基于MDAM-GhostCNN的滚动轴承故障诊断方法

郭俊锋; 谭宝宏; 王智明

doi:10.13700/j.bh.1001-5965.2023.0224

基于MDAM-GhostCNN的滚动轴承故障诊断方法

doi: 10.13700/j.bh.1001-5965.2023.0224

兰州理工大学机电工程学院，兰州 730050

基金项目: 国家自然科学基金(51465034)

详细信息

通讯作者:
E-mail：1426779121@qq.com

中图分类号: TH133.33
计量
- 文章访问数: 327
- HTML全文浏览量: 70
- PDF下载量: 2
- 被引次数: 0
出版历程
- 收稿日期: 2023-05-04
- 录用日期: 2023-06-20
- 网络出版日期: 2023-07-07
- 整期出版日期: 2025-04-30

Fault diagnosis method of rolling bearing based on MDAM-GhostCNN

School of Mechanical and Electronical Engineering，Lanzhou University of Technology，Lanzhou 730050，China

Funds: National Natural Science Foundation of China (51465034)

More Information

Corresponding author: E-mail：1426779121@qq.com

摘要

摘要:
针对传统故障诊断方法特征提取不充分、计算复杂及在变工况下识别准确率低的问题，提出一种基于混合域注意力机制(MDAM)-GhostCNN的滚动轴承故障诊断方法。采用马尔可夫转移场(MTF)将轴承振动信号转化为具有时间相关性的二维特征图；利用Ghost卷积计算精简的优点，构造出GhostCNN；设计一种MDAM，使网络从通道和空间2个维度充分捕获特征信息，实现特征通道间相互依赖的同时让网络有效关注特征空间信息。由此，构建出MDAM-GhostCNN模型。将MTF二维特征图输入到MDAM-GhostCNN模型中进行训练并输出诊断结果。采用凯斯西储大学和江南大学(JNU)轴承数据集进行实验验证，并对其数据集进行加噪处理。结果表明：在变工况下，所建模型有着更高的识别准确率、抗噪性能和泛化性能。
- 滚动轴承 /
- 故障诊断 /
- 马尔可夫转移场 /
- Ghost卷积 /
- 注意力机制
Abstract:
A rolling bearing fault diagnostic approach based on mixture domain attention mechanism (MDAM)-GhostCNN is developed to address the issues of inadequate feature extraction, complicated computation, and low recognition accuracy under varied working conditions in conventional fault detection methods. First of all, the Markov transfer field (MTF) is used to transform the bearing vibration signal into a two-dimensional feature graph with time correlation. Secondly, taking advantage of the simplification of Ghost convolution calculation, GhostCNN is constructed. Then, a MDAM is designed, which makes the network fully capture the feature information from the two dimensions of channel and space, and makes the network pay attention to the feature space information effectively while realizing the interdependence between the feature channels, and construct the MDAM-GhostCNN model. Finally, the MTF two-dimensional feature map is input into the MDAM-GhostCNN model for training and output diagnosis results. Experimental verification and noise processing were performed on the bearing data sets from Jiang Nan University (JNU) and Case Western Reserve University. The results show that under variable working conditions, the constructed model has higher recognition accuracy, noise immunity and generalization performance.
- rolling bearing /
- fault diagnosis /
- Markov transfer field /
- Ghost convolution /
- attention mechanism

HTML全文

图 1 特征生成过程

Figure 1. Feature generation process

下载: 全尺寸图片幻灯片

图 2 改进通道注意力机制

Figure 2. Improved channel attention mechanism

下载: 全尺寸图片幻灯片

图 3 空间注意力机制

Figure 3. Spatial attention mechanism

下载: 全尺寸图片幻灯片

图 4 混合域注意力机制结构

Figure 4. Mixed domain attention mechanism structure

下载: 全尺寸图片幻灯片

图 5 MDAM-GhostCNN故障诊断模型

Figure 5. MDAM-GhostCNN fault diagnosis model

下载: 全尺寸图片幻灯片

图 6 MDAM-GhostCNN模型故障诊断流程图

Figure 6. MDAM-GhostCNN model fault diagnosis flow chart

下载: 全尺寸图片幻灯片

图 7 轴承实验装置示意图

Figure 7. Schematic diagram of bearing experimental device

下载: 全尺寸图片幻灯片

图 8 不同参数对诊断精度的影响

Figure 8. Effects of different parameters on diagnostic accuracy

下载: 全尺寸图片幻灯片

图 9 CWRU^[23]轴承数据集在变负载下不同模型的分类精度

Figure 9. Classification accuracy of CWRU^[23] bearings data set for different models under variable loads

下载: 全尺寸图片幻灯片

图 10 故障分类结果的混淆矩阵

Figure 10. Confusion matrix of fault classification results

下载: 全尺寸图片幻灯片

图 11 不同模型的信噪比

Figure 11. Signal-to-noise ratio of different models

下载: 全尺寸图片幻灯片

图 12 模型训练过程可视化

Figure 12. Visualization of model training process

下载: 全尺寸图片幻灯片

图 13 JNU轴承数据模型训练可视化

Figure 13. Visualization of JNU bearing data model training

下载: 全尺寸图片幻灯片

图 14 JNU数据集轴承变转速下不同模型的分类精度

Figure 14. Classification accuracy of different models in bearing variable working conditions of JNU data set

下载: 全尺寸图片幻灯片

图 15 JNU数据集上不同模型的信噪比

Figure 15. Signal-to-noise ratio of different models on the JNU dataset

下载: 全尺寸图片幻灯片

表 1 MDAM-GhostCNN结构

Table 1. MDAM-GhostCNN structure

特征层	卷积核数量	卷积核大小	输出大小
输入			(128，128，1)
卷积层	32	5×5	(128，128，32)
GN			(128，128，32)
最大池化1	32	2×2	(64，64，32)
Ghost Conv 1	32	3×3	(64，64，32)
GN			(64，64，32)
最大池化2	32	2×2	(32，32，32)
Ghost Conv 2	32	3×3	(32，32，32)
GN			(32，32，32)
最大池化3		2×2	(16，16，32)
MDAM			(16，16，32)
全局平均池化			(1，1，32)
分类器	7		(7)

下载: 导出CSV

表 2 CWRU^[23]滚动轴承数据集

Table 2. CWRU ^[23]rolling bearing data set

损伤位置	标签	损伤直径/mm	样本数量
			数据集A		数据集B		数据集C
			训练集	测试集	训练集	测试集	训练集	测试集
正常	0	0	240	100	240	100	240	100
内圈	1	0.18	240	100	240	100	240	100
内圈	2	0.36	240	100	240	100	240	100
外圈	3	0.18	240	100	240	100	240	100
外圈	4	0.36	240	100	240	100	240	100
滚动体	5	0.18	240	100	240	100	240	100
滚动体	6	0.36	240	100	240	100	240	100

下载: 导出CSV

表 3 不同模型参数量和训练时间

Table 3. Number of Different model parameters and training time

模型	参数量	训练时间/s
本文模型	11500	2.84
MBDS-CNN^[26]	149300	3.81
WKCNN^[24]	26900	3.09
ResNet^[25]	84900	3.45
GhostCNN	700	2.73
ILeNet-5^[27]	40900	3.25

下载: 导出CSV

表 4 JNU轴承数据集说明

Table 4. JNU rolling bearing data set description

损伤位置	标签	样本数量
		工况F1		工况F2		工况F3
		训练集	测试集	训练集	测试集	训练集	测试集
正常	0	240	100	240	100	240	100
内圈	1	240	100	240	100	240	100
外圈	2	240	100	240	100	240	100
滚动体	3	240	100	240	100	240	100

下载: 导出CSV

参考文献(27)

