GmpFemTransientSolver Class Reference

Basic class for solving a transient FEM problem. More...

#include <gmpFemTransientSolver.h>

Inheritance diagram for GmpFemTransientSolver:

Collaboration diagram for GmpFemTransientSolver:

Public Member Functions
	GmpFemTransientSolver (GmElementMesh *mesh, GmSimulationData *simulation, const QList< GmpFemPhysics * > &physics, GmNumSolver *solver, bool nonlinear, const GmpFemSolverOptions &options, const GmLogCategory &logger)
	Constructor. Expects to receive as parameters the mesh we are acting upon, the list of physics objects that will cooperate to create the global stiffness matrix and the solver used to solve the resulting linear system. More...
virtual	~GmpFemTransientSolver ()
	Destructor.
virtual bool	init ()
	Prepares the solver for assembling matrices by creating the assembler object and allocating the needed memory. This operations only needs to be done once if the solver is used in a loop.
bool	runStep (double dt, bool updateDof=false)
	Execute the process. Should be used by linear processes only. More...
bool	runLinearStep (double dt, int iter, bool updateDof=false)
	Execute the process. Should be used by non linear processes only. More...
bool	calcLinearResidual (double dt, double *rnorm, double *maxNodeDiff, double *avgNodeDiff)
	When solving a non linear system by repeated iterations, this function aims to calculate the residual of the actual solution. More...
bool	nonlinear () const
	Returns true if the solver is non linear.
virtual bool	addStateItemsToGroup (GmStateDump *state, int groupId)
	Adds to 'state' the data items that should be saved for this FEM process. Should probably be overriden by derived classes to add their own requirements. More...
Public Member Functions inherited from GmpFemSolver
	GmpFemSolver (GmElementMesh *mesh, GmSimulationData *simulation, const QList< GmpFemPhysics * > &physics, GmNumSolver *solver, const GmpFemSolverOptions &options, const GmLogCategory &logger)
	Constructor. Expects to receive as parameters the mesh we are acting upon, the list of physics objects that will cooperate to create the global stiffness matrix and the solver used to solve the resulting linear system. More...
virtual	~GmpFemSolver (void)
	Destructor.
bool	run ()
	Execute the process. Returns true on success.
bool	calcLinearResidual (double *rnorm, double *maxNodeDiff, double *avgNodeDiff)
	When solving a non linear system by repeated iterations, this function aims to calculate the residual of the actual solution. More...
void	setMatrixCombinerObject (const GmpFemAssemblerMatrixCombiner *combiner)
	Sets the combiner object that will be used by the assembler to merge elemental data from multiple matrices (eg. C and K) into an equivalent matrix. More...
void	setVectorCombinerObject (const GmpFemAssemblerVectorCombiner *combiner)
	Sets the combiner objects that will be used by the assembler to merge elemental data from multiple vectors (eg. Fe and Fi) into an equivalent vector. More...
bool	update (int mode, QString &err)
	Informs the solver of an external change in model dofs, elements or BCs that must be forwarded to the assembler. See the documentation of GmpFemAssembler::update() for more details.
