Simulation of welding shape process in P-GMAW based on neural network models

As one of efficient and good-adaptability welding methods, pulsed gas metal arc welding(P-GMAW) has been applied in industrial production widely. In this paper, the modeling and simulation methods in P-GMAW shape process of low carbon steel were studied. Firstly, a series of BP neural network dynamic models were established for P-GMAW shape process; then stable-state and dynamic simulations were implemented with these modes to reveal the welding form rules in P-GMAW. Meanwhile, this paper proposes a method that using the neural network model to investigate the relationship between the top side weld pool characterized parameters and the backside weld pool width. With the proposed methods, the validity and reliability of the topside weld pool characteristic parameters were verified. The methods and results of these modeling and simulation provide the conditions for exploring the welding shape rules and designing the welding process controllers.