晶圓製造批量優先序分佈與設定之研究

A Study of Priority Distribution and Setting in Wafer Fabs

陳建良(James C. Chen) ； 蕭佳毓(C. Y. Hslao) ； 鄭智強(J. C. Jeng)

《中原學報》 30卷1期 (2002/03) Pp. 33-49

摘要

在晶圓製造廠中，在製品的管理非常重要，因為它影響到系統的平均在製品水準與週期時間。在製品管理的方法之一，是針對不同之批量給予不同之優先序。在實務上，優先序通常分成五個等級（由高至低）：Super Hot、Hot、Rush、Normal、Slow。本研究將探討兩個主題：一是尋找合理之優先序分佈，二是比較動態修改優先序方法與不修改優先序之方法。本文以AutoSched模擬軟體發展一模擬模式以評估各績效指標（週期時間平均值、週期時間標準差、在製品平均值、在製品標準差）。模擬結果顯示動態優先序設定法配合（4％Super Hot,16％Hot,16％Rush,32％Normal,32％Slow）與(10％Super Hot,15％Hot,20％Rush,25％Normal,30％Slow)的優先序分佈可導致最低之週期時間平均值與在製品平均值，並有合理之週期時間標準差與在製品標準差。

關鍵字

晶圓制造 ； 在制品 ； 優先序分布 ； 優先序設定

並列摘要

In a wafer fab, the management of-wafer-in-process (WIP) is important, as it affects the cycle time and average WIP level of the system. One way to manage the WIP is to assign different priorities to different lots. In practice, the priority can be divided into five classes (from high to low): super hot, hot, rush, normal, and slow. This research tries to answer two questions. First, what is a reasonable distribution of these priorities? Second, should a lot keep its priority or dynamically change its priority throughout the fabrication process to result in good system performance?
AutoSched simulation models are developed to evaluate the performance (such as the average and standard deviation of cycle time and WIP) of the combination of priority distributions and priority setting methods. Simulation results show that a dynamic priority setting method combined with priority distributions of 4% super hot, 16% hot, 16% rush, 32% normal, and 32% slow as well as 10% super hot, 15% hot, 20% rush, 25% normal, and 30% slow lead to the best average cycle time and average WIP level and reasonable standard deviation of cycle time and WIP level.

並列關鍵字

wafer fab ； wafer-in-process ； priority distribution ； priority setting