Article ID: | iaor19982202 |
Country: | Netherlands |
Volume: | 76 |
Issue: | 1 |
Start Page Number: | 233 |
End Page Number: | 259 |
Publication Date: | Feb 1998 |
Journal: | Annals of Operations Research |
Authors: | Luh Peter B., Hasegawa Tetsuo, Gou Ling, Zhang Yuanhui, Nagahora Takaaki, Tsuji Makoto, Yoneda Kiyoshi, Kyoya Yuji, Kano Toshiyuki |
Scheduling is a key factor for manufacturing productivity. Effective scheduling can improve on-time delivery of products, reduce inventory, cut lead times, and improve the utilization of bottleneck resources. This study was motivated by the design and implementation of a scheduling system for the manufacturing of Toshiba's gas insulated switchgears. The manufacturing is characterized by significant machine setup times, strict local buffer capacities, the option of choosing a few alternative processing routes, and long horizon as compared to the time resolution required. This problem has been recognized to be extremely difficult because of the combinatorial nature of integer optimization and the large size of the real problem. Our goal is thus to obtain near-optimal schedules with quantifiable quality in a computationally efficient manner. To achieve this goal, a novel integer optimization formulation with a separable structure is developed, and a solution methodology based on a combined Lagrangian relaxation, dynamic programming, and heuristics is developed. The method has been implemented using the object-oriented programming language C++, and numerical testing shows that the method generates high-quality schedules in a timely fashion to achieve on-time delivery of products and low inventory. Through explicit consideration of setups, tanks with the same processing requirements tend to be processed together to avoid excessive setups. The integrated treatment of machines and buffers facilitates the smooth flow of parts through the system. The embedded routing selection mechanism also balances the load among candidate routes. Finally, the newly developed ‘time step reduction technique’ implicitly establishes two time scales to reduce computational requirements without much loss of modeling accuracy and scheduling performance, thereby enabling the resolution of long horizon problems with controllable computational requirements.