標題: SPC與EPC之整合研究 : 利用類神經網路構建製程干擾模式Integrate SPC and EPC: Using Neural Network To Construct Disturbance Model 作者: 許俊欽Jun-Ching Hsu蘇朝墩Chao-Ton Su工業工程與管理學系 關鍵字: 統計製程管制;工程製程管制;類神經網路;時間數列;SPC;EPC;Neural Network;Time series 公開日期: 2000 摘要: 統計製程管制（Statistical Process Control，SPC）與工程製程管制（Engineering Process Control，EPC）為兩種不同監控製程的方式，SPC管制圖的目的在於找出影響製程變異之可歸屬原因並加以去除，EPC的目的在於消除可預期之變異以使製程輸出值接進目標值。雖然SPC與EPC為兩種不同的製程監控技術，但是其兩者目的均是降低製程變異，因此整合SPC與EPC一同監控製程，定能有效的降低製程變異。過去文獻中，構建製程干擾預測模式均利用時間數列(例如ARIMA)的分析技術，但是有鑑於時間數列需經過許多繁雜的統計檢定步驟，因此，本研究利用類神經網路（Neural Networks）構建製程干擾預測模式，並利用此預測干擾模式構建調整方程式以進行EPC回饋調整，最後再整合SPC管制圖找出可歸屬原因並加以去除，以使得製程變異予以最小化。透過案例說明，利用類神經網路所構建之製程干擾預測模式比起時間數列ARIMA所構建之干擾預測模式能更有效的消除可預期之變異。Statistical Process control (SPC) and Engineering Process Control (EPC) are two different ways to monitor the process. The purpose of SPC is to find assignable causes and remove them. EPC is used to reduce the effect of predictable variation to keep the process output on target. Although SPC and EPC represent two different control techniques, their objective is the same, i.e. to reduce process variation. Therefore, combine SPC and EPC can effectively reduce variability. In the past, researchers usually utilized time series (such as ARIMA) to construct the disturbance model. However, constructing a time series model needs a complex statistical testing process. This research employs neural networks to construct the disturbance model. This disturbance model is used to construct an adjustment equation, thereby EPC and SPC can be combined to find assignable causes. A numerical example is analyzed. Results show that our proposed approach outperforms the traditional ARIMA approach. URI: http://140.113.39.130/cdrfb3/record/nctu/#NT890031043http://hdl.handle.net/11536/66525 顯示於類別： 畢業論文