基于跳跃连接-CNN-BiLSTM的轴承故障诊断

doi:10.16180/j.cnki.issn1007-7820.2023.11.009

摘要/Abstract

摘要：

针对轴承故障数据含有噪声等无关成分及大部分轴承故障诊断方法不能充分利用故障数据的问题,文中提出一种基于跳跃连接-卷积神经网络-双向长短时记忆网络的故障诊断模型。利用短时傅里叶变换将原始振动信号转化为二维时频图像,用卷积神经网络和长短时记忆网络分别提取时频图像的空间和时间特征,并结合全连接层实现分类。添加软阈值注意力和跳跃连接结构,并滤除无关成分可充分利用不同网络层级的输出特征。采用MFPT(Machinery Failure Prevention Technology)轴承数据对所提诊断模型进行验证,实验结果表明该模型能够实现99.79%的故障识别准确率。

关键词: 故障诊断, 深度学习, 卷积神经网络, 双向长短时记忆网络, 跳跃连接, 注意力机制, 短时傅里叶变换, 软阈值

Abstract:

In view of the problem that bearing fault data contains irrelevant components such as noise and most bearing fault diagnosis methods cannot make full use of fault data, a fault diagnosis model based on skip connection- convolutional neural network- bidirectional long short-term memory network is proposed. The original vibration signal is converted into a two-dimensional time-frequency image using short-time Fourier transform, and the spatial and temporal features of the time-frequency image are extracted by convolutional neural network and long-short-term memory network respectively, and the classification is realized by combining the fully connected layer. Adding the structure of soft threshold attention and skip connection can make full use of the output features of different network levels while filtering out irrelevant components. The proposed diagnostic model is verified by MFPT(Machinery Failure Prevention Technology) bearing data, and the experimental results show that the proposed model can achieve a fault identification accuracy of 99.79%.

Key words: fault diagnosis, deep learning, convolutional neural network, bidirectional long short-term memory network, skip connection, attention mechanism, short-time Fourier transform, soft threshold

中图分类号:

TP389.1

于广增,张巧灵,周玉蓉. 基于跳跃连接-CNN-BiLSTM的轴承故障诊断[J]. 电子科技, 2023, 36(11): 56-65.

YU Guangzeng,ZHANG Qiaoling,ZHOU Yurong. Bearing Fault Diagnosis Based on SC-CNN-BiLSTM[J]. Electronic Science and Technology, 2023, 36(11): 56-65.

