Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Materials and Design 35 (2012) 647–654

Contents lists available at SciVerse ScienceDirect

Materials and Design


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

A target-based normalization technique for materials selection


Ali Jahan a,⇑, Marjan Bahraminasab a, K.L. Edwards b
a
Faculty of Engineering, Semnan Branch, Islamic Azad University, Semnan, Iran
b
School of Engineering & Applied Science, Aston University, Aston Triangle, Birmingham B4 7ET, UK

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

Article history: Ranking and selection of the optimal material is an important stage in the engineering design process.
Received 23 July 2011 However, most of the methods proposed for ranking in materials selection have tended to focus on cost
Accepted 2 September 2011 and benefit criteria, with target values receiving much less attention in spite of their importance in many
Available online 8 September 2011
practical decision-making problems such as selecting materials to best match the properties of human
tissue in biomedical engineering applications. In response to this perceived gap, the development of a
Keywords: new normalization technique is considered in this paper that provides an extension of the Technique
H. Selection of materials
for Order Preference by Similarity to Ideal Solution (TOPSIS) method and objective weighting in materials
H. Weighting and ranking factors
H. Performance indices
selection. There are four example cases included to validate the accuracy of outcomes from the proposed
model. It is believed that the proposed decision-making model is suitable for linking to material dat-
abases and has the potential to enhance the efficiency of computer-aided materials selection systems.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction engineering design. For instance, high specific stiffness and


strength materials such as polymer composites are used for
Material selection is an integral part of the design process [1,2] advanced sports equipment.
that has interdependency with the associated materials and man- (4) Developing, comparing and ranking alternatives [1,3,6,7], and
ufacturing processes [3,4]. It can radically alter the final design and using soft material requirements to further narrow the field
not only affect the efficiency of production systems and customer of possible candidates down to a few promising materials.
satisfaction, but also impact on environmental issues. Generally, (5) Searching for supporting information about the top-ranked
the materials selection process consists of the following five steps: candidate materials [3], evaluation and decision for the opti-
mal solution [1,2,5,6], and finally verification tests [2].
(1) Definition of the design [2].
(2) Analyzing and translating the design requirements (expressed The theories and methods of formalized material selection have
as constraints and objectives) into material performance indi- been well reviewed by many researchers [8,10–18]. Those research-
ces/properties [2,3,5,6]. ers who have developed multi-criteria decision-making (MCDM)
(3) Comparing the properties required with a materials property methods in material selection have focused on facilitating steps 4
database in order to select a few candidate materials that look and 5 above. These studies include either developing the novel
promising for the application [1–3,5–8]. Screening is based on MCDM techniques or the application of the current ones in ranking
the idea of classifying performance requirements into ‘rigid’ [19–24], normalization [25,26], and weighting process [25,27–29] of
and ‘soft’ requirements [9]. Rigid requirements can be used material selection. In MCDM methods, multiple attribute decision-
for the initial selection of materials to eliminate any unsuitable making (MADM) approaches are more popular for material
groups. For instance, metallic materials are eliminated when selection, for which, the so-called Technique for Order Preference
selecting materials for an electrical insulator. Soft require- by Similarity to Ideal Solution (TOPSIS) [30] is the most prevalent
ments are normally subjected to negotiation and compromise one [13].
and therefore involve more complicated decision-making. Most MCDM studies in the field of materials selection have
Examples of requirements in this group consist of mechanical concentrated on cost criteria, in which smaller values are always
properties, physical properties, and cost, which are normally preferable and benefit criteria whose higher values are desirable.
compared on the basis of their relative importance to an However, target value criteria [31], which are important in different
engineering design applications, have received much less attention.
These type of criteria are particularly significant in biomedical
⇑ Corresponding author. Tel.: +98 2313322034. applications where implant materials should possess similar
E-mail addresses: iranalijahan@yahoo.com, a.jahan@semnaniau.ac.ir (A. Jahan). properties to those of human tissues in order to match the original

0261-3069/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.matdes.2011.09.005
648 A. Jahan et al. / Materials and Design 35 (2012) 647–654

