]> https://gitweb.dealii.org/ - code-gallery.git/commitdiff
Traveling waves 184/head
authormagomedovs <shamil96@mail.ru>
Mon, 1 Jul 2024 17:12:51 +0000 (20:12 +0300)
committermagomedovs <shamil96@mail.ru>
Mon, 1 Jul 2024 17:12:51 +0000 (20:12 +0300)
31 files changed:
TravelingWaves/AuxiliaryFunctions.h [new file with mode: 0644]
TravelingWaves/CMakeLists.txt [new file with mode: 0644]
TravelingWaves/IntegrateSystem.h [new file with mode: 0644]
TravelingWaves/LimitSolution.cc [new file with mode: 0644]
TravelingWaves/LimitSolution.h [new file with mode: 0644]
TravelingWaves/LinearInterpolator.h [new file with mode: 0644]
TravelingWaves/Parameters.cc [new file with mode: 0644]
TravelingWaves/Parameters.h [new file with mode: 0644]
TravelingWaves/ParametersList.prm [new file with mode: 0644]
TravelingWaves/ParametersListDeflagrationFast.prm [new file with mode: 0644]
TravelingWaves/ParametersListDeflagrationSlow.prm [new file with mode: 0644]
TravelingWaves/ParametersListDetonation.prm [new file with mode: 0644]
TravelingWaves/README.md [new file with mode: 0644]
TravelingWaves/Solution.cc [new file with mode: 0644]
TravelingWaves/Solution.h [new file with mode: 0644]
TravelingWaves/TravelingWaveSolver.cc [new file with mode: 0644]
TravelingWaves/TravelingWaveSolver.h [new file with mode: 0644]
TravelingWaves/calculate_profile.cc [new file with mode: 0644]
TravelingWaves/calculate_profile.h [new file with mode: 0644]
TravelingWaves/doc/author [new file with mode: 0644]
TravelingWaves/doc/builds-on [new file with mode: 0644]
TravelingWaves/doc/dependencies [new file with mode: 0644]
TravelingWaves/doc/entry-name [new file with mode: 0644]
TravelingWaves/doc/pics/detonation_delta_0.001.png [new file with mode: 0644]
TravelingWaves/doc/pics/detonation_delta_0.01.png [new file with mode: 0644]
TravelingWaves/doc/pics/fast_deflagration_delta_0.01.png [new file with mode: 0644]
TravelingWaves/doc/pics/slow_deflagration_delta_0.01.png [new file with mode: 0644]
TravelingWaves/doc/pics/sparsity-pattern_p1.png [new file with mode: 0644]
TravelingWaves/doc/tooltip [new file with mode: 0644]
TravelingWaves/main.cc [new file with mode: 0644]
TravelingWaves/plot.py [new file with mode: 0644]

diff --git a/TravelingWaves/AuxiliaryFunctions.h b/TravelingWaves/AuxiliaryFunctions.h
new file mode 100644 (file)
index 0000000..f773070
--- /dev/null
@@ -0,0 +1,36 @@
+#ifndef AUXILIARY_FUNCTIONS
+#define AUXILIARY_FUNCTIONS
+
+#include <vector>
+#include <cmath>
+#include <fstream>
+#include <string>
+#include <cstring>
+
+// Comparison of numbers with a given tolerance.
+template <typename T>
+bool isapprox(const T &a, const T &b, const double tol = 1e-10)
+{
+  return (std::abs( a - b ) < tol);
+}
+
+// Fill the std::vector with the values from the range [interval_begin, interval_end].
+template <typename T>
+void linspace(T interval_begin, T interval_end, std::vector<T> &arr) 
+{
+  const size_t SIZE = arr.size();
+  const T step = (interval_end - interval_begin) / static_cast<T>(SIZE - 1);
+  for (size_t i = 0; i < SIZE; ++i) 
+  {
+    arr[i] = interval_begin + i * step;
+  }
+}
+
+// Check the file existence.
