A study on weld defects classification in gas metal arc welding process using machine learning techniques

Abstract

Gas metal arc welding process (GMAW) is one of arc welding processes commonly used in industries due to its wide range of applications and economic advantages. In GMAW process, the arc interacts with environment leading to weld defects that are realized in post weld non-destructive techniques (NDT). This leads to necessity of in-process monitoring and control of the process to ensure quality by defect free welds. The present work is a preliminary study intended to develop an in-process monitoring system that can identify and classify the defects in GMAW process. In this study, experiments are conducted on tube-to tube butt joints in flat position by varying the process variables such as current, voltage, travel speed and contact-tube-to-workpiece-distance (CTWD). It involves good weld as reference for three types of defects such as porosity, burn through and lack of penetration were considered. The instantaneous current and voltage signals were recorded using acquisition system that is later used in statistical features extraction. Based on statistical features such two classification techniques includes decision tree and support vector machine (SVM) were deployed to classify the defects with reference to good weld and their efficiencies are reported.