標題: 使用灰關聯分析與決策實驗室分析法改善傳統失效模式與效應分析風險優先數Improving the Risk Priority Number of Traditional FMEA by Grey Relational Analysis and DEMATEL 作者: 蔡依庭張永佳張桂琥工業工程與管理學系 關鍵字: 失效模式與效應分析;風險優先數;灰關聯分析;決策實驗室分析法;failure modes and effects analysis (FMEA);risk priority number (RPN);grey relational analysis (GRA);decision making trail and evaluation laboratory (DEMATEL) 公開日期: 2011 摘要: 失效模式與效應分析（Failure Modes and Effects Analysis, FMEA）是一種已被廣泛用於提升產品品質及系統可靠度的方法，常使用風險優先數（Risk Priority Number, RPN）作為判斷失效模式對系統影響程度的依據，而風險優先數為嚴重度（Severity, S）、發生度（Occurrence, O）及可偵測度（Detection, D）三項各為1到10級指標的乘積。但許多學提出風險優先數包含高重複率、假設SOD重要性相同、未遵守順序權重準則，及未考量失效模式與失效原因直接和間接的關聯等缺陷，以至於根據風險優先數大小得到的排序結果可能不符合實際需求。本研究認為造成較多失效的原因應優先被改善，故本研究提出結合灰關聯分析（Grey Relational Analysis, GRA）與決策實驗室分析法（Decision Making Trial and Evaluation Laboratory, DEMATEL），首先利用灰關聯分析修正風險優先數，達到降低重複率且遵守順序權重準則的目的；接著使用決策實驗室分析法找出失效模式及失效原因間直接與間接關係，以達到優先處理造成較多失效的原因的目的。最後本研究將此方法應用於兩個實際案例以證實方法的有效性，並與其他方法比較，可供決策者有較合理的資訊作為參考。Failure modes and effects analysis (FMEA) is wildly used for improving product quality and reliability of systems. It utilizes risk priority number (RPN) as a basis of assessing influence of failures, the product of severity (S), occurrence (O), and detection (D) indicator, which is rated from one to ten respectively. But many scholars point out the shortcomings of RPN such as high reduplication rate, assume equal importance of SOD, not follow ordered weighted rule, and failed to consider the direct and indirect relationship between failure modes and causes of failure, so that the rank by RPN values may dissatisfy actual needs. This study thinks that the cause of more failures should give higher priority, so provides a combination of grey relational analysis (GRA) and decision making trail and evaluation laboratory (DEMATEL). Utilize GRA to modify RPN values for lowering reduplication and following ordered weighted rule first, then apply DEMATEL to explore the direction and indirection relationship in failure modes and causes of failure for giving higher priority to cause more failures. Finally, two real cases are presented to verify effectiveness of the measure proposed by this study, and compared with other methods for providing decision makers more reasonable reference information. URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079933514http://hdl.handle.net/11536/50077 Appears in Collections: Thesis