Hi, I am working on the bouyancy driven flow over isothermal vertical plate , I had generated a simple rectangular geometry with wall ,inlet,outlet and farfield in gambit, since the flow is by natural convection, what boundary conditions should I give for inlet and farfield, I tried giving pressure inlet for inlet, but I am not getting the laminar boundary layer profile when I observe velocity contour in Fluent, I am using boussinesq model for solving this problem.I am getting the above profile if I use velocity inlet condition for inlet by giving zero velocity at the inlet, but I consider the flow to be incompressible, so I think velocity inlet with zero velocity is not the correct boundary condition.

so friends I request you to help me in solving this problem.

sri

First, don't use any velocity boundaries. Use pressure boundaries only. Second, check the temperatures you specify for each boundary. All boundaries except your heated wall should have the same temperature. Third, check your body force specification. Fourth, don't use the Boussinesq approximation. Specify a temperature dependent density instead. Finally, check your mesh. If the fluid cells directly adjacent to the heated wall are smaller than the thickness of the boundary layer, you aren't going to get the correct solution.

Thankyou very much for your responce.I will try doing with non-boussinesq approximation. sri

I think that Evan Rosenbaum's reply is very good.

For natural convection analysis, first of all, couple of meshes(say, at least 3 or more meshes) should be located within the boundary layer in order to anylyze boundary layer profile. Otherwise, the solution is, in general, physically unrealistic. The boundary layer thickness is very thin and it is, in nondimensional form, Ra^(-1/4). I am sure that this concept is the most important one for natural convection analysis because many engineers have been successiful to follow this guide(construct grid net where 3 or more meshes are located within boundary layer). Then, Boussinesq approximation may be good approach. I was always successful with Boussinesque approximation.

Some useful comments : 1) If Ra is very large, say larger than 1.e07, you might solve the problem by time marching algorithm. 2) For natural convection, as far as my experience is concerned, Fluent 4.x is more stable than Fluent 5.x, if you can construct structured grid-net.

In the above, as you know, Ra is the Rayleigh number.

Sincerely, Jinwook

Hi Evan and Jinwook,

I thank you both for your valuable advice, actually in the problem which I am working Ra> 1.0e07, Jinwook as you suggested that for such problems , we have to use time marching algorithm, but I am not aware with that concept. could you please explain me clearly about time marching algorithm and how to use it.I am using fluent5.0 for my work.

Evan can you please tell me if I use a non-boussinesq approximation, how can I make a polynomial function which gives the relationship between temperature and density. I use "air" in the problem.

If you need the data clearly please let me know.

thanks.

I think fluent package is the worest ever for natural convection if you try to set the solution to compressible flow type you will get unrealistic answer whereas you should expect the the incompresible and boussnisq approx are just partialy correct and limited by the approximation. Is there any reason why compressible model should not work?

Using the "Define - Materials" menu select air and there is a drop-down box, directly above the numeric density value, that allows you to select piecewise-linear or polynomial specifications for the density.

Hi friends, I am still facing difficulty in solving the problem which I mentioned previously. The mesh which I have generated in gambit is much refined, will it effects the solution?? Can anyone suggest me why I am unable to solve the problem using boussinesq model .

Thanks.

[rabia]

Hi Evan,

I am also trying to do the problem related to natural convection over a horizontal solar panel which is placed under the solar irradiance. I have made the domain outside the solar panel which is 50 times bigger. I applied the DO solar load model to the solar panel. The boudary condition for solar panel is wall having the temperature of 343 K. While the boundary condition for domain is pressure outlet. The temperature of the domain is 300K and 0 gauge pressure.
Beam direction for solar load is (0,-1,0) and gravity is 0,-9.81,0 .

Kindly anyone help me out why i am getting reversed flow and my solution is also not converging. I am using bousineq approximation.