Identification and Optimal Control of Nonlinear Systems using Sinusoid Neural Network.

Abstract

A neural network can be applied for identification and optimal control of nonlinear systems. However, much information is necessary for identification because of lack of the learning ability of the neural network with the sigmoid function as the output function of the neuron. In this paper, the neural network with the sinusoidal function as the output function of the neuron is proposed. This neural network approximates the nonlinear function well. Using this neural network, good results of identification of 1-degree-of-freedom Duffing-type system are obtained with less information. and optimal control of this system by means of dynamic vibration absorber is shown to be possible through computer simulation.