Data Processing System for NonStationary Processes Based on The Synthesis of Soft Computing Components

Abstract

The problem has been formulated and the scientific and methodological foundations of data processing systems for non-stationary objects based on neural networks, fuzzy set models, fuzzy inference algorithms, and neuro-fuzzy networks have been developed. Are proposed mechanisms for structural and parametric identification, finding a set of terms of linguistic variables, rules of inference, determining the coefficients of fuzzy rules, using the properties of self-adaptation, self-regulation, network organization, and the formation of databases and knowledge. The efficiency of the implemented algorithms was evaluated on the basis of test unimodal and multimodal output functions.