基于深度学习的矿井滚动轴承故障诊断方法

窦桂东; 白艺硕; 王均利; 黄博昊; 阳康

doi:10.13272/j.issn.1671-251x.2023070085

基于深度学习的矿井滚动轴承故障诊断方法

doi: 10.13272/j.issn.1671-251x.2023070085

1.
陕西彬长小庄矿业有限公司，陕西咸阳　713500
2.
中国矿业大学（北京）人工智能学院，北京　100083
3.
陕西彬长矿业集团有限公司，陕西咸阳　712000

基金项目: 国家自然科学基金资助项目（52074305）。

详细信息

作者简介:
窦桂东（1983—），男，山东泰安人，高级工程师，从事煤矿智能化研究和应用工作，E-mail：313934867@qq.com

通讯作者:
白艺硕（1999—），女，河南南阳人，硕士研究生，研究方向为深度学习与故障诊断，E-mail：bys1311@163.com。

中图分类号: TD67
计量
- 文章访问数: 733
- HTML全文浏览量: 74
- PDF下载量: 101
- 被引次数: 0
出版历程
- 收稿日期: 2023-07-24
- 修回日期: 2024-01-12
- 网络出版日期: 2024-01-31

A fault diagnosis method for mine rolling bearings based on deep learning

1.
Shaanxi Binchang Xiaozhuang Mining Co., Ltd., Xianyang 713500, China
2.
School of Artificial Intelligence, China University of Mining and Technology-Beijing, Beijing 100083, China
3.
Shaanxi Binchang Mining Group Co., Ltd., Xianyang 712000, China

摘要

摘要: 针对传统卷积神经网络在煤矿井下等复杂环境中难以充分挖掘数据特征等问题，提出了一种基于马尔可夫转移场（MTF）和双通道多尺度卷积胶囊网络（DMCCN）的矿井滚动轴承故障诊断方法，构建了MTF−DMCCN故障诊断模型。根据MTF和灰度图对原始振动信号进行编码后，采用双通道输入模式连接卷积网络获取浅层特征；将特征图进行融合后输入到胶囊网络，提高模型对空间信息的敏感度；在网络中引入Inception模块，聚焦多尺度特征，加强网络的特征提取能力；通过胶囊层进行向量化处理，实现滚动轴承的故障诊断与分类。消融实验、抗噪性及泛化性实验结果表明：Inception模块、灰度图输入、MTF图像输入均对轴承故障诊断具有正向促进的作用，MTF编码对模型的诊断精度提升最高；MTF−DMCCN模型具有较好的鲁棒性和抗噪声能力；MTF−DMCCN模型具有优异的变转速适应能力，在不同工况条件下具有良好的泛化性能。为进一步验证模型性能，选取格拉姆角差场（GADF）、格拉姆角和场（GASF）、灰度图、MTF等图像编码方式与不同网络相结合，采用辛辛那提大学数据集（IMS）进行对比实验，结果表明，MTF−DMCCN模型能有效识别滚动轴承故障类型，平均故障诊断准确率达99.37%。
- 滚动轴承 /
- 故障诊断 /
- 马尔可夫转移场 /
- 胶囊网络 /
- Inception结构 /
- MTF编码
Abstract: A fault diagnosis method for mine rolling bearings based on Markov transition field(MTF) and dual-channel multi-scale convolutional capsule network (DMCCN) is proposed to address the problem of traditional convolutional neural networks being unable to fully explore data features in complex environments such as coal mines. The MTF-DMCCN fault diagnosis model is constructed. After encoding the original vibration signal based on MTF and grayscale image, a dual channel input mode is used to connect the convolutional network to obtain shallow features. The method inputs the feature maps fusion into the capsule network to improve the sensitivity of the model to spatial information. The method introduces Inception modules into the network to focus on multi-scale features and enhance the network's feature extraction capabilities. Finally, vectorization processing is carried out through the capsule layer to achieve fault diagnosis and classification of rolling bearings. The results of ablation, noise resistance, and generalization experiments show that the Inception module, grayscale image input, and MTF image input all have a positive promoting effect on bearing fault diagnosis. MTF coding has the highest improvement in diagnostic precision of the model. The MTF-DMCCN model has good robustness and noise resistance. The MTF-DMCCN model has excellent adaptability to variable speed and still exhibits good generalization performance under different operating conditions. To further validate the performance of the model, image encoding methods such as Gram angle difference field (GADF), Gram angle sum field (GASF), grayscale image, and MTF are selected and combined with different networks. Comparative experiments are conducted using the University of Cincinnati intelligent maintenance system (IMS). The results show that the MTF-DMCCN model can effectively recognize the type of rolling bearing faults, with an average fault diagnosis accuracy of 99.37%.
- rolling bearings /
- fault diagnosis /
- Markov transition field /
- capsule network /
- Inception structure /
- MTF encoding

