異常製程機台之探勘預測

Abstract

半導體產業已成為台灣主流科技產業。在晶圓的生產過程中，如何避免晶圓暇疵與缺陷等問題，進而提高產品良率與生產產能並降低成本支出，已經成為每家半導體廠所不得不重視的問題。半導體製程技術發展越來越精密，所以先進製程技術的提升扮演非常重要的角色。本研究透過生產資訊系統中的錯誤診斷分析系統來提升製程技術的即時監控。由系統自動產生監控規則即時發佈預警訊息，自動找出根本的原因並應用資料探勘技術，尋找異常製程機台以提高晶圓品質良率。製程與設備人員經由管制圖分析可清楚發現異常情況，進而縮短相關生產人員尋找異常原因之時間，以明確找出機台、反應室、製作方法異常問題，並提升機台的稼動率達到資源充分利用。

The semiconductor industry has become the mainstream technology industry in Taiwan. In the wafer production process, how to avoid issues such as wafer flaws and defects, and to improve product yield and production capacity and reduce costs has become the major efforts of semiconductor factories. The development of semiconductor process technology which is increasingly sophisticated and advanced, play a very important role. This study improves the real-time monitoring of process technology by analyzing production information. The developed system automatically generate system monitoring rules to immediately release alarm messages and automatically determines the underlying reason. This study employs data mining and statistical techniques to identity abnormal process equipments to improve the yield of wafer quality. Through the control char analysis, abnormal process equipment can be identified, to find the cause of exception and shorten the production. The proposed study can clearly identify the abnormal problems of production process and improve the utilization rate of the machine to achieve the full use of resources.