Optimal design of accelerated life tests under model uncertainty

Abstract

The previous works on the optimal design of accelerated life tests are concentrated on the minimizing the variance of certain parameters at the use condition assuming that the relationship between the stress and the mean lifetime is a true one. However, the true relationship is not usually known in practice, and therefore, the experimenter must provide some protection against the likely departures from the assumed model. In this paper, the life time of each test unit is assumed to be exponential. It is further assumed that the experimenter assumes a first order relationship between the stress and the mean lifetime while the true one is a quadratic funciton. Developed is an optimal accelerated life test plan where the MSE of the estimated mean lifetime at the use condition is minimized.