Identification of nonlinear systems based on a genetically trained fuzzy neural network

Abstract EN

This paper presents an intelligent modeling technique that combines the merits of Fuzzy Logic (FL), Neural Networks (NNs) and Genetic Algorithms (GAs), where the GA is used to train a Fuzzy Neural Identifier (FNI) to identify ill-defined dynamical systems using the series-parallel identification model.

The parameters of the FNI (including the input and output scaling factors, the centers and widths of the membership !'unctions (MFs) for the input variables, and the quantization levels of the output variable, that are subjected to constraints on their values by the expert) are modified by the real-coding GA with hybrid selection method and elitism strategy based on minimizing the Mean Square of £٢٢٠٢ (MSE) criterion.

The simulation results for modeling three different nonlinear plants show the effectiveness of this FNI.