ALE modue - mesh movement

[bajsi]

Hi,
I am using CS 3.3.2 (installed from Ubuntu repository) and Salome-Meca 2014.2 as pre/post-processor.
My aim is to simulate valves: closing and opening of shutter and determinating corresponding flow rates. Currently I am struggling a lot with using the ALE module for imposed mesh movement.
As a test I am trying to recreate the tutorial for fluid structure interaction (code-saturne.org/cms/sites/default/files/file_attach/Tutorial/Fluid-structure-interaction.pdf), but I ran into difficulties.

I created a mesh (tetrahedral - as I will need such mesh later) and imposed fixed displacement on the cylinder - time dependent so that it moves with constant velocity (as if constant velocity were imposed).
The back and front boundaries have sliding boundary.

The problem is that the cylinder can only move so much as the size of first layer of mesh elements - afterward elements appear to be so thin so that calculation fails. I tried changing mesh viscosity (from 0.1 to 1e10) but nothing significantly different happened. FYI there may be some discrepancy with symbols for viscosity CS requires symbol mesh_viscosity_1 but when viewing results in Salome I found references to mesh_vi1 that I believe was used in previous versions (I tried defining both - but no cake). What am I doing wrong?
Attached is an image of mesh at start and when calculation fails. Attached are also settings file and mesh file.

[Yvan Fournier]

Hello,

The difference in name is not important, even if it is ugly (fields have both a name, used as a "unique identifier", and a label, which can be changed by the user, and is used in logging and postprocessing.

How fast does the mesh quality degrade ? If increasing the mesh viscosity is not sufficient, you may try reducing the time step (I'm not an ALE specialist, but have heard this recommendation multiple times, so I'll forward that ...)

Regards,

Yvan

[bajsi]

Hi Yvan,

I have tried calculating with different time steps (0.1, 0.01, 0.001 s). In all three cases the results were virtually identical. Calculation fails at 14.5 s of simulated time in all three cases.

What I find very strange is that mesh viscosity, as such, has virtually no influence.

I would really appreciate it, if you could forward my problem to someone who is more experienced with ALE-s.

Best regards,
Niko

[Yvan Fournier]

Hello,

You'll probably have to wait a week or two, but I'll forward this.

Regards,

Yvan

[bajsi]

Hi,

I think I may have figured out what was wrong with my model.
Apparently boundaries defined as walls, inlet, and outlet have unmovable mesh.
When I assigned symmetry on front and back boundaries, the cylinder could move as expected and mesh deformed appropriately.