A cellular automaton approach for the prediction of grain size in grain refined alloys

Grain refinement by inoculation is widely practised in metal casting, in particular for aluminium alloys. During the last decades several modelling techniques have emerged to quantitatively predict solidification microstructures. The prediction of grain size remains however a challenge due to the complex competition between nucleation and growth, and physical phenomena taking place at very different length scales. Models often address the thermal recalescence and sometimes the suppression of nucleation by the solute boundary layers. The possible clustering tendency of the nucleant particles is however usually not considered, although it can play an important influence on the final grain size. This effect is particularly difficult to quantify because it requires a small-scale approach to describe the clusters, while a large computational domain is needed for grain size statistics. A cellular automaton (CA) model tracking the grain envelopes was developed to address this challenge. The spatial distribution of the nucleant particles and the solute diffusion field in the intergranular liquid were represented directly on the CA grid. Predicted grain sizes were first compared successfully with published experimental and modelling data. An investigation of the role of clustering of nucleant particles was then carried out, which demonstrated the potential of the model to address this topic.