BIG PARTICLE VOID FRACTION MODEL PROBLEM (DATA EXCHANGE BETWEEN PROCESSORS IS NOT WORKING)

Submitted by ksbiefr on Mon, 05/22/2017 - 23:55

Hi Developer,

I am using latest CFDEM®coupling version 3.6.0 (released Jan 18, 2017).
I am trying to generate a particle with in a cylindrical channel. The ratio of the pipe diameter (0.04) to particle diameter (0.028 ) is 0.04/0.028 = 1.42. Therefore I have to use the big particle void fraction model.
But, I continuously encounter a problem of data exchange between processors. When one particle moves from one processor to next, it freezes either at the interface between the two processors which is nonphysical. This can be seen from the attached Video, where each processor is represented by different color. Also in a continuous generation, the next new particle will then collide with the frozen one and the whole system crashes. I’ve attached the video and the complete case in the following we-transfer link:

https://we.tl/8Ia9nDrBCW

I would greatly appreciate it your help. Thanks in advance for your time.
regards,
Khurram

Forums:

alice's picture

alice | Tue, 05/23/2017 - 09:29

Hello Khurram,
which solver have you been using? I'd recommend cfdemSolverIB, have a look at the twoSpheresGlowinski case which is provided.
Cheers,
Alice

ksbiefr | Wed, 05/24/2017 - 11:12

Hi Alice,
Thanks for the reply. While the resolved-coupling IB method is accurate but it is computationally expensive. At this moment, I am interested in an approximate method with less computational cost which would take into account the interaction of large sized particles (relative to fluid cells) with the flow.
The big particle method seems to be sufficient for our purpose. But as I mentioned earlier, the only problem which I have encountered with this method is its peculiar behavior as the particle passes processor boundary, i.e. it freezes at the interface between the processors.
With regards to your question concerning the solver, I am using the cfdemSolverPiso along with the big particle void fraction model. All the details and settings of the solver can be found in the case which I’ve attached in the link:

https://we.tl/8Ia9nDrBCW

I would greatly appreciate your help on how to resolve data exchange issue between processors for the big-particle void fraction model.

Regards,
Khurram

alice's picture

alice | Mon, 05/29/2017 - 15:48

Hello Khurram,
unresolved CFD-DEM is not suitable for calculating cases where a single particle covers entire fluid cells (particularly if there are more of them). Therefore the voidfraction algorithms are also not designed to distribute the voidfraction of a particle on more processors at a time in order to save time (no additional inter-processor required in contrast to the resolved case, where this does happen). Usually the benefit of the speed-up is higer than the issues caused by this small inaccuracy. For your case you might either have to extend the particle representation such that there is inter-processor communication during the single calculation steps, or switch to resolved CFD-DEM (which is of course more expensive). Given that unresolved CFD-DEM will cause troubles when using big particles anyways, I'd recommend the latter.
Cheers,
Alice