[Tobi]

Hi all,

I have a question to all of you. Maybe you are more familiar with that problem. The case is very simple. A 2d rectangular with an outlet velocity of 1m/s, p is fixed at the inlet with fixedValue; the walls are zeroGradient for U and p. The passive scalar is set to zeroGradient at each boundary. With setFields I generate an area with values = 1 for the passive scalar (everywhere else its zero).

Case 1 has no mesh motion
Case 2 has mesh motion due to the velocity field at the inlet (flux after the iteration = 0 at the inlet face).

The position of the scalar is the same (as expected) but why is without mesh motion the scalar diffusive? The transport of the scalar is only done by time and convection
(no diffusion term). In the mesh motion I added the mesh flux to the solver, but can this be the reason for less diffusion? But if I check the code:

fvm::ddt(S)
+fvm::div(phi, S)

There should be no diffusion in the case with no mesh motion. S is only transported by the flux, which is everywhere constant. :/

Here you can check the results: www.holzmann-cfd.de/cfd-online/passiveScalar.avi


Any hint is appreciated.
Thanks in advance,
Tobi

[chriss85]

Are you sure there should be no diffusion at all? As far as I know, there is always some artificial, numerical diffusion (though I'm not sure if this requires that an actual laplacian is present in the equation).

I have no idea why the diffusion is smaller with mesh motion, but I have no experience with dynamic meshes at all.

[Tobi]

Hi chriss,

you are right. There is always Diffusion, hence there is only convection. But like you mentioned, I also don't know why I get the better result with mesh motion. Maybe it's due to the fact that the mesh flux reduces the convection term itself. (U-Umesh) which is smaller within the mesh motion case.

[kmooney]

I would say that the moving mesh case has less diffusion because there is near zero relative advection taking place. Like you said, with (U-Urel)~=0, there really isn't much math going on to even allow for numerical diffusion.

[fabian_roesler]

Right, I agree with kmooney. The relative advection in the mesh motion case is smaller and so the artificial diffusion is smaller as well.
What schemes did you apply and have you tried with a less diffusive scheme?

Cheers

Fabian

[Tobi]

Hi Fabian,

at the beginning I was wondering why its less diffusive but the reason for that is exactly what we are talking about. (U - Urel) ~ 0. I also tried with different schemes, second order, 1st order, not much difference.

Thanks for all feedbacks,

[Hannes_Kiel]

See here for an explanation