[moham]

Hi everyone,
I have a question about computing turbulence kinetic energy in fluent 16 when you simulate a problem by using LES model.
I did a simulation using LES and obtained average flow by enabling data sampling for time statistics option and now I want to compute turbulence kinetic energy.there are some terms, RMSE x velocity, RMSE y velocity, RMSE z velocity in fluent and my question is related to these terms.how can i compute TKE by using these terms?does the below relation correct?

k=0.5*((RMSE x velocity)^2+(RMSE y velocity)^2+(RMSE z velocity)^2)

please help me in this way.

Thanks a lot.

[LuckyTran]

Yes that works, but not 100%.

I think the RMS quantities are only the resolved parts and you still need to add the contribution of the filtered subgrid scales to completely capture the total TKE. If you do a search of this forum you might find this answer in an old discussion.

Btw, it's RMS and not RMSE.

[moham]

thanks Lucky. I will do your suggestion.

[moham]

sorry Lucky,

I searched between the forums but i couldn't find any thing. may you help me how compute contribution of the filtered subgrid scales in Fluent?

[Andrea1984]

Hi Moham,

you can estimate the SGS model contribution to the total k. If you are using the Dynamic Kinetic Energy Transport SGS model it is straightforward, since you are solving a transport for k_sgs.

For all the other eddy-viscosity SGS closures, I usually evaluate k_sgs from the SGS viscosity and lenght-scale as

k_sgs = (nu_sgs/(rho*l_sgs))^2

where nu_sgs is the sgs dynamic viscosity, rho is the local fluid density and l_sgs is the subgrid lengthscale. If you are using the Smagorinsky model, l_sgs is given by

l_sgs = min(kappa*d, C_s*delta)

where kappa is the von Karman constant, d is the wall distance, C_s is the Smagorinsky constant and delta is the filter size, which in Fluent is equal to the cubic root of the cell volume. If you are using a different SGS model, then you have to use the relevant expression for l_sgs.

Hope this helps,
Andrea

[moham]

thanks a lot Andrea for your good explanation.

[marialhm]

Originally Posted by Andrea1984

Hi Moham,

you can estimate the SGS model contribution to the total k. If you are using the Dynamic Kinetic Energy Transport SGS model it is straightforward, since you are solving a transport for k_sgs.

For all the other eddy-viscosity SGS closures, I usually evaluate k_sgs from the SGS viscosity and lenght-scale as

k_sgs = (nu_sgs/(rho*l_sgs))^2

where nu_sgs is the sgs dynamic viscosity, rho is the local fluid density and l_sgs is the subgrid lengthscale. If you are using the Smagorinsky model, l_sgs is given by

l_sgs = min(kappa*d, C_s*delta)

where kappa is the von Karman constant, d is the wall distance, C_s is the Smagorinsky constant and delta is the filter size, which in Fluent is equal to the cubic root of the cell volume. If you are using a different SGS model, then you have to use the relevant expression for l_sgs.

Hope this helps,
Andrea

Hi, Andrea, do you know how to use this equations in UDF, because I need to use turbulent kinetic energy k and its dissipation epsilon, do you know how to use them in UDF by LES calculation?

Thanks.

Maria

[Andrea1984]

Hi Maria,

from my experience there is no need to use UDFs for the simple expressions in my post above. You can implement them easily via Custom Field Functions (CFF) in Fluent. The only glitch may be the logical expression "min" in the expression of l_sgs for the Smagorinsky model - you might need to use a UDF for that particular equation if it is necessary. If you are not resolving the near-wall region (e.g. you are using the werner-wengle wall functions) you can estimate l_sgs as C_s*delta, drop the logical operator in the equation and use a CFF for l_sgs as well

Andrea

[marialhm]

Originally Posted by Andrea1984

Hi Maria,

from my experience there is no need to use UDFs for the simple expressions in my post above. You can implement them easily via Custom Field Functions (CFF) in Fluent. The only glitch may be the logical expression "min" in the expression of l_sgs for the Smagorinsky model - you might need to use a UDF for that particular equation if it is necessary. If you are not resolving the near-wall region (e.g. you are using the werner-wengle wall functions) you can estimate l_sgs as C_s*delta, drop the logical operator in the equation and use a CFF for l_sgs as well

Andrea

OK, many thanks. I will try again.

