Synergetic synthesis of a neural network controller for an adaptive control of a nonlinear dynamic plant

Abstract

The paper considered issues the development of a self-organizing controller (SC) based on a neuro-fuzzy network that can approximate a nonlinear function with arbitrary accuracy. The SC in the form of neuro-fuzzy networks, possesses the nonlinear property that allows for an increased range of control over the plant, which imparts adaptive properties to the control systems. To reduce the dimensionality of the plant, it is proposed to split the model of the system into sub models with smaller dimensionality, due to which the duration of training of the neuro-fuzzy network is reduced and asymptotic stability is ensured as a whole. The proposed approach is also applicable to multidimensional control systems of the nonlinear dynamic plants. The simulation results showed that the synthesized SC provides good tracking characteristics, the tracking efficiency is no more than 10%, which meets the requirement of the control system.