feat: MGR Strategy for ALM - Face Bubble Functions for Wedge Elements - Multipoint Integration Rules for Tetrahedra and Triangles #3395

matteofrigo5 · 2024-10-11T23:51:42Z

This PR introduces the following enhancements:

- Two solving strategies for ALM: one utilizing MGR and the other employing AMG on the full system.
- Implementation of face bubble functions for wedge elements.
- An optional feature to utilize Multipoint Integration Rules for both tetrahedra and triangles.

…triangle

rrsettgast

just a few questions

rrsettgast · 2025-01-09T18:43:48Z

src/coreComponents/constitutive/contact/CoulombFriction.hpp

-  tractionTrial[ 0 ] = traction[0] + penalty[0] * dispJump[0] * faceArea;
-  tractionTrial[ 1 ] = traction[1] + penalty[1] * (dispJump[1] - oldDispJump[1]) * faceArea;
-  tractionTrial[ 2 ] = traction[2] + penalty[1] * (dispJump[2] - oldDispJump[2]) * faceArea;
+  tractionTrial[ 0 ] = traction[0] + penalty[0] * dispJump[0];


so this "traction" is now a "traction"?

Yes, it is . Actually, it has always been a "traction". In the previous implementation, dispJump was multiplied by faceArea. Now that dispJump is as it should be, it is used in its original form.

rrsettgast · 2025-01-09T19:09:03Z

src/coreComponents/constitutive/contact/FrictionBase.hpp

@@ -134,7 +133,7 @@ class FrictionBaseUpdates
                               real64 ( & tractionNew )[3],
                               integer & fractureState ) const
  {
-    GEOS_UNUSED_VAR( oldDispJump, dispJump, penalty, traction, faceArea, symmetric, fixedLimitTau,
+    GEOS_UNUSED_VAR( oldDispJump, dispJump, penalty, traction, symmetric, fixedLimitTau,


why is this function defined?

rrsettgast · 2025-01-09T19:10:52Z

src/coreComponents/constitutive/contact/FrictionBase.hpp

                                   arraySlice1d< real64 > const & tractionNew ) const
-  { GEOS_UNUSED_VAR( dispJump, deltaDispJump, penalty, traction, faceArea, tractionNew ); }
+  { GEOS_UNUSED_VAR( dispJump, deltaDispJump, penalty, traction, tractionNew ); }


or this one? why not a pure virtual?

because these are device functions. I am pretty sure that nvcc throws an error if we use a pure virtual. We do this a lot in our kernelWrapper objects (probably with the idea of getting nicer compilation errors) but, tbh, I start thinking that it would be better not to use virtual at all for these.

rrsettgast · 2025-01-09T19:12:02Z

src/coreComponents/finiteElement/elementFormulations/H1_Tetrahedron_Lagrange1_Gauss1.hpp

+#if TETRAHEDRON_QUADRATURE_POINTS == 14
+  constexpr static localIndex numQuadraturePoints = 14;
+#elif TETRAHEDRON_QUADRATURE_POINTS == 5
+  constexpr static localIndex numQuadraturePoints = 5;
+#elif TETRAHEDRON_QUADRATURE_POINTS == 1


This is pretty clunky. There should be separate class for the different integration rules.

I agree. It was just a temporary implementation before discussing the creation of a separate class. Would you like to implement it in this PR, or should we open a specific one for this purpose?

We should redefine this class

template< int NUM_Q_POINTS > class H1_Tetrahedron_Lagrange1_Gauss final : public FiniteElementBase { ... }; using H1_Tetrahedron_Lagrange1_Gauss1 = H1_Tetrahedron_Lagrange1_Gauss<1>; ... using H1_Tetrahedron_Lagrange1_Gauss14 = H1_Tetrahedron_Lagrange1_Gauss<14>;

rrsettgast · 2025-01-09T19:13:54Z

src/coreComponents/finiteElement/elementFormulations/H1_TriangleFace_Lagrange1_Gauss1.hpp

@@ -29,6 +29,8 @@ namespace geos
 namespace finiteElement
 {

+#define TRIANGLE_QUADRATURE_POINTS 4


same comment here about a separate class.

rrsettgast · 2025-01-09T19:15:11Z

src/coreComponents/physicsSolvers/contact/ContactFields.hpp

+               WRITE_AND_READ,
+               "Slip" );
+
+DECLARE_FIELD( tangentialTraction,


Are we storing a "traction" and a "tangentialTraction"?

