feat(materials): add substepper for finite strain materials by alexdummer · Pull Request #32 · MAteRialMOdelingToolbox/Marmot

alexdummer · 2026-01-12T03:34:32Z

No description provided.

Copilot

Pull request overview

This WIP pull request introduces a substepping capability for finite strain material models through a decorator pattern, enabling improved numerical stability and accuracy for large deformation increments through analytical tangent accumulation.

Changes:

Adds a new template decorator class MarmotMaterialFiniteStrainSubstepped that wraps existing material models with time substepping functionality
Implements computeStressWithSensitivities method with default numerical differentiation for sensitivity matrices required by substepping
Fixes a bug in centralDifference where the Jacobian matrix was allocated before knowing the function output size

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 19 comments.

Show a summary per file

File	Description
modules/core/MarmotFiniteStrainMechanicsCore/include/Marmot/MarmotMaterialFiniteStrainSubstepped.h	New template class implementing substepping decorator with analytical tangent accumulation
modules/core/MarmotFiniteStrainMechanicsCore/include/Marmot/MarmotMaterialFiniteStrain.h	Adds StateSensitivities struct and computeStressWithSensitivities method with numerical differentiation
modules/core/MarmotMathCore/src/MarmotNumericalDifferentiation.cpp	Fixes bug in centralDifference by deferring matrix allocation until output size is known
modules/materials/FiniteStrainJ2Plasticity/src/FiniteStrainJ2PlasticityRegistration.cpp	Registers substepped variant of J2 plasticity model
modules/materials/FiniteStrainJ2Plasticity/src/FiniteStrainJ2Plasticity.cpp	Changes initialization method from .eye() to memcpy for consistency
modules/core/MarmotMathCore/module.cmake	Adds public headers for Fastor tensor basics and numerical differentiation
examples/finite-strain-mp-solver/FiniteStrainJ2PlasticitySubstepped/example.cpp	Demonstrates usage of substepped material with mixed loading controls
examples/finite-strain-mp-solver/FiniteStrainJ2PlasticitySubstepped/Makefile	Build configuration for the example

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-01-12T03:40:37Z

modules/core/MarmotFiniteStrainMechanicsCore/include/Marmot/MarmotMaterialFiniteStrain.h

+   * @brief Extended stress update computing sensitivities for substepping.
+   * * The default implementation uses Central Finite Differences.


The documentation has formatting issues with asterisks. Change '* The default implementation' to 'The default implementation' and '* @brief' to '@brief'.

...s/core/MarmotFiniteStrainMechanicsCore/include/Marmot/MarmotMaterialFiniteStrainSubstepped.h

Copilot · 2026-01-12T03:40:38Z

modules/core/MarmotFiniteStrainMechanicsCore/include/Marmot/MarmotMaterialFiniteStrain.h

