欢迎访问《汽车安全与节能学报》，
汽车安全与节能学报

汽车安全与节能学报 ›› 2023, Vol. 14 ›› Issue (2): 232-238.DOI: 10.3969/j.issn.1674-8484.2023.02.010

• 汽车节能与环保 • 上一篇    下一篇

安全行驶下的车用电机轴承的数字孪生故障诊断

张希1(), 廖宇兰2(), 李沁逸1,3, 陈益庆1,3   

  1. 1.广安职业技术学院，广安 638000，中国
    2.海南大学 机电工程学院，海口 570100，中国
    3.兰实大学，巴吞他尼 12000，泰国
  • 收稿日期:2022-10-09 修回日期:2022-12-13 出版日期:2023-04-30 发布日期:2023-04-27
  • 作者简介:张希(1989—)，男(汉)，四川，讲师。E-mail: zhangxi2010_1989@163.com。
    廖宇兰(1967—)，女(汉)，广东，教授。E-mail: liaoyulan@sina.com。
  • 基金资助:
    国家自然科学基金项目资助(52065017);教育部科技发展中心高校产学研创新基金(2018A06035)

Fault diagnosis of vehicle motor-bearings under safe running by digital-twin technology

ZHANG Xi1(), LIAO Yulan2(), LI Qinyi1,3, CHEN Yiqing1,3   

  1. 1. Guang’an Vocational Technical College, Guang’an 638000, China
    2. College of Electromechanical Engineering, Hainan University, Haikou 570100, China
    3. Rangsit University, Baton Tani 12000, Thailand
  • Received:2022-10-09 Revised:2022-12-13 Online:2023-04-30 Published:2023-04-27

RichHTML

2

PDF

58

摘要：

为保证汽车的行车安全，结合深度学习与数字孪生技术，来实时完成对电机轴承的故障诊断。用数字孪生技术，将电机轴承的详细属性转换入虚拟空间，构建具有形状、属性、准则和行为的数字孪生体。用形状和属性模块，来融合电机轴承的对象参数，构建基于故障信号处理、诊断和分类的准则模块；结合行为模块所注入的历史记录，来生成数据完成网络训练。用Matlab 2020a完成数据分析。结果表明：用本文方法对常规信号、内圈型、外圈型和滚动体型的诊断准确率分别达到97.5%、97.8%、96.8%和97.1%。从而，与混合神经网络方法和迁移残差网络方法相比，本文方法存在较优的诊断效果和算法性能。

关键词: 汽车安全, 电机轴承, 故障诊断, 数字孪生, 深度学习, 虚拟空间

Abstract:

To ensure the driving safety of automobiles and timely capture the fault status of motor bearings, a fault diagnosis of motor bearings was completed through the combination of deep learning and digital twin technology. The detailed attributes of the bearings were transformed into virtual space through digital twin technology, and a digital twin body with shape, attributes, criteria, and behavior was constructed. The shape and attribute modules integrated the object parameters of the motor bearings to build a criterion module based on fault signal processing, diagnosis, and classification, and combined the historical records injected by the behavior module to generate data for network training. Data analysis was completed using Matlab 2020a. The results show that the diagnostic accuracies of this method for conventional signals, inner race fault, outer race fault, and rolling element fault are 97.5%, 97.8%, 96.8%, and 97.1%, respectively. Therefore, this method has superior diagnostic effect and algorithmic performance compared with that by using the methods of the hybrid neural network and the migrated residual network.

Key words: automotive safety, motor bearing, fault diagnosis, digital twin, deep learning, virtual space

中图分类号: