| Summary: | 碩士 === 國立中央大學 === 統計研究所 === 101 === In this thesis, we consider the accelerated destructive degradation test (ADDT) in which the degradation characteristic is the major indicator of the product failure. We discuss the products whose degradation characteristic is of lognormal distribution with mean being log-linear in the product of the stress level and square root of the observation time. In ADDT, after we measure the degradation characteristics, the products damage. It is called the destructive measure. We use the Bayesian approach with the aid of Markov chain Monte Carlo method to estimate the parameters. Furthermore, we are interested in reliability inferences at the normal stress level, including the mean time to failure, failure time distribution quantiles, failure time piont, and reliability function. We also use the predictive distribution to predict the failure time. From simulation results, when the prior distribution is very informative, the Bayesian approach on ADDT provides accurate reliability inference.
|
|---|
