[Kdelavy]

Hi,

I am simulating open channel flow using interfoam with periodical boundary conditions. In the simulation, I got this weird alpha water field that it contains many holes in the water phase.

I wonder if anyone has encountered this kind of problem and what is the cause and the fix to it? Thank you.

[Alczem]

Hey,


Are you using localEuler as a ddtScheme? I noticed that sometimes, local time stepping would introduce non realistic bubbles in the liquid phase.


If that's not the case, maybe the flow actually traps some air in the water because of turbulence or an obstacle in your channel?

[Kdelavy]

Originally Posted by Alczem

Hey,


Are you using localEuler as a ddtScheme? I noticed that sometimes, local time stepping would introduce non realistic bubbles in the liquid phase.


If that's not the case, maybe the flow actually traps some air in the water because of turbulence or an obstacle in your channel?

Many thanks for your message. Yes I am using the Euler scheme with sub-cycling for my simulation. The model is the SSG Reynolds stress model. The simulation crushes in the middle of the simulation

If I switch to CrankNicolson then the simulation would crush very soon...

I am simulating a open channel flow and there is no obstacle in the channel. So in theory there should not be air trapped in the water in this case I guess.

Here are my fvschemes, is there something wrong in my choices?

ddtSchemes
{
//default Euler;
default CrankNicolson 0.9;
}

gradSchemes
{

default leastSquares;
grad(k) cellLimited leastSquares 1;
grad(epsilon) cellLimited leastSquares 1;
grad(R) cellLimited leastSquares 1;
}

divSchemes
{
div(rhoPhi,U) Gauss limitedLinearV 1;
div(phi,alpha) Gauss vanLeer;
div(phirb,alpha) Gauss interfaceCompression;
div(phi,k) Gauss upwind;
div(phi,epsilon) Gauss upwind;
div(phi,R) Gauss upwind;
div(R) Gauss linearUpwind;
div(phi,nuTilda) Gauss linearUpwind;
div(((rho*nuEff)*dev2(T(grad(U))))) Gauss linear;
div((rho*R)) Gauss linear;
div(((rho*nu)*dev2(T(grad(U))))) Gauss linear;
}

laplacianSchemes
{
default Gauss linear corrected;
}

interpolationSchemes
{
default linear;
}

snGradSchemes
{
default corrected;
}


Thanks a lot in advance.

[Alczem]

Hey!


The localEuler and Euler schemes are two different things. localEuler is for local time stepping to get a faster pseudo steady-state solution (but losing the physically accurate transient solution) and Euler is for pure transient. I would stick with Euler as long as we are not sure about the numerical settings, and only then switch to a more accurate scheme.


You can also try to switch div(rhoPhi,U) to upwind for the time being, and corrected to uncorrected. Once the problem is solved, change them back one by one.


Why are you using leastSquares rather than Gauss linear btw? I am asking out of curiosity since I do not have a lot of experience using leastSquares. Does the issue persist with Gauss linear?


Last thing I can think of right now, turn off turbulence. If it does not affect anything, try adding "density variable;" to your turbulenceProperties. And there is also a fvOptions function when turbulence is turned on for multiphase flows that might be worth a shot if your issue is turbulence related:


https://www.openfoam.com/documentati...urbulence.html


Keep us posted

[Kdelavy]

Originally Posted by Alczem

Hey!


The localEuler and Euler schemes are two different things. localEuler is for local time stepping to get a faster pseudo steady-state solution (but losing the physically accurate transient solution) and Euler is for pure transient. I would stick with Euler as long as we are not sure about the numerical settings, and only then switch to a more accurate scheme.


You can also try to switch div(rhoPhi,U) to upwind for the time being, and corrected to uncorrected. Once the problem is solved, change them back one by one.


Why are you using leastSquares rather than Gauss linear btw? I am asking out of curiosity since I do not have a lot of experience using leastSquares. Does the issue persist with Gauss linear?


Last thing I can think of right now, turn off turbulence. If it does not affect anything, try adding "density variable;" to your turbulenceProperties. And there is also a fvOptions function when turbulence is turned on for multiphase flows that might be worth a shot if your issue is turbulence related:


https://www.openfoam.com/documentati...urbulence.html


Keep us posted

Hi Alczem,

I have tried to alter the schemes but it seems that those changes would not make much difference. I guess the problem is very much related to turbulence. Actually if I use the standard k-epsilon model the computation is fine. However when I switch to the RSM, it is problematic. As the point to study the turbulence structure, I could not turn turbulence off though...

The reason to choose LSQ rather than Gauss linear is that I thought the LSQ method is more accurate, I have also seen other posts in the forum that says LSQ is preferred than Gauss linear. However in my case, the simulation performance is almost the same with these two methods.

Best regards,
Delavy

[Alczem]

Hey,


I see, sorry I am not familiar with RSM models, hopefully someone will be experienced enough to help you out!

[Kdelavy]

Originally Posted by Alczem

Hey,


I see, sorry I am not familiar with RSM models, hopefully someone will be experienced enough to help you out!

Thank you all the same!