基于动态多尺度注意力机制的数控机床伺服刀架动力头退化状态轻量级识别方法

doi:10.3901/JME.260503

机械工程学报 ›› 2026, Vol. 62 ›› Issue (10): 348-361.doi: 10.3901/JME.260503

• 交叉与前沿 • 上一篇

扫码分享

基于动态多尺度注意力机制的数控机床伺服刀架动力头退化状态轻量级识别方法

许伟洋^1,2, 杨兆军^1,2, 刘志峰^1,2, 何佳龙^1,2, 李国发^1,2, 陈传海^1,2, 刘严^1,2

1. 吉林大学数控装备可靠性教育部重点实验室长春 130022;
2. 吉林大学机械与航空航天工程学院长春 130022

收稿日期:2025-05-13 修回日期:2025-12-03 发布日期:2026-07-29
作者简介:许伟洋,男,1997年出生,博士研究生。主要研究方向为数控机床关键功能部件的智能运维。E-mail:xuwy23@mails.jlu.edu.cn;杨兆军,男,1956年出生,博士,教授,博士研究生导师。主要研究方向为数控机床可靠性理论与技术、数控机床可靠性加速试验技术与装备等。E-mail:yzj@jlu.edu.cn;刘志峰,男,1973年出生,博士,教授,博士研究生导师。主要研究方向为数控机床可靠性理论与技术、高性能装配理论与技术、智能制造技术。E-mail:lzfjlu@jlu.edu.cn;何佳龙(通信作者),男,1989年出生,博士,教授,博士研究生导师。主要研究方向为数控机床可靠性理论与技术、剩余寿命预测与维护等。E-mail:hejl@jlu.edu.cn;李国发,男,1970年出生,博士,教授,博士研究生导师。主要研究方向为数控机床可靠性理论与技术、结构可靠性分析与优化设计等。E-mail:ligf@jlu.edu.cn;陈传海,男,1983年出生,博士,教授,博士研究生导师。主要研究方向为数控机床可靠性理论与技术等。E-mail:cchchina@foxmail.com;刘严,男,1996年出生,博士,博士后。主要研究方向为数控机床可靠性理论与技术、剩余寿命预测等。E-mail:liuy_jlu@jlu.edn.cn
基金资助:
国家自然科学基金(U24B2063、52375497)和吉林省教育厅科学研究(JJKH20250085BS)资助项目。

Lightweight Recognition of Degradation States for the Servo Turret Power Head Based on Dynamic Multi-scale Attention Mechanism

XU Weiyang^1,2, YANG Zhaojun^1,2, LIU Zhifeng^1,2, HE Jialong^1,2, LI Guofa^1,2, CHEN Chuanhai^1,2, LIU Yan^1,2

1. Key Laboratory of CNC Equipment Reliability of Ministry of Education, Jilin University, Changchun 130022;
2. School of Mechanical and Aerospace Engineering, Jilin University, Changchun 130022

Received:2025-05-13 Revised:2025-12-03 Published:2026-07-29

摘要/Abstract

摘要： 针对深度学习模型在数控制造装备退化状态识别时受模型计算量大等的限制，难以满足工业现场边缘计算设备轻量化部署的需求，提出一种基于动态多尺度注意力机制与模型轻量化的伺服刀架动力头退化状态识别方法。首先针对现有模型因特征提取能力有限导致鲁棒性和准确性不足等问题，构建具有不同卷积核大小的卷积层，增强多尺度信息的捕捉能力；通过多层次特征融合，将注意力机制、动态卷积与全连接层输出融合，综合利用各层特征提升模型鲁棒性和准确性；其次，为有效实现模型轻量化，提出最小差异化辅助训练策略，在显著降低模型参数规模的同时保证状态识别的准确率。使用伺服刀架动力头可靠性试验系统数据对所提方法进行了6组试验验证。试验结果表明，所提退化状态识别方法在轻量化前平均准确率可达99.66%，模型参数量减少95.44%(从520 000降至20 000左右)后，准确率仍可保持在94.09%。为实现数控制造装备核心部件退化状态识别方法的轻量化，提供了参考。