HTML全文

图 1 胶囊网络结构

Figure 1. Capsule network structure

下载: 全尺寸图片幻灯片

图 2 动态路由算法结构

Figure 2. Dynamic routing algorithm structure

下载: 全尺寸图片幻灯片

图 3 MTF−DMCCN故障诊断模型

Figure 3. Fault diagnosis model based on Markov transition field(MTF) and dual-channel multi-scale convolutional capsule network(DMCCN)

下载: 全尺寸图片幻灯片

图 4 胶囊传递中的权值更新

Figure 4. Weight update in capsule delivery

下载: 全尺寸图片幻灯片

图 5 不同尺寸的MTF图像

Figure 5. MTF images of different sizes

下载: 全尺寸图片幻灯片

图 6 不同模型的混淆矩阵

Figure 6. Confusion matrix of different models

下载: 全尺寸图片幻灯片

图 7 各模型在不同噪声环境下的混淆雷达图

Figure 7. Confusing radargrams of each model in different noise environment

下载: 全尺寸图片幻灯片

图 8 IMS数据集下不同模型的识别准确率

Figure 8. Recognition accuracy of different models in the IMS dataset

下载: 全尺寸图片幻灯片

表 1 单一工况下轴承故障数据组成

Table 1. Composition of bearing failure data under single operating conditions

样本名称	样本类型	样本个数	标签
IF07	内圈故障	600	0
IF14	内圈故障	600	1
IF21	内圈故障	600	2
OF07	外圈故障	600	3
OF14	外圈故障	600	4
OF21	外圈故障	600	5
BF07	滚动体故障	600	6
BF14	滚动体故障	600	7
BF21	滚动体故障	600	8
VMF	垂直不对中故障	600	9
HMF	水平不对中故障	600	10
N	正常状态	600	11

下载: 导出CSV

表 2 不同工况条件下的数据集参数

Table 2. Dataset parameters under different operating conditions

数据集	电动机负载/kW	电动机转速/（r·min⁻¹）	样本个数
A	0	1 797	7 200
B	0.746	1 772	7 200
C	1.491	1 750	7 200

下载: 导出CSV

表 3 不同模型的识别结果

Table 3. Recognition results of different models

模型	识别准确率/%	运行时间/s
MTF−DMCCN	99.44	156.84
MTF−DCCN	83.72	141.15
MTF−MCCN	94.61	315.29
DMCCN	72.94	105.98

下载: 导出CSV

表 4 变工况下的故障识别准确率

Table 4. Fault recognition accuracy under variable operating conditions %

实验工况	识别准确率
实验工况	实验1	实验2	实验3	实验4	实验5	均值
A→B	81.5	90.6	80.4	80.7	82.3	83.1
A→C	89.2	87.6	88.9	91.3	84.0	88.2
B→A	78.5	71.6	78.1	73.6	79.7	76.3
B→C	80.2	77.1	79.3	83.1	81.8	80.3
C→A	77.6	76.9	77.2	80.9	83.4	79.2
C→B	81.7	91.0	82.3	86.6	80.9	84.5

下载: 导出CSV

参考文献(20)

