DGOperatorImplicit< dim, number, is_viscous > Class Template Reference

Developer Documentation: MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous > Class Template Reference

Developer Documentation

Operator for the matrix-free evaluation of a compressible single-phase flow cutDG formulation for implicit time integration. More...

#include <dg_operator_implicit.hpp>

Inheritance diagram for MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >:

[legend]

Collaboration diagram for MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >:

[legend]

Public Types
using	VectorType = dealii::LinearAlgebra::distributed::Vector< number >

using	ConservedVariables = ConservedVariablesType< dim, number >

using	ConservedVariablesGradient = ConservedVariablesGradientType< dim, number >

Public Member Functions
	DGOperatorImplicit (OperationScratchData< dim, number > &flow_scratch_data)
	Constructor.

void	reinit () override
	Reinitialize the internal data structures.

void	advance_time_step (number time, number time_step) override
	Advances solver by a single time step.

void	add_external_force (std::shared_ptr< ExternalFlowForce< dim, number > > external_force_residuum, std::shared_ptr< ExternalFlowForceJacobian< dim, number > > external_force_jacobian) override

void	compute_system_matrix_from_matrixfree (dealii::TrilinosWrappers::SparseMatrix &sparse_matrix) const
	Compute the matrix representation of the Jacobian.

void	compute_inverse_diagonal_from_matrixfree (VectorType &diagonal) const
	Compute the inverse elements of the diagonal of the Jacobian.

void	apply_jacobian (number time_step, VectorType &dst, const VectorType &src) const
	Compute the result of J*x, where J is the Jacobian.

void	compute_residual (number current_time, number time_step, const VectorType &src, VectorType &dst, const VectorType &old_solution) const
	Compute the negative residual.

void	local_cell_jacobian_kernel (FECellIntegrator< dim, dim+2, number > &delta_phi, const FECellIntegrator< dim, dim+2, number > &phi, unsigned int q_index) const
	Local cell operations at the given quadrature point for computing the Jacobian.

void	local_face_jacobian_kernel (FEFaceIntegrator< dim, dim+2, number > &delta_phi_m, FEFaceIntegrator< dim, dim+2, number > &delta_phi_p, const FEFaceIntegrator< dim, dim+2, number > &phi_m, const FEFaceIntegrator< dim, dim+2, number > &phi_p, unsigned int q_index) const
	Local face operations at the given quadrature point for computing the Jacobian.

void	local_boundary_face_jacobian_kernel (FEFaceIntegrator< dim, dim+2, number > &delta_phi_m, const FEFaceIntegrator< dim, dim+2, number > &phi_m, unsigned int q_index) const
	Local boundary face operations at the given quadrature point for computing the Jacobian.

Public Member Functions inherited from MeltPoolDG::CompressibleFlow::DGOperatorBase< dim, number, n_species >
virtual	~DGOperatorBase ()=default

virtual void	add_external_force (std::shared_ptr< ExternalFlowForce< dim, number, n_species > > external_force_residuum, std::shared_ptr< ExternalFlowForceJacobian< dim, number, n_species > > external_force_jacobian)=0

Private Member Functions
void	apply_jacobian_analytic (const VectorType &src, VectorType &dst) const
	Compute the result of J*x, where J is the Jacobian computed analytically.

void	apply_jacobian_finite_differences (const VectorType &src, VectorType &dst) const
	Compute the result of J*x, where J is the Jacobian approximated by finite differences.

void	local_cell_residual (const dealii::MatrixFree< dim, number > &matrix_free, dealii::LinearAlgebra::distributed::Vector< number > &dst, const dealii::LinearAlgebra::distributed::Vector< number > &src, const std::pair< unsigned int, unsigned int > &cell_range) const
	The local cell applier computing the residual contribution of the cell.

void	local_face_residual (const dealii::MatrixFree< dim, number > &matrix_free, dealii::LinearAlgebra::distributed::Vector< number > &dst, const dealii::LinearAlgebra::distributed::Vector< number > &src, const std::pair< unsigned int, unsigned int > &face_range) const
	The local cell applier computing the residual contribution of the inner faces.

void	local_boundary_face_residual (const dealii::MatrixFree< dim, number > &matrix_free, dealii::LinearAlgebra::distributed::Vector< number > &dst, const dealii::LinearAlgebra::distributed::Vector< number > &src, const std::pair< unsigned int, unsigned int > &face_range) const
	The local cell applier computing the residual contribution of the boundary faces.

