Creating Subsurface Defect Specimens for Deep Learning Analyzing of Radiographic Weld Testing

The purpose of this work is to create more defective specimen for development of deep learning model for non-destructive inspection of welding materials of the workpiece by radiographic testing. Inadequate amounts of defective samples in some rare cases poses a major challenge for the development of defect detection and classification deep learning models. Obtaining actual rare defects from real works is improbable in the industry. Therefore, an approach to create and collect defective welding sample are design and proposed. The specimen design and defect definition are according to international standards.