关于nOuterCorrectors影响计算速度和计算精度的问题

wuyukai

计算一个简单二维多圆柱绕流的动网格算例，网格数量大概40-50万，fvsolution中的nOuterCorrectors和nCorrecters我尝试了不同的值，发现随着该值设置的越来越大，结果才会收敛，最终分别等于20和10。但是这个条件下，计算时间特别久，一个二维40-50万网格的算例要计算将近90个小时，感觉很不合理。想请教一下各位大佬，我是不是哪里设置出错了？

/*--------------------------------*- C++ -*----------------------------------*\
| =========                 |                                                 |
| \\      /  F ield         | OpenFOAM: The Open Source CFD Toolbox           |
|  \\    /   O peration     | Version:  v2006                                 |
|   \\  /    A nd           | Website:  www.openfoam.com                      |
|    \\/     M anipulation  |                                                 |
\*---------------------------------------------------------------------------*/
FoamFile
{
    version     2.0;
    format      ascii;
    class       dictionary;
    location    "system";
    object      fvSolution;
}
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //

solvers
{
    "(p|pcorr)"
    {
        solver          PCG;
        preconditioner        DIC;
        tolerance       1e-04;
        relTol          0.1;
    }
    "(p|pcorr)Final"
    {
        $p;
        relTol          0;
    }
    U
    {
        solver          smoothSolver;
        smoother        symGaussSeidel;
        tolerance       1e-04;
        relTol          0.1;
    }
    UFinal
    {
        $U;
        tolerance       1e-04;
        relTol          0.0;
    }
}


PIMPLE
{
    momentumPredictor   no;
    correctPhi          yes;
    nOuterCorrectors    20;
    nCorrectors         10;
    nNonOrthogonalCorrectors 1;
    ddtCorr             true;

    pRefPoint           (20.0 0.0 0.5);
    pRefValue           0.0;
}

relaxationFactors
{
   fields
   {
   }
   equations
   {
       "U.*"             1;
   }
}
// ************************************************************************* //

/*--------------------------------*- C++ -*----------------------------------*\
| =========                 |                                                 |
| \\      /  F ield         | OpenFOAM: The Open Source CFD Toolbox           |
|  \\    /   O peration     | Version:  v2006                                 |
|   \\  /    A nd           | Website:  www.openfoam.com                      |
|    \\/     M anipulation  |                                                 |
\*---------------------------------------------------------------------------*/
FoamFile
{
    version     2.0;
    format      ascii;
    class       dictionary;
    location    "system";
    object      fvSchemes;
}
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //

ddtSchemes
{
    default         Euler;
}

gradSchemes
{
    default         Gauss linear;
}

divSchemes
{
    default         none;
    div(U)          Gauss linear;
    div(phi,U)      Gauss linearUpwind grad(U);
    div(phi,k)      Gauss linearUpwind grad(U);
    div(phi,K)      Gauss linearUpwind grad(U);
    div(phi,omega)  Gauss linearUpwind grad(U);
    div((nuEff*dev2(T(grad(U))))) Gauss linear;
}

laplacianSchemes
{
    default         Gauss linear corrected;
}

interpolationSchemes
{
    default         linear;
}

snGradSchemes
{
    default         corrected;
}

wallDist
{
    method meshWave;
}


// ************************************************************************* //

波流力

算的什么多圆柱？有多少个圆柱？图形摆出来看看

cccrrryyy

nCorrectors太大了，推荐1-3，最多我见过用到3的。PISO算法本来只有2步压力修正，所以这个值你可以固定，不放心的话用3好了。

nOuterCorrectors你需要仔细的测试一下，这个和时间步大小也有关系。大致范围5-100左右吧。FLUENT那边的推荐值是20-25，我自己不管是在FLUENT还是FOAM测试出来的结果也都差不多，最多会用到30，一般情况下20左右。如果很多外迭代还无法收敛那你需要考虑缩小时间步长，而不是去无限制的去增大外迭代步数。另外你没有给relaxationFactors，为什么？

