SSG Reynolds Stress Turbulent Source on Boundary
I have been trying to input a turbulent boundary source into my model. Very simply I have two cubic fluid domains that share one side as an interface. Water flows in one side of one cube at 1/4m/s through the interface and out the far-side of other cube, such the the inlet and outlet are parallel. I am using the SSG Reynolds Stress Model and have to define the stress tensor coefficients and a diffusion rate.
I want my source to roughly give a 10% turbulent intensity (I). From reading up on the solver theory guide I have been using the following to calculate k, epsilon and uiuj values:
k = (3/2) (U(freestream)*I)^2 units [m^2s^-2]
uiuj = (2/3)k for i=j, 0 otherwise units [m^2s^-2]
epsilon = Rho*C*(k^2/dviscosity*viscosityRatio) units [kg m^-1 s^-3]
where C is 0.1 and viscosityRatio is 10
Then when I come to input these into CFX-Pre the units required (for a total source) are:
uiuj [kg m^2 s^-3] so I assume I have to multiply my k value by my mass flow rate to account for the missing [kg/s] right?
epsilon [kg m^2 s^-4]. Here the missing units are [m^3/s], I assume that its now volume flow rate I have to multiply my epsilon values by since the density has already been taken into account in my original equation?
When I do my calculations assuming:
total mass flow rate =5.5e4 [kg/s],
volume flow rate = 53.76 [m^3/s]
I end up with the following (in the form required by CFX-Pre):
uiui = 1078 [kg m^2 s^-3]
epsilon = 409077 [kg m^2 s^-4]
This just seems wrong to me, is that dissipation rate not way too large?
When I run the simulation, either my solution diverges dramatically and the solver crashes or I see no change in my velocity flow field going through the interface.
Can anyone shed some light on the reason for this? Is it the way I have calculated my values above, or is it the turbulent domain settings for my two cubes? I have left the domain settings to default, should I be changing the epsilon coefficients for the domains too? My geometry is so simple (free slip walls all round, nothing obstructing flow) that any turbulence is very quickly dissipated, when I set the inlet to high turbulence (I=10%, dvRatio=100) then I still see the turbulence for a short distance from the inlet, I see nothing at my turbulent interface.
Any help would be greatly appreciated!
I do not have time to check your maths so I will leave that bit up to you.
But for the convergence difficulties, I recommend you start with a simpler model, maybe a 2-eqn turbulence model and see if that converges. If that is OK then go to RSM with simple default boundary conditions. This way you can tell whether the simulation itself has a problem, it is RSM or is your boundary conditions.
Be aware that RSM models are much harder to converge than 2-eqn turbulence models and are very sensitive to mesh quality.
I have already run through everything exactly as you said: using k-epsilon and it converges easily, witched to SsG with no turbulent interface and got a converged solution albeit after more iterations (as expected). I am looking at spatial/temporal pressure on a disk in turbulent flow so I need the SSG Reynolds Stress model to see the turbulence more clearly and am using a transient solution.
My question isnt about my arithmetic it is about the equations I have used and assumptions I have made about units. Even increasing my viscosity ratio to 100 I still get an eddy dissipation rate which is 10 times larger than my k value. I am asking if that seems physically correct as I see very little turbulence at my interface.
The dissipation is frequently far higher than production - it just means the turbulence is being heavily damped in that area. This is a real effect.
Have a look at the production versus dissipation for the k-e and SSG runs in that area.
|All times are GMT -4. The time now is 00:06.|