Long-term extreme buffeting response investigations for longspan bridges considering uncertain turbulence parameters based on field measurements

Abstract

When bridges become increasingly longer, the dynamic behavior from wind and waves can dominate the structural load effects. To build better knowledge about the dynamic behavior of long-span bridges subjected to turbulent wind in complex terrain, the Norwegian University of Science and Technology has instrumented the Hardanger Bridge with a full-scale measurement system. The Hardanger Bridge measurement program, along with many full-scale measurement efforts on long-span bridges around the world, has shown a large scatter in the measured buffeting response when plotted against the mean wind velocity.

The wind field information defining the basis for design of such bridges, is fundamental to achieve a reliable extreme load definition. In this thesis, the traditional methods for wind field measurements are investigated by comparing the in-situ mast measurements and terrain model wind tunnel tests performed before the design of the Hardanger Bridge with wind measurements along the span of the current bridge.

The full long-term method is recognized as the most accurate way to predict the extreme response of a structure subjected to stochastic dynamic loads. However, in current design practice for buffeting action of long-span bridges this has not yet become the standard way to estimate the extreme responses. In the work presented herein, the long-term extreme buffeting response of the Hardanger Bridge is investigated, considering the turbulence variability effects and the short-term extreme response uncertainty. The findings show that both the turbulence variability and the short-term extreme response uncertainty is very important for the design response of the Hardanger Bridge girder.

Finally, a new algorithm based on Gaussian process regression for long-term extreme response calculations is proposed, exhibiting several attractive qualities in terms of accuracy and computational efficiency.