考量等候時間限制之半導體爐管機台派工法則

Dispatching Rules with Queue-Time-Limit Consideration for Furnace in Wafer Fabrication Factory

蔡建基(Chien-Chi Tsai)

指導教授 ： 林則孟

國立清華大學/工學院/工業工程與工程管理學系碩士在職專班/碩士(2008年)

摘要

半導體晶圓廠為了避免在製品於生產線上等待過久而造成晶圓表面的製程缺陷（Defect），會根據製程與產品特性，在特定製程完成加工之後，制定一段時間來做為等候時間的限制（Queue Time Limit）以維持產品良率，若超過此時間限制晶圓必須重新加工或是報廢。等候時間限制問題若發生在批次工作站（Batch Workstation）影響將更加嚴重，到達批次工作站之晶圓批除了需要等待前一批次加工完畢外，還必須增加集批（Form Batch；等待製程條件（Recipe）相同的批量集結成一批次）的額外等候時間，此加工型態會造成產品的等候時間增長，進而使超出等候時間限制的機率加大導致降低產出量。

而爐管機台同時具有「批次加工」與「等候時間限制」這兩個重要特性，也需要具有集批的等候狀態，因此本研究乃針對爐管區製程討論其特性對生產之影響，並以Look-ahead為基礎，提出於等候時間限制的考量下，兩階段集批策略派工法則（Two-stage Form Batch Strategy；TFBS），目的在預防晶圓在加工過程中，因超過等候時間限制所造成的重工或報廢，除了能節省生產成本，預期能對爐管加工良率（Furnace Line Yield）及訂單達交率（CLIP Rate）有明顯助益。

本研究籍由eM-Plant模擬軟體建構模擬模式，系統由八部清洗機台、七部爐管機台與五種產品組合，各產品之加工時間、等候時間上限與可加工機台不同。實驗設計在三種產品來到率與三種交期鬆弛度的環境因子設定下，探討不同的派工法則對於各項績效指標（訂單達交率、超過等候時間限制比率與爐管加工良率）之表現差異。由模擬結果得知，本研究所提出之兩階段集批策略派工法則，在具有等候時間限制下，可有效的提升訂單達交率與爐管加工良率。

關鍵字

爐管機台 ； 批次加工 ； 等候時間限制 ； 集批 ； 生產線良率

並列摘要

To avoid wafer contamination in semiconductor manufacturing, some process steps have to set up “queue-time-limit” to keep line yield, If a lot exceeds the queue-time-limit, quality issues will occur and the lot needs to be reworked or scrapped. The influences are more serious especially the lots released in front of a batch workstation. The waiting time may be increased due to “form batch” procedure（i.e. waiting for the same recipe）. Besides, this kind of process characteristics will cause high over-queue-time ratio and low throughput.

Once the rework can be reduced, the output qty of the fabrication factories will be increased. Therefore, it is necessary to develop a dispatching rule to determine lot process priority before machine operating in advance. This research focuses on furnace machine area having the two features queue-time-limit and batch process at same time. In this paper, we propose a dispatching rule called “two-stage form batch strategy”(TFBS) with queue-time-limit consideration. Furthermore, TFBS combines “look-ahead” mechanism, which estimates future arrival events according to the information of upstream and downstream on product environment. The objective of TFBS is to prevent the lot from outrunning queue time, decreasing furnace line yield, thus to increase CLIP rate.

The experiment is built and constructed the simulation model by eM-Plant simulation software, made up of eight wet machines, seven furnace machines and five kinds of products on which all products have different processing time, queue time limit and individual furnace machines. In experimental design, three kinds of arrival rate and three level of slack setup on due date, probed into different dispatching rules affected to every performance indices, including CLIP, over-queue-time ratio and furnace line yield. From the result, TFBS performed on CLIP and furnace line yield in very effective improvement.

並列關鍵字

furnace machine ； batch process ； queue-time-limit ； form batch ； line yield

參考文獻

3.陳勝一（2007），「批量加工之排程問題-半導體廠爐管機台為例」，清華大學工業工程與工程管理研究所碩士論文。

12.Ahmadi, J.H., Ahmadi, R.H., Dasu, S., and Tang, C.S.（1992）, Batching and scheduling jobs on batch and discrete processors, Operations Research, 40, 750-763.

13.Bertrand, J.W.M., and Wortmann, J.C.（1981）, Production Control and Information Systems for Component-Manufacturing Shops, Elsevier Publishing Company, Amsterdam.

14.Blackstone, J. H., Phillips, D. T., Hogg and G.l.（1982）,“A state-of-the-art survey of dispatching rules for manufacturing job shop operation.” International Journal of Production Research, Vol.20（1）, pp.27-45.

15.Deb, R. K., and Serfozo, R. F.（1973）, Optimal Control of Batch Service Queues. Advances in Applied Probability, Vol. 5, p340-361.