Automatic Detection and Classification of Laser Welding Defects

Abstract

Laser welding is one of the most common methods used in joining metal materials. However, welding errors are usually encountered during the welding process due to external factors. The detection and correct classification of these welding faults are of great importance for the reliability of the weld and the material which was welded. Traditionally, welding error detection is usually handed by visual inspection. However, visual inspection is an error-prone and slow process. In this study, an image processing and machine learning based method is proposed to automatically detect welding defects. The welded materials are divided into sub-images and each sub-image is examined for defect detection and defect type classification. The study has been examined in 2 aspects as 2-class classification and 3- class classification. As a result of the study, an average of 90% precision is obtained with Logistic Regression (LR) and 91% precision is obtained with the Support Vector Machine (SVM) in the 2-class classification process. As for the 3-class classification, an average of 91% precision is obtained with the LR, and 93% precision values are obtained with the SVM. © 2022 Elsevier B.V., All rights reserved.