[derHandballer73]

Hi all,

I'm currently trying to model a rotating channel flow using FLUENT 17.1. The channel has the size 125.66x20x62.8 (x,y,z) in mm. First of all, i want to get the non rotating case in Agreement with the investigations of Grundestam, Wallin and Johannsson ("Direct numerical simulations of rotating turbulent channel flow")
The boundary conditions for Inlet/Outlet and the side-walls are periodic. To maintain the flow I set a pressure Gradient in x-direction with the value -8.47 Pa/m for the periodic condition. I used a RANS-Simulation with the Reynolds Stress Model (default settings) and the pressure-based solver. As material I used the Fluent Default air properties.

So now I have the following Problem with my results:

If I use a coarse mesh, I will get a x-velocity-Profile at the outlet of my channel looking very similar to the results in the literature.

when i use the mesh, which is used in the literature (way more refined than the coarse mesh), the Profile looks different an the values are lower. I attached some Pictures of the profiles.

I thought there have to be nearly the same results, or even better results with the fine mesh. The Setup is identical for both simulations. Both simulations are using the simple scheme and FOU for TKE TDR and Reynolds Stresses. SOU is used for Energy, Momentum and pressure.

What could be the reason for this behaviour?


Your help is much appriciated!

Regards,

[LuckyTran]

Lower is an understandment... You have 5 m/s on the coarse mesh and 0.25 m/s on the fine mesh. The difference is more than an order of magnitude. Which one is closer to the correct answer?


I'm just going to say you should completely go over the setup again on whichever one is producing the most incorrect result. Probably you messed up a boundary condition or something. There's no reason for the results to be that different.

[derHandballer73]

Hi LuckyTran,

thanks for replying.


The coarse mesh is very close to the literature.

To be honest, I expected this answer. So I have asked a friend to compare the two setups and he also doesn´t found any differences.

[Alisa_W]

What's your simulation type? steady or unsteady？ If it is steady simulation ,you can use SRF model. I have tried it and got the right curve.

Originally Posted by derHandballer73

Hi LuckyTran,

thanks for replying.


The coarse mesh is very close to the literature.

To be honest, I expected this answer. So I have asked a friend to compare the two setups and he also doesn´t found any differences.

[cfdnewb123]

Originally Posted by Alisa_W

What's your simulation type? steady or unsteady？ If it is steady simulation ,you can use SRF model. I have tried it and got the right curve.

May I ask how you get the right curve when rotation is introduced? My solver works fine when rotation is zero but when I introduced rotation, the asymmetrical profile is quite off and less asymmetry from the DNS result. May I clarify how you set up the rotational motion?
For rotation, I introduce Omega=20 rad/s to both frame motion and wall (y=-h or h) at absolute frame. Is this how you did it too?