Fiducial inference-based failure mechanism consistency analysis for accelerated life and degradation tests

Accelerated tests are usually done at the elevated stress levels, while extrapolation is needed to obtain the reliability estimations at the normal stress level. The premise of such extrapolation is that the failure mechanisms at different stress levels are consistent, otherwise, the obtained reliability estimations would be seriously biased. This paper treats the problem of testing the consistency of failure mechanism for the accelerated life and degradation tests, which is formulated as a multiple comparison problem towards the coefficients of variation. Then, a fiducial inference based method is proposed to tackle this problem. Under this framework, fiducial quantities of the coefficients of variation or their monotonic transformations are derived for some extensively used lifetime and degradation models. Simulation studies are conducted for evaluation, whose results indicate that the proposed method outperforms the benchmark methods regarding the type I error and power. Finally, several examples are provided for illustration. © 2022 Elsevier Inc.