Identification of Multichannel Nonlinear Systems Excited by Realisations of Mutivariate Orthogonal Multisine Random Time-Series

Abstract

In the paper nonlinear transformations of multivariate orthogonal multisine random time-series are discussed. A focus on transformations that preserves orthogonality of transformed multivariate orthogonal multisine random time-series elements is given and the following from this property decomposition of multichannel nonlinear dynamic system identification problem into separate single-channel nonlinear dynamic system identification problems is described. This decomposition is illustrated by simulation examples devoted to identification of two-input single-output nonlinear dynamic systems based on observed mixtures of single-channel nonlinear block-oriented dynamic system outputs and the corresponding samples of the used bivariate orthogonal white multisine random excitations.