Académique Documents
Professionnel Documents
Culture Documents
DOI 10.1007/s00170-008-1530-5
ORIGINAL ARTICLE
Received: 6 August 2007 / Accepted: 15 April 2008 / Published online: 3 June 2008
# Springer-Verlag London Limited 2008
Abstract In this article, we consider the facility layout way. FLP has attracted many researchers because of its
problem which combines the objective of minimization of practical benefit and importance for many planning functions.
the total material handling cost and the maximization of A good solution for the facility layout problem would
total closeness rating scores. Multi-objective optimization is contribute to the overall efficiency of operations. A poor
the way to consider the two objectives at the same time. A layout can lead to accumulation of work-in process inventory,
simulated annealing (SA) algorithm is proposed to find the overloading of material handling systems, inefficient setups
non-dominated solution (Pareto optimal) set approximately and longer queues [1]. Therefore, solution of an FLP is a
for the multi-objective facility layout problem we tackle. strategic study to be conducted.
The Pareto optimal sets generated by the proposed Traditionally, there are two approaches for the facility
algorithm was compared with the solutions of the previous layout problem. The first one is the quantitative approach
algorithms for multi-objective facility layout problem. The aiming at minimizing the total material handling cost
results showed that the approximate Pareto optimal sets we between departments based on a distance function. Armour
have found include almost all the previously obtained and Buffa [2] and Buffa et al. [3] can be given as examples
results and many more approximate Pareto optimal solutions. for this approach. The second one is the qualitative
approach aiming at maximizing closeness rating scores
Keywords Facility layout problem . Simulated annealing . between departments based on a closeness function. The
Pareto optimal solution . Multi-objective optimization most important example for this approach can be systematic
layout planning-SLP procedure, suggested by Muther [4].
Detailed reviews of the literature on facility layout problem
were conducted by Kusiak and Heragu [5] and Meller and
1 Introduction Gau [6].
The approaches stated above were used separately to
Facility layout is one of the most important problems for solve the facility layout problem. However both approaches
modern manufacturing systems. Facility layout plays a key have advantages and disadvantages. The models that use
role for companies, and it is an inseparable part of the one of the criteria are not explanatory because there are
manufacturing system design process. A facility layout many factors which affect the facility layout. In order to do
problem (FLP) is about arranging the physical departments a more effective layout design, both criteria have to be
within a facility to help the facility work in a productive considered. Thus, we need to solve the problem by using
multi-objective model.
The purpose of multi-objective facility layout problems
is to generate efficient alternatives which can then be
R. ahin (*) : O. Trkbey presented to the decision-maker (DM) so that he/she can
Department of Industrial Engineering,
select the best facility layout alternative while considering
Faculty of Engineering and Architecture, Gazi University,
06570 Ankara, Turkey conflicting and non-commensurate objectives [7]. In previous
e-mail: rsahin@gazi.edu.tr studies, objectives were combined by many different
1004 Int J Adv Manuf Technol (2009) 41:10031018
due to handling of material. c is taken as equal to the objective problems were converted into a single objective
highest level of material handling between facilities. one by weighting the objectives and an optimizing solution
d) Khare et. al [14], in Eq. (1), defined Aijkl term as was sought. But in this study, qualitative and quantitative
Aijkl=1djl fik+2rikdjl. Definition of the parameters are objectives are handled as different objective functions and
same as given above, and 1 and 2 are the weights non-dominated approximate Pareto optimal solution set is
used to unify the two objectives. constructed through a search procedure. From here on we
will use the Pareto optimal set (POS) in place of
Apart from the studies above, many other works were
approximate Pareto optimal solution set.
carried out on multi-criteria facility layout problem.
Multi-objective facility layout problem (MOFLP) can be
Malakooti and Tsurushima [15] developed an expert system
stated as below:
for multi-objective facility layout problem. Raoot and
Rakshit [16] had a linguistic model approach. In this
model, by fuzzy set approach, they examined the qualitative Min f x ff1 x; f2 xg 5
and quantitative criteria where complete and clear informa-
tion can not be gathered. Suresh and Sahu [17] used
simulated annealing. Chen and Sha [18] and Sha and Chen subject to
[9] conducted studies related to the qualified solution of
multi-criteria facility layout problem. In these studies, by n X
X n X
n X
n
developing dominant index and probability of superiority f 1 x Cijkl djl Xij Xkl 6
concepts, respectively, they tried to find whether one i1 j1 k1 l1
solution is better than the other. Deb and Bhattacharyya
[19] searched for a solution to multi-criteria facility layout
problem by using fuzzy logic. Chen and Sha [20] offered a n X
X n X
n X
n
new heuristic approach by giving more importance by f 2 x Rijkl djl Xij Xkl 7
giving different weights to objectives. i1 j1 k1 l1
Many real life problems have more than one objective. where
These problems are named as multi-objective optimization 1 if facility i is assigned to location j;
Xij
problems. Although the solution of single objective 0 otherwise;
(
optimization problem figures out a single solution, the fik djl ; if facility i is assigned to location j and
solution of multi-objective optimization figures out a set of Cijkl facility k is assigned to location l
0; otherwise
solutions. In this set, a set where none of the solutions (
dominated each other is called non-dominated or Pareto rik djl ; if facility i is assigned to location j
Rijkl and facility k is assigned to location l
optimal solution set. Each solution in this set has equal
0; otherwise
importance. A solution x* is a Pareto optimal if no objective
fik work flow from facility i facility k
function can be improved without worsening at least one
rik closeness rating score of facilities i and k.
