The Effect of Acceleration Signal Length on the Outputs from Modal Identification Methods

Abstract

The damping ratio is a significant factor influencing the dynamic behaviour of a structure. The safety, serviceability and habitability of a structure are all impacted by the damping ratio. Damping ratio does not relate to a unique physical phenomenon like mass or stiffness and in practice, design analysis relies on estimates of damping ratios from empirical measurements of similar structures. Overestimates of the damping ratio arising from uncertainty in estimates can lead to structures experiencing acceleration responses during wind and seismic events that could potentially cause human discomfort. Full-scale testing provides the most accurate insight into the actual damping ratio of a structure. Where full scale testing is performed, acceleration signals are recorded and assessed using modal identification techniques to identify the characteristic modal parameters such as natural frequency and damping ratio. However, this form of testing is not without errors which may arise as a result of response conditions during monitoring, the modal identification methods applied, the duration of acceleration signal processed and the sampling frequency of the acceleration signal.This paper considers real ambient acceleration response signals of different lengths and sampling frequencies obtained from a full-scale monitoring campaign. The influences of signal length and sampling frequency on the natural frequencies and the damping ratios calculated using two different modal identification methods are investigated. Outputs from signal lengths of 12 hours, 1 hour, and 10 minutes are compared as well as sampling frequencies of 20Hz, 10 Hz and 5Hz. The two modal identification methods used are the Bayesian Fast Fourier Transform (BFFT) and a composite method of Analytical Mode Decomposition (AMD) and the Random Decrement Technique (RDT) in which bootstrapping is also performed to identify error in the estimates.