Acoustic characteristic optimal design for railway steel–concrete composite bridge based on the RBFNN-NSGA-II algorithm
Structure-borne noise from railway steel–concrete composite (SCC) bridges presents a persistent challenge due to high sound pressure levels across both low and high frequency ranges. This study establishes a hybrid finite element–statistical energy analysis (FE-SEA) numerical model to predict acoustic radiation from an SCC bridge. Field measurements validate the model with discrepancies of only 0.4 dB and 1.1 dB in overall sound pressure levels. Using uniform design sampling, a high-accuracy radial basis function neural network (RBFNN) surrogate is trained to map cross-sectional parameters to acoustic and cost objectives. The non-dominated sorting genetic algorithm (NSGA-II) then performs multi-objective constrained optimization, generating a Pareto frontier for sound power level (SWL) and material cost. The technique for order preference by similarity to an ideal solution (TOPSIS) selects the optimal parameter combination, achieving a 5 dB reduction in SWL and a 23.9% decrease in material cost. These results demonstrate that strategic cross-sectional adjustments can simultaneously mitigate noise and reduce expenditure, offering a practical framework for acoustic optimization in railway bridge design.