ML_FiniteElements Namespace Reference

Default namespace for ML finite element example. More...

Classes
class	ML_FiniteElements::AbstractGrid
	Abstract interface to access finite element grids. More...
class	ML_FiniteElements::AbstractProblem
	Abstract interface to define linear problems. More...
class	ML_FiniteElements::AbstractQuadrature
	Interfaces for quadrature over elements. More...
class	ML_FiniteElements::AbstractVariational
	Pure virtual class that defines the variational form. More...
class	ML_FiniteElements::GalerkinVariational< T >
	Defines a pure Galerkin variational form of a scalar PDE. More...
class	ML_FiniteElements::HexCubeGrid
	Creates a grid composed by hexahedra in a cube. More...
class	ML_FiniteElements::HexQuadrature
	Quadrature formula on hexahedra. More...
class	ML_FiniteElements::LinearProblem
	Basic implementation of scalar finite element problem. More...
class	ML_FiniteElements::QuadQuadrature
	Quadrature formula on quadrilaterals. More...
class	ML_FiniteElements::QuadRectangleGrid
	Creates a grid with quadrilaterals on a rectangle. More...
class	ML_FiniteElements::SUPGVariational< T >
	SUPG discretization of an advection-diffusion PDE. More...
class	ML_FiniteElements::TetCubeGrid
	Creates a grid with tetrahedral elements in a cube. More...
class	ML_FiniteElements::TetQuadrature
	Quadrature formula on tetrahedra. More...
class	ML_FiniteElements::TRIANGLEGrid
class	ML_FiniteElements::TriangleQuadrature
	Quadrature formula on triangles. More...
class	ML_FiniteElements::TriangleRectangleGrid
	Creates a grid composed by triangles, the domain is a rectangle. More...
Variables
const int	ML_INTERNAL = 0
const int	ML_BOTTOM = -1
const int	ML_RIGHT = -2
const int	ML_TOP = -3
const int	ML_LEFT = -4
const int	ML_FRONT = -5
const int	ML_REAR = -6
const int	ML_DIRICHLET = 1000
const int	ML_NEUMANN = 1001

---

Detailed Description

Default namespace for ML finite element example.

The ML_FiniteElements namespace contains all the classes and the examples that show a possible structure for a finite element code using Epetra and ML.

Scalar second-order, symmetric and non-symmetric PDEs of type

can be discretized using this example. Neumann boundary conditions require minimal changes to the code. must be a 2D or a 3D domain, discretized using triangles, quadrilaterals, tetrahedra or hexahedra. The code can be quite easily extended to tackle vector problems, using the same finite element space for all unknowns.

Two discretizations are available:

• GalerkinVariational is a pure Galerkin approach. Since this approach is unstable for large Peclet numbers, the class defines b = 0.
• SUPGVariational is a SUPG discretization, only for triangles and tetrahedra, and coth formula for tau. It should be used for advective problems.

The code is based on the following pure virtual classes:

• AbstractGrid defines the query methods that each grid class must implement. AbstractGrid is based on the getrow() concept: instead of prescribing a given format, the user must define a set of methods, that will be used by the code to query for the vertices in a local elements, the coordinates of each vertex, the boundary faces, and so on.
• AbstractQuadrature defines the basis and test functions and the quadrature strategy.
• AbstractVariational defines a set of abstract interfaces for the variational form. An implementation of this class constructs the elemental finite element matrix and right-hand side.
• AbstractProblem takes an AbstractGrid, an AbstractVariational, and an AbstractQuadrature in input, then builds the global finite element matrix, and imposes the boundary conditions as specified by the user. It can also compute the norm of the numerical solution, exact solution, and error, using H1, semi-H1 or L2 norms.

The solution can be visualized using MEDIT (see web page http://www.ann.jussieu.fr/~frey/logiciels/medit.html for details and download).

A nice feature of this small example is that some grids can be created on-the-fly, with no need of input files. The following domains can be triangulated:

• In 2D, a rectangle, using classes TriangleRectangleGrid and QuadRectangleGrid.
• In 3D, a cube, using classes TetCubeGrid and HexCubeGrid.

These classes are valid for both serial and parallel runs.

An interface to triangle (see http://www-2.cs.cmu.edu/~quake/triangle.html for more details and download) is under development; the interested user can modify the file ml_TRIANGLEGrid.h. This interface is currently working for serial runs only.

The interested user should look to the following examples, all located in ml/examples/FiniteElements:

• Laplacian3D.cpp describes how to solve the Laplace equation in a unitary cube, discretized using hexahedra. After solution, the norm of the numerical solution, exact solution and error is computed, and the numerical solution is visualized using MEDIT.
• AdvDiff2D.cpp describes how to solve an advection-diffusion problem in a 2D domain, using SUPG.
• MLAPI shows how to interface with the MLAPI classes.
• TestAll is for testing purposed only.

Author:

Marzio Sala, SNL 9214.

Date:

Last updated on Apr-05.

---