Learning Forum

[Prashank]

i am trying to create a density separator .

the bottom inlet on left side is fluid inlet.(+y direction)

the vent in the -z direction is inlet for particles.

the whole is model is tilted 30 degrees to horizontal with fluid inlet on the lower side.

MY IDEA :-

when fluid enter from the left bottomest inlet ,a vortex is created in the cylinder. when particles are injected from the right side, the particles whose density is greater than that of fluid are forced to exit from pressure outlet in the right side[+x direction] while particles whose density is smaller than the fluid is trapped in the central vortex and leaves from the pressure outlet in the left side(+ve z direction).

MY ANSYS MODEL CONDITIONS:-

SOLVER TYPE- pressure based

STEADY STATE ANALYSIS

MODEL- k epsilion, rng, swirl dominated flow

DPM MODEL-Interaction with continuos phase,update dpm source every flow iteration, max no of steps-5000000, step length factor-5

injection type-surface discrete random walk model, random eddy lifetime-2

solution method-second order upwind

 

somehow the calculation stuck after the 9th iteration for like 1 hr.

is anything wrong in the conditions...?

how to solve this model correctly...?

(my system configurations if that help- intel i5, 8 cores).

 

[DrAmine]

Perhaps this large number of integration steps is causing the problem. I would first of all use transient solver with transient particle tracking and as it is s wril dominated flow refrain from using two equation models and either use RMS or Algebraic RSM or something much more high end.

[Prashank]

Thanks sir for your valuable suggestions....
I followed your advice and change my model to transient , reynolds stress (7 equations) linear pressure strain.
My discrete phase model :-
Unsteady particle tracking Track with fluid time step Interaction with continuos phase Update dpm source every flow iterations.


1.After all these settings the scaled residuals are fluctuating and it seems like they will never converge.
2. Particle are not interacting with fluid.( Everything is 0 in particle summary, fates are not printed on console)

[Rob]

If you've switched to transient it'll take some time steps before particles reach anywhere to have a fate reported. If you look at the DPM update how many particles are in the model?

To add, the residuals will give a saw-tooth plot as we aim to converge each time step and then move onto the next one: this is what triggers the spike in the residual plot. 

[Prashank]

Thanks for replying....
I have 1088 particles & I have done 1500 iterations still fate is not reported....

[DrAmine]

Give an adequate update numbers of source terms (dpm update interval), allows the continuous phase to converge and perhaps use smaller time step size.

7 Equation models are expensive but good for this kind of flows but there is a cheaper alternative: Enable beta feature and use EARSM instead (will be available under two k-omega category).

 

Other way is to run the continuous phase first alone, get the velocity field and then only run particles: not really adequate but quick and might be sufficient for small loading and small stokes numbers!! Have a fun!

[Prashank]

I am quite new to simulation.....
Can you elaborate...?

[DrAmine]

Which part of my answer is not clear for you?

 

 

[Prashank]

I am currently running k - omega model... It helps to converge quickly but the main problem of particle injection is not solved. I still does not got any fate of particles & every thing is 0 in the summary.


Honestly I didn't understand what you suggested in the last paragraph...

[DrAmine]

I just suggested to make a quick run: solve Flow without particles. Than freeze the flow field and solve only particles.

 

Now back to the main issue: What yo you mean with particles are not solved and no fates? Please add screenshots of DPM Injection Panel, DPM Settings etc...

[Prashank]

here are screenshots:-

[DrAmine]

What is the stop time?

[Prashank]

10 sec

[DrAmine]

Can you perhaps paste the first two time steps from the console. I would like to see what Fluent reports.

[Prashank]