+    /* std::cout << "deformation.F: \n" << deformation.F << std::endl; */
+    /* std::cout << "stateOld: \n" << Map<const VectorXd>(stateOld.data(), nState) << std::endl; */
+    // Run the actual update once to get Stress and New State (at t_n+1)
+    computeStress( response, tangents, deformation, timeIncrement );
+
+    /* std::cout << "Base Stress after update: \n" << response.tau << std::endl; */
+
+    // If there are no state variables (elastic), sensitivities are zero/empty
+    if ( nState == 0 )
+      return;
+
+    // --- Numerical Differentiation Setup ---
+
+    // Prepare sensitivity matrices
+    sensitivities.dState_dF.resize( nState, 9 );
+    sensitivities.dState_dStateOld.resize( nState, nState );
+    sensitivities.dStress_dStateOld.resize( 9, nState );
+
+    // We need temporary storage for the FD routines to avoid corrupting the actual results
+    // or the input pointers.
+
+    // 2. Compute d(State_new) / d(F_new)
+    // We perturb F, run computeStress (starting from fixed StateOld), and measure State_new.
+    auto func_dState_dF = [&]( const VectorXd& F_vec ) -> VectorXd {
+      Deformation< 3 > defPerturbed;
+      std::memcpy( defPerturbed.F.data(), F_vec.data(), 9 * sizeof( double ) );
+
+      ConstitutiveResponse< 3 > respPert;
+      std::vector< double >     stateTemp = stateOld; // Copy from OLD
+      respPert.stateVars                  = stateTemp.data();
+
+      AlgorithmicModuli< 3 > tanPert; // dummy
+
+      // Run update
+      computeStress( respPert, tanPert, defPerturbed, timeIncrement );
+
+      // Return NEW state
+      Map< VectorXd > res( respPert.stateVars, nState );
+      return res;
+    };
+
+    Map< const VectorXd > F_map( deformation.F.data(), 9 );
+    sensitivities.dState_dF = forwardDifference( func_dState_dF, F_map );
+
+    // print dState_dF if det is nan
+    if ( sensitivities.dState_dF.hasNaN() ) {
+      std::cout << "Computed dState/dF has NaN!" << std::endl;
+      std::cout << "F: \n" << deformation.F << std::endl;
+      std::cout << "StateOld: \n" << Map< const VectorXd >( stateOld.data(), nState ) << std::endl;
+    }
+    /* std::cout << "Computed dState/dF via FD: \n" << sensitivities.dState_dF << std::endl; */
+
+    // 3. Compute d(State_new) / d(State_old) AND d(Stress) / d(State_old)
+    // We perturb State_old, run computeStress (fixed F), and measure State_new and Stress.
+
+    // Since centralDifference returns one matrix, we do this in two passes or combine outputs.
+    // Let's do two passes for clarity, though it doubles the cost of this specific block.
+    // Given the prompt constraints, we use the provided interface.
+
+    // A. d(State_new) / d(State_old)
+    auto func_dState_dStateOld = [&]( const VectorXd& StateOld_vec ) -> VectorXd {
+      ConstitutiveResponse< 3 > respPert;
+      std::vector< double >     stateTemp( nState );
+      std::memcpy( stateTemp.data(), StateOld_vec.data(), nState * sizeof( double ) );
+      respPert.stateVars = stateTemp.data();
+
+      AlgorithmicModuli< 3 > tanPert; // dummy
+
+      // Run update with perturbed old state
+      computeStress( respPert, tanPert, deformation, timeIncrement );
+
+      Map< VectorXd > res( respPert.stateVars, nState );
+      return res;
+    };
+
+    Map< const VectorXd > StateOld_map( stateOld.data(), nState );
+    sensitivities.dState_dStateOld = forwardDifference( func_dState_dStateOld, StateOld_map );
+
+    // print dState_dF if det is nan
+    if ( sensitivities.dState_dStateOld.hasNaN() ) {
+      std::cout << "Computed dState/dStateOld has NaN!" << std::endl;
+      std::cout << "F: \n" << deformation.F << std::endl;
+      std::cout << "StateOld: \n" << Map< const VectorXd >( stateOld.data(), nState ) << std::endl;
+    }
+    // B. d(Stress) / d(State_old)
+    auto func_dStress_dStateOld = [&]( const VectorXd& StateOld_vec ) -> VectorXd {
+      ConstitutiveResponse< 3 > respPert;
+      std::vector< double >     stateTemp( nState );
+      std::memcpy( stateTemp.data(), StateOld_vec.data(), nState * sizeof( double ) );
+      respPert.stateVars = stateTemp.data();
+
+      AlgorithmicModuli< 3 > tanPert; // dummy
+
+      computeStress( respPert, tanPert, deformation, timeIncrement );
+
+      // Return Stress (flattened 9x1)
+      Map< VectorXd > res( respPert.tau.data(), 9 );
+      return res;
+    };
+
+    sensitivities.dStress_dStateOld = forwardDifference( func_dStress_dStateOld, StateOld_map );
+    if ( sensitivities.dStress_dStateOld.hasNaN() ) {
+      std::cout << "Computed dStress/dStateOld has NaN!" << std::endl;
+      std::cout << "F: \n" << deformation.F << std::endl;
+      std::cout << "StateOld: \n" << Map< const VectorXd >( stateOld.data(), nState ) << std::endl;
+    }


Multiple debug print statements are left commented out in production code (lines 110-111, 155-163, 221-227, 255-260, 278-282). While this is a WIP PR, these should either be removed or replaced with a proper logging framework before merging to maintain code cleanliness.

modules/core/MarmotFiniteStrainMechanicsCore/include/Marmot/MarmotMaterialFiniteStrain.h

modules/core/MarmotMathCore/src/MarmotNumericalDifferentiation.cpp

Copilot · 2026-01-12T03:40:41Z

modules/core/MarmotFiniteStrainMechanicsCore/include/Marmot/MarmotMaterialFiniteStrain.h

+  virtual void computeStressWithSensitivities( ConstitutiveResponse< 3 >& response,
+                                               AlgorithmicModuli< 3 >&    tangents,
+                                               StateSensitivities&        sensitivities,
+                                               const Deformation< 3 >&    deformation,
+                                               const TimeIncrement&       timeIncrement ) const


The computeStressWithSensitivities method modifies response.stateVars as a side effect while also computing sensitivities. This could be confusing for users who might expect the method to be const (like computeStress) or who might not expect state updates when computing sensitivities. Consider documenting this behavior clearly in the method documentation, explicitly stating that the method updates the state variables in addition to computing sensitivities.

