杜小磊1,2, 陈志刚1,2, 张楠1, 许旭1,2,等.SST和深度脊波网络在轴承故障诊断中的应用[J].河南理工大学学报（自然科学版）,2020,39(1):75-82.

DU Xiaolei1,2, CHEN Zhigang1,2, ZHANG Nan1, XU Xu1,2,et al.Applications of SST and deep ridgelet network in bearing fault diagnosis[J].Journal of Henan Polytechnic University(Natural Science) ,2020,39(1):75-82.

SST和深度脊波网络在轴承故障诊断中的应用

杜小磊1,2, 陈志刚1,2, 张楠1, 许旭1,2

1.北京建筑大学机电与车辆工程学院，北京 100044；2.北京市建筑安全监测工程技术研究中心，北京 100044

摘要:传统滚动轴承故障诊断方法过度依赖专家经验，故障特征提取与选取困难。为此，提出一种基于同步压缩小波变换（synchrosqueezed wavelet transform ， SST ）和深度脊波网络（deep ridgelet network， DRN）的滚动轴承故障诊断方法。首先，对轴承振动信号进行SST变换，得到信号时频图像；其次，将时频图像进行双向二维主成分分析压缩，然后将其作为DRN的输入，进行自动特征提取和故障识别。试验结果表明，该方法能够对滚动轴承进行多工况和多种故障程度的有效识别，特征提取能力和识别能力优于浅层人工神经网络、支持向量机等传统方法，以及深度信念网络、深度稀疏自编码器等深度学习模型。

关键词:滚动轴承;故障诊断;同步压缩小波变换;深度脊波网络

doi:10.16186/j.cnki.1673-9787.2020.1.10

基金项目:国家自然科学基金资助项目（51605022 ）；北京市教委科技计划项目（SQKM201710016014 ）；北京市优秀人才培养项目 （2013D005017000013 ）；北京市属高校基本科研业务费专项项目（X18217）

收稿日期:2019/04/19

修回日期:2019/05/29

出版日期:2020/01/15

Applications of SST and deep ridgelet network in bearing fault diagnosis

DU Xiaolei1,2, CHEN Zhigang1,2, ZHANG Nan1, XU Xu1,2

1.Beijing University of Civil Engineering and Architecture， Beijing  100044 ， China；2.Beijing Engineering Research Center of Monitoring for Construction Safety， Beijing  100044 ， China

Abstract:The traditional fault diagnosis methods of rolling bearing have such shortcomings as largely dependent on expert prior knowledge， difficulty in fault feature extraction and selection. So that a method based on synchrosqueezed wavelet transforms （ SST ） and deep ridgelet network （ DRN） was proposed. Firstly，the bearing vibration signals were transformed by SST to get the time-frequency images. Secondly，the time-frequency images were compressed by two-directional two dimensional principal components analysis（ TD-2DPCA）.Finally，the compressed time-frequency images were sent to DRN for automatic feature extraction and fault identification. The experimental results showed that the proposed method could effectively identify the bearing faults under multiple working conditions and multiple fault severities. The proposed method had better ability of feature extraction and recognition than traditional methods such as shallow ANN，SVM， and deep learning models such as deep belief network，deep sparse auto-encoder.

Key words:rolling bearing;fault diagnosis;synchrosqueezed wavelet transform;deep ridgelet network

SST和深度脊波网络在轴承故障诊断中的应用_杜小磊.pdf