An AI-Driven Approach: To Detect and Predict Weld Film Irregularities for Enhanced Quality Control

Abstract

Weld quality assurance is a critical aspect of industrial applications, where defects such as porosity, “cracks”, “lack of fusion”, and “slag inclusion” can compromise structural integrity and safety. Traditional “Non-Destructive Testing (NDT)” methods such as Radiographic Testing (RT) rely heavily on manual inspection, which is time-consuming, error-prone, and subjective. Recent advancements in “Artificial Intelligence” (AI) have introduced new possibilities for automated defect detection and prediction.
This study proposes a “Hybrid AI”-driven approach that integrates “Convolutional Neural Networks” (CNNs) for defect detection in radiographic images with “Machine Learning (ML)” algorithms (“Random Forest”, XGBoost, and “Gradient Boosting”) for defect prediction based on welding process parameters. The research utilizes the GDXray dataset for radiographic weld images and historical welding parameters to develop an intelligent defect detection and prediction model.
The results demonstrate that the hybrid AI model outperforms traditional NDT approaches, achieving high accuracy in defect classification while providing predictive insights that allow for proactive quality control. This study contributes to Industry 4.0 applications, improving weld quality management, reducing costs, and enhancing manufacturing efficiency.