[airwarriorg91]

Hello,
I am looking for code to perform DNS of flow around complex geometries at moderate Re (20K -30K). Till now, I have been using Xcompact3D and Nek5000. Xcompact3d is good for flows till Re 5K max otherwise it is comutationally very expensive on the otherhand Nek5000 is fast but meshing around the complex geometry is difficult. So, I am looking for some alternative which is computationally fast as well as doesn’t require meshing.

Thus, I have a few questions regd. that,

1. How much computationally expensive the code is for a DNS (Some example would be highly helpful in understanding)?
2. How acccurate is the method ?

I am studying flow around UAVs through DNS.

Thanks,
Gaurav

[FMDenaro]

Originally Posted by airwarriorg91

Hello,
I am looking for code to perform DNS of flow around complex geometries at moderate Re (20K -30K). Till now, I have been using Xcompact3D and Nek5000. Xcompact3d is good for flows till Re 5K max otherwise it is comutationally very expensive on the otherhand Nek5000 is fast but meshing around the complex geometry is difficult. So, I am looking for some alternative which is computationally fast as well as doesn’t require meshing.

Thus, I have a few questions regd. that,

1. How much computationally expensive the code is for a DNS (Some example would be highly helpful in understanding)?
2. How acccurate is the method ?

I am studying flow around UAVs through DNS.

Thanks,
Gaurav

A DNS formulation depends on the grid resolution not on the type of code. Any consistent code can work (theoretically) in DNS resolution, the issue is just in the correct grid to resolve all turbulence scales.


A different matter is a LES code, there a many aspects to be considered.

[sbaffini]

Originally Posted by airwarriorg91

Hello,
I am looking for code to perform DNS of flow around complex geometries at moderate Re (20K -30K). Till now, I have been using Xcompact3D and Nek5000. Xcompact3d is good for flows till Re 5K max otherwise it is comutationally very expensive on the otherhand Nek5000 is fast but meshing around the complex geometry is difficult. So, I am looking for some alternative which is computationally fast as well as doesn’t require meshing.

Thus, I have a few questions regd. that,

1. How much computationally expensive the code is for a DNS (Some example would be highly helpful in understanding)?
2. How acccurate is the method ?

I am studying flow around UAVs through DNS.

Thanks,
Gaurav

For traditional codes, your best shot as of today is, no doubt, Nek5000. It is the only code I know of which is currently used and supported for DNS of industrial cases. And, let me add a personal opinion, a mesh is so important for LES/DNS that completely automating it, or doing it having only user friendliness in mind is, conceptually speaking, very wrong.

I don't follow as much Incompact, which might as well be used for industrial cases but, my point here is that it isn't, formally speaking, traditional as it uses immersed boundary.

If you are willing to move even forward then from traditional, you could look at Lattice-Boltzmann codes, which also rely on the immersed boundary concept. I'm not into them but, according to their marketing materials, they can do marvels of any kind and complexity, and seem to be on the scale-resolving side of the matter (altough, I have several personal concerns about their definitions of LES and DNS)