It was just for output purposes. Having the absolute value of tangential traction is convenient without needing two separate tangential components.

rrsettgast · 2025-01-09T19:26:10Z

src/coreComponents/physicsSolvers/contact/SolidMechanicsAugmentedLagrangianContact.hpp

+
+    constexpr static char const * iterativePenaltyTFacString() { return "iterPenaltyT"; }
+
+    constexpr static char const * tolJumpDispNFacString() { return "tolJumpN"; }


should these be dependent on the contact parameters?

As far as the default values are concerned, I would say yes. However, it is convenient to have these tolerances to fine-tune the physics solver for the specific problem.

rrsettgast · 2025-01-09T19:27:04Z

src/coreComponents/physicsSolvers/contact/kernels/SolidMechanicsALMSimultaneousKernels.hpp


    for( int i=0; i<numTdofs; ++i )
    {
      stack.tLocal[i] = m_traction( k, i );
-      stack.dispJumpLocal[i] = m_dispJump( k, i );
-      stack.oldDispJumpLocal[i] = m_oldDispJump( k, i );
+      stack.dispJumpLocal[i] = m_dispJump( k, i )*m_faceArea[k];


this seems odd to me that you would need to multiply by an area here.

rrsettgast · 2025-01-09T19:27:48Z

src/coreComponents/physicsSolvers/contact/kernels/SolidMechanicsALMSimultaneousKernels.hpp

-    dispJump[0] = stack.dispJumpLocal[0] * m_faceArea[k];
-    dispJump[1] = ( stack.dispJumpLocal[1] - stack.oldDispJumpLocal[1] ) * m_faceArea[k];
-    dispJump[2] = ( stack.dispJumpLocal[2] - stack.oldDispJumpLocal[2] ) * m_faceArea[k];
+    dispJump[0] = stack.dispJumpLocal[0];


See comment below.

src/coreComponents/physicsSolvers/contact/kernels/SolidMechanicsALMSimultaneousKernels.hpp

…grange1_Gauss.hpp

CusiniM · 2025-01-14T18:29:51Z

src/coreComponents/finiteElement/elementFormulations/H1_Tetrahedron_Lagrange1_Gauss.hpp

 *           |  +__  \_         |   s        2      0   1   0
 *           | /2  \__ \_       |  /         3      0   0   1
 *           |/       \__\      | /          =====  ==  ==  ==
 *           +------------+     *------r
 *          0              1
 *
 */
-class H1_Tetrahedron_Lagrange1_Gauss1 final : public FiniteElementBase
+template< int NUM_Q_POINTS >


Suggested change

template< int NUM_Q_POINTS >

template< typename NUM_Q_POINTS >

and then NUM_Q_POINTS::value to use it andH1_Tetrahedron_Lagrange1_Gauss< std::integral_constant< int, 1 > >

CusiniM · 2025-01-14T18:35:19Z

src/coreComponents/finiteElement/elementFormulations/H1_Tetrahedron_Lagrange1_Gauss.hpp

+    else
+    {
+      GEOS_UNUSED_VAR( q );
+      GEOS_ERROR( "NUM_Q_POINTS not available for this element type" );


I don't know if we need this. I guess ideally we would like it not to compile at all.

Maybe we can add at the top of the class:

static_assert(NUM_Q_POINTS::value == 1 || NUM_Q_POINTS::value == 5 |NUM_Q_POINTS::value == 14, "NUM_Q_POINTS::value must be 1, 5, or 14!");

…GEOS into feature/mfrigo/mgr4alm

codecov · 2025-01-15T18:59:53Z

Codecov Report

Attention: Patch coverage is 19.17808% with 177 lines in your changes missing coverage. Please review.

Project coverage is 56.47%. Comparing base (8f9126c) to head (4f46983).

