ico only running with a crazy small time step
 

I am running a model of flow in a pretty basic channel (2m wide x 4 m in length x 4m tall) and I can only keep the Courant number in check if I run it with a time step of 1e-16.

My block mesh uses a (40 40 40) grid, and going any finer on this means that the computation times takes ages.

With my current time step a simple simulation would take several weeks to run. Does anybody have any further tricks on how to keep the Courant number in check with a reasonable time step?

It seems to be a trade-off between a super fine mesh and a super small time step, both of which mean really long computation times.

Is the max velocity in the domain realistic? You might be having a velocity blow up which is driving down the Co and deltaT down.

thanks for the reply, my max velocity is reasonable, but I'm now wondering if I haven't messed something up in my U file:

dimensions [ 0 1 -1 0 0 0 0 ];

internalField uniform ( 0 0 0 );

boundaryField
{

Wall8 //inlet
{
type fixedValue;
value uniform ( -0.4 0.4 0.1 );
}
Wall6
{
type fixedValue;
value uniform ( 0 0 0 );
}
Wall7
{
type fixedValue;
value uniform ( 0 0 0 );
}
Wall9
{
type fixedValue;
value uniform ( 0 0 0 );
}
Wall10
{
type fixedValue;
value uniform ( 0 0 0 );
}
Wall12
{
type fixedValue;
value uniform ( 0 0 0 );
}
Wall17 //outlet
{
type zeroGradient;
}
Wall20
{
type fixedValue;
value uniform ( 0 0 0 );
}
Wall21
{
type fixedValue;
value uniform ( 0 0 0 );
}
Wall22
{
type fixedValue;
value uniform ( 0 0 0 );
}
Wall27
{
type fixedValue;
value uniform ( 0 0 0 );
}
Wall28
{
type fixedValue;
value uniform ( 0 0 0 );
}
}

I'm having the same problem…

http://www.cfd-online.com/Forums/ope...-cylinder.html

You may be facing a problem with numerical stability. What kind of fvSchemes are you using? A more diffusive/stable convection discretization scheme might help.

If you're looking for a steady-state solution you might be able to get away with running upwind for a little while and switching over to more accurate schemes later on in the run.

Thanks for the suggestion. I will look for examples of the fvScheme that you suggest. I'm currently using the following fvScheme:


ddtSchemes
{
default Euler;
}

gradSchemes
{
default Gauss linear;
grad(p) Gauss linear;
}

divSchemes
{
default none;
div(phi,U) Gauss linear;
}

laplacianSchemes
{
default none;
laplacian(nu,U) Gauss linear corrected;
laplacian((1|A(U)),p) Gauss linear corrected;
}

interpolationSchemes
{
default linear;
interpolate(HbyA) linear;
}

snGradSchemes
{
default corrected;
}

fluxRequired
{
default no;
p ;
}