Design of accelerated degradation tests based on the wiener process model

Abstract

As industrial competition becomes stronger and the development period of product becomes shorter, the adoption of accelerated tests is nearly a necessity for evaluating the lifetime of product. Accelerated degradation tests (ADTs) have been effectively used for this purpose. Before conducting an ADT, such decision variables as stress levels (or stress increasing rates), proportions of units allocated to the stress levels, stress level change times, measurement times and/or total number of test units should be carefully determined in advance to obtain the most precise and efficient estimates of the lifetime information at the use condition.In this thesis, statistically optimal and compromise ADT plans are developed under step- and progressive-stress loading methods. For both cases, the degradation characteristic of test units is assumed to follow the Wiener process. As optimization criteria, the asymptotic variance of the MLE of the q-th quantile of the lifetime distribution at the use condition and the total test cost are adopted. For the step-stress case, stress level change times are allowed to change between successive measurement times. Then, the two stress levels and the stress level change time is optimally determined for a two-step-stress (statistically optimal) ADT plan, and the three stress levels and two stress level change times are optimally determined for a three-step-stress (compromise) ADT plan. Finally, by taking the number of measurements into account, a method for selecting the cost-optimal ADT plan that minimizes the total test cost is developed for both two- and three-step stress cases.In the progressive-stress case, unlike the step-stress case, the type of decision variables depends on the number of measurements (m). For statistically optimal progressive-stress ADT plans, two stress increasing rates and the proportion of units allocated to each stress increasing rate are optimally determined for the case where m = 1, and two measurement ti...