Accepted Manuscript

A comprehensive MCDM-based approach using TOPSIS,

COPRAS and DEA as an auxiliary tool for material selection
problems

Seyed Hadi Mousavi-Nasab, Alireza Sotoudeh-Anvai

PII: S0264-1275(17)30169-7
DOI: doi: 10.1016/j.matdes.2017.02.041
Reference: JMADE 2781
To appear in: Materials & Design
Received date: 19 November 2016
Revised date: 13 February 2017
Accepted date: 14 February 2017

Please cite this article as: Seyed Hadi Mousavi-Nasab, Alireza Sotoudeh-Anvai , A
comprehensive MCDM-based approach using TOPSIS, COPRAS and DEA as an
auxiliary tool for material selection problems. The address for the corresponding author
was captured as affiliation for all authors. Please check if appropriate. Jmade(2016), doi:
10.1016/j.matdes.2017.02.041

This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.
 ACCEPTED MANUSCRIPT

A comprehensive MCDM-based approach using TOPSIS, COPRAS and DEA as an

auxiliary tool for material selection problems

Seyed Hadi Mousavi-Nasab1

T
IP
Department of Industrial Engineering, Kerman Branch, Islamic Azad University,

CR
Kerman, Iran

US
E-mail: [email protected]
AN
M

Alireza Sotoudeh-Anvai
ED

Department of Industrial Engineering, Kerman Branch, Islamic Azad University,

PT

Kerman, Iran
CE

E-mail: [email protected]
AC

Tel.: +98 9123314524

1
Corresponding author, E-mail: [email protected]
1
 ACCEPTED MANUSCRIPT

Abstract

Material selection problem can be interpreted as an intricate MCDM problem. The aim

of this work is to provide a simple and comprehensive MCDM-based framework for solving

T
this problem. First, we review basic characteristics of general material selection problem

IP
under MCDM paradigm. For doing so, we have studied over 60 papers2 published between

CR
2010 and 2016 as a brief continuation of previous review (Jahan et al., 2010) in this field.

On the other hand, many researchers have emphasized in complicated decision problems

US
more than one MCDM methods should be applied to obtain a more trustworthy and safer
AN
decision. Under the scrutiny of over 100 scientific articles, COPRAS and TOPSIS are chosen

for tackling material selection problem in general. It is observed that the suggested
M

approach by integrating these MCDM techniques is simple and effective. Also we examine
ED

the use of DEA as an MCDM tool in material selection problem. It is found that DEA can be

employed to handle this problem by considering a classical remark, but MCDM cannot be
PT

generally replaced by DEA in this area.

CE

Keywords: Materials selection problem, comprehensive approach, COPRAS, TOPSIS, DEA

AC

2
This study demonstrates that a very important number of articles in this field have been published in
“Materials and Design” (published by Elsevier).
2
 ACCEPTED MANUSCRIPT

Acronyms applied in this work

AHP Analytic Hierarchy Process

T
ANP Analytic Network Process

IP
CES Cambridge Engineering Selector

CR
COPRAS Complex Proportional Assessment

COPRAS-G COPRAS with Grey Relations

US
CRS Constant Returns to Scale
AN
DEA Data Envelopment Analysis

DEMATEL Decision-Making Trial and Evaluation Laboratory

M

DM Decision Maker
ED

DMU Decision-Making Unit

DoE Design of Experiment

PT

DSS Decision Support System

CE

EVAMIX Evaluation of Mixed Data

EXPROM2 extended PROMETHEE II

AC

ELECTRE Elimination et Choice Translating Reality

GRA Grey Relation Analysis

GTMA Graph Theory and Matrix Approach

LP Linear Programming

3
 ACCEPTED MANUSCRIPT

MCDM Multi-Criteria Decision Making

MOORA Multi-Objective Optimization on the basis of Ratio

Analysis

NIS Negative-Ideal Solution

T
OCRA Operational Competitiveness Rating Analysis

IP
OR Operational Research

CR
ORESTE Organization, Rangement Et Synthese De Donnes

Relationnelles

US
PIS Positive Ideal Solution
AN
PRCI Pearson’s Rank Correlation Index

PROMETHEE Preference Ranking Organization Method for

M

Enrichment Evaluations
ED

PSI Preference Selection Index

RAM Rang Adjusted Measure

PT

SAW Simple Additive Weighting

CE

SMART Simple Multiattribute Rating Technique

SRCI Spearman’s Rank Correlation Index

AC

TOPSIS Technique for Order of Preference by Similarity to Ideal

Solution

UTA the Utility Additive

VIKOR VlseKriterijumska Optimizacija I Kompromisno Resenje

4
 ACCEPTED MANUSCRIPT

( Serbian term)

VRS Variable Returns to Scale

WPM Weighted Product Method

T
IP
CR
US
AN
M
ED
PT
CE
AC

5
 ACCEPTED MANUSCRIPT

1. Introduction

The material selection problem should be discussed from a strategic viewpoint due to

its very high impact in the design and development process. There are over 80,000

T
materials in the world (Jahan et al., 2010 and Chatterjee et al., 2011). Clearly, the ideal

IP
option cannot be discovered because of the conflicting tradeoffs among selection factors

CR
(Shanian and Savadogo, 2006). Therefore, a DM has to take into account various factors for

choosing the most appropriate material. For example, Kumar and Singal (2015)

US
summarized a list of various decision factors in material selection problem. Usually,
AN
selection of materials for a given product includes two main steps: 1) Investigation of the

product requirements and 2) Selection of the best possible materials. As a result, often DM
M

is in quest of a proper form of DSS to satisfy the necessities of a given problem. Ashby
ED

(2005) presented a valuable taxonomy of approaches for material selection including free

searching using quantitative study, questionnaire strategy, inductive reasoning and analog
PT

method. Ardeshirilajimi et al. (2015) mentioned that there are different material selection
CE

approaches such as Ashby’s methods, MCDM techniques and other optimization (such as

Tao et al., 2016; Sakundarini et al., 2013; Vats, 2015). However, choosing the best material
AC

from available alternatives3 in the presence of conflicting criteria4 has an MCDM

framework. The MCDM paradigm has been considered as the most famous wing of decision

making theory (Bernroider and Stix, 2007). Jahan et al. (2010), who provided a

3
In this paper, we employ the terms “alternative”, “option” and “choice” interchangeably.
4 In this paper, we employ the terms “criterion”, “attribute” and “decision factor” interchangeably.
6
 ACCEPTED MANUSCRIPT

comprehensive review in the field of material selection, pointed out that MCDM methods

have the potential to significantly improve the material selection process. According to

Karande et al. (2013), material selection problem with different units and several

conflicting attributes can be properly tackled employing MCDM techniques. Many MCDM

T
methods such as AHP, ANP, SAW, TOPSIS, VIKOR, COPRAS, PROMETHEE, DEMATEL,

IP
ELECTRE, MOORA, SMART, PSI, WPM, EVAMIX, GTMA, etc. have been initiated to handle

CR
MCDM problems over recent years. Kumar and Singal (2015) provided a list of various

applications of MCDM techniques in material selection problems. Also a classification of

US
such techniques has been studied in Kumar et al. (2017). The majority of these techniques,
AN
if not all, have been used to material selection problem because of its user friendly process

(Das et al., 2016). However, the identification of which MCDM technique is “best” has been
M

a very hard target to be arrived (Sarkis, 2000). Needless to say, each MCDM technique has
ED

its advantages and inadequacies. According to many researches such as Hobbs and Meier

(2000), Mela et al. (2012) and Kou et al. (2014), DM often cannot deduce that one MCDM
PT

technique is better than the others for a given problem. Therefore, as suggested by Løken
CE

(2007), Jahan et al. (2012), Mulliner et al. (2016) and Wang et al. (2016), more than one

MCDM technique should be applied to a given MCDM problem to obtain a more

AC

trustworthy decision. Although the rankings produced by the various MCDM techniques

usually are not equal, they give a very helpful insight to appraise the alternatives. As a

result, in this work we employ efficiently two MCDM techniques and present an MCDM-

based framework for solving general material selection problem. On the other hand, DEA is

7
 ACCEPTED MANUSCRIPT

one of the most popular methods for performance measurement of DMUs that turn several

inputs into various outputs. The key fame of DEA is owing to its aptitude to obtain

comparative efficiency of a DMU without previous weights on inputs and outputs5

(variables). The basic goal of DEA is not to prioritize units and this model determines

T
efficient and inefficient DMUs. However, several studies have discussed about close

IP
connections and similarities between DEA and MCDM methods. According to Stewart

CR
(1996) , Li and Reeves (1999) and Zavadskas et al. (2014), DEA and MCDM, as two of the

fastest growing parts of OR, have been employed to a very extensive types of managerial

US
problems. But to the best of our knowledge, tackling material selection problem using DEA
AN
is relatively novel. Furthermore, the absence of DEA models in the famous literature review

(Jahan et al., 2010) in material selection methods is clear. Therefore, we aim to employ DEA
M

in material selection problems because of its very straightforward framework and intuitive
ED

nature. The final goal of this work is to suggest a comprehensive and simple MCDM-based

procedure on the basis of the distinguishing qualities of three MCDM techniques for general
PT

material selection problem. Briefly, the contribution of our work is fivefold:

CE

1) We provide for the first time an extensive review regarding characteristics of

AC

material selection problems in general practice under MCDM paradigm.

2) We provide a comprehensive review for selecting the appropriate MCDM technique

for general material selection problem.

5
In this paper, we employ the terms “input and output” and “variable” interchangeably.
8
 ACCEPTED MANUSCRIPT

3) According to this study, TOPSIS and COPRAS are selected as the best MCDM

techniques for ranking the alternative materials in general practice.

4) DEA can be used in material selection problem as an auxiliary tool by considering a

rough rule of thumb.

T
5) A comprehensive MCDM-based approach using TOPSIS, COPRAS and DEA for

IP
general material selection problem is introduced.

CR
US
As mentioned, up to now, many MCDM-based methods have been introduced for

choosing the best option or ranking of materials. However, our suggested MCDM-based
AN
approach has been created on the basis of general characteristics of material selection

problem as well as the strengths of three MCDM techniques. Noteworthy, it is the key
M

novelty of this work. Also the suggested method is very straightforward to implement and
ED

no extra calculations are essential. Finally, the obtained ordering of materials using
PT

proposed approach is in very good agreement with those deduced by the past studies.
CE

This article is arranged as follows: in Section 2 we review the relevant literature. Section
AC

3 reviews MCDM framework, TOPSIS, COPRAS, DEA and various comparative advantages of

these methods. In Section 4, we present our suggested MCDM-based approach for material

selection problem in generl. Section 5 presents five real examples to exhibit the

applicability and efficacy of the suggested approach. Conclusion is included in Section 6.

9
 ACCEPTED MANUSCRIPT

2. Literature review

Now, we review some relevant literature on MCDM techniques for material selection

problems and choosing an MCDM technique in this field.

T
IP
2.1. MCDM techniques for material selection problems

CR
According to Jahan et al. (2010), TOPSIS, ELECTRE and AHP have been the most

US
prevalent techniques in material selection context. In general, literature review shows AHP
AN
is one of the very popular MCDM approaches to cope with MCDM problems ( e.g.

Ramanathan and Ganesh, 1995; Özcan et al., 2011; Mansor et al., 2013; Milani et al., 2013;
M

Khorshidi et al., 2013; Mulliner et al., 2016). But when there are many alternatives, the
ED

pairwise comparison approach done by AHP is clearly infeasible. Chatterjee et al. (2011)

pointed out that calculation time of AHP is very high and this technique has a complex
PT

structure in material selection problem. Also according to Hatami-Marbini et al. (2013),

CE

AHP has a controversial frame and suffers from deficiency of transparency in procedure.

Various shortcomings of AHP were discussed by Ramanathan and Ganesh (1995).

AC

Although Mansor et al. (2013) employed AHP in a material selection problem and reported

its good result, the laborious nature of this technique is obvious. Jahan et al. (2010)

mentioned that AHP can evaluate a limited amount of options, which is often not over 15.

But as pointed out by Girubha and Vinodh (2012), in material selection problem the

10
 ACCEPTED MANUSCRIPT

number of options and criteria are often large. For example, Chauhan and Vaish (2013)

studied a material selection problem with 37 alternatives along with various decision

factors. Also İpek et al. (2013) analyzed another material selection problem with 35

options and several criteria. More formally, we can deduce that AHP is not an efficient

T
method in the numerous MCDM problems such as material selection problem.

IP
Notheworthy, when the number of alternatives is increased several MCDM methods will

CR
computationally be difficult. For example, GTMA method, which was initiated by Rao

(2006) as a MCDM tool, has often been limited to a problem with 5 or 6 criteria due to

US
difficult computation procedure (Karande and Chakraborty, 2012). Also Ashby method will
AN
be too difficult when the amount of attributes is increased (Yazdani and Payam, 2015). As

can be seen in Jahan et al. (2010), another popular MCDM technique in material selection
M

problem is ELECTRE. However, Rao (2006) pointed out that ELECTRE can obtain just a
ED

partial ranking of materials. Also as the number of choices rises, the volume of calculations

increased fast. Furthermore, ELECTRE only provides the rank of each alternative and does
PT

not determine numerical score for more understanding of discrepancy between options
CE

(Jahan et al., 2010). Although ELECTRE II provides a complete ranking of the choices, as

mentioned by Athawale and Chakraborty (2011), this technique does not work efficiently
AC

because of mathematical intricacy in selection problems. On the other hand, PROMETHEE

method aims to diminish the limitations with ELECTRE by giving a more obvious approach

(Jahan et al., 2011). Løken (2007) explained that a major difference between ELECTRE III

and PROMETHEE II is in computation process. PROMETHEE II has a clear calculation

11
 ACCEPTED MANUSCRIPT

process whilst a DM finds the computations from ELECTRE III very intricate. Also Peng and

Xiao (2013) reported that PROMETHEE has smaller computational efforts than ELECTRE.

However, PROMETHEE family suffers from various defects. For example, designing the

problem and identifying the weights are key weak points of PROMETHEE (Hatami-Marbini

T
et al., 2013). Moreover, PROMETHEE is not able to provide a perfect ranking of the options.

IP
As mentioned above, based on the work of Jahan et al. (2010), TOPSIS, ELECTRE, and AHP

CR
have been the most widespread MCDM techniques in material selection area.

