Lot Permission Rule in Wafer Fabrication Factories with Queue-Time-Limit Considerations

• Main Authors: Dennis Tu, 涂漢威
• Other Authors: Yi Feng Hung
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/35929258687473423711

碩士 === 國立清華大學 === 工業工程與工程管理學系 === 93 === As far as the wafer fabrication process is concerned, there are many factors affecting the throughput and yield rate. One of them is the “Queue-Time-Limit”. Certain process steps have the queue-time-limit, which means a lot needs to begin the following process step within certain time limit after it completes its operation of the previous process step. If a lot exceeds the queue-time-limit, quality issues will occur and the lot needs to be reworked. This will waste the machine capacity. The characteristics of the wafer fabrication process, like rework, re-entrant, queue-time-limit, and long production cycle time, make the production management difficult. If a lot arrives for a certain process step, the manager can prevent the lot from rework if he can foresee whether the lot will exceed queue-time-limit or not in advance. Once the rework can be reduced, the throughput of the fab will be increased. Therefore, it is necessary to develop a lot permission rule to determine whether a lot should go or not in advance in this complicated production environment. This research focuses on certain process steps having the queue-time-limit between some steps of the whole wafer production steps,. We propose a methodology called simulation-based lot permission rule (SBLP) in a multi-product, multi-route, multi-workstation, multi-machine environment with rework and re-entrant features. The purpose of SBLP is to prevent the lot from rework, saving the capacities of machines, thus to increase throughput. After simulation experiments, the results show that SBLP outruns FCFS in many environment settings like different demand-to-capacity ratios and different queue-time-limit levels. Therefore, we conclude that SBLP has significant effects on throughput.