應用啟發式演算法建構晶圓代工產業之生產規劃系統

Abstract

台灣中小型晶圓代工廠其客戶規模較小，訂單有筆數多但量小的特性，而且對景氣之反應特別快，一有風吹草動，立刻會進行下急單或抽單之動作，因此中小型代工廠之生產規劃彈性及反應能力相對更顯重要。然而生產規劃作業必須整合客戶需求預估、訂單、在製品及產能等相關資訊，在少量多樣之產品組合下，必須運用有效率之工具，才能因應快速變化之市場需求。 本研究結合兩種啟發式演算法，組合生產線循序作業電腦方法及以優先權規則為基礎之啟發式演算法，設計生產規劃作業之生產排程演算法，同時參考派工法則與晶圓投片法則，並依晶圓代工廠需求及特性，設計適當之產能限制模型及晶圓投片規則，以此建構其生產規劃系統，以簡單、快速及有彈性的方法，產生合理可行之生產計劃，除提供經營管理者決策之用，同時做為製造廠之生產目標。 本研究成果可協助中小型晶圓代工廠，大幅縮短生產規劃作業之時間，增加規劃之彈性及效率；提高生產計畫之精細度，由製程微米與產品類別之彙總計劃，精細至訂單與需求配置之產品型號，提昇資訊分析之精細度及準確率；增加規劃晶圓投片期間，由只能預估月總投片量及依人工排定一週內之投片計劃，改善為能安排每日所有訂單及需求配置產品之投片計劃，提昇此計畫之透明度；提高生產規劃作業之頻率，由原只能二週一次正式之生產規劃作業，提高至每週一次，時間間隔縮短，使生產計畫能即時反映需求及生產狀況之變化，提供正確之資訊供經營管理者參考及決策。

In Taiwan, the scale of the customers of small and medium-sized wafer foundry companies are small. The characters of their orders are large number but less volume. Their reactions to the boom are very fast. The actions of delivering urgent orders and canceling orders are as usual. Base on the above reasons, the flexibility and reacting capability of the production planning of the foundry companies are more important. The research combines two heuristic algorithms, COMSOAL and Priority rule-based algorithm, to design a production scheduling algorithm. Meanwhile it refers to dispatching rule and wafer start rule. Also base on the requirements and characters of wafer foundry companies, it designs proper capacity constraint model and wafer start rule to construct production planning system. Using simple, fast and flexible methods to generate reasonable and feasible production plans. The result of the research has several contributions. First, it shortens the execution time of generating a production plan. Second, the resolution of production plan is raised from the summary of manufacturing process and product type to full product name of order and demand allocation. It improves the resolution and accuracy of information analysis. Third, the planning period of wafer start plan is improved from monthly to daily. Last, the frequency of production planning is improved from once a fortnight to once a week. The time interval is shortened. It lets the production plan can timely react to the change of demand and production status and provides the right information for the references and decision of high managements.