Omu_IntSDIRK Class Reference

#include <Omu_IntSDIRK.h>

Inheritance diagram for Omu_IntSDIRK:

[legend]

Collaboration diagram for Omu_IntSDIRK:

[legend]

List of all members.

---

Detailed Description

Singly diagonally implizit Runge Kutta method for (stiff) ordinary differential equations (ODEs).

The implementation is based on Hairer/Wanner: "Solving ordinary differential equations II".

Public Member Functions
	Omu_IntSDIRK ()
	constructor
	~Omu_IntSDIRK ()
	destructor
Implementation of predefined methods.
See also: Omu_Integrator
char *	name ()
	Name of a specific integrator.
virtual void	init (int k, const Omu_StateVec &xc, const Omu_Vec &q, const Omu_DependentVec &Fc, bool sa)
	Initialize integrator, like adaptation of array dimensions and test of pre-conditions, like explicit ODE.
virtual void	solve (int kk, double tstart, double tend, Omu_StateVec &xc, Omu_StateVec &dxc, Omu_Vec &q, Omu_DependentVec &Fc)
	Solve differential equations over one sample period.
Protected Attributes
int	_jac_sbw
	User given semi-bandwidth of Jacobian (default: -1).
bool	_banded
	User specification to allow banded solver (default: false).
bool	_banded_solver
	internal flag for using banded solver
bool	_sparse_solver
	User specification to allow sparse solver (default: false).

---

Member Function Documentation

char* Omu_IntSDIRK::name

(

)

[inline, virtual]

Name of a specific integrator.

Implements Omu_Integrator.

virtual void Omu_IntSDIRK::init	(	int	k,
		const Omu_StateVec &	xc,
		const Omu_Vec &	q,
		const Omu_DependentVec &	Fc,
		bool	sa
	)			[virtual]

Initialize integrator, like adaptation of array dimensions and test of pre-conditions, like explicit ODE.

An explicit ODE is present if Fc.Jdx.is_scalar() and Fc.Jdx[0][0] != 0.0. An algebraic state j, 0<=j<_n can be identified with Fc.Jdx.is_zero_column[j]. Note that the init method should not find consistent initial conditions of a DAE as parameters may jump between sample periods kk afterwards.

Parameters:

	k	stage
	xc	continuous-time states, dim(xc)=_n
	q	sensitivity parameters, dim(q)=_nq
	Fc	pre-initialized data structures for evaluating residuals, dim(Fc)=_n, including structurally analyzed Jacobians, dim(Fc.Jx)=_n._n, dim(Fc.Jxp)=_n._n, dim(Fc.Jq)=_n._nq
	sa	indicates if sensitivity analysis is required

Reimplemented from Omu_Integrator.

virtual void Omu_IntSDIRK::solve	(	int	kk,
		double	tstart,
		double	tend,
		Omu_StateVec &	xc,
		Omu_StateVec &	dxc,
		Omu_Vec &	q,
		Omu_DependentVec &	Fc
	)			[virtual]

Solve differential equations over one sample period.

Additionally solve sensitivity equations if _sa is true. The default implementation calls the depreciated solve method.

Parameters:

	kk	sample period
	tstart	start time for integration
	tend	end time for integration
	xc	continuous-time states at tstart, dim(xc)=_n; sensitivities xc.Sq at tstart, dim(xc.Sq)=_n._nq, if _sa is true
	dxc	pre-allocated time derivatives of continuous-time states, dim(dxc)=_n
	q	sensitivity parameters, dim(q)=_nq
	Fc	pre-allocated argument for residual calls, dim(Fc)=_n, including structurally analyzed Jacobians Fc.Jx, Fc.Jdx, Fc.Jq

Return values:

xc

continuous-time states at tend; sensitivities xc.Sq at tend if _sa is true

Exceptions:

E_CONV

indicate that the integrator failed to converge; may be caught during optimization step length test

Reimplemented from Omu_Integrator.

---

Member Data Documentation

int Omu_IntSDIRK::_jac_sbw [protected]

User given semi-bandwidth of Jacobian (default: -1).

If a value >= 0 is specified, then it is used instead of the automatic detection.

bool Omu_IntSDIRK::_banded [protected]

User specification to allow banded solver (default: false).

Banded solvers are used if _banded is true, _jac_sbw < 0, and if the automatic detection indicates that the problem can be solved more efficiently in this way.

---

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

• Omu_IntSDIRK.h