Solving computational fluid dynamics problems on unstructured grids with distributed parallel processing

Abstract

This paper describes a distributed parallel implementation of a finite element method for simulating incompressible viscous flows. When applied to unstructured finite element meshes, a conventional implementation of this algorithm becomes particularly space demanding for the storage of system matrices. This parallel implementation not only results in a faster computation but is also potentially capable of solving much larger problems as data and system matrix blocks are distributed. Experimental results demonstrating these key features are presented.