[arvindpj]

Just curious, when specifying constant heat flux bc for the wall, I am noticing that the computed wall temperature varies with Y+ values.

I ran test cases of flow inside a heated tube and noticed that for LowRe settings, I had to use a Y+<<1 eg., Y+=0.02 to get mesh converged wall temperature results.

For regular wall function approach, Y+>30, the wall temperatures always varied with Y+ values.

My working fluid is water heated near boiling temperature. My goal is to simulate sub-cooled boiling phenomenon. And most sub-cooled boiling models use special wall functions and hence this conundrum of estimating wall and near wall temperatures accurately.

Cheers,
Jay

[LuckyTran]

Hello. You have rediscovered why everyone always does a mesh sensitivity study.


It's generally much easier to converge fluxes (gradients) than absolute values because the absolute value is the summed effect (or spatially integrated effect) of the fluxes.


For turbulent flows, it's also quite sensitive because of the use of wall models.

[agd]

And to add to what LuckyTran said, heat transfer calculations typically require a much tighter wall spacing than y+ < 1. Most of the work that I have seen from colleagues doesn't converge well until y+ < 0.1 or less.

[lucpotier]

Hello, I'm currently strugling with the same issue. I managed to have y+=1 for my resolved simulation (no wall model here). The friction is good but the wall fluxes, the gradient of temperature seems to be underesolved by a large factor.

Could you explain why the criteria y+=1 ensure good friction but is not sufficient for Temperature gradient ?
Is there a thermal y+ that could be found to ensure thermal boundary layer resolution.

[LuckyTran]

Recall the thickness of the thermal boundary layer in comparison to the hydrodynamic boundary layer for Pr or 1 and slightly less than 1.

As for why some people need silly low y+ values, it has less to do with y+ value of the wall adjacent cell and a lot more to do with the actual resolution, i.e. the number of cells in the boundary layer and their distribution.

[FMDenaro]

Originally Posted by lucpotier

Hello, I'm currently strugling with the same issue. I managed to have y+=1 for my resolved simulation (no wall model here). The friction is good but the wall fluxes, the gradient of temperature seems to be underesolved by a large factor.

Could you explain why the criteria y+=1 ensure good friction but is not sufficient for Temperature gradient ?
Is there a thermal y+ that could be found to ensure thermal boundary layer resolution.

There is a basic confusion in your question: y+ is a function of the wall distance, nothing but a local Re number measured from the wall. That is a measure of the cell distribution in the dynamic BL.

Now you asked for the heat flux, that is a BC that enters into the energy equation. Thus, why do you look at the condition for the dynamic and not the temperature BL? In theory, you should define a counterpart of y+ for that. If the parameter is equivalent to the local Re number, we should talk of a local Peclet number measured from the wall. Only if the fluid has Prandtl number =1 we could assume a similarity between the thermal and the dynamic BL. From a numerical point of view, since we need to prescribe a derivative at the wall, that is y+ (thermal) = 0, you need to have 2-3 cells adjacent to the wall at y+<<1.