Materials and Design 30 (2009) 4043–4053

Contents lists available at ScienceDirect

Materials and Design

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

Selection of materials using compromise ranking and outranking methods

Prasenjit Chatterjee a, Vijay Manikrao Athawale b, Shankar Chakraborty a,*
a
Department of Production Engineering, Jadavpur University, Kolkata 700 032, West Bengal, India
b
Department of Mechanical Engineering, Government Polytechnic, Amravati 444 603, India

a r t i c l e i n f o a b s t r a c t

Article history: Selection of proper materials for different components is one of the most challenging tasks in the design
Received 23 February 2009 and development of products for diverse engineering applications. Materials play a crucial and important
Accepted 11 May 2009 role during the entire design and manufacturing process. Wrong selection of material often leads to huge
Available online 20 May 2009
cost involvement and ultimately drives towards premature component or product failure. So the design-
ers need to identify and select proper materials with specific functionalities in order to obtain the desired
Keywords: output with minimum cost involvement and specific applicability. This paper attempts to solve the mate-
Material selection
rials selection problem using two most potential multi-criteria decision-making (MCDM) approaches and
Multi-criteria decision-making
VIKOR
compares their relative performance for a given material selection application. The first MCDM approach
ELECTRE is ‘Vlse Kriterijumska Optimizacija Kompromisno Resenje’ (VIKOR), a compromise ranking method and
Concordance matrix the other one is ‘ELimination and Et Choice Translating REality’ (ELECTRE), an outranking method. These
Discordance matrix two methods are used to rank the alternative materials, for which several requirements are considered
simultaneously. Two examples are cited in order to demonstrate and validate the effectiveness and flex-
ibility of these two MCDM approaches. In each example, a list of all the possible choices from the best to
the worst suitable materials is obtained taking into account different material selection criteria. The rank-
ings of the selected materials almost corroborate with those as obtained by the past researchers.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction the component and broadly outline the main material characteris-
tics and processing requirements [1–4]. The main design task lies
It is well established that materials play an important role in in comparing the properties of a finite set of materials and select-
engineering design. For design and manufacturing of different ing the best one out of this set. While choosing a new material for
types of mechanical components, knowledge of material proper- an engineering application, the decision maker usually applies trial
ties, cost, design considerations and their influences are manda- and error methods or employs his/her knowledge from previous
tory. A large number of available alternative materials, having experimentations.
complex relationships with various selection parameters (criteria), Various approaches have already been proposed by the past
make the material selection process a challenging task. Selection of researchers to solve the problem of proper material selection.
the most suitable material involves the study of a large number of Wang and Chang [5] proposed a fuzzy multiple criteria decision-
factors, like mechanical, electrical and physical properties, and cost making approach to help selecting the best suited tool steel mate-
considerations of the materials. For mechanical design, the rial for a specific manufacturing application, such as die design, jig
mechanical properties of the materials are given the top priorities. and fixture design. Chen [6] developed a method to solve the tool
The most important mechanical properties usually encountered in steel materials selection problem under fuzzy environment, where
the material selection process are strength, stiffness, toughness, the importance weights of different criteria and the ratings of var-
hardness, density and creep resistance. But there must be some ious alternatives under different criteria were assessed in linguistic
methodological approach so as to assist the decision maker to se- terms using fuzzy numbers. Tretheway et al. [7] developed a
lect the most appropriate material for a given mechanical engi- knowledge structure for materials selection based on material per-
neering application. Perhaps the first step in the material formance and failure analysis. Jee and Kang [8] utilized the tech-
selection process is to specify the performance requirements of nique for order preference by similarity to ideal solution (TOPSIS)
method to rank the alternative materials, while simultaneously
considering several requirements and subsequently employed the
entropy method to evaluate the weight factor for each material
* Corresponding author. Tel.: +91 033 2414 6153 (O)/2548 2655 (R); fax: +91 033
2414 6153. property. Sapuan [9] developed a knowledge-based system (KBS)
E-mail address: [email protected] (S. Chakraborty). in the selection of polymeric-based composite materials. Qian

0261-3069/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.matdes.2009.05.016
4044 P. Chatterjee et al. / Materials and Design 30 (2009) 4043–4053

and Zhao [10] applied the concept of ‘performance index’ to select 2. Compromise ranking method
the best suited material for a given micro-electromechanical sys-
tem (MEMS) design. The selection process was based on matching The VIKOR (the Serbian name is ‘Vlse Kriterijumska Optimizacija
the performance characteristics to the requirements. A series of Kompromisno Resenje’ which means multi-criteria optimiza-
performance indices was deployed to allow a wide range compar- tion (MCO) and compromise solution) method was mainly
ison of materials for several typical sensing and actuating struc- established by Zeleny [20] and later advocated by Opricovic and
tures, and a rapid identification of the alternative materials for a Tzeng [21–24]. This method is developed to solve MCDM problems
given task. Rao [11] presented a methodology for material selec- with conflicting and non-commensurable (attributes with different
tion for a given engineering component using graph theory and units) criteria, assuming that compromise can be acceptable for
matrix approach. A digraph was developed taking into account sev- conflict resolution, when the decision maker wants a solution that
eral material selection criteria and their relative importance for the is the closest to the ideal solution and the alternatives can be eval-
application considered. Shanian and Savadogo [12] adopted an uated according to all the established criteria. It focuses on ranking
‘ELimination and Et Choice Translating REality’ (ELECTRE) based and selecting the best alternative from a set of alternatives with
approach in material selection. While constructing a material conflicting criteria, and on proposing compromise solution (one
selection decision matrix and criteria sensitivity analysis, the ELEC- or more). The compromise solution is a feasible solution, which
TRE method was applied to obtain a more precise material selec- is the closest to the ideal solution and farthest from the nega-
tion for a particular application, including logical ranking of the tive-ideal solution, and a compromise means an agreement estab-
considered materials. Manshadi et al. [13] proposed a novel lished by mutual concessions made between the alternatives. The
numerical technique for materials selection, which was based on following multiple attribute merit for compromise ranking is
the weighting factor approach while combining the non-linear nor- developed from the Lp-metric used in the compromise program-
malization with a modified digital logic method. Chan and Tong ming method [15,25].
[14] presented an integrated methodology of constructing an order
pair of materials and end-of-life product strategy for material ( )1=p
XM
selection using the grey relational analysis approach. Rao [15] pro- Lp;i ¼ ðwj ½ðmij Þmax  mij =½ðmij Þmax  ðmij Þmin Þp
posed an improved compromise ranking method-based logical j¼1

procedure for material selection for a given engineering applica- 1  p  1; i ¼ 1; 2; . . . ; N ð1Þ

