[Eman.]

Hi,

I have a problem with my LES simulation. I have backflow at the outlet of my domain which results in instability and as a result Courant number increases very fast and the code blows up eventually. What is the best outlet b.c for velocity and pressure that are not reflective and allow the eddies to leave the domain without influencing way upstream? For the velocity, I thought of inletOutlet and advective b.c and for the pressure zeroGradient and waveTransmisive. It'd be appreciated if you could share your experience with me.

Thanks

[C-L]

I haven't used it myself but I think 'fixedmeanvalue' might be appropriate for the pressure boundary condition. Have a look through the below thread and see if you can find anyone with similar problems.

fixedMeanValue compilation in OpenFOAM 4.0

[Tobi]

As the first guess I would fix the inflow (based on the fluxes). For that purpose you can use inletOutlet with an inletValue zero. The pressure could be fixed. However, I do not know your simulation settings and which kind of LES you are performing (incompressible / compressible / transonic / subsonic / hyper ...). So you have to decide which BC you want to use. I guess at the moment you have zeroGradient for the velocity. However, this is a common problem (reverse flow) and there are several possibilities.

• BC
• Geometry extension
• etc.

Cheers...

[Eman.]

Originally Posted by C-L

I haven't used it myself but I think 'fixedmeanvalue' might be appropriate for the pressure boundary condition. Have a look through the below thread and see if you can find anyone with similar problems.

Thanks. It sounds like a good idea. I will try that. By the way, do you know if it is the same as fixedMean b.c in OpenFoam 3?

Originally Posted by Tobi

As the first guess I would fix the inflow (based on the fluxes). For that purpose you can use inletOutlet with an inletValue zero. The pressure could be fixed. However, I do not know your simulation settings and which kind of LES you are performing (incompressible / compressible / transonic / subsonic / hyper ...). So you have to decide which BC you want to use. I guess at the moment you have zeroGradient for the velocity. However, this is a common problem (reverse flow) and there are several possibilities.

• BC
• Geometry extension
• etc.

Cheers...

Thanks Tobias. I am doing an incompressible LES with Smagorinsky model and my inflow data is taken from a precursor simulation. You are right, I currently have zeroGradient for both velocity and pressure and it blows up very fast. I will try to use InletOutet for the velocity outlet patch.

[Tobi]

Okay, but keep in mind: Setting p and U at one patch for incompressible cases is not a good thing in the numerical point of view. Good luck.

PS: If you are searching for explanations of the boundary conditions, look into the header files. There is a description

[peob]

Eman,

You might take a look at the following thread...

mixed inflow/outflow downstream boundary condition question

I had stability issue with mixed inflow/outflow at an outflow boundary, and going to a totalPressure boundary has essentially rectified the problem.

[Tobi]

I just want to mention something here. The totalPressureboundary condition should be used with a pressure based velocity boundary condition. However, the main problem in reverse flow is the following. If we have an inlet and an outlet and based on the flow structure, we have a back flow at the outlet, the main problem is that we do not know how the flow should flow inside the domain again. Of course based on the pressure field we can estimate something and hope that this is true but in fact we do not have any information. Simple example:

• Assuming a particle loaded flow. At the outlet we have a reverse flow. However, the particle leaves the domain but in reality it can happen that the particle will re-enter the outlet again based on the reverse flow. But here we are limited because we cannot track the particle after it left the domain and therefore, we cannot say where it would/might re-enter in the outlet. I hope that it was clear.

However, I agree that a totalPressure boundary condition provides a better numerical system because it is a mixed boundary which is zeroGradient for each face where the flux is negative and is adjusted based on the flux if we have inflow in order to provide the correct pressure on the faces (totalPressure - kinetmaticPressure). This is adjusting the inflow in a way that it makes more sense than zeroGradient at all

[huangfei]

Hi Eman,
Could you tell me how did you solve this backflow problem? I encounter the same issue. BC at outlet I set as below:
U:zeroGradient,
p:fixedValue;
k:inletOutlet;
nusgs:zeroGradient;