Simulation Modeling And Performance Optimization Of Re-entrant Manufacturing System

Re-entrant manufacturing system, typified by the semiconductor wafer fabrication process, is the third type manufacturing system. This class of manufacturing system has characteristics of highly re-entrant, hundreds of machines, thousands of processing steps, lots of uncertainty constraints, batch processing, multiple product types and long cycle times etc., which are obviously different from Flow Shops and Job Shops. As the most capital intensive and technologically complex stage in the front end of the semiconductor manufacturing line, the process of wafer fabrication have many both closely related and mutually restricted decision making and control issues from the initial order processing and allocation of resources to the production phase of material planning and equipment scheduling, these complex decision making and control problems can directly influence the whole performance of semiconductor wafer fabrication system. The traditional mathematical analytic model, based on operational research, is hard to obtain the optimal solution within a short time. And yet the simulation optimization method, based on discrete event simulation and optimization algorithm, has strong modeling capability and advanced intelligent algorithm, thus providing a new idea to solve the decision making and control issues of re-entrant manufacturing system.This paper takes the performance optimization issues of re-entrant manufacturing system as a research object, and the optimal target is to increase the mean throughput as well as reduce the mean cycle time. A new Multidimensional Combination Policy (MCP), based on simulation optimization method, is proposed to solve the performance optimization issues of re-entrant manufacturing system which takes the dispatching rule of each workstation, entire and bottleneck workstation constant work in process into account. To be specific, the main research content includes the following two aspects:(1) Re-entrant manufacturing system simulation optimization modeling. With regard to the performance optimization issues of re-entrant manufacturing system, the classic Intel Mini-Fab model is selected as simulation object, and based on the ExtendSim simulation development platform to build the simulation optimization model. This model takes the dispatching rule of each workstation, entire and bottleneck workstation constant work in process as the decision input variables, and the objective function is to maximize the system mean throughput as well as ensure the bottleneck workstation maintain higher utilization rate simultaneously. The simulation optimization modeling can provide the model foundation and technical support for solving the performance optimization issues of re-entrant manufacturing system.(2) Experimental evaluation of MCP which is based on simulation optimization method. In ExtendSim software, the "Optimizer" control is embedded genetic algorithm, so it can be used to design the simulation optimization module of this model. The best MCP, which can show the optimal system performance, will be selected through the overall iterative search of genetic algorithm after running the simulation optimization procedure. For steady-state system, the system should have the minimum mean cycle time under MCP by Little's law. In order to further verify the superiority of MCP, the other two simulation experiments are also designed to assess the different policies presented in preterit articles, including the dispatching rule combine with entire constant work in process policy and the bottleneck workstation constant work in process policy. Comparison on the results of these three simulation experiments indicates that the MCP which is based on simulation optimization method can improve the system mean throughput and remarkably reduce the mean cycle time simultaneously.