Using sensitivity analysis to validate Bayesian networks for airplane subsystem diagnosis

In this paper, we introduce our practice of building Bayesian networks for airplane subsystem diagnosis. We use an efficient sensitivity analysis method to validate our diagnostic models. The algorithm is based on relevance-based decomposition in joint tree computation framework. We also describe how to use sensitivity analysis in model elicitation procedure to validate Bayesian network models based on our practice of model building for airplane subsystem fault isolation.