晶圓廠針測區多工單等級生產規劃系統之設計

Abstract

本文所探討之對象為晶圓製造廠之內部針測區。由於針測區屬於晶圓製造前段製程中的最後一道加工步驟，因此其規劃目標傾向於各產品交期之達交。所以本文在考量多產品等級、以及配合前段製程之產出之情況下，以延遲工單數最少為前提，構建一符合長期目標之生產排程規劃模式。 本文主要分為兩個模組，首先，在粗略產能規劃模組中，考量產品與針測工作站間所存在的製程規格能力(capability)問題，並利用製程規格能力負荷分配演算法(COLA)[39]，平均分配針測工作站之最大可用產能於各個產品，以避免產生工作站間負荷不均的現象；其次，在主生產排程規劃組中，利用先前所得之各項規劃值，估算針測區中各工作站平均到達率，以及產出率等相關參數指標，並以等候理論(M/M/C、M/D/C)之相關數學模型，進行產品等級別生產週期時間的估算，並將所得之結果與產品之原定交期，計算出工單之最晚投料時點；之後，考量系統中所存在之各工單到達之不確定性，以及各工作站所剩餘之產能，制訂出各工單投料順序表等相關規劃值，以利交期的達成，而對於無法於最晚投料時點前到達之工單，在依照其實際投料時點給予修正後交期，藉以提昇公司達交水準。 模擬結果顯示，透過本文粗略產能規劃模組之設計，可快速地估算各工作站所需分配之最適產能負荷，並輔助主生產排程模組之生產週期時間之估算及各項規劃值之制訂，使系統達到有效性及可行性之規劃結果，以供現場生產活動控制做為決策之參考依據。

This paper focuses on the probing area in wafer fabrication. Because wafer probing is the last operation in the front-end of wafer fabrication, the main planning target is due-date oriented. By considering multiple order priority and unstable production environment in the front-end process, the thesis constructs a production scheduling and planning model to satisfy the long-term target and to minimize the number of delay orders. This paper proposes two modules. The rough-cut capacity planning (RCCP) module considers the issue of process capability between orders and probing workstation and applies the capability-oriented loading allocation (COLA) algorithm [39] to allocate capacity of probing workstations to products in order to avoid unbalanced loading among workstations. Based on the planning results obtained, the master production schedule (MPS) module estimates relevant parameters such as average arrival rate to each workstation and throughput rate, etc. Then queuing theory model (M/M/c & M/D/c) is applied to estimate the production cycle time for each product and each rank. Based on the estimated results and the original due date of each order, the latest releasing time of each order can be determined. By considering the uncertainty of order arrival time and the remaining capacity of workstations, release sequence list is prepared for achieving on time delivery. Furthermore, for the orders that arrive after the latest releasing time, a modified due date will be calculated based on the actual releasing time, in order to promote on time delivery. Simulation results show that the proposed RCCP module can efficiently estimate the appropriate capacity loading for each workstation and the MPS module can estimate production cycle time and other relevant parameters. The entire system can provide effective and efficient planning results, which are valuable for decision making in shop floor production control.