DNS and LES in FLUENT
 

While running LES in FLUENT as we run a steady state K-epsilon model for getting a converged solution and once the flow is converged we use LES model to get transient results. Can we follow a similar approach for performing DNS in FLUENT, just by selecting the laminar model after a steady state k-epsilon gets converged?

For a DNS you need a very fine mesh and a very fine time step. Selecting "laminar" alone is not enough for DNS. I don't see any advantage in having a steady state solution as initial guess for a transient DNS. What is the idea behind this?

The idea is to get fully resolved turbulent length and time scales for a flow past a sphere at high Re. I have already run LES for a Re of up to 5000. I am able to reproduce results for pressure coefficient vs theta, drag coefficients, lift coefficients, energy spectra, strouhal number to mention a few. But the problem is I have been told several times by my colleagues that since in LES we model certain terms we cannot say that we have fully resolved the flow just by using fundamental NSE, to have a fully resolved flow it is essential to go for a DNS. The purpose is to get more insights in terms of physics for the above mentioned case. Kindly suggest whether the above purpose could also be fulfilled by LES or should I necessarily go for DNS?

Yes, with LES you have a model for the kinetic energy of the eddies that are smaller than your grid size, the Subgrid Scale Model.

If you want to solve ALL the turbulent effects from NS, you need a very fine mesh that can resolve down to the Kolmogorov Microscales, where kinetic energy is dissipated into heat. https://en.wikipedia.org/wiki/Kolmogorov_microscales

So what you can do is try to calculate the needed grid size and time step size for this DNS approach. It usually is absolutly not doable without a supercomputer.

On the other hand, if you have already good agreement with experimental data from the LES approach, doing a DNS seems academical, doesn't it..