Interference-fit analysis

Peter Halverson
Offline
Senior Member

14 years 8 months ago #4610 by Peter Halverson

Replied by Peter Halverson on topic Re:Interference-fit analysis

Part of the issue is that when you change the mesh from linear to quad in aster it preserves the original mesh groups meaning that it doesn't add the new quad groups.

Les groupes de mailles sont conservés, les groupes de noeuds aussi, comme lors du raffinement
de maillage les noeuds créés ne sont pas introduits dans les groupes de noeuds.

I do not know how to get around this in Code-Aster directly. However, you can do the transform in salome. But that isn't entirely the problem (although it is a big portion of this). I'll need some time for meditation on contact and quad elements. Post edited by: Peter Halverson, at: 2010/08/27 17:15

kwou
Offline
Moderator

14 years 8 months ago #4612 by kwou

Replied by kwou on topic Re:Interference-fit analysis

Hoi Pete

I think (and I hope) you misinterprete the part about preserving the groups etc in CA when you change from linear to quad. The advantage of grouping geometrical entities is exaclty that you can change your mesh but preserve these geometrical entities.
That is why you should -in general- not use nodes (noeuds) directly from the mesh. You could use nodes that are attached to geometry (eg vertices)- they remain unchanged during mesh manipulation.

[code:1]Les groupes de mailles sont conservés, les groupes de noeuds aussi, comme lors du raffinement
de maillage les noeuds créés ne sont pas introduits dans les groupes de noeuds.[/code:1]
Is this extracted from Homard - mesh raffinement?
See above - new nodes donot belong to previously defined node group.
They do however belong to eg a surface if they are created wihtin that surface. This will be the case for changing from linear to quadratic elements. The new (midside) nodes belong to the group as long as they are on the geometrical entity.
At least this is how I understand it.

kind regards - kees Post edited by: Kees Wouters, at: 2010/08/28 10:42

Interest: structural mechanics, solar energy (picture at 'my location' shows too little pv panels)

--
kind regards - kees

theCapitain
Topic Author
Offline
Junior Member

14 years 8 months ago #4615 by theCapitain

Replied by theCapitain on topic Re:Interference-fit analysis

Following Peter's thought, I tried to convert the mesh to quadratic elements form Salome main menu getting a good solution, with excellent agreement with analytical results and a smooth contact surface. Also linear elements works fine, as noticed some posts ago. Instead lin-quad conversion inside Code_Aster leads at the contact surface with “spikes” visible in my first post. It's evident that some "esclave" node penetrates the "maitre" surface … however doesn't look like all the intermediate (added) node misbehave.
For my display problem I had an idea: since the EQUI_NOEU_SIGM field shouldn't contain displacement data, maybe it cannot produce a displacement scaling display. Maybe the deformed shape and scalar map display based on EQUI_NOEU_SIGM field lead to a deformed shape somehow parametrized on stresses. Please, someone let me know if I'm right.
Thanks Post edited by: theCapitain, at: 2010/08/28 20:11

Peter Halverson
Offline
Senior Member

14 years 8 months ago #4616 by Peter Halverson

Replied by Peter Halverson on topic Re:Interference-fit analysis

Kees Wouters wrote:

I think (and I hope) you misinterprete the part about preserving the groups etc in CA when you change from linear to quad. [/quote]

I think that I did. It came from CREA_MAILLAGE (U4.23.02 page 9/14). I did a quick test and noticed that they were definitely part of the group_MA even after changing them into quad elements. Which is what one would expect. The text I posted must only apply to GROUP_NO, which is once again what you would expect. I'm not sure what lead me to believe this. I was getting funny results and ended up converting my meshes to quadratic in salome. Perhaps it was that CA does not create the extra nodes on geometry and Salome can?

@theCaptian

I'm still not sure what you are looking for in displaying the displacements. Do you want an image of the displacement and vmis if so select the timestep under the DEPL field and right click displacement and scalar. Under the scalar field select the appropriate stress Equi_something and then in the next tab select VMIS. The scaling variable only applies to the displacement

JMB
Offline
Elite Member

14 years 8 months ago #4618 by JMB

Replied by JMB on topic Re:Interference-fit analysis

theCapitain wrote:

I tried to convert the mesh to quadratic elements form Salome main menu getting a good solution, with excellent agreement with analytical results and a smooth contact surface.

Hello,

I have not had the time to run you previous model yet, but would you please post your latest, now that you are happy with the results? Thank you.

Regards,
JMB

theCapitain
Topic Author
Offline
Junior Member

14 years 8 months ago #4621 by theCapitain

Replied by theCapitain on topic Re:Interference-fit analysis

Peter Halverson wrote:

Do you want an image of the displacement and vmis if so select the timestep under the DEPL field and right click displacement and scalar. Under the scalar field select the appropriate stress Equi_something and then in the next tab select VMIS. The scaling variable only applies to the displacement

OK, I'm an idiot. I didn't observe that Equi_something can be extracted from the field DEPL so I selected displacement and scalar by right click on the EQUI_NOEU_SIGM field obtaining a strange deformed shape.

JMB wrote:

would you please post your latest, now that you are happy with the results? Thank you.

Here is my last result, the mesh is more graceful (but that isn't the solution) it's created inside Salome with conversion to quadratic. Well, neither that is the solution, linear elements perform well providing a satisfactory mesh density on contact surface, while a coarse mesh leads to convergence difficulties. For example, with the mesh on the image I cannot get a solution using linear elements for non-convergence error.
I'm just doing tests on interference fitted assembly and I started this tread after obtaining curious results using quadratic elements (changing the mesh in Aster). Actually I haven't understood what's the matter.

Moderators: catux

Time to create page: 0.154 seconds

Notice

Interference-fit analysis

Attachments:

Main Menu

Forum

Login Form