在极低标签率情况下，现有的图神经网络（Graph Neural Network, GNN）在图构造时存在节点间的关联信息挖掘不充分等问题。工业生产中，旋转机械常工作在变转速工况下，且标记故障样本代价高昂。针对上述两个问题，本文基于JS相对熵（Jenson's Shannon Divergence，JS相对熵）和动态图注意力网络（Dynamic Graph Attention Network，DGAT），提出了一种熵-图注意力网络,并将其应用于极低标签率下变转速工况的旋转机械半监督故障诊断中。首先，设计了基于JS相对熵的图构造方法，用于充分挖掘GNN中样本间的关联信息。其次，构建基于熵-动态图注意力网络的半监督学习模型，通过动态注意力机制进一步挖掘样本中故障敏感特征。最后，将所提方法在变转速工况下轴承和齿轮箱数据集上进行验证，结果表明所提方法能够在标签率不超过1%的极低情况下准确诊断出旋转机械的不同故障类型，且性能优于其它常用的图神经网络。
Abstract
In the case of very low labeling rates, existing Graph Neural Networks (GNN) have problems such as insufficient mining of correlation information between nodes when constructing graphs. In industrial production, rotating machinery often operates at variable speeds, and labelling fault samples is getting expensive. Aiming at the above two problems, a novel semi-supervised fault diagnosis method based on Jenson's Shannon relative entropy (JS) and dynamic graph attention network (DGAT) for rotating machinery under variable speed conditions is proposed in this paper. Firstly, a graph construction method based on JS is designed, which can fully mine the relevant information between samples. Secondly, a semi-supervised learning model based on entropy-dynamic graph attention networks (E-DGAT) is constructed to further mine the the sensitive features of faults in samples through dynamic attention mechanism. The proposed method is validated on a dataset of bearings and gearbox under variable speed conditions, and the results show that the proposed method is able to accurately diagnose different fault types of rotating machinery with a very low labelling rate of no more than 1%, and the proposed method is better than other commonly used GNN.
关键词
旋转机械 /
故障诊断 /
相对熵 /
图神经网络 /
变转速 /
低标签率
{{custom_keyword}} /
Key words
rotating machinery /
fault diagnosis /
graph neural networks /
relative entropy /
variable speed /
low labelling rate
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 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(09):84-90.
SHAO Haidong, ZHANG Xiaoyang, CHENG Junsheng, et al. Intelligent fault diagnosis of bearing using enhanced deep transfer anto-encoder[M].Journal of Mechanical Engineering,2020,56(9):84–90.
[2] 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(01):234-248
WEN Chenglin, LV Feiya. Review on deep learning based fault diagnosis[J]. Journal of Electronics and Information Technology, 2020, 42(1): 234-248.
[3] ZHAO Rui, YAN Ruqiang, CHEN Zhenghua, et al. Deep learning and its applications to machine health monitoring[J].Mechanical Systems and Signal Processing, 2019, 115: 213-237.
[4] Razavi-Far R, Hallaji E, Farajzadeh-Zanjani M, et al. A Semi-Supervised Diagnostic Framework Based on the Surface Estimation of Faulty Distributions[J]. IEEE Transactions on Industrial Informatics, 2019, 15(3): 1277-1286.
[5] 邢晓松,郭伟.基于改进半监督生成对抗网络的少量标签轴承智能诊断方法[J].振动与冲击,2022,41(22):184-192.
XING Xiaosong, GUO Wei. Intelligent diagnosis methed for bearings with few labelled samples based on an improved semi-supervised learning-based generative adversarial network[J]. Journal of Vibration and Shock.,2022,41(22):184-192.
[6] ZHANG Yongchao, YU Kun, LEI Zihao, et al. Integrated intelligent fault diagnosis approach of offshore wind turbine bearing based on information stream fusion and semi-supervised learning[J]. Expert Systems with Applications, 2023, 232: 1-15.
[7] LIAO Wenlong, Bak-Jensen B, Pilla J. R.i, et al. A Review of Graph Neural Networks and Their Applications in Power Systems. Journal of Modern Power Systems and Clean Energy,2022,10(2): 345-360.
[8] 王博. 基于半监督图卷积的行星齿轮箱故障诊断方法研究[D].哈尔滨：哈尔滨理工大学,2022.
WANG Bo. Research on Fault Diagnosis Method of Planetary Gearboxes based on Semi-GraghConvolution [D].Harbin, China:Harbin University of Science and Technology,2022.
[9] Kavianpour M, Ramezani A, Beheshti M T H. A class alignment method based on graph convolution neural network for bearing fault diagnosis in presence of missing data and changing working conditions[J]. Measurement, 2022, 199: 111536.
[10] XIE Zongliang, CHEN Jinglong, FENG Yong, et al. Semi-supervised multi-scale attention-aware graph convolution network for intelligent fault diagnosis of machine under extremely-limited labeled samples[J]. Journal of Manufacturing Systems, 2022, 64: 561-577.
[11] YAN Shen, SHAO Haidong, XIAO Yiming, et al. Semi-supervised fault diagnosis of machinery using LPS-DGAT under speed fluctuation and extremely low labeled rates[J]. Advanced Engineering Informatics, 2022, 53:1-12.
[12] LI Tianfu, ZHOU Zheng, LI Sinan, et al. The emerging graph neural networks for intelligent fault diagnostics and prognostics: A guideline and a benchmark study. Mechanical Systems and Signal Processing, 2022, 168: 1-37.
[13] YANG Lechang, BI Sifeng, Faes M G R, et al. Bayesian inversion for imprecise probabilistic models using a novel entropy-based uncertainty quantification metric[J]. Mechanical Systems and Signal Processing, 2022, 162: 1-21.
[14] HUANG Huan, Baddour N, LIANG Ming. Bearing vibration data collected under time-varying rotational speed conditions. Data in brief, 2018,21: 1745-1749.
[15] ZHANG Haiqi, LU Guangquan, ZHAN Mengmeng, et al. Semi-supervised classification of graph convolutional networks with Laplacian rank constraints. Neural Processing Letters, 2021,54(4):2645-2656.
[16] Hamilton W, Ying Z, Leskovec . Inductive representation learning on large graphs. Advances in neural information processing systems, 2017,30.
[17] Defferrard M, Bresson X, Vandergheynst P. Convolutional neural networks on graphs with fast localized spectral filtering[J]. Advances in neural information processing systems, 2016, 29
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
PDF(2087 KB)
