Importance Sampling of Re-entrant Queueing Systems

9622084 Deuermeyer本研究致力于发展新的模拟理论和方法，以提高离散模拟在半导体制造中的性能和有效性。计划人员和设计人员对使用周期时间来计划设备采购和调度策略越来越感兴趣；离散事件模拟是目前唯一可用于支持设计活动的分析工具。需要对仿真方法进行具体改进，以便1)更快的执行速度，2)获得周期时间估计器的方法，以及3)特殊的方差减少技术，该技术考虑了预防性维护和罕见故障事件导致的周期时间变异性的增加。采用广义半马尔可夫过程(GSMP)作为建模框架。这一研究将在生成GSMPS样本路径、开发GSMPS重要抽样方案、开发计算周期时间交叉概率的方法以及开发演示理论方法的软件方面提供新的能力。从产品周期时间和生产率以及设施和机器能力的角度来看，这些将大大有助于解决与半导体晶圆厂相关的极其复杂的工艺分析和设计中的苛刻问题。

9622084 Deuermeyer This research focuses on the development of new simulation theory and methods to improve the performance and effectiveness of discrete simulation as it applies to semiconductor manufacturing. Planners and designers are becoming more interested in using cycle time to plan equipment purchases and scheduling policies; discrete-event simulation is the only analysis tool currently available to support design activities. Specific improvements in simulation methodology are needed for 1) faster execution speed, 2) methods to obtain cycle-time estimators, and 3)special variance-reduction technology that accounts for the increase in variability in cycle time caused by preventive maintenance and rare failure events. Generalized Semi-Markov Processes (GSMP) will be used as the modeling framework. This research will provide new capabilities in generating sample paths of GSMPs, developing importance sampling schemes for GSMPs, developing methods for computing crossing probabilities of cycle time and developing software to demonstrate the theoretical methods. These should contribute significantly toward solutions to the demanding problems in analysis and design of the extremely complex processes associated with semiconductor wafer fabs from the point of view of product cycle time and throughput rates, and facility and machine capacity.