CFD Online Discussion Forums

CFD Online Discussion Forums (
-   OpenFOAM Running, Solving & CFD (

andy_andy September 13, 2006 23:11

The DNS solver requires inputs
The DNS solver requires inputs for Ea and K0 in "BOXTURBDICT" and for U0sigma, U0alpha, U0kupper and U0lower in "turbulence properties".

I want to simulate a turbulent flow for a known integral length scale and turbulent rms velocity and then apply it to a combustion event. I assume I can estimate these inputs from this and the thermodynamic state of the fluid. How do I go about that?

I have found some bits around and could have a guess at their meaning, however its always best to ask especially as it takes me weeks to find a solution. I would like to find the exact definitions of these parameters from literature or a website. Can anybody help?

I am happy to update the WIKI once I understand it all. I am sure others having the same problems.

andy_andy September 24, 2006 21:48

I`d welcome any input from oth
I`d welcome any input from others who have been through the same process.

Here is my progress to date.

The turbulence spectrum appears to be based on the following equation.

E(k)= (Ea*(k/k0)**4)exp(-2.0*(k/k0)**0.5)

That was easy enough.

However, the inputs for "turbulenceProperties" appear to be more difficult to define. They appear to be linked to the forcing of turbulence.

I have traced the terms U0sigma, U0alpha, U0kupper and U0lower back to UOprocess.C but cannot find the definition of each of these terms or any literature describing them.

My best guess at the moment is that U0kupper and U0lower are the maximum and minimum wavelengths used to excite the turbulence. I still cannot work out what U0sigma and U0alpha are exactly. Any help would be greatly appreciated.

grtabor September 25, 2006 03:22

UO process uses an Uhlenbeck-O
UO process uses an Uhlenbeck-Ornstein stochastic process to generate a stream of random numbers - or rather not quite random as they are stochastic, and so each number is partly related to the one preceeding it. Sigma and alpha are the two parameters in the UO process, controlling the variance and mean (I think) respectively. Have a look at

for more details. Unfortunately there are various different definitions of the process with different scalings for alpha and sigma, and I'm not clear which reference was used for the original algorithm; I always have to sort through the code to work out the precise definition.


andy_andy November 9, 2006 22:06

Thank you for your help. I
Thank you for your help.

I am still having problems with the DNS though. My solution always becomes unstable. Below is a summary of my problem.

Geometry 5mm x 5mm square 2 "cyclic" patches and an "empty" for the 3rd dimension

256x256 grid size

dT= 0.0001s

nu= 1.59e-5

Ea= 0.00056
K0= 10

UOsigma = 0.1
UOalpha = 0.8
U0Kupper = 300
U0Klower = 7

My K and Epsilon slowly grow until there is no solution. I have tried reducing the timestep and tuning each input to no use. Any ideas?


hjasak November 9, 2006 22:15

Are you by any chance doing a
Are you by any chance doing a 2-D DNS simulation? This will (of course) not work, as the physics is all wrong. If you are doing 3-D as you should, you should not be using empty patch type.


johndeas February 21, 2008 13:15

boxTurbDict seems to never be
boxTurbDict seems to never be accessed. I am a bit confused about the magnitude of the forcing in the current version.

ali March 10, 2008 08:23

Hi, I am trying to create a bo
Hi, I am trying to create a box with initial turbulent. I have the box dimention and Vrms, I don't know how to relate them to Ea and K0. the formula which is presented here for turbulence energy spectrum [E(k)= (Ea*(k/k0)**4)exp(-2.0*(k/k0)**0.5)] is not clear for me, I will be very grateful if some one can clue me in on this problem.

hismother February 13, 2012 05:01

Hi Andy,

are you still working on this. Since now I just begin with DNS and am facing the same problem as yours here. Would be nice, if you can show me how to overcome this.

Thanks in advance,


JackW July 30, 2013 09:51

Hi all,

I know this is old, but I would love to know if anyone got anywhere,

I am guessing that sigma and alpha are the variance and mean, but I do not know how they relate to the taylor Re. This is what I would LOVE to know!

boxTurbDict controls the application boxTurb which generates the initial conditions on velocity.
This I don't think is too important as the forcing will eventually send the simulation to the correct Re.

The upper limit on K seems very large, as in you are forcing too much of the spectrum and this could well be the simulation is blowing up.


All times are GMT -4. The time now is 04:47.