LPG Filling Pump 0.701485077 1 0.6462681 1
IV.8 Low output- LPG Filling Pump 0.90383693 7 0.9249999 9
IV.9 Low output- LPG Filling Pump 0.711530791 2 0.7115379 2
IV.10 Low output- LPG Filling Pump 0.90383693 7 0.9249999 9
IV.11 Low output- LPG Filling Pump 0.90383693 7 0.9249999 9
IV.12 Excessive power consumption- LPG Filling Pump 0.90383693 7 0.9249999 9
IV.13 Excessive power consumption- LPG Filling Pump 0.90383693 7 0.9249999 9
IV.14 Excessive power consumption- LPG Filling Pump 0.807683861 4 0.8644228 7
IV.15 Excessive power consumption- LPG Filling Pump 0.807683861 4 0.7721149 4
IV.16 Pump is Noisy- LPG Filling Pump 0.807683861 4 0.7721149 4
IV.17 Pump is Noisy- LPG Filling Pump 0.754584469 3 0.7306970 3
IV.18 Pump is Noisy- LPG Filling Pump 0.807683861 4 0.7721149 4
IV.19 Pump is Noisy – LPG Filling Pump 0.754584469 3 0.7306970 3
V.1 No Vehicle Fill- LPG Dispensing Pump 0.754584469 3 0.8230049 5
V.2 No Vehicle Fill- LPG Dispensing Pump 0.754584469 3 0.8230049 5
V.3 No Vehicle Fill- LPG Dispensing Pump 0.754584469 3 0.8230049 5
V.4 Low Fuel Flow- LPG Dispensing Pump 0.754584469 3 0.8230049 5
V.5 Low Fuel Flow- LPG Dispensing Pump 0.807683861 4 0.7721149 4
V.6 Low Fuel Flow- LPG Dispensing Pump 0.807683861 4 0.7721149 4
V.7 Pulsing Flow- LPG Dispensing Pump 0.754584469 3 0.8230049 5
V.8 No Fuel Output- LPG Dispensing Pump 0.754584469 3 0.8230049 5
V.9 No Fuel Output- LPG Dispensing Pump 0.850737538 6 0.8835819 8
VI.1 Valve Not Opening 0.90383693 7 0.9249999 9
VI.2 Valve Not Opening – ROV 0.754584469 3 0.7306970 3
VI.3 Valve Not Opening 0.807683861 4 0.8644228 7
VI.4 Valve Closing Very Slowly-ROV 0.754584469 3 0.7306970 3
VII.1 Control system not working 0.90383693 7 0.9249999 9
VII.2 Contactor/ Relay /detector failure 0.711530791 2 0.7115379 2
VII.3 Loose connection-Control system 0.807683861 4 0.7721149 4
VII.4 Program fault 0.850737538 6 0.8835819 8
VII.5 Selector switch wrong selection 0.807683861 4 0.8644228 7
VII.6 Program fault 0.807683861 4 0.7721149 4

b x−a c c−x
∫a xdx + ∫b xdx 1 4.5. Analyzing the failure modes by fuzzy FMEA with GRA
b−a c−b
X∗ = b x−a c c−x
= (a + b + c )
∫ dx +∫ dx 3
a b−a b c−b (8) The failure modes identified by FMEA are incorporated and the risk
prioritization is done as a multi-criteria approach decision making
The conventional RPN values and the fuzzy RPN values in des- process by grey relational approach. The rule base development in the
cending order are given in Table 4. The fuzzy RPN is found to have an case of fuzzy FMEA can be substituted as a multi-criteria approach and
average value of 0.417 (low medium-medium). Fig. 9 shows the rule MCDM techniques may be applied to yield a ranking strategy. Grey
viewer in fuzzy logic toolbox. relational analysis (GRA) is applied to this problem to prioritise and
rank the events in the absence of rule base which is explained below.
Comparative Series is obtained from defuzzified expert opinions

241

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

 A. Maniram Kumar et al.

Table 6
Comparison of ranks between traditional FMEA and fuzzy FMEA approaches- (Failure modes above average RPN Values).

