Research On PF-LBM Model Of Coupled Pores Multi-field Coupling Based On OpenCL Parallel

Numerical simulation can reduce experimental costs and improve experimental safety factors and qualitatively and quantitatively research experiments.However,numerical simulation has problems that cannot be solved in a reasonable time.For this problem,GPU parallel computing can provide effective support.GPU parallel computing is used to accelerate the numerical simulation calculation,it can improve the efficiency of calculations and experiments,and multiple experiments can be completed in a limited time.The acquisition of high-quality metals mainly depends on the effective control of the solidification microstructure and defects of the metal.Among them,the microscopic pores defect is one of the main defects encountered during the alloy solidification process,which will greatly affects the mechanical properties of the material,and reduces the material's tensile strength,toughness,fatigue and so on.The formation of pores in the solidification microstructure involves complex gas-liquid-solid three-phase transition problems.As an effective numerical method for simulating fluid flow,the Lattice Boltzmann Method(LBM)has been proven to be effective simulated complex multiphase flow systems,and are inherent suitable for parallel computing.For the dendrite growth in the solidification microstructure,the phase field method(PF)is currently one of the most promising numerical simulation methods in the solidification structure simulation.It has a clear physical basis,and the simulation results are reliable.There is no need to track the solid-liquid interface,and external fields such as flow field,electric field,and magnetic field can be easily introduced into the phase field control equation.In this paper,the PF model that simulates dendrite growth and the LBM model that simulates flow field are coupled to establish a coupled pore multi-field coupling PF-LBM model.The PF-LBM model has the problems of small simulation scale,low computational efficiency,and unable to complete the solution within a reasonable time.In this paper,the parallelized solution of the established PF-LBM model is realized on the hybrid heterogeneous cluster platform of CPU+single GPU based on OpenCL parallel programming.Further optimize the PF-LBM model,make full use of the CPU + GPU hybrid heterogeneous system to achieve acceleration and improve the computational efficiency.Under the same conditions,the OpenCL-based CPU + GPU heterogeneous system model is up to 62 times faster than the CPU system model.In order to verify the validity of the model,the results of the OpenCL-based CPU + GPU heterogeneous model are compared with the classical theory to verify the correctness of the model,and the solution time is compared with the solution time of the result of the CPU serial program.The comparison verifies the effectiveness of the heterogeneous system.On the OpenCL-based CPU + GPU heterogeneous system,the morphology evolution process of binary alloy multi-dendritic growth and pores formation under isothermal conditions is simulated and solved.The simulation scale has been expanded and the solution speed has been greatly accelerated,and the formation of pores caused by the preferential growth of dendrites,bubbles nucleation,bubbles growth,bubbles coalition,and the interaction of bubbles and dendrites have been discussed and analyzed.The simulation results show that the growth of bubbles will affect the morphology of dendrites.Conversely,the growth of dendrites will also affect the growth and morphology evolution of bubbles.The interaction between dendrites growth and bubbles will lead to the formation of pores.When the length of the simulated region of secondary dendrite arm growth is too short or too long,bubbles are likely to escape.The length of the simulated region where the secondary dendrite arm grows is just enough when the secondary dendrite arm grows to the top of the right side,it will be difficult for bubbles to escape.