Unfortunately, AHP and ELECTRE family are intricate in nature and need the DM’s high

US
interpretation in pairwise assessment (Rao, 2006; Jahan et al., 2010; Özcan et al., 2011;
AN
Jahan et al., 2011; Chatterjee et al., 2011; Karande and Chakraborty, 2012; Hatami-Marbini

et al., 2013).
M
ED

After Jahan et al.’s (2010) review, Chatterjee et al. (2011) studied COPRAS and EVAMIX

for material selection problem. They reported that COPRAS is a very simple technique as
PT

well as has a low calculation time compared to AHP and EVAMAIX. Khabbaz et al. (2009)
CE

applied a simplified fuzzy approach for handling quantitative as well as qualitative material

factors. Maniya and Bhatt (2010) introduced PSI method for material selection problem
AC

and claimed that this method is easier compared to other methods such as TOPSIS and

VIKOR. However, according to Rao and Patel (2010), this claim does not have adequate

validity. Rao and Patel (2010) revealed some shortcomings of previous studies such as

Khabbaz et al. (2009) and Maniya and Bhatt (2010). Also they introduced a new MCDM

12
 ACCEPTED MANUSCRIPT

approach by the objective and the subjective preferences to choose the integrated weights

of decision factors. Cicek et al. (2010) suggested an integrated DSS to suit the appropriate

methods on the basis of various factors such as size and type of a given problem. Yazdani

and Payam (2015) prepared a comparative work by Ashby method, VIKOR and TOPSIS for

T
a given material selection problem and derived a meaningful agreement between these

IP
approaches. Also Chauhan and Vaish (2013) reported a good congruence between TOPSIS

CR
and Ashby technique in a given material selection problem. Çalışkan et al. (2013) suggested

a decision approach including EXPROM2, TOPSIS and VIKOR for solving a material

US
selection problem and found SRCI between TOPSIS and VIKOR is the highest. Jahan et al.
AN
(2012) suggested a new normalization in TOPSIS for material selection problem. They

deduced that this TOPSIS needs less mathematical effort compared to comprehensive
M

VIKOR (Jahan et al., 2011) for efficient decision. Singh et al. (2016) used GRA to obtain a
ED

perfect ranking in material selection problem. According to this study, GRA was proved to

be a powerful technique in selection problem. Noteworthy, GRA procedure is similar to

PT

TOPSIS procedure. But TOPSIS method employs Euclidean distance, whereas GRA exerts
CE

grey relational grade. Milani et al. (2013) studied the use of ANP, as a generalized form of

AHP, in a material selection problem. In that study, the authors considered the
AC

dependencies between the attributes or options. Liu et al. (2014) suggested a hybrid MCDM

framework combining DEMATEL, ANP, and VIKOR to handle selection problem in materials

area. Also Peng and Xiao (2013) employed PROMETHEE incorporated with ANP for a

material selection problem. In their work, ANP was employed to obtain weights and

13
 ACCEPTED MANUSCRIPT

PROMETHEE to prioritize options. However, as pointed out by Yang et al. (2008),

normalization process in classical ANP suffers from a severe defect in practical fields.

By analyzing the literature, we can see numerous good MCDM-based approaches have

T
been introduced in materials selection field, but there is a key need to suggest a

IP
straightforward and efficient mathematical procedure to guide and support the DM in this

CR
area.

US
2.2. Choosing an MCDM technique
AN
Sun and Li (2010) mentioned that more than 70 MCDM methods have been introduced
M

for simplifying the decision procedure. A survey of MCDM approaches was provided by
ED

Zavadskas et al. (2014). However, it is a very challenging subject to choose which MCDM

technique is suitable for a given problem. Voogd (1983) revealed that at least 40% of cases
PT

each MCDM method generated a dissimilar outcome from any other method. According to
CE

Zanakis et al. (1998), theses inconsistencies happen because:

AC

1) MCDM methods employ weights in a different way.

2) MCDM techniques vary in the way to choosing the “best” option.

3) Most MCDM methods try to aggregate the decision factors (objectives) in a different

way.

14
 ACCEPTED MANUSCRIPT

4) A number of MCDM methods use supplementary parameters that influence

solution.

Sun and Li (2010) mentioned that selection of the most proper MCDM procedures is a

T
difficult MCDM problem. Mela et al. (2012) studied several MCDM techniques in the field of

IP
building design. However, they pointed out that the best MCDM technique can hardly be

CR
discovered. Cicek and Celik (2010) suggested a framework to choose a MCDM technique

among AHP, PROMETHEE, SMART, ELECTRE, and TOPSIS in material selection field. But as

US
pointed out by Jahan et al. (2011), the selected MCDM approach using Cicek and Celik’s
AN
procedure usually is not exclusive. According to Løken (2007), the most significant factor in

MCDM context is to discover a MCDM technique that evaluates what it is expected to

M

measure. Also Chatterjee et al. (2011) studied the performance of various MCDM
ED

techniques by considering several factors such as calculation time, ease, transparency, form

of information, etc. As stated earlier, although each method has its own distinguishing
PT

quality, various MCDM procedures can produce incompatible consequences when applied
CE

to the same problem. As a result, different comparative studies of MCDM methods have

been provided in the literature (e.g. Triantaphyllou, 2013; Ceballos et al., 2016, Darji and
AC

Rao, 2014; Sen et al., 2016; Zavadskas et al., 2014; Løken, 2007). For example, Özcan et al.

(2011) compared TOPSIS, AHP and ELECTRE as the most well-known techniques in MCDM

context. According to this study, TOPSIS and ELECTRE generated similar results in location

selection problem. Athawale and Chakraborty (2011) analyzed the performance of 10

15
 ACCEPTED MANUSCRIPT

MCDM techniques such as TOPSIS, ELECTRE II, PROMETHEE II, AHP, etc. in a robot

selection problem and observed that TOPSIS, WPM, and GRA are slightly superior to the

others. Also on the basis of the ranking of the top three choices, PROMETHEE presented the

worst mismatch. Chatterjee and Chakraborty (2012) shown that the results produced by

T
EXPROM2, COPRAS-G, ORESTE and OCRA completely match with the outcome produced by

IP
VIKOR. Paul et al. (2015) reported slight disagreement in the results obtained by TOPSIS

CR
and PPROMETHEE in a selection of 3D printers. Kumar and Ray (2015) mentioned that

EXPROM2 and OCRA are better than ORESTE in a gear material selection problem.

US
Maity and Chakraborty (2015) used PROMETHEE II, COPRAS, TOPSIS, VIKOR, AHP, and
AN
ANP to tackle a tool steel material selection problem. According to this study, a very high

SRCI can be observed between these MCDM methods and PROMETHEE II. Mulliner et al.
M

(2016) deduced that none of MCDM procedures is ideal and proposed over one technique
ED

must be used to a problem to achieve a stronger decision. Also Jahan et al. (2012) pointed

out in intricate material selection problems at least two MCDM methods should
PT

concurrently be exerted for a decision problem. On the other hand, various studies in
CE

MCDM context have emphasized the use of straightforward and understandable techniques

to cope with MCDM problems (Løken, 2007; Bernroider and Stix, 2007; Mulliner et al.,
AC

2016). Mulliner et al. (2016) pointed out that the difficulty of an MCDM procedure may

prevent its use practically. As explained by Chatterjee and Chakraborty (2012), the

difficulty of a MCDM problem will be distinguished by the number of choices, criteria and

form of data (deterministic, stochastic, and so on). Therefore, to decrease the difficulty in

16
 ACCEPTED MANUSCRIPT

choosing the best MCDM method in material selection problem, the calculations should be

straightforward, easy and economical.

3. Preliminaries and distinctive features of TOPSIS, COPRAS and DEA for

T
material selection problem

IP
CR
To make our work self-contained, MCDM framework, TOPSIS, COPRAS, and DEA that will

be employed in the suggested model are reviewed. Furthermore, comprehensive as well as

US
comparative information about these methods on the basis of more than 100 papers is
AN
provided to readers.
M

3.1. MCDM structure

ED

MCDM is a discipline which allows the consideration of disparate as well as conflicting

PT

attributes to prioritize or choose various alternatives to help a DM. An MCDM framework

CE

has usually been described by four elements as follows:

AC

- Criteria (known as attributes or decision factors)

- Alternatives (known as options or choices)

- Weight of attribute (known as comparative significance of criterion)

- Performance ratings of options regarding the criteria

17
 ACCEPTED MANUSCRIPT

The decision matrix as specified in Model (1) depicts options ( A1 , A2 ,..., Am ), attributes (

n
C1 , C2 ,..., Cn ), weights of attributes ( w1 , w2 ,..., wn ) where w
j 1
j  1 and the performance

ratings of choices, i.e. xij ( for i  1,..., m; j  1,..., n) .

T
IP
CR
C1 C2 … Cn

 x11 x1n 

US
A1 x12 ....
 
 x21 x21 .... x2 n 
A2 . . . 
AN
D   W  [w1 , w2 ,..., wn ] (1)
. . . . 
. . . 
 
M

.
 xm1 xm 2 .... xmn 
.
ED

Am
PT
CE

The task of DM is to obtain the best options or to rank alternatives by using a MCDM

technique based on the decision matrix.

AC

18
 ACCEPTED MANUSCRIPT

The first step in most MCDM approaches is normalization of decision matrix. Due to

various units of the attributes, normalization is indispensable for direct comparison.

Various MCDM techniques utilize various normalization methods. Jahan and Edwards

(2015) pointed out that although the majority of normalization approaches appear to be

T
just slight variations of each other, these minor differences may produce major outcomes

IP
on the final decision when choosing materials. Jahan and Edwards (2015) provided a good

CR
review on normalization methods. Also reader can see Shih et al. (2007) and Peng and Xiao

(2013) and references therein for more information.

US
AN
Because the evaluation attributes are not equally significant, determining the suitable

weight for criteria is another key step in MCDM context. But DM opinion can play a very
M

vital role to obtain criteria weights and will influence the quality of results. The majority of
ED

the methods for this task can be classified into two groups (Lotfi and Fallahnejad, 2010):

subjective and objective. Subjective weights are obtained just based on the preference of
PT

DM. But the objective weights are identified by mathematical frames in the absence of DM’s
CE

preferences. As pointed out by Dehghan-Manshadi et al. (2007), Jahan et al. (2011),

Anupam et al (2014), Sotoudeh-Anvari and Sadi-Nezhad (2015), and Wang et al. (2016)
AC

various techniques such as AHP, Shannon’s Entropy, Simos approach, Digital Logic, Z-

numbers approach, Z-transformation approach, DoE, etc. have been employed for

determining criteria weight. Also Athawale et al. (2011) provided a list of various

weighting methods for material selection problems. For example, AHP has been helpful in

19
 ACCEPTED MANUSCRIPT

weighting of the criteria by the subjective evaluations in some material selection problem

(Yousefpour and Rahimi, 2014). Anupam et al. (2014) suggested an efficient Entropy based

TOPSIS for material selection in pulp and papermaking. In Çalışkan et al. (2013) the

criteria weights were obtained by a compromised weighting technique, where Shannon’s

T
Entropy and AHP were integrated. Also in Yazdani and Payam (2015) the criteria weights

IP
were identified by experts directly. More formally, according to Jahan et al. (2011),

CR
specifying weight to attributes must be carried out carefully to avoid bias or generating the

solution you proposed. After normalization and weighting process each MCDM technique

US
employs a special approach to rank alternatives.
AN
3.2. The TOPSIS method
M
ED

TOPSIS is created on selecting the best choice having the shortest Euclidean distance

and the farthest Euclidean distance to the PIS and from the NIS, respectively. The PIS is one
PT

that maximizes the maximizing attributes and minimizes the minimizing attributes, while
CE

the NIS maximizes the minimizing criteria and minimizes the maximizing criteria (Wang et

al., 2016). Literature review shows that TOPSIS, which was introduced by Hwang and Yoon
AC

(1981), is a privileged technique in selection problem in general. For example, Sun and Li

(2010) pointed out that TOPSIS is the best method compared to other MCDM techniques in

an aircraft selection problem. Peng (2015) reported that TOPSIS is the most suitable

technique for a quake vulnerability selection problem. Noteworthy, TOPSIS has a very

20
 ACCEPTED MANUSCRIPT

transparent logic and is useful for qualitative as well as quantitative information. Also it

can be used simply when the amount of options and criteria is large (Özcan et al., 2011;

Jahan et al., 2010; Sotoudeh-Anvari and Sadi-Nezhad, 2015). Evidently, TOPSIS method is

one of the most widely accepted MCDM techniques in material selection problem (Jahan et

T
al., 2010). Liao (2015) reported 31 records about material selection and TOPSIS in the Web

IP
of Science. Based on a comprehensive literature review, Jahan et al. (2010) reported that

CR
TOPSIS is a very efficient MCDM technique for material selection. Also Shanian and

Savadogo (2006), Huang et al. (2011), İpek et al. (2013), Maity and Chakraborty (2013),

US
Anupam et al (2014), Tewari et al. (2015), Kumar and Singal (2015), Kaushik et al. (2015)
AN
and Debnath et al. (2016) pointed out that TOPSIS is very reliable and efficient for tackling

different material selection problems. According to Yousefpour and Rahimi (2014), TOPSIS
M

is a managerial approach that will be useful in material selection problem. Kumar and
ED

Singal (2015) pointed out that TOPSIS is best suited for penstock material selection. Rathod

and Kanzaria (2011) deduced that TOPSIS and fuzzy TOPSIS are appropriate approaches in
PT

material selection problem when the ratings are precise and imprecise. According to
CE

Zanakis et al. (1998), TOPSIS experiences the fewest rank reversals compared to the

several MCDM methods such as ELECTRE and various version of AHP. This phenomenon is
AC

an alteration in the ordering of options when an option is entered or removed. As pointed

out by Dey et al. (2016), this event confuses a DM and must be minimized. Furthermore,

contrary to AHP or SAW, TOPSIS can directly use maximizing as well as minimizing

attributes. As mentioned before, the difficulty of an MCDM procedure can prevent its use in

21
 ACCEPTED MANUSCRIPT

practice. Fortunately, TOPSIS has a very easy calculation procedure (Shih et al., 2007). Dey

et al. (2016) provided a good comparison of TOPSIS with other famous MCDM techniques

such as VIKOR, AHP, etc.

