Power converter fault diagnosis using AI tech
The fault characteristics of IGBT open-circuit signal are unstable and the fault sample types are unbalanced, which leads to the inaccuracy of IGBT open-circuit fault diagnosis. Therefore, two unbalance IGBT open-circuit fault diagnosis methods based on 1D-CNN and 2D-CNN are proposed. The experiment...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis-Master by Coursework |
| Language: | English |
| Published: |
Nanyang Technological University
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/167225 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | The fault characteristics of IGBT open-circuit signal are unstable and the fault sample types are unbalanced, which leads to the inaccuracy of IGBT open-circuit fault diagnosis. Therefore, two unbalance IGBT open-circuit fault diagnosis methods based on 1D-CNN and 2D-CNN are proposed. The experimental model based on 1D-CNN is mainly used for real-time monitoring, and as a contrast experiment, it is highlighted that the method needs to be improved. Then, the continuous wavelet transform (CWT) is used to convert the IGBT open-circuit current signal into time-frequency image, and the non-stationary characteristics of the IGBT open-circuit are described effectively. Secondly, a 2D-CNN fault diagnosis model is established based on time-frequency images, which effectively overcomes the influence of sample imbalance on fault diagnosis accuracy. Finally, the comparison experiment shows that the diagnosis accuracy of IGBT open-circuit fault based on the 2D-CNN model is 99.5%, indicating that the method has a higher accuracy in the classification of IGBT open-circuit state under the unbalanced sample scenario. |
|---|