Is this possible, to change the method of acceleration for the coupling after a certain time instant. Should different precice-configure.xml files be created ?
Thank you in advance !
This is currently not possible, at least not automatically.
What you can always do is to run the case till some time with one configuration, and then restart your simulations with another configuration.
Why do you want to do this? What is your use case?
Thank you for replying.
There is no specific problem in my mind for now. But I thought about cases where the solver (structure solver for example) changes (e.g changes scheme) after a certain time instant (because of the nature of the problem), and I thought about changing the acceleration method according to that.
For example, for the elastic tube tutorial, I thought about changing from IQN-ILS to an Aitken acceleration when the structure uses a linear elastic law instead of the quadratic one and see if it will converge (In a pressure pulse scenario).
Maybe something to consider for our roadmap, then. (@uekerman)
We are only now starting to get into the adaptive realm (with @fsimonis currently working on changing the mesh interface at runtime), this could be an interesting next step, given a use case. Given that IQN-ILS does not work in all cases, this could be interesting.
We are currently discussing a kind of similar problem, in which IQN should fall back to a constant underrelaxation for similar robustness reasons: IQN implementation does not handle starting from zero-state · Issue #1496 · precice/precice · GitHub
1 Like
Thank you ! Looking forward to that.
I am not yet too sure whether there are real use case where changing between IQN and Aitken underrelaxation would be necessary. With filtering, IQN should already be adaptive enough.
Only using IQN-ILS here does not converge? Or is it just an experiment for the sake of it? I think I have not yet understood the motivation.
This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.