基于MSCNN-BiGRU的轴承故障诊断模型  

Bearing Fault Diagnosis Model Based on MSCNN-BiGRU

作　　者：徐紫薇 李彦锋 黄洪钟[1,2] 尉询楷 王浩[3] Xu Ziwei;Li Yanfeng;Huang Hongzhong;Wei Xunkai;Wang Hao(School of Mechanical and Electrical Engineering,University of Electronic Science and Technology of China,Chengdu 611731,China;Center for System Reliability and Safety,University of Electronic Science and Technology of China,Chengdu 611731,China;Beijing Aeronautical Technology Research Center,Beijing 100076,China)

机构地区：[1]电子科技大学机械与电气工程学院,成都611731 [2]电子科技大学系统可靠性与安全性研究中心,成都611731 [3]北京航空工程技术研究中心,北京100076

出　　处：《质量与可靠性》2024年第4期16-24,共9页Quality and Reliability

摘　　要：针对轴承故障诊断中由于信号存在噪声干扰等因素影响,导致诊断结果可信度不高和不准确等问题,提出了基于MSCNN-BiGRU的轴承故障诊断模型。首先,引入多尺度卷积神经网络(MSCNN),分3条支路多尺度并行提取轴承振动信号的退化特征,并且在每条支路卷积后加上双向门控循环单元(BiGRU)进一步提取信号的时序特征;然后,构建MSCNN-BiGRU的故障诊断模型,对轴承进行故障诊断;最后,开展噪声干扰下的消融实验和对比实验,在不同数据集中测试并验证模型的性能与鲁棒性,并与其他方法进行对比分析。结果表明,该模型能有效提高诊断精度,具有较强的泛化能力。A bearing fault diagnosis model based on MSCNN-BiGRU was proposed to solve the problem of low reliability and inaccuracy due to noise interference in bearing fault diagnosis.Firstly,a Multi-scale Convolutional Neural Network(MSCNN)was introduced to extract the degradation characteristics of bearing vibration signals in parallel with three branches,and a Bidirectional Gated Recurrent Unit(BiGRU)was added to the convolutional layer of each branch to further extract the signal temporal characteristics.Then,a fault diagnosis model of MSCNN-BiGRU was constructed for bearing fault diagnosis.Finally,ablation experiments and comparison experiments under noise interference were carried out to test and verify the performance and robustness of the model in different datasets.Comparison and analysis were conducted with other methods.The results show that the model proposed in this paper can effectively improve the diagnostic accuracy and has strong generalization ability.

关 键 词：轴承 故障诊断 多尺度卷积神经网络 双向门控循环单元 

分 类 号：TH133.3[机械工程—机械制造及自动化]