laplacianFoam - solver settings for steady state diffusion problems

Adam_K · June 4, 2021, 07:50

Hello,

I'm trying to simulate the diffusion through a porous cube, with C = 1 at the top and C = 0 at the bottom. The other four sides are zero-flux boundaries. In terms of numerical solutions, I feel like this one should be quite simple, but it doesn't appear to be the case.

1) Is there a set of solver recommendations listed somewhere for steady state diffusion problems? I found a bunch of useful tutorials online (like wolfdynamics) but they typically target more complicated cases with flow and multiple phases.

2) Even for steady state simulations controlDict includes time steps which seem to simply be a way of keeping track of the number of solver iterations. Do I only really need to care about the final residuals? Or is it important to have many steps as well? Is a high number of iterations per step indicative of poor fvSchemes/fvSolution settings or is it mainly indicative that my tolerances are too tight?

Initially I used a GAMG solver but the resulting flux values through my domain did not converge very well. So then I started using the PCG solver with the following fvSolution file.

Code:

/*--------------------------------*- C++ -*----------------------------------*\
| =========                 |                                                 |
| \\      /  F ield         | OpenFOAM: The Open Source CFD Toolbox           |
|  \\    /   O peration     | Version:  v2012                                 |
|   \\  /    A nd           | Website:  www.openfoam.com                      |
|    \\/     M anipulation  |                                                 |
\*---------------------------------------------------------------------------*/
FoamFile
{
    version     2.0;
    format      ascii;
    class       dictionary;
    location    "system";
    object      fvSolution;
}
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
solvers
{
    T
    {
        solver          PCG;
        preconditioner  DIC;
        tolerance       1e-8;
        relTol          0;
    }
}
SIMPLE
{
    nNonOrthogonalCorrectors 1;
}
// ************************************************************************* //

Here's an example of my output. In the 25 steps the final residuals are all very close, but the number of iterations decreases considerably. In a case like this, is there any benefit to going past one or two steps since the final residual doesn't change?

Code:

/*---------------------------------------------------------------------------*\
  =========                 |
  \\      /  F ield         | OpenFOAM: The Open Source CFD Toolbox
   \\    /   O peration     | Website:  https://openfoam.org
    \\  /    A nd           | Version:  8
     \\/     M anipulation  |
\*---------------------------------------------------------------------------*/
Build  : 8-340defec456f
Exec   : laplacianFoam -parallel
Date   : May 27 2021
Time   : 15:33:04
Host   : "srv16-seds"
PID    : 262648
I/O    : uncollated
Case   : /home/user/CUBE/OL1_LC1000
nProcs : 4
Slaves : 
3
(
"srv16-seds.262649"
"srv16-seds.262650"
"srv16-seds.262651"
)

Pstream initialized with:
    floatTransfer      : 0
    nProcsSimpleSum    : 0
    commsType          : nonBlocking
    polling iterations : 0
sigFpe : Enabling floating point exception trapping (FOAM_SIGFPE).
fileModificationChecking : Monitoring run-time modified files using timeStampMaster (fileModificationSkew 10)
allowSystemOperations : Allowing user-supplied system call operations

// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
Create time

Create mesh for time = 0


SIMPLE: No convergence criteria found

Reading field T

Reading transportProperties

Reading diffusivity DT

No finite volume options present

Calculating temperature distribution

Time = 1

DICPCG:  Solving for T, Initial residual = 0.99999999999107569426115560418111, Final residual = 9.7359573231362151803842296593147e-09, No Iterations 674
DICPCG:  Solving for T, Initial residual = 0.29936521364782786580605034032487, Final residual = 9.8227074689240328111439158327405e-09, No Iterations 598
ExecutionTime = 200.78999999999999204192135948688 s  ClockTime = 202 s

