CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   FLUENT (https://www.cfd-online.com/Forums/fluent/)
-   -   Asymmetrical solution using RANS (https://www.cfd-online.com/Forums/fluent/97988-asymmetrical-solution-using-rans.html)

 Marli February 29, 2012 19:15

Asymmetrical solution using RANS

1 Attachment(s)
Hi all, I was wondering if someone could help me with a re-occuring problem I'm having.

I'm modelling external flow around an aircraft wheel (closed rim) (meshed using Pointwise) using k-e RKE model with Enhanced Wall Function and Double Precision. I'm using a Y+=1 value against the wall for the wheel diameter of 1.4m, and reynolds number of 6.71x10^6, 70m/s inlet flow. A while ago I asked a question about the mesh, and in that thread you can see the basic geometry of the mesh:
http://www.cfd-online.com/Forums/mai...t-problem.html
The basic mesh hasn't changed too much from then, so you can get an insight into how I build the mesh. Apart from an initial hand built quarter-section, the rest of the mesh is build up using mirror-copies using the absolute x,y,z planes as mirror planes, as point 0,0,0 is at the centre of the wheel, so as far as I can see my wheel is perfectly aligned to the flow direction (aside from node location rounding variations).

Now to my confusion: My results seem reasonable, my Cp graphs show what I expected, Cl and Cd are also around what I expected. The computations are allowed to run for around 10,000 iterations, and by this time the Residuals have stabilized, with Continuity the highest at around 2.5x10^-4, with the others down to 1x10^-6. However, my results show asymmetry in the wake structures. The picture below shows the iso-surface of vorticity magitude=200, and here you can see how the trailing vortices are not symmetrical either side of the wheel. It does seem though that the top right and bottom left vortices are symmetrical to each other, and visa-versa, joining up to create like an S shape vortex covering the whole of the trailing side of the wheel. So one side vortices are inverted from the other side's.

I have been using a steady RANS setup, using the SIMPLEC pressure scheme, Least Squares Cell Based, and all other Second Order discretization schemes. My understanding is that a bluff body flow such as around this wheel would naturally be unsteady, but since I am using a steady solver, that the solution should be averaged out, and so I would expect a symmetrical solution.

The asymmetrical solution is for a 6-million cell grid (what I considered to be high density), and appears the same phenomena with courser grids, and also when introducing a ground proximity (as I am investigating the effect of wheel ground proximity), although the solution seems to be symmetrical when the wheel is actually in contact with the ground.

So, do you think my asymmetric solution is a correct one, maybe a product of tiny inaccuracies that cause the flow to favour one side of the wheel or another, or is there something I'm doing wrong with the mesh or fluent setup that's causing it? The 10,000 iterations I did seemed sufficient that the residuals and Cl,Cd values were staying constant or having small constant oscillations.

Any help would be much appreciated.

Luke

2 Attachment(s)
Dear Marli,

Thank you for sharing your experience and problem.
I am having a quite same problem as you had. I am just wondering if you have found any reasonable answer to your question that you can share here.

Here is a short explanation of my problem.
I have a 3D flow over a modified NACA airfoil. I have attached the surface mesh of the geometry and showed where I expect to see a symmetry. In the second picture, I have shown the velocity profile downstream the flow. As it can be seen, The solution is not symmetry.

I have been searching over this question for a couple of days but I was not able to come up with a good solution, but here is what I explain. As the flow in my geometry separates after a certain length from the leading edge, presence of the separation in my problem leads to asymmetrical solution. Does this make sense ?

I am using a steady-state solver and k-OmegaSST turbulent model and modeling the sublayer with wall functions. I have second order accurate FV schemes.

I would greatly appreciate any hint.
Thank you