I'm working on a stirrer simulation. When using RANS turbulence modells I used to determine the power number with the turbulent eddy dissipation (ted);

Np=Sum(ted*volume of finite volume) / (n^3 * d^5)

Now I performed a LES and I don't know how to determine the turbulent eddy dissipation. Can anybody help me?

Patrick

PS: In CFX-Pre it was possible to choose the turbulent eddy dissipation as a transient result variable. Unfortunately this variable was not written out (bug in CFX).

Hi Patrick,

For most LES models you don't need to set turbulent eddy dissipation. The LES model directly models the dissipation in larger eddies, and the dissipation in the smaller sub-grid eddies are usually modelled with a simple relation which does not include a modelled dissipation parameter (eg the Smagorinsky SGS model is the most common).

The issue with LES is setting up initial conditions and boundary conditions which contain the larger eddy structures. If you are modelling a stirrer it is likely you do not have any boundary conditions requiring a turbulent specification, but you will have to specify the initial condition. This can usually be overcome by starting with a simple initial condition and running the model until the turbulent structures have reached steady state (in an average sense).

Regards, Glenn

Hi Glenn,

Maybee I didn't explained that clearly; In my problem I need the dissipaion of every finite volume to postprocess the results. That's one way to determine the power number of the stirrer. I also need these data to define a local scale-up criteria for this stirrer.

LES models the large eddies directly, that's clear. Maybe I'm wrong but even with LES it should be possible to determine the local dissipation. Reasonably it would contain the dissipation of the large eddies and the modelled dissipation of the small eddies. But I don't know how.

Regards, Patrick

Go to out "output control" tab in CFX-pre. Click on "Trn Stats" tab (which means transient statistics) and select the varibale you need in the result file e.g. k or eps or u_i u_j (stress tensor). You need to integrate the equation for quite a some in order to compute meaningful averaged quantities.

Try it, it should work otherwise write a bug report to CFX.

cheers VK

Hi,

Thanks, but I'm aware of the CFX-functionality. My problem is the equation. Can you tell me the correct equation?

There are some bugs in CFX but I already reported them. But no one of our CFX supporters was able to tell me the correct equation. (-> They made some proposals - but all wrong).

Thanks for any help.

What I understood from your mail that you donot want to use the dissipation values computed from CFX but want to compute on your own. This wont be very easy in a code where source code is not open. The dissipation tensor can be computed as (written in Latex notation):

\epsilon_ij = \nu \overbar{\partial u_i u_j/\partial x_k \partial x_k}

Once, this tensor is computed, you need to compute the one half of its trace, i.e. 0.5*\epsilon_ii to get the variable ted.

You are probably using a non-orthogonal grid. To accurately compute gradient of fluctutating velocities you need to have Jacobian of coordinate transform. This info is not available in CFXpost. Moreover, if you want to compute dissipation using CFX-post, it will not be possible at all since fluctuating quantities (u_i and u_j etc) are most propabely not stored at every time-step.

You may write a user routine for CFX-pre and rerun the simulations again. But once simulations are over there is no way to compute dissipation unless or until you store each and every infomation which is usally not the case. Best would be to use CFX inbuilt functions or write a small program following the above equation.

Note that dissipation is a time-averaged quantity and you need full time history to compute it.

Hope this helps.

VK

I wanted to use the dissipation values computed by CFX. It's possible to select the variable in CFXpre. I did the run, wrote out the transient data (together with velocities and so) and got a huge amount of data. But the ted was not written out (-> CFX bug). That's why I have to run the simulation again and have to write out all variables to calculate it.

Hi

In section "output results" of CFDpre you have two tabs: trn Results and trn Stats.

Check again carefully that in both tabs you have selected dissipation to be output/computed variable. If you skip ted in one of them (esp. trn Stats), no dissipation data will be written. The reason is that ted is not computed directly in a LES or DNS unlike a RANS. Solver needs to do time-averaging during the course of simulation and for this it is absolutly neccessary that you turn on "trn Stats" feature in addition to "trn Results" in section output control of CFXpre.

If possbile post, COMPLETE section dealing with the output variables of your ccl file on this group.

Hi,

You are right. I didn't set ted in the trn Stats. I wanted to do the statistical postprocessing with another software. I will try this and hope that works. Thanks for your help!

Regards