ABSTRACT
This paper investigates the accuracy of the Hilbert-Huang transform (HHT) in capturing the time-varying frequencies of a short-span bridge traversed by a vehicle travelling at a constant speed. The bridge and vehicle are modelled as a simply supported beam and a quarter-car, respectively. The HHT uses empirical mode decomposition to divide the original signal into mono-component signals, called intrinsic mode functions (IMFs), where the Hilbert transform can extract instantaneous frequencies (IFs). Each IMF is associated with a dominant frequency band, although mode mixing is possible. In order to improve the predicted frequencies, several filters are applied before and after performing the HHT with a threefold purpose: (i) to remove the static component, (ii) to isolate the first mode of vibration, and (iii) to obtain meaningful and denoised IFs. The influences of a localized stiffness loss in the bridge, different vehicle speeds, and three signal-to-noise ratios on the results are discussed.
