Particle Trajectory Parallel Model

PFS (Particle Fluid System) has been widely utilized in many sciences, such as chemical and ecological engineering. As a powerful simulation technology on PFS, the investigation of DPM (Discrete Particle Model) has been a foreland in this domain for reasonable rationale. However, the calculation of discrete particle model has a huge restriction from present PC running speed and memory capacity due to the complexity of PFS, and it has confined to understand more internal mechanics of PFS. In recent years, along with the rapid growth of high performance parallel computing, parallel discrete particle model has a requirement for complex PFS simulation both in application and in theory.In discrete particle model, the computation of fluid is very important as a carrier of particle. With the direction of PCAM, a methodology for parallel algorithms designing, parallel SIMPLER and parallel PISO algorithms are developed in collocated grid based on domain decomposition method. At the same time, parallel time decomposition algorithm is introduced to parallel fluid computation based on time decomposition method. The detailed rules about domain partition and data exchange are also presented. In order to improve employment efficiency of parallel machine, adaptive relaxation factors are used to achieve load balance between sub-domains. On this condition, after analyzing some key elements such as data structure and message passing, parallel discrete particle model is established by separating PFS to some related fluid sub-domains and particle sub-stacks. Computation performance and parallel performance are testified by computing the driven flow in a square cavity and convective-diffusion equations on PC cluster. Bubble fluidization system is simulated to verify the parallel efficiency and expansibility of parallel discrete particle model.The numerical simulation demonstrates that parallel fluid algorithms and parallel discrete particle model can obtain coincidence result with theoretical or experimental result. Especially, well speed-up and expansibility can be acquired by parallel discrete particle model.