基于神经网络的机械臂摩擦力参数辨识方法  

Identification Method of Friction Force Parameters of Manipulator Based on Neural Network

作　　者：赵柏山[1] 刘丹丹 ZHAO Baishan;LIU Dandan(School of Information Science and Engineering,Shenyang University of Technology,Shenyang 110870,China)

机构地区：[1]沈阳工业大学信息科学与工程学院,辽宁沈阳110870

出　　处：《机械制造与自动化》2023年第5期155-158,共4页Machine Building & Automation

摘　　要：非线性摩擦是导致机械臂末端振动、引起跟踪轨迹偏移、控制误差的主要原因之一,因此对机械臂的摩擦力进行估计是非常必要的。在柔性关节机械臂的动力学模型中引入库仑-黏性摩擦力,获取摩擦力与特征信号之间的关系数据,将数据送入卷积神经网络与门控循环单元组合模型中对摩擦力进行参数辨识。分别对关节1处和关节2处的摩擦力进行辨识,辨识精度达到了99%以上。为建立更精确的机械臂模型、提高控制精度提供了理论基础。As onlinear friction is one of the main reasons for the vibration of the end of the manipulator,the deviation of the tracking trajectory and the control error,it is of necessity to estimate the friction force of the manipulator.Introduces the Coulomb-viscous friction force into the dynamic model of the flexible joint manipulator for numerical solution,abtains the data between the friction force and the characteristic signal and feeds the data into the combined model of convolutional neural network and gated recurrent unit for parameter identification of friction force.The friction parameters of joint 1 and joint 2 were identified respectively with the identification accuracy as over 99%,which provides a theoretical basis for establishing a more accurate manipulator model and improving the precision of motion control.

关 键 词：柔性关节机械臂 卷积神经网络 门控循环单元 参数辨识 

分 类 号：TP241[自动化与计算机技术—检测技术与自动化装置]