基于Inception-BLSTM的滚动轴承故障诊断方法研究  

A study on method of rolling bearing fault diagnosis based on Inception-BLSTM

作　　者：赵凯辉[1] 吴思成 李涛 贺才春[4] 查国涛[4] ZHAO Kaihui;WU Sicheng;LI Tao;HE Caichun;ZHA Guotao(College of Electrical and Information Engineering,Hunan University of Technology,Zhuzhou 412007,China;College of Traffic Engineering,Hunan University of Technology,Zhuzhou 412007,China;CRRC Zhuzhou Institute Co.,Ltd.,Zhuzhou 412005,China;Zhuzhou Times New Material Technology Co.,Ltd.,Zhuzhou 412007,China)

机构地区：[1]湖南工业大学电气与信息工程学院,湖南株洲412007 [2]湖南工业大学交通工程学院,湖南株洲412007 [3]中车株洲电力机车研究所有限公司,湖南株洲412005 [4]株洲时代新材料科技股份有限公司,湖南株洲412007

出　　处：《振动与冲击》2021年第17期290-297,共8页Journal of Vibration and Shock

基　　金：国家自然科学基金(61773159);湖南省自然科学基金(2020JJ6083,2020JJ6067,2019JJ40072);湖南省研究生科研创新项目(CX20190861)。

摘　　要：针对传统的滚动轴承故障诊断方法依赖大量先验知识以及容易人为引入误差等缺点,结合Inception模型的多尺度抽象特征提取能力与双向长短时记忆(BLSTM)神经网络序列建模的优势,提出一种基于Inception-BLSTM的滚动轴承故障诊断方法。首先,设计Inception模型从滚动轴承振动信号中提取出多尺度抽象特征。其次,设计BLSTM进一步学习特征信息的时间依赖性。最后,通过全连接层将特征信息映射到对应的故障模式并得出诊断结果。实验结果表明,该方法在多负载场景下的轴承故障识别精度达到了99.6%,具有良好的负载适应性以及抗干扰能力。To solve the defect of the conventional rolling bearing fault diagnosis method requiring a large amount of prior knowledge and easy to introduce error artificially, a method of rolling bearing fault diagnosis based on Inception-BLSTM was proposed, which combining the multiscale deep feature extraction ability of the Inception model and the advantage of bidirectional long short-term memory(BLSTM) neural network in sequence modeling. First, multiscale abstract features from the vibration signal of a rolling bearing were extracted using one-dimensional convolution kernels. Then the BLSTM was used for learning the time-dependence of features. Finally, the feature information was mapped to the corresponding fault mode through the full connection layer and the diagnosis result was obtained. The results show that the identification accuracy of the proposed method in multi-load scenarios is 99.6%, which has good load adaptability and anti-disturbance ability.

关 键 词：滚动轴承 故障诊断 Inception模型 双向长短时记忆(BLSTM) 

分 类 号：TH133.33[机械工程—机械制造及自动化] TH165.3