To solve time stepping problems, i.e. systems of linear ordinary differential equations of the form [A2]{y"} + [A1]{y'} + [A0]{y} = {b}. More...

#include <TimeStepping.h>

Public Member Functions
	TimeStepping ()
	Default constructor.

	TimeStepping (TimeScheme s, real_t time_step=theTimeStep, real_t final_time=theFinalTime)
	Constructor using time discretization data. More...

	~TimeStepping ()
	Destructor.

void	set (TimeScheme s, real_t time_step=theTimeStep, real_t final_time=theFinalTime)
	Define data of the differential equation or system. More...

void	setLinearSolver (LinearSolver &ls)
	Set reference to LinearSolver instance. More...

void	setPDE (Equa &eq, bool nl=false)
	Define partial differential equation to solve. More...

void	setRK4RHS (Vect< real_t > &f)
	Set intermediate right-hand side vector for the Runge-Kutta method. More...

void	setRK3_TVDRHS (Vect< real_t > &f)
	Set intermediate right-hand side vector for the TVD Runge-Kutta 3 method. More...

void	setInitial (Vect< real_t > &u)
	Set initial condition for the system of differential equations. More...

void	setInitial (Vect< real_t > &u, Vect< real_t > &v)
	Set initial condition for a system of differential equations. More...

void	setInitialRHS (Vect< real_t > &f)
	Set initial RHS for a system of differential equations when the used scheme requires it. More...

void	setRHS (Vect< real_t > &b)
	Set right-hand side vector.

void	setRHS (string exp)
	Set right-hand side as defined by a regular expression. More...

void	setBC (Vect< real_t > &u)
	Set vector containing boundary condition to enforce.

void	setBC (int code, string exp)
	Set boundary condition as defined by a regular expression. More...

void	setNewmarkParameters (real_t beta, real_t gamma)
	Define parameters for the Newmark scheme. More...

void	setConstantMatrix ()
	Say that matrix problem is constant. More...

void	setNonConstantMatrix ()
	Say that matrix problem is variable. More...

void	setLinearSolver (Iteration s=DIRECT_SOLVER, Preconditioner p=DIAG_PREC)
	Set linear solver data. More...

void	setNLTerm0 (Vect< real_t > &a0, Matrix< real_t > &A0)
	Set vectors defining a nonlinear first order system of ODEs. More...

void	setNLTerm (Vect< real_t > &a0, Vect< real_t > &a1, Vect< real_t > &a2)
	Set vectors defining a nonlinear second order system of ODEs. More...

real_t	runOneTimeStep ()
	Run one time step. More...

void	run (bool opt=false)
	Run the time stepping procedure. More...

void	Assembly (const Element &el, real_t b, real_t A0, real_t A1, real_t A2=nullptr)
	Assemble element arrays into global matrix and right-hand side. More...

void	SAssembly (const Side &sd, real_t b, real_t A=nullptr)
	Assemble side arrays into global matrix and right-hand side. More...

LinearSolver &	getLSolver ()
	Return LinearSolver instance.

Detailed Description

To solve time stepping problems, i.e. systems of linear ordinary differential equations of the form [A2]{y"} + [A1]{y'} + [A0]{y} = {b}.

Author: Rachid Touzani

Copyright: GNU Lesser Public License

Features:

The system may be first or second order (first and/or second order time derivatives
The following time integration schemes can be used:
- For first order systems: The following schemes are implemented Forward Euler (value: FORWARD_EULER)
  Backward Euler (value: BACKWARD_EULER)
  Crank-Nicolson (value: CRANK_NICOLSON)
  Heun (value: HEUN)
  2nd Order Adams-Bashforth (value: AB2)
  4-th order Runge-Kutta (value: RK4)
  2nd order Backward Differentiation Formula (value: BDF2)
- For second order systems: The following schemes are implemented Newmark (value: NEWMARK)

Constructor & Destructor Documentation

◆ TimeStepping()

TimeStepping	(	TimeScheme	s,
		real_t	time_step = `theTimeStep`,
		real_t	final_time = `theFinalTime`
	)

Constructor using time discretization data.

