cylindrical coordinate system (or not?)

[pilafa]

Hi,

It seems to have a confusion between which coordinate systems are truly available and the ones mentioned in the documentation.

I would like to define a velocity normal to the face of a cylindre. I understand that there is no cylindrical coordinate system to do that. Am I right?
Any alternative to suggest?

Thanks,
Pierre

[mzenker]

Yes there is:

Coordinate System = Axi Symmetric

Then your 2D geometry will be rotated about the y axis.

HTH,

Matthias

[pilafa]

Hi Matthias,

The problem is that I have a 3D geometry, not a 2D axi-symmetric.

Can I still use this CS?

Pierre

[mzenker]

No, but there is the possibility of coordinate transformation in Elmer. Check out chapter 14 "coordinate transformation" of the solver manual.

HTH,

Matthias

[pilafa]

Thanks!
Looks promising....

On the picture you see my original mesh on the left, and what I see after adding :

Code: Select all

Simulation
...
  Coordinate Transformation = String cartesian to cylindrical 

I tried to revert the transformation adding something in the postproc solver but without success....
Any suggestion?

Thanks again
Pierre

[raback]

Hi

I think if you set the transformation to Simulation section it is a global irreversible operator but adding to it to each solver makes it solver specific.

Now unfortunately there is no 3D solver that would be operating in cylindrical coordinates. Cylindrical coordinates is a 2D special case. There is usually no reason in using cylindrical coordinates if there is a symmetry break.

Also if you would use this transformation for postprocessing the trouble is that vector fields are not automatically treated accordingly. It is sort of funny to solve N-S in cartesian coordinates, map the mesh to cylindrical coordinates but still have the result in cartesian coordinates.

The functionality was initially coded for Tokamak fusion plasma simulation where these both coordinate systems were used. However there were also some additional mapping solvers that did not find their way to the ElmerSolver library. Thus I would find it surpricing if these features would be very usable as such.

-Peter

[pilafa]

Hi Peter,

Thank you for this information.

The initial idea was to set the velocity boundary condition normal to a cylindrical surface. But I realize that I could simply set the velocity using Normal-Tangential CS. No?

Code: Select all

Boundary Condition 1
  Name = String "inlet"
  Target Boundaries(1) = 1
  
  Normal-Tangential Velocity = Logical True
  Velocity 1 = Real 1.0
  Velocity 2 = Real 0.0   
  Velocity 3 = Real 0.0

End