基于变分模态提取及轻量级网络的滚动轴承故障诊断

doi:10.13229/j.cnki.jdxbgxb.20231047

摘要/Abstract

摘要：

设计了一种将变分模态提取（VME）与轻量级卷积神经网络（CNN）结合的滚动轴承故障诊断方法，解决了CNN在复杂工业环境下诊断性能低下及参数量庞大的问题。使用VME提取多个传感器收集的振动信号中的期望模态，并构建多传感器灰度特征图，消除信息干扰的同时实现数据融合。在SqueezeNet基础上引入残差结构与超轻量级子空间注意力模块（ULSAM），构建轻量级残差注意力卷积神经网络（LRACNN）。实验结果表明，本文方法在复杂环境下拥有很高的故障识别率和诊断稳定性。

关键词: 故障诊断, 滚动轴承, 卷积神经网络, 注意力机制, 轻量级

Abstract:

A rolling bearing fault diagnosis method combining variational mode extraction （VME） and lightweight convolutional neural network （CNN） was designed to solve the problems of low diagnostic performance of CNN in complex industrial environments as well as the problem of large number of parameters. VME was used to extract the desired modes in the vibration signals collected from multiple sensors and construct the multi-sensor grayscale feature maps to eliminate information interference while enabling data fusion. The residual structure and ultra-lightweight subspace attention module（ULSAM） are introduced on the basis of SqueezeNet to construct a lightweight residual attention convolutional neural network （LRACNN）. The method has a high fault recognition rate and diagnostic stability in complex environments.

Key words: fault diagnosis, rolling bearing, convolutional neural network, attention mechanism, lightweight

中图分类号:

TH17

冯志刚,王首起,于明月. 基于变分模态提取及轻量级网络的滚动轴承故障诊断[J]. 吉林大学学报(工学版), 2025, 55(6): 1883-1891.

Zhi-gang FENG,Shou-qi WANG,Ming-yue YU. Rolling bearing fault diagnosis based on variational mode extraction and lightweight network[J]. Journal of Jilin University(Engineering and Technology Edition), 2025, 55(6): 1883-1891.

图/表 16

图1

图2

图3

图4

图5

图6

表1

图7

图8

表2

不同工况下DE和FE不同故障类型轴承振动信号的中心频率"

标签	$ω d / H z$
	A0		A1		A2		A3
	DE	FE	DE	FE	DE	FE	DE	FE
Nor	1 000	1 000	1 100	1 000	1 100	1 000	2 100	2 100
Ba1	3 300	4 100	3 400	4 100	3 400	4 100	3 400	1 400
Ba2	3 200	4 100	3 300	4 100	3 200	1 400	1 400	1 400
Ba3	3 200	4 100	3 300	400	3 400	1 100	3 300	1 100
In1	3 600	1 500	3 500	1 500	3 500	1 400	3 600	1 400
In2	3 400	4 300	3 500	4 300	3 500	4 300	1 400	1 100
In3	2 900	500	2 900	1 800	2 800	3 300	2 800	3 300
Ou1	3 400	3 300	2 800	3 300	2 800	3 300	3 300	3 300
Ou2	3 400	4 100	3 400	4 100	3 500	4 100	3 400	4 100
Ou3	2 800	4 100	3 400	600	3 500	600	3 400	5 100

