| Peer-Reviewed

Classification Fusion for Bearing Fault Diagnosis with Multi-source Domain Shift

Received: 10 May 2022     Published: 12 May 2022
Views:       Downloads:
Abstract

Bearings are one of the most widely used components of rotating machinery, whose failure can cause serious injuries and economic losses, therefore bearing fault diagnosis is an essential step in maintaining the safe and stable operation of industrial processes. Bearing fault diagnosis aims to detect the bearing fault condition and accurately classify it into a fault category based on sensing signals, such as vibration data. In practical applications, bearings always work in different types of equipment and under various working conditions, leading to performance degradation of diagnosis models due to the domain gap between the training data and the test data. Domain adaptation has been developed to address the domain shift problem in bearing fault diagnosis with demonstrated efficacy. Current domain adaptation models focus on the single-source scenario, by ignoring that sensing data may be collected from multiple sources in practical applications, and then be annotated for mode training. In this situation, it is non-trivial to use the single-source domain adaptation model to address the multi-source domain shift problem, because the domain gap exists among the source domains and the target domain. To solve this problem, we propose a novel bearing fault diagnosis model based on classification decision fusion to address the problem of multi-source domain conversion. Firstly, we train a source-aware fault diagnosis model in each source domain and then use it to predict the fault labels of the target samples. Second, a similarity score between each source domain and target domain is computed based on their feature distributions using local discriminant analysis and Maximum Mean Discrepancy. Finally, the similarity scores are used as domain weights in a proposed classification decision fusion strategy that uses a weighted linear combination process of predicted fault labels to provide the final predicted labels for the target samples. The benefits of the adaptive weighting fusion based on the classification result level, which makes full use of the available data from multiple source domains, measures the differences in distribution between the source and target domains and automatically adjust the weights to improve the diagnostic capability of the target domain. The effectiveness of the proposed method for diagnosing bearing faults under different operating and measurement conditions was verified using a bearing data set provided by Case Western Reserve University.

Published in Automation, Control and Intelligent Systems (Volume 10, Issue 2)
DOI 10.11648/j.acis.20221002.12
Page(s) 17-26
Creative Commons

This is an Open Access article, distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium or format, provided the original work is properly cited.

Copyright

Copyright © The Author(s), 2022. Published by Science Publishing Group

Keywords

Multi-source Domain Transfer, Bearing Fault Diagnosis, Classification Fusion, Knowledge Transfer, Cross-domain Fault Diagnosis

