结合SE-VAE与M1DCNN的小样本数据下轴承故障诊断  

Bearing Fault Diagnosis Under Small Sample Data Based on SE-VAE and M1DCNN

作　　者：李梦男 李琨[1] 叶震 高宏宇 LI Mengnan;LI Kun;YE Zhen;GAO Hongyu(Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学信息工程与自动化学院,昆明650500

出　　处：《机械科学与技术》2024年第5期773-780,共8页Mechanical Science and Technology for Aerospace Engineering

基　　金：国家自然科学基金项目(82160787);昆明理工大学科技园有限公司下达项目(2018KF3)。

摘　　要：针对轴承故障诊断中故障样本数量少导致诊断正确率低的问题,提出了一种基于注意力机制变分自编码器(SE-VAE)和多尺度一维卷积神经网络(M1DCNN)的轴承故障诊断方法。将轴承数据集的训练集输入到SE-VAE中进行训练,生成与训练样本分布相似的生成样本,并添加到训练集中增加训练集的样本数量。将扩充后的训练集输入到M1DCNN中进行训练,随后将训练好的模型应用于测试集,输出故障诊断结果。实验结果表明,所提方法能够在不同负载的小样本轴承故障数据集上取得较好的故障诊断准确率。Aiming at the problem of low diagnostic accuracy caused by the small number of fault samples in bearing fault diagnosis,a new bearing fault diagnosis method based on attention mechanism variation autoencoder(SE-VAE)and multi-scale onedimensional convolutional neural network(M1DCNN)was proposed.Firstly,the training set of bearing data set is input into SE-VAE for training,generated samples with similar distribution to the training samples are obtained and added to the training set to increase the number of samples in the training set.Then,the extended training set is input into M1DCNN for training,and finally the trained model is applied to the test set to output the fault diagnosis results.Experimental results show that the proposed method can achieve better fault diagnosis accuracy on small sample bearing fault data sets with different loads.

关 键 词：轴承故障诊断 变分自编码器 注意力机制 多尺度一维卷积神经网络 小样本 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]