Data-driven model identification and control of the quasi-zero-stiffness system

Jiyuan Jiang,Jie Tang,Ke Sun,Huatao Chen,Yinghui Li,Dengqing Cao
DOI: https://doi.org/10.1007/s11071-024-09811-5
IF: 5.741
2024-06-07
Nonlinear Dynamics
Abstract:Nonlinear vibraion isolation technique is widely employed for vibration suppression. An identification-control integrated method based on data-driven approaches is proposed for solving the optimal control law of a nonlinear time-continuous dynamic system. A dynamic surrogate model of the quasi-zero-stiffness (QZS) vibration isolation system is established by using an identification algorithm combined physical information neural network and Runge–Kutta method with the input and output signals of the original model. Two approximate optimal controllers are trained through the particle swarm optimization with 'loser-out' skill and the twin delayed deep deterministic policy gradient (TD3) in the sense of a self-defined objective function, where controllers communicate with the dynamic surrogate model during the training process. Then, the comprehensive performance in the condition of variable load and Gaussian noise excitation, and the displacement transmissibility are tested on the original model. The results show that the identified surrogate model can accurately reproduce the dynamic characteristics of the original model and the trained controllers are able to accomplish the control tasks successfully with a certain adaptability, further enhancing the low-frequency vibration isolation of the QZS isolator.
engineering, mechanical,mechanics
What problem does this paper attempt to address?