Application handling matrices and assembling the linear system to solve the sensitivity equations. More...

#include <optimization_block_matrix_application.h>

Inheritance diagram for AppFrame::OptimizationBlockMatrixApplication< dim >:

Collaboration diagram for AppFrame::OptimizationBlockMatrixApplication< dim >:

Public Member Functions
	OptimizationBlockMatrixApplication (boost::shared_ptr< AppFrame::ApplicationData > data=boost::shared_ptr< AppFrame::ApplicationData >())
	Constructor for an object, owning its own mesh and dof handler.

	OptimizationBlockMatrixApplication (AppFrame::DoFApplication< dim > &, bool triangulation_only)
	Constructor for an object, borrowing mesh and dof handler from another object.

	~OptimizationBlockMatrixApplication ()
	Destructor.

virtual void	declare_parameters (ParameterHandler &param)
	Declare all parameters that are needed for:

virtual void	set_parameters (const std::vector< std::string > &name_dvar, const std::vector< double > &value_dvar, ParameterHandler &param)
	Function called by optimization loop in order to set the values in the ParameterHandler to the new design parameters.

virtual void	initialize (ParameterHandler &param)
	Initialize application.

virtual void	init_solution (AppFrame::FEVector &src)=0
	Initialize nonlinear solution.

virtual void	responses (std::vector< double > &f, const AppFrame::FEVectors &vectors)
	Post-processing.

virtual void	check_responses ()
	This class is called by responses to make sure that all responses requested are implemented in either cell_responses, global_responses or face_responses.

virtual void	cell_responses (std::vector< double > &resp, const typename DoFApplication< dim >::CellInfo &info, const AppFrame::FEVector &sol)
	This class is used to evaluate all the functionals that require looping over cells.

virtual void	global_responses (std::vector< double > &resp, const AppFrame::FEVector &sol)
	This class is used to evaluate all responses that do not require looping over cells.

virtual void	print_responses (std::vector< double > &resp)
	This function is used to print the responses to screen (deallog)

virtual void	dresponses_dl (std::vector< std::vector< double > > &df_dl, const AppFrame::FEVectors &src)
	Post-processing.

virtual void	cell_dresponses_dl (std::vector< std::vector< double > > &cell_df_dl, const typename DoFApplication< dim >::CellInfo &info, const AppFrame::FEVector &src)
	This class is used to evaluate the derivative of all the functionals that require looping over cells with respect to the design variables.

virtual void	global_dresponses_dl (std::vector< std::vector< double > > &df_dl, const AppFrame::FEVector &src)
	This class is used to evaluate the sensitivities of all responses that do not require looping over cells with respect of the design variables.

virtual void	dresponses_du (std::vector< AppFrame::FEVector > &dst, const AppFrame::FEVectors &src)
	Loop over all cells and assemble the vector \[ \frac{\partial f_m}{\partial u_i} \] that is used to solve the sensitivity equations by using the local matrix integration functions cell_dfunctional_du(), bdry_dfunctional_dlu() and edge_dfunctinal_du() provided by the derived class.

virtual void	cell_dresponses_du (std::vector< AppFrame::FEVector > &df_du, const typename DoFApplication< dim >::CellInfo &info, std::vector< std::vector< double > > &src)
	Integration of local bilinear form.

virtual void	global_dresponses_du (std::vector< AppFrame::FEVector > &df_du, const AppFrame::FEVector &src)
	This class is used to evaluate the sensitivities of all responses that do not require looping over cells with respect of the design variables.

virtual void	dresidual_dlambda (std::vector< AppFrame::FEVector > &dst, const AppFrame::FEVectors &src)
	Loop over all cells and assemble the vector \[ \frac{\partial R_n}{\partial \lambda_k} \] that is used to solve the sensitivity equations by using the local matrix integration functions cell_dresidual_dlambda(), bdry_dresidual_dlambda() and edge_dresidual_dlambda() provided by the derived class.

virtual void	cell_dresidual_dlambda (std::vector< AppFrame::FEVector > &cell_vector, const typename DoFApplication< dim >::CellInfo &cell, std::vector< std::vector< double > > &src)
	Integration of local bilinear form.

void	solve_direct (std::vector< std::vector< double > > &df_dl, const AppFrame::FEVectors &sol)
	Solver in order to obtain the analytical sensitivities for the given objective and constraints using the direct method.

