Acoustic characteristic optimal design for railway steel–concrete composite bridge based on the RBFNN-NSGA-II algorithm

• • The hybrid FE-SEA model predicts overall sound pressure levels with discrepancies of only 0.4 dB and 1.1 dB compared to field measurements, establishing a reliable baseline for acoustic optimization without costly physical prototyping. • • The RBFNN surrogate achieves high accuracy in mapping cross-sectional parameters to acoustic and cost responses, enabling rapid evaluation of thousands of design candidates within the NSGA-II optimization loop. • • The optimized cross-sectional configuration reduces the structure-borne sound power level by 5 dB, a substantial attenuation that corresponds to a ~68% reduction in acoustic energy, directly addressing community noise complaints along railway corridors. • • Material cost decreases by 23.9% relative to the initial design, demonstrating that acoustic improvement does not necessitate increased expenditure; the Pareto frontier reveals trade-offs that can be tuned to project-specific budget constraints.