DCGAN Bearing Fault Diagnosis Method under Unbalanced Samples
-
摘要: 在实际工况下, 轴承可采集到的故障样本分布往往呈现极强的不均衡特性, 该特性对故障诊断精度具有不可忽略的影响。为提高样本不均衡情况下的轴承故障诊断精度, 采用样本生成扩充的思路, 提出一种基于深度卷积生成对抗网络的故障诊断方法。首先针对轴承振动数据信号的特性, 采用快速傅里叶变换使其转化为频域, 并通过归一化进行预处理; 其次利用深度卷积生成对抗网络进行对抗训练, 生成具有真实样本特征的虚拟样本。模型采用衰减学习率并增设Dropout层, 提高了模型生成的效率及真实性。最后, 构建一维卷积神经网络模型完成故障诊断。实验验证结果表明, 提出的方法能有效提高样本不均衡情况下的诊断精度以及诊断稳定性。Abstract: In actual operating conditions, the distribution of fault samples that can be collected by bearings often exhibits extremely strong imbalance characteristics, which has a non-negligible effect on the accuracy of fault diagnosis. In order to improve the accuracy of bearing fault diagnosis in the case of unbalanced samples, a new fault diagnosis method based on deep convolution generation adversarial network is proposed using the idea of sample generation and expansion. First of all, according to the characteristics of the bearing vibration data signal, the fast Fourier transform is used to convert the vibration signal into the frequency domain, and the preprocessing is performed through normalization. Secondly, the deep convolution generation confrontation network is used for confrontation training to generate virtual samples with real sample characteristics. The model uses an attenuated learning rate and adds a Dropout layer, which can improve the efficiency and authenticity of model generation. Finally, a one-dimensional convolution neural network model is constructed to complete the fault diagnosis. Experimental verification results show that the method proposed can effectively improve the diagnostic accuracy and diagnostic stability in the case of sample imbalance.
-
Key words:
- bearing /
- fault diagnosis /
- sample imbalance /
- sample generation
-
表 1 故障类型和总量
类型 故障位置 损伤等级 样本数量 单样本数 FFT处理后 1 无 无 20 2 048 1 024 2 外圈 2 20 2 048 1 024 3 外圈 1 20 2 048 1 024 4 内圈 1 20 2 048 1 024 5 内圈 2 20 2 048 1 024 -
[1] CERRADA M, SÁNCHEZ R V, LI C, et al. A review on data-driven fault severity assessment in rolling bearings[J]. Mechanical Systems and Signal Processing, 2018, 99: 169-196 doi: 10.1016/j.ymssp.2017.06.012 [2] LEI Y G, LIN J, HE Z J, et al. A review on empirical mode decomposition in fault diagnosis of rotating machinery[J]. Mechanical Systems and Signal Processing, 2013, 35(1-2): 108-126 doi: 10.1016/j.ymssp.2012.09.015 [3] YANG Y, YU D J, CHENG J S. A fault diagnosis approach for roller bearing based on IMF envelope spectrum and SVM[J]. Measurement, 2007, 40(9-10): 943-950 doi: 10.1016/j.measurement.2006.10.010 [4] HOU Z R. Rolling bearing fault diagnosis based on wavelet packet and improved BP neural network for wind turbines[J]. Applied Mechanics and Materials, 2013, 347-350: 117-120 doi: 10.4028/www.scientific.net/AMM.347-350.117 [5] MURALIDHARAN V, SUGUMARAN V. A comparative study of Naíve Bayes classifier and Bayes net classifier for fault diagnosis of monoblock centrifugal pump using wavelet analysis[J]. Applied Soft Computing, 2012, 12(8): 2023-2029 doi: 10.1016/j.asoc.2012.03.021 [6] QIN X W, LI Q L, DONG X G, et al. The fault diagnosis of rolling bearing based on ensemble empirical mode decomposition and random forest[J]. Shock and Vibration, 2017, 2017: 2623081 https://www.hindawi.com/journals/sv/2017/2623081/ [7] ZHAO S F, LIANG L, XU G H, et al. Quantitative diagnosis of a spall-like fault of a rolling element bearing by empirical mode decomposition and the approximate entropy method[J]. Mechanical Systems and Signal Processing, 2013, 40(1): 154-177 doi: 10.1016/j.ymssp.2013.04.006 [8] 张扬帆, 张海鹏, 孙俊. 基于Lévy分布的不平衡数据过采样方法[J]. 计算机工程与应用, 2019, 55(16): 150-156 doi: 10.3778/j.issn.1002-8331.1804-0218ZHANG Y F, ZHANG H P, SUN J. Lévy-based oversampling technique for imbalanced datasets[J]. Computer Engineering and Applications, 2019, 55(16): 150-156 (in Chinese) doi: 10.3778/j.issn.1002-8331.1804-0218 [9] 周建伟. 不平衡数据的下采样方法研究[J]. 计算机与数字工程, 2019, 47(9): 2155-2160 doi: 10.3969/j.issn.1672-9722.2019.09.010ZHOU J W. Under-sampling method research in class-imbalanced data[J]. Computer and Digital Engineering, 2019, 47(9): 2155-2160 (in Chinese) doi: 10.3969/j.issn.1672-9722.2019.09.010 [10] 王圆方. 基于层次聚类改进SMOTE的过采样方法[J]. 软件, 2020, 41(2): 201-204 doi: 10.3969/j.issn.1003-6970.2020.02.044WANG Y F. Hybrid algorithm of aggregation hierarchy clustering and SMOTE for oversampling[J]. Computer Engineering & Software, 2020, 41(2): 201-204 (in Chinese) doi: 10.3969/j.issn.1003-6970.2020.02.044 [11] GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial networks[J]. Communications of the ACM, 2020, 63(11): 139-144 doi: 10.1145/3422622 [12] CRESWELL A, WHITE T, DUMOULIN V, et al. Generative adversarial networks: an overview[J]. IEEE Signal Processing Magazine, 2018, 35(1): 53-65 doi: 10.1109/MSP.2017.2765202 [13] JANSSENS O, SLAVKOVIKJ V, VERVISCH B, et al. Convolutional neural network based fault detection for rotating machinery[J]. Journal of Sound and Vibration, 2016, 377: 331-345 doi: 10.1016/j.jsv.2016.05.027 [14] 李阳, 张亚非, 苗壮, 等. 基于t-SNE卷积编码的图像检索方法[J]. 计算机应用研究, 2017, 34(4): 1244-1248, 1264 doi: 10.3969/j.issn.1001-3695.2017.04.065LI Y, ZHANG Y F, MIAO Z, et al. Convolutional encoding image retrieval method based on t-SNE[J]. Application Research of Computers, 2017, 34(4): 1244-1248, 1264 (in Chinese) doi: 10.3969/j.issn.1001-3695.2017.04.065 [15] XU Q T, HUANG G, YUAN Y, et al. An empirical study on evaluation metrics of generative adversarial networks[J]. arXiv: 1806.07755, 2018