T
According to Mulliner et al. (2016), one of the most crucial decision factors in choosing

IP
MCDM procedures is the adaptability with the problem’s goal. Generally, in material

CR
selection problem, the impact of each criterion cannot be separately analyzed and should

usually be considered as a trade-off with other criteria. For example, variation in

US
mechanical or electrochemical factors can modify the preferences of other parameters.
AN
More formally, variations in a criterion may be compensated by other factors (Shanian and

Savadogo, 2006). Fortunately, TOPSIS is a compensatory method and provides clear trade-
M

offs among criteria. The TOPSIS method allows compromise between various decision
ED

factors where a bad effect in one factor may be compensated with the good effect in other

factor (Kaushik et al., 2015). But for example, ELECTRE and PROMETHEE have a non-
PT

compensatory root. Therefore, these advantages make TOPSIS as a key MCDM approach
CE

compared to the majority of MCDM techniques such as AHP and ELECTRE in material

selection context. Moreover, since uncertainty exists in nearly all decision information
AC

several researchers have been considered ambiguity involved in material selection

problems. Fortunately, TOPSIS can very easily be combined or extended with other

concepts of uncertainty such as type-2 fuzzy set, Z-numbers (Sotoudeh-Anvari and Sadi-

Nezhad, 2015), etc. For example, Liao (2015) provided an interval type-2 fuzzy TOPSIS for

22
 ACCEPTED MANUSCRIPT

material selection. Various applications as well as many developments of TOPSIS in

material selection problems can be seen, for example, in Jahan et al. (2012), Mirhedayatian

et al. (2013) and Singh et al. (2016). Finally, TOPSIS can be applied as a benchmark in

material selection problem to evaluate outcomes of other MCDM techniques. The key steps

T
in TOPSIS are explained as below:

IP
CR
Step 1: normalize decision matrix using the Model (2):

US
xij
rij  (i  1, 2,..., m; j  1, 2,..., n)
n (2)
 xij
AN
j 1

Step 2: provide weight to this matrix by:

M
ED

vij  wj  rij i  1,2,..., m; j  1,2,..., n. (3)

PT

Step 3: define the best (ideal) solution and nadir ideal solution as follows:
CE
AC

A*  {v1* , v2* ,...., vn*}  {(max j vij i  I ),(min j vij i  I )}

i  1, 2,..., m; j  1, 2,..., n.
A  {v1 , v2 ,...., vn }  {(min j vij i  I ),(max j vij i  I )} (4)
i  1, 2,..., m; j  1, 2,..., n.

23
 ACCEPTED MANUSCRIPT

Where I  is related to benefit attributes and I  is related to cost attributes.

Step 4: achieve the remoteness of all choices from A and A using Model (5):

T
IP
n
D 
 (v  v j ) 2 i  1, 2,..., m

CR
i ij
j 1
(5)

US
AN
n
D 
i

 (v
j 1
ij  v j )2 i  1, 2,..., m
M

Step 5: a relative closeness to the perfect solution will be gained using Model (6):
ED
PT

Di
CCi*  i  1, 2,..., m
Di  Di (6)
CE
AC

Step 6: prioritize the options by CCi* . The larger CCi* indicates better accomplishment of

options.

24
 ACCEPTED MANUSCRIPT

3.3. The COPRAS method

The COPRAS technique, which is initiated by Zavadskas et al. (1994), works similar to SAW.

T
SAW is the most straightforward and one of the most accepted MCDM methods. Usually,

IP
SAW is employed as a benchmark tool to evaluate outcomes produced from other MCDM

CR
techniques (Wang et al., 2016). Abdullah and Adawiyah (2014) provided a good review on

SAW techniques highlighting the different usages related to useful selection of materials.

US
For example, Anupam et al. (2015) suggested an MCDM-based technique on the basis of
AN
entropy integrated SAW for material selection in pulping and papermaking field. However,

SAW can employ just maximizing attributes, whereas minimizing criteria must be turn into
M

the maximizing factors before use. As mentioned, cost-type factors must be converted in to
ED

benefit-type factors before normalization in SAW, but this operation is not a trivial task. For

example, Millet and Schoner (2005) explained that this transformation in AHP may cause
PT

contradictory outcomes. Fortunately, this inadequacy is removed in COPRAS. More

CE

formally, it is the superiority of COPRAS over SAW (Podvezko, 2011). Also according to

Zavadskas et al. (2014), who provided a good review of MCDM techniques, COPRAS is a fast
AC

developed method to deal with real problems. It should be noted that although Karande

and Chakraborty (2012) mentioned that COPRAS method suffers from a very hard

procedure and MOORA is easy to use, unfortunately the authors did not provide any reason

for this claim. On the contrary, according to Podvezko (2011), Mulliner et al. (2013),

25
 ACCEPTED MANUSCRIPT

Nuuter et al. (2015), Mulliner et al. (2016) and Chatterjee et al. (2011) COPRAS is simple

and even very simple to utilization. Also Karande and Chakraborty (2012) claimed that

COPRAS has been affected by the process of normalization. But normalization method in

both COPRAS and MOORA are exactly the same. On the other words, although MOORA has

T
its own advantages (e.g. remarkable simplicity in use), this technique has no obvious

IP
advantage over COPRAS. Finally, Mulliner et al. (2013), Zavadskas et al. (2014),

CR
Kouchaksaraei et al. (2015) and Turanoglu Bekar et al. (2016) reported that COPRAS

technique has been successfully used to a vast range of real problems. The key steps of

US
COPRAS technique is reviewed as below (Chatterjee et al., 2011):
AN
Step 1: normalize decision matrix using Model (7):
M
ED

xij
rij  m

x (7)
PT

ij
i 1
CE

Step 2: weight normalized matrix as follows:

AC

yij  w j  rij (i  1,2,..., m; j  1,2,..., n) (8)

26
 ACCEPTED MANUSCRIPT

Step 3: calculate sums of weight normalized scores for beneficial as well as cost criteria

using Model (9) as follows:

n
S i   y ij
j 1

T
IP
(9)

CR
n
Si   yij
j 1

US
In Model (9), y ij are related to maximizing criteria and y ij are related to minimizing
AN
criteria. The bigger S  i , the higher is the choice as well as the lower S  i , the higher is that
M

option. In fact, S  i and S  i demonstrate the level of objectives reached by option i .

ED

Step 4: obtain the relative priority of the options, i.e. Qi using Model (10) as follows:
PT
CE

S
AC

i
(10)
Qi  S i  i 1
m
(i  1, 2,..., m)
1
Si 
i 1 Si

27
 ACCEPTED MANUSCRIPT

The alternative with the highest value of Qi ( Qmax ) is the best.

Step 5: compute the level of performance ( U i ) for option i .

T
Finally, U i provides an absolute prioritizing and is calculated using Model (11) as follows:

IP
CR
Qi (11)
Ui   100
Qmax

US
AN
In Model (11), Qmax is the highest relative importance score.
M

3.4. TOPSIS vs. VIKOR

ED
PT

Recently, several researchers such as Jahan et al. (2011), Girubha and Vinodh (2012), Liu

et al. (2013), Vats and Vaish (2013), Jahan and Edwards (2013) and Ishak et al. (2016)
CE

employed VIKOR to select the most appropriate material in different material selection
AC

problems. Although in classical form, VIKOR uses linear normalization and TOPSIS employs

vector normalization, as mentioned by Chatterjee et al. (2011), VIKOR and TOPSIS are

naturally similar. However, due to vector normalization, TOPSIS can have higher

distinguishing power (Mulliner et al., 2016). Antucheviciene et al. (2012) analyzed the

results of fuzzy TOPSIS based on linear and vector normalization for a given selection
28
 ACCEPTED MANUSCRIPT

problem. According to this study, fuzzy TOPSIS based on vector normalization produced

more realistic result. Jahan et al. (2011) mentioned that both VIKOR and TOPSIS are

suitable for tackling a large number of options and attributes in material selection problem.

Çalışkan et al. (2013) deduced that the correlations obtained between VIKOR and TOPSIS

T
are usually high. Also Sharma and Gupta (2015) reported a very good agreement between

IP
VIKOR and TOPSIS in material selection for thin film transistor. But Huang et al. (2009)

CR
warned that VIKOR even in a simple problem may produce an incorrect ranking. Reader

interested in exploring on this serious flaw is referred to Huang et al. (2009), Jahan et al.

US
(2011) and Ceballos et al. (2016). Also VIKOR is not simple to apply and the computation
AN
time of this technique is relatively high (Dey et al., 2016). Notheworthy, VIKOR’s outcome is

influenced by an extra parameter, i.e.  (strategy weight of the maximum set usefulness).
M

Ceballos et al. (2016) warned that VIKOR’s result is very sensitive to this parameter.
ED

Antucheviciene et al. (2011) and Antucheviciene et al. (2012) proved that the ultimate

choice should be considered by TOPSIS and COPRAS rather than VIKOR in a redevelopment
PT

problem. Moreover, Dey et al. (2016) reported that the top ranking of options matches in
CE

the majority of MCDM techniques such as TOPSIS and MOORA except by VIKOR in a

machine tool selection. However, as mentioned by Mela et al. (2012), various fields in
AC

engineering have problems with special specifications that can influence the selection of

MCDM technique. For example, Caterino et al. (2009) studied a comparative performance

of several MCDM techniques for a retrofit problem. In spite of VIKOR did not provide a

complete ranking in Caterino et al. (2009), the authors deduced that TOPSIS and VIKOR are

29
 ACCEPTED MANUSCRIPT

more suitable for dealing with that problem compared with WPM, ELECTRE, PROMETHEE

I, II. Dey et al. (2016) provided a good comparison of TOPSIS with VIKOR in various ranking

problems. More formally, although VIKOR is known as an accepted method in various

MCDM problems, the comparative studies such as Caterino et al. (2009), Sun and Li (2010),

T
Antucheviciene et al. (2011), Athawale and Chakraborty (2011), Chatterjee et al. (2011),

IP
Karande and Chakraborty (2012), Mela et al. (2012), Jahan et al.(2012), Çalışkan et al.

CR
(2013), Kou et al. (2014), Yazdani and Payam (2015), Peng (2015) have not reported any

superior performance of VIKOR over TOPSIS.

US
AN
3.5. The DEA method
M

3.5.1. DEA models

ED

DEA as a LP method appraises the relative efficiency value of a set of DMUs which use
PT

various inputs to create various outputs. Initially, Charnes, Cooper, and Rhodes (1978)
CE

introduced the original DEA framework, called CCR model, to cope with the CRS situation.

Next, Banker, Charnes, and Cooper (1984) developed a novel DEA model, called BCC model,
AC

to tackle the VRS condition. We refer to Cooper et al. (2007) and Podinovski and

Thanassoulis (2007) and references therein for more study about this topics. Moreover,

DEA frameworks can be categorized to maximize outputs (known as output-oriented

30
 ACCEPTED MANUSCRIPT

model) or to minimize inputs (known as input-oriented model). Mathematically, the

output-oriented CCR frame is modeled as follows:

Consider n DMUs that consume m various inputs to create s various outputs. The

T
efficiency value of DMUo i.e. Eo is calculated by the following CCR framework:

IP
CR
s

u y
US
r ro
max Eo  r 1
m

v x
AN
i io
i 1
s

u y r rj
M

s.t. r 1
m
1 j  1,..., n
v x i ij (12)
ED

i 1

ur , vi  0
PT
CE
AC

31
 ACCEPTED MANUSCRIPT

In Model (12), yrj and xij are the r th output and i th input of the j th DMU, respectively.

Also ur and vi are the weights of the r th output and i th input, respectively. To simplify

the computation, Model (12) can be transformed into an LP model as follows:

T
IP
s
max Eo   ur yro

CR
r 1

s.t.

US
m

v x
i 1
i io 1
(13)
s m

u y  v x
AN
r rj i ij 0 j  1,..., n
r 1 i 1

ur , vi  0 r  1, 2,..., s and i  1, 2,...m

M
ED
PT

Model (13) is executed n times to calculate the efficiency value of all DMUs by fitting the

best achievable input and output weights. Finally, when Eo  1 DMUo is efficient and the
CE

efficiency value less than 1 indicates that DMUo is inefficient. Also Model (13) can be
AC

converted into the Model (14) by duality as follows:

32
 ACCEPTED MANUSCRIPT

min  o
s.t.

 x
j
j ij   o xio i  1, 2,..., m

T
 y j rj  yro r  1, 2,..., s

IP
j

j  0 j (14)

CR
Model (14) is known as the “envelopment” model. Because of lesser constraints Model

US
(14) is easier to work out than Model (13). In addition, BCC model can be obtained by
AN
n
adding 
j 1
j  1 as a constraint to CCR model. Noteworthy, up to now different DEA
M

models have been suggested and often these models have been generated different
ED

efficiency values depending on the framework (input or output-oriented, CCR or BCC, etc)

(Podinovski and Thanassoulis, 2007). Readers are referred to Cooper et al. (2007) and
PT

Cook and Seiford (2009) for two nice reviews of DEA models and a comprehensive
CE

discussion on these models.

AC

33
 ACCEPTED MANUSCRIPT

3.5.2. Some advantages of DEA

As mentioned, DEA models have several privileges. For example, DEA is a non-

T
parametric method and does not require a production function for variables. On the other

IP
hand, DEA can be considered to be providing an approximate of the production function

CR
(Stewart, 1996). Also DEA does not need a DM to set weights to inputs and outputs.

Moreover, DEA does not require to any transformation of the variables into one unit

US
(Shabani et al., 2012). Lohtia et al. (2007) reported that DEA is more precise than
AN
“Regression” technique in determining efficient and inefficient units. On the other hand,

DEA is a very potent method for benchmarking. Lee and Kim (2014) believed that
M

benchmarking for inefficient DMUs is the most important advantage of DEA. DEA
ED

determines a group of efficiency units for an inefficient DMU, with guidelines on how much

must be improved to be efficient (Lee and Kim, 2014). Finally, various DEA models have
PT

been successfully and effectively used in different problems. For example, the reader is
CE

referred to Sotoudeh-Anvari et al. (2016) and references therein for more details.
AC

34
 ACCEPTED MANUSCRIPT

3.5.3. Some disadvantages of DEA

The classical DEA has been usually blamed because of two key disadvantages:

T
1) Low discriminating power

IP
2) Idealistic weight dispensation

CR
Briefly, when the number of variables is high in comparison with the number of units, the

US
majority of DMUs are appraised efficient (Cooper et al., 2007). Also idealistic weight
AN
dispensation takes place when some DMUs are estimated as efficient because they have

tremendously large or small weights for some variables (input and/or output) whereas
M

these extreme values are actually undesirable (Li and Reeves, 1999).
ED

As mentioned earlier, the classical DEA models organize DMUs into two groups: efficient
PT

or inefficient and fail to differentiate among efficient units. But a DM is often interested in a
CE

perfect ranking. In DEA, the amount of variables determines the amount of constraints

(Nunamaker, 1985) and as this amount increases, the efficiency values of DMUs will rise.
AC

On the other words, high numbers of variables against the number of DMUs can reduce the

discriminatory ability in DEA. Several researches pointed out that the poor discrimination

is the key weakness of DEA. In order to solve this problem, various methods such as super-

efficiency techniques, multicriteria DEA approaches, benchmark ranking, cross-efficiency

35
 ACCEPTED MANUSCRIPT

techniques, etc. have been suggested. A survey of ranking techniques in DEA was provided

by Adler et al. (2002). Also Angulo-Meza and Lins (2002) presented a good taxonomy of

various techniques for improving discrimination in DEA. Although these methods have

some advantages, from an operational view, most of these approaches increase the volume

T
of calculations. Also these methodologies are not simple to comprehend for a general DM.

