k-epsilon model doesn't shows vortex

najunaju · February 26, 2013, 05:17

Dear all,

I am running bluff body simulation in OpenFoam with various turbulence models. I used both kwsst and les 2D simulation and found after 2.5 sec of simulation it starts forming vortex after the body. drag and lift values are also okay.

But with same mesh and BC, for k-epsiln and k-w turbulence model, I don't know why but it doesn't shows any vortex. I run the simulations for long time, it didn't produce any vortexes after the body.

Does any body have any idea why is it happening? Please help me, your suggesting will expedite my research and I will be grateful to you.

Sincerely,
Naju.

Lieven · February 26, 2013, 09:21

Hi Naju,

It makes perfectly sense that you don't find vortices in the wake of the body when using the k-e and k-w model. With RANS, the effect of such vortices on the flow is fully modelled and translated into an additional turbulent viscosity term in the momentum equation (= additional momentum dissipation). Because of this term, all vortices you would expect are (should be) damped out.
This might sound strange but this in fact the whole purpose of RANS: you model the fluctuating component of the flow, regardless the scale of the fluctuations, and solve for the reynolds averaged components. So the result you get is (an approximation of) the time-averaged flow field.

In LES on the other hand, only the vortices/fluctuations below grid scale are modelled. Therefore, the added viscosity in the momentum equation is relatively small and does not damp out all vortices, hence you find a transient solution.

So first, based on this explanation, and in my modest opinion, it is not fully correct that you find vortices in case of k-w sst.
And second, 2D-LES does not make sense. Turbulence is by definition a 3D effect so I wouldn't trust those results...

Cheers,

L

najunaju · February 26, 2013, 20:55

Dear Lieven,

Greetings!

Thank you very much for your suggestion. Actually, I am new user of OpenFOAM. Also my theoretical background is not so strong. I am studying all the basic theories. By this time I need your help. I read the code in OpenFOam of all these three models,

In kepsioln model eddy viscosity is modeled from cmu, k and, epsilon; in kw, eddy viscosity is modeled from, k and w; in kwsst, eddy viscosity is modeled from a1, k, and gradient(U).

But k, epsilon and w are calculated from mesh or somehow they are calculating, so I can't modify. Only way is to modify coefficient, cmu, a1, alphaK, alphaw, betaStar or sigmaEps etc.

Now, What is your suggestion, Do I need to change the mesh or coefficients to get the vortex? In case coefficient, which parameter should I modify?

Sorry to bother you.

Please, help.

wes1204 · March 9, 2013, 01:52

i think you should make your mesh more fine.

i had same problem several time.
I used pimplefoam and RANS(k-epsilon model).
before using RANS, I tried running case using LES(smagorisky) and it show me vortex.
so I modified this case to RANS in same mesh.
but the case calculated by RANS didn't show me vortex.
then I made mesh more fine in same condition (density of mesh was almost 2times).
finally, RANS model show me vortex.

Lieven · March 9, 2013, 13:41

The fact that the vortices appear is most likely due to unphysical wiggles caused by the discretization scheme. So they are not an accurate representation of the real vortices but the result of numerical instability because of large gradients near your object. You can simply test this by using an upwind scheme for all parameters in fvSchemes. I'm pretty this will make your vortices disappear, regardless the grid resolution.

You should use higher order schemes which can cope with this effect, such as Gamma or SFCD (not claiming that these are the best options, nor that the effect would disappear completely). And again, if you are simulating a steady situation with RANS, e.g. the flow through a channel with objects present, your result should be steady as well.

Cheers,

Lieven

thiagomarinho · July 5, 2023, 03:06

I think this is most likely to the discretization scheme. Keep in mind that laminar flows may also have vortices, ie: von karman street.

For such a case I started with a robust and stable discretization scheme and after stabilizing the simulation I switched to a more acurate one with good results. If you need the particulars I'll need to dig through my notes.

February 26, 2013, 05:17	k-epsilon model doesn't shows vortex	#1
najunaju New Member Md. Naimul Haque Join Date: Jan 2013 Posts: 14 Rep Power: 13	Dear all, I am running bluff body simulation in OpenFoam with various turbulence models. I used both kwsst and les 2D simulation and found after 2.5 sec of simulation it starts forming vortex after the body. drag and lift values are also okay. But with same mesh and BC, for k-epsiln and k-w turbulence model, I don't know why but it doesn't shows any vortex. I run the simulations for long time, it didn't produce any vortexes after the body. Does any body have any idea why is it happening? Please help me, your suggesting will expedite my research and I will be grateful to you. Sincerely, Naju.

