Engineering Failure Analysis 135 (2022) 106163

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Improved FMECA for effective risk management decision making

by failure modes classification under uncertainty
C.M. La Fata a, *, A. Giallanza a, R. Micale b, G. La Scalia a
a
Dipartimento di Ingegneria, Università degli Studi di Palermo, Viale delle Scienze, 90128, Palermo, Italy
b
Dipartimento di Ingegneria, Università degli Studi di Messina, Contrada di Dio, 98166, Messina, Italy

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

Keywords: Failure Mode, Effects, and Criticality Analysis (FMECA) is a proactive reliability and risk man­
FMECA agement technique extensively used in practice to ensure high system performance by prioritising
Interval-valued ELECTRE TRI failure modes. Owing to the limitations of traditional FMECA, multi-criteria decision-making
Failure modes classification
methods have been employed over the past two decades to enhance its effectiveness. To consider
Propulsion system
the vagueness and uncertainty of the FMECA evaluation process, an interval-based extension of
the Elimination et Choice Translating Reality (ELECTRE) TRI method is proposed in the present
paper for the classification of failure modes into risk categories. Therefore, ratings of failure
modes against risk parameters are provided in an interval form by a panel of FMECA experts and
then properly synthesised to provide a group consensus and improve the accuracy of the results.
The relative importance of the risk parameters is also considered. The method is validated using a
numerical example relating to the propulsion system of a fishing vessel.

1. Introduction

Failure Mode and Effects Analysis (FMEA) is a very useful and straightforward technique for system reliability and risk analysis
aimed at identifying potential Failure Modes (FMs) and their causes and effects on system performance [24]. As an extension of FMEA,
Failure Mode, Effects, and Criticality Analysis (FMECA) allows the prioritisation of FMs based on the Risk Priority Number (RPN),
computed by the product of the FM ratings against the risk parameters Severity (S), Occurrence (O), and Detection (D). By referring to
the Standard IEC 60812, S, O, and D are evaluated on a ten-point scale (Table 1), so that the higher the RPN, the higher the criticality of
the FM.
Despite the wide use of conventional FMECA in different contexts [54,43,19], its application has numerous limitations, including
the following [27,10,7,13;6,5;2,4,36,30,56]:

- The computation of RPN by the simple product of S, O, and D ratings is too simplistic and strongly influenced by their variations.
- S, O, and D are equally weighted.
- Diverse S, O, and D values may return the same RPN, although the FM implications are significantly different.

* Corresponding author.
E-mail addresses: concettamanuela.lafata@unipa.it (C.M. La Fata), antonio.giallanza@unipa.it (A. Giallanza), rosa.micale@unime.it (R. Micale),
giada.lascalia@unipa.it (G. La Scalia).

https://doi.org/10.1016/j.engfailanal.2022.106163
Received 3 November 2021; Received in revised form 11 February 2022; Accepted 15 February 2022
Available online 18 February 2022
1350-6307/© 2022 Elsevier Ltd. All rights reserved.
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

- General assessment structure of FMECA is group-oriented, and the FMECA group comprises experts with different levels of
knowledge and experience. Judgments can often be vague and uncertain so that experts are often unlikely able to precisely evaluate
FMs on S, O, and D.

Consequently, a wide variety of enhanced versions of FMECA have been proposed in the literature to overcome the aforementioned
disadvantages [11;12;8]. In this regard, numerous contributions integrate FMECA with Multi-Criteria Decision-Making (MCDM) ap­
proaches to hierarchically decompose the system under investigation into its basic components and to identify the criticality of FMs by
means of prioritisation. Based on the obtained FMs ranking, the Decision Maker (DM) has to decide the criticality threshold above
which proper preventive and/or corrective measures have to be implemented as a matter of urgency, and this choice unavoidably
depends on both the DM expertise and the amount of material, financial, and human resources available. On the other hand, this issue
may be overcome a priori by the classification of FMs into ordered risk categories that allow the DM to promptly analyse and visualise
FMECA results, thus ensuring more efficient actions and decisions [1,9,52]. To address this issue also taking into account the un­
certainty of information and the vagueness of human feelings and recognition in real FMECA contexts, an interval-valued extension of
the ELimination Et Choice Translating REality (ELECTRE) TRI MCDM method is designed in the present paper. ELECTRE TRI [46,57] is
an outranking MCDM method that addresses sorting decision problems where alternatives are assigned to ordered categories. Sorting
problems refer to absolute judgments, namely the intrinsic value of an alternative arises from its comparison with the boundary
profiles of classes properly defined. As a result, the assignment class of an alternative is independent of the others. On the other hand,
ranking problems are based on pairwise comparisons of alternatives against evaluation criteria, and the ranking position of an
alternative derives from its comparison with the others. As a consequence, the insertion of a whatever new alternative may consid­
erably change the ranking results [44;6;40]. In addition, the use of ELECTRE TRI enables the intransitivity of preferences that generally
occurs in practical applications [55,26,16] and allows the DM to consider the relative importance of criteria. By introducing specific
thresholds, the method also provides a way to properly manage the gradual transition of the DM from the indifference to the preference
condition during the evaluation process. In the context of the present paper, alternatives to be categorised are hence the FMs obtained
by FMEA, whereas the evaluation criteria are the risk parameters S, O, and D. Based on the DM expertise, perception, and need to
adequately differentiate the criticality of FMs, risk categories are ordered and defined a priori (i.e. negligible, minor, moderate, sig­
nificant and severe). Therefore, a panel of experts is involved to rate alternatives against criteria. In this regard, FMECA team members
are allowed to provide interval-valued ratings rather than crisp ones owing that intervals are deemed to be the most suitable and
simplest way to deal with decision making problems in uncertain environments [49]. Afterwards, the interval-valued scores of al­
ternatives against criteria are aggregated to provide a group consensus, thus improving the accuracy of the FMECA results. To
demonstrate the validity of the designed interval-based ELECTRE TRI approach to FMECA, it is implemented in the propulsion system
of a fishing vessel operating in the west sea of Sicily.
The paper is organised as follows. Section 2 provides the literature review, and the proposed methodological approach is given in
Section 3. Section 4 details the application case, and conclusions are presented in Section 5.

2. Literature review on FMECA by MCDM

In recent years, several theories and methods have been proposed to enhance the assessment ability of FMECA, overcoming the
main shortcomings of its traditional version. Accordingly, numerous literature reviews on improved FMECA and related applications
have been published [30]; Dabous et al., 2021a, 2021b; [56,29]. These reviews have shown that MCDM methods have been widely
exploited by academicians and researchers to overcome RPN limitations. Owing to its ability to discretize the risk analysis process into
its phases, MCDM represents a very useful decision-aiding support approach for FMECA applications to consider the relative impor­
tance of risk parameters, and to prioritise FMs by using rigorous mathematical models [30]. In this regard, Zammori and Gabbrielli
[58] integrated the Analytic Network Process (ANP) in FMECA to evaluate the interactions among the main failure causes. Kutlu and
Ekmekçioğlu [27] combined the Fuzzy Analytic Hierarchy Approach (FAHP) and the Fuzzy Technique for Order of Preference by
Similarity to Ideal Solution (FTOPSIS) to weigh risk factors and prioritise FMs, respectively. Grey Relational Analysis (GRA) and the
DEcision-MAking Trial and Evaluation Laboratory (DEMATEL) were proposed by Chang et al. [8] to compare the impact of FMs. Liu
et al. [32] designed an intuitionistic fuzzy hybrid TOPSIS, whereas Liu et al. [31] combined interval 2-tuple linguistic variables with

Table 1
Scales of S, O, and D.
S O D Rating

None Remote: failure is nearly impossible Almost certain 1

Very minor Low: relatively few failures Very high 2
Minor High 3
Low Moderate: occasional failures Moderately high 4
Moderate Moderate 5
Significant Low 6
Major High: repeated failures Very low 7
Extreme Remote 8
Serious Very high: failure is almost inevitable Very remote 9
Hazardous without warning Absolutely uncertain 10

