[kingkongjaffa]

Basically 0 CAD models you see with color coding and a mesh are actually accurate.

In order to mesh the geometry for finite element analysis, the geometry virtually always needs to be defeatured.

So the cross sectional CAD model here is a nice curiosity but basically useless for any reverse engineering purposes which is the key reason this stuff is kept secret.

[weinzierl]

In Germany we say "The DIN knows non color". DIN is our standardization organization and informally also how their documents are called.

I did finite element model preparation for a living many year ago and it did not only involve heavy defeaturing but interestingly also remeshing with quads.

Renderers love triangles, FE solvers love boring quads.

[krisoft]

> Renderers love triangles, FE solvers love boring quads.

Btw even in Blender (which is pure visual rendering) people prefer quads. The common wisdom is that you should keep your topology quads with nice rectangle-ish aspect ratios if you can at all. It is not that triangles don't work, but they have a tendency to do visually unpleasant silly things when animated or sculpted or subdivided.

[USiBqidmOOkAqRb]

The "quads only" rationalizations come off as quite cargo-culty.

Sure, edge rings won't have dead ends, and that's useful when adding edge loops to increase detail, but doesn't necessarily mean the topology is of high quality. Using only quads some cursed helix type topology can be constructed.

First thing subdivision will do on non-quads is ortho(1) operator, so a nice averaged vertex in the middle will be added. High side count cylinders will have weird saddles around caps, but that's due to subdivision being wrong tool for the task.

Quads can be abused to look bad too. If a mesh looks bad, it's a bad mesh. If animated mesh looks bad, it's bad rigging and skinning. Knowing "this one industry secret" won't fix those.

1: https://en.wikipedia.org/wiki/Conway_polyhedron_notation#Ori...

[cherryteastain]

Most 3D meshes will have a combination of hexahedral and tetrahedral elements anyway so the surface will be a combination of trangular and quadrilateral elements. Accuracy and convergence wise, it doesn't matter as much as polynomial order/time step size/element size.

[albrewer]

> FE solvers love boring quads

That's because, in a mathematical sense, triangular and tetrahedral meshes aren't able to be as accurate as quickly.

[kingkongjaffa]

yeah there are several adaptive meshing techniques that use a mixture nowadays, imagine a bar with a hole bored through it, you might use quad meshes for the majority of the bar and then switch to tetrahedral meshes close to the bore hole to better model the curved geometry of the hole, and increase the node density in high stress concentration regions for a more accurate simulation.

[thrwooshfem]

Sandia FEM is using the different blocks (colors) to represent different materials. This is pretty common in a multi physics finite element program.

This story is probably nothing interesting because this went through all the public use approvals needed for public presentations and being available on osti.gov.

It is probably just a toy test problem used on a capabilities logo for Sierra. Maybe it comes from some sort of integration test that is easier to run than the actual problem.