4th International Quality Conference

May 19th 2010

Center for Quality, Faculty of Mechanical Engineering, University of Kragujevac

A Fuzzy AHP and TOPSIS for ELV Dismantling

Danijela Tadić 1) Selection

Slavko Arsovski1) Abstract: In this paper, a new method for ELV dismantling
selection is proposed which has purpose to determine the
Miladin Stefanović1) sequence of dismantling of ELV which arrive to
dismantling center. Its solution is of the most importance
Aleksandar Aleksić1) for ELV waste management problem, as it influences the
choice and efficiency of ELV waste management strategies
1) Faculty of Mechanical in dismantling centers. It is realistically posed that the
Engineering, University of choice of locations depends on multiple, rather conflicting
Kragujevac, Serbia criteria. The criteria values can be either crisp or
uncertain. The main contribution of this paper is the
development of methodology for determining the
dismantling order of coming ELV and/ or ELV which are
already in the dismantling centers. The uncertain criteria
values are described by linguistic expressions modeled by
triangular fuzzy numbers. The extension of the fuzzy
Technique for Order Preference by Similarity to Ideal
Solution (TOPSIS) is applied to rank the ELV which are in
the dismantling center.
Keywords: End-of-Life Vehicles selection, dismantling
centers, uncertainties, multi-criteria approaches, fuzzy sets

1. INTRODUCTION has become one of their most important

tasks. Moreover, problem of ELV
The prevention of wastage of vehicles, recycling is not exclusively faced by
the improvement of vehicle dismantling industrialized countries. According to
and their recycling is another policy of (Togawa, 2006), the implementation of
Directive 2000/53/EC (EC (2000) strict product-oriented legislation (Life-
Directive 2000/53/EC) which, among Cycle- Assessment standard is associated
other, defines the obligations and rights of with car-industry) will sooner or later
all partakers in this process: a country as a become of dominant interest in developing
site owner, car-manufacturers, power and countries.
water producers, spare parts Determining dismantling order of
manufacturers, drivers, recycles, etc. The ELV in dismantling center is one of ELV
primal aim of such a policy is to achieve waste management problems. The treated
the economic success of End-of Life problem has become of special importance
Vehicles (ELV) dismantling processes for developing countries, such as Serbia
(target value of ELV recyclability level is where ELV recycling process is in its
about 95%). infancy. The importance of the considered
In the past decade, recycling of ELV problem can be illustrated with the fact
has arisen as a very important issue for that productivity of dismantling centers
car-manufactures worldwide, and the depends in the first place on the chosen
improvement of ELV recycling processes dismantling order of the ELV.
It is mentioned, that selection of

4th IQC May, 19 2010 377

dismantling order of ELV is not a straight experts and managers.
forward task. Many aspects, such as Optimal dismantling order of ELV in
environmental features, social impact dismantling center can be obtained by
assessment, cost considerations, etc. must applying some developed MADM
be accounted for, in order to point to an techniques or by combining different
adequate management of recycling MADM techniques. The widest appliance
processes. in multi-criteria problem for selecting of
The considered problem is a group different items has AHP method and
decision-making problem under multiple TOPSIS method or their combination (Ho,
criteria, and it can be stated as multi- Xu, and Dey, 2009). In AHP method
criteria optimization problem (MADM). (Saaty, 1990) decision-making problem is
The degree of uncertainty, the number of hierarchy structured, the weights criteria
decision makers and nature of criteria and preference of alternatives under each
those have to be taken into account in treated criterion are assigned according to
solving this problem. The different pairwise comparison matrix of considered
decision criteria may vary depending on factors. The elements of these matrixes are
the need of the organization and changing obtained upon the evaluation of decision
of environment. Estimation of criteria makers. These matrixes represent input
weights and uncertain criteria values data for ranking and best alternative
cannot be performed with an exact selection with respect to all treated criteria
numerical value. A more realistic approach and their weights. TOPSIS (Hwang, and
may be to use linguistic assessments Yoon 1981) is based on the best alternative
instead of numerical values. In other selection, which has the shortest distance
words, the all uncertainties which exist in from the positive-ideal alternative and the
considered problem can be described by longest distance from the negative-ideal
linguistic variables (Zadeh, 1975). In this alternative. In the conventional forms of
paper, modeling of these linguistic the AHP and TOPSIS methods, only crisp
variables is based on the fuzzy set theory parameters values have been considered
(Zadeh, 1975, Zimermann, 1996, Pedrycy, for supplier selection, which represents
and Gomide, 1998). their basic defect.
Fuzzy set theory can provide a In many papers, which can be found
valuable tool which copes with three major in the literature, MADM problems are
problematic areas of selection problems: solved by a proposed two-stage method.
imprecision, randomness and ambiguity. At the first stage, AHP or FAHP method is
As far as imprecision is concerned it used to determine the weight of treated
provides a powerful tool to weigh selection criteria and/or the weight of suppliers
criteria importance. As far as ambiguity is (Cebi and Bayraktar, 2003, Wang, et al,
concerned it copes better than other 2004, 2005, Percin, 2006, Chen, Lin, and
methods with the treatment of linguistic Huang, 2006, Chan, and Kumar, 2007, Xia
variables. Fuzzy logic enables us to and Wu , 2007, Torfi, Farahani, and
emulate the human reasoning process and Rezapour, 2009). Secondly, some other
make a decision based on vague or methods are used in order to determine the
imprecise data (Kaur, and Chakrabortyb, best alternative with respect to all treated
2007). criteria, simultaneously, and their relative
The fuzzy set theory resembles human importance.
reasoning in its use of approximate In this paper, we constructed: (1) a
information and uncertainty to generate pairwise comparison matrix of relative
decisions; based on natural languages; importance of considered criteria, and (2) a
provide better communication between pairwise comparison matrix of preference

