axisymetric pb on code saturn

  • guillaume23
  • Topic Author
  • Offline
  • Fresh Boarder
  • Fresh Boarder
More
10 years 9 months ago #4858 by guillaume23
Replied by guillaume23 on topic Re:axisymetric pb on code saturn
hello
I use the caelinux 2009 distribution with salome 5.1.3 and CS 2.0.0 rc1.
i have made different test. I tried to move the mesh, it looks like good. I tried the checking face orientation without succes.
I think it is an axis problem so I will try to explain the method i use with the case issue of www.caelinux.org/wiki/index.php/Contrib:...Flow#Fluent_Tutorial just to know if the method is rigth.

I did the geometry and mesh like in the tutorial.
This tutorial is with an old version of CS.
The axisymetrie is along y axis, the radius along z axis and the extrusion of 2D mesh along x axis. I enter 360 degree and 1 in Dy vector in preiodic boundaries. Then I put the definition of boundary region like in the tutorial. At this step I have a problem because in the new version there is no "axis" choice in nature of boundary region, I put symetry.

So there the axis rotation is bisecting a face because the axis is a edge of the mesh. the same for your second comment, if the box have a 360 degree rotation, i think there is interconnexion in the result.
I made a test with an axisymetrie with 30 degree angle and the preprocessing look like good.

I think my method is wrong, but i don't know exactly where.

Have you got an idea?

I have no listing file when the simulation failled. only listpre. file like this I send in the previous post.

Please Log in or Create an account to join the conversation.

  • guillaume23
  • Topic Author
  • Offline
  • Fresh Boarder
  • Fresh Boarder
More
10 years 9 months ago #4859 by guillaume23
Replied by guillaume23 on topic Re:axisymetric pb on code saturn
hello
I use the caelinux 2009 distribution with salome 5.1.3 and CS 2.0.0 rc1.
i have made different test. I tried to move the mesh, it looks like good. I tried the checking face orientation without succes.
I think it is an axis problem so I will try to explain the method i use with the case issue of www.caelinux.org/wiki/index.php/Contrib:...Flow#Fluent_Tutorial just to know if the method is rigth.

I did the geometry and mesh like in the tutorial.
This tutorial is with an old version of CS.
The axisymetrie is along y axis, the radius along z axis and the extrusion of 2D mesh along x axis. I enter 360 degree and 1 in Dy vector in preiodic boundaries. Then I put the definition of boundary region like in the tutorial. At this step I have a problem because in the new version there is no "axis" choice in nature of boundary region, I put symetry.

So there the axis rotation is bisecting a face because the axis is a edge of the mesh. the same for your second comment, if the box have a 360 degree rotation, i think there is interconnexion in the result.
I made a test with an axisymetrie with 30 degree angle and the preprocessing look like good.

I think my method is wrong, but i don't know exactly where.

Have you got an idea?

I have no listing file when the simulation failled. only listpre. file like this I send in the previous post.

Please Log in or Create an account to join the conversation.

More
10 years 9 months ago #4860 by Yvan Fournier
Replied by Yvan Fournier on topic Re:axisymetric pb on Code_Saturne
Hello,

We do not generally use periodicity but simply symmetries for 2D axisymmetric flows. If you have a velocity component normal to the 2D plane, you would need periodicity, but then I am not sure whether you may call the calculation axisymmetric or simply periodic (which is not really important).

In any case, periodicity of rotation with an angle of 360 degrees has no use: if your mesh was built by rovolving a 2D mesh around an axis with a full 360 degrees, then you have meshed the whole domain, so you have a full 3D mesh (probably with a bad quality mesh near the axis).

For an axisymmetric calculation, 10 to 30 degrees should be enough. Also, one cell thickness is normally enough (otherwise, you are doing a true 3D calculation, not an ayisymmetric 2D calculation). Also, if you have an edge along axis Y for example, and the 2D mesh is in the XY plane, revolving the mesh from -10 degrees around Y to +10 degrees around Y is better the from 0 to 20 degrees, as the XY plane is the median plane, and the Z component should be null (which makes postprocessing simpler).

Finally, note than in Code_Saturne, periodicity is not a boundary condition, but is a true geometric property (this is not perfectly true for periodicity of rotation, as some terms are explicited, making this more similar to boundary conditions).

That is why periodicity is defined in the preprocessing stage, and not with boundary conditions.

Best regards,

Yvan

Please Log in or Create an account to join the conversation.

  • guillaume23
  • Topic Author
  • Offline
  • Fresh Boarder
  • Fresh Boarder
More
10 years 9 months ago #4861 by guillaume23
Replied by guillaume23 on topic Re:axisymetric pb on Code_Saturne
Hello,
thank you for this information. I see the problem become of my geometry. I send a picture of the new mesh i did with you advises just to know if i understand well. I have made some test and preprocesseur and simulation are good. I have some strange solution for velocity near the wall but i have to increase iteration number to see more.

File Attachment:


If I have understand, periodic boundary is for geometry periodicity (I don't see exactly what it is). Axisymetry is defined by symmetry in boundary condition, but when we say "axisymetry" is it obligatory a 360 degree rotation or it can be only a 180 (or else) degree rotation?

Thank you.

Post edited by: guillaume23, at: 2010/10/20 19:41<br /><br />Post edited by: guillaume23, at: 2010/10/20 19:43
Attachments:

Please Log in or Create an account to join the conversation.

More
10 years 9 months ago #4864 by Yvan Fournier
Replied by Yvan Fournier on topic Re:axisymetric pb on Code_Saturne
Hello,

Any angle can do. Theoretically, what is important is that the mesh is built based on a revolution around an axis (so that volume ratios are consistent). Usually, -10 to + 0 or -5 to +5 degrees is OK, and having a domain that is &quot;symmetric&quot; around a plane helps simplify post-processing and interpretation.

The mesh in your screenshot seems OK (though I do not see if it is symmetric relative to an XY, XY, or YZ plane, which you can always attain through an additional rotation if this is not already the case).

Best regards,

Yvan

Please Log in or Create an account to join the conversation.

  • guillaume23
  • Topic Author
  • Offline
  • Fresh Boarder
  • Fresh Boarder
More
10 years 8 months ago #4897 by guillaume23
Replied by guillaume23 on topic Re:axisymetric pb on Code_Saturne
hello,

I think i understand. I have other basic questions but I will send on another post.

Thank you for advices.

Guillaume

Please Log in or Create an account to join the conversation.

Moderators: catux
Time to create page: 0.142 seconds
Powered by Kunena Forum