Effects of initial conditions in operational modal analysis

In this study, the effects of high-amplitude initial conditions on the accuracy of modal parameters, identified from output-only vibration data, are investigated. The influence on the sample output correlation function, which is the basis of most time-domain operational modal analysis techniques, is analyzed first. Then, a numerical simulation is performed to quantify the effect of nonzero initial conditions on the relative accuracy of natural frequencies, damping ratios, and mode shapes. It is shown that, when all identification assumptions are satisfied, high-amplitude initial conditions can significantly reduce the estimation errors, especially for short data records. Finally, a full-scale application is presented where the modal parameters of a six-span high-speed railway bridge are determined from output-only data. The results obtained with two different data sets are compared: The first one consists of the bridge's response to ambient data only, whereas the second one also contains the free vibration recorded immediately after a train passage. Although for most modes the results are similar, it is possible to identify some additional bending and torsion modes from the free vibration data with good accuracy. Copyright © 2013 John Wiley & Sons, Ltd.