Вы находитесь на странице: 1из 19

The Emerald Research Register for this journal is available at The current issue and full text archive

text archive of this journal is available at

www.emeraldinsight.com/researchregister www.emeraldinsight.com/0265-671X.htm

22,9 Systematic failure mode effect
analysis (FMEA) using fuzzy
linguistic modelling
Rajiv Kumar Sharma, Dinesh Kumar and Pradeep Kumar
Indian Institute of Technology (IIT), Roorkee, India
Received 25 September 2004
Revised 8 January 2005
Accepted 27 January 2005
Purpose – To permit the system safety and reliability analysts to evaluate the criticality or risk
associated with item failure modes.
Design/methodology/approach – The factors considered in traditional failure mode and effect
analysis (FMEA) for risk assessment are frequency of occurrence (Sf), severity (S) and detectability
(Sd) of an item failure mode. Because of the subjective and qualitative nature of the information and to
make the analysis more consistent and logical, an approach using fuzzy logic is proposed. In the
proposed approach, these parameters are represented as members of a fuzzy set fuzzified by using
appropriate membership functions and are evaluated in fuzzy inference engine, which makes use of
well-defined rule base and fuzzy logic operations to determine the criticality/riskiness level of the
failure. The fuzzy conclusion is then defuzzified to get risk priority number. The higher the value of
RPN, the greater will be the risk and lower the value of RPN, and the lesser will be the risk. The fuzzy
linguistic assessment model was developed using toolbox platform of MATLAB 6.5 R.13.
Findings – The applicability of the proposed approach is investigated with the help of an illustrative
case study from the paper industry. Fuzzy risk assessment is carried out for prioritizing failure causes of
the hydraulic system, a primary element of the feeding system. The results provide an alternate ranking
to that obtained by the traditional method. It is concluded from the study that the fuzzy logic-based
approach not only resolves the limitations associated with traditional methodology for RPN evaluation
but also permits the experts to combine probability of occurrence (Sf), severity (S) and detectability (Sd) of
failure modes in a more flexible and realistic manner by using their judgement, experience and expertise.
Originality/value – The paper integrates the use of fuzzy logic and expert database with FMEA and
may prove helpful to system safety and reliability analysts while conducting failure mode and effect
analysis to prioritize failures for taking corrective or remedial actions.
Keywords Failure modes and effects analysis, Reliability management, Systems and control theory,
Fuzzy logic
Paper type Research paper

Failure mode and effect analysis (FMEA) is a structured, bottom-up approach that
starts with known potential failure modes at one level and investigates the effect on the
next subsystem level. All complex mechanical systems are composed of several
subsystems which can be further broken down up to a component level (Wang et al.,
1996). A complete FMEA analyses of a system therefore span all the levels in the
International Journal of Quality & hierarchy from bottom to top (Figure 1). FMEA as a formal design methodology was
Reliability Management
Vol. 22 No. 9, 2005
pp. 986-1004 The authors are grateful to the editor and anonymous reviewers for their constructive
q Emerald Group Publishing Limited
suggestions, which have been incorporated substantially to improve the quality and readability
DOI 10.1108/02656710510625248 of the paper.
Fuzzy linguistic


Figure 1.
Hierarchical structure of

first proposed by NASA in year 1963 for their obvious reliability requirements. Since
then, it has been extensively used as a powerful tool for safety and reliability analysis
of products and processes in a wide range of industries particularly, aerospace, nuclear
and automotive industries (Gilchrist, 1993; O’Connor, 2000; Ebeling, 2000). In 1977, it
was adopted and promoted by Ford Motor Company. The Ford procedure extended
FMEA methodology in automotive sector to assess and prioritize potential process and
design- related failures. The main objective of FMEA is to discover and correct the
potential failure problems during the stages of design and production. There are two
phases in FMEA. The first phase is concerned with identification of the potential
failure modes and their effects. It includes defining the potential failures of product’s
component, subassemblies, final assembly and its manufacturing processes and the
second phase is concerned with performing criticality analysis to determine the
severity of failure modes by evaluating and ranking (RPN) the criticality level of each
failure. Figure 2 shows the flow chart revealing general procedure for carrying out
FMEA process. In brief, the ten steps involved are as described as follows:
(1) Identify the system to be analysed. Divide the system into subsystems and /or
assemblies in order to localize the search for components and develop a list of
components for each assembly.
(2) Construct the block diagram of the system. Use structural, (hardware),
functional, combined, master logic diagram and cause and effect diagram to
identify relations among components (Figure 3).
(3) Determine all potential failure modes of each component, their causes and the
effects of failure modes on the immediate function or item, on sub-systems and
the entire system.
(4) Evaluate each failure mode in terms of worst potential consequence (severity).
(5) Identify failure detection methods and compensating provision(s) for each
failure mode.
(6) Estimate the probability of occurrence (Sf) using both qualitative and
quantitative techniques.


