Microstructure-informed modeling of LPBF AlSi10Mg components: understanding the origin of tensile anisotropy through local stress-strain partitioning using comparative crystal plasticity material models

This study quantifies the microstructure controlled tensile anisotropy of nearly fully dense laser powder bed fusion (LPBF) AlSi10Mg. Tensile specimens built at 0°, 45°, and 90° relative to the build direction were selected, enabling a focus on orientation dependent grain morphology and crystallographic alignment. EBSD analysis revealed pronounced differences in grain shape and alignment across orientations, which were reproduced in a statistically representative 3D polycrystalline RVE (100 × 50 × 100 μm³, ∼590 grains) generated using DREAM.3D. Crystal plasticity simulations were performed using DAMASK under periodic boundary conditions using two constitutive frameworks: (i) a phenomenological slip-based model and (ii) a dislocation density-based physics model. The parameters of both models were calibrated against the experimental tensile data through Bayesian optimization with an objective function combining stress-strain and hardening-rate errors. To capture build orientation effects with minimal parameter freedom, only one strength related parameter was varied in each model, while all rate sensitive and hardening parameters were held constant across orientations. The orientation dependent strength parameter was linked to the dimension of Si-rich cellular network rather than treated as a purely empirical fitting term. Both formulations reproduced the experimental anisotropy and predicted higher local stress accumulation in the 0° condition and more uniform strain accommodation in the 90° condition at the same macroscopic strain. The higher flow stress in the 0° condition was associated with the smaller effective Si-network spacing along the loading direction, which increases dislocation-barrier interactions. Conversely, the larger spacing in the 90° condition promoted easier plastic flow. These results demonstrate that tensile anisotropy in LPBF AlSi10Mg can be captured using a large, grain-resolved mesoscale RVE while implicitly incorporating sub-grain Si-network effects through a physically informed strength parameter. The comparison also clarifies the relative capabilities of phenomenological and dislocation density-based formulations for predicting orientation dependent deformation.