• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2024, Vol. 60 ›› Issue (6): 82-90.doi: 10.3901/JME.2024.06.082

• 特邀专栏:数据-知识混合驱动的智能制造系统 • 上一篇    下一篇

扫码分享

基于多尺度-高效通道注意力网络的刀具故障诊断方法

狄子钧1, 袁东风1, 李东阳1, 梁道君1, 周晓天2, 信苗苗3, 曹凤3, 雷腾飞3   

  1. 1. 山东大学信息科学与工程学院 济南 250100;
    2. 山东大学控制科学与工程学院 济南 250061;
    3. 齐鲁理工学院机电工程学院 济南 250200
  • 收稿日期:2023-11-05 修回日期:2024-01-20 出版日期:2024-03-20 发布日期:2024-06-07
  • 通讯作者: 袁东风,男,博士,教授,博士研究生导师。主要研究方向为新一代信息技术,柔性制造。E-mail:dfyuan@sdu.edu.cn
  • 作者简介:狄子钧,男,1997年出生。主要研究方向为深度学习。E-mail:d75du@mail.sdu.edu.cn
  • 基金资助:
    山东省重大科技创新工程“面向机床制造领域的新型自适应生产系统关键技术研发及应用示范”资助项目(2019JZZY010111)。

Tool Fault Diagnosis Method Based on Multiscale-efficient Channel Attention Network

DI Zijun1, YUAN Dongfeng1, LI Dongyang1, LIANG Daojun1, ZHOU Xiaotian2, XIN Miaomiao3, CAO Feng3, LEI Tengfei3   

  1. 1. School of Information Science and Engineering, Shandong University, Jinan 250100;
    2. School of Control Science and Engineering, Shandong University, Jinan 250061;
    3. School of Mechanic and Electronic Engineering, Qilu Institute of Technology, Jinan 250200
  • Received:2023-11-05 Revised:2024-01-20 Online:2024-03-20 Published:2024-06-07

摘要: 目前,生产加工流程正向着智能化迈进,设备的故障诊断及预测性维护在保障企业生产效率,降低生产成本方面起着至关重要的作用。刀具作为数控机床的切削工具,其实时健康状态直接影响着机床的加工效率和产品质量。对刀具磨损状态的精准监测有助于避免因刀具失效导致的产品质量问题。基于此背景,研究一种基于深度学习的刀具故障诊断方法,将高效通道注意力应用到多尺度卷积神经网络中,提出基于多尺度-高效通道注意力网络的刀具故障诊断方法,利用通道特征学习将机床主轴不同方向的振动信号进行自适应的特征融合,从而提升刀具磨损状态诊断精度。此外,设计刀具磨损试验平台,用于采集符合实际生产的数据,在实际生产场景中验证所提算法的性能。试验结果表明,所提出方法较多尺度网络的刀具故障诊断准确率提高4.47%。

关键词: 刀具故障诊断, 多尺度卷积神经网络, 高效通道注意力, 特征融合, 智能制造

Abstract: Currently, the production process is moving towards intelligence, the equipment fault diagnosis and predictive maintenance play a vital role in ensuring the production efficiency and reducing the production cost. As the milling tool of CNC machine tool, the health state of the machine tool directly affects the processing efficiency and product quality. Precise monitoring of tool wear condition is helpful to avoid product quality problems caused by tool failure. Therefore, A tool fault diagnosis method based on deep learning, which applies the efficient channel attention to the multiscale convolutional neural network. The tool fault diagnosis method based on multiscale-efficient channel attention network (MS-ECA Net) is proposed. MS-ECA Net uses channel feature learning to adaptively fuse the vibration signals of the machine tool spindle in different directions, so as to improve the accuracy of tool wear state diagnosis. In addition, in order to verify the performance of the proposed algorithm in the actual production scenario, A tool wear test platform to collect the data in line with the actual production, and uses the data to verify the performance of the proposed algorithm. Experiment results show that the proposed method improves the accuracy of tool fault diagnosis by 4.47%.

Key words: tool fault diagnosis, convolutional neural networks, channel attention, feature fusion, intelligent manufacturing

中图分类号: