[rorymcd98]

So I'm doing 2D LES of a turbulent jet. Here is the vorticity field animation for the less refined mesh:

https://i.imgur.com/YIOs5SF.gif

And the more refined mesh:

https://i.imgur.com/Ai80FOy.gif

As you can see, the turbulent jet sustains itself for much longer in the less refined mesh. Intuitively I would have thought that the more refined mesh is more accurate, however, I don't know if this is true. For my assignment I have to be able to study the stable shear layer, so having a longer jet would be beneficial.

The less refined mesh was from the 'recommended grid sizing' in the assignment. Which mesh should I use then? Is there a good way to justify the less refined mesh?

[FMDenaro]

Originally Posted by rorymcd98

So I'm doing 2D LES of a turbulent jet. Here is the vorticity field animation for the less refined mesh:

https://i.imgur.com/YIOs5SF.gif

And the more refined mesh:

https://i.imgur.com/Ai80FOy.gif

As you can see, the turbulent jet sustains itself for much longer in the less refined mesh. Intuitively I would have thought that the more refined mesh is more accurate, however, I don't know if this is true. For my assignment I have to be able to study the stable shear layer, so having a longer jet would be beneficial.

The less refined mesh was from the 'recommended grid sizing' in the assignment. Which mesh should I use then? Is there a good way to justify the less refined mesh?

Be careful in the LES theory, the filter size changes according to the grid size (for implicit filtering) thus each grid represents a different level of epxected resolution. You cannot simply compare in terms of grid refinement the contour lines but you need to compute the statistics.

For vanishing grid size the LES solution will tend to the DNS solution.

[rorymcd98]

Thanks for the answer. What types of statistics do I need to compute for LES? I initially did a RANS simulation to determine the integral length scale based on kinetic energy, and have targeted at least 5 cell lengths per integral length scale in the bulk of my flow.

I also tried to target resolving 80% of the turbulent kinetic energy in the flow, but I'm having difficulty with this step as the subgrid (modelled) tke in the flow appears to be extremely small in comparison.

[FMDenaro]

Originally Posted by rorymcd98

Thanks for the answer. What types of statistics do I need to compute for LES? I initially did a RANS simulation to determine the integral length scale based on kinetic energy, and have targeted at least 5 cell lengths per integral length scale in the bulk of my flow.

I also tried to target resolving 80% of the turbulent kinetic energy in the flow, but I'm having difficulty with this step as the subgrid (modelled) tke in the flow appears to be extremely small in comparison.

You need RMS, correlations, spectra. But be aware that your case is 2D, you do not have a physical turbulence in your jet flow.

[rorymcd98]

I have a question about turbulence in LES. Often people use Reynolds number as a criteria for turbulent flow, are these same criteria sort of irrelevant for 2D LES or maybe even entirely irrelevant?

Also, since my flow cannot transition to turbulence, does that mean it is laminar flow?

[FMDenaro]

Originally Posted by rorymcd98

I have a question about turbulence in LES. Often people use Reynolds number as a criteria for turbulent flow, are these same criteria sort of irrelevant for 2D LES or maybe even entirely irrelevant?

Also, since my flow cannot transition to turbulence, does that mean it is laminar flow?

Reynolds number value is not unique, it depends on the lenght scale you use. Only in specific case it makes sense to use it for defining a regime.


I do not understand what do you mean for "my flow cannot transition to turbulence"