Você está na página 1de 9

International Journal of Emerging Technology and Advanced Engineering

Website: www.ijetae.com (ISSN 2250-2459,Volume 1, Issue 1, November 2011)


1

Optimization of a Production System
Using Search Method
N. G. Nia
1
, Sh. K. Badrievish
2

1
Institute of Economy and Demography, Academy of Sciences, Dushanbeh, Tajikistan
2
Iranian Polymer &Petrochemical Institute, Tehran, Iran
1
n.garshasbi@ippi.ac.ir
2
kshyms@yandex.ru
Abstract The Kanban production authorization system
has been widely implemented as a control scheme for just-
in-time manufacturing. This system is normally applied in
systems with reliable process, low setup time, static
demand and excess capacity. The use of Kanban avoids
the need for complex information and hierarchical control
systems on the shop floor. In the present work Genetic
Algorithm (GA) approach is used to set number of
Kanbans and lot sizes at each station to achieve best
performance of the system. Productivity of the approach
is studied in term of a predefined achieved performance
value and CPU time which is an important factor to
evaluate an algorithm. System objective is derived by an
experiment and it includes minimizing average Kanban
lot queue and maximizing mean throughput rate.
Although GA is not developed for optimization problem
and it is not guaranteed that GA gives global optimum,
but it is successful in finding a satisfactory solution for the
practical problem, while CPU time used by GA is much
less than many other optimization methods.

Keywords- Genetic Algorithm , Just in time, Optimization.
I. INTRODUCTION
The Japanese market strategy seems to be rooted in
their production system, and the literature is filled with
reports of marketable quality levels, achieved along
with lower costs through higher productivity and very
low in-process inventories, namely Just In Time
Production System. Just-in-time (JIT) is really a
philosophy not a technique as such. It developed in
Japan and spread pretty fast to the Western
industrialized countries. The philosophy is aimed at
preventing all kinds of wastes, whatever their nature
and whenever they may occur. We have to make very
sincere and wholehearted effort to detect waste and
then go all out to eliminate it. The problem is mainly of
detecting the massive waste that lies all around us, most
of which is hidden from our eyes-not hidden in the real
sense of being invisible but the inability of our minds
eyes to identify the things we physically see as waste.
Real production systems are multi-stage production
systems with complicated structure and incorporate
uncertainties in the fluctuation of final demand,
machine breakdowns, defects or unavailable parts and
so on. A just in time (JIT) production system is
invented to aim chiefly at cost reduction by eliminating
unnecessary elements in production [1]. That is, a
subsequent process withdraws necessary parts from a
preceding process at a necessary point in time, and then
the preceding process produces the parts withdrawn by
the subsequent process. Two kinds of Kanbans,
production ordering and withdrawal Kanban are used as
tools to control the production and withdrawal
quantities in the process. The system under study
consists of a single production line that includes three
workstations for production and two material handling
stations in between for transportation. The system is
simulated and optimal number of kanbans at each
station with the use of a local search methods, Genetic
Algorithm are determined.
Setting the number of Kanban at each station has
been a popular research topic and researchers have
proposed several alternative approaches. These studies
fall into two categories: analytical models and
simulation modeling. Berkley B. J. [2] and Krajewski et
al. [3] introduced stochastic into the simulation models
developed in term of demand and processing time
distribution and breakdown. It also differentiates from
other studies in the literature from a statistical analysis
perspective. B. J. Berkley 1996 [4] developed a
simulation model to set container size in two-card
Kanban system. He observed smaller container size
lead to smaller average total inventories, and do not
always lead to poorer average customer service.
Smaller container size can lead to better average
customer service when the cost of greater total setup
time are offset by the benefits of more frequent material