Files with missing lines	Patch %	Lines
...ntact/SolidMechanicsAugmentedLagrangianContact.cpp	0.00%	60 Missing ⚠️
...entFormulations/H1_Tetrahedron_Lagrange1_Gauss.hpp	44.89%	27 Missing ⚠️
.../elementFormulations/H1_Wedge_Lagrange1_Gauss6.hpp	0.00%	27 Missing ⚠️
...Strategies/AugmentedLagrangianContactMechanics.hpp	0.00%	15 Missing ⚠️
...t/kernels/SolidMechanicsALMSimultaneousKernels.hpp	0.00%	9 Missing ⚠️
...ntFormulations/H1_TriangleFace_Lagrange1_Gauss.hpp	65.21%	8 Missing ⚠️
...ntact/SolidMechanicsAugmentedLagrangianContact.hpp	0.00%	8 Missing ⚠️
...ents/finiteElement/FiniteElementDiscretization.cpp	36.36%	7 Missing ⚠️
...omponents/constitutive/contact/CoulombFriction.hpp	0.00%	6 Missing ⚠️
...onents/linearAlgebra/interfaces/hypre/HypreMGR.cpp	0.00%	3 Missing ⚠️
... and 5 more

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #3395      +/-   ##
===========================================
- Coverage    56.53%   56.47%   -0.06%     
===========================================
  Files         1157     1158       +1     
  Lines       100617   100760     +143     
===========================================
+ Hits         56880    56909      +29     
- Misses       43737    43851     +114

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

CusiniM · 2025-01-15T22:55:32Z

src/coreComponents/finiteElement/FiniteElementDiscretization.cpp

+  registerWrapper( viewKeyStruct::useHighOrderQuadratureRuleString(), &m_useHighOrderQuadratureRule ).
+    setInputFlag( InputFlags::OPTIONAL ).
+    setApplyDefaultValue( 0 ).
+    setDescription( "Specifier to indicate whether to use a high order quadrature rule" );


Suggested change

setDescription( "Specifier to indicate whether to use a high order quadrature rule" );

setDescription( "Flag to indicate whether to use a high order quadrature rule or not." );

matteofrigo5 added 4 commits September 29, 2024 16:00

Adding linear solving strategy

00157b3

Adding bubble functions for wedge elements

d9cd65e

Fixed face bubble functions for wedge elements

7cf3f56

Adding parameters to tune the iterative penalty coefficients

e8d69eb

matteofrigo5 self-assigned this Oct 11, 2024

matteofrigo5 requested review from CusiniM, rrsettgast, castelletto1, victorapm and andrea-borio as code owners October 11, 2024 23:51

matteofrigo5 added type: feature New feature or request flag: no rebaseline Does not require rebaseline labels Oct 11, 2024

Adding points for quadrature rule

5a85a40

matteofrigo5 requested a review from jhuang2601 as a code owner November 1, 2024 20:37

matteofrigo5 added 2 commits November 4, 2024 15:39

Scaling dispJump by area

c7d9186

Added var to set the number of quadrature points for tetrahedron and …

69ab77a

…triangle

matteofrigo5 requested a review from Guotong-Ren as a code owner November 13, 2024 23:10

matteofrigo5 changed the title ~~feat: linear solving strategy for ALM and face bubble functions for wedge elements~~ feat: MGR Strategy for ALM - Face Bubble Functions for Wedge Elements - Multipoint Integration Rules for Tetrahedra and Triangle Jan 9, 2025

Develop branch merged and conflicts fixed

b946e59

matteofrigo5 added flag: ready for review flag: requires rebaseline Requires rebaseline branch in integratedTests and removed flag: no rebaseline Does not require rebaseline labels Jan 9, 2025

rrsettgast reviewed Jan 9, 2025

View reviewed changes

matteofrigo5 added 3 commits January 10, 2025 02:51

Cleaning up SolidMechanicsALMSimultaneousKernels.hpp

36c70c5

uncrustify_style

e5ff261

Cleaning up H1_TriangleFace_Lagrange1_Gauss.hpp and H1_Tetrahedron_La…

459aef6

…grange1_Gauss.hpp

matteofrigo5 requested review from wrtobin and untereiner as code owners January 14, 2025 05:48

