3D Cellular Automaton Modeling of Grain Structure Evolution during Laser Powder Bed Fusion and Laser Rescanning in Al-Si Alloys

Abstract

3D Printing, or Additive Manufacturing (AM), enables the fabrication of complex components directly from digital designs through the controlled addition of material. It has great advantages in design flexibility, material efficiency, and customization compared to traditional manufacturing. Among metal AM processes, Laser Powder Bed Fusion (LPBF) has become one of the most widely adopted techniques, which is particularly suitable for producing high-precision, high-performance parts. However, LPBF-fabricated parts often suffer from unsatisfactory mechanical performance, due to undesirable columnar grain structures, anisotropic mechanical properties, and residual stresses under conditions of rapid solidification and intense thermal gradients. Laser Rescanning (LR), an in-process heat treatment strategy that re-traces previously solidified paths, has been proposed to reduce these drawbacks by refining microstructures and improving part quality. This thesis develops a comprehensive three-dimensional Cellular Automaton (CA) model, coupled with Finite Element Analysis (FEA), to investigate grain structure evolution in Al–10Si alloys during LPBF and LR processes. Particular attention is given to the incorporation of nucleation mechanisms, solute redistribution, and eutectic substructure formation, with the goal of understanding and predicting grain refinement under varying processing conditions. This thesis presents a three-part modeling study to investigate grain structure evolution and refinement mechanisms during LPBF and LR.

This thesis presents a three-part modeling study. The first part establishes a baseline simulation framework by coupling a FEA (using a Goldak heat source), with a mesoscale CA grain growth model. Simulations of LPBF and LR under a unidirectional scanning strategy reveal variations in melt pool geometries and thermal profiles at the two scanning stages. The grain structure exhibits a mixture of equiaxed and columnar grains, including fine equiaxed grains near melt pool boundaries due to fusion boundary nucleation, and a columnar-to-equiaxed (CET) transition driven by local thermal conditions from the lower part to the center of the melt pool. Quantitative analysis using Principal Component Analysis (PCA) confirms a reduction in grain size about 20% after LR. This is attributed to the formation of shallower melt pools and extended volume of equiaxed grains resulting from fusion boundary nucleation, as well as increased cooling rates during LR.

Building on this foundation, the second part discusses the effects of varying LR process parameters and scanning strategies. On the one hand, the model successfully predicts grain structure evolution under varying process parameters, demonstrating its capability to capture the effects of different laser powers and scanning speeds. Simulations conducted under two representative LR conditions indicate that lower power and higher speed tend to promote finer grain structures. This highlights the model’s potential for microstructure tailoring across a broader range of processing conditions. On the other hand, a comparative study of scanning strategies, including unidirectional, bidirectional, and island rescanning, demonstrates that strategies with frequent turning points enhance grain refinement and reduce texture intensity. To further examine whether increasing turning frequency continuously improves refinement, systematic variations of scanning lengths and scanning angle rotations are performed. The results show that grain refinement does not vary monotonically with scanning length, but confirm that grain size and orientation can nonetheless be effectively tailored through scanning path design.

The final part of the study applies the developed 3D CA model, extended to incorporate eutectic solidification, to simulate grain and sub-grain structure evolution during laser scanning and rescanning on Al–10Si alloy. A novel CA approach is developed to simulate the multiphase solidification behavior of Al–10Si, capturing both primary α-Al dendrites and eutectic Si-rich networks. It also confirms the grain refinement behavior driven by rescanning. Simulated grain and sub-grain structures align with EBSD and SEM observations. The model also reveals the formation of solute segregation bands, consistent with EDS data, highlighting the critical role of composition redistribution in promoting grain nucleation near melt pool boundaries.

Overall, this thesis establishes a validated and versatile Cellular Automaton-Finite Element (CAFE) model capable of predicting complex grain structure evolution and sub-grain features in LPBF and LR. The model offers insights into solidification mechanisms and provides a foundation for tailoring process parameters to control microstructure in Al–Si alloys. Meanwhile, the model has the potential to be extended to other alloy systems with similar solidification characteristics, enabling broader applications in microstructure engineering for AM (e.g., LPBF).