Time = 2

DICPCG:  Solving for T, Initial residual = 0.04936009407679749216857345572862, Final residual = 9.5588303849891255977202478202187e-09, No Iterations 559
DICPCG:  Solving for T, Initial residual = 0.012411825469636579219456784528575, Final residual = 9.9918096524054039411005939392944e-09, No Iterations 465
ExecutionTime = 361.29000000000002046363078989089 s  ClockTime = 363 s

Time = 3

DICPCG:  Solving for T, Initial residual = 0.0049470722656381055370999888509687, Final residual = 9.909222132449522098222617898667e-09, No Iterations 414
DICPCG:  Solving for T, Initial residual = 0.0028858897183733894092894090022128, Final residual = 9.9931511985862581086117862705191e-09, No Iterations 359
ExecutionTime = 488.35000000000002273736754432321 s  ClockTime = 494 s

Time = 4

DICPCG:  Solving for T, Initial residual = 0.0021913925195753808893017744452436, Final residual = 9.9800311390956302319122818021875e-09, No Iterations 286
DICPCG:  Solving for T, Initial residual = 0.0018585806673575490811228805299038, Final residual = 9.9114819390907541165364761032511e-09, No Iterations 214
ExecutionTime = 570.04999999999995452526491135359 s  ClockTime = 576 s

Time = 5

DICPCG:  Solving for T, Initial residual = 0.0016614318260583058484841645352503, Final residual = 9.8674054518951495158781637420317e-09, No Iterations 92
DICPCG:  Solving for T, Initial residual = 0.0015266092312223946986932121561154, Final residual = 9.9325194049474741141119227337379e-09, No Iterations 32
ExecutionTime = 591.57000000000005002220859751105 s  ClockTime = 598 s

Time = 6

DICPCG:  Solving for T, Initial residual = 0.0014273509883033445172495490282927, Final residual = 9.7150124507147298465936570855545e-09, No Iterations 17
DICPCG:  Solving for T, Initial residual = 0.0013513557145077965092327954010898, Final residual = 9.746098580813178473996809356595e-09, No Iterations 12
ExecutionTime = 597.67999999999994997779140248895 s  ClockTime = 604 s

Time = 7

DICPCG:  Solving for T, Initial residual = 0.0012922577249493555351378404338902, Final residual = 8.765346045313475593952001546031e-09, No Iterations 11
DICPCG:  Solving for T, Initial residual = 0.001245602391909901183672526592261, Final residual = 8.2352667077197453915482681604567e-09, No Iterations 11
ExecutionTime = 602.64999999999997726263245567679 s  ClockTime = 609 s

Time = 8

DICPCG:  Solving for T, Initial residual = 0.0012073769139051989361666183953048, Final residual = 9.5026691000652706992280743651148e-09, No Iterations 10
DICPCG:  Solving for T, Initial residual = 0.0011757369760335395058453711314428, Final residual = 7.2940530228971660421404454768129e-09, No Iterations 11
ExecutionTime = 607.40999999999996816768543794751 s  ClockTime = 614 s

Time = 9

DICPCG:  Solving for T, Initial residual = 0.0011494781238797395842010651278997, Final residual = 8.3643011070344197137123765269096e-09, No Iterations 10
DICPCG:  Solving for T, Initial residual = 0.0011276899527113946625905827048086, Final residual = 9.0911588808574767646151549606304e-09, No Iterations 10
ExecutionTime = 612.10000000000002273736754432321 s  ClockTime = 619 s

Time = 10

DICPCG:  Solving for T, Initial residual = 0.0011095948934148026302631828698964, Final residual = 8.8844444582587322795010770226654e-09, No Iterations 10
DICPCG:  Solving for T, Initial residual = 0.0010946490732921261134630297462422, Final residual = 8.5150157599228302120235789204883e-09, No Iterations 10
ExecutionTime = 616.67999999999994997779140248895 s  ClockTime = 623 s

