基于冲击特征提取的旋转机械智能故障诊断

胡爱军; 孙俊豪; 邢磊; 向玲

doi:10.13224/j.cnki.jasp.20220106

基于冲击特征提取的旋转机械智能故障诊断

doi: 10.13224/j.cnki.jasp.20220106

华北电力大学能源动力与机械工程学院机械工程系河北省电力机械装备健康维护与失效预防实验室，河北保定 071003

基金项目: 国家自然科学基金（52175092）

详细信息

作者简介:
胡爱军（1971-），男，教授、博士生导师，博士，主要从事机械设备状态监测与故障诊断研究。E-mail：bdlaohu@126.com

通讯作者:
孙俊豪（1998-），男，硕士，主要从事旋转机械设备状态监测与诊断研究。E-mail：799388772@qq.com

中图分类号: V240.2；TH132；TH133.33
计量
- 文章访问数: 291
- HTML浏览量: 121
- PDF量: 63
- 被引次数: 0
出版历程
- 收稿日期: 2022-03-04
- 网络出版日期: 2023-04-11

Intelligent fault diagnosis of rotating machinery based on impact feature extraction

Hebei Key Laboratory of Electric Machinery Health Maintenance and Failure Prevention， Department of Mechanical Engineering，School of Energy Power and Mechanical Engineering， North China Electric Power University，Baoding Hebei 071003，China

摘要

摘要:
针对齿轮、轴承故障，提出了基于冲击特征提取胶囊网络的旋转机械智能故障诊断模型。在胶囊网络的构架基础上，将原始故障振动信号作为输入，通过构造首层小波核卷积层，针对性提取冲击故障特征，提高深度学习网络特征提取的可解释性。在小波核卷积层之后扩展一层卷积层，强化首层小波核卷积层提取的特征，将强化的特征经初级胶囊层、数字胶囊层输出分类结果，从而构造了“端到端”的小波卷积胶囊网络模型。通过对各层提取的特征可视化分析，证明了该模型对故障振动信号的冲击特征具有良好的提取能力。3个不同实验平台的数据集验证结果表明不同故障类型、不同故障程度的齿轮及轴承的识别精度最高可达到100%，并具有良好的泛化能力。
- 冲击特征 /
- 胶囊网络 /
- 小波卷积核 /
- 齿轮箱 /
- 滚动轴承
Abstract:
In view of the gear and bearing faults, an intelligent fault diagnosis model of rotating machinery based on impact feature extraction capsule network was proposed. Based on the structure of the capsule network, the original fault vibration signal was taken as the input, and the first wavelet kernel convolution layer was constructed to extract the impact fault features, so as to improve the interpretability of the feature extraction of the deep learning network. After the wavelet kernel convolution layer, a convolution layer was extended to strengthen the features extracted by the first wavelet kernel convolution layer. The enhanced features were processed through the primary and digital capsule layers to output the final diagnosis result. Therefore, an “end-to-end” wavelet convolution capsule network model was constructed. The impact feature extraction ability of the proposed model was verified through the feature visualization analysis for each layer. The datasets verification results of three different experimental platforms show that the recognition accuracy of gears and bearings with different fault types and fault degrees can reach 100% at most, presenting good generalization ability.
- impact feature /
- capsule network /
- wavelet convolution kernel /
- gearbox /
- rolling bearing

HTML

图 1 冲击特征提取胶囊网络模型图

Figure 1. Impact feature extraction capsule network model diagram

下载: 全尺寸图片幻灯片

图 2 智能诊断流程图

Figure 2. Intelligent diagnosis flow chart

下载: 全尺寸图片幻灯片

图 3 华北电力大学齿轮箱实验台

Figure 3. Gear box test bench of North China Electric Power University

下载: 全尺寸图片幻灯片

图 4 齿轮故障类型及其时域波形

Figure 4. Gear fault type and the time domain waveform

下载: 全尺寸图片幻灯片

图 5 华北电力大学齿轮箱数据集CAPS、LW-CAPS、C-CAPS和IF-CAPS模型每层的t-SNE可视化图

Figure 5. Visualization diagram of t-SNE for each layer of CAPS, LW-CAPS, C-CAPS and IF-CAPS models based on gearbox dataset of North China Electric Power University

