Longitudinal Seismic Mitigation and Response Asymmetry of a High-Pier Long-Span Stiff-Skeleton Arch Bridge with Fluid Viscous Dampers

Abstract

The longitudinal seismic response of a high-pier long-span stiff-skeleton arch bridge is spatially asymmetric because unequal pier heights, the fixed–movable bearing arrangement, and arch–pier interaction create nonuniform force transfer paths, even when the installed fluid viscous dampers (FVDs) obey a symmetric velocity-dependent law. This study evaluates that response redistribution and the mitigation achieved by longitudinal FVDs under near-fault motions. A three-dimensional SAP2000 model was established using elastic beam elements for the girder, arch ribs, cap beams, and piers, Plastic (Wen) links for spherical steel damping bearings, foundation springs for pile–soil interaction, and Maxwell-type FVD links. Thirty combinations of damping coefficient and velocity exponent were first screened under an El Centro record scaled to 0.64 g. The selected case (α = 0.3 and C = 2000 kN·s/mα) was then evaluated by paired analyses of models with and without FVDs under 15 records grouped descriptively as short-, medium-, and long-period pulse-like motions and non-pulse motions. Across the paired record set, the mean record-wise reductions were 21.34% for bridge-wide maximum bearing displacement, 27.56% for P2 pier-top displacement, 9.73% for bridge-wide maximum pier-base shear force, and 12.62% for bridge-wide maximum pier-base bending moment. Mean arch rib reductions ranged from 10.00% for axial force to 25.00% for bending moment. Seven of the eight monitored response metrics decreased under all 15 records; arch rib axial force decreased under 14 records and was unchanged under 1. Local force increases nevertheless occurred at several arch-supported piers in the spatial El Centro comparison, demonstrating that global mitigation does not imply spatially symmetric or uniformly beneficial component response. Within the selected record set, the medium-period group produced the largest average demands for several response measures, but this is a sample-specific observation rather than a resonance inference. Removing the extracted velocity pulse component reduced most responses, although the nonlinear original-versus-residual comparison cannot be interpreted as an additive pulse contribution. The paired record set check supports the robustness of the selected FVD case for the investigated sample, but does not establish record-independent optimality.