void	solve_adjoint (std::vector< std::vector< double > > &df_dl, const AppFrame::FEVector &sol)
	Solver in order to obtain the analytical sensitivities for the given objective and constraints using the adjoint method.

unsigned int	get_n_resp () const
	Member function that returns the number of responses.

unsigned int	get_n_dvar () const
	Member function that returns the number of design variables.

std::vector< std::string >	get_name_dvar () const
	Member function that returns the name of the design variables.

std::vector< std::string >	get_name_responses () const
	Member function that returns the name of the responses.

void	print_default_parameter_file ()
	Print the default parameter handler file.

bool	get_bool_transfer_solution ()
	Function to see if we are transferring a solution on a refined grid to the initial coarse grid.

virtual void	set_read_initial_solution (bool &read)

void	set_optimization_parameters (unsigned int &n_dvar, unsigned int &n_resp, std::vector< std::string > &name_design_var, std::vector< std::string > &name_responses)
	This routine assigns the value of n_dvar, n_resp, name_design_var and name_responses to the internal variables n_dvar, n_resp, name_design_var and name_response.

void	set_output_variables (std::vector< std::string > &dakota_name_responses)

Public Member Functions inherited from AppFrame::BlockMatrixApplication< dim >
void	do_assemble (const FEVectors &, typename DoFHandler< dim >::active_cell_iterator begin, typename DoFHandler< dim >::active_cell_iterator end, Threads::Mutex &mutex)
	Multithreaded assemble function called by assemble().

virtual void	post_cell_assemble ()
	A call back function for assemble(), called after all cell matrices have been entered, but before the face matrices are computed.

	BlockMatrixApplication (boost::shared_ptr< ApplicationData > data=boost::shared_ptr< ApplicationData >())
	Constructor for an object, owning its own mesh and dof handler.

	BlockMatrixApplication (DoFApplication< dim > &, bool triangulation_only)
	Constructor for an object, borrowing mesh and dof handler from another object.

void	_initialize (ParameterHandler &param)
	Initialize data of this class.

void	add_vector_for_cell_matrix (std::string name, unsigned int block, unsigned int components, bool values, bool derivatives)
	Add a named vector from data to be handed to the cell_matrix() function via CellInfo::values and CellInfo::derivatives.

void	add_vector_for_flux_matrix (std::string name, unsigned int block, unsigned int components, bool values, bool derivatives)
	Add a named vector from data to be handed to the bdry_matrix() and face_matrix() functions via FaceInfo::values and FaceInfo::derivatives.

void	remesh_matrices ()
	Initialize sparsity patterns and matrices for the new mesh.

virtual void	remesh ()
	Refine grid accordingly to the Refinement options under "Grid Generation" in the parameter file.

void	assemble (const FEVectors &)
	Loop over all cells and assemble the system #matrices by using the local matrix integration functions cell_matrix(), bdry_matrix() and face_matrix() provided by the derived class.

void	assemble_numerically (const FEVectors &src, const double delta=1e-6)
	Compute the Jacobian of the system of equations, \[ J(i,j) = \frac{\partial R_i}{\partial u_j} \] by using forward differences.

void	residual_constraints (FEVector &dst) const
	Redefinition of residual_constraints() in DoFHandler.

virtual void	cell_matrix (MatrixVector &cell_matrices, const typename DoFApplication< dim >::CellInfo &cell)
	Integration of local bilinear form.

virtual void	bdry_matrix (MatrixVector &face_matrices, const typename DoFApplication< dim >::FaceInfo &face)
	Integration of local bilinear form.

virtual void	face_matrix (MatrixVector &matrices11, MatrixVector &matrices12, MatrixVector &matrices21, MatrixVector &matrices22, const typename DoFApplication< dim >::FaceInfo &face1, const typename DoFApplication< dim >::FaceInfo &face2)
	Integration of local bilinear form.

virtual void	dirichlet_bc (std::map< unsigned int, double > &boundary_values) const

Public Member Functions inherited from AppFrame::DoFApplication< dim >
	DoFApplication (boost::shared_ptr< ApplicationData > data=boost::shared_ptr< ApplicationData >(), bool multigrid=false)
	Constructor for an object, owning its own mesh and dof handler.

	DoFApplication (bool multigrid)
	Constructor for an object owning its own mesh and dof handler and creating new ApplicationData.

