[kandelabr]

I have a 3d model of the fluid inside a pump impeller. I use SRFPimpleFoam with these boundary conditions (at the moment I have turbulence modelling turned off):

0/Urel:

internalField   uniform (0 0 0);

boundaryField
{
    Inlet
    {
        type            SRFVelocity;
        inletValue      uniform (5.5 0 0);
        relative        no;
        value           uniform (5.5 0 0);
    }

    Outlet
    {
        type            rotatingPressureInletOutletVelocity;
        omega           constant (439.823 0 0); // 4200 rpm
        value           uniform (0 0 0);
    }

    Walls
    {
        type            fixedValue;
        value           uniform (0 0 0);
    }
}

0/p:

internalField   uniform 0;

boundaryField
{
    Inlet
    {
        type            zeroGradient;
    }

    Outlet
    {
        type            fixedValue;
        value           uniform 0;
    }
    Walls
    {
        type            zeroGradient;
    }
}

The problem is that the solution is not converging. Max courant number jumps to high values very quickly. I guess this has something to do with conditions on the outlet but the problem is that in a rotating reference frame I can't extend outlet further because it all becomes an impeller. Am i correct? Can this be a meshing problem? I use tetrahedral mesh made with Salome.
Any suggestions?

Thanks

[pablo_fil]

Hi kandelabr!

Maybe your time step is too large. Try to enable adaptation with a low target Courant number to find out if you where too far from the appropiate time step value.

Hope it helps!

[kandelabr]

Hello,
thanks for reply.
My initial courant number is as low as 0.5 but in a few timesteps it escalates to 1000+. When I cancel calculation and see intermediate results in paraFoam, there's a strangely high velocity right on the edge of blade on the outlet.

My 'theory' is that I should include part of the fluid outside the impeller. I added a ring on the outlet (see attachment). Walls of the ring don't rotate.

I have updated boundary conditions:

0/Urel:

boundaryField
{
    Inlet
    {
        type            SRFVelocity;
        inletValue      uniform (5.5 0 0);
        relative        no;
        value           uniform (5.5 0 0);
    }

    Outlet
    {
        type            rotatingPressureInletOutletVelocity;
        omega           constant (439.6 0 0);
        value           uniform (0 0 0);

    }

    ImpellerWalls
    {
        type            fixedValue;
        value           uniform (0 0 0);
    }

    VoluteWalls
    {
        type            SRFVelocity;
        inletValue      uniform (0 0 0);
        relative        no;
        value           uniform (0 0 0);
    }
}

0/p:

boundaryField
{
    Inlet
    {
        type            zeroGradient;
    }

    Outlet
    {
        type            fixedValue;
        value           uniform 0;
    }
    "(ImpellerWalls|VoluteWalls)"
    {
        type            zeroGradient;
    }
}

The results are the same though. Something is not right since courant number gets pretty high (before I cancel the whole thing).

Starting time loop

Courant Number mean: 0.000251018 max: 0.498309
Time = 5e-006

PIMPLE: iteration 1
smoothSolver:  Solving for Urelx, Initial residual = 1, Final residual = 0.0589673, No Iterations 2
smoothSolver:  Solving for Urely, Initial residual = 1, Final residual = 0.0353024, No Iterations 2
smoothSolver:  Solving for Urelz, Initial residual = 1, Final residual = 0.0354758, No Iterations 2
GAMG:  Solving for p, Initial residual = 1, Final residual = 0.0251039, No Iterations 10
time step continuity errors : sum local = 3.17346e-005, global = -9.67471e-006, cumulative = -9.67471e-006
GAMG:  Solving for p, Initial residual = 0.0642548, Final residual = 6.36254e-009, No Iterations 28
time step continuity errors : sum local = 3.2037e-009, global = 9.18978e-010, cumulative = -9.67379e-006
ExecutionTime = 21.036 s  ClockTime = 21 s

Courant Number mean: 0.115253 max: 11.8192
Time = 1e-005

PIMPLE: iteration 1
smoothSolver:  Solving for Urelx, Initial residual = 0.295107, Final residual = 0.0172095, No Iterations 2
smoothSolver:  Solving for Urely, Initial residual = 0.325871, Final residual = 0.0193292, No Iterations 2
smoothSolver:  Solving for Urelz, Initial residual = 0.32537, Final residual = 0.0192109, No Iterations 2
GAMG:  Solving for p, Initial residual = 0.0443229, Final residual = 0.00123252, No Iterations 4
time step continuity errors : sum local = 0.000405645, global = 0.000161995, cumulative = 0.000152321
GAMG:  Solving for p, Initial residual = 0.170854, Final residual = 5.22185e-009, No Iterations 31
time step continuity errors : sum local = 8.10986e-010, global = -2.2843e-010, cumulative = 0.000152321
ExecutionTime = 76.17 s  ClockTime = 76 s

