Meta-DAE Fault Diagnosis Based on Prototype Domain Enhancement in Few-Shot

doi:10.12141/j.issn.1000-565X.230715

Abstract

Abstract:

Rolling bearings, as a type of precision mechanical component, are widely used in modern industrial machinery and equipment. It is of great significance to diagnose bearing faults using reasonable methods during bea-ring operation. However, in the actual complex and ever-changing environment, the collection of vibration signals often faces challenges such as limited sample sizes, noise interference, and operating condition variations, resulting in low fault diagnosis accuracy. To address the problem of small-sample rolling bearing fault diagnosis under noise interference and variable operating conditions, this paper proposed a meta-learning denoising model based on prototype domain enhancement (Meta-DAE). Firstly, a small-sample fault dataset based on time-frequency diagrams was constructed, and a deep convolutional generative adversarial network was introduced for data preprocessing to gene-rate a pseudo-sample set with a similar distribution. Then, the fault sample set was input into Meta-DAE for adaptive feature extraction. Meta-DAE adopts a prototype domain enhancement strategy to make prototype points of the same category more closely clustered in the embedding space. At the same time, an encoder with noise reduction performance was constructed, and a target function based on prototype domain enhancement and denoising was designed. By fine-tuning the model under small-sample conditions, the noise robustness and classification accuracy of the model were improved. Experimental results of small-sample fault diagnosis under noise interference and variable operating conditions show that, compared to other models, the proposed model demonstrates strong noise robustness. Under -8 dB strong noise interference, the model achieves a classification accuracy improvement of 35.78% to 57.25% using only 10 samples for fine-tuning.

Key words: few-shot, fault diagnosis, meta learning, prototype domain enhancement, denoising autoencoder

CLC Number:

TH165.3

MA Ping, LIANG Cheng, WANG Cong, LI Xinkai, ZHANG Hongli. Meta-DAE Fault Diagnosis Based on Prototype Domain Enhancement in Few-Shot[J]. Journal of South China University of Technology(Natural Science Edition), 2025, 53(1): 62-73.

Figures/Tables 20

Fig.1

Fig.2

Table 1

Fig.3

Fig.4

Table 2

Fig.5

Table 3

Table 4

Table 5

Classification accuracy of Meta-DAE model under different prototype domain enhancement factors"

$η$	10类-1样本划分策略下的分类准确率/%			10类-5样本划分策略下的分类准确率/%
$η$	样本数为10	样本数为20	样本数为30	样本数为10	样本数为20	样本数为30
0.0	94.56	98.82	99.30	97.56	99.16	99.56
0.2	97.25	99.25	99.73	98.86	99.31	99.76
0.4	98.67	99.55	99.76	99.16	99.37	99.89
0.6	99.16	99.67	99.83	99.64	99.87	99.98
0.8	99.12	99.42	99.86	99.38	99.64	99.92
1.0	98.07	99.25	99.68	99.12	99.52	99.86

Table 5

Fig.6

Table 6

Fig.7

Fig.8

Table 7

Fig.9

Table 8

Table 9

Table 10

Table 11

References 26

1	HU X S， ZHANG K， LIU K L，et al ．Advanced fault diagnosis for lithium-ion battery systems：a review of fault mechanisms，fault features，and diagnosis procedures［J］．IEEE Industrial Electronics Magazine，2020，14（3）：65-91.
2	KANKAR P K， SHARMA S C， HARSHA S P ．Fault diagnosis of ball bearings using machine learning methods［J］．Expert Systems with Applications，2011，38（3）：1876-1886.
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，2020，138：106587/1-39.
4	FENG Y， CHEN J L， XIE J S，et al ．Meta-learning as a promising approach for few-shot cross-domain fault diagnosis：algorithms，applications，and prospects［J］．Knowledge-Based Systems，2022，235：107646/1-27.
5	HOANG D T， KANG J ．A survey on deep learning based bearing fault diagnosis［J］．Neurocomputing，2019，335：327-335.
6	JIA F， LEI Y， GUO L，et al ．A neural network constructed by deep learning technique and its application to intelligent fault diagnosis of machines［J］．Neurocomputing，2018，272：619-628.
7	WEN L， LI X Y， GAO L，et al ．A new convolutional neural network-based data-driven fault diagnosis method［J］．IEEE Transactions on Industrial Electronics，2017，65（7）：5990-5998.
8	ZHANG W， PENF 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（2）：425/1-21.
9	ZHAO M H， ZHONG S S， FU X Y，et al ．Deep residual shrinkage networks for fault diagnosis［J］．IEEE Transactions on Industrial Informatics，2019，16（7）：4681-4690.
10	万若青，张纯，江汇强，等．基于深度自编码器的振动信号盲去噪方法［J］．振动与冲击，2023，42（12）：118-125.
	WAN Ruoqing， ZHANG Chun， JIANG Huiqiang，et al ．A blind denoising method of vibration signals based on a deep autoencoder［J］．Journal of Vibration and Shock，2023，42（12）：118-125.
11	DONG Y J， LI Y Q， ZHENG H L，et al ．A new dynamic model and transfer learning based intelligent fault diagnosis framework for rolling element bearings race faults：solving the small sample problem［J］．ISA Transactions，2022，121：327-348.
12	刘飞，陈仁文，邢凯玲，等．基于迁移学习与深度残差网络的滚动轴承快速故障诊断算法［J］．振动与冲击，2022，41（3）：154-164.
	LIU Fei， CHEN Renwen， XING Kailing，et al ．Fast fault diagnosis algorithm for rolling bearings based on transfer learning and deep residual network［J］．Journal of Vibration and Shock，2022，41（3）：154-164.
13	CHEN H， LIU R N， XIE Z X，et al ．Majorities help minorities：hierarchical structure guided transfer lear-ning for few-shot fault recognition［J］．Pattern Recognition，2022，123：108383/1-15.
14	ZHANG T C， CHEN J L， HE S L，et al ．Prior knowledge-augmented self-supervised feature learning for few-shot intelligent fault diagnosis of machines［J］．IEEE Transactions on Industrial Electronics，2022，69（10）：10573-10584.
15	强睿儒，赵小强．基于格拉姆角差场和生成对抗网络的小样本滚动轴承故障诊断方法［J］．华南理工大学学报（自然科学版），2024，52（10）：64-75.
	QIANG Ruiru， ZHAO Xiaoqiang.Fault diagnosis method for rolling bearings based on Gram angular field and generative adversarial networks［J］．Journal of South China University of Tchnology（Natural Science Edition），2024，52（10）：64-75.
16	胡若晖，张敏，许文鑫．基于DCGAN和DANN网络的滚动轴承跨域故障诊断［J］．振动与冲击，2022，41（6）：21-29.
	HU Ruohui， ZHANG Min， XU Wenxin ．Cross-domain fault diagnosis of rolling element bearings using DCGAN and DANN［J］．Journal of Vibration and Shock，2022，41（6）：21-29.
17	LEI T Y， HU H H， LUO Q Y，et al ．Adaptive meta-learner via gradient similarity for few-shot text classification［EB/OL］．（2022-09-10）［2023-06-20］．.
18	CHEN Y D， GUAN C Y， WEI Z K，et al ．Meta-Delta：a meta-learning system for few-shot image cla-ssification［J］．Proceedings of Machine Learning Research，2021，140：17-28.
19	HOSPEDALES T， ANTREAS A， MICAELL P，et al ．Meta-learning in neural networks：a survey［J］．IEEE Transactions on Pattern Analysis and Machine Intelligence，2021，44（9）：5149-5169.
20	LI C J， LI S B， ZHANG A S，et al ．Meta-learning for few-shot bearing fault diagnosis under complex working conditions［J］．Neurocomputing，2021，439：197-211.
21	WANG D， ZHANG M， XU Y C，et al ．Metric-based meta-learning model for few-shot fault diagnosis under multiple limited data conditions［J］．Mechanical Systems and Signal Processing，2021，155：107510/1-15.
22	FENG Y， CHEN J L， ZHANG T C，et al ．Semi-supervised meta-learning networks with squeeze-and-excitation attention for few-shot fault diagnosis［J］．ISA Transactions，2022，120：383-401.
23	YAN R Q， GAO R X， CHEN X F ．Wavelets for fault diagnosis of rotary machines：a review with applications ［J］．Signal Processing，2014，96：1-15.
24	RADFORD A， METZ L， CHINTALA S ．Unsupervised representation learning with deep convolutional generative adversarial networks［EB/OL］．（2015-11-19）［2023-06-20］．.
25	SNELL J， SWERSKY K， ZEMEL R ．Prototypical networks for few-shot learning［C］∥ Proceedings of the 31st International Conference on Neural Information Processing Systems．Long Beach：ACM，2017：4080-4090.
26	SAUFI S R， AHMAD Z A B， LEONG M S，et al ．Gearbox fault diagnosis using a deep learning model with limited data sample［J］．IEEE Transactions on Industrial Informatics，2020，16（10）：6263-6271.