[1]	张旭辉,潘格格,郭欢欢,等. 基于深度迁移学习的采煤机摇臂部滚动轴承故障诊断方法[J]. 煤炭科学技术,2022,50(4):256-263. ZHANG Xuhui,PAN Gege,GUO Huanhuan,et al. Fault diagnosis method for rolling bearing on shearer arm based on deep transfer learning[J]. Coal Science and Technology,2022,50(4):256-263.
[2]	郭秀才,吴妮,曹鑫. 基于特征融合与DBN的矿用通风机滚动轴承故障诊断[J]. 工矿自动化,2021,47(10):14-20,26. GUO Xiucai,WU Ni,CAO Xin. Fault diagnosis of rolling bearing of mine ventilator based on characteristic fusion and DBN[J]. Industry and Mine Automation,2021,47(10):14-20,26.
[3]	ZHANG Xiaochen,CONG Yiwen,YUAN Zhe,et al. Early fault detection method of rolling bearing based on MCNN and GRU network with an attention mechanism[J]. Shock and Vibration,2021. DOI: 10.1155/2021/6660243.
[4]	ZHENG Zhi,FU Jiuman,LU Chuanqi,et al. Research on rolling bearing fault diagnosis of small dataset based on a new optimal transfer learning network[J]. Measurement,2021,177. DOI: 10.1016/J.MEASUREMENT.2021.109285.
[5]	史志远,滕虎,马驰. 基于多信息融合和卷积神经网络的行星齿轮箱故障诊断[J]. 工矿自动化,2022,48(9):56-62. SHI Zhiyuan,TENG Hu,MA Chi. Fault diagnosis of planetary gearbox based on multi-information fusion and convolutional neural network[J]. Journal of Mine Automation,2022,48(9):56-62.
[6]	姚齐水,别帅帅,余江鸿,等. 一种结合改进Inception V2模块和CBAM的轴承故障诊断方法[J]. 振动工程学报,2022,35(4):949-957. YAO Qishui,BIE Shuaishuai,YU Jianghong,et al. A bearing fault diagnosis method combining improved inception V2 module and CBAM[J]. Journal of Vibration Engineering,2022,35(4):949-957.
[7]	SABOUR S,FROSST N,HINTON G E. Dynamic routing between capsules[EB/OL]. [2023-06-05]. https://arxiv.org/abs/1710.09829.
[8]	王超群,李彬彬,焦斌. 基于门控循环单元胶囊网络的滚动轴承故障诊断[J]. 轴承,2021(5):56-62. WANG Chaoqun,LI Binbin,JIAO Bin. Fault diagnosis for rolling bearings based on capsule network of gated recurrent unit[J]. Bearing,2021(5):56-62.
[9]	CHEN Tianyou,WANG Zhihua,YANG Xiang,et al. A deep capsule neural network with stochastic delta rule for bearing fault diagnosis on raw vibration signals[J]. Measurement,2019,148. DOI: 10.1016/j.measurement.2019.106857.
[10]	WEN Long,LI Xinyu,GAO Liang,et al. A new convolutional neural network-based data-driven fault diagnosis method[J]. IEEE Transactions on Industrial Electronics,2018,65(7):5990-5998. doi: 10.1109/TIE.2017.2774777
[11]	LIANG Pengfei,DENG Chao,WU Jun,et al. Single and simultaneous fault diagnosis of gearbox via a semi-supervised and high-accuracy adversarial learning framework[J]. Knowledge-Based Systems,2020,198. DOI: 10.1016/j.knosys.2020.105895.
[12]	YAN Jialin,KAN Jiangming,LUO Haifeng. Rolling bearing fault diagnosis based on Markov transition field and residual network[J]. Sensors,2022,22(10). DOI: 10.3390/S22103936.
[13]	WANG Mengjiao,WANG Wenjie,ZHANG Xinan,et al. A new fault diagnosis of rolling bearing based on Markov transition field and CNN[J]. Entropy,2022,24(6). DOI: 10.3390/E24060751.
[14]	姜家国,郭曼利. 基于MTF和DenseNet的滚动轴承故障诊断方法[J]. 工矿自动化,2022,48(9):63-68. JIANG Jiaguo,GUO Manli. Fault diagnosis method of rolling bearing based on MTF and DenseNet[J]. Journal of Mine Automation,2022,48(9):63-68.
[15]	赵志宏,李春秀,窦广鉴,等. 基于MTF−CNN的轴承故障诊断研究[J]. 振动与冲击,2023,42(2):126-131. ZHAO Zhihong,LI Chunxiu,DOU Guangjian,et al. Bearing fault diagnosis method based on MTF-CNN[J]. Journal of Vibration and Shock,2023,42(2):126-131.
[16]	瞿红春,朱伟华,高鹏宇,等. 基于注意力循环胶囊网络的滚动轴承故障诊断[J]. 振动. 测试与诊断,2022,42(6):1108-1114,1243. QU Hongchun,ZHU Weihua,GAO Pengyu,et al. Fault diagnosis of rolling bearing based on attention recurrent capsule network[J]. Journal of Vibration,Measurement & Diagnosis,2022,42(6):1108-1114,1243.
[17]	PECHYONKIN M. Understanding Hinton's capsule networks. Part 3. Dynamic routing between capsules[EB/OL]. [2023-06-05]. https://pechyonkin.me/capsules-3/.
[18]	Bearing Data Center of Case Western Reserve University. Seeded fault test data [EB/OL]. [2023-06-05]. https://engineering.case.edu/bearingdatacenter/.
[19]	LEE J,QIU H,YU G,et al. Bearing data set[EB/OL]. [2023-06-05]. https://data.nasa.gov/download/brfb-gzcv/application%2Fzip.
[20]	ZHANG Wei,PENG Gaoliang,LI Chuanhao,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(2). DOI: 10.3390/s17020425.