PARAMETER ESTIMATION OF NON-LINEAR DYNAMICAL SYSTEMS.
A computational scheme is proposed for modeling nonlinear dynamical systems using a polynomial form representation. The model output is expressed as a sum of each state raised to negative or positive integer powers. Subsequently, based solely upon the input-output data, the coefficient of each term is estimated in order to establish system structure. The procedure is illustrated by two examples using analytically generated data in one case and experimentally measured test data in the other case.