International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459,Volume 1, Issue 1, November 2011)
2

handling, that is less frequent station material
starvation, and improved finished-goods part-mix.
In his study container size and number of Kanbans
are varied in tandem so that total in-process inventory
capacity remains constant. Yavuz I. H, et al. [5]
developed a simulation model of a mixed model just-
in-time manufacturing line. They showed the effect of
number of Kanban, number of products, outbound
inventory, demand uncertainty, scheduling rule, degree
of imbalance, number of station on seven different
performance measure. F. T. S. Chan [6] investigates the
effect of varying kanban size on the performance of
just-in-time (JIT) manufacturing systems by using
computer simulation models. Berna Dengiz et al. [7]
used a tabu search algorithm in conjunction with a
simulation model of a JIT system to find the optimum
number of kanbans. G. D. Sivakumar et al. [8]
developed a model of the multi-stage system to
traditional and adaptive kanban system. A GA-based
search is employed to set the parameters of the system.
N. Selvaraj [9] used discrete event simulation software
to determine the optimal number of kanbans in
Extended Kanban Control System. He determined
optimal number of kanbans in such a way that,
throughput is maximum, work in process is low and
machine utilization is high. S. Siha [10] developed a
continuous time Markov model for pull production
system. The various allocation pattern of Kanban
capacity and mean production time over the system
station are studied. B. J. Berkley [11] developed a
decomposition approximation using imbedded Markov
chains for two-card systems with periodic material
handling and erlang processing time. He gave several
examples to show how the approximation could be
used to find the required number of Kanbans, the
required withdrawal cycle time or both. M. G. Mitwash
[12] defined a mathematical model for the problem and
showed to reduce to a simpler problem. Using the
simpler problem, they investigate necessary and
sufficient conditions for feasibility of the problem and
specific choices of Kanban levels. These conditions are
used to define upper and lower bounds on the optimal
number of Kanbans which are employed for each item.
R. P. Philipoom et al. [13] provided an integrated
approach which address number of Kanbans, container
size and final assembly sequence simultaneously. The
solution procedure is used, is called JACKS (JIT
Algorithm for Container, Kanbans and Sequence) and it
is based on an algorithm published previously by them.
K. Ohono et al. [1] considered a JIT production system
with the production ordering and supplier Kanbans
under stochastic demand. A necessary and sufficient
condition is derived under which the JIT production
system has a stationary distribution of the backlogged
demand. An algorithm is devised for determining
optimal numbers of two kinds Kanbans that minimize
an expected average cost per period. S. K.
Chaharsooghi et al. [14] modified a JIT production
system with two kinds of kanban for a supply chain
system. He developed an algorithm to evaluate the
optimal numbers of kanbans and batch sizes of each
plant by minimizing the total cost of a chain.
As it is known the performance function of Kanban
system may have many local optimums, in such a
condition, finding best performance value through
traditional optimization methods almost is not possible.
Search algorithms give us a high chance to escape from
local optimum and move to near global solution.
Implementing of the algorithms is easy, and those have
flexibility of changing system parameters and also their
own parameters.
The application of GA to kanban system is reported
by Kochel et al. [15] and they have showed that their
approach outperforms the heuristic methods of Mitra
and Mitrani [16], [17] and of Wang and Wang [18],
[19]. Shahabudeen et al. [20] have used GA to set the
number of kanbans at each station and the lot size. A
simulation model with a single-card system has been
designed and used for analysis. Paris et al. [21]
proposed a simulation model with GA to optimize the
design options of manufacturing systems. C. Alabas et
al. [22] developed three simulation heuristic procedures
based on Genetic algorithm, Simulated annealing and
Tabu search and compared both with respect to the best
results achieved by each algorithm in a limited time
span and their speed of convergence. P. Shahabudeen et
al. [23] used Genetic algorithm and simulated
annealing-based heuristic to set the design parameters
of adaptive kanban system. In his system simulated
annealing based heuristic produces better solution with
improved computational efficiency.
The above literature review shows the recent works
done in the design of Kanban system. This study is to
set number of kanbans and lotsizes in a JIT production
with the use of GA.
A new objective function is derived based on an
experiment to catch more reliable result, differentiates
the present work from other studies.