Figure 2.
FMEA process flow chart

Figure 3.
Functional block diagram

(7) Calculate the risk priority number (RPN), using relation RPN ¼ S f.S. Sd.
(8) Determine whether corrective action is required or not depending upon the
RPN. If required than identify corrective design or other actions required to
eliminate the causes of failure. The actions may be:
Compensatory to minimize the loss in event of failure occurrence.
Preventive to avoid a failure situation.
(9) Develop recommendations to enhance the system performance. Fuzzy linguistic
(10) Prepare FMEA report by summarizing the analysis in tabular form (Table I). modelling
Traditional methodology of risk assessment in FMEA
Traditionally, criticality or risk assessment in failure mode, effects and criticality
analysis (FMECA) is executed in two ways (Bowles and Pelaez, 1995; Xu and Tang,
2002; Sharma, 2004): 989
(1) By calculating a criticality number (CN).
(2) By developing a risk priority number (RPN).

This is described in US MIL-STD-1629A “Procedures for performing a failure mode,

effects and criticality analysis” (MIL-STD 1629, 1980). Criticality number technique is
used mostly in the high-risk plants, such as nuclear and aerospace industries and RPN
technique is used in consumer goods, manufacturing and process industries.

Criticality number (CN) calculation

Equation (1) is used for calculating criticality number CN number for each item failure
mode i:

CN i ¼ ai bi I i t ð1Þ
where the terms: ai: failure mode ratio; bi: failure-effect probability, li: part failure rate;
and t: operating time.

Risk priority number (RPN) calculation

Equation (2) is used to calculate risk priority number (RPN):

RPN ¼ S f :S:S D ð2Þ

where the terms: Sf: frequency of occurrence; S: severity of its failure effects and SD:
chance of the failure being undetected.

Limitations of traditional approaches

Both the approaches are widely used for risk/criticality assessment in aerospace,
nuclear, automotive and process industries, but they have several
shortcomings/drawbacks in ways in which calculations are made and results are
In the CN calculation, the item criticality number may be underestimated when a
failure mode has multiple effects in different severity categories since only the most
severe effect is used in the calculation.
In RPN analysis, various sets of Sf, S and Sd may produce an identical value,
however, the risk implication may be totally different For instance, consider two
different events having values of Sf ¼ 3, S ¼ 4, Sd ¼ 5 and Sf ¼ 1, S ¼ 10, Sd ¼ 6
respectively. Both these events will have a total RPN value of 60; however, the risk
implications of these two events may not necessarily be the same, which may result in
high-risk events going unnoticed. The other disadvantage of the RPN ranking method
is that it neglects the relative importance among Sf, S and Sd. The three factors are
assumed to have the same importance but in real practical applications the relative


Table I.

Format of FMEA table

System FMEA No.
Subsystem Page.
Component Prepared by
Core team FMEA Date (org.)
results (Rev)

Potential Potential Potential Current

Item/ failure effects of causes of design Recommend Actions
function mode failure Severity failure Occurrence controls Detect actions taken Severity Occurrence Detect RPN
importance among the factors exists. For instance, a failure mode with a very high Fuzzy linguistic
severity, low rate of occurrence, and moderate detectability (say 9, 3, and 5 modelling
respectively) may have a lower RPN (135) than one with all parameters moderate (say
5, 6, and 6 yielding an RPN of 180), even though it should have a higher priority for
corrective action.
Though suitable weights can be assigned to the parameters, in order to make the
assessment more precise and more reliable there is a need to handle the subjective or 991
qualitative information associated with the analysis in consistent and logical manner.
Therefore to address the above listed limitations, an approach using fuzzy linguistic
modeling is proposed.