	DoFApplication (DoFApplication< dim > &, bool triangulation_only, bool multigrid=false)
	Constructor for an object, borrowing mesh and dof handler from another object.

	~DoFApplication ()
	Destructor which deletes owned objects.

void	_initialize (ParameterHandler &param)
	Initialize from parameter values.

void	add_vector_for_transfer (FEVector *)
	Add the vector to be transfered from one mesh to the next.

void	delete_vector_for_transfer ()
	Delete the vector to be transfered from one mesh to the next.

virtual void	remesh_dofs ()
	Initialize dof handler, count the dofs in each block and renumber the dofs.

virtual void	init_vector (FEVector &dst) const
	Reinitialize the BlockVector dst such that it contains block_size.size() blocks.

virtual double	evaluate (const FEVectors &)
	Evaluate a functional during postprocessing such as drag or lift on an aerodynamics problem.

virtual double	estimate (const FEVectors &src)
	Estimate the error.

virtual double	residual (FEVector &dst, const FEVectors &src, bool apply_boundaries=true)
	Loop over all cells to compute the residual.

unsigned int	memory_consumption () const
	Compute the amount of memory needed by this object.

virtual double	cell_estimate (const CellInfo &)
	Integration of estimate inside a cell.

virtual double	bdry_estimate (const FaceInfo &)
	Integration of estimate on a boundary face.

virtual double	face_estimate (const FaceInfo &, const FaceInfo &)
	Integration of estimate on an interior face.

virtual void	cell_residual (FEVector &cell_vector, const CellInfo &cell)
	Integration of local linear form.

virtual void	bdry_residual (FEVector &face_vector, const FaceInfo &face)
	Integration of local linear form.

virtual void	face_residual (FEVector &face_vector1, FEVector &face_vector2, const FaceInfo &face1, const FaceInfo &face2)
	Integration of local linear form.

template<class BOX , class WORKER >
void	integrate_1form (BOX &box, WORKER &local_forms, FEVector &dst, const FEVectors &src)
	The function integrating 1-forms as for instance used by residual().

void	integrate_functional (LocalEstimate< dim > &local_forms, FEVectors &dst, const FEVectors &src)
	The function integrating functionals using local integrators.

void	store_triangulation (Triangulation< dim > &new_tr)
	Function to copy a triangulation object for use after refinement.

void	transfer_solution_to_coarse_mesh (Triangulation< dim > &tr_coarse, AppFrame::FEVector &coarse_solution, AppFrame::FEVector &refined_solution)
	Function to perform the transfer of a solution on a refined grid to the initial coarse grid.

void	read_init_solution (AppFrame::FEVector &dst, bool &good_solution) const
	Function to transfer a solution on a refined grid to the initial coarse grid.

template<class INFO , typename TYPE >
void	fill_local_data (const INFO &info, const std::vector< VectorSelector > &data_vector, std::vector< std::vector< std::vector< TYPE > > > &data) const
	Fill local data vector with values of the stored finite element function vectors.

virtual void	grid_out (const std::string &basename)
	Output the grid used to solve the problem.

virtual void	data_out (const std::string &basename, const FEVectors &src)

virtual void	data_out (const std::string &basename, const FEVectors &src, const std::vector< std::string > &solution_names)

virtual void	data_out (const std::string &basename, const FEVector &solution, const std::vector< std::string > &solution_names, const FEVector &postprocessing=FEVector(), const std::vector< std::string > &postprocessing_names=std::vector< std::string >())
	This function outputs the results of a computation.

Public Member Functions inherited from AppFrame::ApplicationBase
	ApplicationBase (boost::shared_ptr< ApplicationData > ex_data=boost::shared_ptr< ApplicationData >())
	Constructor for an application.

	ApplicationBase (const ApplicationBase &other)
	Copy constructor.

virtual	~ApplicationBase ()
	Virtual destructor.

void	print_parameters_to_file (ParameterHandler &param, const std::string &file_name, const ParameterHandler::OutputStyle &style)
	Print default parameters for the application to a file.

virtual void	start_vector (FEVector &dst, std::string caller) const
	Initialize vector to problem size.

virtual void	solve (FEVector &start, const FEVectors &rhs)
	Solve the system assembled with right hand side `rhs` and return the result in `start`.

virtual void	Tsolve (FEVector &start, const FEVectors &rhs)
	Solve the dual system assembled with right hand side `rhs` and return the result in `start`.

virtual void	grid_out (const std::string &filename) const
	Write the mesh in the format specified by the ParameterHandler.

virtual void	data_out (const std::string &filename, const FEVectors &src, const std::vector< std::string >)

boost::shared_ptr < ApplicationData >	get_data ()
	Get access to the protected variable data.

const boost::shared_ptr < ApplicationData >	get_data () const
	Get read-only access to the protected variable data.

virtual std::string	id () const
	Return a unique identification string for this application.

virtual void	notify (const Event &reason)
	Add a reason for assembling.

virtual void	clear ()
	Reset the application class such that a call to initialize() is possible again and will produce the same result as if the object was fresh.

void	clear_events ()
	Clear all notifications.

