comparison of first and second order tet elements 2

[femac]

Hello, do you have any information or experience about the validation and comparison of the first and second order tet element results? Some body are asserting, that the linear tets are affording more real values. It acts mainly for calculation of die casting aluminium parts.
Thanks! Femac

 

[corus]

Linear tetrahedral elements are of no use at all other than in a thermal analysis, possibly. Some programs actually prevent you from using them in stress analysis.

 

[Theanalyst]

Hello femac,
Try this link, I think it has very useful info that generally applies to any FEA program:

http://www.mscsoftware.com/support/library/conf/auc99/p04299.pdf

Take care.

 

[brad]

WHoever is asserting that first-order tets are giving better results is most certainly wrong. First order tets should be avoided like the plague.

Their formulation is not good, and I don't know of an FEA software company which does not harshly warn against using them.

NAFEMS has some good benchmarks which can be used to evaluate various elements; the easiest thing to do is some very simple tests--cantilever beams, etc--for which you can obtain a known theoretical solution. When you do these tests, you will find that first order tetrahedrals yield very poor results.

Brad

 

[AbuAqel]

Thanks Theanalyst,
It was usefull for me

Ciao
AbuAqel

 

[brad]

I should acknowledge corus's point--in heat transfer first order tets are acceptable (I didn't mean to imply otherwise). My comments were relevant to stress.
Brad

 

[llianhui]

I think first-order tets are giving better results in explicit dynamics.
Any comment?

 

[brad]

First-order tets, explicit or otherwise, are still constant-strain elements and will yield overly-high stiffnesses (and result in poor results) and poor convergence.

Second-order elements in contact can yield spurious contact stresses in many situations. This has to do with the implementation of contact algorithms as applied to 2nd order element formulation. In this case, contact-intensive problems (which often happen in explicit analyses) could yield questionable local contact stresses for second-order elements (although far-field responses should still be reasonable).

First order tets, in contract, may give apparently "cleaner" contact stresses, but the underlying formulational problems still exist, and I would not trust those first-order tetrahedral results.

Some codes have specially-formulated 2nd order elements which address the 2nd order contact issues. In this case, these elements provide not only superior element response but also reasonable contact response.

In short, for explicit non-contact problems, 2nd tets are definitively superior, but for explicit contact problems they may have problems at the contact patch (which is addressed in some codes).

Brad

 

[femac]

To All:
thanks for your very usefull contributions. I made an opinion poll by our customers the european car makers, too. All are same opinion as above. I found out also other comparison literature in mean time.
Below is an adress about a very good comparison of all hexa and all tetra meshes for elastic and elasto-plastic analysis from Brigham University & National Laboraties.

http://research.et.byu.edu/cubit/Benzley,Perry,Merkley,Clark,Sjaardama.pdf

femac