[AliE]

Hello CFD friends,

Today I have for you a practical question on LES. I would like to implement some simple LES models (with no dynamic procedure) in my code (e.g. Smagorinsky or WALE).

After that the filtering procedure has been carried out, an equation formally similar to RANS comes out. The Reynolds stress tensor is approximated:



With modelled somehow. The eddy viscosity term can be associated to the diffusion term in the NS equation, while the diagonal term is assorbed by the pressure term.

Mathematically I am fine with this, but practically I have this question:

While I am solving numerically the problem, the diagonal term is simply assorbed by the pressure and thus "disappear" from my equation or have I to add it to rhs of equations?

Thanks for clarifying me this step!

[FMDenaro]

Originally Posted by AliE

Hello CFD friends,

Today I have for you a practical question on LES. I would like to implement some simple LES models (with no dynamic procedure) in my code (e.g. Smagorinsky or WALE).

After that the filtering procedure has been carried out, an equation formally similar to RANS comes out. The Reynolds stress tensor is approximated:



With modelled somehow. The eddy viscosity term can be associated to the diffusion term in the NS equation, while the diagonal term is assorbed by the pressure term.

Mathematically I am fine with this, but practically I have this question:

While I am solving numerically the problem, the diagonal term is simply assorbed by the pressure and thus "disappear" from my equation or have I to add it to rhs of equations?

Thanks for clarifying me this step!

Yes, it disappears from the practical computation but you have to consider thate the pressure field you compute by solving the elliptic pressure equation is modified implicitly by the presence of such term. But since in incompressible flows you have no thermodynamic meaning in the pressure field, no matter about this additional contribution.

[AliE]

Hello Filippo,

thank you for your kind replay. This is the same conclusion that I have drawn looking at an old LES code, but I was totally unsure about this.

Thus practically only diffusion term changes via a modified eddy viscosity? Am I right?

Since this discussion is open, I will take advantage of your knowledge a little bit more if you are fine with this

I know that the inlet condition is tricky in LES. There are some codes out there that generate for you some turbulet inlet condition according to some parametern but, in your opinion, is it resonable give a uniform inflow and let the les model working in the wake region? Think for example to a stalled airfoil.

[FMDenaro]

Originally Posted by AliE

Hello Filippo,

thank you for your kind replay. This is the same conclusion that I have drawn looking at an old LES code, but I was totally unsure about this.

Thus practically only diffusion term changes via a modified eddy viscosity? Am I right?

Since this discussion is open, I will take advantage of your knowledge a little bit more if you are fine with this

I know that the inlet condition is tricky in LES. There are some codes out there that generate for you some turbulet inlet condition according to some parametern but, in your opinion, is it resonable give a uniform inflow and let the les model working in the wake region? Think for example to a stalled airfoil.

Yes, for an eddy viscosity model (both static and dynamic) only the viscosity is changed.


The inflow conditions in LES (actually also in DNS) are an issue, many different proposals can be found, I cannot say if one can be preferred to others. What you can do is to prescribe a uniform inlet by adding some noise but be aware that you need to add a upward region into the domain to let the flow correlate physically and neglecting the numerical setting before impacting the airflow. Therefore you will add more grid points and consequently increase the computational cost.

[sbaffini]

Originally Posted by FMDenaro

Yes, it disappears from the practical computation but you have to consider thate the pressure field you compute by solving the elliptic pressure equation is modified implicitly by the presence of such term. But since in incompressible flows you have no thermodynamic meaning in the pressure field, no matter about this additional contribution.

Just to mention explicitly that, for a compressible code, or whenever you need the absolute pressure value (e.g., for thermodynamics), you would need to actually subract that turbulent contribution for your purpose.

Nonetheless, it is not uncommon to just don't do that even for compressible flows as the turbulent contribution tends to be a rather small % of the total pressure and it is also affected by modeling errors.


The same also happens, for example, when gravity is used and hydrostatic pressure absorbed in the pressure. In this case, however, it is of paramount importance to actually remove the hydrostatic part before using the pressure.

[AliE]

Ok, loud and clear guys. thanks!