Towards Automatic Crack Detection by Deep Learning and Active Thermography

• Published in: Advances in Computational Intelligence Vol. 11507; pp. 151 - 162
• Main Authors: Moreno, Ramón, Gorostegui-Colinas, Eider, de Uralde, Pablo López, Muniategui, Ander
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2019; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Cracks detection; Deep learning; EBW; Image segmentation; Induction; NDT; Thermography; TIG
• ISBN: 3030205177; 9783030205171
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-20518-8_13

Metal joining processes are crucial in current technological devices. To grant the quality of the weldings is the key to ensure a long life cycle of a component. This work faces crack detection in Electron-Bean Welding (EBW) and Tungsten Inert Gas (TIG) weldings using Inductive Thermography with the aim to substitute traditional Non-Destructive Testing (NDT) inspection techniques. The novel method presented in this work can be divided up into two main phases. The first one corresponds to the thermographic inspection, where the thermographic recordings are reconstructed and processed, whereas the second one deals with cracks detection. Last phase is a Convolutional Neural Network inspired in the well-known VGG model which segments the thermographic information, detecting accurately where the cracks are. The thermographic inspection has been complemented with measurements in an optical microscope, showing a good correlation between the experimental and the prediction of this novel solution.