Accelerated life tests at higher usage rates

Accelerated life tests are extensively used to provide quickly the information about the life distributions of products. Test units are subjected to elevated stresses which yield shorter lives. For some products whose life is defined by usage, e.g., mileage and cycles, test units are also run at higher usage rates (UR) to compress the test time. This paper presents a method for testing products at both higher stress levels, and UR. Censoring time is pre-determined and fixed, while censoring usage is a function of UR. A UR effect model is proposed to describe the dependence of usage to failure (UTF) on UR. The relationship between UTF, and stress and UR is established, and used to estimate the UTF distribution at design stresses and usual UR. The model parameters are estimated by maximum likelihood method. The best compromise test plans, which choose the UR, stress levels, and sample sizes, are devised by minimizing the asymptotic variance of the estimator of a life percentile at design stresses and usual UR. The efficiency, and sensitivity of the test plans are evaluated. The results show that the test plans are efficient, and robust.