| Summary: | 碩士 === 國立臺灣大學 === 工業工程學研究所 === 88 === A full-scale semiconductor wafer manufacturing plant contains more than one hundred types of machines, and the quantity of machines may be as high as 300 to 400. A resource portfolio refers to the makeup, in quantity and type, of the set of processing machines in a plant. Because the development of process and product technologies is very rapid and dynamic in the semiconductor industry, resource portfolio planning in a wafer fabrication plant is interwoven with the problems of product mix, capacity planning, and performance evaluation and is a very complex problem. This paper describes a decision model for determining the optimal resource portfolio that satisfies requirements of product throughput, flow time and performance utility. A procedure is introduced to quickly generate a large solution space of resource portfolios. Each portfolio is evaluated using a queuing capacity model and a bottleneck analysis method to obtain its expected throughput and flow time performance. Finally, a decision model based on utility functions and indifference curves is developed to determine the optimal resource portfolio and its optimal operating condition. In summary, this paper presents a systematic method of portfolio planning for determining the best trade-off between throughput, flow time and investment cost for a manufacturing environment with complex queuing behavior. Industry data obtained from a modern wafer manufacturing plant has been used in this study to develop and test this method.
|
|---|
