Dynamic System Identification Based on Two-Dimensional Autoregressive Model Fitting

Abstract

This report deals with dynamic system identification which is based on two-dimensional autoregressive (AR) model fitting. The method proposed basically utilizes mapping correspondence between modal parameters shown in a z-transformed plane and those in a Laplacian (s-domain) plane. In order to reduce computer consuming time, the EFFT algorithm by Gan is introduced, and satisfactory time reduction can be obtained. As a practical application of the method, the time-dependent behavior of vibration parameter for the robot arm structure is identified and represented by trajectories in the s-characteristic plane.