Simulation And Research Of 3D Dendrite Growth Phase-field Model Based On MPI

With the rapid development of information technology and computational materials science advances,the simulation and research of 3D dendrite growth phase-field model is realized by numerical simulation technology.Phase field as one of the most effective microstructure numerical simulation method using phase field simulation of a binary alloy solidification process has become a hot research direction of the metal material microstructure simulation studies.However,due to the variable phase field model of binary alloy contains numerous and complex,traditional serial programs subject to static memory size of the program itself,cannot expand the size of the field phase,so binary alloy phase field model most still remain in the two-dimensional stage,while the calculation time is long,low computational efficiency has become a three-dimensional simulation of bottlenecks,so how to dramatically increase the size of the field phase,reduce computing time and improve computational efficiency is a binary alloy phase field simulation studies need to be resolved problem.In this paper,the MPI parallel technology is used to solve the problem of the two element alloy phase field model simulation and the following research work is mainly carried out:(1)In this paper,the MPI parallel technology is used to realize the binary alloy phase field model.The parallel simulation results are compared with the serial serial simulation results.The results show that the parallel simulation results are reliable and effective.Simulation scale is expanded from 2503 meshes to 1000 * 1000 * 500 meshes,and large scale simulation of 3D dendrite of a binary alloy is realized.In addition,we analyze the growth state of a large scale of the binary alloy.(2)It shows that the speedup can reach 15.45 times by the group communication and the speedup is higher than 15.45 times speedup of point-to-point communication.The maximum analog scale of group communication is equivalent to that of serial.The maximum analog scale of point to point communication increases with the increase of the processes number;The acceleration effect is best according to the X axis in a variety of MPI domain division and the acceleration effect decreases with the Increase of X axis number.(3)The reliability and validity of the simulation results are verified by the comparison between MPI+Open MP hybrid programming technology and MPI parallel technology.Under the same conditions,the speedup of two parallel technologies are compared.The results show that the acceleration of MPI+Open MP mixed programming technology is higher than that of accelerated compared with pure MPI parallel technology under the same conditions.The following rules can be summed up: the number of threads is even when the number of processes is even and the number of threads is even or odd when the number of processes is odd through the analysis of optimal process number and the number of threads on MPI+Open MP.