RECURSIVE-IDENTIFICATION METHOD FOR THE HAMMERSTEIN MODEL - EXTENSIONTO NONLINEAR MISO SYSTEMS

A new method for identification of nonlinear systems using the Hammers tein model is presented, This is developed along the lines of the basi c Kalman filter and has an important property for on-line identificati on that parameters of the nonlinear static polynomial and those of the linear dynamic transfer function are estimated recursively and separa tely; thus, the number of coefficients to be estimated is minimal, and computational requirements are considerably reduced. This procedure i s extended to nonlinear multi-input single-output (MISO) systems. Two recursive methods are derived. The first one consists of estimating pa rameters of an equivalent realization of the nonlinear MISO representa tion. The second recursive algorithm estimates recursively and separat ely parameters of each submodel of the system, using a new formulation of the input-output difference equation. Both procedures are based on the algorithm developed in the nonlinear single-input single-output ( SISO) case. To illustrate the efficiency of these algorithms, numerica l examples are provided.