中文摘要 本研究主要係針對具有因龐大的設備成本導致過高的維護成本之產業為研究對象,並結合可靠度(Reliability)的觀念與成本效益分析來發展一套以設備製程不良率為基礎之預防維護決策系統。 本研究所考慮的參數有:平均失效間隔時間(MTBF)、生產批量數(N)、預防維護成本(PMC)、單位產品成本(IC)以及單位產品製程時間(PT)…等,期望藉由以上參數來建構預防維護模式以制定製程不良率之上限值(P)為停機維護的閥值(Threshold Value),並利用系統模擬軟體eM-Plant的運算能力來有效執行其計算過程。另外,藉由倒傳遞網路(BPNN)之函數對應能力,建構設備之預防維護決策系統。 最後利用模擬機台生產製程之實驗資料進行數據分析,驗證本維護模式之可行性。且本研究亦驗證出預防維護決策系統具有不需等到設備失效即可預知何時該執行預防維護的能力,有效達到預知性預防(Prevention by Prediction)之特性。 關鍵詞:預防維護、可靠度、不良率、倒傳遞網路、預知性預防
Abstract The purpose of this research is about combining the concepts of reliability and cost-effect analysis to develop a preventive maintenance decision system based on manufacturing process defective rate for those industries with the feature of high maintenance cost for the equipments. Parameters mainly considered in this paper include: Mean Time Between Failure (MTBF), Production Lot (N), Preventive Maintenance Cost (PMC), Item Cost (IC) and Process Time per Unit (PT). We expect not only to build a preventive maintenance model by using these parameters but also to decide the upper limit of process defective rate (P). Once the process defective rate is beyond to this limit, an immediate equipment maintenance has to be carried out. During this study, we use eM-Plant to simulate the proposed model, and use back-propagation neural network to construct the decision system. In this study, data collected from the simulation process are utilized to demonstrate the feasibility of this model. The unique feature of the preventive maintenance decision system is that it can be applied to practical manufacturing process, and use its capability of prevention by prediction to detect the time to maintain in advance. Key Word:Preventive Maintenance , Reliability , Defective Rate , Back-Propagation Neural Network , Prevention by Prediction