Professional Documents
Culture Documents
www.emeraldinsight.com/1741-038X.htm
Optimizing
Optimizing supply chain supply chain
management using fuzzy management
approach
737
N. Gunasekaran
Department of Mechanical Engineering, Kumaraguru College of Technology, Received July 2005
Coimbatore, India Revised January 2006
Accepted February 2006
S. Rathesh
Industrial Engineer, Fine Jewellery (I) Ltd, Mumbai, India
S. Arunachalam
School of Computing and Technology, University of East London,
Dagenham, UK, and
S.C.L. Koh
Management School, University of Sheffield, Sheffield, UK
Abstract
Purpose – The purpose of this paper is to propose a fuzzy multi-criteria decision-making procedure
and it is applied to find a set of optimal solution with respect to the performance of each supplier. This
method with the use of Monte Carlo simulation produces overall desirability level less imprecise and
more realistic than those of the conventional QFD methods for engineering design evaluation.
Design/methodology/approach – A few responses obtained from customers are simulated using a
triangular fuzzy QFD algorithm, Monte Carlo simulation and a multi-objective model to optimise the
total user preferences.
Findings – The proposed approach provides decision-making with an optimal solution less
imprecise in a QFD-based collaborative product design environment.
Research limitations/implications – The proposed approach depends on the few responses and
the random numbers derived from simulation. The random numbers need to be used after passing
them through random number testing methods. The responses obtained from the customer are
considered to be genuine and original.
Originality/value – The triangular fuzzy, Monte Carlo simulation and multi-objective optimisation
are embedded into QFD environment to make the decisions less imprecise than that of conventional
QFD and it is tested for a case study problem. It definitely helps the managers in a collaborative
product design environment.
Keywords Supply chain management, Fuzzy logic, Quality function deployment,
Monte Carlo simulation, Decision making
Paper type Research paper
X
K
w0j ¼ di · r ij ; j ¼ 1; 2; . . . ; n: ð1Þ
i¼1
w0j
wj ¼ ; j ¼ 1; 2; . . . ; n: ð2Þ
X
n
0
wj
j¼1
Mask Operation welding
Electron Tube
Black Matrix
Monitor Users
Requirement
Brightness 9
Linearity 3
Focus 9
Figure 1.
White Balance 1 The HoQ for CRT example
Convergence 3 using traditional QFD
Absolute weight 111 87 27 135 144 504 approach
Relative weight (%) 22 17.3 5.36 26.8 28.6
JMTM Weights so determined represent the importance that the customer indirectly describes
17,6 to each design characteristics. They can be interpreted as the degree of “attention” that
a designer must reserve to each single technical characteristic during the product
development process. The determination of weights by means of equation (1) needs the
knowledge of the degree of importance of each customer requirements (dj), and the
conversion of symbols contained in the relationship matrix into “equivalent” numerical
740 scores (rij). The shaded region shown in Figure 1 is the prioritized HoQ for technical
characteristics and customer needs for the example listed above.
0.5
Figure 2.
Membership functions for
LV 0
1 3 9 x
A membership function of a fuzzy set A ~ relates the possible quantitative values; the Optimizing
linguistic response may take a probability for each value selected. For example, if
x; a; b; c [ R; a , b , c, and R ¼ (2 1 1), the membership function, mAðxÞ; ~ is
supply chain
defined as (Lowen, 1996): management
8
>
> ðx 2 aÞ=ðb 2 aÞ; a # x # b;
<
~
mAðxÞ ¼ ðc 2 xÞ=ðc 2 aÞ; b # x # c; ð3Þ 741
>
>
: 0; otherwise:
Fuzzy QFD using MCS and appropriate membership function is used to quantify each
response and then the responses from all experts are averaged to provide the final value
of the measure (Abebe et al., 2000). To illustrate, consider the membership function
shown in Figure 2 that are mathematically defined in Table I. Now suppose M experts
provide their opinion on the relationship between customer needs i and technical
characteristics j using one of the k LV provided. The relationship score is calculated by:
X
M
MCS triang LVm
k
m¼1
S ij ¼ ; i and j; ð4Þ
M
where MCS½triangðLVm k Þ is the MCS value for the appropriate membership function of
the linguistic variable LVk that was chosen by expert m. The same process is repeated
for each customer needs – technical characteristics to generate the entire set of
relationship scores. The importance of each performance aspect is computed similarly.
X
J
TUPn ¼ RWj WAnj ; n; ð9Þ
j¼1
X
M
MCS LVm
k
WAnj ¼ m¼1 ; n and j: ð10Þ
M
Finally, the normalized TUP for alternative n (NTUPn), is calculated by:
TUPn
NTUPn ¼ ; every n; ð11Þ
X
N
TUPn
n¼1
Multi-objective model
The final stage of the proposed methodology is the multi-objective model. Clearly, no
single model is capable of being applicable to all situations rather a model must be
built that represents the situation. In the following, two objectives are considered that
we assume conflict, namely, maximizing customer value as measured by user
satisfaction and minimizing cost. The objective functions shall be formulated as
follows:
P
(1) Minimize TC ¼ ni¼1 TCi X i
P
(2) Maximize TUS ¼ ni¼1 NTUPi X i
Optimizing
Alternative 1 Alternative 2 Alternative 3 ... Alternative N
supply chain
Req. 1 RW1* WA11 RW1* WA21 RW1* WA31 ... RW1* WAN1 management
Req.
