Skip to content
GitLab
Projects
Groups
Snippets
Help
Loading...
Help
What's new
7
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in
Toggle navigation
Open sidebar
Aland, Sebastian
amdis
Commits
32dfc015
Commit
32dfc015
authored
Aug 08, 2012
by
Praetorius, Simon
Browse files
Options
Browse Files
Download
Email Patches
Plain Diff
tutorial updated
parent
a7553081
Changes
1
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
with
164 additions
and
0 deletions
+164
-0
extensions/tutorial.hpp
extensions/tutorial.hpp
+164
-0
No files found.
extensions/tutorial.hpp
View file @
32dfc015
...
...
@@ -706,8 +706,172 @@ refinement.refine(...) call the mesh is globally refined up to the <c>initial le
function and especially the interface can be represented by the mesh.
*/
//-----------------------------------------------------------
/*! \page boundary_conditions Boundary conditions
For a second order elliptic partial differential equation one has to describe the geometry and prescribe boundary conditions to close
the system. Three types of boundary conditions are very typical: Dirichlet-type, Neumann-type or the Robin-type boundary condition and
the periodic boundary conditions. AMDiS provides interfaces for all these types. In the macro-mesh one has to mark several edges as
boundary edges. During refinement the mark is handed down to the leaf elements. That's the crux of the matter. Boundary conditions are
defined edge-wise, but what to do if two edges with different boundary conditions collide? In the macro-mesh one has to set for each
separate boundary part a boundary number. Depending on this number the one or other condition is preferred for the colliding vertex.
For problems with more than one equation on the same macro-mesh one can construct examples that are solvable with this strategy. Also
there are problem with the periodic boundary condition and mixed finite elements. Neumann boundary conditions in complicated parts of
a boundary need also special consideration. Here some new techniques are introduced to overcome some of the problems.
- Dirichlet boundary conditions
- singular points
- implicit areas
- Periodic boundary conditions
- Implicit Neumann condition
\section dirichlet_bc Dirichlet boundary conditions
There are three methods to enforce Dirichlet boundary condition, i.e. \f$ u = g\; on\; \Gamma \f$. To use these you have to replace the
standard ProblemStat by ExtendedProblemStat. The methods <c>addSingularDirichletBC(...)</c> add a condition at only one single degree
of freedom. For the second method <c>addImplicitDirichletBC(...)</c> you have to pass a function, that has negative values in the
vertices where to enforce the boundary condition, e.g. a signed distance function. The last method <c>addManualDirichletBC(...)</c>
provides an interface to describe parts of a boundary easily.
\subsection singular_dirichlet_bc Singular Dirichlet boundary condition
For example we want to solve the Neumann problem, we have to define an additional condition, since \f$ u \f$ is only defined up to a constant.
\f[ \Delta u = f\quad in\; \Omega,\quad\partial_n u = 0\quad on\; \partial\Omega \f]
One way out of this dilemma is to set a Dirichlet boundary condition on just one point. There are two ways to do this: first you can give a
dofindex for which to set the value, second you can specify a coordinate and the algorithm looks for the nearest degree of freedom to set
the boundary condition there.
\code
template<typename ValueContainer>
void addSingularDirichletBC(DegreeOfFreedom idx, int row, int col,
ValueContainer &values);
template<typename ValueContainer>
void addSingularDirichletBC(WorldVector<double> &pos, int row, int col,
ValueContainer &values);
\endcode
You have to the row and column where to set the boundary condition in the system of differential equations and as last argument you have to give
the value that should be enforced. The <c>ValueContainer</c> can ether be scalar reference, i.e. <c>double&</c>, an
<c>AbstractFunction< double, WorldVector<double> ></c>, or a <c>DOFVector< double ></c>.