收敛不收敛是你自己设置的，不是说收敛了就对，没收敛就错。最终要靠结果去判断，特别是和实验结果的对比。

wuyukai

@cccrrryyy 感谢感谢，我在正式计算之前，算了单个圆柱的，和Oseen公式推出来的理论解能够比上。
您说的方法我试试，由于OpenFoam我也只是初学，没有特别系统的学过，您说的relaxationFactors 我还不太清楚，我去好好了解一下。

wuyukai

@波流力网格差不多长这样

Joann

参考DTCHullMoving算例中nOuterCorrectors为3，每个时间步里迭代3次，获得3个模型中心的运动坐标，那么如果要获得时间-运动的曲线，是不是选择最后一次迭代的坐标比较好呢？

李东岳

log文件贴一下

Joann

@李东岳比如这个时间步里，PIMPLE迭代了3次，分别有三个旋转中心，如果一个时间步就取一个值，是不是迭代最后一次的旋转中心较好？

Courant Number mean: 0.00758176 max: 1.02562
Interface Courant Number mean: 8.04602e-05 max: 0.732939
Time = 8.89

PIMPLE: Iteration 1
forces forces:
    Not including porosity effects
Rigid-body motion of the hull
    Centre of rotation: (3 0 0.189302)
    Orientation: (0.999758 0 -0.0219868 0 1 0 0.0219868 0 0.999758)
    Linear velocity: (0 0 -0.0427599)
    Angular velocity: (0 -0.0134282 0)
GAMG:  Solving for pcorr, Initial residual = 1, Final residual = 0.000933377, No Iterations 20
time step continuity errors : sum local = 1.87363e-10, global = 7.57308e-13, cumulative = -3.15194e-09
smoothSolver:  Solving for alpha.water, Initial residual = 2.27988e-05, Final residual = 6.44668e-12, No Iterations 3
Phase-1 volume fraction = 0.803336  Min(alpha.water) = -3.41348e-13  Max(alpha.water) = 1
Applying the previous iteration compression flux
MULES: Correcting alpha.water
MULES: Correcting alpha.water
MULES: Correcting alpha.water
MULES: Correcting alpha.water
Phase-1 volume fraction = 0.803336  Min(alpha.water) = -9.55996e-09  Max(alpha.water) = 1
GAMG:  Solving for p_rgh, Initial residual = 0.00721682, Final residual = 4.55154e-08, No Iterations 17
time step continuity errors : sum local = 1.0765e-09, global = -2.39228e-12, cumulative = -3.15433e-09
PIMPLE: Iteration 2
forces forces:
    Not including porosity effects
Rigid-body motion of the hull
    Centre of rotation: (3 0 0.189302)
    Orientation: (0.999758 0 -0.0219868 0 1 0 0.0219868 0 0.999758)
    Linear velocity: (0 0 -0.0427459)
    Angular velocity: (0 -0.0134156 0)
GAMG:  Solving for pcorr, Initial residual = 1, Final residual = 0.000949765, No Iterations 19
time step continuity errors : sum local = 1.34488e-10, global = -1.2392e-13, cumulative = -3.15445e-09
smoothSolver:  Solving for alpha.water, Initial residual = 1.58142e-05, Final residual = 6.39378e-12, No Iterations 3
Phase-1 volume fraction = 0.803336  Min(alpha.water) = -3.4121e-13  Max(alpha.water) = 1
Applying the previous iteration compression flux
MULES: Correcting alpha.water
MULES: Correcting alpha.water
MULES: Correcting alpha.water
MULES: Correcting alpha.water
Phase-1 volume fraction = 0.803336  Min(alpha.water) = -8.31726e-10  Max(alpha.water) = 1
GAMG:  Solving for p_rgh, Initial residual = 0.000213947, Final residual = 4.56853e-08, No Iterations 19
time step continuity errors : sum local = 1.07998e-09, global = -2.92566e-12, cumulative = -3.15738e-09
PIMPLE: Iteration 3
forces forces:
    Not including porosity effects
Rigid-body motion of the hull
    Centre of rotation: (3 0 0.189302)
    Orientation: (0.999758 0 -0.0219868 0 1 0 0.0219868 0 0.999758)
    Linear velocity: (0 0 -0.0427505)
    Angular velocity: (0 -0.0134234 0)
GAMG:  Solving for pcorr, Initial residual = 1, Final residual = 0.000923356, No Iterations 22
time step continuity errors : sum local = 1.36977e-10, global = -1.34327e-13, cumulative = -3.15751e-09
smoothSolver:  Solving for alpha.water, Initial residual = 1.57985e-05, Final residual = 6.20267e-12, No Iterations 3
Phase-1 volume fraction = 0.803336  Min(alpha.water) = -3.412e-13  Max(alpha.water) = 1
Applying the previous iteration compression flux
MULES: Correcting alpha.water
MULES: Correcting alpha.water
MULES: Correcting alpha.water
MULES: Correcting alpha.water
Phase-1 volume fraction = 0.803336  Min(alpha.water) = -5.01932e-08  Max(alpha.water) = 1
GAMG:  Solving for p_rgh, Initial residual = 0.000101498, Final residual = 4.62009e-08, No Iterations 19
time step continuity errors : sum local = 1.09272e-09, global = 4.09503e-12, cumulative = -3.15342e-09
smoothSolver:  Solving for omega, Initial residual =2.53013e-05, Final residual = 7.65699e-09, No Iterations 2
smoothSolver:  Solving for k, Initial residual = 5.87558e-05, Final residual = 2.49652e-08, No Iterations 3
ExecutionTime = 27536.4 s  ClockTime = 27737 s

