IDENTIFICATION AND CONTROL OF DYNAMICAL SYSTEMS USING NEURAL NETWORKS

Abstract

A number of studies have been addressed issue related with identification and control of dynamical systems. In spite of such attempts, constructive procedures similar to those available for linear systems are not available for non-linear systems. Hence the identification and control of non-linear systems is a formidable problem. The properties of neural networks, like flexibility, learning ability and ability to approximate any non-linear function, motivate us for their use in this field. In this dissertation identification and control of dynamical systems using multi-layer feedforward neural network are discussed. Plant is simulated using backpropagation algorithm. Identification and control is explored using MATLAB neural network toolbox on computer. Simulation results for identification indicate that the trained neural network behaviour approximates the behaviour of the actual plant. Simulation results for control indicate that when generated control is applied to the plant, the output of the plant follows the reference model output.