[ggfiorillo]

Could it be that you are interpreting your heat transfer coefficient wrong? I am new to CFD but i read somewhere that the heat transfer coefficients that a CFD code will put out is different from what we see in empirical correlations. If your convection coefficients are larger, this may be the case. I tried finding where i read it but could not, so here is my best attempt from memory.

The CFD code looks at the wall heat flux [W/m^2] at a element on a wall, it calculates this some way without knowing the convection coefficient. The wall flux can be set equal to the following:

Q" = h * (Twall - Tnode).

So the CFD code uses this formula to back out h but dividing the heat flux at an element by the temperature difference between the wall temperature and the nodal temperature. This is very different than how we look at heat transfer microscopically. We consider a bulk or free flowing temperature and use this free flow temperature to determine a heat transfer coefficient. The heat flux at the wall is the same regardless of how far out you probe your temperature. So using the free flow temperature, which will have a greater delta from the wall temperature than the CFD's nodal temperature, you end up with smaller convection coefficients as a result and vice versa for what CFD codes calculate.

I have probably not done a great job at explaining, or perhaps this is your case but maybe it will help. In the mean time i will keep looking for the place where i read this.

Cheers!

[ghorrocks]

Good point Giovanni.

My explanation of what you are saying is that CFX does not know the fluid reference temperature to use in the Q" = h(Twall-Tref) equation. So it generates a reference temperature from the conditions near the wall. Most empirical studies of heat transfer coefficients have a clearly defined reference condition (the far field temperature of the inlet temperature), and this is usually very different to the value CFX generates. This means that the heat transfer coefficients generated by CFX can be very different to empirical studies.

One way to fix this is to use the expert parameter which defines the reference temeprature for the heat transfer coefficient calculation. I think it is "htc coefficient reference temperature" or something like that. Then you can manually define the correct reference temperature. Alternately you can take the convective wall heat flux calculated by CFX and work out the h yourself using your defined reference temperature in post processing if you like. If you use either of these approaches you should get heat transfer coefficient values matching empirical studies (assuming you have done an accurate simulation).