A general accelerated life model for step-stress testing

TLDR

The paper proposes a general accelerated life model for step‑stress testing and a likelihood formulation applicable to Weibull and lognormal distributions. The model applies to any life distribution whose scale parameter changes with stress, extends to multiple‑stress and profiled patterns, and includes algorithms for fitting and testing. The model enables convenient interpretation of step‑stress testing results and its practical utility is shown in a case study.

Abstract

In this paper we propose a general accelerated life model for step-stress testing and present a general likelihood function formulation for step-stress models that use both Weibull and lognormal distributions. The proposed model is also applicable to any life distribution in which the stress level only changes the scale parameter of the distribution, and can also be extended to multiple-stress as well as profiled testing patterns. Algorithms for fitting and testing such models are described and illustrated. The model provides a convenient way to interpret the step-stress accelerated life testing results. The practical use of the proposed statistical inference model is demonstrated by a case study.