The use of computational methods to perform inspection tasks in the modern industry has increased in the last years, the fast response and repeatability are the key factor to define a standard in quality control. The visual inspection of components was in charge of highly skilled operators who were searching for defects or failures, but the complexity of the inspection triggered the necessity to migrate to a computational approach. This computer inspection system is composed of the hardware and the algorithm. The hardware includes the lighting and the camera, meanwhile, the algorithm works directly with the failure identification.

In the welding industry, the SMAW (Shielded Metal Arc Welding) consists of using an electrode stick that is melted for the heat produced from an electro-voltaic arc between the electrode and the surface of the base metal. The quality of the welding depends on the quality of the electrode, that is why this project is focused on the visual inspection of the electrodes. The visual defects that compromise the welding quality can be fixed in two categories: the shape of the electrode and the texture of the surface. These two categories contemplate the existence of cracks, porosity and extrusion defects.

The method which delivered the best results (94% of accuracy, 92% of repeatability) used a dark background, colour segmentation, image pre-processing (border enhancement), YOLO network for deep learning and image entropy performed in MATLAB. Additionally, the detection method allows identifying the position of the defects, their size and their type.The location of the defect is determined thanks to the division of the electrodes into smaller parts, if a failure is detected in one or more of them it is possible to know the position and the size of the defect. Meanwhile, the type of defect is the result of analysing the deep learning and entropy results. The collected failure data is presented in the inspection report, which has the potential to become a tool that could modify and enhance the manufacturing processes.