tion. Sharif Ullah and Harib [16] presented an intelligent method
to deal with the material selection problems wherein the design where M is the number of criteria and N is the number of alterna-
configurations, working conditions and design-relevant informa- tives. The mij values (for i = 1,2,. . ., N; j = 1,2,. . .,M) indicate the val-
tion were not precisely known. Shanian and Savadogo [17] ues of criteria for different alternatives. In the VIKOR method, L1,i
compared some of the most widely used multi-criteria decision- and L1,i are used to formulate the ranking measure.
making methods for solving a material selection problem of highly The procedural steps for the VIKOR method are enlisted as
sensitive components involving conflicting and multiple design follows:
objectives. Zhou et al. [18] proposed an integration of artificial neu-
ral networks with genetic algorithms to optimize the multiple  Step 1: Identify the major material selection criteria for a given
objectives of material selection to develop sustainable products. engineering application and short-list the materials on the basis
Khabbaz et al. [19] introduced a fuzzy logic approach to deal with of the identified criteria satisfying the requirements. A quantita-
the qualitative properties of materials and the corresponding fuzzy tive or qualitative value is assigned to each identified criterion to
space for solving different material selection decision-making construct the related decision matrix.
problems.  Step 2:
Although a good amount of research work has already been (a) After short-listing the materials and development of the
carried out by the past researchers on materials selection, as decision matrix, determine the best, (mij)max and the worst,
exhibited in the comprehensive list of products along with their (mij)min values of all the criteria.
associated material selection criteria in Table 1, there is still a (b) The weights or relative importance of the considered criteria
need for a simple as well as systematic approach/mathematical are determined using analytic hierarchy process (AHP) or
tool to guide the decision maker while taking a proper material any other method.
selection decision. Taking decision in the presence of multiple (c) Calculate the values of Ei and Fi.
conflicting criteria is known as multi-criteria decision-making
(MCDM) process. Depending on the type of alternatives, the
X
M
MCDM problems can be divided into continuous or discrete type. Ei ¼ L1;i ¼ wj ½ðmij Þmax  mij =½ðmij Þmax  ðmij Þmin  ð2Þ
A typical MCDM problem consists of three main components, i.e. j¼1
(a) alternatives, (b) criteria, and (c) relative importance (weight)
 
of each criterion. All the elements of a MCDM problem must be F i ¼ L1;i ¼ Maxm of wj ½ðmij Þmax  mij =½ðmij Þmax  ðmij Þmin 
normalized to the same units so that all the possible criteria can j ¼ 1; 2; . . . M ð3Þ
be considered in the decision-making process. In this paper, an
attempt is made to discover the potentiality and applicability
of two almost new MCDM approaches while selecting the most Eq. (2) is applicable to beneficial criteria (whose higher val-
suitable material for a given engineering application. The first ues are desirable for a given application). For non-beneficial
MCDM approach adopted is VIKOR (a compromise ranking meth- criteria (whose lower values are preferable for a given appli-
od) and the other one is ELECTRE (an outranking method). Two cation), [(mij)maxmij] in Eq. (2) is to be replaced by
real-life examples are cited to demonstrate and compare the per- [mij(mij)min]. Hence, for non-beneficial criteria, Eq. (2) can
formance of both these MCDM approaches. The first example be rewritten as:
deals with the selection of the most appropriate material for de-
sign of a flywheel [8], whereas, the second example considers the X
M

choice of the best suited material for design of a sailing-boat Ei ¼ L1;i ¼ wj ½ðmij Þ  ðmij Þmin =½ðmij Þmax  ðmij Þmin  ð4Þ
j¼1
mast [19].
 P. Chatterjee et al. / Materials and Design 30 (2009) 4043–4053 4045

Table 1
List of different material selection criteria as considered by the past researchers.

Author(s) Product/component Material selection criteria

Ljungberg [4] Sustainable products Service temperature, volume resistivity, Young’s modulus, total cost per unit, impact
load
Wang and Chang [5] Tool Non-deforming properties, safety in hardening, toughness, resistance to softening
effect of steel, wear resistance, machinability, material cost
Chen [6] Tool Non-deforming properties, safety in hardening, toughness, resistance to softening
effect of steel, wear resistance, machinability, material cost
Tretheway et al. [7] A coating material Hardness, toughness, Young’s modulus, shear modulus, maximum service
temperature, wear resistance, chemical resistance, dimensions, shape, porosity,
surface roughness, bond strength, biofouling resistance
Jee and Kang [8] Flywheel Fatigue limit, fracture toughness, density, price per unit mass, fragmentability
Sapuan [9] Automotive pedal box system Mechanical, physical and chemical properties, and economic and manufacturing
considerations
Qian and Zhao [10] Micromechanical element Density, Young’s modulus, fracture strength
Shape-changing microactuator Maximum actuation strain, maximum actuation stress, stroke work coefficient,
Young’s modulus, density
Rao [11] A cryogenic storage tank for liquid nitrogen Toughness index, yield strength, Young’s modulus, density, thermal expansion,
thermal conductivity, specific heat
A product to be operated in a high-temperature oxygen-rich Hardness, machinability, cost, corrosion resistance
environment
Shanian and A mass-produced non-heat-treatable cylindrical cover Density, compressive strength, ultimate tensile stress, spring back index, bend force
Savadogo [12] index, static load index, hardness, yield stress, elastic modulus, thermal diffusivity,
thermal conductivity, thickness, cost of base material
Manshadi et al. [13] A cryogenic storage tank for liquid nitrogen Toughness, yield strength, Young’s modulus, density, thermal expansion coefficient,
thermal conductivity, specific heat
A spar for an aircraft wing Price, tensile strength, Young’s modulus, density, compressive strength, creep
resistance
Chan and Tong [14] Dustbin of a vacuum cleaner Density, water absorption, hardness, tensile strength, elongation at break, tensile
modulus, coefficient of thermal expansion
Rao [15] A metallic bipolar plate for polymer electrolyte fuel cell Elastic modulus, density, tensile strength, thermal stress, coefficient of thermal
expansion, thermal conductivity, thermal diffusivity, fracture toughness, resistivity,
cost, corrosion rate, recycle fraction, hydrogen permeability
A product to be operated in a high-temperature Tensile strength, Young’s modulus, density, corrosion resistance
environment
Sharif Ullah and Harib A structural component for a robotic mechanism Cost, stiffness, strength, weight, impact toughness, impact on the environment,
[16] thermal expansion
Shanian and A mass-produced non-heat-treatable cylindrical cover Density, compressive strength, ultimate tensile stress, spring back index, bend force
Savadogo [17] index, static load index, hardness, yield stress, elastic modulus, thermal diffusivity,
thermal conductivity, thickness, cost of base material
Zhou et al. [18] A ‘green’ drink container Young’s modulus, shear modulus, Poisson’s ratio, elastic limit, tensile strength,
compressive strength, hardness, fatigue limit, density, price of raw material, price of
recycle, EI material production, EI disposal (incineration), EI disposal (landfill), EI
recycling, energetic cost, recyclable fraction
Khabbaz et al. [19] A cryogenic storage tank for liquid nitrogen Toughness index, yield strength, Young’s modulus, density, thermal expansion,
thermal conductivity, specific heat
A spar for an aircraft wing Price, tensile strength, Young’s modulus, density, compressive strength, creep
resistance
A sailing-boat mast Specific strength, specific modulus, corrosion resistance, cost

(d) Calculate Pi value. obtained by ranking with Pi measure. The best alternative is
the one having the minimum Pi value.
Pi ¼ mððEi  Ei- min Þ=ðEimax  Ei- min ÞÞ
þ ð1  mÞððF i  F i- min Þ=ðF i- max  F i- min ÞÞ ð5Þ The VIKOR method is an effective MCDM tool, specifically appli-
cable to those situations when the decision maker is not able, or
where Ei-max and Ei-min are the maximum and minimum val- does not know to express his/her preference at the beginning of
ues of Ei respectively, and Fi-max and Fi-min are the maximum system design. The computational procedure of the VIKOR method
and minimum values of Fi respectively. v is introduced as is quite simple and it offers a systematic and logical approach to ar-
weight of the strategy of ‘the majority of attributes’ (or ‘the rive at the best decision. The obtained compromise solution can be
maximum group utility’). The value of v lies in the range of accepted by the decision maker because it provides a maximum
0–1. Normally, the value of v is taken as 0.5. The compromise group utility of the ‘majority’ and a minimum individual regret
can be selected with ‘voting by majority’ (v > 0.5), with of the ‘opponent’. The compromise solutions can be the base for
‘consensus’ (v = 0.5) or with ‘veto’ (v < 0.5). negotiations, involving the decision maker’s preference on criteria
weights [15]. The VIKOR results depend on the ideal solution,
(e) Arrange the alternatives in the ascending order, according to which stands only for the given set of alternatives. Inclusion (or
the values of Pi. Compromise ranking list for a given v can be exclusion) of an alternative can affect the VIKOR ranking of the
4046 P. Chatterjee et al. / Materials and Design 30 (2009) 4043–4053

