Subbotin S. Methods of diagnosis model building on the basis of neuro-fuzzy networks in intelligent diagnosis systems

The purpose of the thesis - the creation of new and improvement of existing methods of building of diagnostic models on the basis of neuro-fuzzy networks in intelligent diagnosis systems for the solving of a problem of rising of quality, level of automation and speed of diagnostic model synthesis based on neuro-fuzzy networks. The object of research - the process of intelligent diagnosis system building. The subject of research - the methods of diagnosis model building in intelligent diagnosis systems based on neuro-fuzzy networks. The new methods and software tools for building of intelligent diagnosis systems that synthesize diagnostic models in a neuro-fuzzy basis with automatic parameter tuning and on the basis of hybrid stochastic search are developed. The model of sample quality and the model of diagnostic neural model quality as well as methods of ample forming are proposed. Their use allows to solve the problem of automation of the construction of intelligent diagnosis systems. The experimental research of the properties and characteristics of the developed methods was provided by the practical problem solving of diagnosis and automatic classification. The recommendations on their using are proposed. The obtained results are enhanced to the practical realization and introduced in enterprises and organizations.