Application of Bayesian networks to predict SMART power semiconductor lifetime

In this paper Bayesian networks are used to model semiconductor lifetime data from a cyclic stress test system. The data of interest is a mixture of log-normal distributions, representing different failure mechanisms and moreover, the data is censored. To understand the complex lifetime behavior, interactions between test settings, geometric designs, material properties and physical parameters of the semiconductor device are modeled by a Bayesian network. For the network's structure and parameter learning statistical toolboxes in MATLAB have been extended and applied. Due to censored observations MCMC simulations are necessary to determine the posterior distribution. For model selection the ARD algorithm and goodness of fit criteria such as marginal likelihoods, Bayes factors, posterior predictive density distributions and SSEPs are used. The results indicate that the application of Bayesian networks to semiconductor reliability provides useful information about the interactions between covariates and serves as a reliable alternative to currently applied methods.