378 D. Tadić, S. Arsovski, M. Stefanović, A. Aleksić

of alternative. The elements of these • Management Team defines the
matrixes are defined as: the relative possible quantity of ELV which comes to
importance/preference of criterion k/a over dismantling center, unit/ time. The
criterion. It appears that the weight assessment of decision makers are is based
determination of criteria is more reliable on: (a) analysis of historical data, on the
when using pairwise comparisons than experience of other dismantling centers,
obtaining them directly, because it is easier (b) professional observation, etc. In
to make a comparison between two criteria practice different approaches are more
than make an overall weight assignment. often combined.
Also, the relative importance of each pair • Management Team defines the
of treated criteria is described by linguistic group of criteria according to which each
expressions by more decision makers. The ELV which is in the dismantling center
fuzzy rating of each decision maker is and should be dismantled is being
modeled by triangular fuzzy numbers. evaluated. The problem of selection of
Weights vector is calculated by using the criteria according to which ELV for
extent analysis method (Chang, 1996). dismantling are evaluated and selected can
The criteria values can be crisp and be observed as an isolated problem. In this
uncertain. Normalization of crisp criteria is paper, we consider the following criteria:
performed according to linear (1) age of vehicle, (2) the need for spare
normalization procedure (Pomerol, and parts (spare parts), (3) possession of
Barba, 2000). special tools for dismantling of ELVs, (4)
A closeness coefficient according to condition of cars, (5) income / unit ELV,
determine rank of alternatives according to and (6) the level of pollution of water and
which we can determine is calculated by soil by the ELV.
using procedure which is defined in • To each defined criterion an
conventional TOPSIS method. organized pair (relative importance, value
The paper is organized as following. joined) is associated.
In Section 2 a multi-criteria approach for • Relative importance of treated
ranking of ELV in dismantling problem is criteria does not depend on ELV for
given. In Section 3, synthetic extent dismantling, and is in most cases hardly
analyses method for calculating the final changes. Generally, the relative
priority weights of criteria and alternative importance of criteria is different and
preference is presented. In Section 4, determined according to knowledge and
principles of modified of TOPSIS method experience of Management Team, and it
are presented. can be stated as the following ways.
Values of defined criteria are determined
for each ELV for dismantling separately.
2. MULTICRITERA In this problem, these values can be crisp
APPROACH FOR SUPPLIER and/or uncertain.
SELECTION
2.2 Notation
2.1 Basic asssumptions
a-alternative (ELV for dismantling),
In fact, ELV for dismantling ranking and a=1,..,A,
selection problem is a group multiple- k-criterion, k=1,..,K,
criteria decision-making problem. A-total number of ELV for dismantling,
Assumptions, under which a model for K’-total number of crisp criteria,
considered problem, are: K-total number of treated criteria,
E-total number of decision makers who are