Sl No Failure modes (Traditional FMEA) Failure modes (Fuzzy FMEA)

RPN Rank FRPN Rank

1 IV.2 No delivery of LPG – LPG Filling Pump (Suction filter/strainer blocked or leaking air) 120 1 IV.7 No delivery of LPG- LPG Filling Pump (Excess internal clearances) 0.6 1
2 IV.7 No delivery of LPG- LPG Filling Pump (Excess internal clearances) 112 2 IV.2 No delivery of LPG – LPG Filling Pump (Suction filter/strainer blocked or leaking air) 0.5 2
3 VI.2 Valve Not Opening – ROV (Solenoid failure) 105 3 VI.2 Valve Not Opening – ROV (Solenoid failure) 0.5 2
4 IV.17 Pump is Noisy- LPG Filling Pump (Damaged 'O' ring seals or gaskets) 96 4 IV.17 Pump is Noisy- LPG Filling Pump (Damaged 'O' ring seals or gaskets) 0.5 2
5 IV.19 Pump is Noisy – LPG Filling Pump (Worn or damaged shaft in seal zone) 96 4 IV.19 Pump is Noisy – LPG Filling Pump (Worn or damaged shaft in seal zone) 0.5 2
6 V.1 No Vehicle Fill-LPG Dispensing Pump (Dispenser malfunctioning or not authorizing) 96 4 V.1 No Vehicle Fill-LPG Dispensing Pump (Dispenser malfunctioning or not authorizing) 0.5 2
7 V.2 No Vehicle Fill-LPG Dispensing Pump (Blocked filters) 96 4 V.2 No Vehicle Fill-LPG Dispensing Pump (Blocked filters) 0.5 2
8 V.3 No Vehicle Fill-LPG Dispensing Pump (Insufficient LPG in tank) 96 4 V.3 No Vehicle Fill-LPG Dispensing Pump (Insufficient LPG in tank) 0.5 2
9 V.4 Low Fuel Flow-LPG Dispensing Pump (Restriction in nozzle) 96 4 V.4 Low Fuel Flow-LPG Dispensing Pump (Restriction in nozzle) 0.5 2
10 VII.2 Contactor/Relay /detector failure (Over loading, mechanical failures, improper current 96 4 VII.2 Contactor/Relay /detector failure (Over loading, mechanical failures, improper current rating 0.5 2
rating of components) of components)
11 III.2 Overfill of storage tank (Human error or instrument failure) 80 5 III.2 Overfill of storage tank (Human error or instrument failure) 0.5 2
12 V.6 Low Fuel Flow-LPG Dispensing Pump (Faulty bypass valve) 75 6 I.4 Piping leak (Mechanical failure/corrosion) 0.5 2

242
13 IV.16 Pump is Noisy- LPG Filling Pump (Bearings worn condition) 72 7 III.4 Unloading hose link leaks (inappropriate connection or Mechanical failure) 0.5 2

https://freepaper.me/t/302654
14 VII.5 Selector switch wrong selection (Operator fault) 72 7 IV.9 Low output- LPG Filling Pump (Cavitations or vaporization on suction side of pump) 0.5 2
15 I.4 Piping leak(Mechanical failure /corrosion) 64 8 V.7 Pulsing Flow-LPG Dispensing Pump (Insufficient LPG in storage tank or excessive inlet 0.5 2
restriction)
16 III.4 Unloading hose link leaks (inappropriate connection or Mechanical failure) 64 8 V.8 No Fuel Output-LPG Dispensing Pump (No power supply to pump or pump contactor fault) 0.5 2
17 IV.9 Low output- LPG Filling Pump (Cavitations or vaporization on suction side of pump) 64 8 VI.4 Valve Closing Very Slowly-ROV (Dirt in valve) 0.5 2
18 V.7 Pulsing Flow-LPG Dispensing Pump (Insufficient LPG in storage tank or excessive inlet 64 8 I.2 Over pressure and failure of storage tank- LPG storage tank (Fill storage on cold day/High 0.47 3
restriction) temperature environment)
19 V.8 No Fuel Output-LPG Dispensing Pump (No power supply to pump or pump contactor 64 8 I.3 Pressure relief device failure- LPG storage tank (Mechanical failure) 0.47 3
fault)

: ‫خودت ترجمه کن‬

20 VI.4 Valve Closing Very Slowly-ROV (Dirt in valve) 64 8 I.5 Failure of Storage tank (External fire near the storage tank by fuel spill) 0.47 3
21 IV.14 Excessive power consumption- LPG Filling Pump (Bearings worn condition) 60 9
22 IV.15 Excessive power consumption- LPG Filling Pump (Air leakage in inlet piping) 60 9
23 IV.18 Pump is Noisy- LPG Filling Pump (Misalignment between pump and driver causing 60 9
excessive shaft distortion)
24 VI.3 Valve Not Opening (No adequate pneumatic pressure) 60 9
25 VII.3 Loose connection-Control system (Vibration/ Arching) 60 9
26 VII.6 Program fault (Wrong program loaded) 60 9
Safety Science 101 (2018) 231–247
A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

Table 7
Ranking comparison between traditional, fuzzy FMEA approaches and with GRA trials (Failures are ordered on conventional RPN ranking).