virtual bool	stateAboutToBeSaved (GmStateDump *state)
virtual bool	stateSaved (GmStateDump *state)
virtual bool	stateAboutToBeLoaded (GmStateDump *state)
virtual bool	stateLoaded (GmStateDump *state)
	Informs physics objects that a load operation was completed giving them an opportunity to adjust its state in case the loaded data is not exactly what the internal state needs to be. It will also call derivedResults() so that previous results can be recovered. In that call, the nonLinearSolver parameter is obtained from the stateLoadedCalcDerivedResultsNonLinear() virtual function.
virtual bool	fillStateControlMapData (QVariantMap *map)
	Method called for filling the state dump control regustered in addStateItemsToGroup()
virtual bool	stateControlMapDataLoaded (QVariantMap *map)
	Method called when the state dump control registered in addStateItemsToGroup() has been loaded.
virtual bool	stateLoadedCalcDerivedResultsNonLinear ()
	Returns the value of the "nonLinearSolver" parameter that will be passed to the calcDerivedResults() call from the standard implementation of stateLoaded()
Public Member Functions inherited from QObject
virtual const QMetaObject *	metaObject () const const
virtual void *	qt_metacast (const char *)
virtual int	qt_metacall (QMetaObject::Call, int, void **)
	QObject (QObject *parent)
virtual bool	event (QEvent *e)
virtual bool	eventFilter (QObject *watched, QEvent *event)
QString	objectName () const const
void	setObjectName (const QString &name)
bool	isWidgetType () const const
bool	isWindowType () const const
bool	signalsBlocked () const const
bool	blockSignals (bool block)
QThread *	thread () const const
void	moveToThread (QThread *targetThread)
int	startTimer (int interval, Qt::TimerType timerType)
int	startTimer (std::chrono::milliseconds time, Qt::TimerType timerType)
void	killTimer (int id)
T	findChild (const QString &name, Qt::FindChildOptions options) const const
QList< T >	findChildren (const QString &name, Qt::FindChildOptions options) const const
QList< T >	findChildren (const QRegExp &regExp, Qt::FindChildOptions options) const const
QList< T >	findChildren (const QRegularExpression &re, Qt::FindChildOptions options) const const
const QObjectList &	children () const const
void	setParent (QObject *parent)
void	installEventFilter (QObject *filterObj)
void	removeEventFilter (QObject *obj)
QMetaObject::Connection	connect (const QObject *sender, const char *signal, const char *method, Qt::ConnectionType type) const const
bool	disconnect (const char *signal, const QObject *receiver, const char *method) const const
bool	disconnect (const QObject *receiver, const char *method) const const
void	dumpObjectTree ()
void	dumpObjectInfo ()
void	dumpObjectTree () const const
void	dumpObjectInfo () const const
bool	setProperty (const char *name, const QVariant &value)
QVariant	property (const char *name) const const
QList< QByteArray >	dynamicPropertyNames () const const
void	destroyed (QObject *obj)
void	objectNameChanged (const QString &objectName)
QObject *	parent () const const
bool	inherits (const char *className) const const
void	deleteLater ()