Time = 11

DICPCG:  Solving for T, Initial residual = 0.0010826748904859536878908521373432, Final residual = 8.3195762452344395818437308839574e-09, No Iterations 10
DICPCG:  Solving for T, Initial residual = 0.0010731954153872067143915547404731, Final residual = 8.0562293940606097488640790420071e-09, No Iterations 10
ExecutionTime = 621.29999999999995452526491135359 s  ClockTime = 628 s

Time = 12

DICPCG:  Solving for T, Initial residual = 0.001066093538563909126279383166036, Final residual = 7.8890377911349672425259712399415e-09, No Iterations 10
DICPCG:  Solving for T, Initial residual = 0.0010613834567342851026433558914164, Final residual = 7.706864278009963131769483209426e-09, No Iterations 10
ExecutionTime = 625.95000000000004547473508864641 s  ClockTime = 633 s

Time = 13

DICPCG:  Solving for T, Initial residual = 0.0010587205574505466343371296034093, Final residual = 9.925345394732363765788595998292e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.0010577118511664704465780051378943, Final residual = 7.6419085909385810127787366135069e-09, No Iterations 10
ExecutionTime = 630.47000000000002728484105318785 s  ClockTime = 637 s

Time = 14

DICPCG:  Solving for T, Initial residual = 0.001057805476903748466421806284643, Final residual = 9.3352777658709296960675103082584e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.0010589031917784710694341310244226, Final residual = 9.4994149884089111388916018292178e-09, No Iterations 9
ExecutionTime = 634.78999999999996362021192908287 s  ClockTime = 641 s

Time = 15

DICPCG:  Solving for T, Initial residual = 0.001060947571672199180731199774641, Final residual = 7.3132647701374290112027792484795e-09, No Iterations 10
DICPCG:  Solving for T, Initial residual = 0.0010638777983996935451449505904975, Final residual = 8.9612383765949922625054858732647e-09, No Iterations 9
ExecutionTime = 639.27999999999997271515894681215 s  ClockTime = 646 s

Time = 16

DICPCG:  Solving for T, Initial residual = 0.0010676714089250536124375834390321, Final residual = 9.1361847543866629127292864048088e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.0010722922104279652803782152403755, Final residual = 7.057993721573911677559808085583e-09, No Iterations 10
ExecutionTime = 643.76999999999998181010596454144 s  ClockTime = 651 s

Time = 17

DICPCG:  Solving for T, Initial residual = 0.0010777213747203905850474692229568, Final residual = 8.7014997708987952197143817922691e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.0010839168463820873355291096729047, Final residual = 8.9290261775459772933884826221808e-09, No Iterations 9
ExecutionTime = 648.08000000000004092726157978177 s  ClockTime = 655 s

Time = 18

DICPCG:  Solving for T, Initial residual = 0.0010908465365125387391831424110933, Final residual = 6.8529259759907971351336162266297e-09, No Iterations 10
DICPCG:  Solving for T, Initial residual = 0.0010985028170533844837919446035812, Final residual = 8.4932865057385959629627079805862e-09, No Iterations 9
ExecutionTime = 652.53999999999996362021192908287 s  ClockTime = 659 s

Time = 19

DICPCG:  Solving for T, Initial residual = 0.0011068681176652969948975835379201, Final residual = 8.744625818748223665747071526623e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.0011159362932743601901086627492532, Final residual = 9.8523988577962435105843527153452e-09, No Iterations 9
ExecutionTime = 656.91999999999995907273842021823 s  ClockTime = 664 s

Time = 20

DICPCG:  Solving for T, Initial residual = 0.0011256943819246672394696062724506, Final residual = 9.7968062836950933207724018757234e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.0011361396297413822904787128464932, Final residual = 9.746827695909840881030629660027e-09, No Iterations 9
ExecutionTime = 661.25 s  ClockTime = 668 s

Time = 21

DICPCG:  Solving for T, Initial residual = 0.0011472620636991796329307025104072, Final residual = 9.7136993136242671375382118597978e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.0011590610956644252418218377087555, Final residual = 9.6817867381137596561430872649234e-09, No Iterations 9
ExecutionTime = 665.57000000000005002220859751105 s  ClockTime = 673 s

