Use DGBase as a Simulation Constraint ROL::Constraint_SimOpt. More...

#include <flow_constraints.hpp>

Inheritance diagram for PHiLiP::FlowConstraints< dim >:

Collaboration diagram for PHiLiP::FlowConstraints< dim >:

Public Member Functions
	FlowConstraints (std::shared_ptr< DGBase< dim, double >> &_dg, std::shared_ptr< BaseParameterization< dim >> _design_parameterization, std::shared_ptr< dealii::TrilinosWrappers::SparseMatrix > precomputed_dXvdXp=nullptr)
	Constructor.

	~FlowConstraints ()
	Destructor.

void	update_1 (const ROL::Vector< double > &des_var_sim, bool flag=true, int iter=-1)
	Update the simulation variables.

void	update_2 (const ROL::Vector< double > &des_var_ctl, bool flag=true, int iter=-1)
	Update the control variables. More...

void	solve (ROL::Vector< double > &constraint_values, ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl, double &) override
	Solve the Simulation Constraint and returns the constraints values. More...

void	value (ROL::Vector< double > &constraint_values, const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl, double &) override
	Returns the current constraint residual given a set of control and simulation variables.

void	applyJacobian_1 (ROL::Vector< double > &output_vector, const ROL::Vector< double > &input_vector, const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl, double &) override
	Applies the Jacobian of the Constraints w. r. t. the simulation variables onto a vector.

void	applyAdjointJacobian_1 (ROL::Vector< double > &output_vector, const ROL::Vector< double > &input_vector, const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl, double &) override
	Applies the Jacobian of the Constraints w. r. t. the simulation variables onto a vector.

void	applyInverseJacobian_1 (ROL::Vector< double > &output_vector, const ROL::Vector< double > &input_vector, const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl, double &) override
	Applies the inverse Jacobian of the Constraints w. r. t. the simulation variables onto a vector.

int	construct_JacobianPreconditioner_1 (const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl)
	Constructs the Jacobian preconditioner.

void	destroy_JacobianPreconditioner_1 ()
	Frees Jacobian preconditioner from memory;.

void	applyInverseJacobianPreconditioner_1 (ROL::Vector< double > &output_vector, const ROL::Vector< double > &input_vector, const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl, double &)
	Applies the inverse Jacobian preconditioner. More...

int	construct_AdjointJacobianPreconditioner_1 (const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl)
	Constructs the Adjoint Jacobian preconditioner.

void	destroy_AdjointJacobianPreconditioner_1 ()
	Frees adjoint Jacobian preconditioner from memory;.

void	applyInverseAdjointJacobianPreconditioner_1 (ROL::Vector< double > &output_vector, const ROL::Vector< double > &input_vector, const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl, double &)
	Applies the inverse Adjoint Jacobian preconditioner. More...

void	applyInverseAdjointJacobian_1 (ROL::Vector< double > &output_vector, const ROL::Vector< double > &input_vector, const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl, double &) override
	Applies the adjoint Jacobian of the Constraints w. r. t. the simulation variables onto a vector.

void	applyJacobian_2 (ROL::Vector< double > &output_vector, const ROL::Vector< double > &input_vector, const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl, double &) override
	Applies the Jacobian of the Constraints w. r. t. the control variables onto a vector.

void	applyAdjointJacobian_2 (ROL::Vector< double > &output_vector, const ROL::Vector< double > &input_vector, const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl, double &) override
	Applies the Jacobian of the Constraints w. r. t. the control variables onto a vector.

void	applyAdjointHessian_11 (ROL::Vector< double > &output_vector, const ROL::Vector< double > &dual, const ROL::Vector< double > &input_vector, const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl, double &tol) override
	Applies the adjoint of the Hessian of the constraints w. r. t. the simulation variables onto a vector. More...

void	applyAdjointHessian_12 (ROL::Vector< double > &output_vector, const ROL::Vector< double > &dual, const ROL::Vector< double > &input_vector, const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl, double &tol) override
	Applies the adjoint of the Hessian of the constraints w. r. t. the simulation variables onto a vector. More...

void	applyAdjointHessian_21 (ROL::Vector< double > &output_vector, const ROL::Vector< double > &dual, const ROL::Vector< double > &input_vector, const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl, double &tol) override
	Applies the adjoint of the Hessian of the constraints w. r. t. the simulation variables onto a vector. More...

void	applyAdjointHessian_22 (ROL::Vector< double > &output_vector, const ROL::Vector< double > &dual, const ROL::Vector< double > &input_vector, const ROL::Vector< double > &des_var_sim, const ROL::Vector< double > &des_var_ctl, double &tol) override
	Applies the adjoint of the Hessian of the constraints w. r. t. the simulation variables onto a vector. More...

