引用本文:付东学,赵希梅.基于神经网络观测器的反推终端滑模位置控制[J].控制理论与应用,2023,40(1):132~138.[点击复制]
FU Dong-xue,ZHAO Xi-mei.Backstepping terminal sliding mode position control based on neural network observer[J].Control Theory and Technology,2023,40(1):132~138.[点击复制]
基于神经网络观测器的反推终端滑模位置控制
Backstepping terminal sliding mode position control based on neural network observer
摘要点击 894  全文点击 297  投稿时间：2021-12-15  修订日期：2023-02-17
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2022.11227
  2023,40(1):132-138
中文关键词  永磁直线同步电机  神经网络  终端滑模控制  观测器  抖振
英文关键词  permanent magnet linear synchronous motor  neural network  terminal sliding mode control  observer  chattering
基金项目  辽宁省自然科学基金计划重点项目(20170540677)资助.
作者单位E-mail
付东学 沈阳工业大学电气工程学院  
赵希梅* 沈阳工业大学电气工程学院 zhaoxm_sut@163.com 
中文摘要
      为了提高永磁直线同步电机(PMLSM)的位置跟踪精度, 本文提出了一种基于神经网络自适应观测器的反 推终端滑模控制(TSMC)方法. 首先, 建立PMLSM的动力学模型. 然后, 利用RBF神经网络的万能逼近特性去逼近系 统中不确定性, 并将逼近后的输出信号输入给自适应观测器进行跟踪目标位置和速度的估计, 补偿由不确定性所导 致的跟踪误差, 进而获得高精度的跟踪性能. 同时反推TSMC方法能够保证系统状态在有限时间内收敛, 有效改善 了系统响应速度和鲁棒性能. 此外, 设计出一种新型饱和函数来改善系统抖振, 并利用Lyapunov稳定性定理进行了 闭环系统稳定性分析. 最后, 通过空载和负载实验证实了该控制方案的有效性.
英文摘要
      In this paper, a backstepping terminal sliding mode control (TSMC) method based on neural network adaptive observer is designed to promote the position tracking accuracy of permanent magnet linear synchronous motor (PMLSM). First, the dynamics model of PMLSM is established. Then, the generalized approximation property of RBF neural network is used to approximate the system uncertainty, and the approximated output signal is fed to the adaptive observer for tracking target position and velocity estimation to compensate the tracking error caused by the uncertainty, and then obtain the high accuracy tracking performance. The backstepping TSMC method also ensures that the system state converges in finite time, which effectively develops the response speed and robustness. In addition, a new saturation function is designed to weaken chattering, and Lyapunov theorem is used to ensure the stability of the closed-loop system. Finally, the effectiveness of the control scheme is verified through no-load and load experiments.