Abstract
The overset grid method can effectively solve the complex flow problem of large-scale relative motion, but it is difficult to guarantee efficiency and accuracy of the overset grids with over 10 million scale and dozens of components. Based on a hybrid overset grid system that supports unstructured, Cartesian, and polyhedral meshes, an automated and MPI process-level parallel overset grid assembly method is developed in this paper. Combining rigid body motion and Unsteady Reynolds-Averaged Navier-Stokes (URANS) method, the method is suitable for simulating large-scale relative motion problems. The assembly procedure of overset grid includes four steps: hole-cutting, overlap-optimization, donor-searching, and interpolation. The process of hole-cutting combines the local hole mapping method and the direct cut method to quickly predict the overlapping relationship between different sub-grids, then realizes accurate hole-cutting; In the process of overlap-optimization, the wall/hole surface distance method is innovatively used to reduce the grid overlapping area, avoiding a large number of donor-searching operations in the traditional wall distance method; The donor-searching process uses the inverse mapping method and stencil walking method to quickly determine the donor, even across the boundaries such as partition boundaries and physical boundaries. This article uses WPFS store separation case and UH60 rotor hover case to verify the reliability and parallel performance of the developed overset grid assembly method. The results show that the assembly method is suitable for dynamic simulation of complex flow fields, and the parallel performance of overset grid assembly can still meet engineering requirements under extreme load imbalance conditions.
Keywords
Get full access to this article
View all access options for this article.
