Intelligent Fault Diagnosis Under Variable Loads Based on the Embedded Gear Fault Mechanism

doi:10.12068/j.issn.1005-3026.2025.20230296

Abstract

Abstract:

Under variable load conditions， machine learning-based gear fault diagnosis models face the challenge of relying on specific target condition samples for training. To overcome this limitation， the feature components in the signal that can reflect the health status of gears and remain invariant to load variations were solved based on the gear fault mechanism， thereby constructing a fault frequency waveform convolution module and embedding it into the convolutional neural network. Additionally， to enhance the network’s feature extraction capability， a multi-scale attention module was introduced. Based on these modules， a variable load gear fault diagnosis model named FWaveNet was constructed and applied to the gear fault dataset from Northeastern University. The results showed that its diagnostic accuracy is significantly better than that of existing models. Through specific signal processing techniques and network architecture design， precise identification of gear health status under load fluctuations is achieved， and a solution for engineering applications in the fault diagnosis of variable load gears is provided.

Key words: deep learning, variable load, fault diagnosis, fault mechanism, gear

CLC Number:

TH 165+3

Bin YU, Hong-chun SUN, Da-yong YE. Intelligent Fault Diagnosis Under Variable Loads Based on the Embedded Gear Fault Mechanism[J]. Journal of Northeastern University(Natural Science), 2025, 46(4): 61-70.

Figures/Tables 13

References 24

1	张佳雄，魏静，张春鹏，等. 高阶调谐齿轮参数设计及动态响应研究［J］.振动工程学报，2020，35（2）：369-378.
	Zhang Jia-xiong， Wei Jing， Zhang Chun-peng， et al. Parameter design and dynamic response study of a high-order tuning gear［J］. Journal of Vibration Engineering， 2020，35（2）：369-378.
2	Li X Y， Kong X W， Liu Z D， et al. A novel framework for early pitting fault diagnosis of rotating machinery based on dilated CNN combined with spatial dropout［J］. IEEE Access， 2021， 9： 29243-29252.
3	Emmanuel S， Yihun Y， Nili A Z， et al. Planetary gear train microcrack detection using vibration data and convolutional neural networks［J］. Neural Computing and Applications， 2021， 33（24）： 17223-17243.
4	张智禹，尹爱军，谭建. 融合注意力机制的改进DBN变工况齿轮箱故障诊断方法［J］. 振动与冲击， 2021， 40（14）： 47-52.
	Zhang Zhi-yu， Yin Ai-jun， Tan Jian. Improved DBN method with attention mechanism for the fault diagnosis of gearboxes under varying working condition［J］. Journal of Vibration and Shock， 2021， 40（14）： 47-52.
5	Zhang R， Tao H Y， Wu L F， et al. Transfer learning with neural networks for bearing fault diagnosis in changing working conditions［J］. IEEE Access， 2017， 5： 14347-14357.
6	Stacke K， Eilertsen G， Unger J， et al. Measuring domain shift for deep learning in histopathology［J］. IEEE Journal of Biomedical and Health Informatics， 2021， 25（2）： 325-336.
7	Li J L， Li X Y， He D， et al. A domain adaptation model for early gear pitting fault diagnosis based on deep transfer learning network［J］. Proceedings of the Institution of Mechanical Engineers， Part O： Journal of Risk and Reliability， 2020， 234（1）： 168-182.
8	Zhang Z W， Chen H H， Li S M， et al. Sparse filtering based domain adaptation for mechanical fault diagnosis［J］. Neurocomputing， 2020， 393： 101-111.
9	赵桐，雷保珍，王训伟，等. 基于迁移学习和特征重用的MIM齿轮缺陷检测［J］. 传感器与微系统， 2021， 40（10）： 129-131， 135.
	Zhao Tong， Lei Bao-zhen， Wang Xun-wei， et al. MIM gear defect detection based on transfer learning and feature reuse［J］. Transducer and Microsystem Technologies， 2021， 40（10）： 129-131， 135.
10	Meng Z， Shi G X， Wang F L. Vibration response and fault characteristics analysis of gear based on time-varying mesh stiffness［J］. Mechanism and Machine Theory， 2020， 148： 103786.
11	Zghal B， Graja O， Dziedziech K， et al. A new modeling of planetary gear set to predict modulation phenomenon［J］. Mechanical Systems and Signal Processing， 2019， 127： 234-261.
12	Luo W， Qiao B J， Shen Z X， et al. A dynamic model to predict modulation sidebands of planetary gear sets with localized planet faults［C］//2018 Prognostics and System Health Management Conference （PHM-Chongqing）. Chongqing， 2018： 1269-1273.
13	Luo H G， Hatch C， Hanna J， et al. Amplitude modulations in planetary gears［J］. Wind Energy， 2014， 17（4）： 505-517.
14	Feng Z P， Zuo M J. Vibration signal models for fault diagnosis of planetary gearboxes［J］. Journal of Sound and Vibration， 2012， 331（22）： 4919-4939.
15	Wang C D， Sun H C， Cao X. Construction of the efficient attention prototypical net based on the time-frequency characterization of vibration signals under noisy small sample［J］. Measurement， 2021， 179： 109412.
16	Zhang D， Chen Y Y， Guo F H， et al. A new interpretable learning method for fault diagnosis of rolling bearings［J］. IEEE Transactions on Instrumentation and Measurement， 2020， 70： 3507010.
17	Zhang Q S， Wu Y N， Zhu S C. Interpretable convolutional neural networks［C］//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City， 2018： 8827-8836.
18	Krizhevsky A， Sutskever I， Hinton G E. ImageNet classification with deep convolutional neural networks［J］. Communications of the ACM， 2017， 60（6）： 84-90.
19	He K M， Zhang X Y， Ren S Q， et al. Deep residual learning for image recognition［C］//2016 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）. Las Vegas， 2016： 770-778.
20	Song X D， Cong Y Y， Song Y F， et al. A bearing fault diagnosis model based on CNN with wide convolution kernels［J］. Journal of Ambient Intelligence and Humanized Computing， 2022， 13（8）： 4041-4056.
21	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.
22	Hao S J， Ge F X， Li Y M， et al. Multisensor bearing fault diagnosis based on one-dimensional convolutional long short-term memory networks［J］. Measurement， 2020， 159： 107802.
23	Wang K， Zhao W， Xu A D， et al. One-dimensional multi-scale domain adaptive network for bearing-fault diagnosis under varying working conditions［J］. Sensors， 2020， 20（21）： 6039.
24	Ganin Y， Ustinova E， Ajakan H， et al. Domain-adversarial training of neural networks［J］.Journal of Machine Learning Research， 2017，17（1）：2096-2030．

