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

Abstract In recent years, the issue of structure-borne noise generated by steel–concrete composite (SCC) bridges has become increasingly severe. To control this noise by adjusting the cross section parameters of SCC bridges, this study first established a numerical model based on the hybrid finite element–statistical energy analysis (FE-SEA) method. The overall sound pressure levels calculated by numerical model are compared with field measurements, showing discrepancies of 0.4 dB and 1.1 dB, respectively. The comparison confirms the accuracy of the numerical model. Then, a high-accuracy radial basis function neural network (RBFNN) was trained using samples generated from the numerical model with uniform design. To achieve greater noise reduction with lower costs, the non-dominated sorting genetic algorithm (NSGA-II) was used for multi-objective constrained optimization, resulting in the Pareto frontier for sound power levels (SWLs) and material cost. Finally, the solution set was evaluated using the technique for order preference by similarity to an ideal solution method, and the optimal combination of cross sectional parameters was obtained. This combination resulted in a 5 dB reduction in the SWL of the structure and a 23.9% reduction in material cost.