改进一维卷积神经网络与双向门控循环单元的轴承故障诊断研究

杨云; 丁磊; 张昊宇

doi:10.13433/j.cnki.1003-8728.20200638

改进一维卷积神经网络与双向门控循环单元的轴承故障诊断研究

doi: 10.13433/j.cnki.1003-8728.20200638

华东交通大学电气与自动化工程学院, 南昌 330033

基金项目:

江西省重点研发计划项目 20202BBEL53008

详细信息

作者简介:
杨云(1972-), 高级实验师, 研究方向为故障诊断、检测技术及自动化, yyang@ecjtu.edu.cn

中图分类号: TH133.33;TH165.3
计量
- 文章访问数: 162
- HTML全文浏览量: 120
- PDF下载量: 15
- 被引次数: 0
出版历程
- 收稿日期: 2021-05-06
- 刊出日期: 2023-04-25

Research on Bearing Fault Diagnosis of Improved One-dimensional Convolutional Neural Network and Bidirectional Gated Recurrent Unit

School of Electrical and Automation Engineering, East China Jiaotong University, Nanchang 330033, China

摘要

摘要: 针对传统智能故障诊断依赖于人工经验进行特征提取和传统卷积神经网络(Convolutional neural networks, CNN)参数过多、训练量过大且无法充分利用时间序列信息的缺点, 提出一种基于改进一维卷积神经网络与双向门控循环单元的深度学习新算法。首先, 该方法利用一维卷积神经网络自提取能力进行特征提取, 同时设计了一个全局均值池化层替换传统卷积神经网络的全连接层, 减少参数数量; 其次, 引入双向门控循环单元学习特征信号中的时间序列关系; 最后, 通过支持向量机替换传统CNN中的Softmax层进行故障分类, 进一步提高诊断的准确率。实验表明, 该方法将诊断的准确率提升至99.8%, 并且加快了诊断的速度。通过与其他方法的对比, 证明了该方法有着更高的准确率, 更快的诊断速度, 更好的鲁棒性。
- 轴承故障诊断 /
- 卷积神经网络 /
- 双向门控循环单元 /
- 支持向量机
Abstract: Aiming at the shortcomings of traditional intelligent fault diagnosis method that relies on manual experience for feature extraction and traditional convolutional neural networks (Convolutional neural networks, CNN) with too many parameters, too much training, and inability to make full use of time series information, a new deep learning algorithm based on an improved one-dimensional convolutional neural network and two-way gated loop unit is proposed. Firstly, the method uses the self-extraction ability of one-dimensional convolutional neural networks for feature extraction, and at the same time, a global mean pooling layer is designed to replace the fully connected layer of traditional convolutional neural networks to reduce the number of parameters. Secondly, the two-way gated loop unit is introduced to learn the time series relationship in the characteristic signal. Finally, the support vector machine is used to replace the traditional Softmax layer in CNN for fault classification, which further improves the accuracy of diagnosis. Experiments show that this method improves the accuracy of diagnosis to 99.8% and speeds up the diagnosis. Through comparison with other methods, it is proved that this method has higher accuracy, faster diagnosis speed and better robustness.
- bearing fault diagnosis /
- convolutional neural network /
- two-way gated recurrent unit /
- support vector machine

HTML全文

图 1 卷积神经网络基本结构

下载: 全尺寸图片幻灯片

图 2 GRU结构图

下载: 全尺寸图片幻灯片

图 3 BiGRU结构图

下载: 全尺寸图片幻灯片

图 4 改进1DCNN-BiGRU-SVM模型结构图

下载: 全尺寸图片幻灯片

图 5 故障诊断流程图

下载: 全尺寸图片幻灯片

图 6 滚动轴承实验台

下载: 全尺寸图片幻灯片

图 7 训练集和验证集损失曲线

下载: 全尺寸图片幻灯片

图 8 测试集和验证集准确率曲线

下载: 全尺寸图片幻灯片

图 9 故障诊断多分类混淆矩阵

下载: 全尺寸图片幻灯片

图 10 准确率对比曲线

下载: 全尺寸图片幻灯片

图 11 损失率对比曲线

下载: 全尺寸图片幻灯片

图 12 t-SNE降维可视化

下载: 全尺寸图片幻灯片

图 13 不同模型在不同负载下故障诊断结果

下载: 全尺寸图片幻灯片

表 1 滚动轴承故障数据集

故障类型		训练集	验证集	测试集	标签
	滚动体	700	200	100	1
0.007/inch	内圈故障	700	200	100	2
	外圈故障	700	200	100	3
	滚动体	700	200	100	4
0.014/inch	内圈故障	700	200	100	5
	外圈故障	700	200	100	6
0.021 /inch	滚动体	700	200	100	7
	内圈故障	700	200	100	8
	外圈故障	700	200	100	9
	正常	700	200	100	0