Public Attributes
dealii::TrilinosWrappers::SparseMatrix	dXvdXp
	Stores the mesh sensitivities.

double	flow_CFL_
	Regularization of the constraint by adding flow_CFL_ times the mass matrix.

Protected Attributes
int	i_out = 1000
	ID used when outputting the flow solution.

int	iupdate = 9000
	ID used when outputting the flow solution.

Private Attributes
const int	mpi_rank
	MPI rank used for printing.

const bool	i_print
	Whether the current processor should print or not.

std::shared_ptr< DGBase< dim, double > >	dg
	Smart pointer to DGBase.

std::shared_ptr< BaseParameterization< dim > >	design_parameterization
	Parameterization which links design variables to the volume nodes.

Parameters::LinearSolverParam	linear_solver_param
	Linear solver parameters. More...

dealii::LinearAlgebra::distributed::Vector< double >	design_var
	Design variables values.

Ifpack_Preconditioner *	jacobian_prec
	Jacobian preconditioner. More...

Ifpack_Preconditioner *	adjoint_jacobian_prec
	Adjoint Jacobian preconditioner. More...

Detailed Description

template<int dim>
class PHiLiP::FlowConstraints< dim >

Use DGBase as a Simulation Constraint ROL::Constraint_SimOpt.

The simulation variables will be the DoFs stored in the DGBase::solution.

The control variables will be some of the the FFD points/directions. The given ffd_design_variables_indices_dim point to the points/directions used as design variables.

Definition at line 26 of file flow_constraints.hpp.

Member Function Documentation

◆ applyAdjointHessian_11()

template<int dim>

void PHiLiP::FlowConstraints< dim >::applyAdjointHessian_11	(	ROL::Vector< double > &	output_vector,
		const ROL::Vector< double > &	dual,
		const ROL::Vector< double > &	input_vector,
		const ROL::Vector< double > &	des_var_sim,
		const ROL::Vector< double > &	des_var_ctl,
		double &	tol
	)

override

Applies the adjoint of the Hessian of the constraints w. r. t. the simulation variables onto a vector.

More specifically, apply

\[ \mathbf{v}_{output} = \left( \sum_i \psi_i \frac{\partial^2 R_i}{\partial u \partial u} \right)^T \mathbf{v}_{input} \]

onto the input_vector to obtain the output_vector

Definition at line 526 of file flow_constraints.cpp.

◆ applyAdjointHessian_12()

template<int dim>

void PHiLiP::FlowConstraints< dim >::applyAdjointHessian_12	(	ROL::Vector< double > &	output_vector,
		const ROL::Vector< double > &	dual,
		const ROL::Vector< double > &	input_vector,
		const ROL::Vector< double > &	des_var_sim,
		const ROL::Vector< double > &	des_var_ctl,
		double &	tol
	)

override

Applies the adjoint of the Hessian of the constraints w. r. t. the simulation variables onto a vector.

More specifically, apply

\[ \mathbf{v}_{output} = \left( \sum_i \psi_i \frac{\partial^2 R_i}{\partial u \partial x} \right)^T \mathbf{v}_{input} \]

onto the input_vector to obtain the output_vector

Definition at line 553 of file flow_constraints.cpp.

◆ applyAdjointHessian_21()

template<int dim>

void PHiLiP::FlowConstraints< dim >::applyAdjointHessian_21	(	ROL::Vector< double > &	output_vector,
		const ROL::Vector< double > &	dual,
		const ROL::Vector< double > &	input_vector,
		const ROL::Vector< double > &	des_var_sim,
		const ROL::Vector< double > &	des_var_ctl,
		double &	tol
	)

override

Applies the adjoint of the Hessian of the constraints w. r. t. the simulation variables onto a vector.

More specifically, apply

\[ \mathbf{v}_{output} = \left( \sum_i \psi_i \frac{\partial^2 R_i}{\partial x \partial u} \right)^T \mathbf{v}_{input} \]

onto the input_vector to obtain the output_vector

Definition at line 608 of file flow_constraints.cpp.

◆ applyAdjointHessian_22()

template<int dim>

void PHiLiP::FlowConstraints< dim >::applyAdjointHessian_22	(	ROL::Vector< double > &	output_vector,
		const ROL::Vector< double > &	dual,
		const ROL::Vector< double > &	input_vector,
		const ROL::Vector< double > &	des_var_sim,
		const ROL::Vector< double > &	des_var_ctl,
		double &	tol
	)

override

Applies the adjoint of the Hessian of the constraints w. r. t. the simulation variables onto a vector.

