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/)
-   -   epsilon and omega free stream BC's (http://www.cfd-online.com/Forums/openfoam-solving/118266-epsilon-omega-free-stream-bcs.html)

immortality May 23, 2013 15:45

epsilon and omega free stream BC's
 
in this page:
http://www.cfd-online.com/Wiki/Turbu...ary_conditions
there are different formulas for epsilon and omega,I want to test with both kOmegaSST and k-epsilon-realizable(does anyone knows how to set it in OF?)
how can find out which is appropriate for OF with your experience?

for epsilon we have:
\epsilon = C_\mu \, \frac{k^\frac{3}{2}}{l}
or
\epsilon = C_\mu^\frac{3}{4} \, \frac{k^\frac{3}{2}}{l}
or
\epsilon = C_\mu \, \frac{\rho \, k^2}{\mu} \, (\frac{\mu_t}{\mu})^{-1}

and for omega we have:
\omega = \frac{\sqrt{k}}{l}
or
\omega = C_\mu^{-\frac{1}{4}} \, \frac{\sqrt{k}}{l}
or
\omega = \frac{\rho \, k}{\mu} \, (\frac{\mu_t}{\mu})^{-1}

immortality May 25, 2013 15:12

Hi again
I saw in page 25 of this thesis that he has used :\frac{3}{2} \, (l \, I \, U)^{2}(I managed this LaTeX formula at last!)

the thesis:
Simulation and validation of
compressible flow in nozzle geometries
and validation of OpenFOAM for this application

is this work and change in formula by adding l is acceptable?
http://s2.picofile.com/file/77781039...rbeit.pdf.html

wyldckat May 25, 2013 16:17

Hi Ehsan,

Quick note - here's another link to said thesis: http://www.rw.ethz.ch/alumni/thesis/...sterarbeit.pdf

From what I can figure out, the author of the thesis went along with the spirit of this whole turbulence stuff.
If you look at the first sections of the wiki page you've mentioned earlier, you'll see in "Modified turbulent viscosity" that the "turbulent length scale" is used for the "modified turbulent viscosity". This is possibly what influenced the author's decision.

As for the wiki, it does have some more helpful hints, namely the one about "Eddy viscosity ratio": http://www.cfd-online.com/Wiki/Eddy_viscosity_ratio
In essence, if you are able to calculate the "turbulent length scale", it's easier to calculate through it, instead of through the "Eddy viscosity ratio".

In case you haven't seen how to calculate the relevant parameters, here's a quote:
As for which calculation to use: my usual rule of thumb is to use the equations I can calculate :D

Best regards,
Bruno

Eloise September 6, 2013 08:49

Hi Ehsan,
I'm also interested in some free-stream turbulence problems. I run a simulation of an object in uniform flow and I would like to perform a sensitivity study on the inlet turbulence intensity level. I've tested several TI levels (3% to 7%), but in all cases the TI drops to about 1% a few cells after the inlet. So whatever TI level I put at the inlet, my object is always in a 1% TI level.
Have you experienced something similar with your simulations? If not, what is the way to maintain TI through the domain?
About my case: the velocity profile is steady and uniform, the mesh is uniform, I use k-omega SST turbulence model.
Best regards,
Elo´se

wyldckat September 7, 2013 07:47

Hi Elo´se,

A few questions regarding your case:
  1. Is it 2D or 3D?
  2. Have you first tested to see if it runs well in laminar flow?
  3. Is it a steady-state or transient simulation?
  4. The turbulence intensity and so on, is it similar throughout the whole domain or is it only near the object that it keeps the low turbulence values?
Best regards,
Bruno

immortality September 7, 2013 08:00

Quote:

Hi Ehsan,
I'm also interested in some free-stream turbulence problems. I run a simulation of an object in uniform flow and I would like to perform a sensitivity study on the inlet turbulence intensity level. I've tested several TI levels (3% to 7%), but in all cases the TI drops to about 1% a few cells after the inlet. So whatever TI level I put at the inlet, my object is always in a 1% TI level.
Have you experienced something similar with your simulations? If not, what is the way to maintain TI through the domain?
About my case: the velocity profile is steady and uniform, the mesh is uniform, I use k-omega SST turbulence model.
Best regards,
Elo´se
Hi Eloise
sorry for delay,how did you notice that TI in internal cells becomes low?

