SimScale CAE Forum

Restarting simulation if convergence isn't reached


#1

Hi All,

I am curious whether it is possible to implement the ability to continue running a simulation that reaches the end time before convergence. For example you are running a steady CFD case and you set the end time to 10000 and the solution reaches that time without hitting the convergence criteria, currently, you have to delete the run and increase the end time and re-run the full simulation. From some experience using OpenFOAM I know this is easily performed with OpenFOAM, so it would seem to me that it shouldn’t be too hard to setup in Simscale. Thanks for your time.

-Ian


#2

@steezorigineez Ian,

very much agreed! To use an existing solution field as the “starting point” for a new simulation run is something we’re working on. However it’s too early to provide a release date on it but I’ll keep you posted on the development of this feature. In the meantime a brief hint (maybe you’re already using this): For most flow simulation analysis types, under “Simulation Control” you can set the option that the simulation shall be initialized with a potential flow solution (see screenshot below).

It’s obviously not a replacement for a real restart feature but can help sometimes to speed up convergence.

Best,

David


#3

Hi @dheiny,

I also wanted to ask this question but I see that it already exists. So is this feature already available? This is not only needed when you run out of time but also if set to little time steps to reach solution. Unfortunately potentialFoam is not a solution here.

Best regards,
Andrzej


#4

Hi Andrzej (@akosior ),

I have the feeling I do say this too often lately to you :slight_smile: - but: not yet. It’s in the works but no release date yet.

As another hint in the meantime, if you’re dealing with steady-state simulations: The Numerics tree item let’s you set convergence criteria / residual tolerances for all solution fields:

That way you can bump up the number of iterations but set a criterion at which it should terminate the simulation / consider it converged. This approach could help at least in some cases.

Best,

David


Correlation issues