A bayesian approach to reliability demonstration for aerospace systems

This paper presents practical methods for utilizing Bayesian techniques to improve the accuracy of reliability demonstration during acquisition of aerospace systems. The approach focuses on simple methods that can be implemented with basic knowledge of Bayesian procedures and without specialist software packages. Conjugate distributions are employed to allow rapid computation of posterior distributions for three cases: (1) probability of success in a binomial process, (2) mean time between failure with an underlying exponential distribution and (3) mean time between failure with an underlying Weibull distribution. Additionally, a method is introduced to allow partially relevant data to be introduced into the analysis with an appropriate weight applied. Possible expansions are considered for cases where more experienced analysts and better tools are available. Finally, examples of the process based on representative data are provided for both a probability of success and mean time between failure problem