Fuzzy risk assessment methodology

The fuzzy risk assessment methodology based on fuzzy sets theory (propounded by
Zadeh (1965)) provides a more flexible and meaningful way to assess risk associated
with component/item failure modes. The parameters, i.e. Sf, S and Sd which are used in
FMEA are fuzzified using appropriate membership functions to determine degree of
membership in each input class. The resulting fuzzy inputs are evaluated in fuzzy
inference engine, which makes use of well-defined rule base consisting of If-Then rules
and fuzzy logic operations to determine criticality/riskiness level of the failure. The
fuzzy conclusion is then defuzzified to get risk priority number. Higher the value of
RPN, greater will be the risk and lower the value of RPN, lesser will be the risk. The
Fuzzy Linguistic assessment model was developed using toolbox platform of
MATLAB 6.5 R.13. The basic system architecture consists of three main modules, i.e.
knowledge base module and user input/output interface module as shown in Figure 4.
The main components associated with the model are: Fuzzification; Fuzzy rule base;
Fuzzy inference system; and Defuzzification. The following paragraphs briefly discuss
each of them.

Fuzzification refers to transformation of crisp inputs into a membership degree, which
expresses how well the input belongs to the linguistically defined terms. Using the
toolbox simulator of Matlab (2000) membership functions are developed for both

Figure 4.
System architecture of
fuzzy linguistic model
IJQRM inputs and output. To represent input variables (Sf, S and Sd) graphically, trapezoidal
22,9 membership functions (Figure 5) are used which are consistent with the definitions of
probability of failure occurrence, severity and non-detectability used in the study as
depicted in Tables II-IV. The interpretation of descriptive terms Remote (R), Low (L),
Moderate (M), High (H) and Very high (VH) used to describe occurrence, the severity
and the non-detectability is presented in Table V (Klir and Yuan, 1995).

Figure 5.
Fuzzy membership
function for linguistic
variables (Sf.S. Sd)

Descriptive assessment of
probability of failure Mean time between failures (MTBF) Score Occurrence rate (%)

Remote . 5 years 1 , 0.01

Table II. Low 2-5 years 2-3 0.01-0.1
Scales used to measure Moderate 1-2 years 4-6 0.1-0.5
probability of failure High 3-6 months 7-8 0.5-1
occurrence Very high , 3 months 9-10 .1

Non-detection Score Likelihood of non-detection (%)

Remote 1 0-5
Low 2 6-15
3 16-25
Moderate 4 26-35
5 36-45
6 46-55
High 7 56-65
Table III.
8 66-75
Scale used to determine
9 76-85
probability of
non-detection Very high 10 86-100

Rank no. Severity effect Meaning

1 Remote Less MTTR .1 hour

2-3 Low MTTR .1 day
Table IV. 4-5-6 Moderate MTTR 1-4 days
Scale used for severity 7-8 High External intervention for repairs
assessment 9-10 Very high Line shut down or production loss
Fuzzy linguistic
term Probability of occurrence (Sf) Severity (S) Non-detectability (Sd) modelling
Remote It would be very unlikely for A failure that has no effect Defect remains undetected
these failures to be observed on the system performance, until the system
once the operator probably will performance degrades to the
not notice extent that the task will not 993
be completed
Low Likely to occur once but A failure that would cause Defect remains undetected
unlikely to occur more slight annoyance to the until the system
frequently operator, but would cause performance is severely
no deterioration to the reduced
Moderate Likely to occur more than A failure that would cause Defect remains undetected
once high degree of operator until the system
dissatisfaction, or that performance is affected
causes noticeable but slight
deterioration in system
High Near certain to occur at least A failure that causes Defect remains undetected
once deterioration in system until inspection or test is
performance and/or leads to carried out.
minor injuries. Table V.
Very high Near certain to occur several A failure that would Failure remains undetected, Interpretation of
times seriously affect the ability to such a defect would almost descriptive terms used for
complete the task or cause certainly be detected during graphical representation
damage, serious injuries or inspection or test of fuzzy membership
death function

To represent output variable, riskiness/priority level graphically (Figure 6) both

triangular and trapezoidal membership functions are used. Multiple experts with
different degree of competencies “C” are used to construct the membership function. The
descriptive terms describing the output membership function are Not important, Minor,
Low, Moderate, Important and Very important. (In Appendix 2, the procedure which
makes use of expertise judgment in designing membership functions is discussed).

