电液负载模拟器的RBF神经网络控制

RBF NEURAL NETWORK CONTROL ON ELECTRO-HYDRAULIC LOAD SIMULATOR

针对传统控制器自适应能力和鲁棒性差、参数调节繁琐等缺陷,提出了神经网络自适应控制方案,同时针对神经网络控制器计算复杂和稳定性证明缺乏理论依据等不足问题,在传统控制理论的基础上设计了基于PID的神经网络控制方案。根据电液负载模拟器的结构特点和目前神经网络控制的发展水平,提出了基于RBF神经网络的PID控制器,并构建了负载模拟器的神经网络控制与基于Jacobian的消扰(消除多余力)结构,仿真和试验证明其具有很好的自适应性和鲁棒性。

A new control structure for load simulator to improve its performance is presented, in which there are three RBF neural networks applied, one is taken as NN PID controller, the second is taken as compensator for adjusting anti-disturbance coefficient dynamically, the third one is use for identifying the plant (NNI). The neural network PID controller is designed by requirement for steady, which overcome the shortcoming of the only neural network controller. A RBF neural network is used to identify the plant, in which the Jacobian matrix is got to use for adjusting disturbance-reduced parameter in the compensating neural network. The simulation and experiment are carried out, which show a good result. This method is of validity and robustness.