Abstract—This paper addresses the interaction effect between the workload and the information accuracy in work-sharing environment on the performance of a serial production line. The work-sharing is organized by Dynamic Line Balancing (DLB) policy. Under this policy, the workers can help each other with some tasks, called shared tasks. We investigated two levels of information accuracy that support the decision of working on or passing the shared tasks to the next buffer. We employed different configurations of workload under these two levels of information accuracy. A measure of workload is proposed to ease the comparison. The results of simulation showed a remarkable influence of workload on the performance and found its patterns as work-in-process (WIP) level changes. These patterns are clearly affected by changing the information accuracy level. Index Terms—Work-sharing, information accuracy, dynamic line balancing, fixed workload. I. INTRODUCTION The trend to producing variety of products with small quantities, reducing the WIP that is necessary to achieve the targeted throughput, increasing the utilization has magnified the importance of the work-sharing. Actually, this concept originally appeared in cell manufacturing which is one of JIT revolution’s outcomes. The highest performance in these manufacturing environments can be achieved as much as workers approach full cross-trained level. Moreover, the inventory is considered an absolute evil that represents the Japanese perspective and their strategy is to eliminate all factors that necessitate the inventory. Whereas, in western production systems it is a necessary evil as it makes production running smoothly [1]. Practically, having fully cross-trained workers is expensive since it consumes time and money. Additionally, the fast change of products will keep the need to continue training in high expenses, and on the other hand, the worker’s speed will be slower with many tasks (full cross-trained) than that with a limited number of tasks (partially cross-trained) with limited amount of WIP. Among different ways of applying work-sharing, we focused in this research on the way that considers the previous point (partially cross-trained), which is referred to as Dynamic Line Balancing (DLB). In DLB, introduced by [2], Manuscript received November 9, 2013; revised January 9, 2014. The authors are with the Industrial Engineering and Management Department, Tokyo Institute of Technology, 2-12-1 Oh-okayama, Meguro-ku, Tokyo 152-8552 Japan (e-mail: {kasmo.s.aa, enkawa.t.aa, Suzuki.s.ag}@m.titech.ac.jp). the workers stay at their stations and there is no movement between stations. Each worker is assigned to fixed tasks of a job that can be done only by him. He can also help the upstream and downstream workers in shared tasks after finishing the fixed ones. A worker chooses to pass on a job with the shared task undone or completes the shared task according to specific rules depending on system information. This way of work-sharing is in somehow similar to fixed task zone chaining introduced by [3] but here the flow of job is unidirectional and no movement between stations. Reference [3] concluded that this design can nearly reach the performance of a fully cross-trained system. Reference [2] found that by using DLB with a half-full buffer (HFB) control rule, the Work-In-Process (WIP) inventory can be reduced and the efficiency can be improved. Reference [4] showed that DLB can increase the efficiency even with no buffer. They used a new model called subtask model. Two types of models in term of information accuracy level used to make a decision of passing or working on the shared task have appeared in the literature; model A (as in [2], [4]) and model B (as in [4]). In model A, the decision is supported by the workcontent in downstream buffer only while in model B it is supported by the workcontent available to the downstream station, which includes the one in the buffer and in the downstream station under processing not completed. Reference [5] proposed a new control rule with the model A, called SRNS (Smallest R No Starvation). It showed that this rule performs well by comparing its resulting performance with the optimal performance. They also compared in a later paper [6] between models A and B, and also between SRNS and HFB with constant-WIP (CONWIP) policy. Reference [7] compared between HFB and 50-50 rule which utilizes the ratio of workcontent at the upstream station to total workload in the system. It also found three factors that have a notable effect on the work-sharing’s results. These factors are preemptability of shared task, granularity of shared task, and processing time variability. In this paper, we studied an interaction influence of two new factors; workload and information accuracy. The workload refers here to the sizes of fixed tasks of adjacent stations. Reference [6] demonstrated by using some examples how DLB can remedy the loss in the efficiency due to imbalanced fixed tasks. Also [7] treated this factor so briefly by one example and found that the performance gets bad when the fixed workload is not balanced. It also used the workload to compare between HFB and 50-50 rules. We studied the workload in a completely different way than that in [7]. The Work-Sharing in a Serial Production Line with Consideration of the Balance of Fixed Workload and the Accuracy of Information Salah Kasmo, Takao Enkawa, and Sadami Suzuki International Journal of Computer Theory and Engineering, Vol. 6, No. 4, August 2014 336 DOI: 10.7763/IJCTE.2014.V6.885