AI-Driven Active Tendon Control Systems for Real-Time Vibration Mitigation in Long-Span Steel Bridges

Long-span steel bridges are susceptible to dynamic vibrations induced by wind, traffic, and seismic activities, which compromise structural integrity and serviceability. Traditional passive and semi-active control systems often lack the adaptability required for real-time mitigation of these complex vibrations. This study proposes an AI-driven active tendon control system, leveraging machine learning (ML) algorithms to optimize tendon forces dynamically, thereby enhancing vibration suppression. A high-fidelity finite element (FE) model of a representative long-span steel bridge is developed using ANSYS. The model incorporates time-dependent loadings, material nonlinearities, and boundary conditions to simulate real-world vibration scenarios. A deep reinforcement learning (DRL) framework is integrated with the FE model to train a control agent, enabling real-time adjustment of tendon actuation forces based on vibration feedback. The agent’s performance is validated through numerical simulations under diverse excitation conditions, including turbulent wind fields and stochastic traffic loads. Results demonstrate a 40% reduction in peak deck accelerations and displacements compared to conventional passive systems, alongside improved computational efficiency for real-time deployment. The system’s robustness is further verified under seismic excitations, showcasing adaptive resonance avoidance. This research advances the integration of AI in civil infrastructure, offering a scalable solution for vibration control in long-span bridges.