Parameters

[in]	s	Choice of the scheme: To be chosen in the enumerated variable TimeScheme (see the presentation of the class)
[in]	time_step	Value of the time step. This value will be modified if an adaptive method is used. The default value for this parameter if the value given by the global variable `theTimeStep`
[in]	final_time	Value of the final time (time starts at 0). The default value for this parameter is the value given by the global variable `theFinalTime`

Member Function Documentation

◆ Assembly()

void Assembly	(	const Element &	el,
		real_t *	b,
		real_t *	A0,
		real_t *	A1,
		real_t *	A2 = `nullptr`
	)

Assemble element arrays into global matrix and right-hand side.

This member function is to be called from finite element equation classes

Parameters

[in]	el	Reference to Element class
[in]	b	Pointer to element right-hand side
[in]	A0	Pointer to matrix of 0-th order term (involving no time derivative)
[in]	A1	Pointer to matrix of first order term (involving time first derivative)
[in]	A2	Pointer to matrix of second order term (involving time second derivative) [Default: `nullptr`]

◆ run()

void run ( bool opt = false )

Run the time stepping procedure.

Parameters

[in] opt Flag to say if problem matrix is constant while time stepping (true) or not (Default value is false)

Note: This argument is not used if the time stepping scheme is explicit

◆ runOneTimeStep()

real_t runOneTimeStep ( )

Run one time step.

Returns: Value of new time step if this one is updated

◆ SAssembly()

void SAssembly	(	const Side &	sd,
		real_t *	b,
		real_t *	A = `nullptr`
	)

Assemble side arrays into global matrix and right-hand side.

This member function is to be called from finite element equation classes

Parameters

[in]	sd	Reference to Side class
[in]	b	Pointer to side right-hand side
[in]	A	Pointer to matrix [Default: `nullptr`]

◆ set()

void set	(	TimeScheme	s,
		real_t	time_step = `theTimeStep`,
		real_t	final_time = `theFinalTime`
	)

Define data of the differential equation or system.

Parameters

[in]	s	Choice of the scheme: To be chosen in the enumerated variable TimeScheme (see the presentation of the class)
[in]	time_step	Value of the time step. This value will be modified if an adaptive method is used. The default value for this parameter if the value given by the global variable `theTimeStep`
[in]	final_time	Value of the final time (time starts at 0). The default value for this parameter is the value given by the global variable `theFinalTime`

◆ setBC()

void setBC	(	int	code,
		string	exp
	)

Set boundary condition as defined by a regular expression.

Parameters

[in]	code	Code for which expression is assigned
[in]	exp	Regular expression to assign as a function of `x`, `y`, `z` and `t`

◆ setConstantMatrix()

void setConstantMatrix ( )

Say that matrix problem is constant.

This is useful if the linear system is solved by a factorization method but has no effect otherwise

◆ setInitial() [1/2]

void setInitial ( Vect< real_t > & u )

Set initial condition for the system of differential equations.

Parameters

[in] u Vector containing initial condition for the unknown

Remarks: If a second-order differential equation is to be solved, use the the same function with two initial vectors (one for the unknown, the second for its time derivative)

◆ setInitial() [2/2]

void setInitial	(	Vect< real_t > &	u,
		Vect< real_t > &	v
	)

Set initial condition for a system of differential equations.

Parameters

[in]	u	Vector containing initial condition for the unknown
[in]	v	Vector containing initial condition for the time derivative of the unknown

Note: This function can be used to provide solution at previous time step if a restarting procedure is used.; This member function is to be used only in the case of a second order system

◆ setInitialRHS()

void setInitialRHS ( Vect< real_t > & f )

Set initial RHS for a system of differential equations when the used scheme requires it.

Giving the right-hand side at initial time is somtimes required for high order methods like Runge-Kutta

Parameters

[in] f Vector containing right-hand side at initial time. This vector is helpful for high order methods

Note: This function can be used to provide solution at previous time step if a restarting procedure is used.

