運用OWA結合DEMATEL法修正傳統RPN方法失效風險的排序

Abstract

失效模式與效應分析(failure mode and effects analysis，FMEA)屬於風險評估的技巧，許多產業用它來評估系統或設施失效所帶來的影響，藉此協助工程師找出產品設計或製造過程中的瑕疵，執行正確的行動預防失效發生。目前大部分的失效模式與效應分析方法使用風險優先數（risk priority number，RPN）來評估失效風險，然而傳統的RPN方法有五個主要缺點：(1)衡量尺度的問題；(2)未考慮權重；(3) RPN值範圍違反統計假設；(4) RPN值有許多的重複數；(5)未考慮系統組成彼此的間接關係。因此，為修正傳統的RPN方法用於評估失效風險的缺點，本研究提出一個由下而上的歸納估計方式，以模糊集配合順序加權平均(ordered weighted averaging，OWA)結合決策實驗室分析法（decision making trial and evaluation laboratory，DEMATEL）重新排定系統失效風險的優先順序，並使用離心幫浦的例子進一步與本研究所提出之方法與傳統RPN法及Seyed-Hosseini et al. 於2006年所提出的方法比較。結果證明使用本研究所提之方法可有效解決傳統RPN法衡量尺度的問題、減少RPN值的重複數，並能得到更正確合理的風險評估結果。個案研究部分，本研究運用所提出之方法評估0.15 µm DRAM蝕刻的製程風險及智慧型電力模組(intelligent power module，IPM)的失效風險，排定其失效風險的優先順序，藉以提供重要的資訊供管理階層決策參考。

Failure mode and effects analysis (FMEA) is used as a risk assessment technique to determine the effects of systems and equipment failures for many industries. It can help engineers find product defects in design or manufacturing processes, and allow them to take corrective actions for preventing the failures. Most current FMEA methods use the risk priority number (RPN) value to evaluate the risk of failure. However, conventional RPN methodology has been criticized to have five main shortcomings as below: (1) problem of the measurement scale; (2) severity, occurrence, and detection are not equally weighted with respect to one another in terms of risk; (3) the RPN scale itself has some non-intuitive statistical properties; (4) the RPN elements have many duplicate numbers; (5) conventional RPN method has not considered indirect relations between components. Therefore, an efficient, simplified algorithm to evaluate the orderings of risk for failure problems is needed. This thesis proposes an inductive bottom-up estimation methodology, which bases on decision making trial and evaluation laboratory (DEMATEL) approach and uses fuzzy sets and ordered weighted averaging (OWA) for reprioritization of failures in a system FMEA. The proposed methodology resolves some of the shortcomings of the conventional RPN method. In numerical verification, an FMEA of the centrifugal pump is presented to further illustrate the proposed approach. After comparing the result that was obtained from the proposed method with the other listing approaches, this research has found that the proposed approach can resolve problem of the measurement scale, reduce more duplicated RPN numbers and get a more accurate, reasonable risk assessment than other existed approaches. For advanced experiment, the proposed method is applied to assess the risk of the 0.15 µm DRAM etching process and the intelligent power module (IPM) as two illustrative practical examples to provide valuable information for the decision makers. As a result, the stability of product and process can be assured.