The process of welding has strong repeatability for the welding of the same parts and the same welding track. A tracking control method for gas tungsten arc welding (GTAW) process based on iterative learning control is proposed. An iterative learning control algorithm for the control of GTAW process is designed based on the dynamic process model of welding. And the convergence of the algorithm is proved. The results of research show that the repeated information of welding can be effectively used by the iterative learning control in the process of welding. After about 60 times of iteration learning, the output of welding system can better track on the desired trajectory to meet the expected control effect. It verifies the effectiveness of the proposed method. The better tracking performance can be acquired by ILC in contrast to PID algorithm and it also can suppress the effect of repetitive external disturbances.