International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459,Volume 1, Issue 1, November 2011)
3

The objective function is defined in the term of
minimizing average Kanban lot queue, maximizing
mean throughput rate and CPU time simultaneously.
II. THEORY AND FORMULATION

Genetic Algorithm

Genetic Algorithm is an optimization algorithm,
which was first suggested by John Holland [24] in his
book, Adaptation in natural and artificial systems. As
the name suggests, the process observed in natural
evolution inspired Genetic Algorithm. They attempt to
mimic this process and utilize them for solving a wide
range of optimization problems. In general, Genetic
Algorithm performs directed random search through a
given set of alternatives with the aim of finding the best
alternative with respect to given criteria goodness.
These criteria are required to be expressed in term of
objective function, which is usually referred to as a
fitness function. Genetic algorithm requires that the set
of alternatives to be searched through be finite. In the
case of an optimization problem which does not satisfy
this requirement, the set involved must be discrete and
an appropriate finite subset must be selected. It is
further required that the alternative be coded in string
of some specific finite length. These strings are called
chromosomes. Genetic algorithm searches for the best
alternative (in the sense of given objective function,
fitness) through chromosomes evolution. Basic steps
in Genetic Algorithm are shown in Figure 1. First, an
initial population of chromosomes is randomly
selected. Then each of the chromosomes in the
population is evaluated in term of its fitness. Next, a
new population of chromosomes is selected from the
given population by giving a greater chance to select
chromosomes with high fitness. This is called natural
selection.
If given stopping criteria are not met, some specific,
genetic-like operations are performed on chromosomes
of the new population. These operations produce new
chromosomes calling offsprings. The same steps of this
process, evaluation and natural selection are then
applied to chromosomes of the resulting population.
The whole process is repeated until given stopping
criteria are met. The best chromosomes in the final
population express the solution.


Fig1. High-level description of Genetic Algorithm
Design of JIT Model
The system under study consists of a single
production line that includes three workstations for
production and two material handling stations in
between for transportation. Then following
assumptions are considered.
1) Two types of job are produced in the system.
2) The rule for satisfying the demand is always
First Come First Serve.
3) Process time for each process is uniformly
distributed.
4) Setup time for one lot is exponentially
distributed.
5) Setup occur when the previous type of product
is different from the current one.
6) Inter arrival time of demands is exponentially
distributed.
7). Quantity of demand is uniformly distributed
with mean 360 and spread 120
8) Lot size ranges as 20, 40, 60, 80, 100, 120
units.
9) There is an infinite supply of raw materials at
the input of the production system.
10) Any Kanban detached at the output of a stage
is immediately available at the upstream stage.

International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459,Volume 1, Issue 1, November 2011)
4



Parameters Used in Simulation Study
Six different cases are considered which have
different system parameters. The parameters of various
cases are given below. Here process times at stations
are for one product and also service times for material
handling are for one lot.
Case 1: In this case interarrival times have exponential
distribution with mean of 120 for A product and 100
for B product.
TABLE 1
SYSTEMS PARAMETERS IN FIRST CASE
Station
No. of
Machines
Process Time
Setup
Time A B
WS1 2 U(0.5, 1.6) U (0.4, 0) Exp (3)
MH1 1 U(12, 0) U (12, 0) -
WS2 3 U(0.4, 0.2) U (0.3, 0.1) Exp (2)
MH2 1 U(12, 0) U(12, 0) -
WS3 2 U(0.3, 0.1) U(0.4, 0.2) Exp (6)

Case 2: In this case interarrival times have exponential
distribution with mean of 120 for A product and 100
for B product.
TABLE 2
SYSTEMS PARAMETERS IN SECOND CASE
Station
No. of
Machines
Process Time
Setup
Time A B
WS1 2 U(0.5, 1.6) U(.4,0) Exp(3)
MH1 1 U(100, 0) U(100, 0) -
WS2 3 U(0.4, 0.2) U(0.3, 0.1) Exp(2)
MH2 1 U(100, 0) U(100, 0) -
WS3 2 U(0.3, 0.1) U(0.4, 0.2) Exp(6)

Case 3: In this case interarrival times have exponential
distribution with mean of 120 for A product and 100
for B product.


