Parallel partitioning algorithm for numerical simulation of gas-liquid two-phase flow during the mold filling process

Abstract

The two-phase simulation of casting filling, which has high accuracy for the velocity field, pressure field, and the gas-liquid interface behavior, has been widely applied to the many fields of materials processing, especially in the prediction of casting filling process. However, a disadvantage of two-phase simulation is the high demand for computation resource, which makes the calculation time-consuming for practical application. In the study, based on the SOLA Level-Set method for gas-liquid two-phase flow during the mold filling process, a new parallel partitioning algorithm standard is proposed by using MPI (Message Passing Interface) parallel programming technique. A region-based algorithm based on dynamic programming is designed and a parallel partition model is established. A parallel program for gas-liquid two-phase flow numerical simulation is developed. A mapping array method is proposed to reduce the communication overhead between sub-domains. The communication steps between sub-domains in the process of solving the complex flow field are designed. Tests on actual castings show that the parallel algorithm has a good performance. The speedup ratio with 4 to 8 processing cores is above 2, and the maximum can reach 4.6.