[4B4] Localisation and segmentation of welding parts in ultrasound scans using deep learning
R Rhéaume
Université du Québec à Trois-Rivières, Canada
Welded joints are an essential part of structures. Joining materials using welds allows for sturdy structures. To ensure the safe use of structures with welded joints, welding defects must be identified and analysed by qualified analysts using phased array ultrasound testing scans. These analyses help to identify defects, determine their criticality and establish future actions such as repairing, periodic monitoring or rejection of the joint. To further refine the analysis, identifying and localising the welded joint itself is necessary to add geometric context to the defects. The criticality of the defects highly depends on their location with respect to the weld. This paper presents a segmentation tool that localises and identifies the weld, its type and the location of defects.
Keywords: deep learning, segmentation, weld parts, indication, phased array ultrasound scans.
Keywords: deep learning, segmentation, weld parts, indication, phased array ultrasound scans.
