Maybe, that's a atupid question: I'm working in an air volume with different temperatures in each surface. This temperature difference must move the air according to natural convection. I have to decide which model of density, conductivity, Cp, and viscosity I will use, as well as the viscous model(because if I use the "air" that fluent has, with all the properties as constants, it doesn't work properly). I need a good-enought model for natural convection. Any suggest?

You first have to chose a model for density variations with temperature. It depend of the temperature difference your are working with. If it's less 15°C (in air) you can used Boussinesq approximation. If it's bigger, you should use incompressible Ideal Gas (you have to specifie the operating density in the operating panel, put the density at the minimum temperature, if you don't specified the operating density, Fluent uses the mean value computed on the whole domain, it can leads to a unphycal flow without convergence on residuals). For other properties, you can use the air material of Fluent and change properties using a picewise linear variation. You just have to specifie value of properties for some temperatures. For steady flow, I take into account properties variations only for conductivity and viscosity. If you temperature difference is not large, people says that is not necessary to take into account properties variation.

Thank you very much! Your answer has been very useful to me.

Maybe, in the future, I will be interested in presurre differences in the volume.May An incompressible Ideal gas model for density variation affect the results? Could it be better to use an ideal gass model?

I'm studing the heat transfer coeficcient in the volume,what viscous model will be better?

Thanks in advance.

There are some different between incompressible ideal gas and boussinesq approximation only if temperature difference is large. Incompressible Ideal GAs : density variation is used in all equations. Boussinesq approximation : density variation is used only for buoancy force. Visicous model : ready user guide www.fluentusers.com Modele depend on your configuration : Ra number, curve of streamline etc etc. Laminar or turbulent flow? If it's laminar : no problem, choose laminar model If it's Turbulent : k-epsilon standard is a high reynolds numbre, k-epsilon RNG can be ok for low reynolds number, be careful on near wall treatment. Reynolds stress is maybe better that k-epsilon, but it leads to a higher CPU (memory and time) (5 equations).