A parallel unstructured mesh adaptation for unsteady compressible flow simulations

A simple and efficient implementation of adaptive mesh refinement (AMR) for distributed-memory parallel computers will be presented. The proposed method involves relatively low complexity in programming and leads to acceptable reduction in computing time. Introducing a lookup table including grid connectivity information and simplified algorithms for AMR, the procedures for reconstructing adaptive grids are carried out in parallel with local data to a large extent. A simple static load balancing scheme was adopted, and the grids are not repartitioned and no data redistribution is performed. A numerical example shows that the proposed implementation of AMR is effective to reduce the computing time for unsteady compressible flow problems on a distributed-memory system.