Development of a Para-AMR algorithm for simulating dendrite growth under convection using a phase-field-lattice Boltzmann method

By combining adaptive mesh refinement and parallel computing, a high performance numerical algorithm was developed to simulate dendrite growth against convection using a phase-field-lattice Boltzmann method. Numerical tests on both 2-D and 3-D dendrite growth cases revealed that, by employing moderate amount of computing resources (10(1)-10(2) parallel processes), this algorithm, without compromising any accuracy, could improve the computational efficiency by 2-3 orders of magnitude, or for most cases shorten the overall elapsed simulation time by 95%, comparing with the normally applied explicit algorithm. Besides, the computational stability or convergence of the algorithm could be maintained even when the local volume fraction of solid approached similar to 100%, which could not be achieved if other implicit algorithms like SIMPLE was employed. (C) 2017 Elsevier B.V. All rights reserved.