层	层的类型	核尺寸	步幅	输出
1_FWave Layer	卷积核	1×9	1	2×1024
2_Multi-scale attention	卷积核	1×1, 1×3, 1×5, 1×7	1	24×1024
2_Pool	最大池化	1×2	2	24×512
3_ Multi-scale attention	卷积核	1×1, 1×3, 1×5, 1×7	1	96×512
3_Pool	最大池化	1×2	2	96×256
4_Conv	卷积核	1×3	1	192×256
4_Pool	最大池化	1×2	2	192×128
5_Conv	卷积核	1×3	2	192×64
GAP	全局平均池化	—	—	192×1
Full connected_1	全连接层	—	—	128
Full connected_2	全连接层	—	—	32
Full connected_3	全连接层	—	—	类别

层	层的类型	核尺寸	步幅	输出
1_FWave Layer	卷积核	1×9	1	2×1024
2_Multi-scale attention	卷积核	1×1, 1×3, 1×5, 1×7	1	24×1024
2_Pool	最大池化	1×2	2	24×512
3_ Multi-scale attention	卷积核	1×1, 1×3, 1×5, 1×7	1	96×512
3_Pool	最大池化	1×2	2	96×256
4_Conv	卷积核	1×3	1	192×256
4_Pool	最大池化	1×2	2	192×128
5_Conv	卷积核	1×3	2	192×64
GAP	全局平均池化	—	—	192×1
Full connected_1	全连接层	—	—	128
Full connected_2	全连接层	—	—	32
Full connected_3	全连接层	—	—	类别

健康状态	样本数
健康状态	A	B	C
健康	1 000	1 000	1 000
齿根裂纹	1 000	1 000	1 000
齿面剥落	1 000	1 000	1 000

健康状态	样本数
健康状态	A	B	C
健康	1 000	1 000	1 000
齿根裂纹	1 000	1 000	1 000
齿面剥落	1 000	1 000	1 000

模型	A→B	A→C	B→A	B→C	C→A	C→B	平均值
Baseline	98.96±0.12	78.41±1.33	99.47±0.1	85.38±0.74	87.91±1.05	91.64±1.55	90.30±0.37
B+FWave	99.96±0.01	87.72±1.48	99.95±0.01	88.93±1.07	95.02±0.70	98.83±0.18	95.07±0.33
B+MA	99.07±0.13	79.22±0.91	99.68±0.05	86.93±0.47	89.25±1.34	92.36±0.66	91.09±0.33
本文方法	99.98±0.01	94.82±0.34	99.97±0.01	98.32±0.27	97.62±0.18	99.45±0.05	98.36±0.14