- 學位論文
資訊粒化技術用於設備失效分析
Applying Information Granulation Technique to Equipment Failure Analysis
摘要
設備是公司重要的資產，高附加價值的設備更需要妥善維護及管理以滿足公司之生產營運需求。設備狀況之良窳深切影響生產系統之產能及產品之品質及良率，所以良好的設備管理是企業在競爭日益激烈的市場環境下生存之重要關鍵。近年來為了改善設備的安全性，增加生產產品的良率，應用網路科技於設備之診斷、維修與監控等技術逐漸受到學術界與實務界之重視。這些診斷與維修行為需要一決策系統來提供失效原因判斷與分類。儲存在資料庫或資料倉儲內的設備狀態歷史紀錄資料量相當龐大，這些大量的設備狀態資料在資料探勘方面有一個相當困難的問題，即是真實設備狀態資料經常為不平衡資料（Imbalance Data），也就是說設備正常資料量遠多於失效狀態之資料，這狀況會影響決策資料之正確性。本研究將透過資訊粒化技術來減少資料量，並增加稀少資料的比例以解決資料不平衡的問題，提供決策者更精確之資訊。 研究中將以台電公司火力發電廠為例.由於電廠中之產出為發電量，發電量對於設備各項參數如溫度、壓力、振動、輪軸轉速之變化亦最為敏感，因此失效模式分析將以發電量變化為主要監測對象，透過資訊粒化縮減資料量，再分別以類神經網路及支援向量機對縮減後資料進行訓練，而分類結果能輔助失效原因診斷，進而針對失效狀況即時作出回應。
並列摘要
Equipments, especially those with substantial adding value, are important assets of a company and need to be carefully maintained. Since most of the industries are highly automated, the condition of machines directly affects the capacity, quality and yield of a production system. As a result, equipment failure mode analysis becomes a crucial issue for equipment providers and shop floor supervisors. The historical information of a machine can be voluminous, and imbalance phenomenon often exists between normal data and failure data. In this study, imbalance data mining, a data mining technique, is adopted to solve this problem. Some practical data are used to verify the feasibility of this method. This study takes the Taipower thermal power plant as its example. Since the parameters of a machine or equipment, such as temperature, pressure, vibration, and turbine shaft speed, exert considerable influence on the power load, Failure Mode and Effect Analysis conducted by this study focuses on monitoring the power load. The method reduces the size of information by using granulation technique and trains the reduced information by neural network and support vector machine. The results of classification can help diagnose the reasons of failures and provide real-time responses to handle the failures.