[mcampbell]

All,

I have been modeling a hydrocyclone using particle tracking in CFX. I have had very good success solving the Eulerian phases (water & air) and have good correlation to the well-known Hsieh hydrocyclone that has been heavily researched and documented including LDV experimental results. FWIW I am utilizing the SST turbulence model with Curvature Correction activated. This limits the computational cost while maintaining adequate accuracy. (Even with this relatively simple model it was no small task to achieve a fully developed hydrocyclone with an air core and reach convergence). My streamlines and velocities are right where I expect them.

Now, I am having difficulty getting relevant particle tracking results. I have worked on this extensively and get results that are directionally correct but I lack accuracy. I am very familiar with the ins and outs of particle tracking within CFX - I am using one way coupling since my volume fraction is relatively low <5% (may be stretching it) with the Schiller Naumann drag model. I have run both transient and steady state solution with similar results. My issue is the particles are not ending up where I expect them to go - too many small particles exit through the underflow when they should be traveling with the bulk of the flow to the overflow (95% of the influent fluid exits through the overflow). I have very good empirical data to compare to. I believe the issue is rooted in the details of the particle tracking, namely the particles are not reversing direction when the streamlines do. I am aware that the particles should not follow the streamline exactly BUT 1-5 micron particles in water should be fairly fixed to the fluid (very high drag) at the acceleration in this situation.

Has anyone figured out how to deal with particle tracking with high curvature? Could this be a mesh issue? An integration issue? Other?

Thanks,

Matt Campbell

[ghorrocks]

Have you done a sensitivity analysis on the key parameters of the particle tracking model - such as number of integration steps per element, maximum particle integration time step?

Also I suspect your particles may need a finer mesh and/or time step than the eularian model to capture the curvature. You might need to repeat the mesh and time step sensitivity analysis with the particle tracking model.

[mcampbell]

ghorrocks,

Thank you very much for the reply! I only now noticed the reply notification in my Junk folder! Anyways, very good suggestions. I was suspicious of the mesh not being able to handle the full behavior of the particle tracking. In the process of investigating this I found the expert parameter 'pt extrap vel for integ' which according to CFX documentation: Uses extrapolation of the particle velocity during particle integration. This may be helpful for cases with streamline curvature.. Activation of this parameter made significant improvements to the results. The way I understand it the particles where being mathematically marched outward due to interpolation and therefore sending them inadvertently to the underflow.

I might achieve similar or even better results using a finer mesh. But, since I work in industry we had to move forward with the best possible information at the time (we had good enough results) - I can't spend 2-5 years like a grad student can...though I may revisit to investigate further.

There was still difficulty with the smallest particles, but for our purposes this was not important - we wanted to know the fate of the larger particles to estimate the total mass retained (tiny particles were irrelevant). In the end we compared to a situation with new geometry and a much higher viscosity carrier fluid. For reference - since you rarely see any closure on these issues- I attached snapshot of a relevant page in the report.

p.s. ghorrocks - I read alot on this forum. Thanks for all your effort and sharing you knowledge!

[ghorrocks]

That looks like pretty reasonable results, so well done. It should be good enough to use as a design tool which sounds like that was the purpose.

Interesting about the particle tracking expert parameter. I have not used that one before but I have had problems with particles not following the flow properly. It sounds useful in these cases.

Glad to hear I have helped!