Hi, Iam trying to simulate turbulent natural convection in a square cavity using Fluent 6.0 the problem specifications are:

1.Square aluminium cavity of area 1 m2 with top and bottom walls being adiabatic, Hot wall temp=320 k,Cold wall temp=300 K.

2.Properties of Air(Boussinesq assumption) being taken as density=1.177,Cp=1007,Viscosity=1.857e-5,K=0.02653,Co-eff of thermal expansion=0.018,dT=20 K,g=9.81 values are chosen such that Ra for the flow is 1e10.

3.Turbulence model:K-E standard,with enhanced wall treatment and full buoyancy effects.Defaults for all model constants.

4.Mesh size=200x200,double sided with ratios=1.02

5.Solver : 2ddp,steady,with default under relaxation parameters.

Under these settings the maximum Surface Nusselt number for the cold wall was obtained as -80.355 while the experimental value was -138,can anyone suggest reason for deviation ?

Please suggest if the problem has been set up properly in Fluent and alternative formulations Thanks, Hemanth

Three things you should check jump to mind.

First, confirm that y+ near the walls is about 1. This is required for the enhanced wall treatment.

Second, consider not using Boussinesq. We have had better results using ideal gas law or density as a function of temperature.

Third, consider using RNG instead of the standard k-e model. I use the standard model whenever possible and it works well for many buoyancy problems, but it does have some weaknesses at low Re.

Hi, Can you tell me if steady state formulation is appropriate for this case or should I be using unsteady solver ?I found that a time step size of 0.1 would be fine.If unsteady solver is the right option then what should be the number of time steps i should employ ? Thanks, Hemanth

Just make sure that ur providing correct input values in the "Reference Panel". Also check that u have defined g correctly. Mahesh

With natural convection you never know. Try steady-state. If it has convergence problems, try unsteady.