Fourier number + problem velocity results

[clalgourdin]

Hello,

During a simulation of a air diffuser in a room, I have stranges velocities.

I have already change mesh parameters and time step parameters, but I would like to know how can I check what is wrong ?

In the listing file the Courant Number is correct and residual part too.

About the Fourier Number or the rate (Nb Courant / Nb Fourier) what is the correct value for that ?

(I have cut the listing file to lighten it)

Thanks for your help.

Claude

[I'm using the version 4.0]

[Yvan Fournier]

Hello,

What type of mesh are you using ? Do you have buoyancy effects ?

Regards,

Yvan

[clalgourdin]

Hello Yvan,

Thanks for your reply,

About the meshing, I used the alogoritm Netgen 3D-2D-1D in Salome.

About the buoyancy, the density is variable according the air temperature.

Regards

Claude

[Yvan Fournier]

Hello,

Could you visualize the mesh quality criteria, and see if the strange spot correlates with that ?

If you have a mesh quality issue, for a simple geometry, using a "ijk" type mesh in SALOME would help. If that is not possible, adding more sweeps in the Code_Saturne resolution might help. Also, is you mesh 2D extruded, or fully 3D ? If you are using symmetry on some faces, extruding a 2D mesh will probably give better results than meshing a thin slice in 3D (if you mesh is not "thin", forget about this remark).

Also, are you using iterative or least-squares gradients ? With tetrahedra, least squares with extended neighborhood might be better.

Regards,

Yvan

[clalgourdin]

Hello Yvan,

about the mesh quality, there are 11 poor meshs on 495 000 cells so the meshing is correct.
Details about the mesh is in the attached file.

I'm using for the gradient calculation method : "Iterative handling of non-orthogonalities", so I will try with the least square method over partial extended cell neighborhood"

Thanks for your help

Claude

[Luciano Garelli]

Hello,

Did you try to reduce the Courant Number to get a Co~1..5?.
Looking at the figure that you posted, I think that you can do a structured mesh using 3D extrusion, doing a mesh for the diffuser, a mesh for the room and then gluing at the interface.

Regards,

Luciano

[clalgourdin]

Hello Luciano,

This is a good remark, I have changed the meshing netgen 3D-2D-1D for the diffuser and hexahedric (hexahedron i,j,k) mesh for the room and assembled it together, I think this is the best way, but the junction is bad and my simulation diverge (conjugate gradient).

I would like to know, if there is a step in Salome to improve the connection between the two meshs ?

Best regards

Claude

[Yvan Fournier]

Hello,

Did you try the non-conformal to conformal joining of Code_Saturne ?

Regards,

Yvan

[Luciano Garelli]

Hello Claude,

What do you refer with "the junction is bad"? Are you doing the face joining in salome (build compound->merge coincident nodes and elements) or in CS (Face joining)??

I don't know how is your geometry, but you can use projection 1D2D in Salome in order to have the same mesh at the interface between both mesh. Also, you have to check in the CS listing the joining information in order to know if all the face were joined.

Regards,

Luciano

[clalgourdin]

Hello Luciano and Yvan,

I tried to do the meshs junction properly with a simple case between 2 cubes.

Firstly, I tried with hexa meshs, and it's work correctly, when I make the junction in C_S or in Salome (build compound).

Now, I'm trying to make the same thing but with complex meshs (NETGEN).

I have successfully made the meshing junction (conformal mesh) thanks to the function projection 1D2D with a sub mesh.

After that, when I am "checking the mesh" in C_S, I have the following mistake "bad connectivity for the cells" (see attached file in French), with a mesh junction in Salome.

I also tried with two meshs with the junction in C_S, but I have also a mistake, with the error in the listing file (attached) : "420 faces have a null distance between centers, For these faces, the weight is set to 0.5"

Thank you for your help.

Claude