<p>In this paper, the competitive growth of columnar grains during selective laser melting of nickel-based superalloys was investigated by combining macroscopic finite element simulations and mesoscopic phase-field simulations based on experimental results. Macroscopic finite element model was developed to investigate the motion of the molten pool with the fluid flow and the heat transfer. The shape of the molten pool from the single-track selective laser melting (SLM) experiment with that from macroscopic simulations was compared. The two-dimensional temperature field of the cross section extracted from finite element simulations was projected into the phase-field mesh for mesoscopic numerical simulations by using the separable cubic interpolation method. In the objective region, the distribution of the temperature field is highly nonlinear as well as the thermal gradient and the cooling rate, and the normal to the liquidus is seen as the main direction of the temperature gradient. Continuous unusual overgrowth events can be identified at the converging grain boundary because the decrease of the primary spacing in the favorably oriented (FO) dendritic array gives rise to the high local undercooling of the FO dendrite at the grain boundary. Predicted results show that the increase of the angle difference between the FO dendritic array and the unfavorably oriented (UO) dendritic array results in more frequent the events of the overgrowth event in the converging cases while the FO dendrites always continuously eliminate the UO dendrite in the diverging cases. Furthermore, it is proved that the interface kinetic attachment plays a crucial role in the competitive growth during selective laser melting process although the solute trapping effect can be neglected.</p>