CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   VOF-what residuals are worth considering "converged"? (http://www.cfd-online.com/Forums/fluent/72600-vof-what-residuals-worth-considering-converged.html)

 cfdiscool February 11, 2010 10:27

VOF-what residuals are worth considering "converged"?

Hallo

My case: 3d VOF 0.2 < all URFs < 0.5 steady state i have 4.2 million hexa cells

I am trying to see the interaction of air and water in high pressure nozzles of very less diameter (0.7mm)

I have set the residuals to 10^-4. I patched the values of 1 for water as secondary phase

1. What powers of residuals can now be considered as valid for a converged solution ? (I already have all cont-vel-k-eps-volfrac residuals less than 10^-3) ?

should i simulate further or can I take these results as valid with 10^-3 residuals?

2.Basic question is whether if a Multiphase problem like this can be treated as "Steady" ?

3.What is normally the criterion to "decide" whether a simulation is converged or not??

:)

 Trev February 11, 2010 12:28

I would have all residuals converged below at least 10-4 and check that your mass flow rate is steady and the mass flux error is around the order of 10-7 kg/s then you can consider your solution converged.

 cfdiscool February 11, 2010 17:33

Hi Neil
Thank u:)

How can I do these in FLUENT ?

 Trev February 11, 2010 17:45

Solve>Monitors>Residual
Solve>Monitors>Surface -->Define, then set report type to mass flowrate

Neil

 cfdiscool February 12, 2010 13:11

Hi Niel

many thanks :)...I would try your suggestion.....

For my VOF model residuals are going better for

cont x y z vel k and vol fraction...to thee orders of 10^-4 in just 100 iterations.....

But epsilon give me the trouble......it converges very slow and after 150 iterations it starts diverging and alsso then vol fraction residuals are raising ....

My case would need at least 10^-5 convergence it seems.....

Should i switch to other discretizition schemes after running the first 100.120 iterations ?

Or is there a way to go get convergence after attaining all residuals 10^-3 ??
:confused:

Should i change to unsteady solution ?

for ex changing urfs ...as previously mentioned i have the constraint due to VOF to keep my residuals between 0.2 and 0.5

thanks

 Trev February 12, 2010 13:37

I would switch to the unsteady solver then you can use what ever discretisation you want and not have troubles with convergence if the time step is set correctly. Firstly though just let the solution run as it is and wait for the residuals to level out and see what order of magnitude they are before doing this. Bear in mind you will need way more iterations than 100-200 for most cases before the boundary conditions have been processed and the solution has evolved.

Neil

 cfdiscool February 13, 2010 22:47

Hi Niel

I tried switching to unsteady and attempted to do with variable time step...

smallest size cell is 0.001 mm ...its a very fine grid and large one....

My time step turns out to be in the order of 10^-8 and i even decreased the minimum step size to 10^-10

in one run FLUENT gave the out put as

courant number is 0 and in other simulation attempt

it gave the following report and quit to iterate:

Updating solution at time level N...
Global Courant Number : 0.28
done.
physical-dt 1.0000e-06
iter continuity x-velocity y-velocity z-velocity k epsilon time/iter
1 1.0000e+00 1.5793e+02 2.2789e-01 2.2312e-02 2.8591e+00 5.6203e+07 0:17:21 19

Internal error at line 8910 in file 'sg_press.c' on Node 0.
couldn't allocate velocity and coefficient array

Did i set the time step too low than FLUENT defaults ...???

my flow velocity is around 220 m/s

thanks

 Philipov February 14, 2010 12:42

hi,
it is always a good idea to use unsteady flow with VOF method... do not be afraid of the solution.... look at the BC and be sure you are not using improper ones like outflow for example....
I have a question: are you sure you have 4.2 millions nodes?!?!? how do you calculate this mesh amount?!?!?!?!
Look at the CFL number.... it must not be so big.....
if you have such velocity (220m/s) and cell size of 0.001.... try to find the step size.... it is about 4E-8s

 cfdiscool February 15, 2010 08:29

Hi Philipov

Thank u for the remarks....

I also would like to go further with a Unsteady VOF solution.....but my problem is following:

After getting some good convergence at 10^-4 my residuals start to diverge and FLUENT also LIMITS the turbulent viscosity ratio...and the number of the cells in which turbulent viscosity ratio is limited increases gradually with the number of iterations, however there is little fall in residuals from this point where turbulent viscosity is limited....

I am sure I have mire than 4.2 million nodes...Its a very complex grid with cell sizes being at the range 0.001 mm

so at such small sizes and complex grid...i wouldn't try Unsteady case which is still complex....

I have set my time step in unsteady case to the order of 10^-8 & changed to variable time stepping with adjusted max and min time stepping defaults , even though my solution does not start after initialization ....

this is bit challenging for me ...and as a beginner it is difficult to get my head around this....

regards

 pranab_jha April 6, 2010 16:09

Hi,
I am running some simulations in a pipe with two-phase flow (air and water). I am also not getting very good convergence with continuity. I am using unsteady solver with no-iterative time adv. How can I get better convergence with the mass-flux. For air, mass conservation is OK but now with water.
Also, I initialize air volume fraction to be 1 inside the pipe. If I put all water (air vf=0), I get divergence along with reverse flow on my pressure outlet boundary.
Do you guys have any suggestions?
Thanks,
Pranab

 All times are GMT -4. The time now is 13:49.