Courant Number mean: 0.114349 max: 13.1679
Time = 1.5e-005

PIMPLE: iteration 1
smoothSolver:  Solving for Urelx, Initial residual = 0.0773901, Final residual = 0.00273242, No Iterations 2
smoothSolver:  Solving for Urely, Initial residual = 0.063731, Final residual = 0.00264391, No Iterations 2
smoothSolver:  Solving for Urelz, Initial residual = 0.0636405, Final residual = 0.00263283, No Iterations 2
GAMG:  Solving for p, Initial residual = 0.685402, Final residual = 0.0158037, No Iterations 2
time step continuity errors : sum local = 0.000249623, global = -3.31512e-005, cumulative = 0.00011917
GAMG:  Solving for p, Initial residual = 0.44843, Final residual = 8.32125e-009, No Iterations 29
time step continuity errors : sum local = 6.59205e-011, global = -1.86083e-011, cumulative = 0.00011917
ExecutionTime = 93.095 s  ClockTime = 93 s

Courant Number mean: 0.114484 max: 22.49
Time = 2e-005

PIMPLE: iteration 1
smoothSolver:  Solving for Urelx, Initial residual = 0.0317763, Final residual = 0.00129583, No Iterations 2
smoothSolver:  Solving for Urely, Initial residual = 0.0257825, Final residual = 0.00123121, No Iterations 2
smoothSolver:  Solving for Urelz, Initial residual = 0.0257636, Final residual = 0.00121398, No Iterations 2
GAMG:  Solving for p, Initial residual = 0.206075, Final residual = 0.00911697, No Iterations 2
time step continuity errors : sum local = 9.355e-005, global = 1.57552e-005, cumulative = 0.000134925
GAMG:  Solving for p, Initial residual = 0.141025, Final residual = 6.86812e-009, No Iterations 31
time step continuity errors : sum local = 6.23933e-011, global = 1.76006e-011, cumulative = 0.000134925
ExecutionTime = 157.508 s  ClockTime = 157 s

Courant Number mean: 0.114674 max: 44.8948
Time = 2.5e-005

PIMPLE: iteration 1
smoothSolver:  Solving for Urelx, Initial residual = 0.0215758, Final residual = 0.00129572, No Iterations 2
smoothSolver:  Solving for Urely, Initial residual = 0.0174309, Final residual = 0.00128903, No Iterations 2
smoothSolver:  Solving for Urelz, Initial residual = 0.0175009, Final residual = 0.00129434, No Iterations 2
GAMG:  Solving for p, Initial residual = 0.264416, Final residual = 0.00736207, No Iterations 3
time step continuity errors : sum local = 5.0336e-005, global = -5.26303e-006, cumulative = 0.000129662
GAMG:  Solving for p, Initial residual = 0.279803, Final residual = 4.17876e-009, No Iterations 31
time step continuity errors : sum local = 2.87213e-011, global = -8.10944e-012, cumulative = 0.000129662
ExecutionTime = 177.742 s  ClockTime = 177 s

Courant Number mean: 0.114891 max: 308.443
Time = 3e-005

PIMPLE: iteration 1
smoothSolver:  Solving for Urelx, Initial residual = 0.0211509, Final residual = 0.000875977, No Iterations 4
smoothSolver:  Solving for Urely, Initial residual = 0.0202869, Final residual = 0.000901536, No Iterations 4
smoothSolver:  Solving for Urelz, Initial residual = 0.0225594, Final residual = 0.00107766, No Iterations 4
GAMG:  Solving for p, Initial residual = 0.396771, Final residual = 0.0149064, No Iterations 3
time step continuity errors : sum local = 0.000129075, global = 2.29825e-006, cumulative = 0.00013196
GAMG:  Solving for p, Initial residual = 0.522485, Final residual = 4.09076e-009, No Iterations 33
time step continuity errors : sum local = 4.17708e-011, global = 1.18071e-011, cumulative = 0.00013196
ExecutionTime = 239.778 s  ClockTime = 239 s

