Вы находитесь на странице: 1из 14

This article was downloaded by: [National Institute of Technology - Calicut] On: 30 October 2012, At: 06:23 Publisher:

Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

International Journal of Production Research


Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/tprs20

U-shaped assembly line balancing problem with genetic algorithm


Rea Kook Hwang , Hiroshi Katayama & Mitsuo Gen
a a a b

Graduate School of Science&Engineering, Waseda University, 3-4-1 Ohkubo, Shinjuku Ward, Tokyo, 169-0072, Japan
b

Graduate School of Information, Production&Systems, Waseda University, 2-7 Hibikino, Wakamatsu-ku, Kitakyushu, Fukuoka, 808-0135, Japan Version of record first published: 01 Jul 2008.

To cite this article: Rea Kook Hwang, Hiroshi Katayama & Mitsuo Gen (2008): U-shaped assembly line balancing problem with genetic algorithm, International Journal of Production Research, 46:16, 4637-4649 To link to this article: http://dx.doi.org/10.1080/00207540701247906

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-andconditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

International Journal of Production Research, Vol. 46, No. 16, 15 August 2008, 46374649

U-shaped assembly line balancing problem with genetic algorithm


REA KOOK HWANG*y, HIROSHI KATAYAMAy and MITSUO GENz
yGraduate School of Science & Engineering, Waseda University, 3-4-1 Ohkubo, Shinjuku Ward, Tokyo, 169-0072, Japan zGraduate School of Information, Production & Systems, Waseda University, 2-7 Hibikino, Wakamatsu-ku, Kitakyushu, Fukuoka, 808-0135, Japan

Downloaded by [National Institute of Technology - Calicut] at 06:23 30 October 2012

(Revision received January 2007) This paper presents a multi-objective genetic algorithm (moGA) to solve the U-shaped assembly line balancing problem (UALBP). As a consequence of introducing the just-in-time (JIT) production principle, it has been recognized that U-shaped assembly line systems offer several benefits over the traditional straight line systems. We consider both the traditional straight line system and the U-shaped assembly line system, thus as an unbiased examination of line efficiency. The performance criteria considered are the number of workstations (the line efficiency) and the variation of workload. The results of experiments show that the proposed model produced as good or even better line efficiency of workstation integration and improved the variation of workload. Keywords: U-shaped Priority-based coding assembly line; Multi-objective genetic algorithm;

1. Introduction Recently, assembly lines have become the primary mechanism of productive systems. Although the problems relative to the assembly line are of a great variety in reality, the literature on the subject has been focusing on a few specific problems. Among the many new production lines, most are being arranged as U-shaped lines rather than straight lines as a consequence of the use of the JIT production principle. JIT is an umbrella term for a number of principles or techniques typical U-shaped lines, pull production control, and quality management (Wemmerlov and Hyer 1989). The purpose is to improve product quality and cost by eliminating all waste in the production system (Brown and Mitchell 1991). In cellular manufacturing, dissimilar but sequentially related machines are clustered near each other to meet the processing needs of a family of products (Wemmerlov and Hyer 1989). In JIT, each cluster or cell is further improved by moving employees, workstations, or both into a U-shaped configuration that increases the possible

*Corresponding author. Email: rkhwang@kata.mgmt.waseda.ac.jp


International Journal of Production Research ISSN 00207543 print/ISSN 1366588X online 2008 Taylor & Francis http://www.tandf.co.uk/journals DOI: 10.1080/00207540701247906

