Research toward a Bayesian Procedure for Calculating System Reliability

The purpose of this paper* is to describe research toward the application of Bayesian statistical methods to the calculation of system reliability The motivation for the suggested procedures stems from (1) the fact that Bayesian methods provide for the formal inclusion of all relevant information in reliability calculations and (2) the suitability of Bayesian methods to problems in which the probability distributions of interest change over time. In this report, suggested procedures are given for obtaining component reliability u estimates based on system test outcomes. Three possible test outcomes are considered: (1) success, (2) failure caused by known components, and (3) failure caused by unknown components. Using this procedure, substitution of new components or modification of the configuration of the system does not invalidate the test data accumulated up to the time of the changes.