Model convergency issue

[hjia1005]

I am running a model calibration using pestpp-ies. With the random realizations of parameters, the steady-state takes a long time to converge. As a solution, I split the original model into a steady-state model and the transient model. I set the OUTER_MAXIMUM as 10000 for the steady-state model. In my workflow, after steady state is done, the result head is read as the strt for the transient model which has a much lower outer_maximum (300) and ATS is also on. But for the convenience of post-processing, I included the steady state as the first stress period in the transient model, and the first stress period normally takes less than 1 second to run as planned. This strategy worked for 99% of the model runs until I saw one failure, see below. I set both OUTER_DVCLOSE and INNER_DVCLOSE as 0.001 for the transient model, but in this model, it refused to converge even with a low value of 5e-6. May I ask how to avoid this and make it converge?

[hjia1005]

For most of the other model runs, it's like this:

[hjia1005]

It seems this is the solution, to update the NO_PTC in the ims file. I will give it a try.

BEGIN options
PRINT_OPTION all
COMPLEXITY complex
NO_PTC first
END options

[RyanConway91]

I have found PCT often causes problems with complex models with multiple timesteps/stress periods. Turning it off is the first thing I try with with convergence issues. It certainly can help in some cases, but not all.