Control of weld pool width and cooling rate in circumferential GTA welding of a pipe by using neural network model

This paper deals with the problem concerning the weld pool width and the cooling rate in circumferential GTA welding of a pipe. In order to obtain the high quality of the weld, it is important to keep the weld pool width and the cooling rate constant. The numerical simulations in the circumferential GTA welding of a pipe are performed by using the finite difference method. A weld pool model is constructed by a neural network. The welding speed and the welding current are determined to obtain the desired pool width and the cooling rate by using the neural network model and the steepest descent method. A new approach is proposed in which the welding is made clockwise and counterclockwise simultaneously from 0° of pipe by using twin heat sources