I'm trying to model natural convection in a vertical tube with a closed warm end and an open cold end (including and excluding axial heat flux in the wall). I suffer serious convergence problems. I tried several appoaches: initialysing a velocity profile - upward at the wall, downward in the centre of the pipe; modelling only 45 degrees of the pipe and periodic boundary conditions; varying the temperature difference between upper and lower bc. so far no physically reasonable results. has anybody some experience in this field or some hints?

Hi,

What Rayliegh number are you operating at? Large scale transient instabilities often occur in natural convection and these can make convergence tricky, especially if you are attempting to get a steady state solution. In fact it is quite possible a steady state solution does not exist.

Try running a transient model.

Glenn Horrocks

Actually I'm not quite sure how to calculate the Ra-number correctly. Like I mentioned I'm modelling free convection in a pipe with a heated closed end and an open cold end under different angles of inclination. If I take the global temperature difference, the length between hot and cold end and sin(angle.of.inclination)*g the Ra-numbers seem to be way to high for the vertical pipe. I'm studying the literature right now. I'm aiming for a steady statesolution and laminar convection.

Hi,

As I said, if you are using a laminar steady state simulation and the flow is really turbulent and/or transient then it won't converge, and if it does it will be rubbish. You need to check that the flow is really laminar and steady state. I'll bet that it is transient.

Glenn Horrocks