Models and Methods for Degradation Processes

To assess the reliability of components from the study of their degradation processes, it is necessary to develop (1) models that are simple, and (2) statistical methods to design experiments and analyze their results. The models must be physically meaningful to increase confidence in the extrapolations in time and in stress required to derive component reliability. The statistical methods must be capable to provide information useful in setting up strategies for burn-in, qualification or certification programs. In particular, the statistical methods must be able to discriminate between the various physical degradation models that may be postulated. In this talk, we describe models and statistical methods which meet the above requirements, and illustrate them with two examples.