A knowledge-based approach to automated flow-field zoning for computational fluid dynamics

Flow-field zoning is becoming popular in computational fluid dynamics (CFD) as a means of alleviating the three-dimensional (3-D) grid-generation bottleneck caused by complex geometries, the need for selective grid refinement, and computer memory and speed limitations. The goals of this study were: (1) to lay the foundation for an automated 3-D zonal grid-generation capability for CFD through the development of a demonstration computer program capable of automatically zoning the flow field of representative 2-D aerodynamic configurations, and(2) to determine the applicability of a knowledge-based programming approach to the domain of flow-field zoning. Several aspects of flow-field zoning make the application of knowledge-based techniques challenging: the need for perceptual information, the role of individual bias in the design and evaluation of zonings, and the fact that the zoning process has been modeled as a constructive, design-type task (for which there are relatively few examples of successful knowledge-based systems in any domain). Engineering solutions to the problems arising from these aspects were developed, and a demonstration system was implemented that can design, generate, and output flow-field zonings for representative 2-D aerodynamic configurations