[LuckyGuy]

Cheers!

I‘m currently performing a study of a nozzle in CFX, which I unfortunately can not share. With applying the orthogonality smoother i’ve ensured that my first cell is perpendicular to the surface. However with my approach the other boundary layer cells are not exactly perpendicular to the surface. Because I didn’t find anything specific for the boundary layer in the solver-theory and in some related papers I’m assuming that this is ok, and the solver still resolves the BL good. Because I’m not sure I wanted to ask if anyone has experience with this or can share me a best practice guide for nozzles (even for the meshing), so I can clear things out.

Have a good day!

Edit: This Video shows pretty accurately what my mesh looks like till 6:10 (https://youtu.be/wfzomnq_Pes)

[ghorrocks]

Different simulations have different requirements for the quality of the mesh. I suspect you are doing a compressible simulation - if there are shock waves present they require a high quality mesh to converge and be accurate (higher quality than the recommendations in the CFX documentation).

So why not work out for yourself if it is a problem? Do a simulation with this mesh, one which is much worse and one which is much better. Compare the results and see if it makes a difference in your case.

[LuckyGuy]

Thanks for your reply!
I thought about that aswell, but my main limitations are that I currently cannot work out a better mesh in the boundary layer in my limited time. The only thing what would come into my mind is that I create a tetra mesh with prism layers, where the layers are perpendicular to the surface. The comparison of these both meshes also has to take into account the different mesh types, which could be difficult if different results are obtained with both meshes.
I wanted to clear the theory behind it, before I invest my time which could be wasted, but I’ll give it a go.

Do you have maybe any recommendations for best practice meshing in a nozzle? I’ve looked through various of papers and guides but did not find anything for heat transfer in a nozzle where the meshing process is described.

[ghorrocks]

You can get shock waves in internal flow as well. It does not have to be an exhaust, you just need a sufficient pressure drop across the throat.

Yes, comparing to a tet/prism mesh sounds like a good idea.

The CFX documentation has recommendations for boundary layer meshing. My comments beyond that are to do a mesh sensitivity study to work out what you actually need for the accuracy you require.

Note that boundary layer modelling of heat transfer is challenging and you might have problems getting accurate answers. This is an inherent problem in the turbulence models.

[LuckyGuy]

Thanks for your advice, I’ll create a hybrid mesh and see if Im able to see any differences.

[Opaque]

If your mesh was created as described in the video, you should be fine and your only way to know if it is any good is doing a mesh sensitivity analysis, i.e. either coarse/refine the mesh further and compare the results.

Keep in mind what "characteristic variable" you want to measure for accuracy, i.e. flow variable or heat transfer variable.

Changing the mesh type, or topology is not a good idea as a mesh sensitivity step. Keep the mesh topology fixed when coarsening/refining the mesh. When you do the comparisons for different meshes you want to account for the spacing differences ONLY. Otherwise, you will have too many parameters to analyze.

[LuckyGuy]

Thanks for the input Opaque! According to the CFX documentation I should have around 10-15 BL cells which are normal to my surface.

Are there any techniques which I could use?

[Opaque]

Questions:
1 - Have you run the simulation with the meshes you got?
2 - Does the simulation converge?
3 - Did you encounter any issues with the initial solutions?
4 - If you refine/coarse the mesh, does the solution changes too much/little?
5 - Have you looked at the turbulence wall parameter requirements for the results, i.e. y+? Is it within the requirements for the model you selected?

Once your solution is no longer sensitive to your mesh, does it compare well with the benchmark data, experimental or analytical?

The meshing guidelines are that, guidelines, not a "law of meshing"; therefore, take them to guide you to set up something useful and later refine the mesh based on your problem requirements.

[LuckyGuy]

Sorry for the late response, but I didn't have time lately.
I'll try to analyze the questions, hopefully this week, and will answer them, I hope this is ok for you.