可靠度評估程序之構建

Abstract

工業界的競爭漸趨激烈，顧客對產品的要求也日趨嚴格，不但希望產 品在使用時有一定的績效，也希望能預先知道產品的可靠度。業者若欲使 其產品具有競爭力，則必須提昇其產品的可靠度，而對產品進行可靠度評 估，則為提昇或維持產品可靠度水準的主要依據。現有文獻中所建議之可 靠度評估流程，均過於簡化不夠完備，且在其評估流程中也未提出適用的 統計方法，而使工程人員在實際運用上有其困難。因此，本研究主要目的 即在建構一套完整、實用的可靠度評估程序，提供評估流程中每一步驟所 應使用到的統計方法，使工程人員能更有效地分析可靠度測試計畫後的數 據資料，找出系統的失效機率分佈，並推導出系統的可靠度。本研究並以 二實例來說明如何應用所建構之評估程序。最後，本研究與三個現有之可 靠度評估程序做一比較，結果顯示本研究所提出之可靠度評估程序較為完 備。關鍵詞： 可靠度，可靠度評估，不可維修系統，可維修系統 The paper presents a procedure for evaluating the reliability of systems.This procedure emphasizes the statistical analyses of failuredata. It suggests proper statistical methods for each stepin the procedure. It covers tests for randomness, outliers and trend. It also includes goodness-of-fit tests (Kolmogorov-Smirnor test and Chi-square test) for testing the distribution andmodel of the failure data, such as Exponnential, Weibull, Normal, Lognormal, Gamma distribution and the power law model. When the failure data fails to fit one of the above distirbutions or the power law model, the procedure suggests that the failure data may come from many different groupsand it should be evaluated separately. In order to obtain a more precise results, theprocedure combines thecurrent data with the historical data which has the same distirbution to evaluate the reliability of the systems. The procedure can provide engineers a clearer picture of evaluating the reliability for systems and it is illustrated by two numerical examples. Finally, the proposedprocedure is compared with three existing evaluation procedures for systems and it is found to be superior. Key Words: Reliability, Reliability Evaluation, NonrepairableSystem, Repairable System