数据不平衡分布下轴承故障诊断方法

doi:10.13229/j.cnki.jdxbgxb20210374

吉林大学学报(工学版) ›› 2022, Vol. 52 ›› Issue (11): 2523-2531.doi: 10.13229/j.cnki.jdxbgxb20210374

• 车辆工程·机械工程 • 上一篇

数据不平衡分布下轴承故障诊断方法

曹洁^1,^2,³(),何智栋¹,余萍^1,³(),王进花^1,^3,⁴

^1.兰州理工大学电气工程与信息工程学院，兰州 730050
^2.甘肃省城市轨道交通工程研究中心，兰州 730050
^3.甘肃省制造信息工程研究中心，兰州 730050
^4.兰州理工大学电气与控制工程国家实验教学中心，兰州 730050

收稿日期:2021-04-26 出版日期:2022-11-01 发布日期:2022-11-16
通讯作者: 余萍 E-mail:caoj@lut.edu.cn;yup@lut.edu.cn
作者简介:曹洁（1966-），女，教授，博士生导师.研究方向：信息融合理论及应用，智能信息处理，智能交通系统的理论及应用. E-mail： caoj@lut.edu.cn
基金资助:
国家自然科学基金项目(61763028);甘肃省自然科学基金项目(20JR5RA463)

Bearing fault diagnosis method under unbalanced data distribution

Jie CAO^1,^2,³(),Zhi-Dong HE¹,Ping YU^1,³(),Jin-hua WANG^1,^3,⁴

^1.College of Electrical & Information Engineering，Lanzhou University of Technology，Lanzhou 730050，China
^2.Engineering Research Center of Urban Railway Transportation of Gansu Province，Lanzhou 730050，China
^3.Engineering Research Center of Manufacturing Information of Gansu Province，Lanzhou 730050，China
^4.National Experimental Teaching Center of Electrical and Control Engineering，Lanzhou University of Technology，Lanzhou 730050，China

Received:2021-04-26 Online:2022-11-01 Published:2022-11-16
Contact: Ping YU E-mail:caoj@lut.edu.cn;yup@lut.edu.cn

摘要/Abstract

摘要：

针对在滚动轴承的故障诊断中数据的不平衡分布会降低模型诊断能力的问题，本文提出一种首层拥有大尺度卷积核的一维卷积神经网络（WKFL-1DCNN）。WKFL-1DCNN首先使用较大的首层卷积核提取故障特征，并在交替的卷积层后添加批标准化（BN）层来调整数据分布；然后使用类平衡损失函数代替交叉熵损失函数来抵消数据不平衡分布给网络造成的影响。实验表明，本文所作改进能够有效提升WKFL-1DCNN在不平衡故障诊断中的表现，其故障诊断能力优于其他对比算法。

关键词: 故障诊断, 不平衡数据分布, 卷积神经网络, 类平衡损失函数, 滚动轴承

Abstract:

Aiming at the problem that the unbalanced distribution of data in the fault diagnosis of rolling bearings will reduce the model's diagnostic ability， a 1DCNN with Width Kernel of First Layer （WKFL-1DCNN） is proposed. WKFL-1DCNN firstly uses a larger kernel in first-layer to extract fault features， and adds a BN（Batch normalization） layer after the alternate convolution layer to adjust the data distribution； then uses the Class-Balanced loss function instead of the Cross-Entropy loss function to offset the impact of the data imbalanced distribution on the network. Experiments show that the improvement method in this paper can effectively improve the performance of WKFL-1DCNN in unbalanced fault diagnosis， and its fault diagnosis ability is better than other comparison algorithms.

Key words: fault diagnosis, unbalanced data distribution, convolutional neural network, Class-Balance loss function, rolling bearing

中图分类号:

TP277

曹洁,何智栋,余萍,王进花. 数据不平衡分布下轴承故障诊断方法[J]. 吉林大学学报(工学版), 2022, 52(11): 2523-2531.

