CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (http://www.cfd-online.com/Forums/openfoam-solving/)
-   -   Relative Tolerance (http://www.cfd-online.com/Forums/openfoam-solving/65033-relative-tolerance.html)

RachelElizabeth June 1, 2009 22:31

Relative Tolerance
 
Hi
I was wondering if someone could help me to understand the details of relTol.

I am unsure which has the stricter convergence criteria relTol=0.1 or relTol=0.001.

I understand that if you set relTol=0 that you force the solution to converge to the solver tolerance, but I am not sure about other values.

Cheers
Rach

hellorishi June 4, 2009 10:56

Hello,

relTol is a parameter to say how much the intial residual should reduce in a particular time step.

e.g. relTol=0, then default tolerence is targetted. (e-6 here)
DICPCG: Solving for p, Initial residual = 1, Final residual = 7.55402e-07, No Iterations 35
...
DICPCG: Solving for p, Initial residual = 0.379232, Final residual = 3.38648e-07, No Iterations 34


e.g. relTol=0.01, then solver only tries to reduce residual by certain factor. This relTol is the order of magnitude reduction. So here Solver tries to reach a value below 1/100 of initial value.
DICPCG: Solving for p, Initial residual = 1, Final residual = 0.00837542, No Iterations 24
...
DICPCG: Solving for p, Initial residual = 0.378888, Final residual = 0.00355735, No Iterations 23

So relTol=0.001 is stricter criteria & requires more iterations. Observe the number of iterations required in the above examples, for the same calculation.


All times are GMT -4. The time now is 00:00.