Nonlinear Identification Of Dual-rate Systems Based On Hammerstein Model

For a nonlinear dual-rate system with fast input and slow output data, which means the output sampling time is longer than the regular input sampling time and the input contains nonlinear property ,we develop two methods to identify the fast single-rate model, one of which is least-squares methods based on polynomial transformation technique and the other is output error methods based on Gauss-Newton algorithm. The output error method is also applied to develop a composition observer for an industrial Propylene column. The main contributions could be summarized as follows:1) For a dual-rate system with nonlinear Hammerstein model, we firstly use the polynomial transformation technique to access a high-order model and then apply least-squares methods to estimate the parameters of the lifted model.2) An output error method based on Gauss-Newton method is proposed to solve the problem whose output sample time is very long. This method reduces the number of parameters to be identified and improve the accuracy of parameter estimation.3) Recursive identification algorithm of dual-rate system based on Gauss-Newton method and recursive identification algorithm based on polynomial transformation technique are proposed and demonstrated with detailed examples.4) Those identification methods are also applied in an industrial Propylene column to build a composition observer .