How to skip the element check in abaqus ?

[opethian]

Hi to all experts,
I have another hot topic for you !!!
I have a question about the element check in abaqus. When i run a model done by another person i got an error message like ErrElemVolSmallNegZero this is because of element quality. It is ok so far but they claim that it is possible to run this model and the model, the model is working ok and there is a way to skip the element check. Do you think that it is possible to skip the element check in abaqus and how? I run the model in abaqus 6.7.4 and i also have other versions of abaqus too lastly it didnt work with 6.5.6 either.
Thanks

 

[johnhors]

Remove the element from the input deck.

 

[opethian]

but i need those elements for the compatibility of my model and i have a lot of errorenous elemnets so i heard from more experienced engineers that there shuld be a way to just sk?p the element quality check. I didnt mesh the model and i have to run this model without and re-mesh. that the problem indeed.

 

[johnhors]

Abaqus as with all commercial solvers will abort a solution when confronted with very badly shaped elements. Such elements ill-condition the stiffness matrix, any solution if you get the solver to proceed is highly questionable, not just in the locality of the bad element(s) but for the entire model!

If you have just a sprinkling of illegal elements within a large tetrahedral mesh (away from regions of interest) then the effect of removing these elements will most likely have a negligible effect on the validity of the solution.

However you have now stated "i have a lot of errorenous elemnets". Unfortunately this renders the model as not just useless but also positively dangerous! Your only correct course of action is to ditch this model and build a good one, at least on moral grounds, and most probably on legal grounds as well!

Just how experienced are those "more experienced engineers" ?

 

[gwolf2]

In this situation I would read the mesh into my pre-processor and assess the number of bad elements. If it is less than 100, say, I would correct them myself manually. This assumes you have access to quick, simple node move and element jacobian check functions in the preprocessor (another reason why I like PATRAN).

If there are more bad elements than you can fix in a few hours it is a bad model and should be viewed with extreme caution. That kind of sloppiness is often a sign of other problems in the model like bad boundary conditions etc.

gwolf

 

[opethian]

i see... u both say right in my opinion but the interesting point is they claim that this model is running there is a possiblity that they gave me an older version of input file or they do it with a purpose i just want to find out they by considering if there is a way to skip the checkk test. ? on the otherhand it is not problem for me to remesh the part.

 

[johnhors]

"it is not problem for me to remesh the part."

Then do that!

NEVER bypass an element check, even if it is possible.

I agree with gwolf2, if the original modeller didn't bother to check the validity of the mesh then I would have little confidence with the rest of it.

I use software which can automatically check and correct element Jacobians by manipulating the mid side node positions (coarsely meshed holes can end up with a facetted appearance looking like a 50 pence coin or thrupenny bit), very often this will make a bad element with excessive curvature become good. However this routine will not revive something like a nearly flat tetrahederal element, in which instance you must reposition one or more vertex nodes and then run the risk of corrupting adjacent elements! I've done this for a few elements in models in the past, but it is better to re-mesh the part.

 

[gwolf2]

John, what software you use?

Yes - it's always mid nodes on fillets and holes because someone can't be bothered to seed these surfaces/curves tidily prior to TET meshing.
gwolf

 

[johnhors]

gwolf2 - I use Roshaz to do the mid-side node movements to improve element Jacobians.