[beaminteraction] Individual augmentation scaling parameters

[isteinbrecher]

@maxfirmbach this should somewhat do what we talked about. We can now scale the augmentation blocks for beam and solid independently. This can be controlled via the parameters AUGMENTATION_SCALING_PARAMETER_BEAM and AUGMENTATION_SCALING_PARAMETER_SOLID which simply scale the penalty terms. The underlying penalty calculation is not changed.

I think it makes sense to do it this way, as a different penalty parameter for beam and solid violates momentum conservation and I prefer not to have that in the main evaluation routine of the penalty terms.

We still have to think about a good way to test this feature.

[Copilot]

Pull request overview

Adds independent augmentation scaling for beam vs. solid DOFs in beam-to-solid volume meshtying by introducing two new input parameters and applying them during augmented Lagrange penalty regularization assembly.

Changes:

• Introduce AUGMENTATION_SCALING_PARAMETER_BEAM / ..._SOLID parameters and plumb them through beam-to-solid params.
• Expose beam/solid DOF row maps from the mortar manager to support selective scaling.
• In indirect assembly, assemble penalty regularization into temporary objects, apply row-wise scaling, and accumulate into the global system.

Reviewed changes

Copilot reviewed 11 out of 11 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
src/core/linalg/src/sparse/4C_linalg_fevector.hpp	Inline-implements FEVector::multiply(...) via Epetra call wrapper.
src/beaminteraction/src/contact/beam_to_solid/4C_beaminteraction_contact_beam_to_solid_params_base.hpp	Adds getters + members for beam/solid augmentation scaling parameters.
src/beaminteraction/src/contact/beam_to_solid/4C_beaminteraction_contact_beam_to_solid_params_base.cpp	Initializes + reads new scaling parameters from input list.
src/beaminteraction/src/contact/beam_to_solid/4C_beaminteraction_contact_beam_to_solid_mortar_manager.hpp	Adds accessors for beam and solid DOF row maps.
src/beaminteraction/src/contact/beam_to_solid/4C_beaminteraction_contact_beam_to_solid_input.cpp	Registers new input parameters with defaults and descriptions.
src/beaminteraction/src/contact/4C_beaminteraction_contact_submodel_evaluator_assembly_manager_indirect.hpp	Extends assembly manager interface to accept meshtying params.
src/beaminteraction/src/contact/4C_beaminteraction_contact_submodel_evaluator_assembly_manager_indirect.cpp	Applies beam/solid-specific scaling to penalty regularization contributions before adding to system.
src/beaminteraction/src/contact/4C_beaminteraction_contact_submodel_evaluator_assembly_manager_direct.hpp	Extends direct assembly manager interface to accept meshtying params.
src/beaminteraction/src/contact/4C_beaminteraction_contact_submodel_evaluator_assembly_manager_direct.cpp	Updates signature to match new interface (params currently unused here).
src/beaminteraction/src/contact/4C_beaminteraction_contact_submodel_evaluator_assembly_manager.hpp	Updates base virtual interface for evaluate_force_stiff(...).
src/beaminteraction/src/contact/4C_beaminteraction_contact_submodel_evaluator.cpp	Passes meshtying params through to assembly managers.

[isteinbrecher]

@maxfirmbach Thanks for the changes!

Do you have an idea how we can actually test this feature? I think with the current test, different values of the scaling parmeters will lead to a passing test, since in the converged state these values don't matter. Should we maybe test all the "augmented" test cases with a single newton step and ConstDis predictor? (Of course it would be nice if we could simply test the global tangent matrix, but for now the other approach should also work).

[isteinbrecher]

Should we add a check at the beginning of this if clause, that this is not a nullptr?

[maxfirmbach]

Yes, I guess this makes sense.

[maxfirmbach]

@maxfirmbach Thanks for the changes!

Do you have an idea how we can actually test this feature? I think with the current test, different values of the scaling parmeters will lead to a passing test, since in the converged state these values don't matter. Should we maybe test all the "augmented" test cases with a single newton step and ConstDis predictor? (Of course it would be nice if we could simply test the global tangent matrix, but for now the other approach should also work).

Yeah, not sure yet ... we could do it that way.