Semi-active vibration control of stay cables using neural networks

This paper describes a study on semi-active vibration control of stay cables by using electro/magneto-rheological (ER/MR) dampers and adopting neural network control technique. An improved neuro-controller, which is derived without need of a reduced-order system model, is first developed for implementing the semi-active control. The neuro-controller is devised following three steps. In step 1, based on a multi-degree-of-freedom finite element model of the cable, an LQG controller is constructed to obtain the optimal feedback fully active control force by assuming complete state observation and solving algebraic Riccati equation. In step 2, a neural network is designed and trained to emulate the performance of the LQG controller. The trained neural network is still a fully active controller but only a few response states are included as network inputs in the training process to simulate incomplete state observation. In this way, the need of system model reduction and an extra state estimator is eliminated. In step 3, the fully active network controller is clipped to achieve the voltage value required to semi-actively control the cable vibration through ER/MR dampers. Both the velocity orientation clipping and the maximum voltage clipping are introduced. After completing the design of the neuro-controller, a numerical example of a 12m-long stay cable specimen connected with a small-size ER damper is provided to verify the control effectiveness of the proposed strategy.