new set of alternatives. In this method, the ranking score of each  Step 4: If the subset has a single element or is small enough to
alternative is derived from an aggregation of all the considered cri- apply value judgment, select the final decision. Otherwise, steps
teria, the weights of the criteria and a balance between total and 2–4 are repeated until a single element or small subset exists.
individual satisfaction. As the VIKOR method employs linear nor-  Step 5: If a full ranking of the alternatives is required, apply an
malization procedure, the normalized values are not dependent extension of the ELECTRE, i.e. ELECTRE II method. Calculate
on the evaluation units of the selection criteria. another two indices as follows [29]:
X
n X
n
3. Outranking method Pure concordance index ðC j Þ ¼ cðj; kÞ  cðk; jÞ ðj – kÞ
k¼1 j¼1

The ELECTRE (ELimination and Et Choice Translating REality) ð9Þ

method, developed by Roy and Vincke [26], is based on multi-attri- X
n X
n
bute utility theory (MAUT) with the intention to improve efficiency Pure discordance index ðDj Þ ¼ dðj; kÞ  dðk; jÞ ðj – kÞ
without affecting the outcome while considering less information k¼1 j¼1

[27]. It is a procedure that sequentially reduces the number of alter- ð10Þ

natives the decision maker is faced within a set of non-dominated
alternatives. The concept of an outranking relation S is introduced
(a) Once these two indices are estimated, obtain two rankings
as a binary relation defined on the set of alternatives A. Given the
on the basis of these indices.
alternatives Aj and Ak, Aj outranks Ak or AjSAk, if given all that is
(b) Determine an average ranking from the two rankings as
known about the two alternatives, there are enough arguments
obtained in step 5(a).
to decide that Aj is at least as good as Ak. The goal of this outranking
(c) Select that alternative which has the best average rank.
method is to find all alternatives that dominate other alternatives
while they cannot be dominated by any other alternative. To find
By interactively changing the threshold values in the ELECTRE
the best alternative, the ELECTRE method also requires the
method, the size of the subset consisting of the promising pair of
knowledge of the weights of all the criteria. Each criterion Ci2C is
alternatives can be changed. ELECTRE I method is used to develop
assigned a subjective weight, wi (the sum of the weights of all
a partial ranking and choose a set of the promising alternatives.
criteria equals to one), and every pair of the alternatives Aj and
ELECTRE II is used for ranking all the alternatives. When an outranking
Ak is assigned a concordance index, c(j,k) given as below:
X creditability between two alternatives is to be measured, ELECTRE
cðj; kÞ ¼ wi ; j; k ¼ 1; 2; . . . ; n; j – k ð6Þ III is adopted where an outranking degree can be established.
g i ðjÞPg i ðkÞ While solving material selection problems, sometimes non-
compensatory aggregation procedures are required where com-
where gi(j) and gi(k) are the normalized measures of performance of
pensation of the loss on a given material selection criterion by a
jth and kth alternative, respectively, with respect to ith criterion in
gain in another one cannot be accepted by the decision maker
the decision matrix. Thus, for an ordered pair of alternatives (Aj,Ak),
[12]. This type of problem can be well tackled using the ELECTRE
the concordance index c(j,k) is the sum of all the weights for those
methods which are proved to be quite fast and flexible based on
criteria where the performance score of Aj is at least as that of Ak.
simple logic having the strength to solve problems in the presence
Clearly, the concordance index lies between 0 and 1. A discordance
of incomparability. The ELECTRE methods are also suitable for link-
index, d(j,k) is also calculated as given below:
ing with computer databases dealing with multi-criteria decision-
dðj; kÞ ¼ 0; if g i ðjÞ P g i ðkÞ; i ¼ 1; 2; . . . ; m making (MCDM) problems.
max ðg i ðkÞ  g i ðjÞÞ From the comparison of results from other compensatory and
g i ðkÞ>g i ðjÞ
¼ otherwise; j; k ¼ 1; 2; . . . ; n; j – k non-compensatory models, it is observed that the ELECTRE methods
max ðjg i ðkÞ  g i ðjÞjÞ seem to show reasonable sensitivity [28]. The criteria weights in
i¼1;...;m
these methods are treated as the ‘coefficients of importance’ and
ð7Þ
moreover, a significantly weak criterion value of an alternative can-
Once these two indices are estimated, an outranking relation S can not directly be compensated for by other good criteria values. On the
be defined as: other hand, the solution mechanism in the ELECTRE methods is not
^ as extreme as the purely non-compensatory methods. The pure con-
Aj SAk if and only if cðj; kÞ P ^c and dðj; kÞ 6 d ð8Þ cordance and discordance indices in the ELECTRE II method incorpo-
^ are the threshold values as set by the decision maker.
where ^c and d rate two extreme opposite relationships, i.e. strong relationship and
If the threshold values are high, it will be more difficult to pass the weak relationship, whereby strong-ranking and weak-ranking are
^ ¼ 0:3 [28]). For an outranking relation
tests (normally, ^c ¼ 0:7 and d deduced to obtain the final ranking. The ELECTRE II evaluation
to be judged as true, both the concordance and discordance indices method is widely considered as an effective and efficient decision
should not violate their corresponding threshold values. When aid with a broad range of applications covering diverse fields of pol-
these two tests are performed for all the pairs of alternatives, the icy-making. As the criteria weights used in the ELECTRE methods
preferred alternatives are those which outrank more than being are usually obtained beforehand, the calculation method used to
outranked. The steps for implementing the ELECTRE method are de- determine these weights exerts great impact upon the final ranking.
scribed as below [27]: Therefore, particular attention is to be placed on the weighting
methodology while using the ELECTRE methods. These methods
 Step 1: Obtain the normalized values of all the criteria. also serve as one alternative for approaching complex choice prob-
 Step 2: Construct the outranking relations by following the con- lems with multiple criteria and multiple participants.
cordance and discordance definitions, and develop a graph rep-
resenting the dominance relations among the alternatives. In 4. Numerical illustrations
this graph, if alternative Aj outranks alternative Ak, then a direc-
ted arc exists from Aj to Ak. In order to demonstrate and validate the applications of the two
 Step 3: Obtain a minimum dominating subset by using the min- above-mentioned MCDM approaches for solving material selection
imum concordance and maximum discordance indices. problems, the following two real-life examples are cited.
 P. Chatterjee et al. / Materials and Design 30 (2009) 4043–4053 4047

4.1. Example 1 Table 3

Ei, Fi and Pi values for example 1.

