Learning Forum

[Adrian Sieradzki]

Has anyone else encountered problem that causes DPM to crash fluent when it has to do too many iterations? Seems like problem with rendering particles, because after tracking finishes nothing appears in the viewport and fluent becomes stuck?

Looks like the issue is related to low flow rates where particle positions are very close to each other between time-steps. Unfortunately I am observing that changing "Step Length Factor" in DPM doesn't seem to do anything?

[Rob]

How may parcels (particles) are in the system, and how many steps did you set in the DPM panel? 

[Adrian Sieradzki]

it's steady time, ~100 particles max steps allowed 50 000 

[Rob]

Drop the max steps to about 5000 and see how it behaves. Best guess is you have a few particles that aren't moving but are trying to use all 50k integration steps and that gives the appearance of a crash. 

[Adrian Sieradzki]

If I drop the number of steps then my partiles don't reach the outlet. Could it be particles are "not moving" since I see the message that DPM iteration was complete and all particles escaped but nothing shows in the viewport and ansys becomes unresponsie. 

[Rob]

Could also be the graphics catching up. How does it behave if you Track rather than Display? If that works Display but skip most of the tracks. A picture may help to figure out what's going on. 

[Adrian Sieradzki]

 

Here are my particle tracks near outlet

 

Decreasing flow rate/increasing particle size leads to the unresponsive behaviour, as can be seen the particle tracks are so close together they are overlapping. I am running this on a PC with RTX4090 so not sure how it could be not strong enough to render out all particles 

Using "track" i.e. only running the DPM does not result in unresponsive behaviour.

[Rob]

That looks like a load of parcels following the same track. It may be a result of the angle/slice but the mesh doesn't look overly well resolved either. 

[Adrian Sieradzki]

No, I am simulating an effect where particles get foused into narrow streamline based on diameter. That's the point of my experiment. Why would you say mesh doesn't look well resolved? My cells are too large? 

[Rob]

Cells are large relative to the channel. You need good near wall resolution for the y+ checks, but what is never made clear in academic courses is the need to resolve the rest of the flow too!  I'd prefer a much more gradual jump in cell size from the inflation. Plot a contour of velocity on the centre plane with node values off - how does it compare to node values on? 

DPM particles can only see the wall (for collisions etc) when in the near wall cell so you may be underestimating wall contact. 

[Adrian Sieradzki]

 

I think it must be the angle, I have ~20 cells across channel, 4x along height 5x along width which is not horrible. I tried 10 times the resolution but this makes the problem of viewport crashing/becoming unresponsive even worse - since I have even more particles now. Seems something horrible is up with graphics/renderer. 

 

[Rob]

That's a possibility, is the model transient, or just the particles? Note, 4 by 5 on the surface isn't well resolved! 

[Adrian Sieradzki]

Just the particles. 
Is there a way to avoid every n-th particle in a trace? 

[Rob]

Have a look at Skip option, mid-right on the Particle Tracks panel.