A Bayesian methodology for assessing reliability during product development

Abstract

Discussed are two issues of concern to the defense acquisition community: how to develop a reasonably accurate assessment of weapons system reliability in a small sample environment, and, in the absence of extensive testing, how to assess the impact of corrective action on reliability growth. A stepwise process is described for analyzing failure data derived from various sources: subassembly level, system level (factory), system level (field) and operational testing. Bayes theory is applied in a sequential manner to the various levels of testing. The prior-distribution and updating procedures at each level involve using engineering judgement to evaluate the relevance of the various kinds of tests, the significance of failures observed, and the effectiveness of corrective actions. The application of the Bayesian process sets the language and format that provide a framework for gathering, organizing, and incorporating the expert knowledge and consensus of the entire engineering team into the assessment of reliability growth.<>