Bayesian inference model for step-stress accelerated life testing with type-II censoring

Abstract

In this paper we present a Bayes inference model for a simple step-stress accelerated life test (SSALT) using type-II censored samples. We assume that the failure times at each stress are exponentially distributed with a mean that is a log-linear function of the natural stress level, and derive a likelihood function for the SSALT model under type-II censoring. We integrate the engineering knowledge into the prior distribution of the parameters in log-linear function, and through a Siegel-gamma distribution conjugation we can derive the posterior distribution for the parameters of interest. Applying Bayes approach to SSALT, the statistical precision of parameter inference is improved and the required number of samples is reduced.