other objective functions. Mathematically, Pareto optimal
djl distance from location j to location l.
solution is defined as: x*S is Pareto optimal solution if
there is no solution for xS where fi(x) fi(x*) i i=1,2,3,k Equation 5 indicates that we are minimizing both f1(x)
and fi(x)<fi(x*) at least for one i. In previous studies, multi- and f2(x). f1(x) and f2(x) show the total material handling
1006 Int J Adv Manuf Technol (2009) 41:10031018
cost, and the total closeness rating score between facilities, Step 4 Make the iteration counter 0 at each temperature
respectively. In f2(x), weighted distance cost, which level: il=0
represents closeness rating score, is considered as a penalty Step 5 Increase the iteration counter one at each temper-
to be minimized similar to Khare et al. [14]. ature level: il=il+1.
Facility layout problem is a combinatorial optimization Step 6 Choose two facilities randomly (ij) and find a
problem in NP-complete sense. For this reason, finding an neighbour solution by swapping their places. Then
optimal solution for a facility layout problem of a real life calculate the function of the neighbour solution. At
model is difficult due to increase in required computation the same time, calculate weighted sum of the two
time. In literature, heuristic methods are commonly used. objective function values (Eil) using the given
weights.
START
Input parameters
T = Tin , el = 0
Create Sin
Calculate E0, Ecost, Ecloseness
Sbest = Sc = Sin
Ebest = E0
il = 0
il = il+1
Generate Sil from Sc
Calculate Eil, Ecost_il, Ecloseness_il
Y
Sil Pareto
optimal? Update Pareto List
E = Eil - Ec
Y Y
Sc = Sil, Ec = Eil
rnd e(E / T)
N N
Ec < Ebest il NIET
Y Y
Sbest = Sc el = el + 1
Ebest = Ec Tel+1 = Tel*
N
el elmax
STOP
1008 Int J Adv Manuf Technol (2009) 41:10031018
130
12, 15, 20 and 30 departments. The type of layout is
120
known for all test problems. These layouts are given in
Score
110
Table 1.
See Fig. 2 for an example of a 12-department problem. 100
The algorithm is run twice for each test problem. In the first 90
one, to combine the objectives at the stage of accepting 80
180
190
200
210
220
230
240
250
260
550
580
610
640
670
700
730
760
790
820
850
Material Handling Cost
in dataset 1, comparisons for previous studies and efficient Fig. 4 Pareto solutions and efficient frontier for a 12-department test
frontier are given in Appendix 1. Comparison for an eight- problem
department test problem in dataset 1 is given in Table 4.
The proposed SA algorithm found five Pareto solutions for
Note: Material handling cost and closeness rating scores for D+S, F+C, H+T and S+C are the values found by the proposed method that is
recommended to them according to their solution. The layouts obtained by D+S are directly taken from F+C
1010 Int J Adv Manuf Technol (2009) 41:10031018
this problem. All layouts obtained by Dutta and Sahu [11], the layouts obtained by the other methods are dominated
except for one layout, are dominated by found Pareto by found Pareto solutions, found Pareto solutions set
solutions. The layouts proposed by Foetenberry and Cox contains some of them. But any solution produced by the
[12], and Sha and Chen [9] are included in found Pareto other methods dominates proposed Pareto solutions.
solutions. For the test problem of 12 departments in dataset 2, a
Comparison for a 12-department test problem in POS which consist of 24 Pareto solutions was found.
dataset 1 is given in Table 5. The proposed SA algorithm Figure 4 shows the solutions in this POS and efficient
found 20 Pareto solutions for this problem. While some of frontier. Obtained Pareto solutions are compared with the
Note: Material handling cost and closeness rating scores for D+S, and S+S, are the values found by the proposed method that is recommended to
them, according to their solution. The layouts obtained by D+S are directly taken from S+S.