关键词: 退化状态识别, 动态多尺度卷积, 注意力机制, 模型轻量化, 伺服刀架动力头

Abstract: Addressing the limitations of deep learning models in recognizing degradation states of CNC manufacturing equipment, such as high computational requirements, which hinder their deployment on lightweight edge computing devices in industrial settings, this paper proposes a method for degradation state recognition of servo turret power heads based on a dynamic multi-scale attention mechanism and model lightweighting. To address the limited feature extraction ability of existing models, we use convolutional layers with different kernel sizes to capture multi-scale information. Multi-level feature fusion, combining attention mechanisms, dynamic convolutions, and fully connected layers, enhances the model’s robustness and accuracy. To achieve lightweighting, a minimum differentiation-assisted training strategy is introduced, significantly reducing parameter size while maintaining predictive performance. Six sets of experiments were conducted to verify the proposed method using data from the reliability testing system of the servo turret power head. Experimental results show that the proposed method achieves an average accuracy of 99.66% before lightweighting, and can still maintain an accuracy rate of 94.09% after reducing the parameter size by 95.44% (from about 520 000 to about 20 000 parameters). This provides a reference for lightweight monitoring of the degradation status in CNC machine tools.

Key words: degradation state recognition, dynamic multi-scale convolution, attention mechanism, lightweight, servo turret power head

中图分类号:

TH17

许伟洋, 杨兆军, 刘志峰, 何佳龙, 李国发, 陈传海, 刘严. 基于动态多尺度注意力机制的数控机床伺服刀架动力头退化状态轻量级识别方法[J]. 机械工程学报, 2026, 62(10): 348-361.

XU Weiyang, YANG Zhaojun, LIU Zhifeng, HE Jialong, LI Guofa, CHEN Chuanhai, LIU Yan. Lightweight Recognition of Degradation States for the Servo Turret Power Head Based on Dynamic Multi-scale Attention Mechanism[J]. Journal of Mechanical Engineering, 2026, 62(10): 348-361.

