(3.238.96.184) 您好!臺灣時間:2021/05/08 22:27
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果

詳目顯示:::

我願授權國圖
: 
twitterline
研究生:林胤堯
論文名稱:應用統計製程管制與類神經網路監控多注頭製程之研究
指導教授:邵曰仁邵曰仁引用關係李天行李天行引用關係
學位類別:碩士
校院名稱:輔仁大學
系所名稱:應用統計研究所
學門:數學及統計學門
學類:統計學類
論文種類:學術論文
畢業學年度:89
語文別:中文
論文頁數:76
中文關鍵詞:多注頭製程串檢驗法Shewhart管制圖類神經網路平均運行長度
外文關鍵詞:Multiple Stream ProcessRun TestShewhart ControlNeural NetworksAverage Run Lengthchart
相關次數:
  • 被引用被引用:0
  • 點閱點閱:123
  • 評分評分:系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:1
  今日產業界面對的環境是一個競爭激烈、趨向全球化及顧客導向的市場,大量、快速、廉價的生產是許多產業生存的必備條件;因此,各行各業無不尋求更好的生產方式以縮減生產成本,這不但關係企業是否能在激烈的競爭中脫穎而出,也關係著是否能達成顧客對產品品質的要求,由於多注頭製程的效率快速、產量較多,眾多業者採用此類製程,所以如何有效的監督多注頭製程是一項非常重要的課題。
  本文使用了統計製程管制技術中的串檢驗方法(Run Test)以及Shewhart管制圖方法監控多注頭製程,Run Test方法對單一注頭位移的偵測較敏銳,而Shewhart管制圖則對整體製程失控的監控效能較佳,但兩種管制方法各有其限制與不足之處。
  本文進而利用類神經網路中的倒轉遞網路模式,形成新的管制方法,並彌補統計製程管制技術中的Run Test方法以及Shewhart管制圖方法的不足之處。而為了評估所提方法之偵測能力,本實驗除選擇三個注頭、五個注頭及七個注頭等三種多注頭之製程,並針對數種注頭位移不同的型態,利用電腦模擬評估建購模式偵測個別注失控的正確比率,且同時將其結果與使用傳統統計製程中之Run Test方法及Shewhart管制圖方法之正確比率進行比較,了解何者於偵測失控時所需之平均運行長度(Average Run Length; ARL)較大。本文的研究發現:類神經網路方法可以明確地判斷出多注頭製程中,哪一個或哪幾個注頭失控,無論是哪一種多注頭製程,類神經網路方法的偵測能力均比Run Test方法以及Shewhart管制圖方法好,偵測製程失控所需的ARL十分地短,顯示應用類神經網路的新的方案擁有較佳的偵測能力。
  Nowadays, the capability to rapidly produce products in large quantities and low costs are essential requirements for industries to complete in such a highly competitive and customer-oriented environment. And all industries are trying their best to seek better production methods in order to reduce their production costs. Among them, the multiple stream process is getting more and more attention since it can satisfy the above-mentioned requirements. Therefore the monitoring and detecting of assignable causes in a multiple stream process is an important issue for both the researcher and industry processes. This study applies the methods of Run Test and Shewhart control chart to monitor the multiple stream process. This study uses a set of simulations to compare the performance of these two methods when the multiple steam process is out of control. The research findings indicate that while the Run Test method is more sensitive when one individual filling head has gone wrong, the performance of a modified Shewhart control chart is more effective when all of the filling heads are out of control. But both of them are restricted and lack enough detecting capability in some situations. In order to solve the issue of the above-mentioned drawbacks, this research proposes an alternative approach in monitoring the multiple stream process. By the ability of modeling complicated system and its generalization capability, the artificial neural networks (NNs) is used to make up the inadequacy of the traditional Run Test and Shewhart control charts, A set of simulations are used to compare the performance of NNs, Run Test and Shewhart control chart when the Multiple stream process is out of control. The case considered in the simulations are 3 filling, 5 filling and 7 fillings heads of a multiple stream process. The research findings indicate that the proposed NNs are more sensitive in detecting the out-of-control signal in terms average run length (ARL) than the alternative approaches.
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
系統版面圖檔 系統版面圖檔