TABLE 3
SYSTEMS PARAMETERS IN THIRD CASE
Station
No. of
Machines
Process Time
Setup
Time A B
WS1 2 U(0.5, 1.6) U(0.4, 0) Exp(3)
MH1 1 U(1, 0) U(1, 0) -
WS2 3 U(0.4, 0.2) U(0.3, 0.1) Exp(2)
MH2 1 U(1, 0) U(1, 0) -
WS3 2 U(0.3, 0.1) U(0.4, 0.2) Exp(6)

Case 4: In this case interarrival times have exponential
distribution with mean of 12 for "A" product and 10 for
"B" product. Other parameters are same as first case.
Case 5: In this case interarrival times have exponential
distribution with mean of 120 for A product and 100
for B product.
TABLE 4
SYSTEMS PARAMETERS IN FIFTH CASE
Station
No. of
Machines
Process Time
Setup
Time A B
WS1 2 U(5, 16) U(4, 0) Exp(3)
MH1 1 U(12, 0) U(12, 0) -
WS2 3 U(4, 2) U(3, 1) Exp(2)
MH2 1 U(12, 0) U(12, 0) -
WS3 2 U(3, 1) U(4, 2) Exp(6)

Case 6: In this case, all parameters are same as fifth
case, but there is no any setup time.
Formulation of Multi-Objective Criteria
A Performance measure of the system
System performance characteristics of mean
throughput rate and total number of Kanbans lots
waiting in queue are considered. The objective of the
system is to maximize throughput rate and minimize
the total number of Kanban lots waiting in queue.


International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459,Volume 1, Issue 1, November 2011)
5

B Mean throughput rate (MTR)
Cumulative mean throughput rate is defined as the
ratio of total satisfied demand to total generated
demand. Our objective is to maximize this measure:
1
1
MTR=
n
i
i
n
i
i
SD
GD
=
=


(1)
where
i
SD is satisfied demand during ith request,
i
GD
is Generated demand during ith request and n Number
of demand requests during simulation

C Aggregate Average Kanban Queue (AAKQ)
Aggregate average Kanban queue is the sum of the
average number of Kanban lots waiting in the queue at
all stations, during simulation. AAKQ is defined as
follows:
1
AAKQ=
n
j
j
I lot size
=


(2)
where ij is number of kanbans in the queue during jth
request.

Objective Function
Number of Kanbans and lot size directly affect the
performance defined above. The performance measure
can be expressed as a function of number of Kanbans at
each station and lot size. The multi-objective function
turn as maximizing mean throughput rate while
minimizing aggregate average Kanban lots queue.
Since the individual measure are moving in opposite
directions, when summed up as it, there is a chance of
good solution in each case to be nullified resulting in
poor performance measure. Hence a unified function
has to be constructed to get both the performance
measure in unidirectional. For this reason an
experiment is conducted and presented later, to
redesign AAKQ. Redesigned AAKQ (RAAKQ) is
defined as follows:
1
3816
RAAKQ
3816
n
j
j
I lot size
=

=


(3)
The combined objective function is:
{ } U max f MTR, RAAKQ = (


(4)
The above constructed objective function is used to
find the overall performance measure called Objective
value to measure the system effectiveness. The
objective function is to maximize objective value.

