Damping estimation of large wind-sensitive structures

Abstract The adequacy of two identification techniques to accurately estimate the modal damping ratios (MDRs) from ambient vibration records is assessed for long-span suspension bridges with eigen frequencies close to or below 0.1 Hz. The first method is an automated covariance driven stochastic subspace identification (SSI-COV) algorithm and the second one is an automated Frequency domain decomposition algorithm (AFDD). The bias and the dispersion of the identified MDRs are first assessed using simulated bridge vibration records and then investigated using full-scale acceleration data obtained on the Lysefjord Bridge, Norway. The simulated data showed that record durations of 60 min provided up to three times more accurate estimates of the MDRs than if the more common record duration of 10 min is used. For the full-scale acceleration records and a duration of 60 min, the MDR estimates showed a relatively large dispersion with a non-Gaussian distribution, suggesting that the median value of the MDR may be more reliable than the arithmetic mean. The AFDD algorithm was observed to estimate the MDRs with a larger bias than the SSI-COV method. This suggests that the frequency-domain based approach is not well suited for the modal parameters identification of long suspension bridges with eigen-frequencies around and below 0.1 Hz.