+inline bool file_exists(const std::string &filename) 
+{
+  std::ifstream f(filename.c_str());
+  return f.good();
+}
+
+#endif
diff --git a/TravelingWaves/CMakeLists.txt b/TravelingWaves/CMakeLists.txt
new file mode 100644 (file)
index 0000000..dfc346e
--- /dev/null
@@ -0,0 +1,43 @@
+##
+#  CMake script for the TravelingWaves program:
+##
+
+# The name of the project and target:
+SET(TARGET "main")
+
+SET(TARGET_SRC
+  ${TARGET}.cc calculate_profile.cc Solution.cc TravelingWaveSolver.cc Parameters.cc LimitSolution.cc 
+  )
+
+CMAKE_MINIMUM_REQUIRED(VERSION 3.13.4)
+
+FIND_PACKAGE(deal.II 9.5.0
+  HINTS ${deal.II_DIR} ${DEAL_II_DIR} ../ ../../ $ENV{DEAL_II_DIR}
+  )
+IF(NOT ${deal.II_FOUND})
+  MESSAGE(FATAL_ERROR "\n"
+    "*** Could not locate a (sufficiently recent) version of deal.II. ***\n\n"
+    "You may want to either pass a flag -DDEAL_II_DIR=/path/to/deal.II to cmake\n"
+    "or set an environment variable \"DEAL_II_DIR\" that contains this path."
+    )
+ENDIF()
+
+#
+# Are all dependencies fulfilled?
+#
+IF(NOT (DEAL_II_WITH_UMFPACK AND DEAL_II_WITH_SUNDIALS)) # keep in one line
+  MESSAGE(FATAL_ERROR "
+    Error! This program requires a deal.II library that was configured with the following options:
+        DEAL_II_WITH_UMFPACK  = ON
+        DEAL_II_WITH_SUNDIALS = ON
+    However, the deal.II library found at ${DEAL_II_PATH} was configured with these options:
+        DEAL_II_WITH_UMFPACK  = ${DEAL_II_WITH_UMFPACK}
+        DEAL_II_WITH_SUNDIALS = ${DEAL_II_WITH_SUNDIALS}
+    This conflicts with the requirements."
+    )
+ENDIF()
+
+DEAL_II_INITIALIZE_CACHED_VARIABLES()
+PROJECT(${TARGET})
+
+DEAL_II_INVOKE_AUTOPILOT()
diff --git a/TravelingWaves/IntegrateSystem.h b/TravelingWaves/IntegrateSystem.h
new file mode 100644 (file)
index 0000000..1078293
--- /dev/null
@@ -0,0 +1,103 @@
+
+#ifndef INTEGRATE_SYSTEM
+#define INTEGRATE_SYSTEM
+
+#include <boost/numeric/odeint.hpp>
+
+#include <iostream>
+#include <fstream>
+#include <string>
+
+template <typename state_T, typename time_T>
+void SaveSolutionIntoFile(const std::vector<state_T>& x_vec, const std::vector<time_T>& t_vec, std::string filename="output_ode_sol.txt") 
+{
+  if (!x_vec.empty() && !t_vec.empty()) 
+  {
+    std::ofstream output(filename);
+    output << std::setprecision(16);
+
+    size_t dim = x_vec[0].size();
+    for (size_t i = 0; i < t_vec.size(); ++i) 
+    {
+      output << std::fixed << t_vec[i];
+      for (size_t j = 0; j < dim; ++j) 
+      {
+        output << std::scientific << " " << x_vec[i][j];
+      }
+      output << "\n";
+    }
+    output.close();
+  } 
+  else 
+  {
+    std::cout << "Solution is not saved into file.\n";
+  }
+}
+
+// type of RK integrator 
+enum class Integrator_Type 
+{
+  dopri5,
+  cash_karp54,
+  fehlberg78
+};
+
+// Observer
+template <typename state_type>
+class Push_back_state_time
+{
+public:
+  std::vector<state_type>& m_states;
+  std::vector<double>& m_times;
+
+  Push_back_state_time(std::vector<state_type>& states, std::vector<double>& times)
+    : m_states(states), m_times(times) 
+  {}
+
+  void operator() (const state_type& x, double t) 
+  {
+    m_states.push_back(x);
+    m_times.push_back(t);
+  }
+};
+
+
+// Integrate system at specified points. 
