Hi, I have the same problem as two other users in this forum. Since answers are rare I try it once more: I try to model a heat sink in free convection. To test various geometries I want to cut one fin from the middle of the heat sink and use periodic boundary conditions for the left and right borders to keep the grid small. So far no problems. To get a free buoyancy driven flow I selected pressure-inlet at the bottom and pressure-outlet at the top of the domain. Both got the pressure value 0 Pa. The fins are at constant termperature. But the solution dont want to be stable. I get an upstream next to my isothermal fin wall but also an downstream inbetween two fins. That seems to be not realistic since the gravity is pointing downwards. Modelling the same problem in forced convection had no problems, is easy to converge and agree with physics. But I have no idea why free convection doesnt work. I would appreciate any advice or experience.

Thanks. Jan

Hi Jan,

I solved that kind of problem with an electrical analogy, which is the following:

It takes a voltage drop to get an electric current flowing. So, it also takes a pressure gradient to get a convective flow to move.

I had that kind of difficulty while modeling an industrial oven with one inlet and one outlet. Even if the two of them where exposed to air there was a pressure gradient that I did not define well, so the flow was never taking place. As soon as I played with the pressure gradient things started to change.

There is something else related to this kind of problems, but first try that out and then we'll get into more detailed aspects related to Fluent.

Best regards

Carlos Ortiz

First of all you dont give any periodic boundary condtions, take only one fin and try to solve the problem. And in natural convection meshing plays an important rold in getting satisfactory results. For this you have to take boundary layer mesh at the walls. Go for some fine mesh. And also, you have to select proper time step ( below 0.2) and the solution should converge in each time step. Initially it will take lot of iterations to converge but after that it will converge with in less time. Try in this way, if you are satisfied with the results you can go for periodic-boundary condtions.

Venu Gopal