Block-oriented identification of Hammerstein/Wiener-models using the RLS-algorithm

Abstract

A new method for identifying Hammerstein/Wiener models is proposed, where the non-linear characteristic is modelled by a piecewise linear function and the linear dynamic part is represented by a linear transfer function. The parameters of the overall model are estimated by recursive least-squares (RLS), therefore enabling the algorithm to be implemented on-line. Using appropriate validity functions the edges of the piecewise linear characteristic may be smoothed out in order to achieve a better fit. An adaptive algorithm for optimal partitioning of the local linear models is also proposed on the basis of statistical analysis. Using the new approach an efficient and accurate model can be identified.