ZZS improved stress

Andrea

Mecway plot only stress tensor components when ZZS improved stress is activated (no Von Mises - no principal stresses)

Victor

Noted. Actually I didn't even know ZZS improved stress existed.

Victor

I'm working on this for version 8 now. Can you tell me if this way would be suitable?

Von Mises, principal and Tresca stresses would be generated using the ZZ stress components if they exist in the solution and the regular stress components otherwise. There would be a little note on them indicating that they came from ZZ stress if they did.

An alternative would be separate von Mises and principal stresses for ZZ that appear alongside the regular ones. I don't like this because it would be crowded but would it have some use, perhaps to compare them?

EDIT: A 3rd way would be a switch to choose between ZZ or regular stress before solving. Then the solution looks the same either way but all stresses are ZZ if your chose that.

Do you use ZZ stress because it's more accurate or to estimate error?

Andrea

I use it as an error estimator. Accuracy depends from mesh convergence.
If you have a good mesh ZZS stress and classic stress are very closed, so ZZS display a little difference.
Where the difference is higher you can evaluate if to refine mesh.
For a good mesh I take as a reference level +/- 5% (Generally 5% is the threshold for acceptance stress difference with estimator) (For ideal stress the difference is obviously positive)

For my opinion you can leave the actual situation adding ZZS vMises and Tresca. At least you can remove ZZS differences of the stress tensor and PS so under ZZS there will be only 2 values (vMises, Tresca). For my opinion could be enough for an error estimator

Regards

Atached example of ZZS for a good mesh

Victor

Since you're always going to do further calculations on it, maybe instead of showing any ZZ stress values, it could calculate the error itself and just show that? Or error + 6 ZZ stress components so you can verify it by hand.

I wonder though, why people don't use ZZ stress for everything. It sounds like it's generally more accurate than ordinary node averaged stress. The original 1992 Zienkiewicz-Zhu paper recommends FEA software should do that!

Andrea

Sincerly I know ZZS since Dhondt inserted it on CCX (from version 2.6, if I remember well).
Why he decided to confine it as an error estimator and not as a standard stress I don't know and I haven't the knowledge to understand.

Better thing is to ask directly to Dhondt!

Using ZZS often you will receive this warning! (see picture)

Victor

OK, thanks for your help. I'll have it optionally generate the ZZ von Mises, Tresca and principal stresses as well as the tensor components. And off by default because of these odd errors and not working with tet elements.

German

Victor

It is not a good idea to "switch to choose between ZZ or regular stress before solving." because for error estimation you need both, normal Von Mises and ZZS Von Mises.

Victor

OK, it'll have both at the same time.