4638

R. K. Hwang et al.

interaction among employees (Brown and Mitchell 1991). While it is in a JIT production environment that the U-shaped line takes its most complete form, it becomes an optimal production principle when the U-shaped line is in place. The U-shaped line complements the JIT principle by providing more alternatives. It is able to assign tasks to workstations (operators) better than comparable straight-line systems because operators can handle adjacent tasks, as well as tasks on both sides of the U-shaped line. Furthermore, as another advantage, the U-shaped line simultaneously maximizes both the use of operational work space and operator communication and trust, such that machines take up less space and workers are closer to one another. Although it is compact and effective, it is not perfectly efficient due to a lack of workload balance. Aside from the many advantages of the U-shaped line system, there are still issues that need to be addressed as to how to improve its efficiency. This paper illustrates a potential solution to assembly line workload balance, whether it be in a U-shaped line or a straight line, since workload balance seems to be the most important element of both lines. It is, however, the U-shaped line and JIT principle that has given rise to this problem because it is such a compact and continuous operation that it constantly demands a solution for balancing line ergonomics. Since the JIT and U-shaped line system is still a cutting edge integrative design, it also encompasses newer problems, as well, which when solved will also complement traditional line system designs. Realistically at this juncture, the problem now surfaced can be equally attributable to both types of lines and an overall solution to workload balance. Therefore, both types of lines will be included in the following examination as containing similar problems and thus will be beneficiaries of the elemental need for a balanced workload solution. Nowadays, evolutionary approaches have been developed to solve this problem. For example, genetic algorithm (GA) based approaches have been applied successfully in many research fields over the last decade (Gen and Cheng 1997, Runarsson and Jonsson 1999). The proposed GA point of concern is a search algorithm based on the principles of evolution and natural genetics. In this case, the GA combines the exploitation of the past results with the new areas of space search exploration. By using survival of the fittest techniques, combined with a structured yet randomized information exchange, a GA can mimic some of the innovative attributes of a human search. We used the multi-objective genetic algorithm (moGA) with the priority-based encoding method (Gen and Cheng 2000), which is the knowledge of how to handle the problem of producing encoding, to treat the precedence constraints efficiently. This is done because there is scope for improvement in the variation of workload with former priority-based genetic algorithm (PGA). For implementing our proposed algorithm, we have two objectives: minimizing workstations (maximizing line efficiency) and minimizing the variation of workload. The rest of this paper is organized as follows: section 2 defines the problem of assembly line balancing; section 3 describes the proposed framework of proposed design procedure; section 4 considers the numerical experiments involving the proposed method; and conclusions and necessary future research are finally discussed in section 5.

Downloaded by [National Institute of Technology - Calicut] at 06:23 30 October 2012

U-shaped assembly line balancing problem with genetic algorithm

4639

2. Assembly line balancing problem Assembly line balancing problems fall into the NP-hard class of combinatorial optimization problems. An assembly line consists of workstations i 1, . . . , m arranged along a conveyor belt or similar mechanical material-handling equipment. The jobs are consecutively launched down the line and are moved from station to station. At each station, certain operations are repeatedly performed regarding the cycle time (maximum or average time available for each work cycle). The decision problem of optimally partitioning (balancing) the assembly work among the workstations with respect to some objective is known as the simple assembly line balancing problem (SALBP). Manufacturing a product on an assembly line requires partitioning the total amount of work into a set of elementary operations named tasks V {1, . . . , n}. Performing a task j takes a task time tj and requires certain machine equipment and/or skills of workers. Precedence constraints between the tasks have to be observed so that necessary technological and organizational conditions can be satisfied. These elements can be summarized and visualized by a precedence diagram with assembly network. It contains a node for each task, node weights for the task times, and arcs for the precedence constraints. Assembly line balancing problems (ALBP) are classified into two types, type I _ and type II (Hackman et al. 1989). In type I problems, the cycle time, assembly tasks time, and precedence requirements are given and minimize the number of workstations. These type I challenges generally occur when designing new assembly lines, where the purpose is to achieve a forecast demand for the number of _ workstations to be reduced. In type II problems, when the number of workstations is fixed, the dilemma is to minimize the cycle time. Solving this problem will maximize _ the production rate. This type II problem occurs when the organization wants to produce the optimum number of items by using a fixed number of workstations without adding new machines. However, this paper considers the type I problem with workload variation. Furthermore, we assume that each station is able to produce any product in any cycle. Figure 1 shows a precedence diagram with n 11 tasks having task times between 1 and 7 (time units). The precedence constraints for, for example, task 7 express that its processing requires that tasks 3, 4, 5 (direct predecessors) and 1 (indirect predecessor) be completed. The other way around, task 7 must be completed before its (direct and indirect) successors 9 and 11 can be started. Any type of SALBP consists of finding a feasible line balance, i.e., an assignment of each task to a station such that the precedence constraints and further restrictions are fulfilled. The set Si of tasks assigned to a station i (1, . . . , m) constitutes its P station load. The cumulated task time tSi j2Si tj is called the station time. When a fixed common cycle time c is given, a line balance is only feasible if no station time exceeds c. In the case of t(Si)5c, the station i has an idle time of c t(Si) time units in each cycle. For the example of figure 1, a feasible line balance with cycle time c 10 and m 6 stations is given by the station loads S1 {1, 2, 6}, S2 {4, 5}, S3 {3, 7}, S4 {8}, S5 {9, 10}, S6 {11}. While no idle time occurs in stations 1 and 5, stations 2, 3, 4 and 6 show idle times of 2, 2, 4 and 6, respectively as shown in figure 2.

