dear cfd-friends, i'm wondering if DNS (= direct numerical simulation, meaning calcualtion of turbulent flow without ANY modelling) is possible with a commercial package as FLUENT. i know that mesh size is determined by kolmogorov's length scale. did anybody try it anytime ?

yup. Google scholar: dns fluent. In 3D naturally.

WOW!!! How can we have a direct numerical simulation using FLUENT? Is it possible using a special version of the software?

I think if we use LES with a very fine numerical grid, we can obtain a result that approaches to the result obtained using DNS. Is it correct?

I have about 7 years experience in working with FLUENT and very interested to have some information about DNS using Fluent.

Thanks

We can obtain DNS solution if we use LES. However in order to minimise the effects of turbulence model, our grid will have to be fine enough to resolve the kolomogorov length scales. I personally believe we do not need to waste our computer resource and go to the extreme of resolving every fluctuation in our flow. If you understand the physics of LES you would prefer to stick to it which would give as good results as DNS and much at lower cost.

Mani