Failure ID Failures FMEA Fuzzy FMEA GRA-I GRA –II

Equal weights (0.3333) for (S, O, D) Weighing factors (0.21, 0.26, 0.53)

RPN Rank FRPN Rank Grey relation Rank Grey relation Rank

IV.10 Low output- LPG Filling Pump 120 1 0.5 2 0.90384 7 0.925 9
IV.15 Excessive power consumption- LPG Filling Pump 112 2 0.6 1 0.80768 4 0.77211 4
VI.2 Valve Not Opening – ROV 105 3 0.5 2 0.75458 3 0.7307 3
IV.7 No delivery of LPG- LPG Filling Pump 96 4 0.5 2 0.70149 1 0.64627 1
IV.9 Low output- LPG Filling Pump 96 4 0.5 2 0.71153 2 0.71154 2
V.1 No Vehicle Fill- LPG Dispensing Pump 96 4 0.5 2 0.75458 3 0.823 5
V.2 No Vehicle Fill- LPG Dispensing Pump 96 4 0.5 2 0.75458 3 0.823 5
V.3 No Vehicle Fill- LPG Dispensing Pump 96 4 0.5 2 0.75458 3 0.823 5
V.4 Low Fuel Flow- LPG Dispensing Pump 96 4 0.5 2 0.75458 3 0.823 5
VII.2 Contactor/ Relay /detector failure 96 4 0.5 2 0.71153 2 0.71154 2
III.2 Overfill of storage tank 80 5 0.5 2 0.75458 3 0.823 5
V.6 Low Fuel Flow- LPG Dispensing Pump 75 6 0.4 4 0.80768 4 0.77211 4
IV.6 No delivery of LPG- LPG Filling Pump 72 7 0.4 4 0.90384 7 0.925 9
VII.5 Selector switch wrong selection 72 7 0.4 4 0.80768 4 0.86442 7
I.4 Piping leak 64 8 0.5 2 0.75458 3 0.823 5
III.4 Unloading hose link leaks 64 8 0.5 2 0.75458 3 0.823 5
IV.17 Pump is Noisy- LPG Filling Pump 64 8 0.5 2 0.75458 3 0.7307 3
V.7 Pulsing Flow- LPG Dispensing Pump 64 8 0.5 2 0.75458 3 0.823 5
V.8 No Fuel Output- LPG Dispensing Pump 64 8 0.5 2 0.75458 3 0.823 5
VI.4 Valve Closing Very Slowly-ROV 64 8 0.5 2 0.75458 3 0.7307 3
IV.4 No delivery of LPG- LPG Filling Pump 60 9 0.4 4 0.85074 6 0.88358 8
IV.5 No delivery of LPG- LPG Filling Pump 60 9 0.4 4 0.90384 7 0.925 9
IV.8 Low output- LPG Filling Pump 60 9 0.4 4 0.90384 7 0.925 9
VI.3 Valve Not Opening 60 9 0.4 4 0.80768 4 0.86442 7
VII.3 Loose connection-Control system 60 9 0.4 4 0.80768 4 0.77211 4
VII.6 Program fault 60 9 0.4 4 0.80768 4 0.77211 4
II.1 Drive away while connected to dispenser 50 10 0.4 4 0.80768 4 0.86442 7
II.2 Hose failure- LPG Dispenser 50 10 0.4 4 0.80768 4 0.86442 7
I.6 Piping failure- LPG Dispenser 48 11 0.4 4 0.85074 6 0.88358 8
II. 4 Vehicle pressure relief device leaks- LPG Dispenser 48 11 0.4 4 0.90384 7 0.925 9
III.1 Vehicle tank isolation valve leaks- LPG storage tank Filling 48 11 0.4 4 0.85074 6 0.88358 8
Process
IV.11 Low output- LPG Filling Pump 48 11 0.4 4 0.90384 7 0.925 9
IV.12 Excessive power consumption- LPG Filling Pump 48 11 0.4 4 0.90384 7 0.925 9
V.9 No Fuel Output- LPG Dispensing Pump 48 11 0.4 4 0.85074 6 0.88358 8
VII.4 Program fault 48 11 0.4 4 0.85074 6 0.88358 8
II.5 Nozzles leaks after disconnect- Dispenser 40 12 0.4 4 0.80768 4 0.86442 7
III.5 Release from connecting hose- LPG storage tank Filling 40 12 0.4 4 0.80768 4 0.86442 7
Process
V.5 Low Fuel Flow- LPG Dispensing Pump 40 12 0.4 4 0.80768 4 0.77211 4
I.2 Over pressure and failure of storage tank- LPG storage tank 36 13 0.47 3 0.82478 5 0.86333 6
I.3 Pressure relief device failure LPGstorage tank 36 13 0.47 3 0.82478 5 0.86333 6
I.5 Failure of Storage tank 36 13 0.47 3 0.82478 5 0.86333 6
IV.3 No delivery of LPG- LPG Filling Pump 36 13 0.3 5 0.85074 6 0.88358 8
III.3 LPG transferring trailer bullet leak 32 14 0.4 4 0.85074 6 0.88358 8
IV.1 No delivery of LPG- LPG Filling Pump 30 15 0.3 5 0.90384 7 0.925 9
IV.14 Excessive power consumption- LPG Filling Pump 30 15 0.3 5 0.80768 4 0.86442 7
VI.1 Valve Not Opening 30 15 0.3 5 0.90384 7 0.925 9
I.1 Pressure relief malfunction- LPG storage tank 24 16 0.3 5 0.90384 7 0.925 9
II.3 Leak in connection- LPG Dispenser 24 16 0.3 5 0.80768 4 0.86442 7
IV.13 Excessive power consumption- LPG Filling Pump 24 16 0.3 5 0.90384 7 0.925 9
IV.18 Pump is Noisy- LPG Filling Pump 24 16 0.3 5 0.80768 4 0.77211 4
IV.19 Pump is Noisy – LPG Filling Pump 24 16 0.3 5 0.75458 3 0.7307 3
IV.2 No delivery of LPG – LPG Filling Pump 24 16 0.3 5 0.75458 3 0.823 5
VII.1 Control system not working 24 16 0.3 5 0.90384 7 0.925 9
IV.16 Pump is Noisy- LPG Filling Pump 16 17 0.3 5 0.80768 4 0.77211 4

