Are swept meshes always better?

[KevinDeSmet]

I have a fairly simple question actually, that I'm sure will result in a not-so-simple answers for all analysis cases.
My question is the following: are swept meshes always better? In terms of analysis time, convergence and results accuracy.

A good example would be a hat stiffener beam which is being meshes with shell elements.
Does one prefer a random triangular mesh or does one prefer an ordered swept quad mesh that follows the general shape of the beam?

 

[rstupplebeen]

For rectangular or boxy geometry like you suggest quad and hex are almost always better.

The more difficult question arises with complex geometry like organic shapes. Here sometimes hex meshes are near impossible to achieve.

If using triangles or tets I usually run with second order. Infact Abaqus has changed this to the default.

In the end the mesh needs good aspect ratios and mesh densities where needed. I usually test several and pick the one with the best:

cost (time) vs reward (accuracy)

I hope this helps.

Rob Stupplebeen

http://www.optimaldevice.com

https://sites.google.com/site/robertkstupplebeen/

 

[rb1957]

all good advice from Rob ...

use quad meshs whenever you can.
avoid TET4s as you'd avoid TRIAs

another day in paradise, or is paradise one day closer ?

 

[KevinDeSmet]

I'm just suspicuous that a neatly swept quad mesh on a geometry such as a hat stiffener beam will artifically constrain the model?
In the case that the material is an isotropic one, such as steel, titanium or aluminum.

 

[rb1957]

if you use a TRIA3 mesh you might align the inherent element stiffness that might affect some of the results, but i don't see anything major (other than pave meshing with TRIA3s).

btw, i used TRIA3 intentionally ... you wouldn't mesh a sheet metal part with TETs, would you ?

another day in paradise, or is paradise one day closer ?

 

[inline6]

TRIA3 and TET4 are terrible elements