Criteria for comparison of automatic mesh generation methods

Many different schemes have been proposed for automatic mesh generation. Some of these are described in this issue. In order to consider their relative merits it is necessary to identify the criteria which are of importance to us.

This paper does not carry out the comparison of methods, but lays the foundation for such a comparison by suggesting nine criteria, based on the view of the meshing process as a function from geometry to mesh. The first questions emerging from such an approach, the domain and range, deal with the universe of shapes, plus the other data necessary, and with the meshes generated. The other questions, of locality, continuity, orientation independence, representation-independence, idempotence, and of local density control, are of particular relevance in the context of analysis being used as a design tool.