E.g.: (L, H, L) gives (0.25, 0.75, 0.25). The optimal level of parameters considered from the Eq. (3) and used to calculate the grey relationship
determines the standard series. The safe operation of every system coefficient. The correlation coefficient is calculated from Eq. (4).
needs the least of occurrence probability, severity and non-detection. Careful selection of the weighting coefficient is necessary to get the
The lowest level values are considered and standard series is the desired degree of relation. Equal weightage of 0.333 is given to occurrence,
outcome for the failure modes which is the defuzzified lowest linguistic severity and detection and weights used by Arabian-Hoseynabadi et al.
score i.e. (AN, AN, AN) = (0.066667, 0.066667, 0.066667). (2010), i.e., (0.21, 0.26, and 0.53) for (S, O, D) is also considered for
The difference between the comparative and standard series is comparison. The failure modes are ranked from the ascending order i.e.

243

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

smaller the degree of relation, then stronger is the effect of the cause. The fuzzy rule base allocates similar RPN values for several failure
The failure modes and ranks are given in Table 5. modes. This can be used to cluster the failure modes in order of priority.
When GRA approach is compared with rule base results, the order of
5. Results and discussion priority is found to vary in the top and the bottom most group of
priority failure modes. We can notice LPG low output (FM IV.10) is
FMEA plays a vital role to remedy the gap between the process found to have lower priority in GRA and a higher priority in rule base
technology and operational practices by suggesting proactive main- approach. And failure modes like IV.19, IV.18, IV.16, IV.2, IV.14, and
tenance regime and producing combined precautions. At this point, the II.3 are found to have higher importance in GRA approach.
core issue specific to this failure mode is to reduce/prevent the risky Among GRA trials, similar results are obtained with exception to
situations of faults in LPG filling pump, ROVs, dispensers, etc. For in- few failure modes whose order of priority is rearranged. In the above
stance, the current maintenance regime (planned preventive main- Table 7 results were found to have pronounced the importance or
tenance) can be enhanced with relevance to the operational control/ weightage to the three factors occurrence, severity and non detection.
planning perspective. It means that the in-service hours can be recon- The ranks are found to have rearranged based on the weightage. The
sidered in accordance with the condition monitoring. This operational dissimilar weighting factors are applied there is a noticeable re-
control can be valued with technology integration into existing control arrangement of priority of failure modes which pronounces their im-
systems. An additional unit that continuously monitors performance portance. It can be of ease to user to retune the approach by altering the
parameters (i.e. Inlet/Outlet flow rate, pressures, noise levels) can be weights, in case of demands arising due to modifications or alterations.
integrated. Consequently, the FMEA of the system aid us to produce GRA approach can be used in the case of absence of predefined in-
precautions both in system design and operational levels to prevent the ference rules which require a lot of expertise otherwise.
risks of failure mode. The similar approach can be followed for the
other components and their corresponding failure modes subsequently. 6. Conclusion
Based on the results, safety improvement actions for different stages of
processes were suggested. Fuzzy FMEA is applied to an auto LPG refuelling station in this
The results depict the average RPN values (Traditional- work. The potential failure modes were identified and analyzed for
56.63 & Fuzzy-0.417). It is observed that about 26 failures and 20 their effects, root causes, and detection methods. The RPN and FRPN
