[Arvin]

Hi everbody,
I am simulating natural convection over a horizontal heated plate. The geometry is basically a tank.I have used boundary layer mesh near the heated plate, but I dont know if it is fine enough. How do i check that?

I have used temperature dependent function of density instead of Boussinesq approximation ( working fluid is water). I have not changed under relaxation factors. There is no change in velocity and temperature of water (0.0017 m/s and 306K) even after 20,000 iterations. Solution doesn't converge. Please help me with your suggestions.Thanks!!

[flotus1]

There is no change in velocity and temperature of water (0.0017 m/s and 306K)

If the velocities and temperatures remain constant throughout the domain, the solution is actually converged.

[Arvin]

But, I hear people generally use Boussinesq approximation for a natural convection problem. Is it ok to use density as a function of temperature? If so, what is an ideal expression that you would suggest?
And I am supplying heat flux to the bottom plate, so is it necessary to specify the heat generation rate also in BC?
Thanks for your interest. I would appreciate if you can help me with these questions.

[flotus1]

Boussinesq approximation is often used because it is a very good approximation when the density changes are small.
Since you are dealing with water, you can definitely simplify your model with the approximation instead of a temperature-dependent density.

I can not see what is on your screen, so I have no idea what you mean by "bottom plate" and "BC".

[Arvin]

I am simulating by supplying heat flux for producing high temperature source at the bottom, do I need to give heat generation rate along with the heat flux?

[Zaktatir]

Boussinesq is quite a good approximation if and only if alpha*deltaT is smaller than one alpha being your thermal expansion coefficient. You may use the incompressible ideal gas keeping the dependency of your density only on temperature. Water may be however treated as incompressible but not a gas and therefore use first of all Boussinesq and check the variation of your density.

Mesh has to be fine enough and use low-Re turbulence model. There are a lot of things to check but start with that at first..

Keep it real

[flotus1]

That depends on what exactly you are trying to simulate. I really can not know.

[Arvin]

What do you mean by low Re turbulence model? I cannot find it anywhere. And how do we know deltaT value for checking that expression?

[Arvin]

I am trying to find the boundary layer thickness developed at the bottom.