The first example deals with the selection of the most appropri- Materials Ei Fi Pi Rank
ate material for design of a flywheel [8] which is a device to store 300M 0.7658 0.3644 0.9515 9
kinetic energy as used in automobiles, urban subway trains, mass 2024-T3 0.7431 0.4000 0.9844 10
transit buses, wind-power generators, etc. Despite its many advan- 7050-T73651 0.7339 0.3799 0.9507 8
tages, relative poor energy storage and a chance of catastrophic Ti–6Al–4V 0.5599 0.3582 0.7356 6
E glass–epoxy FRP 0.6698 0.3856 0.9138 7
failure hinder its applications. The main requirements in a flywheel S glass–epoxy FRP 0.4280 0.3186 0.6538 5
design are to store the maximum amount of kinetic energy per unit Carbon–epoxy FRP 0.3221 0.1431 0.3408 2
mass and to ensure against premature failure due to fatigue or brit- Kevlar 29–epoxy FRP 0.3690 0.2636 0.5377 4
tle fracture. If the nature of failure is fatigue, the performance in- Kevlar 49–epoxy FRP 0.0479 0.0333 0 1
Boron–epoxy FRP 0.4804 0.2000 0.5287 3
dex will be rlimit/q (where, rlimit is the fatigue limit of the
material and q is the material density). This signifies that the high-
er the value of rlimit/q, the lower the weight of the material for a
given fatigue strength and consequently, the kinetic energy per
unit mass of the flywheel will be higher. For failure due to brittle
Table 4
fracture, fracture toughness (KIC) of the material will be the perfor- Ranking of materials for example 1 for different values of v.
mance measure. If the flywheel breaks into small pieces at final
failure, the hazard will be much reduced. Thus, the fragmentability Pi (v = 0.1) Materials Pi (v = 0.9) Materials

of the flywheel material is an important property from the safety 0.9128 300M (7) 0.9903 300M (10)
point of view. Hence, four criteria are considered, i.e. fatigue limit 0.9969 2024-T3 (10) 0.9720 2024-T3 (9)
0.9464 7050-T73651 (8) 0.9549 7050-T73651 (8)
(rlimit/q), fracture toughness (KIC/q), price per unit mass and frag-
0.8559 Ti–6Al–4V (6) 0.6152 Ti–6Al–4V (6)
mentability. Among these four criteria, fatigue limit, fracture 0.9514 E glass–epoxy FRP (9) 0.8761 E glass–epoxy FRP (7)
toughness and fragmentability are the beneficial attributes where 0.7531 S glass–epoxy FRP (5) 0.5544 S glass–epoxy FRP (4)
higher values are desirable, and price/mass is a non-beneficial 0.3077 Carbon–epoxy FRP (2) 0.3738 Carbon–epoxy FRP (2)
0.6100 Kevlar 29–epoxy FRP (4) 0.4654 Kevlar 29–epoxy FRP (3)
attribute where smaller value is always preferable. Ten materials
0 Kevlar 49–epoxy FRP (1) 0 Kevlar 49–epoxy FRP (1)
comprising four metals and six unidirectional fiber-reinforced- 0.4693 Boron–epoxy FRP (3) 0.5878 Boron–epoxy FRP (5)
epoxy composites are considered as the alternatives [8]. Thus,
the MCDM problem consists of 10 alternative materials and 4
material selection criteria, as shown in Table 2.

4.1.1. VIKOR method While calculating the Pi values, the value of v is usually taken as
This material selection problem for flywheel design is first 0.5 [15], but actually its value lies between 0 and 1. Table 4 shows
solved using the VIKOR method. At first, the best and the worst val- the comprise rankings of the alternative materials for flywheel de-
ues of all the criteria are identified. Jee and Kang [8] employed the sign for two extreme values of v = 0.1 and v = 0.9. In both these cases,
entropy method to determine the weights of the considered crite- the best choice of material (Kevlar 49–epoxy FRP) does not change,
ria. They also took into account four cases of subjective weights for although the ranking of the alternative materials changes slightly.
the criteria. The first case, where w1 = 0.4, w2 = 0.3, w3 = 0.2 and It is a customary practice to take the value of v as 0.5 in the
w4 = 0.1, is used here for the VIKOR method-based analysis. Now, VIKOR method-based analyses, but the results of Table 4 exhibit
the values of Ei and Fi are calculated using Eqs. (2) or (4) and (3) that the choice of the best suited material is not affected by the
respectively, as given in Table 3. Table 3 also exhibits the values varying values of v, although the ranking of the other materials
of Pi for v = 0.5 and the compromise ranking list of the considered may change. This is due to the uniqueness of the mathematical
alternative materials. The candidate materials are arranged in treatment as involved in the VIKOR method.
ascending order, according to the values of Pi. The best choice of
material for flywheel design is material 9 (Kevlar 49–epoxy FRP). 4.1.2. ELECTRE method
Carbon–epoxy FRP is the second choice and the last choice is mate- Now, the same problem of selecting the most appropriate mate-
rial 2 (2024-T3). Jee and Kang [8] obtained a ranking of the alterna- rial for flywheel design is solved using the ELECTRE method. At
tive materials as 5-9-7-6-8-3-4-2-1-10 (for case 1), whereas, using first, the original decision matrix, as shown in Table 2, is normal-
the VIKOR method, the compromise ranking of materials is ized [27] and is given in Table 5.
9-10-8-6-7-5-2-4-1-3. It is observed that in the VIKOR method, Then using Eq. (6), the concordance index values are calculated
the best choice of material for flywheel design remains the same. as follows:

Table 2
Quantitative data for example 1 [8].

Sl. no. Materials rlimit/q KIC/q Price/mass Fragmentability

1 300M 100 8.6125 4200 Poor (3)
2 2024-T3 49.6454 13.4752 2100 Poor (3)
3 7050-T73651 78.0142 12.5532 2100 Poor (3)
4 Ti–6Al–4V 108.8795 26.0042 10,500 Poor (3)
5 E glass–epoxy FRP 70 10 2735 Excellent (9)
6 S glass–epoxy FRP 165 25 4095 Excellent (9)
7 Carbon–epoxy FRP 440.2516 22.0126 35,470 Fairly good (7)
8 Kevlar 29–epoxy FRP 242.8571 28.5714 11,000 Fairly good (7)
9 Kevlar 49–epoxy FRP 616.4384 34.2466 25,000 Fairly good (7)
10 Boron–epoxy FRP 500 23 315,000 Good (5)
4048 P. Chatterjee et al. / Materials and Design 30 (2009) 4043–4053

cð1; 2Þ ¼ 0:40 þ 0:5  0:10 ¼ 0:45 Now, suppose that, judging all the values in both the concor-
cð1; 3Þ ¼ 0:40 þ 0:5  0:10 ¼ 0:45 dance and discordance matrices, the decision maker has specified
a minimum concordance of 0.70 and a maximum discordance of
cð1; 4 ¼ 0:20 þ 0:5  0:10 ¼ 0:25
0.30, i.e. c(j,k) > 0.70 and d(j,k) < 0.30. With these specifications
cð1; 5Þ ¼ 0:40 and comparing all the concordance and discordance values with
cð1; 6Þ ¼ 0 their respective thresholds, a graph is constructed, as shown in
cð1; 7Þ ¼ 0:20 Fig. 1. The directed path which appears in the graph is determined
cð1; 8Þ ¼ 0:20 by the set of indices that simultaneously satisfy both these require-
ments. This index is (A9,A7).
cð1; 9Þ ¼ 0:20
Using this graph, the decision maker can determine the best
cð1; 10Þ ¼ 0:20 choice of material by eliminating the other nodes. The direction
While calculating the concordance index values, if there are ties be- of the arrow determines which alternative outranks the others.
tween the alternatives, they would receive one half of the weight In Fig. 1, material 9 outranks material 7 and hence, Kevlar
[27]. The complete set of indices is represented by the concordance 49–epoxy FRP is the best and Carbon–epoxy FRP is the second best
matrix, as given in Table 6. material to choose among the ten alternatives. These results ex-
For the discordance indices, the following calculations are made actly match with those of the VIKOR method based analysis and
using Eq. (7): also with the result as obtained by Jee and Kang [8].