IP
On the other words, most of these approaches are not flawless and in order to suitable use

CR
of them a DM must be completely familiar with DEA field. For example, according to Doyle

and Green (1994), having several solutions to best weights in DEA reduces the utility of

US
cross-efficiency approach. Moreover, cross-efficiency is applicable when the amount of
AN
DMUs is rather low (Li and Reeves, 1999).
M

As mentioned, the poor discrimination may take place when DM has few DMUs to
ED

evaluate in relation to the amount of variables in a given problem. But this case is not the

only one. For example, even where a lot of DMUs and few variables are involved, if a
PT

number of units have an uncommon combination of variables and we employ a radial DEA
CE

model then these DMUs will just be appraised with each other resulting poor

discrimination for those DMUs (Podinovski and Thanassoulis, 2007). It is important to note
AC

that CCR and BCC are known as radial models, namely inputs (in the input-oriented

framework) are proportionally decreased whereas outputs stay fixed. Also there are

various non-radial DEA models such as RAM. For a comprehensive discussion on these

models, please see Cooper et al. (2007) and Cook and Seiford (2009).

36
 ACCEPTED MANUSCRIPT

On the other hand, as mentioned, DEA lets weight flexibility to achieve the best

efficiency. But one of the key flaws of DEA is its excessively much weight flexibility. For

example, a number of significant inputs and/or outputs may be allocated weights of zero.

T
This distribution of weights may be unacceptable because it is not in agreement with DM

IP
judgment (Lins et al., 2007; Bernroider and Stix, 2007; Mecit and Alp, 2013). Although in

CR
DEA the weights are limited, namely the ratio of weighted outputs to weighted inputs does

not transgress 1, there are no a certain limitation on how much weight may be assigned on

US
any variable compared with the others (Liu and Peng, 2008). Therefore, to achieve the
AN
highest efficiency, we can see that a DMU allocates all of its weight on just a few inputs

and/or outputs. To overcome this problem, various weight restriction approaches such as
M

cone ratio, assurance region, etc. have been suggested. Unfortunately, the majority of the
ED

above techniques is complex and can be infeasible. For example, Mecit and Alp (2013)

suggested a new model of restricted DEA by correlation coefficients. But as pointed out by
PT

Unsal and Örkcü (2016), Mecit and Alp’s (2013) model loses its capability for negative
CE

correlations in a number of cases. Moreover, weight restrictions are often assessed on the

basis of value judgments. This issue produces an unsubstantiated extension of the

AC

production technology and as a result generates a serious shortcoming. On the other

words, this approach can nullify the importance of efficiency as an improvement index

(Atici and Podinovski, 2015). Finally, according to Li and Reeves (1999), two

37
 ACCEPTED MANUSCRIPT

aforementioned key drawbacks in DEA are inter-related and often take place at the same

time.

As mentioned, we aim to apply DEA as a weight-independent technique because of its

T
very straightforward framework and intuitive nature for real problems. For example,

IP
Athawale et al. (2011) employed UTA as a weight-independent MCDM method to choose

CR
the best choice in a material selection problem. Although UTA has a relatively sound

framework, this method is not computationally simple. Moreover, this approach is

US
influenced by a supplementary parameter (  ).
AN
As mentioned, various DEA models have been introduced to tackle the disadvantages of
M

classical DEA. But from computational viewpoint, these methods need meticulous
ED

mathematics from DM. Also DM needs to have a very deep background in DEA context. For

example, please see Dyson et al. (2001), Adler et al. (2002), Angulo-Meza and Lins (2002),
PT

Lins et al. (2007), Cooper et al. (2007), Cook and Seiford (2009), Podinovski and
CE

Thanassoulis (2007), Mecit and Alp (2013) and Unsal and Örkcü (2016) for details about

the intricacy as well as difficulty of these methods. Generally spoken, the use of the
AC

aforementioned techniques is not economical and reliable in material selection problem.

38
 ACCEPTED MANUSCRIPT

3.5.4. The rule of thumb in DEA

As mentioned, the high numbers of variables in comparison with the number of units can

lessen the discrimination in DEA. On the other words, according to Cooper et al. (2007), it is

T
advisable that the number of alternatives exceed several times the number of variables to

IP
obtain a reliable outcome. A “rule of thumb” in DEA is to select the number of alternatives

CR
as follows (Cooper et al., 2007):

US
n  max{ m  s , 3  (m  s) } (15)
AN
In Model (15), n is the number of DMUs, m is the number of inputs and s is the number of
M

outputs.
ED

As stated above, according to Li and Reeves (1999), two aforementioned drawbacks in

PT

DEA are inter-related as well as often take place at the same time. On the other hand, in
CE

order to correct use of ranking techniques and weight restriction methods a DM must be

completely familiar with DEA context. Clearly, it is too hard for nearly all general DMs. As a
AC

result, we can deduce that the aforementioned approaches are not applicable in material

selection problem, particularly for a general DM. Therefore, we will try to alleviate those

two flaws of DEA by considering Model (15).

39
 ACCEPTED MANUSCRIPT

3.6. DEA as a MCDM technique

Ramanathan (2006) explained that the aim of DEA is to obtain the efficiency score of a

DMU by comparing how creatively this DMU transforms inputs into outputs. However, as

T
pointed out by Bouyssou (1999), while DEA was suggested for evaluating the efficiency of

IP
DMUs, it is playing a wider role as an MCDM tool now. In the literature, various ways of

CR
using DEA as an MCDM approach have been suggested. As mentioned by Sarkis (2000), a

methodological link is to consider maximizing attributes as outputs, minimizing attributes

US
as inputs and DMUs as alternatives in MCDM and DEA, respectively. It should be noted that
AN
in MCDM context, the attribute whose bigger value is better called maximizing6 criterion

such as quality and the attribute whose lower value is better called minimizing7 criterion
M

such as corrosion rate (Jahan et al., 2011 and Chatterjee et al., 2011). By using this
ED

resemblance, at first Stewart (1996) considered a link between DEA and MCDM. However,

Dyson et al. (2001) addressed various pitfalls of DEA in practical use. These disadvantages
PT

can generate serious pitfalls especially when DEA is applied as MCDM technique. As
CE

mentioned earlier, various techniques have been suggested to tackle these weaknesses.

However, nearly all of the aforementioned methods appear to be difficult as well as

AC

knowledge intensive. For example, Wang and Luo (2006) suggested a combined technique

using DEA and TOPSIS to improve the discrimination of DEA. Then, Mirhedayatian et al.

6
Also known as “beneficial” or “positive” factor
7
Also known as “non-beneficial”, “cost” or “negative” factor
40
 ACCEPTED MANUSCRIPT

(2013) applied Wang and Luo’s method in a material selection problem. But as pointed out

by Wu (2006), Wang and Luo’ method suffers from various drawbacks.

In nearly all MCDM techniques, the reasonable weight is allocated to each decision

T
attribute by DM (using subjective or objective methods) to aggregate several attributes into

IP
a single factor. Obviously, theses weights have a very big effect on the outcomes of MCDM

CR
methods. However, Wu (2006) pointed out that the superiority of DEA is that this

technique needs no aspect of judgment and its decision on efficiency is built just on the

US
data and model. Kou et al. (2014) mentioned that DEA has a relative privilege to MCDM
AN
methods because DEA can determine efficient alternatives with the least prior information.

Also Opricovic and Tzeng (2003) explained the DEA may be a helpful tool in MCDM field,
M

mainly when a DM is not capable to allocate the weights of decision factors at the start of
ED

design. But Doyle (1995) warned that DEA has been recognized as a lazy expert's technique

in MCDM context. Furthermore, as pointed out by Opricovic and Tzeng (2003), often DEA
PT

does not give a set of weights that represent the priority of a DM. On the other words, DEA
CE

is a technique different from MCDM methods and only in some situations can cautiously be

employed as a MCDM tool. This topic will be discussed in the next section.
AC

41
 ACCEPTED MANUSCRIPT

4. The proposed framework

Now in this section, we suggest a MCDM-based procedure using TOPSIS, COPRAS and DEA

for general material selection problem.

T
IP
The assessment process begins from the selection of crucial criteria for a given material

CR
selection problem. More formally, in this step, DM identifies the conditions that a product

must withstand and then converts these properties into required material selection factors

US
(Huang et al., 2011). On the other hand, as pointed out by Das et al. (2016), in material
AN
selection problem, material alternatives can be selected using “Ashby Material Selection

Chart8” (Ashby, 2005). In Ashby’s approach, nearly all mechanical characteristics of a

M

design are considered. Also as stated by Shanian and Savadogo (2006), DM can utilize the
ED

CES9 software, which is presented by Cambridge University and Ashby, for obtaining the

suitable alternatives and their specifications. Although Ashby’s approach is known as one
PT

of the most effective technique for material selection, this approach provides just a list of
CE

candidates to DM to select from. Therefore, in this stage bias may happen. Fortunately, this

problem can be reduced by using MCDM methods. After determining the material selection
AC

criteria, determining the suitable weight for attributes (as explained in sub-section 3.1) and

providing a list of accessible materials, the suggested comprehensive framework using

three MCDM techniques can be employed to choose the best material. As pointed out

8
also known as “Material Properties Chart”
9
http://www.grantadesign.com, Accessed October 2016
42
 ACCEPTED MANUSCRIPT

earlier, no single MCDM approach is preferred as the most appropriate for all forms of

material selection problem. On the other words, we cannot derive that one MCDM

technique is better or more efficient than the others. Thus, many researchers have

suggested that over one technique must be applied to a given problem to achieve a

T
stronger and safer decision. First, in order to find the most suitable MCDM techniques, we

IP
review some broad properties of general material selection problem on the basis of an

CR
extensive literature review as follows:

- The number of options and criteria in material selection problem is usually large.

US
Therefore, choosing the most suitable option has heavy computations. To diminish
AN
the difficulty, the calculation process should be simple and sound (Chatterjee et al.,

2011 and Girubha and Vinodh, 2012).

M

- In material selection problem, the impact of each criterion should often (but not
ED

always) be considered as a trade-off regarding other criteria. Therefore, to handle

this problem we require a compensatory method (Shanian and Savadogo, 2006).

PT

- In material selection problem, both forms of criteria, i.e. benefit-type and cost-type
CE

are used (Jee and Kang, 2000; Jahan et al., 2011; Chauhan and Vaish, 2013;

Yazdani and Payam, 2015).

AC

- In material selection problem, quantitative and qualitative data are usually used

(Liao, 2015; Jahan and Edwards, 2013).

- In material selection problem, complete ranking of materials is often required

(Chatterjee et al. 2011; Singh et al., 2016).

43
 ACCEPTED MANUSCRIPT

As pointed out by Cicek et al. (2010), although numerous MCDM techniques have been

suggested to tackle material selection problems, there is still an inevitable need to apply

proper procedure in agreement with the problem nature. Therefore, on the basis of the

T
general aspects of material selection problem, we can use TOPSIS and COPRAS for solving

IP
this type of problem in general because:

CR
- TOPSIS and COPRAS are transparent methods and have sound logic (Shih et al.,

US
2007; Podvezko, 2011; Das et al., 2012; Nuuter et al., 2015; Mulliner et al., 2013).
AN
- TOPSIS and COPRAS have a very straightforward computation procedure that can

be without difficulty programmed (Shih et al., 2007; Chatterjee et al., 2011;

M

Mulliner et al., 2013; Yazdani and Payam, 2015; Nuuter et al., 2015; Mulliner et al.,
ED

2016).

- TOPSIS and COPRAS can be performed without difficulty when the number of
PT

options and attributes is very large (Shanian and Savadogo, 2006; Jahan et al.,
CE

2010; Chatterjee et al., 2011; Özcan et al., 2011; Mulliner et al., 2013; Yazdani and

Payam, 2015).
AC

- TOPSIS and COPRAS are easily applied for positive (benefit-type) as well as

negative (cost-type) attributes with one procedure (Podvezko, 2011; Chatterjee et

al., 2011; Özcan et al., 2011; Mulliner et al., 2013).

44
 ACCEPTED MANUSCRIPT

- TOPSIS and COPRAS provide a perfect ranking of options (Özcan et al., 2011;

Mulliner et al., 2013; Mulliner et al., 2016).

- TOPSIS and COPRAS are compensatory method (Shanian and Savadogo, 2006;

Kaushik et al., 2015; Mulliner et al., 2016).

T
- TOPSIS and COPRAS are not influenced by any supplementary parameter (Shih et

IP
al., 2007; Antucheviciene et al., 2011; Chatterjee et al. 2011; Mulliner et al., 2016).

CR
- TOPSIS and COPRAS can handle quantitative as well as qualitative factors (Jahan

et al., 2010; Podvezko, 2011; Mulliner et al., 2013).

US
AN
Consequently, in general, TOPSIS and COPRAS can be applied for material selection. On the

other words, these two MCDM techniques can well be matched with real material selection
M

problems. Hence, at first, DMs can apply TOPSIS and COPRAS to general material selection
ED

problem, next compare and evaluate their results and finally make the ultimate decision

easily. Obviously, the suggested approach is very easy to understand and very simple to
PT

execution because it does not need wide familiarity with mathematics or MCDM context.
CE
AC

45
 ACCEPTED MANUSCRIPT

On the other hand, by considering Model (15), classical DEA can be employed as an

auxiliary tool in this filed because (Charnes et al., 1978; Banker et al., 1984; Sarkis, 2000;

Dyson et al. 2001; Angulo-Meza and Lins, 2002; Opricovic and Tzeng, 2003; Ramanathan

2006; Cooper et al., 2007; Lins et al., 2007; Cook and Seiford, 2009; Mecit and Alp, 2013;

T
Atici and Podinovski, 2015; Sotoudeh-Anvari et al., 2016; Yousefi and Hadi-Vencheh,

IP
2016; Sadi-Nezhad and Sotoudeh-Anvari, 2016):

CR
- DEA has a very sound logic.

US
- DEA has a very straightforward computation procedure that can be easily
AN
programmed

- DEA can simply handle benefit criteria (outputs) as well as cost criteria (inputs)
M

with one model. It should be noted that in some MCDM problems there is only
ED

benefit (or cost) criteria. For example, if just benefit criteria are taken into

account, the pure output DEA model without input can easily be employed (Lee
PT

and Kim, 2014).

CE

- DEA is not influenced by any additional parameter.

- DEA can handle quantitative and qualitative variables.

AC

46
 ACCEPTED MANUSCRIPT

But as mentioned above, when using classical DEA as a MCDM tool various problems can

occur. In fact, we use classical DEA only as an auxiliary tool in material selection problem

because using dissimilar sets of weights by each DMU is very deceptive in a ranking

problem. Clearly, this flexibility of weights attenuates the comparison among units on a

T
common foundation. More formally, it is obvious that to make a meaningful comparison of

IP
ranking approaches identical weights must be considered for inputs and outputs (decision

CR
factors), but via DEA each unit is free to select its optimal weights in order to reach the best

efficiency score. Although various methods attempt to obtain a common set of weights in

US
order to achieve comparable results by DEA, similar to ranking methods or weight
AN
restriction techniques, to proper utilization of these methods a DM must be an expert in

DEA field. For example, Liu and Peng (2008) suggested an LP-based approach to find a
M

common set of weights for units. But Ramezani-Tarkhorani et al. (2014) pointed out this
ED

technique has various drawbacks. For example, Liu and Peng’s (2008) approach provided

no assurance to confirm that has a distinctive solution. Also it is vital to mentioned that
PT

although “Software” to use these DEA models may be available, if a DM does not
CE

comprehend what is taking place within a complex model, he/she discerns this approach

like a black box. Consequently, the outcome can be that the DM does not trust in the
AC

suggestion from that approach (Løken, 2007). Hence, we employ DEA in classical form in

material selection problem because traditional DEA model is very simple to understand

and use for general DM.

