Generic transaction/cache data; contains no constitutive state. More...

#include "users_modules/eshelbian_plasticity/src/IncrementalOptimization.hpp"

Collaboration diagram for EshelbianPlasticity::IncrementalOptimizationContext:

Public Member Functions
MoFEMErrorCode	rollbackTrial ()

Public Attributes
boost::shared_ptr< IncrementalOptimizationProblem >	problem

SmartPetscObj< Vec >	committedControl

SmartPetscObj< Vec >	candidateControl

SmartPetscObj< Vec >	gradient

SmartPetscObj< Vec >	objectiveGradient

SmartPetscObj< DM >	constraintDM

SmartPetscObj< Vec >	inequalityConstraints

SmartPetscObj< Mat >	inequalityJacobian

SmartPetscObj< Vec >	equilibratedState

SmartPetscObj< Vec >	referenceControl

SmartPetscObj< Vec >	baselineState

SmartPetscObj< Vec >	lastValidState

SmartPetscObj< Vec >	cachedControl

SmartPetscObj< Vec >	cachedState

double	baselineConservativeValue = 0

IncrementalObjectiveEvaluation	cachedEvaluation

EquilibriumSolveFunction	stateSolve

PetscInt	stateSolveCount = 0

PetscInt	rejectedTrialCount = 0

double	equilibriumSeconds = 0

double	gradientAssemblySeconds = 0

double	equilibriumTolerance = 1e-8

bool	baselineValid = false

bool	cacheValid = false

Detailed Description

Generic transaction/cache data; contains no constitutive state.

Terminology used below:

Control: variables selected by TAO. For plasticity its tensor part is Delta H^p, the increment of H^p = log(F^p), which is the logarithmic plastic strain in the symmetric, zero-plastic-spin model. Its scalar part is the epigraph coordinate Delta kappa; committed kappa_n remains in physics-owned mesh storage sharing the same field layout.
Mechanical state: equilibrium unknowns, such as displacement, solved at fixed control by the nested TS/SNES solve.
Trial: one control proposed by TAO and evaluated without changing committed material history.
Rejected trial: a trial for which the nested state solve fails or its residual exceeds equilibriumTolerance; its changes are discarded.
Committed data: the accepted rollback point and physics-owned material history; trial evaluations must not modify them.
Rollback: restore the accepted control and last converged mechanical state after rejecting a trial.
Reference: the zero-increment control at the beginning of a load step.
Baseline: the equilibrated state and conservative value at the reference control; the latter is subtracted to form the energy increment.
Last valid state: the most recent equilibrated trial state, used as the starting point for the next nested solve and as its recovery state.
Cached evaluation: results retained for an identical repeated control, avoiding another equilibrium solve.
Commit: after final validation, atomically update material history and make the accepted state the starting point of the next load step.

Definition at line 151 of file IncrementalOptimization.hpp.

Member Function Documentation

◆ rollbackTrial()

MoFEMErrorCode EshelbianPlasticity::IncrementalOptimizationContext::rollbackTrial ( )

Definition at line 14 of file IncrementalOptimization.cpp.

                                                             {
  MoFEMFunctionBegin;
  CHKERR VecCopy(committedControl, candidateControl);
  CHKERR VecGhostUpdateBegin(candidateControl, INSERT_VALUES, SCATTER_FORWARD);
  CHKERR VecGhostUpdateEnd(candidateControl, INSERT_VALUES, SCATTER_FORWARD);
  MoFEMFunctionReturn(0);
}

Member Data Documentation

◆ baselineConservativeValue

double EshelbianPlasticity::IncrementalOptimizationContext::baselineConservativeValue = 0

Conservative value at referenceControl, subtracted from the objective.

Definition at line 183 of file IncrementalOptimization.hpp.

◆ baselineState

SmartPetscObj<Vec> EshelbianPlasticity::IncrementalOptimizationContext::baselineState

Equilibrated-state snapshot corresponding to referenceControl.

Definition at line 175 of file IncrementalOptimization.hpp.

◆ baselineValid

bool EshelbianPlasticity::IncrementalOptimizationContext::baselineValid = false

Whether the reference objective value and state are initialized.

Definition at line 201 of file IncrementalOptimization.hpp.

◆ cachedControl

SmartPetscObj<Vec> EshelbianPlasticity::IncrementalOptimizationContext::cachedControl

Control associated with cachedState and cachedEvaluation.

Definition at line 179 of file IncrementalOptimization.hpp.

◆ cachedEvaluation

IncrementalObjectiveEvaluation EshelbianPlasticity::IncrementalOptimizationContext::cachedEvaluation

Values and optional conservative gradient for cachedControl/cachedState. The complete objective gradient is a mutable TAO workspace, not cached.

Definition at line 186 of file IncrementalOptimization.hpp.

