在半導體產業中,下游的封裝廠由於進入的門檻不論是技術或資金都遠較上游晶圓廠來得低,且需求的人力較高。再者,由於產業的專業分工與競爭壓力下,下游的封裝廠經營越顯艱困獲利下滑。如何發揮有效的機台的使用率使其獲利增加,將是競爭的不二法門。本文主要是探討半導體下游封裝廠,依據封裝廠前段的銲線製程,以ID3決策樹演算法為基礎,其目的在於將產品型號本身與銲線設備之限制加以考慮之後,將資料分類;進行產品與銲線機台的選配。本研究發展一對多機台分派演算法求解與實務案例做驗證,以判斷產能是否滿足客戶的需求;以供管理階層做迅速的決策。 本研究所發展機台派工模式滿足公司之目標:(1)當產能大於需求時,目標在保留高階機台數量最多,使固定週期內預留的產能最大化;增加可承諾數量(Available-to-Promise,簡稱ATP),以供日後需求使用下投資金額最低;(2)當產能小於需求時,在善用有限產能使利潤最大化或投資機台數最少化下縮短計算時間。 本研究實際應用於派工作業上,並與過去之機台派工與使用規劃模式,做產品製造週期時間內「產能剩餘率」、「最大獲利」與「投資金額」之比較。實務案例說明新的派工模式有顯著的改善。證明所構建之機台派工與使用規劃模型,比工廠的現行的作業有較佳的績效。
To manage a semiconductor assembly factory is highly challenging due to numerous competitors and huge investment on machines with risk. Improving the utilization of expensive machines is thus one of the most key factors to increase profitability of the company. This study focuses on the issues regarding the machine selection and resource allocation of the wire bond process in front end process of the assembly factory. We use ID3 algorithm to establish a decision tree which is use to build the matching relationship of the machine categories and product types. Furthermore, this research develops a task allocation heuristic to determine the capacity use and to fulfill the customer expectation. As compared to the conventional model used in practice, the proposal model has achieved the following goals -- (1) When the supply of capacity is greater than demand, the model retains the largest quantity of high-level machines. (2)When the capacity smaller than the demand, the model consumes the least computational time.