+template <typename ODE_obj_T, typename state_type, typename Iterable_type>
+void IntegrateSystemAtTimePoints(
+  std::vector<state_type>& x_vec, std::vector<double>& t_vec, const Iterable_type& iterable_time_span,
+  const ODE_obj_T& ode_system_obj, 
+  state_type& x, const double dt,
+  Integrator_Type integrator_type=Integrator_Type::dopri5, bool save_to_file_flag=false,
+  const double abs_er_tol=1.0e-15, const double rel_er_tol=1.0e-15
+  )
+{
+  using namespace boost::numeric::odeint;
+
+  if (integrator_type == Integrator_Type::dopri5) 
+  {
+    typedef runge_kutta_dopri5< state_type > error_stepper_type;
+    integrate_times( make_controlled< error_stepper_type >(abs_er_tol, rel_er_tol),
+              ode_system_obj, x, iterable_time_span.begin(), iterable_time_span.end(), dt, Push_back_state_time< state_type >(x_vec, t_vec) );
+  } 
+  else if (integrator_type == Integrator_Type::cash_karp54) 
+  {
+    typedef runge_kutta_cash_karp54< state_type > error_stepper_type;
+    integrate_times( make_controlled< error_stepper_type >(abs_er_tol, rel_er_tol),
+              ode_system_obj, x, iterable_time_span.begin(), iterable_time_span.end(), dt, Push_back_state_time< state_type >(x_vec, t_vec) );
+  }
+  else 
+  { // Integrator_Type::fehlberg78
+    typedef runge_kutta_fehlberg78< state_type > error_stepper_type;
+    integrate_times( make_controlled< error_stepper_type >(abs_er_tol, rel_er_tol),
+              ode_system_obj, x, iterable_time_span.begin(), iterable_time_span.end(), dt, Push_back_state_time< state_type >(x_vec, t_vec) );
+  }
+
+  if (save_to_file_flag) 
+  {
+    SaveSolutionIntoFile(x_vec, t_vec);
+  } 
+
+}
+
+#endif
\ No newline at end of file
diff --git a/TravelingWaves/LimitSolution.cc b/TravelingWaves/LimitSolution.cc
new file mode 100644 (file)
index 0000000..039bba7
--- /dev/null
@@ -0,0 +1,75 @@
+#include "LimitSolution.h"
+
+namespace TravelingWave
+{
+
+  LimitSolution::LimitSolution(const Parameters &parameters, const double ilambda_0, const double iu_0, const double iT_0, const double iroot_sign)
+    : params(parameters)
+    , problem(params.problem)
+    , wave_speed(problem.wave_speed_init)
+    , lambda_0(ilambda_0)
+    , u_0(iu_0)
+    , T_0(iT_0)
+    , root_sign(iroot_sign)
+  {
+    calculate_constants_A_B();
+  }
+
+  double LimitSolution::omega_func(const double lambda, const double T) const
+  {
+    return problem.k * (1. - lambda) * std::exp(-problem.theta / T);
+  }
+
+  void LimitSolution::operator() (const state_type &x , state_type &dxdt , const double /* t */)
+  {
+    dxdt[0] = -1. / wave_speed * omega_func(x[0], T_func(x[0]));
+  }
+
+  double LimitSolution::u_func(const double lambda) const
+  {
+    double coef = 2 * (wave_speed - 1) / problem.epsilon - 1;
+    return (coef + root_sign * std::sqrt(coef * coef - 4 * (problem.q * lambda + B - 2 * A / problem.epsilon))) / 2;
+  }
+
+  double LimitSolution::T_func(const double lambda) const
+  {
+    return u_func(lambda) + problem.q * lambda + B;
+  }
+
+  void LimitSolution::calculate_constants_A_B()
+  {
+    B = T_0 - u_0;
+    A = u_0 * (1 - wave_speed) + problem.epsilon * (u_0 * u_0  + T_0) / 2;
+  }
+
+  void LimitSolution::set_wave_speed(double iwave_speed)
+  {
+    wave_speed = iwave_speed;
+    calculate_constants_A_B();
+  }
+
+  void LimitSolution::calculate_u_T_omega()
+  {
+    if (!t_vec.empty() && !lambda_vec.empty())
+    {
+      u_vec.resize(lambda_vec.size());
+      T_vec.resize(lambda_vec.size());
+      omega_vec.resize(lambda_vec.size());
+      for (unsigned int i = 0; i < lambda_vec.size(); ++i)
+      {
+        u_vec[i].resize(1);
+        T_vec[i].resize(1);
+        omega_vec[i].resize(1);
+
+        u_vec[i][0] = u_func(lambda_vec[i][0]);
+        T_vec[i][0] = T_func(lambda_vec[i][0]);
+        omega_vec[i][0] = omega_func(lambda_vec[i][0], T_vec[i][0]);
+      }
+    }
+    else
+    {
+      std::cout << "t_vec or lambda_vec vector is empty!" << std::endl;
+    }
+  }
+
+} // namespace TravelingWave
\ No newline at end of file
