DIAGNOSTIC BAYESIAN NETWORKS

The paper presents a causal-probabilistic approach to the technical di agnosis in which the solution of the technical diagnostic problem is c onsidered as a probabilistic inference on a special kind of Bayesian n etworks called Diagnostic Bayesian Networks. A mechanism of probabilis tic inference and an algorithm for inference control are described. It is proved that a diagnostic problem represented by a singly connected Diagnostic Bayesian network can be decomposed to a sequence of subpro blems with directed tree or multitree topology which are exactly solve d in the sense of minimizing the average number of executed tests. The applications of the approach and future trends are briefly discussed.