Enhancing Welding Training With An AI System For Welding Skill Assessment

Visual assessment of welds is crucial in welding training, yet traditional instructor-led evaluations are prone to subjectivity and inconsistent feedback. This research introduces an innovative artificial intelligence (AI)-driven system for objective evaluation and analysis of welding skills through visual inspection. The system utilizes deep learning models, incorporating Vision Transformer for welding skill classification and YOLO (You Only Look Once) for precise weld defect detection. Experimental validation demonstrates the effectiveness of the Vision Transformer, achieving an 89% accuracy in classifying welding skills based on visual inspection, while the YOLO model attained an F1-score of 0.8 in weld defect detection. This AI-powered approach has the potential to enhance the objectivity, consistency, and efficiency of welding skill assessment in educational settings.