Detailed Description

Time discretized first order implicit quasi-linear ODE; to be solved using the Newton-Raphson method for resolving nonlinearities.

See also: ODESystemTag::FirstOrderImplicitQuasilinear

Definition at line 57 of file TimeDiscretizedODESystem.h.

#include <TimeDiscretizedODESystem.h>

Inheritance diagram for NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >:

[legend]

Collaboration diagram for NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >:

[legend]

Public Types
using	ODE = ODESystem< ODETag, NonlinearSolverTag::Newton >
	The type of ODE.

using	MatTrans = MatrixTranslator< ODETag >

using	TimeDisc = TimeDiscretization
	A shortcut for a general time discretization scheme.

Public Member Functions
	TimeDiscretizedODESystem (const int process_id, ODE &ode, TimeDisc &time_discretization)

	~TimeDiscretizedODESystem () override

void	assemble (std::vector< GlobalVector * > const &x_new_timestep, std::vector< GlobalVector * > const &x_prev, int const process_id) override

std::vector< GlobalIndexType >	getIndicesOfResiduumWithoutInitialCompensation () const override

void	getResidual (GlobalVector const &x_new_timestep, GlobalVector const &x_prev, GlobalVector &res) const override

void	getJacobian (GlobalMatrix &Jac) const override

void	computeKnownSolutions (GlobalVector const &x, int const process_id) override

void	applyKnownSolutions (GlobalVector &x) const override

void	applyKnownSolutionsNewton (GlobalMatrix &Jac, GlobalVector &res, GlobalVector &minus_delta_x) const override

void	updateConstraints (GlobalVector &lower, GlobalVector &upper, int const process_id) override

bool	isLinear () const override

void	preIteration (const unsigned iter, GlobalVector const &x) override

IterationResult	postIteration (GlobalVector const &x) override

TimeDisc &	getTimeDiscretization () override
	Exposes the used time discretization scheme.

MathLib::MatrixSpecifications	getMatrixSpecifications (const int process_id) const override

virtual TimeDiscretization &	getTimeDiscretization ()=0
	Exposes the used time discretization scheme.

Static Public Attributes
static const ODESystemTag	ODETag
	A tag indicating the type of ODE.

Private Types
using	Index = MathLib::MatrixVectorTraits< GlobalMatrix >::Index

Private Attributes
ODE &	_ode
	ode the ODE being wrapped

TimeDisc &	_time_disc
	the time discretization to being used

std::unique_ptr< MatTrans >	_mat_trans
	the object used to compute the matrix/vector for the nonlinear solver

std::vector< NumLib::IndexValueVector< Index > > const *	_known_solutions
	stores precomputed values for known solutions

GlobalMatrix *	_Jac
	the Jacobian of the residual

GlobalMatrix *	_M
	Matrix \( M \).

GlobalMatrix *	_K
	Matrix \( K \).

GlobalVector *	_b
	Matrix \( b \).

std::size_t	_Jac_id = 0u
	ID of the `_Jac` matrix.

std::size_t	_M_id = 0u
	ID of the `_M` matrix.

std::size_t	_K_id = 0u
	ID of the `_K` matrix.

std::size_t	_b_id = 0u
	ID of the `_b` vector.

std::size_t	_xdot_id = 0u
	ID of the vector storing xdot in intermediate computations.

std::vector< std::size_t >	_xdot_ids

Member Typedef Documentation

◆ Index

using NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::Index = MathLib::MatrixVectorTraits<GlobalMatrix>::Index

private

Definition at line 143 of file TimeDiscretizedODESystem.h.

◆ MatTrans

using NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::MatTrans = MatrixTranslator<ODETag>

The auxiliary class that computes the matrix/vector used by the nonlinear solver.

Definition at line 70 of file TimeDiscretizedODESystem.h.

◆ ODE

using NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::ODE = ODESystem<ODETag, NonlinearSolverTag::Newton>

The type of ODE.

Definition at line 67 of file TimeDiscretizedODESystem.h.

◆ TimeDisc

using NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::TimeDisc = TimeDiscretization

A shortcut for a general time discretization scheme.

Definition at line 72 of file TimeDiscretizedODESystem.h.

Constructor & Destructor Documentation

◆ TimeDiscretizedODESystem()

NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::TimeDiscretizedODESystem	(	const int	process_id,
		ODE &	ode,
		TimeDisc &	time_discretization
	)

explicit

Constructs a new instance.

