Accelerated Life Tests: Designs Comparison within a B ayesian Framework

Abstract

Abstract Accelerated life testing (ALT) is the set of procedures used to reduce the time needed to obtain information related to life characteristics of an item, material or part of interest. The importance of the design of these ALTs is paramount to obtain inferential life estimators with, for example, minimum variance. Typically, for a set ALT, step‐regimen optimal stress levels and duration are determined in the ALT design phase. Herein we focus on the comparison of different ALT designs (fixed stress, profile ALT, progressive step‐stress ALT, and regressive ALT) within a single Bayesian inference framework. We analyze the preposterior variance of the use stress reliability based on a single failure over the course of the ALT for these different ALT designs. To the best of our knowledge Miller and Nelson in their 1983 article titled “Optimum simple step‐stress plans for accelerated life testing” in IEEE Transactions on Reliability R–32 , 59–65, were among the earliest to claim similarity of asymptotic estimator variance when comparing step‐stress testing ALT with fixed stress ALT. The results in this article shed some light on this claim within a Bayesian context.