I am very thankfull for your input I will try to figure it out 🙂

Maybe its possible when the eddys dissipat very fast, so if the Reynoldsnumber is higher it should work better, but I am not sure about this

I was wondering how one would write Production term when doing a DNS calculation? Should it be something like this?

The last part is grad(U)

And Ubar is the mean velocity

You plan to do this calculations as postprocessing?

and you are missing the averaging on the tensor

P_ij=

I wouldn't do it like that: it's expensive during calculations (you have to calculate grad, and hold P)

I would just save and use the ergodicity principle to obtain the fluctuations

Does it make really that difference for the calculation time when youre not looking for performance? I once implemented a calculation for the turublent heatflux and there where no big difference in the calculation time. And I used a Mesh with more then 20 Millionen not just a case thats too small to recognize.

@User what do you mean with mean veolicty? like UMean of the cell or the complete simulation?

> Does it make really that difference for the calculation time when youre not looking for performance? I once implemented a calculation for the turublent heatflux and there where no big difference in the calculation time. And I used a Mesh with more then 20 Millionen not just a case thats too small to recognize. @User I am always looking for performance in FOAM, I don't say the way proposed by @User is incorrect, it's just slow. Besides we don't know what kind of computer he/she has

Hi @slopezcastano , let me answer your questions. I plan to do do a runtime computation. I have a channel flow and have all the code setup to get the mean statistics and dump them to a file. That's the aim to compute the TKE budget. By Umean I mean the average value in homogeneous direction at a given z index and averaged over time. I am testing everything on my Xeon chip server and deploy the calculations on a cray system with MPI using container.

Looking forward to a fruitful discussion, sorry about the late response, the sun rose a while back in Chicago!

which Re_tau?

Right now testing on Re_tau= 500 and for my case the maximum should be < 1000

@User yeah you are right, I understand the importance of the performance and I didnt wanted to say you are wrong. Just sometimes when I want to calculate something I dont want to look at the performance and only want the results 😄

If you ask me, I would create a coded function object with the following fields:

from these you can calculate the budget, using the ergodicity principle

So you are saying just using time as the averaging direction?

no, you can perfectly average your fields long periodic and symmetric directions, as well as time

well, yes

I guess the answer to this would be long

what I meant is that by corollary of the E. Principle you can claim that: = - ^2

* create the following fields

I see I'll have a look at that. Just curious how is the coded one different from a in place implementation? All I do is loop over mesh and average over directions.

> I see I'll have a look at that. Just curious how is the coded one different from a in place implementation? All I do is loop over mesh and average over directions. @User Don't do this during run time, then your code will really get slow

> * create the following fields @slopezcastano can you specify a time averaging window with a repeating period? I have an oscillating pressure gradient and would like to average at different wave phases in direction and time etc.