[1]	HAN T, LIU C, YANG W G, et al. Deep transfer network with joint distribution adaptation: a new intelligent fault diagnosis framework for industry application[J]. ISA Transactions, 2020, 97: 269-281. doi: 10.1016/j.isatra.2019.08.012
[2]	HE S Y, HU D Y, YU G, et al. Trackside acoustic detection of axle bearing fault using wavelet domain moving beamforming method[J]. Applied Acoustics, 2022, 195: 108851. doi: 10.1016/j.apacoust.2022.108851
[3]	HOU M X, SHI H T. Stator-winding incipient shorted-turn fault detection for motor system in motorized spindle using modified interval observers[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 3505716.
[4]	LIU R N, YANG B Y, ZIO E, et al. Artificial intelligence for fault diagnosis of rotating machinery: a review[J]. Mechanical Systems and Signal Processing, 2018, 108: 33-47. doi: 10.1016/j.ymssp.2018.02.016
[5]	袁彩艳, 孙洁娣, 温江涛, 等. 多域信息融合结合改进残差密集网络的轴承故障诊断[J]. 振动与冲击, 2022, 41(4): 200-208. YUAN C Y, SUN J D, WEN J T, et al. Bearing fault diagnosis based on information fusion and improved residual dense networks[J]. Journal of Vibration and Shock, 2022, 41(4): 200-208(in Chinese).
[6]	HOANG D T, KANG H J. A survey on deep lelarning based bearing fault diagnosis[J]. Neurocomputing, 2019, 335: 327-335. doi: 10.1016/j.neucom.2018.06.078
[7]	SCHOLARPEDIA G E H. Deep belief networks[J]. Scholarpedia the Peer-Reviewed Open-Access Encyclopedia, 2009, 4(5): 5947.
[8]	RUMELHART D E, HINTON G E, WILLIAMS R J. Learning representations by back-propagating errors[J]. Nature, 1986, 323: 533-536. doi: 10.1038/323533a0
[9]	XU Y, LI Z X, WANG S Q, et al. A hybrid deep-learning model for fault diagnosis of rolling bearings[J]. Measurement, 2021, 169: 108502. doi: 10.1016/j.measurement.2020.108502
[10]	ZHANG Z Q, YANG Q Y, ZI Y Y. Multi-scale and multi-pooling sparse filtering: a simple and effective representation learning method for intelligent fault diagnosis[J]. Neurocomputing, 2021, 451: 138-151. doi: 10.1016/j.neucom.2021.04.066
[11]	CHEN Z Y, MAURICIO A, LI W H, et al. A deep learning method for bearing fault diagnosis based on cyclic spectral coherence and convolutional neural networks[J]. Mechanical Systems and Signal Processing, 2020, 140: 106683. doi: 10.1016/j.ymssp.2020.106683
[12]	XU Z F, LI C, YANG Y. Fault diagnosis of rolling bearing of wind turbines based on the variational mode decomposition and deep convolutional neural networks[J]. Applied Soft Computing, 2020, 95: 106515. doi: 10.1016/j.asoc.2020.106515
[13]	YE M Y, YAN X A, CHEN N, et al. Intelligent fault diagnosis of rolling bearing using variational mode extraction and improved one-dimensional convolutional neural network[J]. Applied Acoustics, 2023, 202: 109143.
[14]	刘俊锋, 俞翔, 万海波, 等. 基于 MFMD 和 Transformer-CNN 的滚动轴承故障诊断方法[J]. 航空动力学报, 2023, 38(6): 1446-1456. LIU J F, YU X, WAN H B, et al. Fault diagnosis method of rolling bearing using MFMD and Transformer-CNN[J]. Journal of Aerospace Power, 2023, 38(6): 1446-1456(in Chinese) .
[15]	FU W L, JIANG X H, LI B L, et al. Rolling bearing fault diagnosis based on 2D time-frequency images and data augmentation technique[J]. Measurement Science and Technology, 2023, 34(4): 045005. doi: 10.1088/1361-6501/acabdb
[16]	金江涛, 许子非, 李春, 等. 基于卷积双向长短期记忆网络与混沌理论的滚动轴承故障诊断[J]. 振动与冲击, 2022, 41(17): 160-169. JIN J T, XU Z F, LI C, et al. Fault diagnosis of rolling bearing based on CCNN-BiLSTMN method[J]. Journal of Vibration and Shock, 2022, 41(17): 160-169(in Chinese).
[17]	MAO G, ZHANG Z Z, QIAO B, et al. Fusion domain-adaptation CNN driven by images and vibration signals for fault diagnosis of gearbox cross-working conditions[J]. Entropy, 2022, 24(1): 119. doi: 10.3390/e24010119
[18]	WANG M J, WANG W J, ZHANG X N, et al. A new fault diagnosis of rolling bearing based on Markov transition field and CNN[J]. Entropy, 2022, 24(6): 751. doi: 10.3390/e24060751
[19]	HAN K, WANG Y H, TIAN Q, et al. GhostNet: more features from cheap operations[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2020: 1577-1586.
[20]	HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE Press, 2018: 7132-7141.
[21]	HUANG Z L, WANG X G, HUANG L C, et al. CCNet: criss-cross attention for semantic segmentation[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE Press, 2019: 603-612.
[22]	XU G W, LIU M, JIANG Z F, et al. Online fault diagnosis method based on transfer convolutional neural networks[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69(2): 509-520. doi: 10.1109/TIM.2019.2902003
[23]	Case Western Reserve University. Case Western Reserve University bearing data[EB/OL]. (2018-05-22) [2023-04-28]. https://gitcode.com/open-source-toolkit/1cd2d/?utm_source=tools_gitcode&index=bottom&type=card&webUrl.
[24]	SONG X D, CONG Y Y, SONG Y F, et al. A bearing fault diagnosis model based on CNN with wide convolution kernels[J]. Journal of Ambient Intelligence and Humanized Computing, 2022, 13(8): 4041-4056. doi: 10.1007/s12652-021-03177-x
[25]	HAO X Y, ZHENG Y, LU L, et al. Research on intelligent fault diagnosis of rolling bearing based on improved deep residual network[J]. Applied Sciences, 2021, 11(22): 10889. doi: 10.3390/app112210889
[26]	刘恒畅, 姚德臣, 杨建伟, 等. 基于多分支深度可分离卷积神经网络的滚动轴承故障诊断研究[J]. 振动与冲击, 2021, 40(10): 95-102. LIU H C, YAO D C, YANG J W, et al. Fault diagnosis of rolling bearings based on multi-branch deep separable convolutional neural networks[J]. Vibration and Shock, 2021, 40(10): 95-102(in Chinese) .
[27]	吴晨芳, 杨世锡, 黄海舟, 等. 一种基于改进的LeNet-5模型滚动轴承故障诊断方法研究[J]. 振动与冲击, 2021, 40(12): 55-61. WU C F, YANG S X, HUANG H Z, et al. An improved fault diagnosis method of rolling bearings based on LeNet-5[J]. Journal of Vibration and Shock, 2021, 40(12): 55-61(in Chinese).