Downloaded by [National Institute of Technology - Calicut] at 06:23 30 October 2012

4640

R. K. Hwang et al.
2: Execution time 2 7 6 1 1 5 5 3 4 5 3 7 9 2 6 6 8 5 10 4 11

Downloaded by [National Institute of Technology - Calicut] at 06:23 30 October 2012

Figure 1.

Precedence diagram with assembly network (Jackson 1956).

S1: 10 1 2 6

S2: 8 4 5 idle 3

S3: 8 7 idle 8

S4: 6 idle 9

S5: 10 10 11

S6: 4 idle

Figure 2.

Completed line assignment for straight assembly line.

Of course UALBP can undergo the same generalizations as SALBP with respect to cost-oriented objectives, paralleling, workstation selection, and so on. Every solution feasible for SALBP is feasible for UALBP as well, because a U-shaped line does not need to include crossover stations. However, the optimal UALBP solution may have an improved line efficiency compared to the optimal SALBP solution due to the increased possibilities of combining tasks to station loads. Applying the example of figure 1 to UALBP with given c 10, an optimal solution with m 5 can be found, which is shown in figure 3. The first station starts executing task 1 but gets back every task at the end of the production process to perform task 11. It can obtain a feasible line balance with cycle time c 10 and m 5 stations given by the station loads S1 {1, 11}, S2 {2, 4, 5}, S3 {6, 7, 9}, S4 {3, 10}, S5 {8}. While no idle time occurs in stations 1, 2, 3 and 4, station 5 shows an idle time of 4.

3. The proposed multi-objective genetic algorithm 3.1 Genetic representation We proposed the operation of the priority-based GA using the precedence diagram with an assembly network as shown in figure 1. The proposed representation process can be thought of as consisting of two parts: the assignment of tasks to workstations and the task operation ordering within a workstation. To work effectively with precedence constraints, we used the priority-based encoding method (Gen and Cheng 2000). The position of a gene was used to represent a task node, and the value of the gene was used to represent the priority of the task node for constructing a task sequence among candidates. As is the proposed encoding method, first randomly generate the initial chromosome as shown in procedure 1 (figure 4). Each chromosome position is called a gene. Each gene will use the priority of nodes in an assembly network.

U-shaped assembly line balancing problem with genetic algorithm

4641

1 S1: 10

6 S3: 10

S4: 10 S5: 6

S2: 10

11

10

idle

Downloaded by [National Institute of Technology - Calicut] at 06:23 30 October 2012

Figure 3.

Completed line assignment for U-shaped assembly line.

Figure 4.

An example priority-based encoding procedure.

This encoding method easily verifies any permutation of the encoding to correspond to the sequences, so that most existing genetic operators can easily be applied to the encoding. Suppose we want to assign n tasks to m workstations by using the above mentioned chromosome. We shall use the precedence diagram of assembly network as shown in figure 1 to illustrate. Initially, we try to find a node for the first position. Nodes 1 and 11 are eligible for the first position, and therefore suitable for the start node. The priorities of nodes 1 and 11 are 7 and 3 respectively so task 1, having the highest priority, is placed into the task sequence TS. For the second position the possible nodes are 2, 3, 4, 5 and 11 which have priorities 2, 10, 4, 5 and 3 respectively, so task 3 is placed next into the task sequence TS. We repeat these steps until we obtain a complete task sequence TS {1, 3, 5, 4, 7, 9, 11, 2, 6, 8, 10}, which is the priority-based decoding procedure 2 (figure 5). In the next step, we assign tasks to workstations from the above task sequence TS, which corresponds to procedure 3 in figure 6. In the example, we obtained a feasible line balance with cycle time c 7 and m 7 stations given by the station loads S1 {1, 5}, S2 {2, 3}, S3 {4}, S4 {7, 11}, S5 {6, 9}, S6 {8}, S7 {10}. While no idle time occurs in stations 1, 2, 3, 4 and 5 stations 6 and 7 show idle times of 1 and 2, respectively. Table 1 describes the trace table for the overall procedure.

