电子学报 ›› 2015, Vol. 43 ›› Issue (7): 1315-1319.DOI: 10.3969/j.issn.0372-2112.2015.07.010

• 学术论文 • 上一篇    下一篇

基于HBF神经网络的自适应观测器

闻新1, 张兴旺1, 张威2   

  1. 1. 沈阳航空航天大学航空航天工程学部, 辽宁沈阳 110136;
    2. 北京石油化工学院信息工程系, 北京 102617
  • 收稿日期:2014-04-23 修回日期:2014-10-13 出版日期:2015-07-25 发布日期:2015-07-25
  • 通讯作者: 张兴旺
  • 作者简介:闻 新 男,1961年3月生,辽宁沈阳人,博士生导师,现任南京航空航天大学航天控制系主任.长期从事航天器总体设计工作及航天器智能故障诊断工作.曾主持国家921工程"神舟飞船故障模拟与仿真实验室"建设、组织完成若干国家"863"项目和总装支撑预研项目,曾主持完成国家发改委"高精度卫星导航产品产业化"等项目.曾担任中国航天科工集团公司研发中心主任,中心副总师、主任和总指挥等职务. E-mail:wen_xin2004@126.com

Adaptive Observer Based on HBF Neural Networks

WEN Xin1, ZHANG Xing-wang1, ZHANG Wei2   

  1. 1. Faculty of Aerospace Engineering, Shenyang Aerospace University, Shenyang, Liaoning 110136, China;
    2. College of Information Engineering, Beijing Institute of Petrochemical Technology, Beijing 102617, China
  • Received:2014-04-23 Revised:2014-10-13 Online:2015-07-25 Published:2015-07-25

摘要:

传统的RBF(Radial Basis Function)神经元基函数通常把高斯类型与单一宽度作为每个神经元的激活函数,这些特性限制了网络神经元的性能,特别是在处理复杂的非线性建模问题上.为了克服这个限制,本文应用了具有类似RBF网络,但激活函数不同-超基函数HBF(Hyper Basis Function)的网络.结合RBF网络,分析了HBF网络的结构、基函数形式及基函数对网络的影响,利用决策树算法计算了网络中心.在此基础上,提出了一种基于HBF神经网络的自适应观测器设计方法,并通过引入Lyapunov函数,证明了这种观测器设计方法的稳定性;最后通过仿真验证了这种HBF神经网络观测器能很好地观测系统的状态值.

关键词: 状态估计, HBF(Hyper Basis Function)神经网络, 非线性系统, 决策树

Abstract:

Conventional RBF neuron network is usually based on Gaussian activation function with single width for each activation function.This feature restricts neuron performance for modeling the complex nonlinear problems.To accommodate limitation of a single scale,this paper applies neural network with similar but yet different activation function—HBF(Hyper Basis Function).The state for nonlinear systems is estimated by using HBF neural networks.Combined with RBF (Radial Basis Function) networks,the structure of networks,the form of its basis functions and its influence on HBF(Hyper Basis Function) are analyzed.Decision tree algorithm is used to determine the network center.Then a design method of adaptive observer based on HBF neural networks is proposed.The Lyapunov function is introduced to prove the stability and the conditions of the bounded error of the observer.And this HBF neural network is turned out to observe system state very well by simulation.

Key words: state estimation, HBF(hyper basis function) neural networks, nonlinear system, decision tree

中图分类号: