Identification of polyperiodic nonlinear systems

Abstract

A method for identifying the Fourier-Volterra kernels of polyperiodic nonlinear systems by crosscorrelating the system output with prescribed nonlinear transformations of a prescribed random input is introduced. A frequency-domain counterpart of this method, which can utilize a computationally efficient FFT algorithm to compute cross-spectra of the system output and prescribed nonlinear transformations of the input, is also introduced. With complex inputs, the methods apply to infinite-order systems with infinite memory. For real inputs, they apply only to finite-order systems.