基于神经网络的机械部件数控加工误差控制  

Neural Network-Based Error Control in CNC Machining of Mechanical Components

作　　者：林世南 LIN Shinan(Institute of Intelligent Manufacturing,Huazhong University of Science and Technology,Quanzhou,Fujian 362000,China)

机构地区：[1]泉州华中科技大学智能制造研究院,福建泉州362000

出　　处：《自动化应用》2024年第23期36-39,共4页Automation Application

摘　　要：详细介绍了一种基于双目视觉系统和基于神经网络的机器人铣削加工误差补偿方法。利用PowerVision1500D双目视觉仪器,实现对机器人铣削加工中刀具端位置误差的精确测量。通过配置在工件和机器人主轴上的编码标志点,双目视觉系统能够实时获取和计算刀具TCP在工件坐标系下的实际位置。进一步,采用粒子群优化的反向传播神经网络建立预测模型,用于模拟和预测机器人位姿与刀具TCP位置误差之间的关系,从而提前规划和调整铣削轨迹。最后,在航空铝7075材料部件上进行了综合误差补偿方法的实验验证,结果表明该补偿策略显著提高了加工精度和效率。This paper describes in detail a robot milling machining error compensation method based on binocular vision system and neural network-based.The PowerVision1500D binocular vision instrument is used to achieve accurate measurement of tool end position error in robotic milling machining.Through the coded marker points configured on the workpiece and robot spindle,the binocular vision system is able to acquire and calculate the actual position of the tool TCP under the workpiece coordinate system in real time.Further,a prediction model was developed using a particle swarm optimised back-propagation neural network for simulating and predicting the relationship between the robot′s positional attitude and the positional error of the tool TCP,so as to plan and adjust the milling trajectory in advance.Finally,the integrated error compensation method is experimentally validated on an aerospace aluminium 7075 material component,and the results show that the compensation strategy significantly improves the machining accuracy and efficiency.

关 键 词：机器人铣削 双目视觉系统 误差预测模型 神经网络 

分 类 号：TH133[机械工程—机械制造及自动化] TP183[自动化与计算机技术—控制理论与控制工程]