Integrator Class Reference

The Integrator class is basically a wrapper of the DomainBase class with regard to some status changing methods. More...

#include <Integrator.h>

Inheritance diagram for Integrator:

Collaboration diagram for Integrator:

Public Types
enum class	Type { Implicit , Explicit }

Public Member Functions
	Integrator (unsigned=0)
void	set_domain (const std::weak_ptr< DomainBase > &)
shared_ptr< DomainBase >	get_domain () const
virtual int	initialize ()
virtual Type	type () const
void	set_time_step_switch (bool)
bool	allow_to_change_time_step () const
void	set_matrix_assembled_switch ()
bool	matrix_is_assembled () const
virtual bool	time_independent_matrix () const
int	process_load ()
virtual int	process_constraint ()
int	process_criterion () const
int	process_modifier () const
int	process_load_resistance ()
virtual int	process_constraint_resistance ()
void	record () const
virtual void	assemble_resistance ()
virtual void	assemble_matrix ()
virtual void	assemble_effective_matrix ()
virtual vec	get_force_residual ()
virtual vec	get_displacement_residual ()
vec	get_auxiliary_residual () const
virtual sp_mat	get_reference_load ()
virtual const vec &	get_trial_displacement () const
void	update_load () const
void	update_constraint () const
void	update_trial_load_factor (double) const
void	update_trial_load_factor (const vec &) const
virtual void	update_from_ninja ()
void	update_trial_time (double)
virtual void	update_incre_time (double)
virtual int	update_trial_status (bool)
int	sync_status (bool)
virtual int	update_internal (const mat &)
mat	solve (const mat &)
mat	solve (const sp_mat &)
mat	solve (mat &&)
mat	solve (sp_mat &&)
virtual int	solve (mat &, const mat &)
virtual int	solve (mat &, const sp_mat &)
virtual int	solve (mat &, mat &&)
virtual int	solve (mat &, sp_mat &&)
void	erase_machine_error (vec &) const
void	stage_and_commit_status ()
void	stage_status () const
virtual void	commit_status ()
virtual void	clear_status ()
virtual void	reset_status ()
virtual vec	from_incre_velocity (const vec &, const uvec &)
virtual vec	from_incre_acceleration (const vec &, const uvec &)
virtual vec	from_total_velocity (const vec &, const uvec &)
virtual vec	from_total_acceleration (const vec &, const uvec &)
vec	from_incre_velocity (double, const uvec &)
vec	from_incre_acceleration (double, const uvec &)
vec	from_total_velocity (double, const uvec &)
vec	from_total_acceleration (double, const uvec &)
Public Member Functions inherited from UniqueTag
	UniqueTag (const UniqueTag &)=delete
	UniqueTag (UniqueTag &&)=delete
UniqueTag &	operator= (const UniqueTag &)=delete
UniqueTag &	operator= (UniqueTag &&)=delete
	~UniqueTag () override=default
	Tag (unsigned=0)
	Tag (const Tag &)=default
	Tag (Tag &&) noexcept=default
Public Member Functions inherited from Tag
	Tag (unsigned=0)
	Tag (const Tag &)=default
	Tag (Tag &&) noexcept=default
Tag &	operator= (const Tag &)=delete
Tag &	operator= (Tag &&)=delete
virtual	~Tag ()=default
void	set_tag (unsigned) const
unsigned	get_tag () const
void	enable ()
void	disable ()
void	guard ()
void	unguard ()
bool	is_active () const
bool	is_guarded () const
virtual void	print ()

Protected Member Functions
virtual void	update_parameter (double)
virtual int	process_load_impl (bool)
virtual int	process_constraint_impl (bool)
virtual bool	has_corrector () const
virtual int	correct_trial_status ()

Detailed Description

The Integrator class is basically a wrapper of the DomainBase class with regard to some status changing methods.

