The Research And Implementation Of The Parallelization Of ELBM Algorithm

As the basic subject for studying of fluid motion,fluid mechanics has three mathematical models: microscopic molecular model,mesoscopic kinetic model and macroscopic continuous model.Mesoscopic model not only has the microscopic model assumptions fewer features,but also has the advantage of macroeconomic models which does not care about the details of the molecular motion,so mesoscopic model is widely used in processing of multi-scale and multi-physics complex flow problems.In the numerical simulation methods of mesoscopic model,lattice gas cellular automata exists statistical noise,complex collision operator and does not meet the Galilean invariance and other shortcomings;lattice Boltzmann method overcomes the shortcomings of lattice gas cellular automata,but exists numerical instability problems in high Reynold;entropic lattice Boltzmann method inherits the advantages of lattice Boltzmann method,possesses a clear physical process,simple calculation,easy programming,good locality,the advantages of good parallelism and scalability,etc.,and is suitable for calculation of large-scale flow problem,which has overcome the numerical instability problem under the limit of the monotonicity and the minimality of Boltzmann H function and has a wider range of applications.The evolution of the entropic lattice Boltzmann method includes the solving of relaxation parameters,collision and propagation process,the calculation is more intensive and has a higher demand on computer performance.Because of bottlenecks of single-core performance,parallel computing has become the main way of improving computing performance.Parallel computers has go through the development that from a single nuclear to multi-core,multi-processor and further to the coprocessor,programming model also comes from Open Mp,MPI and to CUDA,parallel computing has become an important means to reduce the scientific computing time and expand simulation scale.In this paper,the basic principle of entropic lattice Boltzmann method and parallel computing has been learn ed deeply,according to the serial entropic Lattice Boltzmann method,parallel algorithms which are based on Multi-core and GPU are designed,and the numerical simulation on the lid-driven cavity flow using the parallel algorithms has been done,the main work of this paper is as follows:1)Considering the serial program of the entropic lattice Boltzmann method which takes long time in simulation and has low computational efficiency,the parallel algorithms of the entropic lattice Boltzmann method based on Multi-core and GPU are designed with the characteristics of the Open Mp and GPU programming model.2)According to the characteristics of the lid-driven cavity flow,the parallel algorithms based on Multi-core and GPU were implemented on the Dawning server W580,the performance of the parallel algorithms is analyzed and compared under the the different mesh size,different number of iterations and different types of precision,experimental results show that the parallel algorithms have good parallel speedup and scalability,expanding the simulation scale of the entropic lattice Boltzmann method.3)We use Tecplot software to visualize the lid-driven cavity flow simulation results in post-processing,respectively,drawing up the flow velocity vector and the fluid streamlines under different mesh size and different Reynolds,and also the speed graph through the geometric center axis,the results show that the simulation results are in accordance with theoretical analysis,simulation results with Fluent software were compared to verify the correctness of the model.