本文针对神经网络模型在有标签样本数量较少的情况下，容易产生网络过拟合、故障诊断精度低、不能充分利用大量无标签样本数据等问题，提出一种基于连续小波变换和教师-学生网络的半监督学习方法用于旋转机械的故障诊断。该方法以改进的LeNet5卷积神经网络模型为基础，建立具有相同结构和初始化参数的学生网络模型和教师网络模型。首先，将旋转机械振动信号进行连续小波变换，将其转换为三维时频图像。接着，利用教师模型的预测结果生成伪标签，将这些伪标签和真实标签结合起来，训练学生网络。同时，通过指数加权移动平均算法更新教师网络模型参数。实验结果表明，相对于纯监督学习模型，所提出的算法能够在有标签样本数量较少的情况下显著提高模型训练过程的稳定性和故障诊断的精度。
Abstract
This paper proposes a semi-supervised learning method based on continuous wavelet transform and teacher-student network for fault diagnosis of rotating machinery, aiming to address the problems of network overfitting, low fault diagnosis accuracy, and underutilization of large amounts of unlabeled data when neural network models are used with limited labeled samples. The method is based on an improved Lenet-5 convolutional neural network model, which establishes a student network model and a teacher network model with the same structure and initialization parameters. First, the vibration signal of rotating machinery is transformed by continuous wavelet transform into a three-dimensional time-frequency image. Then, pseudo-labels are generated using the prediction results of the teacher model, and these pseudo-labels are combined with the real labels to train the student network. At the same time, the teacher network model parameters are updated using exponential weighted moving average algorithm. The experimental results show that compared with the pure supervised learning model, the proposed algorithm can significantly improve the stability of the model training process and the accuracy of fault diagnosis with limited labeled samples.
关键词
故障诊断 /
旋转机械 /
连续小波变换 /
半监督学习
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] Lei J, Liu C, Jiang D. Fault diagnosis of wind turbine based on Long Short-term memory networks[J]. Renewable energy, 2019, 133: 422-432.
[2] Zhao R, Wang D, Yan R, et al. Machine health monitoring using local feature-based gated recurrent unit networks[J]. IEEE Transactions on Industrial Electronics, 2017, 65(2): 1539-1548.
[3] 史红梅,郑畅畅,司瑾，等.基于动态加权的多尺度残差网络旋转机械故障诊断算法[J].振动与冲击,2022,41(23):67-74+93.
SHI Hong-mei, ZHENG Chang-chang, SI Jin, et al. Fault diagnosis algorithm of rotating machinery based on dynamic weighted multiscale residual network[J]. Journal of vibration and shock, 2022,41(23):67-74+93.
[4] Mo Z, Zhang Z, Tsui K L. The variational kernel-based 1-D convolutional neural network for machinery fault diagnosis[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 1-10.
[5] Yoo Y, Baek J G. A novel image feature for the remaining useful lifetime prediction of bearings based on continuous wavelet transform and convolutional neural network[J]. Applied Sciences, 2018, 8(7): 1102.
[6] Wu J, Zhao Z, Sun C, et al. Few-shot transfer learning for intelligent fault diagnosis of machine[J]. Measurement, 2020, 166: 108202.
[7] 苏元浩,孟良,许同乐，等.不平衡数据集下优化WGAN的风电机组齿轮箱故障诊断方法[J].太阳能学报,2022,43(11):148-155.
SU Yuan-hao, MENG Liang, XU Tong-le, et al. Wind turbine gearbox fault diagnosis method for optimized WGAN with unbalanced data sets[J]. Acta energiae solaris sinica,2022,43(11):148-155.
[8] Wen L, Gao L, Li X. A new deep transfer learning based on sparse auto-encoder for fault diagnosis[J]. IEEE Transactions on systems, man, and cybernetics: systems, 2017, 49(1): 136-144.
[9] Shao H, Xia M, Han G, et al. Intelligent fault diagnosis of rotor-bearing system under varying working conditions with modified transfer convolutional neural network and thermal images[J]. IEEE Transactions on Industrial Informatics, 2020, 17(5): 3488-3496.
[10] 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 Xiao-song, GUO Wei. Intelligent diagnosis method 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.
[11] 李亦佳,王静,王正方，等.基于多重同步压缩变换的微震信号去噪方法研究[J].应用基础与工程科学学报,2022,30(02):486-500.
LI Yi-jia, WANG Jing, WANG Zheng-fang, et al. Research on de-noising method of microseismic signal based on multi-synchronous compression transform[J]. Journal of Basic Science and Engineering, 2022,30(02):486-500.
[12] 李盛前,张小帆.基于小波变换的双阈值水下焊缝图像边缘检测研究[J].机床与液压,2022,50(23):173-178.
LI Sheng-qian, ZHANG Xiao-fan. Research on double-threshold edge detection for under-water welding seam image based on wavelet transform[J]. Machine tool & hydraulics, ,2022,50(23):173-178.
[13] 邓飞跃,唐贵基.基于时间-小波能量谱样本熵的滚动轴承智能诊断方法[J].振动与冲击,2017,36(09):28-34.
DENG Fei-yue, TANG Gui-ji. An intelligent method for rolling element bearing fault diagnosis based on time-wavelet energy spectrum sample entropy[J]. Journal of vibration and shock, 2017,36(09):28-34.
[14] LeCun Y, Bottou L, Bengio Y & Haffner P. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE. 86(11): 2278 - 2324.
[15] Ioffe S, Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift[C]. International conference on machine learning. pmlr, 2015: 448-456.
[16] KINGMA D P, LEI BA J. Adam: a method for stochastic optimization[J]. Computer Science, 2014:1-15.
[17] Laine S, Aila T. Temporal ensembling for semi-supervised learning[J]. arXiv:1610.02242, 2016.
[18] Van der Maaten L, Hinton G. Visualizing data using t-SNE[J]. Journal of machine learning research, 2008, 9(11)
{{custom_fnGroup.title_cn}}
脚注
{{custom_fn.content}}
PDF(1583 KB)
