`SolutionStrategy.after_nonlinear_failure` saves time step data causing weird visualization

[vlipovac]

The base solution strategy at line 477 calls self.save_data_time_step() after a time step failes to converge.
If the time manager then goes back in time and reduces dt, and converges, it results in a weird visualization where the simulation jumps back and forth in time.

weird.mp4

To me this is unexpected behavior, useful only for debugging maybe.
Should this be removed? Or at least made optional?

[jwboth]

I do not think this is intended. I also realized a similar thing. The exporting counter is increased even though the time step has not passed convergence. In my view, this is a bug and can be declared as such.

[pschultzendorff]

Just my two cents: save_data_time_step also calls solver_statistics.save in addition to exporting data. Removing self.save_data_time_step() from after_nonlinear_failure would eliminate any trace of failed time steps. Then one could no longer assess some common solver metrics such as total number of nonlinear iterations.

[vlipovac]

Yes, that is important. The data extracted by solver_statistics could use a facelift in terms of flexibility how data is saved and stored, and when.

[jwboth]

I agree (having implemented some small fixes myself due to some limitations). Since several people make use of the solver statistics, I suggest a common effort. It would start with a consolidation of ideas and then the upgrade of the code. Get Outlook for Android<https://aka.ms/AAb9ysg>

…

________________________________ From: Veljko Lipovac ***@***.***> Sent: Monday, May 19, 2025 10:32:47 AM To: pmgbergen/porepy ***@***.***> Cc: Jakub Wiktor Both ***@***.***>; Comment ***@***.***> Subject: Re: [pmgbergen/porepy] `SolutionStrategy.after_nonlinear_failure` saves time step data causing weird visualization (Issue #1421) [https://avatars.githubusercontent.com/u/68282540?s=20&v=4]vlipovac left a comment (pmgbergen/porepy#1421)<#1421 (comment)> Yes, that is important. The data extracted by solver_statistics could use a facelift in terms of flexibility how data is saved and stored, and when — Reply to this email directly, view it on GitHub<#1421 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ARR52V5KXBFN2QH7JJ3VKYD27GJK7AVCNFSM6AAAAAB4GXF4AGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDQOJQGEYTQNRYHA>. You are receiving this because you commented.Message ID: ***@***.***> [External email] Make sure you recognize the sender's email address before you click links, open attachments, or get involved in financial transactions. Contact IT-support BRITA if you have any questions.