Intelligent vibration control of tensile cable based on deep reinforcement learning

Vibration of tensile cables commonly occurs in the engineering structures such as cable-stayed bridges, which may have negative effect on the driving comfort and safety, and further lead to the fatigue problems of the structure. This paper proposes a semi-active control strategy based on deep reinforcement learning and magneto-rheological (MR) damper, for the vibration control of tensile cables, and the corresponding algorithm have been developed. Through the interaction with the cable-damper system, the intelligent agent can evaluate every possible control parameter and achieve an effective control policy. Then, according to the real-time vibration state, the agent would determine an action current for the MR damper and thus change the damping coefficient of the damper, further make influences on the vibrating cable. Basically, the proposed strategy realizes the model-free semi-active control of cable vibrations. To validate the effectiveness of the proposed semi-active control strategy, a scale model test has been conducted, where the test cases of passive and semi-active control strategies are carried out and compared. Results show that, the semi-active control shows a prior performance in vibration reduction compared to the passive control strategy, with regard to the vibration profile, the vibration energy, as well as the energy dissipation of MR damper.