[Kestas]

I want to simulate free vortex flow. (ex. tornado, dust devil etc.)
I am simulating with ANSYS FLUENT 12





As you see secondary flow is forming quite well, but but conservation of angular momentum doesn't appear so swirl doesn't form.
Flow from sides(pressure-inlet) is set to be directional( -0.95 tangent, -0.05 radial).
I have tried transient and steady simulation but results are the same.


Here is some pictures of how it should look like:

from http://www.blwtl.uwo.ca/Public/Tornados.aspx
http://www.blwtl.uwo.ca/User/Doc/Tor...Simulation.pdf



from http://vortexengine.ca/cfd.shtml

dust devil http://www.youtube.com/watch?v=AwKc7H5gIWo>

So my question is: what can cause problem with conservation of angular momentum?



Actually I am studying computer science, but have great interest in CFD, meteorology and aviation(soaring).

[Kestas]

anybody can't help??

[Kestas]

... I succeeded
CFD is really interesting

[ahlo7]

Congratulations!
Where you made changes that let it works?

Thanks

Ahlo

[Kestas]

I simulated dust devil. Just give enough initial swirl using UDF, and use ground heat as heat source(no secondary flow pipe as in previous pictures).
Also there are few nuances related to simulation mathematical model - most meso-scale meteorogical simulations use LES with Smagorinsky-Lilly as SGS model, and atmospheric thermodynamics - initial vorticity, dry adiabatic lapse rate.
Also mesh quality must be good in regions of large gradients.
You can find much info about LES modelling of atmospheric convection on scientific articles websites(as springerlink).

[Charles Chandler]

Can you post some graphics of the results, such as the flow lines in an elevation view? I'm not a student of CFD, but I am a student of tornado theory, and I'm trying to track down the reasons for the difficulties in modeling tornadoes with standard fluid dynamics, using real-world parameters. The ground is not a relevant heat source -- the real-world situation is purely a matter of low pressure within the mesocyclone, and it's my understanding that the swirl ratio necessary to get a tornadic vortex to latch onto the boundary is not present in the mesocyclone. Furthermore, setting up high angular momentum just in the surface inflow is unrealistic. There has been a lot of work on experimentally-derived algorithms and coefficients, but why do we have to doctor the simulations to match the observed phenomena -- why doesn't standard CFD return an exact solution without being told the correct answer?

Any help here would be greatly appreciated.

[Kestas]

Currently I have another project.
http://www.stud.ktu.lt/~kesmaci/about%20vortex%20project%20short.pdf
http://www.stud.ktu.lt/~kesmaci/

This project has two phases:
1) simulation of dust devil test stand
2) recreation of a dust devil in field conditions

If you are interested too, write here. We can work together. There are need for computing resources.

[eplanting]

Ah, wow, the original post is old, too bad simulation requires someone to refine the mesh based on a judgement call, the software should be able to automatically do that. Too many users of FEA and CFD just use the statement "it's not science it's art" which isn't true, it's science, you refine the mesh by n^2 points until you get convergence as set by theoretical limits between simulated models. Anyway, it'd be a help to people who were ignorant of convergence to suddenly see an option that they didn't know about to refine areas of high gradients in the software package, i.e. leave the computer on all day to continue to refine and save results from each simulation. But oh well, I'm just someone who ansys didn't hire, thanks.

[Kestas]

Thanks for reply, eplanting, so how to refine mesh in central region, I found this option only in old versions of Ansys, does v13 have it?

Originally Posted by eplanting

Ah, wow, the original post is old, too bad simulation requires someone to refine the mesh based on a judgement call, the software should be able to automatically do that. Too many users of FEA and CFD just use the statement "it's not science it's art" which isn't true, it's science, you refine the mesh by n^2 points until you get convergence as set by theoretical limits between simulated models. Anyway, it'd be a help to people who were ignorant of convergence to suddenly see an option that they didn't know about to refine areas of high gradients in the software package, i.e. leave the computer on all day to continue to refine and save results from each simulation. But oh well, I'm just someone who ansys didn't hire, thanks.

[Centurion2011]

cool stuff. Congrats

[Kestas]

Hello, much time passed, but i decided to finish this project, more precisely to do transient simulation, to make it look like real world.
Now I got new PC, and ansys 13.
But first of all, I have one question - I want to increase mesh quality around center axis of simulation domain. How do you recommend to do this ? Any kind of meshing controls ?


Thanks,
Kestas