[FMDenaro]

I posted this question on the RG site (https://www.researchgate.net/post/UR...ompared_to_LES) and I would receive some opinions from you about this issue.


Thanks

[FMDenaro]

The topic seems involving more researchers and gets many answers, I hope will be partecipated by new ideas

[arjun]

i wish i was knowledgeable;e enough to add something on the topic.

[FMDenaro]

Originally Posted by arjun

i wish i was knowledgeable;e enough to add something on the topic.

The core of the discussion is quite simple, why URANS does not converge to RANS if the flow is statistically steady?

[flotus1]

It doesn't? Even if we use a time step size >> time scale of the slowest vortices?

[FMDenaro]

Originally Posted by flotus1

It doesn't? Even if we use a time step size >> time scale of the slowest vortices?

I see that always an unsteady solutions develop... indeed they perform a time averaging process to get a statistically meaningful solution...
And that make no sense for me

[sbaffini]

As I already mentioned on RG, what if the steady solution does not even exists for the given RANS model (for whatever reason)? What's the meaning of this URANS solution?

[FMDenaro]

Originally Posted by sbaffini

As I already mentioned on RG, what if the steady solution does not even exists for the given RANS model (for whatever reason)? What's the meaning of this URANS solution?

I agree. For example, the solution over a sphere at a critical Re number. It is statistically steady at any Re number, so whay RANS does not converge under certain circumstances?

[FMDenaro]

I add to this old post the link for the note I wrote about this topic.


https://www.researchgate.net/publica...ocal_averaging

[LuckyTran]

Filippo, I read through it in detail and found it therapeutic. Having discussed it with the community here over the years, we are obviously in alignment.

I thank you for the presentation on pg 15. We need to perpetuate this and make more people aware that all the equations are the freakin' same, only the underlying models are different. I am personally sick of hearing that RANS breaks down when there is flow separation, blah, blah, blah. RANS cannot be the issue, because DNS is the same equation! It's the model that needs to be addressed!

There is a need for a call-to-action to folks working with spatially and temporally resolved data (DNS'ers, LES'ers, and PIV'ers) to do the ensemble averaging of fields and reexamine how the turbulence models might fit into this system. Unfortunately, most DNS'ers and PIV'ers are to busy chasing POD structures and turbulence spectra to give a damn. This analysis is not being actively pursued at any level (not in practical cases nor canonical cases like plane shear layers).

I strongly believe we are on the cusp of a breakthrough with the ensemble-averaged framework serving as the key vector. Consider for example the piston-driven unsteadiness in your example but have this piston oscillate at frequencies as high as the Kolmogorov scales. This would clearly break the local time-averaged URANS. However, I think most people would agree, that periodic forcing at a fixed frequency, does not cause turbulence to act in any bizarre new way. In the ensemble-averaging framework, there is no need for a temporal scale separation, and it is here therefore that I expect these ensemble average statistics to elucidate what these closure models need to look like for URANS and the understanding can then be readily applied to local time filtering. I continue to emphasize here that temporal scale separation is not necessary, as long as one understands that the ensemble averaging operator will decompose the instantaneous velocity into a temporally coherent part and a randomly fluctuating turbulent part. In my opinion, the classical unsteady phenomenon (like vortex shedding) falls into this coherent category and that's why it magically shows up in SRANS because, as you point out numerous times, no explicit averaging is applied to the equations being solved. The averaging is implied.

[FMDenaro]

Originally Posted by LuckyTran

Filippo, I read through it in detail and found it therapeutic. Having discussed it with the community here over the years, we are obviously in alignment.

I thank you for the presentation on pg 15. We need to perpetuate this and make more people aware that all the equations are the freakin' same, only the underlying models are different. I am personally sick of hearing that RANS breaks down when there is flow separation, blah, blah, blah. RANS cannot be the issue, because DNS is the same equation! It's the model that needs to be addressed!

There is a need for a call-to-action to folks working with spatially and temporally resolved data (DNS'ers, LES'ers, and PIV'ers) to do the ensemble averaging of fields and reexamine how the turbulence models might fit into this system. Unfortunately, most DNS'ers and PIV'ers are to busy chasing POD structures and turbulence spectra to give a damn. This analysis is not being actively pursued at any level (not in practical cases nor canonical cases like plane shear layers).

I strongly believe we are on the cusp of a breakthrough with the ensemble-averaged framework serving as the key vector. Consider for example the piston-driven unsteadiness in your example but have this piston oscillate at frequencies as high as the Kolmogorov scales. This would clearly break the local time-averaged URANS. However, I think most people would agree, that periodic forcing at a fixed frequency, does not cause turbulence to act in any bizarre new way. In the ensemble-averaging framework, there is no need for a temporal scale separation, and it is here therefore that I expect these ensemble average statistics to elucidate what these closure models need to look like for URANS and the understanding can then be readily applied to local time filtering. I continue to emphasize here that temporal scale separation is not necessary, as long as one understands that the ensemble averaging operator will decompose the instantaneous velocity into a temporally coherent part and a randomly fluctuating turbulent part. In my opinion, the classical unsteady phenomenon (like vortex shedding) falls into this coherent category and that's why it magically shows up in SRANS because, as you point out numerous times, no explicit averaging is applied to the equations being solved. The averaging is implied.

I totally agree, of course.

I had to cover all the fundamental literature to see how the topic is arrived to be so confused.

Maybe, in future, a key point would be in developing a closure model where the meaning of the variable (ensemble averaging, time-filtering, etc.) is explictly present by means of the proper operator.
I think that a line to get a unified framework between URANS and LES (space-time) formulation is to not commute the operator in the equations.

Thus, some idea in the Pruett formulation could be introduced.
This is a research that, at present, has no interest in the community, probabily because it would not attract funds...
I will continue to write some ideas in my note in the next months.