Int J Adv Manuf Technol (2009) 41:10031018 1011
ones found by Dutta and Sahu (D+S) [11] and Suresh and with the solutions of previous studies in the literature.
Sahu (S+S) [17]. The Pareto solutions and results of the Results showed that proposed method generates better
comparisons are given in Table 3. Dutta and Sahu [11] and solutions. Additionally previous studies offered single
Suresh and Sahu [17] found the best solutions for each solution under certain circumstances, however with this
weight, by giving different weights to objective values. As study, for the multi-objective facility layout problem, more
it can be seen in Table 3, only the solution for =0.5 found than one Pareto solution is recommended to the decision-
by Suresh and Sahu [17] is among the Pareto solutions and maker. Decision-maker can decide on different solutions
the rest of the solutions found both by Dutta and Sahu [11] according to todays changing economic conditions and
and Suresh and Sahu [17] are dominated by the Pareto choose one solution among the set of Pareto optimal
solutions found with the proposed method. Pareto solutions solutions. This will enable the decision-maker to make
set of 15-, 20-, and 30-department test problems in dataset 2 better, more effective and flexible decisions depending on
and comparisons are given in Appendix 2. Comparison for the market circumstances. For future studies, considering
a 15-department test problem in dataset 2 is given in more than two objectives, using other meta-heuristic
Table 6. The proposed SA algorithm found 23 Pareto methods (genetic algorithm, tabu search, ant colony
solutions for a 15-department test problem. As it can be algorithm etc.), or forming hybrid methods can be
seen in Table 6, the solution for =0.7 found by Suresh considered. Additionally it will be suitable to develop
and Sahu [17] dominates the Pareto solution 1 obtained by solution procedure for the problems with dynamic, fuzzy
the proposed SA algorithm. But the rest of the solutions or stochastic data.
found both by Dutta and Sahu [11] and Suresh and Sahu
[17] are dominated by the Pareto solutions found with the
proposed method.
In Table 7, comparison for a 20-department test problem Appendix 1
in dataset 2 is given. For this test problem, the proposed SA
algorithm found 39 Pareto solutions, but four of them is Note: Material handling cost and closeness rating scores for
dominated by the solutions found by Suresh and Sahu [17]. D+S, F+C, H+T and S+C, are the values found by the
The solution for =0.9 found by Suresh and Sahu [17] and proposed method that is recommended to them according to
all solutions found by Dutta and Sahu [11] are dominated by their solution. The layouts obtained by D+S are directly
the Pareto solutions found with the proposed method. taken from F+C.
Comparison for a 30-department test problem in dataset 2
is given in Table 8. The proposed SA algorithm found 33
Pareto solutions for a 30-department test problem. As it can
be seen in Table 8, the solutions for =0.6, =0.8, and =
0.9 found by Suresh and Sahu [17] dominates the Pareto
solutions 1, 2, 3, and 4 obtained by the proposed SA
algorithm.
Pareto Solutions
Closeness Rating Score
290
280
6 Summary and conclusions 270
260
250
In this study, SA algorithm, which finds Pareto solutions 240
for the minimization of material handling cost and 230
220
maximization of closeness rating score between depart- 210
ments, is presented for multi-objective facility layout 200
300
320
340
360
380
400
420
440
460
Note: Material handling cost and closeness rating scores for D+S, and S+S, are the values found by the proposed method that is recommended to
them according to their solution. The layouts obtained by D+S are directly taken from S+S
Pareto Solutions
650
Closeness Rating Score
630
610
590
570
550
530
510
490
470
450
2500
2550
2600
2650
2700
2750
2800
2850
2900
2950
3000
Pareto Solutions
Closeness Rating Score
1950
1900
1850
1800
Appendix 2 1750
1700
Note: Material handling cost and closeness rating scores for
1100
1150
1200
1250
1300
1350
1400
1450
1500
1550
D+S and S+S are the values found by the proposed method
Material Handling Cost
that is recommended to them according to their solution. Fig. 7 Pareto solutions and efficient frontier for a 15-department test
The layouts obtained by D+S are directly taken from S+S. problem
1014 Int J Adv Manuf Technol (2009) 41:10031018
Pareto Solutions
4000
Closeness Rating Score
3950
3900
3850
3800
3750
3700
3650
3600
3550
3500
2600
2700
2800
2900
3000
3100
3200
3300
3400
3500
3600
6400
6570
6740
6910
7080
7250
7420
Pareto Solutions
7590
7760
7930
8100
Fig. 9 Pareto solutions and efficient frontier for a 30-department test
Int J Adv Manuf Technol (2009) 41:10031018
Table 8 Comparison of a 30-department test problem