Neuro–fuzzy network for identification and control of nonlinear systems

Abstract

    Unnecessary rule can be eliminated from a rule base and an insignificant variable from a learned rule can be removed using the parameters of learned GRBF network when identified nonlinear system. Jang proposed the recurrent neuro-fuzzy network RNFN which have been shown to possess good function of identification, long-range prediction and control capability for dynamic nonlinear systems. This paper combines the advantages of the two networks to build fuzzy-neural network which identify and control for a complex objects. Fuzzy-neural network with generalized fuzzy function of GFM and recurrent neuro-fuzzy network is used for modeling or forecasting system. Then the results of fuzzy neural network is used to build a long-range predictive controller based on GPC algorithm to control nonlinear dynamics object which is multiple input-single output MISO.