Natural convection in a square volume (2D)
 

Hello World.

I described the following problem in the main forum as well. But there I tried to compute it with COMSOL. This time I simulated with FLUENT and had some other problems.

Problem Geometry is just a square. The problem has two adiabat boundary conditions at the top and at the bottom. On the boundarys to the right and to the left I created temperature-conditions. The left wall is 100 K hotter than the other. Concerning the flow there are 4 similar conditons: walls (u=v=0) on all boundarys.

In materials I declared the boussinesq approximation. But still I have to specify a density.

When solving the problem I get a contour of Stream Function that looks like the one I have from a reverence solution. But when I display the density it is constant all over the area. But how is that? It has to follow the boussinesq law?

Case, Data and Mesh are stored here: http://rapidshare.com/files/87129064/square.rar.html

Re: Natural convection in a square volume (2D)
 

Please consult section 13.2.5 of the user guide. Specifically, "Boussinesq model treats density as a constant value in all solved equations, except for the buoyancy term in the momentum equation".

In addition, temperature difference of 100 degrees may be too big for this model.

Re: Natural convection in a square volume (2D)
 

Ok, Thanks. I will look into the user guide. But why should the temperature difference be too big?

As far as I know these kind of problems are specified by Prandtl and Rayleigh numbers.

The problem is described in REDDY,GARTLIN (2001). There you can find reverence solutions for some Rayleigh numbers (page 278-279). There is an online-edition of this book at this place: http://books.google.com/books?hl=de&...LTcI#PPA278,M1

If I choose all variables so they match one of the dimensonless numbers I thing 100 K may not be too big ...

Re: Natural convection in a square volume (2D)
 

The Boussinesq approximation is valid when beta*(T-T0) << 1 (this is stated in the Fluent User's Guide). If your T is too large, this may not hold. Please check if your model is within the valid range.

Re: Natural convection in a square volume (2D)
 

Well it is not. Especially on the "hot side" ... But the solution for velocities and stream function are pretty good compared with the reverence solution I mentioned above ...

Re: Natural convection in a square volume (2D)
 

For bousinissque approximation the temperature difference should be at max. 35 degree C.