Abstract: To address the problem of contour error affecting the machining accuracy of CNC machine tool feeding system in the process of high precision machining,a contour error prediction and pre-compensation method based on the combination of digital twin model and cross-coupling was proposed.Aiming at the multi-domain cross-coupling characteristics of the CNC feeding system,a joint simulation mechanism model of mechatronics was constructed.At the same time,based on the Convolutional Neural Network with Bidirectional Long Short Term Memory (CNN-BiLSTM) that integrated the Squeeze-and-Excitation networks (SE) and cross-coupling compensation,the data driven compensation model was proposed to predict and pre-compensate contour error,which together constituted a digital twin model of the feed system.The contour error was predicted by the data-driven model,and the digital twin model was pre-compensated based on the prediction results.Finally,the coupled control gain obtained from the digital twin model's predicted contour error and pre-compensation was updated to the control part of the actual two-axis feeding system,which improved the contour control performance.The reliability of the contour error prediction and pre-compensation was verified by the dual-axis feeding system,and the effectiveness of the actual contour error compensation was also verified by using the virtual-real synchronization experiments,and the relative errors were all kept below 6% using this method compared with the pre-compensation.The results showed that the proposed method could effectively predict and control the contour error,and could be applicated in the field of precision motion control.

Key words: digital twin, feed system, attention mechanism, contour error, error compensation

摘要： 针对数控机床进给系统的轮廓误差影响加工精度问题,提出一种基于数字孪生模型和交叉耦合相结合的轮廓误差预测与预补偿方法。该方法针对数控进给系统的多领域交叉耦合的特点,首先构建了机电一体化的联合仿真机理模型,同时基于融合了SE注意力机制(SE)的卷积循环神经网络(CNN-BiLSTM)和交叉耦合补偿构建了数据驱动补偿模型,二者共同构成了进给系统数字孪生模型。其次,通过数据驱动模型预测轮廓误差,同时基于预测结果对数字孪生模型进行预补偿。最后将数字孪生模型预估的轮廓误差与预补偿得到的耦合控制增益更新到实际进给系统的控制部分,提高了轮廓的控制性能。通过实验在进给系统上验证了轮廓误差预测与预补偿的可靠性,同时利用虚实同步实验验证了实际轮廓误差补偿的有效性。与补偿前相比,相对误差均保持在6%以下。结果表明:该方法能够有效的预测并控制轮廓误差,在精密运动控制领域有良好的应用前景。

关键词: 数字孪生, 进给系统, 注意力机制, 轮廓误差, 误差补偿