Page Menu
Home
c4science
Search
Configure Global Search
Log In
Files
F91720032
boundary_condition_tmpl.hh
No One
Temporary
Actions
Download File
Edit File
Delete File
View Transforms
Subscribe
Mute Notifications
Award Token
Subscribers
None
File Metadata
Details
File Info
Storage
Attached
Created
Wed, Nov 13, 20:11
Size
10 KB
Mime Type
text/x-c
Expires
Fri, Nov 15, 20:11 (1 d, 23 h)
Engine
blob
Format
Raw Data
Handle
21289770
Attached To
rAKA akantu
boundary_condition_tmpl.hh
View Options
/**
* @file boundary_condition_tmpl.hh
*
* @author Dana Christen <dana.christen@gmail.com>
* @author Nicolas Richart <nicolas.richart@epfl.ch>
*
* @date creation: Fri May 03 2013
* @date last modification: Mon Jun 23 2014
*
* @brief XXX
*
* @section LICENSE
*
* Copyright (©) 2014 EPFL (Ecole Polytechnique Fédérale de Lausanne)
* Laboratory (LSMS - Laboratoire de Simulation en Mécanique des Solides)
*
* Akantu is free software: you can redistribute it and/or modify it under the
* terms of the GNU Lesser General Public License as published by the Free
* Software Foundation, either version 3 of the License, or (at your option) any
* later version.
*
* Akantu is distributed in the hope that it will be useful, but WITHOUT ANY
* WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
* A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
* details.
*
* You should have received a copy of the GNU Lesser General Public License
* along with Akantu. If not, see <http://www.gnu.org/licenses/>.
*
*/
/* -------------------------------------------------------------------------- */
#include "element_group.hh"
/* -------------------------------------------------------------------------- */
__BEGIN_AKANTU__
/* -------------------------------------------------------------------------- */
template<typename ModelType>
void BoundaryCondition<ModelType>::initBC(ModelType & model, Array<Real> & primal, Array<Real> & dual)
{
this->model = &model;
this->primal = &primal;
this->dual = &dual;
if(this->model->getSpatialDimension() > 1)
this->model->initFEEngineBoundary();
}
/* -------------------------------------------------------------------------- */
template<typename ModelType>
void BoundaryCondition<ModelType>::initBC(ModelType & model,
Array<Real> & primal,
Array<Real> & primal_increment,
Array<Real> & dual)
{
this->initBC(model, primal, dual);
this->primal_increment = &primal_increment;
}
/* -------------------------------------------------------------------------- */
/* Partial specialization for DIRICHLET functors */
template<typename ModelType>
template<typename FunctorType>
struct BoundaryCondition<ModelType>::TemplateFunctionWrapper<FunctorType, BC::Functor::_dirichlet> {
static inline void applyBC(const FunctorType & func,
const ElementGroup & group,
BoundaryCondition<ModelType> & bc_instance) {
ModelType & model = bc_instance.getModel();
Array<Real> & primal = bc_instance.getPrimal();
const Array<Real> & coords = model.getMesh().getNodes();
Array<bool> & boundary_flags = model.getBlockedDOFs();
UInt dim = model.getMesh().getSpatialDimension();
Array<Real>::vector_iterator primal_iter = primal.begin(primal.getNbComponent());
Array<Real>::const_vector_iterator coords_iter = coords.begin(dim);
Array<bool>::vector_iterator flags_iter = boundary_flags.begin(boundary_flags.getNbComponent());
for(ElementGroup::const_node_iterator nodes_it(group.node_begin());
nodes_it!= group.node_end();
++nodes_it) {
UInt n = *nodes_it;
Vector<bool> flag(flags_iter[n]);
Vector<Real> primal(primal_iter[n]);
Vector<Real> coords(coords_iter[n]);
func(n, flag, primal, coords);
}
}
};
/* -------------------------------------------------------------------------- */
/* Partial specialization for NEUMANN functors */
template<typename ModelType>
template<typename FunctorType>
struct BoundaryCondition<ModelType>::TemplateFunctionWrapper<FunctorType, BC::Functor::_neumann> {
static inline void applyBC(const FunctorType & func,
const ElementGroup & group,
BoundaryCondition<ModelType> & bc_instance) {
UInt dim = bc_instance.getModel().getSpatialDimension();
switch(dim) {
case 1: { AKANTU_DEBUG_TO_IMPLEMENT(); break; }
case 2:
case 3: {
applyBC(func, group, bc_instance, _not_ghost);
applyBC(func, group, bc_instance, _ghost);
break;
}
}
}
static inline void applyBC(const FunctorType & func,
const ElementGroup & group,
BoundaryCondition<ModelType> & bc_instance,
GhostType ghost_type) {
ModelType & model = bc_instance.getModel();
Array<Real> & dual = bc_instance.getDual();
const Mesh & mesh = model.getMesh();
const Array<Real> & nodes_coords = mesh.getNodes();
const FEEngine & fem_boundary = model.getFEEngineBoundary();
UInt dim = model.getSpatialDimension();
UInt nb_degree_of_freedom = dual.getNbComponent();
QuadraturePoint quad_point;
quad_point.ghost_type = ghost_type;
ElementGroup::type_iterator type_it = group.firstType(dim - 1, ghost_type);
