Elmer Discussion Forum

Hello,
I'm trying to simulate the transportation of gas inside volumes. I tried to base myself on the tutorials, but I'm not sure I did it right.
I putted all walls on no slip, an entrance and a exit for the gas as in the tutorial. For the simulation I used the transient and BDF setups. You can find in attachment the case.sif and the result illustrated with paraview (the volume is almost cylindrical, I need to succeed to use the 3d because it's just the first step. Later I'll use non symmetric volumes).
First of all you can see that in the bottlenecks the gas does not seem to go through. Is it just the representation that fail to use unique tetrahedrons or there's a problem?
Then the vectors does not look correct to me, they seem to go directly against the walls, without really following it.

The question is simply, do I use correctly the Navier-Stokes simulation, given in the attached case.sif?

Thank you very much for your help

Hello,

Looking at your sif - a few comments. Air Flow inlet velocity is 10 m/s, so the flow could be turbulent: use a turbulence model, for eg. kepsilon. Also remove NS Convect = False from Equation 1 - you need convection. Didn't quite understand why you used two materials - air and ideal gas. If there is only one fluid, i.e, air, please use navier-stokes and k-epsilon solvers with just one coupled non-linear iteration of both solvers.

-kumar

Hi

I think that for ideal gas you should probably use the bubble stabilization. It is more generic while the standard stabilization assumes more strict form of the equation.

-Peter

Hello,
Thanks for your answers,
To put convection helped, I already got a more convincing result, but not yet very different.
For the k-epsilon, I'm quite ignorant. I tried to figure it out, especially from the elmersolvermanual.pdf, but it's not described.
I tried quickly a naive approach, putting it active an launching it just to see what it gives. I just got NaN as Relative change values and it finishes on an ERROR.
I obviously use it wrongly (sorry, I'm really a beginner in FEM). Can you give me some indications to use it correctly please? thank you very much.

PS: Changing to bubble stabilization didn't have significative effect; the case.sif is the same except for NS Convect

I'm back after the summer, I tried what you suggested and put here the solution, thanks a lot, it helped a lot.

So the main trick was to add the k-epsilon in the equation. You can see it in the attached case.sif.

Thanks again for your help.