图/表 16

图1

图2

图3

图4

表1

图5

图6

表2

图7

图8

表3

表4

图9

图10

图11

表5

参考文献 27

[1]	张雪英, 栾忠权, 刘秀丽. 基于深度学习的滚动轴承故障诊断研究综述[J]. 设备管理与维修, 2017, 37(18):130-133.
	Zhang Xueying, Luan Zhongquan, Liu Xiuli. Summary of research on fault diagnosis of rolling bearing based on deep learning[J]. Plant Maintenance Engineering, 2017, 37 (18):130-133.
[2]	Boudiaf A, Moussaoui A, Dahane A, et al. A comparative study of various methods of bearing faults diagnosis using the Case Western Reserve University data[J]. Journal of Failure Analysis and Prevention, 2016, 16(2):271-284. doi: 10.1007/s11668-016-0080-7
[3]	沙美妤, 刘利国. 基于振动信号的轴承故障诊断技术综述[J]. 轴承, 2015(9):59-63.
	Sha Meiyu, Liu Liguo. Review on fault diagnosis technology for bearings based on vibration signal[J]. Bearing, 2015(9):59-63.
[4]	王兴龙, 郑近德, 潘海洋, 等. 基于MED与自相关谱峭度图的滚动轴承故障诊断方法[J]. 振动与冲击, 2020, 39(18):118-124.
	Wang Xinglong, Zheng Jinde, Pan Haiyang, et al. Fault diagnosis method for rolling bearings based on minimum entropy deconvolution and autograms[J]. Journal of Vibration and Shock, 2020, 39(18):118-124.
[5]	詹瀛鱼, 程良伦, 王涛. 解相关多频率经验模态分解的故障诊断性能优化方法[J]. 振动与冲击, 2020, 39(1):115-122.
	Zhan Yingyu, Cheng Lianglun, Wang Tao. Fault diagnosis performance optimization method based on decorrelation multi-frequency EMD[J]. Journal of Vibration and Shock, 2020, 39(1):115-122.
[6]	Wang W J, McFadden P D. Application of wavelets togearbox vibration signals for fault detection[J]. Journal of Sound and Vibration, 1996, 192(5):927-939. doi: 10.1006/jsvi.1996.0226
[7]	李恒, 张氢, 秦仙蓉, 等. 基于短时傅里叶变换和卷积神经网络的轴承故障诊断方法[J]. 振动与冲击, 2018, 37(19):124-131.
	Li Heng, Zhang Qing, Qin Xianrong, et al. Fault diagnosismethod for rolling bearings based on short-time Fourier transform and convolution neural network[J]. Journal of Vibration and Shock, 2018, 37(19):124-131.
[8]	王恒迪, 邓四二, 杨建玺, 等. 基于参数优化变分模态分解的滚动轴承早期故障诊断[J]. 振动与冲击, 2020, 39(23):38-46.
	Wang Hengdi, Deng Sier, Yang Jianxi, et al. Incipient fault diagnosis of rolling bearing based on VMD with parameters optimized[J]. Journal of Vibration and Shock, 2020, 39(23):38-46.
[9]	郑圆, 胡建中, 贾民平, 等. 一种基于参数优化变分模态分解的滚动轴承故障特征提取方法[J]. 振动与冲击, 2020, 39(21):195-202.
	Zheng Yuan, Hu Jianzhong, Jia Minping, et al. A method for rolling bearing fault feature extraction based on parametric optimization VMD[J]. Journal of Vibration and Shock, 2020, 39(21):195-202.
[10]	王奉涛, 薛宇航, 王雷, 等. 基于流形学习的滚动轴承故障盲源分离方法[J]. 振动、测试与诊断, 2020, 40(1):43-47.
	Wang Fengtao, Xue Yuhang, Wang Lei, et al. Blind source separation method for rolling bearing faults based on manifold learning[J]. Journal of Vibration,Measurement & Diagnosis, 2020, 40(1):43-47.
[11]	刘秀丽, 徐小力, 吴国新, 等. 基于变分模态分解的故障弱信息提取方法[J]. 华中科技大学学报(自然科学版), 2020, 48(7):1-6.
	Liu Xiuli, Xu Xiaoli, Wu Guoxin, et al. Extraction method of weak fault information based on variational mode decomposition[J]. Journal of Huazhong University of Science and Technology(Natural Science Edition), 2020, 48(7):1-6.
[12]	Bi F, Li X, Liu C, et al. Knock detection based on the optimized variational mode decomposition[J]. Measurement, 2019, 140(7):1-13. doi: 10.1016/j.measurement.2019.03.042
[13]	Hinton G E, Salakhutdinov R R. Reducing the dimensionality of data with neural networks[J]. Science, 2006, 313(5786):504-507. doi: 10.1126/science.1127647 pmid: 16873662
[14]	邓佳林, 邹益胜, 张笑璐, 等. 一种改进CNN在轴承故障诊断中的应用[J]. 现代制造工程, 2020, 41(4):142-147.
	Deng Jialin, Zou Yisheng, Zhang Xiaolu, et al. Application of an improved CNN in fault diagnosis of bearings[J]. Modern Manufacturing Engineering, 2020, 41(4):142-147.
[15]	赵小强, 张青青. 改进Alexnet的滚动轴承变工况故障诊断方法[J]. 振动、测试与诊断, 2020, 40(3):472-480.
	Zhao Xiaoqiang, Zhang Qingqing. Improved Alexnet based fault diagnosis method for rolling bearing undervariable conditions[J]. Journal of Vibration,Measurement & Diagnosis, 2020, 40(3):472-480.
[16]	肖雄, 王健翔, 张勇军, 等. 一种用于轴承故障诊断的二维卷积神经网络优化方法[J]. 中国电机工程学报, 2019, 39(15):4558-4568.
	Xiao Xiong, Wang Jianxiang, Zhang Yongjun, et al. A two-dimensional convolutional neural network optimization method for bearing fault diagnosis[J]. Proceedings of the CSEE, 2019, 39(15):4558-4568.
[17]	史光宇, 徐健, 杨强. 基于卷积神经网络的风电机组轴承机械故障智能诊断方法[J]. 华北电力大学学报(自然科学版), 2020, 47(4):71-79.
	Shi Guangyu, Xu Jian, Yang Qiang. Intelligent fault diagnosis on wind turbine bearing based on convolutional neural network[J]. Journal of North China Electric Power University(Natural Science Edition), 2020, 47(4):71-79.
[18]	刘炳集, 熊邦书, 欧巧凤, 等. 基于时频图和CNN的滚动轴承故障诊断[J]. 南昌航空大学学报(自然科学版), 2018, 32(2):86-91.
	Liu Bingji, Xiong Bangshu, Ou Qiaofeng, et al. Fault diagnosis of rolling bearing based on time-frequency representations and CNN[J]. Journal of Nanchang Hangkong University(Natural Sciences), 2018, 32(2):86-91.
[19]	范宇雪, 王江文, 梅桂明, 等. 基于Bi-LSTM的小样本滚动轴承故障诊断方法研究[J]. 噪声与振动控制, 2020, 40(4):103-108.
	Fan Yuxue, Wang Jiangwen, Mei Guiming, et al. Rolling bearing fault diagnosis method based on Bi-LSTM under less samples condition[J]. Noise and Vibration Control, 2020, 40(4):103-108.
[20]	Wang J, Wang D, Wang S, et al. Fault diagnosis of bearings based on multi-sensor information fusion and 2D convolutional neural network[J]. IEEE Access, 2021, 9(7):23717-23725. doi: 10.1109/Access.6287639
[21]	张龙, 甄灿壮, 熊国良, 等. 基于深度时频特征的机车轴承故障诊断[J]. 交通运输工程学报, 2021, 21(6):247-258.
	Zhang Long, Zhen Canzhuang, Xiong Guoliang, et al. Locomotive bearing fault diagnosis based on deep time-frequency features[J]. Journal of Traffic and Transportation Engineering, 2021, 21(6):247-258.
[22]	Xue F, Zhang W, Xue F, et al. A novel intelligent fault diagnosis method of rolling bearing based on two-stream feature fusion convolutional neural network[J]. Measurement, 2021, 176(10):109-226.
[23]	闫书豪, 乔美英. 基于一维WConv-BiLSTM的轴承故障诊断算法[J]. 电子科技, 2021, 34(4):75-82.
	Yan Shuhao, Qiao Meiying. Bearing fault diagnosis algorithm based on one-dimensional WConv-BiLSTM[J]. Electronic Science and Technology, 2021, 34(4):75-82.
[24]	张训杰, 张敏, 李贤均. 基于二维图像和CNN-BiGRU网络的滚动轴承故障模式识别[J]. 振动与冲击, 2021, 40(23):194-201.
	Zhang Xunjie, Zhang Min, Li Xianjun. Rolling bearing fault mode recognition based on 2D image and CNN-BiGRU[J]. Journal of Vibration and Shock, 2021, 40(23):194-201.
[25]	Bechhoefer E. Condition based maintenance fault database for testing of diagnostic and prognostics algorithms[EB/OL].(2020-02-27)[2021-11-15]https://mfpt.org/fault-data-sets/.
[26]	Woo S, Park J, Lee J, et al. CBAM:Convolutional block attention module[C]. Munich: European Conference on Computer Vision, 2008:3-19.
[27]	Zhao M, Zhong S, Fu X, et al. Deep residual shrinkage networks for fault diagnosis[J]. IEEE Transactions on Industrial Informatics, 2020, 16(7):4681-4690. doi: 10.1109/TII.9424

