Optimal Scheduling Of Re-Entrant Manufacturing System In Uncertain Environment

Re-entrant lines are a class of complex production systems abstracted from semiconductor and film manufacturing systems. With the development of information technology and microelectronic industry, re-entrant lines have drawn much attention of academic and practical arenas. Because of the re-entrant, lots strive to the access to the equipments, meanwhile, differences of technics, quantities and types of production are all making the scheduling of re-entrant lines harder than other manufacturing systems.Semiconductor wafer fabrication is a typical class of complex manufacturing system with re-entrant being its feature. The optimization of semiconductor wafer fabrication has attracted huge attention, for increasingly heated competition in the high-investment, high-profit and high-risk semiconductor manuafcutring industy. Involving the re-entrant process flows, complexity, uncertainty, multi-constrain and multi-objective of a semiconductor production line, this paper proposes control and optimal scheduling approaches separately on releasing, lot scheduling and maintenance scheduling considering the advantages of intelligent optimization methods: 1 Fuzzy Petri Net (FPN) is used for releasing control of semiconductor wafer fabrication. Based on the releasing FPN model and the real-time data collected from the line, the constructed releasing mechanism can help to determine suitable releasing action to be adopted at different conditions and the throughput of the wafer fabrication can be maximized.2 A simulated annealing algorithm (SA) embedded search strategy is researched for scheduling semiconductor fabrication line. Considering the conciseness of rules and the powerful search capability of SA, the coding of the strategy represents a combination of lot releasing policies and scheduling policies, and the approach is finally used to search for the optimal combination of scheduling rules.3 The adaptive neuro-fuzzy inference system (ANFIS) is applied to build a failure prediction model. Embedded the failure prediction model into the original maintenance scheduling, a new maintenance scheduling mechanism is constructed which will help us to enhance the predictive power of the system failures and improve the performance of the line.The research results show that:The aforesaid control and optimal scheduling strategies can promote better system performance. The proposed approaches are feasible and effective which can be referenced to solve the practical scheduling problems of re-entrant manufacturing systems.