AdaptiveSubstepper.h

Go to the documentation of this file.

/* ---------------------------------------------------------------------
 *                                       _
 *  _ __ ___   __ _ _ __ _ __ ___   ___ | |_
 * | '_ ` _ \ / _` | '__| '_ ` _ \ / _ \| __|
 * | | | | | | (_| | |  | | | | | | (_) | |_
 * |_| |_| |_|\__,_|_|  |_| |_| |_|\___/ \__|
 *
 * Unit of Strength of Materials and Structural Analysis
 * University of Innsbruck,
 * 2020 - today
 *
 * festigkeitslehre@uibk.ac.at
 *
 * Matthias Neuner matthias.neuner@uibk.ac.at
 *
 * This file is part of the MAteRialMOdellingToolbox (marmot).
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * The full text of the license can be found in the file LICENSE.md at
 * the top level directory of marmot.
 * ---------------------------------------------------------------------
 */
 
#pragma once
#include "Marmot/MarmotJournal.h"
#include "Marmot/MarmotTypedefs.h"
 
namespace Marmot::NumericalAlgorithms {
 
  template < size_t materialTangentSize, size_t nIntegrationDependentStateVars >
  class AdaptiveSubstepper {
  public:
    typedef Eigen::Matrix< double, materialTangentSize, materialTangentSize > TangentSizedMatrix;
    typedef Eigen::Matrix< double, nIntegrationDependentStateVars, 1 > IntegrationStateVector;
 
    AdaptiveSubstepper( double          initialStepSize,
                        double          minimumStepSize,
                        double          maxScaleUpFactor,
                        double          scaleDownFactor,
                        double          integrationErrorTolerance,
                        int             nPassesToIncrease,
                        const Matrix6d& Cel );
 
    void setConvergedProgress( const Marmot::Vector6d& stressOld, const IntegrationStateVector& stateVarsOld );
    bool isFinished();
    double getNextSubstep();
    int getNumberOfSubsteps();
    bool finishSubstep( const Marmot::Vector6d&       resultStress,
                        const TangentSizedMatrix&     dXdY,
                        const IntegrationStateVector& stateVars );
    void finishElasticSubstep( const Marmot::Vector6d& resultStress );
    bool discardSubstep();
    bool repeatSubstep( double decrementationFactor );
 
    void getConvergedProgress( Marmot::Vector6d& stress, IntegrationStateVector& stateVars );
    Matrix6d getCurrentTangentOperator();
    void getResults( Marmot::Vector6d& stress, Matrix6d& consistentTangent, IntegrationStateVector& stateVars );
 
  private:
    const double initialStepSize, minimumStepSize, maxScaleUpFactor, scaleDownFactor, integrationErrorTolerance;
    const int    nPassesToIncrease;
    const bool   ignoreErrorToleranceOnMinimumStepSize;
 
    double currentProgress;
    double currentSubstepSize;
    int    passedSubsteps;
    int    substepIndex;
 
    Marmot::Vector6d       stressProgress;
    IntegrationStateVector stateProgress;
    TangentSizedMatrix     consistentTangentProgress;
 
    // temporal storages, which are used until a cycle full/half/half has finished successfully
    Marmot::Vector6d       stressProgressHalfTemp, stressProgressFullTemp;
    IntegrationStateVector stateProgressHalfTemp, stateProgressFullTemp;
    TangentSizedMatrix     consistentTangentProgressHalfTemp, consistentTangentProgressFullTemp;
 
    TangentSizedMatrix elasticTangent;
 
    enum SubsteppingState { FullStep, FirstHalfStep, SecondHalfStep };
    SubsteppingState currentState;
 
    bool acceptSubstepWithFullStepOnly();
    bool splitCurrentSubstep();
  };
} // namespace Marmot::NumericalAlgorithms
 
namespace Marmot::NumericalAlgorithms {
  template < size_t n, size_t nState >
  AdaptiveSubstepper< n, nState >::AdaptiveSubstepper( double          initialStepSize,
                                                       double          minimumStepSize,
                                                       double          maxScaleUpFactor,
                                                       double          scaleDownFactor,
                                                       double          integrationErrorTolerance,
                                                       int             nPassesToIncrease,
                                                       const Matrix6d& Cel )
    : initialStepSize( initialStepSize ),
      minimumStepSize( minimumStepSize ),
      maxScaleUpFactor( maxScaleUpFactor ),
      scaleDownFactor( scaleDownFactor ),
      integrationErrorTolerance( integrationErrorTolerance ),
      nPassesToIncrease( nPassesToIncrease ),
      ignoreErrorToleranceOnMinimumStepSize( true ),
      currentProgress( 0.0 ),
      currentSubstepSize( initialStepSize ),
      passedSubsteps( 0 ),
      substepIndex( -1 )
  {
    elasticTangent                       = TangentSizedMatrix::Identity();
    elasticTangent.topLeftCorner( 6, 6 ) = Cel;
 