failures are above the average values by the traditional and fuzzy rule scores have also been computed, with the results analyzed and dis-
base method respectively and are compared in Table 6. The fuzzy ap- cussed. The usage of linguistic terms allows the experts to dispense a
proach demonstrates a wider concern in considering the failures. With a significant value for the risk factors considered. Thus the identified
resemblance of the traditional approach, the fuzzy approach presents a failure modes are not overlooked (owing to a low RPN) when taken into
little reallocation in the prioritisation of the failure modes by applying account for the precedence of attention.
the proposed methodology. The proposed approach using fuzzy rule base could be suitable for
The failure ranking comparison of failure modes above average RPN use in screening phase during a formal safety assessment (FSA) where
values demonstrate fuzzy approach considers the uncertainty and am- only a relative ranking order is required basically to differentiate fail-
biguity of assessment process and provides more stable solution. In both ures with low and high risk levels.
the approaches, the failures associated with LPG filling pump, ROV and The grey approach allows the experts to fix on important factor for
dispensing pumps are found to peak the ranks. The fuzzy approach the study, the results may provide precious inferences for the decision
categorizes into five sets of priority ranks. It is noted that the failure making process. Moreover the approach can be applied for a number of
modes IV.14, IV.15, IV.18, VI.3, VII.3, VII.6, VII.5, V.6, and IV.16 are industries like petrochemical, chemical industries, etc.
found to be above average RPN in traditional approach but are not The advantages of FMEA coupled with fuzzy approach to a LPG
present in fuzzy approach. The failure modes I.2, I.3, I.5 are observed to refuelling station can be summarised as follows:
be with FRPN value more than average but are not present in tradi-
tional approach. They refer to storage tank overpressure, pressure relief • Expert knowledge and experience are systematically combined for
failure and tank failure. Owing to the severity of those failures, the use in an FMEA study.
fuzzy approach prioritises to a third category of importance. This re- • Experts deliver their opinion more realistically as linguistic terms.
covers the accuracy of ranking by applying fuzzy rule base to FMEA. • The fuzzy-FMEA can be extended with a number of risk factors not
Fuzzy FMEA by GRA approach enables the users to realistically restricting to three.
process experts judgments (linguistic terms) and consequently improves • An ability to handle various types of uncertainties, such as im-
the relevance of the FMEA technique in ALDS. The rule base is sub- precision, incompleteness, fuzziness of quantitative risk assessments
stituted by means of a grey relational approach to achieve the priority which formulates the approach more flexible, more practical and
ranking. The importance weights of failure modes are engaged into more realistic. The reason is that fuzzy method can overcome the
deliberation, which makes the projected FMEA methodology more uncertainty of input data and it’s imprecision.
practical, realistic and more flexible. FM IV.7 (Suction filter/strainer • GRA approach can be used in the case of absence of predefined
blockage) is found to get the top priority. FM IV.9 (Cavitations or va- inference rules which require a lot of expertise otherwise. It can be
porisation in suction side of pump) and VII.2 (Contactor/ Relay /de- easily tuned to suit system modifications or to new systems which
tector failure) are following the top failure mode. increases the applicability of the approach.

Appendix A

244

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

Table A1
Rule base for fuzzy output.

