Free surface flow in counter current limitation

j.zamorano · April 17, 2013, 10:57

Hello, I am trying to do a simulation of 2 flows (water and air) flowing counter currently in a horizontal hot leg. I have problems of convergence with steady state, so now I am trying a transient one. I have defined constant velocity inlets for both water and air, an outlet for the water with pressure condition, and an opening for the air. Even with the transient simulation, it is becoming really hard to get convergence. Can you give me any advice? I have selected a 1ms timestep. Thanks.

ghorrocks · April 17, 2013, 19:48

Please post an image of what you are modelling and the CCL or output file.

j.zamorano · April 22, 2013, 04:52

http://www.subirimagenes.com/otros-f...-8398648.html]

[/URL]

Left: air constant velocity inlet
Bottom: water constant velocity inlet
Bottom: water outlet (pressure difference 0 Pa)
Up: air opening (dinamic pressure 0 Pa)

Solver control:
Transient, 1 sec total time, 1 ms timestep, Upwind, second order backward euler, 1E-04 residual target.

Free surface model: air and water continuous fluids, 2F model for drag coefficient, homogeneous turbulence, k-epsilon model

j.zamorano · April 22, 2013, 04:57

Is this enough information? If you need the output file or the CCL please give me your e-mail, and I will send it to you.

Thanks

ghorrocks · April 22, 2013, 10:00

Why have you chose a 1ms time step? Unless you have actually done a sensitivity analysis on that time step size and determined it is what you need then it is bound to be wrong.

So let the solver work the time step out for you. Use adaptive timestepping, homing in on 3-5 coeff loops per iteration.

Are you using surface tension?

Why are you using a highly inaccurate spatial discretisation scheme (upwind) but a very accurate temporal scheme (2nd order back eular)? If you just want quick and dirty results then run 1st order in time as well. If you want an accurate answer you will need to run a better spatial discretisation than that.

JuPa · April 22, 2013, 10:28

If you let the solver work out the time step via adaptive time stepping I've found the time step it selects is usually prohibitively small, such as 1e-8 s (that is if you converge to 3-5 loops).

ghorrocks · April 22, 2013, 19:25

That is because it needs small time steps! There are very few simulations where this does not find the correct time step.

The only exception to this is when you are happy to sacrifice accuracy to just "muddle" through a tricky bit of the simulation and don't care about accuracy. An example could be when you first start the simulation up, or when an aggressive action occurs which you do not care about the details (eg a fast combustion event).

If this slows the simulation too much then
* Check that your mesh is not too fine
* Check your convergence is not too tight
* Get a faster computer
* Get more computers and run parallel.

j.zamorano · April 24, 2013, 10:34

I am new in CFX, What parameters should I take to make adaptive timestepping? Max Courant number?

ghorrocks · April 25, 2013, 08:04

No. Courant number only has a tenuous link to convergence in CFX as it is an implicit solver.

The recommended approach for adaptive time stepping is homing in on 3-5 coeff loops per iteration.

j.zamorano · April 25, 2013, 11:31

I am not using surface tension forces, should I? I think the diameter is too big and this forces are too small (0.05 m). I am using free surface model, should I change it to mixture model?

Thank you so much

ghorrocks · April 25, 2013, 19:49

Only use surface tension if it is important for the flow you are modelling. It is an expensive model and will slow you down a lot, so only use it if you have to.

j.zamorano · April 26, 2013, 07:33

http://www.subirimagenes.com/imagen-...-8405900.html]

[/URL]

This is happening to me when I try to do a transient simulation. Does anyone know what happen? The program says something about overflow

ghorrocks · April 26, 2013, 08:23

This is a floating point error (which then caused lots of other errors including crashing the parallel run) - this is an FAQ: http://www.cfd-online.com/Wiki/Ansys...do_about_it.3F

j.zamorano · April 30, 2013, 07:44

I am having quite good results with steady state simulations. Now, I am doing transient ones. I want to reach the second state of the picture by improving the gas velocity. But, although the function for the gas velocity inlet is well defined, when I plot the results it shows everytime the first state (left). Can be this because problems of convergence? What should I do?

http://www.subirimagenes.com/otros-t...-8411961.html]

[/URL]

ghorrocks · April 30, 2013, 20:32

I do not understand your problem. What are you seeing and what do you expect to happen?

April 17, 2013, 10:57	Free surface flow in counter current limitation	#1
j.zamorano New Member Javier Join Date: Apr 2013 Posts: 7 Rep Power: 13	Hello, I am trying to do a simulation of 2 flows (water and air) flowing counter currently in a horizontal hot leg. I have problems of convergence with steady state, so now I am trying a transient one. I have defined constant velocity inlets for both water and air, an outlet for the water with pressure condition, and an opening for the air. Even with the transient simulation, it is becoming really hard to get convergence. Can you give me any advice? I have selected a 1ms timestep. Thanks.

