Identification of Multivariable Hammerstein CARMA System Using Special Test Signals

Abstract

Abstract The multivariable Hammerstein controlled autoregressive moving average (CARMA) system contains the sum of bilinear parameter vectors in the identification model, which is difficult to be transformed into a regression form for direct identification. A two-stage identification technique is developed in this article. By using multiple sets of special test signals, this chapter achieves separable identification for the linear subsystem and the nonlinear subsystem. Then, using the property of binary signal input to nonlinear part, parameters of noise model and the linear subsystem can be calculated by recursive extended least squares (RELS) method. To implement the off-line identification, the RELS is used to identify the nonlinear subsystem parameters. Finally, simulation examples illustrate the validity of proposed approach in identifying multivariable Hammerstein CARMA system.