A TIME AND FREQUENCY DOMAIN PROCEDURE FOR IDENTIFICATION OF STRUCTURAL DYNAMIC MODELS

摘要

A system identification procedure is presented which uses a time domain least squares curve fit to extract system poles and a frequency domain curve fit to extract system modal vectors. The time domain pole extraction relies on an asymptotically large Hankel matrix to converge the modal frequencies and damping ratios. The modal vectors are then extracted using the converged system poles in a linear least squares curve fit in the frequency domain. An important component of the frequency domain curve fit is that it includes an improper term as an unknown of the curve fit. This compensates for the fact that most modal vibration data is significantly influenced by modes outside the bandwidth of the data collection procedure. The resulting minimal order model achieves excellent time and frequency domain accuracy without the need for a nonlinear regression. An application to experimental test data is presented to demonstrate the accuracy of the method.