Evaluation of defects in the GMAW robotic welding process through color analysis

In industry, the evaluation of the quality of welded joints is important to meet the required quality and safety standards. In a general context, identifying defects and porosities in the welding process requires a comprehensive inspection using different methods, such as non-destructive and destructive tests. In addition, welding inspectors require special training and certifications to perform efficient welding inspection. In this paper, we propose to analyze welding through computer vision techniques as an auxiliary tool for identifying defects in coatings. It is proposed to perform a color analysis for the segmentation of defects generated during the welding process found in ASTM A36 steel plates coated in stainless steel 308, using the GMAW robotic process. Automation through computer vision offers an alternative tool to identify and classify welding defects by color segmentation in a database of micrographs. At the end of the process, it is concluded that the HSV and YCbCr color spaces achieve better se mentation of defects in the samples, such as lack of fusion and porosities.