4th IQC May, 19 2010 379

the members of the Management Team 3. MODELLING OF
~ e UNCERTAINNTIES
W kk ' - a triangular fuzzy number
All the criteria for evaluating ELV for
⎛ x; l e , m e , u e ⎞
⎜ kk ' ⎟ describing the dismantling usually do not have the same
⎝ kk ' kk ' ⎠ relative importance, and does not depend
fuzzy rating of relative importance of each on ELV for dismantling. Also, it can be
pair of considered criteria of each decision considered as unchangeable during the
maker, considered period of time. Determining
~ the importance and alternative preferences
W kk ' - a triangular fuzzy number
( )
under uncertain criteria involve a high
x; l ' , m ' , u ' describing the degree subjective judgment and individual
kk kk kk preferences of decision makers. We think
relative importance of each pair of treated that the judgment of each pair of treated
criteria, (k=1,..,K), criteria, regarding each pair of the
wk -the relative importance of criterion k, considered alternatives best suits human-
k=1,..,K, decision nature (analogously AHP
fak -cardinal value of criterion k for method). In conventional AHP, the
alternative a, k=1,.., ; a=1,..,A, pairwise comparison is established using a
n
f ak - normalized value of fak, k = 1,.., K ' ; standard integer scale [1-9]. Value 1
denotes that item i (criterion k and
a=1,.., A,
alternative a) is important as criterion , and
~e value 9 denotes that item i is extremely
P aa ' - a triangular fuzzy number more important/preferential that item
⎛ y; l e , m e , u e ⎞ i' , i ≠ i' .
⎜ aa ' ⎟ describing the
⎝ aa ' aa ' ⎠ Using the discrete scale of AHP is
fuzzy rating of preference of each pair of simple and easy but it is not sufficient to
alternative under criterion k, take into account the uncertainty
associated with the mapping of one’s
k = K ' + 1,..., K of each decision maker, perception to a number (Kwong, and Bai,
~ 2003). Decision makers express their
P aa ' - a triangular fuzzy number
(
y; l ' , m ' , u '
aa aa aa
)
describing the
judgments far better by using linguistic
expressions than by representing in terms
of precise numbers. It feels more confident
preference of each pair of treated to give interval judgments than fixed value
alternative under uncertain criterion k judgments.
criteria, k = K ' + 1,..., K , In this paper, the fuzzy rating of each
p ak - the preference of alternative a under decision maker is described by linguistic
expressions which can be represented as
criterion k, a=1,..,A; k = K ' + 1,..., K , triangular fuzzy number
v k+ - positive-ideal value of criterion k, ~ e
W ii' = ⎛⎜ x; l iie ' , m e ' , u e ' ⎞⎟ with the
k=1,..,K , ⎝ ii ii ⎠
vk− - negative-ideal value of criterion k, lower and upper bounds l e ' , u e ' and
k=1,…,K ii ii
ca - a closeness coefficient for alternative modal value m e ' , respectively. The
a, a=1,..,A. ii

380 D. Tadić, S. Arsovski, M. Stefanović, A. Aleksić

greater u e ' − l e ' , the fuzzier the degree. strongly importance, very strongly
ii ii importance, and very strongly importance.
Values in the domain of these triangular These linguistic expressions are modeled
fuzzy numbers belong to real set into by triangular fuzzy numbers which are
interval [1-9]. Value in domain of each given in the following way:
these five fuzzy numbers has the same equally ~
meaning as value of standard scale which important/prefe R E = ( x; 1, 1, 2 )
is given in conventional AHP. rence
If strong relative importance of
moderately ~
criterion k ' over criterion k holds, then R M = ( x ; 1 .5 , 3 , 4 .5 )
important/prefe
pairwise comparison scale can be rence
represented by the fuzzy number
−1 ⎛ 1 strongly ~
~ ⎛~ ⎞ 1 1 ⎞ V S = (x; 3.5 , 5 , 6.5 )
W ii' = ⎜⎜ W i'i ⎟⎟ =⎜ , , ⎟. important/prefe
⎜u m' l' ⎟ rence
⎝ ⎠ ⎝ i'i ii ii ⎠
very strongly ~
If i = i ' then relative importance important/prefe R VS = (x; 6 ,7 , 8 )
criterion k/preferentnost alternative a over rence
criterion k ' /prema alternativi a ' is very strongly ~
represented by single point 1 which is important/prefe R VVS = (x; 7.5 , 9 , 9 )
triangular fuzzy number (1,1,1). rence
The aggregated fuzzy rating of
relative importance of each pair of As one can see there are three basic
considered criteria/preferentnost of each terms low, medium and high importance.
pair of treated alternatives must include the Two additional terms with the word very
fuzzy rating of all decision makers. The are obtained by moving strict boundaries.
aggregated fuzzy rating can be defined as: All five terms are given in Fig. 1.
These triangular fuzzy numbers, as the
l ' = min l e ' simplest shape of membership functions
ii e=1,..,E ii
are shown in Fig. 1.
1 E e
m '
ii
=
E e=1 ii
∑
⋅ m ' , u ' = max u e '
ii e=1,..,E ii
1,2