.. 2 RW
.. 2* WA12 RW
.. 2* WA22 ..RW2* WA32 .. . . RW
.. 2* WAN2
. . . . .. .
Req. j RWj* WA1j RWj* WA2j RWj* WA3j ... RWj* WANj
X n X n Xn X n
743
TUPn RWj *WA1j RWj *WA2j RWj *WA3j ... RWj *WAij
i¼1 X n i¼1 X n i¼1 X n i¼1 X n
NTUPn TUP1 = TUPn TUP2 = TUPn TUP3 = TUPn ... TUPN = TUPn Table II.
i¼1 i¼1 i¼1 i¼1 Total user preferences
where TUS is the total user satisfaction, TC the total cost, TUPn the TUPs for vendor n,
TCn the total cost of vendor n and:
(
1 if vendor n is selected;
Xn ¼
0 otherwise:
X
n
such that TCi X i # TCmaximum ; X [ S; X n [ ð0; 1Þ for every n;
i¼1
where TCmaximum is the maximum threshold value of TC, S the feasible region.
A solution to this problem yields the maximum value of TUS for the threshold value
of TC. A second sub-problem is now formulated and solved in which TUSminimum is
used as the threshold value of TUS and TC is minimized. The most obvious initial
value of TUSminimum is the optimal value formed by solving the first sub-problem:
X
n
Min TC ¼ TCi X i
i¼1
X
n
such that NTUPi X i $ TUSminimum ; X [ S; X n [ ð0; 1Þ for every n;
i¼1
The resulting solution represents the best choice for the decision-maker. The model is
actually used as a quantitative tool to support this complex decision-making process
rather than providing the solution. To facilitate this, the second sub-problem is
repeatedly solved as the threshold value for the primary objective is incrementally
relaxed. In the above example alternative solutions would be generated by repeatedly
JMTM solving second sub-problem as the value of TUSminimum is decreased. The decision
maker is then presented a series of compromise solutions from which he or she can
17,6 proceed to impose other decision criteria.
An illustrative example
Let us consider the simple case of a design of cathode ray tube (CRT) with supply chain
744 network. The authors have made an attempt to determine the technical design
characteristics prioritization form the customer point of view. Figure 3 shows the
supply chain network with reliability as one major factor. Figures 4 and 5 show
the HoQ for the CRT product. It represents, the degree of importance is based on the
number of responses from customers as shown in Table III.
Figure 3.
Reliability consideration
in SCM
Internal Reliability Assurance Testing
Electron Tube
Black Matrix
Monitor Users
Requirement
Brightness 6.66 5.67 2 53.99
Linearity 3.99 2 3.98 5
Figure 4. Focus 6.34 4.68 4.33 3.01 6.99
The HoQ for CRT example White Balance 1.66 4.33
using fuzzy QFD approach Convergence 4.07 1.93 4.35 6.67 3.01
(stage-I) Absolute weight 75.4 42.5 46.9 79.5 103 347.4
Relative weight (%) 21.7 12.2 13.5 22.9 29.7
n h io
dbrightness ¼ 3 MCS triang LVm ; Optimizing
strong
supply chain
MCS triang LVm
moderatek ¼ 7:007; 6:986; 7:003; 4:314 management
The value of 7.007, 6.986, and 7.003 are obtained from MCS using triangular
distribution for “Strong” shown in Figure 2 while 4.314 is the value from the
745
distribution for “Moderate”.
0:004 þ 0:3ð2:693Þ
Gð7:007; xi Þ ¼ ¼ 0:9794;
0:021 þ 0:3ð2:693Þ
Similarly grey relation is found for all pairs of importance of the performance aspect,
and the pair’s having highest relation are taken for averaging as show below:
Aging Test
Total Absolute weight
Degree of importance
STEM - C
SHORT
PIN - C
COEK
MIX
Vendors
Criterion V1 V2 V3 V4 V5 V6
Mask operation and welding 0.935 0.935 1.532 0.927 0.942 0.938
Black matrix 0.528 0.854 0.525 0.855 0.531 0.854
Fluorescent paint 0.225 0.225 0.584 0.225 0.225 0.225
Electron gun sealing 1.604 1.613 1.616 1.605 1.606 1.606
Electron tube 2.091 1.285 2.081 0.495 1.282 2.082
COCK 1.216 1.207 0.461 1.206 1.215 0.466
SHORT 0.035 0.035 0.092 0.035 0.035 0.092
MIK 1.807 1.792 1.106 1.113 1.796 1.794
PIN – C 0.758 0.291 0.755 1.217 0.755 0.752
STEM – C 1.895 1.167 1.883 1.889 1.883 0.448
Table IV. Total user preferences 11.09 9.404 10.64 9.567 10.27 9.257
NTUPs and TUPs value Normalized TUP 0.184 0.156 0.177 0.159 0.171 0.154
Multi-objective modeling Optimizing
For this example, it is assumed that the costs of the product by alternative vendors, as
per the details in Table V.