diff --git a/TravelingWaves/LimitSolution.h b/TravelingWaves/LimitSolution.h
new file mode 100644 (file)
index 0000000..e5dcc20
--- /dev/null
@@ -0,0 +1,47 @@
+#ifndef LIMIT_SOLUTION
+#define LIMIT_SOLUTION
+
+#include "Parameters.h"
+#include <iostream>
+#include <vector>
+
+namespace TravelingWave
+{
+  typedef std::vector< double > state_type;
+
+  class LimitSolution
+  {
+  public:
+    LimitSolution(const Parameters &parameters, const double ilambda_0, const double iu_0, const double iT_0, const double root_sign = 1.);
+
+    void operator() (const state_type &x , state_type &dxdt , const double /* t */);
+    void calculate_u_T_omega();
+    void set_wave_speed(double iwave_speed);
+
+    std::vector<double> t_vec;
+    std::vector<state_type> omega_vec;
+    std::vector<state_type> lambda_vec;
+    std::vector<state_type> u_vec;
+    std::vector<state_type> T_vec;
+
+  private:
+    double omega_func(const double lambda, const double T) const;
+    double u_func(const double lambda) const;
+    double T_func(const double lambda) const;
+
+    void calculate_constants_A_B();
+
+    const Parameters &params;
+    const Problem    &problem;
+    double wave_speed;
+
+    const double lambda_0, u_0, T_0;  // Initial values.
+    double A, B;                      // Integration constants.
+
+    const double root_sign;           // Plus or minus one.
+  };
+
+
+} // namespace TravelingWave
+
+#endif
\ No newline at end of file
diff --git a/TravelingWaves/LinearInterpolator.h b/TravelingWaves/LinearInterpolator.h
new file mode 100644 (file)
index 0000000..7c2720d
--- /dev/null
@@ -0,0 +1,59 @@
+#ifndef LINEAR_INTERPOLATOR
+#define LINEAR_INTERPOLATOR
+
+#include <cmath>
+#include <algorithm>
+#include <vector>
+
+// Linear interpolation class
+template <typename Number_Type>
+class LinearInterpolator
+{
+public:
+  LinearInterpolator(const std::vector<Number_Type> &ix_points, const std::vector<Number_Type> &iy_points);
+  Number_Type value(const Number_Type x) const;
+
+private:
+  const std::vector<Number_Type> x_points;     // Must be an increasing sequence, i.e. x[i] < x[i+1]
+  const std::vector<Number_Type> y_points;
+};
+
+template <typename Number_Type>
+LinearInterpolator<Number_Type>::LinearInterpolator(const std::vector<Number_Type> &ix_points, const std::vector<Number_Type> &iy_points)
+  : x_points(ix_points)
+  , y_points(iy_points)
+{}
+
+template <typename Number_Type>
+Number_Type LinearInterpolator<Number_Type>::value(const Number_Type x) const
+{
+  Number_Type res = 0.;
+
+  auto lower = std::lower_bound(x_points.begin(), x_points.end(), x);
+  unsigned int right_index = 0;
+  unsigned int left_index = 0;
+  if (lower == x_points.begin())
+  {
+    res = y_points[0];
+  }
+  else if (lower == x_points.end())
+  {
+    res = y_points[x_points.size()-1];
+  }
+  else
+  {
+    right_index = lower - x_points.begin();
+    left_index = right_index - 1;
+
+    Number_Type y_2 = y_points[right_index];
+    Number_Type y_1 = y_points[left_index];
+    Number_Type x_2 = x_points[right_index];
+    Number_Type x_1 = x_points[left_index];
+
+    res = (y_2 - y_1) / (x_2 - x_1) * (x - x_1) + y_1;
+  }
+
+  return res;
+}
+
+#endif
\ No newline at end of file
diff --git a/TravelingWaves/Parameters.cc b/TravelingWaves/Parameters.cc
new file mode 100644 (file)
index 0000000..ed55a0a
--- /dev/null
@@ -0,0 +1,52 @@
+#include "Parameters.h"
+
+namespace TravelingWave
+{
+  using namespace dealii;
+  
+  Problem::Problem() 
+    : ParameterAcceptor("Problem")
+  {
+         add_parameter("delta", delta = 0.01); 
+       add_parameter("epsilon", epsilon = 0.1);            
+       add_parameter("Prandtl number", Pr = 0.75);        
+       add_parameter("Lewis number", Le = 1.0);      
+       add_parameter("Constant of reaction rate", k = 1.0);
+       add_parameter("Activation energy", theta = 1.65);
+       add_parameter("Heat release", q = 1.7);
+    add_parameter("Ignition Temperature", T_ign = 1.0);
+    add_parameter("Type of the wave (deflagration / detonation)", wave_type = 1);   // 0 for "deflagration"; 1 for "detonation".