下载: 全尺寸图片幻灯片

图 6 凯斯轴承实验台

Figure 6. Case bearing test bench

下载: 全尺寸图片幻灯片

图 7 不同来源的两组数据在4种模型下准确率对比图

Figure 7. Comparison of accuracy of two groups of data from different sources under four models

下载: 全尺寸图片幻灯片

图 8 东南大学齿轮箱数据集IF-CAPS模型的t-SNE聚类结果

Figure 8. t-SNE clustering diagram of IF-CAPS model based on the gearbox dataset of Southeast University

下载: 全尺寸图片幻灯片

图 9 凯斯西储大学轴承数据集IF-CAPS模型的t-SNE聚类结果

Figure 9. t-SNE clustering diagram of IF-CAPS model based on the bearing dataset of Case Western Reserve University

下载: 全尺寸图片幻灯片

表 1 华北电力大学齿轮箱样本集

Table 1. Gearbox sample set of North China Electric Power University

标签	故障类型	训练样本数	测试样本数
0	正常状态	420	105
1	大齿轮点蚀故障	420	105
2	大齿轮断齿故障	420	105
3	小齿轮磨损故障	420	105
4	大齿轮断齿+小齿轮磨损	420	105
5	大齿轮点蚀+小齿轮磨损	420	105

下载: 导出CSV

表 2 IF-CAPS模型具体参数

Table 2. Specific parameters of IF-CAPS model

层类型	核尺寸	步长	输出尺寸
小波核卷积层	15×1	1	64×64×387
卷积层2	15×1	1	64×256×373
初级胶囊层	9×1	2	5856×8
数字胶囊层			6×16

下载: 导出CSV

表 3 3种模型测试精度

Table 3. Test accuracy of three models

方法	精度/%	收敛迭代次数
CNN	88.57	80
CAPS	91.90	100
LW-CAPS	98.64	45
IF-CAPS	100.00	20

下载: 导出CSV

表 4 东南大学齿轮箱样本集

Table 4. Gearbox sample set of Southeast University

标签	故障类型	训练样本数	测试样本数
0	正常状态	1 000	200
1	缺齿故障	1 000	200
2	齿根断裂故障	1 000	200
3	齿面磨损故障	1 000	200
4	裂纹故障	1 000	200

下载: 导出CSV

表 5 美国凯斯西储大学轴承样本集

Table 5. Case Western Reserve University bearing sample set

标签	故障类型	故障直径	训练样本数	测试样本数
0	正常状态	0	480	120
1	内圈故障	0.007	480	120
2	内圈故障	0.014	480	120
3	内圈故障	0.021	480	120
4	外圈故障	0.007	480	120
5	外圈故障	0.014	480	120
6	外圈故障	0.021	480	120
7	滚动体故障	0.007	480	120
8	滚动体故障	0.014	480	120
9	滚动体故障	0.021	480	120

下载: 导出CSV

表 6 凯斯轴承数据集下不同模型精确度对比

Table 6. Accuracy comparison of different models under Case bearing data set

方法	精度/%
文献[10]	95.50
文献[16]	94.10
文献[24]	94.63
IF-CAPS	100.00

下载: 导出CSV

参考文献(24)

