[x86_64leon]

Dear OF users


Has anyone already experienced passive scalar losses when using MRF ?



Let me give more details :

1) We run a simpleFoam agitated vessel case using MRF (the case is indeed a tank with a standard Rushton turbine, well known by chemists). The MRF result is a steady state result.

2) Then we run the scalarTransportFoam solver using the MRF hydrodynamic fields from former step. The hydrodynamic field is frozen. The passive scalar T is initialized as a sphere and we let the scalar be convected and diffused over the domain. note that scalarTransportFoam is an unsteady solver. This step 2) is a classical way to obtain mixing time in a stirred tank.



Then, we compute the passive scalar volume average during the step 2) using the following function object :

scalar_volAverage
{
type volFieldValue;
libs (fieldFunctionObjects);

// What to compute ?
regionType all;
operation volAverage;
fields (T);

// Write controls
writeControl timeStep;
writeInterval 10;
writeFields false;
log false; // Should we write to stdout ?
}



The result is that the value of this volume integral diminishes by orders of magnitude ... This means that the "total mass" of passive scalar is not conserved. We have checked for passive scalar flow rate through each of the walls (the geometry is closed; there is no inlet nor outlet) and the scalar flow rates are all null ...


Any idea of what is going wrong there ?

Best

PS : We have also tried to use the pimpleFoam solver with frozenFlow=yes and functionObject scalarTransport. The result is the same : Mass loss of passive scalar !

[x86_64leon]

Update to this "strange" behaviour of non conservation of passive scalar "mass" in my computations. It is related to the fluid viscosity....
My problem arises with water (nu = 1e-6 in transportProperties) but disappears with a more viscous fluid.... (nu=2.81e-5, correspondig to glycerol).
Any idea of what's going on there ?
Could it be a machine precision issue ?

[x86_64leon]

The problem is very certainly an interaction between MRF and the frozenFlow flag ...


If I use the following settings for PIMPLE (along with the FO scalarTransport for passive scalar transport), with an MRF computation, then the scalar is NOT conserved :

PIMPLE
{
frozenFlow yes;
momentumPredictor yes;
correctPhi yes;
nOuterCorrectors 0;
nCorrectors 0;
nNonOrthogonalCorrectors 1;
pRefCell 0;
pRefValue 0;
}

On the other hand, if I no longer use a frozen flow (U and p will change at each time step), then the passive scalar is perfectly conserved :
PIMPLE
{
frozenFlow no;
momentumPredictor yes;
correctPhi yes;
nOuterCorrectors 1;
nCorrectors 2;
nNonOrthogonalCorrectors 1;
pRefCell 0;
pRefValue 0;
}


Are there issues when running with a frozen flow ?
Of course, computational time is much higher when the flow field U,p needs to be updated ... !!