Bayesian optimization of tuned mass damper for vibration control of the suspension footbridge with human–vehicle–bridge interaction

Abstract

Tuned mass damper (TMD) is a crucial device for mitigating bridge vibrations induced by dynamic loads. This study presented an analytical solution for the dynamic response of a bridge equipped with TMDs under human–vehicle–bridge interaction (HVBI). Then, a finite element (FE) model and a Bayesian optimization (BO) framework were developed to enhance the efficiency of the vibration control. Following this, comprehensive analyses, including time history and frequency domain assessments, were conducted to evaluate the system’s performance. Finally, the effectiveness and robustness of the optimized TMD were compared to those of the unoptimized version. The results indicate that during the BO process, the optimized results of the objective function tend to stabilize after 300 trials. Considering calculation cost and optimization effectiveness, setting the optimization rate at 15%∼17% is more appropriate. The optimized TMD exhibited a vibration reduction rate of 15.14% greater than that of the unoptimized one. Additionally, the vibration reduction rates of the optimized TMD case across step frequencies consistently exceeded those of the unoptimized one. Optimizing the TMD parameters using BO improves the energy dissipation rate across different frequencies and positions of the bridge.