toughness_anisotropy_jump.py

[pedrolimasi]

Description

• This script uses a predictor-corrector frontAdvancing scheme
• Failure happens during attempt time-step
• Specifically the following section is to blame:

https://github.com/geoEnergyLab-EPFL/pyFrac/blob/pedro/dev/src/time_step/ts_solution.py#L371-L375

# preventing infinite or not effective loops
if (k >= sim_properties.maxFrontItrs and norm > 0.026) or k > 100:
    if norm > 10 * sim_properties.tolFractFront or k > 200:
        exitstatus = 6
        return exitstatus, None

• In particular, I'm inclined to blame the magic number 0.026, as I can see that, at that point:
  • k == 25 == sim_properties.maxFrontItrs and
  • norm == 0.00919...
• This results in the code reaching maxFrontItrs, but not exiting as it should.
• Script breaks for lack of space to store the norm history for this iteration (since it reserved space for 25 front iterations only)

Context

• This example takes 2 hours on debug-mode on gelpc1 to reach the error.

[pedrolimasi]

Closed by

• 67751d0
• 84cc47b