An accelerated optimization method for parallel generation of adaptive Cartesian grids
Abstract
<p indent="0mm">The efficiency and quality of computational mesh generation directly determine the quality of numerical simulation results. The Cartesian mesh has received widespread attention due to its good adaptability, high degree of automation, and strong ability to handle complex shapes. However, for complex shapes, the extremely large mesh volume makes it difficult to achieve satisfactory mesh generation efficiency. Therefore, it is necessary to develop 3D Cartesian mesh parallel generation technology that is versatile, efficient, and highly scalable. In this paper, a 3D Cartesian mesh parallel generation optimization method is proposed based on the KD tree (K-Dimensional Tree). First, based on the geometric information of complex shapes in 3D flow problems, a set of surface triangles is used as input to generate a bounding box enclosing the shape, which is then used to construct the KD tree. Second, an accelerated optimization algorithm for wall distance calculation and intersecting cell judgment is developed based on the constructed KD tree, along with an algorithm that accelerates the classification of inner and outer cells by performing a lookup on the projected points in the KD tree. Finally, the effectiveness of the adaptive Cartesian mesh parallel generation algorithm proposed in this paper is verified by using models of a sphere, cube, M6 wing, and Su-27 aircraft. The numerical results show that the optimization method for intersection judgment and wall distance calculation proposed in this paper can improve efficiency by two orders of magnitude compared to the traditional method, while the efficiency of inner and outer cell classification can be improved by three orders of magnitude. This fully demonstrates that the proposed optimization method can generate adaptive Cartesian meshes required for flow field calculations in a fast, robust, and automatic manner.</p>
