An Auxiliary Classifier Generative Adversarial Network Based Fault Diagnosis for Analog Circuit
To solve the analog fault diagnosis problem with fewer samples, a transformer based auxiliary classifier generative adversarial network (ACGAN) is investigated for circuit fault diagnosis by constructing both generator and discriminator in ACGAN with pure transformer components. The transformer has...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2023-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10216968/ |
| _version_ | 1797739776818806784 |
|---|---|
| author | Yongqiang Zheng Dongqing Wang |
| author_facet | Yongqiang Zheng Dongqing Wang |
| author_sort | Yongqiang Zheng |
| collection | DOAJ |
| description | To solve the analog fault diagnosis problem with fewer samples, a transformer based auxiliary classifier generative adversarial network (ACGAN) is investigated for circuit fault diagnosis by constructing both generator and discriminator in ACGAN with pure transformer components. The transformer has high model generality due to its weak inductive bias, but it also increases the risk of overfitting on small datasets. Therefore, we use ACGAN to generate sample data to enrich the dataset and mitigate the overfitting. However, ACGAN is severely unstable during the training period, for this reason, a confidence mechanism for the discriminator is added to improve the classification accuracy and a new layer normalization method for the generator is studied to avoid the loss of conditional information. Take the sallen-key filter circuit and the biquad high-pass filter circuit as the experiment objects. The experiment results indicate that the transformer based ACGAN diagnosis method can effectively improve diagnosis accuracy, reach 96.22% and 95.35%, respectively. |
| first_indexed | 2024-03-12T14:03:02Z |
| format | Article |
| id | doaj.art-b5c4262e55ea42d5a0d23827890301dc |
| institution | Directory Open Access Journal |
| issn | 2169-3536 |
| language | English |
| last_indexed | 2024-03-12T14:03:02Z |
| publishDate | 2023-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj.art-b5c4262e55ea42d5a0d23827890301dc2023-08-21T23:00:21ZengIEEEIEEE Access2169-35362023-01-0111868248683310.1109/ACCESS.2023.330526110216968An Auxiliary Classifier Generative Adversarial Network Based Fault Diagnosis for Analog CircuitYongqiang Zheng0https://orcid.org/0000-0003-3391-0353Dongqing Wang1https://orcid.org/0000-0001-8856-4289College of Electrical Engineering, Qingdao University, Qingdao, ChinaCollege of Electrical Engineering, Qingdao University, Qingdao, ChinaTo solve the analog fault diagnosis problem with fewer samples, a transformer based auxiliary classifier generative adversarial network (ACGAN) is investigated for circuit fault diagnosis by constructing both generator and discriminator in ACGAN with pure transformer components. The transformer has high model generality due to its weak inductive bias, but it also increases the risk of overfitting on small datasets. Therefore, we use ACGAN to generate sample data to enrich the dataset and mitigate the overfitting. However, ACGAN is severely unstable during the training period, for this reason, a confidence mechanism for the discriminator is added to improve the classification accuracy and a new layer normalization method for the generator is studied to avoid the loss of conditional information. Take the sallen-key filter circuit and the biquad high-pass filter circuit as the experiment objects. The experiment results indicate that the transformer based ACGAN diagnosis method can effectively improve diagnosis accuracy, reach 96.22% and 95.35%, respectively.https://ieeexplore.ieee.org/document/10216968/Transformer; analog circuitneural networksfault diagnosisauxiliary classifier generative adversarial network (ACGAN) |
| spellingShingle | Yongqiang Zheng Dongqing Wang An Auxiliary Classifier Generative Adversarial Network Based Fault Diagnosis for Analog Circuit IEEE Access Transformer; analog circuit neural networks fault diagnosis auxiliary classifier generative adversarial network (ACGAN) |
| title | An Auxiliary Classifier Generative Adversarial Network Based Fault Diagnosis for Analog Circuit |
| title_full | An Auxiliary Classifier Generative Adversarial Network Based Fault Diagnosis for Analog Circuit |
| title_fullStr | An Auxiliary Classifier Generative Adversarial Network Based Fault Diagnosis for Analog Circuit |
| title_full_unstemmed | An Auxiliary Classifier Generative Adversarial Network Based Fault Diagnosis for Analog Circuit |
| title_short | An Auxiliary Classifier Generative Adversarial Network Based Fault Diagnosis for Analog Circuit |
| title_sort | auxiliary classifier generative adversarial network based fault diagnosis for analog circuit |
| topic | Transformer; analog circuit neural networks fault diagnosis auxiliary classifier generative adversarial network (ACGAN) |
| url | https://ieeexplore.ieee.org/document/10216968/ |
| work_keys_str_mv | AT yongqiangzheng anauxiliaryclassifiergenerativeadversarialnetworkbasedfaultdiagnosisforanalogcircuit AT dongqingwang anauxiliaryclassifiergenerativeadversarialnetworkbasedfaultdiagnosisforanalogcircuit AT yongqiangzheng auxiliaryclassifiergenerativeadversarialnetworkbasedfaultdiagnosisforanalogcircuit AT dongqingwang auxiliaryclassifiergenerativeadversarialnetworkbasedfaultdiagnosisforanalogcircuit |