透過您的圖書館登入
IP:3.147.73.35
  • 學位論文

以類神經網路建立偵測自我相關製程平均值偏移和參數估計之雙邊管制法

A Neural Network-based Two-Sided Control Procedure for Monitoring and Characterizing the Mean Shifts of Autocorrelated Process

指導教授 : 鄭春生
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


傳統統計製程管制圖均在製程數據服從常態分配且符合獨立性的假設下發展而成。然而,在現今之科技發展迅速的帶動下,許多製程已採用自動化之生產及檢驗方式,但此種方式將導致製程數據間具有高度之自我相關性,若仍以傳統管制圖來監控製程將造成錯誤警報明顯增加。 過去之研究顯示,偵測自我相關製程中之偏移變異,應用殘差管制法具有較佳之效益,但預測模型之參數估計誤差將明顯增加其判斷誤差。而在類神經網路之管制方法上,大多數的學者所提出之方法只能處理單一自我相關係數下之製程,而實際製程之自我相關係數將隨時間或批量之差異而改變。因此,針對不同之自我相關係數下的製程數據,就必須有多個類神經網路才能達到全面性之監控效益。 為了有效改善過去管制方法之缺點,並能正確且快速偵測製程之偏移狀況,本研究以多個不同自我相關係數所獲得之數據,來建構以類神經網路為基之管制法。本研究亦利用類神經網路發展一估計偏移量之系統,當所提管制法偵測出製程發生偏移時,可進一步估計其偏移量,藉以提供資訊給工程人員立即調整製程參數,並使製程能快速地回復到管制狀態內。本研究中,以平均連串長度以及平均絕對誤差百分比作為評估之績效指標。研究顯示,類神經網路管制法之績效較傳統EWMA管制法佳,另外,再以模擬產生之製程數據,驗證本研究所發展的類神經網路管制法確實有實務應用之可行性。 本研究所提出之類神經網路管制法具有三項優良特性,第一、無論製程處於何種程度之自我相關係數下,都能具有良好的偵測效益;第二、所提出之管制程序可有效避免殘差管制法中預測模型之參數估計誤差之缺點。第三、如同傳統管制法,本研究所提出之類神經網路管制法對於平均值向上及向下偏移均具有相同之偵測能力。

並列摘要


In the standard applications of statistical process control, a state of statistical control is identified with a process generating independent and identically distributed random variables. In fact, the assumption of uncorrelated or independent observations is not even approximately satisfied in some manufacturing processes. A scheme of neural network is based on a function of multiple correlation coefficients instead of single correlation coefficient on traditional control charts. They are used to eliminate disadvantages of control charts and detect the process mean shift quickly and correctly. The main idea of neural network-based approach is to monitor the process mean shifts and detect process mean shifts and estimate the magnitude of shifts. We use the back-propagation neural network to solve these problems. A process control method will be more effective if the adjusted magnitude can be estimated correctly in this research of process mean changed. The performance of neural networks are evaluated by the average run lengths (ARL) and mean absolute percent errors (MAPE) using simulation. From simulation data of this model and developed results, the neural network method strongly outperforms EWMA charts and can be wildly adapted to many production processes. In this study, a neural network-based control method offers three main advantages. First, this method offers good detecting ability for most and most of the processes under any level of autocorrelation coefficient. Second, the proposed control procedures effectively eliminate errors and disadvantages of parameter estimating on residual control chart. Last, neural network control method is capable of detecting process of mean upwards and downward shift as well as traditional control charts.

參考文獻


42. 萬維君,「應用類神經網路於製程平均值變化之偵測及參數之估計」,元智大學工業工程研究所碩士論文,2001。
39. 吳聰宏,「類神經網路應用在品質管制中相關性製程數據之管制」,元智大學工業工程研究所碩士論文,1994。
41. 鄭春生、鄭靜彥,「以類神經網路辨識製程個別值數據之平均值、變異數及相關性的變化」,品質學報,6,29-43 (1999)。
2. Alwan, L. C. and Roberts, H. V. “ Time-series modeling for statistical process control,” Journal of Business and Economic Statistics, 6, 87-95 (1988).
4. Chang, S. I., and Aw, C. A. “A neural fuzzy control chart for detection and classify process mean shifts,” International Journal of Production Research, 34, 2265-2278 (1996).

被引用紀錄


邱玉琳(2007)。應用類神經網路與支援向量機於多變量自我相關製程變異來源之辨識〔碩士論文,元智大學〕。華藝線上圖書館。https://doi.org/10.6838/YZU.2007.00216

延伸閱讀