Automated detection of soldering splashes using YOLOv5 algorithm

Abstract

This paper deals with automated visual inspection of electronic boards in serial manufacturing of power electronics devices. Soldering splashes generated in the relevant phases of manufacturing can decrease the quality, parameters and lifetime of hybrid power semiconductor modules. Soldering splashes can occur in restricted area of electronic board and must be removed. Automated inspection is provided using neural network YOLO trained on image dataset of electronic boards acquired by authors in SEMIKRON Slovakia company. Implemented method will lead to higher reliability of manufacturing process.