Time = 22

DICPCG:  Solving for T, Initial residual = 0.0011715315234211794073043710895377, Final residual = 9.6586736034028382467561365306435e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.0011846738895478236587577303140506, Final residual = 9.6407293764852469080778942000023e-09, No Iterations 9
ExecutionTime = 669.87999999999999545252649113536 s  ClockTime = 677 s

Time = 23

DICPCG:  Solving for T, Initial residual = 0.0011984888612897154867853455328941, Final residual = 9.6288876976836497699403070072051e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.0012129765554697276644380288246339, Final residual = 9.6220355587989789020312954227927e-09, No Iterations 9
ExecutionTime = 674.23000000000001818989403545856 s  ClockTime = 681 s

Time = 24

DICPCG:  Solving for T, Initial residual = 0.0012281366253050461385259994884223, Final residual = 9.6215856387426752108226702341585e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.0012439737911124472500429272159295, Final residual = 9.6302432487258157729989954350981e-09, No Iterations 9
ExecutionTime = 678.55999999999994543031789362431 s  ClockTime = 686 s

Time = 25

DICPCG:  Solving for T, Initial residual = 0.0012604911384544083968739514034496, Final residual = 9.6440262237010657485857656858255e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.0012776885952470796222607551939632, Final residual = 9.6624218731269942531164125103585e-09, No Iterations 9
ExecutionTime = 682.87999999999999545252649113536 s  ClockTime = 690 s

End

Finalising parallel run

The above solution (aside from the question of the necessity of 25 steps) does seem to work fine on a 100mm X 100 mm X 100 mm cube but when I increase the length to 1600 mm the number of iterations hits 1000 and the final residual is many orders of magnitude larger. Does it seem like I'm doing something wrong or is this to be expected with a higher aspect ratio mesh (going from 1:1:1: to 16:1:1)?

Code:

/*---------------------------------------------------------------------------*\
  =========                 |
  \\      /  F ield         | OpenFOAM: The Open Source CFD Toolbox
   \\    /   O peration     | Website:  https://openfoam.org
    \\  /    A nd           | Version:  8
     \\/     M anipulation  |
\*---------------------------------------------------------------------------*/
Build  : 8-340defec456f
Exec   : laplacianFoam -parallel
Date   : Jun 03 2021
Time   : 18:17:25
Host   : "srv16-seds"
PID    : 572296
I/O    : uncollated
Case   : /home/user/CASES/OL1_LC1000
nProcs : 4
Slaves : 
3
(
"srv16-seds.572297"
"srv16-seds.572298"
"srv16-seds.572299"
)

Pstream initialized with:
    floatTransfer      : 0
    nProcsSimpleSum    : 0
    commsType          : nonBlocking
    polling iterations : 0
sigFpe : Enabling floating point exception trapping (FOAM_SIGFPE).
fileModificationChecking : Monitoring run-time modified files using timeStampMaster (fileModificationSkew 10)
allowSystemOperations : Allowing user-supplied system call operations

// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
Create time

Create mesh for time = 0


SIMPLE: No convergence criteria found

Reading field T

Reading transportProperties

Reading diffusivity DT

No finite volume options present

Calculating temperature distribution

Time = 1

DICPCG:  Solving for T, Initial residual = 1, Final residual = 0.0013376263303212366888306217305171, No Iterations 1000
DICPCG:  Solving for T, Initial residual = 0.29183194732282119465693881465995, Final residual = 0.00026922573679248184164852775168697, No Iterations 1000
ExecutionTime = 10962.19000000000050931703299284 s  ClockTime = 10981 s

Time = 2

DICPCG:  Solving for T, Initial residual = 0.077903245197562878687769227781246, Final residual = 2.9009004950946739692348511074904e-05, No Iterations 1000
DICPCG:  Solving for T, Initial residual = 0.029377252753838793003504648027047, Final residual = 1.7210094286604748071815851329092e-05, No Iterations 1000
ExecutionTime = 22106.639999999999417923390865326 s  ClockTime = 22146 s

