73 PIMPLE
 74 {
 75     momentumPredictor yes;
 76     nOuterCorrectors 15;
 77     nCorrectors     4;
 78     nNonOrthogonalCorrectors 1;
 79     rhoMin          0.1;
 80     rhoMax          5.0;
 81     pMinFactor      0.5;
 82     pMaxFactor      2.0;
 83 
 84 }

空间和时间离散格式如下

 18 ddtSchemes
 19 {
 20     default         backward;
 21 }
 22 
 23 gradSchemes
 24 {
 25     default         Gauss linear;
 26     limited         cellLimited Gauss linear 1;
 27     grad(U)         $limited;
 28     grad(k)         $limited;
 29     grad(kU)        $limited;
 30     grad(omega)     $limited;
 31     grad(omegaU)    $limited;
 32 }

 33
 34 divSchemes
 35 {
 36     default         none;
 37     div(phi,U)      Gauss linearUpwind limited;
 38 
 39     turbulence      Gauss linearUpwind limited;
 40     energy          Gauss linearUpwind limited;
 41 
 42     div(phi,k)      $turbulence;  
 43     div(phi,kU)     $turbulence;
 44     div(phi,omega)  $turbulence;
 45     div(phi,omegaU) $turbulence;
 46 
 47     div(phi,e)      $energy;
 48     div(phi,h)      $energy;
 49     div(phi,K)      $energy;
 50     div(phi,Ekp)    $energy;
 51 
 52     div(phiv,p)     Gauss upwind;
 53     div((phi|interpolate(rho)),p)   Gauss upwind;
 54 
 55     div(phi,ReThetat)  $turbulence;
 56     div(phi,gammaInt)  $turbulence;
 57 
 58     div(B)          Gauss linear;
 59     div(meshPhi,p)  Gauss linear;
 60 
 61     div((nuEff*dev2(T(grad(U))))) Gauss linear;
 62     div(((rho*nuEff)*dev2(T(grad(U))))) Gauss linear;
 63     div(nonlinearStress) Gauss linear;
 64 
}

网格是blockMesh做的，比较简单，因为要保证边界层密度所以纵横比比较大。请问导致发散的主要原因是什么？什么方法可以改善？我个人不希望再降低时间步长，因为目前已经很小了，采用pimple算法也是由于它能够在较大的库郎数下计算。谢谢各位！～