Introduce stop condition for the FluxMatcher iterations?

[avdeevag]

I noticed that sometimes the flux matching solution is significantly better with less number of iterations than we have in default settings (see illustration below). If it is not too difficult maybe some conditions (based on the reached tolerance for total residual?) can be implemented to the FluxMatcher to stop on the optimal number of iterations?

Solution with 300 iterations for the ITER case

versus default 500 iterations

Script to reproduce these cases

chk = FUSE.Checkpoint();

ini, act=FUSE.case_parameters(:ITER, init_from=:scalars, boundary_from=:scalars);

dd = IMAS.dd();
FUSE.init(dd, ini, act);

chk[:init] = dd,ini,act;

R6="sat2_es_mastu_azf+1";

dd_R1,ini_R1,act_R1 = chk[:init];

#----------------------

act_R1.ActorTGLF.user_specified_model = R6

act_R1.ActorCoreTransport.model = :FluxMatcher
#act_R1.ActorFluxMatcher.max_iterations = :300
#act_R1.ActorFluxMatcher.optimizer_algorithm = :simple
display(act_R1.ActorFluxMatcher)

FUSE.ActorPFdesign(dd_R1, act_R1);
FUSE.ActorStationaryPlasma(dd_R1, act_R1);

FUSE.plot_plasma_overview(dd_R1)

[orso82]

@avdeevag could you please try running with act_R1.ActorFluxMatcher.do_plot=true and report how the convergence plot looks?

[avdeevag]

If I do act_R1.ActorFluxMatcher.do_plot=true it does not produce any plots when i run FUSE.ActorStationaryPlasma(dd_R1, act_R1); So, I run FUSE.ActorFluxMatcher(dd_R1, act_R1, do_plot=true) before the SationaryPlasma

and after

[TimSlendebroek]

Looks like we take collect(res.zero),

But like you discvoered maybe when it fails it doesnt' take the best one and uses the last