void	local_cell_jacobian (const dealii::MatrixFree< dim, number > &matrix_free, VectorType &dst, const VectorType &src, const std::pair< unsigned int, unsigned int > &cell_range) const
	Computes the cell contribution of the Jacobian.

void	local_face_jacobian (const dealii::MatrixFree< dim, number > &matrix_free, VectorType &dst, const VectorType &src, const std::pair< unsigned int, unsigned int > &face_range) const
	Computes the inner face contribution of the Jacobian.

void	local_boundary_face_jacobian (const dealii::MatrixFree< dim, number > &matrix_free, dealii::LinearAlgebra::distributed::Vector< number > &dst, const dealii::LinearAlgebra::distributed::Vector< number > &src, const std::pair< unsigned int, unsigned int > &face_range) const
	Computes the boundary face contribution of the Jacobian.

Private Attributes
const VectorType *	time_integrator_old_solution = nullptr

VectorType	disturbed_residual

number	current_time_step

OperationScratchData< dim, number > &	flow_scratch_data
	Scratch data for compressible flows.

TimeIntegration::BDFIntegrator< dim, number >	time_integrator
	Time integrator class used for the time integration.

ConvectiveKernels< dim, number >	convective_terms
	Object for the convective term evaluations.

ViscousKernels< dim, number >	viscous_terms
	Object for the viscous term evaluations.

std::vector< std::shared_ptr< ExternalFlowForce< dim, number > > >	external_forces_residual

std::vector< std::shared_ptr< ExternalFlowForceJacobian< dim, number > > >	external_forces_jacobian

Detailed Description

template<int dim, typename number, bool is_viscous = true>
class MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >

Operator for the matrix-free evaluation of a compressible single-phase flow cutDG formulation for implicit time integration.

Template Parameters

dim	Dimension of the considered simulation case.
number	Floating point format type.
is_viscous	Indicates whether the flow is viscous.

Member Typedef Documentation

◆ ConservedVariables

template<int dim, typename number , bool is_viscous = true>

using MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::ConservedVariables = ConservedVariablesType<dim, number>

◆ ConservedVariablesGradient

template<int dim, typename number , bool is_viscous = true>

using MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::ConservedVariablesGradient = ConservedVariablesGradientType<dim, number>

◆ VectorType

template<int dim, typename number , bool is_viscous = true>

using MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::VectorType = dealii::LinearAlgebra::distributed::Vector<number>

Constructor & Destructor Documentation

◆ DGOperatorImplicit()

template<int dim, typename number , bool is_viscous>

MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::DGOperatorImplicit ( OperationScratchData< dim, number > & flow_scratch_data )

explicit

Constructor.

Parameters

flow_scratch_data Reference to the flow scratch data object (usually owned by the corresponding operation class).

Member Function Documentation

◆ add_external_force()

template<int dim, typename number , bool is_viscous>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::add_external_force	(	std::shared_ptr< ExternalFlowForce< dim, number > >	external_force_residuum,
		std::shared_ptr< ExternalFlowForceJacobian< dim, number > >	external_force_jacobian
	)

override

◆ advance_time_step()

template<int dim, typename number , bool is_viscous>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::advance_time_step	(	number	time,
		number	time_step
	)

overridevirtual

Advances solver by a single time step.

This function performs a single implicit time step of size time_step starting from the solution at time time.

Note: The function does not take care about updating the solution history object or similar operations which are not directly related to the integration. It only advances the solution by a single time step starting from the current solution in the solution history object of the flow_scratch_data object.

Implements MeltPoolDG::CompressibleFlow::DGOperatorBase< dim, number, n_species >.

◆ apply_jacobian()

template<int dim, typename number , bool is_viscous>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::apply_jacobian	(	number	time_step,
		VectorType &	dst,
		const VectorType &	src
	)		const

Compute the result of J*x, where J is the Jacobian.

The method on how to compute/approximate the Jacobian is defined by the user in the compressible flow data.

Parameters

src	Source vector x with which the Jacobian gets multiplied.
dst	Location at which the result of J*x is stored.

Exceptions

Exception if the layout of the two given vectors src and dst are not identical.

Note: This function assumes that the function set_stage_constants() has been called in advance.

