半導體廠兩種廠房設計方案之評估

Abstract

為了使半導體廠能大規模產出，有兩種廠房設計方案可選擇；一種是建造一個大型晶圓廠，另一種是建造兩個可相互支援產能的小型晶圓廠，故本研究提出了兩個研究議題比較這兩種設計的優劣：第一個議題的研究重點在探討單廠區的最適「運輸軌道層數」設計，過去幾乎沒有人探討。此議題利用等候網路模式為基礎，在既定的目標產出下，求解單廠區最適運輸軌道層數。實例驗證說明大規模產出的晶圓廠需要兩層或以上的運輸軌道系統才能符合經濟效益。第二個議題則探討雙廠區跨廠途程規劃的議題，為先前研究的延伸，本研究則提出更有效率的方法解決跨廠途程規劃問題。本研究依據上述兩種設計方案，在特定產出的情境下作成本/利潤分析。實例驗證說明雙廠區可跨廠途程規劃的方式適合大規模產出生產（例如，月產出90K），而單廠區多層運輸軌道的設計則適用於小規模產出（例如，月產出45K）。

To provide a large-scale throughput system in semiconductor manufacturing, there are two design options. One is building a single large fab (or factory). The other is building two smaller fabs and requesting them to support each other in capacity. To effectively compare the two design options, this research examines two topics. The first topic addresses a transportation issue for the single-fab design option, which has been rarely noticed in literature. A queuing-based method is developed to determine optimal transportation layers for a fab with a designated throughput. Numerical experiments indicate that a large-scale fab indeed may need two or more layers of transportation system in order to produce wafers more economically. The second topic addresses the cross-fab routing decision for the two-fab design option. Extending from a prior study, we propose a much more efficient method to solve the cross-fab routing problem. Based on the two proposed methods, we compare the effectiveness of the two design options in various designated throughputs. Numerical experiments indicate that the two-fab option is better for a large-scale throughput (e.g., 90K wafers per month), while the single-fab option is better for a small-scale throughput (e.g, 45K wafers/month).