引用本文:王建成,李强亚,刘涛,谭永红,阎帅.梯度提升最小二乘支持向量回归的压电执行器磁滞特性建模[J].控制理论与应用,2024,41(9):1692~1697.[点击复制]
WANG Jian-cheng,LI Qiang-ya,LIU Tao,TAN Yong-hong,YAN Shuai.Hysteresis characteristics modeling of piezoelectric actuator by gradient boosting least-squares support vector regression[J].Control Theory and Technology,2024,41(9):1692~1697.[点击复制]
梯度提升最小二乘支持向量回归的压电执行器磁滞特性建模
Hysteresis characteristics modeling of piezoelectric actuator by gradient boosting least-squares support vector regression
摘要点击 1805  全文点击 26  投稿时间：2022-08-10  修订日期：2024-05-07
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2023.20713
  2024,41(9):1692-1697
中文关键词  压电执行器  磁滞效应  磁滞算子  最小二乘支持向量机  可保证收敛粒子群算法  梯度提升
英文关键词  piezoelectric actuator  hysteresis effect  hysteretic operator  least-squares support vector machine  guaranteed convergence particle swarm optimization  gradient boosting
基金项目  国家自然科学基金项目(62327807, 62361136585), 教育部重点基地平台科研专题项目(DUT21LAB113)资助.
作者单位E-mail
王建成 大连理工大学 wangjcdl@163.com 
李强亚 大连理工大学  
刘涛* 大连理工大学 tliu@dlut.edu.cn 
谭永红 上海师范大学  
阎帅 大连理工大学  
中文摘要
      针对用于精密运动定位的压电执行器具有磁滞效应的问题, 本文提出一种基于梯度提升最小二乘支持向量回归(GB-LSSVR)的建模方法. 首先, 通过引入磁滞算子构造拓展的输入空间, 将磁滞的多值映射转换为一对一映射. 然后, 建立基于GB-LSSVR的磁滞模型, 设计可保证收敛粒子群算法(GCPSO)对GB-LSSVR模型参数进行优化. 最后, 将所提出方法用于实际预测一个压电执行器的位移. 结果表明, 该方法相对于经典的最小二乘支持向量回归(LSSVR)和截断最小二乘支持向量回归(T-LSSVR)算法, 能得到更加准确的结果.
英文摘要
      Concerning the problem of hysteresis effect related to piezoelectric actuators used for precise motion positioning, a modeling method is proposed based on the gradient boosting least-squares support vector regression (GB-LSSVR). Firstly, an expanded input space is constructed by introducing a hysteretic operator, such that the multi-valued mapping of hysteresis is transformed into a one-to-one mapping. Then the hysteresis model is established based on the GB-LSSVR, of which the parameters are optimized by the guaranteed convergence particle swarm optimization (GCPSO) algorithm. Finally, the proposed method is applied to practically predict the displacement of a piezoelectric actuator. The results show that the proposed method could obtain more accurate result compared to the classical algorithms of least-squares support vector regression and truncated least-squares support vector regression.