表2

图9

表3

表4

图10

表5

表6

参考文献 15

[1]	Yang D, Karimi H R, Sun K. Residual wide-kernel deep convolutional auto-encoder for intelligent rotating machinery fault diagnosis with limited samples[J]. Neural Networks, 2021, 14: 133-144.
[2]	陈晓雷, 孙永峰, 李策, 等. 基于卷积神经网络和双向长短期记忆的稳定抗噪声滚动轴承故障诊断[J]. 吉林大学学报: 工学版, 2022, 52(2): 296-309.
	Chen Xiao-lei, Sun Yong-feng, Li Ce, et al. Stable anti⁃noise fault diagnosis of rolling bearing based on CNN⁃BiLSTM[J]. Journal of Jilin University (Engineering and Technology Edition), 2022, 52(2): 296-309.
[3]	Gu J, Peng Y, Lu H, et al. A novel fault diagnosis method of rotating machinery via VMD, CWT and improved CNN[J]. Measurement, 2022, 200: No.111635.
[4]	Yao D, Liu H, Yang J, et al. A lightweight neural network with strong robustness for bearing fault diagnosis[J]. Measurement, 2020, 159: No.107756.
[5]	Bai R, Xu Q, Meng Z, et al. Rolling bearing fault diagnosis based on multi-channel convolution neural network and multi-scale clipping fusion data augmentation[J]. Measurement, 2021, 184: No.109885.
[6]	Zhong H, Lv Y, Yuan R, et al. Bearing fault diagnosis using transfer learning and self-attention ensemble lightweight convolutional neural network[J]. Neurocomputing, 2022, 501: 765-777.
[7]	Wang S, Feng Z. Multi-sensor fusion rolling bearing intelligent fault diagnosis based on VMD and ultra-lightweight GoogLeNet in industrial environments[J]. Digital Signal Processing, 2024, 145: No.104306.
[8]	王军, 张维通, 闫正兵, 等. 基于迁移QCNN的孪生网络轴承故障诊断方法[J]. 计算机测量与控制, 2024, 32(4): 6-12.
	Wang Jun, Zhang Wei-tong, Yan Zheng-bing, et al. Twin network-based bearing fault diagnosis method with transfer QCNN[J]. Computer Measurement & Control, 2024, 32(4): 6-12.
[9]	Nazari M, Sakhaei S M. Variational mode extraction: A new efficient method to derive respiratory signals from ECG[J]. IEEE Journal of Biomedical and Health Informatics, 2018, 22(4): 1059-1067.
[10]	Saini R, Jha N K, Das B, et al. ULSAM: Ultra-ightweight ubspace ttention odule for ompact onvolutional eural etworks[C]∥2020 IEEE Winter Conference on Applications of Computer Vision (WACV), Snowmass, CO, USA, 2020: 1616-1625.
[11]	Ye M, Yan X, 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: No.109143.
[12]	Smith W A, Randall R B. 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.
[13]	Wang Y, Yan J, Sun Q, et al. Bearing intelligent fault diagnosis in the industrial internet of things context: a lightweight convolutional neural network[J]. IEEE Access, 2020, 8: 87329-87340.
[14]	Ling L, Wu Q, Huang K, et al. A lightweight bearing fault diagnosis method based on multi-channel depthwise separable convolutional neural network[J]. Electronics, 2022, 11(24): No.4110.
[15]	Xu Z, Mei X, Wang X, et al. Fault diagnosis of wind turbine bearing using a multi-scale convolutional neural network with bidirectional long short term memory and weighted majority voting for multi-sensors[J]. Renewable Energy, 2022, 182: 615-626.

