Comparison of 2D unstructured grid generation techniques

Abstract

Description of Techniques Two-dimensional unstructured grid generation techQuadtree Methods niques are evaluated based on the criteria of grid qualThe encoding technique involves recursively ity, efficiency, robustness, maintainability, and ease of subdividing existing quadrilateral cells until the desired extension to three dimensions. a description of the methods investigated, including tion of the technique includes the following tasks, quadtree, advancing front, and Delaunay schemes. Further, grids generated about three configurations, a crew escape capsule, a multi-element airfoil, and a hypersonic aircraft/store configuration are presented and used to evaluate the methods. The Delaunay sweepline method proved to he the most efficient and effective technique based on the present evaluation. The study includes level of refinement is achieved, A typical implements1. Enclose the entire domain of interest with a quadrilateral cell. 2. Establish criteria for controlling the subdivision. 3. Test each quadrilateral cell to determine if it is to be subdivided. If the quad requires refinement, it is divided into either two or four quads depending on the character of the subdivision function. Introduction Unstructured grid generation offers the potential to automatically discretize extremely complex domains. The following study examines three classes of 2-D unstructured grid generation schemes: quadtree, advancing front, and Delaunay techniques in terms of the resulting grid quality and their efficiency, robustness, maintainability, and ease of extension to 3-D. First, each unstructured grid generation technique is described. Next, grids about the three configurations are presented. And finally, the primary characteristics of each method are discussed. *Engineer Senior, CFD Group. Member AIAA tconsultsnt, Member AIAA $Graduate Student, Member AIAA 4. Repeat the process until the desired length scale is resolved in each region of the domain. The primary disadvantage of directly applying the previous process is the lack of boundary conformity in the final decomposition, Le. nonvertical or nonhorizontal boundaries are represented with a series of reentrant corners. One method, proposed by Yerry and ShephardJl] modifies the procedure to accept shapes produced by trimming the corners of quad elements. Further, producing the final mesh requires additional logic to create adequate transitions between quadrants a t different levels of subdivision, ensure reasonable cell aspect ratios, provide control over node density, and force the boundary nodes to conform exactly to the boundaries. The present study does not use the quadtree encoding procedure to produce the final triangulations; however,