The next example shows how to use this methods:
\code
int main(int argc, char* argv[])
{
AMDiS::init(argc, argv);
// ===== create and init the scalar problem =====
ExtendedProblemStat ellipt("ellipt");
ellipt.initialize(INIT_ALL);
// === create adapt info ===
AdaptInfo adaptInfo("ellipt->adapt", ellipt.getNumComponents());
AdaptStationary adapt("ellipt->adapt", ellipt, adaptInfo);
// ===== create matrix operator =====
Operator matrixOperator(ellipt.getFeSpace());
matrixOperator.addTerm(new Simple_SOT);
ellipt.addMatrixOperator(matrixOperator, 0, 0);
// ===== create rhs operator =====
Operator rhsOperator(ellipt.getFeSpace());
rhsOperator.addTerm(new Simple_ZOT(1.0));
ellipt.addVectorOperator(rhsOperator, 0);
// ===== add singular dirichlet boundaty condition =====
WorldVector<double> p; p.set(0.0);
double zero = 0.0;
ellipt.addSingularDirichletBC(p, 0, 0, zero);
// ===== start simulation =====
adapt.adapt();
ellipt.writeFiles(adaptInfo, true);
AMDiS::finalize();
}
\endcode
The solution will look like
\image html elliptSingular.png "Singular Dirichlet condition in the bottom left corner" width=300px
\subsection implicit_dirichlet_bc Implicit Dirichlet boundary condition
Assume you want to solve an equation in a "complicated" domain, that can be described by a signed distance function. Then
you can use a Diffuse-Domain approach or the implicit dirichlet boundary conditions defined in ExtendedProblemStat.h. Two
interfaces are available. You can set a DOFVector that holds the values of the signed distance function, or you can specify
an AbstractFunction:
\code
template<typename ValueContainer>
void addImplicitDirichletBC(DOFVector<double> &signedDist,
int row, int col,
ValueContainer &values);
template<typename ValueContainer>
void addImplicitDirichletBC(AbstractFunction<double, WorldVector<double> > &signedDist,
int row, int col,
ValueContainer &values);
\endcode
Similar to the singular Dirichlet conditions the value container can be a scalar, an AbstractFunction, or a DOFVector. In the next Example
the signed distance functions describes a disk area. At the boundary the Dirichlet condition should be enforced. The algorithm now not only
set the Dirichlet values at the boundary, but in the whole area with distance values less than zero. So a simple algebraic equation is solved
in the outer part of the complicated domain. Near the boundary the mesh should be refined, since one wants to resolve the boundary vertices
very near to the implicit interface.
\code
int main(int argc, char* argv[])
{
AMDiS::init(argc, argv);
// ===== create and init the scalar problem =====
ExtendedProblemStat ellipt("ellipt");
ellipt.initialize(INIT_ALL);
// === create adapt info ===
AdaptInfo adaptInfo("ellipt->adapt", ellipt.getNumComponents());
AdaptStationary adapt("ellipt->adapt", ellipt, adaptInfo);
// ===== signedDist function that describes the geometry =====
AbstractFunction<double, WorldVector<double> > *signedDistFct = new InverseCircle(1.0);
// ===== refine mesh at interface =====
SignedDistRefinement refFunction(ellipt.getMesh());
RefinementLevelCoords2 refinement(
ellipt.getFeSpace(),
&refFunction,
signedDistFct);
refinement.refine(9);
// ===== create matrix operator =====
Operator matrixOperator(ellipt.getFeSpace());
matrixOperator.addTerm(new Simple_SOT);
ellipt.addMatrixOperator(matrixOperator, 0, 0);
// ===== create rhs operator =====
Operator rhsOperator(ellipt.getFeSpace());
rhsOperator.addTerm(new Simple_ZOT(1.0));
ellipt.addVectorOperator(rhsOperator, 0);
// ===== add boundary conditions =====
double zero = 0.0;
ellipt.addImplicitDirichletBC(*signedDistFct, 0, 0, zero);
// ===== start simulation =====
adapt.adapt();
ellipt.writeFiles(adaptInfo, true);
AMDiS::finalize();
}
\endcode
The solution is shown in the next pictures. On the left one can see the mesh locally refined on the boundary of a circle:
\image html elliptImplicit.png "Implicit Dirichlet condition at the boundary of a circle" width=300px
*/
#endif // AMDIS_EXTENSIONS_TUTORIAL_INCLUDE
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment