Statistical Characteristic for Detecting Weld Penetration Defects in Gas-Metal Arc Welding

Abstract

In robotic arc welding, many disturbances often result in weld defects, such as incomplete penetration and burn-through. The development of an intelligent monitoring system for weld defects in real time during the robotic arc welding process is of great significance. In this paper, robotics-based gas-metal arc welding experiments are carried out on the mild steel testpieces with ‘V’-type groove. Two disturbances are intentionally introduced to produce two kinds of weld penetration defect, i.e. incomplete penetration and burn-through. A through-the-arc sensing method is used to capture the transient values of the welding voltage and current. The raw data of the captured welding current and voltage are processed, and the statistical characteristic M10 is extracted to correlate the welding conditions to the weld penetration information. This lays the foundation for intelligent monitoring of weld quality in robotic arc welding.