Estimation of an N-L-N Hammerstein-Wiener model (Q1614436)

Parametric identification of a noisy nonlinear SISO Hammerstein-Wiener dynamical model is considered. A relaxation recursive identification scheme to determine approximate model parameters is proposed. The nonlinear static characteristics are parameterized by using cubic splines, and the linear dynamics is represented by a high-order ARX model. This yields the identification error bilinear-in-parameters. The convergence of the scheme to the best model parameters is examined. Simulation examples illustrate the effectiveness of the method.