Fuzzy rule base

Skilled human analysts often have good, intitutive knowledge of behavior of system
and the risks involved in various types of failures. To express this knowledge fuzzy

Figure 6.
Graphical representation
of fuzzy priority
membership function
IJQRM rules provide a natural platform for abstracting information based on expert’s
22,9 judgement and engineering knowledge. Expert’s knowledge and expertise about the
interaction between various failure modes and their effects is represented in form of
fuzzy “If-Then” rules. “If” refers to an antecedent that is compared to the inputs, and
“Then” refers to a consequent, which is the result/output (Ross, 1995; Terano et al.,
1987; Zimmermann, 1996).
994 For instance:
RI ; if x is M i then y is N i i ¼ 1; 2; 3 . . . K ð3Þ
. x is the input (antecedent) linguistic variable.
Mi are the antecedent linguistic constants (qualitatively defined functions).
y is the output (consequent) linguistic variable.
Ni are the consequent linguistic constants.
All the rules that have any truth in their antecedent will fire and contributes towards
the fuzzy conclusion set. The format of rules framed in the study is shown in Figure 7.

Fuzzy inference system (FIS)

Fuzzy inference mechanism is based on the compositional rule of inference proposed
by Zadeh. By using the inference mechanism an output fuzzy set is obtained from
the rules and the input variables. For instance, a fuzzy rule expressed by equation
(3) is represented by a fuzzy relation R: (X x Y), which is computed by using
equation (4):
mR ðx; yÞ ¼ I ½mA ðxÞ; mB ÞyÞ ð4Þ
where the operator I can be either an implication or a conjunction operator (a t-norm).
From the literature (Zimmermann, 1996) it is observed that there are two most common
types of inference systems frequently used; the max-min inference and the max-prod
inference method. Examples of t-norms are the minimum, oftenly called “mamdani
implication” and the product, called the Larsen implication. In the study mamdani’s
max-min inference method is used. The min operator is used for the conjunction of the

Figure 7.
Format of rules framed on
fuzzy inference system
rule and for the implication function and the max operator is used for the aggregation Fuzzy linguistic
of the fuzzy sets. The compositional rule of inference proposed by Zadeh results in
equation (5):

mB0 ðyÞ ¼ max minðbk ; mBk ðyÞÞ; ð5Þ

bk ¼ minai;k
ai;k ¼ sup min (mA0 ðxi Þ; mA ðxÞÞ
For instance, in Figure 8, schematic representation of the fuzzy reasoning
mechanism with two rules is presented. First, the numerical input variables
(occurence, severity) are fuzzifed using appropriate membership functions.Then the
min operator is used for the conjunction and for the implication operations. The
outputs (individual fuzzy sets) are aggregated by using the max operator and
finally, the aggregated output is defuzzifed to obtain a crisp value. In the study the
Mamdani’s fuzzy inference method is used because it is typically used in modeling
human expert knowledge. All the parameters adopted in mamdani model to
generate FIS system are presented in Table VI. (In Appendix 1 the algorithm used
in Mamdani’s inference method is presented (Tsoukalas and Uhrig, 1997; Terano
et al., 1987)).