Parameters

process_id	ID of the ODE to be solved.
ode	the ODE to be wrapped.
time_discretization	the time discretization to be used.

Definition at line 46 of file TimeDiscretizedODESystem.cpp.

    : _ode(ode),
      _time_disc(time_discretization),
      _mat_trans(createMatrixTranslator<ODETag>(time_discretization))
{
    _Jac = &NumLib::GlobalMatrixProvider::provider.getMatrix(
        _ode.getMatrixSpecifications(process_id), _Jac_id);
    _M = &NumLib::GlobalMatrixProvider::provider.getMatrix(
        _ode.getMatrixSpecifications(process_id), _M_id);
    _K = &NumLib::GlobalMatrixProvider::provider.getMatrix(
        _ode.getMatrixSpecifications(process_id), _K_id);
    _b = &NumLib::GlobalVectorProvider::provider.getVector(
        _ode.getMatrixSpecifications(process_id), _b_id);
}

References NumLib::MatrixProvider::getMatrix(), NumLib::VectorProvider::getVector(), NumLib::GlobalVectorProvider::provider, and NumLib::GlobalMatrixProvider::provider.

◆ ~TimeDiscretizedODESystem()

NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::~TimeDiscretizedODESystem ( )

override

Definition at line 65 of file TimeDiscretizedODESystem.cpp.

{
    NumLib::GlobalMatrixProvider::provider.releaseMatrix(*_Jac);
    NumLib::GlobalMatrixProvider::provider.releaseMatrix(*_M);
    NumLib::GlobalMatrixProvider::provider.releaseMatrix(*_K);
    NumLib::GlobalVectorProvider::provider.releaseVector(*_b);
}

References NumLib::GlobalVectorProvider::provider, NumLib::GlobalMatrixProvider::provider, NumLib::MatrixProvider::releaseMatrix(), and NumLib::VectorProvider::releaseVector().

Member Function Documentation

◆ applyKnownSolutions()

void NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::applyKnownSolutions ( GlobalVector & x ) const

override

Definition at line 160 of file TimeDiscretizedODESystem.cpp.

{
    ::detail::applyKnownSolutions(_known_solutions, x);
}

References detail::applyKnownSolutions().

◆ applyKnownSolutionsNewton()

void NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::applyKnownSolutionsNewton	(	GlobalMatrix &	Jac,
		GlobalVector &	res,
		GlobalVector &	minus_delta_x
	)		const

override

Definition at line 166 of file TimeDiscretizedODESystem.cpp.

{
    if (!_known_solutions)
    {
        return;
    }
 
    using IndexType = MathLib::MatrixVectorTraits<GlobalMatrix>::Index;
    std::vector<IndexType> ids;
    for (auto const& bc : *_known_solutions)
    {
        ids.insert(end(ids), begin(bc.ids), end(bc.ids));
    }
 
    // For the Newton method the values must be zero
    std::vector<double> values(ids.size(), 0);
    MathLib::applyKnownSolution(Jac, res, minus_delta_x, ids, values);
}

References MathLib::applyKnownSolution().

◆ assemble()

void NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::assemble	(	std::vector< GlobalVector * > const &	x_new_timestep,
		std::vector< GlobalVector * > const &	x_prev,
		int const	process_id
	)

override

Definition at line 74 of file TimeDiscretizedODESystem.cpp.

{
    namespace LinAlg = MathLib::LinAlg;
 
    auto const t = _time_disc.getCurrentTime();
    auto const dt = _time_disc.getCurrentTimeIncrement();
    auto const& x_curr = *x_new_timestep[process_id];
 
    std::vector<GlobalVector*> xdot(x_new_timestep.size());
    _xdot_ids.resize(x_new_timestep.size());
    for (std::size_t i = 0; i < xdot.size(); i++)
    {
        xdot[i] =
            &NumLib::GlobalVectorProvider::provider.getVector(_xdot_ids[i]);
        _time_disc.getXdot(*x_new_timestep[i], *x_prev[i], *xdot[i]);
    }
 
    _M->setZero();
    _K->setZero();
    _b->setZero();
    _Jac->setZero();
 