4642

R. K. Hwang et al.

Downloaded by [National Institute of Technology - Calicut] at 06:23 30 October 2012

Figure 5.

Priority-based decoding procedure.

Figure 6.

Priority-based decoding procedure.

3.2 Crossover Here the position-based crossover operator, by the weight mapping crossover (WMX) (Hwang et al. 2006) that we used, can be viewed as a two-point crossover of a real number string and a remapping by order of different real number strings. This WMX operator is shown in figure 7.

3.3 Mutation We used the swap mutation operator, in which two positions are selected at random and their contents are swapped as shown in figure 8.

U-shaped assembly line balancing problem with genetic algorithm Table 1. j 0 11 1 4 2 5 6 9 7 10 3  S {1, 11} {2, 3, 4, 5, 11} {2, 4, 5, 11} {2, 4, 11} {2, 7, 11} {2, 9, 11} {2, 11} {2, 10} {6, 10} {8, 10} {10} Trace table of example for figure 1. v( j ) v(1) 7, v(11) 3 v(2) 2, v(3) 10, v(4) 4, v(5) 5, v(11) 3 v(2) 2, v(4) 4, v(5) 5, v(11) 3 v(2) 2, v(4) 4, v(11) 3 v(2) 2, v(7) 9, v(11) 3 v(2) 2, v(9) 6, v(11) 3 v(2) 2, v(11) 3 v(2) 2, v(10) 1 v(6) 8, v(10) 1 v(8) 11, v(10) 1 v(10) 1 j*(tj) 1(6) 3(5) 5(1) 4(7) 7(3) 9(5) 11(4) 2(2) 6(2) 8(6) 10(5) Si

4643

Downloaded by [National Institute of Technology - Calicut] at 06:23 30 October 2012

S1:{l} 6(l) S2:{3} 5(2) S1:{l, 5} 7(0) S3:{4} 7(0) S4:{7} 3(4) S5:{9} 5(2) S4:{7, 11} 7(0) S2:{3, 2} 7(0) S5:{9, 6} 7(0) S6:{8} 6(1) S7:{10} 5(2)

step 1: select the substring at random parent 1 : parent 2 : 7 3 2 6 10 10 4 9 5 8

substring selected 8 5 9 10 11 2 6 7 1 11 3 4

step 2: determine mapping relationship 3 8 5 9 10 11 2 6 7 8 2 1 4 2 7 3 3 5 1 2 4 1 2 9 4

9 11 6

8 11 6

3 1 4 5 10 2

7 10 2

step 3: legalize offspring with mapping relationship offspring 1 : offspring 2 : 7 3 2 6 10 10 4 9 5 8 8 5 11 7 6 10 9 2 1 11 3 4

Figure 7.

Two point-based weight mapping crossover (WMX) operator.

Exchanging points

parent :

10

11

offspring :

11

10

Figure 8.

Swap mutation operator.

4644

R. K. Hwang et al.

3.4 Evaluation function and selection Because of the long-term effect of balancing decisions, the objectives to be carried out have to be carefully chosen with consideration of the strategic goals of the enterprise. From an economic point of view, cost and profit related objectives should be considered. However, measuring and predicting the cost of running a line over months or years and the profits achieved by selling the products assembled, is rather complicated and error-prone. A usual surrogate objective consists of maximizing the line utilization. This is measured by the line efficiency as the productive fraction of the lines total operating time, and it directly depends on the cycle time c and the number of stations m (Scholl and Becker 2006). The evaluation function is essentially the objective function for the problem. It provides a means of evaluating the search nodes, and it also controls the selection process. Here, the calculation of evaluation functions is composed of two factors, line efficiency and variation of workload. First fitness function ( f1): maximizes the line efficiency (E ): f1 E tsum : m c 1

Downloaded by [National Institute of Technology - Calicut] at 06:23 30 October 2012

