Deep Domain Adaptation Approach Using an Improved Parallel Residual Network for Cross-Domain Bearing Fault Diagnosis
Recently, bearing fault diagnosis based on transfer learning (TL) has been a hot topic, which has attracted widespread interest due to its ability to adapt bearing fault datasets with different feature distributions. However, existing research suffer from low diagnosis efficiency and poor generaliza...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2024-01-01
|
| Series: | Shock and Vibration |
| Online Access: | http://dx.doi.org/10.1155/2024/7262611 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832546491571372032 |
|---|---|
| author | Jiezhou Huang |
| author_facet | Jiezhou Huang |
| author_sort | Jiezhou Huang |
| collection | DOAJ |
| description | Recently, bearing fault diagnosis based on transfer learning (TL) has been a hot topic, which has attracted widespread interest due to its ability to adapt bearing fault datasets with different feature distributions. However, existing research suffer from low diagnosis efficiency and poor generalization capabilities. Therefore, an improved parallel residual network-domain adaptation (IPRN-DA) method for bearing fault diagnosis is proposed in this paper, which is to address these challenges. Firstly, a parallel residual block (PRB) is designed to extract critical features that can fully characterize the original signals without significantly increasing the model parameters and more attention is paid to them. Secondly, a hybrid attention mechanism (HAM) is constructed to adaptively integrate channel and spatial features to enhance fault feature information. Finally, multikernel maximum mean discrepancy (MK-MMD) is employed to measure the distribution difference between the source and target domains in advanced feature extraction and predicted label spaces, implementing high-precision bearing transfer diagnosis. Rolling bearing datasets from Case Western Reserve University (CWRU) and Jiangnan University (JNU) are used to validate the effectiveness of the presented method. Experimental results illustrate that the algorithm can extract domain-invariant features for different cross-domain diagnosis tasks and thus improve fault diagnosis accuracy. |
| format | Article |
| id | doaj-art-cf47c2ee362c4fcb9b85f94e991c535c |
| institution | Kabale University |
| issn | 1875-9203 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | Wiley |
| record_format | Article |
| series | Shock and Vibration |
| spelling | doaj-art-cf47c2ee362c4fcb9b85f94e991c535c2025-02-03T06:51:38ZengWileyShock and Vibration1875-92032024-01-01202410.1155/2024/7262611Deep Domain Adaptation Approach Using an Improved Parallel Residual Network for Cross-Domain Bearing Fault DiagnosisJiezhou Huang0Guangzhou Sinorobot Intelligent Technology Co., Ltd.Recently, bearing fault diagnosis based on transfer learning (TL) has been a hot topic, which has attracted widespread interest due to its ability to adapt bearing fault datasets with different feature distributions. However, existing research suffer from low diagnosis efficiency and poor generalization capabilities. Therefore, an improved parallel residual network-domain adaptation (IPRN-DA) method for bearing fault diagnosis is proposed in this paper, which is to address these challenges. Firstly, a parallel residual block (PRB) is designed to extract critical features that can fully characterize the original signals without significantly increasing the model parameters and more attention is paid to them. Secondly, a hybrid attention mechanism (HAM) is constructed to adaptively integrate channel and spatial features to enhance fault feature information. Finally, multikernel maximum mean discrepancy (MK-MMD) is employed to measure the distribution difference between the source and target domains in advanced feature extraction and predicted label spaces, implementing high-precision bearing transfer diagnosis. Rolling bearing datasets from Case Western Reserve University (CWRU) and Jiangnan University (JNU) are used to validate the effectiveness of the presented method. Experimental results illustrate that the algorithm can extract domain-invariant features for different cross-domain diagnosis tasks and thus improve fault diagnosis accuracy.http://dx.doi.org/10.1155/2024/7262611 |
| spellingShingle | Jiezhou Huang Deep Domain Adaptation Approach Using an Improved Parallel Residual Network for Cross-Domain Bearing Fault Diagnosis Shock and Vibration |
| title | Deep Domain Adaptation Approach Using an Improved Parallel Residual Network for Cross-Domain Bearing Fault Diagnosis |
| title_full | Deep Domain Adaptation Approach Using an Improved Parallel Residual Network for Cross-Domain Bearing Fault Diagnosis |
| title_fullStr | Deep Domain Adaptation Approach Using an Improved Parallel Residual Network for Cross-Domain Bearing Fault Diagnosis |
| title_full_unstemmed | Deep Domain Adaptation Approach Using an Improved Parallel Residual Network for Cross-Domain Bearing Fault Diagnosis |
| title_short | Deep Domain Adaptation Approach Using an Improved Parallel Residual Network for Cross-Domain Bearing Fault Diagnosis |
| title_sort | deep domain adaptation approach using an improved parallel residual network for cross domain bearing fault diagnosis |
| url | http://dx.doi.org/10.1155/2024/7262611 |
| work_keys_str_mv | AT jiezhouhuang deepdomainadaptationapproachusinganimprovedparallelresidualnetworkforcrossdomainbearingfaultdiagnosis |