Geo/12964 u quasistatic p euler #13022
Conversation
|
question out of curiosity, what changed in the settlement test? since now its not failing after this correction of the euler scheme? |
There was a problem hiding this comment.
Thank you for picking this up and making the schemes more understandable and consistent! I don't have any comments or suggestions for the code itself. I do seem to remember that the results for the settlement tests were slightly different before. Does it converge slower now (and is that why the minimum amount of iterations was increased) and if so, why? Also, does the same happen for the python/c++ settlement tests in settlement.py?
| "residual_relative_tolerance": 0.0001, | ||
| "residual_absolute_tolerance": 1E-09, | ||
| "min_iterations": 6, | ||
| "min_iterations": 10, |
There was a problem hiding this comment.
What is the reason we need to increase the min number of iterations?
There was a problem hiding this comment.
Not having the unbalance from RHS terms with velocity gives a quite different RHS. That triggers a difference in convergence behaviour as it is more difficult to iterate a small unbalance away than it is to iterate on a big unbalance. Hence the increase of the maximum number of iterations, to arrive at the same number of steps ( with the existing setting it scaled down the step size once, resulting in the same final displacements reached in an extra step.
No changes from the settlement workflow test ( this uses elastic material throughout i.s.o. a UDSM with ABC material behaviour, which makes it easier to converge. )
…asistatic scheme.
WPK4FEM
force-pushed
the
geo/12964_u_quasistatic_p_euler
branch
from
8b17922 to
56149f9
Compare
📝 Description
A quasistatic scheme should not fill velocity and acceleration but leave those at zero.
Adapted unit tests and integration test.