dð1; 2Þ : C 1 ¼ 0:0209  0:0422 ¼ 0:0213; C 2 ¼ 0:0662  0:0423 ¼ 0:0239; C 3 ¼ 0:1105  0:1099 ¼ 0:0006;
C 4 ¼ 0:0536  0:0536 ¼ 0
dð1; 2Þ ¼ maxð0:0239; 0:0006Þ= max jð0:0213; 0:0239; 0:0006; 0Þj ¼ 1
dð1; 3Þ : C 1 ¼ 0:0329  0:0422 ¼ 0:0093; C 2 ¼ 0:0617  0:0423 ¼ 0:0194; C 3 ¼ 0:1105  0:1099 ¼ 0:0006;
C 4 ¼ 0:0536  0:0536 ¼ 0
dð1; 3Þ ¼ maxð0:0194; 0:0006Þ= max jð0:0093; 0:0194; 0:0006; 0Þj ¼ 1
dð1; 4Þ : C 1 ¼ 0:0459  0:0422 ¼ 0:0037; C 2 ¼ 0:1278  0:0423 ¼ 0:0855; C 3 ¼ 0:1083  0:1099 ¼ 0:0016;
C 4 ¼ 0:0536  0:0536 ¼ 0
dð1; 4Þ ¼ maxð0:0037; 0:0855Þ= max jð0:0037; 0:0855; 0:0016; 0Þj ¼ 1
dð1; 5Þ : C 1 ¼ 0:0295  0:0422 ¼ 0:0127; C 2 ¼ 0:0491  0:0423 ¼ 0:0068; C 3 ¼ 0:1104  0:1099 ¼ 0:0005;
C 4 ¼ 0:1607  0:0536 ¼ 0:1071
dð1; 5Þ ¼ maxð0:0068; 0:0005; 0:1071Þ= max jð0:0127; 0:0068; 0:0005; 0:1071Þj ¼ 1
dð1; 6Þ : C 1 ¼ 0:0696  0:0422 ¼ 0:0274; C 2 ¼ 0:1229  0:0423 ¼ 0:0806; C 3 ¼ 0:1100  0:1099 ¼ 0:0001;
C 4 ¼ 0:1607  0:0536 ¼ 0:1071
dð1; 6Þ ¼ maxð0:0274; 0:0806; 0:0001; 0:1071Þ= max jð0:0274; 0:0806; 0:0001; 0:1071Þj ¼ 1
dð1; 7Þ : C 1 ¼ 0:1857  0:0422 ¼ 0:1435; C 2 ¼ 0:1082  0:0423 ¼ 0:0659; C 3 ¼ 0:1015  0:1099 ¼ 0:0084;
C 4 ¼ 0:1250  0:0536 ¼ 0:0714
dð1; 7Þ ¼ maxð0:1435; 0:0659; 0:0714Þ= max jð0:1435; 0:0659; 0:0084; 0:0714Þj ¼ 1
dð1; 8Þ : C 1 ¼ 0:1024  0:0422 ¼ 0:0602; C 2 ¼ 0:1404  0:0423 ¼ 0:0981; C 3 ¼ 0:1081  0:1099 ¼ 0:0018;
C 4 ¼ 0:1250  0:0536 ¼ 0:0714
dð1; 8Þ ¼ maxð0:0602; 0:0981; 0:0714Þ= max jð0:0602; 0:0981; 0:0018; 0:0714Þj ¼ 1
dð1; 9Þ : C 1 ¼ 0:2599  0:0422 ¼ 0:2177; C 2 ¼ 0:1683  0:0423 ¼ 0:1260; C 3 ¼ 0:1044  0:1099 ¼ 0:0055;
C 4 ¼ 0:1250  0:0536 ¼ 0:0714
dð1; 9Þ ¼ maxð0:2177; 0:1260; 0:0714Þ= max jð0:2177; 0:1260; 0:0055; 0:0714Þj ¼ 1
dð1; 10Þ : C 1 ¼ 0:2109  0:0422 ¼ 0:1687; C 2 ¼ 0:1683  0:0423 ¼ 0:1260; C 3 ¼ 0:0262  0:1099 ¼ 0:0837;
C 4 ¼ 0:0893  0:0536 ¼ 0:0357
dð1; 10Þ ¼ maxð0:1687; 0:1260; 0:0357Þ= max jð0:1687; 0:1260; 0:0837; 0:0357Þj ¼ 1
dð2; 1Þ : C 1 ¼ 0:0422  0:0209 ¼ 0:0213; C 2 ¼ 0:0423  0:0662 ¼ 0:0239; C 3 ¼ 0:1099  0:1105 ¼ 0:0006;
C 4 ¼ 0:0536  0:0536 ¼ 0
dð2; 1Þ ¼ maxð0:0213Þ= max jð0:0213; 0:0239; 0:0006; 0Þj ¼ 0:891
dð3; 1Þ : C 1 ¼ 0:0422  0:0329 ¼ 0:0093; C 2 ¼ 0:0423  0:0617 ¼ 0:0194; C 3 ¼ 0:1099  0:1105 ¼ 0:0006;
C 4 ¼ 0:0536  0:0536 ¼ 0
dð3; 1Þ ¼ maxð0:0093Þ= max jð0:0093; 0:0194; 0:0006; 0Þj ¼ 0:479 and so on:

where Ci is ith material selection criteria.The complete set of indi- Using the ELECTRE method, it cannot be said how much mate-
ces is represented by the discordance matrix, as given in Table 7. rial 9 outranks material 7. This method is useful for selecting the
 P. Chatterjee et al. / Materials and Design 30 (2009) 4043–4053 4049

Table 5
Normalized decision matrix for example 1.

Sl. no. Materials rlimit/q KIC/q Price/mass Fragmentability

1 300M 0.0422 0.0423 0.1099 0.0536
2 2024-T3 0.0209 0.0662 0.1105 0.0536
3 7050-T73651 0.0329 0.0617 0.1105 0.0536
4 Ti–6Al–4V 0.0459 0.1278 0.1083 0.0536
5 E glass–epoxy FRP 0.0295 0.0491 0.1104 0.1607
6 S glass–epoxy FRP 0.0696 0.1229 0.1100 0.1607
7 Carbon–epoxy FRP 0.1857 0.1082 0.1015 0.1250
8 Kevlar 29–epoxy FRP 0.1024 0.1404 0.1081 0.1250
9 Kevlar 49–epoxy FRP 0.2599 0.1683 0.1044 0.1250
10 Boron–epoxy FRP 0.2109 0.1130 0.0262 0.0893

Table 6
Concordance matrix for example 1.

Alternatives A1 A2 A3 A4 A5 A6 A7 A8 A9 A10
A1 – 0.45 0.45 0.25 0.40 0 0.20 0.20 0.20 0.20
A2 0.55 – 0.45 0.25 0.50 0.20 0.20 0.20 0.20 0.20
A3 0.55 0.55 – 0.25 0.90 0.20 0.20 0.20 0.20 0.20
A4 0.75 0.75 0.75 – 0.70 0.30 0.50 0.20 0.20 0.50
A5 0.60 0.50 0.10 0.30 – 0.25 0.30 0.30 0.30 0.30
A6 1 0.80 0.80 0.70 0.75 – 0.60 0.30 0.30 0.60
A7 0.80 0.80 0.80 0.50 0.70 0.40 – 0.45 0.05 0.30
A8 0.80 0.80 0.80 0.80 0.70 0.70 0.55 – 0.25 0.60
A9 0.80 0.80 0.80 0.80 0.70 0.70 0.95 0.75 – 1
A10 0.80 0.80 0.80 0.50 0.70 0.40 0.70 0.40 0 –

Table 7
Discordance matrix for example 1.

