bounding epsilon blow up
 

It is a case using simpleFoam solver with K-epsilon Turbulence model.

the fellowing result will blow up. Is there anyone can help me to find out where the problem comes from? BC setting('epsilo' attached)?

Time = 0.001
DILUPBiCG: Solving for Ux, Initial residual = 0.64936, Final residual = 0.0285001, No Iterations 1
DILUPBiCG: Solving for Uy, Initial residual = 0.527993, Final residual = 0.0462129, No Iterations 1
DILUPBiCG: Solving for Uz, Initial residual = 0.815427, Final residual = 0.0168725, No Iterations 2
DICPCG: Solving for p, Initial residual = 1.42064e-15, Final residual = 1.42064e-15, No Iterations 0
time step continuity errors : sum local = 6.86127e+27, global = 7.64441e+11, cumulative = -2.2062e+23
DILUPBiCG: Solving for epsilon, Initial residual = 4.35678e-07, Final residual = 4.35678e-07, No Iterations 0
bounding epsilon, min: 3.3635e-22 max: 1.18516e+87 average: 7.77262e+80
DILUPBiCG: Solving for k, Initial residual = 5.98618e-09, Final residual = 5.98618e-09, No Iterations 0
ExecutionTime = 193.75 s ClockTime = 195 s



BC setting:

epsilo:

wall
{
type epsilonWallFunction;
Cmu 0.09;
kappa 0.41;
E 9.8;
value uniform 0.0012;
}

inlet
{
type fixedValue;
value uniform 0.0012;
}

outlet
{
type zeroGradient;
}

I'm new to FOAM.
I will be very appreciated for your help!

Maybe a draft of the geometry, details of problem and

system/controlDict
system/fvSchemes
system/dvSolution
0/*

dictionaries can give us a better idea how to solve the blowing up.

Regards.

Try setting the k and e schemes to upwind rather than linear in the fvSchemes directory. I've had problems achieving convergence using second order schemes for the turbulence models.

In my case it was a great help to reduce the relaxation factors. I had the same problem. The relaxation factors are responsible for the search of a the start value in the calculation of the parameters depending of the result in the last time step. If you reduce them, the start value is not so far away to the last result but you´ll need more time for the calculation.

Try it. You find the relaxationFactors in fvSolution and then reduce all factors 0.1 down. If it doesn´t help, try 0.2.

It could be the geometry, too. But try it.
Hope, it helps.
greetz phil

Anyone has a solution to that problem?
 

Hi,

I have similar problems as described in the original post. Both the bounding epsilon and bounding k blow up some time after starting a PIMPLE RNGkEpsilon. Tried changing the discretization schemes to upwind, experimenting with under relaxation factors, time-step etc.

Initially the simulation blew up straight after the beginning (i.e. first ~5 time steps), with initial k-epsilon values defined as very small. I calculated the initial conditions using formulas in OpenFoam user guide (https://www.openfoam.com/documentati...k-epsilon.html). I'm trying to simulate an external flow on an ocean seabed geometry (in .vtk format). k=0.0096, epsilon= 2.81e-7

Then I changed these initial values to some, bigger fixed values, corresponding to much lower Re and not really making any sense (just an experimental guess) i.e. k and epsilon = 0.00325 at inlets and internal field and 0.325 at the geometry walls. Simulation blew up much later in sim time (30 s = many time steps) in comparison to educated guesses about initial conditions I made above.

I use adjustableRunTime to control Co number. Also, the simulation worked with much lower refinement, lower-resolution mesh before. Is it a mesh problem?

Thanks for any help.