Performance Evaluation of Vision Inspection System for MIG Welding Defects

Metal Inert Gas (MIG) welding is one of the major metal joining process used to fabricate many engineered artifacts and structure such as cars, ships, space shuttles and pipe lines. Flaws resulted from welding operations are detrimental to the integrity of the fabricated artifacts or structure. Although the welding process is carried out as manually or automatically, flaws are formed during the welding operations.These flaws include lack of fusion, porosities, cracks, lack of penetration, excess weld, insufficient weld, inclusions, gas holes etc. To maintain the desirable level of structural integrity, welds must be inspected according to the established standards. In this paper, a machine vision system is introduced to extract the various features of the MIG welded joint by capturing image through CCD camera with proper illumination, and then various image processing techniques and classifier is used to calssify the defects accoriding to the international standards. This vision system is connected to the host computer and classification is done by artificial neural network based on predefined one. In this proposed method, a comparison is made between the accuracy of the single image by turn on four zones LEDs of the illumination at a time with the accuracy of the multiple images by the zones LEDs are turned on, one after the other This proposed method enables overall accuracy of the four zones of the images as 95% from the 40 samples of the welded images and finally parameters are used to evaluate the performance of the proposed system.