[blackbow]

Dear CFD-Community

Info: My PDF File was to large for the forum. I uploaded it here: ParallelPlates-CFDOnline.pdf

I’m interested in the heat flux coefficient over the length of the bottom wall. The heat flux depends strongly on the turbulence boundary layer. For the model validation I calculated this values with a nusselt-correlation for laminar and turbulent cases. And that’s now the point where I am confused:
I calculated for this case a reynolds number of around 45’000. In a paper I found a critical Reynolds-number for parallel plates of 2’288. According to this numbers the flow should be clearly turbulent. I did the following 2 cases:

- When I calculate with 10% Inlet Turbulence and SST I get quite a good fit with my Nusselt-calculation for the turbulent case (see 2. page in the attached document). So far so good.
- But: I’m also able to just switch of the SST and change to laminar with exactly the same case setting. The convergence of the MAX U-Mom (inlet flow direction) seems to be worse than in the turbulent case (leads after some time to 10^-4) – the other values are comparable. And: The result of this laminar simulation fits nearly perfectly the Nusselt-calculation for the laminar case.
So why is the laminar solver able to calculate this flow, which is according to the reynolds number strongly turbulent? Might the value of the critical reynolds number be wrong? In the Ansys documentation it is of course described, that it will be very difficult to converge a turbulent flow with a laminar solver. Why is it in my case possible to calculate it in this way?

Hint: In the plot of the turbulent kinetic energy with the Inlet=10% intensity on the symmetry plane, it is visible, that this value is decreasing in the middle of the two plates quite fast. Only at the wall there is a turbulent boundary layer with higher values. Does this mean, that there is no turbulence except of the BL? (See last picture in the attached document)
Thanks a lot!
Dani

[ghorrocks]

Your linked file is an exe. I do not open exe files I have downloaded. But it looks like your PDF is too long to be reasonable for the forum anyway.

You seem to say that when you run it turbulent you get the expected turbulent heat transfer, but when you run it laminar you get the laminar heat transfer even though the flow is not laminar.

So your question seems to be why did the laminar case converge when the flow is turbulent? There is a hint in where you say that the laminar case converged slowly and only to 10E-4 - this suggests the solver is having a harder time converging due to numerical instability. This instability is probably due to the flow starting to go turbulent as small transient features form. If you refined the mesh or used smaller time steps these will probably get bigger and convergence even harder to get.

So the reason your laminar simulation converged was that there was enough dissipation in your mesh and solver for it to converge anyway (at least partly converge).

As for your final sentence: For low Re turbulence like this the turbulence is generated in the shear layer next to the wall. In the core of the flow there is no turbulence generation so it dissipates. This is as expected. The turbulence spreads from the boundary layer into the core of the flow.