    _ode.preAssemble(t, dt, x_curr);
    try
    {
        _ode.assembleWithJacobian(t, dt, x_new_timestep, xdot, process_id, *_M,
                                  *_K, *_b, *_Jac);
    }
    catch (AssemblyException const&)
    {
        for (auto& v : xdot)
        {
            NumLib::GlobalVectorProvider::provider.releaseVector(*v);
        }
        throw;
    }
 
    LinAlg::finalizeAssembly(*_M);
    LinAlg::finalizeAssembly(*_K);
    LinAlg::finalizeAssembly(*_b);
    MathLib::LinAlg::finalizeAssembly(*_Jac);
 
    for (auto& v : xdot)
    {
        NumLib::GlobalVectorProvider::provider.releaseVector(*v);
    }
}

References MathLib::LinAlg::finalizeAssembly(), NumLib::VectorProvider::getVector(), NumLib::GlobalVectorProvider::provider, NumLib::VectorProvider::releaseVector(), and MathLib::EigenVector::setZero().

◆ computeKnownSolutions()

void NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::computeKnownSolutions	(	GlobalVector const &	x,
		int const	process_id
	)

override

Definition at line 151 of file TimeDiscretizedODESystem.cpp.

{
    _known_solutions =
        _ode.getKnownSolutions(_time_disc.getCurrentTime(), x, process_id);
}

◆ getIndicesOfResiduumWithoutInitialCompensation()

std::vector< GlobalIndexType > NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::getIndicesOfResiduumWithoutInitialCompensation ( ) const

inlineoverride

Returns: The global indices for the entries of the global residuum vector that do not need initial non-equilibrium compensation.

Definition at line 92 of file TimeDiscretizedODESystem.h.

    {
        return _ode.getIndicesOfResiduumWithoutInitialCompensation();
    }

◆ getJacobian()

void NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::getJacobian ( GlobalMatrix & Jac ) const

override

Definition at line 144 of file TimeDiscretizedODESystem.cpp.

{
    _mat_trans->computeJacobian(*_Jac, Jac);
}

◆ getMatrixSpecifications()

MathLib::MatrixSpecifications NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::getMatrixSpecifications ( const int process_id ) const

inlineoverride

Definition at line 130 of file TimeDiscretizedODESystem.h.

    {
        return _ode.getMatrixSpecifications(process_id);
    }

◆ getResidual()

void NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::getResidual	(	GlobalVector const &	x_new_timestep,
		GlobalVector const &	x_prev,
		GlobalVector &	res
	)		const

override

Definition at line 127 of file TimeDiscretizedODESystem.cpp.

{
    // TODO Maybe the duplicate calculation of xdot here and in assembleJacobian
    //      can be optimized. However, that would make the interface a bit more
    //      fragile.
    auto& xdot = NumLib::GlobalVectorProvider::provider.getVector(_xdot_id);
    _time_disc.getXdot(x_new_timestep, x_prev, xdot);
 
    _mat_trans->computeResidual(*_M, *_K, *_b, x_new_timestep, xdot, res);
 
    NumLib::GlobalVectorProvider::provider.releaseVector(xdot);
}

References NumLib::VectorProvider::getVector(), NumLib::GlobalVectorProvider::provider, and NumLib::VectorProvider::releaseVector().

◆ getTimeDiscretization()

TimeDisc & NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::getTimeDiscretization ( )

inlineoverridevirtual

Exposes the used time discretization scheme.

Implements NumLib::TimeDiscretizedODESystemBase< NonlinearSolverTag::Newton >.

Definition at line 129 of file TimeDiscretizedODESystem.h.

129{ return _time_disc; }

◆ isLinear()

bool NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::isLinear ( ) const

inlineoverride

Definition at line 117 of file TimeDiscretizedODESystem.h.

117{ return _ode.isLinear(); }

◆ postIteration()

IterationResult NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::postIteration ( GlobalVector const & x )

inlineoverride

Definition at line 124 of file TimeDiscretizedODESystem.h.

    {
        return _ode.postIteration(x);
    }

◆ preIteration()

void NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::preIteration	(	const unsigned	iter,
		GlobalVector const &	x
	)

inlineoverride

Definition at line 119 of file TimeDiscretizedODESystem.h.

    {
        _ode.preIteration(iter, x);
    }

◆ updateConstraints()

void NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::updateConstraints	(	GlobalVector &	lower,
		GlobalVector &	upper,
		int const	process_id
	)

inlineoverride

Definition at line 111 of file TimeDiscretizedODESystem.h.