Static Public Attributes
static const AppFrame::Event	sensitivity_analysis
	The Event set by OptimizationMatrixApplication if if we want to compute the sensitivities and a new matrix should be assembled.

Protected Types
typedef void(*	CELL_Dvalues )(std::vector< AppFrame::FEVector > &, const typename DoFApplication< dim >::CellInfo &, std::vector< std::vector< double > > &)
	Definition of a pointer to a function that is called in a loop over cells to compute the derivatives of either residual or responses with respect to design variables.

Protected Member Functions
void	print_dresponses_dl (std::vector< std::vector< double > > pdf_pdl)
	Auxiliary routine to print the values df_dl This routine should be called once df_df is assembled.

void	print_dresponses_du (std::vector< AppFrame::FEVector > df_du)
	Auxiliary routine to print the values of df_du This routine should be called once df_du is assembled.

void	dfunction (std::vector< AppFrame::FEVector > &dst, const AppFrame::FEVectors &src, bool dfunctional_du, bool dresidual_dlambda)
	This is an auxiliary function called by dresidual_dlambda and dfunctional_du.

Protected Attributes
unsigned int	n_dvar
	Number of design variables.

unsigned int	n_obj
	Number of objective functions.

unsigned int	n_resp
	Number of responses = n_obj + n_con.

std::vector< std::string >	name_design_var
	Member that stores the name of the design variables.

std::vector< std::string >	name_responses
	Member that stores the name of the responses, i.e.

std::vector< std::string >	name_output_var
	Member that stores the name of the output variables, These names will be written to name_responses if optimization is not being used and ignored otherwise.

bool	output_coarse_solution
	Decision variable for whether the solution is to be output for transfer.

bool	read_in_initial_solution
	Decision variable for whether the initial solution should be read from file.

bool	optimization
	Decision variable for whether the application is being used for optimization.

Protected Attributes inherited from AppFrame::BlockMatrixApplication< dim >
Table< 2, DoFTools::Coupling >	cell_couplings
	Couplings through cell bilinear forms.

Table< 2, DoFTools::Coupling >	flux_couplings
	Couplings through flux bilinear forms.

boost::shared_ptr< Quadrature < dim > >	quadrature_assemble_cell
	Quadrature rule for matrix assembling on cells.

boost::shared_ptr< Quadrature < dim-1 > >	quadrature_assemble_face
	Quadrature rule for matrix assembling on faces.

BlockSparseMatrix< double >	matrix
	Storage for the actual matrix.

FilteredMatrix< BlockVector < double > >	filtered_matrix
	Wrapper for the actual matrix that contains all the boundary conditions.

Protected Attributes inherited from AppFrame::DoFApplication< dim >
std::map< unsigned int, double >	boundary_constraints
	List of all nodes constrained by a strong boundary condition, together with a value to be assigned.

ConstraintMatrix	hanging_node_constraints
	Constraint Matrix object.

boost::shared_ptr< Mapping< dim > >	mapping
	The mapping used for the transformation of mesh cells.

unsigned int	mapping_degree
	Degree used for polynomial mapping of arbitrary order.

boost::shared_ptr < Triangulation< dim > >	tr
	Pointer to the Triangulation object.

boost::shared_ptr < FiniteElement< dim > >	element
	The finite element used in dof.

boost::shared_ptr< DoFHandler < dim > >	dof
	Pointer to the DoFHandler object.

boost::shared_ptr < MGDoFHandler< dim > >	mg_dof
	Pointer to the MGDoFHandler object.

MeshWorker::BlockInfo	block_info

Vector< float >	cell_errors
	The result of error estimation by cell.

Vector< float >	face_errors
	The result of error estimation by face.

GridOut	g_out
	The object for writing grids.