下载: 导出CSV

表 2 改进1DCNN-BiGRU-SVM模型诊断结果

故障序号	精确率	召回率	F₁调和均值	样本
0	1.00	1.00	1.00	100
1	1.00	1.00	1.00	100
2	1.00	1.00	1.00	100
3	1.00	0.98	0.99	100
4	1.00	1.00	1.00	100
5	0.98	1.00	0.99	100
6	1.00	1.00	1.00	100
7	1.00	1.00	1.00	100
8	1.00	1.00	1.00	100
9	1.00	1.00	1.00	100
平均值/总数	0.998	0.998	0.998	1 000

下载: 导出CSV

表 3 模型参数数量与诊断结果

名称	改进 1DCNN-BiGRU-SVM	传统 1DCNN-BiGRU-Softmax
Conv1d	51 300	51 300
batch	400	400
max pooling	0	0
dropout	0	0
Conv1d	20 100	20 100
batch	400	400
max pooling	0	0
dropout	0	0
flatten	0	0
reshape	0	0
BiGRU	1 872	1 872
activation	0	0
Global maxpool	0	-
dense	-	1 400
activation	0	0
dense	70	2 010
SVM	0	-
Softmax	-	0
总数	74 142	77 482
准确率	0.998	0.986
诊断时间/s	57.002 243	72.002 416

下载: 导出CSV

表 4 不同模型在不同负载下故障诊断结果

名称	0 kW	0.735 kW	1.470 kW	2.205 kW
SVM	0.708	0.729	0.736	0.722
VMD排列熵+SVM	0.962	0.958	0.947	0.973
传统1DCNN	0.947	0.953	0.943	0.964
传统 1DCNN-BiGRU-Softmax	0.974	0.987	0.979	0.984
改进 1DCNN-BiGRU-SVM	0.996	0.992	0.994	0.998

下载: 导出CSV

参考文献(19)

