Automated domain decomposition for computational fluid dynamics

Abstract Automation of flow-field zoning in two-dimensions is an important step towards easing the three-dimensional grid generation bottleneck in computational fluid dynamics. A knowledge-based approach works well, but several aspects of flow-field zoning make the use of such an approach challenging. A proposed model and language to describe the process of zoning a flow field are presented, followed by a discussion of the implementation of EZGrid, a knowledge-based two-dimensional (2-D) flow-field zoner. Results are shown for representative 2-D aerodynamic configurations. Finally, an approach to the evaluation of flow-field zonings is described and used to compare the performance of EZGrid with that of a human expert.