Under-predicting heat transfer
 

Hello, I am trying to simulate the natural convection of heat from a hot pipe in a room.
I have done experimental testing to determine the convective coefficient and the CFD sim correlates nicely if I specify this coefficient on the external boundary.
I have now tried to model an ambient region around the pipe and let the CFD take care of the heat convection to see what the results are.
The simulation is under-predicting the heat transfer, the experimental data has a convective coefficient of 19W/mK and the CFD results in values around 11W/mK.

I am in the middle of performing a study on the volume of the ambient region to determine how much of an effect it has, so far increasing the volume does increase the coefficient but I will reach the computing limit before it gets close to experimental.
The ambient region boundaries has a very large convective coefficient (50000W/mK) in an attempt to try and model it as a sort of far field.

I am using K Epsilon in the ambient region with a (Xu) Buoyancy driven two layer type model. The Y+ next to the pipe is all less than 2. The density is a polynomial that I have entered from a reference book and im using Sulerland law for thermal conductivity and dynamic viscosity.

I would be very very grateful for any help or previous experience in this type of case.

Thanks a lot.