Modeling data distribution for two-phase flow problems by weighted graphs.

Abstract

Two-phase flow problems occur in a wide variety of engineering problems, for instance when an oil drop interacts in surrounding water. Efficient numerical simulations of such complex phenomena typically involve adaptive techniques and parallel computing. The distribution of data among processes is formulated by a graph partitioning problem. We introduce edge and node weights specifically targeted toward a parallel finite element solver for two-phase flow problems in three space dimensions. Numerical results using up to 256 processes are reported comparing different node weighting strategies.