Figure 8.
Schematic representation
of the fuzzy reasoning
22,9 [System] Name ¼ ‘FRP’
Name ¼ ‘FMEA Range ¼ [0 1]
Type ¼ ‘mamdani’ NumMFs ¼ 6
Version ¼ 2.0 MF1 ¼ ‘N.Imp’:‘trapmf’, 0 0 0.10.2]
NumInputs ¼ 3 MF2 ¼ ‘Minor’:‘trimf’, [0.1 0.22 0.40]
996 NumOutputs ¼ 1 MF3 ¼ ‘Low’:‘trimf’, [0.258 0.39 0.544]
NumRules ¼ 27 MF4 ¼ ‘Moderate’:‘trimf’,[0.433 0.591 0.77]
And Method ¼ ‘min’ MF5 ¼ ‘Imp’:‘trimf’,[0.647 0.758 0.898]
Or Method ¼ ‘max’ MF6 ¼ ‘V.Imp’:‘trapmf’,[0.80 0.90 1.0 1.0]
Imp Method ¼ ‘min’ [Rules]
AggMethod ¼ ‘max’ 4 5 5, 6 (1):1
DefuzzMethod ¼ ‘centroid’ 4 4 5, 4 (1):1
[Input1] 4 5 4, 5 (1):1
Name ¼ ‘Occurrence’ 4 4 3, 5 (1):1
Range ¼ [0 10] 4 5 3, 5 (1):1
NumMFs ¼ 5 5 5 5, 6 (1):1
MF1 ¼ ‘Remote’:‘trapmf’, 0 0 1.0 2.0] 5 4 4, 5 (1):1
MF2 ¼ ‘Low’:‘trapmf’, [1.0 2.0 3.0 4.0] 5 3 5, 5 (1):1
MF3 ¼ ‘Moderate’:‘trapmf’, [3.0 4.0 6.0 7.0] 5 2 5, 5 (1):1
MF4 ¼ ‘High’:‘trapmf’, [6.0 7.0 8 9.0] 5 1 4, 3 (1):1
MF5 ¼ ‘V.High’:‘trapmf’, [8.0 9.0 10.010.0] 3 4 4, 5 (1):1
[Input2] 3 5 5, 5 (1):1
Name ¼ ‘Non Detectability’ 3 4 4, 4 (1):1
Range ¼ [0 10] 3 4 5, 4 (1):1
NumMFs ¼ 5 3 3 4, 4 (1):1
MF1 ¼ ‘Remote’:‘trapmf’, [0 0 1.0 2.0] 3 4 3, 3 (1):1
MF2 ¼ ‘Low’:‘trapmf’, [1.0 2.0 3.0 4.0] 3 3 3, 2 (1):1
MF3 ¼ ‘Moderate’:‘trapmf’, [3.0 4.0 6.0 7.0] 2 4 3, 2 (1):1
MF4 ¼ ‘High’:‘trapmf’, [6.0 7.0 8.0 9.0] 2 3 3, 2 (1):1
MF5 ¼ ‘V.High’:‘trapmf’, [8.0 9.0 10.0 10.0] 2 2 3, 1 (1):1
[Input3] 2 2 2, 1 (1):1
Name ¼ ‘Severity’ 2 4 4, 4 (1):1
Range ¼ [0 10] 2 3 4, 2 (1):1
NumMFs ¼ 5 1 2 3, 1 (1):1
MF1 ¼ ‘Remote’:‘trapmf’, 0 0 1.0 2.0] 1 3 3, 1 (1):1
MF2 ¼ ‘Low’:‘trapmf’, [1.0 2.0 3.0 4.0] 1 1 1, 1 (1):1
Table VI. MF3 ¼ ‘Moderate’:‘trapmf’, [3.0 4.0 6.0 7.0] 1 2 2, 1 (1):1
Listing of information on MF4 ¼ ‘High’:‘trapmf’, [6.0 7.0 8.0 9.0]
FIS MF5 ¼ ‘V.High’:‘trapmf’, [8.0 9.0 10.0 10.0]

Finally, defuzzification is done to obtain crisp ranking from the fuzzy conclusion set,
which is used to express the riskiness/criticality level of the failure so that corrective or
remedial actions can be prioritized accordingly. There are many defuzzification
methods available in literature but most commonly used are Chen’s ranking (1985),
Yager’s centroidal (1980) and Tran and Duckstein (2002) methods. The criteria used to
select suitable defuzzification method are disambiguity (result in unique value),
plausibility (lie approximately in the middle of the area), and computational simplicity
(Ross, 1995). In the study, Yager’s centroidal method is used for defuzzification which Fuzzy linguistic
is given by equation (6): modelling
mB 0 ð yÞy:dy
Defuzzified value ¼ R ð6Þ
mB 0 ð yÞdy
where B0 is the output fuzzy set, and mB0 i is the membership function.

Illustrative case
To demonstrate the application of proposed approach for carrying out system FMEA,
an industrial case study from a paper mill is proposed. There are many functional units
in a paper mill such as feeding, pulp preparation, pulp washing, screening, bleaching
and preparation of paper. Due to high level of saturation and increase in number of
breakdown maintenance interventions, it is decided to conduct failure mode analysis of
sub-systems in one of the main functioning unit i.e. feeding system. The feeding
system consists of three sub systems (Kumar et al., 1988), namely:
Sub system 1. the Blower, for pushing the wood chips through pipe;
Sub system 2. the chain conveyor and bucket conveyor for transporting and lifting
the chips to feed the digester;
Sub system 3. the stand-by unit (for sub system 2) when there is a failure in sub
system 2 stand-by unit is switched on which feeds the digester.
The effect of each possible cause of failure of sub system 2, which consists of chain,
conveyor and bucket conveyor, is evaluated in terms of probability of occurrence (Sf),
probability of non-detection (Sd) and severity of failure (S). The brief description of
methodology used to calculate the scores related to failure of occurrence, likelihood of
non-detection of failure and severity of failure of SS2 is as followed.