Jie CAO,Zhi-Dong HE,Ping YU,Jin-hua WANG. Bearing fault diagnosis method under unbalanced data distribution[J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(11): 2523-2531.

图/表 12

图1

图2

图3

表1

表2

表3

图4

图5

表4

图6

表5

表6

参考文献 25

1	Tang D H, Hee J K. A survey on deep learning based bearing fault diagnosis[J]. Neurocomputing, 2019, 335: 327-335.
2	Szegedy C, Ioffe S, Vanhoucke V, et al. Inception-v 4, inception-resnet and the impact of residual connections on learning[J/OL]. (2020-08-23).
3	Mikolov T, Karafiát M, Burget L, et al. Recurrent neural network based language model[C]∥Interspeech, Conference of the International Speech Communication Association, Makuhari, Japan, 2015: 1045-1048.
4	钟辉, 康恒, 吕颖达, 等. 基于注意力卷积神经网络的图像篡改定位算法[J]. 吉林大学学报: 工学版, 2021, 51(5): 1838-1844.
	Zhong Hui, Kang Heng, Ying-da Lyu, et al. Image manipulation localization algorithm based on channel attention convolutional neural networks[J]. Journal of Jilin University(Engineering and Technology Edition), 2021, 51(5): 1838-1844.
5	Xia M, Li T, Liu L Z, et al. Intelligent fault diagnosis approach with unsupervised feature learning by stacked denoising autoencoder[J]. IET Science, Measurement & Technology, 2017, 11(6): 687-695.
6	Zhang W, Li C H, Peng G L, et al. A deep convolutional neural network with new training methods for bearing fault diagnosis under noisy environment and different working load[J]. Mechanical Systems and Signal Processing, 2018, 100: 439-453.
7	Shao H, Jiang H, Zhao H, et al. An enhancement deep feature fusion method for rotating machinery fault diagnosis[J].Knowledge-Based Systems, 2017, 119: 200-220.
8	Levent E. Bearing fault detection by one-dimensional convolutional neural networks[J]. Mathematical Problems in Engineering, 2017,2017: 1-9.
9	Pan J, Zi Y Y, Chen J L, et al. LiftingNet: a novel deep learning network with layerwise feature learning from noisy mechanical data for fault classification[J]. IEEE Trans Ind Electron, 2018, 65(6): 4973-4982.
10	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.
11	Eren L, Ince T, Kiranyaz S. A generic intelligent bearing fault diagnosis system using compact adaptive 1D CNN classifier[J]. Signal Process Syst, 2019, 91(2): 179-189.
12	邓飞跃, 吕浩洋, 顾晓辉, 等. 基于轻量化神经网络Shuffle-SENet的高速动车组轴箱轴承故障诊断方法[J]. 吉林大学学报: 工学版, 2022, 52(2):474-482.
	Deng Fei-yue, Hao-yang Lyu, Gu Xiao-hui, et al. Fault diagnosis of high-speed train axle bearing based on a lightweight neural network Shuffle-SENet[J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(2): 474-482.
13	Wang X, Mao D X, Li X D. Bearing fault diagnosis based on vibro-acoustic data fusion and 1D-CNN network[J]. Measurement, 2021, 173(6): 108518.
14	Lessmeier C, Kimotho J K, Zimmer D, et al. Condition monitoring of bearing damage in electromechanical drive systems by using motor current signals of electric motors: a benchmark data set for data-driven classification[C]∥European Conference of the Prognostics and Health Management Society, Bilbao, Spain, 2016:1-17.
15	Jia F, Lei Y, Lu N, et al. Deep normalized convolutional neural network for imbalanced fault classification of machinery and its understanding via visualization[J]. Mechanical Systems and Signal Processing, 2018, 110: 349-367.
16	Huang S Z, Tang J, Dai J Y, et al. 1DCNN fault diagnosis based on cubic spline interpolation pooling[J]. Shock and Vibration, 2020(2/3): 1-13.
17	Kang B Y, Xie S N, Marcus R, et al. Decoupling representation and classifier for long-tailed recognition[J/OL]. [2020-10-21].
18	Cao K D, Wei C, Adrien G, et al. Learning imbalanced datasets with label-distribution-aware margin loss[J/OL]. [2020-06-18]. 0/arXiv.1906.07413
19	Cui Y, Jia M L, Lin T Y, et al. Class-balanced loss based on effective number of samples[C]∥IEEE Conference on Computer Vision and Pattern Recognition, Long Beach, USA, 2019: 9268-9277.
20	张明德, 卢建华, 马婧华. 基于多尺度卷积策略CNN的滚动轴承故障诊断[J]. 重庆理工大学学报:自然科学, 2020, 34(6): 102-110.
	Zhang Ming-de, Lu Jian-hua, Ma Jing-hua. Fault diagnosis of rolling bearing based on multi-scale convolution strategy CNN[J]. Journal of Chongqing University of Technology(Natural Science), 2020, 34(6): 102-110.
21	Sergey I, Christian S. Batch normalization: accelerating deep network training by reducing internal covariate shift[J/OL]. [2020-08-11]. 8550/arXiv.1502.03167
22	Bjorck J, Gomes C, Selman B. Understanding batch normalization[J/OL]. [2020-06-01]. / 10.48550/arXiv.1806.02375 doi: 10.48550/arXiv.1806.02375
23	Kingma D P, Ba J. Adam: a method for stochastic optimization[J/OL]. [2020-12-22].
24	Zhang W, Peng G L, Li C H, et al. A new deep learning model for fault diagnosis with good anti-noise and domain adaptation ability on raw vibration signals[J]. Sensors, 2017, 17(3): 425-445.
25	Jiang G Q, He H B, Yan J, et al. Multiscale convolutional neural networks for fault diagnosis of wind turbine gearbox[J]. IEEE Transactions on Industrial Electronics, 2019, 66(4): 3196-3207.