form and function [32]. Furthermore, there exist many examples of Table 2 compares the proposed technique with non-monotonic
target values, or benchmarks in other applications of MCDM [33]. and linear-max–min normalization methods using a sample bene-
The use of MCDM is particularly important when the application is fit and cost criterion. It is clear from both criteria C1 and C2 that all
complex or advanced, or the use of leading edge technology is in- normalized values in non-monotonic technique are either near to 1
volved. Intuition alone is insufficient in all but the simplest of cases, or zero, and it means that this method cannot properly normalize
and the support provided by targeted MCDM is beneficial to obtain the data. It also shows that the result of normalized values in the
the optimal choice of material. It is envisaged that the technique will target-based normalization approach is the same as those for the
be particularly useful when the materials and/or application are new linear max–min method.
and experience of past performance is not readily available.
Although comprehensive VIKOR [31] was developed to address 2.1. Improving the current weighting models using the suggested
this issue, there is still a need to enhance the capability of MCDM normalization technique
methods based on the specific nature of material selection problems
[34]. However, recently some methods [28,35] have been proposed Weighting methods, which try to define the importance of cri-
for objective weighting of criteria in material selection that either do teria, are categorized into subjective, objective, and integrated
not have any provision for engineering design states in which satis- methods. The subjective methods determine the weight of attri-
fying the target values are desirable or suffer from shortcomings. As butes solely based on preference information of attributes given
a consequence, this has prompted the development of TOPSIS and by expert evaluation. Thus, it can be according to the previous
enhancing the capability of objective weighting methods in material experience and designer’s preferences. The objective methods ob-
selection through a new normalization technique. tain the weights only based on the known data of the problem
and are useful when a decision-maker is nonexistent. Although
2. Proposed target-based normalization method there is no lack of subjective weighting techniques in material
selection [25,29,38,39], there is a need for objective weighting,
A MCDM problem with finite possibilities can be expressed in especially when material selection is based on target values of cri-
a matrix format. It includes possible alternatives (materials) Ai teria. For instance, the recently proposed objective weighting
(i = 1, . . . , m), which decision makers have to choose, criteria methods by Maniya and Bhatt [35] do not have any provision for
(material properties) cj (j = 1, . . . , n), relative importance of criteria weighting when target values of criteria are important. Further-
(or weights) wj, and elements of xij which is the rating of alterna- more, the suggested weighting technique by Rao and Patel [28]
tive i with respect to criterion j as shown in Table 1. which is based on standard deviation (Eq. (3)), may lead to mis-
Normalization deal with cost and benefit criteria in most of leading due to lack of the normalization process.
MADM techniques to obtain comparable scales. Shih et al. [36] re- X
n
viewed normalization methods in TOPSIS [30] including vector wj ¼ rj rj j ¼ 1; . . . ; n ð3Þ
dimensionless technique, three types of linear dimensionless j¼1
technique, and non-monotonic normalization. Milani et al. [37]
compared vector normalization with four types of linear dimension- This is shown in Table 3 which demonstrates criterion of density
less techniques in TOPSIS. Among the current dimensionless tech- with two scales. It is clear that the same data with different scales
niques in TOPSIS, non-monotonic normalization [36] can cover makes large differences on calculating of standard deviation and
target values of criteria as well as cost and benefit criteria, but in this consequently on objective weighting of criteria. It proves the neces-
technique, standard deviation has been used in the normalization sity of normalizing data based on Eq. (1). It means standard devia-
formula which can yield misleading results. tion must be calculated based on Eq. (4) before applying Eq. (3) to
Thus, a novel technique is proposed as follow (Eq. (1)), for both obtain objective weights.
objective weighting of criteria and ranking of materials using TOPSIS sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pm 2
method, to cover situations in which approaching the target values i¼1 ðr ij  r j Þ
are vital in decision making process.
rj ¼ j ¼ 1; . . . ; n ð4Þ
m
jxij  T j j
r ij ¼ 1  n o n o ð1Þ
Max xij ; T j  Min xmin
max
ij ; T j 2.2. Extended model of TOPSIS according to suggested
normalization technique
where Tj is either the most favorable element (xij) or the target value
in criteria j. The proposed normalization method can be called TOPSIS technique is based on the principle that the optimal
extension of linear max–min method (Eq. (2)) [36]. point should have the shortest distance from the positive ideal solu-
8 tion (PIS) and the farthest from the negative ideal solution (NIS).
> xij xmin
> ij
< xmax xmin for benefit criteria
ij ij
Therefore, this method is suitable for risk avoidance designer(s), be-
r ij ¼ ð2Þ cause the designer(s) might like to have a decision which not only
>
> xmax
ij
xij
: xmax xmin for cost criteria makes as much profit as possible, but also avoids as much risk as
ij ij
possible [40]. TOPSIS has been used predominantly in materials
selection due to its superior characteristics [13], therefore in this
paper, this technique has been developed for the selection of mate-
Table 1
A typical multiple attribute decision problem.
rials based on target values of criteria. The proposed steps for the
extended version of TOPSIS can be expressed as follows:
w1 w2 ... wn
C1 C2 ... Cn
(1) Convert the raw measures xij into the standardized mea-
A1 x11 x12 ... x1n sures rij according to the proposed normalization technique
A2 x21 x22 ... x2n
A3 x31 x32 ... x3n
in Eq. (1).
.. .. .. .. (2) Develop a set of importance weights (wj) for criteria.
. . . .
(3) Multiply the columns of the normalized decision matrix by
Am xm1 xm2 ... xmn
the associated weights:
A. Jahan et al. / Materials and Design 35 (2012) 647–654 649

Table 2
Comparing the proposed technique with non-monotonic and linear-max–min normalization methods using a sample benefit and cost criterion.

