[sbaffini]

Dear all,

with this post i want to promote the availability of large eddy simulation (LES) data for the turbulent channel flow at Re_tau = 590, following the well known DNS database of Moser, Kim and Mansour.

The database, which is available at:

http://www.cfdlab.polimi.it/Research/LESinItaly.tar.gz

has been produced within the LESinItaly initiative, an Italian collaborative framework on LES.

Different research groups produced the results with different codes, in particular:

- In house staggered 2nd order Finite Difference
- In house cell-centered 2nd order Finite Volume
- In house hybrid 4th order/pseudo-spectral solver
- Open-Source FV codes Code_Saturne and OpenFOAM
- Commercial FV codes Fluent and TransAT

Each group has provided results with its own SGS model(s) and a no model computation on two separate grids (differing in the wall-normal resolution). You can find additional information in the README files accompanying the database.

Regards

[cfdnewbie]

Great, we have some computations of that channel as well, thank you for the link and for compiling the data. One question: Can the results be cited (for comparison?) and if yes, how ? Would you have a bibtex for example?

Thanks again!

[FMDenaro]

Originally Posted by cfdnewbie

Great, we have some computations of that channel as well, thank you for the link and for compiling the data. One question: Can the results be cited (for comparison?) and if yes, how ? Would you have a bibtex for example?

Thanks again!

the README file should contain the infos ...

[cfdnewbie]

Allright, thanks a lot!

[FMDenaro]

we are currently revising the full paper submitted on Meccanica

[sbaffini]

I just want to add that, if this post gets lost, you can find again the link to the database in my blog page (in few days actually, as it is not updated yet):

http://www.cfd-online.com/Forums/blogs/sbaffini/

[huangxianbei]

Originally Posted by sbaffini

Dear all,

with this post i want to promote the availability of large eddy simulation (LES) data for the turbulent channel flow at Re_tau = 590, following the well known DNS database of Moser, Kim and Mansour.

The database, which is available at:

http://www.cfdlab.polimi.it/Research/LESinItaly.tar.gz

has been produced within the LESinItaly initiative, an Italian collaborative framework on LES.

Different research groups produced the results with different codes, in particular:

- In house staggered 2nd order Finite Difference
- In house cell-centered 2nd order Finite Volume
- In house hybrid 4th order/pseudo-spectral solver
- Open-Source FV codes Code_Saturne and OpenFOAM
- Commercial FV codes Fluent and TransAT

Each group has provided results with its own SGS model(s) and a no model computation on two separate grids (differing in the wall-normal resolution). You can find additional information in the README files accompanying the database.

Regards

Glad to see you again The Re=10935,right? I see that in the example of Nek5000 , the Ret=590 also, so I just want to confirm that.

[sbaffini]

Hi Huang,

yes, that's the Reynolds according to the DNS. However, as all the codes run with a fixed pressure gradient, the resulting bulk Re numbers for this database will vary from code to code and case to case. Still, they are generally close to this value.

This activity started back in 2009 and has nothing to do with Nek5000, which i started using the last summer. However, according to the user file of the example (http://trac.mcs.anl.gov/projects/nek...urbChannel.usr) it is the same flow, but computed with a fixed mass flow rate.

[huangxianbei]

Originally Posted by sbaffini

Hi Huang,

yes, that's the Reynolds according to the DNS. However, as all the codes run with a fixed pressure gradient, the resulting bulk Re numbers for this database will vary from code to code and case to case. Still, they are generally close to this value.

This activity started back in 2009 and has nothing to do with Nek5000, which i started using the last summer. However, according to the user file of the example (http://trac.mcs.anl.gov/projects/nek...urbChannel.usr) it is the same flow, but computed with a fixed mass flow rate.

Thank you very much. The mass flow rate is fixed in the example,I'm curious about the fixed gradp and mass flow rate, can this make the result a difference? As I have done last several months, the gradp is fixed using OpenFOAM

[sbaffini]

Well, this is (in my opinion) an interesting question indeed. Apparently, this topic has been addressed "privately" in all the labs working on DNS and only recently some details on the topic are coming out in the papers. The universal answer to the question is that it doesn't matter, the results are exactly the same if you fix the pressure gradient or the mass flow rate (corresponding to the same pressure gradient).

While i had the exact same answer from some tests, i admit that i am still not very happy on this matter. One can think of the fixed mass flow case as a fixed pressure gradient case with a super-imposed, null on average, fluctuating forcing. Having the same results means that such fluctuating component has no direct effect on the mean flow.

This might depend from the scale separation between the input (the fluctuating forcing is still constant in space) and the output (the fluctuating forcing balances the local fluctuations on the wall shear stress to keep the mass flow constant), which is very large. But, as the flow governing equations are not the same, i don't expect the effect to be exactly null for all the statistics.

[huangxianbei]

Originally Posted by sbaffini

Well, this is (in my opinion) an interesting question indeed. Apparently, this topic has been addressed "privately" in all the labs working on DNS and only recently some details on the topic are coming out in the papers. The universal answer to the question is that it doesn't matter, the results are exactly the same if you fix the pressure gradient or the mass flow rate (corresponding to the same pressure gradient).

While i had the exact same answer from some tests, i admit that i am still not very happy on this matter. One can think of the fixed mass flow case as a fixed pressure gradient case with a super-imposed, null on average, fluctuating forcing. Having the same results means that such fluctuating component has no direct effect on the mean flow.

This might depend from the scale separation between the input (the fluctuating forcing is still constant in space) and the output (the fluctuating forcing balances the local fluctuations on the wall shear stress to keep the mass flow constant), which is very large. But, as the flow governing equations are not the same, i don't expect the effect to be exactly null for all the statistics.

Yes, when gradp is fixed, the term should be explicitly appear in the equation.As you said, the effect shows almost null in the test cases, so I think it's due to the same Reynolds number(almost,not exactly). This dimensionless number truly depicts the flow characteristic in this way, which can be like this: The same Re flow in channel means the same statistics.