Alternatives A1 A2 A3 A4 A5 A6 A7 A8 A9 A10
A1 – 1 1 1 1 1 1 1 1 1
A2 0.892 – 1 1 1 1 1 1 1 1
A3 0.479 0.418 – 1 1 1 1 1 1 1
A4 0.020 0.037 0.035 – 1 1 1 1 1 1
A5 0.121 0.158 0.112 0.735 – 1 1 1 1 1
A6 0 0.005 0.005 0.047 0.004 – 1 0.935 1 1
A7 0.059 0.055 0.059 0.141 0.229 0.308 – 0.388 1 0.335
A8 0.018 0.002 0.030 0.002 0.391 1 1 – 1 1
A9 0.026 0.029 0.027 0.018 0.155 0.188 0 0.024 – 0
A10 0.497 0.445 0.475 0.497 0.465 0.595 1 0.755 1 –

alternative outranks the others and no decision can be made

[27]. The ELECTRE (or ELECTRE I) method only finds the outranking
A9 A7
relationship between the two best alternatives, whereas, the
ranking of all the alternatives can be available in the ELECTRE II,
ELECTRE III and ELECTRE IV methods [30].
Fig. 1. Resulting graph for example 1.
In order to determine the full ranking of the alternative materi-
als for flywheel design, the ELECTRE II method is now applied and
best set of alternatives that outranks the others and also finding the corresponding pure concordance and discordance indices for
the best alternative in that set. It is sometimes possible that no all the alternative materials are calculated using Eqs. (9) and

Table 8
Ranking of materials for example 1 using ELECTRE II.

Materials Pure concordance index Initial rank Pure discordance index Initial rank Average rank Final rank
300M 4.3 10 6.8876 10 10 10
2024-T3 3.5 9 6.7189 9 9 9
7050-T73651 2.5 7 5.1536 8 7.5 8
Ti–6Al–4V 0.3 6 1.6519 7 6.5 6
E glass–epoxy FRP 3.1 8 0.8827 6 7 7
S glass–epoxy FRP 2.7 3 3.0950 3 3 3
Carbon–epoxy FRP 0.6 5 5.4275 2 3.5 2
Kevlar 29–epoxy FRP 3.0 2 2.6314 4 3 4
Kevlar 49–epoxy FRP 5.6 1 8.5361 1 1 1
Boron–epoxy FRP 1.2 4 1.6047 5 4.5 5
4050 P. Chatterjee et al. / Materials and Design 30 (2009) 4043–4053

10 Table 10
Jee and Kang Ei, Fi and Pi values for example 2.
9
VIKOR Materials Ei Fi Pi Rank
8
ELECTRE II AISI 1020 0.9789 0.3274 0.9868 15
7 AISI 1040 0.9705 0.3190 0.9620 13
Rank

6 ASTM A242 type 1 0.97302 0.3239 0.9751 14

AISI 4130 0.69225 0.3161 0.7870 5
5
AISI 316 0.8012 0.3330 0.8926 11
4 AISI 416 heat treated 0.7599 0.3159 0.8273 7
AISI 431 heat treated 0.7522 0.3089 0.8062 6
3
AA 6061 T6 0.8374 0.3274 0.9012 12
2 AA 2024 T6 0.8132 0.3249 0.8809 9
1 AA 2014 T6 0.8122 0.3274 0.8860 10
AA 7075 T6 0.7938 0.3266 0.8730 8
1 2 3 4 5 6 7 8 9 10
Ti–6Al–4V 0.5463 0.3129 0.6912 4
Material Epoxy–70% glass fabric 0.4108 0.3097 0.6017 3
Epoxy–63% carbon fabric 0.1517 0.1205 0 1
Fig. 2. Comparative ranking of materials for example 1. Epoxy–62% aramid fabric 0.3449 0.3137 0.5713 2

(10), respectively. These index values along with the initial, aver-
age and final rankings are shown in Table 8. The complete ranking rankings (Spearman’s rank correlation coefficient, rs = 0.8909),
of the materials as obtained using the ELECTRE II method is which shows the potentiality of both these MCDM methods in
10-9-8-6-7-3-2-4-1-5. solving such type of material selection problems.
For this material selection problem for flywheel design, both
these MCDM methods give the top rank to the same material 4.2. Example 2
(Kevlar 49–epoxy FRP), although there are small variations in the
rankings as obtained using these two methods. Fig. 2 shows the pic- The second example is related to the selection of material for a
torial representation and comparison of the rankings of these two sailing-boat mast in the form of a hollow cylinder [19]. For such a
methods with respect to the ranking as derived by Jee and Kang component, in addition to high yield strength and high elastic
[8]. There exists a very high rank correlation between these two modulus, a material with low specific density due to weight

Table 9
Quantitative data for example 2 [19].

Sl. no. Materials Specific strength (MPa) Specific modulus (GPa) Corrosion resistance Cost category
1 AISI 1020 35.9 26.9 1 (Poor) 5 (Very low)
2 AISI 1040 51.3 26.9 1 (Poor) 5 (Very low)
3 ASTM A242 type 1 42.3 27.2 1 (Poor) 5 (Very low)
4 AISI 4130 194.9 27.2 4 (V. good) 3 (Moderate)
5 AISI 316 25.6 25.1 4 (V. good) 3 (Moderate)
6 AISI 416 heat treated 57.1 28.1 4 (V. good) 3 (Moderate)
7 AISI 431 heat treated 71.4 28.1 4 (V. good) 3 (Moderate)
8 AA 6061 T6 101.9 25.8 3 (Good) 4 (Low)
9 AA 2024 T6 141.9 26.1 3 (Good) 4 (Low)
10 AA 2014 T6 148.2 25.8 3 (Good) 4 (Low)
11 AA 7075 T6 180.4 25.9 3 (Good) 4 (Low)
12 Ti–6Al–4V 208.7 27.6 5 (Excellent) 1 (Very high)
13 Epoxy–70% glass fabric 604.8 28 4 (V. good) 2 (High)
14 Epoxy–63% carbon fabric 416.2 66.5 4 (V. good) 1 (Very high)
15 Epoxy–62% aramid fabric 637.7 27.5 4 (V. good) 1 (Very high)

Table 11
Ranking of materials for example 2 for different v values.
Pi (v = 0.1) Materials Pi (v = 0.9) Materials
0.9763 AISI 1020 (14) 0.9974 AISI 1020 (15)
0.9398 AISI 1040 (8) 0.9843 AISI 1040 (13)
0.9608 ASTM A242 type 1 (13) 0.9893 ASTM A242 type 1 (14)
0.8938 AISI 4130 (6) 0.6801 AISI 4130 (5)
0.9785 AISI 316 (15) 0.8067 AISI 316 (11)
0.9009 AISI 416 heat treated (7) 0.7537 AISI 416 heat treated (7)
0.8704 AISI 431 heat treated (5) 0.7419 AISI 431 heat treated (6)
0.9590 AA 6061 T6 (12) 0.8433 AA 6061 T6 (12)
0.9459 AA 2024 T6 (9) 0.8159 AA 2024 T6 (9)
0.9560 AA 2014 T6 (11) 0.8160 AA 2014 T6 (10)
0.9504 AA 7075 T6 (10) 0.7956 AA 7075 T6 (8)
0.8625 Ti–6Al–4V (4) 0.5198 Ti–6Al–4V (4)
0.8325 Epoxy–70% glass fabric (2) 0.3709 Epoxy–70% glass fabric (3)
0 Epoxy–63% carbon fabric (1) 0 Epoxy–63% carbon fabric (1)
0.8415 Epoxy–62% aramid fabric (3) 0.3011 Epoxy–62% aramid fabric (2)
 P. Chatterjee et al. / Materials and Design 30 (2009) 4043–4053 4051

Table 12
Normalized decision matrix for example 2.