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


C1(Benefit, T = 629) 271 251 244 269 629 295 306 292 267 232 204 237
Non-monotonic 0.01 0.00 0.00 0.00 1.00 0.01 0.01 0.01 0.00 0.00 0.00 0.00
Linear-max–min 0.16 0.11 0.09 0.15 1.00 0.21 0.24 0.21 0.15 0.07 0.00 0.08
Target-based 0.16 0.11 0.09 0.15 1.00 0.21 0.24 0.21 0.15 0.07 0.00 0.08
C2 (Cost, T = 0.3) 0.7 0.7 0.7 0.7 0.9 0.75 0.75 0.75 0.75 0.7 0.65 0.3
Non-monotonic 0.02 0.02 0.02 0.02 0.00 0.01 0.01 0.01 0.01 0.02 0.04 1.00
Linear-max–min 0.33 0.33 0.33 0.33 0.00 0.25 0.25 0.25 0.25 0.33 0.42 1.00
Target-based 0.33 0.33 0.33 0.33 0.00 0.25 0.25 0.25 0.25 0.33 0.42 1.00

Table 3
Necessity of normalization before calculating of standard deviation in Rao and Patel method [28].

Mat 1 Mat 2 Mat 3 Mat 4 Mat 5 Standard deviation (r)


Without normalization Density (kg/m3) 5 70 220 1600 3400 1463.4
Density (gr/cm3) 0.005 0.07 0.22 1.6 3.4 1.4634
After normalization Density (kg/m3) 0 0.02 0.06 0.47 1 0.43
Density (gr/cm3) 0 0.02 0.06 0.47 1 0.43

V ij ¼ r ij wj ; j ¼ 1; 2; 3; . . . ; n; i ¼ 1; 2; 3; . . . ; m ð5Þ ples are considered. The first example, which includes the target
value of criteria, illustrates the improved weighing procedure
(4) Identify the PIS.
  and extended version of TOPSIS in material selection. Examples
fV þ1 ; V þ2 ; V þ3 ; . . . ; V þn g ¼ ðMax V ij ji ¼ 1; . . . ; m 2–3 attempt to prove the validity of the extended TOPSIS and the
i
last example shows how the new method overcomes the weak
(5) Identify the NIS. point of the original TOPSIS.
 
fV 1 ; V 2 ; V 3 ; . . . ; V n g ¼ ðMin V ij ji ¼ 1; . . . ; m
i
3.1. Example 1
(6) Develop a distance measure for each alternative to both
ideal (D+) and nadir (D) using the following equations: Electrical insulating materials offer a very high resistance to the
!0:5 flow of electric current so that they can be used to confine the cur-
X
n
Dþi ¼ ðV ij  V þj Þ2 i ¼ 1; 2; 3; . . . ; m ð6Þ rent in its proper path along the conductor. The functions of insu-
j¼1 lators are diverse and depend on the design requirements and
service conditions. The selection of insulation materials for com-
!0:5 puter cables is considered, with the main requirements for this
X
n
Di ¼ ðV ij  V j Þ2 i ¼ 1; 2; 3; . . . ; m ð7Þ selection summarized. In this study, the cost criteria are dielectric
j¼1 constant, relative cost and dissipation factor, the benefit criteria
are dielectric strength and volume resistance and finally the target
(7) Calculate the relative closeness to the ideal solution accord- value is thermal expansion (2.3  105 K) criterion due to compat-
ing to the following equation: ibility requirements. Table 4 shows the candidate materials, crite-
Di ria, and objectives of designer [43].
Ci ¼ ; i ¼ 1; 2; 3; . . . ; m; 0 < Ci < 1 ð8Þ
Di þ Dþi In this example, the details of the proposed method are illus-
(8) Rank alternatives by maximizing the ratio in Step 7. The lar- trated step by step as shown below:
ger the index value, the better the performance of the Step (1) The normalized data according to proposed linear nor-
alternative. malization method are shown in Table 5. The most favorable val-
Therefor the TOPSIS method is extended by developing a meth- ues of criteria are 78000, 1  1018, 0.0002, 2.1, 2.3  105, and 1,
odology for solving material selection problems when approaching respectively.
target values of criteria. Also, steps 4 and 5 in the extended TOPSIS Step (2) Subjective weight was extracted from Farag [43] and
are simplified compared to the original TOPSIS, because of the objective weight was obtained according to Eqs. (1), (3), and (4).
advantages of the proposed normalization procedure. The deci- Table 6 shows the final weight, which is the combination of subjec-
sion-making method suggested in this paper seems to be suitable tive and objective weights based on the average function.
for linking to spreadsheet software and might help to address the Step (3) Multiply the columns of the normalized decision matrix
limitation of computer-aided materials selection systems [13] by the associated weights.
and enhance efficiency of digital tools [41] in material selection. Step (4) Identify the PIS.
Empowering materials databases with decision-making tools are
V þ ¼ ð0:1847; 0:2758; 0:1444; 0:1148; 0:0662; 0:1892Þ
of significance because the available set of materials is rapidly
growing both in type and number [42]. Step (5) Identify the NIS.

