Hello,

I want to know if FLUENT can perform Direct Numerical Simulation of turbulent flows. Regards Rajani

Hi,

It is not possible to perform DNS simulation in Fluent because of the fact that DNS requires too much computational effort.

Regards,

what if u use a lot of cells for a small geometry and solve equations without turbulence models- isn't that performing a DNS?

you can perform DNS on fluent by simply using the laminar model to solve the complete navier stokes equations. the turbulence terms are inherent in them! and as Alice has noted, an insane number of grid points within your domain will start to resemble DNS. if you have the computing resources to throw at such a problem then go for it.

nandu

absolutely, just make sure your grid is fine enough to resolve the kolmogorov scales, and turn on the unsteady laminar equations, and there you go. Then, choose a time-step representative of the turbulent time-scales in your domain, something on the order of 10-10 seconds or smaller, and let it rip.

Now, you may need 10000 CPU's running in parallel, about 10,000,000 GB's of memory, and you'll need to run billions of time-steps just to simulate a few seconds of real time, but in principle, it can be done. (Some sarcasm of course embedded in that).

[Joseph Piasecki]

Is ANSYS use for DNS possble with ordinary resources?

[beer]

Hi
Still not
I have seen a DNS a few months ago running for a few weeks on round about 1000 CPU`s to comute a region smaller than a finger nail. That was a parallel FORTRAN Code and ANSYS is much slower than such a research code. Even if Antons answer is from 2004, he is still right and it is not really possible to run a DNS with ANSYS for common Rynolds numbers.

Greetings

[VictorLP]

Hello. I want to simulate primary jet breakup using VOF+Level Set multiphase model. There is no suitable RANS turbulence model for my case. I tried a number but result is not realistic. Much beter result i have obtained with "laminar". But i need beter, closer to experimental data. The average cell size i have is 2e-6m to resolve fractions with size about 10e-6. My case is 2D axisymmetric. I have tried 3D LES calculation with same cell size - result is close to 2D Laminar but computing time much much bigger because 1 million elements became 16 million.

May be 2D solution with fine grid (close to DNS sizes) will be more suitable than 3D LES?
How can i estimate necessary size of cells and timestep for good similarity to DNS solution? I understand that it should be Kolmogorov size and turbulence time scale, but i don't know how practically to calculate this scales.

[ekha]

Hi, after 15 years of your comment, did researchers perform Direct Numerical Simulation of turbulent flows using Fluent ?

[Tariq Javaid]

Originally Posted by Anton
;110917

absolutely, just make sure your grid is fine enough to resolve the kolmogorov scales, and turn on the unsteady laminar equations, and there you go. Then, choose a time-step representative of the turbulent time-scales in your domain, something on the order of 10-10 seconds or smaller, and let it rip.

Now, you may need 10000 CPU's running in parallel, about 10,000,000 GB's of memory, and you'll need to run billions of time-steps just to simulate a few seconds of real time, but in principle, it can be done. (Some sarcasm of course embedded in that).

How to make sure that refined mesh for DNS resolves Kolmogorov length scales? And should these length scales be resolved throughout the domain?