基于改进深度残差网络的旋转机械故障诊断

doi:10.12305/j.issn.1001-506X.2022.06.34

摘要/Abstract

摘要：

针对旋转机械工况复杂多变、有标签样本不足而导致的故障特征提取困难等问题, 提出了一种用于旋转机械故障诊断的改进深度残差网络(improved deep residual network, IDRN)。首先, 采集旋转机械一维振动信号进行数据预处理; 然后, 在深度残差网络的基础上引入了长短时记忆(long short-term memory, LSTM)网络, 其中, LSTM网络可以有效捕捉故障的时序信息; 在残差块中引入Dropout层提高了故障诊断的精度和收敛速度; 最后在轴承与齿轮数据集上验证本文提出方法的有效性。实验结果表明, 该方法在堆叠多层网络模型时, 没有出现明显的网络退化现象, 与当前广泛使用的几种诊断方法进行对比实验, 表现出了较高的平均诊断精度和良好的适用性。

关键词: 故障诊断, 改进深度残差网络, 长短时记忆网络, Dropout层

Abstract:

An improved deep residual network (IDRN) for fault diagnosis of rotating machinery is proposed to solve the problems of fault feature extraction difficulty caused by complex and variable working conditions and insufficient samples of labels. Firstly, one-dimensional vibration signals of rotating machinery are collected for data preprocessing. Then, long short-term memory (LSTM) network is introduced on the basis of the deep residual network, in which the time-series information of faults could be captured effectively.The Dropout layer is introduced into the residual block to improve the accuracy and convergence speed of fault diagnosis. Finally, the validity of the proposed method is verified on the data sets of bearings and gears.Experimental results show that there is no obvious network degradation phenomenon when the proposed method is used to stack multi-layer network models. Compared with several widely used diagnostic methods, the proposed method shows higher average diagnostic accuracy and good applicability.

Key words: fault diagnosis, improved deep residual network (IDRN), long short-term memory (LSTM) network, Dropout layer

中图分类号:

TH17

侯召国, 王华伟, 周良, 付强. 基于改进深度残差网络的旋转机械故障诊断[J]. 系统工程与电子技术, 2022, 44(6): 2051-2059.

Zhaoguo HOU, Huawei WANG, Liang ZHOU, Qiang FU. Fault diagnosis of rotating machinery based on improved deep residual network[J]. Systems Engineering and Electronics, 2022, 44(6): 2051-2059.