47
 ACCEPTED MANUSCRIPT

Please bear in mind that to use DEA in this section, instead of organizing the decision

attributes into benefit criteria or cost criteria, we consider them as outputs and inputs,

respectively as depicted in Fig.1.

T
IP
Decision context

CR
MCDM context DEA context

US
Alternatives DMUs

Benefit criteria Outputs

AN
Cost criteria Inputs
Inputs and outputs weights
M

Criteria weights by
by model
DM
ED
PT

Fig. 1. MCDM and DEA in the decision context (Sarkis,

CE

2000)
AC

48
 ACCEPTED MANUSCRIPT

As stated above, unlike MCDM techniques, DM does not provide the criteria weights in

classical DEA. However, the DEA model suggests a group of weights in order to each DMU

achieves its maximum feasible efficiency. Therefore DEA can determine the wasteful DMUs

that have poor performance even when receiving their best imaginable weights (Angulo-

T
Meza and Lins, 2002). However, we cannot conclude that inefficient alternative by DEA

IP
will be the worst alternative by any other method in material selection problem (as will be

CR
shown in Examples 1 and 3). On the other hand, in nearly all MCDM approaches, weight is

determined to each decision factor which depends on its significance to the

US
accomplishment of the system. Needless to say, the ranking of alternatives is very sensitive
AN
to the variations in criteria weights. But because of various backgrounds, the criteria

weights determined differs from DM to DM and this can produce different prioritizing.
M

However, as pointed out by Opricovic and Tzeng (2003), often DEA cannot provide a set of
ED

weights that represent the priority of a DM. Nevertheless, in various real decision

problems, a DM is incapable to weight attributes at the beginning of design due to the lack
PT

of complete information. In such cases, the use of DEA will be very useful. More formally,
CE

the use of DEA can be advantageous by considering Model (15), especially when a DM is not

capable to clarify criteria weights at the start of planning in a material selection problem.
AC

Furthermore, DEA is still helpful in situations where criteria weights are determined by

DM. Clearly, the results from COPRAS and TOPSIS are on the basis of weighted criteria and

the result of DEA is regardless of the DM’s preferences. Therefore, these three outcomes

provide a wider insight for DM about the decision factors and the weights of them. Finally,

49
 ACCEPTED MANUSCRIPT

as pointed out by Wang et al. (2016) and Kou et al. (2014), the more efficient techniques

involved in the decision procedure, the more reasonable the outcomes will be. As a result,

the ranking by three MCDM techniques, namely COPRAS, TOPSIS, and DEA is much more

truthful than ranking of materials by only one MCDM technique. The flowchart of the

T
proposed procedure can be seen in Fig. 2. Noteworthy, in the suggested approach, we use

IP
CRS model of DEA because discrimination of this model is superior to VRS framework in

CR
general (Podinovski and Thanassoulis, 2007). On the other words, any CCR-efficient unit is

BCC-efficient either. Let us point out that TOPSIS, COPRAS and DEA can easily

US
be coded by MATLAB or other software such as Python software.
AN
M
ED
PT
CE
AC

50
 ACCEPTED MANUSCRIPT

Specify the material selection problem

Determine the attributes as well as list the alternative material

T
Form the decision matrix

IP
CR
Assign the attributes weights

US
Obtain the rank of alternatives using COPRAS and TOPSIS
AN
No
The number of choices is greater
than max {m*s, 3*(m+s)}
M

Yes
ED

Use DEA as a auxiliary tool along with TOPSIS and COPRAS

PT

Analyze the results

CE

Obtain final selection decision

AC

Fig. 2. The general proposed framework

51
 ACCEPTED MANUSCRIPT

In some conditions, when a DM cannot exactly weight the criteria and the number of

alternatives and criteria do not satisfy Model (15), DM can use DoE techniques, trial and

error methods, or employ previous experiences to obtain the criteria weights as depicted in

Fig. 3. Then COPRAS and TOPSIS can be used for an initial decision. Nothworthy, DoE is a

T
statistical technique employed to obtain the various effects (independent and interactive)

IP
of different criteria on a system concurrently. For example, in order to minimize the

CR
influence of DM preference in decision process, Wang et al. (2016) applied DoE for weight

identification in MCDM fields. Furthermore, it should be noted that when several material

US
attributes must be considered and the criteria weights are not lucid, identification of the
AN
weights will be fundamentally intuitive. As a result, the reliability of an MCDM method is

extremely diminished. Hence, a DM must be allowed to modify his/her evaluations after

M

he/she obtain the first outcome from the decision procedure. To overcome this difficulty,
ED

some effective techniques such as Dehghan-Manshadi et al. (2007) have been introduced in

material selection problems.

PT
CE
AC

52
 ACCEPTED MANUSCRIPT

Specify the material selection problem

Determine the attributes as well as list the alternative material

T
Form the decision matrix

IP
CR
No
The number of choices is greater
than max {m*s, 3*(m+s)}

US
Yes

Apply DEA as an auxiliary tool and obtain the initial Weight criteria by DoE or trial and error methods
ranking
AN
Analyze the result Apply COPRAS and TOPSIS and provide an initial ranking
M
ED

Obtain initial selection decision on the basis of DM judgment

PT
CE

Fig. 3. The proposed framework when DM cannot weight the attributes at

the start of design

AC

53
 ACCEPTED MANUSCRIPT

To determine the suitability of different MCDM approaches to cope with a material

selection problem, their results are compared by various techniques of non-parametrical

correlation. For example, SRCI is executed for analyzing the statistical significance of

resemblance between different pair of rankings (Antucheviciene et al., 2011; Chatterjee et

T
al., 2011; Athawale and Chakraborty, 2011; Kumar and Ray, 2015 and Dey et al., 2016).

IP
Also PRCI can be estimated between two alternatives which show the linear level of

CR
dependence (Chauhan and Vaish, 2013). Although the intermediate rankings usually

mismatch among the used MCDM techniques, in the majority of material selection problem

US
the top n ranked materials are important. For example, Athawale and Chakraborty (2011)
AN
considered agreement between the top three-ranked choices as a decision index. We refer

to Antucheviciene et al. (2011), Antucheviciene et al. (2012), Chauhan and Vaish (2013),
M

Peng (2015) and Dey et al. (2016) for details on this subject. Since the goal of this study is
ED

to evaluate ranks of options obtained by different MCDM techniques, Athawale and

Chakraborty’s (2011) approach, namely consideration of the top n ranked (based on the
PT

preferences of a DM) alternatives may be useful. It is important to mention that in spite of

CE

the considerable growth in MCDM fields, the validity of the ranking results stay still an

open subject (Hobbs and Meier, 2000). Finally, as can be seen in the suggested model, DM
AC

does not require deep background in MCDM context or mathematics for this decision

procedure.

54
 ACCEPTED MANUSCRIPT

5. Numerical analysis

In this section five real-world examples are presented to exhibit the applicability and

efficiency of the suggested procedures. First of all, the malfunction of classical DEA to give

T
enough discrimination among a small amount of options is revealed by a real material

IP
selection problem.

CR
Example 1: this example, which adopted from Jee and Kang (2000), considers the material

US
selection for a flywheel. Some specifications are necessary for flywheel as follows:
AN
a) Performance indicator
M

b) Fracture toughness
ED

c) Fragmentability (a safety factor)

d) Price
PT
CE

In Table 1, the 10 alternatives and 4 attributes for this material selection problem are

demonstrated. In these decision factors “Performance indicator”, “Fracture toughness” and

AC

“Fragmentability” (shown by (+) in Table 1) are the beneficial criteria where higher scores

are better. Also the price is a non-beneficial criterion (shown by (-) in Table 1) where

smaller rate is preferred. Using Entropy technique weights of the attributes are obtained

wa  0.4, wb  0.3, wc  0.2, and wd  0.1.

55
 ACCEPTED MANUSCRIPT

Table 1

Data (decision matrix) for Example 1

T
No Materials Attributes

IP
CR
rlimit/q (+) KIC/q (+) Fragmentability(+)

Price(-)

US
1 300M 100 8.6125 3 4200
AN
2 2024-T3 49.6454 13.475 3 2100

3 7050-T73651 78.0142 12.5532 3 2100

M

4 Ti-6AL-4 V 108.8795 26.0042 3 10500

ED

5 E glass-epoxy 70 10 9 2735

FRP
PT

6 S glass-epoxy 165 25 9 4095

CE

FRP

7 Carbon–epoxy 440.2516 22.0126 7 35470

AC

FRP

8 Kevlar 29–epoxy 242.8571 28.5714 7 11000

FRP

9 Kevlar 49–epoxy 616.4384 34.2466 7 25000

56
 ACCEPTED MANUSCRIPT

FRP

10 Boron–epoxy 500 23 5 315000

FRP

T
Now, DEA (CCR), ELECTRE II, VIKOR, COPRAS and TOPSIS are exerted to aforementioned

IP
data. In Table 2, the orders of the choices by these techniques are tabulated.

CR
Table 2

US
DEA, VIKOR, TOPSIS and ELECTRE II rankings for Example 1
AN
No Materials Efficiency Ranking by Ranking Ranking Ranking Ranking
M

score DEA (CCR) by by by by

ED

VIKOR TOPSIS ELECTRE COPRAS

II
PT

1 300M 0.59 4 9 9 10 10
CE

2 2024-T3 1 1 10 10 9 9

3 7050-T73651 0.97 2 8 8 8 8
AC

4 Ti-6AL-4 V 0.38 6 6 6 6 7

5 E glass-epoxy 1 1 7 7 7 6

FRP

6 S glass-epoxy 1 1 5 5 3 4

57
 ACCEPTED MANUSCRIPT

FRP

7 Carbon–epoxy 0.3 7 2 2 2 3

FRP

8 Kevlar 29– 0.54 5 4 4 4 5

T
epoxy FRP

IP
9 Kevlar 49– 0.61 3 1 1 1 1

CR
epoxy FRP

10 Boron–epoxy 0.03 8 3 3 5 2

US
FRP
AN
First of all, we can see the number of DMUs and variables do not satisfy Model (15). This
M

event may reduce the discrimination power of DEA in this problem. Clearly, the top-ranked
ED

alternative as produced by COPRAS, VIKOR, TOPSIS and ELECTRE II completely matches.

Mathemathically, the SRCI between TOPSIS and COPRAS is 0.951 and that between TOPSIS
PT

and VIKOR is 1. Noteworthy, SRCI lies between +1 and -1. +1 denotes direct and positive
CE

(perfect agreement) relationship and -1 denotes inverse and negative relationship (perfect

disagreement) between the ranks. Also Karande and Chakraborty (2012) reported a good
AC

correlation between TOPSIS and MOORA in this case. As a result, by considering all MCDM

methods the most appropriate material for flywheel is “Kevlar 49–epoxy FRP”. But when

DEA is used to solve this case, “Kevlar 49–epoxy FRP” (alternative 9) is an inefficient

choice. Also using the majority of MCDM techniques “2024-T3” (alternative 2) is one of the

58
 ACCEPTED MANUSCRIPT

worst options. But according to DEA, this choice is efficient. Obviously, this example shows

that DEA does not work correctly for material selection problem in general. Also as

discussed, DEA is not generally a suitable technique as a screening tool in material

selection. For example, Mehdiabadi et al. (2013) used DEA for screening of options to

T
decrease the choices to a handy number in a material selection. Then, they employed

IP
TOPSIS to prioritize the remaining efficient options. But by Mehdiabadi et al.’s approach

CR
material 9 (the best option in this case) will be omitted in the first step, i.e. in screening

step. Therefore, we can generally deduce that DEA is not useful as a screening tool in

US
MCDM problems. Finally, from Table 2, the intermediate rankings of the options slightly
AN
mismatch among the used MCDM techniques. Fig. 4 compares the ranking of COPRAS and

TOPSIS.
M
ED

TOPSIS

11 COPRAS
10
9
PT

8
7
6
Rank
5
CE

4
3
2
1
AC

0
1 2 3 4 5 6 7 8 9 10
Alternative materials

Fig. 4. Comparative

rankings for example 1

59
 ACCEPTED MANUSCRIPT

Example 2: this example which adopted from Khabbaz et al. (2009), considers the material

selection for a sailing-boat mast. According to Khabbaz et al. (2009) a number of

specifications are necessary for sailing-boat mast as follows:

T
IP
a) Specific strength (benefit factor)

CR
b) Specific modulus(benefit factor)

c) Corrosion resistance (benefit factor)

US
d) Cost category (cost factor)
AN
In Table 3, the 15 alternatives and 4 criteria for this material selection problem are
M

presented.
ED
PT
CE
AC

60
 ACCEPTED MANUSCRIPT

Table 3

Data (decision matrix) for Example 2

No Materials Specific Specific Corrosion Cost

T
strength modulus resistance category

IP
(+) (+) (+) (-)

CR
1 AISI 1020 35.9 26.9 1 5

2 AISI 1040 51.3 26.9 1 5

US
ASTM A242 type
AN
3 1 42.3 27.2 1 5

4 AISI 4130 194.9 27.2 4 3

M

5 AISI 316 25.6 25.1 4 3

ED

AISI 416 heat

6 treated 57.1 28.1 4 3

PT

AISI 431 heat

CE

7 treated 71.4 28.1 4 3

8 AA 6061 T6 101.9 25.8 3 4

AC

9 AA 2024 T6 141.9 26.1 3 4

10 AA 2014 T6 148.2 25.8 3 4

11 AA 7075 T6 180.4 25.9 3 4

12 Ti–6Al–4V 208.7 27.6 5 1

61
 ACCEPTED MANUSCRIPT

Epoxy–70% glass

13 fabric 604.8 28 4 2

Epoxy–63%

14 carbon fabric 416.2 66.5 4 1

T
Epoxy–62%

IP
15 aramid fabric 637.7 27.5 4 1

CR
Using digital logic technique weights of these attributes are obtained as

US
wa  0.333, wb  0.333, wc  0.125, wd  0.208 .
AN
Now, DEA, TOPSIS, COPRAS and VIKOR are applied to this data. The ordering of choices
M

by the aforementioned methods is shown in Table 4.