    {
        _ode.updateConstraints(lower, upper, process_id);
    }

Member Data Documentation

◆ _b

GlobalVector* NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::_b

private

Matrix \( b \).

Definition at line 150 of file TimeDiscretizedODESystem.h.

◆ _b_id

std::size_t NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::_b_id = 0u

private

ID of the _b vector.

Definition at line 155 of file TimeDiscretizedODESystem.h.

◆ _Jac

GlobalMatrix* NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::_Jac

private

the Jacobian of the residual

Definition at line 147 of file TimeDiscretizedODESystem.h.

◆ _Jac_id

std::size_t NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::_Jac_id = 0u

private

ID of the _Jac matrix.

Definition at line 152 of file TimeDiscretizedODESystem.h.

◆ _K

GlobalMatrix* NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::_K

private

Matrix \( K \).

Definition at line 149 of file TimeDiscretizedODESystem.h.

◆ _K_id

std::size_t NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::_K_id = 0u

private

ID of the _K matrix.

Definition at line 154 of file TimeDiscretizedODESystem.h.

◆ _known_solutions

std::vector<NumLib::IndexValueVector<Index> > const* NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::_known_solutions

private

Initial value:

=

nullptr

stores precomputed values for known solutions

Definition at line 144 of file TimeDiscretizedODESystem.h.

◆ _M

GlobalMatrix* NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::_M

private

Matrix \( M \).

Definition at line 148 of file TimeDiscretizedODESystem.h.

◆ _M_id

std::size_t NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::_M_id = 0u

private

ID of the _M matrix.

Definition at line 153 of file TimeDiscretizedODESystem.h.

◆ _mat_trans

std::unique_ptr<MatTrans> NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::_mat_trans

private

the object used to compute the matrix/vector for the nonlinear solver

Definition at line 141 of file TimeDiscretizedODESystem.h.

◆ _ode

ODE& NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::_ode

private

ode the ODE being wrapped

Definition at line 137 of file TimeDiscretizedODESystem.h.

◆ _time_disc

TimeDisc& NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::_time_disc

private

the time discretization to being used

Definition at line 138 of file TimeDiscretizedODESystem.h.

◆ _xdot_id

std::size_t NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::_xdot_id = 0u

mutableprivate

ID of the vector storing xdot in intermediate computations.

Definition at line 158 of file TimeDiscretizedODESystem.h.

◆ _xdot_ids

std::vector<std::size_t> NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::_xdot_ids

mutableprivate

Definition at line 159 of file TimeDiscretizedODESystem.h.

◆ ODETag

const ODESystemTag NumLib::TimeDiscretizedODESystem< ODESystemTag::FirstOrderImplicitQuasilinear, NonlinearSolverTag::Newton >::ODETag

static

Initial value:

=

ODESystemTag::FirstOrderImplicitQuasilinear

NumLib::ODESystemTag::FirstOrderImplicitQuasilinear

@ FirstOrderImplicitQuasilinear

A tag indicating the type of ODE.

Definition at line 63 of file TimeDiscretizedODESystem.h.

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

NumLib/ODESolver/TimeDiscretizedODESystem.h
NumLib/ODESolver/TimeDiscretizedODESystem.cpp

Detailed Description

Public Types

Public Member Functions

Static Public Attributes

Private Types

Private Attributes

Member Typedef Documentation

◆ Index

◆ MatTrans

◆ ODE

◆ TimeDisc

Constructor & Destructor Documentation

◆ TimeDiscretizedODESystem()

◆ ~TimeDiscretizedODESystem()

Member Function Documentation

◆ applyKnownSolutions()

◆ applyKnownSolutionsNewton()

◆ assemble()

◆ computeKnownSolutions()

◆ getIndicesOfResiduumWithoutInitialCompensation()

◆ getJacobian()

◆ getMatrixSpecifications()

◆ getResidual()

◆ getTimeDiscretization()

◆ isLinear()

◆ postIteration()

◆ preIteration()

◆ updateConstraints()

Member Data Documentation

◆ _b

◆ _b_id

◆ _Jac

◆ _Jac_id

◆ _K

◆ _K_id

◆ _known_solutions

◆ _M

◆ _M_id

◆ _mat_trans

◆ _ode

◆ _time_disc

◆ _xdot_id

◆ _xdot_ids

◆ ODETag