[mneben]

Dear Foamers,

for multiphase flows with disperse particles the literature recommends the usage of a source term in the energy equation (Multiphase flows with droplets and particles, Crowe et.al., p. 445). This term shall mention the work due to body forces acting on the particle, mainly the work due to the drag force.
To get this I have to multiply the momentum source term with each cell’s number averaged particle velocity.
Now there is the problem:
How do I get an appropriate inner product of the momentum equation source term with a volVectorField?

An Example:
The momentum equation of the reactingParcelFoam-solver has on the left side a momentum source called parcels.SU(U).
Additionally one needs a volVectorField Up which corresponds to a averaged particle velocity in each cell.
Now I thought lets do something like Up&parcels.SU(U) and add this to the right side of the energy equation, but this fails because parcels.SU(U)is a Foam::tmp<Foam::fvVectorMatrix>.


Greetings

Matthias

[dkxls]

Originally Posted by mneben

Now I thought lets do something like Up&parcels.SU(U) and add this to the left side of the energy equation, but this fails because because parcels.SU(U)is a Foam::tmp<Foam::fvVectorMatrix>.

Jep you cannot take the inner product of a volVectorField and fvVectorMatrix.

But you can get UTrans, which is a DimensionedField, and use that one to compute what you want since you are not interested in the boundary field anyways.
Have a look at:

src/lagrangian/intermediate/clouds/Templates/KinematicCloud/KinematicCloudI.H

-Armin

[mneben]

Hello Armin,

thanks for your quick response. According to your recommendation I checked UTrans, especially the function calcVelocity written in KinematicParcel.C:

dUTrans += dt*(Feff.Sp()*(Ures.average() - Uc_) - Fcp.Su());

whereas Utrans will be calculated by
td.cloud().UTrans()[cellI] += np0*dUTrans

When there is no lift force Fcp.SU() should be zero and UTrans should be the drag force times the timestep.

To check this I created a test case with only one particle that is accelerated by the drag force. I added in uncoupledKinematicParcelFoam some lines


Info<<min(kinematicCloud.UTrans())<<endl;
Info<<min(kinematicCloud.UCoeff())<<endl;
Info<<max(kinematicCloud.UTrans())<<endl;
Info<<max(kinematicCloud.UCoeff())<<endl;

and switched coupled in thermoCloudProperties to true.
But I only got the following lines:


min(kinematicCloud:UTrans) [1 1 -1 0 0 0 0] (0 0 0)
min(kinematicCloud:UCoeff) [1 0 0 0 0 0 0] 0

max(kinematicCloud:UTrans) [1 1 -1 0 0 0 0] (0 0 0)
max(kinematicCloud:UCoeff) [1 0 0 0 0 0 0] 1.37330258e-09

UTrans is zero and has not the expected value, whereas UCoeff is correct (compared with the analytic solution of the drag force).

Why is Utrans zero? Where is my mistake?

Greetings
Matthias

[dkxls]

Originally Posted by mneben

Why is Utrans zero? Where is my mistake?

Sorry, I cannot help you right much further since I would have to dig through the code and see what is really done there and I don't have time for that.

A couple of hints though:

• Are you running time-accurate or steady state?
• If you use a fully explicit coupling between Lagrangian and continuous phase, the source term in the momentum equation is (see KinematicCloudI.H):

  fvm.source() = -UTrans()/(this->db().time().deltaT());

[vonboett]

Hi Matthias,

could you get new insights? I try to introduce the drag of all parcels in a cell as a bodyforce (like gravity) into interFoam, based on UncoupledKinematicParcelFoam. The one-way coupling works fine, but getting the momentum sources into the UEqn is hard.

[mneben]

Hello vonboett,

a good reference for your question is the source code of the DPMFoam solver.

Here we can find the following lines:

fvVectorMatrix cloudSU(kinematicCloud.SU(Uc));

fvVectorMatrix UcEqn
(
fvm::ddt(alphac, Uc) + fvm::div(alphaPhic, Uc)
- fvm::Sp(fvc::ddt(alphac) + fvc::div(alphaPhic), Uc)
+ continuousPhaseTurbulence->divDevRhoReff(Uc)
==
(1.0/rhoc)*cloudSU
);

Greetings
Matthias

[vonboett]

Hi Matthias,
I can really recommend
http://www.cfd-online.com/Forums/ope...tml#post547057
I managed to get the UEqn working but as soon as I add the explicit source of drag & buoyancy I'm getting unstable. Anyway, at least I can model one-way coupling including collisions