Abstract:

Model-based diagnosis describes a system using the structure and behavior model of the system. However, there exists uncertainty in model-based diagnosis because of the coupling relations among the faulty components. This paper incorporates Bayesian network into the framework of model-based diagnosis method and develops a Bayesian network model in terms of observations to solve uncertainty. The posterior probabilities of diagnoses are computed using the Bayesian network model to determine the most probable faulty component. An attitude control system of twoaxis satellite was taken as an example to illustrate and validate the proposed method.

Key words: computer application, model-based fault diagnosis, uncertainty, Bayesian network, posterior probability