Research On The Calculation Method Of Three-dimensional Dendrite Growth Under Convection Based On LBM+OpenMP

Phase field method as a kind of computing technology, can be used to simulate the formation of the microstructure directly, is an effective tool for simulating complex solidification interface structure. With the continuous development of the phase field method, in order to more faithfully reproduce the solidification microstructure evolution process, it’s necessary to expand the simulation study of two-dimensional plane to the three-dimensional space, and in addition to the phase field and temperature field, the flow field and other fields also need to be introduced.Therefore, the complex models of multi-field coupling and the larger scales of simulation requirements has become a bottleneck in the study of computation. Serial program has been far can not solve computationally intensive and inefficient,small-scale simulation and other key issues. Therefore, to explore the reliable and efficient numerical methods is the key to the success of the subject. This paper researchs on the calculation method of three-dimensional dendrite growth under convection based on LBM+Open MP, and the main research work are as follows:(1) Coupling phase field model of dendrite growth with the LBM(Lattice Boltzmann method), a phase field-LBM model(PF-LBM) that can simulate 3D dendritic non-isothermal growth of undercooling pure substance melt in forced flow is established in this paper. Taking high-purity succinonitrile(SCN) as an example, the effect of flow on 3D dendrite growth is studied.(2) On multi-core processor platforms, the parallel computing method of flow affected three-dimensional dendritic growth is designed based on Open MP standard. It reduces the data communications. So that the problems of calculating multi-field coupling PF-LBM model can be solved.(3) The calculation efficiency between LBM and SOLA algorithm is compared under the same condition. The results show that, for the PF-LBM model, the effect of the forced flow on 3D dendrite growth morphology and the dendrite asymmetric growth can be quantitatively simulated in a large scale, and the dendrite growth law is consistent with the crystallization theory. Then, the LBM algorithm overcomes the weakness of large calculation amount and slow computation speed for the traditional algorithm, and it achieves a speedup of 11.7 when the calculation region size is 2003.(4) In the process of experiment, the calculation efficiency of serial algorithm and parallel algorithm based on Open MP for LBM-PF model application program wascompared. Results show that parallel algorithm of three-dimensional dendrite growth under convection based on Open MP framework is in an effective way of performance optimization on multi-core platform. Compared with the serial algorithm of LBM,after parallel optimization, it can take full advantage of multi-core CPU resources to improve the computational efficiency. The parallel speedup can reach 5.62 on the eight-core CPU.(5) The parallel computing performance under different processor cores was analyzed. Results show that the performance of parallel computing will not improve indefinitely with the unlimited increase of processor cores number. Instead, the efficiency of parallel execution is the highest when the number of CPU cores is relatively small.