晶圓廠在製品流程時間之實證模式建構

Abstract

目前晶圓代工廠的服務範圍已經擴大到設計支援、光罩製作、晶圓製造、晶圓測試封裝，以及產品問題的分析。也就是說製造部門需要與工程部門有更密切的配合。製造部門不僅僅是注重機台的工作效率，需要能夠更進一步掌握在製品的流動行為，以獲取穩定及可接受的達交率。半導體製造轉型成為製造服務業，重點擺在符合顧客的需求，因而業者所面臨的挑戰包括了有效率的工程設計與生產製造的整合，服務品質與進度的監控，以及達到穩定及可信賴的達交績效。針對製造服務業的新形式，不同於以往的製造業，需要建立一套適用於現今環境的生產管制方法。 對於半導體製造供應鏈來說，上游的下單顧客期望他的訂單能夠準確及時地滿足終端市場的需求，顧客在意的是他訂單目前的狀態，以便調整他的市場策略。 因此，半導體製造業者評估顧客滿意度的關鍵績效指標分為兩項：交期排程的正確性以及訂單流程時間。 對於這兩項績效指標，本研究以系統模擬的方法建構處於產能滿載的晶圓廠行為模型，分析晶圓廠中在製品的流動行為，並以產品流程時間的平均天數與標準差天數這組數值來描述工廠的行為，研究結果顯示可以利用常態分佈來表示產品流程時間的機率分佈，以之作為績效評估指標，可以用來監控與管制晶圓廠的達交績效。另一個核心議題為探討系統利用率與流程時間的關連性，結果顯示兩者的關係式為指數函數。根據此資訊，晶圓廠能提供晶片設計商在目前產能利用率下，產品流程時間預測的參考，並作為下單時間的調整依據。

The services of the foundry cover mask design, fabrication, assembly, final test, and trouble shooting. In other words, the manufacturing department and the engineering department need to be more cooperated. To attain robust and acceptable delivery performance, the manufacturing department should not only focus on the machine utilization, but also on the WIP behavior. As claimed to be manufacturing service, the niche is to meet the customer requirement. The challenge will be to overcome the controllability and predictability problems, service monitoring and control of the business orders. In IC supply chain, a chip provider pays attention to the current state of released orders in order to meet the market requirement. Hence, on customer satisfaction, there are two key performance indexes at the foundry: delivery schedule accuracy and order fulfillment lead time. For these two indexes, a fully loaded foundry model based on simulation program has been established in this thesis to analyze the WIP behavior. The research is to obtain the mean and variance of the flow time. The distribution of the flow time has been validated and it seems to be the normal distribution. Another core issue is to analyze the relationship between system utilization and flow time; our result shows that the relationship follows an exponential function. Based on it, foundry can provide IC design house the forecasting flow time with the current capacity utilization, and as the adjustment reference of order release time.