基于迁移学习的齿轮箱开放集跨工况故障诊断

马翔; 许庶; 尚芃超; 马剑; 周汝志

doi:10.13700/j.bh.1001-5965.2022.0719

基于迁移学习的齿轮箱开放集跨工况故障诊断

doi: 10.13700/j.bh.1001-5965.2022.0719

马翔^{1, 2},
许庶^{1, 2},
尚芃超¹,
马剑^{1, 2, 3, 4, ,},
周汝志⁵

1.
北京航空航天大学可靠性与系统工程学院，北京 100191
2.
北京航空航天大学可靠性与环境工程技术重点实验室，北京 100191
3.
北京航空航天大学可靠性工程研究所，北京 100191
4.
北京航空航天大学国际创新研究院，杭州 311115
5.
上海卫星工程研究所，上海 201109

基金项目: 中央高校基本科研业务费专项资金(YWF-22-L-516)；民用飞机专项科研(MJ-2018-Y-58)

详细信息

通讯作者:
E-mail：09977@buaa.edu.cn

中图分类号: V240.2
计量
- 文章访问数: 509
- HTML全文浏览量: 259
- PDF下载量: 52
- 被引次数: 0
出版历程
- 收稿日期: 2022-08-17
- 录用日期: 2022-09-23
- 网络出版日期: 2022-12-28
- 整期出版日期: 2024-05-29

Fault diagnosis of gearbox under open set and cross working condition based on transfer learning

MA Xiang^{1, 2},
XU Shu^{1, 2},
SHANG Pengchao¹,
MA Jian^{1, 2, 3, 4
, ,},
ZHOU Ruzhi⁵

1.
School of Reliability and Systems Engineering，Beihang University，Beijing 100191，China
2.
Key Laboratory of Reliability and Environmental Engineering Technology，Beihang University，Beijing 100191，China
3.
Institute of Reliability Engineering，Beihang University，Beijing 100191，China
4.
International Innovation Institute，Beihang University， Hangzhou 311115，China
5.
Shanghai Institute of Satellite Engineering，Shanghai 201109，China

Funds: The Fundamental Research Funds for the Central Universities (YWF-22-L-516); Special Research on Civil Aircraft (MJ-2018-Y-58)

More Information

Corresponding author: E-mail：09977@buaa.edu.cn

摘要

摘要:
随着工业与航空航天技术的不断发展，齿轮箱等旋转机械的工况与故障模式逐渐趋于多样化、复杂化，可靠性与安全性问题日益突出，大量工况数据缺乏故障标签，且不同工况间故障模式不对称，迫切需要研究有效的故障诊断方法。以齿轮箱为案例验证对象，设置跨工况和开放集故障诊断场景，针对目标工况故障标签匮乏的问题，提出利用迁移学习将源工况的知识迁移到目标工况，利用交叉熵分类损失函数对已知故障类型进行识别的方法；针对跨工况条件下故障模式不对称的开放集问题，提出利用卷积神经网络提取工况间的相似数据特征，利用二分类损失函数对目标工况的已知类与未知类进行分类的方法。提出联合损失函数，训练诊断模型，实现故障特征从源域到目标域的联合迁移。案例分析结果表明：所提方法能够实现开放集情况下的跨工况故障诊断，且平均诊断准确度在90%以上。
- 旋转机械 /
- 数据驱动 /
- 开放集 /
- 迁移学习 /
- 故障诊断
Abstract:
With the continuous development of industry and aerospace technology, the working conditions and failure modes of rotating machinery are becoming increasingly diversified and complex, and reliability and safety problems are becoming increasingly prominent. It is critical to research efficient fault detection techniques since many working condition data lack fault labels and the failure modes amongst various working conditions are asymmetric. Take the gearbox as the case verification object, set up cross-working conditions, and open set fault diagnosis scenarios. A method is proposed to address the issue of lacking fault labels under the target working condition. It takes into account the ability of migration learning to facilitate cross-domain knowledge application. Specifically, migration learning is used to transfer knowledge from the source working condition to the target working condition, and the cross entropy classification loss function is used to identify known fault types. However, transfer learning has the problem that the greater the field difference is, the worse the effect is. It is difficult to solve the open set problem of asymmetric fault modes under cross-working conditions. In order to identify the known and unknown classes of target working circumstances, a method utilizing a convolutional neural network to extract similar data characteristics between working conditions is proposed. The two classification loss functions are then used in this process. The joint loss function is proposed to train the diagnosis model and realize the joint migration of fault features from the source domain to the target domain. The results of the case analysis show that the method can realize cross-working condition fault diagnosis under an open set, and the average diagnostic accuracy is more than 90%.
- rotating machinery /
- data-driven /
- open set /
- transfer learning /
- fault diagnosis