V  ¼ ð0; 0; 0; 0; 0; 0; 0Þ
3. Verification of the new normalization method
Steps (6), (7), and (8): The distance from the ideal and nadir
The rankings of the selected materials generally have been sup- ideal alternatives, the relative closeness to the ideal solution, and
ported with those obtained by the past researchers. Therefore, in the ranking orders of materials by the proposed method are dem-
order to express and validate the proposed approach, four exam- onstrated in Table 7.
650 A. Jahan et al. / Materials and Design 35 (2012) 647–654

Table 4
Decision matrix for insulating materials (example 1).

Objectives of design Max Max Min Min Target value Min


No. Materials Dielectric strength Volume resistance Dissipation factor (60 Hz) Dielectric constant Thermal expansion Relative cost
(V/mm) (ohm/cm) (60 Hz) (105/°C)
1 PTFE 14,820 1  1018 0.0002 2.1 9.5 4.5
2 CTFE 21,450 1  1018 0.0012 2.7 14.4 9
3 ETFE 78,000 1  1016 0.0006 2.6 9 8.5
4 Polyphenylene oxide 20,475 1  1017 0.0006 2.6 6.5 2.6
5 Polysulfone 16,575 1  1014 0.0010 3.1 5.6 3.5
6 Polypropylene 21,450 1  1016 0.0005 2.2 8.6 1

Table 5
Normalized data for example 1.

No. Materials Dielectric strength Volume resistance Dissipation factor Dielectric constant Thermal expansion Relative cost
(V/mm) (ohm/cm) (60 Hz) (60 Hz) (105/°C)
1 PTFE 0.0000 1.0000 1.0000 1.0000 0.4050 0.5625
2 CTFE 0.1049 1.0000 0.0000 0.4000 0.0000 0.0000
3 ETFE 1.0000 0.0099 0.6000 0.5000 0.4463 0.0625
4 Polyphenylene oxide 0.0895 0.0999 0.6000 0.5000 0.6529 0.8000
5 Polysulfone 0.0278 0.0000 0.2000 0.0000 0.7273 0.6875
6 Polypropylene 0.1049 0.0099 0.7000 0.9000 0.4793 1.0000

Table 6
Subjective, objective and combined weights for insulating materials (example 1).

Dielectric strength Volume resistance Dissipation factor Dielectric constant Thermal expansion Relative cost
(V/mm) (ohm/cm) (60 Hz) (60 Hz) (105/°C)
Subjective weight 0.2 0.33 0.13 0.07 0.07 0.2
Objective weight 0.1694 0.2215 0.1588 0.1596 0.1122 0.1784
Final weight 0.1847 0.2758 0.1444 0.1148 0.0911 0.1892

The result obtained by the proposed technique is almost in predominantly used for the verification of different MCDM meth-
agreement with ranking orders of materials using comprehensive ods in material selection [19,20,25,26,29]. Low temperature tanks
VIKOR. Farag solved the same problem earlier using the limits on are used in many industrial applications when storage and trans-
properties (LOP) approach [9,43]. The revised and right formula portation of a liquid may be preferred to a gas. Candidate materials
of LOP [31] is presented here (Eq. (9)) again to correct typing mis- for the storage tank in addition to having good weld-ability and
take in the previous research. Where Lj, Uj and Tj are lower limit, process-ability, lower density, specific heat, smaller thermal expan-
upper limit, and target values of criteria, respectively. sion coefficient, thermal conductivity, and adequate toughness at
nl
X  
Lj X nu
r ij X
nt  Minðrij ; T j Þ  the intended operating temperature should be sufficiently strong
mi ¼ wj þ wj þ wj   1 i ¼ 1; 2; 3; . . . ; m and stiff. The list of candidate materials, criteria, weights and objec-
j
r ij j
U j j
Maxðr ij ; T j Þ
tives are shown in Table 8. In order to make a comparison between
ð9Þ the proposed model and other methods, the same weights are con-
Furthermore, LOP is based on simple additive weighting (SAW) sidered by those others applied.
technique, while Simanaviciene and Ustinovichius [44] showed that The normalized data according to the proposed linear normali-
TOPSIS is more sensitive than SAW method to the values of criteria. zation method are presented in Table 9. The distance from the ideal
Accordingly, it can be highlighted that the LOP and its corrected ver- and nadir ideal alternatives, and the relative closeness to the ideal
sion are not as reliable as the extended TOPSIS method. solution, are demonstrated in Table 10.
It is observed from Table 10 that material 3 (SS 301-FH) has the
3.2. Example 2 highest score and is thus ranked first. The second best alternative is
material 5 (Ti–6Al–4 V) and the last ranked alternative is material
This example is a well-known case of a cryogenic storage tank 7 (70Cu–30Zn). This result shows 89% (Spearman’s rank correlation
for low temperature application (Fig. 1), that has been coefficient) agreement with those obtained by Rao and Davim [29]

Table 7
The results of the suggested method and comprehensive VIKOR in example 1.

