[shock77]

Hi,


I am running a simulation of a highly underexpanded jet. The simulation runs stable when laminar. If I try to use a turbulence model, the simulation crashes. I have tracked the values for k, omega and nut and I think the simulation crashes because I get very high turbulent viscosities ( up to 130). The strange thing is, that the high viscosities are in areas far away from the jet and near the walls. Thats why I think wrong boundary conditions lead to those artefacts.


My walls are slip, because I am not interested in the boundary layer, which is very small and would need much effort to resolve. I am interested in the shear layer of the underexpanded jet.


My boundary conditions for k, omega, nut and alphat are:


k

• Inlet: turbulentIntensityKineticEnergyInlet
• Outlet: zeroGradient
• Walls: slip

omega

• Inlet: turbulentMixingLengthFrequencyInlet
• Outlet: zeroGradient
• Walls: slip

nut

• Inlet: calculated
• Outlet: calculated
• Walls: slip

alphat

• Inlet: calculated
• Outlet: calculated
• Walls: slip

I think slip is not the right boundary condition. Unfortunately I dont know which works best with slip walls, since I usually have noSlip at the wall and use wallfunctions with it.

[Santiago]

Originally Posted by shock77

Hi,


I am running a simulation of a highly underexpanded jet. The simulation runs stable when laminar. If I try to use a turbulence model, the simulation crashes. I have tracked the values for k, omega and nut and I think the simulation crashes because I get very high turbulent viscosities ( up to 130). The strange thing is, that the high viscosities are in areas far away from the jet and near the walls. Thats why I think wrong boundary conditions lead to those artefacts.


My walls are slip, because I am not interested in the boundary layer, which is very small and would need much effort to resolve. I am interested in the shear layer of the underexpanded jet.


My boundary conditions for k, omega, nut and alphat are:


k

• Inlet: turbulentIntensityKineticEnergyInlet
• Outlet: zeroGradient
• Walls: slip

omega

• Inlet: turbulentMixingLengthFrequencyInlet
• Outlet: zeroGradient
• Walls: slip

nut

• Inlet: calculated
• Outlet: calculated
• Walls: slip

alphat

• Inlet: calculated
• Outlet: calculated
• Walls: slip

I think slip is not the right boundary condition. Unfortunately I dont know which works best with slip walls, since I usually have noSlip at the wall and use wallfunctions with it.

Well, if you have a slip wall, then you don't have momentum gradients of any kind there, hence, turbulence can only be transported but not produced. Furthermore, the no-penetration condition doesn't allow for k/omega to be convected out of the domain. This only leaves diffusive transport which, by setting free-Slip, you're also killing off... You have a local variation of k/omega not being balanced locally by any other operator...

[shock77]

Hi Santiago,


so basically what I am doing is to accumulate my turbulent quantities at the boundaries, right? So in order to stabilize my simulation I should zeroGradient at the walls for k, omega, nut and alphat?

[Santiago]

Originally Posted by shock77

Hi Santiago,


so basically what I am doing is to accumulate my turbulent quantities at the boundaries, right? So in order to stabilize my simulation I should zeroGradient at the walls for k, omega, nut and alphat?

I'd say that in a theoretical free-stream, k = 0. Omega on the other hand is not zero. A value can be estimated, from the lenght scales and viscosity of the fluid's flow.

Depending on the quality of your grid, you might need to clip the viscosity as well, to, say, 5000 times the molecular viscosity...

[shock77]

So you suggest a fixedValue = 0 for k and a fixedValue = x for omega. I have read this in many cases, that this has been done.


But what do you mean with clipping the viscosity? What boundary condition would you suggest for nut?

[Santiago]

Originally Posted by shock77

So you suggest a fixedValue = 0 for k and a fixedValue = x for omega. I have read this in many cases, that this has been done.


But what do you mean with clipping the viscosity? What boundary condition would you suggest for nut?

Clipping, as in bounding the field to a required range.

nut should't have boundary conditions, as you are not solving a PDE for it in this case. Perhaps you need to impose wall functions on the walls.

[shock77]

I am applying Boundary Conditions on nut to get the field for postProcessing.
I have tried using wallfunctions aswell, but nut still rises above 130.


I have tried wallfunctions + slip and wallfunctions + noslip. Both with the same result.