III. EXPERIMENTAL DESIGN AND RESULTS

Experiment 1. Drive of the RAAKQ
In this experiment, it is tried to redesign AAKQ as a
maximization objective function. For a runlength of
10000 units of time, various lot sizes, and various
numbers of Kanbans, average Kanban queue is found;
the result is given in Table 5. Based on maximum
Kanban queue that is observed AAKQ is redesigned.
This redesigned AAKQ is tested in C column that
whether it is able to give sufficient variety for AAKQ
or not.
TABLE 5
EXPERIMENT TO DRIVE A MAXIMIZATION OBJECTIVE
FUNCTION FOR AAKQ
NOK
LOT SIZE
20
LOT SIZE
60
A B C A B C
1 2.2 44 0.9885 1.5 90 0.9764
3 10.4 208 0.9455 5.0 300 0.9214
5 18.5 370 0.9030 9.5 570 0.8506
8 34.5 690 0.8192 16.6 996 0.7390
10 44.6 892 0.7662 22 1320 0.6541
13 54.9 1098 0.7123 26.9 1614 0.5770
15 66 1320 0.6541 31.5 1890 0.5047
NOK
LOT SIZE
120
A B C
1 1.9 228 0.9402
3 4.3 516 0.8648
5 7.8 936 0.7547
8 14 1680 0.5597
10 17.6 2112 0.4465
13 22.9 2780 0.2715
15 26.5 3180 0.1667
A= Average Kanban Queue Length B= ALot Size
C= Redesigned AAKQ Value (RAAKQ)
* RAAKQ= 1-B/(31801.2)

The range of calculated RAAKQ in the table is
recommending the use of redesigned AAKQ.

International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459,Volume 1, Issue 1, November 2011)
6

Experiment 2.Effect of number of Kanbans and Lot size

This experiment shows the effect of number of
Kanbans and lot sizes on the MTR and the RAAKQ.
These effects are illustrated in Figure 2 to 9. As it is
observed by increasing number of Kanbans the
RAAKQ is decreased rapidly. So in our study we limit
the number of Kanbans to fifteen. Increasing lot sizes
causes in increasing the MTR and at the same time
decreasing the RAAKQ. So the lot sizes are not limited.


Fig 2 Effect of Number of Kanbans on RAAKQ and MTR
Lot Size = 20



Fig 3 Effect of Number of Kanbans on RAAKQ and MTR
Lot Size = 60



Fig 4 Effect of Number of Kanbans on RAAKQ and MTR
Lot Size = 100



Fig 5 Effect of Number of Kanbans on RAAKQ and MTR
Lot Size = 120

Fig 6 Effect of Lot Size on RAAKQ and MTR, No. of
Kanbans = 1



International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459,Volume 1, Issue 1, November 2011)
7


Fig 7 Effect of Lot Size on RAAKQ and MTR, No. of
Kanbans = 5



Fig 8 Effect of Lot Size on RAAKQ and MTR, No. of
Kanbans = 10



Fig 9 Effect of Lot Size on RAAKQ and MTR, No. of
Kanbans = 15

Experiment 3.
The genetic algorithm is performed for the defined
system in the six different cases. The GA is done six
times for each case. Table 6 and Table 7 show the
objective values and average used time (CPU time).
TABLE 6
OBJECTIVE VALUES AND AVERAGE SYSTEM TIMES FOR
VARIOUS CASES OBTAINED BY GENETIC ALGORITHM
CASE
GA
Trial 1 Trial 2 Trial 3 Trial 4
Obj-val Obj-val Obj-val Obj-val
1 1.7744 1.7434 1.7652 1.5674
2 0.9006 1.0333 0.9307 0.8549
3 1.9069 1.6642 1.9069 1.6467
4 0.8997 0.9433 0.8868 0.9040
5 0.9691 1.0109 1.0022 0.9018
6 0.9847 1.0574 0.9688 1.0571
CASE
GA
Trial 5 Trial 6 Average
Time Obj-val Obj-val
1 1.5473 1.7398 0.0994
2 0.9513 0.9188 0.0365
3 1.4914 1.6178 0.1353
4 0.9732 0.8555 0.1681
5 0.9985 0.9704 0.0252
6 1.0065 1.0341 0.0290

