Hi. I´m trying to model heat transfer in a fluid domain (naphta). The domain is a semicircle cut by a plane in the bottom, and it has two fixed temperatures in the bottom (low) and in the top (high).The walls are adiabatic. Watching this I expect to have a little effect of natural convection that increases the flux (respect of conduction). But when I include the k-e turbulence model (I do it because cfx recomends this when the Rayleigh Number is bigger than 10^10) the flux becomes extremely big (out of what can be expected). It seems to converge, so I don´t know which could be the problem. Can anybody help me??

Hi,

How can the walls be fixed temperature and adiabatic?

The error could be due to numerical issues (not yet converged, not mesh independent, wrong differencing scheme) or physical issues (unsuitable turbulence model, running steady state when it should be transient or missing some other important physics). We cannot be more specific unless you provide more details.

Glenn Horrocks

Glenn, the top and bottom wall have fixed temperatures. The others (sorry I didn´t say it before) are adiabatic. The turbulence model is the recomended by the manual (k-e for natural convection) and we are running it in steady state. We are using constant density and the correction by Boussinesq. Also we are now trying to solve it in transient, but we think this could be solved in steady state. Thanks. Juan Martin.

Hi,

(Sorry about the wait, just come back from holidays)

Many natural convection models have large scale flow instabilities, even when they at first glance appear steady state. You may need to run transient to properly model this effect.

Some other issues to consider are listed in my previous post.

Glenn Horrocks