A Study of Sample Size Analysis to Support Reliability Test Programming

樣本數對可靠度參數估計影響之研究

蔡有藤(Yuo-Tern Tsai) ； 王國雄(Kuo-Shong Wang)

《中國機械工程學刊》 28卷3期 (2007/06) Pp. 315-323

https://doi.org/10.29979/JCSME.200706.0008

摘要

本文研究不同樣品大小及數據型態對估計可靠度參數影響，內容包含可靠度數據收集、非參數化及參數化可靠度估計方法介紹；針對參數化估算模型，建立二參數偉伯分佈最大相似估算法，分析不同失效數據及截尾數據對估算精度影響，建立可靠度試驗時間、成本估算模式，分析不同精度要求下所需之樣本數、試驗時間、試驗成本，以為規劃可靠度試驗的依據。

並列摘要

The paper investigates the influences of different sample sizes to estimate reliability parameters to support reliability test programming. The methods of data collection are first introduced for constructing the basis of statistical reliability analysis. The analyzed approaches including the nonparametric methods and the parametric methods are reported so that the reliability-related functions can be appropriately modeled. Focusing on parametric methods, the maximum likelihood estimator (MLE) is used for estimating the reliability parameters. Two parameter's Weibull distribution is adopted as an example to investigate the estimated accuracy for the parameters at different sample sizes. According to the investigated results, the proper sample size and the truncated scale for the tested data can be validly determined no sooner than the expected precision given. Moreover, a model of estimating the length of testing is derived based on the distribution of failure data of product. Integrating the estimated results in sample size, truncated scale and test length, the needed total cost for reliability test can be accurately evaluated. As a result, the optimization for reliability test can be programmed by minimizing the total cost.

並列關鍵字

reliability analysis ； parameter estimation and sample size