And the second fitness function ( f2): minimizes the variation of workload (V ): s Pm 2 i1 Ui aver f2 V , 2 m P where aver m Ui =m is an average utilization of total workstations and i1 Ui tSi =tSm max is the utilization of the workstation Si. i1 Genetic algorithms have been applied mainly to a single objective, but many real-world problems involve multiple objectives. When a single objective GA is applied to a multi-objective optimization problem, multiple objective functions should be combined into scalar fitness functions. For the assembly line balancing problem, we consider evaluation factors such as line efficiency (E ) and the variation of workload (V ) simultaneously. In this paper, we combine adaptive weight approach (AWA) (Gen and Cheng 2000) that utilizes some useful information from the current population to readjust weights for obtaining a search pressure toward a positive ideal point. For the examined solutions at each generation, we define two extreme points (maximum: f , minimum: f ) as: f f max , f max 3 1 2 f f min , f min , 1 2 4

where f max and f min are the maximal and minimal values for the qth objective as q q defined by the following equations: f max maxf fq Vk g, q
j

q 1, 2 q 1, 2:

5 6

f min minf fq Vk g, q

U-shaped assembly line balancing problem with genetic algorithm


f2 Minimal rectangle containing all current solutions (f1min, f2max) f + Maximum extreme point (f1max, f2min)

4645

Positive ideal point Whole criteria space

f2max f

f2min

Adaptive moving line

Downloaded by [National Institute of Technology - Calicut] at 06:23 30 October 2012

Minimum Subspace extreme point corresponding to current solutions f1min f1max f1

Figure 9.

Adaptive weights and adaptive hyper plane.

The equation driven above is a hyper plane defined by the following extreme points in current solutions: max f min f1 2 min 7 f1 f max : 2 Adaptive moving line defined by the extreme points f max , f min and f min , f max are 2 1 1 2 shown figure 9. The weighted-sum objective function for a given chromosome Vk is given by the following equation: FVk
2 X q1

wq fq Vk

2 X

fq Vk , f max f min q q1 q

where wq is adaptive weight for objective q: wq f max q 1 , f min q q 1, 2: 9

The evaluation function is designed as follows: evalVk 1 FVk


2 q1

1  : fq Vk =f max f min q q

10

We use the roulette wheel selection (Holland 1975), which is a method to reproduce a new generation proportional to the fitness of each individual. In this procedure, the solutions are placed on a roulette wheel where the section of the wheel for a better solution is larger than for a poorer solution. It is called the roulette wheel selection because the selection technique of the parent selection is that each individual is given a

4646

R. K. Hwang et al.

chance to become a parent in proportion to its fitness evaluation; the best chances of selecting a parent can be produced by a spinning roulette wheel with the size of its slots for each parent being proportional to their fitness. Obviously those with the largest fitness (and slot sizes) have more of a chance to be chosen, which is, needless to say, analogous to the notion of the survival of the fittest in humans and animals.

4. Numerical experiment

Downloaded by [National Institute of Technology - Calicut] at 06:23 30 October 2012

Several well-known test problems considered by Talbot et al. (1986) are solved using the proposed multi-objective genetic algorithm. We demonstrate using both line types; a straight line and a U-shaped line, and describe the performance of the line efficiency and the variation of workload in table 2. By using two types of fitness function we can check the effectiveness of multi-objective functions. The genetic algorithm applied the following parameters throughout the simulations: . . . . . Population size: Maximum generation: Crossover probability: Mutation probability: Terminating condition: popSize 100 maxGen 1000 pC 0.7 pM 0.3 100 generations with same fitness

The results of the experimental test are presented in table 2. We compared the multi-objective (EV ) with the single objective (E ). We also compared results of the proposed multi-objective genetic algorithm with those in the previous literature (Integer Programming and Maximum Ranked Positional Weight Method). The average CPU time (20 iterations) for the proposed moGA is shown in figure 10. As can be seen, the U-shaped line with the proposed moGA takes little execution time because of the increased number of generations. The execution time depends on the number of generations and population size. Although the U-shaped line with proposed moGA takes more CPU time, the CPU time requirement of the proposed moGA is reasonable and workable. As table 2 shows, the proposed multi-objective genetic algorithm performs well throughout on the test data. The bold characters indicate the results of improvement. In cases of small size problems, we obtained small improvement of line balance, but in many cases of large size problems, the variation of workload improves. Especially, the U-shaped line performed line balancing (variation of workload) well because of the increased possibilities of combining tasks to station loads. In view of the results achieved so far, the proposed moGA improved the variation of workload positively. Thus, the more confident and better variation of workload means an improved line balance has taken place.

