Coupling FE Scalar Field Variable to FV Kernels #29106

csdechant · 2024-11-20T15:29:11Z

Reason

This PR will enable the coupling of FE scalar field variable values and gradients to FV kernels. This PR directly addresses issue #15062

Design

This PR includes the following:

Introduces adCoupledGradientFace, which returns gradient of a coupled variable at an element face for FV Kernel.
Tests the coupling of FE scalar field variable values to FVElementKernel and FVFluxKernel.
Tests the coupling of FE scalar field variable gradients to FVElementKernel and FVFluxKernel.
Tests the coupling of FV variable gradients to FVElementKernel and FVFluxKernel.
Introduces convergence studies for all new kernel objects as heavy tests.

Impact

This PR will allow for great degree of user flexibility when solving multiphysics problems where some physics are favorable towards FE and other physics are favorable towards FE (e.g. coupling FE electromagnetics to FV conducting fluids).

moosebuild · 2024-11-20T17:42:24Z

Job Documentation, step Docs: sync website on a947630 wanted to post the following:

View the site here

This comment will be updated on new commits.

lindsayad · 2024-11-20T22:19:59Z

when you say scalar, you mean scalar field right? Because MOOSE has scalar variables whose values are independent of position on the mesh

csdechant · 2024-11-20T23:48:08Z

@lindsayad you're absolutely correct. I use the term scalar to denote that vector variables (like Nédélec variables) are not considered in this PR, but I forgot about MOOSE's Scalar variable and kernel objects. I renamed the PR to clarify scalar field.

lindsayad · 2024-11-22T16:44:46Z

I figured but just wanted to check. I'll start reviewing once more tests are passing

GiudGiud

this looks promising
can you add a section in the Coupleable.md (most likely?) on coupling FE variables in FV objects

GiudGiud · 2024-11-22T04:04:41Z

framework/include/base/Assembly.h

@@ -283,6 +283,12 @@ class Assembly
   */
  const QBase * const & qRuleFace() const { return constify_ref(_current_qrule_face); }

+  /**
+   * Returns the reference to the current FV quadrature being used on a current face


Suggested change

* Returns the reference to the current FV quadrature being used on a current face

* Returns the reference to a pointer to the current FV quadrature being used on a current face

GiudGiud · 2024-11-22T04:04:56Z

framework/include/base/Assembly.h

@@ -691,7 +697,7 @@ class Assembly
   */
  void reinit(const Elem * elem, unsigned int side, const std::vector<Point> & reference_points);

-  void reinitFVFace(const FaceInfo & fi);
+  void reinitFVFace(const FaceInfo & fi, bool areFE);


docstring, what does areFE mean

GiudGiud · 2024-11-22T16:50:15Z

framework/include/interfaces/Coupleable.h

@@ -630,6 +630,17 @@ class Coupleable
  virtual const ArrayVariableValue & coupledArrayValueOlder(const std::string & var_name,
                                                            unsigned int comp = 0) const;

+  /**
+   * Returns gradient of a coupled variable at an element face for FV Kernel / BC coupling


ideally coupleable does not mention "FV" as all routines would work for all variables? (just need to be implemented)

GiudGiud · 2024-11-22T16:51:18Z

framework/include/interfaces/NeighborCoupleable.h

+  /// Whether the MooseObject is a finite volume object
+  const bool _is_fv;


Suggested change

/// Whether the MooseObject is a finite volume object

const bool _is_fv;

use the Coupleable is_fv attribute (make it protected if it s not)

GiudGiud · 2024-11-22T16:51:48Z

framework/include/loops/ComputeFVFluxThread.h

+  // FE coupled variables. If there are some FE variables, then FE reinit is
+  // necessary.
+  bool areFE = false;
+  for (auto var : _fv_vars)


Suggested change

for (auto var : _fv_vars)

for (auto var : _vars)

GiudGiud · 2024-11-22T20:25:10Z

framework/src/variables/MooseVariableData.C

+}
+
+// NOTES: GhostValues for FE variable with equal the Face average value
+// of the neighbor side.


