Deep Learning-Based Weld Contour and Defect Detection from Micrographs of Laser Beam Welded Semi-Finished Products

Laser beam welding is used in many areas of industry and research. There are strategies approaches to further improve the weld seam properties laser welding. Metallography often needed evaluate welded seams. Typically, images examined evaluated by experts. The evaluation process qualitatively provides welds. Particularly times when artificial intelligence being more processes, quantization that could previously only be determined gaining importance. In this contribution, we propose use deep learning perform semantic segmentation micrographs complex achieve automatic detection properties. A dataset created containing 282 labeled images. training performed with DeepLabv3+. trained model achieves a value around 95% for contour 76.88% mean intersection over union (mIoU).