By default, the Step object calls DomainBase(Workshop) object to update displacement/resistance/stiffness independently. When it comes to dynamic analysis (time integration is involved), it is necessary to compute the equivalent load/stiffness by combining several quantities.

The Integrator object is acting like an agent between Workshop and Step, that can modify corresponding quantities to account for dynamic effect.

Author

tlc

Date

27/08/2017

Version

0.1.2

Member Enumeration Documentation

Type

enum class Integrator::Type

strong

Enumerator
Implicit
Explicit

Constructor & Destructor Documentation

Integrator()

Integrator::Integrator ( unsigned  T = 0 )

explicit

Member Function Documentation

allow_to_change_time_step()

bool Integrator::allow_to_change_time_step

(

)

const

Some time integration methods (multistep methods) require time step to be constant (for at least some consecutive steps). Call this method in solvers to determine whether it is allowed to change time step.

assemble_effective_matrix()

void Integrator::assemble_effective_matrix ( )

virtual

Assemble the global effective matrix A in AX=B. For FEM applications, it is often a linear combination of stiffness, mass, damping and geometry matrices.

Reimplemented in BatheTwoStep, GeneralizedAlpha, GSSSS, LeeNewmark, LeeNewmarkFull, LeeNewmarkIterative, Newmark, OALTS, WilsonPenzienNewmark, and ExplicitIntegrator.

assemble_matrix()

void Integrator::assemble_matrix ( )

virtual

Assemble global matrices such as stiffness, mass, damping and geometry matrices. This method should come with the companion method assemble_effective_matrix().

Reimplemented in LeeNewmarkIterative, NonviscousNewmark, ImplicitIntegrator, and ExplicitIntegrator.

Here is the caller graph for this function:

assemble_resistance()

void Integrator::assemble_resistance ( )

virtual

Reimplemented in GeneralizedAlphaExplicit, BatheTwoStep, GeneralizedAlpha, GSSSS, LeeNewmark, Newmark, NonviscousNewmark, OALTS, RayleighNewmark, WilsonPenzienNewmark, and ExplicitIntegrator.

Here is the caller graph for this function:

clear_status()

void Integrator::clear_status ( )

virtual

Reimplemented in LeeNewmarkBase, BatheExplicit, GERKN, BatheTwoStep, NonviscousNewmark, OALTS, and WilsonPenzienNewmark.

Here is the caller graph for this function:

commit_status()

void Integrator::commit_status ( )

virtual

Reimplemented in LeeNewmarkBase, BatheExplicit, GERKN, BatheTwoStep, NonviscousNewmark, OALTS, and WilsonPenzienNewmark.

Here is the caller graph for this function:

correct_trial_status()

int Integrator::correct_trial_status ( )

protectedvirtual

Correct the trial status. This method is called when the integrator has a corrector. It is used to correct the trial status after the computation. The default implementation does nothing and returns success. Override this method to implement the corrector.

Reimplemented in BatheExplicit, GeneralizedAlphaExplicit, GERKN, and GSSE.

Here is the caller graph for this function:

erase_machine_error()

void Integrator::erase_machine_error

(

vec &

ninja

)

const

Avoid machine error accumulation. The penalty method can apply homogeneous constraints approximately. The corresponding DoF shall be set to zero after solving the system.

Here is the call graph for this function:

from_incre_acceleration() [1/2]

vec Integrator::from_incre_acceleration ( const vec &  , const uvec &  encoding  )

virtual

When external loads are applied, they can be applied in forms of displacement/velocity/acceleration. The time integration methods, by default, form effective stiffness matrices in displacement domain. That is, in AX=B, A is the effective stiffness matrix and X is the displacement increment. Thus, loads in velocity/acceleration must be converted to displacement. This cannot be done arbitrarily due to compatibility issues. This method takes acceleration increment and converts it to TOTAL displacement.

