[suryakaundinya]

Hello everyone,

I have a question about using the turbulentDigitalFilterInlet boundary condition for defining a turbulent velocity field in LES simulations. This is available in v1906. Do anyone know how to define the integral length scale tensor in the model.

In the code I have seen that it is define in the local patch co-ordinate system. If that is the case, when my patch is normal to either x,y or z-axis then the tensor should be something like (L1 0 0 0 L2 0 0 0 L3). But according to the code the tensor cannot have 0 values for any of the components.

Can someone please elaborate on how to define this parameter?

[HPE]

I think L definition is as follows:

(Luux Luuy Luuz Lvvx Lvvy Lvvz Lwwx Lwwy Lwwz)

That is, for example, for the first Luux: u-u correlation in the x-direction; for the second item Luuy: u-u correlation in the y-direction and so on.

To my knowledge, for turbulent fields, this set of correlations cannot be zero. The code then seems to be guarded from such input to reduce potential user errors. Zero correlation means complete randomness, and turbulence is not random, yet stochastic.

Hope this helps.

[suryakaundinya]

Hi,


I think it does not represent the velocity correlation. The code description says that it represents the "integral length scale" as sshown below:

L is the Integral length-scale set (9-comp):{e1,e2,e3}{u,v,w} [m]
L = (<Lxu> <Lxv> <Lxw> <Lyu> <Lyv> <Lyw> <Lzu> <Lzv> <Lzw>)

It is also mentioned that the "integral length scale set input is in the local patch
coordinate system". Also it is specified that "L_e1u, Le1v, Le2w, should always correspond to the length scales that are in association with the convective mean flow direction."

From these I think that Lx, Ly, Lz are integral length scales in different directions. And e1, e2, e3 are the local patch co-ordinates. So my question is how to define the these components where patch is parallel to a specific axis.



With regards,
Surya.

[HPE]

Integral length scale there is only the pesudo-integration of the two-point correlation function of velocity in a given direction. They contain the same information.

[HPE]

"From these I think that Lx, Ly, Lz are integral length scales in different directions. And e1, e2, e3 are the local patch co-ordinates. So my question is how to define the these components where patch is parallel to a specific axis."

e1 is your mean convection direction, say x. e2 and e3 are the orthogonal directions to e1, say y and z.

Now say, you clockwise rotate your mesh 60 degrees around the global x-axis by keeping the L input the same. In the local coordinate system, which does not feel the rotation, if you measure Ls, you will get the input back. But if you do the integral scale measurement in the rotated mesh on global coordinate system, you will get different set of Ls in the global directions. That's it.

[suryakaundinya]

Hi HPE,

Thanks a lot for the information. I might sound stupid but can you please answer one last question?

For a case where I do not have DNS data , how do we calculate the two-point correlation? Or if I know the integral length scale how to calculate the two-point correlation?

With regards,
Surya.

[mr35diamonds]

If you do not have DNS data, then you cannot compute the two-point correlation. The integral length scale is defined as a length scale that corresponds to the correlation time of a turbulent cascade. In principle, if you know the spectral scaling, you can get the structure function and estimate how the correlation function scales, but you cannot get exact values. You may try simulating without turbulent inflow BC and see how that goes.