Combined Parametric-Nonparametric Identification of Block-Oriented Systems

Hammerstein system -- Wiener system -- Wiener-Hammerstein (sandwich) system -- Large-scale interconnected systems -- Structure detection and model order selection -- Time-varying systems -- Simulation studies -- Summary.;This book considers a problem of block-oriented nonlinear dynamic system identification in the presence of random disturbances. This class of systems includes various interconnections of linear dynamic blocks and static nonlinear elements, e.g., Hammerstein system, Wiener system, Wiener-Hammerstein ("sandwich") system and additive NARMAX systems with feedback. Interconnecting signals are not accessible for measurement. The combined parametric-nonparametric algorithms, proposed in the book, can be selected dependently on the prior knowledge of the system and signals. Most of them are based on the decomposition of the complex system identification task into simpler local sub-problems by using non-parametric (kernel or orthogonal) regression estimation. In the parametric stage, the generalized least squares or the instrumental variables technique is commonly applied to cope with correlated excitations. Limit properties of the algorithms have been shown analytically and illustrated in simple experiments. Front Matter....Pages 1-13 Introduction....Pages 1-21 Hammerstein System....Pages 23-85 Wiener System....Pages 87-102 Wiener-Hammerstein (Sandwich) System....Pages 103-111 Additive NARMAX System....Pages 113-136 Large-Scale Interconnected Systems....Pages 137-147 Structure Detection and Model Order Selection....Pages 149-169 Time-Varying Systems....Pages 171-180 Simulation Studies....Pages 181-187 Summary....Pages 189-191 Back Matter....Pages 193-238