[kasper1301]

Hi,

I have been running twoPhaseEulerFoam on a bubble column successfully with the mixtureKEpsilon turbulence models for both phases. However, if I switch to continousGasKEpsilon for the gas phase and LaheyKEpsilon for the liquid phase, the solver crashes after just a few timesteps. Does anyone know how to remedy this?

Additional info:
Initial turbulence is zero, hence set to 10^-8
Inlet turbulence is low, ~10^-3
Co < 0.25

Regards,
Kasper

[RodriguezFatz]

How exactly does the solver crash?

[kasper1301]

Floating point exception, the problem seems to be that both k and epsilon becomes unreasonably large very quickly.

[RodriguezFatz]

1) I don't know twoPhaseEulerFoam, but can you set the number of iterations during each time step, like in pimpleFoam?
2) Are your initial conditions (zero turbulence, ...) physically meaningfull or just random? I had problems when I chose bad values of k and omega as initial conditions.

[kasper1301]

1) Yes, that is possible. It just seems weird that it works without any problems with the mixtureKEpsilon model, and that it crashes with the other turbulence models.
2) The initial condition is a tank filled with water with no movement, so the turbulence should be zero (set to 10^-8 to avoid numerical issues). The inlet conditions are calculated according to [1, 2].

[1] http://aerojet.engr.ucdavis.edu/flue...ug/node217.htm
[2] http://support.esi-cfd.com/esi-users/turb_parameters/

[RodriguezFatz]

to 1) -> I tryed to get a simple piece of pipe running with different turbulence models. Just 5 diameter long and periodic boundary conditions. There are some k-epsilon models I couldn't get running with all different schemes and initial settings. So don't worry...

Did you try to just increase the inner iterations? Did you try to have some more dissipative schemes, such as some upwind (how does your fvSchemes look like)?

[kasper1301]

I just used the same schemes as in the RAS-tutorial with twoPhaseEulerFoam, haven't had time to experiment with different schemes yet.

divSchemes
{
    default                     none;

    div(phi,alpha.methane)          Gauss vanLeer;
    div(phir,alpha.methane)         Gauss vanLeer;

    div(phi.methane,kappai.methane)     Gauss vanLeer;

    "div\(alphaPhi.*,U.*\)"     Gauss limitedLinearV 1;
    "div\(phi.*,U.*\)"          Gauss limitedLinearV 1;
    "div\(alphaPhi.*,.*rho.*\)" Gauss linear;

    "div\(alphaPhi.*,(h|e).*\)" Gauss limitedLinear 1;
    "div\(alphaPhi.*,(K.*|p)\)" Gauss limitedLinear 1;

    "div\(alphaPhi.*,(k|epsilon).*\)"  Gauss limitedLinear 1;
    "div\(phim,(k|epsilon)m\)"    Gauss limitedLinear 1;
    
    "div\(\(\(alpha.*nuEff.*\)*dev2\(T\(grad\(U.*\)\)\)\)\)" Gauss linear;
}

Increasing the number of inner iterations may solve the problem, but I'm guessing the simulations will be slowed down (it currently takes ~40 hours), so I will probably just use the mixtureKEpsilon model

[farzadpolytechnic]

hi kasper
i have same problem with turbulece model. when i change my turbulence model from mixtureKEpsolon to other allowable turbulence model like LaheyKEpsilon solver crash after some time steps.
so could you find the reason? how can solve this problem?

best regards
farzad