Rolling bearing fault diagnosis based on correlation frequency loss Wasserstein generative adversarial network
-
摘要: 针对滚动轴承故障样本不平衡导致故障诊断性能下降的问题，提出一种基于相关性频率损失生成对抗网络(generative adversarial network, GAN)的滚动轴承故障诊断方法。通过计算真实数据与生成数据的互相关性，定量分析二者在频域上的差异，并指导模型训练，提高生成数据与真实数据的相似性，突破数据中难以合成的故障频率，最后采用卷积神经网络进行故障诊断。利用凯斯西储大学轴承数据集对所提方法进行实验验证，在不平衡比例为10∶1、20∶1和50∶1情况下，相关性频率损失生成对抗网络的样本生成能力明显高于其他方法，显著提高了故障识别率，验证了该方法的有效性和优越性。Abstract: To address the decline in rolling bearing fault diagnosis performance caused by imbalanced fault samples, a fault diagnosis method based on correlation frequency loss generative adversarial network (GAN) was proposed. The mutual correlation between real and generated data was calculated to quantitatively analyze their differences in the frequency domain. which was used to guide model training, improve the similarity between the generated and real data, and overcome the difficulty of synthesizing fault frequencies in the data. Finally, a convolutional neural network (CNN) was employed for fault diagnosis. The proposed method is experimentally validated using the Case Western Reserve University bearing dataset. Under imbalance ratios of 10:1, 20:1 and 50:1, the sample generation capability of the correlation frequency loss GAN was significantly higher than that of other methods, leading to a marked improvement in the fault recognition rate. This results verifies the effectiveness and superiority of the proposed method.
-
图 2 基于CFL-WGAN网络的轴承故障诊断流程图
Figure 2. Bearing fault diagnosis flow chart based on CFL-WGAN network
图 4 GAN、WGAN和CFL-WGAN生成数据与真实数据频谱对比图
Figure 4. Spectrum comparison of data generated by GAN、WGAN、CFL-WGAN and real data
图 5 不平衡比例50∶1条件下生成模型进行故障诊断后生成数据质量折线图
Figure 5. Line chart of generated data quality for fault diagnosis under condition of 50∶1 imbalance ratio
表 1 CWRU数据集实验任务设计表
Table 1. CWRU dataset experimental task design table
故障类型 故障直径
（英寸in）标签 训练集样本数 测试集
样本数10 ∶1 20 ∶1 50 ∶1 正常 — 0 100 100 100 100 内圈故障1 0.007 1 10 5 2 100 内圈故障2 0.014 2 10 5 2 100 内圈故障3 0.021 3 10 5 2 100 外圈故障1 0.007 4 10 5 2 100 外圈故障2 0.014 5 10 5 2 100 外圈故障3 0.021 6 10 5 2 100 滚动体故障1 0.007 7 10 5 2 100 滚动体故障2 0.014 8 10 5 2 100 滚动体故障3 0.021 9 10 5 2 100 
下载: 导出CSV
表 2 不平衡比例20∶1时生成数据与真实数据的互相关性
Table 2. Correlation between generated and real data under 20∶1 imbalance ratio
数据来源 真实样本 GAN VAE-GAN WGAN CFL-WGAN 互相关性 0.8079 0.7860 0.7705 0.7893 0.7934
下载: 导出CSV
表 3 不同不平衡比例时生成模型生成数据的诊断精度
Table 3. Diagnostic accuracy of data generated by generative models under different imbalance ratios
单位：% 不平衡比例 GAN VAE-GAN WGAN CFL-WGAN 10∶1 99.54 99.43 99.64 99.73 20∶1 98.47 97.61 98.84 99.30 50∶1 94.60 95.75 96.92 97.72
下载: 导出CSV
-
[1] MARTIN-DIAZ I, MORINIGO-SOTELO D, DUQUE-PEREZ O, et al. Early fault detection in induction motors using adaboost with im- balanced small data and optimized sampling[J] . IEEE Transactions on Industry Applications, 2020, 20(23): 14337 − 14346. doi: 10.1109/JSEN.2020.3008177 [2] 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. doi: 10.1109/TII.2020.2967822 [3] 郭玉荣, 茅健, 赵嫚. 基于CNN与注意力BiLSTM的轴承剩余使用寿命预测方法[J] . 上海工程技术大学学报, 2023, 37(1): 96 − 104. doi: 10.3969/j.issn.1009-444X.2023.01.015 [4] JIA F, LEI Y G, LU N, et al. Deep normalized convolutional neural network for imbalanced fault classification of machinery and its understanding via visualization[J] . Mechanical Systems and Signal Processing, 2018, 110: 349 − 367. doi: 10.1016/j.ymssp.2018.03.025 [5] 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. doi: 10.1016/j.ymssp.2019.106587 [6] ZHANG R F, LIU Y X. Research on development and application of support vector machine - Transformer fault diagnosis[C] //Proceedings of the International Symposium on Big Data and Artificial Intelligence. Hong Kong, China: Association for Computing Machinery, 2018: 262 − 268. [7] GOODFELLOW I J, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial nets[C] //Proceedings of the 28th International Conference on Neural Information Processing Systems. Montreal: MIT Press, 2014: 2672 − 2680. [8] RADFORD A, METZ L, CHINTALA S. Unsupervised representation learning with deep convolutional generative adversarial networks[C] //Proceedings of the 4th International Conference on Learning Representations. San Juan: ICLR, 2016: 1 − 16. [9] ARJOVSKY M, CHINTALA S, BOTTOU L. Wasserstein GAN[J] . arXiv:1701.07875, 2017. [10] QIN Z H, HUANG F G, PAN J F, et al. Improved generative adversarial network for bearing fault diagnosis with a small number of data and unbalanced data[J] . Symmetry, 2024, 16(3): 358. doi: 10.3390/sym16030358 [11] BAI G L, SUN W, CAO C, et al. GAN-based bearing fault diagnosis method for short and imbalanced vibration signal[J] . IEEE Sensors Journal, 2024, 24(2): 1894 − 1904. doi: 10.1109/JSEN.2023.3337278 -
下载:
点击查看大图
计量
- 文章访问数: 13
- HTML全文浏览量: 7
- PDF下载量: 1
- 被引次数: 0
