静电悬浮的NARX网络自适应逆控制  

Adaptive Inverse Control for Electrostatic Suspension Using NARX Network

作　　者：颜诗源[1] 张克志[2] 钱峰[2] 席涛[1] 张胜修[1] 

机构地区：[1]第二炮兵工程学院303室,陕西西安710025 [2]上海交通大学导航与控制研究所,上海200240

出　　处：《控制工程》2010年第6期774-777,共4页Control Engineering of China

基　　金："十一五"国防预研项目(51309050201)

摘　　要：静电悬浮控制系统中存在建模不准确及对象扰动,传统控制器只能在动态控制精度和扰动消除性能之间折衷;为了克服其对控制器精度的影响,研究了带扰动消除的自适应逆控制算法。以非线性自回归动态神经网络进行正模型、逆模型以及扰动消除控制器的实时辨识,利用基于遗传算法的改进粒子群算法进行神经网络的更新,以提高自适应收敛速度和精度。设计了基于DSP与PC的仿真环境,分别部署静电悬浮虚拟被控对象和自适应逆控制算法,实现对控制算法的实时验证。结果表明所设计的控制结构与算法可以实现对静电悬浮的稳定控制与扰动消除。利用PC和相应的I/O接口,以及所部署的实时控制算法可以实现快速控制原型,为控制器的工程实现提供基础。To the problem that model uncertainty and plant disturbances are inevitable to electrostatic suspension control system,and the traditional controllers are compromise between dynamic performance and disturbance canceling,an adaptive inverse control algorithm with disturbance cancelling is introduced to improve the suspension control accuracy.NARX neural networks are used to identify the plant model,inverse model and disturbance cancelling in real-time.A particle swarm optimization updating algorithm based on genetic algorithm is used to improve the adapting speed and accuracy.The virtual plant to be controlled and the adaptive inverse control algorithm are laid out on DSP and PC separately,to testify the control algorithm in real-time.The results show that the control structure and algorithm have stable control performance and disturbance cancelling ability.The PC with appropriate I/O and the real-time algorithm can set as the control prototype,and have significance to the realization of the control system.

关 键 词：静电悬浮 NARX神经网络 自适应逆控制 实时仿真 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]