Optimal planning and inference for sequential accelerated life testing with two or more experimental factors

In this paper, we consider the planning (design) stage as well as the inference (analysis) stage of ALT. One of the most important step is to design a statistically sound ALT test plan such as deciding on the number of factors for the test, the levels of each factor, sample size, censoring type, etc. The test results often need to be extrapolated through a physics -based and statistical model to obtain the reliability prediction at use level. The traditional design of experiments approach is inadequate for ALT planning and analysis since the test would yield censored data, non-occurrence of failure modes at some stress levels and so on. In addition, most of previous work on designing ALT plans focused on ALTs with single stress. Escobar et al. designed and analyzed ALTs with two or more experimental factors, assuming a linear model without interaction between factors.