Multi-objective optimization and multicriteria decision-making for vibro-acoustic design of long-span steel-concrete composite bridges

To mitigate the structure-borne noise of long-span steel-concrete composite bridges (LSCBs) while maintaining structural economy, this study proposes an integrated vibro-acoustic optimization framework. First, a hybrid FE-SEA model is established and validated against field measurements, demonstrating high accuracy with overall sound pressure levels (OSPLs) deviations limited to only 0.7–1.2 dB. To alleviate the computational burden associated with repeated numerical simulations, a highly accurate back-propagation neural network (BPNN) surrogate model (R2 = 0.99) is trained using FE-SEA-generated samples, accelerating the structure-borne noise prediction process by a factor of 23,375. Subsequently, the surrogate model is coupled with the non-dominated sorting genetic algorithm II (NSGA-II) to optimize OSPL, cost, and mass simultaneously. Ultimately, to select optimal solutions from this Pareto set, a dynamic decision-making framework is established to facilitate the flexible selection under conflicting objectives. This framework enables rapid, flexible multi-objective optimization, providing practical guidance for LSCB design under varying engineering requirements.