Sl. no. Materials Specific strength Specific modulus Corrosion resistance Cost category
1 AISI 1020 0.0123 0.0607 0.0208 0.0640
2 AISI 1040 0.0176 0.0607 0.0208 0.0640
3 ASTM A242 type 1 0.0145 0.0614 0.0208 0.0640
4 AISI 4130 0.0668 0.0614 0.0833 0.0670
5 AISI 316 0.0088 0.0567 0.0833 0.0670
6 AISI 416 heat treated 0.0196 0.0635 0.0833 0.0670
7 AISI 431 heat treated 0.0245 0.0635 0.0833 0.0670
8 AA 6061 T6 0.0350 0.0583 0.0625 0.0655
9 AA 2024 T6 0.0486 0.0589 0.0625 0.0655
10 AA 2014 T6 0.0508 0.0583 0.0625 0.0655
11 AA 7075 T6 0.0618 0.0585 0.0625 0.0655
12 Ti–6Al–4V 0.0715 0.0623 0.1042 0.0699
13 Epoxy–70% glass fabric 0.2072 0.0632 0.0833 0.0684
14 Epoxy–63% carbon fabric 0.1426 0.1502 0.0833 0.0699
15 Epoxy–62% aramid fabric 0.2185 0.0621 0.0833 0.0699

Table 13
Concordance matrix for example 2.

Alternatives A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 A11 A12 A13 A14 A15

A1 – 0.333 0.167 0 0.666 0 0 0.333 0.333 0.333 0.333 0 0 0 0
A2 0.666 – 0.500 0 0.666 0 0 0.333 0.333 0.333 0.333 0 0 0 0
A3 0.833 0.500 – 0.167 0.666 0 0 0.333 0.333 0.333 0.333 0 0 0 0
A4 1 1 0.833 – 0.833 0.833 0.833 1 1 1 1 0 0.063 0.063 0.063
A5 0.333 0.333 0.333 0.167 – 0.333 0.333 0.666 0.333 0.333 0.333 0 0.063 0.063 0.063
A6 1 1 1 0.167 0.666 – 0.500 0.666 0.666 0.666 0.666 0.333 0.400 0.063 0.400
A7 1 1 1 0.167 0.666 0.500 – 0.666 0.666 0.666 0.666 0.333 0.396 0.063 0.396
A8 0.666 0.666 0.666 0 0.333 0.333 0.333 – 0.167 0.333 0.167 0 0 0 0
A9 0.666 0.666 0.666 0 0.666 0.333 0.333 0.833 – 0.500 0.500 0 0 0 0
A10 0.666 0.999 0.666 0 0.666 0.333 0.333 0.666 0.500 – 0.167 0 0 0 0
A11 0.666 0.999 0.666 0 0.666 0.333 0.333 0.833 0.500 0.833 – 0 0 0 0
A12 1 1 1 1 1 0.666 0.666 1 1 1 1 – 0.333 0.230 0.562
A13 1 1 1 0.937 0.937 0.600 0.604 1 1 1 1 0.666 – 0.396 0.396
A14 1 1 1 0.937 0.937 0.937 0.937 1 1 1 1 0.770 0.604 – 0.500
A15 1 1 1 0.937 0.937 0.600 0.604 1 1 1 1 0.437 0.604 0.500 –

Table 14
Discordance matrix for example 2.

Alternatives A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 A11 A12 A13 A14 A15

A1 – 1 1 1 1 1 1 1 1 1 1 1 1 1 1
A2 0 – 0.226 1 1 1 1 1 1 1 1 1 1 1 1
A3 0 1 – 1 1 1 1 1 1 1 1 1 1 1 1
A4 0 0 0 – 0 0.042 0.047 0 0 0 0 1 1 1 1
A5 0.065 0.141 0.091 1 – 1 1 1 1 1 1 1 1 1 1
A6 0 0 0 1 0 – 1 0.738 1 1 1 1 1 1 1
A7 0 0 0 1 0 0 – 0.502 1 1 1 1 1 1 1
A8 0.060 0.060 0.077 1 0.797 1 1 – 1 1 1 1 1 1 1
A9 0.043 0.043 0.060 1 0.523 0.717 0.863 1 – 1 1 1 1 1 1
A10 0.060 0.060 0.077 1 0.495 0.667 0.791 0 0.333 – 1 1 1 1 1
A11 0.046 0.052 0.061 1 0.392 0.493 0.557 0 0.038 0 – 1 1 1 1
A12 0 0 0 0 0 0.021 0.023 0 0 0 0 – 1 1 1
A13 0 0 0 0 0 0.001 0.001 0 0 0 0 0.153 – 1 1
A14 0 0 0 0 0 0 0 0 0 0 0 0.237 0.743 – 0.862
A15 0 0 0 0 0 0.007 0.007 0 0 0 0 0.141 0.098 1 –

limitations should be considered. Moreover, as the component is

subjected to mechanical impacts and spray of water, high fracture A14 A12
toughness as well as good corrosion resistance for the material are
also required. Khabbaz et al. [19] considered four criteria, i.e.
specific strength, specific modulus, corrosion resistance and cost
category, and 15 alternative materials while choosing the best A13
material for the sailing-boat mast. The list of the alternative mate-
rials along with their corresponding criteria values are given in
Table 9. Among these four criteria, specific strength, specific Fig. 3. Resulting graph for example 2.
4052 P. Chatterjee et al. / Materials and Design 30 (2009) 4043–4053

Table 15
Ranking of materials for example 2 using ELECTRE II.

Materials Pure concordance index Initial rank Pure discordance index Initial rank Average rank Final rank
AISI 1020 9 14 14 15 14.5 15
AISI 1040 8.333 13 9.9 13 13 14
ASTM A242 type 1 7 9 11 14 11.5 12
AISI 4130 5.039 1 5.9 5 3 3
AISI 316 6.620 8 6.1 12 10 10
AISI 416 heat treated 2.388 2 2.8 9 5.5 7
AISI 431 heat treated 2.376 3 0.2 7 5 5
AA 6061 T6 9.666 15 4.8 11 13 13
AA 2024 T6 8.177 11 2.9 10 10.5 11
AA 2014 T6 8.333 12 0.5 8 10 9
AA 7075 T6 7.500 10 2.4 6 8 8
Ti–6Al–4V 1.545 7 8.5 4 5.5 6
Epoxy–70% glass fabric 1.467 6 10 3 4.5 4
Epoxy–63% carbon fabric 0.389 4 13 2 3 1
Epoxy–62% aramid fabric 1.389 5 13 1 3 2

modulus and corrosion resistance are the beneficial attributes, 15

whereas cost is a non-beneficial attribute. Khabbaz et al.
13 VIKOR
4.2.1. VIKOR method
11 ELECTRE II
While solving this problem using the VIKOR method, at first, the
best and the worst values of all the criteria are identified. Khabbaz 9
Rank

et al. [19] determined the weights for the four criteria as

w1 = 0.333, w2 = 0.333, w3 = 0.125 and w4 = 0.208 using a modified 7
digital logic method. Using Eqs. (2), or (4), (3), and (5), the corre- 5
sponding Ei, Fi and Pi values for the 15 alternative materials are
computed, as given in Table 10. 3
From the compromise ranking of the alternative materials for
design of a hollow cylinder mast for a sailing-boat, as given in 1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Table 10, it is observed that material 14 (epoxy–63% carbon fabric) Material
is the best choice, followed by material 15 (epoxy–62% aramid
fabric). Material 1 (AISI 1020) is the worst choice. Khabbaz et al. Fig. 4. Comparative ranking of materials for example 2.
[19] obtained a ranking of the alternative materials as
15-14-13-4-12-11-10-8-6-7-5-9-2-1-3, whereas a compromise
ranking of 15-13-14-5-11-7-6-12-9-10-8-4-3-1-2 is achieved remains the same and a very high Spearman’s rank correlation
using the VIKOR method. Table 11 shows the compromise rankings coefficient value of 0.9625 between the rankings shows the
of the alternative materials for two extreme values of v = 0.1 and similarity of applicability of these two methods. Fig. 4 exhibits
v = 0.9. In both the cases, the choice of the best and the worst mate- the comparative rankings of the alternative materials as derived
rials for the hollow sailing-boat mast design remain almost the using the VIKOR and ELECTRE II methods with respect to that
same. obtained by Khabbaz et al. [19].