[enayath]

Hello Everybody,

Did anyone of you figure it out what RMSE is exactly?
Does it fully cover the TKE for the resolved scales?

Thank you,
Hooman

[LuckyTran]

Originally Posted by enayath

Hello Everybody,

Did anyone of you figure it out what RMSE is exactly?
Does it fully cover the TKE for the resolved scales?

Thank you,
Hooman

The RMS velocity only gives you the filtered velocities so you will end up calculating only the resolved part of the turbulent kinetic energy.

You need to add the subgrid-scale kinetic energy to get the total turbulent kinetic energy.

[enayath]

Thank you Lucky,

Yeah...I was looking for the resolved part of TKE...I just need to know RMSE is kinda equal to RMS (in the resolved scales) as in the Fluent user guide, it is written it is the RMS error...not sure what does that mean... In other words, I need to make sure that RMSE can be used to define the resolved part of TKE by the following equation:
TKE , resolved=0.5*(RMSEx*RMSEx+RMSEy*RMSEy+RMSEz*RMSEz)

Thank you,
Hooman

[LuckyTran]

Originally Posted by enayath

Thank you Lucky,

Yeah...I was looking for the resolved part of TKE...I just need to know RMSE is kinda equal to RMS (in the resolved scales) as in the Fluent user guide, it is written it is the RMS error...not sure what does that mean... In other words, I need to make sure that RMSE can be used to define the resolved part of TKE by the following equation:
TKE , resolved=0.5*(RMSEx*RMSEx+RMSEy*RMSEy+RMSEz*RMSEz)

Thank you,
Hooman

The answer is yes, it does!

The confusion is whether you are talking about velocity or velocity fluctuations. I was always confused. The mean velocity is non-zero. The RMS of velocity will be biased by the mean velocity. The RMSE of the velocity is actually the RMS of the velocity fluctuation.

When you say turbulent kinetic energy, you are usually referring to the RMS of the fluctuations only, which is equivalent to RMSE of velocity.

So yes, the RMSE of the resolved velocities is what you need to calculate the resolved part of the turbulent kinetic energy. Usually the question is whether the fluent variables are only the resolved part or the total because they are not well documented and RMS is much easier to understand. I wasn't paying attention to RMS vs RMSE in my responses.

[FoxInCFD]

I've been searching in this forum for answers about this and I want to recommend this thread and add some details.

Firstly I want to stress that so far, Fluent 2021 R1 still doesn't compute k_sgs in Smagorinsky, WALE, WMLES and WMLES S-Ω. Only the Dynamic Kinetic Energy Subgrid-Scale Model computes k_sgs.

The version of Fluent 2022 R1 has a beta feature that allow you to estimate the LES quality by estimating the subgrid turbulent kinetic energy as
k_sgs = (µ_tSGS/(0.3*ρ))*S
Where µ_tSGS is subgrid-scale eddy viscosity, rho is density and S is strainrate magnitude.

Furthermore, in OpenFoam this field is available in WALE and Smagorinsky models.
In the Smagorinsky model they compute k_sgs by calculating the solution of a quadratic equation https://www.openfoam.com/documentati...agorinsky.html



In the WALE model a different formula is applied and can be checked in the source code of WALE.c https://www.openfoam.com/documentati...8C_source.html

Therefore, if you want to do the same in Fluent, you will need to implement a UDF to return the subgrid kinetic energy.
If anyone ever implemented this UDF I think it would be very useful for anyone using Fluent and using the available subgrid-scale models.

Originally Posted by Andrea1984

Hi Moham,

you can estimate the SGS model contribution to the total k. If you are using the Dynamic Kinetic Energy Transport SGS model it is straightforward, since you are solving a transport for k_sgs.

For all the other eddy-viscosity SGS closures, I usually evaluate k_sgs from the SGS viscosity and lenght-scale as

k_sgs = (nu_sgs/(rho*l_sgs))^2

where nu_sgs is the sgs dynamic viscosity, rho is the local fluid density and l_sgs is the subgrid lengthscale. If you are using the Smagorinsky model, l_sgs is given by

l_sgs = min(kappa*d, C_s*delta)

where kappa is the von Karman constant, d is the wall distance, C_s is the Smagorinsky constant and delta is the filter size, which in Fluent is equal to the cubic root of the cell volume. If you are using a different SGS model, then you have to use the relevant expression for l_sgs.

Hope this helps,
Andrea