+
+    add_parameter("Type of boundary condition for the temperature at the right boundary", T_r_bc_type = 1);   // 0 for "Neumann" (deflagration); 1 for "Dirichlet" (detonation).
+
+    add_parameter("T_left", T_left = 5.3);    // Dirichlet boundary condition.
+    add_parameter("T_right", T_right = 0.9);  // For detonation waves the value serves as a Dirichlet boundary condition. For deflagration waves it serves for construction of the piecewise constant initial guess.
+    add_parameter("u_left", u_left = -0.2);   // For detonation waves the value is ignored. For deflagration waves it serves for construction of the piecewise constant initial guess.
+    add_parameter("u_right", u_right = 0.);   // Dirichlet boundary condition.
+
+    add_parameter("Initial guess for the wave speed", wave_speed_init = 1.2);   // For detonation waves the value is ignored. For deflagration waves it serves as an initial guess for the wave speed.
+  }
+
+  FiniteElements::FiniteElements()
+    : ParameterAcceptor("Finite elements")
+  {
+    add_parameter("Polynomial degree", poly_degree = 1);
+    add_parameter("Number of quadrature points", quadrature_points_number = 3);
+  }
+
+  Mesh::Mesh()
+         : ParameterAcceptor("Mesh")
+  {
+         add_parameter("Interval left boundary", interval_left = -50.0);  
+       add_parameter("Interval right boundary", interval_right = 20.0);
+       add_parameter<unsigned int>("Refinements number", refinements_number = 10);
+       add_parameter("Adaptive mesh refinement", adaptive = 1);    // 1 for adaptive; 0 for global.
+  }
+
+  Solver::Solver()
+    : ParameterAcceptor("Solver")
+  {
+    add_parameter("Tolerance", tol = 1e-10);
+  }
+
+} // namespace TravelingWave
diff --git a/TravelingWaves/Parameters.h b/TravelingWaves/Parameters.h
new file mode 100644 (file)
index 0000000..2dd02c6
--- /dev/null
@@ -0,0 +1,61 @@
+#ifndef PARAMETERS
+#define PARAMETERS
+
+#include <deal.II/base/parameter_acceptor.h>
+
+namespace TravelingWave
+{
+  using namespace dealii;
+
+  struct Problem : ParameterAcceptor
+  {
+    Problem();
+
+    double delta, epsilon;
+    double Pr, Le;
+    double k, theta, q;
+    double T_ign;
+    int wave_type;
+    int T_r_bc_type;
+    double T_left, T_right;
+    double u_left, u_right;
+    
+    double wave_speed_init;
+  };
+
+  struct FiniteElements : ParameterAcceptor
+  {
+         FiniteElements();
+
+    unsigned int poly_degree;
+    unsigned int quadrature_points_number;
+  };
+
+  struct Mesh : ParameterAcceptor
+  {
+    Mesh();
+    
+    double interval_left;
+    double interval_right;
+    unsigned int refinements_number;
+    int adaptive;
+  };
+
+  struct Solver : ParameterAcceptor
+  {
+    Solver();
+
+    double       tol;
+  };
+
+  struct Parameters
+  {
+    Problem           problem;
+    FiniteElements             fe;
+    Mesh                                     mesh;
+    Solver                   solver;
+  };
+
+} // namespace TravelingWave
+
+#endif
diff --git a/TravelingWaves/ParametersList.prm b/TravelingWaves/ParametersList.prm
new file mode 100644 (file)
index 0000000..7092dbb
--- /dev/null
@@ -0,0 +1,40 @@
+#      List of dimensionless parameters
+
+subsection Problem
+  set delta                         = 0.01
+  set epsilon                       = 0.1
+  set Prandtl number                = 0.75    # Pr
+  set Lewis number                  = 1.0     # Le
+  set Constant of reaction rate     = 1.0     # k
+  set Activation energy             = 1.65    # theta
+  set Heat release                  = 1.7     # q
+  set Ignition Temperature          = 1.0     # T_ign
+
+  set Type of the wave (deflagration / detonation) = 0    # 0 for "deflagration"; 1 for "detonation".
