Accuracy problem

Zato_Ichi · March 22, 2011, 13:21

Hello everyone ! I have a quite strange problem with solving of two equations:

Code:

fvVectorMatrix uEqn
    (
        fvm::ddt(U) ==  (fvc::div(sigma) - fvc::grad(p))/rho
    );

    uEqn.solve() ;
    
    pU = p*U ; 
    
    fvScalarMatrix EEqn
    (
        fvm::ddt(E) == -fvc::div(pU)/rho
    ); 
    EEqn.solve() ;

initial condition for U is uniform (0 0 0) for E uniform 215250, for p i made a step-like distribution with funkySetFields :

Code:

gap
    {
     field p; 
     expression "2e05";
     condition "(pos().x)<=(pi)";
         keepPatches 1; 
    }

Solution domain is one-dimensional with 10000 cells. The problem is that second equation gives that E changes throughout all solution domain with a small fluctuations about e-12, and after each time step this errors accumulates.
Here 1st 100 values of E :

Code:

215250
215250
215250
215249.999999999971
215250
215250
215250
215249.999999999971
215250
215249.999999999971
215250
215250
215250
215249.999999999971
215250.000000000029
215250
215250
215250
215249.999999999971
215249.999999999971
215249.999999999971
215250
215249.999999999971
215250
215250
215250
215250
215250
215249.999999999971
215250.000000000029
215249.999999999971
215250
215250
215249.999999999971
215249.999999999971
215249.999999999971
215249.999999999971
215249.999999999971
215249.999999999971
215249.999999999971
215250
215250
215250
215249.999999999971
215250
215250
215249.999999999971
215250
215249.999999999971
215250
215250
215249.999999999971
215249.999999999971
215250
215250
215249.999999999971
215249.999999999971
215249.999999999971
215249.999999999971
215249.999999999971
215249.999999999971
215249.999999999971
215250
215249.999999999971
215250
215249.999999999971
215249.999999999971
215249.999999999971
215249.999999999971
215249.999999999971
215250
215250
215250
215250
215249.999999999971
215249.999999999971
215250
215250
215250
215250
215249.999999999971
215250
215250
215249.999999999971
215249.999999999971
215250
215250
215250
215250
215249.999999999971
215249.999999999971
215250
215250
215250
215250
215249.999999999971
215250
215250
215249.999999999971
215250
215249.999999999971

I use Gauss linear scheme for div(pU) term. I've tried various linear solvers with difference tolerance values, but it doesn't help.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
UDF compiling problem	Wouter	Fluent UDF and Scheme Programming	6	June 6, 2012 04:43
Fluent accuracy and boundary conditions	Paolo Lampitella	FLUENT	0	June 12, 2008 06:25
Incoherent problem table in hollow-fiber spinning	Gianni	FLUENT	0	April 5, 2008 10:33
problem with using colocated code	Jack	Main CFD Forum	0	December 15, 2002 00:15
GAMBIT meshing problem	Gauthier Lambert	Main CFD Forum	1	August 3, 2000 09:22