A study of damping errors in correlation-driven stochastic realizations using short data sets

Abstract

A numerical study, motivated by applications in structural engineering, is conducted to investigate the effects of using short data sets for the identification of system damping in correlation-driven stochastic realizations. Sets of single-degree-of-freedom systems are excited with white noise and the eigensystem realization algorithm, with the aid of the modal confidence factor, is employed for the identification scheme. The study reveals information regarding the effects of sample size and Hankel matrix dimension on the resulting estimates of system damping. Results illustrate that significant bias is associated with using short data sets in a correlation-driven framework; in particular errors are seen to increase for high frequency, high-damping systems. Some of this error can be attributed to the inclusion of problematic regions of the correlation function in addition to leakage from the correlation estimation.