DataOut< dim, DoFHandler< dim > >	d_out
	The object for writing data.

std::vector < DataComponentInterpretation::DataComponentInterpretation >	data_interpretation
	Vector for adjusting output to vector data.

std::string	refinement
	Refinement parameter from parameter file.

unsigned int	initial_refinement
	Initial refinement from parameter file.

double	refinement_threshold
	Refinement threshold for adaptive method from parameter file.

double	coarsening_threshold
	Coarsening threshold for adaptive method from parameter file.

bool	sort_cuthill
	Flag for sorting with Cuthill McKee algorithm.

Point< dim >	sort_direction
	Direction for downstream sorting.

std::vector< FEVector * >	transfer_vectors
	List of vector names to be transfered from one grid to the next.

Quadrature< dim >	quadrature_residual_cell
	Quadrature rule for residual computation on cells.

Quadrature< dim-1 >	quadrature_residual_bdry
	Quadrature rule for residual computation on boundary faces.

Quadrature< dim-1 >	quadrature_residual_face
	Quadrature rule for residual computation on faces.

bool	boundary_fluxes
	Extend the integration loops in assemble() and residual() also to boundary faces.

bool	interior_fluxes
	Extend the integration loops in assemble() and residual() also to interior faces.

Protected Attributes inherited from AppFrame::ApplicationBase
boost::shared_ptr < ApplicationData >	data
	Object for auxiliary data.

Event	notifications
	Accumulate reasons for reassembling here.

Additional Inherited Members
Public Types inherited from AppFrame::DoFApplication< dim >
typedef MeshWorker::InfoObjects::IntegrationInfo < dim, FEValuesBase< dim > >	CellInfo
	This is a redefinition of CellInfo in order to call MeshWorker objects instead of the previous implementation in AppFrame.

typedef MeshWorker::InfoObjects::IntegrationInfo < dim, FEFaceValuesBase< dim > >	FaceInfo
	This is a redefinition of CellInfo in order to call MeshWorker objects instead of the previous implementation in AppFrame.

Public Attributes inherited from AppFrame::DoFApplication< dim >
unsigned int	verbosity
	Controls verbosity of certain functions.

std::vector < DataComponentInterpretation::DataComponentInterpretation >	solution_interpretations
	`solution_interpretations` identifies whether some `solution_names` are scalars or parts of a vector.

std::vector < DataComponentInterpretation::DataComponentInterpretation >	postprocessing_interpretations
	`postprocessing_interpretations` identifies whether some `postprocessing_names` are scalars or parts of a vector.

bool	output_materials_and_levels
	`output_materials_and_levels` if `true` then visualized, otherwise suppressed.

Vector< double >	output_materials
	Vector that will be used by `data_out` to store the material ids.

Vector< double >	output_levels
	Vector that will be used by `data_out` to store the refinement levels at each cell.

Detailed Description

template<int dim>
class AppFrame::OptimizationBlockMatrixApplication< dim >

Application handling matrices and assembling the linear system to solve the sensitivity equations.

This class inherits the information on Triangulation and DoFHandler from its base class BlockMatrixApplication. It adds information on the linear system for sensitivity analysis

Author: Marc Secanell, 2006

Member Typedef Documentation

template<int dim>

typedef void(* AppFrame::OptimizationBlockMatrixApplication< dim >::CELL_Dvalues)(std::vector< AppFrame::FEVector > &, const typename DoFApplication< dim >::CellInfo &, std::vector< std::vector< double > > &)

protected

Definition of a pointer to a function that is called in a loop over cells to compute the derivatives of either residual or responses with respect to design variables.

Constructor & Destructor Documentation

template<int dim>

AppFrame::OptimizationBlockMatrixApplication< dim >::OptimizationBlockMatrixApplication ( boost::shared_ptr< AppFrame::ApplicationData > data = boost::shared_ptr< AppFrame::ApplicationData >() )

Constructor for an object, owning its own mesh and dof handler.

If data is null, a new handler for application data is generated.

see constructor of DoFApplication.

template<int dim>

AppFrame::OptimizationBlockMatrixApplication< dim >::OptimizationBlockMatrixApplication	(	AppFrame::DoFApplication< dim > &	,
		bool	triangulation_only
	)

Constructor for an object, borrowing mesh and dof handler from another object.

