基于神经网络的拉索主动控制力的实现

拉索容易在外部激励下发生大幅振动,给拉索提供轴向控制力进行拉索主动控制是一种新的方法.本文拉索主动控制策略采用经典二次型线性最优控制,由于LQR控制的关键是根据状态矩阵即时求出Riccati方程,但是Riccati方程是一个矩阵非线性方程,其阶数高,变量间又相互耦合,求解十分困难,为了减小控制力输出滞后,更快的求解出控制力,更好的应用于拉索实时控制,本文基于神经网络具有很强的学习能力和非线性逼近能力,根据大量实验数据采用神经网络方法来预测下一步拉索振动状态所对应的控制力,并进行了仿真,证实了其有效性.

Cable Active Control Force Implementation Based on Neural Network

Cable easily happened large amplitude vibration under external excitation,providing the axial control force to cable active control is a kind of new method. In this paper,the cable active control strategy is classical quadratic linear optimal control.Because the key of the LQR control is to calculate Riccati equation according to the real-time state matrix,but the Riccati equation is a nonlinear matrix equation,and it have a high order,mutual coupling between variables,so it is very difficult to solve. In order to reduce lag control output force and fast to solve the out of control force,we want to apply to real-time control better. In this paper,based on the neural network has strong learning ability and nonlinear approximation capability. According to a lot of experimental data using the neural network method to predict the cable vibration state of the control force,and a simulation is carried out,this paper proves its effectiveness.