导出引用管理器 EndNote|Reference Manager|ProCite|BibTeX|RefWorks

链接本文: http://www.cjmenet.com.cn/CN/10.3901/JME.260503

http://www.cjmenet.com.cn/CN/Y2026/V62/I10/348

参考文献

[1] 杨兆军，何佳龙，刘志峰，等. 数控机床可靠性技术新进展[J]. 机械工程学报，2023，59(19)：152-163. YANG Zhaojun，HE Jialong，LIU Zhifeng，et al. Recent progress in reliability technology of CNC machine tools[J]. Journal of Mechanical Engineering，2023，59(19)：152-163.
[2] 杨兆军，陈传海，陈菲，等. 数控机床可靠性技术的研究进展[J]. 机械工程学报，2013，49(20)：130-139. YANG Zhaojun，CHEN Chuanhai，CHEN Fei，et al. Progress in the research of reliability technology of machine tools[J]. Journal of Mechanical Engineering，2013，49(20)：130-139.
[3] HE J，WU C，SHEN W，et al. A model for remaining useful life interval prediction of servo turret power head system of turn-milling center under time-varying operating conditions[J]. Computers & Industrial Engineering，2024，197：110592.
[4] 严彤彤，王冬，彭志科，等. 基于谱幅融合广义健康指数的可解释装备退化评估优化模型研究进展[J]. 机械工程学报，2024，60(18)：1-16. YAN Tongtong，WANG Dong，PENG Zhike，et al. Advances in interpretable machine degradation assessment optimization models based on spectral amplitude fusion aided generalized health indices[J]. Journal of Mechanical Engineering，2024，60(18)：1-16.
[5] JIAO Jinyang，ZHAO Ming，LIN Jing，et al. A comprehensive review on convolutional neural network in machine fault diagnosis[J]. Neurocomputing，2020，417：36-63.
[6] ZHU Junjun，JIANG Quansheng，SHEN Yehu，et al. Application of recurrent neural network to mechanical fault diagnosis：A review[J]. Journal of Mechanical Science and Technology，2022，36(2)：527-542.
[7] LÜ Haixin，CHEN Jinglong，PAN Tongyang，et al. Attention mechanism in intelligent fault diagnosis of machinery：A review of technique and application[J]. Measurement，2022，199：111594.
[8] ZHOU Jianghong，QI Junyu，CHEN Dingliang，et al. Continuous remaining useful life prediction by self-guided attention convolutional neural network and memory consciousness adjustment[J]. IEEE internet of Things Journal，2024，11(19)：31947-31958.
[9] CAO Yudong，DING Yifei，JIA Minping，et al. A novel temporal convolutional network with residual self-attention mechanism for remaining useful life prediction of rolling bearings[J]. Reliability Engineering & System Safety，2021，215：107813.
[10] JIANG Guangjun，DUAN Zhengwei，ZHAO Qi，et al. Remaining useful life prediction of rolling bearings based on TCN-MSA[J]. Measurement Science and Technology，2023，35(2)：025125.
[11] LIN Lin，WU Jinlei，FU Song，et al. Channel attention & temporal attention based temporal convolutional network： A dual attention framework for remaining useful life prediction of the aircraft engines[J]. Advanced Engineering Informatics，2024，60：102372.
[12] JIA Chunchun，HE Hongwen，ZHOU Jiaming，et al. A performance degradation prediction model for PEMFC based on bi-directional long short-term memory and multi-head self-attention mechanism[J]. International Journal of Hydrogen Energy，2024，60：133-146.
[13] CHENG Yu，WANG Duo，ZHOU Pan，et al. Model compression and acceleration for deep neural networks：The principles, progress, and challenges[J]. IEEE Signal Processing Magazine，2018，35(1)：126-136.
[14] 刘翰儒，陈桂映，许泽林，等. 基于SERepVGG-A2的轻量化卷积神经网络在熔池状态识别中的应用[J]. 机械工程学报，2025，61(3)：440-448. LIU Hanru，CHEN Guiying，XU Zelin，et al. Application of lightweight convolutional neural network method based on SERepVGG-A2 in melt pool state recognition[J]. Journal of Mechanical Engineering，2025，61(3)：440-448.
[15] JIANG Lingli，SHI Changzhi，SHENG Heshan，et al. Lightweight CNN architecture design for rolling bearing fault diagnosis[J]. Measurement Science and Technology，2024，35(12)：126142.
[16] CHEN Xi，WANG Hui，LU Siliang，et al. Bearing remaining useful life prediction using federated learning with Taylor-expansion network pruning[J]. IEEE Transactions on Instrumentation and Measurement，2023，72：3512610.
[17] YAN Shen，SHAO Haidong，WANG Jie，et al. LiConvFormer：A lightweight fault diagnosis framework using separable multiscale convolution and broadcast self-attention[J]. Expert Systems with Applications，2024，237：121338.
[18] WANG Bin，XIONG Yongcheng，TAN Liguo. A high-precision aero-engine bearing fault diagnosis based on spatial enhancement convolution and vision transformer[J]. IEEE Transactions on Instrumentation and Measurement，2024，74：2501915.
[19] QIAN Chenhui，YANG Zhaojun，HE Jialong，et al. A new indirect transfer fault diagnosis method based on feature separation[J]. Knowledge-Based Systems，2024，300：112058.
[20] GUO Menghao，LIU Zhengning，MU Taijiang，et al. Beyond self-attention：External attention using two linear layers for visual tasks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence，2022，45(5)：5436-5447.
[21] WANG Qilong，WU Banggu，ZHU Pengfei，et al. ECA-Net：Efficient channel attention for deep convolutional neural networks[C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition，June 14-19，2020，Online：IEEE，2020：11534-11542.
[22] HU Jie，SHEN Li，SUN Gang. Squeeze-and-excitation networks[C]//IEEE Conference on Computer Vision and Pattern Recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，June 18-23, 2018，Salt Lake City，UT：IEEE，2018：7132-7141.
[23] ZHANG Hu，ZU Keke，LU Jian，et al. EPSANet：An efficient pyramid squeeze attention block on convolutional neural network[C]//16th Asian Conference on Computer Vision (ACCV). Computer Vision – ACCV 2022. Lecture Notes in Computer Science，December 4-8, 2022，Macao, Peoples R China：Springer，2023，13843：541-557.

