| Summary: | 碩士 === 國立成功大學 === 化學工程學系碩博士班 === 100 === Semiconductor manufacturing industry is a technology and capital-intensive industry. As feature sizes shrink and wafer sizes increase, the manufacture should use intricate run-by-run control methods to improve the product quality and tools utilization of any production facility. In this thesis, the semiconductor manufacturing process is treated as multiple product input and multiple M/M/1/K queuing systems operating network. Thus, other than the run-by-run control, product quality and efficiency can be improved by the three dispatching controls in queuing network: Routing, Admission and Scheduling. It is assumed that drift noise phenomena exist in machine disturbances, the variance computation equation for white noise production process has been derived under Double-EWMA run-by-run control.Forthermore, it has been verified that the product quality is indeed influence by dispatching control. Based on these assumptions, a mixed integer nonlinear programming model (MINLP) is formulated to determine the optimal dispatching policies for maximizing process capability or minimizing average system time. Numerical simulation procedure is also devised to confirm the validity of the resulting dispatching model.
|
|---|