◆ apply_jacobian_analytic()

template<int dim, typename number , bool is_viscous>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::apply_jacobian_analytic	(	const VectorType &	src,
		VectorType &	dst
	)		const

private

Compute the result of J*x, where J is the Jacobian computed analytically.

Parameters

src	Source vector x with which the Jacobian gets multiplied.
dst	Location at which the result of J*x is stored.

Exceptions

Exception if the layout of the two given vectors src and dst are not identical.

◆ apply_jacobian_finite_differences()

template<int dim, typename number , bool is_viscous>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::apply_jacobian_finite_differences	(	const VectorType &	src,
		VectorType &	dst
	)		const

private

Compute the result of J*x, where J is the Jacobian approximated by finite differences.

Parameters

src	Source vector x with which the Jacobian gets multiplied.
dst	Location at which the result of J*x is stored.

Exceptions

Exception if the layout of the two given vectors src and dst are not identical.

◆ compute_inverse_diagonal_from_matrixfree()

template<int dim, typename number , bool is_viscous>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::compute_inverse_diagonal_from_matrixfree ( VectorType & diagonal ) const

Compute the inverse elements of the diagonal of the Jacobian.

Parameters

diagonal Vector in which the inverse elements of the diagonal are stored.

◆ compute_residual()

template<int dim, typename number , bool is_viscous>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::compute_residual	(	number	current_time,
		number	time_step,
		const VectorType &	src,
		VectorType &	dst,
		const VectorType &	old_solution
	)		const

Compute the negative residual.

Compute the negative residual, i.e. -(y'-F(y)) where y' is the temporal derivative of the primary variables and F is the sum of all fluxes occuring in the compressible Navier-Stokes equations (right-hand side).

Parameters

current_time	Current physical time.
src	Current solution vector used to compute the residual.
dst	Vector in which the residual is stored.

Exceptions

Assert if the layout of the two given vectors src and dst are not identical.

Note: This function assumes that the function set_stage_constants() has been called in advance.

◆ compute_system_matrix_from_matrixfree()

template<int dim, typename number , bool is_viscous>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::compute_system_matrix_from_matrixfree ( dealii::TrilinosWrappers::SparseMatrix & sparse_matrix ) const

Compute the matrix representation of the Jacobian.

Parameters

sparse_matrix Spars matrix in which the resulting matrix representation is stored.

◆ local_boundary_face_jacobian()

template<int dim, typename number , bool is_viscous = true>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::local_boundary_face_jacobian	(	const dealii::MatrixFree< dim, number > &	matrix_free,
		dealii::LinearAlgebra::distributed::Vector< number > &	dst,
		const dealii::LinearAlgebra::distributed::Vector< number > &	src,
		const std::pair< unsigned int, unsigned int > &	face_range
	)		const

private

Computes the boundary face contribution of the Jacobian.

Computes the contribution of the boundary faces to the product of the Jacobian and the provided source vector for a specified face range. The current solution of the primary variables, required for the Jacobian computation, is retrieved from the solution_history object.

Parameters

matrix_free	The matrix-free object utilized by the applier.
dst	The destination vector to which the computed result is added.
src	The source vector to be multiplied with the Jacobian.
face_range	The range of faces considered by the applier.

◆ local_boundary_face_jacobian_kernel()

template<int dim, typename number , bool is_viscous = true>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::local_boundary_face_jacobian_kernel	(	FEFaceIntegrator< dim, dim+2, number > &	delta_phi_m,
		const FEFaceIntegrator< dim, dim+2, number > &	phi_m,
		unsigned int	q_index
	)		const

Local boundary face operations at the given quadrature point for computing the Jacobian.

Parameters

delta_phi_m	Face integrator for the change in the primary variables on the inner face. Quadrature point distributions are added to this integrator.
phi_m	Cell integrator for the primary varibales on the inner face.
q_index	Quadrature point index.

◆ local_boundary_face_residual()

template<int dim, typename number , bool is_viscous = true>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::local_boundary_face_residual	(	const dealii::MatrixFree< dim, number > &	matrix_free,
		dealii::LinearAlgebra::distributed::Vector< number > &	dst,
		const dealii::LinearAlgebra::distributed::Vector< number > &	src,
		const std::pair< unsigned int, unsigned int > &	face_range
	)		const

private

The local cell applier computing the residual contribution of the boundary faces.