2
 Table 2
FMECA studies in the literature.
Reference Method Risk parameters Goal Multi- Uncertainty of Sector/Case study

C.M. La Fata et al.

group in FM ratings
FM ratings

Zammorri & ANP S, O and D, while S consists of three sub-parameters, i.e. FMs ranking Control system of a power
Gabrielli, damages, production and maintenance costs shovel
2011
[27] FAHP and FTOPSIS S, O and D FMs ranking √ √ Automotive industry
[8] GRA and DEMATEL S, O and D FMs ranking Thin-film transistor liquid
crystal display cell process from
a professional LCD
manufacturer
[32] Intuitionistic fuzzy hybrid TOPSIS S, O and D FMs ranking √ √ Colour super twisted nematic
(CSTN)
[31] Interval 2-tuple linguistic variables and S, O and D FMs ranking √ √ Tertiary care hospital with an
GRA empirical application on a C-arm
X-ray machine
[34] FAHP, entropy approach and fuzzy S, O and D FMs ranking √ √ Tertiary care university hospital
VIKOR with an empirical application on
a general anaesthesia process
[38] FlowSort group decision support system S, O and D FMs sorting Plastic bottles manufacturing by
a blow-moulding process
[33] Cloud Model Theory and PROMETHEE S, O and D FMs ranking √ √ Hospital emergency department
[6] ELECTRE TRI S, O and D FMs sorting Dairy manufacturing industry
[2] Intuitionistic FMEA-WASPAS S, O, D, cost, duration of exposure, and system safety factors Ranking of both FMs √ √ Medium voltage cell assembly
and corrective- line
preventive strategies
[3] AHP and FTOPSIS O, time of operation and modality of execution of FMs ranking √ Street-cleaning vehicle
3

maintenance actions
[20] FBWM and Z-MOORA S, O and D FMs ranking √ √ Automotive industry
[35] BWM and GRA S, O, D and expected cost FMs ranking √ √ Electronics company of
smartphone manufacturing
[21] AHP and ERVD S, O and D. In addition, S consists of two sub-parameters, i.e. FMs ranking Squirrel-cage induction motors
SHE and operational severity. SHE comprises safety, health of a leading steel process plant
and environment, while operational severity comprises P-F
interval, maintenance cost, production lost and age of
equipment
[23] Fuzzy BWM and fuzzy Bayesian network O, D and S, while S consists of product flexibility, product FMs ranking √ Plastic production
colour and form, and product strength
[37] DEMATEL, SAW, VIKOR, COPRAS, GRA S, O, D, expected cost and environmental protection FMs ranking √ Machine tool manufacturing

Engineering Failure Analysis 135 (2022) 106163

and TOPSIS awareness company
[59] FAHP and FTOPSIS O, D and S, while S comprises the sub-criteria project cost, Project risks ranking √ √ Agricultural projects
time, and quality
[60] Linguistic neutroscopic numbers, S, O and D FMs ranking √ √ Water gasification system
modified PROMETHEE based on regret
theory, maximizing deviation model and
TOPSIS
[22] Interval-valued spherical weighted S, O, D, cost, prevention, and effectiveness of OHS measures FMs ranking √ √ Marble manufacturing facility
arithmetic mean operator (IVSWAM) and
interval-valued spherical fuzzy extension
of TOPSIS (IVSF-TOPSIS)
[28] AHP and VIKOR S, O and Human Error Probability (HEP) FMs ranking Wooden products
manufacturing company
[45] FAHP, Design Structure Matrix, BWM and S, O and D, while S consists of three sub-parameters, i.e. FMs ranking √ √ Centrifugal pump
FTOPSIS security, cost and environment
Proposed Interval-valued ELECTRE TRI S, O and D FMs sorting √ √ Propulsion system
approch
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

GRA to optimise the effectiveness of FMEA. Referring to a general anaesthesia process, Liu et al. [34] weighed risk factors using a
combined FAHP and entropy approach to consider the uncertainty and vagueness of human subjective perception and experience.
Subsequently, the prioritisation of the FMs was obtained by using the Fuzzy VIšekriterijumsko KOmpromisno Rangiranje (FVIKOR)
method. The FlowSort group decision support system was proposed by Lolli et al. [38] to sort FMs into priority classes. Liu et al. [33]
used Preference Ranking Organization METHod for Enrichment of Evaluations (PROMETHEE) to find out the criticality of FMs in a
hospital emergency department. Considering a single DM context and referring to a dairy manufacturing industry, Certa et al. (2017b)
implemented the traditional ELECTRE TRI technique to categorise FMs. Carpitella et al. [3] suggested FTOPSIS to prioritise the FMs of
a street-cleaning vehicle. Three evaluation risk parameters were considered, of which two were specifically related to maintenance. Lo
and Liou [35] combined the Best Worst Method (BWM) and GRA to address the uncertainty of information in FMECA applications.
With respect to the automotive spare parts industry, Ghoushchi et al. [20] first used fuzzy BWM to get the relative importance of risk
parameters and to rank FMs by Multi-Objective Optimisation using Ratio Analysis based on Z-number theory (Z-MOORA). Referring to
a medium voltage cell assembly line, Can [2] combined intuitionistic FMECA and Weighted Aggregated Sum Product Assessment
(WASPAS) to rank both FMs and corrective-preventive strategies, also considering cost, duration of exposure, and system safety

FMECA

FMECA Group

Identification of failure modes,

their causes and effects

Elicitation of preference
Elicitation of interval-valued
parameters (i.e. thresholds,
ratings of failure modes against
profiles and relative importance
O, S and D and aggregation Interval-valued ELECTRE
of risk factors )
TRI

Classification of failure modes

Correction NO
FMECA Report
required?

YES

Implement corrective measures

Fig. 1. Methodological approach.

4
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

parameters. Gugaliya et al. [21] computed criteria weights by using AHP, whereas the FMs of a process plant induction motor were
prioritised by the Election based on Relative Value Distance (ERVD). Lo et al. [37] considered further criteria besides S, O, and D and
then proposed the DEcision-MAking Trial and Evaluation Laboratory (DEMATEL) to compute their influential weights. In addition,
different MCDM methods (e.g., Simple Additive Weighting - SAW, VIKOR, and GRA) were applied to develop different rankings, and
TOPSIS was finally used to integrate the results. As regards the plastic production sector, Gul et al. [23] proposed an improved FMECA
with fuzzy BWM and a fuzzy Bayesian network. Zhu et al. [60] combined PROMETHEE, regret theory, and TOPSIS to assess the risk
priority of the FMs of a water gasification system. Aiming to support risk management for an agricultural project, Zandi et al. [59]
defined the severity criterion as constituted by three sub-criteria (i.e., project cost, time, and quality) and used the FAHP to weigh
them. The final risk ranking was obtained by FTOPSIS. Mzougui and Felsoufi [45] first used the FAHP to weigh all team members on
the basis of their knowledge. Subsequently, the Authors combined BWM and FTOPSIS to weigh the risk parameters and prioritise FMs,
respectively. In La Fata et al. [28], the prioritisation of the occupational risks of a Sicilian manufacturing company was performed by
the combination of AHP and VIKOR. Three risk parameters were considered of which one specifically referred to the contribution of
human factors to the occurrence of accidents. Gul and Ak [22] used fuzzy sets to model the vagueness and uncertainty of DMs and then
proposed TOPSIS to determine the priorities of failures. As a novelty to traditional RPN, cost, prevention, and effectiveness risk pa­
rameters were considered in addition to S, O, and D. FMECA studies mentioned in the literature review are listed in Table 2, according
to their contributions in the field.

3. Methodological approach

The interval-based extension of ELECTRE TRI to FMECA is described in the present section and is synthesised in Fig. 1.

