Abstract
The identification of nonlinear dynamical systems still is a non-trivial procedure. Signal based methods that retrieve basic information about the system prior to detailed identification can provide valuable assistance in this task. In the paper the detection and characterization of nonlinearities as well as the estimation of the system order are discussed. Regarding the first topic, a new method is presented that is based on the Method of Internal Harmonics Cross-Correlation by Dimentberg and Sokolov but provides additional information about the nature of the nonlinearity and uses nonlinear instead of linear correlation. Concerning the second topic, a method is presented that, in contrast to most existing methods for nonlinear systems, incorporates a random input signal to promote the excitation of all system states. Both methods are illustrated with numerical examples.
Bayesian recognition of a moving object information-measuring system state: a priori information weight correction