| High performance computing has penetrated into various research fields as an effective and reliable computational technique,and is now widely used in materials engineering.The production of high-quality metals depends on the effective control of microstructure and defects during solidification,which is influenced by a variety of factors and can be studied quantitatively by adopting numerical simulations that can significantly reduce the cost required for traditional experiments.The solidification of alloys involves a complex transformation between gas/liquid/solid phases,and porosity is one of the main factors causing solidification defects.The phase field method(PF)is applied to the numerical simulation of solidification dynamics and fluid dynamics without showing the tracking distinction interface and easily combining other external fields.Lattice Boltzmann method(LBM)can effectively simulate complex and variable fluid flow and is applied to multiphase flow systems such as gases and liquids.Therefore,this paper combines the phase field and magnetic field to establish a kinetic model to study bubbles,coupling the PF model and LBM model to establish a Shan-Chen multiphase flow coupled stomatal PF-LBM model,simulating the motion of bubbles in the fluid under the influence of magnetic field and the change of dendrite and bubble motion during the solidification of the alloy.The boundary of the LBM method is easy to handle,which is very suitable for parallel computing,for the problems of large computation and low computational efficiency of the coupled PF-LBM model,the OpenCL parallel programming framework is used to realize the parallel solution of the model.The main research contents and conclusions are as follows:(1)The bubble dynamics model is developed by coupling the continuity equation,momentum equation,phase field equation and magnetic field equation that control the fluid flow,the rising process and morphological changes of bubbles in a fluid under the influence of vertical and horizontal magnetic fields are studied.The results show that the rise of the bubble is accelerated after the application of a vertical magnetic field and decelerated after the addition of a horizontal magnetic field,the bubble is stretched along the direction of the magnetic field lines,and the degree of deformation is proportional to the strength of the magnetic scalar.(2)The evolution of the microstructure of dendrites is simulated using a binary alloy model based on the PF method,and the effects of anisotropy and solute concentration on solidifying dendrites are investigated,the correctness of the motion of bubbles in the liquid phase based on Shan-Chen multiphase flow of LBM is validated,the feasibility of the coupled PF-LBM model is verified by the gas/liquid/solid of threephase interface wetting problem.The coupled growth of dendrites and bubbles in Al-4.0wt% Cu microstructure is simulated,and the natural phenomenon of bubbles from nucleation to growth to merging is shown to be a synergistic competition between bubbles and dendrites.(3)OpenCL-based heterogeneous programming platform is used to solve and compute the coupled PF-LBM model.The model is further optimized by changing the data storage method,designing synchronous operations and removing data dependencies to fully utilize the computing power of CPU+GPU.Compare the solution time of OpenCL-based heterogeneous platform and CPU serial computing system,and analyze the parallel computing results under different conditions and models.The results show that the computation time for parallel solving under the OpenCL-based CPU+GPU heterogeneous platform is significantly reduced,reaching a speedup ratio of up to 20 times for the simulation area of 1000×1500 in the oriented dendrite and up to 22 times for the simulation area of 1500×1500 in the isometric dendrite compared to the CPU computing platform.The comparison between solution times verifies the effectiveness of accelerated parallelism in heterogeneous systems. |
PDF Full Text Request