3.1. Interval-valued ELECTRE TRI method

ELECTRE TRI is a sorting MCDM technique developed by Roy [46] to assign alternatives to a set of ordered classes defined a priori.
Boundaries between classes are referred to as reference profiles and they may be calculated by proper elicitation techniques or sug­
gested by experts [53,14]. The assignment of an alternative to a class is derived from a comparison of its scores against criteria with the
reference profiles [44]. Therefore, let j be the jth criterion belonging to the set J of criteria and let (C1, C2, …, Cc) be the c ordered
classes, where C1 and Cc are the worst (i.e. severe) and the best (i.e. negligible) classes respectively. For the generic category Cf | f = (1,
2, …, c), bf− 1 and bf are the lower and upper reference profiles that limit the class itself (Fig. 2).
The first step of ELECTRE TRI is the development of the outranking relations Sr. For every alternative ai and reference profile bf, the
outranking relation Sr (i.e., “ai Sr bf”) means that “ai is at least as good as bf.” The development of the outranking relation “ai Sr bf” relies
on the comparison between the alternative and the reference profiles (and vice versa) to verify the concordance (i.e., a sufficient
majority of criteria support the statement “ai Sr bf”) and discordance principles (i.e., the minority of criteria are not strong enough to
deny the assertion “ai Sr bf”) [44]. For every couple of alternative ai and reference profile bf | f = 1, 2, …, (c − 1) and vice versa, the
validation of the outranking relation is based on the computation of the partial and global concordance indices and of the discordance
index. Afterwards, the credibility index σ(ai, bf) ∈ [0,1] (σ (bf, ai), respectively) is defined. It denotes the degree of credibility of the
assertion “ai Sr bf” (“bf Sr ai,” respectively). In this regard, the following preference parameters need to be set [48,47,18;6;41;42]:

- Relative importance wj (i.e., weight) of the jth criterion, ∀j ∈ J.

- Score rj(ai) of the alternative ai against the criterion j, ∀ai |i = (1, 2, ..., I) and ∀j ∈ J.

Criterion

1
2
3
C1 C2 Cf Cf+1 Cc

J-1
J
b0 b1 b2 bf-1 bf bc
Profile

Fig. 2. Ordered classes.

5
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

- Discrimination thresholds qj, pj, and vj [17]. On a specific criterion j, they are used to model situations in which the difference
between the scores associated with the alternative ai and the profile bf (and vice versa) may
a) be compatible with the indifference between the alternative ai and the profile bf (indifference threshold qj),
b) justify the preference in favour of the alternative ai with respect to the profile bf and vice versa (preference threshold pj), or
c) be interpreted as a hesitation between opting for a preference or an indifference between the alternative ai and the profile bf
(veto threshold vj).

In the context of the present study, FMs identified by the involved panel of experts are the alternatives to be assigned, whereas the
evaluation criteria are the parameters S, O, and D. Owing to the unavoidable uncertainty and vagueness of information and human
feeling in real-life applications of FMECA, every FM is supposed to be evaluated by precise or interval values by the involved panel of N
[ ]
experts. Let rj,k (ai ) = rj (ai ); rj (ai ) be the score of ai against j provided by the expert k in an interval form. Subsequently, ratings
k
arising from the N involved experts have to be synthesised for every criterion and alternative. To this end, the weighted arithmetic
mean is used in intervals in the context of the present study. Because zk is the weight of expert k, the aggregated interval-valued score of
[ ]
ai on criterion j, that is,rj (ai ) = rj (ai ); rj (ai ) , is computed as follows (Eq. (1)):
∑N [ ]
[ ] k=1 zk ⋅ rj (ai ); rj (ai ) k
rj (ai ) = rj (ai ); rj (ai ) = ∑N . (1)
k=1 zk

In this regard, the main interval operators are given in Appendix A; the reader may refer to Jaulin et al. [25] for more details.
Accordingly, the formulation of the ELECTRE TRI indices must be fixed. For every criterion j | j=(1, 2, …, J) and profile bf | f = 1, 2, …,
( ) ( )
(c − 1), let rj bf be the precise score of the profile bf against criterion j. rj bf , thresholds, and wj are assumed as crisp values that may
( )
be represented by degenerate intervals. Moreover, let’s indicate the interval-valued difference between rj(ai) and rj(bf) by δj ai , bf =

Table 3
Interval-valued partial concordance index.
Condition Formula
{ {
δj ⩽qj ( ) Cj = 1 (2)
Cj ai ; bf =
δj ⩽qj Cj = 1

{ ⎧
δj ⩽qj r (b ) − rj (ai ) + pj (3)
⎨ Cj = j f
⎪
⎪
qj < δj ⩽pj (
Cj ai ; bf
)
= pj − qj
⎪
⎪
⎩ C =1 j

{ ⎧
qj ⩽δj < pj ⎪
⎪ rj (bf ) − rj (ai ) + pj (4)
⎪ Cj =
⎪
⎪
δj ⩽pj ( ) ⎨ pj − qj
Cj ai ; bf =
⎪
⎪
⎪ rj (bf ) − rj (ai ) + pj
⎩ Cj =
⎪
⎪
pj − qj

{ ⎧
qj ⩽δj ⩽pj ⎪ Cj = 0 (5)
⎪
⎨
δj > pj (
Cj ai ; bf
)
= rj (bf ) − rj (ai ) + pj
⎩ Cj =
⎪
⎪
pj − qj

{ {
δj ⩾pj ( ) Cj = 0 (6)
Cj ai ; bf =
δj ⩾pj Cj = 0

{ {
δj ⩽qj ( ) Cj = 0 (7)
Cj ai ; bf =
δj ⩾pj Cj = 1

6
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

[ ( ) ( )]
δj ai , bf ; δj ai , bf . When a decreasing preference versus of criteria is considered, “ai Sr bf” is developed as follows:

( ) [ ( ) ( )] (
a) The interval-valued partial concordance index Cj ai , bf = Cj ai , bf ; Cj ai , bf is computed under criterion j. To do this, Cj ai ,
) [ ( ) ( )] ( ) [ ( ) ( )] ( ) [ ] ( ) [ ]
bf = Cj ai , bf ; Cj ai , bf and δj ai , bf = δj ai , bf ; δj ai , bf are replaced with Cj ai , bf = Cj ; Cj and δj ai , bf = δj ; δj ,
respectively, for simplification, and Equations (2)–(7) are used (Table 3).
[ ]
b) The interval-valued global concordance index C(ai , bf ) = C(ai , bf ); C(ai , bf ) is computed (Eq. (8)). It represents the extent to which
ratings of ai and bf against all criteria support the outranking relation “ai Sr bf”:
∑J
j=1 wj ⋅ Cj (ai , bf )
C(ai , bf ) = ∑J (8)
j=1 wj

( ) [ ( ) ( )]
c) The interval-valued discordance index Dj ai , bf = Dj ai , bf ; Dj ai , bf is computed under criterion j. It represents the extent to
( ) [ ( ) ( )] ( )
which criterion j does not support the assertion “ai Sr bf.” To compute this index, Dj ai , bf = Dj ai , bf ; Dj ai , bf and δj ai , bf =
[ ( ) ( ) ] ( ) [ ] ( ) [ ]
δj ai , bf ; δj ai , bf are replaced with Dj ai , bf = Dj ; Dj and δj a, bf = δj ; δj , respectively, for simplification, and Equations
(9)–(14) are used (Table 4).
[ ]
d) The credibility index σ(ai , bf ) = σ (ai , bf ); σ (ai , bf ) of the outranking relation is calculated using the equation (15), where J*
comprises those criteria for which Dj (ai , bf )⩾C(ai , bf ).

