Intelligent vibration reduction algorithm of cable based on deep reinforcement learning

Here, to reduce dependence of vibration control algorithm on dynamic model accuracy oF cable-MR damper system, a model-free vibration reduction algorithm based on deep reinforcement learning theory was proposed. In this method, the interaction between control module and environment was used to realize adaptive semi-active control of cable vibration, and the voltage applied by the damper was adjusted online according to response states of cable specific points to reduce feedback control requirements. To verify the effectiveness of the proposed intelligent control algorithm, Galerkin method was used to establish cable-MK damper environment model, and control effects of viscous damper multimodal control, bang-bang control and deep reinforcement learning control on cable's wind-induced vibration were analyzed contrastively by taking cable vibration reduction design of a real bridge as an example. The results showed that under random wind load, the deep reinforcement learning control algorithm can not only realize model-free vibration control of cable, but also its control effect can be better than those of viscous damper's multi-modal control and MR damper's bang-bang control. © 2022 Chinese Vibration Engineering Society. All rights reserved.