let's move this inside the routine

GiudGiud · 2024-11-22T20:30:27Z

framework/src/variables/MooseVariableFE.C

+    _element_data->computeGhostValuesFace(fi, *_neighbor_data);
+  }
+  else
+    mooseError("FE to FV coupling broken.");


Suggested change

mooseError("FE to FV coupling broken.");

mooseAssert(false, "FE to FV coupling broken.");

We dont expect to hit this in production

GiudGiud · 2024-11-22T20:30:56Z

framework/src/variables/MooseVariableFE.C

+void
+MooseVariableFE<OutputType>::computeAdGradFaceAvg(const FaceInfo & /*fi*/)
+{
+  mooseError("computeAdGradFaceAvg(const FaceInfo & fi) are only for MooseVariableFE<Real>.");


Suggested change

mooseError("computeAdGradFaceAvg(const FaceInfo & fi) are only for MooseVariableFE<Real>.");

mooseError("computeAdGradFaceAvg(const FaceInfo & fi) are only implemented for MooseVariableFE<Real>.");

GiudGiud · 2024-11-22T20:31:13Z

framework/src/variables/MooseVariableFE.C

+VectorValue<ADReal>
+MooseVariableFE<OutputType>::adGradSln(const FaceInfo & /*fi*/, const Moose::StateArg & /*state*/)
+{
+  mooseError("adGradSln(const FaceInfo & fi) are only for MooseVariableFE<Real>.");


Suggested change

mooseError("adGradSln(const FaceInfo & fi) are only for MooseVariableFE<Real>.");

mooseError("adGradSln(const FaceInfo & fi) are only implemented for MooseVariableFE<Real>.");

GiudGiud · 2024-11-22T20:32:26Z

framework/src/variables/MooseVariableFE.C

+
+  _ad_grad_u_face = adGradSlnAvg();
+  _ad_grad_neighbor_u_face = adGradSlnAvgNeighbor();
+  _ad_grad_face_avg = (_ad_grad_u_face[0] + _ad_grad_neighbor_u_face[0]) / 2.0;


this is for a FE variable right

can you think of a reason to not have equal weights?

GiudGiud · 2024-11-22T20:44:23Z

framework/src/variables/MooseVariableData.C

+  _ad_zero = 0;
+  unsigned int nqp = _current_qrule->n_points();
+  // NOTES: This seems like a lot of if statements. Maybe there is better way
+  // to seperate face vs element averaging...


you could split it in computeADElementAveraging and computeADFaceAveraging

GiudGiud · 2024-11-22T20:44:28Z

framework/src/variables/MooseVariableData.C

+void
+MooseVariableData<OutputType>::computeADAveraging()
+{
+  _ad_zero = 0;


do we need this?

GiudGiud · 2024-11-22T20:45:27Z

framework/include/base/Assembly.h

+  /// FV quadrature rule used on faces
+  QBase * _current_FV_qrule_face;
  /// The current arbitrary quadrature rule used on element faces


if there is another solution here than doubling up the attribute, it'd be good

GiudGiud · 2024-11-23T14:02:22Z

Maybe some high level questions,

could we then use any routine from Coupleable in a FVKernel and get the expected computations or do we have to pick only the new ones (for elemental and flux kernels respectively). If the latter is there a way to enforce this?
Why is the average value preferred over computing the centroid value? Is there some theory around this? I guess for the volume terms this leads to a conservative coupling. What about faces?

Addresses idaholab#15062

…drature points during a FE/FV simulation

moosebuild · 2024-12-04T21:05:47Z

Job Precheck, step Clang format on d1db6ab wanted to post the following:

Your code requires style changes.

A patch was auto generated and copied here
You can directly apply the patch by running, in the top level of your repository:

curl -s https://mooseframework.inl.gov/docs/PRs/29106/clang_format/style.patch | git apply -v

Alternatively, with your repository up to date and in the top level of your repository:

git clang-format 586fdf59d3f0efb469a3717233db6d0ed2634848

moosebuild · 2024-12-05T00:15:02Z

Job Coverage, step Generate coverage on a947630 wanted to post the following:

Framework coverage

	586fdf	#29106 a94763
	Total	Total	+/-	New
Rate	85.15%	85.15%	-0.00%	84.08%
Hits	107980	108138	+158	169
Misses	18827	18858	+31	32

