Data-Driven Time-Varying Eigensystem Realization Algorithm With Data-Correlation

Different data-driven system identification methods and algorithms in the field of structures have been developed, analyzed, and tested for modal parameter identification but inherent nonlinearities and noise in the data can drastically limit the application of such methods in order to adequately describe the real system behavior. While classical state-space realization techniques are, in essence, a least-squares fit to the pulse response measurements, introducing output auto-correlation and cross-correlations over a defined number of lag values has the potential to temper the effect of noise. This paper introduces a data-correlation approach to the time-varying eigensystem realization algorithm (TVERA/DC). As motivational cases to support this new method, the identification of the vibrational characteristics of a space structure is considered as well as the dynamical identification of a mechanical system with time-varying angular velocity.