Welding defects are known as abnormalities produced in the specified welding material suitable to incorrect welding execution or incorrect welding model. Weld defects are classified as internal weld defects and External weld defects. Internal weld defect occurs inside the weld metal whereas external weld defect occurs at the surface of the weld metal. Inner welding abnormalities are named as slag inclusion, unfinished fusion, necklace cracking, unfinished diffusion and Toe crack etc. External weld defects such as spatter, porosity, crater, under cut etc. Welding defect can be detected only by using intelligent man power and eliminated using high current at the starting and proper filler material. In order to substitute man power, the Novel feature extraction algorithm which detects the weld defect with more accuracy is proposed. For this ultrasonic S-scan images of Toe crack which is one of the internal cold crack which is being maximum occurring while welding or manufacturing any industrial machines or space craft’s is used. Further analysis is done using the algorithms Such as HOG (Histogram of oriented gradients), and GLCM parameters. Neural network which uses the machine learning algorithm (K-nearest neighbor) is trained using the unique features extracted from the algorithms individually to prove its accuracy and novelty.
Volume 12 | 07-Special Issue
Pages: 1934-1943
DOI: 10.5373/JARDCS/V12SP7/20202307