5. Conclusion In this paper, we presented a multi-objective genetic algorithm using the priority-based coding method to improve a variation of workload on assembly line production systems. A moGA using the priority-based coding method was

Downloaded by [National Institute of Technology - Calicut] at 06:23 30 October 2012

Table 2. Straight line Proposed GA Fitness function (E ) E V No. of St. E V E V No. of St. No. of No. of St. St. No. of St. Fitness function (EV ) Int. Pro. MRPW solution solution Fitness function (E ) Proposed GA Fitness function (EV ) E V U-Shaped line

Solutions for comparison problems.

Int. Pro. solution

Problem

No. of Cycle tasks time

No. of St.

No. of St.

Mitchell

21

Heskia

28

Sawyer

30

Kibridge

45

Tonge

70

U-shaped assembly line balancing problem with genetic algorithm

Arcus

83

Arcus

111

14 15 21 138 205 324 27 33 54 79 92 184 176 364 468 5853 6842 8412 10 816 5755 10 027 10 743 17 067 0.042 0.090 0.000 0.022 0.001 0.153 0.035 0.046 0.030 0.004 0.000 0.000 0.031 0.039 0.014 0.019 0.043 0.038 0.160 0.044 0.030 0.035 0.006

8 8 5 8 5 4 13 11 7 7 6 3 21 10 8 13 12 10 8 27 16 15 9

8 8 5 8 5 4 13 11 7 7 6 3 21 10 8 13 12 10 8 27 16 15 9

93.7% 87.5% 100.0% 92.7% 99.9% 79.0% 92.3% 89.2% 85.7% 99.8% 100.0% 100.0% 95.0% 96.4% 93.8% 92.4% 92.2% 90.0% 87.5% 96.8% 93.7% 93.3% 97.9%

0.057 0.090 0.000 0.054 0.001 0.329 0.107 0.109 0.233 0.004 0.000 0.000 0.052 0.053 0.129 0.092 0.129 0.160 0.211 0.049 0.093 0.123 0.033

8 8 5 8 5 4 13 11 7 7 6 3 21 10 8 13 12 10 8 27 16 15 9

93.7% 87.5% 100.0% 92.7% 99.9% 79.0% 92.3% 89.2% 85.7% 99.8% 100.0% 100.0% 95.0% 96.4% 93.8% 92.4% 92.2% 90.0% 87.5% 96.8% 93.7% 93.3% 97.9%

8 8 5 8 5 4 13 10 6 7 6 3 21 10 8 13 12 10 8 27 16 15 9

8 8 6 8 6 4 14 11 7 8 7 4 21 10 8 14 12 10 8 27 16 15 9

8 8 5 8 5 4 13 10 6 7 6 3 21 10 8 13 12 10 8 27 16 15 9

93.7% 87.5% 100.0% 92.7% 99.9% 79.0% 92.3% 98.1% 100.0% 99.8% 100.0% 100.0% 95.0% 96.4% 93.8% 92.4% 92.2% 90.0% 87.5% 96.8% 93.7% 93.3% 97.9%

0.055 0.139 0.000 0.112 0.001 0.332 0.071 0.024 0.000 0.004 0.000 0.000 0.043 0.023 0.063 0.061 0.097 0.123 0.200 0.036 0.074 0.063 0.019

8 8 5 8 5 4 13 10 6 7 6 3 21 10 8 13 12 10 8 27 16 15 9

93.7% 87.5% 100.0% 92.7% 99.9% 79.0% 92.3% 98.1% 100.0% 99.8% 100.0% 100.0% 95.0% 96.4% 93.8% 92.4% 92.2% 90.0% 87.5% 96.8% 93.7% 93.3% 97.9%

0.023 0.023 0.000 0.007 0.001 0.062 0.023 0.014 0.000 0.004 0.000 0.000 0.029 0.009 0.004 0.014 0.019 0.038 0.089 0.019 0.029 0.031 0.004

4647

Note: the proposed GA implemented on Intel(R) Pentium 4 with 1 GB RAM and 3.20 GHz speed. Int. Pro.: Integer Programming, MRPW: Maximum Ranked Positional Weight.