1
membership functions

The relative importance of each pair 0,8

equallly importance
moderately importance
of considered criteria/preferentnost of each 0,6
strongly importance

pair of treated alternatives is described by very strongly importance

very very strongly
0,4
triangular fuzzy number importance

~
( )
0,2

W ii' = l ', m ', u ' with the lower 0

ii ii ii 0 2 4 6 8 10
criteria importance

and upper bounds l ' , u ' and modal

ii ii Figure 1 – Five triangular fuzzy
value m ' , respectively. numbers that describe relative importance
ii of criteria
In this paper, the fuzzy rating of each
decision maker can be described by using Weights vector of considered criteria
a five linguistic expressions: equally and vector of preferentnosti alternative
importance, moderately importance, under uncertain criteria are calculated by

4th IQC May, 19 2010 381

applying the concept of extent analysis v k− :
(Chang, 1996).
(a) for a crisp criterion criterion k,
4. PRINCIPLES OF MODIFIED k=1,..., K '
TOPSIS METHOD v +k = max f ak n
, v −k = min f ak n
a =1,..,A a =1,..,A
TOPSIS method is based on choosing
the best alternative, which has the shortest
distance from the positive-ideal solution, (b) for uncertain criterion criterion k,
and the longest distance from the negative-
ideal solution (Hwang, and Yoon, 1981).
k = K ' + 1,.., K
In this section, a systematic approach v +k = max p ak , v −k = min p ak
to extend the TOPSIS is proposed to solve a =1,..,A a =1,..,A
the supplier selection problem under
uncertainties.
The algorithm of the proposed method Step 5. Calculate distance of each ELV
is realized in the following steps: for dismantling a, a=1,..,A, from positive-
Step 1. Calculation of weights vector
~M
of considered criteria by applying ideal solution, d a and negative-ideal
procedure which is presented in Section 3.
Step 2. Calculation of preferentnosti ~m
alternativa under uncertain criterion k, solution, d a , are calculated:
'
k = K + 1,.., K . M K' K + ~

Step 3. Calculation of normalized

da = ∑ w k ⋅ v +k − f akn + ∑ w k ⋅ v k − p ak
k =1 k = K ' +1
values for crisp criteria:
(a) benefit-type of criteria
~m K' K −
f ak
n
f ak =
A
da = ∑ w k ⋅ v k− − f ak
n
+ ∑ w k ⋅ v k − p ak
k =1 k = K ' +1
∑ f ak
a =1 Step 6. A closeness coefficient,
c a (a = 1,.., A ) is obtained as:
(b) cost type-criteria
m
da
n f − f min ca =
f ak = 1 − ak m M
f max da + da

f min = min f ak ,
a =1,...,A
max 5. THE CONCLUSIONS
f = max f ak
a =1,...,A
The experience of developed countries
Step 4. Determining of positive-ideal all over the world points out the neccessity
of developing recycling industry and
solution, v k+ , and negative-ideal solution,
integrating it with other industries with the
aim of maintaining the market

382 D. Tadić, S. Arsovski, M. Stefanović, A. Aleksić

competitiveness. This effect is reaching for handling fuzzy AHP. TOPSIS method
developing countries, and Serbia will be is used for ranking ELV in dismantling
affected sooner or latter. Also, the demand center.
for ELV waste management has a growing It is shown that fuzzy sets are suitable
trend, which means that Serbia, having a for modeling of the uncertain input data in
car industry, should react promply and the considered ELV waste management
undertake all neccessary actions that problem that are subjectively estimated.
would enable the development of ELV The developed fuzzy models are flexible:
recyling processes. (1) they include and operate with both
It can be concluded that determining precise and imprecise specific data, (2) all
the optimal sequence of ELV dismantling the changes, as the changes in the number
which are located in dismantling center of criteria or its relative importance, or
depends on many different criteria, such number of possible locations and
as: economic group criteia, social group membership functions shape of fuzzy
criteria, environmental group crietria, etc. numbers can be easily incorporated into
These criteria are very often in conflict. the model, and (3) fuzzy model could be
The considered criteria have a different modified for solving different waste
relative importance and values criteria can management problems.
be crisp or uncertain. Further research will be carried out
In this paper, weights vector and primarily in the analysis of fuzzy
alternatives preferences vectors under developed model sensitivity to different
uncertain criteria are obtained by applying shapes of fuzzy terms membership
extent analysis approach for the synthetic functions which appear in the model.
extent values of the pairwise comparison .