Eloise September 9, 2013 02:54

1 Attachment(s)
Quote:

Originally Posted by wyldckat (Post 450364)
  1. Is it 2D or 3D?
  2. Have you first tested to see if it runs well in laminar flow?
  3. Is it a steady-state or transient simulation?
  4. The turbulence intensity and so on, is it similar throughout the whole domain or is it only near the object that it keeps the low turbulence values?

Hi Bruno,

1. It is a 3D case
2. I ran the case in laminar and haven't had any specific problem. Do you think of something specific which should be checked in the laminar case? I'll have a look at the paper, thanks for the link.
3. It is a steady-state simulation
4. The TI decreases in the very first cells after the inlet. Around the object, the TI increases again. (see attached image, which is a cut of the domain) Ehsan: that's how I detected the issue.

Regards,
Elo´se

immortality September 9, 2013 05:30

How did you define turbulence intensity field?is there any postProcessing utility for that?

Eloise September 9, 2013 06:53

Quote:

Originally Posted by immortality (Post 450619)
How did you define turbulence intensity field?is there any postProcessing utility for that?

This page shows you how to get it from the k field. You can then use the "calculator" in paraView to define a new variable in your field.

Elo´se

immortality September 9, 2013 07:33

how you have defined TI in inlet?
you should set it to a constant so that can define k in inlet not inverse.

Eloise September 9, 2013 08:47

Quote:

Originally Posted by immortality (Post 450662)
how you have defined TI in inlet?
you should set it to a constant so that can define k in inlet not inverse.

I compute the k and omega values corresponding to the desired TI level and use the "fixedValue" for k and omega at the inlet. Do you think the "turbulentIntensityKineticEnergyInlet" for k and "turbulentMixingLengthFrequencyInlet" for omega makes any difference? (as described here)

Elo´se

immortality September 9, 2013 11:28

is velocity constant in inlet?if it varies during run and you have set a fixed k and omega it might makes some error,I don't know how you set your BC's,could you attach them?

wyldckat September 9, 2013 14:32

Greetings to all!

@Elo´se: Well, the only way to maintain the turbulence intensity is if you introduce small obstacles, such as wires or something like that.

Because there are a few details in play here and all of them are related to the solver doing what it was told to do. More specifically:
  • There's a lot of empty space between the inlet and the obstacle. It's natural that any and all vortices will fade away, given enough space.
    • Unless you inject with a very high speed and high turbulence intensity.
  • The "fvSchemes" might be using too much 1st order div schemes. Or in other words, if all div equations are solved with 1st order, you'll see a very quick smoothing of the results. This is why it usually converges and converges fast!
  • The RAS turbulence models are only representative of turbulence, they usually don't simulate the micro-vortices themselves. For that, you would need a much finer mesh + LES or DNS.

Which reminds me... have you ever seen those old fluid dynamics/mechanics videos showing how turbulence work and so on? I can't remember where I kept the link for them...
Anyway, one of them shows the effect of a wire inducing turbulence.

Best regards,
Bruno

Eloise September 10, 2013 03:38

Quote:

Originally Posted by immortality (Post 450693)
is velocity constant in inlet?if it varies during run and you have set a fixed k and omega it might makes some error,I don't know how you set your BC's,could you attach them?

Nothing fancy in my BC. Constant velocity, k and omega, and nothing varies in time. I guess it is the same as in the motorbike tutorial.

@Bruno: Thanks for the advices on fvSchemes. I'll have a look at that too. I guessed that if there is nothing to create velocity gradients and maintain turbulence, then the turbulence would fade out. Or at least that's how I understood my problem. I asked about it here because I was wondering if others observed the same and if something could be done about it :)


All times are GMT -4. The time now is 01:04.