[1]	刘东东,程卫东,万广通. 基于故障特征趋势线模板的滚动轴承故障诊断[J]. 机械工程学报,2017,53(9): 83-91. doi: 10.3901/JME.2017.09.083 LIU Dongdong,CHENG Weidong,WAN Guangtong. Rolling bearing fault diagnosis based on fault characteristic trend line template[J]. Journal of Mechanical Engineering,2017,53(9): 83-91. (in Chinese) doi: 10.3901/JME.2017.09.083
[2]	陈是扦,彭志科,周鹏. 信号分解及其在机械故障诊断中的应用研究综述[J]. 机械工程学报,2020,56(17): 91-107. doi: 10.3901/JME.2020.17.091 CHEN Shiqian,PENG Zhike,ZHOU Peng. Review of signal decomposition theory and its applications in machine fault diagnosis[J]. Journal of Mechanical Engineering,2020,56(17): 91-107. (in Chinese) doi: 10.3901/JME.2020.17.091
[3]	童靳于,罗金,郑近德. 基于增强深度自编码网络的滚动轴承故障诊断方法[J]. 中国机械工程,2021,32(21): 2617-2624. TONG Jinyu,LUO Jin,ZHENG Jinde. Fault diagnosis method of rolling bearing based on enhanced depth self coding network[J]. China Mechanical Engineering,2021,32(21): 2617-2624. (in Chinese)
[4]	杨宇,罗鹏,甘磊,等. SADBN及其在滚动轴承故障分类识别中的应用[J]. 振动与冲击,2019,38(15): 11-16, 26. doi: 10.13465/j.cnki.jvs.2019.15.002 YANG Yu,LUO Peng,GAN Lei,et al. SADBN and its application in rolling bearing fault classification and identification[J]. Vibration and Shock,2019,38(15): 11-16, 26. (in Chinese) doi: 10.13465/j.cnki.jvs.2019.15.002
[5]	胡茑庆,陈徽鹏,程哲,等. 基于经验模态分解和深度卷积神经网络的行星齿轮箱故障诊断方法[J]. 机械工程学报,2019,55(7): 9-18. doi: 10.3901/JME.2019.07.009 HU Fuqing,CHEN Huipeng,CHENG Zhe,et al. Fault diagnosis method of planetary gearbox based on empirical mode decomposition and deep convolution neural network[J]. Journal of Mechanical Engineering,2019,55(7): 9-18. (in Chinese) doi: 10.3901/JME.2019.07.009
[6]	李俊,刘永葆,余又红. 卷积神经网络和峭度在轴承故障诊断中的应用[J]. 航空动力学报,2019,34(11): 2423-2431. LI Jun,LIU Yongbao,YU Youhong. Application of convolution neural network and kurtosis in bearing fault diagnosis[J]. Journal of Aerospace Power,2019,34(11): 2423-2431. (in Chinese)
[7]	张淑清,胡永涛,姜安琦,等. 基于双树复小波和深度信念网络的轴承故障诊断[J]. 中国机械工程,2017,28(5): 532-536, 543. doi: 10.3969/j.issn.1004-132X.2017.05.005 ZHANG Shuqing,HU Yongtao,JIANG Anqi,et al. Bearing fault diagnosis based on dual tree complex wavelet and deep belief network[J]. China Mechanical Engineering,2017,28(5): 532-536, 543. (in Chinese) doi: 10.3969/j.issn.1004-132X.2017.05.005
[8]	宫文峰,陈辉,张泽辉,等. 基于改进卷积神经网络的滚动轴承智能故障诊断研究[J]. 振动工程学报,2020,33(2): 400-413. GONG Wenfeng,CHEN Hui,ZHANG Zehui,et al. Research on intelligent fault diagnosis of rolling bearing based on improved convolutional neural network[J]. Journal of Vibration Engineering,2020,33(2): 400-413. (in Chinese)
[9]	时培明,梁凯,赵娜,等. 基于深度学习特征提取和粒子群支持向量机状态识别的齿轮智能故障诊断[J]. 中国机械工程,2017,28(9): 1056-1061, 1068. doi: 10.3969/j.issn.1004-132X.2017.09.009 SHI Peiming,LIANG Kai,ZHAO Na,et al. Gear intelligent fault diagnosis based on deep learning feature extraction and particle swarm support vector machine state recognition[J]. China Mechanical Engineering,2017,28(9): 1056-1061, 1068. (in Chinese) doi: 10.3969/j.issn.1004-132X.2017.09.009
[10]	杨平,苏燕辰. 基于卷积门控循环网络的滚动轴承故障诊断[J]. 航空动力学报,2019,34(11): 2432-2439. YANG Ping,SU Yanchen. Rolling bearing fault diagnosis based on convolution gated cyclic network[J]. Journal of Aerospace Power,2019,34(11): 2432-2439. (in Chinese)
[11]	卞景艺,刘秀丽,徐小力,等. 基于多尺度深度卷积神经网络的故障诊断方法[J]. 振动与冲击,2021,40(18): 204-211. BIAN Jingyi,LIU Xiuli,XU Xiaoli,et al. Fault diagnosis method based on multi-scale depth convolution neural network[J]. Vibration and Shock,2021,40(18): 204-211. (in Chinese)
