Added ComponentsConvergence #31039

joshuahansel · 2025-07-15T13:24:30Z

Added class ComponentsConvergence that checks convergence of all components
Added new method Component::getNonlinearConvergence() to return that Component's convergence, if any
Have not actually overridden the above method for any Component yet, just introducing concept for review

joshuahansel · 2025-07-15T13:39:51Z

@lindsayad @GiudGiud @licharlot For anyone interested, I'm putting up this PR for the concept to be reviewed. This is how I'm planning to handle convergence for Components.

I'm still working on the pieces I plan to use for the individual convergence checks. Here's generally what I'm thinking might be "good" criteria:

Max iterate change of each variable (say, {p, T, vel}) in that Component's block(s) less than some tolerance (provided by parameters like max_change_T)
Require a loose absolute residual check to mitigate false convergence due to possible convergence stalling (in which case condition 1 would report convergence). Do our best to normalize the residual to a standard error. For example, do some kind of nondimensionalization, and correct any domain size or mesh size biases. For example, for drhoA/dt, normalize by rho_ref A (and then mesh size bias).

GiudGiud · 2025-07-15T14:59:09Z

i like the idea. I think this mirrors the component-scaling that was done in SAM but without modifying the nonlinear solve performance

moosebuild · 2025-07-15T15:47:58Z

Job Documentation, step Docs: sync website on 75bfb3d wanted to post the following:

View the site here

This comment will be updated on new commits.

moosebuild · 2025-07-15T17:36:32Z

Job Coverage, step Generate coverage on 75bfb3d wanted to post the following:

Framework coverage

	bb0a08	#31039 75bfb3
	Total	Total	+/-	New
Rate	86.02%	86.02%	-0.00%	100.00%
Hits	124559	124560	+1	5
Misses	20244	20245	+1	0

Diff coverage report

Full coverage report

Modules coverage

Thermal hydraulics

	bb0a08	#31039 75bfb3
	Total	Total	+/-	New
Rate	88.58%	88.73%	+0.15%	100.00%
Hits	14730	15025	+295	308
Misses	1899	1908	+9	0

Diff coverage report

Full coverage report

Full coverage reports

Reports

This comment will be updated on new commits.

lindsayad

Looks sensical to me

GiudGiud

Not a formal review, just an early curious look with some suggestions. I can see how great this is going to be now! This should bring the long-sought robustness to THM. or at least help

GiudGiud · 2025-10-13T10:00:56Z

framework/src/problems/FEProblemBase.C

 }

+bool
+FEProblemBase::hasMaterial(const std::string & name, const THREAD_ID tid)


Suggested change

FEProblemBase::hasMaterial(const std::string & name, const THREAD_ID tid)

FEProblemBase::hasActiveMaterial(const std::string & name, const THREAD_ID tid)

considering the implementation below

GiudGiud · 2025-10-13T10:02:05Z

modules/thermal_hydraulics/doc/content/source/convergence/ComponentsConvergence.md

+
+This [IterationCountConvergence.md] can be used with the [Components system](Components/index.md) to check the convergence of all Components that return a Convergence object with `getNonlinearConvergence()`. The inner convergence check returns its convergence status as follows:
+
+- `CONVERGED`: All Components returning a Convergence object are `CONVERGED`


Suggested change

- `CONVERGED`: All Components returning a Convergence object are `CONVERGED`

- `CONVERGED`: All Convergence objects created by Components are `CONVERGED`

same for the other?

modules/thermal_hydraulics/doc/content/source/convergence/FlowChannel1PhaseConvergence.md

GiudGiud · 2025-10-13T10:04:35Z

...les/thermal_hydraulics/doc/content/source/functormaterials/FlowModel1PhaseFunctorMaterial.md

@@ -0,0 +1,14 @@
+# (AD)FlowModel1PhaseFunctorMaterial
+
+This [functor material](/FunctorMaterials/index.md) computes the following quantities


Suggested change

This [functor material](/FunctorMaterials/index.md) computes the following quantities

This [functor material](/FunctorMaterials/index.md) creates functor material properties for the following quantities

can you add something like "the names of the functor material properties" are specified by ...