◆ cachedState

SmartPetscObj<Vec> EshelbianPlasticity::IncrementalOptimizationContext::cachedState

Equilibrated state associated with cachedControl.

Definition at line 181 of file IncrementalOptimization.hpp.

◆ cacheValid

bool EshelbianPlasticity::IncrementalOptimizationContext::cacheValid = false

Whether cachedControl exactly identifies a reusable evaluation.

Definition at line 203 of file IncrementalOptimization.hpp.

◆ candidateControl

SmartPetscObj<Vec> EshelbianPlasticity::IncrementalOptimizationContext::candidateControl

Current TAO trial control, synchronized before physics evaluation.

Definition at line 158 of file IncrementalOptimization.hpp.

◆ committedControl

SmartPetscObj<Vec> EshelbianPlasticity::IncrementalOptimizationContext::committedControl

Last accepted control; source for rollback after a rejected trial.

Definition at line 156 of file IncrementalOptimization.hpp.

◆ constraintDM

SmartPetscObj<DM> EshelbianPlasticity::IncrementalOptimizationContext::constraintDM

Rectangular problem mapping controls to inequality constraints.

Definition at line 164 of file IncrementalOptimization.hpp.

◆ equilibratedState

SmartPetscObj<Vec> EshelbianPlasticity::IncrementalOptimizationContext::equilibratedState

Mechanical state equilibrated for candidateControl.

Definition at line 171 of file IncrementalOptimization.hpp.

◆ equilibriumSeconds

double EshelbianPlasticity::IncrementalOptimizationContext::equilibriumSeconds = 0

Accumulated wall time spent in nested equilibrium solves.

Definition at line 195 of file IncrementalOptimization.hpp.

◆ equilibriumTolerance

double EshelbianPlasticity::IncrementalOptimizationContext::equilibriumTolerance = 1e-8

Maximum accepted residual norm from a nested equilibrium solve.

Definition at line 199 of file IncrementalOptimization.hpp.

◆ gradient

SmartPetscObj<Vec> EshelbianPlasticity::IncrementalOptimizationContext::gradient

Reduced conservative gradient evaluated at candidateControl.

Definition at line 160 of file IncrementalOptimization.hpp.

◆ gradientAssemblySeconds

double EshelbianPlasticity::IncrementalOptimizationContext::gradientAssemblySeconds = 0

Accumulated wall time spent assembling conservative gradients.

Definition at line 197 of file IncrementalOptimization.hpp.

◆ inequalityConstraints

SmartPetscObj<Vec> EshelbianPlasticity::IncrementalOptimizationContext::inequalityConstraints

Values of the TAO inequalities, using the convention c_i >= 0.

Definition at line 166 of file IncrementalOptimization.hpp.

◆ inequalityJacobian

SmartPetscObj<Mat> EshelbianPlasticity::IncrementalOptimizationContext::inequalityJacobian

Jacobian of the inequality map with respect to the control.

Definition at line 168 of file IncrementalOptimization.hpp.

◆ lastValidState

SmartPetscObj<Vec> EshelbianPlasticity::IncrementalOptimizationContext::lastValidState

Latest converged state, used to warm-start or restore trial solves.

Definition at line 177 of file IncrementalOptimization.hpp.

◆ objectiveGradient

SmartPetscObj<Vec> EshelbianPlasticity::IncrementalOptimizationContext::objectiveGradient

Complete objective gradient, including local incremental resistance.

Definition at line 162 of file IncrementalOptimization.hpp.

◆ problem

boost::shared_ptr<IncrementalOptimizationProblem> EshelbianPlasticity::IncrementalOptimizationContext::problem

Physics-specific callbacks used by the generic optimiser.

Definition at line 153 of file IncrementalOptimization.hpp.

◆ referenceControl

SmartPetscObj<Vec> EshelbianPlasticity::IncrementalOptimizationContext::referenceControl

Control defining zero increment at the start of the load step.

Definition at line 173 of file IncrementalOptimization.hpp.

◆ rejectedTrialCount

PetscInt EshelbianPlasticity::IncrementalOptimizationContext::rejectedTrialCount = 0

Number of TAO trials rejected because equilibrium was not obtained.

Definition at line 193 of file IncrementalOptimization.hpp.

◆ stateSolve

EquilibriumSolveFunction EshelbianPlasticity::IncrementalOptimizationContext::stateSolve

Fixed-control state solve callback.

Definition at line 189 of file IncrementalOptimization.hpp.

◆ stateSolveCount

PetscInt EshelbianPlasticity::IncrementalOptimizationContext::stateSolveCount = 0

Number of nested equilibrium solves performed in this context.

Definition at line 191 of file IncrementalOptimization.hpp.

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

users_modules/eshelbian_plasticity/src/IncrementalOptimization.hpp
users_modules/eshelbian_plasticity/src/impl/IncrementalOptimization.cpp

Public Member Functions

Public Attributes