How to speed up a reactingFoam simulatin?
 

Hey everybody,


I am working on the reactindFoam case. Trying to simulate a flame. I am using a wedge grid with 10k cells. I am simulation with a adjusted time step:


adjustTimeStep yes;
maxCo 0.1;
maxDeltaT 5e-3;


With this settings i got a time step 2.1e-05. . . 2.6e-05. That seems to me a lil bit low and the case needed a view hours to converge. So, I increased maxCo to 0.5 and was hopping to speed up the simulation up. The time step i got now was 1.2e-04. . . 1.3e-04. Setting maxCo to 0.6 gave me a janaf-error. The next thing i was trying to do was to set adjustTimeStep to no and deltaT to a higher number than 2e-04. This gave me a janaf-error once again.


A simulation of the same flame with CFX converge with a time step of 1e-2. That makes the CFX-simulation much more faster than in OpenFOAM.


Do anybody know how it's possible to get a stabel simulation with a higher time step?


Does anybody know what the initialChemicalTimeStep in the constant/chemistryProperties directionary is and how this effects deltaT?


regards, ilja

HI ilja,

i guess you allready know some of my answers, but here we go:

- decompose and run in parallel
- if it runs stabel you can reduce the number of correctors. (system/fvScemes) eg :
PISO
{
nOuterCorrectors 1;
nCorrectors 1;
nNonOrthogonalCorrectors 1;
momentumPredictor off;
}
- if you reduce the number of species for which you solve the equation you could also gain some performance.

the initialChemicalTimeStep is the time step which the chemistry-solver uses.
The flowfield and the chemistry are solved by "different parts" of the solver "reactingFoam" therefore thy dont need neccessarily to have the same timestep. The initialChemicalTimeStep should be of the magnitude of time of chemical reaction-> something like e-7.
(hope this is correct)


regards