Welcome to the forums. Please post in English or French.

You are not logged in.

#1 2021-06-25 10:10:00

Rethu
Member
Registered: 2021-06-09
Posts: 5

Mesh sensitivity, convergence in code_aster and Ansys

Greetings!

I am trying to reproduce a mechanical problem in code_aster and Ansys, to compare the capabilities of the two environment. Looking through the results, it is questionable for me, if I created the right model in code_aster, as the difference from the target value is enormous compared to the Ansys results.

The problem is a point load of a hemisphere, modelled with shell elements, more details enclosed.

I am enclosing the general model description, the comaprison excel, as well as the command and mesh file of one of the code_aster cases (mesh size 32x32x32, QUAD4 elements, DKT modelisation). Am I doing some mistakes while building up the model? Or is this kind of difference expected? The convergence of refining the mesh is perfectly observable, but in Ansys it is much 'faster' and precise, even when using QUAD4 elements in a coarse mesh.

Thank you for your help in advance!

Rethu


Attachments:
p18-3_radial_point_load_on_a_hemisphere.zip, Size: 190.38 KiB, Downloads: 68

Offline

#2 2021-06-25 12:42:40

sameer21101970
Member
Registered: 2019-09-06
Posts: 353

Re: Mesh sensitivity, convergence in code_aster and Ansys

run below .comm file and reply.


Attachments:
p18-3comm.comm, Size: 2.69 KiB, Downloads: 63

Offline

#3 2021-06-25 13:15:29

Rethu
Member
Registered: 2021-06-09
Posts: 5

Re: Mesh sensitivity, convergence in code_aster and Ansys

Dear Sameer,

Thank you for your reply. Unfortunately, the results are the same with your command as well.

Rethu

Offline

#4 2021-06-25 13:22:41

mf
Member
Registered: 2019-06-18
Posts: 261

Re: Mesh sensitivity, convergence in code_aster and Ansys

Hello,

you didn't use 2nd order elements in CA, did you (your model is stiffer)? What about the elements in Ansys? Are they 2nd order (I am pretty sure...)? I believe in Ansys there is the option of first order elements that suppress the hourglass effect (thus they are softer than the 1st order elements in CA)? Although, I must say, I am not very experienced with shells....

Mario.

Last edited by mf (2021-06-25 13:24:17)

Offline

#5 2021-06-25 13:51:16

Rethu
Member
Registered: 2021-06-09
Posts: 5

Re: Mesh sensitivity, convergence in code_aster and Ansys

Dear Mario,

Thank you for your answer!

In the excel you can see that I did multiple simulations, in some cases with second order, in some cases with first order elements, but during one case I used the same element type for both Ansys and code_aster. With the second order elements (QUAD9 bi-quadliteral and COQUE_3D modelisation) the difference is acceptable in my opinion, but when I use first order elements (in both Ansys and code_aster, DKT modelisation and QUAD4 elements) the results are more than unusual. I tried to build the models as identical as possible to have an opportunity to compare the two environment.

Rethu

Last edited by Rethu (2021-06-25 13:52:13)

Offline

#6 2021-06-25 13:54:55

mf
Member
Registered: 2019-06-18
Posts: 261

Re: Mesh sensitivity, convergence in code_aster and Ansys

Hello,
as a simple rule, if you ask me: never use 1st order in mechanical. Don't even think about it. Use them during building your mechanical model to reduce calculation time and for thermal models only.

If everything else is identical, a few percent difference are also too much. Maybe there is something else different...

Mario.

EDIT: as I tried to describe in my last post: the first order in Ansys may be the ones with hourglass suppression, so they may not be the same as the first order in CA.

Last edited by mf (2021-06-25 13:58:12)

Offline

#7 2021-06-28 09:34:22

Rethu
Member
Registered: 2021-06-09
Posts: 5

Re: Mesh sensitivity, convergence in code_aster and Ansys

Dear Mario,

I didn't get the point of your previous reply, but I think I understand now. I will check more deeply about how Ansys controls the hourglass effect, and if there is an option to ignore this control. On the other hand, is it possible to somehow control the hourglass effect for linear shell elements in CA?

I also understand that it is not advised to use linear elements at all, but unfortunately it is out of question, I have to use both linear and second order elements for the comparison.

Rethu

Offline

#8 2021-06-28 11:14:26

mf
Member
Registered: 2019-06-18
Posts: 261

Re: Mesh sensitivity, convergence in code_aster and Ansys

Hello,

as far as I know, CA doesn't have such linear elements.

Mario.

Offline