Automated Identification of Sheet Metal Parts Using Deep Learning Techniques

Accurate identification and classification of sheet metal parts are crucial for optimizing production workflows, ensuring quality control, and reducing operational costs in manufacturing industries. Traditional identification methods, such as barcoding, labeling, marking, and inkjet printing, have several limitations, including label deterioration, misplacement, and errors caused by manual handling. This study presents a deep learning-based automated approach for identifying sheet metal parts, significantly improving efficiency, reliability, and scalability. The methodology involves converting 2D drawings into an image format, applying data augmentation techniques, and training a Convolutional Neural Network (CNN) using the InceptionV3 architecture. Furthermore, a semantic segmentation model based on U-Net is used to isolate sheet metal parts from their background, ensuring accurate classification. This study presents an integrated framework that combines classification and segmentation techniques specifically tailored for the automated identification of sheet metal parts in industrial environments. The proposed system achieves an outstanding classification accuracy with a Top-1 accuracy of 96%, Top-3 accuracy of 99.36%, and Top-5 accuracy of 100%. By automating the identification process, the proposed system eliminates the need for manual intervention, minimizes errors, and enhances traceability, making it a promising solution for smart manufacturing environments. The results demonstrate the potential of deep learning models in industrial automation and quality assurance processes in the sheet metal industry.