Learning Forum

[TONY ARHUR]

DPM injection.

I am run a CFD-DPM simulation in fluent, I am encountering a situation where by after certain time-step the “Advancing DPM injection” does not progress forwards and will be stuck at that time step for days although the solution has converged.

Please is there anyway I can overcome this as this happens just after less than 10% of simulation.

[Prashanth]

Hello,

How many particles is in your model before it gets stuck? Does your model have recirculation zones? 

[TONY ARHUR]

The model has a rotating zones.
There were 26178 particles before it got stuck

[Prashanth]

Okay, does these many particles congregate at a location due to these rotating zones? If so, you need to do some UDF work and take care of the very high particle count at localized zones by getting them trapped or removing them etc.

[TONY ARHUR]

Looking at the track animation I have, the particles have not yet entered the rotation zones, they are still at the inlet zone. Also how do I reduce the number of particles. It seems to me no matter how low I reduce the mass flow rate of the DPM injection I get the same number of particles being produced.

[Prashanth]

You can use single or group injections to customize the injection locations better.

[TONY ARHUR]

i have develop this code to  be compiled in fluent. i get this error message (

C sources: ['PARTICLES.c']

lld-link: error: undefined symbol: begin_particle_loop

>>> referenced by PARTICLES.obj:(particle_limit_source)

 

lld-link: error: undefined symbol: end_particle_loop

>>> referenced by PARTICLES.obj:(particle_limit_source)

scons: *** [libudf.dll] Error 1).

find attached if you could please help me with error .

 

[Rob]

The parcel count is linked to the injection, particle timestep and stochastic tries (which shouldn't be active in transient mode). It has nothing to do with the injection mass: that alters the parcel weight (mass) but not the number (read up on parcel theory). The solver has always mixed particle and parcel terms, and this then confuses anyone not aware of this. 

26k is not an excessive number, so shouldn't cause problems. 

What are you actually trying to model as that'll help figure out how to solve it. Messing with particle count via a UDF is probably not going to end well. 

[TONY ARHUR]

I am modelling 3-way particle-particle, particle-fluid and particle-geometry collision in twin-screw granulation 

[Rob]

As in screw conveyor? Does the fluid do that much to the particle flow? 

[TONY ARHUR]

Yes please, screw conveying.

please I don't get what you mean by "does the fluid do that much to the particle flow" but what I meant by the particle-fluid is to say the injection is in continuous flow with the fluid flow

[Rob]

In a granular screw conveyor the particles are typically flowing from a hopper or similar. The fluid (ie space between the particles) typically doesn't do much unless it's a very viscous paste.  If you can ignore the fluid part then Rocky may be a better option, or full Eulerian in Fluent. I'd not use DDPM here as you're unlikely to use the DPM part anywhere in the domain: the volume fraction is high enough to switch to the Eulerian part pretty much everywhere. 

[TONY ARHUR]

Thank you very much.

but what I intended doing was to have viscous fluid interacting with the particles. Believe you saying that is not possible in fluent?

[Rob]

OK, so Eulerian-Granular. DDPM is generally for tracking from dispersed to a bit less dispersed. Eulerian covers most problems but you need to understand how it works. Rocky will do most granular related applications; if the particles are small it will probably be OK here. 

[TONY ARHUR]

thanks.

currently I am working 0.2mm of particle size. I think it should be ok and fluent.

[Rob]

Size isn't as important as size relative to the cells. How big are the cells in the screw?