Probability of occurrence of failure (Sf)

Probability of occurrence of failure is evaluated as a function of mean time between
failures. The data related to mean time between failures of SS2 is obtained from previous
historical records, maintenance log-books and is then integrated with the experience of
maintenance personnel. For instance, if MTBF of component is between three to six
months then probability of occurrence of failure is high (occurrence rate 0.5-1 per cent)
with the score ranging between 7-8. Table II presents the descriptive assessment of
probability of failure occurrence with corresponding MTBF and scores assigned.

Probability of non-detection of failures (Sd)

The chance of detecting a failure cause or mechanism depends on various factors such
as ability of operator or maintenance personnel to detect failure through naked eye or
by periodical inspection or with the help of machine diagnostic aids such as automatic
controls, alarms and sensors. For instance, probability of non-detection of failure of a
component through naked eye is say, 0-5 per cent is ranked 1 with non-detectability
remote. The values of Sd for various failure causes reported in the study are evaluated
according to the score reported in Table III.
IJQRM Severity of failure (S)
22,9 Severity of failure is assessed by the possible outcome of failure effect on the system
performance. The severity of effect may be regarded as remote, moderate or very high.
In the study the data related to mean time to repair (MTTR) obtained from historical
records, supported with the experience of maintenance personnel is used to obtain
score for severity. For instance, if MTTR of facility/component is less than, say one
998 hour, effect may be regarded as remote. If external intervention is required for repairs,
or MTTR exceeds four days, then effect may be regarded as high and if system
degrades resulting in line shut down/production stoppage, then the severity may be
regarded as very high. Apart from the parameters listed in Table IV, the economics
(associated with maintenance, spares and manpower) and safety aspects can also be
used to obtain more reliable score.

Results and conclusions

Table VII presents the traditional FMEA analysis for the hydraulic system. The
numerical values of FMEA parameters, i.e. Sf, S and Sd are obtained by using the
methodology discussed above (this makes use of maintenance records integrated with
the knowledge of maintenance people/experts). Then RPN number for each failure
cause is evaluated by multiplying the factor scores using equation (2) The resulting
RPN number and priority ranking is presented in the last two columns of Table VII.
To investigate the applicability of proposed approach for prioritizing failure causes
of hydraulic system, Fuzzy risk assessment is carried out by adopting a four-step
procedure (i.e. fuzzification, rule formation and evaluation and defuzzification)
discussed previously in detail. The results so obtained are presented in Table VIII. It is
depicted in Table VIII that the ranking of priority of various failure causes obtained
from the traditional FMEA is altered (CauseG . CauseE . CauseA . CauseF .
CauseD . CauseB . CauseC). For instance, Cause G turned out to be one of the most
critical failure causes in terms of RPN, while, after conducting fuzzy criticality
assessment using FIS it ranks only at the second place, with a score of 0.653. At the
same time, Cause F becomes the most critical one, with an overall score of 0.766. The
order of failure causes obtained by fuzzy critical assessment approach (FCAM) is
(CauseF . CauseG . CauseB . CauseD . CauseE . CauseA . CauseC.)
For instance, the FCAM output for failure Cause A (with inputs S f ¼ 7, S ¼ 7 and
S d ¼ 6) is 0.543. And cause C (with inputs S f ¼ 3, S ¼ 5 and S d ¼ 5) is 0.247, as
shown in Figure 9 (a) and (b). The mapping of inputs (Sf), (Sd) and (S) to the output
(FRPN) through the linguistic If-Then rules adopted in the study is represented using a
control surface plots (Figure 10 (a)(b) and (c)). The plots help to examine the
consistency of the rules framed in FIS. The surface displays the dependency of the
output as a function of the inputs – that is, it illustrates the entire span of the output set
based on the entire span of the input(s) set. These three-dimensional plots represent
very well a two-input and one-output system. Since in the study we have used three
inputs, i.e. (Sf), (S) and (Sd) so the surface plot can be represented with a group of
surfaces keeping one of the input variables stable.
Thus, the uncertainty in the traditional assessment is solved using fuzzy linguistic
modeling. The use of fuzzy set approach, confirms that how fuzzy linguistic
assessment methodology which makes use of membership functions, a well defined
fuzzy rule base and an inference system can enhance and improve the understanding of
Potential Potential cause of
Component Function failure mode Potential effect of failure failure *S
*S *S