Time = 3

DICPCG:  Solving for T, Initial residual = 0.019212631418907503283444171415795, Final residual = 2.882887164641591866957480091993e-06, No Iterations 1000
DICPCG:  Solving for T, Initial residual = 0.017580080110607066573580326007686, Final residual = 1.7682425868997719388880712418355e-06, No Iterations 1000
ExecutionTime = 33438.139999999999417923390865326 s  ClockTime = 33497 s

Time = 4

Tobi · June 7, 2021, 09:29

Very strange behavior. Can you share your case?

By the way ... the initial residual is the important one. The final shows us, if the linear solver does find a solution of the matrix system we are solving at the moment.

Adam_K · June 7, 2021, 09:56

Quote:

Originally Posted by Tobi

Very strange behavior. Can you share your case?

Sure, I've attached the case files + .brep + .geo file (I mesh with gmsh) since I assume that uploading the actual mesh will be too big.

Quote:

Originally Posted by Tobi

By the way ... the initial residual is the important one. The final shows us, if the linear solver does find a solution of the matrix system we are solving at the moment.

Thanks for the clarification, looks like I've got some more background reading to do.

Tobi · June 7, 2021, 11:04

Please at in the source file of the laplacian.C file:

Code:

            fvOptions.constrain(TEqn);
            TEqn.relax();
            TEqn.solve();
            fvOptions.correct(T);

and recompile the application (wmake). In addition, insert in the fvSolution file the following at the end:

Code:

relaxationFactors
{
    equations
    {
        T     0.8;
    }
}

I guess it is too stiff and we need under-relaxation as the time-dependent term is missing (which is a limiter in the solution).

Adam_K · June 8, 2021, 10:05

Thanks for the tips!

I assume that you meant laplacianFoam.c in openfoam8/applications/solvers/basic/laplacianFoam.c.

I was not able to save the file because it was read only. So I did the following based on this thread.

Code:

cp -r $FOAM_APP $FOAM_RUN

Replaced FOAM_APPBIN in "$FOAM_RUN/applications/solvers/basic/laplacianFoam/Make
/files" by FOAM_USER_APPBIN. Then ran wmake.

It seems to run faster now (12 vs 33 seconds in the example below), although the initial residual after 25 steps is 5 times larger.

Here without relaxation:

Code:

/*---------------------------------------------------------------------------*\
  =========                 |
  \\      /  F ield         | OpenFOAM: The Open Source CFD Toolbox
   \\    /   O peration     | Website:  https://openfoam.org
    \\  /    A nd           | Version:  8
     \\/     M anipulation  |
\*---------------------------------------------------------------------------*/
Build  : 8-340defec456f
Exec   : laplacianFoam
Date   : Jun 08 2021
Time   : 11:20:30
Host   : "srv16-seds"
PID    : 757820
I/O    : uncollated
Case   : /home/user/testNoRelax
nProcs : 1
sigFpe : Enabling floating point exception trapping (FOAM_SIGFPE).
fileModificationChecking : Monitoring run-time modified files using timeStampMaster (fileModificationSkew 10)
allowSystemOperations : Allowing user-supplied system call operations

// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
Create time

Create mesh for time = 0


SIMPLE: No convergence criteria found

Reading field T

Reading transportProperties

Reading diffusivity DT

No finite volume options present

Calculating temperature distribution

Time = 1

DICPCG:  Solving for T, Initial residual = 0.99999999999946209694456911165616, Final residual = 9.3941622433799702091526345553406e-09, No Iterations 252
DICPCG:  Solving for T, Initial residual = 0.29041453461302718652703447332897, Final residual = 8.9072139109022076936970487514265e-09, No Iterations 220
ExecutionTime = 7.790000000000000035527136788005 s  ClockTime = 8 s

