[starz]

I have a horizontal wall without any thickness in between the flow field, which means the faces are defined as a no-slip wall.

I use LES with Werner-Wengle wall functions. My y+ is around 200. After I got the results, I export the data on a vertical line that passed through the horizontal wall. I export both the node and cell center values. Then I check that node value exact at the location of the wall. And I find the velocity is 0, but mean velocity is not 0. While the mean velocity on the face should also be 0.

In RANS with wall functions, the mean velocity of the node at the wall is always = 0.

Here, I plot the mean velocity from LES. The location of the wall is marked as yellow, which is clearly not =0.


I am curious about how ANSYS Fluent saves the data when using LES and RANS. Does LES solve and save the velocity at the cell center instead of the node, therefore the node value at the wall is not 0 by interpolation?

Hope I could get some reference and information.

Best regards
Many thanks
Starz

[LuckyTran]

The data structure is identical for LES or RANS. Values are at cell centers. Values at faces are reconstructed/interpolated and don't need to be saved.


I can't comment on this zero wall thickness internal boundary with wall functions giving non-zero mean velocity.

[vinerm]

What values are reported if velocity contours or plots are plotted directly for the walls? If those are NOT 0, then there is certainly some issue, either with the implementation of the wall-function or the wall-function itself. What happens if you do not enable Werner-Wengle?

[starz]

Originally Posted by LuckyTran

The data structure is identical for LES or RANS. Values are at cell centers. Values at faces are reconstructed/interpolated and don't need to be saved.


I can't comment on this zero wall thickness internal boundary with wall functions giving non-zero mean velocity.

Many thanks. So do both velocity and pressure are saves at cell centers?
If so, by interpolation, the velocity on the zero thickness wall can be non-zero, right?

[starz]

Originally Posted by vinerm

What values are reported if velocity contours or plots are plotted directly for the walls? If those are NOT 0, then there is certainly some issue, either with the implementation of the wall-function or the wall-function itself. What happens if you do not enable Werner-Wengle?

Sorry, I made a mistake here. Actually I find the velocity (instantaneous) is 0. However, the mean velocity that averaged from a lot of time steps, is not 0. While the mean velocity on the face should also be 0

Here below I plot the velocity (instantaneous ). There are walls with thickness at y= 3,6,9. And the velocity is 0. But for the mean velocity that averaged from a lot of time steps, it is not 0 (as shown in the figure in the post). So I suspect that it is because of how FLUENT save and interpolate the data,
_20200222115815.png

[vinerm]

Fluent solves the equations at the cell-center and time-averaging is also performed at the cell centered values. If a value is 0, its mean is also supposed to be 0. So, the discrepancy you are observing could be due to the following reason.

When instantaneous values are plotted, Fluent shows the node values as per the applied condition, i.e., no-slip. However, there is no such condition available to revert to while accessing RMS or Mean values. Those values are based on interpolation from the mean values at the cell-centers.

[LuckyTran]

When you plot a value of a flow variable on a surface, Fluent plots the values of the wall adjacent cells (the cell centered values) unless you specifically choose to plot node values, and node values are interpolated.

Originally Posted by starz

Actually I find the velocity (instantaneous) is 0. However, the mean velocity that averaged from a lot of time steps, is not 0. While the mean velocity on the face should also be 0

1) If you average a bunch of zeros you get zero.
2) What is being averaged is the non-zero cell centered velocity, not the interpolated-to-zero velocity.