The SABRE Method: Designing a Reliability Test Program for an Artillery Fired Atomic Projectile (AFAP).

摘要

The SABRE method (Simulation Approach to Bayesian Reliability Evaluations) is a straightforward analytical approach and is presented utilizing mathematical modeling, Monte Carlo techniques, and Bayesian statistics. The objective was to design a test program of minimum sample size for an artillery-fired atomic projectile (AFAP) presently being developed. Engineers and mathematicians constructed a subcomponent level mathematical model of the system. A 'NO-KNOWLEDGE' prior (Beta) distribution is assumed for each data point in the model and a series of Monte Carlo experiments are performed, each utilizing different sample sizes and assumed failures. Each Monte Carlo produces a Bayesian system posterior distribution as in the Tri-Service approach. All posteriors are compared against design requirements, and the sample size which produced the best posterior was selected. (Author)