Reimplemented in GeneralizedAlphaExplicit, GERKN, GSSE, BatheTwoStep, GeneralizedAlpha, GSSSS, Newmark, OALTS, and ExplicitIntegrator.

Here is the caller graph for this function:

from_incre_acceleration() [2/2]

vec Integrator::from_incre_acceleration	(	double	magnitude,
		const uvec &	encoding
	)

A simplified version similar to from_incre_acceleration(const vec&, const uvec&). It assumes all DoFs share the same magnitude.

Here is the call graph for this function:

from_incre_velocity() [1/2]

vec Integrator::from_incre_velocity ( const vec &  , const uvec &  encoding  )

virtual

When external loads are applied, they can be applied in forms of displacement/velocity/acceleration. The time integration methods, by default, form effective stiffness matrices in displacement domain. That is, in AX=B, A is the effective stiffness matrix and X is the displacement increment. Thus, loads in velocity/acceleration must be converted to displacement. This cannot be done arbitrarily due to compatibility issues. This method takes velocity increment and converts it to TOTAL displacement.

Reimplemented in BatheTwoStep, GeneralizedAlpha, GSSSS, Newmark, OALTS, and ExplicitIntegrator.

Here is the caller graph for this function:

from_incre_velocity() [2/2]

vec Integrator::from_incre_velocity	(	double	magnitude,
		const uvec &	encoding
	)

A simplified version similar to from_incre_velocity(const vec&, const uvec&). It assumes all DoFs share the same magnitude.

Here is the call graph for this function:

from_total_acceleration() [1/2]

vec Integrator::from_total_acceleration ( const vec &  total_acceleration, const uvec &  encoding  )

virtual

Reimplemented in GeneralizedAlphaExplicit, GERKN, GSSE, BatheTwoStep, OALTS, and ExplicitIntegrator.

Here is the call graph for this function:

Here is the caller graph for this function:

from_total_acceleration() [2/2]

vec Integrator::from_total_acceleration	(	double	magnitude,
		const uvec &	encoding
	)

Here is the call graph for this function:

from_total_velocity() [1/2]

vec Integrator::from_total_velocity ( const vec &  total_velocity, const uvec &  encoding  )

virtual

Reimplemented in BatheTwoStep, and OALTS.

Here is the call graph for this function:

Here is the caller graph for this function:

from_total_velocity() [2/2]

vec Integrator::from_total_velocity	(	double	magnitude,
		const uvec &	encoding
	)

Here is the call graph for this function:

get_auxiliary_residual()

vec Integrator::get_auxiliary_residual

(

)

const

Assemble the global residual vector due to nonlinear constraints implemented via the multiplier method.

Here is the call graph for this function:

get_displacement_residual()

vec Integrator::get_displacement_residual ( )

virtual

Assemble the global residual vector in displacement-controlled solving schemes. Apart from the global resistance and external load vectors, the reference load vector shall also be considered.

Reimplemented in LeeNewmarkBase, GeneralizedAlphaExplicit, GeneralizedAlpha, and GSSSS.

Here is the caller graph for this function:

get_domain()

shared_ptr< DomainBase > Integrator::get_domain

(

)

const

Here is the caller graph for this function:

get_force_residual()

vec Integrator::get_force_residual ( )

virtual

Assemble the global residual vector in load-controlled solving schemes.

Reimplemented in LeeNewmarkBase, GeneralizedAlphaExplicit, GeneralizedAlpha, and GSSSS.

Here is the caller graph for this function:

get_reference_load()

sp_mat Integrator::get_reference_load ( )

virtual

Reimplemented in GeneralizedAlphaExplicit, GeneralizedAlpha, and GSSSS.

Here is the caller graph for this function:

get_trial_displacement()

const vec & Integrator::get_trial_displacement ( ) const

virtual

Reimplemented in ExplicitIntegrator.

has_corrector()

bool Integrator::has_corrector ( ) const

protectedvirtual

Indicate whether the integrator has a corrector. Some time integration methods adopt predictor-corrector type scheme. The final committed state is different from the one used in computation. Override this method to indicate whether the integrator has a corrector. If it returns true, the correct_trial_status() method will be called thus has to be implemented.

