[selvam2487]

Dear Friends,

I am investigating fluid mixing in T junction flows using LES in ANSYS CFX. I have a problem with my simulation. I wanted to do my LES in central differencing scheme. But the solution I get is very unrealistic. These are my input flow conditions

Hot inlet:Temperature - 393 K,mass flow - 400 g/s
Cold inlet:Temperature - 298 K,mass flow - 100 g/s

But the temperature range is 200 K - 450 K in the output result file. I have checked almost everything that I know of to find the source of error. But still central differencing scheme does not work for my LES. The mesh quality parameters are as follows:

Min. Angle - 30°
Aspect ration - 10
volume change - 5

I have attached some pictures which show the inputs that I gave for my simulations in the dropbox link below. Kindly have a look at it and please help me to correct my mistake. Thank you for your help in advance.

https://www.dropbox.com/sh/fai92g0od...f/MTv1KYjnWp#/

P.S: Even if I use Bounded Central difference scheme, the unrealistic temperatures remain the same.


Regards,
Karthick Selvam.

[ghorrocks]

What happens when you run a well bounded scheme like high resolution or upwinding?

[selvam2487]

Dear Glenn,

When using upwind scheme, the temperature is within limits and the results are ok. The same goes for high resolution scheme too. The results are physical. But when performing the LES using central differencing scheme, the results become very unphysical.

Regards,
Karthick

[flotus1]

Does your mesh contain non-Hexahedral cells?
With all commercial CFD solvers I used so far, central differences for LES could only be used for hexahedral meshes.
With tet or poly cells, the results were always unrealistic.

[ghorrocks]

OK, so it looks like you have a boundedness problem. Alex's comment is important - mesh quality is vital.

But also have a look at your results using the bounded schemes like high resolution. If the results are reasonable (ie turbulence spectrums look OK, global results are reasonable) then you can consider using it.

[selvam2487]

Dear Glenn and Alex,

I am using a hexahedral mesh for my LES. I have attached the pictures of mesh type and mesh quality parameters. I have searched for the mistakes that you mentioned in my mesh (like hexahedral elements, mesh quality etc.) and I found that they are within reasonable limits. I am also attaching the link for the blocking of my T junction so that you can have a look at my mesh quality and detect any obvious or other mistakes that I did not look carefully into.

https://www.dropbox.com/s/szf9j5r4n8...%20blocking.7z

Thank you.
Regards,
Karthick

[ghorrocks]

"Reasonable" limits for some simulations might not be reasonable for this application. I suspect you will find this case is much more sensitivie to mesh quality than normal, so your "reasonable limits" are not applicable.

I woudl recommend you do a study of mesh quality versus simulation quality. Generate a dummy geometry with similar physics which is easy to mesh with high quality elements (preferably perfectly orthogonal hexes) Then produce a series of meshes, starting from perfectly orthogonal/aspect ratio=1 and stepping down in mesh quality by deliberately distorting the mesh. I bet if you do the same simulation on this series of meshes that it runs fine on the perfectly orthogonal/aspect ratio=1 mesh and you will fine boundedness errors start increasing as the mesh quality decreases. And I bet you will find the error is unacceptable well before the "reasonable limits" you have quoted.

This test will take a bit of time to set up and run, but it will be very valuable for your simulation and you will know the mesh quality you need and I bet you will learn a lot about CFD in the process. Sounds worthwhile to me.

[selvam2487]

Dear Glenn,

Thanks for the information. I will do the mesh quality study which you have mentioned and will observe what happens with decreasing quality of the mesh.

Regards,
Karthick

[enayath]

Any update selvam2487?

Could you figure it out? what was the problem?