統計製程管制 (SPC) 與工程製程管制 (EPC) 為兩種品質改善之策略。SPC主要是在找出影響製程變異之可歸屬原因,並加以去除,來降低製程之變異性。EPC管制法主要在於定期地透過調整製程之可控制變數,來消除影響製程之干擾,以使製程變異減至最少。 傳統SPC管制法的使用是假設取樣的品質數據間彼此獨立。EPC管制法則是應用於製程數據間存在相關性之情況。EPC管制法之應用有賴於建立自我相關性結構。 近年來,整合統計製程管制與工程製程管制的策略深獲產業界之興趣,透過整合SPC與EPC,使得製程管制的效果更臻理想。本研究發展兩個類神經網路來整合SPC與EPC。第一個類神經網路是用來建立具有自我相關性之結構,以去除製程中之干擾。接著以第二個類神經網路來做為監控製程之工具,相當於傳統管制圖法。本研究是以模擬方法,估計平均連串長度 (ARL) 來評估類神經網路之效益。經由深入之分析及比較,研究結果顯示類神經網路優於傳統的方法。 關鍵詞:統計製程管制、工程製程管制、類神經網路、平均連串長度
Statistical process control (SPC) and engineering process control (EPC) are two strategies for quality improvement. SPC seeks to reduce process variability by detecting and eliminating assignable causes of variation. On the other hand, EPC seeks to minimize variability by reversing the effect of process disturbance through regular adjustments to manipulable process variables. SPC is traditionally applied to processes where successive observations are statistically independent. EPC is usually applied to processes in which successive observations are related over time. The application of EPC relies on the auto-correlated structure. Recently, the integrated SPC/EPC scheme is gaining interests in industries. Superior control can be achieved through the application of integrated SPC/EPC scheme. In this research, two neural networks had been developed to model the combined SPC/EPC scheme. The first neural network was employed to construct the auto-correlated structure to cancel out the disturbance. The second neural network acted as a monitoring tool similar to traditional control charts. The performance of the proposed neural networks was evaluated by the average run lengths (ARL’s) using simulation. The results show that the proposed neural networks outperform traditional approach. Keywords: statistical process control, engineering process control, neural networks, average run lengths