September 24, 2024
Radiographic testing is the most widely used non-destructive method for detecting
discontinuities in industrial piping welds. However, human interpretation of radiographic films is
both time-consuming and demands a high level of expertise. Despite the significant progress in
deep learning techniques in related fields, such as medical radiography, previous research efforts
focusing on weld discontinuities have been limited by a lack of training data and the insufficient
representation of real-world conditions.
This paper presents a comprehensive system that automatically detects welding zones, evaluates
film quality, and classifies weld discontinuities in piping processes. The proposed framework
exhibits superior generalization capabilities, transcending the constraints of a specific industry or
piping size. Key advantages of our approach include increased accuracy, fast processing times,
and automated interpretation of welding films across a broad spectrum of image qualities. As a
result, it achieves outstanding detection and classification performance, providing significant
benefits for welding inspection and quality assessment.
For more details : View article document
