Experimental validation of melt pool temperature and height prediction model for robotic welding

Abstract

This paper develops a control-oriented model to predict fillet melt-pool temperature and bead height in robotic Gas Metal Arc Welding (GMAW) and experimentally validates model’s predictions. The model is derived from first principles through coupled mass- and energy-balance equations that describe the evolution of molten-pool height and temperature using a lumped-parameter formulation suitable for real-time control. Model coefficients describing geometric behavior are calibrated from post-weld measurements of steady-state bead height trends. The model was experimentally validated on a UR10e collaborative robot executing straight fillet welds using two wire diameters (0.889 mm and 1.5875 mm) across voltages of 20-34 V and travel speeds of 8-25 mm/s. A short-wavelength infrared camera captures the in-situ thermal fields while post-weld photographs provide ground-truth geometry for validation. An image-processing pipeline converts the thermal frames into per-frame height estimates that track post-weld profiles with frame-wise deviations of approximately 3.5–10%. The model predicts that temperature increases with voltage and reinforcement decreases with travel speed, matching experimental trends with low error. For pool temperature, the mean absolute and root mean squared errors are 21.7/26.1 $°$C for the 0.889 mm wire and 13.4/16.3 $°$C for the 1.5875 mm wire. For bead height, the corresponding errors are 0.375/0.477 mm and 0.527/0.667 mm.

These results confirm that the physics-based model captures the coupled thermal and geometric behavior of fillet welding with accuracy suitable for model-based estimation and control.