Reimplemented in BatheExplicit, GeneralizedAlphaExplicit, GERKN, and GSSE.

Here is the caller graph for this function:

initialize()

int Integrator::initialize ( )

virtual

Reimplemented in LeeNewmark, LeeNewmarkBase, LeeNewmarkFull, LeeNewmarkIterative, NonviscousNewmark, and WilsonPenzienNewmark.

Here is the caller graph for this function:

matrix_is_assembled()

bool Integrator::matrix_is_assembled

(

)

const

process_constraint()

int Integrator::process_constraint ( )

virtual

The main task of this method is to apply constraints (of various forms implemented in various methods). Combinations of different types need to be considered: 1) homogeneous, 2) inhomogeneous, 3) linear, 4) nonlinear. Combinations of different methods need to be considered: 1) penalty, 2) multiplier. On exit, the global stiffness matrix should be updated, the global residual vector should be updated.

Reimplemented in LeeNewmark, LeeNewmarkFull, LeeNewmarkIterative, and WilsonPenzienNewmark.

Here is the call graph for this function:

Here is the caller graph for this function:

process_constraint_impl()

int Integrator::process_constraint_impl ( bool  full )

protectedvirtual

Reimplemented in GeneralizedAlphaExplicit, GERKN, GSSE, GeneralizedAlpha, and GSSSS.

Here is the caller graph for this function:

process_constraint_resistance()

int Integrator::process_constraint_resistance ( )

virtual

This method is similar to process_constraint(), but it only updates the global residual vector. The global stiffness matrix is not touched as in some solving schemes, the global stiffness matrix is only assembled and factorised once at the beginning. Subsequent iterations do not assemble the global stiffness matrix again and reuse the factorised matrix. In this case, the factorised matrix cannot be modified.

Reimplemented in LeeNewmark, LeeNewmarkFull, and LeeNewmarkIterative.

Here is the call graph for this function:

Here is the caller graph for this function:

process_criterion()

int Integrator::process_criterion

(

)

const

process_load()

int Integrator::process_load

(

)

Here is the call graph for this function:

process_load_impl()

int Integrator::process_load_impl ( bool  full )

protectedvirtual

Reimplemented in GeneralizedAlphaExplicit, GERKN, GSSE, GeneralizedAlpha, and GSSSS.

Here is the caller graph for this function:

process_load_resistance()

int Integrator::process_load_resistance

(

)

Here is the call graph for this function:

process_modifier()

int Integrator::process_modifier

(

)

const

Here is the caller graph for this function:

record()

void Integrator::record

(

)

const

reset_status()

void Integrator::reset_status ( )

virtual

Reimplemented in LeeNewmarkBase, and WilsonPenzienNewmark.

Here is the caller graph for this function:

set_domain()

void Integrator::set_domain

(

const std::weak_ptr< DomainBase > &

D

)

set_matrix_assembled_switch()

void Integrator::set_matrix_assembled_switch

(

)

set_time_step_switch()

void Integrator::set_time_step_switch

(

bool

T

)

Here is the caller graph for this function:

solve() [1/8]

mat Integrator::solve

(

const mat &

B

)

Here is the call graph for this function:

Here is the caller graph for this function:

solve() [2/8]

mat Integrator::solve

(

const sp_mat &

B

)

Here is the call graph for this function:

solve() [3/8]

mat Integrator::solve

(

mat &&

B

)

Here is the call graph for this function:

solve() [4/8]

int Integrator::solve ( mat &  X, const mat &  B  )

virtual

Reimplemented in LeeNewmarkBase, WilsonPenzienNewmark, and ExplicitIntegrator.

