Large Scale Parallel Computing Of Particle Laden Shear Flows By The Direct Numerical Simulation

With the development of the large scale high performance parallel computers on the hardware and software, it provides more opportunities on the direct numerical simulation of particle laden flows. In order to further investigate the problems facing for the international scientists, such as particle modulation, preferential concentration, and interactions between particles and fluid on mass and momentum transfer, the author conducts a theoretical research on the parallel algorithms for the particle laden flows. With the development of the parallel computing technologies, more accurate simulating methods will be proposed in future, and the information of the intricate mechanism in particle laden multiphase flow will be more obvious in future. In addition, it will significantly dependent on the large scale high performance parallel computing. So it is vital important to further conduct some research on the large scale parallel algorithms for the particle laden flows.The most urgent problem for the large scale parallel computing of particle laden flows is the parallel skills used for the particles. Working for about one year, the author successfully conducts a large scale parallel computing for the particle laden jet flow by the use of Eulerian-Lagrangian method. The information for the particles which cross the boundary of the computational domain is sent by the MPI. Two-way interactions are considered in this parallel computing, and the particles collisions are solved by the hard-sphere model in the parallel computing. The total number of particles used in our simulation is about20million without the idea of computational particles. The parallel computing code programmed by the author is strictly validated by the inspecting of the parallel efficiency and simulation results are compared with experimental data.The author conducts some research on the three dimensional non-reflecting boundary conditions based on the recent development in this field. It takes about three years for this research. The explicit derivations for the three dimensional non-reflecting boundary conditions are presented in this thesis, and the treating methods for the face, edge and corner are also given in order to give a clear understanding on how to use the theory of non-reflecting boundary conditions for practical simulation. The author has done several numerical tests in order to get the convergent simulation results. In addition, a large scale parallel computing with500million grids for the round jet flow is performed at the Shanghai Supercomputer Center (SSC). The three dimensional non-reflecting boundary conditions used in our numerical simulation are considered to be more stable on the point of numerical simulation.The author also conducts some research on the particle modulation on the fluid vortex structures and turbulence properties based on the above mentioned codes. In order to properly use the point force method for the Eularian-Lagrangian simulation of particle laden flows, the author carefully inspect the Kolmogorov length scale of the carrier fluid, and the ratio of the diameter of particles to the Kolmogorov length scale. The simulations results are compared extensively with experimental data. The information on the small scale modulation by the particles will be very useful on the further research on the intricate mechanism of particle modulation considering the difficulty of experimental measurement.The author has done some research on the round jet flow by the DNS based on the above mentioned code. The main purpose of this research is to investigate the entrainment, the evolution of turbulent properties, kinetic energy budget and Reynolds stress budget. This simulation is computational extensive with a total of500million grids, from the core region to the far-field self-similar region. The turbulence properties obtained by the author will give valuable information on the intricate mechanism of round jet development.