A Multi-Hazard Safety Evaluation Framework for a Submerged Bridge using Machine Learning Model

Abstract

This study proposes a submerged bridge safety evaluation process against seismic and flood hazards. Due to the uncertainties in the scours, seismic hazard, and structural performance for a given seismic excitation are inevitable, reliability analysis is adopted. A machine-learning based scour risk curve, which is established by the multivariate adaptive regression splines (MARS) incorporated with firefly algorithm (FA), is built to reflect the flood hazard. The seismic hazard is measured using a code-based probabilistic seismic hazard curve. A series of nonlinear time-history analyses are performed to determine the structural performance under different peak-ground-acceleration values. Displacement ductility is used to measure the bridge performance under attacks of both hazards. The influence of the immersed water depth on a bridges performance is investigated. A case study, in which the nonlinear behaviors in concrete (including core and cover areas), steel bar and soil are included in a bridge model, is conducted to illustrate the proposed methodology and the structural performances with added mass are investigated to show the submerged water effect. According to the results obtained, highly variability of seismic performances is observed and it is important to include the immersed water depth to capture the seismic capacity of a bridge if the submerged bridge depth is great.