Reliability Evaluation under Accelerated Degradation Testing with Recovery Capability Considered

Abstract

Degradation data always contains useful information on life and reliability, even if no failure has occurred. By applying additional stress, accelerated degradation testing(ADT) can shorten testing time. Many factors can affect the measured degradation data. Besides the degradation process itself and measurement errors, the measurement methods will also have impact on the degradation level. This paper considers the impact of data under two different data collection methods on the results of lifetime inference. Different statistical models are applied to the data from two measurement methods, both of which are based on a linear regression model. Likelihood inference for these models are developed. Simulation is used to demonstrate the effect of data collection methods on the inference results.