4648
Time (minutes) 1.2

R. K. Hwang et al.

STRAIGHT LINE 1.0

U-SHAPED LINE

0.8 CPU time

0.6

Downloaded by [National Institute of Technology - Calicut] at 06:23 30 October 2012

0.4

0.2

0
14 15 21 138 205 324 27 33 54 79 92 184 176 364 468 5853 6842 8412 10816 5755 10027 10743 17067

Mitchell 21

Heskia 28

Sawyer 30

Kibridge 45 Problem sets

Tonge 70

Arcus 83

Arcus 111

Figure 10. CPU time of problem sets for the moGA.

proposed for solving the U-shaped assembly line problem. Priority-based coding is a key factor of how to encode a problem solution into a chromosome that conditions all subsequent steps in genetic algorithms. We described both a straight and U-shaped assembly line balancing problem. The proposed priority-based encoding handled the problem of how to produce a chromosome to work efficiently, we used a weight mapping crossover (WMX). In a numerical experiment, we used several well-known task graphs for the ALBP, and we compared proposed moGA with the previous literature. From the result it was confirmed that the genetic algorithm can provide workable solutions whereas the optimal UALBP solution has an improved line efficiency compared to the optimal SALBP solution in some cases. Again, this is due to the increased possibilities of combining tasks to station loads. Moreover, the proposed moGA confirmed that it, as a priority-based genetic approach, is able to find a better solution focusing on working out a large proportion of problems to make significant balancing improvements. The simulation has shown that it is possible to solve generally formulated problems as well as tightly defined problems. Finally, the proposed approach is able to find numerous feasible solutions in one round of genetic approach simulation. Also, the coding method can be applied to other distribution scheduling problems. All in all, this will help the designer to choose a suitable alternative for assembly line design and be able to construct a flexible assembly operating system. Most importantly with the use of our approach we hope a designer will be able to input the calculation of the organizations objectives precisely of long-term workload balance without further imbalance. Thus far, we have made progress, and although other academics may research where we depart, as of now we plan to move forward and re-situate our work in a new model to test its limits yet again. In the next step of our study we plan to move the focus of many items, including the balancing with sequencing problem, into the framework of a mixed-model assembly line system.

U-shaped assembly line balancing problem with genetic algorithm

4649

References
Brown, K.A. and Mitchell, T.R., A comparison of just-in-time and batch manufacturing: the roles of performance obstacles. Acad. Manage. J., 1991, 43, 906917. Gen, M. and Cheng, R., Genetic Algorithms & Engineering Design, 1997 (Wiley: New York). Gen, M. and Cheng, R., Genetic Algorithms & Engineering Optimization, 2000 (Wiley: New York). Hackman, S.T., Magazine, M.J. and Wee, T.S., Fast, effective algorithms for simple assembly line balancing problems. J. Oper. Res., 1989, 37, 916924. Holland, J., Adaptation in Natural and Artificial Systems, 1975 (University of Michigan Press: Ann Arbor, MI). Hwang, R.K., Gen, M. and Katayama, H., A performance evaluation of multiprocessor scheduling with genetic algorithm. Asia Pac. Manage. Rev., 2006, 11, 6772. Jackson, J.R., A computing procedure for a line balancing problem. Manage. Sci., 1956, 2, 261271. Runarsson, T.P. and Jonsson, M.T., Genetic production systems for intelligent problem solving. J. Intell. Manuf., 1999, 10, 181186. Scholl, A. and Becker, C., A survey on problems and methods in generalized assembly line balancing. Eur. J. Oper. Res., 2006, 168, 694715. Talbot, F.B., Patterson, J.H. and Gehrlein, W.V., A comparative evaluation of heuristic line balancing techniques. Manage. Sci., 1986, 32, 430454. Wemmerlov, U. and Hyer, N.L., Cellular manufacturing in the U.S. industry: a survey of users. Int. J. Prod. Res., 1989, 27, 15111530.

Downloaded by [National Institute of Technology - Calicut] at 06:23 30 October 2012

Вам также может понравиться