Unveiling the influence of printing surfaces in powder bed fusion electron beam melting through multiphysics simulation

Controlling internal defects within as-built parts is one of the great interests in the additive manufacturing field. In this study, we explore the powder spreading and defect evolution mechanisms on realistic printing surfaces through a comprehensive multiphysics simulation. The efficacy of a flat surface criterion for internal defect elimination was verified using a machine learning approach. The steady layer thickness in the electron beam melting process was estimated for different printing surfaces using the simulated powder bed density obtained through a realistic discrete element method model. The steady layer thickness was greater on the flat printing surface compared to the rough surface due to high consolidation shrinkage. Monte-Carlo simulation revealed that electron backscattering is more pronounced on peaks of a rough surface than on a powder bed, due to the limited reabsorption of reflected electrons. The influence of the printing surface on melt pool stability and internal defect evolution was investigated using thermo-fluid dynamic simulations. Under identical process conditions, the molten pool surface exhibited greater stability on a rough printing surface than on a flat one, due to enhanced fluid flow. The flat printing surface resulted in lack of fusion defects < 100 μm in the external side region due to suppressed heat accumulation and a large steady layer thickness. Periodic deep valleys on rough surface can cause coarse defects < 200 μm in the external side region, as the melt pool depth is insufficient to match the increased local layer thickness in the valleys. Therefore, it was demonstrated that the printing surface must be considered to optimize outermost defects in as-built parts produced by the powder bed fusion electron beam melting process. • Effectiveness of a flat surface criterion for defect optimization was verified using a machine learning approach • Powder bed quality on rough surfaces improved due to the structural trapping of particles. • Electron backscattering was reduced on the powder bed through the reabsorption of reflected electrons. • Outermost defects were minimized using kernel SVM optimization with a criterion of 25 μm < Sq < 35 μm