TABLE 7
SYSTEM TIMES AND AVERAGE OBJECTIVE VALUES FOR
VARIOUS CASES OBTAINED BY GENETIC ALGORITHM
CASE
GA
Trial 1 Trial 2 Trial 3 Trial 4
System
Time
System
Time
System
Time
System
Time
1 0.0951 0.0941 0.0972 0.1360
2 0.0363 0.0357 0.0375 0.0381
3 0.1810 0.0955 0.1702 0.1488
4 0.1448 0.1566 0.1777 0.1885
5 0.0254 0.0262 0.0225 0.0214
6 0.0236 0.0325 0.0225 0.0252
CASE
GA
Trial 5 Trial 6
Average
Objective Value
System
Time
System
Time
1 0.0890 0.0852 1.6896
2 0.0350 0.0380 0.9316
3 0.1086 0.1079 1.7056
4 0.1630 0.1781 0.9104
5 0.0244 0.0312 0.9755
6 0.0403 0.0298 1.0181

International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459,Volume 1, Issue 1, November 2011)
8


IV. CONCLUSION

Simulation model of a two card Kanban system has
been developed. An objective function consisting of
Kanban size, waiting time and throughput rate has been
developed. Number of Kanbans at each station and lot
sizes are considered as two decision parameters, to
improve objective function. Genetic Algorithm have
been employed to achieve an improved solution. Effect
of number of Kanbans and lot sizes on the two
objective functions (mean throughput rate, MTR, and
redesigned aggregate average Kanban queue, RAAKQ,
) is studied. The study concluded that for the present
system as the number of kanbans or lotsizes icreased,
the MTR increased, while RAKKQ decreased.
With the use of GA in a very low time most
important decision in a JIT production system is made.
It means that for six different cases with various system
parameters number of kanbans and lotsizes at each
station is calculated. As it is known in multiobjective
optimization problems there exist a number of solutions
which are optimum in some sense. The problems
discussed in the present work, also have many
satisfactory answers which provide another advantage
of using GA to make the decision. GA is able to
produce a huge number of answers to the problem and
the user can select the most appropriate solution.
Among a number of solutions for each cases a complete
solution for each case using GA are given in the Table
8.
TABLE 8
IMPROVED SOLUTIONS FOR VARIOUS CASES USING GA
KANBANS
GA
CASE
1
CASE
2
CASE
3
A B A B A B
POK-WS3 2 8 11 6 2 8
POK-WS2 4 8 6 10 4 8
POK-WS1 4 7 8 5 4 7
WK-WS32 14 4 11 6 14 4
WK-WS21 4 7 8 10 4 7
LS 2 3 3 3 2 3
OBJ_VAL 1.7744 1.0333 1.9069
KANBANS
GA
CASE
4
CASE
5
CASE
6
A B A B A B
POK-WS3 2 8 4 6 4 4
POK-WS2 4 8 4 5 6 6
POK-WS1 4 7 11 12 8 9
WK-WS32 14 4 5 9 4 8
WK-WS21 4 7 12 11 8 9
LS 2 3 3 3 6 4
OBJ_VAL 0.9433 1.0109 1.0571
A: Product A B: Product B
LS: Lot Size
POK: Production Order Kanban
WK: Withdrawal Kanban WS: Work Station




References

[1] K. Ohno 1995. Optimal numbers of two kinds of Kanbans in a
JIT production, Prod. Research, Vol. 33, No.5, pp. 1387-1401.
[2] Berkley B. J, 1992. A review of the Kanban production control
research literature, Production and Operation Management, Vol.
1, No. 4, pp. 392-411.
[3] Krajewski L. J, King B. E, Ritzman L. P and Wong D. S, 1987.
Kanban, MRP, and shaping the manufacturing environment,
Management Science, Vol. 33, No. 1, pp. 39-57.
[4] Berkley B. J., 1996. A simulation study of container size in
two-card Kanban system, International Journal of Production
Research, Vol. 34, No. 12, pp. 3417- 3445.
[5] Yavuz I. H. and Satir A, 1995. A Kanban based simulation
study of a mixed model just-in-time manufacturing line,
International Journal of Production Research, Vol. 33, No. 4,
pp. 1027-1048.
[6] F. T. S. Chan, 2001. Effect of kanban size on just-in-time
manufacturing systems, Journal of Materials Processing
Technology, Vol. 116, No. 2-3, pp. 146-160.
[7] Berna Dengiz & Cigdem Alabas, 2000. Simulation optimization
using tabu search, Proceeding of the 2000 Winter Simulation
Conference, Vol. 1, pp. 805 810.
[8] G. D. Sivakumar and P. Shahabudeen, 2008. Design of multi-
stage adaptive kanban system, The International Journal of
Advanced Manufacturing Technology, Vol. 38, No. 3-4, pp.
321-336.

