Adaptive Mesh Refinement for Parallel Finite Element Simulations

Benjamin Kirk, University of Texas

This poster will present some recent developments in the application of adaptive mesh technology to parallel finite element simulations. Adaptive mesh refinement (AMR) has the promise of evolving an optimal grid for a particular application rather than putting the burden on the analysts a-priori. AMR techniques have been successfully applied in many areas in recent years, but much of the success has been limited to serial applications. AMR is complicated in parallel by the inevitable load-balance and partitioning problems. This poster will describe some of the techniques that have been recently developed for efficiently handling AMR on general, unstructured grids in 3D. Data structures and partitioning algorithms will be discussed, and results from a number of transport applications will be presented.

Abstract Author(s): Benjamin S. Kirk