Hammerstein model identification by multilayer feedforward neural networks

A new method for the identification of the nonlinear Hammerstein model, consisting of a static linearity in cascade with a linear dynamic part, is introduced. The static nonlinearity is modelled by a multilayer feedforward neural network (MFNN) and the linear part is modelled by an autoregressive moving average (ARMA) model. A recursive algorithm is developed to update the weights of the MFNN and the parameters of the ARMA. The new method makes use of the well-known nonlinear mapping ability of MFNN and avoids the restrictive assumptions of the previous identification methods. Two numerical examples are presented to illustrate the performance of the developed model and recursive algorithm.