Sl No. Occurrence Severity (Not) detection Fuzzy output

1 Almost none Almost none Almost none None

2 Almost none Almost none Low None
3 Almost none Almost none Medium Very Low
4 Almost none Almost none High Very Low
5 Almost none Almost none Very high Low
6 Almost none Low Almost none None
7 Almost none Low Low Very Low
8 Almost none Low Medium Very Low
9 Almost none Low High Low
10 Almost none Low Very high High Low
11 Almost none Medium Almost none Very Low
12 Almost none Medium Low Very Low
13 Almost none Medium Medium Low
14 Almost none Medium High High Low
15 Almost none Medium Very high Low Medium
16 Almost none High Almost none Very Low
17 Almost none High Low Low
18 Almost none High Medium High Low
19 Almost none High High Medium
20 Almost none High Very high Medium
21 Almost none Very high Almost none Low
22 Almost none Very high Low High Low
23 Almost none Very high Medium Low Medium
24 Almost none Very high High Medium
25 Almost none Very high Very high High Medium
26 Low Almost none Almost none None
27 Low Almost none Low Very Low
28 Low Almost none Medium Very Low
29 Low Almost none High Low
30 Low Almost none Very high High Low
31 Low Low Almost none Very Low
32 Low Low Low Very Low
33 Low Low Medium Low
34 Low Low High High Low
35 Low Low Very high Low Medium
36 Low Medium Almost none Very Low
37 Low Medium Low Low
38 Low Medium Medium High Low
39 Low Medium High Low Medium
40 Low Medium Very high Medium
41 Low High Almost none Low
42 Low High Low High Low
43 Low High Medium Low Medium
44 Low High High Medium
45 Low High Very high High Medium
46 Low Very high Almost none High Low
47 Low Very high Low Low Medium
48 Low Very high Medium Medium
49 Low Very high High High Medium
50 Low Very high Very high Low High
51 Medium Almost none Almost none Very Low
52 Medium Almost none Low Very Low
53 Medium Almost none Medium Low
54 Medium Almost none High High Low
55 Medium Almost none Very high Low Medium
56 Medium Low Almost none Very Low
57 Medium Low Low Low
58 Medium Low Medium High Low
59 Medium Low High Low Medium
60 Medium Low Very high Medium
61 Medium Medium Almost none Low
62 Medium Medium Low High Low
63 Medium Medium Medium Low Medium
64 Medium Medium High Medium
65 Medium Medium Very high High Medium
66 Medium High Almost none High Low
67 Medium High Low Low Medium
68 Medium High Medium Medium
69 Medium High High High Medium
70 Medium High Very high Low High
71 Medium Very high Almost none Low Medium
72 Medium Very high Low Medium
73 Medium Very high Medium High Medium
74 Medium Very high High Low High
(continued on next page)

245

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

Table A1 (continued)

Sl No. Occurrence Severity (Not) detection Fuzzy output

75 Medium Very high Very high High

76 High Almost none Almost none Very Low
77 High Almost none Low Low
78 High Almost none Medium High Low
79 High Almost none High Low Medium
80 High Almost none Very high Medium
81 High Low Almost none Low
82 High Low Low High Low
83 High Low Medium Low Medium
84 High Low High Medium
85 High Low Very high High Medium
86 High Medium Almost none High Low
87 High Medium Low Low Medium
88 High Medium Medium Medium
89 High Medium High High Medium
90 High Medium Very high Low High
91 High High Almost none Low Medium
92 High High Low Medium
93 High High Medium High Medium
94 High High High Low High
95 High High Very high High
96 High Very high Almost none Medium
97 High Very high Low High Medium
98 High Very high Medium Low High
99 High Very high High High
100 High Very high Very high Very High
101 Very high Almost none Almost none Low
102 Very high Almost none Low High Low
103 Very high Almost none Medium Low Medium
104 Very high Almost none High Medium
105 Very high Almost none Very high High Medium
106 Very high Low Almost none High Low
107 Very high Low Low Low Medium
108 Very high Low Medium Medium
109 Very high Low High High Medium
110 Very high Low Very high Low High
111 Very high Medium Almost none Low Medium
112 Very high Medium Low Medium
113 Very high Medium Medium High Medium
114 Very high Medium High Low High
115 Very high Medium Very high High
116 Very high High Almost none Medium
117 Very high High Low High Medium
118 Very high High Medium Low High
119 Very high High High High
120 Very high High Very high Very High
121 Very high Very high Almost none High Medium
122 Very high Very high Low Low High
123 Very high Very high Medium High
124 Very high Very high High Very High
125 Very high Very high Very high Very High

References December 20, 2014) < http://www.iocl.com/products/lpgspecifications.pdf > .

