نوع مقاله : مقاله پژوهشی
عنوان مقاله English
نویسندگان English
This study introduces a novel deep learning-based approach for the automated and accurate detection of surface defects in piston rings, utilizing convolutional neural networks (CNNs). A high-quality dataset of defective and non-defective piston ring images was collected and meticulously annotated. We trained a U-Net model on this dataset, which effectively segments defective regions, achieving an impressive Intersection over Union (IoU) score of 84% and a loss of 0.15. These results demonstrate the model’s high precision in identifying various defect types. A comparative analysis with traditional defect detection methods underscores the superiority of our deep learning approach in both accuracy and processing speed. Additionally, the experimental findings align closely with theoretical predictions from prior research, further validating the proposed model. This research represents a significant advancement in automotive quality control, offering the potential to reduce manufacturing costs associated with defective components.
کلیدواژهها English
