Belief networks and diagnostic modelling

A general approach of methodology related to diagnostic of machinery is based on passive and/or active diagnostic experiments. Objects of this research are often represented by few examples, which make experimental verification of applied models difficult. It requires that specialists' opinions have to be taken into account as an additional knowledge. In the paper particular attention is paid to a special role that can be played in technical diagnostics by belief networks. A new mixed approach was proposed. Inverse models co-operates with special reasoning systems leaking use of belief networks. A general introduction to inverse models and belief networks was also presented. One explained that quality of diagnostic belief networks depends on quality of parameters being introduced on the basis of specialists' opinions. Effectiveness of the application of these networks depends on successful solutions of particular optimisation tasks related to their verification and validation. Possibilities for improving these networks were also discussed.