Field Measurements and Time-Domain Buffeting Analysis of a Long-Span Suspension Bridge in a Mountainous Area

Stochastic responses of a long-span suspension bridge located in a mountainous area are obtained through long-term monitoring. Measured wind data are analyzed, including mean wind speeds, wind directions, wind angles of attack, turbulence intensities, and wind spectra. The results show mean wind directions are distributed in a region largely perpendicular to the bridge axis; turbulence intensities and wind angles of attack decrease rapidly as the mean wind velocity increases. Observed wind properties show obvious discrepancies as compared with those recommended by the Chinese wind-resistant design specification for highway bridges (Chinese specification for short), in terms of wind spectra, wind angles of attack, and turbulence intensity as well. The monitored structural responses, resulting from a combination of wind and vehicle actions, are decomposed into samples with respect to mean wind speeds. While samples are poorly correlated due to vehicle-induced responses, the bottom–edge samples, of which the vehicle-induced effects are minimized statistically, show obvious correlation with the wind speeds. A finite-element model of the full bridge is established, and time-domain buffeting simulations are performed according to wind field parameters advised by the Chinese specification. The results show that, while sharing the same trend with respect to the mean wind speed, the calculated values are generally 22.1–71.5% less than those observed. Observed data indicate that, as the wind yaw angles increase, the trend of the bottom-edge samples becomes progressively less relevant to the mean wind speeds and less correlated to the numerical simulations. Multiple factors potentially responsible for discrepancies between observed and numerical simulations are analyzed. It is shown that the gap between observed and simulated is reduced significantly when the turbulence properties adopted in simulations are adjusted according to observations in the field. On the other hand, while spectra of observed and simulated signals share the same group of resonant frequencies, they differ significantly in spectral values corresponding to dominant frequencies, which reveals the fact that damping properties of structural modes taking part in vibrations might be far more complicated than those adopted in numerical simulations.