相关文章 15

[1]	申铉京,张雪峰,王玉,金玉波. 像素级卷积神经网络多聚焦图像融合算法[J]. 吉林大学学报(工学版), 2022, 52(8): 1857-1864.
[2]	高明华,杨璨. 基于改进卷积神经网络的交通目标检测方法[J]. 吉林大学学报(工学版), 2022, 52(6): 1353-1361.
[3]	杨怀江,王二帅,隋永新,闫丰,周跃. 简化型残差结构和快速深度残差网络[J]. 吉林大学学报(工学版), 2022, 52(6): 1413-1421.
[4]	王学智,李清亮,李文辉. 融合迁移学习的土壤湿度预测时空模型[J]. 吉林大学学报(工学版), 2022, 52(3): 675-683.
[5]	李向军,涂洁莹,赵志宾. 基于多尺度融合卷积神经网络的熔解曲线有效性分类[J]. 吉林大学学报(工学版), 2022, 52(3): 633-639.
[6]	李先通,全威,王华,孙鹏程,安鹏进,满永兴. 基于时空特征深度学习模型的路径行程时间预测[J]. 吉林大学学报(工学版), 2022, 52(3): 557-563.
[7]	曹洁,马佳林,黄黛麟,余萍. 一种基于多通道马尔可夫变迁场的故障诊断方法[J]. 吉林大学学报(工学版), 2022, 52(2): 491-496.
[8]	罗巍,卢博,陈菲,马腾. 基于PSO-SVM及时序环节的数控刀架故障诊断方法[J]. 吉林大学学报(工学版), 2022, 52(2): 392-399.
[9]	高文志,王彦军,王欣伟,张攀,李勇,董阳. 基于卷积神经网络的柴油机失火故障实时诊断[J]. 吉林大学学报(工学版), 2022, 52(2): 417-424.
[10]	王进花,胡佳伟,曹洁,黄涛. 基于自适应变分模态分解和集成极限学习机的滚动轴承多故障诊断[J]. 吉林大学学报(工学版), 2022, 52(2): 318-328.
[11]	董绍江,朱朋,裴雪武,李洋,胡小林. 基于子领域自适应的变工况下滚动轴承故障诊断[J]. 吉林大学学报(工学版), 2022, 52(2): 288-295.
[12]	邓飞跃,吕浩洋,顾晓辉,郝如江. 基于轻量化神经网络Shuffle⁃SENet的高速动车组轴箱轴承故障诊断方法[J]. 吉林大学学报(工学版), 2022, 52(2): 474-482.
[13]	张龙,徐天鹏,王朝兵,易剑昱,甄灿壮. 基于卷积门控循环网络的齿轮箱故障诊断[J]. 吉林大学学报(工学版), 2022, 52(2): 368-376.
[14]	陈晓雷,孙永峰,李策,林冬梅. 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.
[15]	陈菲,杨峥,张志成,罗巍. 面向无标签数据的旋转机械故障诊断方法[J]. 吉林大学学报(工学版), 2022, 52(11): 2514-2522.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

