Adaptive modeling of the weld pool geometry in gas tungsten arc welding

Skilled welders can observe the weld pool and control the welding process accordingly. This implies that an advanced control system could be developed to control the welding process by emulating the estimation and decision making process of the human welder. In this paper the modeling of the 3D weld pool geometry in GTAW process and this model based predictive control of the pool geometry is investigated. An innovative GTAW 3D vision sensing system capable of measuring the weld pool characteristic parameters in real-time is utilized. Dynamic experiments are conducted to establish the model of the front-side weld pool geometrical characteristic parameters in relation to the welding current and speed. Linear model is first constructed as an average model over certain range of welding parameters. Adaptive model is then adopted to represent the dynamic and time-varying welding process. Further, the adaptive model based predictive control algorithm is developed and an analytical solution of the predictive controller is derived. Simulation results indicate the effectiveness of the proposed controller for tracking various set-points and under varying welding conditions. In the future the welding experiments will be conducted to further verify the proposed controller.