[Morlind]

Hi all,
I've been at this for a few months now. I have a number of vehicle models which have either known aero data or reasonable guesses. For some reason I consistently get poor balance numbers (ie. 80% front where it should be 30%) or blown up models.

First I tried the Ahmed model. I struggled with this using both simpleFoam and pisoFoam, all based on the motorbike tutorial. I tried KomegaSST and realizableKE. In the end I managed to get an accurate result by modifying the discretization scheme (according to recommendations from WolfDynamics online tips, huge help!).

So, yay for correlation! Now just slot the car model in... unfortunately not. I have a GT car which is almost entirely sealed (ie. no engine bay complications) and it yields very far front balance even though it has a diffuser. I have an open wheel model which should have more vortices, so presume additional complications, and it is still showing 70-85% front balance and lift only 40-50% of the tunnel numbers. I have tried each with various turbulence and discretization settings but to no avail.


I am not well versed in OpenFoam and all the various methods for manipulation. I used Fluent some years ago and so I've been deciphering things along the way. I am usually successful at problem solving this type of thing but I can't seem to make any appreciable change to the result.


Can anyone point me in the correct direction here? Is there something glaring that either I have missed or which is known to be inaccurate coming from the tutorials?

Here is an example of what my open wheel model looks like using pisoFoam and LES turbulence model. It is obvious that the flow is overly turbulent and also not attached where it should be. I also don't think this is the best model for this work. I beleive RKE is better but I consistently blew up the solver when trying that even with conservative discretization settings. For the record my mesh quality is not terrible. Highest skewness is in the 5 range, all else is good. Peak non-orthogonality is around 64.


The short question is this : Am I using the correct combination of solver and turbulence model? If so, is it an exercise in tuning the schemes? If not, which should I be using?



edit - of course I am happy to post whatever files are relevant to the issue. I have tried so many types of model that I simply don't know which to start with.

[Morlind]

No one has a suggestion?


fvSolution looks like this:
/*--------------------------------*- C++ -*----------------------------------*\
| ========= | |
| \\ / F ield | OpenFOAM: The Open Source CFD Toolbox |
| \\ / O peration | Version: 5 |
| \\ / A nd | Web: www.OpenFOAM.org |
| \\/ M anipulation | |
\*---------------------------------------------------------------------------*/
FoamFile
{
version 2.0;
format ascii;
class dictionary;
location "system";
object fvSolution;
}
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //

solvers
{
p
{
solver GAMG;
tolerance 1e-6;
relTol 0.05;

smoother GaussSeidel;


nCellsInCoarsestLevel 50;
}

pFinal
{
$p;
relTol 0;
}

"(U|k|B|nuTilda)"
{
solver smoothSolver;
smoother symGaussSeidel;
tolerance 1e-7;
relTol 0;
}
}

potentialFlow
{
nNonOrthogonalCorrectors 10;
}


PISO
{
nCorrectors 3;
nNonOrthogonalCorrectors 1;
}

relaxationFactors
{
"U.*" 1;
"nuTilda.*" 1;
}


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




fvSchemes:
/*--------------------------------*- C++ -*----------------------------------*\
| ========= | |
| \\ / F ield | OpenFOAM: The Open Source CFD Toolbox |
| \\ / O peration | Version: 5 |
| \\ / A nd | Web: www.OpenFOAM.org |
| \\/ M anipulation | |
\*---------------------------------------------------------------------------*/
FoamFile
{
version 2.0;
format ascii;
class dictionary;
location "system";
object fvSchemes;
}
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //

ddtSchemes
{
default backward;
}

d2dt2Schemes
{
}

gradSchemes
{
default Gauss linear;

grad(nuTilda) cellLimited Gauss linear 0.5;
grad(U) cellLimited Gauss linear 0.5;
}

divSchemes
{
default none;

div(phi,U) Gauss linearUpwind Grad(U);
div(phi,k) Gauss linearUpwind;
div(phi,nuTilda) Gauss linearUpwind;

div((nuEff*dev2(T(grad(U))))) Gauss linear;
}

laplacianSchemes
{
default Gauss linear limited 0.77;
}

interpolationSchemes
{
default linear;
}

snGradSchemes
{
default limited 0.77;
}

wallDist
{
method meshWave;
}