HTML全文

图 1 封闭集和开放集迁移学习分类任务场景对比

Figure 1. Scenario comparison of closed-set and open-set migration learning classification tasks

下载: 全尺寸图片幻灯片

图 2 基于J-CNN的开放集跨工况迁移故障诊断方法流程

Figure 2. J-CNN-based open-set cross-condition migration fault diagnosis methodology process

下载: 全尺寸图片幻灯片

图 3 基于J-CNN的开放集跨工况迁移故障诊断模型结构示意图

Figure 3. Schematic structure of open-set cross-condition migration fault diagnosis model based on J-CNN

下载: 全尺寸图片幻灯片

图 4 二级减速齿轮箱组成结构

Figure 4. Composition of secondary reduction gearbox

下载: 全尺寸图片幻灯片

图 5 35～40 Hz不同目标域工况数据占比诊断结果

Figure 5. Diagnostic results of percentage of working condition data in different target domains from 35 Hz to 40 Hz

下载: 全尺寸图片幻灯片

表 1 减速齿轮箱不同故障模式注入情况对比

Table 1. Comparison of different failure mode injections in reduction gearboxes

齿轮箱部件	故障模式	是否正常	故障1	故障2	故障3
齿轮	32T	正常	裂纹	正常	正常
	96T	正常	正常	正常	正常
	48T	正常	偏心	偏心	正常
	80T	正常	正常	断齿	正常
轴	输入	正常	正常	正常	不平衡
轴	输出	正常	正常	正常	正常
轴承	轴承1	正常	正常	滚动体	正常
	轴承2	正常	正常	正常	滚动体
	轴承3	正常	正常	正常	外圈
	轴承4	正常	正常	正常	正常
	轴承5	正常	正常	正常	正常
	轴承6	正常	正常	正常	正常

下载: 导出CSV

表 2 基于DANN的开放集故障诊断任务设置及诊断结果

Table 2. DANN-based open-set fault diagnosis task setup and diagnostic results

任务	跨工况情况	源域故障模式数	目标域故障模式数	准确度/％
P1	30～35 Hz	5	6	89.51
P2		4		85.64
P3		3		89.64
P4		2		78.85
P5		1		76.25
P6	35～30 Hz	5	6	73.04
P7		4		72.16
P8		3		83.57
P9		2		80.15
P10		1		81.93

下载: 导出CSV

参考文献(22)

