Modeling of Accelerated Degradation Tests with Interval Failure Time Data

摘要

The benefit of conducting accelerated degradation testing (ADT) is to predict the reliability of a highly reliable product without waiting for actual failures. Numerous stochastic models and statistical inference methods have been developed to analyze ADT data. In many situations, actual failure times associated with the interested degradation processes may also be observed during tests. However, such useful information has not been well studied in previous work on modeling ADT. This paper addresses different methods for analyzing ADT with interval failure times. A numerical study demonstrates that an appropriate method must be carefully selected based on the sample sizes of ADT data and interval failure times in order to achieve precise reliability estimates.