Enhanced Metal Sheet Quality Control Using CNN and SVM Integration

This study introduces an innovative approach to detecting and classifying defects in large metal sheets by combining Convolutional Neural Networks (CNN) with Support Vector Machines (SVM). The research focuses on four specific types of flaws: wavy edges, one-sided wavy edges, coil breaks, and central buckles. The proposed hybrid model delivers remarkable results, achieving an overall accuracy of 99% across all defect types. It consistently surpasses 96% in accuracy, recall, and F1-scores for each defect category, demonstrating its reliability and effectiveness. With macro, weighted, and micro averages all reaching 97.01%, the model exhibits a wellbalanced performance. This outstanding accuracy and consistency underscore the potential of integrating CNNs with SVMs for robust quality control in metal sheet manufacturing.