I guess you need to use a listing of THM.h to get the names

I'll add the names of the properties manually

GiudGiud · 2025-10-13T10:05:32Z

modules/thermal_hydraulics/doc/content/source/postprocessors/THMNormalizedVariableStep.md

+This post-processor normalizes [ElementExtremeFunctorValue.md] by a provided
+constant, for use by [FlowChannel1PhaseConvergence.md].


the name of that PP does not really convey this function. Could be renamed imo

I struggled with making it a reasonable length, but I'm open to suggestions. I decided to name it based on concept

why not just add a scaling factor for ElementExtremeFunctorValue ?

I'll do that

GiudGiud · 2025-10-13T11:20:05Z

modules/thermal_hydraulics/src/base/FlowModelSinglePhase.C

+  const std::vector<std::pair<std::string, Real>> var_norm_pairs{
+      {THM::PRESSURE, _p_ref}, {THM::TEMPERATURE, _T_ref}, {THM::VELOCITY, _vel_ref}};
+  for (const auto & var_norm_pair : var_norm_pairs)
+    addTHMNormalizedVariableStep(var_norm_pair.first, var_norm_pair.second);


this does not seem 1 phase or THM specific? Maybe it should be in a ConvergneceUtils

The step PP itself you mean? Not sure what you're suggesting with ConvergenceUtils. I had considered modifying ElementExtremeFunctorValue but didn't want to "pollute" it. I also considered just using an existing PP to scale, but didn't want the 2nd PP existing (it'd probably be ok though).

GiudGiud · 2025-10-13T11:22:43Z

modules/thermal_hydraulics/src/convergence/ComponentsConvergence.C

+      return Convergence::MooseConvergenceStatus::DIVERGED;
+    else if (status == Convergence::MooseConvergenceStatus::ITERATING)
+      all_converged = false;
+  }


maybe check else == converged

so we are future proof if a 4th option is created for MooseConvergenceStatus::

an assert should be enough to do that

I feel pretty good about keeping these 3 options into eternity - to me it feels like a basic design of the system

even basic design can change

eternity sounds like a very long time. and who knows where AI-driven completely automated development will take moose..

GiudGiud · 2025-10-13T11:23:47Z

modules/thermal_hydraulics/src/functormaterials/FlowModel1PhaseFunctorMaterial.C

+  addPressureFunctorProperty<true>();
+  addPressureFunctorProperty<false>();


do you need both? what for?

I need to take better notes... I'll put some comments in the code after I refresh my memory. But basically error messages would come up regarding mismatch of AD and non-AD.

ok let s document who s requesting this stuff in both flavors. that might be a hidden issue

Looks like only the non-AD is needed, so I'll just add that for now. I'm guessing that earlier, I was trying to leverage the functor property in some other places.

GiudGiud · 2025-10-13T11:24:20Z

modules/thermal_hydraulics/include/functormaterials/FlowModel1PhaseFunctorMaterial.h

+template <bool is_ad>
+void
+FlowModel1PhaseFunctorMaterial::addPressureFunctorProperty()


I think all these templates should move to the source, as you are not using them from another file

GiudGiud · 2025-10-13T11:28:38Z

modules/thermal_hydraulics/src/postprocessors/Normalized1PhaseResidualNorm.C

+
+  const auto variable = getParam<VariableName>("variable");
+  if (variable == "rhoA")
+    return rho_ref * A_ref * h_min;


is there a reason for h_min rather than L_component?
is the residual proportional to 1/dx for THM solves?

Right, it's targeted at the residual for each nonlinear equation, which is proportional to dx (finite volume).

the residual for each cell. how do we get the residual for the whole equation? Is an operation more like infinite norm or sum of absolute residuals?

The base class computes a discrete norm (l1, l2, or linf) of the element residuals for that variable. I choose linf.

Refs idaholab#31038

Closes idaholab#31038

Refs idaholab#31038

Closes idaholab#31038

moosebuild · 2025-11-03T20:08:37Z

Job Test, step Results summary on 75bfb3d wanted to post the following:

Framework test summary

Compared against bb0a08b in job civet.inl.gov/job/3353620.

Removed tests

Added tests

Run time changes

Modules test summary

Compared against bb0a08b in job civet.inl.gov/job/3353620.