If triangulation_only is true, only the triangulation is borrowed.

see constructor of DoFApplication.

template<int dim>

AppFrame::OptimizationBlockMatrixApplication< dim >::~OptimizationBlockMatrixApplication ( )

inline

Destructor.

Member Function Documentation

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::cell_dresidual_dlambda	(	std::vector< AppFrame::FEVector > &	cell_vector,
		const typename DoFApplication< dim >::CellInfo &	cell,
		std::vector< std::vector< double > > &	src
	)

virtual

Integration of local bilinear form.

Reimplemented in FuelCell::Application::AppCathode< dim >.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::cell_dresponses_dl	(	std::vector< std::vector< double > > &	cell_df_dl,
		const typename DoFApplication< dim >::CellInfo &	info,
		const AppFrame::FEVector &	src
	)

virtual

This class is used to evaluate the derivative of all the functionals that require looping over cells with respect to the design variables.

This class is called by responses to evaluate the response at each cell.

Reimplemented in FuelCell::Application::AppCathode< dim >, and FuelCell::Application::AppPemfc< dim >.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::cell_dresponses_du	(	std::vector< AppFrame::FEVector > &	df_du,
		const typename DoFApplication< dim >::CellInfo &	info,
		std::vector< std::vector< double > > &	src
	)

virtual

Integration of local bilinear form.

Reimplemented in FuelCell::Application::AppCathode< dim >, and FuelCell::Application::AppPemfc< dim >.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::cell_responses	(	std::vector< double > &	resp,
		const typename DoFApplication< dim >::CellInfo &	info,
		const AppFrame::FEVector &	sol
	)

virtual

This class is used to evaluate all the functionals that require looping over cells.

This class is called by responses to evaluate the response at each cell.

Reimplemented in FuelCell::Application::AppCathode< dim >, FuelCell::Application::AppPemfc< dim >, FuelCell::Application::AppLaplace< dim >, and FuelCell::Application::AppReadMesh< dim >.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::check_responses ( )

virtual

This class is called by responses to make sure that all responses requested are implemented in either cell_responses, global_responses or face_responses.

Reimplemented in FuelCell::Application::AppCathode< dim >, and FuelCell::Application::AppPemfc< dim >.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::declare_parameters ( ParameterHandler & param )

virtual

Declare all parameters that are needed for:

the computation of the equation coefficients
the control of the linear system solution
...

Reimplemented from AppFrame::BlockMatrixApplication< dim >.

Reimplemented in FuelCell::Application::AppCathode< dim >, FuelCell::Application::AppPemfc< dim >, FuelCell::Application::AppReadMesh< dim >, and FuelCell::Application::AppLaplace< dim >.

Referenced by AppFrame::OptimizationBlockMatrixApplication< dim >::print_default_parameter_file().

Here is the caller graph for this function:

template<int dim>

void AppFrame::OptimizationBlockMatrixApplication< dim >::dfunction	(	std::vector< AppFrame::FEVector > &	dst,
		const AppFrame::FEVectors &	src,
		bool	dfunctional_du,
		bool	dresidual_dlambda
	)

protected

This is an auxiliary function called by dresidual_dlambda and dfunctional_du.

Since both functions have to do the same thing, loop over the cells, but they call a different member function for cell assembly, I use a flag to change from one to the other and then I call this generic function. With this dresidual_dlambda and dfunctional_du have the same interface and the changes happen to a protected class.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::dresidual_dlambda	(	std::vector< AppFrame::FEVector > &	dst,
		const AppFrame::FEVectors &	src
	)

virtual

Loop over all cells and assemble the vector

\[ \frac{\partial R_n}{\partial \lambda_k} \]

that is used to solve the sensitivity equations by using the local matrix integration functions cell_dresidual_dlambda(), bdry_dresidual_dlambda() and edge_dresidual_dlambda() provided by the derived class.

The vector index refers to k and the FEVector index refers to n.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::dresponses_dl	(	std::vector< std::vector< double > > &	df_dl,
		const AppFrame::FEVectors &	src
	)

virtual

Post-processing.

Evaluate the partial derivative of all functionals necessary such as the objective function and the constraints of the optimization problem In this member class we obtain the term:

\[ \frac{\partial f_m}{\partial \lambda_k} \]

Note: size of df_dl is for f num_objectives + num_constraints and for lambda = num_design_var

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::dresponses_du	(	std::vector< AppFrame::FEVector > &	dst,
		const AppFrame::FEVectors &	src
	)

virtual

Loop over all cells and assemble the vector

\[ \frac{\partial f_m}{\partial u_i} \]

that is used to solve the sensitivity equations by using the local matrix integration functions cell_dfunctional_du(), bdry_dfunctional_dlu() and edge_dfunctinal_du() provided by the derived class.