No. Materials Dþ
i
D
i Proposed method Comprehensive VIKOR
Closeness to ideal solution Rank Qi (m = 0.5) Rank
1 PTFE 0.2045 0.3504 0.6315 1 0.0000 1
2 CTFE 0.3052 0.2802 0.4787 2 0.3838 2
3 ETFE 0.3366 0.2161 0.3910 4 0.8325 5
4 Polyphenylene oxide 0.3130 0.1957 0.3846 5 0.7591 4
5 Polysulfone 0.3719 0.1489 0.2859 6 1.0000 6
6 Polypropylene 0.3231 0.2429 0.4291 3 0.7410 3
A. Jahan et al. / Materials and Design 35 (2012) 647–654 651

store kinetic energy in automobiles, urban subway trains, mass


transit buses, wind-power generators, etc. The performance index
of rlimit/q (where, rlimit is the fatigue limit of the material and q
is the material density), fracture toughness (KIC) of the material
(performance measure of failure due to brittle fracture), the frag-
mentability of the flywheel material, and price per unit mass are
the criteria which have been considered in this case. Table 11
shows the decision matrix including criteria, objectives of criteria,
candidate materials, and weights (importance of criteria). The nor-
malized values are presented in Table 12. The distance from the
ideal and nadir ideal alternatives and the closeness to the ideal
solution are shown in Table 13.
This problem earlier was solved by Chatterjee et al. [46] using
the ELECTER and VIKOR methods, and by Jahan et al. [22] using
the linear assignment technique. They only applied subjective
weight for ranking, while Jee and Kang [45] used combination of
subjective and objective weights. In order to have an exact com-
parison, the same weight are considered by those Jee and Kang
[45] applied. Table 13 compares ranking orders of materials by
extended TOPSIS, original TOPSIS, and TOPSIS with linear-sum nor-
malization approach [45]. Spearman’s rank correlation coefficient
shows 95% agreement between the extended TOPSIS and result
of TOPSIS with linear-sum normalization method. Also ranking
orders of alternatives by extended TOPSIS are exactly the same
Fig. 1. Typical cryogenic storage tank for liquefied nitrogen gas.
as original TOPSIS, which demonstrates high efficiency of the pro-
posed model.
in original TOPSIS, which suggests that the proposed method can
be successfully applied to material selection problems. The most 3.4. Example 4
striking advantage that emerges from the above example is that
the proposed method can also be used successfully in problems Using this example, it is shown that the proposed target-based
that only consider cost and beneficial attributes. normalization method will improve the decision-making process,
while in the original TOPSIS approach the vector normalization
3.3. Example 3 technique was widely used [36,47,48]. Four alternatives and two
criteria, which have the same importance, are considered in this
The third example deals with the selection of the most suitable example (Table 14). According to original TOPSIS, A3 is the best alter-
material for the design of a flywheel [45], which is a device used to native and after that A4, A2 and A1. However, this is clearly wrong.

Table 8
Decision matrix for material selection of storage tank (example 2).

Objectives Max Max Max Min Min Min Min


Weight 0.28 0.14 0.05 0.24 0.19 0.05 0.05
No. Materials Toughness Yield strength Young’s Density Thermal Thermal Specific
indexa (MPa) modulus (GPa) g/cm3 expansionb conductivityc heatd
1 Al 2024-T6 75.5 420 74.2 2.8 21.4 0.37 0.16
2 Al 5052-O 95 91 70 2.68 22.1 0.33 0.16
3 SS 301-FH 770 1365 189 7.9 16.9 0.04 0.08
4 SS 310–3AH 187 1120 210 7.9 14.4 0.03 0.08
5 Ti–6Al–4 V 179 875 112 4.43 9.4 0.016 0.09
6 Inconel 718 239 1190 217 8.51 11.5 0.31 0.07
7 70Cu–30Zn 273 200 112 8.53 19.9 0.29 0.06
a
Toughness index, TI, is based on UTS, yield strength YS, and ductility e, at 196 °C. T = (UTS + YS) e/2.
b
Thermal expansion coefficient is given in 106/°C. The values are averaged between RT and 196 °C.
c
Thermal conductivity is given in cal/cm2/cm/°C/s.
d
Specific heat is given in cal/g/°C. The values are averaged between RT and 196 °C.

Table 9
Normalized data for example 2.

No. Materials Toughness index Yield strength (MPa) Young’s modulus (GPa) Density g/cm3 Thermal expansion Thermal conductivity Specific heat
1 Al 2024-T6 0.00 0.258 0.029 0.979 0.055 0 0
2 Al 5052-O 0.03 0 0 1 0 0.113 0
3 SS 301-FH 1.00 1 0.81 0.108 0.409 0.9322 0.8
4 SS 310–3AH 0.16 0.808 0.952 0.108 0.606 0.9605 0.8
5 Ti–6Al–4 V 0.15 0.615 0.286 0.701 1 1 0.7
6 Inconel 718 0.24 0.863 1 0.003 0.835 0.1695 0.9
7 70Cu–30Zn 0.28 0.086 0.286 0 0.173 0.226 1
652 A. Jahan et al. / Materials and Design 35 (2012) 647–654

Table 10
Result of suggested method in example 2.

