[ulfu]

Hello forum,
my advisor has tasked me with coming up with a list of properties a subgrid model for large eddy simulation should have - and I could not really find many. Here are my toughts:


- it should be dissipative, but not too much
- should turn off in laminar flow
- should go to zero at the wall
- should allow backscatter


however, this all seems so trivial... I am struggling to find other properties... what about symmetries for example? How do we know what the right amount of dissipation is? Are there other mathematical properties the model should have?



I would also be grateful if you could point me to a textbook or paper about this topic!
Thanks a lot, have a nice sunday


ulfu

[FMDenaro]

Originally Posted by ulfu

Hello forum,
my advisor has tasked me with coming up with a list of properties a subgrid model for large eddy simulation should have - and I could not really find many. Here are my toughts:


- it should be dissipative, but not too much
- should turn off in laminar flow
- should go to zero at the wall
- should allow backscatter


however, this all seems so trivial... I am struggling to find other properties... what about symmetries for example? How do we know what the right amount of dissipation is? Are there other mathematical properties the model should have?



I would also be grateful if you could point me to a textbook or paper about this topic!
Thanks a lot, have a nice sunday


ulfu

Textbooks I suggest

1) Sagaut
2) Pope

[Andrea1984]

In Nicoud et al. (2011), Using singular values to build a subgrid-scale
model for large eddy simulations, they have developed the so-called Sigma SGS closure based on four desirable property of the differential operator D_m, which is used to evaluate the SGS viscosity as

nu_sgs= (C_m Delta)^2 * D_m(U)

The four properties are:
1. Return a positive quantity which involves only locally defined
velocity gradients
2. Reproduce the correct O(y^3) near-wall asymptotic behaviour
3. Return zero for any two-component or two-dimensional flows
4. Return zero for axisymmetric or isotropic expansion=contraction

The list is definitely not complete, but I think it's a good starting point.

Andrea

[sbaffini]

Originally Posted by Andrea1984

In Nicoud et al. (2011), Using singular values to build a subgrid-scale
model for large eddy simulations, they have developed the so-called Sigma SGS closure based on four desirable property of the differential operator D_m, which is used to evaluate the SGS viscosity as

nu_sgs= (C_m Delta)^2 * D_m(U)

The four properties are:
1. Return a positive quantity which involves only locally defined
velocity gradients
2. Reproduce the correct O(y^3) near-wall asymptotic behaviour
3. Return zero for any two-component or two-dimensional flows
4. Return zero for axisymmetric or isotropic expansion=contraction

The list is definitely not complete, but I think it's a good starting point.

Andrea

Bert Vreman also worked, previously, on something along the same line of this. His work can be found here:

http://www.vremanresearch.nl/Vreman-...bgridmodel.pdf

Roel Verstappen has worked on some additional concepts (check latest works):

https://scholar.google.com/citations...1P0AAAAJ&hl=en

Yet, beware, these reasonings are not even close to be the whole picture in LES SGS modeling.

[FMDenaro]

I don't know the educational level of this requested task, but I would consider only in a second effort the level of complexity required by the mathematical constraints detailed in those papers.


The first task should be in understanding the hystoricaal multilevel filtering approach implied in the (exact) mathematical Germano identity and the approximation introduced then for the dynamic eddy viscosity model.


And, I would also suggest to consider the idea implied in the ILES formulation, that is the no-model LES. The local truncation error is the first "SGS model" one should understand.