Identification Algorithm for Wiener Nonlinear Systems Based on Adaptive Neural Fuzzy Networks

This paper discusses identification method of a class of Wiener nonlinear systems based on adaptive neural fuzzy networks. The Wiener system in current research is comprised of a dynamic linear block modelled by transfer function and a static nonlinear block approximated by a five-layer adaptive neural fuzzy networks. The special test signals involving separable signals and random signals are designed to achieve separation identification of the Wiener system. Firstly, we estimate the unknown parameters of the transfer function by applying the correlation analysis technique depending on separable signals input-output measurements. Secondly, employing the cluster method and the recursive least squares algorithm to estimate the adaptive neural fuzzy networks. The simulation results demonstrate that the developed identification approach can identify with effect the Wiener systems based on adaptive neural fuzzy networks and the adaptive neural fuzzy networks can well model the nonlinearity of the Wiener system.