Augmented Lagrangian (stick mode) - InterfaceKernelBase - Bubble Functions #3130

matteofrigo5 · 2024-05-17T01:02:52Z

This PR introduces the following three features:

Development of kernelBaseInterface: This new interface is designed to launch kernels that perform operations on interface elements.
Implementation of Bubble Element and Face Basis Functions: These functions are used for stabilizing the first-order displacement cell constant pressure formulation in contact mechanics (and the formulation can be easily extended to flow problems).
Augmented Lagrangian Method for Contact Mechanics: This initial implementation focuses on the stick mode. Subsequent PRs will address the slip and open modes once this PR is merged.

codecov · 2024-05-17T22:44:33Z

Codecov Report

Attention: Patch coverage is 0.54795% with 1089 lines in your changes missing coverage. Please review.

Project coverage is 55.71%. Comparing base (9134c42) to head (057f692).
Report is 99 commits behind head on develop.

Files with missing lines	Patch %	Lines
...ntact/SolidMechanicsAugmentedLagrangianContact.cpp	0.00%	591 Missing ⚠️
...hysicsSolvers/contact/SolidMechanicsALMKernels.hpp	0.00%	106 Missing ⚠️
...Solvers/contact/SolidMechanicsALMBubbleKernels.hpp	0.00%	98 Missing ⚠️
...ers/contact/SolidMechanicsALMJumpUpdateKernels.hpp	0.00%	53 Missing ⚠️
...csSolvers/contact/SolidMechanicsALMKernelsBase.hpp	0.00%	46 Missing ⚠️
...iteElement/kernelInterface/InterfaceKernelBase.hpp	0.00%	39 Missing ⚠️
...niteElement/elementFormulations/LagrangeBasis1.hpp	0.00%	37 Missing ⚠️
...ntFormulations/H1_Tetrahedron_Lagrange1_Gauss1.hpp	0.00%	35 Missing ⚠️
...Solvers/contact/SolidMechanicsALMKernelsHelper.hpp	0.00%	17 Missing ⚠️
...lations/H1_Hexahedron_Lagrange1_GaussLegendre2.hpp	0.00%	15 Missing ⚠️
... and 8 more

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #3130      +/-   ##
===========================================
- Coverage    56.41%   55.71%   -0.71%     
===========================================
  Files         1023     1031       +8     
  Lines        86596    87698    +1102     
===========================================
+ Hits         48857    48863       +6     
- Misses       37739    38835    +1096

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

CusiniM · 2024-06-10T20:46:37Z

src/coreComponents/finiteElement/elementFormulations/H1_Hexahedron_Lagrange1_GaussLegendre2.hpp

@@ -75,6 +75,10 @@ class H1_Hexahedron_Lagrange1_GaussLegendre2 final : public FiniteElementBase
 public:
  /// The number of nodes/support points per element.
  constexpr static localIndex numNodes = LagrangeBasis1::TensorProduct3D::numSupportPoints;
+
+  /// The number of nodes/support points per element.


Fixed the comment with: faces/support point

CusiniM · 2024-06-10T20:47:23Z

src/coreComponents/finiteElement/elementFormulations/H1_Hexahedron_Lagrange1_GaussLegendre2.hpp

@@ -209,6 +213,42 @@ class H1_Hexahedron_Lagrange1_GaussLegendre2 final : public FiniteElementBase
    return calcN( q, N );
  }

+  /**
+   * @brief Calculate shape functions values for each support face at a


what's a support face? A face to which we are associating a bubble function?

Yes, a face to which we are associating a bubble function. I have modified this comment with: Calculate face bubble functions values for each face at a given point in the parent space.

CusiniM · 2024-06-10T20:52:09Z

src/coreComponents/finiteElement/elementFormulations/H1_TriangleFace_Lagrange1_Gauss1.hpp

+   */
+  GEOS_HOST_DEVICE
+  GEOS_FORCE_INLINE
+  static void calcBubbleN( real64 const (&pointCoord)[2],


it seems that for 3D elements the bubble functions live on the faces whereas for 2D elements there is one per element. This is clearly how we are using them for contact but if we want this to be more generic we will need element wise bubbles for the 3D cells elements, right?

There are two types of bubble functions both for 3D and 2D elements:

face bubble functions, which live on the faces/edges, for 3D/2D respectively;

element bubble functions, which live on the elements.
In the code, when I refer to the first type, I have added Face in the function name.
Hence, the answer to your question is yes, in the more generic case we will need element wise bubbles functions for the 3D and also face bubble functions for 2D elements.

CusiniM · 2024-06-10T21:02:34Z

src/coreComponents/mesh/CellElementSubRegion.hpp

+  array1d< localIndex > m_bubbleCells;
+
+  /// List of face ids which are shared with the interface elements (useful for bubble elements)
+  array2d< localIndex > m_toFaceElements;


I am not sure I fully understand these 2 lists. They also don't seem to be initialized nor registered as wrappers. I would do that. Make sure you set the sizeFromParent to false.

it seems to be that m_bubbleCells just contains the indices of all 3D cells that are connected to a 2D fracture cell. I would say:
"List of the elements adjacent to a faceElement (useful for bubble elements)"

The second list is a bit more confusing. What's the size of that array2d? Are you trying to store the localIndex of each face connected to the faceElement ? That map already exists in the FaceElementSubRegion so I would use that one.

Yes, you are right. I have to register the two lists as wrappers.

The second list is a 2D array containing the localIndex of each face in both the physical and parent element spaces, which are connected to the faceElement. I can't use the map in the FaceElementSubRegion because I need to access this list during the loop over the cells adjacent to a faceElement.

CusiniM · 2024-06-10T21:07:10Z

src/coreComponents/physicsSolvers/contact/SolidMechanicsALMBubbleKernels.hpp

+    static constexpr int numUdofs = numNodesPerElem * 3;
+
+    /// The number of jump dofs per element.
+    static constexpr int numBubbleUdofs = numFacesPerElem * 3;


are you currently assuming that all faces get a bubble?

Yes, I'm. But I assemble in the global matrix only the contribute of the face bubble connected to the fracture.

CusiniM · 2024-06-10T21:15:24Z