模型	网络层	操作	激活函数
生成器	Input（1，100）
	T-Conv2d（100，512）	BN	ReLU
	T-Conv2d（512，256）	BN	ReLU
	T-Conv2d（256，128）	BN	ReLU
	T-Conv2d（128，64）	BN	ReLU
	T-Conv2d（64，3）	BN	ReLU
判别器	Conv2d（3，64）		LeakyReLU
	Conv2d（64，128）	BN	LeakyReLU
	Conv2d（128，256）	BN	LeakyReLU
	Conv2d（256，512）	BN	LeakyReLU
	Conv2d（512，1）		Sigmoid

模型	网络层	操作	激活函数
编码器	Conv2d（3，32）	BN
	Conv2d（32，64）	BN Maxpool	ReLU
	Conv2d（64，64）	BN	ReLU
	Conv2d（64，64）	BN	ReLU
	Conv2d（64，128）	BN	ReLU
	Conv2d（128，128）	BN Maxpool	ReLU
解码器	T-Conv2d（128，128）	BN	ReLU
	T-Conv2d（128，64）	BN	ReLU
	T-Conv2d（64，64）	BN	ReLU
	T-Conv2d（64，32）	BN	ReLU
	T-Conv2d（32，16）	BN	ReLU
	T-Conv2d（16，3）		Sigmoid

轴承状态	故障尺寸/mm	标签号
正常NC	0.000 0	0
IF	0.177 8	1
	0.355 6	2
	0.553 4	3
OF	0.177 8	4
	0.355 6	5
	0.553 4	6
RF	0.177 8	7
	0.355 6	8
	0.553 4	9

工况	数据集	样本数
工况	数据集	NC	IF-7	IF-14	IF-21	OF-7	OF-14	OF-21	RF-7	RF-14	RF-21
A	训练集	90	90	90	90	90	90	90	90	90	90
A	测试数	210	210	210	210	210	210	210	210	210	210
B	训练集	90	90	90	90	90	90	90	90	90	90
B	测试集	210	210	210	210	210	210	210	210	210	210
C	训练集	90	90	90	90	90	90	90	90	90	90
C	测试集	210	210	210	210	210	210	210	210	210	210
D	训练集	90	90	90	90	90	90	90	90	90	90
D	测试集	210	210	210	210	210	210	210	210	210	210

样本数	模型	分类准确率/%
样本数	模型	无噪声	噪声γ = 0 dB	噪声γ = -4 dB	噪声γ = -8 dB
10	ResNet	90.79	71.78	58.79	39.87
	LeNet	89.43	79.11	62.89	23.22
	ProtoNet	97.67	74.40	65.42	38.95
	LeNet-ProtNet	93.36	48.26	41.80	22.20
	ResNet-ProtoNet	96.20	72.14	35.40	18.40
	Meta-DAE	99.66	88.67	74.25	75.65
20	ResNet	96.62	81.86	70.36	48.38
	LeNet	94.36	81.75	67.13	41.25
	ProtoNet	98.92	82.36	68.22	45.11
	LeNet-ProtNet	96.54	62.20	65.33	21.80
	ResNet-ProtoNet	98.89	84.42	63.00	20.60
	Meta-DAE	99.87	92.60	81.50	83.16
30	ResNet	98.56	85.21	74.39	58.42
	LeNet	97.58	84.79	72.78	46.33
	ProtoNet	99.42	85.82	71.34	56.78
	LeNet-ProtNet	98.72	71.45	69.26	26.13
	ResNet-ProtoNet	99.64	87.52	68.93	24.32
	Meta-DAE	99.98	93.25	82.45	85.67