基于TET-MResNet的通风机轴承故障诊断  

Fault Diagnosis of Ventilator Bearing Based on TET-MResNet

作　　者：胥海东 罗川 王彦召 周强 苏莉君 Xu Haidong;Luo Chuan;Wang Yanzhao;Zhou Qiang;Su Lijun(Maiduoshan Coal Mine,Ningxia Coal Industry Co.,Ltd.,CHN Energy Group,Yinchuan 750408,China;Chongqing Research Institute,China Coal Technology and Engineering Group,Chongqing 400039,China)

机构地区：[1]国家能源集团宁夏煤业有限责任公司麦垛山煤矿,银川750408 [2]中煤科工集团重庆研究院有限公司,重庆400039

出　　处：《煤矿机械》2024年第9期149-152,共4页Coal Mine Machinery

基　　金：中煤科工集团重庆研究院有限公司自立项目(2022ZDXM01)。

摘　　要：针对矿用通风机滚动轴承早期故障信号难以有效提取进而导致故障诊断精度低的问题,提出了一种基于瞬态提取变换(TET)和改进残差网络(ResNet)的故障诊断方法(TET-MResNet)。利用TET提取信号的瞬态时频特征作为特征输入,针对时频特征的高稀疏性,为从全局角度丰富特征信息,加入多尺度特征提取(MFE)模块对ResNet进行改进,最后将故障数据输入模型中进行分类识别。实验结果表明,该方法取得了99.96%的识别准确率,可以实现通风机轴承故障的有效诊断。Aiming at the problem that early fault signals of mining fan rolling bearings are difficult to effectively extract,resulting in low fault diagnosis accuracy,a fault diagnosis method(TET-MResNet)based on transient extraction transform(TET)and improved residual network(ResNet)was proposed.TET was used to extract the transient time-frequency features of the signal as feature input.In view of the high sparsity of time-frequency features,in order to enrich feature information from a global perspective,a multi-scale feature extraction(MFE)module was added to improve ResNet.Finally,Fault data was input into the model for classification and identification.Experimental results show that this method achieves a recognition accuracy of 99.96%,and effective diagnosis of fan bearing faults can be achieved.

关 键 词：故障诊断 瞬态信号 时频变换 深度学习 

分 类 号：TD724[矿业工程—矿井通风与安全]