References
[1] ZHENG Jinde, WANG Xinglong, PAN Haiyang, TONG Jinyu, LIU Qingyun. Rolling Bearing Fault Diagnosis Method Based on Adaptive Autogram. China Mechanical Engineering, 2020, 39 (18): 118-124+131.
[2] ZHENG Jinde, PAN Haiyang, CHENG Junsheng. Rolling bearing fault detection and diagnosis based on composite multiscale fuzzy entropy and ensemble support vector machines. Mechanical Systems and Signal Processing. 2016.09.010:85.10.1016/j.ymssp.
[3] WANG Dong. K-nearest neighbors based methods for identification of different gear crack levels under different motor speeds and loads: Revisited. Mechanical Systems and Signal Processing. 2015.10.007:70.10.1016/j.ymssp.
[4] Zhou JM, Lai HW, Zhang L, Li P. Evaluation of bearing performance degradation based on EMD and logistic regression. Mechanical Design and Research, 2016, 32 (05): 72-75+79.
[5] Jinde, Zheng, Pan, Haiyang, Junsheng, Cheng. (2017). Rolling bearing fault detection and diagnosis based on composite multiscale fuzzy entropy and ensemble support vector machines. Mechanical Systems and Signal Processing. 85. 10.1016/j.ymssp.2016.09.010.
[6] W. Li, F. Li, Y. Luo, P. Wang, and J. Sun, "Deep Domain Adaptive Object Detection: a Survey," 2020 IEEE Symposium Series on Computational Intelligence (SSCI), 2020, pp. 1808-1813, DOI: 10.1109/SSCI47803.2020.9308604.
[7] Yao, Yi, Doretto, Gianfranco. (2010). Boosting for transfer learning with multiple sources. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition. 1855-1862. 10.1109/CVPR.2010.5539857.
[8] Sinno Jialin Pan, Ivor W TSANG, James T KWOK, Qiang YANG. Domain adaptation via transfer component analysis. IEEE Trans Neural Netw. 2011; 22 (2): 199-210.
[9] Kang, S., Hu, M., Wang, Y., Xie, J., Mikulovich, V. I. (2019). Fault Diagnosis Method of a Rolling Bearing Under Variable Working Conditions Based on Feature Transfer Learning. Zhongguo Dianji Gongcheng Xuebao/Proceedings of the Chinese Society of Electrical Engineering. 39. 764-772. 10.13334/j.0258-8013.pcsee.180130.
[10] Yunguang Gao, Shunbo Liu, Feng Li, Zhiguo Liu. Fault detection and diagnosis method for cooling dehumidifier based on LS-SVM NARX model. International Journal of Refrigeration, 2016, 61.
[11] ZHENG Huailiang, WANG Rixin, YANG Yuantao, LI Yuqing, XU Minqiang. Intelligent Fault Identification Based on Multisource Domain Generalization Towards Actual Diagnosis Scenario. IEEE Transactions on Industrial Electronics, 2020 (67): 1293-1304.
[12] A Kernel Two-Sample Test. Arthur Gretton, Karsten M. Borgwardt, Malte J. Rasch, Bernhard Schölkopf, Alexander Smola; 13 (25): 723−773, 2012.
[13] YAN Xunshi, SUN Zhe, ZHAO Jingjing, SHI Zhengang, ZHANG Chen-an. Fault diagnosis of rotating machinery equipped with multiple sensors using space-time fragments. Journal of Sound and Vibration, 2019, 456: 49-64.
[14] Shen Fei, CHEN Chao, Ruqiang YAN, Robert X. GAO. Bearing fault diagnosis based on SVD feature extraction and transfer learning classification. 2015 Prognostics and System Health Management Conference (PHM), 2015: 1-6.
[15] CHUANG Sun, MA Meng, ZHAO Zhibin, TIAN Shaohua, YAN Ruqiang, CHEN Xuefeng. Deep Transfer Learning Based on Sparse Autoencoder for Remaining Useful Life Prediction of Tool in Manufacturing. IEEE Transactions on Industrial Informatics, 2019 (15): 2416-2425.
[16] ZHANG Ran, TAO Hongyang, WU Lifeng, GUAN Yong R. Transfer Learning With Neural Networks for Bearing Fault Diagnosis in Changing Working Conditions. IEEE Access, vol. 2017: 14347-14357.
[17] Raghuraman GOPALAN; Ruonan LI; Rama CHELLAPPA. Domain adaptation for object recognition: An unsupervised approach. 2011 International Conference on Computer Vision, 2011: 999-1006.
[18] Boqing GONG, Yuan SHI, Fei SHA, Kristen GRAUMAN. Geodesic flow kernel for unsupervised domain adaptation. 2012 IEEE Conference on Computer Vision and Pattern Recognition, 2012: 2066-2073.
[19] YANG, Hong-bai, ZHANG, Jiang-an, CHEN, Lei-lei, ZHANG, Hong-li, LIU, Shu-lin. Fault Diagnosis of Reciprocating Compressor Based on Convolutional Neural Networks with Multisource Raw Vibration Signals. Mathematical Problems in Engineering, 2019.
[20] SMITH, Wade, ROBERT B. Randall. Rolling Element Bearing Diagnostics Using the Case Western Reserve University Data: A Benchmark Study. Mechanical Systems and Signal Processing. 2015: 64-65.
Cite This Article
  • APA Style

    Huang Mu-sheng, Wu Song-song, Zheng Shi-yuan, Yu Xi, Zhuang Jia-yang, et al. (2022). Classification Fusion for Bearing Fault Diagnosis with Multi-source Domain Shift. Automation, Control and Intelligent Systems, 10(2), 17-26. https://doi.org/10.11648/j.acis.20221002.12

    Copy | Download

    ACS Style

    Huang Mu-sheng; Wu Song-song; Zheng Shi-yuan; Yu Xi; Zhuang Jia-yang, et al. Classification Fusion for Bearing Fault Diagnosis with Multi-source Domain Shift. Autom. Control Intell. Syst. 2022, 10(2), 17-26. doi: 10.11648/j.acis.20221002.12

    Copy | Download

    AMA Style

    Huang Mu-sheng, Wu Song-song, Zheng Shi-yuan, Yu Xi, Zhuang Jia-yang, et al. Classification Fusion for Bearing Fault Diagnosis with Multi-source Domain Shift. Autom Control Intell Syst. 2022;10(2):17-26. doi: 10.11648/j.acis.20221002.12

