Grid-Quality Measures and Error Estimates

Numerical Grid Generation in Computational Field Simulations, The International Society of Grid Generation, pp.799-808, September 2000

MESHING
RESEARCH
CORNER

Derlon Chu
Powertrain Operation Engine Engineering
Ford Motor Company
dchu@ford.com

Abstract

In CFD, the quality of the mesh is known to affect significantly the accuracy of the computed solutions. In this paper, four new grid-quality measures are presented that can be used to evaluate both structured and unstructured meshes. These measures extract information from the computed solutions as scalar and as vector fields and link them to the mesh geometry. The usefulness of these measures was evaluated by examining how well they correlate to errors in solutions computed on less than ideal grids. This is accomplished by establishing and interrogating a database made up of two steady, incompressible, laminar flow problems that have complicated 2D and 3D flow features and a series of solutions generated on a variety of meshes from "poor" to "good," including grid-independent ones. Results obtained show the grid-quality measures developed to correlate reasonably well with relative errors and change in relative errors, defined relative to the grid-independent solutions.