April 22, 2013, 04:52		#3
j.zamorano New Member Javier Join Date: Apr 2013 Posts: 7 Rep Power: 13	http://www.subirimagenes.com/otros-f...-8398648.html][/URL] Left: air constant velocity inlet Bottom: water constant velocity inlet Bottom: water outlet (pressure difference 0 Pa) Up: air opening (dinamic pressure 0 Pa) Solver control: Transient, 1 sec total time, 1 ms timestep, Upwind, second order backward euler, 1E-04 residual target. Free surface model: air and water continuous fluids, 2F model for drag coefficient, homogeneous turbulence, k-epsilon model Last edited by j.zamorano; April 22, 2013 at 04:54. Reason: Can´t see the picture

April 26, 2013, 07:33		#12
j.zamorano New Member Javier Join Date: Apr 2013 Posts: 7 Rep Power: 13	http://www.subirimagenes.com/imagen-...-8405900.html][/URL] This is happening to me when I try to do a transient simulation. Does anyone know what happen? The program says something about overflow

April 30, 2013, 07:44		#14
j.zamorano New Member Javier Join Date: Apr 2013 Posts: 7 Rep Power: 13	I am having quite good results with steady state simulations. Now, I am doing transient ones. I want to reach the second state of the picture by improving the gas velocity. But, although the function for the gas velocity inlet is well defined, when I plot the results it shows everytime the first state (left). Can be this because problems of convergence? What should I do? http://www.subirimagenes.com/otros-t...-8411961.html][/URL]

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Thermocapillary free surface flow	zakifoam	OpenFOAM Running, Solving & CFD	10	December 12, 2016 12:44
CFX gravity driven free surface flow tutorial	mechovator	CFX	37	July 27, 2009 11:28
free surface flow in non-inertial reference frame	Tiedingg	FLOW-3D	1	February 26, 2009 20:51
Free surface vortex flow	Guillaume	CFX	3	August 25, 2005 21:52
Free surface flow simulation question	Jane Luo	FLUENT	1	April 23, 2004 05:26

April 17, 2013, 19:48		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	Please post an image of what you are modelling and the CCL or output file.

April 22, 2013, 04:57		#4
j.zamorano New Member Javier Join Date: Apr 2013 Posts: 7 Rep Power: 13	Is this enough information? If you need the output file or the CCL please give me your e-mail, and I will send it to you. Thanks

April 22, 2013, 10:00		#5
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	Why have you chose a 1ms time step? Unless you have actually done a sensitivity analysis on that time step size and determined it is what you need then it is bound to be wrong. So let the solver work the time step out for you. Use adaptive timestepping, homing in on 3-5 coeff loops per iteration. Are you using surface tension? Why are you using a highly inaccurate spatial discretisation scheme (upwind) but a very accurate temporal scheme (2nd order back eular)? If you just want quick and dirty results then run 1st order in time as well. If you want an accurate answer you will need to run a better spatial discretisation than that.

April 22, 2013, 10:28		#6
JuPa Senior Member Mr CFD Join Date: Jun 2012 Location: Britain Posts: 361 Rep Power: 15	If you let the solver work out the time step via adaptive time stepping I've found the time step it selects is usually prohibitively small, such as 1e-8 s (that is if you converge to 3-5 loops).

April 22, 2013, 19:25		#7
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	That is because it needs small time steps! There are very few simulations where this does not find the correct time step. The only exception to this is when you are happy to sacrifice accuracy to just "muddle" through a tricky bit of the simulation and don't care about accuracy. An example could be when you first start the simulation up, or when an aggressive action occurs which you do not care about the details (eg a fast combustion event). If this slows the simulation too much then * Check that your mesh is not too fine * Check your convergence is not too tight * Get a faster computer * Get more computers and run parallel.

April 24, 2013, 10:34		#8
j.zamorano New Member Javier Join Date: Apr 2013 Posts: 7 Rep Power: 13	I am new in CFX, What parameters should I take to make adaptive timestepping? Max Courant number?

April 25, 2013, 08:04		#9
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	No. Courant number only has a tenuous link to convergence in CFX as it is an implicit solver. The recommended approach for adaptive time stepping is homing in on 3-5 coeff loops per iteration.

April 25, 2013, 11:31		#10
j.zamorano New Member Javier Join Date: Apr 2013 Posts: 7 Rep Power: 13	I am not using surface tension forces, should I? I think the diameter is too big and this forces are too small (0.05 m). I am using free surface model, should I change it to mixture model? Thank you so much

April 25, 2013, 19:49		#11
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	Only use surface tension if it is important for the flow you are modelling. It is an expensive model and will slow you down a lot, so only use it if you have to.

April 26, 2013, 08:23		#13
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	This is a floating point error (which then caused lots of other errors including crashing the parallel run) - this is an FAQ: http://www.cfd-online.com/Wiki/Ansys...do_about_it.3F

April 30, 2013, 20:32		#15
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	I do not understand your problem. What are you seeing and what do you expect to happen?