[lasherwc]

Hi:

I'm running twoPhaseEulerFoam and I need to use a very small time step to keep the Courant number below 1, and it is taking a very long time to solve. I am not interested in transient behavior, only steady state. Is there a way to change the fvSolutions settings so that I can use a larger time step? For example, I tried setting nOuterCorrectors to 50 (having read about this on the OpenFOAM wiki) but that didn't work.

Thanks,

Bill

[einstein_zee]

Originally Posted by lasherwc

Hi:

I'm running twoPhaseEulerFoam and I need to use a very small time step to keep the Courant number below 1, and it is taking a very long time to solve. I am not interested in transient behavior, only steady state. Is there a way to change the fvSolutions settings so that I can use a larger time step? For example, I tried setting nOuterCorrectors to 50 (having read about this on the OpenFOAM wiki) but that didn't work.

Thanks,

Bill

Hi Bill,

I think there is a trade off between using big time steps (with increasing nOuterCorrectors and at the same time maybe decreasing under relaxation factors) and using small time steps (with decreased number of outer correctors or increased values for urfs.). And this means big time steps requires more time due to increased outer correctors or due to more iterations because of lower urfs. On the other hand you have the same condition when you need to use smaller time steps. What I'm trying to say is that there is no general rule for such questions and it is of course case dependent (based on the physics involved in the problem) which means you need to make trials and errors.

Hope it helps...

[lasherwc]

Thanks so much for responding! So as I understand it I need to play around with nCorrectors, nOuterCorrectors, relaxation factors, convergence criteria, and time step - is that correct? Anything else I should experiment with?

[einstein_zee]

Originally Posted by lasherwc

Thanks so much for responding! So as I understand it I need to play around with nCorrectors, nOuterCorrectors, relaxation factors, convergence criteria, and time step - is that correct? Anything else I should experiment with?

If you are using PIMPLE algorithm, YES. But, we already have some good hints about how to use these algorithms in an efficient manner (you can even find threads here regarding that). And, one more key element is the discretization schemes which has to be selected with care based on your specific needs. If you are not very experienced (like me ) it is good to find some reference people in your field to learn the rules first.

[piu58]

It may be that the total calculation time is decreased with decreasing max Courant number. Hosein explained why: It may be that the solver needs a lot of calculation power for the stabilizing the result for this large timestep. A smaller timestep is faster stabilized and the total computer power is reduced. I often found a maxCo of around 0.1 optimal. If there are large differences in tehe element sizes, even lower.

[lasherwc]

Thanks so much - you've both been very helpful!

[shereez234]

If you dont want to resolve the smallest time scales transientSimpleFoam (if interested search transientSimpleFoam by H Jasak) will enable you to use reallylarge courant numbers.. it is SIMPLE algorithm (no PISO) marching in time.
You can see what has been done and implement this algorithm to any solver of your choice.

[lasherwc]

Thanks, will look into it!