No. Materials Dþ
i
D
i Closeness to ideal solution Ranking orders
Proposed method Original TOPSIS
1 Al 2024-T6 0.358884 0.238076 0.398814 4 5
2 Al 5052-O 0.369369 0.240213 0.394062 5 6
3 SS 301-FH 0.242186 0.33186 0.578107 1 1
4 SS 310–3AH 0.328041 0.18683 0.362868 6 4
5 Ti–6Al–4 V 0.257291 0.278404 0.519706 2 2
6 Inconel 718 0.324957 0.221014 0.404809 3 3
7 70Cu–30Zn 0.376937 0.10102 0.211358 7 7

Table 11
Decision matrix for material selection of flywheel (example 3).

Objectives of design Max Max Max Min


Weight 0.296369 0.059602 0.02135 0.622679
No. Materials rlimit/q KIC/q Fragmentability Price/mass
1 300 M 100 8.6125 3 4200
2 2024-T3 49.6454 13.4752 3 2100
3 7050-T73651 78.0142 12.5532 3 2100
4 Ti-6AL-4 V 108.8795 26.0042 3 10,500
5 E glass–epoxy FRP 70 10 9 2735
6 S glass–epoxy FRP 165 25 9 4095
7 Carbon-epoxy FRP 440.2516 22.0126 7 35,470
8 Kevlar 29-epoxy FRP 242.8571 28.5714 7 11,000
9 Kevlar 49-epoxy FRP 616.4384 34.2466 7 25,000
10 Boron-epoxy FRP 500 23 5 315,000

Table 12 a problem solving process [5]. The specification normally contains


Normalized data for example 3. the essential requirements for a design, with materials’ properties
No. Materials rlimit/q KIC/q Fragmentability Price/mass forming an integral part for mechanical products. These require-
1 300 M 0.09 0.00 0.00 0.99
ments are translated, where possible, into quantifiable targets,
2 2024-T3 0.00 0.19 0.00 1.00 e.g. stiffness, strength, etc. that must be met for an effective design
3 7050-T73651 0.05 0.15 0.00 1.00 solution [3]. However, these targets often conflict with each other,
4 Ti-6AL-4 V 0.10 0.68 0.00 0.97 leading to complexity and compromise in decision-making and as
5 E glass–epoxy FRP 0.04 0.05 1.00 1.00
a consequence lead to sub-optimal solutions [11]. This paper has
6 S glass–epoxy FRP 0.20 0.64 1.00 0.99
7 Carbon-epoxy FRP 0.69 0.52 0.67 0.89 critically investigated the plethora of mathematical techniques,
8 Kevlar 29-epoxy FRP 0.34 0.78 0.67 0.97 some implemented successfully as computer-based tools, for sup-
9 Kevlar 49-epoxy FRP 1.00 1.00 0.67 0.93 porting decision-making. These techniques and tools, used on their
10 Boron-epoxy FRP 0.79 0.56 0.33 0.00 own or together, have limitations for solving materials selection
problems. In materials selection, there is a fundamental need to
The reason is that obviously A4 should be better than A3 since the to- be able to compare several disparate material property targets
tal utility level of A4 is higher than that of A3. It shows that the TOPSIS simultaneously [18]. The proposed target-based normalization
method with vector normalization cannot always distinguish be- technique helps to address this problem directly and therefore im-
tween alternatives correctly while the extended one with target- proves the potential for achieving more optimal solutions.
based normalization method demonstrates the true rank. A decision matrix of materials that includes cost, benefit, and
target criteria must be normalized in order to make different ele-
4. Discussion ments of the decision-making matrix comparable. The comprehen-
sive VIKOR [31] was developed to overcome the shortcomings of
It has been seen that target setting and use is a natural part of the LOP technique [9,43] for addressing target criteria in materials
decision-making in the problem solving process. Designing is such selection. Although comprehensive VIKOR can be used effectively

Table 13
Result of suggested method in example 3.

No. Materials Dþ
i
D
i Closeness to ideal solution Ranking orders
Extended TOPSIS Original TOPSIS TOPSIS with linear-sum normalization [45]
1 300 M 0.2771 0.6170 0.6901 6 6 5
2 2024-T3 0.3010 0.6228 0.6741 9 9 9
3 7050-T73651 0.2869 0.6229 0.6847 7 7 7
4 Ti-6AL-4 V 0.2689 0.6061 0.6927 5 5 6
5 E glass–epoxy FRP 0.2901 0.6232 0.6824 8 8 8
6 S glass–epoxy FRP 0.2381 0.6208 0.7228 4 4 3
7 Carbon-epoxy FRP 0.1183 0.5917 0.8334 2 2 4
8 Kevlar 29-epoxy FRP 0.1971 0.6143 0.7571 3 3 2
9 Kevlar 49-epoxy FRP 0.0442 0.6534 0.9367 1 1 1
10 Boron-epoxy FRP 0.6265 0.2366 0.2741 10 10 10
A. Jahan et al. / Materials and Design 35 (2012) 647–654 653

