[Meshtying] cannot handle Dirichlet BCs at interface in new time integration

[georghammerl]

This issue was originally created by @mayrmt in https://gitlab.lrz.de/baci/baci/-/issues/149 long time ago

Expectations

Dirichlet boundary conditions can be applied at the meshtying interface.

Current Behavior

The input allows for imposing Dirichlet BCs (i.e. prescribe DBC on master side of the interface, leave slave side unconstrained). However, the nonlinear solver does not converge. After a few Newton steps, the STRUCTURE-F-Norm stagnates.

Motivation and Context

Imposition of DBCs on interfaces has been possible in the old time integration framework. Since this is an essential capability, we want to have it available in the new structure framework as well.

At first glance, I see two possible sources for this error:

• The residual vector used for convergence checking does not reflect the DBCs properly.
• DBCs are applied during a call to runPreApplyJacobianInverse() on the structure model evaluator. Afterwards, the meshtying condensation is performed in the meshtying model evaluators runPreApplyJacobianInverse(). This step might mess with the boundary conditions that have been previously applied.

Definition of Done

• Enable imposition of DBCs on meshtying interfaces.
• Add test case to cover that functionality.
• Double check if imposition of DBCs on slave interface is still throwing an error.

Possible Solution

Related Issues

Additional Information

As far as I know, we always apply DBCs first during the call to runPreApplyJacobianInverse() on the structure model evaluator and then do various modifications to the linear system through runPreApplyJacobianInverse() of all other model evaluators. We do not have control or any specific strategy to define the order of all these calls to runPreApplyJacobianInverse().

Interested Parties

@mayrmt