摘要
针对工业机器人夹持工件进行磨削时的力控制问题,提出一种基于神经网络算法的机器人力控制方法,搭建一套工业机器人磨削系统,并在Visual Studio软件环境下开发了相应的上位机软件。通过分析神经网络算法的原理,设计神经网络结构,使用从实际磨削过程中获得的训练数据对神经网络进行训练;将力传感器实时采集的力信号输出给训练好的神经网络模型,预测出机器人磨削加工的轨迹修正值并传给机器人,对磨削轨迹进行实时修正,从而实现工业机器人的间接力控制。最后,在搭建的工业机器人磨削系统上进行了力跟踪实验和钛合金试件磨削实验,验证了所提出的力控制方法和机器人磨削系统的有效性和实用性。
Aiming at the problem of force control when industrial robots clamped workpiece for grinding,a robot force control method based on neural network algorithm was proposed,a set of industrial robot grinding system was built,and the corresponding upper computer software was developed under the environment of Visual Studio.By analyzing the principle of neural network algorithm,the structure of neural network was designed,and the training data obtained from actual grinding process was used for neural network training.The force signal collected in real time by the force sensor was output to the trained neural network model,the trajectory correction value of the robot grinding machining was predicted and passed to the robot,to real-time correct grinding trajectory,so as to realize the indirect force control of the industrial robot.Finally,the force tracking experiment and titanium alloy grinding experiment were carried out on the industrial robot grinding system,which verified the validity and practicability of the proposed force control method and the robot grinding system.
作者
周帅
刘晓鸣
ZHOU Shuai;LIU Xiaoming(College of Mechanical&Electrical Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing Jiangsu 210016,China)
出处
《机床与液压》
北大核心
2023年第17期21-25,共5页
Machine Tool & Hydraulics
关键词
工业机器人
力控制
神经网络
磨削系统
Industrial robot
Force control
Neural network
Grinding system
作者简介
周帅(1999-),男,硕士研究生,研究方向为智能机器人、机器人力控制。E-mail:zshuai@nuaa.edu.cn;通信作者:刘晓鸣(1958-),男,博士,教授,主要研究方向为机器人系统、自动化设备制造。E-mail:grindliu@hotmail.com。
