Structure-borne noise of a steel box railway bridge: Unified FEM–BEM prediction, field validation, and the speed-dependent transition of acoustic radiation patterns

The rapid expansion of railway systems has intensified concerns regarding the structure-borne noise (SBN) of bridges, which is governed by the coupled train–track–bridge interaction. Conventional prediction approaches typically rely on two or more separate software platforms, in most of which the structural response is computed in a finite-element package and then exported into an independent acoustic solver, introducing a data-mapping error. This study developed a unified single-platform FEM–BEM framework, in which the structural and acoustic domains are solved within the same explicit transient environment, avoiding intermediate mesh remapping and preserving the time-resolved excitation associated with moving wheel–rail loads. The framework was validated against steel-plate vibration and sound radiation experiments in an anechoic chamber, yielding a mesh design guideline of 12 elements per bending wavelength, with relative errors below 1.7% in the dominant structural frequencies and below 6% in the corresponding acoustic eigenfrequencies. A full-scale model of a 180-m steel box-girder bridge was then developed, and its predictions matched the field-measured dominant acoustic peaks at 50–63 Hz to within approximately 2 dB. This explicit, time-resolved formulation allowed a parametric analysis showing that the spatial distribution of the radiated sound field evolves with train speed, shifting from a pattern concentrated near the girder ends at low speeds toward a more spatially extended pattern at high speeds. These results demonstrate that local vibration peaks do not correspond to the most severe far-field noise impacts, underscoring the need to account for the speed-dependent excitation mechanism and distance-attenuation characteristics—rather than vibration amplitude alone—when predicting and mitigating SBN in railway bridge engineering.