基于数字孪生的多轴数控机床轮廓误差预测研究  

Contour error prediction of multi-axis CNC machines based on digital twin technology

作　　者：赵冠 赵巍[1] 魏国家 张宇[1] 王威 梁晓阳 ZHAO Guan;ZHAO Wei;WEI Guojia;ZHANG Yu;WANG Wei;LIANG Xiaoyang(School of Mechanical Engineering,Tianjin University of Technology and Education,Tianjin 300222,China;Department of Mechanical and Electrical Engineering,Liaoning Provincial College of Communications,Shenyang 110122,China)

机构地区：[1]天津职业技术师范大学机械工程学院,天津300222 [2]辽宁省交通高等专科学校机电系,沈阳110122

出　　处：《天津职业技术师范大学学报》2024年第4期1-7,共7页Journal of Tianjin University of Technology and Education

基　　金：辽宁省教育厅2022年基本科研项目(LJKMZ20222111)。

摘　　要：为提高五轴数控机床在复杂加工环境下的轮廓误差预测精度,以机床实际加工数据和关键参数为基础,融合数字孪生技术和深度学习神经网络算法,构建一个具有互操作性和闭环性的高保真孪生预测模型。利用S形试件上选定参考点的法向轮廓误差数据及一系列关键参数作为历史数据进行神经网络模型训练。训练完成的模型被嵌入到机床的数字孪生体中,封装成完整的数字孪生轮廓误差预测模型,并在Unity 3D环境下实现加工过程的高度可视化。该模型不仅能够有效预测和指导五轴机床在加工过程中的轮廓误差补偿,进一步提高模型的实用性与交互性,而且还展现了数字孪生技术和深度学习算法在高精度机床控制领域的巨大潜力。实验结果表明:模型最大预测误差小于6μm,最小预测误差为0.9μm,模型能够精准预测和补偿机床的轮廓误差,验证了该方法在复杂加工场景下的有效性和优越性。To enhance the contour error prediction accuracy of five-axis CNC machines in complex machining environments,a high-fidelity twin prediction model featuring interoperability and closed-loop capabilities was successfully constructed by integrating digital twin technology and deep learning neural network algorithms,based on actual machining data and key parameters of the machine.Normal contour error data from selected reference points on an S-shaped workpiece and a series of key parameters were used as historical data to train the neural network model.The trained model was embedded into the machine′s digital twin and encapsulated into a complete digital twin contour error prediction model,achieving high levels of visualization for the machining process within the Unity 3D environment.This model not only effectively predicts and guides contour error compensation during the machining process of five-axis machines,further enhancing its practicality and interactivity,but also demonstrates the tremendous potential of digital twin technology and deep learning algorithms in high-precision machine control.Experimental results show that the model′s maximum prediction error is less than 6 microns,with a minimum prediction error of 0.9 microns,proving the model′s ability to accurately predict and compensate for the machine′s contour errors,validating the effectiveness and superiority of this approach in complex machining scenarios.

关 键 词：轮廓误差 数字孪生 神经网络 误差预测 

分 类 号：TG659[金属学及工艺—金属切削加工及机床]