Cellular automata-lattice Boltzmann model for polycrystalline solidification with motion of numerous dendrites

•A cellular automata-lattice Boltzmann model for polycrystal solidification is built.•The growth and motion of kinematic dendrites using a dynamic grid scheme.•The impulse-based method is used for computing the collision of dendrites.•The mesh dependence of dendritic growth under Cartesian grid is greatly reduced.•The GPU parallel code is developed using the CUDA C++ language. A GPU-accelerated cellular automata-lattice Boltzmann combinatorial model is developed for calculating the preferred growth, movement, and collision behavior of equiaxed crystals in supercooled melts of binary alloys. For moving dendrites, the growth is computed in a dynamic grid that grows with the body, and continuous movement is achieved by moving the dynamic grid. The impulse-based method is used for the collision of dendrites to calculate the post-collision velocity. Each module of the model was rigorously benchmarked, proving that the model has good computational accuracy and efficiency. The model was used for modeling the solidification of an Al-3 wt% Cu alloy, simulating the growth of abundant kinematic equiaxed crystals in a rotating flow and the falling and stacking of dendrites in droves and subsequent grain growth during the columnar to equiaxed transition, respectively.