Vibration signals are important indicators for evaluating the current motor state and are often used to analyze and judge the potential faults of motors. However, due to the sparse distribution of fault signals, the current deep learning-based methods are more likely to predict stable states accurately rather than fault states. To address the above problem, a novel fault diagnosis model for motors based on regularized enhanced graph convolutional neural networks was proposed by researching the motor fault diagnosis task. With the assistance of the encoded motor state correlations and the introduced regularized enhanced structure, the model can improve its prediction ability for various fault types. The experimental results show that compared with traditional deep learning methods, the proposed method has good accuracy improvement on the motor fault diagnosis dataset, verifying the effectiveness of the proposed model. Graph convolutional neural networks can mine the intrinsic correlations between long-range signals in signal sequences. By combining with the regularized structure, they have broad application prospects in various industrial time-series tasks.
WANGYifan, KexuanA, LIUKaiwei, et al. Research on fault diagnosis method of switch machine based on deep learning[J]. Journal of Test and Measurement Technology, 2023, 37(2): 106-111. (in Chinese)
CHENJian, XUTingliang, HUANGZhi, et al. Fault diagnosis of rolling bearings based on acoustic signals[J]. Journal of Vibration and Shock, 2023, 42(21): 237-244. (in Chinese)
YANGHeng, SHENYanxia, TANYongqiang, et al. Fault diagnosis and fault-tolerant control of current sensors in PMSM based on MPCC[J]. Small & Special Electrical Machines, 2024, 52(4): 60-64. (in Chinese)
[7]
THANAPALASINGAMT, BERKELL V, BLOEMP, et al. Relational graph convolutional networks: a closer look[J/OL]. Computer Science: Machine Learning, 2021. arXiv: 2107.10015.
SUNHao, KONGLingsong. Review of fault diagnosis methods for cage motor rotor broken bars based on fault characteristic analysis[J]. Small & Special Electrical Machines, 2023, 51(7): 57-66. (in Chinese)
MAOJingkun, QINYing, LIDeheng. Asynchronous alternating current motor fault diagnosis method based on 2DCNN-VAE-Transformer[J/OL]. Journal of Tianjin University of Technology, 2024: 1-9. in Chinese)
[12]
HUSARIF, SESHADRINATHJ. Sensitive inter-tum fault identifcation in induction motors using deep learning based methods[C]//2020 IEEE International Conference on Power Electronics, Smart Grid and Renewable Energy (PESGRE2020), 2020: 1-6.
[13]
ZHANGL, DENGJ. A relational graph convolution network-based smart risk recognition model for financial transactions[J]. Journal of Circuits, Systems and Computers, 2024, 33(16): 2450293.
[14]
LIUC, SILAMUW, LIY. A Chinese–Kazakh translation method that combines data augmentation and R-drop regularization[J]. Applied Sciences, 2023, 13(19): 10589.
ZHANGHaochen, QUHongjun, NIUXueying, et al. Research on entity identification model of electric power centralized control security hidden danger by integrating attention mechanism[J]. Process Automation Instrumentation, 2023, 44(10): 55-59. (in Chinese)
[17]
王跃龙. 笼型异步电动机多故障智能诊断及分离方法的研究[D]. 太原: 太原理工大学, 2017.
[18]
WANGM, YUJ, LENGH, et al. Bearing fault detection by using graph autoencoder and ensemble learning[J]. Scientific Reports, 2024, 14(1): 5206.