本研究是以倒傳遞類神經網路為基礎,發展一個製程管制法,用來監視並 區分製程平均值及變異性之變化。本研究考慮之異常情形包含:(1)管制 平均值向上或向下移動;(2)製程變異性改變;(3)製程平均值向上(或向 下)移動及製程變異性改變。 本研究發展的類神經網路有單一型及整合型兩種。單一型網路是依照樣本 大小分別建立,而整合型網路則是以一個網路應用於不同樣本大小之情況 。在研究中,我們也針對發展類神經網路之設計策略進行深入之探討。類 神經網路之偵測成效是以正確辨認率和平均連串長度來評估。模擬之結果 顯示本研究發展之類神經網路在辨認率上優於過去之研究。
In this research, a control procedure based on artificial neural networks for distinguishing the changes in mean and variance was developed. The out-of-control conditions considered in this research includes (1) an upward or downward shift in the process mean; (2) a change in the process variability; (3) a shift (up or down) in the process mean along with a change in the process variability. The types of neural network-based control procedures are developed in this research. One is developed based on different sample sizes, the other is developed independent sample size, The design strategies are also investigation in this research. The performance of the proposed neural network has been evaluated based on the recognition accuracy and the average run length (ARL). Simulation results show that the proposed control procedure is better than other research in terms of recognition accuracy.