隨機規劃法在半導體機台組態規劃之應用

Abstract

半導體製造的機台組態規劃問題是指依據產品需求與製程要求，考量投資報酬與風險等因素，以決定工廠內機台類別與數量的規劃程序，所面臨的困難主要在於產品需求的動態變化、複雜的製造途程、以及設備採購前置時間所綜合造成的不確定性。機台組態規劃常見的方法有公式試算、整數規劃、以及隨機規劃，文獻上還未見這些方法的比較分析，因此本文探討隨機規劃法在機台組態規劃的應用價值。隨機規劃法是一種適用於不確定性問題的數學規劃法，其效益與不確定性的程度有關，本文分兩大部分，第一部份是以產業的需求與途程實務資料，先量測產品類別、企業、工廠三個階層所面臨的需求變動特性，以界定需求的不確定程度，第二部份評估隨機規劃法在機台組態決策的適用性，並證實該方法確能有顯著效益。隨機規劃法還未為產業所普遍採用，本文建議在面對需求不確定的產業環境，晶圓製造廠值得採用這個方法。

Tool portfolio planning is an important task in semiconductor manufacturing in which the types and quantities of processing tools are determined. The challenging facing tool portfolio planning is the uncertainty in product demand and process technology, complicated process routing and long procurement lead-time. Methods commonly used for tool portfolio planning include static computation using formulas, integer programming, and stochastic programming. There have been no studies on comparing the relative merits of these methods in application to tool portfolio planning. The paper is made up of two parts. In the first part, we calibrate the uncertainties of demand in several product segments: in the aggregate demand of a company and in the output of several factories. Stochastic programming is a method that is suitable for problems with uncertain factors; its utility will depend on the level of uncertainty. We evaluate the economic value of using stochastic programming in the second part of the paper. We show that the benefits are very significant and thus recommend that stochastic programming be used in tool portfolio planning in the semiconductor manufacturing industry.