基于聚类与梯度混合学习算法RBF神经网络的电液伺服系统建模及仿真  

Modeling and Simulation of Electrohydraulic Servo System Based on RBF Neural Network Trained by a Clustering and Gradient Hybrid Algorithm

作　　者：吴潇潇[1] 陈机林[1] 王力[1] 

机构地区：[1]南京理工大学机械工程学院,江苏南京210094

出　　处：《电气自动化》2011年第1期20-22,共3页Electrical Automation

摘　　要：针对电液伺服系统固有的流量—压力等非线性因素使得传统的机理建模法难以获得系统的精确数学模型,该文采用RBF神经网络对某型号武器扫雷犁电液伺服系统建模并进行了仿真。由于神经网络学习时间较长且不易收敛,采用聚类与梯度训练相结合的混合学习算法对RBF神经网络进行训练。首先使用聚类方法对学习样本进行聚类,确定隐含层结构,然后用梯度训练法对确定的网络结构进行训练,仿真实验验证了该混合学习算法的有效性。As it＇s hard to get the exact model of the electrohydraulic servo system using first-principle method due to the nonlinear causations such as the inherent flux-pressure relation,RBF neural network was used for the modeling and simulation of the electrohydraulic servo system of a mine sweeping plough in a weapon system.Since the neural network had slow training speed and poor convergence,a hybrid algorithm based on clustering and gradient training was proposed to train the network.The clustering algorithm was used for the clustering of the swatch and the structuring of the crytic network layer.The gradient algorithm was used to train the network.The simulation results showed the validity of the algorithm.

关 键 词：RBF神经网络 电液伺服系统 系统辨识 聚类算法 梯度训练算法 

分 类 号：TP275[自动化与计算机技术—检测技术与自动化装置] TP183[自动化与计算机技术—控制科学与工程]