基于深度域适应迁移学习的滚动轴承故障诊断方法研究

Research on fault diagnosis method of rolling bearing based on adaptive migration learning in depth domain

在实际工业生产中,滚动轴承故障数据稀疏,构建的小样本数据集易导致模型过拟合,需要进行数据增强构成训练域。由于机械设备在运行中存在工况时变性,导致训练数据和测试数据之间存在差异,进而造成轴承故障诊断准确率不高,故提出一种基于生成对抗网络和多核最大均值差异(MK-MMD)的深度域适应迁移学习方法。文中通过有限元模拟获得大量带有标签的仿真故障数据,在此基础上构建深度域适应迁移学习的故障诊断模型,并利用凯斯西储大学的轴承数据集对轴承故障诊断模型进行测试。此外,通过与卷积神经网络(CNN)、深度适配网络(DAN)和深度域自适应网络(DANN)的故障诊断结果对比,证明基于有限元模拟的数据增强和迁移学习的轴承故障诊断方法可有效提高轴承故障诊断的准确度。

In actual industrial production,the fault data of rolling bearings are sparse,and the small sample data set is easy to lead to over-fitting of the model,so it is necessary to enhance the data to form a training domain.The working conditions of mechanical equipment will change in real time,which leads to the difference between training data and test data,and the accuracy of bearing fault diagnosis is not high.Therefore,an adaptive migration learning method in depth domain based on generating countermeasure network and multi-core maximum mean difference(MK-MMD)is proposed.In this paper,a large number of labeled simulation fault data were obtained by finite element simulation,and a fault diagnosis model with adaptive transfer learning in depth domain was constructed,and the bearing fault diagnosis model was tested by using Case Western Reserve University’s bearing data set.In addition,compared with the fault diagnosis results of convolutional neural network(CNN),depth adaptive network(DAN)and depth domain adaptive network(DANN),it was proved that the bearing fault diagnosis method based on data enhancement and migration learning of finite element simulation can effectively improve the accuracy of bearing fault diagnosis.