Bubble Column Simulation: Different Turbulence Models different results

zobekenobe · July 7, 2012, 06:00

I'm simulating a Bubble Column; 0.15m by 0.9m, Single Point Sparger, hole diameter 1.96e-3m. I tried the simulation with three different turbulence models; K-eps, RSM and LES.

Superficial gas velocity 20mmps.

Turbulence Nature of the Results/Error
Model

K-eps = The bubble shoots right to the top and it appears as if jet has been sent through.

RSM = Fatal error in linear solver, the Water.turbulence eddy viscosity frequency goes out of bounds.

LES = The simulation runs well and the bubble rises as expected.

Can anyone please help me understand the probable errors in the simulation and the reason for their occurrence.

Thanks

ghorrocks · July 8, 2012, 08:38

* There are lots of multiphase parameters you should consider, like dispersion forces and things like that. That might get the simple turbulence model runs working.
* Try the SST turbulence model. It is the best all-round turbulence model these days, it is better than k-e.
* RSM models are always very hard to get to converge. Make sure your mesh is really good quality, your time step small enough and your initial condition suitable. But even then RSM is hard to get working.
* If LES is working well then why not just stick with that one?

zobekenobe · January 2, 2013, 06:51

Thanks ghorrocks....I know its a late reply...but I finally got the simulations to run without crashing. RSM was tough to run but reducing the time step did help. The RANS model hasnt given good profiles compared with the LES model.

zobekenobe · January 2, 2013, 07:07

These are the profiles I have obtained for Multipoint sparger incorporating the K eps and RANS models for simulation. Bubble Size 5mm, column dimensions 0.15m by 1m.

Any ..absolutely any comments would dearly be appreciated

Thanks

ghorrocks · January 3, 2013, 19:25

This looks typical for RSM - RSM predicts additional vorticies and lots of little features, and k-e predicts a single big vortex with something approaching solid body rotation.

oj.bulmer · January 28, 2013, 10:02

What was the rationale for the simulation, if I may ask? If it is just about the traces of bubble, then did you use Langrangian particle tracking, or it was for bubble size distribution etc using population balance model (MUSIG etc)? Which forces did you enable on bubbles? From the picture it looks like Drag and Lift (since the diameter is small) are a must. You may consider dispersion force as well, if you are lucky enough to have economical convergence...

EDITED>>
Adaptive timestepping really helps, since typically it is difficult to estimate the right timescales... I am quite surprised at your selection of RSM/LES though, considering the quality of mesh these models demand. Throw in the complexity of multiphase physics and its a nightmare in terms of cost and possibility of convergence..

July 7, 2012, 06:00	Bubble Column Simulation: Different Turbulence Models different results	#1
zobekenobe Member zobekenobe Join Date: Mar 2009 Location: Dublin, Ireland Posts: 72 Rep Power: 17	I'm simulating a Bubble Column; 0.15m by 0.9m, Single Point Sparger, hole diameter 1.96e-3m. I tried the simulation with three different turbulence models; K-eps, RSM and LES. Superficial gas velocity 20mmps. Turbulence Nature of the Results/Error Model K-eps = The bubble shoots right to the top and it appears as if jet has been sent through. RSM = Fatal error in linear solver, the Water.turbulence eddy viscosity frequency goes out of bounds. LES = The simulation runs well and the bubble rises as expected. Can anyone please help me understand the probable errors in the simulation and the reason for their occurrence. Thanks

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Turbulence models	Reza Imani	Main CFD Forum	7	April 8, 2016 10:15
Water subcooled boiling	Attesz	CFX	7	January 5, 2013 04:32
bubble collapse simulation	ndabir	FLUENT	2	July 31, 2012 13:59
CFX bubble simulation with free surface model	adma	CFX	6	February 3, 2006 12:17
Turbulence Intensity Vs CFD Simulation	Apple L S Chan	Main CFD Forum	3	December 15, 1998 18:28

July 8, 2012, 08:38		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	* There are lots of multiphase parameters you should consider, like dispersion forces and things like that. That might get the simple turbulence model runs working. * Try the SST turbulence model. It is the best all-round turbulence model these days, it is better than k-e. * RSM models are always very hard to get to converge. Make sure your mesh is really good quality, your time step small enough and your initial condition suitable. But even then RSM is hard to get working. * If LES is working well then why not just stick with that one?

January 2, 2013, 06:51		#3
zobekenobe Member zobekenobe Join Date: Mar 2009 Location: Dublin, Ireland Posts: 72 Rep Power: 17	Thanks ghorrocks....I know its a late reply...but I finally got the simulations to run without crashing. RSM was tough to run but reducing the time step did help. The RANS model hasnt given good profiles compared with the LES model.

January 3, 2013, 19:25		#5
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	This looks typical for RSM - RSM predicts additional vorticies and lots of little features, and k-e predicts a single big vortex with something approaching solid body rotation.

January 28, 2013, 10:02		#6
oj.bulmer Senior Member OJ Join Date: Apr 2012 Location: United Kindom Posts: 473 Rep Power: 20	What was the rationale for the simulation, if I may ask? If it is just about the traces of bubble, then did you use Langrangian particle tracking, or it was for bubble size distribution etc using population balance model (MUSIG etc)? Which forces did you enable on bubbles? From the picture it looks like Drag and Lift (since the diameter is small) are a must. You may consider dispersion force as well, if you are lucky enough to have economical convergence... EDITED>> Adaptive timestepping really helps, since typically it is difficult to estimate the right timescales... I am quite surprised at your selection of RSM/LES though, considering the quality of mesh these models demand. Throw in the complexity of multiphase physics and its a nightmare in terms of cost and possibility of convergence..