ED
PT
CE
AC

62
 ACCEPTED MANUSCRIPT

Table 4

DEA and VIKOR rankings for Example 2

No Materials Efficiency Ranking Ranking Ranking Ranking

T
score by by by by

IP
DEA VIKOR COPRAS TOPSIS

CR
(CCR)

1 AISI 1020 0.0809 10 15 15 15

US
2 AISI 1040 0.0809 10 13 13 13
AN
3 ASTM A242 type 1 0.0818 9 14 14 14

4 AISI 4130 0.274 4 5 5 5

M

5 AISI 316 0.271 5 11 12 12

ED

AISI 416 heat 10 10

6 treated 0.275 3 7
PT

AISI 431 heat 9 9

CE

7 treated 0.275 3 6

8 AA 6061 T6 0.1604 8 12 11 11
AC

9 AA 2024 T6 0.16074 7 9 8 8

10 AA 2014 T6 0.1604 8 10 7 7

11 AA 7075 T6 0.16079 6 8 6 6

12 Ti–6Al–4V 1 1 4 4 4

63
 ACCEPTED MANUSCRIPT

Epoxy–70% glass 3 3

13 fabric 0.49 2 3

Epoxy–63% carbon 1 1

14 fabric 1 1 1

T
Epoxy–62% aramid 2 2

IP
15 fabric 1 1 2

CR
First of all, although the number of options, beneficial attributes and non-beneficial

US
attributes satisfy Model (15) ( 15  max 3  (3  1),(3 1) ), 3 out of 15 options have been
AN
estimated effective. Bear in mind that all efficient DMUs have equal score, but as pointed

out earlier, it is not correct to say that they have the same performance in real practice.
M

According to DEA, materials 12, 14 and 15 are efficient and materials 1 and 2 are the last
ED

two-ranked choices. Using COPRAS, TOPSIS, VIKOR and fuzzy logic method (Khabbaz et al.,

2009) “Epoxy-63% carbon fabric” (material 14) is the best and “AISI 1020” (material 1) is
PT

the worst option. Therefore, DEA by considering Model (15) can be used as an auxiliary
CE

tool in material selection problem. However, DEA cannot provide a complete ranking of

materials in this case. Please note that although Model (15) provided a very helpful
AC

guideline for DM who is interested in using DEA, as pointed out by Podinovski and

Thanassoulis (2007), this “rule of thumb” is not valid to all models in use. Let us recall that

according to Angulo-Meza and Lins (2002) and Atici and Podinovski (2015), the

discrimination of DEA depends on the amount of variables with respect to the number of

64
 ACCEPTED MANUSCRIPT

DMUs, the form of returns to scale i.e. CRS and VRS and the data that is under the

examination. Therefore, even with regard to Model (15), DEA only can be used as an

auxiliary tool in material selection problems along with other MCDM methods. Finally, from

Table 4 the results obtained by TOPSIS and COPRAS completely match. It is seen that the

T
top four-ranked and the last three-ranked alternatives perfectly match by VIKOR and

IP
COPRAS. Mathematically, the SRCI between COPRAS and TOPSIS is 1. Also the SRCI

CR
between COPRAS and VIKOR is 0.939 and that between TOPSIS and VIKOR is 0.939. On the

other words, very good correlations between TOPSIS and COPRAS, VIKOR and COPRAS and

US
VIKOR and TOPSIS can be seen in this case.
AN
Example 3: this example which adopted from Yazdani and Payam (2015) is about material
M

selection in low voltage systems. For this selection problem three attributes are considered
ED

E
as Ca  E and Cc  e (as cost-type criterion) and Cb  (as benefit-type criterion)

PT

where E and  are Young’s modulus and the density respectively also  e is the electrical
CE

resistivity. In Table 5, the 14 material alternatives and 3 criteria for this material selection

problem are shown. The weights of these decision factors are subjectively determined by
AC

experts as wa  0.375, wb  0.15, wc  0.475 (Yazdani and Payam, 2015).

65
 ACCEPTED MANUSCRIPT

Table 5

Data (decision matrix) for Example 3

No. Materials E (-) E  e (-)

(+)


T
IP
1 Al 2.82E-

CR
8.37 5.09 08

2 Ag 1.59E-

US
9.11 2.81 08
AN
3 Cr 1.29E-

16.7 6.23 07
M

4 Pd 1.05E-
ED

11 3.17 07

5 Pt 1.05E-
PT

12.96 2.8 07
CE

6 Cu 1.68E-

10.82 3.61 08
AC

7 Sn 1.15E-

7.07 2.61 07

8 Pb 2.08E-

4 1.19 07

66
 ACCEPTED MANUSCRIPT

9 Ti 4.20E-

10.77 5.07 07

10 Ni 6.99E-

13.89 4.66 08

T
11 Au 2.44E-

IP
8.37 1.9 08

CR
12 Co 6.24E-

14.46 4.85 08

US
13 Fe 9.61E-
AN
14.53 5.18 08

14 W 5.28E-
M

20.27 4.62 08
ED

Now, DEA, VIKOR, TOPSIS and COPRAS are applied to this data. The ordering of choices
PT

by them is shown in Table 6.

CE
AC

67
 ACCEPTED MANUSCRIPT

Table 6

DEA, VIKOR and TOPSIS rankings for Example 3

No. Materials Efficiency Ranking by DEA Ranking Ranking Ranking

T
score (CCR) by by by

IP
TOPSIS VIKOR COPRAS

CR
1 Al 1 1 1 1 2

2 Ag 0.849 2 2 2 1

US
3 Cr 0.613 4 12 12 11
AN
4 Pd 0.473 10 8 6 10

5 Pt 0.355 13 10 10 12
M

6 Cu 1 1 4 4 4
ED

7 Sn 0.607 5 7 5 7

8 Pb 0.489 9 13 9 13
PT

9 Ti 0.774 3 14 14 14
CE

10 Ni 0.5516 7 6 7 6

11 Au 0.420 12 3 3 3
AC

12 Co 0.5515 8 5 8 5

13 Fe 0.586 6 9 11 8

14 W 0.462 11 11 13 9

68
 ACCEPTED MANUSCRIPT

First of all, although the number of options, beneficial attributes and non-beneficial

attributes satisfy Model (15) ( 14  max 3  (2  1),(2 1) ), 2 out of 14 options have been

estimated effective. On the other words, similar to previous example, DEA cannot provide a

complete ranking of materials in this case. According to DEA, materials 1 and 6 are efficient.

T
Also using TOPSIS and VIKOR, “Al” (material 1) is the best. However, according to COPRAS,

IP
the best choice is “Ag” (material 2). Clearly, although the results of TOPSIS and COPRAS in

CR
the majority of the alternatives are exactly the same, the best choice is different.

Fortunately, it is observed that this disagreement is not confusing and based on

US
the preference of DM is justifiable. Finally, from Table 6, the SRCI between TOPSIS and
AN
COPRAS is 0.964 and that between TOPSIS and VIKOR is 0.907. Fig. 5 compares the ranking

of COPRAS and TOPSIS.

M
ED

TOPSIS

15 COPRAS
14
13
PT

12
11
10
9
Rank
8
CE

7
6
5
4
3
AC

2
1
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Alternative materials

Fig. 5. Comparative rankings for example 3

69
 ACCEPTED MANUSCRIPT

Example 4: this example which adopted from Yazdani and Payam (2015) is about material

selection in another aspect of the low voltage systems. Two attributes are considered as

f
E (cost-type criterion) and (benefit-type criterion) where E and  f are Young’s
E

T
modulus and failure strength respectively. The weights of attributes are subjectively

IP
determined by Yazdani and Payam (2015) as wa  0.5, wb  0.5 . In Table 7, the 16

CR
alternatives and 2 criteria for this material selection problem are shown.

US
AN
M
ED
PT
CE
AC

70
 ACCEPTED MANUSCRIPT

Table 7

Data (decision matrix) for Example 4

No Materials E () f Efficiency Ranking Ranking Ranking Ranking

()
E

T
score by DEA by by by

IP
(CCR) TOPSIS VIKOR COPRAS

CR
1 Diamond 34.64 0.83 0.00111 16 16 16 16

2 Si3N4 17.97 3.1 0.00802 14 13 14 13

US
3 Al 8.37 4.29 0.02384 9 8 9 9
AN
4 SiO2 8.54 13.7 0.07460 5 5 5 6

5 Polymmide 2.83 5 0.08216 4 7 7 3

M

6 Ti 10.77 4.31 0.01861 11 10 11 11

ED

7 Ni 13.89 2.59 0.00867 13 12 12 12

8 PVDF 1.52 21.74 0.66514 2 2 2 1

PT

9 Au 8.37 4.29 0.02385 8 8 9 8

CE

10 Ni–Fe 10.95 13.33 0.05661 7 6 6 7

11 SiC 21.21 4.44 0.00974 12 14 13 14

AC

12 Al2O3 16.58 7.27 0.02039 10 11 8 10

13 Quartz 10.34 15.89 0.07147 6 4 4 5

14 W 20.27 1.7 0.00390 15 15 15 15

15 PMMA 1.55 33.33 1.00000 1 1 1 2

71
 ACCEPTED MANUSCRIPT

16 Si 12.56 25 0.09256 3 3 3 4

Now DEA, COPRAS, TOPSIS, and VIKOR are applied to aforementioned data. As can be

seen in Table 7, the top three-ranked and the last two-ranked materials by DEA, TOPSIS

T
and VIKOR completely match. Mathematically, we can see a high SRCI between DEA and

IP
TOPSIS (0.966), DEA and VIKOR (0.969), and very high correlation between DEA and

CR
COPRAS (0.982) as well as VIKOR and TOPSIS (0.979). Obviously, in this case, the number

of attributes and alternatives satisfy Model (15). Hence, we can see DEA works very

US
reliable as a MCDM tool in this example. Consequently, on the basis of these four cases and
AN
Cooper et al.’s (2007) recommendation, it is advisable that the number of alternatives

exceed several times the number of criteria TOPSIS

to obtain the reliable results using DEA as a
M

COPRAS
DEA

MCDM tool for material selection problem. However, as mentioned several times, the
ED

results of DEA can just be interpreted with great caution. Fig. 6 compares the ranking of

COPRAS, TOPSIS and DEA.

PT
CE
AC

72
 ACCEPTED MANUSCRIPT

17
16
15
14
13
12
11
10
Rank 9
8
7

T
6
5
4

IP
3
2
1
0

CR
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Alternative materials

US
Fig. 6. Comparative rankings for example 4
AN
Example 5: this example which adopted from Dehghan-Manshadi et al. (2007) is about
M

material selection in a cryogenic tank. Seven attributes are considered as follows:

ED

(a) Toughness indicator

PT

(b) Yield strength

CE

(c) Young’s modulus

(d) Density
AC

(e) Thermal expansion index

(f) Thermal conductivity

(g) Specific heat

73
 ACCEPTED MANUSCRIPT

The attributes weights are determined by digital logic technique as

wa  0.214, wb  0.1555, wc  0.119, wd  0.190, we  0.155, w f  0.083, wg  0.083 . In

Table 8, the 7 alternatives and 7 criteria for this material selection problem are shown.

T
Table 8

IP
Data for example 5

CR
N Materia toughne Yield Young’s Density Thermal Thermal Specif

o ls ss strength( modulus(
US
(-) expansi ic
AN
index(+) +) +) on index Conductivit heat

(-) y(-) (-)

M

1 Al 75.5 420 74.2 2.8 21.4 0.37 0.16

ED

2024-

26
PT

2 Al5052 95 91 70 2.68 22.1 0.33 0.16

CE

-0

3 SS 301- 770 1365 189 7.9 16.9 0.04 0.08

AC

FH

4 SS 310- 187 1120 210 7.9 14.4 0.03 0.08

3AH

5 Ti–6Al– 179 875 112 4.43 9.4 0.016 0.09

74
 ACCEPTED MANUSCRIPT

4V

6 Inconel 239 1190 217 8.51 11.5 0.31 0.07

718

7 70Cu– 273 200 112 8.53 19.9 0.29 0.06

T
30Zn

IP
CR
Now, various methods such as DEA, COPRAS, TOPSIS, Dehghan-Manshadi et al.’s (2007)

approach, and MOORA are employed to aforementioned data. Using Dehghan-Manshadi et

US
al.’s (2007) approach, the ordering of the materials is generated as 3-5-4-6-7-1-2. COPRAS
AN
method provides a ranking as 3-5-6-4-7-1-2. Also TOPSIS gives a ranking as 3-6-4-5-7-1-2.

The ordering of 3-7-6-4-5-2-1 is obtained by MOORA. The SRCI between Dehghan-

M

Manshadi et al.’s (2007) approach and TOPSIS is very high (0.964), and that between
ED

Dehghan-Manshadi et al.’s (2007) approach and COPRAS is 0.857. Also The SRCI between

TOPSIS and COPRAS is 0.892 and that between TOPSIS and MOORA is 0.785. Clearly, the
PT

best option is material 3, i.e. “SS 301-FH” by all aforementioned MCDM methods. However,
CE

using DEA, 5 out of 7 alternatives, i.e.1, 3, 4, 5, and 6 have been estimated effective. Clearly,

alternative 3 is among them. According to Model (15), this outcome is not unexpected. But
AC

although according to DEA, alternative 1 is an efficient option, according to all MCDM

methods, this choice is among the worst alternatives. Again, we can see DEA is not a

genuine tool for screening and ranking in material selection problem in general. Finally,

although Karande and Chakraborty (2012) reported a good correlation between TOPSIS

75
 ACCEPTED MANUSCRIPT

and MOORA in a material selection problem, the SRCI between these two techniques is low

in this case. Fig. 7 compares the ranking of COPRAS and TOPSIS.

TOPSIS
8
COPRAS
7

T
6

IP
5
Rank 4

CR
3
2
1

US
0
1 2 3 4 5 6 7
Alternative materials
AN
M

Fig. 7. Comparative rankings for example 5

ED

5.1. A discussion
PT
CE

Athawale et al. (2011) claimed that for obtaining the best decision, always weight-

independent MCDM approaches are preferred. Also Wu (2006) mentioned that the
AC

excellence of DEA is that this technique needs no aspect of judgment and its decision is

built just on the data. However, we believe that both these claims are questionable. As

stated, unlike the majority of MCDM techniques such as VIKOR and TOPSIS, in DEA an

expert does not weight the inputs and outputs and the model weights the variables so that

76
 ACCEPTED MANUSCRIPT

all DMUs obtain the highest efficiency. On the other words, it can be fitting the big weights

to inputs or outputs with less significance and fitting the small weights to key inputs or

outputs. Let us point out that the prior information about the significance of the inputs and

outputs can be considered into the some DEA models. Reader is referred to Angulo-Meza

T
and Lins (2002) for details about this subject. However, most of these methods are

IP
complex to understand and implementation. Moreover, often the aforementioned

CR
approaches are not flawless. As pointed out before, choosing a good material for a specific

purpose is a very expensive process. In addition, the application of the extended DEA

US
methods to improve discrimination is not easy too. Fortunately, the simplicity of suggested
AN
approach guides a general DM in making a practical decision without requiring extensive

expertise.
M
ED

On the other hand, as pointed out by Rao and Patel (2010), screening is one of the

very important steps in the material selection problem. It should be noted that the term
PT

“screening” and “preprocess” are not the same in MCDM context. Screening techniques
CE

reduce the alternatives to an operational number for subsequent assessment. Needless to

say, after screening usually another approach should be executed to obtain a full ranking.
AC

But preprocess operation is carried out only for initial insight or more insight about the

criteria weights and their approximate impacts in final ranking. Although Opricovic and

Tzeng (2003) claimed that DEA can be generally used as a screening tool and as a

preprocess tool in MCDM, we demonstrated in Examples 1 and 5 that DEA does not work

77
 ACCEPTED MANUSCRIPT

appropriately as screening tool in general. But by considering Model (15), DEA can be

suitable for the initial insight or preprocess. Also by considering Model (15), DEA can be

useful in material selection along with other techniques such as TOPSIS and COPRAS.

