GPU-accelerated three-dimensional large-scale simulation of dendrite growth for Ti6Al4V alloy based on multi-component phase-field model

Abstract

Ti6Al4V is one of the most widely used ternary alloys in additive manufacturing. Its mechanical properties are highly determined by the prior β phase. However, simulation of the ternary alloy solidified microstructure is limited for the lack of available methods. In this study, we carried out two-dimensional (2D) and three-dimensional (3D) simulations based on a new multi-component (MC) phase-field (PF) model by extending Karma’s binary PF theory. The correctness and accuracy of the MC model are validated by the Gibbs-Thomson relation at dendrite tip. With the equipment multiple GPUs, the highest speedup ratio more than 150 is obtained compared with serial programming running on one of the top CPUs and simulations with billion nodes can be completed within acceptable time. Using this model as benchmark, previous Ti6Al4V dendrite evolution using pseudo-binary PF model is investigated and found magnifying the driving force and growth artificially. 3D large-scale directional solidification simulations also shed light on dendrite merging which leads coarse primary dendrites. We believe that the GPU-accelerated MC PF framework provides an accurate and efficient insight in understanding the solidifying evolution for Ti6Al4V.