Unstructured Mesh Generation Using Automatic Point Insertion and Local Refinement

Abstract

This paper is concerned with designing efficient algorithms for generating high-quality two-dimensional unstructured mesh. Because the most vital factor influencing the quality of mesh generations is point placement, we first propose a background quadtree to represent the density distribution within the computing do main that allows us to place well-spaced points. We then adopt the Bowyer-Watson Delaunay triangulation algorithm along with Steiner point insertion and local refinements. We identify constraints for placing new points that allow us to improve mesh quality incrementally. We also give procedures for performing local refinements which allow us to reduce obtuse angles and thus improve the mesh quality by reducing the aspect ration and area ratio. A running example dealing with a blunt body in some flow domain is also presented to illustrate our method. Then we implement these algorithms to a three elements airfoil configuration.