基于无监督变分自编码器的电机轴承故障识别研究  

Research on Motor Bearing Fault Identification Based on Unsupervised Variation Autoencoder

作　　者：汤瑞 Tang Rui(Department of Automotive Engineering,Kaifeng Technician College,Kaifeng 475000,China)

机构地区：[1]开封技师学院汽车工程系,河南开封475000

出　　处：《防爆电机》2025年第2期13-16,共4页Explosion-proof Electric Machine

基　　金：河南省高等学校重点科研项目(23B510012)。

摘　　要：电机的运行稳定性可以通过监测轴承的振动情况进行判定。为了进一步提高电机轴承故障识别能力,设计了一种基于无监督自编码器的电机轴承故障识别方法。在特征加强下使得数据长度大幅减小,在很大程度上增强了算法的识别效率及精准性。研究结果表明:与无监督自编码器相比,无监督变分自编码器将变分结果当作全新样本的故障辨识准确率更低,表现出来更大的优越性在最大偏移点数取值范畴逐渐扩大时,辨识准确率不断提高,在其数值为150左右可实现接近100%的辨别准确率结果。该研究有助于提高电机轴承的故障识别能力,也可拓宽到其它的机械传动领域。The running stability of electrical motor can be determined by monitoring the vibration condition of the bearings.In order to further improve the identification ability of motor bearing fault,a method of identifying motor bearing fault by the unsupervised autoencoder is designed,which greatly reduces the data length with the enhancement of features and enhances the identification efficiency and accuracy of the algorithm in a large extent.The research results show that the unsupervised variation autoencoder has lower fault identification accuracy compared with the unsupervised autoencoder when the variation results are taken as the new samples.It has a greater advantages that the identification accuracy is constantly improved when the value range of the maximum offset points is gradually expanded.When the value is about 150,the identification accuracy can be close to 100%.This research result is helpful to improve the fault identification ability of motor bearings,and can also be extended to other mechanical transmission fields.

关 键 词：电机轴承 无监督自编码器 故障识别 偏移点 

分 类 号：TM307[电气工程—电机]