Dear CFD users,

I have a conjugate heat transfer problem in a complex internal flow geometry. The analysis procedure I have used is to keep the convection study separated from the conduction study, due to limitations in computational resources, and also due to the fact that mechanical integrity analysis will anyway be performed separately (with a different code) for this problem.

For the post processing of the convection study I have up till now used T_ref = Tbulk = a global temperature, which I know will not work if the fluid heat is entering and exiting the solid at the same time (at different locations) due to negative HTC's. The benefit of this approach is however that the grid dependence of the HTC is eliminated and the solution is roughly independent of the wall boundary condition (I use an "average" constant temperature at the wall). The other way would be to use the approach that is standard in most CFD codes; the near wall node temperature as a reference temperature (T_ref = T_nwn). But then the solution is probably grid dependent and surely strongly dependent on the wall boundary condition (i.e. specified temperature level in my case). What would You prefer?

regards Andreas

Andreas,

You can define the htc and the bulk temperature in different ways. If you are solving internal flow, you should take the average temperature on the section. You should take total temperature if the flow is highly compressible. If you are solving external flow, you should take the free-stream-temperature, in the region where the effects of the boundary layer are negligible. I have seen bulk temperatures taken as the adibatic wall temperature (that is, you compute with adiabatic walls, and use the wall temperature in following runs). You should make sure that you use always the same definition.

Peter

Hi Peter,

thanks for the reply.

How big would you think the error could be by using the global bulk temperature and an average wall temperature for the convection calculation? I guess that if I change the wall temperature a couple of hundred degrees locally in wall this might affect the flow field, but not too seriously (perhaps increase local Reynolds numbers 10-20%). And I would nevertheless not be able to see much of sublayer effects (changed fluid properties near wall) since I am using wall functions to calculate the qwall (it's a turbulent flow simulated with the k-epsilon standard model). Would you say that I am totally off, or just making reasonable engineering calculations?

looking forward to hear from you again

Andreas

Andreas,

The definition of the htc for instance in FLUENT is different to the one you can find in literature: The point is to define properly the bulk fluid temperature. I computed the htc with the mass-averaged total temperature over one section and found differences with the standard definition of htc in FLUENT up to 40%. It has been found that the choice of the turbulence model shouldn't introduce large differences in the results. My opinion is that if you need to match your results against empirical correlations (Sieder-tate, Dittus-Boelter), it is better to define the htc with the mass-averaged total temperature rather than taking the htc distribution from FLUENT. An interesting point that I found is that by modifying the metal temperature, say +/-200K you can still match fairly well empircal correlations. On the other hand, if inlet and metal temperatures are very close, the agreement with correlation disappears. Any explanation for that?

Thanks

Peter