Comparing the coupling schemes¶
The three coupling schemes of tutorial 2, compared on the same problems at several time steps:
monolithic: the contraction model stepped inside the mechanics Newton iteration;
segregated: the naive staggered scheme, which is not convergent once the active stiffness exceeds the passive stiffness;
stabilized: segregated plus Regazzoni & Quarteroni’s consistent stabilization term.
The scripts¶
| Script | What it does |
|---|---|
element.py | One-element studies with prescribed inputs and no EP: static-caisplit, static-zetasplit and dynamic-zetasplit, each against a monolithic reference at a much smaller time step |
run.py | Runs the slab or BiV under every scheme and time step, one subprocess per run, into <root>/<geometry>/<split>/<scheme>/dt<dt>/. A re-run skips the runs that finished |
report.py | Reads such a folder of runs and writes report.md and figures: the error of each run against the monolithic reference, and the observed orders of convergence |
record.py, metrics.py | The per-step recorder the slab and BiV write steps.csv with, and the error metrics |
Running¶
cd examples/scheme_comparison
python3 element.py --output-dir results
python3 run.py --study slab --root results
python3 report.py --root resultsrun.py --dry-run lists the runs without starting them. The naive scheme is expected
to fail on some runs, and its failures are part of the result.