CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   Possible turbulence modelling bug in SRF solvers (http://www.cfd-online.com/Forums/openfoam-solving/100607-possible-turbulence-modelling-bug-srf-solvers.html)

otm April 27, 2012 09:14

Possible turbulence modelling bug in SRF solvers
 
Hi!

I'm working a lot with development of machines that rotates. Hence, I need to use the SRF solvers provided in OF (have not had time to use the MRF solvers yet).

My issue is that I'm not sure that the effects of system rotation, on the turbulence are included properly in the SRFSimpleFoam and SRFPimpleFoam solvers. As far as I can see the turbulence model is using the "Urel" field to calculate the tensors used in the modelling (but I should admit that I often have a hard time tracking down what the code does since there are so many different source files involved). In a linear eddy viscosity model this doesn't matter because only the strain rate tensor ((dUi/dxj+dUj/dxi)/2), which is invariant to rotation is used in the modelling of the Reynolds stresses. However in e.g. the LRR model the rotation rate tensor ((dUi/dxj-dUj/dxi)/2), which is not invariant to rotation is used.


A good test case for turbulence models in rotating frames is the fully developed flow in a pipe rotating around its axis. The flow is three dimensional but still very simple. The azimuthal velocity component (in the stationary frame) should have a parabolic shaped profile in the radial direction if the Reynolds stresses are predicted correctly. With a linear eddy viscosity model the azimuthal velocity will correspond to solid body rotation (i.e. a straight line).

I have tested the LRR model with the SRFSimpleFoam solver in the rotating pipe and did not get the correct profile. So it seems as if there is a bug (or at least an inconsistent method) in the SRF solvers.

Next i will run the pipe case in a stationary frame (with a moving wall). If I can get the right profiles in this case with the LRR model, I think it can be concluded that the implementation in the SRF solvers is wrong.

I welcome input from other users on this issue, if you agree with me or not. And if I have misunderstood something, please inform me.

I'm running OF-2.1.x.

BR
Olle

alberto April 28, 2012 02:12

As a suggestion, if the problem is confirmed, please report it on mantis: http://www.openfoam.org/mantisbt/main_page.php

Thanks,

vkrastev April 30, 2012 05:52

That's quite interesting...Any news about this issue? From a rapid look on Mantis' page it seems it hasn't been reported yet

Regards

V.

otm May 29, 2012 04:03

Update:

I still think that it could be an inadequate implementation of the turbulence models in the SRF-solvers. However, the way I wanted to show this did not work out. The reason is that in order to correctly predict the parabolic profile of the azimuthal velocity in the rotating pipe one needs to have a model for the inter-component transfer between the Reynolds stresses i.e. use a RSTM or EARSM. Unfortunately it turns out that the Reynolds stress models implemented in OF (LRR and Launder Gibson) are based on the simplified "isotropization of production" model (often referred to as IP-LRR) for the inter-component transfer, and this model is known to be unable of predicting the azimuthal velocity profile correctly.

So, the possible bug may not have a big influence until there are more advanced models available in OF.

BR
Olle


All times are GMT -4. The time now is 21:33.