International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459,Volume 1, Issue 1, November 2011)
9

[9] N. Selvaraj, 2009. Determining the Number of Kanbans in
EKCS: A Simulation Modeling Approach, Proceedings of the
International Multi Conference of Engineers and Computer
Scientists, Vol. 2, IMECS 2009, March 18-20, 2009, Hong
Kong (a).
[10] Siha S, 1994. The pull production system: modeling and
characteristics, International Journal of Production Research,
Vol. 32, No. 4, pp. 933-949.
[11] Berkley B. J., 1994. Testing minimum performance levels for
Kanban-controlled lines, International Journal of Production
Research, Vol. 32, No. 1, pp. 93-109.
[12] Mitwasi M. G and Askin R. G, 1994. Production planning for a
multi-item, single-stage Kanban system, International Journal of
Production Research, Vol. 32, No. 5, pp. 1173-1195.

[13] Philipoom R. P, Rees L. P and Taylor B. W III, 1996.
Simultaneously determining the number of Kanbans container
size and the final assembly sequence of product in a just-in-time
shop, International Journal of production Research, , Vol. 34,
No. 1, pp. 51-69.
[14] S. K. Chaharsooghi and A. Sajedinia, 2010. Determination of
Number of Kanbans and Batch Sizes in a JIT Supply Chain
System, Scientia Iranica: Transaction on Industrial Engineering,
Vol. 17, No. 2, pp. 143-149.
[15] Kochel P and Nielander U, 2002. Kanban optimization by
simulation and evolution, Production Planning and Control,
Vol. 13, No. 8, pp.725734.
[16] Mitra D and Mitrani I, 1990. Analysis of a kanban discipline for
cell coordination in production lines I, Management Science,
Vol. 36, No. 12, pp. 15481566.
[17] Mitra D and Mitrani I, 1991. Analysis of a Kanban discipline
for cell coordination in production lines II, Stochastic demands,
Operations Research, Vol. 35, No. 5, pp. 807823.
[18] Wang H and Wang H. P, 1990. Determining number of
kanbans: Step toward non-stock production, International
Journal of Production Research, Vol. 28, pp. 21012115.
[19] Wang, H and Wang H. P, 1991. Optimum number of kanbans
between two adjacent workstations in a JIT system,
International Journal of Production Economics, Vol. 22, No. 3,
pp. 179188.
[20] Shahabudeen P and Krishnaiah K, 1999. Design of bi-criteria
kanban system using Genetic Algorithm, International Journal
of Management and System, Vol. 15, pp. 257274.
[21] Paris J. L, Tautou-Guillaume L, & Pierreval H, 2001. Dealing
with design options in the optimization of manufacturing
systems an evolutionary approach, International Journal of
Production Research, Vol. 39, No. 6, pp. 10811094.

[22] C. Alabas, F. Altiparmak and B Dengiz, 2002. A comparison of
the performance of artificial intelligence techniques for
optimizing the number of kanbans, Journal of Operational
Research Society, Vol. 53, pp. 907-914 (c).
[23] P. Shahabudeen and G. D. Sivakumar, 2008. Algorithm for the
design of single-stage adaptive kanban system, Computers &
Industrial Engineering, Vol. 54, pp. 800820 (b).
[24] Holland J. H, 1975. Adaptation in natural and artificial systems,
Ann Abor, University of Michigan

Você também pode gostar