Bowles, J.B., Peláez, C.E., 1995. Fuzzy logic prioritization of failures in a system failure
mode, effects and criticality analysis. Reliab. Eng. Syst. Saf. 50, 203–213.
Abdelgawad, M., Fayek, A., 2011. Risk management in the construction industry using Braglia, M., 2000. MAFMA: multi-attribute failure mode analysis. Int. J. Qual. Reliab.
combined fuzzy FMEA and fuzzy AHP. J. Constr. Eng. Manage. 136 (9), 1028–1036. Manage. 17 (9), 1017–1033.
Al-shanini, Ali, Ahmad, Arshad, Khan, Faisal, 2014. Accident modeling and analysis in Braglia, M., Frosolini, M., Montanan, R., 2003. Fuzzy TOPSIS approach for failure mode,
process industries. J. Loss Prev. Process Ind. 32, 319–334. effects and criticality analysis. Qual. Reliab. Eng. Int. 19 (5), 425–443.
Almannai, B., Greenough, R., Kay, J., 2008. A decision support tool based on QFD and Chandima Ratnayake, R.M., 2014. Application of a fuzzy inference system for functional
FMEA for the selection of manufacturing automation technologies. Robot. Comput.- failure risk rank estimation: RBM of rotating equipment and instrumentation. J. Loss
Integr. Manuf. 24 (4), 501–507. Prev. Process Ind. 29, 216–224.
Shahriar, Anjuman, Sadiq, Rehan, Tesfamariam, Solomon, 2012. Risk analysis for oil & Chin, K.S., Chan, A., Yang, Y.B., 2008. Development of a Fuzzy FMEA based product
gas pipelines: a sustainability assessment approach using fuzzy based bow-tie ana- design system. Int. J. Adv. Manuf. Technol. 36 (7–8), 633–649.
lysis. J. Loss Prev. Process Ind. 25, 505–523. Sa’idi, Elham, Anvaripour, Bagher, Jaderi, Fereshteh, Nabhani, Nader, 2014. Fuzzy risk
Apurva Chandra. Indian LPG Market Prospects, World LPG Forum, 2010. Madrid. modeling of process operations in the oil and gas refineries. J. Loss Prevent. Process
Retrieved December 30, 2014 from petrofed. http://winwinhosting.net/upload/ Ind., 30, 63–73.
Apurva_Chandra.pdf. Gargama, H., Chaturvedi, S.K., 2011. Criticality assessment models for failure mode ef-
Aqlan, F., Mustafa, E., 2014. Integrating lean principles and fuzzy bow-tie analysis for risk fects and criticality analysis using fuzzy logic. IEEE Trans. Reliab. 60 (1), 102–110.
assessment in chemical industry. J. Loss Prev. Process Ind. 29, 39–48. Gilchrist, W., 1993. Modelling failure modes and effects analysis. Int. J. Qual. Reliab.
Arabian-Hoseynabadi, H., Oraee, H., Tavner, P.J., 2010. Failure modes and effects ana- Manage. 10 (5), 16–23.
lysis (FMEA) for wind turbines. Int. J. Electr. Power Energy Syst. 32 (7), 817–824. Guimarães, A.C.F., Lapa, C.M.F., 2004a. Fuzzy FMEA applied to PWR chemical and vo-
Mentes, Ayhan, Ozen, Emre, 2015. A hybrid risk analysis method for a yacht fuel system lume control system. Prog. Nucl. Energy 44 (3), 191–213.
safety. Saf. Sci. 79, 94–104. Guimarães, A.C.F., Lapa, C.M.F., 2004b. Effects analysis fuzzy inference system in nuclear
BIS 14861. LPG Specification, Indian Oil Corporation Limited website (retrieved problems using approximate reasoning. Ann. Nucl. Energy 31 (1), 107–115.

