[pilafa]

Hi,

I am solving this problem (see image attached):

- Filling of a tank with water. The tank is initially full, we just fill it with new water (water level is constant)
- Axi-symmetric
- No free-surface modeled
- inlet section's surface is equal to outer surface (Ain=Aout)

My question is about the boundary conditions to apply.

I first tried the traditional "velocity prescribed at inlet" and "P=0 at oulet", but it is very difficult to have a good mass balance (The output velocity that I observe is not equal to the inlet velocity prescribed).

I then tried the BC showed on the image. The convergence is much easier, but I am not sure if it's "legal".

- Inlet: Vr
- Outet: -Vr (it should be that way in order to respect mass balance)
- top surface: this surface is at "open air" (but I don't consider free surface). I put P=0, and Vz=0 to prevent the fluid to exit the domain

Do you think that this set of BC is valid?

Thanks,
Pierre

[FMDenaro]

What do you mean for Vr, is a constant value prescribed on all the nodes constituting the inlet and outlet faces? Apart from the fact that a uniform velocity profile is physically questionable (particularly as outflow), it does respect the mass flow and it must work. Thus, I think you have some bug in the code (or I do not understand well your setup...)

[pilafa]

Hi FMDenaro,

What do you mean for Vr, is a constant value prescribed on all the nodes constituting the inlet and outlet faces?

Yes.

Apart from the fact that a uniform velocity profile is physically questionable (particularly as outflow), it does respect the mass flow and it must work

You are right. I guess a parabolic function should be better.

Thus, I think you have some bug in the code (or I do not understand well your setup...)

Actually, the convergence is quite good with the BC I proposed. I never say it didn't work
It's just that I never used this configuration before and wasn't sure if that make sense or not.

Thanks!
Pierre

[FMDenaro]

Originally Posted by pilafa

Hi FMDenaro,



Yes.



You are right. I guess a parabolic function should be better.



Actually, the convergence is quite good with the BC I proposed. I never say it didn't work
It's just that I never used this configuration before and wasn't sure if that make sense or not.

Thanks!
Pierre

from a mathematical point of view the problem is well posed, however the numerical solution you get is physically acceptable only if you assume inviscid flow conditions

[pilafa]

..mmm...interesting.

If I follow you correctly, you say that prescribing the inlet AND outlet velocity would be physically valid only for inviscid flow (no viscosity). Correct?

I don't understand why

Could help me with that?

[FMDenaro]

Originally Posted by pilafa

..mmm...interesting.

If I follow you correctly, you say that prescribing the inlet AND outlet velocity would be physically valid only for inviscid flow (no viscosity). Correct?

I don't understand why

Could help me with that?

well, the fact that an outflow velocity profile is exactly the same of the inflow can happen also for some particular viscous flows but for very restricted conditions. For example the Poiseuille solution, mathematically valid for any Reynolds number but physically not stable just at low Re number.
On the other hand, an inviscid flow hypothesis cancels the shear wall effects and makes more acceptable such profile due to the energy reversibility (no energy dissipation, no entropy production)

[pilafa]

ok, I see. Thanks for this explanation. I agree with you.