...s/core/MarmotFiniteStrainMechanicsCore/include/Marmot/MarmotMaterialFiniteStrainSubstepped.h

examples/finite-strain-mp-solver/FiniteStrainJ2PlasticitySubstepped/example.cpp

…rmotMaterialFiniteStrainSubstepped.h Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

…rmotMaterialFiniteStrain.h Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

….cpp Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

…rmotMaterialFiniteStrainSubstepped.h Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

…epped/example.cpp Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

…rmotMaterialFiniteStrainSubstepped.h Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

…epped/example.cpp Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

examples/finite-strain-mp-solver/FiniteStrainJ2PlasticitySubstepped/Makefile

matthiasneuner · 2026-01-13T09:18:41Z

modules/core/MarmotFiniteStrainMechanicsCore/include/Marmot/MarmotMaterialFiniteStrain.h

+  /**
+   * @struct StateSensitivities
+   * @brief Sensitivity matrices required for analytical substepping.
+   * All matrices represent derivatives of flattened arrays (Row-Major Fastor layout).


Attention: Eigen itself is not Row-Major;

modules/core/MarmotFiniteStrainMechanicsCore/include/Marmot/MarmotMaterialFiniteStrain.h

...s/core/MarmotFiniteStrainMechanicsCore/include/Marmot/MarmotMaterialFiniteStrainSubstepped.h

add substepper for finite strain materials

314d371

alexdummer requested a review from Copilot January 12, 2026 03:34

alexdummer assigned matthiasneuner and alexdummer Jan 12, 2026

alexdummer marked this pull request as draft January 12, 2026 03:34

Copilot started reviewing on behalf of alexdummer January 12, 2026 03:34 View session

Copilot AI reviewed Jan 12, 2026

View reviewed changes

alexdummer and others added 8 commits January 11, 2026 21:59

Update modules/core/MarmotFiniteStrainMechanicsCore/include/Marmot/Ma…

7af2ce3

…rmotMaterialFiniteStrainSubstepped.h Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Update modules/core/MarmotFiniteStrainMechanicsCore/include/Marmot/Ma…

e6b9171

…rmotMaterialFiniteStrain.h Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Update modules/core/MarmotMathCore/src/MarmotNumericalDifferentiation…

79c404a

….cpp Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Update modules/core/MarmotFiniteStrainMechanicsCore/include/Marmot/Ma…

790ce78

…rmotMaterialFiniteStrainSubstepped.h Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Update examples/finite-strain-mp-solver/FiniteStrainJ2PlasticitySubst…

54e7f59

…epped/example.cpp Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Update modules/core/MarmotFiniteStrainMechanicsCore/include/Marmot/Ma…

f4e53d6

…rmotMaterialFiniteStrainSubstepped.h Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Update modules/core/MarmotFiniteStrainMechanicsCore/include/Marmot/Ma…

0707d8b

…rmotMaterialFiniteStrainSubstepped.h Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Update examples/finite-strain-mp-solver/FiniteStrainJ2PlasticitySubst…

8f3a230

…epped/example.cpp Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

matthiasneuner reviewed Jan 13, 2026

View reviewed changes

alexdummer and others added 4 commits January 14, 2026 13:33

change to cmake

ca9f405

move extended stress update to substepper class

c2a3ec6

remove unused import and fix type

c4996dc

Merge branch 'next_v26.05' into finite-strain-substepper-only

66cc149

alexdummer marked this pull request as ready for review February 25, 2026 00:58

alexdummer changed the title ~~[WIP] feat(materials): add substepper for finite strain materials~~ feat(materials): add substepper for finite strain materials Feb 25, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(materials): add substepper for finite strain materials#32

feat(materials): add substepper for finite strain materials#32
alexdummer wants to merge 13 commits intoMAteRialMOdelingToolbox:next_v26.05from
alexdummer:finite-strain-substepper-only

alexdummer commented Jan 12, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Jan 12, 2026

Uh oh!

Uh oh!

Copilot AI Jan 12, 2026

Uh oh!

Uh oh!

Uh oh!

Copilot AI Jan 12, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

matthiasneuner Jan 13, 2026

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

		* @brief Extended stress update computing sensitivities for substepping.
		* * The default implementation uses Central Finite Differences.

Conversation

alexdummer commented Jan 12, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Jan 12, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Copilot AI Jan 12, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Copilot AI Jan 12, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

matthiasneuner Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants