[jbeske]

Hello dear fellow Foamers,

I am at a loss.

I am working on a centrifugal pump case and due to very low volume flow values I have very small dimensions for my pump.

It is a closed channel pump, rotating at 6000rpm with an inlet of 4mm in height (outside diameter - inside diameter) and at the outer radius it is only 1.5mm high.

This leads to very low Reynolds Numbers across the channel, between 1e5 and 1.5e5 .

If I try to get my y+ over 30 I get very little cells across my outlet profile. With y+=100 there would be only 3 cells, which is way to little to use a regular k-epsilon-model, am I right?

Now Openfoam has several turbulence models available, such as the k-omega, k-omegaSST and whatnot. I have no trouble getting my y+ below 1, but I am confused about the different models.

Does anybody have experience with low Reynolds-Numbers?
What would be an appropriate turbulence model to use?

I'd be very grateful for any tips or help

[piu58]

If you use Large Eddy Simulation you make a mesh as fine as you want. The finer the mesh the smaller the eddies resolved by the simulation. If the get very fine you come in the near of DNS (which might be possible in you small geometry).

[jbeske]

Thanks for your fast response!

According to the script of my numerical simulations course, A Reynolds number of around 12 000 still needs about 2000 grid cells across my canal profile.
It also needs a siginificant amount of timesteps to calculate, and I don't really have many resources at my hand, computer-wise.

So I doubt, even if it works with DNS, that I could calculate the simulation in a reasonable fashion :/

Do you maybe know of any turbulence models that make use of a better wall function or no wall function at all?

Because of the small dimensions I have in my model, I think it consists mainly of boundary layer region :/

[piu58]

LES and RANS work different concerning wall functions.

In RANS there is no way to cover turbulence in a direct way. All ha to be modeled. Therefore you need your wall distance, because the models require it and work in a wrong way with larger or smaller yPlus.

LES simulates the turbulence in large eddies in a direct way. It is allowed here to use small element sizes and a small yPlus. The amount of turbulence which has to be modeled with wall functions gets smaller than. All in all the turbulence is covered. Part of it by wall functions and part of it by the direct simulation.

~

You may try LES/dynamicEqn. It uses wall functions of course, but you don't have to set any coefficients.

[jbeske]

Thanks I will try that.

That only leaves the problem of validation.

Do you know where I could ask for or find comparable cases, to validate my results?

Of course, my case is rather special, so there probably won't be something equal, but I'd take anything that goes in the right direction at the moment.

[piu58]

If you don't have any experience with simulations I recommend well studied geometries like channel flow or flow past a sphere. You may choose a very small geometry for that to study the wall functions.

[jbeske]

Okay I will do that.

Thanks again for taking the time to help me