[ancsa]

As I found no solution on the forum to the strange behaviour of the NonlinearKEShih turbulence model I tried to compare the results for a boundary layer flow for which the analytical components or the Reynolds stress tensor can be derived and thus compared.

The analytical solution in my opinion for a boundary layer where grad(U) is only dU/dz is attached as nonlinearStressforBL.gif. The test case is also attached, together with the results plotted.

The inlet was defined by timeVaryingMappedFixedValue to maintain the flow, simpleFoam is used for solution, coded functions in controlDict to obtain fvc::grad(U) and R utility to obtain the Reynolds stress tensor, changing constant/RASProperties from kEpslion to NonlinearKEShih. The inlet values are from wind tunnel measurents.

What I see is
- R_{11} and R_{33} seem to be swapped

I think the problem is that fvc::grad(U) is transposed. I.e. dU/dz appeares in the 10th column of the sample file, instead of the 6th. Using transposed grad(U) exactly results analytically in swapping those terms.

- R_{22} is identical to the linear version which it should not be

The missing dirac terms of the implementation seem to cause the problem as again the analytical result without them is R_{22nonlin} = R_{22lin}.


Maybe there is something I did not take into account and the implementations are correct, I would be grateful if someone could point me to that direction as now this model seems to perform badly for the simpler flow I could think of.

Thanks
Aniko

[ingcorra]

You cannot believe how much you helped me with this post. THE VELOCITY GRADIENT IS TRANSPOSED. (and I even knew it)

A simple minus sign in front of the vorticity tensor and finally the turbulence model I was struggling to make work since a pair of months works flawlessly.

THANK YOU SO MUCH.