Identification of fractional chaotic system parameters

Abstract

In this work, a technique is introduced for parameter identification of fractional order chaotic systems. Features are extracted, from chaotic system outputs obtained for different system parameters, using discrete Fourier transform (DFT), power spectral density (PSD), and wavelets transform (WT). Artificial neural networks (ANN) are then trained on these features to predict the fractional chaotic system parameters. A fractional chaotic oscillator model is used through this work to demonstrate the developed technique. Numerical results show that recurrent Jordan–Elman neural networks with features obtained by the PSD estimate via Welch functions give adequate identification accuracy compared to other techniques.