Table 14 [5] Dieter GE. Engineering design: a materials and processing approach. Boston,
Comparing ranking orders of alternatives in original TOPSIS and extended one USA: McGraw-Hill; 1983.
(example 4). [6] Van Kesteren IEH, Kandachar PV, Stappers PJ. Activities in selecting materials
by product designers. In: Proceedings of the international conference on
Objectives Max Max Results advanced design and manufacture. Harbin, China; 2006.
[7] Jalham IS. Decision-making integrated information technology (IIT) approach
Weight 0.5 0.5
for material selection. Int J Comput Appl Technol 2006;25:65–71.
Alternatives C1 C2 Original TOPSIS Extended TOPSIS [8] Giaccobi S, Kromm FX, Wargnier H, Danis M. Filtration in materials selection
Closeness Rank Closeness Rank and multi-materials design. Mater Des 2009;31:1842–7.
[9] Farag MM. Materials and process selection in engineering. Elsevier Science &
A1 20 5 0.1157 4 0.1241 4 Technology; 1979.
A2 47 10 0.3060 3 0.3277 3 [10] Edwards KL. Strategic substitution of new materials for old: applications in
A3 5 87 0.5316 1 0.5000 2 automotive product development. Mater Des 2004;25:529–33.
A4 87 15 0.5032 2 0.5343 1 [11] Edwards KL. Selecting materials for optimum use in engineering components.
Preference ranking A3 > A4 > A2 > A1 A4 > A3 > A2 > A1 Mater Des 2005;26:469–73.
[12] Cather H, Morris R, Philip M, Rose C. Materials selection. Design
engineering. Oxford: Butterworth-Heinemann; 2001. p. 190–224.
in such situations, this approach needs more mathematical tools [13] Jahan A, Ismail MY, Sapuan SM, Mustapha F. Material screening and choosing
for effective decision-making. Also, in complex materials selection methods – a review. Mater Des 2010;31:696–705.
[14] Ullah AMMS, Harib KH. An intelligent method for selecting optimal materials
problems, at least two ranking techniques must be used simulta-
and its application. Adv Eng Inform 2008;22:473–83.
neously for optimum decision-making. The values of material [15] Johnson M, Kirchain R. Quantifying the effects of parts consolidation and
properties can be either obtained by different measurement sys- development costs on material selection decisions: a process-based costing
approach. Int J Product Econ 2009;119:174–86.
tems or demonstrated using various scales, but the normalized
[16] Farag MM. Quantitative methods of materials substitution: application to
data must be the same. This is because the data used in the deci- automotive components. Mater Des 2008;29:374–80.
sion matrix not only affects the weighting system criteria, but also [17] Ashby M, Johnson K. A structure for material selection. Materials and design.
influences the accuracy of the final ranking of alternatives. Non- 2nd ed. Oxford: Butterworth-Heinemann; 2010. p. 122–41.
[18] Edwards KL, Deng YM. Supporting design decision-making when applying
monotonic normalization [36], which addresses target criteria, materials in combination. Mater Des 2007;28:1288–97.
cannot properly show the variation of data, while the proposed [19] Sarfaraz Khabbaz R, Dehghan Manshadi B, Abedian A, Mahmudi R. A simplified
normalization technique improves the objective weighting proce- fuzzy logic approach for materials selection in mechanical engineering design.
Mater Des 2009;30:687–97.
dure as well as the exactness of ranking in the original TOPSIS. [20] Chatterjee P, Athawale VM, Chakraborty S. Materials selection using complex
The target-based normalization technique has been validated by proportional assessment and evaluation of mixed data methods. Mater Des
solving a broad range of representative materials selection problems 2011;32:851–60.
[21] Lan YP, Meng QC, Li F, Xu GX, Guan ZD. Aircraft design material-selection
found in mechanical engineering product design. This also demon- method based on MAUT theory. Hangkong Cailiao Xuebao/J Aeronaut Mater
strates its versatility in use and intrinsic value to a more effective de- 2010;30:88–94.
sign process, from concept to detail design. The examples have [22] Jahan A, Ismail MY, Mustapha F, Sapuan SM. Material selection based on
ordinal data. Mater Des 2010;31:3180–7.
shown how variations in data type and weightings criteria can be [23] Huang H, Zhang L, Liu Z, Sutherland JW. Multi-criteria decision making and
systematically considered, allowing the potential for scaling up to uncertainty analysis for materials selection in environmentally conscious
more complex problems [24]. From a design perspective, the im- design. The Int J Adv Manuf Technol 2010:1–12.
[24] Cicek K, Celik M. Multiple attribute decision making solution to material
proved accuracy of the final ranking of alternative materials, using
selection problem based on modified fuzzy axiomatic design-model selection
the technique described, imparts confidence in decision-making interface algorithm. Mater Des 2010;31:2129–33.
and hence reduces the incidences of costly mistakes. [25] Dehghan-Manshadi B, Mahmudi H, Abedian A, Mahmudi R. A novel method
for materials selection in mechanical design: Combination of non-linear
normalization and a modified digital logic method. Mater Des 2007;28:
5. Conclusion 8–15.
[26] Fayazbakhsh K, Abedian A, Manshadi BD, Khabbaz RS. Introducing a novel
It has been shown that the proposed normalization method for method for materials selection in mechanical design using Z-transformation in
statistics for normalization of material properties. Mater Des
materials selection is not only able to address the beneficial and 2009;30:4396–404.
non-beneficial criteria effectively, but also the target values of cri- [27] Shanian A, Milani AS, Carson C, Abeyaratne RC. A new application of ELECTRE
teria. Accordingly, the TOPSIS method was extended and its steps III and revised Simos’ procedure for group material selection under weighting
uncertainty. Knowledge-Based Syst 2008;21:709–20.
were updated. Also, it has been shown that the suggested method [28] Rao RV, Patel BK. A subjective and objective integrated multiple attribute
overcomes situations in which the current version of TOPSIS fails decision making method for material selection. Mater Des 2010;31:4738–47.
to select the optimum alternative. It can therefore be used to help [29] Rao RV, Davim JP. A decision-making framework model for material selection
using a combined multiple attribute decision-making method. Int J Adv Manuf
make better decisions. Further, the objective-weighting system Technol 2008;35:751–60.
was developed for situations with target values in MADM, and its [30] Hwang CL, Yoon K. Multiple attribute decision making-methods and
weak points addressed for material selection. The suggested model applications. Berlin: Springer Verlag; 1981.
[31] Jahan A, Mustapha F, Ismail MY, Sapuan SM, Bahraminasab M. A
was illustrated and its use validated with a range of typical mate-
comprehensive VIKOR method for material selection. Mater Des
rials selection problems. It is therefore expected to be used in a 2011;32:1215–21.
wide variety of applications of decision-making, especially during [32] Bahraminasab M, Jahan A. Material selection for femoral component of total
the engineering design process. It also appears that the proposed knee replacement using comprehensive VIKOR. Mater Des 2011;32:4471–7.
[33] Chen Y, Kilgour DM, Hipel KW. An extreme-distance approach to multiple
new procedure is as suitable as comprehensive VIKOR for linking criteria ranking. Mathe Comput Modell 2011;53:646–58.
to materials databases as a means of enhancing the efficiency of [34] Cicek K, Celik M, Topcu YI. An integrated decision aid extension to material
computer-based materials selection tools. selection problem. Mater Des 2010;31:4398–402.
[35] Maniya K, Bhatt MG. A selection of material using a novel type decision-
making method: preference selection index method. Mater Des
References 2010;31:1785–9.
[36] Shih HS, Shyur HJ, Lee ES. An extension of TOPSIS for group decision making.
[1] Farag MM. Quantitative methods of materials selection. In: Kutz M, editor. Mathe Comput Modell 2007;45:801–13.
Handbook of Materials Selection; 2002. [37] Milani AS, Shanian A, Madoliat R, Nemes JA. The effect of normalization norms
[2] Chiner M. Planning of expert systems for materials selection. Mater Des in multiple attribute decision making models: a case study in gear material
1988;9:195–203. selection. Struct Multidiscipl Optimiz 2005;29:312–8.
[3] Ashby MF, Brechet YJM, Cebon D, Salvo L. Selection strategies for materials and [38] Saaty TL. How to make a decision: the analytic hierarchy process. Euro J Operat
processes. Mater Des 2004;25:51–67. Res 1990;48:9–26.
[4] Lovatt AM, Shercliff HR. Manufacturing process selection in engineering [39] Rao RV. A decision making methodology for material selection using an
design. Part 1: the role of process selection. Mater Des 1998;19:205–15. improved compromise ranking method. Mater Des 2008;29:1949–54.
654 A. Jahan et al. / Materials and Design 35 (2012) 647–654

