Classification of Spot-welded Joints in Laser Thermography Data using\n Convolutional Neural Networks

Abstract

Spot welding is a crucial process step in various industries. However,\nclassification of spot welding quality is still a tedious process due to the\ncomplexity and sensitivity of the test material, which drain conventional\napproaches to its limits. In this paper, we propose an approach for quality\ninspection of spot weldings using images from laser thermography data.We\npropose data preparation approaches based on the underlying physics of spot\nwelded joints, heated with pulsed laser thermography by analyzing the intensity\nover time and derive dedicated data filters to generate training datasets.\nSubsequently, we utilize convolutional neural networks to classify weld quality\nand compare the performance of different models against each other. We achieve\ncompetitive results in terms of classifying the different welding quality\nclasses compared to traditional approaches, reaching an accuracy of more than\n95 percent. Finally, we explore the effect of different augmentation methods.\n