Protected Member Functions
virtual bool	initElementSets (GmNumSolver *solver)
	Reimplements GmpFemSolver::initElementSets() to include the C matrix in the set of calculated matrices by physics objects.
virtual bool	cleanup ()
	Dealocates memory and sets allocated resources to NULL. As a convenience, returns false.
bool	step (double dt, int iter, bool updateDof, bool updateBcs)
bool	istep (double dt, int iter, bool updateDof, bool updateBcs)
	Worker function for both runStep() and runLinearStep()
bool	estep (double dt, bool updateDof, bool updateBcs)
bool	prepareLinearSystem (double dt)
	Fills the matrix K and the vector Fe with the linear system resulting from applying an implicit schema to the transient temperature equation. More...
bool	prepareExplicitLinearSystem (double dt)
bool	GuyanReductionSolve ()
Protected Member Functions inherited from GmpFemSolver
bool	initSolver (GmpFemAssembler::FixedDofMode assemblerMode, bool assemblerReverseMapping, bool enableFastUpdate)
	Basic implementation of the init function receiving as parameters the configuration options that are sent to the assembler. More...
virtual bool	initResultAttributes (QString prefix)
	Helpper function used to register the set of result attributes managed by the fem solver. To avoid name clashes, all of them should prepend the given prefix (usually "fem") to the attribute name. More...
GmpFemPhysics::FemResultType	prepareMatrices (bool skipFixedBcs=false)
	Auxilliary function used to fill the K, and f matrices / vectors along with any other matrices/vectors stored in _matSet and _vecSet. More...
GmpFemPhysics::FemResultType	saveElementData (int iter)
	Auxilliary function used to save to the configured file data from the set of element matrices and vectors. In this process, physics are called again to fill element matrices and vectors (according to the active matrices and vectors in _vecSet and _matSet) but the results are NOT added to the global matrices.
bool	solveLinearSystem (bool xFilled)
	Solves the linear system K.x = f taking K and f from the single equivalent matrix/vector or from matrix K / vector f, depending on the assembler configuration. Results are saved in _x. Prints the used matrices / result as configured in _printOptions. More...
double	timeConvert (double val)
	Auxiliary function created for being used by derived classes that converts a time value given in the current simulation time unit and returns that value converted to the physics expected unit.
GmpFemPhysics::FemResultType	traverseElements ()
	Fills the stiffness matrix _K and the force vector _f by traversing elements asking physics for the local element matrix / force vector. More...
GmpFemPhysics::FemResultType	traverseBoundaryElements ()
	Simillar to traverseElements(), this function steps through each element that belongs to an edge or face boundary condition, asking physics for additional contributions to the stiffness matrix _K and the force vector _f.
GmpFemPhysics::FemResultType	traverseExternalLoads ()
	Simillar to traverseElements(), this function steps through each element associated to an external load force, asking physics for additional contributions to the force vector _f.
GmpFemPhysics::FemResultType	traverseContactBoundaries ()
	Simillar to traverseElements(), this function steps through each possible contact from the given contact boundary conditions and asks the physics for additional contributions to the solver matrices and vectors. This is different from the other element based loops since the returned matrices are not restricted to a single element, but can combine nodes from several ones.
GmpFemPhysics::FemResultType	traverseElementsForSaving (FILE *f, int iter)
	Similar to traverseElements() but instead of adding the elements to the assembler, saves the element data to the given file. More...
GmpFemPhysics::FemResultType	fillElementData (const GmElement *e, int physIndex)
	Helper function used by traverseElements() to get data for a single element calling phys->fillElementData()
GmpFemPhysics::FemResultType	fillElementBoundaryData (const GmElement *e, int physIndex, const GmBoundaryCondition *bc, int bcIndex, int bcListIndex, int border, const GmCellBoundary *b)
	Helper function used by traverseBoundaryElements() to get data for a single element calling phys->fillElementDataForBc()
GmpFemPhysics::FemResultType	fillContactData (GmMatrixDof &localDofMap, const GmContactBoundaryCondition *cbc, int physIndex, int index1, int index2)
	Helper function used by traverseContactBoundaries() to get data for a single contact pair calling phys->fillContactData()
bool	addFixedForces ()
	Adds fixed nodal forces, as seen by each physics, to the global vector _f.
virtual bool	applyFixedBoundaryConditions (int *numFixed)
	Apply Dirichlet (fixed) boundary conditions to the global equation system. More...
bool	collectFixedBcs (QVector< bool > &fixedRows, QVector< double > &fixedValues, QList< int > &dofIndex)
	Fills the set of vectors received as parameters with the complete set of fixed boundary conditions, as seen by the full set of physics in use, checking for possible conflicts in conditions. More...
virtual void	collectGlobalContactPairs (const GmContactBoundaryCondition *cbc, QList< QPair< int, int > > &contactPairs)
	Global search rule used to define contact surfaces. Should be implemented to allow global contact conditions, filling contactPairs with the indices in cbc of the in-contact surfaces.
void	printElementData (const GmElement *e, int ndof, const int *dofMapping, const GmpFemPhysics *p, const GmBoundaryCondition *bc=NULL, const GmContactBoundaryCondition *cbc=NULL, int loadId1=-1, int id2=-1)
	Print element dof mapping, local matrices and vectors depending on the current solver print options. More...
void	printGlobalData (bool linear)
	Print global matrices and vectors, either after assembling or before solving the linear system.
Protected Member Functions inherited from QObject
QObject *	sender () const const
int	senderSignalIndex () const const
int	receivers (const char *signal) const const
bool	isSignalConnected (const QMetaMethod &signal) const const
virtual void	timerEvent (QTimerEvent *event)
virtual void	childEvent (QChildEvent *event)
virtual void	customEvent (QEvent *event)
virtual void	connectNotify (const QMetaMethod &signal)
virtual void	disconnectNotify (const QMetaMethod &signal)

