CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   right boundary conditions (

Michael June 9, 1999 04:29

right boundary conditions
Hello everyone,

I have a question about the right choice of boundary conditions. I have started with a simple air flow in a room with a heated cube on the floor. The inflow is from the ceiling downwards specified. The outflow should be over the whole floor. (expept the cube's area) If I use a zero-gradient condition on the floor the solution converges quite well. But unfortunately there's inflow through the floor caused be the free convection at the cube's side walls. To avoid this unrealistic inflow I have tried to specify the velocities at the floor as well and used a little area of zero-gradient condition in a corner of the room for the mass flow difference. But the solution did oscilate with high frequence and the residuals did not decrease either. (in comparision the first case) Could someone explain what happened or give me an advice for choosing the right BC's for my problem.

Thanks in advance.


Patrick Godon June 9, 1999 10:26

Re: right boundary conditions
Hi there,

according to what you write it seems quite physical that due to the convective currents adjacent to the wall of the heating cube, an inflow would take place from the floor (an open boundary) upwards. What's wrong with that? This solution you obtain seems OK to me.

Cheers, PG.

Michael June 9, 1999 10:59

Re: right boundary conditions
Hi Patrick,

in the experiment that I want to simulate the inflowing air from the ceiling left the room through a porous floor over the whole area driven a ventilation unit below the floor. So there was no inflow from the floor due to the natural convection at the cube. To avoid this inflow I changed the BC at the floor from zero-gradient to a reverse inflow BC and specified all variables. (One problem might be that I have to specify the temperature at the floor, which I don't know in advance) Up to temperature differences of 10 K (Inflow-Cube) everything is fine. But e.g. at 30 K I get those before mentioned oscilations in my solution. Transient approach with small time steps were not successful.

Thanks for your comment,


Patrick Godon June 9, 1999 11:30

Re: right boundary conditions
Thanks for giving more details.

Did your oscillations occur next to the heated box?

or on the floor?

On the floor the conditions are on the velocity or on the momentum ?

I guess the conditions should be imposed on the momentum, since you want to conserve the flux of momentum from the ceiling through the floor. For sure the flux of momentum through the floor is not uniform in space.


Ken Light June 9, 1999 12:52

Re: right boundary conditions
Wouldn't a lower static pressure BC below the floor be more realistic? If the ventilation unit is under the floor, then I assume that is what is ultimately causing the air flow in the room.

John C. Chien June 9, 1999 12:59

Re: right boundary conditions
(1). Your exit boundary is located in the region where there is a possible flow reversal due to strong natural convection from heated cube. (2). Two extreme case solutions. (3). Forget about the floor and move the boundary down to the solid floor with exit conditions applied at the side of the sub-floor space where the suction source is located. You may not like this approach. (4). Cover the floor area near the heated cube with solid wall to eliminate the flow reversal the the floor exit boundary location. This should work. (5). The most accurate method is to cover the modeling of the porous floor as part of the interanl structure, with the actual exit boundary located as the first approach in (3). This is because the porous floor will redirect the flow near it based on the porosity value of the floor. (6). Always move the computational inlet and exit further away from the main region of activity, in this case is the interaction of heated cube flow and the flow near the porous floor. It is not a simple problem to handle. And nobody has said that fluid dynamic problem is simple to solve with a CFD code. Good luck.

Patrick Godon June 9, 1999 13:07

Re: right boundary conditions
Yes, since the flow is compressible. Then only after averaging in time and space should the momentum be conserved. In theory all the variables can be prescribed (given rho, P, v) and then used to treat the boundary conditions using the method of characteristics, which are actually the quantities that propagate through the boundary.

All times are GMT -4. The time now is 16:16.