GlobalFluxInterface has a silent, undocumented, default tight coupling

[drewj-usnctech]

I'm working on an internal tight coupling workflow that should have converged, but didn't. After some digging, I found the GlobalFluxInterface sets an undocumented default tight coupler, even if you don't request to converge on keff

armi/armi/physics/neutronics/globalFlux/globalFluxInterface.py

Lines 65 to 77 in 24e36fe

	self._setTightCouplingDefaults()
	def _setTightCouplingDefaults(self):
	"""Enable tight coupling defaults for the interface.
	- allows users to set tightCoupling: true in settings without
	having to specify the specific tightCouplingSettings for this interface.
	- this is splt off from self.__init__ for testing
	"""
	if self.coupler is None and self.cs["tightCoupling"]:
	self.coupler = interfaces.TightCoupler(
	"keff", 1.0e-4, self.cs["tightCouplingMaxNumIters"]
	)

I understand why a default tight coupler for global flux makes sense. In most cases, this is probably what you want to do. But the behavior isn't documented, at least not in the tight coupling docs, which led to my confusion. And it's not included in any logging so the log file wasn't super helpful.

The thing that tipped me off was the addition of the keff convergence table in addition to the field I cared about. So there was something about converging on keff, but I had no idea where it came from

[albeanth]

But the behavior isn't documented, at least not in the tight coupling docs, which led to my confusion

Yeah that's fair. Feel free to make an addition to the docs, if you'd like! I'm happy to review and help with that as needed. Unfortunately, my plate is currently full and I cannot take that on myself.

And it's not included in any logging so the log file wasn't super helpful.
The thing that tipped me off was the addition of the keff convergence table in addition to the field I cared about. So there was something about converging on keff, but I had no idea where it came from

At the conclusion of each tight coupling iteration, there is a tight coupling convergence summary table (as you've seen). The idea was that each interface that is being used for measuring convergence is listed along with the parameter being used to measure convergence of said interface. So my hope was that the interface listed would point a user/dev in that direction. If you have suggestions on how to improve that, I think we'd welcome it!

[john-science]

@drewj-usnctech I agree with Tony about the tight coupling summary table. But... sure, better documentation. That's totally fair.

Do you want to take a crack at that documentation? Or should I?