Modelling and prediction of forced oscillation in two-dimensional non-linear system using Levenberg-Marquardt approximation based neural network

摘要

This paper focuses on the prediction of forced oscillation in a two dimensional non-linear self-oscillating system subjected to high frequency deterministic dither at one of the input using DF techniques. With these data, a neural network (NN) model is created using Levenberg-Marquardt approximation and the oscillation parameters at various dither magnitude have been determined. The comparison with the results of digital simulation as well as simulation results of MATLAB/SIMULINK illustrate the accuracy of DF technique, the neural network model and the potential benefits of incorporating NN as a part of control system architecture.