使用主成分分析及貝氏網路方法於離子植入製程之錯誤偵測與診斷

Abstract

本論文使用資料礦析技術中之統計方法，發展離子植入製程之錯誤偵測與錯誤診斷系統。在偵測系統中，我們分別使用貝氏分類法、自然貝氏分類法與主成分分析結合自然貝氏分類法來建立三種偵測系統。而錯誤診斷系統上則是使用貝氏網路方法與貝氏定理來建立。我們並以實際離子植入製程中由SECSII所量測之資料來驗證此偵測與診斷系統之性能。其結果證明此二系統皆具有滿意的錯誤偵測與診斷效果。我們預期此偵測與診斷系統 可顯著提升離子植入製程的穩定性、使用率、製程良率與可靠度。

In this thesis, the statistial approach of the data mining techniques is used to develop a fault detection and diagnosis system for the ion-implantation process. Three fault detection systems are developed using, respectively, the Bayesian, the naive Bayesian, and the principal component analysis(PCA) with the naive Bayesian. The fault diagnosis system is designed based on the Bayesian network and Bayes' theorem. The performance of fault detection and diagnosis is evaluated by the real data of ion-implantation process obtained via SECSII. Experimental results demonstrate that the success rate of fault detection and diagnosis system is satisfactory. This sytem is expected to increase the stability, utility rate, yield, and reliability of the ion-implantation process.