results different according mesh size

Hello,

I 'm just starting using salomé with CAElinux. I'm very happy of that. But i have problem : I import my 3D from my CAD software using STEP file. I do all the things needed to mesh / calculate with ASTK with success. All right. But if I recalculate the mesh with a different value (0.5 instead of 0.2 for example) the value of deformation is different : 123 instead of 140. Is it correct that the result differs according to th mesh tolerancy ? If yes, how do we know when the result is correct ?

P.S Sorry for mypoor english

Hi,

It is my understanding that with linear elastic FEA it is necessary to do successive runs (i.e. many runs), refining the mesh each time. When the results of the last run and the latest run are close enough within the desired accuracy you have your solution.

I would be interested to here others options on this, as in some locations (i.e. stress concentrations/notches) you may never get convergence in this way.

hello,

in the case that You use 3D-elements, it is very recommanded to use elements with midside nodes. This will improve the convergence to the correct solution much better than using a big and bigger number of simple (linear) 3D-elements (as HEXA8 or TET4). I found this by trying out and discussed with people of EDF who confermed this. Some codes (as Nastran) converge with much less linear 3D-elements, because their element-functions has bubble-functions built in (and may be for some other reasons). Compared to this, the linear 3D-elemnets of Code-Aster are \"built\" very simple. This is not a deficiency of quality, but is nessesary for a good behavior in nonlinear calculations (where Nastran-elements are very bad).

If You have meshed with linear 3D-elements, You can change Your mesh in the command file by using MODI_MAILLAGE. There are tutorials in CAELinux for doing this.

If You are interested for a beam-example demonstrating the convergence and comparing with the exact solution, let me know Your e-mail-adress, You find mine at www.code-aster.de

Good luck, Johannes

Thanks for your answers.
The way of using midside elements seems to be interesting,because even with a coarse value for meshing it is very long with my 3D models (wich are full of unuseful details for FEA...). Can we directly mesh with this kind of elements using SALOME or do we have to do it with EFICAS ?
I'm very beginner, i had a quick look at the tutorials, but in fact i didn't find a case near mine.Wich tuto contains the explanations for the MODI_MAILLAGE ?
Thanks

Sorry, I wrote something wrong. The command You must add ist not MODI_MAILLAGE, but CREA_MAILLAGE. The 2 examples in CAELinux are:

- /opt/helpers/Templates/LinStatics3D.comm, and
- /opt/helpers/Templates/LinStatics3DQuad.comm

The explanation text You find (only at the most recent CAELinux which is CAELinuxBeta3b) at:

- /opt/helpers/docs/tutorials/JMLtutorials/caelinux4.pdf

good luck! Johannes

To Your questions:

\"my 3D models (wich are full of unuseful details for FEA...)\" -> it is useful to simplify the geometry before importing it into Salome to exclude unuseful details which could blow up your FE-Model

\"Can we directly mesh with this kind of elements using SALOME\" -> no
\"or do we have to do it with EFICAS ?\" -> yes