Parameters

matrix_free	Matrix free object on which the applier works on.
dst	Destination vector to which the result is added.
src	Current solution.
face_range	Face range which is considered in the applier.

◆ local_cell_jacobian()

template<int dim, typename number , bool is_viscous = true>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::local_cell_jacobian	(	const dealii::MatrixFree< dim, number > &	matrix_free,
		VectorType &	dst,
		const VectorType &	src,
		const std::pair< unsigned int, unsigned int > &	cell_range
	)		const

private

Computes the cell contribution of the Jacobian.

Computes the contribution of the cells to the product of the Jacobian and the provided source vector for a specified cell range. The current solution of the primary variables, required for the Jacobian computation, is retrieved from the solution_history object.

Parameters

matrix_free	The matrix-free object utilized by the applier.
dst	The destination vector to which the computed result is added.
src	The source vector to be multiplied with the Jacobian.
cell_range	The range of cells considered by the applier.

◆ local_cell_jacobian_kernel()

template<int dim, typename number , bool is_viscous>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::local_cell_jacobian_kernel	(	FECellIntegrator< dim, dim+2, number > &	delta_phi,
		const FECellIntegrator< dim, dim+2, number > &	phi,
		unsigned int	q_index
	)		const

Local cell operations at the given quadrature point for computing the Jacobian.

Parameters

delta_phi	Cell integrator for the change in the primary variables. Quadrature point distributions are added to this integrator.
phi	Cell integrator for the primary variables.
q_index	Quadrature point index.
cell_batch_id	ID of the current cell batch.

◆ local_cell_residual()

template<int dim, typename number , bool is_viscous = true>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::local_cell_residual	(	const dealii::MatrixFree< dim, number > &	matrix_free,
		dealii::LinearAlgebra::distributed::Vector< number > &	dst,
		const dealii::LinearAlgebra::distributed::Vector< number > &	src,
		const std::pair< unsigned int, unsigned int > &	cell_range
	)		const

private

The local cell applier computing the residual contribution of the cell.

Parameters

matrix_free	Matrix free object on which the applier works on.
dst	Destination vector to which the result is added.
src	Current solution.
cell_range	Cell range which is considered in the applier.

◆ local_face_jacobian()

template<int dim, typename number , bool is_viscous = true>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::local_face_jacobian	(	const dealii::MatrixFree< dim, number > &	matrix_free,
		VectorType &	dst,
		const VectorType &	src,
		const std::pair< unsigned int, unsigned int > &	face_range
	)		const

private

Computes the inner face contribution of the Jacobian.

Computes the contribution of the inner faces to the product of the Jacobian and the provided source vector for a specified face range. The current solution of the primary variables, required for the Jacobian computation, is retrieved from the solution_history object.

Parameters

matrix_free	The matrix-free object utilized by the applier.
dst	The destination vector to which the computed result is added.
src	The source vector to be multiplied with the Jacobian.
face_range	The range of faces considered by the applier.

◆ local_face_jacobian_kernel()

template<int dim, typename number , bool is_viscous = true>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::local_face_jacobian_kernel	(	FEFaceIntegrator< dim, dim+2, number > &	delta_phi_m,
		FEFaceIntegrator< dim, dim+2, number > &	delta_phi_p,
		const FEFaceIntegrator< dim, dim+2, number > &	phi_m,
		const FEFaceIntegrator< dim, dim+2, number > &	phi_p,
		unsigned int	q_index
	)		const

Local face operations at the given quadrature point for computing the Jacobian.

Parameters

delta_phi_m	Face integrator for the change in the primary variables on the inner face. Quadrature point distributions are added to this integrator.
delta_phi_p	Face integrator for the change in the primary variables on the outer face. Quadrature point distributions are added to this integrator.
phi_m	Cell integrator for the primary varibales on the inner face.
phi_p	Cell integrator for the primary varibales on the outer face.
q_index	Quadrature point index.

◆ local_face_residual()

template<int dim, typename number , bool is_viscous = true>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::local_face_residual	(	const dealii::MatrixFree< dim, number > &	matrix_free,
		dealii::LinearAlgebra::distributed::Vector< number > &	dst,
		const dealii::LinearAlgebra::distributed::Vector< number > &	src,
		const std::pair< unsigned int, unsigned int > &	face_range
	)		const

private

The local cell applier computing the residual contribution of the inner faces.