图/表 17

图1

图2

图3

图4

图5

表1

表2

表3

表4

图6

图7

图8

图9

图10

图11

表5

表6

参考文献 30

1	WU X Y , ZHANG Y , CHENG C M , et al. A hybrid classification autoencoder for semi-supervised fault diagnosis in rotating machinery[J]. Mechanical Systems and Signal Processing, 2021, 149, 107327. doi: 10.1016/j.ymssp.2020.107327
2	GUO J C , ZHEN D , LI H Y , et al. Fault feature extraction for rolling element bearing diagnosis based on a multi-stage noise reduction method[J]. Measurement, 2019, 139, 226- 235. doi: 10.1016/j.measurement.2019.02.072
3	LEI Y G , YANG B , JIANG X W , et al. Applications of machine learning to machine fault diagnosis: a review and roadmap[J]. Mechanical Systems and Signal Processing, 2020, 138, 106587. doi: 10.1016/j.ymssp.2019.106587
4	邵海东, 张笑阳, 程军圣, 等. 基于提升深度迁移自动编码器的轴承智能故障诊断[J]. 机械工程学报, 2020, 56 (9): 84- 90.
	SHAO H D , ZHANG X Y , CHENG J S , et al. Intelligent fault dia-gnosis of bearings based on lifting depth migration autoencoder[J]. Journal of Mechanical Engineering, 2020, 56 (9): 84- 90.
5	SHAO H D , JIANG H K , ZHANG H Z , et al. Rolling bearing fault feature learning using improved convolutional deep belief network with compressed sensing[J]. Mechanical Systems and Signal Processing, 2018, 100, 743- 765. doi: 10.1016/j.ymssp.2017.08.002
6	LI X Y , LI J L , ZHAO C Y , et al. Gear pitting fault diagnosis with mixed operating conditions based on adaptive 1D separable convolution with residual connection[J]. Mechanical Systems and Signal Processing, 2020, 142, 106740. doi: 10.1016/j.ymssp.2020.106740
7	ZHANG K , TANG B P , DENG L , et al. A hybrid attention improved ResNet based fault diagnosis method of wind turbines gearbox[J]. Measurement, 2021, 179, 109491. doi: 10.1016/j.measurement.2021.109491
8	HOANG D T , KANG H J . Rolling element bearing fault diagnosis using convolutional neural network and vibration image[J]. Cognitive Systems Research, 2019, 53, 42- 50. doi: 10.1016/j.cogsys.2018.03.002
9	YU J B , ZHOU X K . One-dimensional residual convolutional autoencoder based feature learning for gearbox fault diagnosis[J]. IEEE Trans.on Industrial Informatics, 2020, 16 (10): 6347- 6358. doi: 10.1109/TII.2020.2966326
10	MAO W T , FENG W S , LIU Y M , et al. A new deep auto-encoder method with fusing discriminant information for bearing fault diagnosis[J]. Mechanical Systems and Signal Processing, 2021, 150, 107233. doi: 10.1016/j.ymssp.2020.107233
11	LU C , WANG Z Y , ZHOU B . Intelligent fault diagnosis of rolling bearing using hierarchical convolutional network based health state classification[J]. Advanced Engineering Informatics, 2017, 32, 139- 151. doi: 10.1016/j.aei.2017.02.005
12	JIA F , LEI Y G , LIN J , et al. Deep neural networks: a promising tool for fault characteristic mining and intelligent diagnosis of rotating machinery with massive data[J]. Mechanical Systems and Signal Processing, 2016, 72/73, 303- 315. doi: 10.1016/j.ymssp.2015.10.025
13	陈国梁, 岳夏, 周超, 等. 机械故障诊断技术机遇与挑战[J]. 机电工程技术, 2020, 49 (10): 1- 4. doi: 10.3969/j.issn.1009-9492.2020.10.001
	CHEN G L , YUE X , ZHOU C , et al. Opportunity and challenge to the development of mechanical fault diagnosis technology[J]. Mechanical & Electrical Engineering Technology, 2020, 49 (10): 1- 4. doi: 10.3969/j.issn.1009-9492.2020.10.001
14	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proc. of the IEEE Computer Vision and Pattern Recognition, 2016: 770-778.
15	田强, 贾小宁. 基于深度残差神经网络的车标识别[J]. 吉林大学学报, 2021, 59 (2): 319- 324.
	TIAN Q , JIA X N . Vehicle sign recognition based on deep residual neural network[J]. Journal of Jilin University, 2021, 59 (2): 319- 324.
16	HINTON G , SRIVASTAVA N , KRIZHEVSKY A , et al. Improving neural networks by preventing co-adaptation of feature detectors[J]. Computer Science, 2012, 3 (4): 212- 223.
17	SRIVASTAVA N , HINTON G , KRIZHEVSKY A , et al. Dropout: a simple way to prevent neural networks from overfitting[J]. Journal of Machine Learning Research, 2014, 15 (1): 1929- 1958.
18	CHEN L T , XU G H , TAO T F , et al. Deep residual network for identifying bearing fault location and fault severity concurrently[J]. IEEE Access, 2020, 8, 168026- 168035. doi: 10.1109/ACCESS.2020.3023970
19	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, 425. doi: 10.3390/s17020425
20	陈保家, 陈学力, 沈保明, 等. CNN-LSTM深度神经网络在滚动轴承故障诊断中的应用[J]. 西安交通大学学报, 2021, 55 (6): 28- 36.
	CHEN B J , CHEN X L , SHEN B M , et al. Application of CNN-LSTM deep neural network in rolling bearing fault diagnosis[J]. Journal of Xi'an Jiaotong University, 2021, 55 (6): 28- 36.
21	张强, 黄挺, 陆效农, 等. 基于CNN-LSTM深度学习方法及多属性时序数据的故障诊断方法. 中国, 201811472378.6[P]. 2018-12-04.
	ZHANG Q, HUANG T, LU X N, et al. Fault diagnosis method based on CNN-LSTM deep learning method and multi-attribute time-series data. China, 201811472378.6[P]. 2018-12-04.
22	WANG Y M , CHENG L . A combination of residual and long-short-term memory networks for bearing fault diagnosis based on time-series model analysis[J]. Measurement Science and Technology, 2021, 32, 015904. doi: 10.1088/1361-6501/abaa1e
23	JIAO J Y , ZHAO M , LIN J , et al. Residual joint adaptation adversarial network for intelligent transfer fault diagnosis[J]. Mechanical Systems and Signal Processing, 2020, 145, 106962. doi: 10.1016/j.ymssp.2020.106962
24	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
25	SHAO S Y , MCALEER S , YAN R Q , et al. Highly accurate machine fault diagnosis using deep transfer learning[J]. IEEE Trans.on Industrial Informatics, 2019, 15 (4): 2446- 2455. doi: 10.1109/TII.2018.2864759
26	LI X , LI X , MA H . Deep representation clustering-based fault diagnosis method with unsupervised data applied to rotating machinery[J]. Mechanical Systems and Signal Processing, 2020, 143, 106825. doi: 10.1016/j.ymssp.2020.106825
27	XIANG Z , ZHANG X N , ZHANG W W , et al. Fault diagnosis of rolling bearing under fluctuating speed and variable load based on TCO spectrum and stacking auto-encoder[J]. Measurement, 2019, 138, 162- 174. doi: 10.1016/j.measurement.2019.01.063
28	JIANG W , ZHOU J Z , LIU H , et al. A multi-step progressive fault diagnosis method for rolling element bearing based on energy entropy theoryand hybrid ensemble auto-encoder[J]. ISA Transactions, 2019, 87, 235- 250. doi: 10.1016/j.isatra.2018.11.044
29	ZHOU Q , LI Y B , TIAN Y , et al. A novel method based on nonlinear auto-regression neural network and convolutional neural network for imbalanced fault diagnosis of rotating machinery[J]. Measurement, 2020, 161, 107880. doi: 10.1016/j.measurement.2020.107880
30	MENG Z , ZHAN X Y , LI J , et al. An enhancement denoising autoencoder for rolling bearing fault diagnosis[J]. Measurement, 2018, 130, 448- 454. doi: 10.1016/j.measurement.2018.08.010

