A Decomposition Heuristics Based on Multi-Bottleneck Machines for Large-Scale Job Shop Scheduling Problems
Yingni Zhai,Changjun Liu,Wei Chu,Ruifeng Guo,Cunliang Liu
DOI: https://doi.org/10.3926/jiem.1206
2014-01-01
Journal of Industrial Engineering and Management
Abstract:Purpose: A decomposition heuristics based on multi-bottleneck machines for large-scale job shop scheduling problems (JSP) is proposed.Design/methodology/approach: In the algorithm, a number of sub-problems are constructed by iteratively decomposing the large-scale JSP according to the process route of each job. And then the solution of the large-scale JSP can be obtained by iteratively solving the sub-problems. In order to improve the sub-problems' solving efficiency and the solution quality, a detection method for multi-bottleneck machines based on critical path is proposed. Therewith the unscheduled operations can be decomposed into bottleneck operations and non-bottleneck operations. According to the principle of “Bottleneck leads the performance of the whole manufacturing system” in TOC (Theory Of Constraints), the bottleneck operations are scheduled by genetic algorithm for high solution quality, and the non-bottleneck operations are scheduled by dispatching rules for the improvement of the solving efficiency.Findings: In the process of the sub-problems' construction, partial operations in the previous scheduled sub-problem are divided into the successive sub-problem for re-optimization. This strategy can improve the solution quality of the algorithm. In the process of solving the sub-problems, the strategy that evaluating the chromosome's fitness by predicting the global scheduling objective value can improve the solution quality.Research limitations/implications: In this research, there are some assumptions which reduce the complexity of the large-scale scheduling problem. They are as follows: The processing route of each job is predetermined, and the processing time of each operation is fixed. There is no machine breakdown, and no preemption of the operations is allowed. The assumptions should be considered if the algorithm is used in the actual job shop.Originality/value: The research provides an efficient scheduling method for the large-scale job shops, and will be helpful for the discrete manufacturing industry for improving the production efficiency and effectiveness.