應用修正灰預測模型與ACUSUM-C管制圖建構具自我相關製程之管制流程

Abstract

在今日競爭激烈的市場，增進產品製程管制的能力進而製造出穩定且高品質的產品，使其能保持競爭優勢甚至領先世界各國其他競爭者，對於製造廠商而言是非常重要的課題。統計製程管制(Statistical Process Control, SPC)中的管制圖已被廣泛地應用於工廠內部製程的監控與改善。然而，在現今高科技產業中，許多製程是屬於連續性製程，即生產環境是高度自動化之連續生產，因而使得製程輸出值間普遍存在自我相關(autocorrelation)，違反了管制圖資料須獨立之假設，導致管制圖容易出現假警報的現象，嚴重影響到管制圖的偵測績效。本研究針對製程產出間具自我相關之情況，利用修正灰預測模型與適應性累積和管制圖(ACUSUM-C)建構一套有效之線上(in-line)管制流程，以改善自我相關製程造成過多假警報的現象，且同時可偵測製程中可能發生的大、小偏移。本研究最後利用一個實例與模擬案例說明本研究方法的有效性及可行性。

In today's competitive market, enhancing product quality becomes an important issue for manufacturers. Statistical Process Control (SPC) has been widely employed in many industries. In order to control the process, the control chart is utilized to detect the assignable cause of process shift effectively. However, if the process has strong autocorrelation, the performance of control charts will be affected and false alarms will be increased. In this study, modified grey model and ACUSUM-C control chart are utilized to construct an effective modified procedure to eliminate the impact of autocorrelation in autocorrelated process. Furthermore, both of the small and large mean shifts can simultaneously be detected using the proposed procedure. Finally, a real case and a simulated case are utilized to demonstrate the effectiveness of the proposed procedure.