Rational Model Data Smoothers and Identification Algorithms.

This study presents a new algorithm for nonlinear rational model identification. The new algorithm consists of a two-step procedure: a nonlinear rational function smoother is initially designed and used to smooth the data, system identification is then performed based on the smoothed signal. By using the smoothed signal instead of the raw data, the severe noise problems, which arise in the rational model identification, are avoided. The new approach significantly simplifies the procedure for dynamic nonlinear rational model identification, compared with earlier estimators and provides unbiased estimates with the same degree of accuracy.