Flexure problem with ASTER

Did you use the wizard? If yes, check if your mesh is linear or quadratic. I made a benchmarck and I was surprised on how big the difference is beetwen linear and quadratic mesh.

The default mesh in salome is linear, but you can trasform it in quadratic in two ways: in the comme file or simply with the gui in the mesh module inside salome.

Using this approach, I obtained exactly the same results that a collegue obtained using Ansys.

See here:
caelinux.com/CMS/index.php?option=com_jo...;id=2529&catid=4

Alessandro wrote:

Did you use the wizard? If yes, check if your mesh is linear or quadratic. I made a benchmarck and I was surprised on how big the difference is beetwen linear and quadratic mesh.

The default mesh in salome is linear, but you can trasform it in quadratic in two ways: in the comme file or simply with the gui in the mesh module inside salome.

Using this approach, I obtained exactly the same results that a collegue obtained using Ansys.

See here:
caelinux.com/CMS/index.php?option=com_jo...;id=2529&catid=4

Oh thanks, I didn't had this detail in mind. I checked in Advanced Mesh Info and the elements of the mesh are Linear.
Could you be more specific on how to change them in quandratic please ? Or point me a page where I can read more about it ?
I'm looking already on how to modify, but it would be faster if I had a base to start from.

I can't access to salome right now, but let's if my memory is still good :

Mesh module--> menu "modify" or something similar-->trasform to/from quadratic

Anyway it will be easy for you to find the command, or look for it in the on-line help.

Have a good work.

Thanks, i found it and re-created my new groups.
But the problem/model should change from MedLin (I suppose it is linear) too some Quadratic problem, no ?

Boloni wrote:

Thanks, i found it and re-created my new groups.
But the problem/model should change from MedLin (I suppose it is linear) too some Quadratic problem, no ?

Are you italian???? The "no" at the end of your sentences is too funny (I'm italian too!).

I'm not very expert on this particular fact, but things are that if you use a quadratic model, the code behaves making parabolic interpolation inside every element, and this lead the simulation to be more accurate. The price to pay is a big increment of the number of nodes, and consequentely a big increment in the memory and time consumption. In any case, as Joel Cugnoni said to me, just forget to use linear meshes to simulate structural problems.

No I'm not italian but I have italian friends.

I tried to do a LinearStatics3D problem but I get some errors of this type:
"Vérifier la définition du modèle (AFFE_MODELE) et assurez-vous que les types de mailles du maillage (SEG2, TRIA3, QUAD4) sont compatibles avec votre modélisation"
and
"Le calcul de l'option: NSPG_NBVA n'est possible pour aucun de types d'éléments du LIGREL"

I created nodal groups and imposed 0 displacement BC and also applied a pression on a group of facets.

I'm using FEM in electromagnetism and i kind of know about shape fonctions for 1st and 2nd order elements. What I'm planning to do is to make structural calculations with ASTER and electromagnetic computations with an another code, then if I can't use the same mesh to couple them, it will be kind of problematic.
But now, I'm still at the learning and veryfing of ASTER, phase.
Thanks for any tip you will give me