為了有效提升產品的生產品質,管理者會針對生產線進行監控及調整,而統計製程管制和工程製程管制是一般最常用的方法。在資料不具自我相關的情況下,SPC整合EPC將可有效地辨識可歸屬原因並將之消除,然而當資料具有自我相關性的情況下,SPC整合EPC將有發生假警訊之虞。因此,本研究以資料分群的概念利用成長型階層式自組織映射圖網路來建構相關性製程的監控模式,以協助管理者即時判斷製程的可歸屬原因,並達到即時調整的目標。研究結果發現,本模式除了可即時判斷製程的可歸屬原因外,在明顯的干擾時,相對於其他方法,亦能有效地辨識出結果。
Generally speaking, Managers address themselves to monitor and adjust process in order to effectively enhance the quality of product. Statistical Process Control (SPC) and Engineer Process Control (EPC) have been widely applied. When the data is without attribution of self-correlated, integrating SPC and EPC can effectively discriminate the assignable causes and remove them. However, when the data is with attribution of self-correlated, there is a problem that integration of SPC and EPC may inform a false alarm. In this paper, the concept of clustering is applied to solve this problem. Considering the hierarchical property of observed data in real world, growing hierarchical self-organizing map (GHSOM) is adopted. The proposed model can discriminate the assignable causes and adjust abnormal status immediately. The experimental results reveal that our proposed method is effective and efficient for disturbance identification in correlated process data when disturbance is significant.