[last_jedi]

Dear All,

I understand that this topic has already been explained, but there are still some questions regarding the use of icoFoam for turbulent DNS.

1)First of all except Re is, there any other change that should be conducted in this solver in order to achieve turbulent DNS??

2)I know that a small timestep should be implemented but what order should that be?

(trying to model turbulent DNS channel flow)

[Santiago]

A more fundamental question would be: Why DNS with OF?

I mean, you could perfectly do it but is not the right tool for that...

I correct myself, you could EVENTUALLY do it.

[last_jedi]

The answer is, that i would like to test a code that i will later implement in OF and study turbulence prediction. So i can not use any commercial solver.

[Santiago]

Originally Posted by last_jedi

The answer is, that i would like to test a code that i will later implement in OF and study turbulence prediction. So i can not use any commercial solver.

I never said to use any FVM commercial software for DNS. All evidence to the contrary, I would discourage its use for DNS for the same reasons (and some more) I discourage the use of OF.

The reasons are too vast to list them here, but I will give you 2 of them, having pisoFoam as a baseline:

1. The primary property you're interested in when doing DNS is ENERGY conservation, as understood in functional analysis (not talking about thermo here). The primitive formulation of the NSE only conserve mass discretely (velocity is projected into solenoidal space using a "corrector" step), so it does for the energy! Velocity-vorticity formulations of the NSE equations, often used in Discontinuous Galerkin, Spectral Element, methods conserve mass by definition, so conservation of energy is only dependant on the spatio-temporal schemes used.

2. PISO is a formulation that, strictly speaking, is inconsistent in the formulation of the convective term. This is a fact that is recognized even by Issa in his article. This problem is usually mitigated by using low Courant numbers, in my experience with PISO Co<=0.1. You should use a different algorithm in OF if your prime interest is to resolve turbulence (having in mind what i said in 1.)


There are excellent spectral/DG codes available in the wild, not commercial, for free. One salient example is NEK5000.

To conclude, I'm not making an argument against OF or FLUENT or anything of the sorts. I'm saying that you should use the right tool for the right job.