    Copy | Download

  • @article{10.11648/j.acis.20221002.12,
      author = {Huang Mu-sheng and Wu Song-song and Zheng Shi-yuan and Yu Xi and Zhuang Jia-yang and Jing Xiao-yuan},
      title = {Classification Fusion for Bearing Fault Diagnosis with Multi-source Domain Shift},
      journal = {Automation, Control and Intelligent Systems},
      volume = {10},
      number = {2},
      pages = {17-26},
      doi = {10.11648/j.acis.20221002.12},
      url = {https://doi.org/10.11648/j.acis.20221002.12},
      eprint = {https://article.sciencepublishinggroup.com/pdf/10.11648.j.acis.20221002.12},
      abstract = {Bearings are one of the most widely used components of rotating machinery, whose failure can cause serious injuries and economic losses, therefore bearing fault diagnosis is an essential step in maintaining the safe and stable operation of industrial processes. Bearing fault diagnosis aims to detect the bearing fault condition and accurately classify it into a fault category based on sensing signals, such as vibration data. In practical applications, bearings always work in different types of equipment and under various working conditions, leading to performance degradation of diagnosis models due to the domain gap between the training data and the test data. Domain adaptation has been developed to address the domain shift problem in bearing fault diagnosis with demonstrated efficacy. Current domain adaptation models focus on the single-source scenario, by ignoring that sensing data may be collected from multiple sources in practical applications, and then be annotated for mode training. In this situation, it is non-trivial to use the single-source domain adaptation model to address the multi-source domain shift problem, because the domain gap exists among the source domains and the target domain. To solve this problem, we propose a novel bearing fault diagnosis model based on classification decision fusion to address the problem of multi-source domain conversion. Firstly, we train a source-aware fault diagnosis model in each source domain and then use it to predict the fault labels of the target samples. Second, a similarity score between each source domain and target domain is computed based on their feature distributions using local discriminant analysis and Maximum Mean Discrepancy. Finally, the similarity scores are used as domain weights in a proposed classification decision fusion strategy that uses a weighted linear combination process of predicted fault labels to provide the final predicted labels for the target samples. The benefits of the adaptive weighting fusion based on the classification result level, which makes full use of the available data from multiple source domains, measures the differences in distribution between the source and target domains and automatically adjust the weights to improve the diagnostic capability of the target domain. The effectiveness of the proposed method for diagnosing bearing faults under different operating and measurement conditions was verified using a bearing data set provided by Case Western Reserve University.},
     year = {2022}
    }
    

    Copy | Download

  • TY  - JOUR
    T1  - Classification Fusion for Bearing Fault Diagnosis with Multi-source Domain Shift
    AU  - Huang Mu-sheng
    AU  - Wu Song-song
    AU  - Zheng Shi-yuan
    AU  - Yu Xi
    AU  - Zhuang Jia-yang
    AU  - Jing Xiao-yuan
    Y1  - 2022/05/12
    PY  - 2022
    N1  - https://doi.org/10.11648/j.acis.20221002.12
    DO  - 10.11648/j.acis.20221002.12
    T2  - Automation, Control and Intelligent Systems
    JF  - Automation, Control and Intelligent Systems
    JO  - Automation, Control and Intelligent Systems
    SP  - 17
    EP  - 26
    PB  - Science Publishing Group
    SN  - 2328-5591
    UR  - https://doi.org/10.11648/j.acis.20221002.12
    AB  - Bearings are one of the most widely used components of rotating machinery, whose failure can cause serious injuries and economic losses, therefore bearing fault diagnosis is an essential step in maintaining the safe and stable operation of industrial processes. Bearing fault diagnosis aims to detect the bearing fault condition and accurately classify it into a fault category based on sensing signals, such as vibration data. In practical applications, bearings always work in different types of equipment and under various working conditions, leading to performance degradation of diagnosis models due to the domain gap between the training data and the test data. Domain adaptation has been developed to address the domain shift problem in bearing fault diagnosis with demonstrated efficacy. Current domain adaptation models focus on the single-source scenario, by ignoring that sensing data may be collected from multiple sources in practical applications, and then be annotated for mode training. In this situation, it is non-trivial to use the single-source domain adaptation model to address the multi-source domain shift problem, because the domain gap exists among the source domains and the target domain. To solve this problem, we propose a novel bearing fault diagnosis model based on classification decision fusion to address the problem of multi-source domain conversion. Firstly, we train a source-aware fault diagnosis model in each source domain and then use it to predict the fault labels of the target samples. Second, a similarity score between each source domain and target domain is computed based on their feature distributions using local discriminant analysis and Maximum Mean Discrepancy. Finally, the similarity scores are used as domain weights in a proposed classification decision fusion strategy that uses a weighted linear combination process of predicted fault labels to provide the final predicted labels for the target samples. The benefits of the adaptive weighting fusion based on the classification result level, which makes full use of the available data from multiple source domains, measures the differences in distribution between the source and target domains and automatically adjust the weights to improve the diagnostic capability of the target domain. The effectiveness of the proposed method for diagnosing bearing faults under different operating and measurement conditions was verified using a bearing data set provided by Case Western Reserve University.
    VL  - 10
    IS  - 2
    ER  - 

    Copy | Download

Author Information
  • School of Computer Science, Guangdong University of Petrochemical Technology, Maoming, China

  • School of Computer Science, Guangdong University of Petrochemical Technology, Maoming, China

  • School of Computer Science, Guangdong University of Petrochemical Technology, Maoming, China

  • School of Computer Science, Guangdong University of Petrochemical Technology, Maoming, China

  • School of Computer Science, Guangdong University of Petrochemical Technology, Maoming, China

  • School of Computer Science, Guangdong University of Petrochemical Technology, Maoming, China

  • Sections