+  set Type of boundary condition for the temperature at the right boundary = 0    # 0 for "Neumann" (deflagration); 1 for "Dirichlet" (detonation).
+  set T_left    = 2.3       # Dirichlet boundary condition.
+  set T_right   = 0.9       # For detonation waves the value serves as a Dirichlet boundary condition. For deflagration waves it serves for construction of the piecewise constant initial guess.
+
+  set u_left    = -0.2      # For detonation waves the value is ignored. For deflagration waves it serves for construction of the piecewise constant initial guess.
+  set u_right   = 0.1       # Dirichlet boundary condition.
+
+  set Initial guess for the wave speed = 0.2    # For detonation waves the value is ignored. For deflagration waves it serves as an initial guess for the wave speed.
+
+end
+
+subsection Finite elements
+  set Polynomial degree             = 1
+  set Number of quadrature points   = 3
+end
+
+subsection Mesh
+  set Interval left boundary        = -50      
+  set Interval right boundary       = 20
+  set Refinements number            = 10
+  set Adaptive mesh refinement      = 1     # 1 for adaptive; 0 for global.
+end
+
+subsection Solver
+  set Tolerance                     = 1e-10
+end
diff --git a/TravelingWaves/ParametersListDeflagrationFast.prm b/TravelingWaves/ParametersListDeflagrationFast.prm
new file mode 100644 (file)
index 0000000..91ad5b2
--- /dev/null
@@ -0,0 +1,40 @@
+#      List of dimensionless parameters
+
+subsection Problem
+  set delta                         = 0.01
+  set epsilon                       = 0.1
+  set Prandtl number                = 0.75    # Pr
+  set Lewis number                  = 1.0     # Le
+  set Constant of reaction rate     = 1.0     # k
+  set Activation energy             = 1.65    # theta
+  set Heat release                  = 1.7     # q
+  set Ignition Temperature          = 1.0     # T_ign
+
+  set Type of the wave (deflagration / detonation) = 0    # 0 for "deflagration"; 1 for "detonation".
+  set Type of boundary condition for the temperature at the right boundary = 0    # 0 for "Neumann" (deflagration); 1 for "Dirichlet" (detonation).
+  set T_left    = 2.3       # Dirichlet boundary condition.
+  set T_right   = 0.9       # For detonation waves the value serves as a Dirichlet boundary condition. For deflagration waves it serves for construction of the piecewise constant initial guess.
+
+  set u_left    = -0.2      # For detonation waves the value is ignored. For deflagration waves it serves for construction of the piecewise constant initial guess.
+  set u_right   = 0.1       # Dirichlet boundary condition.
+
+  set Initial guess for the wave speed = 0.9    # For detonation waves the value is ignored. For deflagration waves it serves as an initial guess for the wave speed.
+
+end
+
+subsection Finite elements
+  set Polynomial degree             = 1
+  set Number of quadrature points   = 3
+end
+
+subsection Mesh
+  set Interval left boundary        = -50      
+  set Interval right boundary       = 20
+  set Refinements number            = 10
+  set Adaptive mesh refinement      = 1     # 1 for adaptive; 0 for global.
+end
+
+subsection Solver
+  set Tolerance                     = 1e-10
+end
diff --git a/TravelingWaves/ParametersListDeflagrationSlow.prm b/TravelingWaves/ParametersListDeflagrationSlow.prm
new file mode 100644 (file)
index 0000000..7092dbb
--- /dev/null
@@ -0,0 +1,40 @@
+#      List of dimensionless parameters
+
+subsection Problem
+  set delta                         = 0.01
+  set epsilon                       = 0.1
+  set Prandtl number                = 0.75    # Pr
+  set Lewis number                  = 1.0     # Le
+  set Constant of reaction rate     = 1.0     # k
+  set Activation energy             = 1.65    # theta
+  set Heat release                  = 1.7     # q
+  set Ignition Temperature          = 1.0     # T_ign
+
+  set Type of the wave (deflagration / detonation) = 0    # 0 for "deflagration"; 1 for "detonation".
