Transfer learning with inception ResNet-based model for rolling bearing fault diagnosis
With the development of information technology and sensor technology, people have paid more attention to data-driven fault diagnosis. As one of the commonly used methods in fault diagnosis, deep learning has achieved significant results. However, in engineering practice, the insufficient number of l...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
The Japan Society of Mechanical Engineers
2022-06-01
|
| Series: | Journal of Advanced Mechanical Design, Systems, and Manufacturing |
| Subjects: | |
| Online Access: | https://www.jstage.jst.go.jp/article/jamdsm/16/2/16_2022jamdsm0023/_pdf/-char/en |
| _version_ | 1811225550886273024 |
|---|---|
| author | Yongbao LIU Jun LI Qijie LI Qiang WANG |
| author_facet | Yongbao LIU Jun LI Qijie LI Qiang WANG |
| author_sort | Yongbao LIU |
| collection | DOAJ |
| description | With the development of information technology and sensor technology, people have paid more attention to data-driven fault diagnosis. As one of the commonly used methods in fault diagnosis, deep learning has achieved significant results. However, in engineering practice, the insufficient number of labeled samples for fault diagnosis and the poor targeting of extracted features lead to a limited structural depth of deep learning models and inadequate model training, limiting the diagnostic accuracy of fault diagnosis. A novel fault diagnosis method is proposed in this paper by implementing model-based transfer learning in the Inception-ResNet-v2 model. Firstly, the process applies a signal-to-image transformation method in the feature extraction stage to merge the frequency weighted energy operator (FWEO), kurtosis, and raw vibration signals into RGB images as the input dataset for diagnosing the type of rolling bearing faults. Secondly, a new combined transfer learning and Inception-ResNet-v2 CNN model (TL-IRCNN) is proposed for rolling bearing fault diagnosis under minor sample conditions. Finally, The performance of the proposed method was validated using the motor bearing dataset from Case Western Reserve University (CWRU) and the rolling bearing dataset from a local laboratory. The results show that the proposed TL-IRCNN method achieves high fault classification accuracy under minor sample conditions in bearing diagnosis. |
| first_indexed | 2024-04-12T09:09:12Z |
| format | Article |
| id | doaj.art-6daa0f09e32841de9c3d4c734c3e307d |
| institution | Directory Open Access Journal |
| issn | 1881-3054 |
| language | English |
| last_indexed | 2024-04-12T09:09:12Z |
| publishDate | 2022-06-01 |
| publisher | The Japan Society of Mechanical Engineers |
| record_format | Article |
| series | Journal of Advanced Mechanical Design, Systems, and Manufacturing |
| spelling | doaj.art-6daa0f09e32841de9c3d4c734c3e307d2022-12-22T03:39:00ZengThe Japan Society of Mechanical EngineersJournal of Advanced Mechanical Design, Systems, and Manufacturing1881-30542022-06-01162JAMDSM0023JAMDSM002310.1299/jamdsm.2022jamdsm0023jamdsmTransfer learning with inception ResNet-based model for rolling bearing fault diagnosisYongbao LIU0Jun LI1Qijie LI2Qiang WANG3Department of Power Engineering, Naval University of EngineeringDepartment of Power Engineering, Naval University of EngineeringDepartment of Power Engineering, Naval University of EngineeringDepartment of Power Engineering, Naval University of EngineeringWith the development of information technology and sensor technology, people have paid more attention to data-driven fault diagnosis. As one of the commonly used methods in fault diagnosis, deep learning has achieved significant results. However, in engineering practice, the insufficient number of labeled samples for fault diagnosis and the poor targeting of extracted features lead to a limited structural depth of deep learning models and inadequate model training, limiting the diagnostic accuracy of fault diagnosis. A novel fault diagnosis method is proposed in this paper by implementing model-based transfer learning in the Inception-ResNet-v2 model. Firstly, the process applies a signal-to-image transformation method in the feature extraction stage to merge the frequency weighted energy operator (FWEO), kurtosis, and raw vibration signals into RGB images as the input dataset for diagnosing the type of rolling bearing faults. Secondly, a new combined transfer learning and Inception-ResNet-v2 CNN model (TL-IRCNN) is proposed for rolling bearing fault diagnosis under minor sample conditions. Finally, The performance of the proposed method was validated using the motor bearing dataset from Case Western Reserve University (CWRU) and the rolling bearing dataset from a local laboratory. The results show that the proposed TL-IRCNN method achieves high fault classification accuracy under minor sample conditions in bearing diagnosis.https://www.jstage.jst.go.jp/article/jamdsm/16/2/16_2022jamdsm0023/_pdf/-char/enfault diagnosisdeep learningconvolution neural networktransfer learningenergy operator |
| spellingShingle | Yongbao LIU Jun LI Qijie LI Qiang WANG Transfer learning with inception ResNet-based model for rolling bearing fault diagnosis Journal of Advanced Mechanical Design, Systems, and Manufacturing fault diagnosis deep learning convolution neural network transfer learning energy operator |
| title | Transfer learning with inception ResNet-based model for rolling bearing fault diagnosis |
| title_full | Transfer learning with inception ResNet-based model for rolling bearing fault diagnosis |
| title_fullStr | Transfer learning with inception ResNet-based model for rolling bearing fault diagnosis |
| title_full_unstemmed | Transfer learning with inception ResNet-based model for rolling bearing fault diagnosis |
| title_short | Transfer learning with inception ResNet-based model for rolling bearing fault diagnosis |
| title_sort | transfer learning with inception resnet based model for rolling bearing fault diagnosis |
| topic | fault diagnosis deep learning convolution neural network transfer learning energy operator |
| url | https://www.jstage.jst.go.jp/article/jamdsm/16/2/16_2022jamdsm0023/_pdf/-char/en |
| work_keys_str_mv | AT yongbaoliu transferlearningwithinceptionresnetbasedmodelforrollingbearingfaultdiagnosis AT junli transferlearningwithinceptionresnetbasedmodelforrollingbearingfaultdiagnosis AT qijieli transferlearningwithinceptionresnetbasedmodelforrollingbearingfaultdiagnosis AT qiangwang transferlearningwithinceptionresnetbasedmodelforrollingbearingfaultdiagnosis |