Epoxy-related defect detection on pcb of wireless earbuds with transfer learning
Due to the increased manufacturing of wireless earbuds, the semiconductor industry's requirement for PCBs has increased drastically. As the manufacturing of PCBs grows, there is a need to improve the quality control process of the PCB, especially in the defect detection phase, by filtering out...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Year Project / Dissertation / Thesis |
| Published: |
2023
|
| Subjects: | |
| Online Access: | http://eprints.utar.edu.my/6107/1/SE_2005657__YinKarKin.pdf http://eprints.utar.edu.my/6107/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Due to the increased manufacturing of wireless earbuds, the semiconductor industry's requirement for PCBs has increased drastically. As the manufacturing of PCBs grows, there is a need to improve the quality control process of the PCB, especially in the defect detection phase, by filtering out any defective PCBs and stopping them from being used in the manufacturing of wireless earbuds. This study evaluated three deep learning models that could perform defect detection for epoxy-related defects on the PCB of wireless earbuds with at least 90% accuracy. Transfer learning was applied to three pre-trained image classification deep learning models: ResNet50, Xception, and InceptionV3. The models were trained on a real-world PCB dataset provided by ASPL Malaysia after preprocessing the dataset images using OpenCV. ‘Epoxy Overflow on Die’ and ‘Epoxy Overflow on LED’ defects were detected by ResNet50 with an accuracy of 97.3% and 94.0% respectively, while Xception achieved an accuracy of 98.0% in detecting ‘Epoxy on Die’ and ‘FM on Die’ on the testing dataset. |
|---|