Powder-bed fusion, whether using a laser or an electron beam, is a highly localized process with a very small melt pool dimension compared to the component scale. Therefore, the computational cost of the standard numerical studies such as the Finite Element Method (FEM) to assess a full-range temperature history of a relatively big part is not feasible. The multiscale simulation technique can greatly reduce the computation time by capturing the far-field temperature profile with lower degrees of discretization [1], [2]. Therefore, a multiscale FEM thermal analysis with optimized spatial and temporal discretizations was developed in this work to examine temperature history at a point of interest (POI) in a cubic sample of AlSi10Mg (Figure. 1a). The model was prepared using Abaqus special-purpose techniques for additive manufacturing [3] to set up progressive element activation, heating by the moving heat source, and cooling via convection and radiation. 

The obtained temperature field can be imported into successive temperature-dependent analyses such as mechanical and thermo-metallurgical models. In this work, it was used to capture the initial temperature for a high-fidelity melt pool scale solidification-aware thermal analysis (Figure 1b). An accurate melt pool scale analysis is required when performing thermal analysis to be coupled with solidification simulation. Thus, two effective features were incorporated into the in-house Cellular Automata Finite Difference (CAFD) analysis. The first feature was developed based on coupled rough and fine FD analyses and Rosenthal’s equation to ensure realistic boundary conditions. The second feature was a fully-coupled thermal-solidification analysis within the main domain for a correct release of latent heat of fusion based on the non-equilibrium solidification conditions. The results revealed a significantly improved agreement between simulation and experimental microstructures and textures. Furthermore, it was shown that this method does not require a calibrated Goldak heat source to manipulate the melt pool length; instead, a symmetric Gaussian model could produce a consistent melt profile.