基于动态多尺度注意力机制的数控机床伺服刀架动力头退化状态轻量级识别方法

Lightweight Recognition of Degradation States for the Servo Turret Power Head Based on Dynamic Multi-scale Attention Mechanism

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	王裕祥, 高阳, 蒋恩超, 董娜, 陈旭东, 彭凡, 王立闻. 基于GAF-MACNN的风机叶片健康状态监测方法[J]. 机械工程学报, 2026, 62(7): 139-149.
[2]	贾康康, 都军民, 陈飞宇, 时建纬, 侯国义, 李成. 基于卷积神经网络注意力机制模块的复合材料损伤自动识别研究[J]. 机械工程学报, 2026, 62(6): 247-256.
[3]	石逸涵, 张旭, 庄存波, 刘金山, 王家修, 孙连胜. 集成双重注意力机制CNN-LSTM时空网络的离散车间生产瓶颈预测[J]. 机械工程学报, 2026, 62(5): 49-60.
[4]	吴艳灵, 汤宝平, 邓蕾, 付豪. 低通筛选优化神经架构搜索的风电齿轮箱边缘侧故障诊断方法[J]. 机械工程学报, 2025, 61(7): 361-372.
[5]	刘翰儒, 陈桂映, 许泽林, 彭世通, 郭嘉楠, 刘伟嵬, 王奉涛. 基于SERepVGG-A2的轻量化卷积神经网络在熔池状态识别中的应用[J]. 机械工程学报, 2025, 61(3): 440-448.
[6]	宋强, 王闻杰, 贺晓斌, 夏裕俊, 李永兵. 点焊质量诊断模型的泛化性能研究及改进[J]. 机械工程学报, 2025, 61(18): 64-75.
[7]	林天骄, 宋浏阳, 崔玲丽, 王华庆. 非理想数据下基于时序对比注意力模型的寿命预测方法[J]. 机械工程学报, 2025, 61(18): 27-37.
[8]	练秋酉, 郑少武, 涂新奎, 李巍华. 基于体素特征注意力的交通锥桶点云目标检测算法[J]. 机械工程学报, 2025, 61(16): 239-249.
[9]	王志颖, 李天福, 许文纲, 孙闯, 张军辉, 徐兵, 严如强. 降噪混合注意力变分自编码器及其在轴向柱塞泵故障诊断中的应用[J]. 机械工程学报, 2024, 60(4): 167-177.
[10]	付玲, 佘玲娟, 颜镀镭, 张鹏, 龙湘云. 基于内嵌物理信息与注意力机制BiLSTM神经网络的臂架系统疲劳损伤预测模型[J]. 机械工程学报, 2024, 60(13): 205-215.
[11]	李玖法, 邹博文, 任玥. 考虑车-路交互作用的车辆轨迹预测算法研究[J]. 机械工程学报, 2024, 60(10): 76-85.
[12]	李志兵, 肖健梅, 王锡淮. 基于多粒度NRS和改进Bi-LSTM的电力系统暂态稳定评估^*[J]. 电气工程学报, 2023, 18(3): 232-241.
[13]	余晓霞, 汤宝平, 王伟影, 吴宣勇, 李彪. 复杂工况条件下多头注意力双向长短时记忆网络的风电机组缺失数据修复方法研究[J]. 机械工程学报, 2023, 59(14): 1-9.
[14]	卞文彬, 邓艾东, 刘东川, 赵敏, 刘洋, 李晶. 基于改进深度残差收缩网络的风电机组滚动轴承故障诊断方法[J]. 机械工程学报, 2023, 59(12): 202-214.
[15]	吴彤, 李冰锋, 费树岷, 连东辉. 基于改进FCOS算法的架空输电线路防振锤检测^*[J]. 电气工程学报, 2023, 18(1): 143-152.