Statistical Analysis Of The Accelerated Destructive Degradation Model

High technology and competition brings more and more high-quality products. Degradation data analysis is a powerful tool for reliability assessment. However, for some applications, the degradation measurement process destroys or changes the phys-ical/mechanical characteristics of test units, so that only one meaningful measurement can be taken on each unit. An accelerated degradation test with such degradation data is called an accelerated destructive degradation test. In this article we discuss the statistical analysis of the accelerated destructive degradation test.Chapter 1 presents the design of the accelerated destructive degradation test and a class of models for accelerated destructive degradation test. In chapter 2 and 3, the max-imum likelihood estimation and Bayesian estimation of the unknown parameters based on full sample and right censored sample are discussed. In the Bayesian inference, two kinds of different prior distribution are respectively discussed to compare the estimation precision. One is non-informative prior distribution and the other is conjugate prior dis-tribution. The numerical simulation results demonstrates that Bayesian estimation is robust in the destructive degradation data analysis.In addition, on account of the difficulty to directly sampling from the posterior dis-tribution, numerical simulation is applied to do the estimation through Gibbs sampling and ergodic mean is used to estimate the unknown parameters.