Table 4
Interval-valued discordance index.
Condition Formula
{ {
δj ⩽pj ( ) Dj = 0 (9)
Dj ai ; bf =
δj ⩽pj Dj = 0

{ ⎧
δj ⩽pj ⎪ Dj = 0 (10)
⎪
⎨
pj < δj ⩽vj (
Dj ai ; bf
)
= rj (ai ) − rj (bf ) − pj
⎩ Dj =
⎪
⎪
vj − pj

{ ⎧
pj ⩽δj ⩽vj ⎪
⎪ rj (ai ) − rj (bf ) − pj (11)
⎪ Dj =
⎪
⎪
δj ⩽vj ( ) ⎨ vj − pj
Dj ai ; bf =
⎪
⎪
⎪ rj (ai ) − rj (bf ) − pj
⎩ Dj =
⎪
⎪
vj − pj

{ ⎧
pj < δj ⩽vj r (a ) − rj (bf ) − pj (12)
⎨ Dj = j i
⎪
⎪
δ j > vj (
Dj ai ; bf
)
= vj − pj
⎪
⎪
⎩ D =1 j

{ {
δj ⩾vj ( ) Dj = 1 (13)
Dj ai ; bf =
δj ⩾vj Dj = 1

{ {
δj ⩽pj ( ) Dj = 0 (14)
Dj ai ; bf =
δj ⩾vj Dj = 1

7
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

∏1 − Dj (ai , bf )
σ(ai , bf ) = C(ai , bf ) ⋅ , (15)
j∈J *
1 − C(ai , bf )

Afterwards, the preference relation between ai and bf is determined by the comparison of both σ(ai, bf) and σ (bf, ai) with the cutting
level λ, the latter defined as the smallest credibility index to adhere to the outranking relation “ai Sr bf. ” λ belongs to the range [0.5; 1]
∑
and must be > [1 − (max(wj)/ Jj=1 wj )] [48,39]. Finally, the following binary relationships may occur (Fig. 3):

- (ai I bf) ⇔ (ai Sr bf) and (bf Sr ai),

- (ai ≻ bf) ⇔ (ai Sr bf) and not (bf Sr ai),
- (bf ≻ ai) ⇔ not (ai Sr bf) and (bf Sr ai), and
- (ai R bf) ⇔ not (ai Sr bf) and not (bf Sr ai),

where ≻, I, and R represent the preference, indifference, and incomparability relations, respectively.
The higher the cutting level λ, the more difficult for the relation “ai Sr bf” to hold and vice versa [51]. Correspondingly, the number
of incomparability relations increases.
From the obtained binary relations, the alternatives are assigned to classes. In this regard, the pessimistic and optimistic assignment
procedures are proposed. The pessimistic procedure compares the alternative ai with the profile bf, starting from the highest class Cc. If
bf is the first profile for which “ai Sr bf,” the procedure assigns the alternative ai to class Cf+1. Starting from the lowest class C1, the
optimistic procedure compares the alternative ai with the profile bf. If bf is the first profile for which “bf ≻ ai,” the procedure assigns the
alternative ai to class Cf. The procedures return different assignment results if incomparability relations occur. In this case, the
pessimistic assignment procedure leads to assign alternatives to lower classes, whereas the optimistic one assigns alternatives to higher
classes.

4. Case study: Results and discussion

The designed interval-valued ELECTRE TRI-based approach to FMECA is implemented in the real case described in Certa et al. [5].
Referring to the propulsion system of a fishing vessel operating on the west seaside of Sicily, Certa et al. [5] proposed an enhanced
version of FMECA based on Dempster–Shafer theory [50,15]. Involving a team of three (N = 3) experts (i.e., technician, ship owner,
and mechanical engineer), 28 FMs were identified (see Appendix B) and prioritized, disregarding the weights of risk parameters and
experts. One of the main contributions of Certa et al. (2017a) was the collection of a comprehensive list of real interval-valued ratings
of FMs against S, O, and D. Therefore, ratings of alternatives (i.e., failure modes) against criteria (i.e., S, O, and D) (Table 5) provided
for every expert by Certa et al. (2017a) are used to validate the proposed method. Evaluated on a ten-point scale (§ Table 1), S, O, and D
have to be minimised.
The three involved experts are equally weighed; namely, zk is set equal to 1/3 for k = (1, 2, 3). The resulting interval ratings of FMs
against criteria from Equation (1) are synthetized in Table 6.
Five classes of risk are defined a priori, from the worst (i.e., Class 1 - severe) to the best (i.e., Class 5 - negligible). Reference profiles
of classes (bf) along with indifference (qj) and preference (pj) thresholds related to every criterion are listed in Table 7. However, no
veto is introduced here to avoid the non-compensatory behaviour of the method. As a consequence, discordance indices become zero
values, and the concordance indices are equal to the credibility indices.
The criteria are equally weighed in the first analysed scenario (i.e., scenario 1); namely, the criteria weights vector is (1/3, 1/3, 1/
3). Subsequently, the cutting level λ is set to 0.7, 0.8, and 0.9, respectively, and the pessimistic rule is used to assign FMs to risk
categories. By using the optimistic rule, the existence of incomparability relations would return different classification results, thus
assigning FMs to lower risk classes. However, the pessimistic assignment rule is preferred here for a more conservative perspective. The
obtained assignment results from running a Visual Basic Macro on an Intel(R) Core(TM) i7 16 GB RAM computer are presented in

Fig. 3. Binary relations.

8
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

Table 5
Ratings of alternatives on S, O, and D.
Item FM S O D

Expert Expert Expert

1 2 3 1 2 3 1 2 3

1.1 Energy shut-off [8; 9] [7; 8] 8 [5; 6] 6 [5; 6] [1; 3] [2; 4] 3

1.2 Engine shut-down [8; 9] 8 [7; 8] [5; 6] [5; 6] 5 [1; 3] 2 [2; 3]
1.3.1 Level control failure [2; 3] [2; 3] 3 [2; 3] [3; 4] 3 [6; 8] 7 [6; 7]
1.3.2.1 Pump blockage [5; 6] 5 [5; 6] [3; 4] 4 [4–5] [8; 9] [7; 8] 8
1.3.2.2 Filter choking [3; 4] [4; 5] 4 [3; 4] [3; 4] 3 [7; 8] [6; 7] 8
1.3.2.3 Filter choking [4; 5] 5 [5; 6] [3; 4] 3 [3; 4] [7; 8] 8 [8; 9]
1.3.3.1 No fuel pumping to the high-pressure [7; 8] [7; 8] 8 [6; 7] 6 [5; 6] [3; 4] 4 [4; 5]
pump
1.3.3.2 Filter choking [6; 7] 7 [7; 8] [3–4] [4; 5] 3 [4; 5] [4; 5] 5
1.3.3.3 No fuel pumping to the engine [8; 9] [7; 8] 8 [6; 7] 7 [5–6] [3; 4] [4; 5] 4
1.3.3.4 Choking [8; 9] 9 [8; 9] [7; 8] [6; 7] 7 [2; 3] 2 [1; 2]
1.4.1 Filter choking [3; 4] [3; 4] 3 [3; 4] [4; 5] 4 [8; 9] 8 [7; 9]
1.4.2 Palette damage [8; 9] 8 [7; 8] [6; 7] 7 [6; 7] [2; 4] [2; 3] 3
1.4.3 Palette damage [8; 9] [8; 9] 8 [6; 7] [5–6] 7 [2; 3] [3–4] 3
1.4.2 Component blockage [8; 9] 9 [7; 8] [6; 7] 6 [6; 7] [2; 4] 3 [2; 3]
1.4.3
1.5.1 Reduced flow into the cooling system [7; 8] [7; 9] 8 [7; 8] 7 [7; 8] [6; 7] 7 [5; 6]
1.5.2.1 Fluids mixing [7; 8] 7 [7; 8] [6; 7] [5–6] 6 [3; 4] [4; 5] 4
1.5.2.2 No oil pumping to the engine [8; 9] [7; 9] 8 [7; 8] 7 [7; 8] [6; 7] [5; 6] 6
1.5.2.3 Filter chocking [4; 5] 5 [4; 5] [3; 4] [2; 3] 4 [6; 7] 6 [5; 6]
1.5.3.1 Fluids mixing [6; 7] [5; 6] 7 [6; 7] [7; 8] 6 [3; 4] 3 [3; 4]
1.5.4.1 Empty tank (no water) [4; 5] 5 [4; 6] [2; 3] 2 [3; 4] [7; 8] [6; 7] 7
1.5.4.2 No pumping [4; 6] [5; 6] 5 [7; 8] [7; 8] 7 [6; 7] [7; 8] 7
1.5.4.3 Fluids mixing [6; 7] 6 [5; 6] [6; 7] 7 [6; 7] [3; 4] 3 [4; 5]
1.5.5.1 Fluids mixing [6; 7] [7; 8] 6 [6; 7] 6 [5; 6] [3; 4] 4 [4; 5]
1.5.5.2 No pumping [8; 9] 9 [8; 9] [7; 8] [7; 8] 7 [7; 8] [6; 8] 7
1.5.5.3 Filter choking [6; 7] [7; 8] 6 [3; 4] 3 [4; 5] [6; 7] [6; 7] 7
1.6.1 Transmission failure [8; 9] 5 [7; 8] [8; 9] [7; 8] 9 [3; 4] 4 [2; 4]
1.6.2 Roll wear [6; 7] [5; 6] 6 [7; 8] [7; 8] 8 [3; 5] 4 [4; 5]
1.6.3 Chipping [5; 7] 6 [6–7] [8; 9] 9 [7; 8] [3; 4] [3; 5] 4

