[agustinvo]

Hi,


I recently wanted to test the Reynolds Stress models of OpenFOAM on different channel flows in 2D flows as I have done for V&V of eddy viscosity models.


My surprise came up when their performance was dissapointing. Velocity is underestimated by far! What I have also seen is that for 2D simulations, the Rzz stress tensor component is also resolved (while Uz is not). Do you think that could affect as well to the solution. Are these models only developed for 3D simulations?


Thank you in advance for your answer!





PS: please see in attachment the uPlus velocity for a Re_tau=395 channel flow

[HPE]

Please refer below for others to replicate your issue.

[agustinvo]

Hi,


I attach the test case to be run with simpleFoam (first please use blockMesh).


There is a python script to he run from the case folder

python3 python/plotUPlus.py

once the simulation is done.


https://drive.google.com/open?id=1D2...Bt7ptt3OfUkzuA

[HPE]

I think youve forgotten to attach it.

[agustinvo]

Thank you! Now it should be available via Google Drive

[HPE]

I havent run ur case yet; however, noticed that your y1+ is less than 1. LRR is, as far as I know, a high Reynolds number turbulence model, that is, below the onset of inertial layer is not formulated by LRR. You may need to go for a new grid where the first wall normal grid point is within the inertial sublayer, and to use a set of wall functions.

Also I still need to verify that your setup actually reflects ReTau=395, and your normalisation is correct. Sometimes, these become tricky even though look like trivial.

[agustinvo]

Hi,


indeed, the simulation was set with a low Reynolds k-epsilon model (that is why the 100000 folder), and I reached the desired Re_tau.


The first cell height has a y+=0.5.

[HPE]

Sorry didnt get the last entry of yours.

Is the problem due to LRR or low-Re kEpsilon (exactly which kEpsilon variant in OF since "kEpsilon" is high-Re itself)?

EDIT: Sorry. I dont know where LRR appeared. . I think I thought you are using a Reynolds Stress model (but eddy-viscosity model instead you used), thats why.. If you inform which kEpsilon variant you have used, it would be useful before downloading your case (no PC for few days, sorry).