基于时频图与改进ResNet-18网络的滚动轴承故障诊断  

Bearing Fault Diagnosis Basedon Time-frequency Diagramand Improved Resnet-18 Network

作　　者：郝江涛 邓宇轩 朱命国 杨纪民 HAO Jiangtao;DENG Yuxuan;ZHU Mingguo;YANG Jimin(School of Mechanical and Electrical Engineering,Henan Institute of Science and Technology,Xinxiang 453000,China;School of Journalism and Communication,Pingdingshan University,Pingdingshan 467000,China;Henan Vibration Machinery Product Quality Supervision and Inspection Center,Xinxiang 453700,China)

机构地区：[1]河南科技学院机电工程学院,河南新乡453000 [2]平顶山学院新闻与传播学院,河南平顶山467000 [3]河南省振动机械产品质量监督检验中心,河南新乡453700

出　　处：《新乡学院学报》2025年第3期72-76,共5页Journal of Xinxiang University

摘　　要：针对传统的滚动轴承故障检测方法准确率和效率低下的问题,提出一种基于时频图和改进ResNet-18网络的滚动轴承故障检测方法。首先由连续小波变换将滚动轴承的一维故障信号转换为时频图,其次选择ResNet-18网络作为骨干网络,同时引入Ghost模块和SE模块对ResNet-18网络进行改进,使网络在更加轻量化的同时能够提高诊断准确率,最后使用凯斯西储大学的公开轴承数据集对模型进行验证。研究结果表明,所提出模型的故障诊断准确率高。To solve the problem of low accuracy and efficiency of traditional bearing fault detection methods,a bearing fault detection method based on time-frequency graph and improved ResNet-18 network was proposed.The continuous wavelet transform convertedthe one-dimensional fault signal of the rolling bearing into a time-frequency graph.To avoid the problem of gradient vanishing during training,the ResNet-18 network was selected as the backbone network,and the Ghost module and the SE module were introduced to improve the ResNet-18 network,so that the network can be lightweight and improve the diagnostic accuracy.Finally,the modelwas validated by using the public bearing dataset of Case Western Reserve University,and the results show that the proposed modelhad highaccuracyin fault diagnosis.

关 键 词：轴承故障诊断 ResNet网络 Ghost模块 压缩激励机制 

分 类 号：TH133.33[机械工程—机械制造及自动化]