基于RBF网络监督和单神经元自适应控制器  

The Research of Controller Based on RBF Neural Network and Single Neuron Algorithm

作　　者：唐亚鸣[1] 刘飞[1] 

机构地区：[1]河海大学机电工程学院,江苏常州213022

出　　处：《机械设计与制造》2013年第8期32-33,37,共3页Machinery Design & Manufacture

基　　金：中央高校基本科研业务费资助项目(2009B29914)

摘　　要：RBF神经网络是一种3层前向局部逼近的神经网络,因而采用RBF网络可大大加快学习速度并避免局部极小问题,适合于实时控制的要求。将RBF网络监督与单神经元自适应结合组成控制算法,成为新的控制结构RBF-单神经元控制器,经过与RBF网络监督PI控制结果的比较,得出RBF-单神经元控制响应更快误差更小等优点。根据需要RBF-单神经元控制器可以变成单纯的神经元控制器,提高了控制的灵活性。通过对伺服电机的角度控制仿真实验,验证了提出的方法可行性与可靠性。RBF neural network is a kind of 3 layers to the local approximation,which can greatly accelerate the speed of learning and avoid the local minima problem.It is suitable for the requirement of real-time control.RBF neural network supervising algorithm and single neuron adaptive algorithm are combined to create a RBF-single neuron controller for the known system.Compared with the RBF-PI controller,RBF-single neuron controller’s response speed is faster and the value of error is less during the controlling.The simulation experiment based on the angle controlling of the AC-Servo motor with matlab validates the practicability and security of RBF-single neuron controller.

关 键 词：RBF 网络监督 单神经元自适应 五次多项式曲线 

分 类 号：TH16[机械工程—机械制造及自动化] TM571.65[电气工程—电器]