T
Finally, it should be noted that as pointed out by Sarkis (2000), there is no admitted

IP
technique for making a comparison between DEA and other MCDM methods and the

CR
usefulness of DEA can just be assessed on how well it correlates with the MCDM methods

used.

US
AN
6. Conclusion
M

In spite of MCDM methods capabilities, development of a comprehensive MCDM-based

ED

framework for handling material selection problem is a challenge. In this paper, we

reviewed the recent MCDM studies on material selection problems as a brief continuation
PT

of previous review in this field. From the literature, it was cleared that MCDM framework
CE

can suitably be matched with the majority of real material selection problems. On the other

hand, it is a very popular approach that in complicated materials selection problem more
AC

than one MCDM methods to be applied for the best decision. On the basis of an extensive

literature review, TOPSIS and COPRAS were selected as two very efficient MCDM methods

for solving general material selection problem due to its outstanding features and rational

correlation with the previous studies. On the other hand, we examined the use of DEA as

78
 ACCEPTED MANUSCRIPT

MCDM tool in material selection problem. According to this study, on the basis of the

famous rule of thumb, DEA can be employed as an auxiliary tool to tackle a material

selection problem. Based on this paper, DM is recommended to employ TOPSIS and

COPRAS for material selection problem in general. Also DM can use classical DEA by

T
considering a rule of thumb as an auxiliary tool, particularly when he/she is not capable to

IP
assign criteria weights at the start of design. Application of uncertainty theories as well as

CR
utilization of target value factors in the suggested procedure can be various themes of

future study.

US
AN
Acknowledgements
M

This research has been financially supported by Kerman Branch, Islamic Azad University,
ED

Kerman, Iran. The authors greatly appreciate for this support. Also the authors are very

grateful to the editor and all the anonymous referees for their constructive comments,
PT

which have been very helpful in improving this work.

CE
AC

79
 ACCEPTED MANUSCRIPT

References

[1] L. Abdullah, C.W.R. Adawiyah, Simple additive weighting methods of multi criteria

decision making and applications: A decade review, International Journal of Information

T
Processing and Management, 5 (2014) 39-49

IP
CR
[2] N. Adler, L. Friedman, Z. Sinuany-Stern, Review of ranking methods in the data

envelopment analysis context, European Journal of Operational Research 140 (2002) 249-

US
265
AN
[3] L. Angulo-Meza, M.P.E. Lins, Review of methods for increasing discrimination in data
M

envelopment analysis, Annals of Operations Research 116(2002)225–242

ED

[4] J. Antucheviciene, A. Zakarevicius, E.K. Zavadskas, Measuring Congruence of Ranking

PT

Results Applying Particular MCDM Methods, Informatica 22 (2011) 319–338

CE

[5] J. Antucheviciene, E.K. Zavadskas, A. Zakarevičius, Ranking redevelopment decisions of

AC

derelict buildings and analysis of ranking results, Economic Computation and Economic

Cybernetics Studies and Research 46 (2012) 37–62

80
 ACCEPTED MANUSCRIPT

[6] K. Anupam, P. S. Lal, V. Bist, A.K. Sharma, V. Swaroop, Raw material selection for pulping

and papermaking using TOPSIS multiple criteria decision making design, Environmental

Progress & Sustainable Energy 33 (2014) 1034-1041

T
[7] K. Anupam, V Swaroop, A.K Sharma, P. S. Lal, V. Bist, Sustainable raw material selection

IP
for pulp and paper using SAW multiple criteria decision making design, IPPTA Journal 27

CR
(2015) 67-76

US
[8] A. Ardeshirilajimi, A. Aghanouri, A. Abedian, A. Milani, An exponential placement
AN
method for materials selection, The International Journal of Advanced Manufacturing

Technology 78 (2015) 641–650

M
ED

[9] M.F. Ashby, Materials Selection in Mechanical Design, 3rd edn, Pergamon Press, New

York (2005)
PT
CE

[10] V. M. Athawale, S. Chakraborty, A comparative study on the ranking performance of

some multi-criteria decision-making methods for industrial robot selection, International

AC

Journal of Industrial Engineering Computations 2(4) (2011) 831-850

81
 ACCEPTED MANUSCRIPT

[11] V.M Athawale, R. Kumar, S. Chakraborty, Decision making for material selection using

the UTA method, The International Journal of Advanced Manufacturing Technology 57

(2011) 11–22

T
[12] K.B Atici, V. V Podinovski, Using data envelopment analysis for the assessment of

IP
technical efficiency of units with different specializations: An application to agriculture,

CR
Omega 54 (2015) 72-83

US
[13] R.D. Banker, A. Charnes and W.W. Cooper, Some models for estimating technical and
AN
scale inefficiency in data envelopment analysis, Management Science 30 (1984) 1078–1092
M

[14] E. Bernroider, V. Stix, A method using weight restrictions in data envelopment analysis
ED

for ranking and validity issues in decision making, Computers & Operations Research 34

(2007) 2637 – 2647

PT
CE

[15] D. Bouyssou, Using DEA as a tool for MCDM: some remarks, Journal of the Operational

Research Society 50 (1999) 974-978

AC

[16] H. Çalışkan, B. Kurşuncu, C. Kurbanoğlu, Ş.Y. Güven, Material selection for the tool

holder working under hard milling conditions using different multi criteria decision

making methods, Materials & Design 45 (2013) 473-479

82
 ACCEPTED MANUSCRIPT

[17] N. Caterino, I. Iervolino, G. Manfredi, Comparative analysis of multi‐criteria

decision‐making methods for seismic structural retrofitting, Computer-Aided Civil and

Infrastructure Engineering 24 (2009) 432–445

T
[18]B. Ceballos, MT Lamata, DA Pelta, A comparative analysis of multi-criteria decision-

IP
making methods, Progress in Artificial Intelligence 5 (2016) 315-322

CR
[19] K. Cicek, M. Celik, Y.I. Topcu, An integrated decision aid extension to material selection

US
problem, Materials & Design 31 (2010) 4398-4402
AN
[20] A. Charnes, W.W. Cooper and E.L. Rhodes, Measuring the efficiency of decision making
M

units, European Journal of Operational Research 2 (1978) 429–444

ED

[21] P. Chatterjee, V.M Athawale, S. Chakraborty, Materials selection using complex

PT

proportional assessment and evaluation of mixed data methods, Materials & Design 32
CE

(2011) 851-860
AC

[22] P. Chatterjee, S. Chakraborty, Material selection using preferential ranking methods,

Materials & Design 35 (2012) 384-393

83
 ACCEPTED MANUSCRIPT

[23] A. Chauhan, R. Vaish, Hard coating material selection using multi-criteria

decision making, Materials & Design 44 (2013) 240-245

[24] W.D. Cook, L.M. Seiford, Data envelopment analysis (DEA) – Thirty years on, European

T
Journal of Operational Research 192 (2009) 1-17

IP
CR
[25] W.W. Cooper, L.M. Seiford, K. Tone, Data Envelopment Analysis: A Comprehensive Text

with Models, Applications, References and DEA-Solver Software, 2nd ed., Springer, Berlin

US
(2007)
AN
[26] V. P. Darji, R. V. Rao, Intelligent Multi Criteria Decision Making Methods for Material
M

Selection in Sugar Industry, Procedia Materials Science 5 ( 2014 ) 2585 – 2594

ED

[27] D. Das, S. Bhattacharya, B. Sarkar, Decision-based design-driven material selection: A

PT

normative-prescriptive approach for simultaneous selection of material and geometric

CE

variables in gear design, Materials & Design 92 (2016) 787-793

AC

[28] A. Debnath, P. Bhattacharjee, S. Chakraborty, Selection of optimal aluminum

alloy using TOPSIS method under fuzzy environment, Journal of Intelligent & Fuzzy

Systems (2016) DOI: 10.3233/JIFS-161049

84
 ACCEPTED MANUSCRIPT

[29] B. Dehghan-Manshadi, H. Mahmudi, A. Abedian, A novel method for materials selection

in mechanical design: Combination of non-linear normalization and a modified digital logic

method, Materials & Design 28 (2007) 8–15

T
[30] B. Dey, B. Bairagi, B. Sarkar, S.K Sanyal, Multi objective performance analysis: A novel

IP
multi-criteria decision making approach for a supply chain, Computers & Industrial

CR
Engineering 94 (2016) 105-124

US
[31] J. R. Doyle, Multiattribute choice for the lazy decision maker: Let the alternatives
AN
decide, Organizational Behavior and Human Decision Processing 62 (1995) 87-100
M

[32] J. R. Doyle, R. Green, Efficiency and cross-efficiency in data envelopment analysis:

ED

Derivatives, meanings and uses, Journal of the Operational Research Society 45 (1994),

567– 578.
PT
CE

[33] R.G. Dyson, R. Allen, A.S Camanho, V.V Podinovski, Pitfalls and protocols in DEA,

European Journal of Operational Research 132 (2001) 245-259

AC

[34] R.J. Girubha, S. Vinodh, Application of fuzzy VIKOR and environmental impact analysis

for material selection of an automotive component, Materials & Design 37 (2012) 478-486

85
 ACCEPTED MANUSCRIPT

[35] A. Hatami-Marbini, M. Tavana, M. Moradi, F. Kangi, A fuzzy group Electre method for

safety and health assessment in hazardous waste recycling facilities, Safety Science 51

(2013) 414–426

T
[36] B.F. Hobbs, P. Meier, Energy Decisions and the Environment: A Guide to the Use of

IP
Multicriteria Methods. Boston: Kluwer Academic Publishers (2000)

CR
[37] H. Huang, L. Zhang, Z. Liu, J.W Sutherland, Multi-criteria decision making and

US
uncertainty analysis for materials selection in environmentally conscious design, The
AN
International Journal of Advanced Manufacturing Technology 52 (2011) 421–432
M

[38] J.J Huang, G.H Tzeng, H.H Liu, A revised VIKOR model for multiple criteria decision
ED

making-The Perspective of Regret Theory, Cutting-Edge Research Topics on Multiple

Criteria Decision Making (2009) 761-768

PT
CE

[39] C. L. Hwang, K. Yoon, Multiple attribute decision making: Methods and applications,

New York: Springer-Verlag (1981).

AC

[40] M. İpek, İ.H. Selvi, F. Findik, O. Torkul, I.H. Cedimoğlu, An expert system based material

selection approach to manufacturing, Materials & Design 47 (2013) 331-340

86
 ACCEPTED MANUSCRIPT

[41] N. M Ishak, S. D Malingam, M. R Mansor, Selection of natural fibre reinforced

composites using fuzzy VIKOR for car front hood, International Journal of Materials

and Product Technology 53 (2016) DOI: 10.1504/IJMPT.2016.079205

T
[42] A. Jahan, M.Y. Ismail, S.M. Sapuan, F. Mustapha, Material screening and choosing

IP
methods–a review, Materials & Design 31 (2010) 696-705

CR
[43] A. Jahan, F. Mustapha, M.Y Ismail, S.M Sapuan, A comprehensive VIKOR method for

US
material selection, Materials & Design 32 (2011) 1215-1221
AN
[44] A. Jahan, M. Bahraminasab, K.L Edwards, A target-based normalization technique for
M

materials selection, Materials & Design 35 (2012) 647-654

ED

[45] A. Jahan, K. L Edwards, VIKOR method for material selection problems with interval
PT

numbers and target-based criteria, Materials & Design 47 (2013) 759-765

CE

[46] A. Jahan, K.L Edwards, A state-of-the-art survey on the influence of normalization

AC

techniques in ranking: improving the materials selection process in engineering design,

Materials & Design 65 (2015) 335-342

87
 ACCEPTED MANUSCRIPT

[47] D.H. Jee, K.J. Kang, A method for optimal material selection aided with decision making

theory, Materials & Design 21 (2000) 199-206

[48] P. Karande, S. Chakraborty, Application of multi-objective optimization on the basis of

T
ratio analysis (MOORA) method for materials selection, Materials & Design 37 (2012) 317-

IP
324

CR
[49] P. Karande, S.K Gauri, S. Chakraborty, Applications of utility concept and desirability

US
function for materials selection, Materials & Design 45 (2013) 349–358
AN
[50] P. Kaushik, S. Tyagi, B.P Thapliyal, K. Anupam; An approach for selection of coating
M

color formulation using TOPSIS Multi Criteria Decision Making Design, TAPPSA Journal 6
ED

(2015) 20-25
PT

[51] R. S. Khabbaz, B.D. Manshadi, A. Abedian, R. Mahmudi, A simplified fuzzy logic

CE

approach for materials selection in mechanical engineering design, Materials & Design 30

(2009) 687-697
AC

[52] R. H Kouchaksaraei, S.H. Zolfani, M. Golabchi, Glasshouse locating based on SWARA-

COPRAS approach. International Journal of Strategic Property Management, 19(2015) 111-

122

88
 ACCEPTED MANUSCRIPT

[53] G. Kou, Y. Peng, G. Wang, Evaluation of clustering algorithms for financial risk analysis

using MCDM methods, Information Sciences 275 (2014) 1–12

[54] R Kumar, A Ray, Optimal selection of material: an eclectic decision, Journal of The

T
Institution of Engineers (India): Series C 96 (2015) 29-33

IP
CR
[55] R. Kumar, S.K. Singal, Penstock material selection in small hydropower plants using

MADM methods, Renewable and Sustainable Energy Reviews 52 (2015) 240–255

US
AN
[56] A. Kumar, B Sah, A.R Singh, Y. Deng, X. He, a review of multi criteria decision making

(MCDM) towards sustainable renewable energy development, Renewable and Sustainable

M

Energy Reviews 69 (2017) 596–609

ED

[57] H. Lee, C. Kim, Benchmarking of service quality with data envelopment analysis,
PT

Expert Systems with Applications 41 (2014) 3761–3768

CE

[58] X. B Li, G.R Reeves, A multiple criteria approach to data envelopment analysis,
AC

