Creating high quality meshes is an ongoing challenge. The problem is complicated by the fact that there many measures of this quality. For instance, FEA preprocessors have for quite some time now been able to produce meshes composed of very regular elements – quads that are close to squares, triangles that are close to equilateral, etc. However, these meshes typically look like this:
For architectural applications this is not great, primarily because the flow of the mesh edges is highly irregular. This affects structural considerations (i.e. where to place the structural support), and also the visual aesthetics. So, I was playing around with a very large mesh composed of these regular elements and messy edges. I found that its dual (i.e. the mesh defined by connecting the centres of adjacent faces), looks much more convincing:
Image produced in collaboration with LOH ARCHITECTS Associates Inc. 2012
This vast difference in the needs of FEA meshes on the one hand, and architectural meshes on the other begs the question of how to synthesize the two approaches. A mesh that at one scale defines panel and structural layout, and at a much smaller scale defines FEA elements could have implications on the growing problem of managing distinct models during the design process.