Large Eddy Simulation

[Mohammad Hashemi]

Hello Everyone, 

We would like to model a supersonic jet using LES models such as dynamic Smagorinsky or WALE in Basilisk. Is the implement of these models different on octtree mesh in comparison with uniform mesh? Did anyone implement any LES model on Basilisk??

Thanks, 

Mohammad  

[Wojciech Aniszewski]

It's not as much connected to the Octree structure as it is to the AMR method itself (point/block): varying grid size.

Yes, to ensure consistency, the implementation of LES on AMR should be different than that on uniform
grids, and for more than one reason. If you do the grid=filter approach, one such reason would be commutation errors,
introduced due to the fact that on uniform grids certain terms vanish, while they will appear at the refinement boundaries,
once delta (grid size) changes. Other reason would be inconsistencies in the TKE transport; there's a good coverage of that here
https://doi.org/10.1016/j.compfluid.2018.12.012

I'm not aware of anything already published using Basilisk, although there are some unpublished attempts.
w

--
/^..^\ Wojciech (Wojtek) ANISZEWSKI ► Post-doctoral Researcher
( (••) ) [Fr: vôitek anichévsky] ► Sorbonne University
(|)_._(|)~ [Eng: voyteck aanishevsky] ► Institut ∂'Alembert
GPG key Web [English] https://tinyurl.com/y73t4xsg [old! 2015]
ID:AC66485E [Polish ] https://tinyurl.com/yckark7n [old! 2014]

[Antoon van Hooft]

Hallo Mohammad,

Even in the absence of any turbulence model, I'd say the Tangaroa example qualifies as an LES,

http://basilisk.fr/src/examples/tangaroa.c

Its not a too bad idea to let diffusive numerical schemes dissipate the second-order-moment carrying structures as the are "cascaded" (resolved) towards the grid scale.

For the dynamic Smagorinsky model see Arne Abockmann's sandbox,

http://basilisk.fr/sandbox/abockmann/Dynamic_Smagorinsky.h

At the risk of overly emphasizing my own efforts,

There is the Vreman SGS model,

http://basilisk.fr/sandbox/Antoonvh/vreman.h

which may be implemented via

http://basilisk.fr/sandbox/Antoonvh/SGS.h

Or via Arne's improved version:

http://basilisk.fr/sandbox/abockmann/SGS.h 

A test regarding Kolmogorov's scaling law for the second-order structure function is presented here:

http://basilisk.fr/sandbox/Antoonvh/isotropicLES.c

It failed or passed, depending on your standards.

Note that with the "preliminary result disclaimer", there is a page in this work dealing with Basilisk + LES:

https://link.springer.com/article/10.1007/s10546-018-0335-9

Finally, I am working on a sandbox page (and accompanying manuscript) presenting these LES results:

https://vimeo.com/292329175

The soundtrack is composed by Debussy and the methods will be uploaded soon.

I share the concerns of Wojciech, but go with the philosophy that the fidelity of an LES should come from the resolved large eddies. 

Antoon

[Arne Bøckmann]

A big disclaimer on the Dyn. Smag implementation; it is clumsy and completely unverified.

Your simulations are really intriguing, Antoon!  Knowing very little about meteorology, what is that invisible ceiling the thermal bubbles collide with?

Some sort of boundary layer?  Does the motion stop there due to condensation?  Or is it related to the stability of the atmosphere at different vertical temperature gradients?

BR, Arne

[Antoon van Hooft]

Thank you Arne, well spotted on the apparent ceiling,

>Or is it related to the stability of the atmosphere at different vertical temperature gradients?

Yes, the slice in the rear reveals that temperature-gradient-carrying rising plumes accumulate to form a stably stratified (inversion) layer on top of the convection cells.  

Notably, this gradient sharpening process may also be captured in a simple 1D model, even when the vertical (turbulent) transport is modeled to be diffusive.

https://vimeo.com/290115624

Antoon