Adaptive time stepping / getting residual from failed nonlinear solve

Question

I'm ultimately trying to implement adaptive time stepping into model I've created and I was wondering if

1) there is a standard example for adaptive timestepping with fenics? (I've looked but can't find one) or
2) is there a way to have a non linear solver return a residual instead of raising a non-convergence error?

any suggestions are greatly appreciated.

Also I'm using the same type of solver as the example problem in the link below, but open to using an arguably "better one"

https://fenicsproject.org/documentation/dolfin/dev/python/demo/documented/cahn-hilliard/python/documentation.html

Answer 1

1) Not as far as I'm aware. This depends on how you've discretised in time. If you're implementing a "finite differencing" Euler-type theta scheme, there is a crude description of a method here.

2) try solver.parameters["error_on_nonconvergence"] = False

Comments

The solver.parameters["error_on_nonconvergence"] = False is exactly what I was looking for!!!

I'm familiar with the theory of how adaptive timestepping works, I was hoping a for an example to see if there were any non-obvious intricacies with performing it.