PenaltyDirichletBC

Description

PenaltyDirichletBC is a NodalBC used for enforcing Dirichlet boundary conditions which differs from the DirichletBC class in the way in which it handles the enforcement. It is appropriate for partial differential equations (PDEs) in the form

\begin{aligned} -\nabla^2 u &= f && \quad \in \Omega \\ u &= g && \quad \in \partial \Omega_D \\ \frac{\partial u}{\partial n} &= h && \quad \in \partial \Omega_N \end{aligned}

Instead of imposing the Dirichlet condition directly on the basis by replacing the equations associated with those degrees of freedom (DOFs) by the auxiliary equation , the PenaltyDirichletBC is based on the variational statement: find such that (1) holds for every . In Eq. 2, is a user-selected parameter which must be taken small enough to ensure that on . The user-selectable class parameter penalty corresponds to , and must be chosen large enough to ensure good agreement with the Dirichlet data, but not so large that the resulting Jacobian becomes ill-conditioned, resulting in failed solves and overall accuracy losses.

Benefits of the penatly-based approach include simplified Dirichlet boundary condition enforcement for non-Lagrange finite element bases, maintaining the symmetry (if any) of the original problem, and avoiding the need to zero out contributions from other rows in a special post-assembly step. Integrating by parts "in reverse" from Eq. 2, one obtains

(2)

We therefore recover a "perturbed" version of the original problem with the flux boundary condition

(3)

replacing the original Dirichlet boundary condition. It has been shown Juntunen and Stenberg (2009) that in order for the solution to this perturbed problem to converge to the solution of the original problem in the limit as , the penalty parameter must depend on the mesh size, and that as we refine the mesh, the problem becomes increasingly ill-conditioned. A related method for imposing Dirichlet boundary conditions, known as Nitsche's method Juntunen and Stenberg (2009), does not suffer from the same ill-conditioning issues, and is slated for inclusion in MOOSE some time in the future.

References

1. M. Juntunen and R. Stenberg. Nitsche's method for general boundary conditions. Mathematics of Computation, 78(267):1353–1374, July 2009. URL: http://dx.doi.org/10.1090/S0025-5718-08-02183-2.[BibTeX]

   @article{juntunen2009nitsche,
       author = "Juntunen, M. and Stenberg, R.",
       title = "{Nitsche's method for general boundary conditions}",
       journal = "Mathematics of Computation",
       month = "July",
       year = "2009",
       volume = "78",
       number = "267",
       pages = "1353--1374",
       url = "http://dx.doi.org/10.1090/S0025-5718-08-02183-2"
   }
   

Example Input Syntax

[BCs]
  active = 'bc_all'
  [./bc_all]
    type = PenaltyDirichletBC
    variable = u
    value = 0
    boundary = 'top left right bottom'
    penalty = 1e5
  [../]
[]

Input Parameters