编号	网络层	输出维度
1	Input Layer	(32, 1, 224, 224)
2	Conv2D	(32, 32, 224, 224)
3	MaxPooling2D	(32, 32, 112, 112)
4	Conv2D	(32, 64, 112, 112)
5	MaxPooling2D	(32, 64, 56, 56)
6	软阈值注意力模块	(32, 64, 56, 56)
7	Flatten	(32, 200 704)
8	Linear	(32, 64)
9	Flatten	(32, 64, 3 136)
10	Bidirectional (LSTM)	(32, 64)
11	Add	(32, 64)
12	Dropout	(32, 64)
13	Linear	(32, 256)
14	Dropout	(32, 256)
15	Linear	(32, 6)

编号	轴承类型	训练集	测试集	验证集
1	正常状态	672	68	100
2	外圈故障	672	68	100
3	内圈故障	672	68	100
4	中间轴轴承	160	20	20
5	油泵轴轴承	160	20	20
6	行星轴承	160	20	20
-	总数	2 496	264	360

编号	准确率/%	编号	准确率/%	平均准确率/%
1	100.00	6	99.74
2	99.74	7	100.00
3	99.48	8	99.74	99.79
4	99.22	9	100.00
5	100.00	10	100.00

编号	模型	准确率/%
1	CNN-BiLSTM	95.31
2	CNN-BiLSTM(跳跃连接)	97.40
3	CNN-BiLSTM(软阈值注意力)	97.92
4	SC-CNN-BiLSTM	99.79

编号	模型	准确率/%
1	CNN	92.42
2	CNN-BiLSTM	95.31
3	CNN-RF	94.80
4	SC-CNN-BiLSTM	99.79