基于残差网络的汽车轴承“亚健康”诊断技术  

Diagnosis Technology for“Sub health”of Automotive Bearings Based on Residual Networks

作　　者：潘汉平 张计军 葛家琪 黄俊蓉 PAN Hanping;ZHANG Jijun;GE Jiaqi;HUANG Junrong(Guangxi ecological engineering Vocational and Technical College,Liuzhou,Guangxi Zhuang Autonomous Region,545004 China)

机构地区：[1]广西生态工程职业技术学院,广西柳州545004

出　　处：《汽车知识》2023年第8期157-160,共4页Auto Know

基　　金：2023年度广西高校中青年教师科研基础能力提升项目一般项目“基于深度学习的电子诊断技术的新能源汽车检修技术研究”(项目编号:2023KY1264)。

摘　　要：为了实现汽车轴承故障的自动诊断,设计了一种卷积层数较少的残差卷积神经网络模型。利用pytorch框架搭建了一个六层的残差卷积神经网络,以采集的汽车轴承振动数据集为研究对象。其在测试集上的准确率达到100%,相比于传统机器学习中的支持向量机算法,省去了人工特征提取的步骤,极大地提高了技术落地应用的效率。提出的方法对汽车轴承“亚健康”状态的诊断有很好的实际应用价值。In order to achieve automatic diagnosis of automotive bearing faults,a residual convolutional neural network model with fewer convolutional layers was designed.A six layer residual convolutional neural network was constructed using the Pytorch framework,with the collected automotive bearing vibration dataset as the research object.Its accuracy on the test set reaches 100%,which saves the step of manual feature extraction compared to traditional support vector machine algorithms in machine learning,greatly improving the efficiency of technology implementation and application.The proposed method has good practical application value for diagnosing the“sub health”state of automotive bearings.

关 键 词：残差网络 汽车轴承 卷积神经网络 亚健康 

分 类 号：U467[机械工程—车辆工程]