在工程中,大型设备和精密仪器在运行时,产生的振动通常会对周围环境造成影响;因此,研究抑制振动的隔振系统及控制方法具有重要意义.为了扩大隔振范围以隔振大型物体,设计了具有多个电磁隔振单元的并联电磁隔振系统,并提出了 一种融合Q学习的非线性模型预测控制(Nonlinear Model Predictive Con-trol,NMPC)方法实时调控该多隔振单元系统,以提高系统的隔振性能.基于电磁力、线圈电流和电磁铁间距三者的非线性关系建立了并联电磁隔振系统的动力学方程及状态方程,在此基础上设计了 NMPC控制器.其中,利用Q学习方法确定了预测范围,从而避免计算量过大或预测模型不准确的问题;同时,Q学习方法能够优化NMPC方法的目标函数中的权重矩阵V和R.仿真和实验结果表明,在所提出的融合Q学习的NMPC方法控制下的多隔振单元并联系统在外界扰动下,振动幅度显著减小,系统平稳性大大提高.
Control Study of Parallel System With Multiple Vibration Isolation Units Based on Nmpc and Q-Learning
Vibration generated by large equipment and precision instruments during operation usually affects the surrounding environment.It is vital to study vibration isolation systems and control methods to suppress vibration.To broaden the vibration isolation range for large-scale objects,this study designs a parallel electromagnetic vibration isolation system equipped with multiple units and proposes a real-time control strategy for the multi-unit system,integrating Q-learning with nonlinear model predictive control(NMPC)to enhance the system's vibration isolation performance.The dynamic and state equations of the system are established based on the nonlinear relation among electromagnetic force,coil current and the gap between electromagnets,leading to the design of the NMPC controller.The Q-learning method is employed to determine the prediction step length,thereby circumventing issues related to excessive computation or inaccurate prediction models.Simultaneously,the Q-learning method optimizes the weight matrices V and R in the objective function of the NMPC strategy.Simulation and experimental results indicate that the proposed NMPC method,which incorporates Q-learning,significantly diminishes the vibration amplitude and improves the stability performance of the parallel system with multiple vibration isolation units under external disturbances.
nonlinear model predictive controlq-learningactive controlvibration isolationparameter optimization
NETL
NSTL
万方数据