supply chain
The team decides that NTUPs is the highest priority objective and TC is the management
secondary objective; however, TC must not exceed Rs. 5,000. Since, the NTUP is the
primary objective, the first sub-problem is:
747
X
6
Max TUS ¼ NTUPi X i
i¼1
X
6
Such that TCi X i # 5; 000; X [ S; X n [ ð0; 1Þ; i ¼ 1; . . . ; 6
i¼1
The optimal solution to this problems is TUS * ¼ 0.184 for X * ¼ vendor 1. The TC for
vendor 1 is Rs. 4,500. Solving the second sub-problem, minimize TC while keeping
TUs . 0.184 yields vendor 1 as optimal, so the minimum value of TUS is now
relaxed. The first level considered is TUSminimum ¼ 0.154, so the problem becomes:
X
6
Min TC ¼ TCi X i
i¼1
X
6
Such that NTUPi X i $ 0:154; X [ S; X n [ ð0; 1Þ; i ¼ 1; . . . ; 6:
i¼1
The optimal solution for this problem is TC * ¼ Rs. 2,150, X * ¼ vendor 3 and the
associated TUS ¼ 0.177. By further relaxing the TUSminimum, alternative solutions are
generated, some of which are illustrated in Table VI. The NTUP and TC decide the
vendor to be selected. Even though, the vendor 6 quotes less cost, the NTUP is
significantly less. And hence, the proposed methodology provides support in
decision-making for selecting his vendor.
Conclusion
To systemize the engineering design process, reduce cost and leading time, many
methods are emerging now a day, which are different from the conventional one.
References
Abebe, A.J., Guinot, V. and Solomatine, D.P. (2000), “Fuzzy alpha-cut vs Monte Carlo techniques
in assessing uncertainty in model parameters”, Proc. of 4-th International Conference on
Hydroinfomatics, Iowa City.
Appelqvist, P., Lehtonen, J.M. and Kokkonen, J. (2004), “Modelling in product and supply chain
design: literature survey and case study”, Journal of Manufacturing Technology
Management, Vol. 15 No. 7, pp. 675-86.
Franceschini, F. and Rosseto, S. (2002), “QFD: an interactive algorithm for the prioritization of
product’s technical design characteristics”, Integrated Manufacturing Systems, Vol. 13,
pp. 69-75.
Jain, V., Tiwari, M.K. and Chan, F.T.S. (2004), “Evaluation of the supplier performance using an
evolutionary fuzzy-based approach”, Journal of Manufacturing Technology Management,
Vol. 15 No. 8, pp. 735-44.
Lowen, R. (1996), Fuzzy Set Theory, Kluwer Academic Publishers, London.
Vanegas, L.V. and Labib, A.W. (2001a), “Application of new fuzzy-weighted average (NFWA)
method to engineering design evaluation”, International Journal of Production Research,
Vol. 39, pp. 1147-62.
Vanegas, L.V. and Labib, A.W. (2001b), “A fuzzy quality function deployment (FQFD) model for
deriving optimum targets”, International Journal of Production Research, Vol. 39,
pp. 99-120.
Further reading
Bojadziev, G. and Bojadziev, B. (1997), Fuzzy Logic for Business, Finance and Management,
World Scientific Publication, Singapore.
Erol, I. and Ferrel, W.G. Jr (2003), “A methodology for selection problems with multiple Optimizing
conflicting objectives and both qualitative and quantitative criteria”, International journal
of Production Economics, Vol. 86, pp. 187-99. supply chain
Khoo, L.P. and Ho, N.C. (1996), “Framework of a fuzzy quality function deployment system”, management
International Journal of Production Research, Vol. 34 No. 2, pp. 299-311.
Kumar, R. and Midha, P.S. (2001), “A QFD based methodology for evaluating a company PDM
requirements for collaborative product development”, Industrial Management & Data 749
Systems, Vol. 10 No. 3, pp. 126-31.
Ohdar, R. and Ray, P.K. (2004), “Performance measurement and evaluation of suppliers in supply
chain: an evolutionary fuzzy-based approach”, Journal of Manufacturing Technology
Management, Vol. 15 No. 8, pp. 723-34.
Power, D. (2005), “Implementation and use of B2B-enabling technologies: five manufacturing
cases”, Journal of Manufacturing Technology Management, Vol. 16 No. 5, pp. 554-72.
Ross, T.J. (1995), Fuzzy Logic with Engineering Applications, McGraw-Hill International Edition,
Singapore.
Wang, J. (1999), “Fuzzy out ranking approach to prioritize design requirements in quality
function deployment”, International Journal of Production Research, Vol. 37 No. 4,
pp. 899-916.
Weber, C., Werner, H. and Deubel, T. (2003), “A different view on product data
management/product life-cycle management and its future potentials”, Journal of
Engineering Design, Vol. 14 No. 4, pp. 447-64.
Corresponding author
N. Gunasekaran can be contacted at: guna_kct_cbe_tn_in@yahoo.com