A Random Onset Model for Degradation of High-Reliability Systems

Weapons stockpiles are expected to have high reliability over time, but prudence demands regular testing to detect detrimental aging effects and maintain confidence that reliability is high. We present a model, called RADAR, in which a stockpile has high initial reliability that may begin declining at any time. RADAR provides a framework for answering questions about how confidence in continued high reliability can change as a result of reduced sampling, discovery of failed units, and information about when a unit failed. Supplemental materials (available on the Technometrics web site) provide lemmas used in the proof of Theorem 1, details of the Markov chain Monte Carlo algorithm, and additional examples.