More specifically, apply

\[ \mathbf{v}_{output} = \left( \sum_i \psi_i \frac{\partial^2 R_i}{\partial x \partial x} \right)^T \mathbf{v}_{input} \]

onto the input_vector to obtain the output_vector

Parameters

[out]	output_vector	Resulting vector \( \mathbf{v}_{output} \in \mathbb{R}^{N_{ctl}} \)
[in]	dual	Lagrange multiplier associated with constraint vector \( \boldsymbol{\psi} \in \mathbb{R}^{N_{residual}} \)
[in]	input_vector	Input vector \( \mathbf{v}_{input} \in \mathbb{R}^{N_{ctl}} \)
[in]	des_var_sim	Simulation variables \( \mathbf{v}_{input} \in \mathbb{R}^{N_{flow}} \)
[in]	des_var_ctl	Simulation variables \( \mathbf{v}_{input} \in \mathbb{R}^{N_{ctl}} \)
[in]	tol	Tolerance, not used. From virtual ROL function.

Definition at line 667 of file flow_constraints.cpp.

◆ applyInverseAdjointJacobianPreconditioner_1()

template<int dim>

void PHiLiP::FlowConstraints< dim >::applyInverseAdjointJacobianPreconditioner_1	(	ROL::Vector< double > &	output_vector,
		const ROL::Vector< double > &	input_vector,
		const ROL::Vector< double > &	des_var_sim,
		const ROL::Vector< double > &	des_var_ctl,
		double &
	)

Applies the inverse Adjoint Jacobian preconditioner.

construct_AdjointJacobianPreconditioner_1 needs to be called beforehand.

Definition at line 348 of file flow_constraints.cpp.

◆ applyInverseJacobianPreconditioner_1()

template<int dim>

void PHiLiP::FlowConstraints< dim >::applyInverseJacobianPreconditioner_1	(	ROL::Vector< double > &	output_vector,
		const ROL::Vector< double > &	input_vector,
		const ROL::Vector< double > &	des_var_sim,
		const ROL::Vector< double > &	des_var_ctl,
		double &
	)

Applies the inverse Jacobian preconditioner.

construct_JacobianPreconditioner_1 needs to be called beforehand.

Definition at line 316 of file flow_constraints.cpp.

◆ solve()

template<int dim>

void PHiLiP::FlowConstraints< dim >::solve	(	ROL::Vector< double > &	constraint_values,
		ROL::Vector< double > &	des_var_sim,
		const ROL::Vector< double > &	des_var_ctl,
		double &	tol
	)

override

Solve the Simulation Constraint and returns the constraints values.

In this case, we use the ODESolver to solve the steady state solution of the flow given the geometry.

Definition at line 96 of file flow_constraints.cpp.

◆ update_2()

template<int dim>

void PHiLiP::FlowConstraints< dim >::update_2	(	const ROL::Vector< double > &	des_var_ctl,
		bool	flag = `true`,
		int	iter = `-1`
	)

Update the control variables.

Update FFD from design variables and then deforms the mesh.

Definition at line 81 of file flow_constraints.cpp.

Member Data Documentation

◆ adjoint_jacobian_prec

template<int dim>

Ifpack_Preconditioner* PHiLiP::FlowConstraints< dim >::adjoint_jacobian_prec

private

Adjoint Jacobian preconditioner.

Currently uses ILUT

Definition at line 51 of file flow_constraints.hpp.

◆ jacobian_prec

template<int dim>

Ifpack_Preconditioner* PHiLiP::FlowConstraints< dim >::jacobian_prec

private

Jacobian preconditioner.

Currently uses ILUT

Definition at line 48 of file flow_constraints.hpp.

◆ linear_solver_param

template<int dim>

Parameters::LinearSolverParam PHiLiP::FlowConstraints< dim >::linear_solver_param

private

Linear solver parameters.

Currently set such that the linear systems are fully solved

Definition at line 41 of file flow_constraints.hpp.

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

src/optimization/flow_constraints.hpp
src/optimization/flow_constraints.cpp

Public Member Functions

Public Attributes

Protected Attributes

Private Attributes

Detailed Description

template<int dim> class PHiLiP::FlowConstraints< dim >

Member Function Documentation

◆ applyAdjointHessian_11()

◆ applyAdjointHessian_12()

◆ applyAdjointHessian_21()

◆ applyAdjointHessian_22()

◆ applyInverseAdjointJacobianPreconditioner_1()

◆ applyInverseJacobianPreconditioner_1()

◆ solve()

◆ update_2()

Member Data Documentation

◆ adjoint_jacobian_prec

◆ jacobian_prec

◆ linear_solver_param

template<int dim>
class PHiLiP::FlowConstraints< dim >