[1]	XIA M, LI T, XU L, et al. Fault diagnosis for rotating machinery using multiple sensors and convolutional neural networks[J]. IEEE/ASME Transactions on Mechatronics, 2018, 23(1): 101-110. doi: 10.1109/TMECH.2017.2728371
[2]	肖雄, 肖宇雄, 张勇军, 等. 基于二维灰度图的数据增强方法在电机轴承故障诊断的应用研究[J]. 中国电机工程学报, 2021, 41(2): 738-748. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGDC202102032.htm XIAO X, XIAO Y X, ZHANG Y J, et al. Research on the application of the data augmentation method based on 2D gray pixel images in the fault diagnosis of motor bearing[J]. Proceedings of the CSEE, 2021, 41(2): 738-748. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZGDC202102032.htm
[3]	SHAO H D, JIANG H K, LIN Y, et al. A novel method for intelligent fault diagnosis of rolling bearings using ensemble deep auto-encoders[J]. Mechanical Systems and Signal Processing, 2018, 102: 278-297. doi: 10.1016/j.ymssp.2017.09.026
[4]	杨慧斌. 滚动轴承故障诊断中的特征提取与选择方法[D]. 株洲: 湖南工业大学, 2011. YANG H B. Features extraction and selection in rolling bearing fault diagnosis[D]. Zhuzhou: Hunan University of Technology, 2011. (in Chinese)
[5]	QU J X, ZHANG Z S, GONG T. A novel intelligent method for mechanical fault diagnosis based on dual-tree complex wavelet packet transform and multiple classifier fusion[J]. Neurocomputing, 2016, 171: 837-853. doi: 10.1016/j.neucom.2015.07.020
[6]	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. http://www.cnki.com.cn/Article/CJFDTotal-ZDCJ201819021.htm LI H, ZHANG Q, QIN X R, et al. Fault diagnosis method for rolling bearings based on short-time Fourier transform and convolution neural network[J]. Journal of Vibration and Shock, 2018, 37(19): 124-131. (in Chinese) http://www.cnki.com.cn/Article/CJFDTotal-ZDCJ201819021.htm
[7]	徐可, 陈宗海, 张陈斌, 等. 基于经验模态分解和支持向量机的滚动轴承故障诊断[J]. 控制理论与应用, 2019, 36(6): 915-922. doi: 10.7641/CTA.2018.80257 XU K, CHEN Z H, ZHANG C B, et al. Rolling bearing fault diagnosis based on empirical mode decomposition and support vector machine[J]. Control Theory & Applications, 2019, 36(6): 915-922. (in Chinese) doi: 10.7641/CTA.2018.80257
[8]	VERSTRAETE D, FERRADA A, DROGUETT E L, et al. Deep learning enabled fault diagnosis using time-frequency image analysis of rolling element bearings[J]. Shock and Vibration, 2017, 2017: 5067651.
[9]	HARMOUCHE J, DELPHA C, DIALLO D. Incipient fault detection and diagnosis based on Kullback-Leibler divergence using principal component analysis: Part Ⅱ[J]. Signal Processing, 2015, 109: 334-344. doi: 10.1016/j.sigpro.2014.06.023
[10]	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
[11]	CHENG J, WANG P S, LI G, et al. Recent advances in efficient computation of deep convolutional neural networks[J]. Frontiers of Information Technology & Electronic Engineering, 2018, 19(1): 64-77.
[12]	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. https://www.cnki.com.cn/Article/CJFDTOTAL-ZDGC202002021.htm GONG W F, CHEN H, ZHANG Z H, et al. Intelligent fault diagnosis for rolling bearing based on improved convolutional neural network[J]. Journal of Vibration Engineering, 2020, 33(2): 400-413. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZDGC202002021.htm
[13]	孙洁娣, 毛新茹, 温江涛, 等. 深度卷积长短期记忆网络的轴承故障诊断[J]. 机械科学与技术, 2021, 40(7): 1091-1099. doi: 10.13433/j.cnki.1003-8728.20200170 SUN J D, MAO X R, WEN J T, et al. Bearing fault diagnosis using deep CNN and LSTM[J]. Mechanical Science and Technology for Aerospace Engineering, 2021, 40(7): 1091-1099. (in Chinese) doi: 10.13433/j.cnki.1003-8728.20200170
[14]	GU J X, WANG Z H, KUEN J, et al. Recent advances in convolutional neural networks[J]. Pattern Recognition, 2018, 77: 354-377. https://www.sciencedirect.com/science/article/pii/S0031320317304120
[15]	杨丽, 吴雨茜, 王俊丽, 等. 循环神经网络研究综述[J]. 计算机应用, 2018, 38(S2): 1-6+26. YANG L, WU Y Q, WANG J L, et al. Research on recurrent neural network[J]. Journal of Computer Applications, 2018, 38(S2): 1-6+26. (in Chinese)
[16]	魏永合, 王明华. 基于EEMD和SVM的滚动轴承退化状态识别[J]. 计算机集成制造系统, 2015, 21(9): 2475-2483. doi: 10.13196/j.cims.2015.09.024 WEI Y H, WANG M H. Degradation state recognition of rolling bearing based on EEMD and SVM[J]. Computer Integrated Manufacturing Systems, 2015, 21(9): 2475-2483. (in Chinese) doi: 10.13196/j.cims.2015.09.024
[17]	SUN W Q, ZHAO R, YAN R Q, et al. Convolutional discriminative feature learning for induction motor fault diagnosis[J]. IEEE Transactions on Industrial Informatics, 2017, 13(3): 1350-1359. https://ieeexplore.ieee.org/document/7862893
[18]	宫文峰, 陈辉, WANG D W, 等. 基于改进CNN-GAP-SVM的船舶电力变换器快速故障诊断方法[J]. 计算机集成制造系统, 2022, 28(5): 1370-1384. https://www.cnki.com.cn/Article/CJFDTOTAL-JSJJ202205009.htm GONG W F, CHEN H, WANG D W, et al. Fast fault diagnosis method of marine electrical converter based on improved CNN-GAP-SVM algorithn[J]. Computer Integrated Manufacturing Systems, 2022, 28(5): 1370-1384. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JSJJ202205009.htm
[19]	葛阳, 郭兰中, 牛曙光, 等. 基于t-SNE和LSTM的旋转机械剩余寿命预测[J]. 振动与冲击, 2020, 39(7): 223-231+273. GE Y, GUO L Z, NIU S G, et al. Prediction of remaining useful life based on t-SNE and LSTM for rotating machinery[J]. Journal of Vibration and Shock, 2020, 39(7): 223-231+273. (in Chinese)