李东岳

多贴几个时间步，完整的，我再看下。

咦，看错了， @wuyukai 你的log文件贴下

luofq-sysu

@李东岳李老师，借楼问一下：滑移网格计算，每三个时间步会出现pcorr iteration 1000，可能出现的原因在哪方面？网格checkMesh无问题，可能会是GAMGPCG这些格式吗？

李东岳

@luofq-sysu 对，你换一下PCG，很有可能会改善

luofq-sysu

@李东岳谢谢李老师。1. solver.pcorr改成PCG，虽然还是出现iteration 1000，但确实快了很多。2.尝试将pcorr的tolerance从 1e-06改成1e-05，还是会出现 iteration 1000。这个有没有可能跟网格有关呢？整体计算速度慢的原因还有P_rgh也挺费事的

李东岳

@luofq-sysu

这个有没有可能跟网格有关呢

有关。

整体计算速度慢的原因还有P_rgh也挺费事的

网格多少。网格多的话就是那么慢咯

zhanghan

fvSolution中修改每一项的tolerance，我的case是bubbleColumnEvaporatingReacting，自带的tolerance是1e-20，我修改为1e-10后计算的非常快，而且计算结果几乎相同。

FoamFile
{
    version     2.0;
    format      ascii;
    class       dictionary;
    location    "system";
    object      fvSolution;
}
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //

solvers
{
    "alpha.*"
    {
        nAlphaCorr      1;
        nAlphaSubCycles 3;
    }

    p_rgh
    {
        solver          GAMG;
        smoother        DIC;
        tolerance       1e-10;
        relTol          0.001;
    }

    p_rghFinal
    {
        $p_rgh;
        relTol          0;
    }

    "U.*"
    {
        solver          smoothSolver;
        smoother        symGaussSeidel;
        tolerance       1e-10;
        relTol          0;
        minIter         1;
    }

    "(e|h).*"
    {
        solver          smoothSolver;
        smoother        symGaussSeidel;
        tolerance       1e-10;
        relTol          0;
        minIter         1;
    }

    "(k|epsilon|Theta).*"
    {
        solver          smoothSolver;
        smoother        symGaussSeidel;
        tolerance       1e-10;
        relTol          0;
        minIter         1;
    }

    Yi
    {
        solver          smoothSolver;
        smoother        symGaussSeidel;
        tolerance       1e-10;
        relTol          0;
        minIter         1;
        residualAlpha   1e-8;
    }
}

PIMPLE
{
    nOuterCorrectors 1;
    nCorrectors      1;
    nNonOrthogonalCorrectors 0;
}

relaxationFactors
{
    equations
    {
        ".*"            0.3;
    }
}


// ************************************************************************* //