电液伺服加载系统的神经网络逆控制

针对液压加载系统及试件的特殊要求，采用基于径向基函数网络（RBFN）的逆控制器给出了该NARMA模型及逆模型存在的条件。该方法使用实际系统的输入输出信号以及径向基函数网络实现系统建模，并利用系统神经网络模型离线训练系统的逆动态作为控制器，以克服由于实际系统所受的扰动而可能引起的控制器（逆模型）不收敛，实际控制表明该系统对期望加载轨线具有良好的跟踪能力，同时对系统干扰和不确定性具有较强的鲁棒性。

NN Inverse Control for Electro-hydraulic Servo Load System

A kind of inverse controller based on RBFN is adopted in this paper to satisfy the special requests of the electro-hydraulic servo load system and its object, then the existent conditions of the NARMA model and its inverse model is proposed. The method models the system by using RBFN, which only uses the input and output signal, and then the NN inverse model, which is also the controller, is trained by the system NN model offline to ensure its astringency that may be destroyed by the system noise. Practice control shows that the output can track the desire trajectory exactly, and the robustness against some uncertainty and noises is proved.