◆ setLinearSolver() [1/2]

void setLinearSolver	(	Iteration	s = `DIRECT_SOLVER`,
		Preconditioner	p = `DIAG_PREC`
	)

Set linear solver data.

Parameters

[in]	s	Solver identification parameter. To be chosen in the enumeration variable Iteration: DIRECT_SOLVER, CG_SOLVER, CGS_SOLVER, BICG_SOLVER, BICG_STAB_SOLVER, GMRES_SOLVER, QMR_SOLVER [Default: `DIRECT_SOLVER`]
[in]	p	Preconditioner identification parameter. To be chosen in the enumeration variable Preconditioner: IDENT_PREC, DIAG_PREC, ILU_PREC [Default: `DIAG_PREC`]

Note: The argument p has no effect if the solver is DIRECT_SOLVER

◆ setLinearSolver() [2/2]

void setLinearSolver ( LinearSolver & ls )

Set reference to LinearSolver instance.

Parameters

[in] ls Reference to LinearSolver instance

◆ setNewmarkParameters()

void setNewmarkParameters	(	real_t	beta,
		real_t	gamma
	)

Define parameters for the Newmark scheme.

Parameters

[in]	beta	Parameter beta [Default: `0.25`]
[in]	gamma	Parameter gamma [Default: `0.5`]

◆ setNLTerm()

void setNLTerm	(	Vect< real_t > &	a0,
		Vect< real_t > &	a1,
		Vect< real_t > &	a2
	)

Set vectors defining a nonlinear second order system of ODEs.

The ODE system has the form a2(u)'' + a1(u)' + a0(u) = 0

Parameters

[in]	a0	Reference to Vect instance defining the 0-th order term
[in]	a1	Reference to Vect instance defining the first order term
[in]	a2	Reference to Vect instance defining the second order term

◆ setNLTerm0()

void setNLTerm0	(	Vect< real_t > &	a0,
		Matrix< real_t > &	A0
	)

Set vectors defining a nonlinear first order system of ODEs.

The ODE system has the form a1(u)' + a0(u) = 0

Parameters

[in]	a0	Reference to Vect instance defining the 0-th order term
[in]	A0	Reference to Matrix instance

◆ setNonConstantMatrix()

void setNonConstantMatrix ( )

Say that matrix problem is variable.

This is useful if the linear system is solved by a factorization method but has no effect otherwise

◆ setPDE()

void setPDE	(	Equa &	eq,
		bool	nl = `false`
	)

Define partial differential equation to solve.

The used equation class must have been constructed using the Mesh instance

Parameters

[in]	eq	Reference to equation instance
[in]	nl	Toggle to say if the considered equation is linear [Default: `0`] or not

◆ setRHS()

void setRHS ( string exp )

Set right-hand side as defined by a regular expression.

Parameters

[in] exp Regular expression as a function of x, y, z and t

◆ setRK3_TVDRHS()

void setRK3_TVDRHS ( Vect< real_t > & f )

Set intermediate right-hand side vector for the TVD Runge-Kutta 3 method.

Parameters

[in] f Vector containing the RHS

◆ setRK4RHS()

void setRK4RHS ( Vect< real_t > & f )

Set intermediate right-hand side vector for the Runge-Kutta method.

Parameters

[in] f Vector containing the RHS

Public Member Functions

Detailed Description

Constructor & Destructor Documentation

◆ TimeStepping()

Member Function Documentation

◆ Assembly()

◆ run()

◆ runOneTimeStep()

◆ SAssembly()

◆ set()

◆ setBC()

◆ setConstantMatrix()

◆ setInitial() [1/2]

◆ setInitial() [2/2]

◆ setInitialRHS()

◆ setLinearSolver() [1/2]

◆ setLinearSolver() [2/2]

◆ setNewmarkParameters()

◆ setNLTerm()

◆ setNLTerm0()

◆ setNonConstantMatrix()

◆ setPDE()

◆ setRHS()

◆ setRK3_TVDRHS()

◆ setRK4RHS()