Solenoid valve Act as energizer, i.e. to control Breaking Piston fails to execute the Burning of magnet 7 6 7 294 3
piston stroke movement (Cause A)
Pressure regulator Control or regulate pressure Breaking Pressure out of range Mechanical stresses 5 4 7 140 6
(Cause B)
Hydraulic pistons To carry out movements Blow-by Loss of fluid Breaking of seal 3 5 5 75 7
(Cause C)
Breaking Piston fails to execute the Breaking of piston rod 3 7 8 168 5
movement (Cause D)
Hydraulic gear case Maintain the fluid pressure Breaking, Lack of adequate oil pressure Pump wear (Cause E) 8 7 7 392 2
clogging for movement
Leakage Loss of fluid Blow-by (Cause F) 4 6 10 240 4
Pressure tubes Breaking Lack of adequate oil pressure Corrosion (Cause G) 6 9 10 540 1
Notes: *(Sf): probability of occurrence; *(S): severity; *(SD): non-detectability
Fuzzy linguistic

element: hydraulic

(System: feeding; Primary
subsystem2 (SS2)
primary element of
Traditional FMEA of
Table VII.
IJQRM the dynamics of a complex problems in which decisions are to be made from imprecise,
22,9 vague and subjective information. It is concluded from the study that fuzzy logic-based
approach not only resolves the limitations associated with traditional methodology for
RPN evaluation of failure causes in reliability analysis of system but also offers added
advantages such as:
Both quantitative data and, vague or imprecisely defined qualitative
1000 information, can be used in criticality/risk assessment in a consistent manner.

Potential failure Traditional RPN Ranking Fuzzy (FIS) RPN

cause output (traditional) output Ranking (fuzzy)

Cause A 294 3 0.543 6

Cause B 140 6 0.620 3
Table VIII. Cause C 75 7 0.247 7
Comparison of fuzzy (FIS) Cause D 168 5 0.603 4
output and traditional Cause E 392 2 0.596 5
RPN output for various Cause F 240 4 0.766 1
failure causes Cause G 540 1 0.653 2

Figure 9.
(a) and (b) Representation
of fuzzy critical system
output for failure Cause A
and Cause C
Fuzzy linguistic


Figure 10.
Control surface plots
It permits the experts to combine probability of occurrence (Sf), severity (S) and
22,9 detectability (Sd) of failure modes in more flexible and realistic manner by using
their judgment, experience and expertise.

1002 Bowles, J.B. and Pelaez, C.E. (1995), “Application of fuzzy logic to reliability engineering”,
Proceedings of the IEEE, Vol. 83 No. 3, pp. 435-49.
Chen, S.H. (1985), “Ranking fuzzy numbers with maximizing and minimizing set”, Fuzzy Sets and
Systems, Vol. 17 No. 2, pp. 113-29.
Ebeling, C. (2000), An Introduction to Reliability and Maintainability Engineering, Tata
McGraw-Hill Company Ltd, New York, NY.
Gilchrist, W. (1993), “Modeling failure mode and effect analysis”, International Journal of Quality
& Reliability Management, Vol. 10 No. 5, pp. 16-23.
Klir, G.J. and Yuan, B. (1995), Fuzzy Sets and Fuzzy Logic: Theory and Application, Prentice-Hall,
Englewood Cliffs, NJ.
MIL-STD 1629 (1980), Military Standard Procedures for Performing a Failure Mode, Effects and
Criticality Analysis, Dept of Defense, Washington, DC.
O’Connor, P.D.T. (2000), Practical Reliability Engineering, Heyden, London.
Ross, T.J. (1995), Fuzzy Logic with Engineering Applications, McGraw-Hill, New York, NY.
Sharma, R. (2004), “Fuzzy logic methodology to prioritize failure causes in FMEA”, Proceedings
of International Conference on Emerging Technologies, ICET- 2004, Allied Publishers,
New Delhi, pp. 298-306.
Terano, T., Asai, K. and Sugeno, M. (1987), Fuzzy Systems Theory and its Application, Academic
Press, San Diego, CA.
Tran, L.T. and Duckstein, L. (2002), “Comparison of fuzzy numbers using fuzzy distance
measure”, Fuzzy Sets and Systems, Vol. 130 No. 3, pp. 331-41.
Tsoukalas, L.H. and Uhrig, R.E. (1997), Fuzzy and Neural Applications in Engineering, John Wiley
& Sons, New York, NY.
Wang, J., Yang, J.B. and Sen, P. (1996), “Safety analysis and synthesis using fuzzy sets and
evidential reasoning”, Reliability Engineering and System Safety, Vol. 47, pp. 103-18.
Xu, K. and Tang, L.C. (2002), “Fuzzy assessment of FMEA for engine systems”, Reliability
Engineering & System Safety, Vol. 75, pp. 17-29.
Yager, R.R. (1980), “A general class of fuzzy connectives”, Fuzzy Sets and Systems, Vol. 130 No. 3,
pp. 331-41.
Zadeh, L. (1965), “Fuzzy sets”, IEEE Information and Control, Vol. 8, pp. 338-53.
Zimmermann, H. (1996), Fuzzy Set Theory and its Applications, 3rd ed., Kluwer Academic
Publishers, London.