Private Attributes
bool	_implicit
	Is this a implicit or explicit scheme ?
bool	_nonlinear
	Is this a non linear problem ?
GmVector	_oldx
	Previous state of the x vector used for non linear problems;.

Additional Inherited Members
Public Types inherited from GmpFemSolver
enum	PrintOptions {   PRINT_ELEMENT_MATRICES = 0x001, PRINT_ELEMENT_VECTORS = 0x002, PRINT_ELEMENT_DOF_MAPPING = 0x004, PRINT_EQ_MATRIX = 0x008,   PRINT_EQ_VECTOR = 0x010, PRINT_GLOBAL_MATRICES = 0x020, PRINT_GLOBAL_VECTORS = 0x040, PRINT_LINEAR_MATRIX = 0x080,   PRINT_LINEAR_VECTOR = 0x100, PRINT_LINEAR_RESULT = 0x200 }
	Flag options used to compose the printOptions constructor parameter. More...
Static Public Member Functions inherited from QObject
QString	tr (const char *sourceText, const char *disambiguation, int n)
QString	trUtf8 (const char *sourceText, const char *disambiguation, int n)
QMetaObject::Connection	connect (const QObject *sender, const char *signal, const QObject *receiver, const char *method, Qt::ConnectionType type)
QMetaObject::Connection	connect (const QObject *sender, const QMetaMethod &signal, const QObject *receiver, const QMetaMethod &method, Qt::ConnectionType type)
QMetaObject::Connection	connect (const QObject *sender, PointerToMemberFunction signal, const QObject *receiver, PointerToMemberFunction method, Qt::ConnectionType type)
QMetaObject::Connection	connect (const QObject *sender, PointerToMemberFunction signal, Functor functor)
QMetaObject::Connection	connect (const QObject *sender, PointerToMemberFunction signal, const QObject *context, Functor functor, Qt::ConnectionType type)
bool	disconnect (const QObject *sender, const char *signal, const QObject *receiver, const char *method)
bool	disconnect (const QObject *sender, const QMetaMethod &signal, const QObject *receiver, const QMetaMethod &method)
bool	disconnect (const QMetaObject::Connection &connection)
bool	disconnect (const QObject *sender, PointerToMemberFunction signal, const QObject *receiver, PointerToMemberFunction method)
Protected Types inherited from GmpFemSolver
enum	DisabledWarnings {   PrescribedForceForInvalidDof, PrescribedForceForFixedDof, FixedBcForInvalidDof, ConflictingFixedBcValue,   NumDisabledWarnings }
	Enums describing the set of warnings that can be disabled by simulation options. More...
Protected Slots inherited from GmpFemSolver
virtual void	meshChanged ()
	Slot called when the underlying mesh has been changed. More...
Protected Attributes inherited from GmpFemSolver
int	_solverId
	A unique index for this solver object used by the result attribute auto prefix feature.
GmElementMesh *	_mesh
	Mesh with the nodes and cells data.
GmSimulationData *	_simulation
	Simulation Data.
QList< GmpFemPhysics * >	_physicsList
	List of physics that will cooperate to generate the system response.
GmNumSolver *	_solver
	Numeric solver that will be used to solve the set of linear equations.
const GmLogCategory &	_logger
	Logger for eventual messages.
QMutex	_printMutex
	Mutex for serializing calls to printElementData when traversing elements in parallel.
bool	_enabledWarn [NumDisabledWarnings]
	Vector storing which warnings are enabled.
GmpFemSolverOptions	_solverOptions
	The set of solver options.
UnitConverter *	_timeConv
	Unit converter from the currentTimeUnit to the physics unit. Used only by derived solvers.
GmpFemAssembler *	_assembler
	Assembler used for adding element contributions to vectors and matrices.
const GmpFemAssemblerMatrixCombiner *	_matCombiner
	The combiner object used with the assembler to merge data from multiple elemental matrices.
const GmpFemAssemblerVectorCombiner *	_vecCombiner
	The combiner object used with the assembler to merge data from multiple elemental vectors.
GmpFemVectorSet	_vecSet
	Set of global/element vectors filled by physics.
GmpFemMatrixSet	_matSet
	Set of global/element matrices filled by physics.
GmVector	_x
	State vector.
GmVector	_r
	Residual vector.
unsigned	_preCount
	Number of times that prepareMatrices() was called.
unsigned	_runCount
	Number of times that run() was called.
QVariantMap	_controlMap
	Variant map used for storing dump control data for the solver.
int	_iterResAttr
	The registered id for the "fem.iter" result attribute.
int	_iterResErrAttr
	The registered id for the "fem.iterErr" result attribute.
Properties inherited from QObject
	objectName

Detailed Description

Basic class for solving a transient FEM problem.

Constructor & Destructor Documentation

GmpFemTransientSolver()

GmpFemTransientSolver::GmpFemTransientSolver	(	GmElementMesh *	mesh,
		GmSimulationData *	simulation,
		const QList< GmpFemPhysics * > &	physics,
		GmNumSolver *	solver,
		bool	nonlinear,
		const GmpFemSolverOptions &	options,
		const GmLogCategory &	logger
	)

Constructor. Expects to receive as parameters the mesh we are acting upon, the list of physics objects that will cooperate to create the global stiffness matrix and the solver used to solve the resulting linear system.

The nonlinear parameter should be set to true if the problem is non linear and will be solved by multiple calls to runLinearStep().

