Planning accelerated life tests with two stress variables

Abstract

Most of the previous works on designing accelerated life test plans were concerned with the case where a single stress is employed for acceleration. In this thesis, we develop optimal accelerated life test plans when two stresses are involved with possible interaction between them. The lifetimes of test items are assumed to follow an exponential or a Weibull distribution. The intermittent inspection as well as the continuous inspection is assumed. A factorial arrangement of test points is considered for an efficient utilization of equipment, and the low level of each stress and the proportion of test items allocated to each test point are determined such that the asymptotic variance of the maximum likelihood estimator of the qth quantile of lifetime distribution at the use condition or of an acceleration factor is minimized. Patterns of optimal plans are identified through computational experiments. Computational results indicate that the loss in efficiency due to intermittent inspection is not severe, which is an encouraging result in terms of testing effort. Also relative performances of the single- and the two-stress cases are compared in terms of relative efficiency and nonestimability. The restriction on the arrangement of test points is relaxed and statistically optimal plans without the restriction are developed. Comparing the restricted and unrestricted optimal test plans, we found that the restriction do not significantly influence the statistical efficiency.