基于MGPI模型的SMA柔性驱动自适应NN控制
Adaptive neural network control for SMA actuating flexible system based on MGPI hysteresis model
摘要点击 65  全文点击 19  投稿时间:2021-02-25  修订日期:2021-07-01
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2021.10161
  2022,39(4):721-729
中文关键词  形状记忆合金  回滞  自适应神经网络控制
英文关键词  shape memory alloy  hysteresis  adaptive NN control
基金项目  国家自然科学基金项目(62073146), 广东省自然科学基金项目(2018A030313331, 2021A1515011851), 广州市科技计划基础与应用基础研究项目 (202102080435)资助.
作者单位E-mail
冯颖 华南理工大学 zhdfengying@gmail.com 
梁明威 华南理工大学  
中文摘要
      形状记忆合金(SMA)作为一类仿人肌肉驱动的智能柔性驱动材料, 在机器人及高端制造等领域逐步得到 应用, 但由于SMA的热力学效应, 造成输入输出之间存在强饱和回滞非线性, 从而影响了驱动性能. 此外在引入负 载后, SMA柔性驱动部件输出性能表现出更为复杂的驱动特性. 因此, 如何有效抑制带载条件下SMA柔性驱动部件 强饱和非线性影响, 成为提升驱动性能的关键. 针对此问题, 本文重点研究带载条件下SMA柔性驱动部件的建模及 驱动控制算法. 针对SMA驱动部件中的强饱和非线性特性, 本文提出一类修正(MGPI)回滞模型来进行表征. 通过设 定线性输入形状函数, 不仅有效解析表征SMA驱动部件中的饱和回滞非线性, 并且便于控制器设计. 基于MGPI模 型, 考虑柔性驱动部件的动态特性, 本文提出了带载条件下的SMA柔性驱动部件的自适应神经网络控制算法, 实现 考虑内部非线性和外部干扰条件下的驱动精度有效提升, 并保证全局稳定性.
英文摘要
      As a class of smart materials to be utilized as artificial muscles, SMAs have gradually been applied in the robotics and advanced manufacturing areas. However, due to the thermomechanical effects in SMA materials, the strong saturated hysteresis nonlinearities existing in the input-output relationship of SMA driving components will degrade the actuating accuracy. Besides, the output behavior of the SMA driving components is also affected by changes in the loads, presenting the more complex actuating performance. Therefore, the key to improve the actuating performance is to restrain the negative effects originating from the strong hysteresis nonlinearities with loads. Addressing this challenge, the modeling and actuating control algorithms for the SMA driving components under the loading conditions are discussed in this paper. For the strong saturated hysteresis property, a class of modified generalized Prandtl-Ishlinskii (MGPI) hysteresis model is proposed to describe this special feature. By setting the linear input shape function, the proposed MGPI hysteresis model can represent analytically the saturated hysteresis feature accurately and show the facilitation for the controller design. Based on the MGPI hysteresis model, considering the dynamics of the SMA actuators, an adaptive neural network control algorithm is discussed in this paper for the SMA driving components with loads. As an effective solution to problems of internal nonlinearity and external disturbance, the global stability of the closed-loop systems is ensured, and the actuating performance is guaranteed by the proposed method.