The research on vibration active control method based on nonlinear reduced-order functions and predictive control

This paper explores the application of predictive control methods for actively managing structural vibrations and proposes a model predictive controller based on low-dimensional nonlinear reduced-order functions. By training neural networks to identify and train reduced-order nonlinear functions, the feasibility of this approach in controlling complex structures has been verified. The results indicate that the neural network model accurately predicts dynamic responses of the system, while the designed controller exhibits outstanding performance in vibration control across multiple directions for tower cranes. The enhanced controller achieves faster computation by simplifying the predictive model through extended state-space modeling and neural network retraining. Nevertheless, it maintains satisfactory control effectiveness. This method offers new perspectives on reduced-order control of high-dimensional linear and nonlinear systems and shows potential for further enhancing the computational efficiency of model predictive controllers in future research.