[12]	RAVANELLI M, BENGIO Y. Speaker recognition from raw waveform with sincnet [C]//2018 IEEE Spoken Language Technology Workshop (SLT). Athens, Greece: IEEE, 2018: 1021-1028.
[13]	SABOUR S, FROSST N, GEOFFREY E H. Dynamic routing between capsules[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook, NY, US: Curran Associates, 2017: 3859-3869.
[14]	孙岩,彭高亮. 改进胶囊网络的滚动轴承故障诊断方法[J]. 哈尔滨工业大学学报,2021,53(1): 23-28. doi: 10.11918/202004163 SUN Yan,PENG Gaoliang. Fault diagnosis method of rolling bearing based on improved capsule network[J]. Journal of Harbin University of Technology,2021,53(1): 23-28. (in Chinese) doi: 10.11918/202004163
[15]	王超群,李彬彬,焦斌. 基于门控循环单元胶囊网络的滚动轴承故障诊断[J]. 轴承,2021,64(5): 56-62. doi: 10.19533/j.issn1000-3762.2021.05.010 WANG Chaoqun,LI Binbin,JIAO Bin. Fault diagnosis of rolling bearing based on gated circulating unit capsule network[J]. Bearing,2021,64(5): 56-62. (in Chinese) doi: 10.19533/j.issn1000-3762.2021.05.010
[16]	梁海涛,王立纲,王亮. 基于ARCN模型的轴承故障诊断[J]. 航空动力学报,2021,36(9): 1793-1803. doi: 10.13224/j.cnki.jasp.20210186 LIANG Haitao,WANG Ligang,WANG Liang,et al. Bearing fault diagnosis based on ARCN model[J]. Journal of Aerospace Power,2021,36(9): 1793-1803. (in Chinese) doi: 10.13224/j.cnki.jasp.20210186
[17]	顾晓辉,杨绍普,刘永强,等. 基于多目标交叉熵优化的轮对轴承故障特征提取方法[J]. 机械工程学报,2018,54(4): 285-292. doi: 10.3901/JME.2018.04.285 GU Xiaohui,YANG Shaopu,LIU Yongqiang,et al. Fault feature extraction method of wheelset bearing based on multi-objective cross entropy optimization[J]. Journal of Mechanical Engineering,2018,54(4): 285-292. (in Chinese) doi: 10.3901/JME.2018.04.285
[18]	臧怀刚,刘子豪,李玉奎. 基于形态滤波和Laplace小波的轴承故障诊断[J]. 中国机械工程,2016,27(9): 1198-1203. doi: 10.3969/j.issn.1004-132X.2016.09.010 ZANG Huaigang,LIU Zihao,LI Yukui. Bearing fault diagnosis based on morphological filtering and Laplace wavelet[J]. China Mechanical Engineering,2016,27(9): 1198-1203. (in Chinese) doi: 10.3969/j.issn.1004-132X.2016.09.010
[19]	WANG X,ZHANG W. Anti-occlusion face recognition algorithm based on a deep convolutional neural network[J]. Computers and Electrical Engineering,2021,96: 107461. doi: 10.1016/j.compeleceng.2021.107461
[20]	叶壮,余建波. 基于多通道加权卷积神经网络的齿轮箱振动信号特征提取[J]. 机械工程学报,2021,57(1): 110-120. doi: 10.3901/JME.2021.01.110 YE Zhuang,YU Jianbo. Feature extraction of gearbox vibration signal based on multi-channel weighted convolutional neural network[J]. Journal of Mechanical Engineering,2021,57(1): 110-120. (in Chinese) doi: 10.3901/JME.2021.01.110
[21]	KINGMA D, BA J. Adam: a method for stochastic optimization[R]. San Diego, CA, US: 3rd International Conference for Learning Representations, 2015.
[22]	SHAO S Y,MCALEER S,YAN R,et al. Highly accurate machine fault diagnosis using deep transfer learning[J]. IEEE Transactions on Industrial Informatics,2019,15(4): 2446-2455. doi: 10.1109/TII.2018.2864759
[23]	SMITH W,RANDALL R. Rolling element bearing diagnostics using the Case Western Reserve University data: a benchmark study[J]. Mechanical Systems and Signal Processing,2015,64/65: 100-131. doi: 10.1016/j.ymssp.2015.04.021
[24]	余萍,曹洁. 优化堆叠降噪自动编码器滚动轴承故障诊断[J]. 太阳能学报,2021,42(11): 307-314. doi: 10.19912/j.0254-0096.tynxb.2019-1302 YU Ping,CAO Jie. Optimized stacking noise reduction automatic encoder rolling bearing fault diagnosis[J]. Journal of Solar Energy,2021,42(11): 307-314. (in Chinese) doi: 10.19912/j.0254-0096.tynxb.2019-1302