[mickfromlyon]

Hi,

I'm currently trying to simulate a simple 2D natural convection of a vertical plate. The attached diagram sums up the simulation.

The air has constant parameters except for the density which is temperature dependent. Gravity is on. Solver is second order and ‘Body force weighted’. Solution has converged. I tried two viscous models that give me rather different results (for the heat source)

k-e (realizable/enhanced wall treatment/thermal effects):
Total heat flux: 257 W/m2
Heat transfer coefficient : 3.58 W/m2-k
Y+ : 1.46

k-w (standard):
Total heat flux: 309 W/m2
Heat transfer coefficient : 4.30 W/m2-k
Y+ : 1.40

Based on Rayleight number and analytical formulas, the problem is indeed turbulent and the heat transfer coefficient should be something like h = 5.28 W/m2-K ==> 316.8W/m2. Fluent turbulence notes also suggest that k-omega shows “superior performance for wall-bounded boundary layer”.

Based on these results, can I assume that k-w is more suitable for this kind of simulation? And more generally, to simulate air cooling of a fairly hot body? Or did I do something wrong? Thanks

[LuckyTran]

I'm not too surprised they don't quite match, since you're using fully turbulent models to predict what should be a laminar-to-turbulent transition problem. But some housekeeping is in order to make sure we don't do stupid things.

I remember for some k-epsilon models (I don't know about realizable k-e), the buoyancy term in the epsilon equation is off by default and the user must manually turn it on.

I would urge you to do a grid convergence study before making any conclusions. You might even find that the result is worse with a finer mesh.