Determining Performance with Limited Testing when Reliability and Confidence are Mandated.

Dolin, R. M. ; Treml, C. A.

Technical Information Center Oak Ridge Tennessee

关键词: Confidence;    Reliability;    Bayesian binomial hypothesis;    Mathematical derivations;    Failure load intervals;   

RP-ID  :  DE2001763371

学科分类：工程和技术（综合）

美国|英语

来源: National Technical Reports Library

【 摘 要 】

In an environment of increased cost, regulation, and accountability the ability to do unlimited testing to verify the performance parameters of a product are diminishing. At the same time liability and maintenance concerns are requiring that clearly established operating conditions be defined. Because of this, product developers often encounter situations where reliability, and the confidence in that reliability, are mandated but only a limited number of test can be performed to validate these mandated conditions. A methodology is presented for establishing performance parameters in these situations. The methodology is based on Bayesian Hypothesis Testing theory. It is shown how an iterative hypothesis can be defined, Within the a Bayesian framework equations are derived for computing a necessary prior, determining the minimum number to test needing to be performed, establishing how many failures can be observed for a given number of tests, and inferring an overall reliability distribution.