REFERENCES:

[1] Cebi, F., Bayraktar, D. 2003. An integrated appraoch for supplier selection, Logistic
Information Management 16 395-400.
[2] Chan, S.T.F., Kumar, N. 2007. Global supplier development considering risk factors using
fuzzy extended AHP-based approach. Int. J. of Production Research 46 417-431.
[3] Chang, D.Y. 1992. Applications of the extent analysis method on fuzzy AHP, European J.
of Operational Research 95 649-655.
[4] Chen, C.T., Lin, C.T., Huang, S.F. 2006. A fuzzy approach for supplier evaluation and
selection in supply chain management, Int. J. Production Economics 102 289-301.
[5] EC (2000) Directive 2000/53/EC of the European Parliament and of the Council of 18
September 2000 on end-of life vehicles, The Official Journal of the European
Communities, L269, 34-42.
[6] Ho, W., Xu, X., Dey, K.P. 2009. Multi-criteria decision making approaches for supplier
evaluation and selection: A literature review, European J. of Operational Research, doi:
10.1016/j.ejor.2009.05.009.
[7] Hwang, C.L., Yoon, K. 1981. Multiple Attribute Decision Making-Methods and
Applications, Springer-Verlag, Heidelberg.
[8] Kaur, P.& Chakrabortyb, S. (2007) A New Approach to Vendor Selection Problem with
Impact Factor as an Indirect Measure of Quality, Journal of Modern mathematics and
Statistics, 1, 1-8.
[9] Kwong, C.K., Bai, H. 2003. Determining the importance weights for the customer
requirements in QFD using a fuzzy AHP with an extent analysis approach, IIE
Transakcions 35 (7) 619-626.

4th IQC May, 19 2010 383

[10] Pedrycy, W.& Gomide, F. (1998) An introduction to fuzzy sets, Analysis and Design, MIT
Press, Cambridge Massachusetts.
[11] Pomerol, J.C., Barba, R. 2000. Multicriteria Decision Management: Principles and
Practice. Kluwer Academic Publishers.
[12] Percin, S. 2006. An application of the integrated AHP-PGP model in supplier selection,
Measuring Business Excellence 10 34-49.
[13] Saaty, T.L. 1990. How to make a decision: The Analytic Hierarchy Process, European J.
Oper., 48 9-26.
[14] Torfi, F., Farahani, Z.R., Rezapour, S. 2009. Fuzzy AHP to determine the relative weights
of evaluation criteria and Fuzzy TOPSIS to rank the alternatives, Applied Soft Computing,
doi: 10.1016/j.asoc.2009.08.021.
[15] Togawa, K. (2006) Background of the autompbile recycling law enactment in Japan. In:
Proc. 6th International Automovile Recycling Congress, Amsterdam, Netherlands, 15-17.
[16] Wang, G., et al. 2004. Product-driven supply chain selection using integrated multi-criteria
decision-making methodology, International Journal of Production Economics 91 1-15.
[17] Wang, G., et al. 2005. Manufacturing supply chain design and evaluation, International
Journal of Advanced Manufacturing Technology 25 93-100.
[18] Xia, W., Wu, Z. 2007. Supplier selection with multiple criteria in volume discount
environments, OMEGA-International Journal of Management Science 35 494-504.
[19] Zadeh, L.A. 1975. The Concept of a Liguistic Variable and its Application to Approximate
reasoning. Information Scinence 2.
[20] Zimmermann, J.H. (1996) Fuzzy set Theory and its applications, Kluwer Nijhoff
Publising, USA.
[21] Drago Dubrovski,Fuzzy Logic And Management Mistakes In Contemporary Corporate
Decision-Making , International Journal for Quality Research, Volume 3, Number 4, 2009

384 D. Tadić, S. Arsovski, M. Stefanović, A. Aleksić