Weld defect identification and characterization in radiographic images using deep learning: Review

Abstract Defects in the welds degrade quality of weld. Weld defect identification is a challenging task industry because wide range weld imperfections. detection using radiographic images an effective technique for achieving good shipbuilding and aerospace applications. Foreign inclusions, cracks pores are examples welding joint Several appropriate computer-based image processing techniques have made defects possible. It imperfection can show various sizes, shapes, contrasts locations radiography images. The accuracy this inspection process more dependent on external factors also time-consuming. Automatic needed by analyzing obtained directly from digital systems. This paper uses unique image-based approach to small batch X-ray imaging datasets investigate potential solution identification. article compares deep learning network's performance parameter hyper-parameter combinations. Also it traditional approaches manual method, feature-based identification, finally - based several types industrial comparative analysis concludes that learning-based achieved as compared conventional techniques. research highlights few challenges future directions area.