• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2016, Vol. 52 ›› Issue (3): 143-150.doi: 10.3901/JME.2016.03.143

• 数字化设计与制造 • 上一篇    下一篇

基于RLS-DE算法的多变量径向磁轴承系统辨识

魏彤1, 2, 3,  田双彪1, 2, 3   

  1. 1. 北京航空航天大学仪器科学与光电工程学院  北京  100191;
    2. 北京航空航天大学新型惯性仪表与导航系统技术国防重点学科实验室  北京  100191;
    3. 北京航空航天大学惯性技术重点实验室  北京  100191
  • 收稿日期:2015-03-09 修回日期:2015-10-23 出版日期:2016-02-05 发布日期:2016-02-05
  • 通讯作者: 田双彪,男,1986年出生。主要研究方向为磁轴承控制。E-mail:shuangbiao.1@163.com
  • 作者简介:魏彤,男,1972年出生,博士,副教授,硕士研究生导师。主要研究方向为磁轴承控制、磁悬浮控制力矩陀螺研制等。 E-mail:weitong3000@sina.com
  • 基金资助:
    国家自然科学基金资助项目(61174134)

The Identification of Multivariable Radial Magnetic Bearing System Based on RLS-DE Algorithm

WEI Tong1, 2, 3,  TIAN Shuangbiao1, 2, 3   

  1. 1. School of Instrumentation Science & Opto-electronics Engineering, Beihang University, Beijing 100191;
    2. Fundamental Science on Novel Inertial Instrument & Navigation System Technology Laboratory, Beihang University, Beijing 100191;
    3. Science and Technology on Inertial Laboratory, Beihang University, Beijing 100191
  • Received:2015-03-09 Revised:2015-10-23 Online:2016-02-05 Published:2016-02-05

摘要: 磁轴承的系统模型是提高其控制精度、稳定性及可靠性的基础,为了准确获取多变量径向磁轴承的系统模型,提出一种基于“递推最小二乘-差分进化”算法的辨识方法。该方法在剔除与转子转速同频的不平衡量的基础上,首先采用递推最小二乘法对径向磁轴承系统模型进行初步辨识,之后在初步辨识得到的模型参数的小范围内初始化差分进化算法的种群,通过反复进行差分进化算法的变异、交叉和选择操作,直至得到系统模型的最优参数。对该方法进行了仿真和试验验证,仿真结果表明,在递推最小二乘法辨识的基础上,通过差分进化算法小范围搜索得到的辨识模型输出误差的方差下降了92.86%;试验结果表明输出误差的方差能下降80.13%。验证了该方法高精度辨识径向磁轴承系统模型的有效性。

关键词: 辨识, 差分进化算法, 磁轴承, 递推最小二乘法

Abstract: System model of magnetic bearing is the basis for improving control precision, stability and reliability. To achieve the exact model of multivariable radial magnetic bearing system, a method based on recursive least squares differential evolution (RLS-DE) algorithm is proposed. Based on eliminating the synchronous periodic vibration with rotary speed of rotor, recursive least squares (RLS) meheod is applied into elementarily identifying system model of radial magnetic bearing. Furthermore, initialized population of the differential evolution (DE) algorithm on a small scale of model parameters of elementary identification, and the optimal parameters of system model are achieved by repetitive mutation, crossover and selection of the DE algorithm. Based on identifying of RLS algorithm, the simulation indicates the variance of output error of identification model achieved by searching on a small scale is reduced by 92.86%, and the experiment indicates the variance of output is reduced by 80.13%. Results of simulation and experiment show the validity of proposed algorithm in identifying accurately system model of radial magnetic bearing.

Key words: differential evolution algorithm, identification, magnetic bearing, recursive least squares method

中图分类号: