半導體後段製程產出之預警機制

Abstract

摘要 目前半導體業者為了提升本身的服務品質，各個代工廠紛紛以策略聯盟的方式結合，提供IC設計公司整套產品製造的服務，此種服務模式在半導體業界稱之為一元化服務(Turn Key Service)。一元化服務有很多種不同的型態，其中一種型態為IC設計公司向晶圓代工廠下單，晶圓代工廠則提供從晶圓製造、針測、封裝、到最終測試的全套服務。在這樣服務體系內，後段製造廠的生產系統是否具高彈性，以及是否能快速、準確地完成達交承諾，則是晶圓代工廠執行一元化服務的競爭優勢。 然而，以往晶圓代工廠的IE/PC人員監控後段製造廠交期進度是經由後段製造廠定期所回饋的在製品資訊，再憑藉經驗法則來評估交期是否延遲，並沒有一套明顯的管理機制予以掌控，故往往有延遲的情形發生。有鑑於此，本論文目的在於建立一套產出預警機制，其包含了產出預估模組、產能負荷評估模組、延遲預警判斷原則與延遲資訊回饋系統。本論文期望能藉由此預警機制幫助晶圓代工廠的管理者即時做正確的決策，不至於反應過慢，而造成產品交期延誤、產出量落後等嚴重問題；更期望能藉由預警的措施進一步替整個IC製造供應鏈締造更高的服務價值與競爭力。最後，經由與Vigilance、GLT兩系統整合，確實可透過Web畫面讓晶圓代工廠的管理者即時瞭解後段製造廠目前的產出狀況，做出即時的反應與處理，防範延遲的發生。

To advance in service quality for IC design house, many foundries are confederate to provide a package service, called “Turn Key Service”. There are many definitions of Turn Key Service. One kind of them is that foundries provide an integrated service including wafer fabrication, wafer probing, IC packaging, and IC final testing. In such a system, the higher flexibility semiconductor back-end provides, the more competitiveness the corresponding foundry Fab has. However, planners of foundries may not have a good management mechanism to monitor and control production progress of semiconductor back-end. They normally used their experience to identify whether products could be delivered on time or not. As a result, this research is purposed to develop an exception management mechanism to monitor outputs in the semiconductor back-end supply chain. This mechanism consists of output prediction module, capacity evaluation module, exception detection rules, and feedback system. We intend to help planners of foundries make the right response once a possible exceptional event will occur so that a higher value for the entire semiconductor supply chain can be obtained. We integrate another two management systems, Global Lot Tracking and Vigilance, to show the exception management mechanism that can provide on-time information for planners of foundries to prevent products from being delayed.