[RodriguezFatz]

Edit: I just made some pictures of your residuals over iteration / time. This looks ugly. I will post the picture.

[RodriguezFatz]

output.png
This is how the residuals of your case look like.

[tigger]

Philipp,
The solver I use is 'interFoam' and it is multiphase(water, air in this case) and unsteady solver.
interFoam use VOF(Volume of fluid) method to capture water interface(Freesurface).
In VOF, if I understands it correctly, solve 1 transport equation and 1 continuity equation that containing Alpha ( If Alpha is 0 : cell is gas phase(air), 1 : cell is full of water and value between 0 and 1 is interface..
I'll try under-relaxation factor on alpha!

The problem I solve is a floating square box with freestream..(Water flow past a square box)

Residuals in the picture you uploaded shows sudden rise of continuity..I don't know why but there should be something wrong..

by the way, which tool you used to draw residual plot? Did you used GNU plot?
If so, can I see the script you used?

Many thanks,
Jason

[RodriguezFatz]

Hi,

Yes it is gnuplot.
Copy the code to a file called "plotscript" in the same directory as your log file is.
You can run it with "gnuplot plotscript" during runtime, the plot will redone every 1 second.
If you want to plot a picture just delete the "#" in the first two lines and comment out the last lines "reread" and "pause 1".
Everything should be self-explanatory, but if you don't understand it, just ask...

#set terminal png size 800,600 enhanced font "Helvetica,12"
#set output 'output.png'
set logscale y
set title "Residuals"
set ylabel 'Residual'
set xlabel 'Iteration'
set key outside
set grid
set yrange[1e-14:10]

nCorrectors=3
nNonOrthogonalCorrectors=1

nCont = nCorrectors
nP = (nNonOrthogonalCorrectors+1)*nCont
nPSkip = nP-1
nContSkip = nCont-1



plot "< cat log | grep 'Solving for alpha.water' | cut -d' ' -f9 | tr -d ','" title 'alpha' with linespoints,\
"< cat log | grep 'Solving for k' | cut -d' ' -f9 | tr -d ','" title 'k' with linespoints,\
"< cat log | grep 'Solving for epsilon' | cut -d' ' -f9 | tr -d ','" title 'epsilon' with linespoints,\
"< cat log | grep 'time step continuity errors :' | cut -d' ' -f9 | tr -d ','" every nCont::nContSkip title 'continuity' with linespoints,\
"< cat log | grep 'Solving for p_rgh' | cut -d' ' -f9 | tr -d ','" every nP::nPSkip title 'p' with linespoints lc 8
pause 1
#xmax = GPVAL_DATA_X_MAX+2
#xmin = xmax-20
#set xrange [xmin:xmax]
reread

If you change the solver settings you must always also change the plotscript (OuterCorrections,...)

[RodriguezFatz]

The sudden rise comes from the next time step (every 15 iterations). This is quite normal for PIMPLE.
It is not good, that the residuals don't converge appreciably. Is there any change that you can run a steady state solution prior? Such as SIMPLE for initialization?

[tigger]

Philipp,

The scrip you gave me works well!
Thanks!

Well, I've never tried SIMPLE algorithm for initialization..

[Elham]

Hi,

I had a very awfull number of iteration for p-rgh, around 190, and after changing relTol from 0 to 0.001 it is great, around 10.

Thanks for your advice.

[RodriguezFatz]

Great, but does it converge now?

[Elham]

Unfortunately, I didn't pay attention to the second attempt for solving p-rgh which is still have high No iteration.

MULES: Solving for alpha.water
MULES: Solving for alpha.water
MULES: Solving for alpha.water
Phase-1 volume fraction = 2.111971e-05 Min(alpha1) = 0 Max(alpha1) = 0.9974017
GAMG: Solving for p_rgh, Initial residual = 0.05853901, Final residual = 5.498377e-05, No Iterations 6
time step continuity errors : sum local = 3.274181e-09, global = 1.346159e-14, cumulative = -6.08748e-09
GAMG: Solving for p_rgh, Initial residual = 0.0001254836, Final residual = 9.72051e-07, No Iterations 80
time step continuity errors : sum local = 5.814538e-11, global = 5.793274e-17, cumulative = -6.08748e-09
ExecutionTime = 80707.64 s ClockTime = 81027 s

[RodriguezFatz]

You could try a different smoother for the GAMG.

p
    {
        solver           GAMG;
        tolerance        1e-12;
        relTol           0.1;
        smoother         DICGaussSeidel;
        nPreSweeps       0;
        nPostSweeps      1;
        nFinestSweeps    2;
        scaleCorrection  true;
        directSolveCoarsestLevel false;
        cacheAgglomeration on;
        agglomerator     faceAreaPair;
        nCellsInCoarsestLevel 500;
        mergeLevels      1;
        maxIter         100;
    }

I meant the PIMPLE algorithm, not the linear solvers, sorry. Does PIMPLE converge now?