246

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬

A. Maniram Kumar et al. Safety Science 101 (2018) 231–247

Guimarães, A.C.F., Lapa, C.M.F., 2007. Fuzzy inference to risk assessment on nuclear Sharma, R.K., Kumar, D., Kumar, P., 2005. Systematic failure mode effect analysis
engineering systems. Appl. Soft Comput. 7, 17–28. (FMEA) using fuzzy linguistic modelling. Int. J. Qual. Reliab. Manage. 22 (9),
Liu, H-Tien, Tsai, Yi-lin, 2012. A fuzzy risk assessment approach for occupational hazards 986–1004.
in the construction industry. Saf. Sci. 50, 1067–1078. Sharma, R.K., Kumar, D., Kumar, P., 2008. Predicting uncertain behavior of industrial
Liu, Hu-Chen, Liu, Long, Liu, Nan, 2013. Risk evaluation approaches in failure mode and system. Appl. Soft Comput. 8, 96–109.
effects analysis: a literature review. Expert Syst. Appl. 40, 828–838. Rajakarunakaran, Sivaprakasam, Maniram Kumar, A., Arumuga Prabhu, V., 2015.
Ju-Long, Deng, 1982. Control problems of grey systems. Syst. Control Lett. 1 (5), Applications of fuzzy faulty tree analysis and expert elicitation for evaluation of risks
288–294. in LPG refuelling station. J. Loss Prev. Process Ind. 33, 109–123.
Keskin, G.A., Ozkan.C, C., 2009. An alternative evaluation of FMEA: fuzzy ART algorithm. Sugeno, M., 1999. Fuzzy Modelling and Control, first ed. CRC Press, Florida, USA.
Qual. Reliab. Eng. Int. 25 (6), 647–661. Tay, K.M., Lim, C.P., 2006. Fuzzy FMEA with a guided rules reduction system for
Maniram Kumar, A., Rajakarunakaran, S., Arumuga Prabu, V., 2015. Human reliability prioritization of failures. Int. J. Qual. Reliab. Manage. 23 (8), 1047–1066.
analysis by cognitive approach for unloading process in an ALDS (auto LPG dispen- Tay, K.M., Lim, C.P., 2010. Enhancing the failure mode and effect analysis methodology
sing station), Ind. J. Sci. Technol. 8(26). with fuzzy inference techniques. J. Intell. Fuzzy Syst. 21 (1–2), 135–146.
Li, Pe-cheng, Chen, Guo-hua, Dai, Li-cao, Li, Zhang, 2010. Fuzzy logic-based approach for Wang, X., Chan, H.K., Yee, R., Diaz- Rainey, I., 2011. A two-stage fuzzy-AHP model for
identifying the risk importance of human error. Saf. Sci. 48 (7), 902–913. risk assessment of implementing green initiatives in the fashion supply chain. Int. J.
Wei, Liang, Lai-bin, Zhang, Zhao-hui, Wang, Li-xiang, Duan, 2009. Using fuzzy method to Prod. Econ. http://dx.doi.org/10.1016/j.ijpe.2011.03.021.
evaluate safety condition of big diesel engine. J. Loss Prev. Process Ind. 22, 928–933. Wang, Y.M., Chin, K.S., Poon, G.K.K., Yang, J.B., 2009. Risk evaluation in failure mode
Guo, Lijie, Gao, Jinji, Yang, Jianfeng, Kang, Jianxin, 2009. Criticality evaluation of pet- and effects analysis using fuzzy weighted geometric mean. Expert Syst. Appl. 36 (2),
rochemical equipment based on fuzzy comprehensive evaluation and a BP neural 1195–1207.
network. J. Loss Prev. Process Ind. 22, 469–476. Xu, K., Tang, L.C., Xie, M., Ho, S.L., Zhu, M.L., 2002. Fuzzy assessment of FMEA for
Liu, H.-C., Liu, L., Bian, Q.-H., Lin, Q.-L., Dong, N., Xu, P.-C., 2011. Failure mode and engine systems. Reliab. Eng. Syst. Saf. 75 (1), 17–29.
effects analysis using fuzzy evidential reasoning approach and grey theory. Expert Yang, Z., Bonsall, S., Wang, J., 2008. Fuzzy rule-based Bayesian reasoning approach for
Syst. Appl. 38, 4403–4415. prioritization of failures in FMEA. IEEE Trans. Reliab. 57 (3), 517–528.
Kumru, Mesut, Kumru, Pınar Yıldız, 2013. Fuzzy FMEA application to improve pur- Yeh, R.H., Hsieh, H., 2007. Fuzzy assessment of FMEA for a sewage plant. J. Chinese Inst.
chasing process in a public hospital. Appl Soft Comput 13, 721–733. Indus. Eng. 24 (6), 505–512.
Pelaez, C.E., Bowles. J.B, 1996. Using fuzzy cognitive maps as a system model for failure Huang, Yimiao, Ma, Guowei, Li, Jingde, Hao, Hong, 2015. Confidence-based quantitative
modes and effects analysis. Inform. Sci. 88(1:4), 177–199. risk analysis for offshore accidental hydrocarbon release events. J. Loss Prev. Process
Pillay, A., Wang, J., 2003. Modified failure mode and effects analysis using approximate Ind. 35, 117–124.
reasoning. Reliab. Eng. Syst. Saf. 79, 69–85. Chen, Zhen, Wu, Xiaona, Qin, Jianguo, 2014. Risk assessment of an oxygen-enhanced
Zhou, Qingji, Thai, Vinh V., 2016. Fuzzy and grey theories in failure mode and effect combustor using a structural model based on the FMEA and fuzzy fault tree. J. Loss
analysis for tanker equipment failure prediction. Saf. Sci. 83, 74–79. Prev. Process Ind. 32, 349–357.

247

https://freepaper.me/t/302654 : ‫خودت ترجمه کن‬