Table 6
Aggregated ratings of alternatives on O, S, and D.
FM O S D

1.1 5.3333 6.0000 7.6667 8.3333 2.0000 3.3333

1.2 5.0000 5.6667 7.6667 8.3333 1.6667 2.6667

1.3.1 2.6667 3.3333 2.3333 3.0000 6.3333 7.3333
1.3.2.1 3.6667 4.3333 5.0000 5.6667 7.6667 8.3333
1.3.2.2 3.0000 3.6667 3.6667 4.3333 7.0000 7.6667
1.3.2.3 3.0000 3.6667 4.6667 5.3333 7.6667 8.3333
1.3.3.1 5.6667 6.3333 7.3333 8.0000 3.6667 4.3333
1.3.3.2 3.3333 4.0000 6.6667 7.3333 4.3333 5.0000
1.3.3.3 6.0000 6.6667 7.6667 8.3333 3.6667 4.3333
1.3.3.4 6.6667 7.3333 8.3333 9.0000 1.6667 2.3333
1.4.1 3.6667 4.3333 3.0000 3.6667 7.6667 8.6667
1.4.2 6.3333 7.0000 7.6667 8.3333 2.3333 3.3333
1.4.3 6.0000 6.6667 8.0000 8.6667 2.6667 3.3333
1.4.2–1.4.3 6.0000 6.6667 8.0000 8.6667 2.3333 3.3333
1.5.1 7.0000 7.6667 7.3333 8.3333 6.0000 6.6667
1.5.2.1 5.6667 6.3333 7.0000 7.6667 3.6667 4.3333
1.5.2.2 7.0000 7.6667 7.6667 8.6667 5.6667 6.3333
1.5.2.3 3.0000 3.6667 4.3333 5.0000 5.6667 6.3333
1.5.3.1 6.3333 7.0000 6.0000 6.6667 3.0000 3.6667
1.5.4.1 2.3333 3.0000 4.3333 5.3333 6.6667 7.3333
1.5.4.2 7.0000 7.6667 4.6667 5.6667 6.6667 7.3333
1.5.4.3 6.3333 7.0000 5.6667 6.3333 3.3333 4.0000
1.5.5.1 5.6667 6.3333 6.3333 7.0000 3.6667 4.3333
1.5.5.2 7.0000 7.6667 8.3333 9.0000 6.6667 7.6667
1.5.5.3 3.3333 4.0000 6.3333 7.0000 6.3333 7.0000
1.6.1 8.0000 8.6667 6.6667 7.3333 3.0000 4.0000
1.6.2 7.3333 8.0000 5.6667 6.3333 3.6667 4.6667
1.6.3 8.0000 8.6667 5.6667 6.6667 3.3333 4.3333

9
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

Table 7
Thresholds and reference profiles.
O S D

b1 8 8 8
b2 6 6 6
b3 4 4 4
b4 2 2 2
qj 1 1 1
pj 3 3 3
Class

Fig. 4a. Scenario 1, (wO, wS, wD) = (1/3, 1/3, 1/3).

Fig. 4a. For all cutting levels, no FM is assigned to class 1, while 1.3.3.4, 1.5.1, 1.5.2.2, 1.5.5.2, and 1.6.1 are assigned to class 2 thus
resulting the most critical FMs. Therefore, the assignment of these FMs to class 2 is stable for different values of λ. Similarly, 1.3.1,
1.3.3.2, 1.5.3.1, 1.5.4.1, 1.5.4.3, 1.5.5.1, and 1.5.5.3 are always in class 3. On the other hand, the assignment class of the other FMs
depends on the cutting level. As expected, the method tends to assign FMs to higher risk classes (i.e. worst) for increasing values of λ
because of the increasing number of incomparability relations.
A sensitivity analysis of criteria weights is also performed (Fig. 4b–4d). Comparison of the obtained results (Fig. 4a–4d) shows that
1.5.5.2 is the most critical FM because it is assigned to class 2 for all scenarios and levels of λ. Similarly, 1.5.3.1, 1.5.4.3, 1.5.5.1 and
1.5.5.3 are always in class 3. Therefore, the assignments of 1.5.5.2, 1.5.3.1, 1.5.4.3, 1.5.5.1, and 1.5.5.3 are robust, and the highest
criticality of 1.5.5.2 (the only one always in class 2 for whatever scenario and cutting level λ) is in line with the results of Certa et al.
(2017a). Instead, the assignment results of the other FMs depend on the considered scenario and/or cutting level, but they are
consistent with the input data. Specifically, high scores against a criterion lead to the assignment of the related FMs to progressively
worse classes as the weight of the criterion increases. In addition, the higher λ, the higher the risk category of FMs, as expected. For
instance, 1.3.1 ranges from class 3 to class 5 for λ = 0.7, whereas it is assigned to class 3 in all scenarios for λ = 0.8 and λ = 0.9. Given
the low aggregated rating of 1.3.1 against S, the assignment to class 5 occurs in scenario 2, where the highest weight is given to S (i.e.,
wS = 0.5). For λ = 0.7, 1.3.1 is assigned to class 4 in the third scenario, where wO = 0.5. As 1.3.1 is the only FM ranging from classes 3 to
5, it is a candidate to be the least critical, consistent with the results of Certa et al. (2017a). In scenarios 1, 3, and 4, 1.1 and 1.2 move
from class 2 to class 3 to decrease λ. Only in the second scenario, 1.1 and 1.2 are always in class 2. This is consistent with the high
aggregated ratings of these FMs against the risk parameter S. Consequently, the assignment categories of 1.1. and 1.2 are affected by
wS. In contrast, the assignment of 1.3.2.1 is strongly influenced by the weight of D, such that the higher the value of wD, the lower its
class. In addition, the class of 1.3.2.3 is consistent with the input data; namely, it is in class 2 for all values of λ in scenario 4, where the
highest weight is given to D.
The following Fig. 5a–5c visualize the percentage of FMs assigned to risk classes for every analysed scenario and cutting level λ. As

10
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

Class

Fig. 4b. Scenario 2, (wO, wS, wD) = (0.25, 0.5, 0.25).

Class

Fig. 4c. Scenario 3, (wO, wS, wD) = (0.5, 0.25, 0.25).

shown, the highest differentiation among the FMs criticality is obtained for λ = 0.7. Therefore, focusing on λ = 0.7, Fig. 6 displays the
variation of FMs assignment class in the investigated scenarios.
One of the main limitation of the case study is clearly the assumption of equally weighted experts. In fact, the application case is
mainly aimed at highlighting the applicability and validity of the designed method, and the comparison among the various scenarios is
performed to draw the criticality of the FMs. However, in real cases, the DM should assess exactly both the experts’ and criteria

11
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

Class

Fig. 4d. Scenario 4, (wO, wS, wD) = (0.25, 0.25, 0.5).

