[ariaballack]

hi everybody
I am writing a solver which needs two dependent parameters to be solver parallel. Actually c=f(D) and then D must be solved using already solved c. in other words, c and D have to be solved after ther one.but there is something wrong with the term "dimensionedScalar"
here is an abstract of the solver:


fvScalarMatrix cEqn
(
fvm::ddt(c)
+ fvm::div(phi, c)
- fvm::laplacian(Dc, c)
);

cEqn.relax();

cEqn.solve();



// Diffusionskoeffizient D(c,T)

dimensionedScalar g = 1.0/(1.0 + Rho1/Rho2 * (1.0/c - 1.0) );

Info << "g=" << f.value() << " " << f.dimensions() << endl;

Dc = D0 * g * ................

---------------------
in which, g at each time step is a function of c that has been already calculated in previous time step.
how should I loop them?
thanks

[Lieven]

Hi Ariaballack,

You try to convert a volScalarField (I presume) c into a dimensionedScalar g. This is simply not possible. If c is a field, g must be as well. So change dimensionedScalar in

volScalarField g("g",1.0/(1.0 + Rho1/Rho2 * (1.0/c - 1.0) ));

Just two more remarks:
* I don't really know what 'c' exactly represents but it seams to me that in principle 'c' can be 0 locally. If this is the case 1/c might result in a floating point error. But this is quite easy to solve by writing c in the numerator
* From dimensional point of view, the expression for g is kind of strange because you subtract a dimensionless field (1.0) from a dimensioned field (1/c) (I'm pretty sure this will give an error when you run the code).

Cheers,

L