Time = 2

DICPCG:  Solving for T, Initial residual = 0.051972157937314272801998527029355, Final residual = 9.8446714508909861498172047165021e-09, No Iterations 206
DICPCG:  Solving for T, Initial residual = 0.016404898425008588441764700860404, Final residual = 9.7689199185977901698379693528999e-09, No Iterations 167
ExecutionTime = 13.83999999999999985789145284798 s  ClockTime = 14 s

Time = 3

DICPCG:  Solving for T, Initial residual = 0.008710286809118606560220499090974, Final residual = 9.7184480990647138006463984531355e-09, No Iterations 149
DICPCG:  Solving for T, Initial residual = 0.006196627359666518480851848948987, Final residual = 9.5917588970019467713552674555631e-09, No Iterations 135
ExecutionTime = 18.48000000000000042632564145606 s  ClockTime = 19 s

Time = 4

DICPCG:  Solving for T, Initial residual = 0.0050824877528735275256099868101956, Final residual = 9.9325600564494361335043619959904e-09, No Iterations 119
DICPCG:  Solving for T, Initial residual = 0.0044185282788387113608874479098176, Final residual = 9.9924352665122720477183177801607e-09, No Iterations 60
ExecutionTime = 21.48999999999999843680598132778 s  ClockTime = 22 s

Time = 5

...

Time = 21

DICPCG:  Solving for T, Initial residual = 0.00093303476365616012933035250398461, Final residual = 7.2291354382065740724956122301506e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.00090917770396538148420378933067809, Final residual = 7.0569503182932826831494063960729e-09, No Iterations 9
ExecutionTime = 31.37000000000000099475983006414 s  ClockTime = 32 s

Time = 22

DICPCG:  Solving for T, Initial residual = 0.00088608552134905149266080881531593, Final residual = 6.8869026565399615341733985862324e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.00086384993993195114889149799353163, Final residual = 6.7273561699599664429890091886608e-09, No Iterations 9
ExecutionTime = 31.8500000000000014210854715202 s  ClockTime = 32 s

Time = 23

DICPCG:  Solving for T, Initial residual = 0.00084231028520622916195420559759555, Final residual = 6.5680744046397447726836210386402e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.000821479538727757350492775234585, Final residual = 6.4177296108417369699678447646403e-09, No Iterations 9
ExecutionTime = 32.310000000000002273736754432321 s  ClockTime = 33 s

Time = 24

DICPCG:  Solving for T, Initial residual = 0.00080130841386544617889475805583288, Final residual = 6.2670486114671524639597905987753e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.00078203228372547334192738111013909, Final residual = 6.1249004959002893651056657832628e-09, No Iterations 9
ExecutionTime = 32.78000000000000113686837721616 s  ClockTime = 33 s

Time = 25

DICPCG:  Solving for T, Initial residual = 0.00076331279279923743182056927736312, Final residual = 5.9820402596463485177285687667461e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.00074538567135008673529561473714011, Final residual = 5.8474953906457247514119391674798e-09, No Iterations 9
ExecutionTime = 33.259999999999998010480339871719 s  ClockTime = 33 s

End

Here with relaxation:

Code:

/*---------------------------------------------------------------------------*\
  =========                 |
  \\      /  F ield         | OpenFOAM: The Open Source CFD Toolbox
   \\    /   O peration     | Website:  https://openfoam.org
    \\  /    A nd           | Version:  8
     \\/     M anipulation  |
\*---------------------------------------------------------------------------*/
Build  : 8-340defec456f
Exec   : laplacianFoam
Date   : Jun 08 2021
Time   : 11:22:32
Host   : "srv16-seds"
PID    : 757889
I/O    : uncollated
Case   : /home/user/test
nProcs : 1
sigFpe : Enabling floating point exception trapping (FOAM_SIGFPE).
fileModificationChecking : Monitoring run-time modified files using timeStampMaster (fileModificationSkew 10)
allowSystemOperations : Allowing user-supplied system call operations

// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
Create time

Create mesh for time = 0


SIMPLE: No convergence criteria found

Reading field T

Reading transportProperties

Reading diffusivity DT

No finite volume options present

Calculating temperature distribution

Time = 1

DICPCG:  Solving for T, Initial residual = 0.99999999996874677776759199332446, Final residual = 8.8496482942733626673009660816398e-09, No Iterations 11
DICPCG:  Solving for T, Initial residual = 0.19543329461575437155929080290662, Final residual = 2.0585641600321122840042101947468e-09, No Iterations 11
ExecutionTime = 0.78000000000000002664535259100376 s  ClockTime = 1 s

Time = 2

DICPCG:  Solving for T, Initial residual = 0.1006244947085474100711977030187, Final residual = 4.1127489003026880821653010842504e-09, No Iterations 10
DICPCG:  Solving for T, Initial residual = 0.067795502872944279904032782724244, Final residual = 2.9501797847286392794605311228453e-09, No Iterations 10
ExecutionTime = 1.3000000000000000444089209850063 s  ClockTime = 2 s

Time = 3

DICPCG:  Solving for T, Initial residual = 0.050424405476877633136556511317394, Final residual = 1.982804737723231519840403418403e-09, No Iterations 10
DICPCG:  Solving for T, Initial residual = 0.039881397990589102786618269647079, Final residual = 8.6591988770661996594947036479453e-09, No Iterations 9
ExecutionTime = 1.8000000000000000444089209850063 s  ClockTime = 2 s

Time = 4

DICPCG:  Solving for T, Initial residual = 0.032879924207762763199713873518704, Final residual = 6.8781960995812208787371059171434e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.027889144238455992191871857244223, Final residual = 5.6553883176248379435887428195139e-09, No Iterations 9
ExecutionTime = 2.290000000000000035527136788005 s  ClockTime = 2 s

Time = 5

DICPCG:  Solving for T, Initial residual = 0.024140937991051145128373889292561, Final residual = 4.7737373389323962805594811216296e-09, No Iterations 9
DICPCG:  Solving for T, Initial residual = 0.021268055373218387738010193288574, Final residual = 4.1233172868907930064703193818055e-09, No Iterations 9
ExecutionTime = 2.7700000000000000177635683940025 s  ClockTime = 3 s

...

Time = 21

DICPCG:  Solving for T, Initial residual = 0.0046820965781367409122570855117829, Final residual = 5.0277044066751759810488688995193e-09, No Iterations 8
DICPCG:  Solving for T, Initial residual = 0.0045712059814553328632169559853082, Final residual = 4.9511166906100507681988477409925e-09, No Iterations 8
ExecutionTime = 10.16999999999999992894572642399 s  ClockTime = 10 s

Time = 22

DICPCG:  Solving for T, Initial residual = 0.0044655666030541368155537540474143, Final residual = 4.878484250625371291884908089018e-09, No Iterations 8
DICPCG:  Solving for T, Initial residual = 0.0043648094662825601824929222516403, Final residual = 4.8096077017395777760449464104109e-09, No Iterations 8
ExecutionTime = 10.63000000000000078159700933611 s  ClockTime = 11 s

Time = 23

DICPCG:  Solving for T, Initial residual = 0.0042685994508113236922675071127742, Final residual = 4.7444234087666250395576231365605e-09, No Iterations 8
DICPCG:  Solving for T, Initial residual = 0.004176631980765625057339107684129, Final residual = 4.6829679227423145062876648383847e-09, No Iterations 8
ExecutionTime = 11.08000000000000007105427357601 s  ClockTime = 11 s

Time = 24

DICPCG:  Solving for T, Initial residual = 0.0040886276886142360054154920590008, Final residual = 4.6251601516086179586614514000747e-09, No Iterations 8
DICPCG:  Solving for T, Initial residual = 0.004004332917892248523694664896766, Final residual = 4.5712454425893547619251703977537e-09, No Iterations 8
ExecutionTime = 11.53999999999999914734871708788 s  ClockTime = 12 s

