[windmill]

Hello all

I'm modelling a wind tunnel which has a converging-diverging nozzle upstream of the test section. At the moment, the flow almost reaches the target of Mach 2, but what really stumps me is that it becomes sonic not at the throat but a couple of cm downstream.

I've attached some screenshots, as well as the report from CFX-post. If anyone has any ideas I would be grateful for your support.

Thanks
Max

[diamondx]

i think you need a more refined mesh near the wall. can we take a look at your mesh

[ghorrocks]

http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

[windmill]

Here is my mesh. I didn't do anything fancy; just imported my geometry and hit "generate mesh".

I did try improving the mesh for the throat area (see edited_mesh), but got a zero divide error in the first iteration (even with a ridiculously small timestep). That being said, I wasn't entirely sure what I was doing when I redid the mesh and just went with the first improvement I got. I'm currently trying to refine the mesh, keeping the quadrilateral elements. I'm also going to set the RMS residual to below 1e-05.

[Far]

Here is my mesh. I didn't do anything fancy; just imported my geometry and hit "generate mesh".

Thats not the good idea for supersonic simulation. Dont make any fancy mesh, but at least resolve it in stream wise, radially and boundary layer. Specially you need good refinement in the throat and downstream.

[windmill]

Originally Posted by Far

Dont make any fancy mesh, but at least resolve it in stream wise, radially and boundary layer.

I don't quite understand what you mean there. Could you expand on that a bit please?

[diamondx]

due to viscosity, velocity is always equal to zero a the wall (boundary layer), but not like in you picture.that layer should be very small. the more you get to the center of your pipe, the velocity get higher. in order to catch that boundary layer, you need a good resolution near to wall, you have to refine. how about an hexa mesh, you geometry is not that difficult

[windmill]

I tried a hexa mesh (or as close as it would give me) and the boundary layer was doing strange things. I went back to a quad mesh and made it as refined as it would let me without breaking/stopping due to license restricions (5mm - attached).

Now I'm getting close to a M2.5 shock in the tunnel, which is NOT supposed to be happening. I'm not sure what to do on the physics side. Setting a supersonic outlet doesn't work, putting a pressure on the inlet (as opposed to normal velocity which I'm doing now) doesn't work.....

[Far]

try ICEM CFD HEXA

[Far]

we can help you out to get the decent mesh