A Study on the Algorithm for Determining Back Bead Generation in GMA Welding Using Deep Learning

In arc welding, the generation of back bead is considered as one of the main factors that determine the mechanical characteristics of the welded structure. The existence or shape of the beads can be observed by destructive inspection, which cuts the cross section, or non-destructive inspection, which uses visual or ultrasonic waves. In recent manufacturing processes, the demands of high quality and factory automation are continuously presented, and the importance of research on a more efficient real time diagnosis system that can reduce the time and cost for the detection of structural defects of welds is becoming higher. In this study, an algorithm is developed to determine the back bead generation in real time by using current and voltage signals measured in real time by applying deep learning which is one of artificial intelligence techniques. The result proposes a system to determine whether back bead is generated or not, using deep neural network, a type of deep learning.