Results at gauss/integration points. Please help. Urgent!

[ankushaggarwal2]

Hello,
Is it possible to get the results at the gauss/integration points directly, instead of the results at nodes or elements (which are interpolation of results at a few gauss points).

I need to know this urgently. Thank you very much.

 

[cbrn]

Hi,
yes, it is:
ERESX,NO
directly copies the gaussian (integration) points results to the nodes, without extrapolation through element shape functions.

Regards

 

[ankushaggarwal2]

But isn't there any other option?
This command should be used before solving the problem and I have already solved the problem.
And I want to know some method where I can directly see the results at the gauss points, not the nodes or elements.

 

[Stringmaker]

Why don't you look at an elemental solution then? If you are interested in convergence you can use the ERNORM command to calculate the difference between the nodal solution at that at Gauss points. What type of elements are you using?

 

[ankushaggarwal2]

In element solution, the result is the average of all the gauss points.I am using brick 8 node element.
I can't get you on the ERNORM thing. Can you please explain.

 

[Stringmaker]

Are you using fully integrated elements? If you are strain should vary linearly across the element as should stress. Looking at the error norm allows you to compare the highest nodal stress value to that of the highest integration point...hence is a measure of mesh convergence. Try ERNORM,ON in /POST1 to display this. Why exactly do you care about stresses at the Gauss points? Don't you care more about what's happening on the surface of your elements rather than internally?

 

[ankushaggarwal2]

Yes I am using full integration. Can I use ERNORM after getting the solution?

Actually, my guide works with another software ABAQUS, where we can get the results at the gauss points. As the results at gauss points are the direct results of solution, she insists to see the results of my calculation at the guass points too.
If you could help me!

 

[cbrn]

Hi,
yes I can understand the point, but as Stringmaker says I don't see why you couldn't "reverse" the problem to a more "readable" fashion and express everything as nodal results... Bear in mind that what you "see" in post-processing are elements / nodes, never gaussian points, so the stress field you would see even with ERESX,NO would be artificial (copy of values betwen non-coincident locations... in my opinion, much more misleading as the extrapolation using the elem shape functions!).
However, ERNORM works in /POST1, while ERESX works only in /PREP7 or /SOLU. If the solution doesn't take weeks to complete, you'd better re-run with ERESX,NO...

Regards

 

[pja]

If your "guide" is so intent on comparing apples to apples make her extrapolate the solution from the Gauss points to the nodes.....She should be able to do this as postprocessing step.

If all she is concerned with is convergence than as the others say ENORM should do the trick.

 

[ankushaggarwal2]

Thanks for all the suggestions.

Could you also please tell me, for the mesh convergence check using ENORM, what value of structural error energy and absolute maximum stress variation can be taken as acceptable results.

For example in my case,
The absolute stress variation is between 0.3 and 36
And, the structural error energy is between 1e-08 to 0.4e-04.

Thanks.