    consistentTangentProgress         = TangentSizedMatrix::Zero();
    consistentTangentProgressFullTemp = TangentSizedMatrix::Zero();
    consistentTangentProgressHalfTemp = TangentSizedMatrix::Zero();
 
    stressProgressHalfTemp.setZero();
    stressProgressFullTemp.setZero();
    stateProgressHalfTemp.setZero();
    stateProgressFullTemp.setZero();
    consistentTangentProgressHalfTemp.setZero();
    consistentTangentProgressFullTemp.setZero();
 
    currentState = FullStep;
  }
 
  template < size_t n, size_t nState >
  void AdaptiveSubstepper< n, nState >::setConvergedProgress( const Marmot::Vector6d&       stressOld,
                                                              const IntegrationStateVector& stateVarsOld )
  {
    stressProgress = stressOld;
    stateProgress  = stateVarsOld;
  }
 
  template < size_t n, size_t nState >
  bool AdaptiveSubstepper< n, nState >::isFinished()
  {
    return ( ( 1.0 - currentProgress ) <= 2e-16 && currentState == FullStep );
  }
 
  template < size_t n, size_t nState >
  double AdaptiveSubstepper< n, nState >::getNextSubstep()
  {
    switch ( currentState ) {
    case FullStep: {
      const double remainingProgress = 1.0 - currentProgress;
      if ( remainingProgress < currentSubstepSize )
        currentSubstepSize = remainingProgress;
      substepIndex++;
      return currentSubstepSize;
      break;
    }
    case FirstHalfStep: {
      return 0.5 * currentSubstepSize;
      break;
    }
    case SecondHalfStep: {
      return 0.5 * currentSubstepSize;
      break;
    }
    default: return -1.0;
    }
  }
  template < size_t n, size_t nState >
  void AdaptiveSubstepper< n, nState >::getConvergedProgress( Marmot::Vector6d&       stress,
                                                              IntegrationStateVector& stateVars )
  {
    switch ( currentState ) {
    case FullStep: {
      stress    = stressProgress;
      stateVars = stateProgress;
      break;
    }
    case FirstHalfStep: {
      stress    = stressProgress;
      stateVars = stateProgress;
      break;
    }
    case SecondHalfStep: {
      stress    = stressProgressHalfTemp;
      stateVars = stateProgressHalfTemp;
      break;
    }
    }
  }
 
  template < size_t n, size_t nState >
  bool AdaptiveSubstepper< n, nState >::discardSubstep()
  {
    passedSubsteps = 0;
    switch ( currentState ) {
    case FullStep: {
      currentSubstepSize *= scaleDownFactor; // we use the scale factor only here
      break;
    }
    // these cases should actually never happen, as the full step has already converged!
    case FirstHalfStep:
      MarmotJournal::warningToMSG( "UMAT: warning, 1th half sub step has not converged after already "
                                   "converged full step" );
      return acceptSubstepWithFullStepOnly();
 
    case SecondHalfStep:
      MarmotJournal::warningToMSG( "UMAT: warning, 2th half sub step has not converged after already "
                                   "converged full step" );
      return acceptSubstepWithFullStepOnly();
    }
 
    currentState = FullStep;
 
    if ( currentSubstepSize < minimumStepSize )
      return MarmotJournal::warningToMSG( "UMAT: Substepper: Minimal stepzsize reached" );
    else
      return true;
  }
 
  template < size_t n, size_t nState >
  bool AdaptiveSubstepper< n, nState >::repeatSubstep( double factorNew )
  {
    currentState   = FullStep;
    passedSubsteps = 0;
 
    currentSubstepSize *= factorNew; // we use the scale factor only here
 
    if ( currentSubstepSize < minimumStepSize )
      return MarmotJournal::warningToMSG( "UMAT: Substepper: Minimal stepzsize reached" );
    else
      return true;
  }
 
  template < size_t n, size_t nState >
  bool AdaptiveSubstepper< n, nState >::finishSubstep( const Marmot::Vector6d&       resultStress,
                                                       const TangentSizedMatrix&     dXdY,
                                                       const IntegrationStateVector& stateVars )
  {
    if ( currentState == FullStep ) {
      stressProgressFullTemp            = resultStress;
      stateProgressFullTemp             = stateVars;
      consistentTangentProgressFullTemp = consistentTangentProgress;
      consistentTangentProgressFullTemp += currentSubstepSize * elasticTangent;
      consistentTangentProgressFullTemp.applyOnTheLeft( dXdY );
      currentState = FirstHalfStep;
      return true;
    }
    else if ( currentState == FirstHalfStep ) {
 
      stressProgressHalfTemp            = resultStress;
      stateProgressHalfTemp             = stateVars;
      consistentTangentProgressHalfTemp = consistentTangentProgress;
      consistentTangentProgressHalfTemp += 0.5 * currentSubstepSize * elasticTangent;
      consistentTangentProgressHalfTemp.applyOnTheLeft( dXdY );
      currentState = SecondHalfStep;
      return true;
    }
 
    else if ( currentState == SecondHalfStep ) {
      // error Estimation
 
      currentState             = FullStep;
      const double error       = ( resultStress - stressProgressFullTemp ).norm();
      const double errorRatio  = error / integrationErrorTolerance;
      double       scaleFactor = 1.0;
      if ( errorRatio > 1e-10 )
        scaleFactor = 0.9 * std::sqrt( 1. / errorRatio );
 
