晶圓代工廠機台組合決策方法

Abstract

一個晶圓代工廠的機台組合指的是工廠對產能的規劃及配置，換言之，工廠要決定機台種類及機台數量。對晶圓代工廠來說，因為機台非常昂貴，機台組合決策是一個相當重要的問題。以往的文獻大多將焦點放在，如何在預算限制下，最大化工廠產出量。然而，在晶圓代工廠中，週期時間是一個相當重要的績效指標，卻很少有文獻探討如何在及台組合決策中考慮週期時間限制的問題。 本研究的目標是建構晶圓代工廠機台組合及數量的決策機制。本機制的輸入條件包括(1) 產品組合或需求型態，(2) 產出限制或最低產出率，(3) 產品週期時間限制或最長平均週期時間。本研究的結果會產生一組機台組合及每個工作站的機台數量。這個機台組合不僅能夠在既定的產品組合之下，滿足上述兩個生產績效限制條件，還能夠使購置機台的成本最小化。 本文運用基本的空間搜尋演算法，配合四種簡單的規劃方法，找出晶圓代工廠應有的機台組合。測試結果發現，在這四種方法中，最佳和最差的機台組合在成本上的差距，可能會高達30億台幣以上。

A machine-mix configuration denotes a factory capacity implementation plan which describes the number of machines and machine types for the factory. Determining machine-mix configuration is a very important problem for an IC foundry because the machines are quite expensive. Most previous studies on machine-mix configuration focus on developing methods for maximize throughput under budget constraints. In an IC foundry, short cycle time is a very important justification criterion, yet rarely included in previous relevant studies. The objective of this research is to develop a decision-making mechanism for designing the machine-mix configuration for an IC foundry. The input to the decision-making mechanism involves (1) product mix or demand pattern; (2) throughput constraint or the lowest production output rate; (3) cycle time constraints or the longest mean cycle time. The output tends to give a set of machine types and corresponding machine numbers which would satisfy the above two constraints in producing the given product mix, under the criterion of minimizing total investment. A basic design method together with four embellishments have been developed and tested. Testing results show that among the four proposed machine-mix configurations, the difference between the best and the worst solutions can be around 10 Million US dollars.