基于改进卷积神经网络的数控机床旋转轴热误差建模方法
PDF下载 (220)马剑超,项四通*.基于改进卷积神经网络的数控机床旋转轴热误差建模方法[J].宁波大学学报(理工版),2023,36(2):108-115.DOI:
MA Jianchao,XIANG Sitong*.Thermal error modeling method for rotary axis of CNC machine tool based on improved convolutional neural network[J].Journal of Ningbo University(Natural Science & Engineering Edition),2023,36(2):108-115.DOI:
| Title: | Thermal error modeling method for rotary axis of CNC machine tool based on improved convolutional neural network |
| 作者: | 马剑超, 项四通* |
| Author(s): | MA Jianchao, XIANG Sitong* |
| 关键词: | 五轴数控机床; 旋转轴; 热误差; 角度定位误差; 改进卷积神经网络 |
| Keywords: | five-axis CNC machine tool; rotary axis; thermal error; angular positioning error; improved convolutional neural network |
| 分类号: | TH164 |
| 文献标识码: | A |
| 摘要: | 为对五轴数控机床旋转轴的热误差进行更精确地预测, 解决变工况条件下预测精度不佳与热误差数据获取困难的问题, 提出了基于改进卷积神经网络的热误差建模方法. 采用激光干涉仪与热成像仪采集不同温度下的角度定位误差与热图像, 对热误差进行傅里叶函数拟合, 将预测目标由不同角度下的热误差转变为拟合函数参数. 在VGG网络模型架构上, 引入SKNet注意力机制, 以提高模型对变工况下的热图像特征提取水平, 并采用全局平均池化代替全连接层, 以改善过拟合情况. 将建立的模型用于热误差预测, 结果表明, 旋转轴热误差预测RMSE在升温状态下为8.36″, 降温条件下为9.57″, 预测精度达90%以上, 优于普通卷积神经网络模型. 结果证实了所提方法在旋转轴热误差建模中的有效性. |
| Abstract: | To solve the problems of the poor prediction accuracy and the difficulty in thermal error data collecting under various working conditions, a thermal error modeling method based on improved convolutional neural network is proposed for more accurately predicting the thermal error of the five-axis CNC machine tool rotary axis. A laser interferometer and thermal imager are used to collect the angular positioning errors and thermal images of the rotary axis at different temperatures. Through Fourier function fitting, the thermal error prediction under different motion angles is transformed into the prediction of the parameters of the fitting function. Based on the VGG network architecture, the SKNet attention mechanism is introduced to improve the feature extraction ability of the model for different motion situations of the rotary axis, and the global average pooling is adopted to avoid over fitting. The established model is used for thermal error prediction, and the results show that the RMSE of the thermal error prediction is 8.36" under the heating condition and 9.57" under the cooling condition, and the prediction accuracy is over 90%, which is better than the traditional convolutional neural network model. The results confirm the effectiveness of the proposed method in the thermal error modeling of the rotary axis. |
| 参考文献 /References: | [1] Wu C Y, Xiang S T, Xiang W S. Spindle thermal error prediction approach based on thermal infrared images: A deep learning method[J]. Journal of Manufacturing Systems, 2021, 59:67-80. [2] Yang J, Shi H, Feng B, et al. Thermal error modeling and compensation for a high-speed motorized spindle[J]. The International Journal of Advanced Manufacturing Technology, 2015, 77:1005-1017. [3] Wu C, Xiang S, Xiang W. Thermal error modeling of rotary axis based on convolutional neural network[J]. Journal of Manufacturing Science and Engineering, 2021, 143(5):1-13. [4] Bitar-Nehme E, Mayer J R R. Thermal volumetric effects under axes cycling using an invar R-test device and reference length[J]. International Journal of Machine Tools and Manufacture, 2016, 105:14-22. [5] 刘阔, 韩伟, 王永青, 等. 数控机床进给轴热误差补偿技术研究综述[J]. 机械工程学报, 2021, 57(3):156-173. [6] Wegener K, Gebhardt M, Mayr J, et al. Thermal issues in 5-axis machine tools[C]. The Proceedings of MTTRF 2013 Annual Meeting. Eidgenssische Technische, Hochschule Zürich, 2014. [7] Ibaraki S, Inui H, Hong C F, et al. On-machine identification of rotary axis location errors under thermal influence by spindle rotation[J]. Precision Engineering, 2019, 55:42-47. [8] Yang Z Y, Sun M L, Li W Q, et al. Modified Elman network for thermal deformation compensation modeling in machine tools[J]. The International Journal of Advanced Manufacturing Technology, 2011, 54:669-676. [9] Chang C W, Kang Y, Chen Y W, et al. Thermal deformation prediction in machine tools by using neural network[M]//Neural Information Processing. Berlin, Heidelberg: Spinger Berlin Heidelberg, 2006, 4233:850- 859. [10] 夏军勇, 胡友民, 吴波, 等. 热弹性效应分析与机床进给系统热动态特性建模[J]. 机械工程学报, 2010, 46(15): 191-198. [11] Gebhardt M, Ess M, Weikert S, et al. Phenomenological compensation of thermally caused position and orientation errors of rotary axes[J]. Journal of Manufacturing Processes, 2013, 15(4):452-459. [12] Ibaraki S, Oyama C, Otsubo H. Construction of an error map of rotary axes on a five-axis machining center by static R-test[J]. International Journal of Machine Tools and Manufacture, 2011, 51(3):190-200. [13] He Z Y, Fu J Z, Zhang L C, et al. A new error measurement method to identify all six error parameters of a rotational axis of a machine tool[J]. International Journal of Machine Tools and Manufacture, 2015, 88:1-8. [14] Szegedy C, Liu W, Jia Y Q, et al. Going deeper with convolutions[C]//2015 IEEE Conference on Computer Vision and Pattern Recognition, Boston, MA, USA. IEEE, 2015:1-9. [15] Newell A, Yang K, Deng J. Stacked hourglass networks for human pose estimation[C]//Computer Vison – ECCV 2016, 2016:483-499. [16] Baranwal S, Khandelwal S, Arora A. Deep learning convolutional neural network for apple leaves disease detection[C]//International Conference on Sustainable Computing in Science, Technology & Management, 2019: 1-8. [17] Li X, Wang W H, Hu X L, et al. Selective kernel network[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA. IEEE, 2019:510-519. |
| 备注/Memo: | 收稿日期: 2022－03－28. 宁波大学学报（理工版）网址: http://journallg.nbu.edu.cn/ 基金项目: 国家自然科学基金（52175470）; 浙江省自然科学基金（LY20E050005）. 宁波大学学报（理工版）网址：http://journallg.nbu.edu.cn/第一作者: 马剑超（1998－）, 男, 浙江温州人, 在读硕士研究生, 主要研究方向: 五轴数控机床. E-mail: w15757400310@163.com *通信作者: 项四通（1989－）, 男, 浙江绍兴人, 副教授, 主要研究方向: 五轴数控机床. E-mail: xiangsitong@nbu.edu.cn |
