Hex vs Tets

[spacecase]

I am a newbie analyst looking to gather the opinion of this great forum. When using solid elements I have been mainly sticking with hex elements, but with some added complexities to my current project’s design I have been leaning towards using tets with midside nodes. What is everyone’s take on hex vs tets? Are there some general issues to watch out for when using tets?

Thank you in advance!

 

[Stringmaker]

Spacecase,
I'm an intermediate level analyst so the advice I offer may be trumped or further refined by the many here which have more experience than I.

One thing about tet meshes is that you will typically end up with a model that has significantly more DOF's than you would with a hand built hex mesh. This is something that you'll want to keep in mind when you have a project which may push the limits of your computational resources (whatever they may be).

There is no reason why second order tets shouldn't produce what you consider to be acceptable results with the appropriate mesh. As with anything new you will want to err on the side of caution, test analytical solutions against experimental and theoretical solutions, and refine your mesh to the point where you see a mesh convergence in your models. Over time you'll gain the first hand knowledge of what is or isn't a satisfactory mesh for your application.

Hope this helps,
-Brian

 

[johnhors]

You will use tet elements to mesh complex geometry imported from CAD that is impractical and too time consuming to mesh with hex elements. CAD geometry in all probability will have many hidden "nasties" that are not immediately visible or obvious, by nasties I mean sliver faces and faces with very acute angles. If this geometry is meshed "as is" then you will generate elements with very poor shape quality. Increasing mesh density can improve element quality on sliver faces but no amount of mesh refinement can stop bad elements being created at a sharp angle. However refining the mesh on every sliver face can very quickly lead to a model being generated of colossal proportions. This is probably the primary issue when using tet meshes, the elements themselves are fine and in tests which compare them against hex elements on simple geometric models they perform very well, (provided you don't use linear four node tets !!). Tet elements that are highly contorted and squeezed to fit the geometry will produce spurious results (despite the claims of some vendors!). It is best to use a mesher that can either "jump over" the slivers or one that can remove and clean up difficult geometry before the meshing takes place.

There are plenty of analysts about who still insist on hex meshing everything, but with the abundance of available CAD geometry they are a dying breed.

 

[prost]

In principle there should be no difference in the numerical accuracy and reliability of the hex meshes relative to the tet meshes. In practice (unfortunately I have as yet not found mathematical proof of this), hex meshes seem to work better numerically (better convergence rates, fewer wiggles or anomalies in the engineering data such as stresses) then tet meshes. All things being equal, I normally mesh as much as possible with hex elements, while trying to use as few tets as possible (preferably for low stress regions, such as occur on crack faces, but not near crack tips).

Also in practice, you can just as easily goof up the analysis with hex meshes as with tet meshes.

 

[GregLocock]

Are you primarily interetsed in deflections, stiffness and modes, or are you after stresses?

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[spacecase]

Greg, I am primarly performing stress and normal modes analysis.

When working with a situation where the stresses are mainly from bending, is there a rule of thumb for the number of tets a long the thickness? For hex elements I've heard at least four should be used.

Thanks again!

 

[GregLocock]

If you are interetsed in stresses then the elements 'near' the stresses of interest need to be well formed.

For dynamics and stiffness then you can be a lot more casual about element quality, the most important thing is to get the mass right and make sure the mechanisms and bcs are OK.

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.