solve() [5/8]

int Integrator::solve ( mat &  X, const sp_mat &  B  )

virtual

Reimplemented in LeeNewmarkBase, WilsonPenzienNewmark, and ExplicitIntegrator.

solve() [6/8]

int Integrator::solve ( mat &  X, mat &&  B  )

virtual

Reimplemented in LeeNewmarkBase, WilsonPenzienNewmark, and ExplicitIntegrator.

solve() [7/8]

int Integrator::solve ( mat &  X, sp_mat &&  B  )

virtual

Reimplemented in LeeNewmarkBase, WilsonPenzienNewmark, and ExplicitIntegrator.

solve() [8/8]

mat Integrator::solve

(

sp_mat &&

B

)

Here is the call graph for this function:

stage_and_commit_status()

void Integrator::stage_and_commit_status

(

)

Here is the call graph for this function:

stage_status()

void Integrator::stage_status

(

)

const

Here is the caller graph for this function:

sync_status()

int Integrator::sync_status

(

bool

only_correct

)

When a new displacement increment is computed, it is added to global displacement vector. At this moment, nodal and elemental quantities are all computed from the previous displacement vector, directly committing the new results causes out-of-sync issue. Some algorithms use predictor-corrector type scheme, which means the converged quantities are different from the committed quantities. This method is in charge of syncing quantities between global and local quantities by updating nodal/elemental quantities using the committed quantities.

Here is the call graph for this function:

time_independent_matrix()

bool Integrator::time_independent_matrix ( ) const

virtual

Indicate whether the matrix is independent of time, that is, the system does not change with time. This helps to determine whether the matrix needs to be reassembled. For single-step methods, it is typically true. For multistep methods, it is typically false.

Reimplemented in BatheTwoStep, and OALTS.

type()

virtual Type Integrator::type ( ) const

inlinevirtual

Reimplemented in ImplicitIntegrator, and ExplicitIntegrator.

update_constraint()

void Integrator::update_constraint

(

)

const

update_from_ninja()

void Integrator::update_from_ninja ( )

virtual

Reimplemented in ExplicitIntegrator.

Here is the call graph for this function:

update_incre_time()

void Integrator::update_incre_time ( double  T )

virtual

Reimplemented in BatheExplicit, GERKN, and BatheTwoStep.

Here is the call graph for this function:

update_internal()

int Integrator::update_internal ( const mat &  )

virtual

Some algorithms solve a system which differs from the original one. The size of the problem changes thus the computed increment contains additional internal quantities. This method updates internal quantities stored in those integrators.

Reimplemented in LeeNewmarkBase.

update_load()

void Integrator::update_load

(

)

const

update_parameter()

void Integrator::update_parameter ( double  )

protectedvirtual

When time step changes, some parameters may need to be updated.

Reimplemented in BatheExplicit, GeneralizedAlphaExplicit, GERKN, GSSE, Tchamwa, BatheTwoStep, GeneralizedAlpha, GSSSS, Newmark, NonviscousNewmark, and OALTS.

Here is the caller graph for this function:

update_trial_load_factor() [1/2]

void Integrator::update_trial_load_factor

(

const vec &

lambda

)

const

Here is the call graph for this function:

update_trial_load_factor() [2/2]

void Integrator::update_trial_load_factor

(

double

lambda

)

const

Here is the call graph for this function:

Here is the caller graph for this function:

update_trial_status()

int Integrator::update_trial_status ( bool  detect_trivial )

virtual

Reimplemented in BatheExplicit, GeneralizedAlphaExplicit, GERKN, GSSE, Tchamwa, BatheTwoStep, GeneralizedAlpha, GSSSS, Newmark, and OALTS.

Here is the call graph for this function:

Here is the caller graph for this function:

update_trial_time()

void Integrator::update_trial_time

(

double

T

)

Here is the call graph for this function:

---

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

• Solver/Integrator/Integrator.h
• Solver/Integrator/Integrator.cpp