USE OF THE DIRICHLET PROCESS FOR RELIABILITY-ANALYSIS

Recently, many researchers apply the non-parametric Bayesian approach to predict the reliability of highly complex electronic systems. The Dirichlet process is the most common model for the non-parametric Bayesian analysis. The Kuo's simulation procedure [6] for Dirichlet process under a variance reduction techniques introduced in Chien and Kuo (1994) [2] is applied for a Weibull-distributed system. Optimal burn-in time is determined given the cost parameters. A model, the percentage of good items in a lot, is used to explain when the Dirichlet process is not a proper choice.