半導體製造廠機台組合與採購時程之研究

Abstract

一般而言，半導體製造廠的投資金額均甚為龐大常在數百億元之譜，且投資金額中有極大比例是用於購置生產設備，維持各型生產設備適當的數量即能在資金成本、機台閒置與瓶頸製程間取得平衡並促進生產流程順暢與提高生產效率，另一方面，若規劃不當使部份加工站含有過多或過早引進的機台將造成資金浪費、折舊攤提、技術過時以及機台閒置等問題，反之，配置過少的機台卻可能引起瓶頸製程導致生產流程遲滯、在製品堆積及生產效率降低等不利影響，因此，正確的機台組合決策深深影響到半導體製造公司的經營績效與獲利性。本論文即著眼於此，而提出在半導體製造廠在既定之機台投資預算下購進各型生產設備之最適機台組合的解析方法，即在特定之產品組合下找出能最佳化系統目標績效的機台組合，並得以最小化機台投資成本。同時，本論文亦探討一規劃期間 (planning horizon) 內預設之產品混製組合擴展路徑中機台最佳採購時程的問題，亦即在已決定出規劃期間內終期各型設備所需的最適總機台組合下，根據各期或各時程 (time bucket) 之產品混製組合的產能需求來逐步適時適量地購進機台以讓投資資金得以做最有效的運用。而此兩種決策模型皆以本論文所發展之半導體製造系統的系統績效衡量法為基礎。 關鍵字：半導體製造廠、等候網路、製造系統模型、系統績效衡量、機台組合、產品混製組合擴展路徑、採購時程。

In general, the money investing at the semiconductor wafer fabrication is enormous. The greater part of the investment is spent in purchasing equipment. Under the improper machines investment strategy, it will result in the fund waste and machines idleness if there were some little equipment that has more quantity than required. On the other hand, it will bring about the bottleneck processes to give rise to the tardy production flow, WIP piling up and production efficiency reduction if there were some little equipment that has less quantity than required. Therefore, making accurate decision for machine set would deeply influence business performance and profitability of semiconductor corporation. In view of this, this thesis wants to present an analytic approach that how decides to purchase the most appropriate for a set of machine types subjected to a given product mix and the fixed budget. Through the analytic process, one can optimize the system performance and minimize the investment cost. This thesis also studies the problem of optimal machines purchase scheduling. That is, given a default product mix expansion path at each period within planning horizon and under have decided total machines combination needed for the planning horizon, the analytic process suggested by this thesis would like to gradually and timely introduce machines according to different capacity and performances requirement for the separate periods in order to most efficiently make use of the investment fund. Both of the decision models are based on a system performance measure method of semiconductor foundry, which is developed by this thesis. Key words: semiconductor foundry, queueing network, manufacturing system model, system performance measure, machine set, product mix expansion path, purchase scheduling.