Practical Application of Bayesian Reliability Demonstration

This paper builds on the work of Yates & Mosleh [1] in developing a Bayesian reliability demonstration methodology for Australian defence that is simple to employ yet effective in minimizing test resources, maximizing test accuracy and ultimately minimizing financial and operational risk. The additional work is required primarily to resolve issues with respect to understanding the influence of the various parameters and test planning. The paper presents a sensitivity analysis on the parameters of BRDT´s, which establishes that the parameters of the Prior are the most significant source of error in the analysis. It also shows that the results are not particularly sensitive to variations in weight. A number of test planning methodologies for both exponential MTBF and P5 problems are then considered, with the conclusion that the sequential BRDT approach outlined in [2] is the most appropriate to meet the requirements. On the basis of these findings, the paper discusses a potential BRDT process for Australian Defence based on sequential methods and the work of [I]. This will require specification of maximum weights for each evidence class and acceptable risks in the contractual documents, as well as the required reliability characteristics. It will also necessitate formal pre-posterior analysis to ensure the Prior is appropriate and may also require the contractor to seek out additional evidence to satisfy the risk goals.