Diff coverage report

Full coverage report

Modules coverage

Coverage did not change

Full coverage reports

Reports

Warnings

framework new line coverage rate 84.08% is less than the suggested 90.0%

This comment will be updated on new commits.

csdechant requested review from lindsayad, GiudGiud and roystgnr as code owners November 20, 2024 15:29

csdechant changed the title ~~Coupling FE Scalar Variable to FV Kernels~~ Coupling FE Scalar Field Variable to FV Kernels Nov 20, 2024

GiudGiud requested changes Nov 22, 2024

View reviewed changes

GiudGiud reviewed Nov 22, 2024

View reviewed changes

lindsayad self-assigned this Nov 22, 2024

csdechant added 5 commits December 4, 2024 10:58

Editing FE-to-FV coupling based on current MOOSE version

a19f0d8

Clean up Doc string

db178c4

Updating FE-to-FV Coupling tests with heavy convergence testing

12fc65b

Addresses idaholab#15062

Removing unicode characters

19a4164

Editing resizeMaterialData in ComputeFVFluxThread to only pull FE qua…

d1db6ab

…drature points during a FE/FV simulation

csdechant force-pushed the FEtoFVCoupling-Rebased branch from 8bc7e25 to d1db6ab Compare December 4, 2024 21:00

Addressing Clang-Format

a947630

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Coupling FE Scalar Field Variable to FV Kernels #29106

Coupling FE Scalar Field Variable to FV Kernels #29106

csdechant commented Nov 20, 2024 •

edited

Loading

moosebuild commented Nov 20, 2024 •

edited

Loading

lindsayad commented Nov 20, 2024

csdechant commented Nov 20, 2024 •

edited

Loading

lindsayad commented Nov 22, 2024

GiudGiud left a comment

GiudGiud Nov 22, 2024

GiudGiud Nov 22, 2024

GiudGiud Nov 22, 2024

GiudGiud Nov 22, 2024

GiudGiud Nov 22, 2024

GiudGiud Nov 22, 2024

GiudGiud Nov 22, 2024

GiudGiud Nov 22, 2024

GiudGiud Nov 22, 2024

GiudGiud Nov 22, 2024

GiudGiud Nov 22, 2024

GiudGiud Nov 22, 2024

GiudGiud Nov 22, 2024

GiudGiud commented Nov 23, 2024 •

edited

Loading

moosebuild commented Dec 4, 2024

moosebuild commented Dec 5, 2024

	* Returns the reference to the current FV quadrature being used on a current face
	* Returns the reference to a pointer to the current FV quadrature being used on a current face

		/// Whether the MooseObject is a finite volume object
		const bool _is_fv;

	mooseError("FE to FV coupling broken.");
	mooseAssert(false, "FE to FV coupling broken.");

	mooseError("computeAdGradFaceAvg(const FaceInfo & fi) are only for MooseVariableFE<Real>.");
	mooseError("computeAdGradFaceAvg(const FaceInfo & fi) are only implemented for MooseVariableFE<Real>.");

	mooseError("adGradSln(const FaceInfo & fi) are only for MooseVariableFE<Real>.");
	mooseError("adGradSln(const FaceInfo & fi) are only implemented for MooseVariableFE<Real>.");

Coupling FE Scalar Field Variable to FV Kernels #29106

Are you sure you want to change the base?

Coupling FE Scalar Field Variable to FV Kernels #29106

Conversation

csdechant commented Nov 20, 2024 • edited Loading

Reason

Design

Impact

moosebuild commented Nov 20, 2024 • edited Loading

lindsayad commented Nov 20, 2024

csdechant commented Nov 20, 2024 • edited Loading

lindsayad commented Nov 22, 2024

GiudGiud left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

GiudGiud commented Nov 23, 2024 • edited Loading

moosebuild commented Dec 4, 2024

moosebuild commented Dec 5, 2024

Framework coverage

Modules coverage

Full coverage reports

Warnings

csdechant commented Nov 20, 2024 •

edited

Loading

moosebuild commented Nov 20, 2024 •

edited

Loading

csdechant commented Nov 20, 2024 •

edited

Loading

GiudGiud commented Nov 23, 2024 •

edited

Loading