[40] Opricovic S, Tzeng GH. Compromise solution by MCDM methods: a [45] Jee DH, Kang KJ. A method for optimal material selection aided with decision
comparative analysis of VIKOR and TOPSIS. Euro J Operat Res making theory. Mater Des 2000;21:199–206.
2004;156:445–55. [46] Chatterjee P, Athawale VM, Chakraborty S. Selection of materials
[41] Ramalhete PS, Senos AMR, Aguiar C. Digital tools for material selection in using compromise ranking and outranking methods. Mater Des 2009;30:
product design. Mater Des 2010;31:2275–87. 4043–53.
[42] Roth R, Field F, Clark J. Materials selection and multi-attribute utility analysis. J [47] Rathod MK, Kanzaria HV. A methodological concept for phase change material
Computer-Aided Mater Des 1994;1:325–42. selection based on multiple criteria decision analysis with and without fuzzy
[43] Farag M. Materials selection for engineering design. New York: Prentice-Hall; environment. Mater Des 2011;32:3578–85.
1997. [48] Thakker A, Jarvis J, Buggy M, Sahed A. A novel approach to materials selection
[44] Simanaviciene R, Ustinovichius L. Sensitivity analysis for multiple criteria strategy case study: Wave energy extraction impulse turbine blade. Mater Des
decision making methods: TOPSIS and SAW. Procedia - Social Behav Sci 2008;29:1973–80.
2010;2:7743–54.

You might also like