Artificial neural networks present a fast growing computing technique for many fields of applications including gas turbine diagnostics. This paper examines the network based on radial basis functions (radial basis network) and applied to recognise gas path faults. To assess diagnosis efficiency of the radial basis network (RBNs), it is compared with a multilayer perceptron. During the comparison, input data are the same for the both networks; however comparative calculations are repeated for different variations of these data allowing to draw more general conclusions on the RBN applicability. The objectives are to tune the networks, to compare them, and to assess efficiency of the RBN on basis of a probabilistic criterion of correct fault recognition. The comparison results show that the RBN is a perspective technique for gas turbine diagnosis. Key words: gas turbine, gas path diagnosis, radial basis network, probability of correct fault recognition.
