Knowledge-Based Lot-Splitting Optimization Method for Flexible Job Shops Considering Energy Consumption

The development of lot streaming technology in flexible production systems has greatly benefited the aerospace, semiconductor, automotive, pharmaceutical and other discrete manufacturing industries. In this paper, we address a lot-splitting optimization problem of flexible job shop. To handle the issues of large performance gap and low searching efficiency of optimal scheme caused by strong randomness of lot-splitting, a knowledge-based method is proposed. First, under the equal lot-splitting strategy and variable minimum sub-lots constraint, a lot-splitting scheduling model of flexible job shop is constructed. Second, the effectiveness of lot-splitting schemes is assessed by appropriate data-driven algorithms according to the different scales of the problem. Moreover, a neighborhood search strategy based on the number of sub-lots is introduced to improve the search ability of the method. Then, by analyzing the characteristics of the problem, a parameter adjustment strategy based on problem-specific knowledge is presented to balance the search time and performance. Next, sensitivity analyses are carried out to calibrate the parameters based on the design of experiment Taguchi method. Finally, extensive experiments are conducted to evaluate the effectiveness of the proposed method based on the benchmark problems with single or multiple objectives. The results show that the proposed method can obtain high-quality optimized lot-splitting schemes. Note to Practitioners —In discrete manufacturing industries, the flexible production mode in multi-variety and small-batch has become the mainstream. The flexible job shop lot-splitting optimization problem in such a mode is of great significance to practitioners. We mathematically characterize the lot-splitting problem in a flexible job shop considering energy consumption and solve it through a new general multi-objective optimization framework by integrating the lot-splitting and job shop scheduling that is activated from problem-specific knowledge. Existing lot-splitting optimization approaches fail to solve the concerned problem. This work introduces appropriate data-driven algorithms to assess the scheduling schemes after lot-splitting. By updating the minimum lot-size of jobs and the archive set, the search of the optimal lot-splitting scheme is directional. Furthermore, the global and local iterations of the method are adjusted according to the problem size and the number of sub-lots to balance the performance and search time. The results of comparative study show that the proposed method outperforms other algorithms well and can greatly help practitioners manage the lot-splitting scheme of jobs.