Example of Diagnostic Inference Based on Uncertain and Partly Inconsistent Data with Application of the Approximate Statement Network

The paper deals with diagnostic inference based on uncertain and simultaneously partly inconsistent data obtained, e.g. from different sensors. Such cases are very common in diagnostic practice and therefore there is a necessity to deal with them. Interesting approach to solving that kind of tasks consists in an application of the approximate statement network which represents the mutual relations between statements treated as necessary and sufficient conditions. The paper shows an example of applying a diagnostic model represented as the approximate statement network, to inference about technical state of a chosen object. The model was constructed in the REx system which also makes possible creating Bayesian and multimodal networks. Advantages and disadvantages concerning both constructing and using approximate statement networks were briefly described on the basis of obtained results. It seems that presented example shows the possibility of improving supervision systems, especially in regard to the complicated technical objects, by giving a mechanism of avoiding a confusion while making of diagnosis.