[Sushobhan]

Hello everyone,

I've been using OpenFOAM for some months now and am fairly confident with the isothermal solver simpleFoam. Now, I'm trying to use buoyantBoussinesqSimpleFoam to solve a temperature field with the Boussinesq approximation, and I've seem to hit a dead end. I looked around the forum and Googled it, but I could not find an answer to this problem (most examples with the solver, including OpenFOAM's tutorial, are related to natural convection).

Here is what I am trying to do: I have air flowing over a heated horizontal plate of length L = 1.0 m (solving a 2D problem). The value of Re_L was set to 3000 to make the flow turbulent. I set the height of the domain to 5*L i.e., the top far-field boundary is "far away". So I have these faces: an inlet, and outlet, the heated plated, the top atmosphere at 5*L (symmetry), and also the empty faces on the sides for 2D simulation. I created a structured mesh with a finer mesh size near the plate using Gmsh and imported that into an OpenFOAM case. It passes all the checkMesh checks.

Here's what should happen: The air flows in at 0.03 m/s and 300 K from the inlet and, due to the noSlip condition and 310 K fixed temperature at the plate, develops a boundary layer for both velocity and temperature, whose thickness should grow over the length of the plate.

Here's what is happening: I am using a realizable k-eps model for turbulence with wall functions applied at the plate. When I solve the case with just simpleFoam (i.e., only for the velocity and pressure fields), I get a very good solution in about 300 iterations (convergence set to 1e-6). However, when I use buoyantBoussinesqSimpleFoam, with the same 0/ files for U, k, epslion, and nut and adding files in 0/ for T and alphat (I use the same conditions for p_rgh as for p), I get a terrible solution after as much as 6000 iterations (20x the number as with simpleFoam). My velocity accelerates to non-physical values (something like 100x the boundary condition!), and p_rgh and temperature also increase non-physically, and at positions in the domain where nothing should be happening i.e., near the far-field boundary.

Interestingly enough, if I set either g or beta to zero, the solutions looks fine, although obviously it is missing some of the physics. I was mostly following the settings from the hotRoom tutorial, so I'm surprised the solution is so poor despite the simpleFoam solution working very well! I've tried a finer mesh, lowering Re_L into the laminar range, and using the converged simpleFoam solution for U and p as my initial conditions for the buoyant flow problem - same problem every single time. Could anyone help me figure out what I am doing wrong?

I've uploaded my case file here if you'd like to see - just run either simpleFoam or buoyantBoussinesqSimpleFoam. The names of the walls are pretty self-explanatory as well. I am using OpenFOAM-4.0.

Thank you!