FE Topologies and Reliabilities

[Spirit]

Hi pals, here is my problem.
We are considering modeling a complicated structure (different thicknesses) with solid elements (NASTRAN solver). We would be oriented towards bricks, but the difficulties in meshing our geometry in this fashion seem to be too big. We are thus considering using parabolic tetrahedra, but we have no experience on the reliability of these elements in NASTRAN or other codes.
Any of you can give any references/suggestions? An indication on an automatic brick mesher would be greatly appreciated as well... Thanks in advance.

 

[brad]

I know of no really useful (truly) automated brick meshers. Good meshing can often be done with bricks in preprocessors, but it's not automeshing to the degree that can be done for tets.

I have done a fair amount of study into the problem which you are trying to resolve--to use bricks or parabolic tets. Unfortunately, I cannot provide you the details of the study (as it was with a former employer). However, the conclusion was that an equivalent mesh of 2nd order tets gave BETTER answers than first order bricks. By equivalent mesh, I mean that the characteristic element length was the same. Unfortunately, to get an equivalent characteristic element length, one has to have a larger number of dofs for 2nd order tets. If your machines can stomach the computational hit, go with the tets. The engineering time can literally drop by an order of magnitude (we saw modeling time go from 40 hours to 4 hours for complex structures when we compared engineering time).

When we conducted this study, my employer was utilizing first order hexahedrals for analyses; today (due to our conclusions) first order hexahedrals are almost never utilized at that company because of the human overhead (which far outweighs the computational overhead hit).

FYI--Nastran was one of the codes which we examined, so these results are certainly applicable to you.

Details on some of these things--possibly do a search on NAFEMS (

http://www.nafems.org).

NAFEMS is a British governmental agency, the National Agency for Finite Element Methods and Standards. Don't just look on their website; search on the internet for references to them. The reason for this is that NAFEMS has extensive benchmarks which are used by many manufacturers in testing their element formulations. You may be able to find some discussions that are publicly available regarding this type of study.

Brad

 

[Spirit]

Hello bradh, thanks for answering.
My actual experience on the topic corresponds to what you have reported about bricks and 2nd order tets, I was just looking for some other assessment performed on different structures, so here you come!
Thanks again.