Conjugate heat transfer, thermal boundary conditions
 

Dear all,

I am fairly new to CFD, I hope the description of my problem is clear. I am using Fluent (Ansys 12).

My model is a room with dimensions 3.9x4.75x2.7m. At opposite sides of the room there is one inlet (average velocity 0.075 m/s) and one oulet (average velocity 0.28 m/s). The air coming into the room has a temperature of 289 K, also used as backflow temperature at the outlet.
I want to model the air flow pattern in the room, and the convective heat transfer at the surfaces. I also want to know the resulting temperature gradients over the surfaces and take into account the influence of the thermal mass of the walls, ceiling and floor. For this, I use shell conduction for all the walls, ceiling and floor with user-defined material and thicknesses of 0.1 to 0.2 m. As an external boundary condition I have to take either constant temperature or constant heat flux (other boundary conditions assume zero wall thickness).

Now I have three questions:

- Is it correct to use shell conduction in this situation? The main references I find where shell conduction is used, are for very thin structures as the fins in heat exchangers, not for heavy walls.

- To choose a temperature boundary condition at the inner side of the shell (so the side away from the flow domain), ideally, I would like to use the wall temperatures calculated by Fluent in the following way, to approximate adjacent identical rooms:
-> to assign the surface temperature of the floor to the inner side of the shell of the ceiling
-> to assign the surface temperature of the ceiling to the inner side of the shell of the floor
-> for the walls, I can simply couple the two sides of the walls
-> one of the walls is facing the exterior environment, so a constant temperature of 289K can be assigned
Do I need a UDF for this, or can I do this in Fluent directly? Is this a sufficient boundary information to solve the flow?

- Is this a steady-state problem? I am including buoyancy in my model, and am having trouble to make it converge. I make a first approximation by reducing the gravity to 0.098 and first order scheme. This simulation runs smoothly to convergence and results are nice symmetric. If, as recommended by the Fluent manual, I continue the calculation with normal gravity, the residuals get stuck and start oscillating around a certain value. I wonder of I may need a transient simulation?

Thanks for any help!

Sarah
PhD-student, KULeuven

One quick idea as per convergence issues:

You could step up gravity to 0.981, then to 9.81. Also, try lowering the inlet flows (maybe 50%?) and stepping those up as you get converged solutions.