Removed tests

Added tests

Test	Time (s)
`thermal_hydraulics/test:components/flow_channel_1phase.err:components_convergence_scaling`	1.62
`thermal_hydraulics/test:components/flow_channel_1phase.components_convergence`	1.04

Run time changes

Test	Base (s)	Head (s)	+/-
`thermal_hydraulics/test:problems/brayton_cycle.open`	4.23	6.89	+62.77%
`thermal_hydraulics/test:components/inlet_velocity_t_1phase.clg:velocity_t_3eqn`	3.16	4.95	+56.79%
`thermal_hydraulics/test:problems/area_constriction.without_junction`	3.18	4.92	+54.89%
`stochastic_tools/test:web_server_control.stochastic_control/normal`	3.30	5.03	+52.64%
`thermal_hydraulics/test:components/inlet_stagnation_p_t_1phase.phy:p0T0_3eqn`	2.35	3.57	+52.08%

joshuahansel self-assigned this Jul 15, 2025

lindsayad reviewed Jul 15, 2025

View reviewed changes

joshuahansel force-pushed the components-convergence branch from 21be217 to f64f739 Compare August 28, 2025 20:08

joshuahansel force-pushed the components-convergence branch 5 times, most recently from a74a5c2 to 7945b8a Compare October 10, 2025 16:43

GiudGiud reviewed Oct 13, 2025

View reviewed changes

joshuahansel added 12 commits November 3, 2025 12:52

Added ComponentsConvergence

ddff61a

Refs idaholab#31038

Added ADRealToGenericReal()

71dee31

Refs idaholab#31038

Do not trim string in Convergence::verboseOutput()

cfd9914

Closes idaholab#31038

Added end point and h_min methods to DiscreteLineSegmentInterface

cf1c813

Refs idaholab#31038

Added functor material name method to THMNames

8585e12

Refs idaholab#31038

Added FlowModel1PhaseFunctorMaterial

a1e5139

Refs idaholab#31038

Added scale parameter to ElementExtremeFunctorValue

3a4b1b0

Refs idaholab#31038

Added Normalized1PhaseResidualNorm

dc3a4f1

Refs idaholab#31038

Added FlowChannel1PhaseConvergence

f58bda6

Refs idaholab#31038

Added convergence capability to FlowChannel1Phase

69026fa

Refs idaholab#31038

Implemented getNonlinearConvergence for some components

9fa2a2a

Refs idaholab#31038

Added tests for FlowChannel1PhaseConvergence

75bfb3d

Closes idaholab#31038

joshuahansel force-pushed the components-convergence branch from 7945b8a to 75bfb3d Compare November 3, 2025 18:57

	FEProblemBase::hasMaterial(const std::string & name, const THREAD_ID tid)
	FEProblemBase::hasActiveMaterial(const std::string & name, const THREAD_ID tid)


		This [IterationCountConvergence.md] can be used with the [Components system](Components/index.md) to check the convergence of all Components that return a Convergence object with `getNonlinearConvergence()`. The inner convergence check returns its convergence status as follows:

		- `CONVERGED`: All Components returning a Convergence object are `CONVERGED`

	- `CONVERGED`: All Components returning a Convergence object are `CONVERGED`
	- `CONVERGED`: All Convergence objects created by Components are `CONVERGED`

		@@ -0,0 +1,14 @@
		# (AD)FlowModel1PhaseFunctorMaterial

		This [functor material](/FunctorMaterials/index.md) computes the following quantities

		This post-processor normalizes [ElementExtremeFunctorValue.md] by a provided
		constant, for use by [FlowChannel1PhaseConvergence.md].

		addPressureFunctorProperty<true>();
		addPressureFunctorProperty<false>();

Added ComponentsConvergence #31039

Are you sure you want to change the base?

Added ComponentsConvergence #31039

Conversation

joshuahansel commented Jul 15, 2025

Uh oh!

joshuahansel commented Jul 15, 2025

Uh oh!

GiudGiud commented Jul 15, 2025

Uh oh!

moosebuild commented Jul 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

moosebuild commented Jul 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Framework coverage

Modules coverage

Thermal hydraulics

Full coverage reports

Uh oh!

lindsayad left a comment

Choose a reason for hiding this comment

Uh oh!

GiudGiud left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

moosebuild commented Nov 3, 2025

Framework test summary

Removed tests

Added tests

moosebuild commented Jul 15, 2025 •

edited

Loading

moosebuild commented Jul 15, 2025 •

edited

Loading

GiudGiud left a comment •

edited

Loading