Time = 25

DICPCG:  Solving for T, Initial residual = 0.003923514738339924264043379054101, Final residual = 4.5211003078209630404462088964536e-09, No Iterations 8
DICPCG:  Solving for T, Initial residual = 0.0038459594557208845629070470550914, Final residual = 4.4746326395357796834751908056013e-09, No Iterations 8
ExecutionTime = 11.99000000000000021316282072803 s  ClockTime = 12 s

End

Tobi · June 8, 2021, 15:33

Yes that is obvious that it runs faster and that the initial residual is not as small due to the fact that we under-relax the matrix system. That means, we put some part from the matrix A to the source vector b. The result is: more explicit but the diagonal dominance of the matrix A is larger which helps the matrix solver to find the solution faster - indicated to the No iterations which are less for relaxed systems. I just rechecked your case:

Code:

laplacianSchemes
{
    default         none;
    laplacian(DT,T) Gauss linear orthogonal;
}

interpolationSchemes
{
    default         linear;
}

snGradSchemes
{
    default         corrected;
}

Question: Do you have a 100 % hexahedral mesh? if yes, why do you set corrected for the surface gradient calculation and if not, your laplace scheme is wrong.

June 7, 2021, 09:29		#2
Tobi Super Moderator Tobias Holzmann Join Date: Oct 2010 Location: Bad Wörishofen Posts: 2,711 Blog Entries: 6 Rep Power: 52	Very strange behavior. Can you share your case? By the way ... the initial residual is the important one. The final shows us, if the linear solver does find a solution of the matrix system we are solving at the moment. __________________ Keep foaming, Tobias Holzmann

June 7, 2021, 11:04		#4
Tobi Super Moderator Tobias Holzmann Join Date: Oct 2010 Location: Bad Wörishofen Posts: 2,711 Blog Entries: 6 Rep Power: 52	Please at in the source file of the laplacian.C file: Code: fvOptions.constrain(TEqn); TEqn.relax(); TEqn.solve(); fvOptions.correct(T); and recompile the application (wmake). In addition, insert in the fvSolution file the following at the end: Code: relaxationFactors { equations { T 0.8; } } I guess it is too stiff and we need under-relaxation as the time-dependent term is missing (which is a limiter in the solution). croecker likes this. __________________ Keep foaming, Tobias Holzmann

June 8, 2021, 15:33		#6
Tobi Super Moderator Tobias Holzmann Join Date: Oct 2010 Location: Bad Wörishofen Posts: 2,711 Blog Entries: 6 Rep Power: 52	Yes that is obvious that it runs faster and that the initial residual is not as small due to the fact that we under-relax the matrix system. That means, we put some part from the matrix A to the source vector b. The result is: more explicit but the diagonal dominance of the matrix A is larger which helps the matrix solver to find the solution faster - indicated to the No iterations which are less for relaxed systems. I just rechecked your case: Code: laplacianSchemes { default none; laplacian(DT,T) Gauss linear orthogonal; } interpolationSchemes { default linear; } snGradSchemes { default corrected; } Question: Do you have a 100 % hexahedral mesh? if yes, why do you set corrected for the surface gradient calculation and if not, your laplace scheme is wrong. __________________ Keep foaming, Tobias Holzmann

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Solver settings: Coupled, multi-grid, control-volume-based solver	ocyee	CFX	4	May 1, 2020 08:20
PEMFC model with FLUENT	brahimchoice	FLUENT	22	April 19, 2020 16:44
SU2 Inviscid Hypersonic Solver Settings	jclegg	SU2	2	April 10, 2019 12:35
fluent divergence for no reason	sufjanst	FLUENT	2	March 23, 2016 17:08
Ansys CFX problem: unexpected very high temperatures in premix laminar combustion	faizan_habib7	CFX	4	February 1, 2016 18:00