Design and Research of Intelligent Weld Defect Detection System

In the welding process, defects such as cracks, porosity, incomplete fusion, incomplete penetration, and slag inclusion may occur due to welding technology, environmental factors, and other influences, which directly affect the service life and performance of the welded parts in all aspects. Traditional manual weld inspection methods are inefficient, costly, and susceptible to subjective influences. In this paper, we propose an intelligent weld seam inspection system, which first collects weld seam defect sample images, builds a 256∗256 pixel defect sample library, and then uses a random forest algorithm to establish a defect recognition model. After the weld seam image is identified by the model, the defect area can be automatically located on the image and the defect type can be displayed. Experiments have shown that the system is highly accurate in identifying weld defects and can be widely used in the machine building industry and the electrical and electronics industry.