[fatirishman53]

I've been trying to develop a precursor simulation of the turbulent ABL. However, I don't think I'm getting very good results and was hoping someone might be able to point out what I'm doing wrong.

Here's the setup (mostly based off Tim Stovall's master's thesis):

version: OpenFOAM-3.0.0
simulationType: LES
solver: pimpleFoam
domain: 1250m x 800m x 800m, (x, y, z), (streamwise, vertical, horizontal)
mesh: 125 x 200 (grading factor 10) x 80
viscosity: 1.55e-5 m2/s
fvOptions: meanVelocityForce with Ubar = (3 0 0)

B.C.'s:
top: symmetryPlane
sides: symmetryPlane
inlet: cyclic
outlet: cyclic
bottom: wall
U - fixedValue; uniform ( 0 0 0 );
p - zeroGradient;
nut - fixedValue; uniform 1.55e-5; (have used nutkWallFunction and nutUWallFunction)
k - fixedValue; uniform 0;

SGS:
Smagorinsky and kEqn

Wall Functions:
As pictured: nutkWallFunction
Also have tried: nutUWallFunction

I have also run a couple versions using the nutkWallFunction and vanDriest damping. However, all of my results are very similar.

I initially ran this simulation to develop turbulence and monitored the average fluctuations at the domain center in order to determine when it became statistically stable. Then, the initial turbulence was run for 5000 seconds (approx. 10 flow-through times), and probes were taken at 10 Hz. Attached are the post-processed results.

For the nondimensionalized semi-log plots (u+ vs. y+) I used sciPy's curve_fit while iterating through values for u* in order to best fit a function of the form A(ln(x * u* / nu)) + B. The results are in a textbox on each plot and were probed from a vertical line through the center of the domain. My hopes were to have data whose curve fit resulted in a von Karman (kappa) = 0.41 and B = 5.2. However, as you can see, the results are way off. I also got the same curves at the lower and upper y+ regions. At first, I considered these regions to be the sub-viscous layer and outer region. However, the lower y+ values do not match u+ = y+... not even close.

I also included images of my plots for auto-correlation based off probes at the domain center point. I don't understand why my x-component's auto-correlation does not decrease with increased time lag. Instead, it seems to remain perfectly correlated (unless I messed up my Python script).

One other question, which one-equation eddy viscosity model is the kEqn SGS model based off of?

Any help would be immensely appreciated. Thanks in advance.

[fatirishman53]

Here are the log-linear plots for the nutUWallFunction implementation and the nutkWallFunction with vanDriest damping. The auto-correlation plots are very similar to the ones posted previously.

The only things I can think of are:
1. Problem with boundary conditions. Perhaps cyclic sides and a slip top would yield better results.
2. Changing the k boundary condition at the bottom wall to zeroGradient as suggested by de Villers
3. The meanVelocityForce is having an adverse affect on the log layer and auto-correlation (interfering with dissipation???)

[fatirishman53]

Somebody? Anybody? Pretty please....

[lebc]

Hi Matt,

I'm also studying the ABL on OF, but considering k-epsilon turbulence model...

Either way, now I'm using some reference articles, maybe you can have an insight from one of them (of course, if you don't know them already!):

"On the use of the k–epsilon model in commercial CFD software to model the neutral atmospheric boundary layer" D.M. Hargreaves, N.G. Wright

"Appropriate boundary conditions for computational wind engineering models revisited" P.J. Richards, S.E.Norris

"Atmospheric Turbulence Effects on Wind-Turbine Wakes: An LES Study" Yu-Ting Wu and Fernando Port´e-Agel

Hope I can help with this!

Best Regards,
Luis

[Nicole Yu]

Originally Posted by fatirishman53

Somebody? Anybody? Pretty please....

did you finally solve the problem? I have the same questions