[tsherk86]

Hello,
I've been working on creating a mesh for a cylindrical annulus, but have had no luck finding out how to extends the mesh points INSIDE the annulus so that there are grid points to solve the solution inside the cylinder rather than just on the outside. I can't seem to find out how to do this anywhere. Am I missing something? OpenFOAM wouldn't solve all the way through the cylinder if I only had grid points on the exteriors of the annulus would it? i.e., I believe the fluid flow would appear the same the entire way through the cylinder along the z direction as I have no mesh on the inside connecting the boundaries.

Thanks for your help!

-Trevor

[tsherk86]

Here is a picture of my mesh so far

[tsherk86]

And here is my blockMesh file:
blockMeshDict

[cedric_duprat]

I Trevor,

just small questions, what do you want to do ?

- solve Navier-Stokes equation inside the "internal"pipe ? (but no mesh ...)
- solve Navier-Stokes inside the annulus part ?

well, where is your fluid, ect ...
Can you tell me a little bit more just for me to help you (if I can) because, if you need to solve a cylindrical annulus, I don't understand why you need solution in the internal pipe ...

Cedric

[eugene]

To make a cylinder, you need 12 blocks. 4 for the core region and 8 around the outside.

When joined together, the 4 core blocks will expose 8 facets in the radial direction. Seen end on, you will distort them to look like an octagon. To each of these 8 facets you then attach one of the 8 outer blocks to interface with the outer cylinder.

[tsherk86]

Hello Cedric,

The fluid is inside the annulus, so between the inner wall and the outer wall. I wish to solve Navier-Stokes inside the annulus part which is empty of mesh points. My understanding of simulation and modeling is that when solving these equations, the solution is iterated over the mesh points. Therefore, to model the fluid inside the annulus I must have grid points inside the annulus and not just on the walls. Is this correct? If so, I am wondering how to connect the inner and outer walls with a grid inside the annulus.

The problem is the differentially heated rotating annulus problem. Perhaps you've heard of this classic experiment.

Thanks!

-Trevor

[hsieh]

Hi, Trevor,

It does look like you have grid points inside domain you are interested. I run your blockMeshDict and plotted the mesh and a clip of the mesh.



Pei

[hsieh]

Hi, Trevor,

It does look like you have grid points inside domain you are interested. I run your blockMeshDict and plotted the mesh and a clip of the mesh.




Pei

[cedric_duprat]

Hi trevor,

As pei show us, it seems that you aldready have some points inside the annulus region.

These points correspond to the line :
hex (0 1 2 3 16 17 18 19) (10 10 10) simpleGrading (1 1 1)
of the blockMeshDict.
For exemple here, the (10 10 10) is the number of points in each direction (x, y z) for the first block.
The same for the other.
If you are still not sure (I hope you are now) you can still do a checkMesh of your case and you should see the number of nodes, cells, ect ... of your geometry

[tsherk86]

Sorry, I'm not convinced still. Paraview must fill in those lines so that it appears to have mesh points inside. I have done the clip filter with a simple box too and it shows the same effect - making it appear to have grid points all the way through, but they don't. Try doing the clip filter at, say, a 45 degree angle and you will see what I mean. Those lines you see in Pei's picture are created by ParaView for some reason.

Thanks,
Trevor

[hsieh]

Hi, Trevor,

OK, this is a cut out of the original mesh. It will hard to believe that there are no nodes inside the domain.



Pei

[hsieh]

Hi, Trevor,

Forgot to mention this:

Look in constant/polyMesh/Points file, it lists 9680 points. This is 11*11*80 - matches the number of points you specified.

Pei

[tsherk86]

Hello Pei,

Well, after seeing that picture I now believe you are right. Thanks! I will now try my simulation.

I just assumed the points weren't inside because of Paraview actually. When I would look at the mesh in Paraview and zoom inside, there was no inside grid. Also, when I would run the tutorial cases, the fluid flow would appear the same through all slices of the geometry. I guess they must not have been intended to be 3D cases. This is great! I hope you are right.

Thank you very much for your time and assistance.


-Trevor

[cedric_duprat]

Hi Trevor and Pei

"I hope you are right" .... no doubt, we are :o)

Pei, nice picture the last one, I'm really curious to know how you managed to get that with paraView ?
Did you decompose the mesh for parallel run and it correspond to one processor mesh ?

Thanks,

Cedric

[hsieh]

Hi, Cedic,

In paraview 3.2.1,

1. click "Edit", select "Select Cells Through"
2. select the area of mesh you are interested using the mouse.
3. Click "Filter", then, "Alphabetical", then, select "Extract Selection".
4. click "Copy Active Selection".
5. click "Apply"
6. click "Display"
7. under "Style", select "Surface With Edges" under Representation.

Pei
PS: hope I did not miss anything.

[cedric_duprat]

Hi Pei,

Thank you, I'll try with the old version of paraView first.
And maybe with the new one in a couple of weeks.

Cedric

[tsherk86]

Hello again!
I am having some trouble inputting the equations for the model into OpenFOAM. Specifically, for dU/dt (U is the vector fluid velocity), I need to calculate r, the radius ((x^2+y^2)^(1/2)), and the cylindrical unit vector r, i_r, which, when expressed in terms of cartesian coordintes, I THINK is cos(arctan(y/x))*i_x + sin(arctan(y/x))*i_y, where x and y are the usual cartesian coordinates.

I think I can use mag(U) to calculate r (?), but from my reading I believe OpenFOAM treats tensors as a unit, and does not look at individual components, such as x and y... anyhow, I am not sure how to do handle finding x and y. I tried U.component(0) and U. component(1) for x and y, and get a floating point exception. So I gather my other problem is that U.y is sometimes 0...

Any suggestions would be greatly appreciated!!

Thanks,
Trevor

[tsherk86]

Never mind! I think I figured out another way to do the problem.

[T.D.]

Hi Guys, i am new in openfoam, my question is that: if i want to the solve the navier stokes eqns in cylindrical coordinates, i must just define my coordinates system to cylindrical one (r, theta,z)? if so, what about the operators like "laplacian","divergence", for example: div(phi), they will be automatically used in cylindrical coordinates? or we must write new definitions for them in cylind. coord.?
thanks a lot