European Journal of Operational Research 115 (1999) 507-517

[59] T.W. Liao, Two interval type 2 fuzzy TOPSIS material selection methods, Materials &

Design 88 (2015) 1088-1099

89
 ACCEPTED MANUSCRIPT

[60] M.P.E. Lins, A.CM da Silva, C.A.K Lovell, Avoiding infeasibility in DEA models with

weight restrictions, European Journal of Operational Research 181 (2007) 956–966

[61] F.H.F. Liu, H.H. Peng, Ranking of units on the DEA frontier with common weight,

T
Computers & Operations Research 35 (2008) 1624–1637

IP
CR
[62] H. C. Liu, L.X. Mao, Z.Y. Zhang, P. Li, Induced aggregation operators in the VIKOR

method and its application in material selection, Applied Mathematical Modelling 37

US
(2013) 6325-6338
AN
[63] H. C Liu, J.X You, L. Zhen, X. J Fan, A novel hybrid multiple criteria decision making
M

model for material selection with target-based criteria, Materials & Design 60 (2014) 380-
ED

390
PT

[64] R. Lohtia, N. Donthu, I. Yaveroglu, Evaluating the efficiency of Internet banner

CE

advertisements, Journal of Business Research 60 (2007) 365–370

AC

[65] E. Løken, Use of multicriteria decision analysis methods for energy planning problems,

Renewable and Sustainable Energy Reviews 11 (2007) 1584–1595

90
 ACCEPTED MANUSCRIPT

[66] F. H. Lotfi, R. Fallahnejad, Imprecise Shannon’s entropy and multiple attribute decision

making, Entropy 12 (2010) 53–62

[67] S. R. Maity, S. Chakraborty, Grinding wheel abrasive material selection using fuzzy

T
TOPSIS method, Materials and Manufacturing Processes, 4(2013) 408–417

IP
CR
[68] S. R Maity, S. Chakraborty, Tool steel material selection using PROMETHEE II method,

The International Journal of Advanced Manufacturing Technology 78 (2015) 1537–1547

US
AN
[69] K. Maniya, M.G. Bhatt, A selection of material using a novel type decision-making

method: preference selection index method, Materials & Design 31 (2010) 1785-1789
M
ED

[70] M.R Mansor, S.M Sapuan, E.S Zainudin, A.A Nuraini, Hybrid natural and glass fibers

reinforced polymer composites material selection using Analytical Hierarchy Process for
PT

automotive brake lever design, Materials & Design 51 (2013) 484-492

CE

[71] E. D. Mecit, I. Alp, A new proposed model of restricted data envelopment analysis by
AC

correlation coefficients, Applied Mathematical Modelling 37 (2013) 3407–3425

[72] A. Mehdiabadi, A. Rohani, S. Amirabdollahiyan, Ranking industries using a hybrid of

dea-topsis, Decision Science Letters 2 (2013) 251–256

91
 ACCEPTED MANUSCRIPT

[73] K. Mela, T. Tiainen, M. Heinisuo, Comparative study of multiple criteria decision

making methods for building design, Advanced Engineering Informatics 26 (4) (2012)

716–726.

T
[74] A. S. Milani, A. Shanian, C. Lynam, T. Scarinci, An application of the analytic network

IP
process in multiple criteria material selection, Materials & Design 44 (2013) 622-632

CR
[75] I. Millet, B. Schoner, Incorporating negative values into the analytic hierarchy process,

US
Computers and Operations Research 32 (2005) 3163–3173.
AN
[76] S.M Mirhedayatian, S.E Vahdat, M.J Jelodar, R.F Saen, Welding process selection for
M

repairing nodular cast iron engine block by integrated fuzzy data envelopment analysis and
ED

TOPSIS approaches, Materials & Design 43 (2013) 272-282

PT

[77] E. Mulliner, K. Smallbone, V. Maliene, An assessment of sustainable housing

CE

affordability using a multiple criteria decision making method, Omega 41 (2013) 270–279
AC

[78] E. Mulliner, N. Malys, V. Maliene, Comparative analysis of MCDM methods for the

assessment of sustainable housing affordability, Omega 59 (2016) 146-156

92
 ACCEPTED MANUSCRIPT

[79] T. Nunamaker, Using data envelopment analysis to measure the efficiency of non-

profit organizations: A critical evaluation, Managerial and Decision Economics 6 (1985)

50–58

T
[80] T. Nuuter, I. Lill, L. Tupenaite, Comparison of housing market sustainability in

IP
European countries based on multiple criteria assessment, Land Use Policy 42 (2015) 642–

CR
651

US
[81] S. Opricovic, G.H Tzeng, Comparing DEA and MCDM method, Multi-Objective
AN
Programming and Goal Programming (2003) 227-232
M

[82] D. Paul, P. Agarwal, G. Mondal, D. Banerjee, A comparative analysis of different hybrid

ED

MCDM techniques considering a case of selection of 3D printers, Management Science

Letters 5 (2015) 695-708

PT
CE

[83] T. Özcan, N. Çelebi, Ş. Esnaf, Comparative analysis of multi-criteria decision making

methodologies and implementation of a warehouse location selection problem, Expert

AC

Systems with Applications 38 (2011) 9773-9779

[84] Y. Peng, Regional earthquake vulnerability assessment using a combination of MCDM

methods, Annals of Operations Research 234 (2015) 95-110

93
 ACCEPTED MANUSCRIPT

[85] A. H. Peng, X.M. Xiao, Material selection using PROMETHEE combined with analytic

network process under hybrid environment, Materials & Design 47 (2013) 643-652

[86] V.V Podinovski, E. Thanassoulis, Improving discrimination in data envelopment

T
analysis: Some practical suggestions, Journal of Productivity Analysis 28 (2007) 117–126

IP
CR
[87] V. Podvezko, The comparative analysis of MCDA methods SAW and COPRAS,

Engineering Economics 22(2011) 134-146

US
AN
[88] R. Ramanathan, Data envelopment analysis for weight derivation and aggregation in

the analytic hierarchy process, Computers & Operations Research 33 (2006) 1289–1307
M
ED

[89] R. Ramanathan, L.S. Ganesh, Energy resource allocation incorporating qualitative and

quantitative criteria: an integrated model using goal programming and AHP, Socio-
PT

Economic Planning Sciences 29(1995) 197–218

CE

[90] S. Ramezani-Tarkhorani, M. Khodabakhshi, S. Mehrabian, F. Nuri-Bahmani, Ranking

AC

decision-making units using common weighs in DEA, Applied Mathematical Modelling, 38

(2014) 3890-3896

94
 ACCEPTED MANUSCRIPT

[91] R. V. Rao, A material selection model using graph theory and matrix approach,

Materials Science and Engineering 431 (2006) 248–255

[92] R.V. Rao, B.K. Patel, A subjective and objective integrated multiple attribute decision

T
making method for material selection, Materials & Design 31 (2010) 4738-4747

IP
CR
[93] M.K. Rathod, H.V. Kanzaria, A methodological concept for phase change material

selection based on multiple criteria decision analysis with and without fuzzy environment,

US
Materials & Design 32 (2011) 3578-3585
AN
[94] S. Sadi-Nezhad, A. Sotoudeh-Anvari, A new Data Envelopment Analysis under
M

uncertain environment with respect to fuzziness and an estimation of reliability,

ED

OPSEARCH 53 (2016) 103-115

PT

[95] N. Sakundarini, Z. Taha, S.H. Abdul-Rashid, Multi-objective optimization for

CE

high recyclability material selection using genetic algorithm, The International

Journal of Advanced Manufacturing Technology 68 (2013) 1441-1451

AC

[96] J. Sarkis, A comparative analysis of DEA as a discrete alternative multiple criteria

decision tool, European Journal of Operational Research 123 (2000) 543-557

95
 ACCEPTED MANUSCRIPT

[97] B. Sen, P. Bhattacharjee, U. K Mandal, A comparative study of some prominent multi

criteria decision making methods for connecting rod material selection, Perspectives in

Science 8 (2016) 547—549

T
[98] A. Shabani, R.F. Saen, S.M.R. Torabipour, A new benchmarking approach in Cold Chain,

IP
Applied Mathematical Modelling 36 (2012), 212–224

CR
[99] A. Shanian, O. Savadogo, TOPSIS multiple-criteria decision support analysis for

US
material selection of metallic bipolar plates for polymer electrolyte fuel cell, Journal of
AN
Power Sources 159 (2006) 1095–1104
M

[100] P. Sharma, N. Gupta, Investigation on material selection for gate dielectric in

ED

nanocrystalline silicon (nc-Si) top-gated thin film transistor (TFT) using Ashby's, VIKOR

and TOPSIS, Journal of Materials Science: Materials in Electronics 26 (2015) 9607-9613

PT
CE

[101] H.S Shih, H.J Shyur, E. S Lee, An extension of TOPSIS for group decision making,

Mathematical and Computer Modelling 45 (2007) 801–813

AC

[102] T. Singh, A. Patnaik, R. Chauhan, Optimization of tribological properties of cement

kiln dust-filled brake pad using grey relation analysis, Materials & Design 89 (2016) 1335-

1342

96
 ACCEPTED MANUSCRIPT

[103] A. Sotoudeh-Anvari, S. Sadi-Nezhad, A new approach based on the level of reliability

of information to determine the relative weights of criteria in fuzzy TOPSIS, International

Journal of Applied Decision Sciences 8 (2) (2015) 164–178.

T
[104] A. Sotoudeh-Anvari, E. Najafi, S. Sadi-Nezhad, A new data envelopment analysis in

IP
fully fuzzy environment on the base of the degree of certainty of information, Journal of

CR
Intelligent & Fuzzy Systems 30 (2016) 3131–3142

US
[105] T. J. Stewart, Relationships between data envelopment analysis and multicriteria
AN
decision analysis, Journal of the Operational Research Society 47 (1996) 654-665
M

[106] X. Sun, Y. Li, an Intelligent Multi-Criteria Decision Support System for Systems Design,
ED

10th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference (2010) 1-

11. http://dx.doi.org/10.2514/6.2010-9222
PT
CE

[107] F. Tao, LN Bi, Y Zuo, AYC Nee, A hybrid group leader algorithm for green

material selection with energy consideration in product design, CIRP Annals-

AC

Manufacturing Technology 65 (2016) 9-12

97
 ACCEPTED MANUSCRIPT

[108] P. C Tewari, U. Prakash, D. Khanduja, Ranking of Sintered Material for High

Loaded Automobile Application by Applying Entropy-TOPSIS Method, Materials

Today: Proceedings 2 ( 2015 ) 2375 – 2379

T
[109] E. Triantaphyllou, Multi-Criteria Decision Making Methods: A Comparative Study,

IP
Kluwer Academic Publishers, Dordrecht (2013)

CR
[110] E. Turanoglu Bekar, M. Cakmakci, C. Kahraman Fuzzy COPRAS method for

US
performance measurement in total productive maintenance: a comparative
AN
analysis, Journal of Business Economics and Management 17 (2016)663-684
M

[111] M. G Unsal, H. H Örkcü, A note for weight restriction in the ARIII model based on
ED

correlation coefficients, Applied Mathematical Modelling 40 (2016) 10820–10827

PT

[112] G. Vats, R. Vaish, Piezoelectric material selection for transducers under fuzzy
CE

environment, Journal of Advanced Ceramics 2(2013) 141–148

AC

[113] G. Vats, R. Vaish, C. R Bowen, Selection of ferroelectric ceramics for

transducers and electrical energy storage devices, International Journal of Applied

Ceramic Technology 12 (2015) 1-7

98
 ACCEPTED MANUSCRIPT

[114] H. Voogd, Multicriteria evaluation for urban and regional planning, Pion Ltd, (1983)

[115] Y.M Wang, Y. Luo, DEA efficiency assessment using ideal and anti-ideal decision

making units, Applied Mathematics and Computation 173 (2006) 902-915

T
IP
[116] P. Wang, Z. Zhu, Y. Wang, A novel hybrid MCDM model combining the SAW, TOPSIS

CR
and GRA methods based on experimental design, Information Sciences 345 (2016) 27–45

US
[117] D. Wu, A note on DEA efficiency assessment using ideal point: An improvement of
AN
Wang and Luo’s model, Applied Mathematics and Computation 183 (2006) 819–830
M

[118] Y.P.O Yang, HM Shieh, JD Leu, G.H Tzeng, A novel hybrid MCDM model combined with
ED

DEMATEL and ANP with applications, International Journal of Operations Research

5(2008) 160-168
PT
CE

[119] M. Yazdani, A.F. Payam, A comparative study on material selection of

microelectromechanical systems electrostatic actuators using Ashby, VIKOR and TOPSIS,

AC

Materials & Design 65 (2015) 328-334

[120] A. Yousefi, A. Hadi-Vencheh, Selecting Six Sigma projects: MCDM or DEA?, Journal of

Modelling in Management 11 (2016) 309 - 325

99
 ACCEPTED MANUSCRIPT

[121] M. Yousefpour, A. Rahimi, Characterization and selection of optimal parameters to

achieve the best tribological performance of the electrodeposited Cr nanocomposite

coating, Materials & Design 54 (2014) 382-389

T
[122] S. H. Zanakis, A. Solomon, N. Wishart, S. Dublish, Multi-attribute decision making: a

IP
simulation comparison of select methods, European Journal of Operational Research,

CR
107(3) (1998) 507–529

US
[123] E.K. Zavadskas, A. Kaklauskas, V. Sarka, The new method of multicriteria complex
AN
proportional assessment of projects, Technological and Economic Development of

Economy 1 (1994) 131–139

M
ED

[124] E.K Zavadskas, Z. Turskis, S. Kildienė, State of art surveys

of overviews on MCDM/MADM methods, and Economic Development of Economy, 20

PT

(2014) 165-179
CE
AC

100
 ACCEPTED MANUSCRIPT

TOPSIS and COPRAS as the

best multi-criteria decision
making (MCDM) techniques
for material selection
problem.

T
IP
Classical DEA as an auxiliary General material selection problem
tool for material selection under MCDM paradigm

CR
problem

US
By considering n  max m  s,3  (m  s) , n is the
number of alternatives, m is the number of inputs and
AN
s is the number of outputs.
M

TOPSIS
17 COPRAS
16 DEA
15
14
ED

13
12
11
10
Rank 9
8
PT

7
6
5
4
3
2
CE

1
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Alternative materials
AC

Comparative rankings for a real case

Graphical abstract

101
 ACCEPTED MANUSCRIPT

Highlights

 TOPSIS and COPRAS are chosen as the best multi-criteria decision making (MCDM)

techniques for ranking the alternative materials in general practice.

T
 An excellent agreement between COPRAS and TOPSIS method is observed in this

IP
field.

CR
 DEA can be used in material selection problem as an auxiliary tool by considering a

classic rough rule of thumb, which represents the relation between the number of

US
alternatives and the number variables.
AN
M
ED
PT
CE
AC

102