Courant Number mean: 0.116034 max: 1059.57
Time = 3.5e-005

PIMPLE: iteration 1
smoothSolver:  Solving for Urelx, Initial residual = 0.0594116, Final residual = 0.00524483, No Iterations 4
smoothSolver:  Solving for Urely, Initial residual = 0.0595924, Final residual = 0.00385022, No Iterations 6
smoothSolver:  Solving for Urelz, Initial residual = 0.0786362, Final residual = 0.00630619, No Iterations 6
GAMG:  Solving for p, Initial residual = 0.678854, Final residual = 0.0287991, No Iterations 3
time step continuity errors : sum local = 0.000422026, global = -1.63291e-006, cumulative = 0.000130327
GAMG:  Solving for p, Initial residual = 0.676814, Final residual = 8.72489e-009, No Iterations 28
time step continuity errors : sum local = 2.19735e-010, global = -6.7754e-011, cumulative = 0.000130327
ExecutionTime = 254.929 s  ClockTime = 255 s

Courant Number mean: 0.120919 max: 4742.82
Time = 4e-005

PIMPLE: iteration 1
smoothSolver:  Solving for Urelx, Initial residual = 0.262302, Final residual = 0.0131629, No Iterations 6
smoothSolver:  Solving for Urely, Initial residual = 0.331618, Final residual = 0.0246462, No Iterations 10
smoothSolver:  Solving for Urelz, Initial residual = 0.41908, Final residual = 0.0382986, No Iterations 10
GAMG:  Solving for p, Initial residual = 0.806774, Final residual = 0.020975, No Iterations 4
time step continuity errors : sum local = 0.00081054, global = -2.52914e-006, cumulative = 0.000127798
GAMG:  Solving for p, Initial residual = 0.555316, Final residual = 8.76215e-009, No Iterations 30
time step continuity errors : sum local = 9.50579e-010, global = -2.7071e-010, cumulative = 0.000127798
ExecutionTime = 312.559 s  ClockTime = 312 s

Courant Number mean: 0.152711 max: 18586.1
Time = 4.5e-005

The results are also plain wrong. The mesh seems OK, though...

Any comments?

Thanks!

[pablo_fil]

Wow, that mesh looks really cool.

I'm not expert in SRF and that stuff in OF, but I think you should take a look at this:
https://github.com/OpenFOAM/OpenFOAM...ontrolDict#L48

In LINE 48 you can see the adjustTimeStep parameter and maxCo, which you should keep below a reasonable value (<1?)

Maybe it helps

[KateEisenhower]

Hi Nejc,

My first step would be to get the initial residuals lower. It seems like you have only 1 pimple iteration per timestep.

Keep us updated! Best regards,

Kate

[kandelabr]

Hello Pablo, Kate,

I have turned time step adjustment on and it doesn't really help, it just makes time step 1000 times smaller. I also changed nCorrectors and some other settings so that now residuals are 1e-10 or even less and time step has still dropped 1000 times. I'll just leave it be and see what comes out.

I'll be back with a report

Thanks (for now)

[kandelabr]

Well, it turns out that although residuals are now much lower the solution still doesn't converge. I suspect boundary conditions are wrong.

In plain english, I guess that's my case:
- inlet: uniform velocity in absolute coordinates, pressure with zero gradient
- impeller walls: zero velocity in relative coords, zero gradient pressure
- volute walls: zero velocity in absolute coordinates, zero gradient pressure
- outlet: ??? (zero gradient? free stream? rotating <something something>)?

I'm obviously having trouble with outlet BCs. If there's something on the web (or in books) I could learn from about that, I'd be very happy.

[jherb]

Did you try nOuterCorrectors >> 1 with underrelaxation?

[kandelabr]

Yes I have. I had to check, though. And doing that I found pRefCell and pRefValue in my PIMPLE settings... I copied that from another case and missed that... I guess that was the problem.

It seems to work now, without turbulence modelling. I guess I'll have to further investigate boundary conditionsfor k-epsilon.

Thanks!

[kera]

can you post your checkMesh results? As I see you are using a tetra mesh for your problem. I had the same problem before and switching from tetra to hexa had actually solved my problem. Maybe you can try that aswell.

Regards,
Ricky

[kandelabr]

Hello,

checkMesh says the mesh is OK. I have two options now, either try hexahedral mesh with salome, use snappyHexMesh or do a MRF simulation with modelled volute as well.

This moment i'm busy with another project but I'll report here whatever I discover.

Thanks for now.