λ=0.7

Class 5 SCENARIO 4
SCENARIO 3
Class 4
SCENARIO 2
Class 3 SCENARIO 1

Class 2

Class 1

0% 20% 40% 60% 80% 100%

Fig. 5a. Percentage of FMs assigned to classes for λ = 0.7.

weights, so that the resulting FM assignment would be the only one reflecting the DM needs and knowledge of the context under
investigation. Nevertheless, the design of innovative and structured weighting methods to estimate the weights of experts more
precisely is still an open research issue. In FMECA studies, the relative importance of experts is less common compared with the risk
factor weights, and subjective methods have been mainly used so far. To reduce bias or inaccuracy, more research efforts are still
needed. As concerns risk parameters, the inclusion of just O, S and D represents a further limitation. Under the 4.0 perspective, the
ongoing digital transformation is progressively moving industries towards intelligent and predictive maintenance policies. By
continuous monitoring, current technologies more and more allow to implement proper measures to prevent sudden machine/plant
down times. This way, considerable economic savings may be obtained on maintenance costs, maximizing plants availability as well.
With this recognition, the criticality of FMs should be also based on maintenance related facets, so that FMECA studies should consider
further maintenance based criteria besides the traditional O, S and D.

12
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

λ=0.8

Class 5 SCENARIO 4
SCENARIO 3
Class 4
SCENARIO 2

Class 3 SCENARIO 1

Class 2

Class 1

0% 20% 40% 60% 80% 100%

Fig. 5b. Percentage of FMs assigned to classes for λ = 0.8.

λ=0.9

Class 5 SCENARIO 4
SCENARIO 3
Class 4
SCENARIO 2

Class 3 SCENARIO 1

Class 2

Class 1

0% 20% 40% 60% 80% 100%

Fig. 5c. Percentage of FMs assigned to classes for λ = 0.9.

5. Conclusions

An MCDM-based approach to FMECA is proposed to classify FMs into predefined and ordered risk classes, properly addressing the
uncertainty and vagueness that affect the input data. In this regard, the general assessment structure of FMECA is group-oriented, and
the FMECA group comprises experts with different levels of knowledge and experience. Their evaluations usually have some vagueness
and uncertainty, and thus they are often unlikely able to precisely evaluate FMs against the risk parameters. However, experts are more
confident in rating FMs through intervals that better represent their actual perception and knowledge. Accordingly, an interval-valued
extension of the ELECTRE TRI is designed. To the best of the authors knowledge, the classification of FMs also considering the un­
certainty of input data in a multi-group context has not addressed in the literature yet. Therefore, ratings of FMs against risk parameters
are provided by a panel of experts in interval form and then aggregated to provide a group consensus. Differently from the traditional
RPN approach, the proposed one considers both the relative importance of risk factors and the DM imprecision and uncertainty when
assigning a FM to a specific risk category. In this regard, ELECTRE TRI uses proper thresholds to approximate the gradual transition of
the DM from the indifference to the preference state during the evaluation process. Moreover, the classification results make it easier to
identify the FMs by paying attention to priority, namely the ones belonging to higher risk classes. In addition, the method allows
identifying the criterion that has the highest impact on the assignment of a FM to a specific category, thus effectively supporting the
DM when choosing actions to be taken to reduce risks. The effectiveness and feasibility of the proposed method is demonstrated by its
implementation in the case study of Certa et al. [5].
A possible future line of research may address the inclusion of additional risk factors mainly related to the maintenance aspects of
failures. In addition, a further line of development may concern the design of a novel and effective experts weighting method.

13
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

1.1 SCENARIO 1
5
1.6.3 5 1.2 SCENARIO 2
1.6.2 1.3.1
4 SCENARIO 3
1.6.1 3 4 3 3 1.3.2.1 SCENARIO 4
3 3
1.5.5.3 3 2 3 2 2
3 4 1.3.2.2
3 2 2 3
1.5.5.2 2 4 1.3.2.3
3
2 2
1.5.5.1 1 1.3.3.1
3 3

0
1.5.4.3 3 3 4 1.3.3.2
2
3 3
1.5.4.2 2 1.3.3.3
3 2 3
1.5.4.1 4 3
2 2
3 1.3.3.4
2
2 2
3 3 4
1.5.3.1 1.4.1
3 3
3 3
1.5.2.2 1.4.2
1.5.2.1 1.4.3
1.5.1 1.4.2 1.4.3

Fig. 6. FMs assignment class variation for λ = 0.7 in different scenarios.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

Appendix A. . Interval arithmetic

Let A be interval [a; a], with a and a belonging to the set of the real numbers R. Algebraic operations on real intervals can be
expressed in terms of ordinary arithmetic operations on their bounds. Therefore, being B the interval [b; b], arithmetic operations
between X and Y are performed as follows (Eqs. (A.1) and (A.4)).
[ ]
A + B = a + b; a + ̃
b (A.1)

[ ]
A − B = a − b; a − ̃
b (A.2)

[ ( ) ( )]
A ⋅ B = min a ⋅ b, a ⋅ b, a ⋅ ̃
b, a ⋅ b ; max a ⋅ b, a ⋅ b, a ⋅ ̃
b, a ⋅ b (A.3)

[ ( ) ( )]
A ÷ B = min a ÷ b, a ÷ b, a ÷ ̃
b, a ÷ b ; max a ÷ b, a ÷ b, a ÷ ̃
b, a ÷ b (A.4)

The Eqs. (A.5) and (A.6) synthetize the minimum and maximum operators.

min(A; B) = [min(a; b); min(a; b) ] (A.5)

max(A; B) = [max(a; b); max(a; b) ] (A.6)

Based on the previous equations, intervals are ordered as follows (Eqs. A.7-A.8).
A < B ⇔ max(A; B) = B (A.7)
or
A < B ⇔ min(A; B) = A (A.8)
Two real intervals are finally sorted as follows (A.9).

[a; a]⩽[b; b] ⇔ a⩽banda⩽b (A.9)

14
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

Appendix B. . FMEA analysis

Item Component FM Cause Effect

1.1 Engine starting Energy shut-off - Lack of electricity supply No engine starting
system - Oxide deposit
1.2 Engine block Engine shut-down - Valves, driveshaft, or Stop of the fishing vessel
piston damage
1.3.1 Fuel tank Level control failure Electric component failure Reduction of the fishing vessel range
1.3.2.1 Recycle pump Pump blockage - Lack of electricity supply - Damage to the filtering system
(fuel) - Bearings wear - Increase of fuel impurities
1.3.2.2 Fuel filter Filter choking - Fuel impurities - Increase of impurities contained in the
- Lack of maintenance fuel
- Low engine performance
1.3.2.3 Water filter Filter choking Lack of maintenance - High % of H2O inside the fuel
- Possible engine shut-off
1.3.3.1 Low-pressure No fuel pumping to the high- Lack of electricity supply - Low engine performance
pump pressurepump - Possible engine shut-off
1.3.3.2 Fuel filter Filter choking - Fuel impurities No fuel to the engine
- Lack of maintenance
1.3.3.3 High-pressure No fuel pumping to the engine - Rotor damage Engine shut-off
pump - Bearings wear
1.3.3.4 Electro injector Choking - Fuel impurities Low engine performance
- Electric component failure
1.4.1 Air filter Filter choking Lack of maintenance - Low fresh-air pressure
- Possible damage of heat exchanger #2
1.4.2 Turbine Palette damage - Exhausted gas impurities Low engine performance
- Lack of maintenance
1.4.3 Compressor Palette damage - Fresh air impurities Low engine performance
- Lack of maintenance
1.4.2 Turbocharger Component blockage - No driveshaft lubrication - Low engine performance
1.4.3 - Bearings wear - Increase of the engine’s smoke and noise
1.5.1 Seawater Reduced flow into the cooling system - Rotor damage - Possible damage of the cooling system
pump - Bearings wear - Increase of the engine temperature
1.5.2.1 Heat exchanger #1 Fluids mixing Perforation of the heat Increase of the oil engine temperature
(oil–seawater) exchanger
1.5.2.2 Motor oil No oil pumping to the engine Lack of transmission Reduction of the motor oil pressure
pump
1.5.2.3 Motor oil filter Filter chocking - Fluid impurities Bad engine lubrication
- Lack of maintenance
1.5.3.1 Heat exchanger #2 Fluids mixing Perforation of the heat Increase of the fresh-air temperature
(fresh air–seawater) exchanger
(intercooler)
1.5.4.1 Fresh water Empty tank (no water) - Level control failure Increase of the motor oil temperature
tank - Perforation of the tank
1.5.4.2 Fresh water No pumping - Lack of transmission - Possible damage of heat exchanger #3
pump - Bearings wear - Increase of the engine temperature
1.5.4.3 Heat exchanger #3 Fluids mixing Perforation of the heat Increase of the fresh-water temperature
(fresh exchanger
water–seawater)
1.5.5.1 Heat exchanger #4 Fluids mixing Perforation of the heat Increase of the gear-inverter oil
(inverter exchanger temperature
oil–seawater)
1.5.5.2 Gear-inverter oil No pumping Lack of transmission Low pressure of the gear-inverter oil
pump
1.5.5.3 Gear-inverter oil Filter choking - Fluid impurities Bad inverter lubrication
filter - Lack of maintenance
1.6.1 Gear inverter Transmission failure - Mechanical failures - Possible damage of the gear inverter
- Lack of lubrication - Stop of the fishing vessel
1.6.2 Shaft supports Roll wear - Lack of lubrication Propeller shaft not aligned
- Lack of cooling
1.6.3 Propeller Chipping Cavitation - Low propeller performance
- Increase of the propeller noise

References

[1] O. Cailloux, P. Meyer, V. Mousseau, Eliciting ELECTRE TRI category limits for a group of decision makers, Eur. J. Oper. Res. 223 (1) (2012) 133–140.

15
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

[2] G.F. Can, An intutionistic approach based on failure mode and effect analysis for prioritizing corrective and preventive strategies, Hum. Factors Ergonomics
Manufacturing 28 (3) (2018) 130–147.
[3] S. Carpitella, A. Certa, J. Izquierdo, C.M. La Fata, A combined multi-criteria approach to support FMECA analyses: a real-world case, Reliab. Eng. Syst. Saf. 169
(2018) 394–402.
[4] M. Catelani, L. Ciani, M. Venzi, Failure modes, mechanisms and effect analysis on temperature redundant sensor stage, Reliab. Eng. Syst. Saf. 180 (2018)
425–433.
[5] A. Certa, F. Hopps, R. Inghilleri, C.M. La Fata, A Dempster-Shafer Theory-based approach to the Failure Mode, Effects and Criticality Analysis (FMECA) under
epistemic uncertainty: application to the propulsion system of a fishing vessel, Reliab. Eng. Syst. Saf. 159 (2017) 69–79.
[6] A. Certa, M. Enea, G.M. Galante, C.M. La Fata, ELECTRE TRI-based approach to the failure modes classification on the basis of risk parameters: an alternative to
the risk priority number, Comput. Ind. Eng. 108 (2017) 100–110.
[7] K.C. Chai, C.H. Jong, K.M. Tay, C.P. Lim, A perceptual computing-based method to prioritize failure modes in failure mode and effect analysis and its application
to edible bird nest farming, Appl. Soft Comput. 49 (2016) 734–747.
[8] K.-H. Chang, Y.-C. Chang, I.-T. Tsai, Enhancing FMEA assessment by integrating grey relational analysis and the decision making trial and evaluation laboratory
approach, Eng. Fail. Anal. 31 (2013) 211–224.
[9] W.L. Chang, K.M. Tay, C.P. Lim, Clustering and visualization of failure modes using an evolving tree, Expert Syst. Appl. 42 (20) (2015) 7235–7244.
[10] G. Curcurù, G.M. Galante, C.M. La Fata, A bottom-up procedure to calculate the Top Event probability in presence of epistemic uncertainty, in: Proceedings of
the 11th International Probabilistic Safety Assessment and Management Conference and the Annual European Safety and Reliability Conference 2012, PSAM11
ESREL 2012, 6729-6735. Helsinki, Finland, 2012.
[11] S. Abu Dabous, F. Ibrahim, S. Feroz, I. Alsyouf, Integration of failure mode, effects, and criticality analysis with multi-criteria decision-making in engineering
applications: Part I - Manufacturing industry, Eng. Fail. Anal. 122 (2021) 105264, https://doi.org/10.1016/j.engfailanal.2021.105264.
[12] S.A. Dabous, F. Ibrahim, S. Feroz, I. Alsyouf, Integration of failure mode, effects, and criticality analysis with multi-criteria decision-making in engineering
applications: Part II – Non manufacturing industry, Eng. Fail. Anal. 122 (2021), 105296.
[13] C. Dağsuyu, E. Göçmen, M. Narlı, A. Kokangül, Classical and fuzzy FMEA risk analysis in a sterilization unit, Comput. Ind. Eng. 101 (2016) 286–294.
[14] S. Damart, L.C. Dias, V. Mousseau, Supporting groups in sorting decisions: Methodology and use of a multi-criteria aggregation/disaggregation DSS, Decis.
Support Syst. 43 (4) (2007) 1464–1475.
[15] A.P. Dempster, Upper and lower probabilities induced by a multivalued mapping, Ann. Math. Stat. 38 (2) (1967) 325–339.
[16] M.S.M.M. Emamat, C.M.d.M. Mota, M.R. Mehregan, M.R. Sadeghi Moghadam, P. Nemery, Using ELECTRE-TRI and FlowSort methods in a stock portfolio
selection context, Financial Innovation 8 (1) (2022), https://doi.org/10.1186/s40854-021-00318-1.
[17] J. Figueira, S. Greco, M. Ehrgott, Multiple Criteria Decision Analysis, Springer, State of the Art Surveys, 2016.
[18] J. Figueira, B. Roy, Determining the weights of criteria in the ELECTRE type methods with a revised Simos’ procedure, Eur. J. Oper. Res. 139 (2) (2002)
317–326.
[19] R. Ghani, Integration of FMECA and statistical analysis for predictive maintenance, J. Appl. Res. Technol. Eng. 2 (2021) 33–37.
[20] S.J. Ghoushchi, S. Yousefi, M. Khazaeili, An extended FMEA approach based on the Z-MOORA and fuzzy BWM for prioritization of failures, Appl. Soft Comput.
81 (2018), 105505.
[21] A. Gugaliya, S. Boral, V.N.A. Naikan, A hybrid decision making framework for modified failure mode effects and criticality analysis: A case study on process
plant induction motors, Int. J. Quality Reliability Manage. 36 (8) (2019) 1266–1283.
[22] M. Gul, M.F. Ak, A modified failure modes and effects analysis using interval-valued spherical fuzzy extension of TOPSIS method: case study in a marble
manufacturing facility, Soft. Comput. 25 (8) (2021) 6157–6178.
[23] M. Gul, M. Yucesan, E. Celik, A manufacturing failure mode and effect analysis based on fuzzy and probabilistic risk analysis, Appl. Soft Comput. 96 (2020).
[24] IEC 60812: Analysis techniques for system reliability – Procedure for failure mode and effects analysis (FMEA), International Electrotechnical Commission,
Switzerland; 2006.
[25] L. Jaulin, M. Kieffer, O. Didrit, E. Walter, Applied interval analysis – With examples in parameter and state estimation, robust control and robotics, Springer-
Verlag, London, 2001.
[26] A. Klimenko, Intransitivity in Theory and in the Real World, Entropy 17 (12) (2015) 4364–4412.
[27] A.C. Kutlu, M. Ekmekçioğlu, Fuzzy failure modes and effects analysis by using fuzzy TOPSIS-based fuzzy AHP, Expert Syst. Appl. 39 (1) (2012) 61–67.
[28] C.M. La Fata, A. Giallanza, R. Micale, G. La Scalia, Ranking of occupational health and safety risks by a multi-criteria perspective: Inclusion of human factors and
application of VIKOR, Saf. Sci. 138 (2021) 105234, https://doi.org/10.1016/j.ssci.2021.105234.
[29] C.M. La Fata, A. Giallanza, R. Micale, G. La Scalia, A Structured Methodology for the Safety Key Performance Indicator Prioritization: A Case Study, in:
Proceedings of 5th International Conference on System Reliability and Safety, ICSRS 2021, Palermo, Italy, pp. 143-147, ISBN: 978-6654-0048-0, 2022.
[30] H.-C. Liu, X.-Q. Chen, C.-Y. Duan, Y.-M. Wang, Failure mode and effect analysis using multi-criteria decision making methods: A systematic literature review,
Comput. Ind. Eng. 135 (2019) 881–897.
[31] H.C. Liu, P. Li, J.X. You, Y.Z. Chen, A novel approach for FMEA: Combination of interval 2-tuple linguistic variables and grey relational analysis, Qual. Reliab.
Eng. Int. 31 (5) (2015) 761–772.
[32] H.-C. Liu, J.-X. You, M.-M. Shan, L.-N. Shao, Failure mode and effects analysis using intuitionistic fuzzy hybrid TOPSIS approach, Soft. Comput. 19 (4) (2015)
1085–1098.
[33] H.-C. Liu, Z. Li, W. Song, Q. Su, Failure mode and effect analysis using cloud model theory and PROMETHEE method, IEEE Trans. Reliab. 66 (4) (2017)
1058–1072.
[34] H.-C. Liu, J.-X. You, X.-Y. You, M.-M. Shan, A novel approach for failure mode and effects analysis using combination weighting and fuzzy VIKOR method, Appl.
Soft Comput. 28 (2015) 579–588.
[35] H.-W. Lo, J.J.H. Liou, A novel multiple-criteria decision-making-based FMEA model for risk assessment, Appl. Soft Comput. 73 (2018) 684–696.
[36] H.-W. Lo, J.J.H. Liou, C.-N. Huang, Y.-C. Chuang, A novel failure mode and effect analysis model for machine tool risk analysis, Reliab. Eng. Syst. Saf. 183
(2019) 173–183.
[37] H.-W. Lo, W. Shiue, J.J.H. Liou, G.-H. Tzeng, A hybrid MCDM-based FMEA model for identification of critical failure modes in manufacturing, Soft. Comput. 24
(20) (2020) 15733–15745.
[38] F. Lolli, A. Ishizaka, R. Gamberini, B. Rimini, M. Messori, FlowSort-GDSS - A novel group multi-criteria decision support system for sorting problems with
application to the FMEA, Expert Syst. Appl. 42 (17–18) (2015) 6342–6349.
[39] M.M. Merad, T. Verdel, B. Roy, S. Kouniali, Use of multi-criteria decision-aids for risk zoning and management of large area subjected to mining-induced
hazards, Tunn. Undergr. Space Technol. 19 (2) (2004) 125–138.
[40] R. Micale, A. Giallanza, G. Li Puma, G. La Scalia, A proposal for sorting projects portfolio through the ELECTRE TRI method focused on the European strategy,
J. Modern Project Manage. 7 (3) (2019) 1–26.
[41] R. Micale, C.M. La Fata, G. La Scalia, A combined interval-valued ELECTRE TRI and TOPSIS approach for solving the storage location assignment problem,
Comput. Ind. Eng. 135 (2019) 199–210.
[42] P.P. Miglietta, R. Micale, R. Sciortino, T. Caruso, A. Giallanza, G. La Scalia, The sustainability of olive orchard planting management for different harvesting
techniques: An integrated methodology, J. Cleaner Prod. 238 (2019) 117989, https://doi.org/10.1016/j.jclepro.2019.117989.
[43] J.K. Mohanty, P. Dash, P. Pradhan, FMECA analysis and condition monitoring of critical equipment in super thermal power plant, Int. J. Syst. Assurance Eng.
Manage. 11 (2020) 583–599.
[44] V. Mousseau, R. Slowinski, P. Zielniewicz, A user-oriented implementation of the ELECTRE-TRI method integrating preference elicitation support, Comput.
Oper. Res. 27 (7-8) (2000) 757–777.
[45] I. Mzougui, Z.E. Felsoufi, A modified method to improve failure analysis, Int. J. Syst. Assurance Eng. Manage. 12 (2) (2021) 231–244.

