A0368
Title: Parameter estimation for dual-stress accelerated life testing models using approximate Bayesian computation
Authors: Neill Smit - North-West University (South Africa) [presenting]
Lizanne Raubenheimer - Rhodes University (South Africa)
Abstract: Accelerated life testing can be used to estimate the life characteristics of high-reliability products, especially where conventional reliability estimation is not possible due to time and cost constraints. In an accelerated life test, products are exposed to more severe than their normal operating conditions by applying stressors to induce early failures. A time transformation function, which is a functional relationship between the parameters of the life distribution and the accelerated stressors, can be used to estimate the life characteristics of the products under their normal operating conditions. The resulting models are often complicated, and classical parameter estimation is not always possible. Bayesian accelerated life testing models are considered for some widely used life distributions. The generalized Eyring relationship is used as the time transformation function, which incorporates one thermal stressor and one non-thermal stressor. A likelihood-free method, using approximate Bayesian computation, is investigated for parameter estimation. The approximate Bayesian computation method is compared to classical methods, such as maximum likelihood estimation, in a simulation study.
