[Nek5000-users] Number of Elements per Characteristic Length for DNS

[nek5000-users at lists.mcs.anl.gov]

Hi,
typically, in wall bounded turbulence you would design the mesh based on 
inner scaled quantities (i.e. based on the viscous length scale l*). 
This is what we have done in the paper by El Khoury, in accordance to 
studies in channel flow. TOwards the centre, one can then also use 
Kolmogorov units (not sure whether Taylor units are so useful). I am 
quite convinced that the El Khoury simulations are well-resolved DNS, 
which can also be seen when looking at spectra.

Philipp

On 2018-07-11 20:58, nek5000-users at lists.mcs.anl.gov wrote:
> Hello All,
> 
> I am currently investigating the turbulence in a pipe with periodic 
> boundary conditions. I am working on building my mesh and was planning 
> on using Re^(3/4) as the number of elements per characteristic length, 
> but I have found differing methods in literature detailing the number of 
> elements used for DNS. One paper in particular (G.K. El Khoury et al, 
> Direct Numerical Simulation of Turbulent Pipe Flow at Moderately High 
> Reynolds Numbers,  Flow Turbulence Combust, 2013) uses Nek5000 to 
> simulate turbulent flow in a pipe at different Reynolds numbers. The 
> number of elements calculated in this paper seem to be in between the 
> LES and DNS suggested number of elements (element size between the 
> Taylor microscale and the Kolmogorov scale). Turbulent Flows by Pope 
> however suggests a number of elements that is slightly larger than 
> Re^(3/4), and is dependent on the Taylor microscale Reynolds number.
> 
> Any suggestions on the correct way to determine the number of spectral 
> elements in a characteristic length would be appreciated.
> 
> Regards,
> 
> Tommy
> 
> 
> 
> _______________________________________________
> Nek5000-users mailing list
> Nek5000-users at lists.mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users
>