The options parameter contains the set of parameters sent to the solver.

Member Function Documentation

addStateItemsToGroup()

bool GmpFemTransientSolver::addStateItemsToGroup ( GmStateDump *  state, int  groupId  )

virtual

Adds to 'state' the data items that should be saved for this FEM process. Should probably be overriden by derived classes to add their own requirements.

Auxiliar variant map dump item that notifies us for loading and saving data to the map

< Our "father" solver object

Reimplemented from GmpFemSolver.

calcLinearResidual()

bool GmpFemTransientSolver::calcLinearResidual	(	double	dt,
		double *	rnorm,
		double *	maxNodeDiff,
		double *	avgNodeDiff
	)

When solving a non linear system by repeated iterations, this function aims to calculate the residual of the actual solution.

Consider a model K(x).x = f and an initial state x0, the model can be iterated so that we have K1 = K(x0), x1 = inv(K1).f, K2 = K(x1), ... More generally, we have Ki = K(xi-1), xi = inv(Ki).f.

With that model, our residual vector ri will be equal to K(xi).xi - f and the returned values: rnorm: The level two norm of the vector ri, ||ri|| maxNodeDiff: The maximum value of each entry in ri (makes sense only if all dof are comparable). Calculated only if maxNodeDiff != NULL. avgNodeDiff: The average value of each entry in ri (makes sense only if all dof are comparable) Calculated only if avgNodeDiff != NULL.

IMPORTANT: Keep in mind that this function expects to find in _x the result of the last call to runLinearStep() and will by itself fill a new version of _K and _f, calculated based on this new value of _x.

The function returns false on errors (while evaluating K(x)).

prepareExplicitLinearSystem()

bool GmpFemTransientSolver::prepareExplicitLinearSystem ( double  dt )

protected

An auxiliar class used by the assembler for to construct two Global Matrix C and K. (Explicit Methods)

Constructor. Automatically register the combiner on the solver

Enable localMatricesToGlobalVectorContribution() storing returned values in the equivalent global vector

An auxiliar combiner class used for multiplying local vector values by dt.

Destructor. Automatically removes the combiner from the solver

The dt * f_n+1 term

prepareLinearSystem()

bool GmpFemTransientSolver::prepareLinearSystem ( double  dt )

protected

Fills the matrix K and the vector Fe with the linear system resulting from applying an implicit schema to the transient temperature equation.

Implicit schema: (C + dt*K) * x_n+1 = C * x_n + dt * f_n+1 |-—K-—| |-x-| |------—Fe-----—|

If the solver is non linear, uses _oldx instead of _x for the x_n vector

An auxiliar class for combining data from matrices C and K into an equivalent matrix that will be used by the assembler

Constructor. Automatically registers the combiner on the solver

Destructor. Automatically removes the combiner from the solver

The C + dt * K term

Enable localMatricesToGlobalVectorContribution() storing returned values in the equivalent global vector

The C * x_n term

An auxiliar combiner class used for multiplying local vector values by dt.

Constructor. Automatically registers the combiner on the solver

Destructor. Automatically removes the combiner from the solver

The dt * f_n+1 term

runLinearStep()

bool GmpFemTransientSolver::runLinearStep	(	double	dt,
		int	iter,
		bool	updateDof = false
	)

Execute the process. Should be used by non linear processes only.

Important: This function does not update the current simulation time since this should be done only after the last step iteration.

Parameters

dt	The interval that we are going forward in time, in the unit specified by the current time unit. Will be converted to the unit expected by physics.
iter	The number of the nonlinear iteration (1 for the first one)
updateDof	If true means that we need to reload the _x vector from the state vars. This might be needed in rare occasions if the values where changed externally.

Returns

Returns true on success, false on errors.

runStep()

bool GmpFemTransientSolver::runStep	(	double	dt,
		bool	updateDof = false
	)

Execute the process. Should be used by linear processes only.

Important: This function does not update the current simulation time so that its behaviour is compatible with runLinearStep()

Parameters

dt	The interval that we are going forward in time, in the unit specified by the current time unit. Will be converted to the unit expected by physics.
updateDof	If true means that we need to reload the _x vector from the state vars. This might be needed in rare occasions if the values where changed externally.

Returns

Returns true on success, false on errors.

---

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

• inc/gmpFemTransientSolver.h
• src/gmpFemTransientSolver.cpp