Estimates of an Upper Limit of the Number of Parameters in Nonlinear Model Structures

Paper in proceeding, 2009

An approach to estimate the plausible number of parameters in a nonlinear model in an identification problem is considered. By suggesting a series of experiments using periodic input signal, estimates of the disturbance variance and the nonlinear distortion can be formulated. With these estimates an expression for a reasonable number of parameters is obtained. This is useful help when a user has to choose between different types of nonlinear model structure which differ largely in their number of parameters.