Questions on CTETRA

[Anonymous]

Hello,

I generally use hex elements wherever possible.  However, I have a project right now where I have an assembly of  35 or 40 parts, many of which are complex.  Simplifying the CAD solids and then slicing them up in Femap is taking far more time and money than my Client has.  So, I would like to drop back to CTETRAs on some of the parts, which don't need the slicing and usually mesh pretty easily.

Not having much recent experience with CTETRAs, here are my questions:

1.  Years ago, many formulations of tet elements with midside nodes had problems when used on contact surfaces. Is this still a problem with the current 10 node CTETRA?  I see that Ansys and some others no longer have this problem, but can't find the info on the Nei element.

2.  4 node tet elements used to be far too stiff and severely underpredicted both deflection and stress. As such, they were a good thing to avoid.  Is the current formulation of the 4 node CTETRA still problemmatic?  Are there any studies of mesh density v. accuracy for the Nei element compared to theoretical?

3.  Is the Nei formulation of the tet a constant stress element?  I note that Ansys quite a while ago went to a formulation for the tet element that is essentially a degenerate brick, and offers improved performance.

4.  Can anyone offer any Nei-specific CTETRA guidance regarding the tradeoff between higher density of 4 node CTETRAs v. 10 node formulation?  The node count gets out of hand quickly with a tet mesh.

Test cases can answer the above, but that would be a project in itself.   I'm hoping for some guidance to save time.

Thanks!

Andrew   

[Andrew.Sartorelli · Accepted solution]

Hi Andrew,

1. I'm not aware of any issues with mid-side nodes on contacts surfaces in Autodesk Nastran.

2. Yes, 4 node tets are inherently stiffer than hexes or 10 node tets. When working with a tet mesh I always recommend using mid-side nodes. In Nastran In-CAD we only have tet meshing for solid elements and we get very accurate results. As always, one should perform a mesh refinement study to determine if one has converged on a result.

3. I'm unsure of this, but I will check with the development team.

4. Personally, I always recommend to customers that they should be using parabolic 10-node elements rather than linear 4-node elements.

[Anonymous]

Hi Andrew,

Thanks for your reply,

I ran some quick tests, and my conclusion is that 10 node tets are still the only tets to use.  The node count gets high in a hurry, though.

Autodesk should try to come up with a hex mesher that doesn't require slicing, dicing and massaging to come up with a decent mesh.  That would be a breakthrough.  The Femap hex mesher works, but it's a massive time sink for any part that isn't, more or less, dirt simple.

Best regards,

Andrew