      // saturations
      if ( scaleFactor < 1e-1 )
        scaleFactor = 1e-1;
      if ( scaleFactor * currentSubstepSize < minimumStepSize )
        scaleFactor = minimumStepSize / currentSubstepSize;
      if ( scaleFactor > maxScaleUpFactor )
        scaleFactor = maxScaleUpFactor;
      if ( scaleFactor > 10 )
        scaleFactor = 10;
 
      // Error large than tolerance?
      if ( error > integrationErrorTolerance ) {
        passedSubsteps = 0;
        if ( errorRatio < 2 ) {
          return splitCurrentSubstep();
        }
        else {
          return repeatSubstep( scaleFactor );
        }
      }
      else {
        stressProgressHalfTemp = resultStress;
        stateProgressHalfTemp  = stateVars;
        consistentTangentProgressHalfTemp += 0.5 * currentSubstepSize * elasticTangent;
        consistentTangentProgressHalfTemp.applyOnTheLeft( dXdY );
 
        consistentTangentProgress = 2 * consistentTangentProgressHalfTemp - consistentTangentProgressFullTemp;
        stressProgress            = 2 * stressProgressHalfTemp - stressProgressFullTemp;
        stateProgress             = 2 * stateProgressHalfTemp - stateProgressFullTemp;
 
        currentProgress += currentSubstepSize;
 
        passedSubsteps++;
        currentSubstepSize *= scaleFactor;
 
        return true;
      }
    }
    else
      return false;
  }
 
  template < size_t n, size_t nState >
  void AdaptiveSubstepper< n, nState >::finishElasticSubstep( const Marmot::Vector6d& newStress )
  {
    switch ( currentState ) {
    case FullStep: {
      // this means, that the complete current cycle is already successfull,
      // as the two half steps must also be elastic!
      consistentTangentProgress += currentSubstepSize * elasticTangent;
      stressProgress = newStress;
      // no need for two half steps if full step was already elastic
      currentProgress += currentSubstepSize;
      currentState = FullStep;
      passedSubsteps++;
      break;
    }
 
    case FirstHalfStep: {
      consistentTangentProgressHalfTemp += 0.5 * currentSubstepSize * elasticTangent;
      stressProgressHalfTemp = newStress;
      currentState           = SecondHalfStep;
      break;
    }
    case SecondHalfStep: {
      acceptSubstepWithFullStepOnly();
      break;
    }
    }
  }
 
  template < size_t n, size_t nState >
  void AdaptiveSubstepper< n, nState >::getResults( Marmot::Vector6d&       stress,
                                                    Matrix6d&               consistentTangentOperator,
                                                    IntegrationStateVector& stateVars )
  {
    stress                    = stressProgress;
    stateVars                 = stateProgress;
    consistentTangentOperator = consistentTangentProgress.topLeftCorner( 6, 6 );
  }
 
  template < size_t n, size_t nState >
  bool AdaptiveSubstepper< n, nState >::acceptSubstepWithFullStepOnly()
  {
    consistentTangentProgress = consistentTangentProgressFullTemp;
    stressProgress            = stressProgressFullTemp;
    stateProgress             = stateProgressFullTemp;
 
    currentProgress += currentSubstepSize;
    currentState = FullStep;
 
    return true;
  }
 
  template < size_t n, size_t nState >
  bool AdaptiveSubstepper< n, nState >::splitCurrentSubstep()
  {
    if ( currentSubstepSize < 2 * minimumStepSize ) {
      if ( ignoreErrorToleranceOnMinimumStepSize )
        return acceptSubstepWithFullStepOnly();
      else
        return false;
    }
 
    consistentTangentProgressFullTemp = consistentTangentProgressHalfTemp;
    stressProgressFullTemp            = stressProgressHalfTemp;
    stateProgressFullTemp             = stateProgressHalfTemp;
    currentSubstepSize *= 0.5;
    currentState = FirstHalfStep;
 
    return true;
  }
  template < size_t n, size_t nState >
  int AdaptiveSubstepper< n, nState >::getNumberOfSubsteps()
  {
    return substepIndex;
  }
} // namespace Marmot::NumericalAlgorithms