Parameters

matrix_free	Matrix free object on which the applier works on.
dst	Destination vector to which the result is added.
src	Current solution.
face_range	Face range which is considered in the applier.

◆ reinit()

template<int dim, typename number , bool is_viscous>

void MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::reinit ( )

overridevirtual

Reinitialize the internal data structures.

The reinitialization includes setting a new required size for the solution history object according to the demands of the used time integrator.

Implements MeltPoolDG::CompressibleFlow::DGOperatorBase< dim, number, n_species >.

Member Data Documentation

◆ convective_terms

template<int dim, typename number , bool is_viscous = true>

ConvectiveKernels<dim, number> MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::convective_terms

private

Object for the convective term evaluations.

◆ current_time_step

template<int dim, typename number , bool is_viscous = true>

number MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::current_time_step

mutableprivate

Current time step size. This needs to be stored as this value is required by the local cell appliers.

◆ disturbed_residual

template<int dim, typename number , bool is_viscous = true>

VectorType MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::disturbed_residual

mutableprivate

This vector is used in the approximation of the jacobian by finite differences. Here the vector stores the residual with a disturbed input.

◆ external_forces_jacobian

template<int dim, typename number , bool is_viscous = true>

std::vector<std::shared_ptr<ExternalFlowForceJacobian<dim, number> > > MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::external_forces_jacobian

private

This set of pointers may hold a list of external fluid force contributions to the jacobian (e.g., gravity, or user-defined source terms)

◆ external_forces_residual

template<int dim, typename number , bool is_viscous = true>

std::vector<std::shared_ptr<ExternalFlowForce<dim, number> > > MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::external_forces_residual

private

This set of pointers may hold a list of external fluid force contributions to the residuum (e.g., gravity, or user-defined source terms)

◆ flow_scratch_data

template<int dim, typename number , bool is_viscous = true>

OperationScratchData<dim, number>& MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::flow_scratch_data

private

Scratch data for compressible flows.

◆ time_integrator

template<int dim, typename number , bool is_viscous = true>

TimeIntegration::BDFIntegrator<dim, number> MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::time_integrator

private

Time integrator class used for the time integration.

◆ time_integrator_old_solution

template<int dim, typename number , bool is_viscous = true>

const VectorType* MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::time_integrator_old_solution = nullptr

mutableprivate

This vector is used to compute the temporal derivative y' in the residual computation by y'=(current_solution-old_solution)/dt. We do not take the old_solution directly from solution_history as providing the option to pass a (fictional) old solution can have performance advantages, e.g. in the bdf time integration. It can be set by the function set_stage_constants().

◆ viscous_terms

template<int dim, typename number , bool is_viscous = true>

ViscousKernels<dim, number> MeltPoolDG::CompressibleFlow::DGOperatorImplicit< dim, number, is_viscous >::viscous_terms

private

Object for the viscous term evaluations.

The documentation for this class was generated from the following files:

include/meltpooldg/compressible_flow/dg_operator_implicit.hpp
source/compressible_flow/dg_operator_implicit.cpp

DGOperatorImplicit&lt; dim, number, is_viscous &gt; Class Template Reference

Public Types

Public Member Functions

Private Member Functions

Private Attributes

Detailed Description

Member Typedef Documentation

◆ ConservedVariables

◆ ConservedVariablesGradient

◆ VectorType

Constructor & Destructor Documentation

◆ DGOperatorImplicit()

Member Function Documentation

◆ add_external_force()

◆ advance_time_step()

◆ apply_jacobian()

◆ apply_jacobian_analytic()

◆ apply_jacobian_finite_differences()

◆ compute_inverse_diagonal_from_matrixfree()

◆ compute_residual()

◆ compute_system_matrix_from_matrixfree()

◆ local_boundary_face_jacobian()

◆ local_boundary_face_jacobian_kernel()

◆ local_boundary_face_residual()

◆ local_cell_jacobian()

◆ local_cell_jacobian_kernel()

◆ local_cell_residual()

◆ local_face_jacobian()

◆ local_face_jacobian_kernel()

◆ local_face_residual()

◆ reinit()

Member Data Documentation

◆ convective_terms

◆ current_time_step

◆ disturbed_residual

◆ external_forces_jacobian

◆ external_forces_residual

◆ flow_scratch_data

◆ time_integrator

◆ time_integrator_old_solution

◆ viscous_terms

DGOperatorImplicit< dim, number, is_viscous > Class Template Reference