Identification of Parametric (NARMAX) Models from Estimated Volterra Kernels
A method for Nonlinear Auto-Regressive Moving-Average with exogenous input (NARMAX) model identification is proposed that is based on Volterra kernel estimates obtained from input-output data. The method relies on the fundamental relations between Volterra kernels and the parameters of NARMAX models, which are derived using “generalized harmonic balance”. The method identifies different order terms of the NARMAX model separately (up to 3rd-order), allowing easier determination of the structure of the NARMAX model and yielding better parameter estimates. Simulation results are compared with the prediction-error stepwise-regression estimation algorithm introduced by Billings and Voon (1986) and show that the proposed method yields more accurate parameter estimates in the presence of noise.
KeywordsVolterra Model eXogenous Input Volterra Kernel Accurate Parameter Estimate Good Parameter Estimate
Unable to display preview. Download preview PDF.
- 2.Korenberg, M.J. (1987) Functional expansions, parallel cascades and nonlinear difference equations. In: Advanced Methods of Physiological System Modeling, Volume I, V.Z. Marmarelis (ed.), Biomedical Simulations Resource, University of Southern California, Los Angeles, California, pp. 221–240.Google Scholar