晶圓廠達交率導向派工法

Abstract

以接單為主的晶圓代工廠為了滿足顧客的交期，達交率一直是相當重要的績效指標。過去已有許多文獻探討如何以良好的派工法則來提高達交率，然而這些派工法則大多沒有考慮機台「綁機」的限制，因為這種限制是近幾年半導體加工技術越來越精密所產生的一個現象。近年來，Wu等人從生產線平衡的整體觀，考慮機台綁機的限制下，發展出LBSA派工法則。當晶圓的生產途程較短時，LBSA派工法在達交率上表現良好，然而當生產途程較長時，達交率卻大幅降低。有鑑於此，本研究擬修正LBSA派工方法，期使該方法所能應用的產品族範圍擴大。模擬實驗結果驗證出來，本研究所發展的派工演算法在各種長短途程的邏輯產品族(logic product family)下，其達交率均顯著優於Wu等人的派工方法。

Hit-rate, the percentage of on-time completion, is a very important performance measure in a make-to-order fab. This paper presents a dispatching algorithm for a make-to-order semiconductor fab, with machine-dedication feature. This feature imposes a constraint on the production route due to the advance of manufacturing technology, and was rarely addressed in previous relevant literature. A dispatching algorithm, called line balance starvation avoidance (LBSA), was recently developed for a fab with machine-dedication feature. The LBSA algorithm shows very good hit-rate performance for short-routing products but not so well for long-routing products. This paper develops a method to enhance the LBSA algorithm to achieve a high hit-rate performance in both short-routing and long-routing products.