相关文章 15

[1]	杜睿山,王紫珊. 基于时空注意力的多视角人脸表情识别算法[J]. 吉林大学学报(工学版), 2025, 55(6): 2097-2102.
[2]	薛雅丽,俞潼安,崔闪,周李尊. 基于级联嵌套U-Net的红外小目标检测[J]. 吉林大学学报(工学版), 2025, 55(5): 1714-1721.
[3]	才华,王玉瑶,付强,马智勇,王伟刚,张晨洁. 基于注意力机制和特征融合的语义分割网络[J]. 吉林大学学报(工学版), 2025, 55(4): 1384-1395.
[4]	张河山,范梦伟,谭鑫,郑展骥,寇立明,徐进. 基于改进YOLOX的无人机航拍图像密集小目标车辆检测[J]. 吉林大学学报(工学版), 2025, 55(4): 1307-1318.
[5]	李扬,李现国,苗长云,徐晟. 基于双分支通道先验和Retinex的低照度图像增强算法[J]. 吉林大学学报(工学版), 2025, 55(3): 1028-1036.
[6]	张兰芳,李根泽,刘婷宇,余博. 局部多车影响下跟驰行为机理及建模[J]. 吉林大学学报(工学版), 2025, 55(3): 963-973.
[7]	王娜,崔月磊,李杨,王子从. 基于小波包对数能量图的滚动轴承故障诊断方法[J]. 吉林大学学报(工学版), 2025, 55(2): 494-502.
[8]	汪豪,赵彬,刘国华. 基于时间和运动增强的视频动作识别[J]. 吉林大学学报(工学版), 2025, 55(1): 339-346.
[9]	胡宏宇,张争光,曲优,蔡沐雨,高菲,高镇海. 基于双分支和可变形卷积网络的驾驶员行为识别方法[J]. 吉林大学学报(工学版), 2025, 55(1): 93-104.
[10]	郭晓然,王铁君,闫悦. 基于局部注意力和本地远程监督的实体关系抽取方法[J]. 吉林大学学报(工学版), 2025, 55(1): 307-315.
[11]	刘元宁,臧子楠,张浩,刘震. 基于深度学习的核糖核酸二级结构预测方法[J]. 吉林大学学报(工学版), 2025, 55(1): 297-306.
[12]	李路,宋均琦,朱明,谭鹤群,周玉凡,孙超奇,周铖钰. 基于RGHS图像增强和改进YOLOv5网络的黄颡鱼目标提取[J]. 吉林大学学报(工学版), 2024, 54(9): 2638-2645.
[13]	赵宏伟,武鸿,马克,李海. 基于知识蒸馏的图像分类框架[J]. 吉林大学学报(工学版), 2024, 54(8): 2307-2312.
[14]	张锦洲,姬世青,谭创. 融合卷积神经网络和双边滤波的相贯线焊缝提取算法[J]. 吉林大学学报(工学版), 2024, 54(8): 2313-2318.
[15]	余萍,赵康,曹洁. 基于优化A-BiLSTM的滚动轴承故障诊断[J]. 吉林大学学报(工学版), 2024, 54(8): 2156-2166.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

标签	故障类型	传感器	负载/HP	故障直径/mm	数据长度	样本数量
Nor	健康	DE和FE	0~3	—	1 024	200
Ba1	滚动体			0.177 8	1 024	200
Ba2	滚动体			0.355 6	1 024	200
Ba3	滚动体			0.533 4	1 024	200
In1	内圈			0.177 8	1 024	200
In2	内圈			0.355 6	1 024	200
In3	内圈			0.533 4	1 024	200
Ou1	外圈			0.177 8	1 024	200
Ou2	外圈			0.355 6	1 024	200
Ou3	外圈			0.533 4	1 024	200

方法	VME	VMD	EMD	ITD	WPD
正确率/%	100	99.50	97.55	98.80	99.25
时间/s	0.131	2.057	0.182	0.148	0.168

模型	正确率/%	训练时间/s	模型参数/MB
LRACNN	100	73.40	0.045
SqueezeNet	99.200	83.80	0.057
LCNN^［13］	97.386	1 623.00	15.522
ShuffleNet^［13］	92.094	962.00	4.385
MCDS-CNN^［14］	99.790	123.62	1.580

模型	SNR/dB
模型	-4	-2	0	2	4	6
LRACNN	93.30	94.65	96.50	97.10	97.75	98.65
SqueezeNet	91.60	92.85	94.60	96.65	96.70	97.40
CNN+ Attention^［2］	83.96	87.57	90.33	92.82	93.69	95.08
AAnNet^［2］	81.49	88.22	95.44	96.68	97.39	98.33
MSCNN- BiLSTM^［15］	87.00	93.00	96.70	97.50	99.24	99.24

负载/HP	正确率/%
负载/HP	无干扰	50 Hz干扰	60 Hz干扰
0	99.90	99.85	99.75
1	100	99.95	99.90
2	99.95	99.90	99.75
3	99.95	99.65	99.70