Parallel Unstructured Grid Generation And Numerical Computation

The study on how to quickly generate unstructured grids is of great significance for the fluid flow calculation. In order to save the time of unstructured mesh generation, the main work of this paper is to develop the technique of parallel unstructured grid generation. With the effective parallel computation algorithm, the simulation of complex inviscid flow is finally achieved more efficiently and quickly.In this paper, we firstly improve the R .Lohner's wavefront domain-splitting algorithm so that the subgrids and their boundaries are more favorable for grid generation. A new optimization strategy of subdomain's boundary is then presented in order to improve the smoothness of boundaries and the quality of grids. After subdividing the initial meshes by using the above domain-splitting algorithm, we smooth the grid by the parallel method. In this phase, smoothing "subjunctive" boundary points needs mutual communication of subdomains . Next, we also improve the conditions in the paper [1], [13] of receiving new points and elements in the course of grid generation in the subdomain and present a new strategy of receiving new elements only and refusing new points during the course of grid generation of the interface, which can spare much time. The results obtained on the PVM environment demonstrate high effectiveness of the algorithms.Based on the above algorithm of parallel unstructured grid generation, a parallel computing method is applied to a 2-D Euler solver for transonic and supersonic flow on a PVM parallel environment. Jameson finite volume scheme and four-stage Rung-Kutta time-stepping method are employed in the process of the zonal parallel computation of Euler equations. The "wavefront" domain decomposition method developed above is used in the division of flow field domain. The quantity of physics variables on the "subjunctive" boundary elements needs mutual communication of subdomains during the simulation. The results of numerical experiments, the statistics of speedup ratio and parallel efficiency all show the success of parallel unstructured grid generation method and parallel computation algorithm.