Layer	首层卷积核尺寸
1	Conv 64-s8	Conv 96-s8	Conv 128-s8	Conv 64-s8Padding=32
2	Conv 9-s2	Conv 9-s2	Conv 9-s2	Conv 9-s2
2	MaxP 2×2	MaxP 2×2	MaxP 2×2	MaxP 2×2
3	Conv 9-s2	Conv 8-s2	Conv 7-s2	Conv 7-s2
3	MaxP 2×2	MaxP 2×2	MaxP 2×2	MaxP 2×2
4	Conv 5-s2	Conv 5-s2	Conv 5-s2	Conv 5-s2
4	MaxP 2×1	MaxP 2×1	MaxP 2×1	MaxP 2×1
5	F 256-50-14	F 256-50-14	F 256-50-14	F 256-50-14
5	softmax	softmax	softmax	softmax

尺寸	宏精准率/%	宏召回率/%	宏F1-score/%	准确率/%
64	94.02	93.59	93.66	93.98
96	94.64	94.53	94.53	94.87
128	95.71	95.61	95.63	95.84
192	95.25	94.79	94.85	95.13

数据集	训练精度/%	测试精度/%
WKFL-1DCNN	100.0	99.8
WDCNN	100.0	99.7
MSCNN	100.0	99.5
cRT	99.3	98.2

状态及标签	训练集样本数	测试集样本数
Normal（0）	1000 160 160 60 60 60	440
KA04+KA22（1）		360
KI04+KI14（2）		360
KA15（3）		300
KA16（4）		300
KA30（5）		300
KB23（6）	60 60 60 60 60 60 60	300
KB24（7）		300
KB27（8）		300
KI16（9）		300
KI17（10）		300
KI18（11）		300
KI21（12）		300

数据集	训练精度/%	测试精度/%
WKFL-1DCNN	100.0	94.4
WDCNN	100.0	92.1
MSCNN	99.5	89.5
cRT	100.0	90.9

数据不平衡分布下轴承故障诊断方法

Bearing fault diagnosis method under unbalanced data distribution

RICH HTML

PDF (PC)

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 25

相关文章 15

Metrics

本文评价

推荐阅读 0

状态及标签	训练集样本数		测试集样本数
状态及标签	数据集A	数据集B	数据集C	数据集D
正常（9）	520	500	220	200
In-0.007（0）	520	80	180	200
Ball-0.007（1）	520	80	180	200
Out@6-0.007（2）	520	80	180	200
In-0.014（3）	520	30	150	200
Ball-0.014（4）	520	30	150	200
Out@6-0.014（5）	520	30	150	200
In-0.021（6）	520	30	150	200
Ball-0.021（7）	520	30	150	200
Out@6-0.021（8）	520	30	150	200
Out@3-0.007（10）	520	30	150	200
Out@3-0.021（11）	520	30	150	200
Out@12-0.007（12）	520	30	150	200
Out@12-0.021（13）	520	30	150	200