Fefelov A. The models and methods of technical diagnosis on the basis of artificial immune systems and Bayesian networks.

The thesis deals with the problem of development and implementation of fault detection and isolation systems. The main goals of such system are: detection of anomalies during operation of complex technical systems in real time; search for type of a failure and it's localization in a case of incomplete, inexact and inconsistent information environment; prediction of its future technical state. The review had been made and analysis of existing fault detection methods was performed. A new multilevel diagnostic model is offered. The new generalized technology of artificial immune systems construction to deal with the problems of fault detection and isolation is offered. This technology has a distinctive feature of universality of application and allows development of mathematical description of parameter's drift and detection of anomalies. The approach to detection of anomalies during technical system operation which uses mechanisms of negative selection and immune network is proposed. The new method and algorithm of detection a location and type of a failure with Bayesian networks of a modified structure and information-cost criterion is developed. A new information technology of synthesis and adjustment neural networks by means of artificial immune systems is created. This technology is used to solve the problem of prediction for a drift of technical object parameters. The new software architecture is developed and computer based information-analytical system for solving problems of fault detection and isolation was created.