Fragility surrogate models for coastal bridges in hurricane prone zones

•Reliable fluid–structure interaction model for coastal bridges under hurricane events.•Capturing the important responses and deck unseating failure mode.•Surrogate models for prediction of deck unseating failure mode under hurricane events.•Modified wave and surge load models to reduce the computational burden for reliability assessment.

Simply supported multi-span bridges, that comprise the majority of bridges along the US Gulf Coast, are susceptible to the deck unseating failure mode under extreme wave and surge conditions cause by hurricanes. This paper explores central issues for accurate and efficient vulnerability assessment of such bridges under hurricane loads, including the validity of alternative surrogate models for probabilistic performance prediction as well as the modeling uncertainty introduced by adopting simplified loading profiles in the vulnerability assessment. First, a fluid–structure interaction (FSI) model that is capable of capturing the unseating mode of failure is presented to evaluate bridge deck displacements for fragility assessment. Then, different statistical learning techniques are compared to develop surrogate models of bridge fragility using the FSI model in order to reduce the computational expense of developing fragility surfaces, or statements of failure probability conditioned on hazard intensity. Additionally, due to the intricacies in developing FSI models and their significant computational burden, the use of simplified wave load profiles are explored for probabilistic performance assessment. Specifically, a modified wave load model is presented that adjusts existing estimates of wave forces on bridge decks based on insights from the FSI model. Finally, surrogate models using the modified wave loads are compared with those based on FSI analysis to quantify any additional modeling uncertainty introduced and provide further guidance for fragility assessment of coastal bridges.