BPNN-predicted fault permanent dislocation and seismic response analyses of long-span suspension bridges crossing dip-slip faults

Fault dislocations may lead to permanent ground rupture displacements, causing significant damage to bridges spanning fault lines. Accurate prediction of permanent displacement due to fault dislocations is essential for assessing the seismic resilience of bridges across faults. This study introduces a highly precise BPNN (Backpropagation Neural Network)-based model for predicting permanent fault rupture displacements at the ground surface. The model produces artificial ground motion, known as 'fault-crossing ground motion', through the integration of low-frequency (LF) and high-frequency (HF) components. The research centers on a 538-m long-span steel truss suspension bridge, employing a comprehensive finite element (FE) model developed with ABAQUS software. The study methodically investigates the impact of permanent fault rupture displacement, fault-crossing effects, and angles on the seismic behavior of the suspension bridge. Research findings indicate that the BPNN-based approach precisely predicts permanent fault rupture displacements due to fault movements. The position of the fault line relative to the bridge, along with the fault-crossing angle and permanent fault rupture displacement, significantly influences the bridge's seismic responses. Notably, as permanent fault rupture displacements increase, the magnitudes of dynamic responses vary considerably. Longitudinally, the amplification of displacement, shear force, and moment in the towers is more pronounced than transversely. Furthermore, a gradual increase in shear force at the tower base poses a significant risk of shear failure. LF components predominantly determine the requirements for internal forces and displacements. This study presents a direct and dependable method for artificially synthesizing ground motion for the seismic analysis of fault-crossing suspension bridges, underscoring the influence of fault-crossing angles and locations on seismic responses.