[1]	SIAHPOUR S, LI X, LEE J. Deep learning-based cross-sensor domain adaptation for fault diagnosis of electro-mechanical actuators[J]. International Journal of Dynamics and Control, 2020, 8: 1054-1062. doi: 10.1007/s40435-020-00669-0
[2]	MOSLEM A, JASKARAN S, LI X, et al. Cross-domain gearbox diagnostics under variable working conditions with deep convolutional transfer learning[J]. Journal of Vibration and Control, 2020, 27(7-8): 854-864.
[3]	ZHANG W, LI X, MA H, et al. Open set domain adaptation in machinery fault diagnostics using instance-level weighted adversarial learning[J]. IEEE Transactions on Industrial Informatics, 2021, 17(11): 7445-7455. doi: 10.1109/TII.2021.3054651
[4]	陈宏, 邓芳明, 吴翔, 等. 基于梯度提升决策树的电力电子电路故障诊断[J]. 测控技术, 2017, 36(5): 9-12. doi: 10.3969/j.issn.1000-8829.2017.05.003 CHEN H, DENG F M, WU X, et al. Fault diagnosis of power electronics circuit based on gradient boosting decision tree[J]. Measurement & Control Technology, 2017, 36(5): 9-12(in Chinese). doi: 10.3969/j.issn.1000-8829.2017.05.003
[5]	王廷轩, 刘韬, 王振亚, 等. 融合迁移卷积神经网络的跨域滚动轴承故障诊断[J]. 电子测量技术, 2021, 44(10): 167-174. WANG T X, LIU T, WANG Z Y, et al. Cross-domain rolling bearing fault diagnosis based on convolutional neural network combined with transfer learning[J]. Electronic Measurement Technology, 2021, 44(10): 167-174(in Chinese).
[6]	洪晓翠, 段礼祥, 杨晓光, 等. 智能优化算法在机械故障诊断领域的应用综述[J]. 测控技术, 2021, 40(7): 1-8. HONG X C, DUAN L X, YANG X G, et al. Review on the application of intelligent optimization algorithms in mechanical fault diagnosis[J]. Measurement & Control Technology, 2021, 40(7): 1-8(in Chinese).
[7]	张根保, 李浩, 冉琰, 等. 一种用于轴承故障诊断的迁移学习模型[J]. 吉林大学学报(工学版), 2020, 50(5): 1617-1626. ZHANG G B, LI H, RAN Y, et al. A transfer learning model for bearing fault diagnosis[J]. Journal of Jilin University (Engineering and Technology Edition), 2020, 50(5): 1617-1626(in Chinese).
[8]	刘应东, 刘韬, 李华, 等. 变工况轴承的联合分布适应迁移故障诊断[J]. 电子测量与仪器学报, 2021, 35(5): 69-75. LIU Y D, LIU T, LI H, et al. Transfer fault diagnosis of bearings under variable working conditions based on joint distribution adaptation[J]. Journal of Electronic Measurement and Instrumentation, 2021, 35(5): 69-75(in Chinese).
[9]	毕爽. 基于深度迁移学习的风电机组齿轮箱故障诊断方法研究[D]. 长春: 长春大学, 2021. BI S. Research on fault diagnosis method of wind turbine gearbox based on deep transfer learning[D]. Changchun: Changchun University, 2021(in Chinese).
[10]	陈仁祥, 杨星, 胡小林, 等. 深度置信网络迁移学习的行星齿轮箱故障诊断方法[J]. 振动与冲击, 2021, 40(1): 127-133. CHEN R X, YANG X, HU X L, et al. Planetary gearbox fault diagnosis method based on deep belief network transfer learning[J]. Journal of Vibration and Shock, 2021, 40(1): 127-133(in Chinese).
[11]	CAO P, ZHANG S L, TANG J. Preprocessing-free gear fault diagnosis using small datasets with deep convolutional neural network-based transfer learning[J]. IEEE Access, 2018, 6: 26241-26253. doi: 10.1109/ACCESS.2018.2837621
[12]	MA P, ZHANG H L, FAN W H, et al. A novel bearing fault diagnosis method based on 2D image representation and transfer learning-convolutional neural network[J]. Measurement Science and Technology, 2019, 30: 055402. doi: 10.1088/1361-6501/ab0793
[13]	CHEN Z Y, HE G L, LI J P, et al. Domain adversarial transfer network for cross-domain fault diagnosis of rotary machinery[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 69(11): 8702-8712. doi: 10.1109/TIM.2020.2995441
[14]	ZHAO C, SHEN W M. Dual adversarial network for cross-domain open set fault diagnosis[J]. Reliability Engineering & System Safety, 2022, 221: 108358.
[15]	文成林, 吕菲亚. 基于深度学习的故障诊断方法综述[J]. 电子与信息学报, 2020, 42(1): 234-248. doi: 10.11999/JEIT190715 WEN C L, LV F Y. Review on deep learning based fault diagnosis[J]. Journal of Electronics & Information Technology, 2020, 42(1): 234-248(in Chinese). doi: 10.11999/JEIT190715
[16]	CHEN W Q, QIU Y N, FENG Y H, et al. Diagnosis of wind tur-bine faults with transfer learning algorithms[J]. Renewable Energy, 2021, 163: 2053-2067. doi: 10.1016/j.renene.2020.10.121
[17]	PENG C, LI L L, CHEN Q, et al. A fault diagnosis method for rolling bearings based on parameter transfer learning under imbalance data sets[J]. Energies, 2021, 14(4): 1-18.
[18]	BUSTO P P, GALL J. Open set domain adaptation[C]//Proceedings of the IEEE International Conference on Computer Vision. Piscataway: IEEE Press, 2017: 754-763.
[19]	LI X, ZHANG W, DING Q, et al. Intelligent rotating machinery fault diagnosis based on deep learning using data augmentation[J]. Journal of Intelligent Manufacturing, 2020, 31(2): 433-452. doi: 10.1007/s10845-018-1456-1
[20]	TANG S N, YUAN S Q, ZHU Y. Deep learning-based intelligent fault diagnosis methods toward rotating machinery[J]. IEEE Access, 2020, 8: 9335-9346. doi: 10.1109/ACCESS.2019.2963092
[21]	LI F, TANG T J, TANG B P, et al. Deep convolution domain-adversarial transfer learning for fault diagnosis of rolling bearings[J]. Measurement, 2021, 169: 108339. doi: 10.1016/j.measurement.2020.108339
[22]	郭瑜, 童丽娜, 倪旭明. 基于改进卷积神经网络的电力通信网故障诊断研究[J]. 计算机测量与控制, 2022, 30(2): 24-30. GUO Y, TONG L N, YAN X M. Research on fault diagnosis of electric power communication network based on improved convolutional neural network[J]. Computer Measurement & Control, 2022, 30(2): 24-30(in Chinese).