February 26, 2013, 09:21		#2
Lieven Senior Member Lieven Join Date: Dec 2011 Location: Leuven, Belgium Posts: 299 Rep Power: 22	Hi Naju, It makes perfectly sense that you don't find vortices in the wake of the body when using the k-e and k-w model. With RANS, the effect of such vortices on the flow is fully modelled and translated into an additional turbulent viscosity term in the momentum equation (= additional momentum dissipation). Because of this term, all vortices you would expect are (should be) damped out. This might sound strange but this in fact the whole purpose of RANS: you model the fluctuating component of the flow, regardless the scale of the fluctuations, and solve for the reynolds averaged components. So the result you get is (an approximation of) the time-averaged flow field. In LES on the other hand, only the vortices/fluctuations below grid scale are modelled. Therefore, the added viscosity in the momentum equation is relatively small and does not damp out all vortices, hence you find a transient solution. So first, based on this explanation, and in my modest opinion, it is not fully correct that you find vortices in case of k-w sst. And second, 2D-LES does not make sense. Turbulence is by definition a 3D effect so I wouldn't trust those results... Cheers, L iqbalsk8, najunaju, FrankFlow and 2 others like this.

March 9, 2013, 01:52	mesh problem	#4
wes1204 New Member Baek, Donghae Join Date: Jan 2013 Location: Seoul Posts: 24 Rep Power: 13	i think you should make your mesh more fine. i had same problem several time. I used pimplefoam and RANS(k-epsilon model). before using RANS, I tried running case using LES(smagorisky) and it show me vortex. so I modified this case to RANS in same mesh. but the case calculated by RANS didn't show me vortex. then I made mesh more fine in same condition (density of mesh was almost 2times). finally, RANS model show me vortex. najunaju likes this.

March 9, 2013, 13:41		#5
Lieven Senior Member Lieven Join Date: Dec 2011 Location: Leuven, Belgium Posts: 299 Rep Power: 22	The fact that the vortices appear is most likely due to unphysical wiggles caused by the discretization scheme. So they are not an accurate representation of the real vortices but the result of numerical instability because of large gradients near your object. You can simply test this by using an upwind scheme for all parameters in fvSchemes. I'm pretty this will make your vortices disappear, regardless the grid resolution. You should use higher order schemes which can cope with this effect, such as Gamma or SFCD (not claiming that these are the best options, nor that the effect would disappear completely). And again, if you are simulating a steady situation with RANS, e.g. the flow through a channel with objects present, your result should be steady as well. Cheers, Lieven babakflame, najunaju and FrankFlow like this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
BC for epsilon in NaganoTagawa k-epsilon model	nikesh	OpenFOAM Pre-Processing	0	February 16, 2013 08:15
k and epsilon were hard to converge in multiphase model of Fluent	Yanlong Li	ANSYS	0	January 2, 2013 05:25
k and epsilon discretization of RNG model	mehran	Fidelity CFD	0	January 24, 2009 00:01
How to model vortex of supersonic wing tip	Bernard	FLUENT	0	January 29, 2007 03:20
DPM model w/ Wave model - errors in documentation	HS	FLUENT	0	April 12, 2006 04:37

February 26, 2013, 20:55		#3
najunaju New Member Md. Naimul Haque Join Date: Jan 2013 Posts: 14 Rep Power: 13	Dear Lieven, Greetings! Thank you very much for your suggestion. Actually, I am new user of OpenFOAM. Also my theoretical background is not so strong. I am studying all the basic theories. By this time I need your help. I read the code in OpenFOam of all these three models, In kepsioln model eddy viscosity is modeled from cmu, k and, epsilon; in kw, eddy viscosity is modeled from, k and w; in kwsst, eddy viscosity is modeled from a1, k, and gradient(U). But k, epsilon and w are calculated from mesh or somehow they are calculating, so I can't modify. Only way is to modify coefficient, cmu, a1, alphaK, alphaw, betaStar or sigmaEps etc. Now, What is your suggestion, Do I need to change the mesh or coefficients to get the vortex? In case coefficient, which parameter should I modify? Sorry to bother you. Please, help.

July 5, 2023, 03:06		#6
thiagomarinho New Member Thiago Marinho Join Date: Nov 2014 Posts: 8 Rep Power: 11	I think this is most likely to the discretization scheme. Keep in mind that laminar flows may also have vortices, ie: von karman street. For such a case I started with a robust and stable discretization scheme and after stabilizing the simulation I switched to a more acurate one with good results. If you need the particulars I'll need to dig through my notes.