The vector index refers to m and the BlockVector index refers to i.

template<int dim>

bool AppFrame::OptimizationBlockMatrixApplication< dim >::get_bool_transfer_solution ( )

inline

Function to see if we are transferring a solution on a refined grid to the initial coarse grid.

References AppFrame::OptimizationBlockMatrixApplication< dim >::output_coarse_solution.

template<int dim>

unsigned int AppFrame::OptimizationBlockMatrixApplication< dim >::get_n_dvar ( ) const

Member function that returns the number of design variables.

template<int dim>

unsigned int AppFrame::OptimizationBlockMatrixApplication< dim >::get_n_resp ( ) const

Member function that returns the number of responses.

template<int dim>

std::vector<std::string> AppFrame::OptimizationBlockMatrixApplication< dim >::get_name_dvar ( ) const

inline

Member function that returns the name of the design variables.

References AppFrame::OptimizationBlockMatrixApplication< dim >::name_design_var.

template<int dim>

std::vector<std::string> AppFrame::OptimizationBlockMatrixApplication< dim >::get_name_responses ( ) const

inline

Member function that returns the name of the responses.

References AppFrame::OptimizationBlockMatrixApplication< dim >::name_responses.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::global_dresponses_dl	(	std::vector< std::vector< double > > &	df_dl,
		const AppFrame::FEVector &	src
	)

virtual

This class is used to evaluate the sensitivities of all responses that do not require looping over cells with respect of the design variables.

An example of one of this types of constraints is the solid volume fraction.

Reimplemented in FuelCell::Application::AppCathode< dim >, and FuelCell::Application::AppPemfc< dim >.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::global_dresponses_du	(	std::vector< AppFrame::FEVector > &	df_du,
		const AppFrame::FEVector &	src
	)

virtual

This class is used to evaluate the sensitivities of all responses that do not require looping over cells with respect of the design variables.

An example of one of this types of constraints is the solid volume fraction.

Reimplemented in FuelCell::Application::AppCathode< dim >, and FuelCell::Application::AppPemfc< dim >.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::global_responses	(	std::vector< double > &	resp,
		const AppFrame::FEVector &	sol
	)

virtual

This class is used to evaluate all responses that do not require looping over cells.

An example of one of this types of constraints is the solid volume fraction.

Reimplemented in FuelCell::Application::AppCathode< dim >, FuelCell::Application::AppPemfc< dim >, FuelCell::Application::AppLaplace< dim >, and FuelCell::Application::AppReadMesh< dim >.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::init_solution ( AppFrame::FEVector & src )

pure virtual

Initialize nonlinear solution.

Implemented in FuelCell::Application::AppCathode< dim >, FuelCell::Application::AppPemfc< dim >, FuelCell::Application::AppLaplace< dim >, and FuelCell::Application::AppReadMesh< dim >.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::initialize ( ParameterHandler & param )

virtual

Initialize application.

Reimplemented from AppFrame::BlockMatrixApplication< dim >.

Reimplemented in FuelCell::Application::AppCathode< dim >, FuelCell::Application::AppPemfc< dim >, FuelCell::Application::AppLaplace< dim >, and FuelCell::Application::AppReadMesh< dim >.

template<int dim>

void AppFrame::OptimizationBlockMatrixApplication< dim >::print_default_parameter_file ( )

inline

Print the default parameter handler file.

References AppFrame::OptimizationBlockMatrixApplication< dim >::declare_parameters().

Here is the call graph for this function:

template<int dim>

void AppFrame::OptimizationBlockMatrixApplication< dim >::print_dresponses_dl ( std::vector< std::vector< double > > pdf_pdl )

inlineprotected

Auxiliary routine to print the values df_dl This routine should be called once df_df is assembled.

This is done on solve_direct

References AppFrame::OptimizationBlockMatrixApplication< dim >::n_dvar, and AppFrame::OptimizationBlockMatrixApplication< dim >::n_resp.

template<int dim>

void AppFrame::OptimizationBlockMatrixApplication< dim >::print_dresponses_du ( std::vector< AppFrame::FEVector > df_du )

inlineprotected

Auxiliary routine to print the values of df_du This routine should be called once df_du is assembled.

