Research And Application Of SPH Parallel Algorithm Based On Particle Decomposition

The Smoothed Particle Hydrodynamics(SPH)method is a Lagrangian meshless method,which was first proposed by Lucy and Gingold et al.in 1977.Compared with grid-based numerical methods,the SPH method is completely grid-independent and has the advantages of Lagrangian properties,mass point properties and adaptive properties.Therefore,it is very suitable for numerical simulation of complex interface problems such as large deformation and free surface flow.However,the SPH method suffers from the problems of large computational volume and time consuming when dealing with free surface flow problems,so it is necessary to parallelize it.The details in this thesis are as follows:(1)In this thesis,certain numerical techniques of the SPH method are compared and improved to resolve the problem of low computational efficiency of the SPH method when dealing with free surface flow problems.First,a linked list search algorithm is developed to search the neighboring particles.Through numerical simulations of the two-dimensional droplet impinging on the solid wall surface problem,the results indicate that the SPH method based on the chain table search method is more efficient.Second,an enhanced boundary treatment method coupled by solid wall particles and imaginary particles outside the solid wall boundary is proposed.The results indicate that this method can not only prevent the fluid particles near the solid wall boundary from penetrating the solid wall surface,but also significantly reduce the computational time consumed by the SPH numerical simulation.(2)In this thesis,the SPH parallel algorithm based on particle decomposition is proposed to resolve the problems of large computation and time consuming of SPH method.The basic idea of this parallel algorithm is to distribute all particles equally to each process for computation,and each time-step communication only calls the send,receive and broadcast functions once.Therefore,the programming is easy to implement and has a good scalability.The parallel algorithm is applied to simulate the two-dimensional dam-break problem and the two-dimensional droplet-impacting solid wall surface problem,and the results indicate that the particle decomposition-based SPH parallel algorithm proposed in this thesis can not only effectively simulate the free surface flow problems,but also reduce the computational time of the SPH simulation on a large scale.(3)In this thesis,the enhanced boundary treatment method proposed in Chapter 3 and the parallel algorithm proposed in Chapter 4 are extended and applied to the numerical simulation of three-dimensional problems to solve the problem that the numerical simulation of SPH method is limited to two-dimensional space.The relationships between the number of processes,the number of particles,the parallel efficiency and the acceleration ratio are analyzed for the three-dimensional dam-break problem and the three-dimensional droplet impact liquid film problem,respectively.The results indicate that the enhanced boundary treatment method and the particle decomposition-based parallel algorithm proposed in this thesis for the SPH method can significantly reduce the computational time consumed by the numerical simulation when dealing with large-scale free surface flow problems.This thesis is based on the Message Passing Interface(MPI)parallel programming platform.The SPH parallel algorithm based on particle decomposition is designed using C++as the programming language for algorithm implementation.The parallel algorithm is applied to numerical simulations of typical two-dimensional and three-dimensional dam-break and droplet problems,and the relationship between the number of processes,the number of particles,the parallel efficiency and the acceleration ratio is analyzed.The results reveal that the maximum acceleration ratio can be more than 30 when the number of particles is greater than one million.Therefore,the SPH parallel algorithm based on particle decomposition proposed in this thesis can provide an efficient and accurate computational tool for the numerical simulation of large-scale three-dimensional problems.