Enhancing PAUT Inspection With Machine Vision and AI for Intelligent Structural Integrity Assessment

Abstract

Abstract A variety of flaws and defects may occur during the welding processes of critical structural members used in infrastructures. If these flaws exist in a considerable size and quantity, they significantly influence the quality and structural integrity of the infrastructure. The characteristics of these flaws such as size, location and distribution influence decisions such as acceptance, repair, or rejection of the welded components. Since these decisions have significant time and cost consequences for any industrial project, they must be made based on the highest accuracy and measurement reliability possible. Traditional welding inspection methods are often labor-intensive, subjective, and limited by human expertise and accessibility in complex or hazardous environments. This implies a crucial need for an advanced NDE technique that provides accuracy, repeatability and efficiency. This research seeks to overcome these limitations by developing an automated inspection framework that combines robotic scanning with AI-driven data analysis and Machine Vision (MV), enabling real-time, high precision detection and characterization of welding flaws. The integration of robotics and AI in welding inspection will significantly improve efficiency, repeatability, and accuracy in flaw detection, ultimately contributing to the safety and quality assurance of welded structures across industries such as aerospace, automotive, and civil infrastructure.