Further reading
Ben-Daya, M. and Raouf, A. (1996), “A revised failure mode and effect analysis model”,
International Journal of Quality & Reliability Management, Vol. 13 No. 1, pp. 43-7.
Driankov, D., Hellendoorn, H. and Reinfrank, M. (1993), An Introduction to Fuzzy Control,
Springer-Verlag, Berlin.
Ford Motor Company (1988), “Potential failure mode and effects analysis in design (design Fuzzy linguistic
FMECA) and for manufacturing and assembly process (process FMECA): instruction
manual”, internal report, September, Ford Motor Company, Detroit, MI. modelling
Kara-Zatri, C., Keller, A.Z. and Fleming, P.V. (1992), “A smart failure mode and effect analysis
package”, Proceedings of Annual Reliability and Maintainability Symposium, pp. 414-21.
Kruse, R., Gebhardt, J. and Klawonn, F. (1994), Foundations of Fuzzy Systems, John Wiley &
Sons, Chichester. 1003
Kumar, D., Singh, I.P. and Singh, J. (1988), “Reliability analysis of feeding system in paper
industry”, Microelectronics and Reliability, Vol. 2 No. 28, pp. 213-15.
Stamatis, D.H. (1995), Failure Mode and Effects Analysis – FMEA from Theory to Execution,
ASQC Quality Press, New York, NY.
Teng, S.H. and Ho, S.Y. (1996), “Failure mode and effects analysis: an integrated approach for
product design and process control”, International Journal of Quality & Reliability
Management, Vol. 13 No. 5, pp. 8-26.

Appendix 1
Fuzzy inference system (Mamdani type.) Max-min inference algorithm
k ¼ number of rules
For (i ¼ 1; i ¼ k; iþ þ) {/ * Calculate the degree of fulfillment bi of the antecedent part of the
rules */

bi ¼ mAi1 ðx1 Þ ^ mA i2 ðx2 Þ ^ · · · ^ mAipðXpÞ

/ * Derive the output fuzzy set b0i using the minimum t-norm */

mB0 i ð yÞ ¼ bi ^ mBi ð yÞ

/ * Compute the aggregated output fuzzy set by taking the maximum of the individual conclusion
mB0 i */

mB0 ð yÞ ¼ maxi ¼ 1; 2. . .kmB0 i ð yÞÞ}

Appendix 2
Assume that n experts are asked for some x[ X to evaluate the proposition “x belongs to A” as
either true or false where (A is a fuzzy set on X that represents a linguistic term associated with
linguistic variable).
Given a particular element x[ X,
Let ai(x) denote the answer of expert i (i[ N n ).
Let us assume that
ai(x) ¼ 1, when the proposition valued by expert i is true; and
ai(x) ¼ 0, when the proposition valued by expert i is false.
Then probabilistic interpretation of constructed membership function is given by equation as:

ai ðxÞ
AðxÞ ¼
IJQRM When the experts have different degree of competencies, Ci, with regard to the model being
constructed, Equation is modified as:
AðxÞ ¼ C i ai ðxÞ
where C i ¼ 1 and the value of Ci depends upon the knowledge of expert and their expertise in
1004 i¼1
the relevant field.

FMEA Failure mode effect analysis
RPN Risk priority number
FRPN Fuzzy risk priority number
MTBF Mean time between failures
MTTR Mean time to repair.
Sf Frequency of occurrence
S Severity of failure
Sd Non-detectability of failure
CN Criticality number