Weld penetration control in gas tungsten arc welding (GTAW) process

In this paper a predictive control system is developed to control the weld penetration in GTAW process. An innovative 3D vision based sensing system is used to measure the weld pool characteristic parameters in real-time. Weld penetration specified by the back-side bead width as output feedback is calculated by a neuro-fuzzy model using weld pool characteristic parameters as its inputs. Dynamic linear model that correlates the penetration status to the welding inputs (current and speed) is constructed and further improved by incorporating a nonlinear operating point. A linear model based predictive control algorithm is proposed and an analytical solution is derived. Welding experiments confirm that the developed control system is effective in achieving the desired weld penetration state under different initial conditions.