Heat Transfer Modeling of Friction Stir Welding Using PINNs

In this study, we explored the influence of rotational speed on the heat transfer mechanism in friction stir welding (FSW) using physics-informed neural networks (PINNs). The PINN model effectively incorporates prior physical knowledge, such as governing equations, boundary conditions, and initial conditions, to comprehend and forecast intricate physical behaviors within the welding process. The results indicated a significant correlation between rotational speed, friction coefficient, heat generation, and the thermal diffusion of the system. The findings from the PINN learning emphasized the role of the friction coefficient and the translational movement tilt angle of the tool in affecting the temperature profile and weld quality. Despite constraints within a two-dimensional domain, the PINN-based approach emerges as a promising tool for understanding the intricate heat transfer mechanisms in FSW, and it has potential applications in future research for three-dimensional welding modeling between dissimilar materials.