Publication: Detection of solder joint defects on printed circuit board using faster region-based convolutional neural networks
| datacite.subject.fos | oecd::Engineering and technology::Electrical engineering, Electronic engineering, Information engineering::Electrical and electronic engineering | |
| dc.contributor.author | Danish Amsyar bin Mohd Yusof | |
| dc.date.accessioned | 2025-03-27T07:59:27Z | |
| dc.date.available | 2025-03-27T07:59:27Z | |
| dc.date.issued | 2023-08 | |
| dc.description.abstract | This thesis presents a comprehensive study on the development and implementation of a solder joint detection system using the Faster R-CNN framework with a ResNet50 backbone. Solder joint detection plays a crucial role in quality control and inspection within the electronics manufacturing industry. The proposed system leverages deep learning techniques to accurately identify and localize various types of solder joints, including normal, bridging, dirty, and incomplete solder joints. The research begins with the construction of a diverse and annotated dataset, comprising a wide range of solder joint defect scenarios. A detailed analysis of the dataset is performed, followed by the implementation and training of the Faster R-CNN model with a ResNet50 backbone. The training process encompasses an investigation of training loss, learning rate, and validation accuracy to ensure optimal model performance. Experimental results demonstrate the efficacy of the proposed system in detecting and classifying solder joint defects. Evaluation metrics, including average precision, are employed to quantify the model's performance for each solder joint class. The system showcases high accuracy in detecting normal and bridging solder joints, while also exhibiting promising results for incomplete and dirty solder joints. Furthermore, future work is suggested to further improve the system. This includes expanding the dataset to encompass more variations and defects, exploring transfer learning techniques for improved model initialization, employing augmentation techniques for better generalization, extending the system to detect additional defect classes, optimizing the system for real-time deployment, considering edge computing for on-site inspection, and integrating the system with existing quality control processes. Overall, this thesis contributes to the field of solder joint inspection by presenting a robust and automated approach using deep learning techniques. The proposed system has the potential to revolutionize quality assurance practices in electronics manufacturing, ensuring higher product quality, reliability, and efficiency. | |
| dc.identifier.uri | https://erepo.usm.my/handle/123456789/21385 | |
| dc.language.iso | en | |
| dc.title | Detection of solder joint defects on printed circuit board using faster region-based convolutional neural networks | |
| dc.type | Resource Types::text::report::technical report | |
| dspace.entity.type | Publication | |
| oairecerif.author.affiliation | Universiti Sains Malaysia |