ElementGroup::type_iterator type_end = group.lastType (dim - 1, ghost_type);
// Loop over the boundary element types
for(; type_it != type_end; ++type_it) {
const Array<UInt> & element_ids = group.getElements(*type_it, ghost_type);
Array<UInt>::const_scalar_iterator elem_iter = element_ids.begin();
Array<UInt>::const_scalar_iterator elem_iter_end = element_ids.end();
UInt nb_quad_points = fem_boundary.getNbQuadraturePoints(*type_it, ghost_type);
UInt nb_elements = element_ids.getSize();
UInt nb_nodes_per_element = mesh.getNbNodesPerElement(*type_it);
Array<Real> * dual_before_integ = new Array<Real>(nb_elements * nb_quad_points,
nb_degree_of_freedom,
0.);
Array<Real> * quad_coords = new Array<Real>(nb_elements * nb_quad_points, dim);
const Array<Real> & normals_on_quad =
fem_boundary.getNormalsOnQuadPoints(*type_it, ghost_type);
fem_boundary.interpolateOnQuadraturePoints(nodes_coords, *quad_coords,
dim,
*type_it, ghost_type,
element_ids);
Array<Real>::const_vector_iterator normals_begin = normals_on_quad.begin(dim);
Array<Real>::const_vector_iterator normals_iter;
Array<Real>::const_vector_iterator quad_coords_iter = quad_coords->begin(dim);
Array<Real>::vector_iterator dual_iter = dual_before_integ->begin(nb_degree_of_freedom);
quad_point.type = *type_it;
for(; elem_iter != elem_iter_end; ++elem_iter) {
UInt el = *elem_iter;
quad_point.element = el;
normals_iter = normals_begin + el * nb_quad_points;
for(UInt q(0); q < nb_quad_points; ++q) {
quad_point.num_point = q;
func(quad_point,
*dual_iter,
*quad_coords_iter,
*normals_iter);
++dual_iter;
++quad_coords_iter;
++normals_iter;
}
}
delete quad_coords;
/* -------------------------------------------------------------------- */
// Initialization of iterators
Array<Real>::matrix_iterator dual_iter_mat =
dual_before_integ->begin(nb_degree_of_freedom,1);
elem_iter = element_ids.begin();
Array<Real>::const_matrix_iterator shapes_iter_begin =
fem_boundary.getShapes(*type_it, ghost_type).begin(1, nb_nodes_per_element);
Array<Real> * dual_by_shapes =
new Array<Real>(nb_elements*nb_quad_points, nb_degree_of_freedom*nb_nodes_per_element);
Array<Real>::matrix_iterator dual_by_shapes_iter =
dual_by_shapes->begin(nb_degree_of_freedom, nb_nodes_per_element);
Array<Real>::const_matrix_iterator shapes_iter;
/* -------------------------------------------------------------------- */
// Loop computing dual x shapes
for(; elem_iter != elem_iter_end; ++elem_iter) {
shapes_iter = shapes_iter_begin + *elem_iter*nb_quad_points;
for(UInt q(0); q < nb_quad_points; ++q,
++dual_iter_mat, ++dual_by_shapes_iter, ++shapes_iter) {
dual_by_shapes_iter->mul<false, false>(*dual_iter_mat, *shapes_iter);
}
}
delete dual_before_integ;
Array<Real> * dual_by_shapes_integ =
new Array<Real>(nb_elements, nb_degree_of_freedom*nb_nodes_per_element);
fem_boundary.integrate(*dual_by_shapes,
*dual_by_shapes_integ,
nb_degree_of_freedom*nb_nodes_per_element,
*type_it,
ghost_type,
element_ids);
delete dual_by_shapes;
// assemble the result into force vector
fem_boundary.assembleArray(*dual_by_shapes_integ,
dual,
model.getDOFSynchronizer().getLocalDOFEquationNumbers(),
nb_degree_of_freedom,
*type_it,
ghost_type,
element_ids);
delete dual_by_shapes_integ;
}
}
};
/* -------------------------------------------------------------------------- */
template<typename ModelType>
template<typename FunctorType>
inline void BoundaryCondition<ModelType>::applyBC(const FunctorType & func) {
GroupManager::const_element_group_iterator bit = model->getMesh().getGroupManager().element_group_begin();
GroupManager::const_element_group_iterator bend = model->getMesh().getGroupManager().element_group_end();
for(; bit != bend; ++bit) applyBC(func, *bit);
}
/* -------------------------------------------------------------------------- */
template<typename ModelType>
template<typename FunctorType>
inline void BoundaryCondition<ModelType>::applyBC(const FunctorType & func,
const std::string & group_name) {
try {
const ElementGroup & element_group = model->getMesh().getElementGroup(group_name);
applyBC(func, element_group);
} catch(akantu::debug::Exception e) {
AKANTU_EXCEPTION("Error applying a boundary condition onto \""
<< group_name << "\"! [" << e.what() <<"]");
}
}
/* -------------------------------------------------------------------------- */
template<typename ModelType>
template<typename FunctorType>
inline void BoundaryCondition<ModelType>::applyBC(const FunctorType & func,
const ElementGroup & element_group) {
#if !defined(AKANTU_NDEBUG)
if(element_group.getDimension() != model->getSpatialDimension() - 1)
AKANTU_DEBUG_WARNING("The group " << element_group.getName()
<< " does not contain only boundaries elements");
#endif
TemplateFunctionWrapper<FunctorType>::applyBC(func, element_group, *this);
}
__END_AKANTU__
Event Timeline
Log In to Comment