This is done on solve_direct

References AppFrame::OptimizationBlockMatrixApplication< dim >::n_resp.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::print_responses ( std::vector< double > & resp )

virtual

This function is used to print the responses to screen (deallog)

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::responses	(	std::vector< double > &	f,
		const AppFrame::FEVectors &	vectors
	)

virtual

Post-processing.

Evaluate all responses, usually functionals, necessary such as the objective function and the constraints of the optimization problem

Note: size of f is num_objectives + num_constraints

template<int dim>

void AppFrame::OptimizationBlockMatrixApplication< dim >::set_optimization_parameters	(	unsigned int &	n_dvar,
		unsigned int &	n_resp,
		std::vector< std::string > &	name_design_var,
		std::vector< std::string > &	name_responses
	)

This routine assigns the value of n_dvar, n_resp, name_design_var and name_responses to the internal variables n_dvar, n_resp, name_design_var and name_response.

template<int dim>

void AppFrame::OptimizationBlockMatrixApplication< dim >::set_output_variables ( std::vector< std::string > & dakota_name_responses )

inline

References AppFrame::OptimizationBlockMatrixApplication< dim >::name_responses.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::set_parameters	(	const std::vector< std::string > &	name_dvar,
		const std::vector< double > &	value_dvar,
		ParameterHandler &	param
	)

virtual

Function called by optimization loop in order to set the values in the ParameterHandler to the new design parameters.

Since ParameterHandler depends on the problem we are solving, set_parameters() is set at the most inner loop of the application.

Reimplemented in FuelCell::Application::AppCathode< dim >, FuelCell::Application::AppPemfc< dim >, and FuelCell::Application::AppLaplace< dim >.

template<int dim>

virtual void AppFrame::OptimizationBlockMatrixApplication< dim >::set_read_initial_solution ( bool & read )

inlinevirtual

References AppFrame::OptimizationBlockMatrixApplication< dim >::read_in_initial_solution.

template<int dim>

void AppFrame::OptimizationBlockMatrixApplication< dim >::solve_adjoint	(	std::vector< std::vector< double > > &	df_dl,
		const AppFrame::FEVector &	sol
	)

Solver in order to obtain the analytical sensitivities for the given objective and constraints using the adjoint method.

This member function calls the different assemble member functions and solves the system of equations

Note: The adjoint method is the most computationally efficient method if the number of number of design variables is larger than the number of objectives and constraints

template<int dim>

void AppFrame::OptimizationBlockMatrixApplication< dim >::solve_direct	(	std::vector< std::vector< double > > &	df_dl,
		const AppFrame::FEVectors &	sol
	)

Solver in order to obtain the analytical sensitivities for the given objective and constraints using the direct method.

This member function calls the different assemble member functions and solves the system of equations

Note: The direct method is the most computationally efficient method if the number of objectives and constraints is larger than the number of design variables

Member Data Documentation

template<int dim>

unsigned int AppFrame::OptimizationBlockMatrixApplication< dim >::n_dvar

protected

Number of design variables.

Referenced by AppFrame::OptimizationBlockMatrixApplication< dim >::print_dresponses_dl().

template<int dim>

unsigned int AppFrame::OptimizationBlockMatrixApplication< dim >::n_obj

protected

Number of objective functions.

template<int dim>

unsigned int AppFrame::OptimizationBlockMatrixApplication< dim >::n_resp

protected

Number of responses = n_obj + n_con.

Referenced by AppFrame::OptimizationBlockMatrixApplication< dim >::print_dresponses_dl(), and AppFrame::OptimizationBlockMatrixApplication< dim >::print_dresponses_du().

template<int dim>

std::vector<std::string> AppFrame::OptimizationBlockMatrixApplication< dim >::name_design_var

protected

Member that stores the name of the design variables.

Referenced by AppFrame::OptimizationBlockMatrixApplication< dim >::get_name_dvar().

template<int dim>

std::vector<std::string> AppFrame::OptimizationBlockMatrixApplication< dim >::name_output_var

protected

Member that stores the name of the output variables, These names will be written to name_responses if optimization is not being used and ignored otherwise.

template<int dim>

std::vector<std::string> AppFrame::OptimizationBlockMatrixApplication< dim >::name_responses

protected

Member that stores the name of the responses, i.e.

objectives and constraints.