16
C.M. La Fata et al. Engineering Failure Analysis 135 (2022) 106163

[46] B. Roy, Classement et choix en presence de points de vue multiples (la method ELECTRE), Revue Informatique et Recherche Operationnelle 8 (1968) 57–75.
[47] B. Roy, Multicriteria for decision aiding, Kluwer Academic Publishers, London, 1996.
[48] B. Roy, D. Bouyssou, Aide multicritère à la decision: Méthodes et Cas, Economica, Paris, 1993.
[49] M.K. Sayadi, M. Heydari, K. Shahanaghi, Extension of VIKOR method for decision making problem with interval numbers, Appl. Math. Model. 33 (5) (2009)
2257–2262.
[50] G. Shafer (Ed.), A Mathematical Theory of Evidence, Princeton University Press, 1976.
[51] M. Takougang, S. Aimé, M. Pirlo, S. Yonkeu, B. Some, Assessing the response to land degradation risk: The case of the Loulouka catchment basin in Burkina Faso,
in: Evaluation and decision models with multiple criteria: Case studies, Springer-Verlag, Berlin Heidelberg, 2015, pp. 341–399.
[52] K.M. Tay, C.H. Jong, C.P. Lim, A clustering-based failure mode and effect analysis model and its application to the edible bird nest industry, Neural Comput.
Appl. 26 (3) (2015) 551–560.
[53] A.N. The, V. Mousseau, Using assignment examples to infer category limits for the ELECTRE TRI method, J. Multi-Criteria Decision Anal. 11 (1) (2002) 29–43.
[54] T. Tinga, W. Tiddens, F. Amoiralis, M. Politis. Predictive maintenance of maritime systems: models and challenges, in: European safety and reliability conference
ESREL 2017, 2017.
[55] A. Tversky, Intransitivity of preferences, Psychol. Rev. 76 (1) (1969) 31–48.
[56] Z. Wu, W. Liu, W. Nie, Literature review and prospect of the development and application of FMEA in manufacturing industry, Int. J. Adv. Manuf. Technol. 112
(5-6) (2021) 1409–1436.
[57] W. Yu, Aide multicritere a la decision dans le cadre de la problematique dutri: Methodes et applications. Ph.D, Universite Paris Dauphine, Thesis, LAMSADE,
1992.
[58] F. Zammori, R. Gabbrielli, ANP/RPN: A multi criteria evaluation of the risk priority number, Qual. Reliab. Eng. Int. 28 (2011) 85–104.
[59] P. Zandi, M. Rahmani, M. Khanian, A. Mosavi, Agricultural Risk Management Using Fuzzy TOPSIS Analytical Hierarchy Process (AHP) and Failure Mode and
Effects Analysis (FMEA), Agriculture 10 (11) (2020) 504.
[60] J. Zhu, B. Shuai, G. Li, K.-S. Chin, R. Wang, Failure mode and effect analysis using regret theory and PROMETHEE under linguistic neutrosophic context, J. Loss
Prev. Process Ind. 64 (2020) 104048, https://doi.org/10.1016/j.jlp.2020.104048.

17