CusiniM reviewed Jan 14, 2025

View reviewed changes

matteofrigo5 added 2 commits January 14, 2025 13:47

Fixing some review comments

5cd89d9

Fixing compilation issues

907dcae

matteofrigo5 requested review from paveltomin, frankfeifan and ryar9534 as code owners January 15, 2025 00:09

matteofrigo5 and others added 4 commits January 14, 2025 16:10

Merge branch 'develop' into feature/mfrigo/mgr4alm

37d2396

Fixed bug (int-real type casting)

ab95729

Merge branch 'feature/mfrigo/mgr4alm' of https://github.com/GEOS-DEV/…

9e0270f

…GEOS into feature/mfrigo/mgr4alm

uncrustify

53eed41

matteofrigo5 added ci: run CUDA builds Allows to triggers (costly) CUDA jobs ci: run code coverage enables running of the code coverage CI jobs labels Jan 15, 2025

matteofrigo5 added ci: run integrated tests Allows to run the integrated tests in GEOS CI and removed flag: requires rebaseline Requires rebaseline branch in integratedTests labels Jan 15, 2025

CusiniM reviewed Jan 15, 2025

View reviewed changes

CusiniM approved these changes Jan 15, 2025

View reviewed changes

matteofrigo5 added the flag: no rebaseline Does not require rebaseline label Jan 16, 2025

Merge branch 'develop' into feature/mfrigo/mgr4alm

4f46983

matteofrigo5 added flag: requires rebaseline Requires rebaseline branch in integratedTests and removed flag: no rebaseline Does not require rebaseline labels Jan 16, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: MGR Strategy for ALM - Face Bubble Functions for Wedge Elements - Multipoint Integration Rules for Tetrahedra and Triangles #3395

feat: MGR Strategy for ALM - Face Bubble Functions for Wedge Elements - Multipoint Integration Rules for Tetrahedra and Triangles #3395

matteofrigo5 commented Oct 11, 2024 •

edited

Loading

rrsettgast left a comment

rrsettgast Jan 9, 2025

matteofrigo5 Jan 10, 2025

rrsettgast Jan 9, 2025

rrsettgast Jan 9, 2025

CusiniM Jan 9, 2025

rrsettgast Jan 9, 2025

matteofrigo5 Jan 10, 2025

rrsettgast Jan 10, 2025

rrsettgast Jan 9, 2025

rrsettgast Jan 9, 2025

matteofrigo5 Jan 10, 2025

rrsettgast Jan 9, 2025

matteofrigo5 Jan 10, 2025

rrsettgast Jan 9, 2025

rrsettgast Jan 9, 2025

matteofrigo5 Jan 10, 2025

CusiniM Jan 14, 2025

CusiniM Jan 14, 2025

CusiniM Jan 14, 2025 •

edited

Loading

codecov bot commented Jan 15, 2025 •

edited

Loading

CusiniM Jan 15, 2025


		constexpr static char const * iterativePenaltyTFacString() { return "iterPenaltyT"; }

		constexpr static char const * tolJumpDispNFacString() { return "tolJumpN"; }

	template< int NUM_Q_POINTS >
	template< typename NUM_Q_POINTS >

	setDescription( "Specifier to indicate whether to use a high order quadrature rule" );
	setDescription( "Flag to indicate whether to use a high order quadrature rule or not." );

feat: MGR Strategy for ALM - Face Bubble Functions for Wedge Elements - Multipoint Integration Rules for Tetrahedra and Triangles #3395

Are you sure you want to change the base?

feat: MGR Strategy for ALM - Face Bubble Functions for Wedge Elements - Multipoint Integration Rules for Tetrahedra and Triangles #3395

Conversation

matteofrigo5 commented Oct 11, 2024 • edited Loading

rrsettgast 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

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

CusiniM Jan 14, 2025 • edited Loading

Choose a reason for hiding this comment

codecov bot commented Jan 15, 2025 • edited Loading

Codecov Report

Choose a reason for hiding this comment

matteofrigo5 commented Oct 11, 2024 •

edited

Loading

CusiniM Jan 14, 2025 •

edited

Loading

codecov bot commented Jan 15, 2025 •

edited

Loading