故障标号	故障直径/(″)	电机负荷/HP	电机转速/rpm	故障位置	样本长度	训练/验证/测试
N	无	0	1 797	无	2 048	700/200/100
F1	0.021	0	1 797	内圈	2 048	700/200/100
F2	0.021	0	1 797	滚珠	2 048	700/200/100
F3	0.021	0	1 797	外圈6点钟	2 048	700/200/100
F4	0.021	0	1 797	外圈3点钟	2 048	700/200/100
F5	0.021	0	1 797	外圈12点钟	2 048	700/200/100

故障标号	故障直径/(″)	电机负荷/HP	电机转速/rpm	故障位置	样本长度	训练/验证/测试
N′	无	0	1 797	无	2 048	700/200/100
F1′	0.007	1	1 772	内圈	2 048	700/200/100
F2′	0.014	2	1 750	滚珠	2 048	700/200/100
F3′	0.021	3	1 730	外圈6点钟	2 048	700/200/100

故障标号	故障类型	电机负荷	样本长度	训练/验证/测试
H	无	20 Hz-0 V	2 048	700/200/100
G1	剥落	20 Hz-0 V	2 048	700/200/100
G2	断齿	20 Hz-0 V	2 048	700/200/100
G3	裂纹	20 Hz-0 V	2 048	700/200/100
G4	磨损	20 Hz-0 V	2 048	700/200/100

模型结构	模型参数	激活函数	输入大小	输出大小
输入层	-	-	(None, 2 048, 1)	(None, 2 048, 1)
卷积层	(32, 3)	ReLU	(None, 2 048, 1)	(None, 2 046, 32)
卷积层	(32, 3)	ReLU	(None, 2 046, 32)	(None, 2 044, 32)
最大池化层	3	-	(None, 2 044, 32)	(None, 681, 32)
LSTM网络层	32	Tanh	(None, 681, 32)	(None, 681, 32)
改进残差块1	(32, 3)	ReLU	(None, 681, 32)	(None, 681, 32)
改进残差块2	(32, 3)	ReLU	(None, 681, 32)	(None, 681, 32)
改进残差块3	(32, 3)	ReLU	(None, 681, 32)	(None, 681, 32)
卷积层	(32, 3)	ReLU	(None, 681, 32)	(None, 679, 32)
全局平均池化层	-	-	(None, 679, 32)	(None, 32)
全连接层	64	ReLU	(None, 32)	(None, 64)
丢弃层	0.25	-	(None, 64)	(None, 64)
分类层	6	Softmax	(None, 64)	(None, 6)

模型	单工况参数设置
BPNN	网络结构为[32,6],训练周期为20,学习率为0.001,优化器采用随机梯度下降
MLP	网络结构为[64,32,16,6],训练周期为20,学习率为0.001,优化器为Adam
SAE	网络结构为[64,32,16,10,16,32,64,6],训练周期为20,学习率为0.001,优化器为Adam
CNN	网络包含3个卷积层,每个卷积层滤波器尺寸分别为[16,32,64],卷积核大小分别为[64,3,3],池化步长为2,训练周期为20,学习率为0.001,优化器为Adam
RNN	网络结构为[32,32,16,32,6],训练周期为20,学习率为0.001,优化器为Adam
DRN	网络结构包含3个残差块,每个残差块包含2个卷积层和1个Dropout层,每个卷积层滤波器尺寸为32,卷积核大小为3,训练周期为20,学习率为0.001,优化器为Adam