4.2.2. ELECTRE method

In order to solve this problem using the ELECTRE method, the 5. Conclusions
original decision matrix, shown in Table 9, is normalized [27]
and exhibited in Table 12. The two cited examples demonstrate the potentiality, applica-
Now, using Eqs. (6) and (7), the concordance and discordance bility and simplicity of both the compromise ranking and outran-
indices are calculated, as represented in Tables 13 and 14, respec- king methods in solving material selection decision-making
tively. In this problem, say, the decision maker has fixed a mini- problems, involving qualitative as well as quantitative criteria.
mum concordance of 0.60 and a maximum discordance of 0.40, Both these methods can incorporate the decision maker’s prefer-
i.e. c(j,k) > 0.60 and d(j,k) < 0.40. Fig. 3 shows the outranking rela- ences regarding the relative importance of different criteria. The
tionships between the dominating material alternatives. It can be measures of the quantitative and qualitative criteria and their rel-
said that material sets (A14,A12) and (A14,A13) simultaneously ative importance are used together to rank the alternatives, provid-
satisfy both these threshold requirements. In this graph, material ing a better evaluation of the alternatives. The VIKOR method can
14 outranks materials 12 and 13. Hence, epoxy–63% carbon fabric make a compromise ranking among the alternative materials,
is the best choice of material followed by epoxy–70% glass fabric whereas the ELECTRE method is able to search out the best and
and Ti–6Al–4V. This result corroborates with that observed using the next best materials from a given set of alternatives and the
the VIKOR method and also with that as obtained by Khabbaz ELECTRE II method can give a full ranking of the alternative mate-
et al. [19]. rials as considered for a specific engineering application. It is also
Table 15 shows the average and final rankings of the alternative observed that both the VIKOR and ELECTRE II methods are quite
materials when the ELECTRE II method is used. The full ranking is simple to implement involving a large reduction of mathematics
obtained as 15-14-12-3-10-7-5-13-11-9-8-6-4-1-2. as compared to the other conventional material selection methods.
In both these two MCDM methods, the best choice of material The results derived using both these MCDM methods show an
(epoxy–63% carbon fabric) for the sailing-boat mast design excellent correlation with those obtained by the past researchers,
 P. Chatterjee et al. / Materials and Design 30 (2009) 4043–4053 4053

which specifically prove the global applicability of these two meth- [14] Chan JWK, Tong TKL. Multi-criteria material selections and end-of-life product
strategy: grey relational approach. Mater Des 2007;28:1539–46.
ods while solving such type of complex material selection prob-
[15] Rao RV. A decision making methodology for material selection using an
lems. The compromise ranking and outranking methods can also improved compromise ranking method. Mater Des 2008;29:1949–54.
be used for any type of decision-making problem, involving any [16] Sharif Ullah AMM, Harib KH. An intelligent method for selecting optimal
number of quantitative and qualitative criteria and any number materials and its application. Adv Eng Inform 2008;22:473–83.
[17] Shanian A, Savadogo O. A methodological concept for material selection of
of alternatives. highly sensitive components based on multiple criteria decision analysis.
Expert Sys Appl 2009;36:1362–70.
References [18] Zhou C-C, Yin G-F, Hu X-B. Multi-objective optimization of material selection
for sustainable products: artificial neural networks and genetic algorithm
approach. Mater Des 2009;30:1209–15.
[1] Ljungberg LY, Edwards KL. Design, materials selection and marketing of
[19] Khabbaz RS, Manshadi BD, Abedin A, Mahmudi R. A simplified fuzzy logic
successful products. Mater Des 2003;24:519–29.
approach for materials selection in mechanical engineering design. Mater Des
[2] Edwards KL. Selecting materials for optimum use in engineering components.
2009;30:687–97.
Mater Des 2005;26:469–72.
[20] Zeleny M. Multiple criteria decision making. New York: McGraw Hill; 2002.
[3] Deng Y-M, Edwards KL. The role of materials identification and selection in
[21] Opricovic S, Tzeng GH. Multicriteria planning of post-earthquake sustainable
engineering design. Mater Des 2007;28:131–9.
reconstruction. Comput Aid Civ Infrastruct Eng 2002;17:211–20.
[4] Ljungberg LY. Materials selection and design for development of sustainable
[22] Opricovic S, Tzeng GH. Fuzzy multicriteria model for post-earthquake land-use
products. Mater Des 2007;28:466–79.
planning. Nat Hazards Rev 2003;4:59–64.
[5] Wang M-JJ, Chang T-C. Tool steel materials selection under fuzzy environment.
[23] Opricovic S, Tzeng GH. Compromise solution by MCDM methods: a
Fuzzy Sets Sys 1995;72:263–70.
comparative analysis of VIKOR and TOPSIS. Eur J Oper Res 2004;156:445–55.
[6] Chen S-M. A new method for tool steel materials under fuzzy environment.
[24] Opricovic S, Tzeng GH. Extended VIKOR method in comparison with
Fuzzy Sets Sys 1997;92:265–74.
outranking methods. Eur J Oper Res 2007;178:514–29.
[7] Tretheway KR, Wood RJK, Puget Y, Roberge PR. Development of a knowledge-
[25] Rao RV. Decision making in the manufacturing environment using graph theory
based system for materials management. Mater Des 1998;19:39–56.
and fuzzy multiple attribute decision making methods. London: Springer-
[8] Jee D-H, Kang K-J. A method for optimal material selection aided with decision
Verlag; 2007.
making theory. Mater Des 2000;21:199–206.
[26] Roy B, Vincke P. Multi-criteria analysis: survey and new directions. Eur J Oper
[9] Sapuan SM. A knowledge-based system for materials selection in mechanical
Res 1981;8:207–18.
engineering design. Mater Des 2001;22:687–95.
[27] Cho KT. Multicriteria decision methods: an attempt to evaluate and unify.
[10] Qian J, Zhao Y-P. Materials selection in mechanical design for microsensors
Math Comput Model 2003;37:1099–119.
and microactuators. Mater Des 2002;23:619–25.
[28] Milani AS, Shanian A, El-Lahham C. Using different ELECTRE methods in
[11] Rao RV. A material selection model using graph theory and matrix approach.
strategic planning in the presence of human behavioral resistance. J Appl Math
Mater Sci Eng A 2006;431:248–55.
Dec Scs 2006:1–19. Article ID 10936.
[12] Shanian A, Savadogo O. A material selection model based on the concept of
[29] Hunjak T. Mathematical foundations of the methods for multicriteria decision
multiple attribute decision making. Mater Des 2006;27:329–37.
making. Math Commun 1997;2:161–9.
[13] Manshadi BD, Mahmudi H, Abedian A, Mahmudi R. A novel method
[30] Almeida AT. Multicriteria modelling of repair contract based on utility and
for materials selection in mechanical design: combination of non-
ELECTRE I method with dependability and service quality criteria. Ann Oper
linear linearization and a modified digital logic method. Mater Des
Res 2005;138:113–26.
2007;28:8–15.