Semiconductor Manufacturing Plants: Design of Efficient Operating Policies and Performance Analysis

9403571库玛半导体制造是一个复杂的过程，涉及数以百计的加工步骤。晶圆多次返回到许多工具，导致高度可重入的工艺流程。这项建议涉及工厂的有效运行、分析和设计问题。其目标是在两个重要方面推进最先进的技术：(1)优化周期-时间性能，(2)半导体制造厂的性能评估。这一建议追求的是一种研究某些随机变量平均值约束的方法。它开发了一个获得性能界限的程序化程序，并举例说明了它在各种制造厂模型和调度策略中的应用。这种方法可以用来建立系统背后的马尔可夫链的稳定性，即性能递归，甚至矩的有界性。该提案中提出的研究是应用程序驱动的，并使用了从模拟、启发式和软件开发到理论的各种工具。目标是推进半导体晶圆厂调度的最新技术水平，以及拟议方法的理论基础。***

9403571 Kumar Semiconductor manufacturing is a complex process, involving hundreds of steps of processing. Wafer return several times to many tools, leading to a highly re-entrant process flow. This proposal addresses the problems of efficient operation, analysis and design of plants. The goal is to advance the state-of-the art on two important fronts: (1) Optimization of Cycle-Time Performance, and (2) Performance Evaluation of Semiconductor Manufacturing Plants. This proposal pursues an approach of studying the constraints on the mean value of certain random variables. It develops a programmatic procedure for obtaining performance bounds, and illustrates its application to various fab models and scheduling policies. This approach can be used to establish the stability, i.e. performance recurrence, and even the boundedness of moments, of the Markov chain underlying the system. The research proposed in this proposal is application driven, and uses a variety of tools ranging from simulations, heuristics, and software development, to theory. The goal is to advance the state of the art in scheduling semiconductor fabs, as well as the theoretical foundation of the proposed approach. ***