## Fluids

#### Turbulence quantities for RANS and LES

• aitor.amatriain
Subscriber

To whom it may concern,

I am implementing a custom model in ANSYS Fluent and it includes a source term. However, this source term (including some adjusted coefficients) have been deduced from LES simulations, and depend on turbulence quantities as turbulent kinetic energy and turbulent viscosity

I know that k_RANS > k_LES (k_LES=k_SGS) and mu_t_RANS>mu_t_LES. However, is there any reference that gives an estimation of the difference between these values? An order of magnitude will be enough. Of course, I imagine that the possible comparison would depend on the size of the grid of LES simulations.

Thank you in advance,

Aitor

• DrAmine
Ansys Employee
I am not sure If I understood it correctly but you might calculate the Pope Criterion which is nothing else bit the ration of modelled K.E of SGS Eddies to the sum of all Kinetic Energy of all eddies. Pope suggests the ratio to be less than 0.2 to resolve 80% of the total kinetic Energy.
• aitor.amatriain
Subscriber
Thank you for your response.
I see that PC = k_SGS/(k_SGS+k_RES)=k_SGS/k. However, I am not performing LES simulations.
Let's say that you have a UDS equation with no diffusion and a source term equal to k_SGS. Let's also assume that k_SGS is such that PC = 0.8.
Now let's move to RANS in order to solve the mentioned equation. The unsteady and convective terms are equal, but what about k_SGS? If I put k_RANS=k_SGS, then my source term is much higher than the one in LES (as k_RANS >k_SGS)and the results are not correct. Could it be approximated as k_RANS = 0.8*k_SGS?
Thank you

• DrAmine
Ansys Employee
I think it is better if you share here the original problem say whatever it's deduced from LES. I am sure that one cannot make a one to one correlation but you might correlate the influence of the non-resolved part on the whole part and from that you might extrapolate what happens if you do not resolve any thing and you model everything (RANS).
• aitor.amatriain
Subscriber
As the correlation is not one to one, may uncertainties arise and I do not think that this is the right model to use.
Thank you for your help