+  set Type of boundary condition for the temperature at the right boundary = 0    # 0 for "Neumann" (deflagration); 1 for "Dirichlet" (detonation).
+  set T_left    = 2.3       # Dirichlet boundary condition.
+  set T_right   = 0.9       # For detonation waves the value serves as a Dirichlet boundary condition. For deflagration waves it serves for construction of the piecewise constant initial guess.
+
+  set u_left    = -0.2      # For detonation waves the value is ignored. For deflagration waves it serves for construction of the piecewise constant initial guess.
+  set u_right   = 0.1       # Dirichlet boundary condition.
+
+  set Initial guess for the wave speed = 0.2    # For detonation waves the value is ignored. For deflagration waves it serves as an initial guess for the wave speed.
+
+end
+
+subsection Finite elements
+  set Polynomial degree             = 1
+  set Number of quadrature points   = 3
+end
+
+subsection Mesh
+  set Interval left boundary        = -50      
+  set Interval right boundary       = 20
+  set Refinements number            = 10
+  set Adaptive mesh refinement      = 1     # 1 for adaptive; 0 for global.
+end
+
+subsection Solver
+  set Tolerance                     = 1e-10
+end
diff --git a/TravelingWaves/ParametersListDetonation.prm b/TravelingWaves/ParametersListDetonation.prm
new file mode 100644 (file)
index 0000000..5d10f08
--- /dev/null
@@ -0,0 +1,40 @@
+#      List of dimensionless parameters
+
+subsection Problem
+  set delta                         = 0.01
+  set epsilon                       = 0.1
+  set Prandtl number                = 0.75    # Pr
+  set Lewis number                  = 1.0     # Le
+  set Constant of reaction rate     = 1.0     # k
+  set Activation energy             = 1.65    # theta
+  set Heat release                  = 1.7     # q
+  set Ignition Temperature          = 1.0     # T_ign
+
+  set Type of the wave (deflagration / detonation) = 1    # 0 for "deflagration"; 1 for "detonation".
+  set Type of boundary condition for the temperature at the right boundary = 1    # 0 for "Neumann" (deflagration); 1 for "Dirichlet" (detonation).
+  set T_left    = 5.3       # Dirichlet boundary condition.
+  set T_right   = 0.9       # For detonation waves the value serves as a Dirichlet boundary condition. For deflagration waves it serves for construction of the piecewise constant initial guess.
+
+  set u_left    = -0.2      # For detonation waves the value is ignored. For deflagration waves it serves for construction of the piecewise constant initial guess.
+  set u_right   = 0.       # Dirichlet boundary condition.
+
+  set Initial guess for the wave speed = 0.2    # For detonation waves the value is ignored. For deflagration waves it serves as an initial guess for the wave speed.
+
+end
+
+subsection Finite elements
+  set Polynomial degree             = 1
+  set Number of quadrature points   = 3
+end
+
+subsection Mesh
+  set Interval left boundary        = -50      
+  set Interval right boundary       = 20
+  set Refinements number            = 10
+  set Adaptive mesh refinement      = 1     # 1 for adaptive; 0 for global.
+end
+
+subsection Solver
+  set Tolerance                     = 1e-10
+end
diff --git a/TravelingWaves/README.md b/TravelingWaves/README.md
new file mode 100644 (file)
index 0000000..3c6c842
--- /dev/null
@@ -0,0 +1,273 @@
+# Traveling-wave solutions of a qualitative model for combustion waves
+
+This program demonstrates the use of adaptive finite elements to compute traveling-wave solutions of partial differential equations. One of the challenges in solving this type of problem is the presence of an unknown wave speed. Below we show how to overcome this.
+
+
+## Building, compiling and running
+To run the program, enter the following commands from the current directory:
+```
+mkdir build && cd build
+cmake ..
+make
+./main ../ParametersList.prm
+```
+
+## Problem
+To illustrate the algorithm for the computation of traveling-wave profiles, we consider a combustion model described in [1]. In a moving reference frame of a steadily propagating wave we have the following nondimensionalized system:
+@f{align*}{
+    - c u_{\xi} + (1 + \epsilon u) u_{\xi} &= -\dfrac{\epsilon}{2} T_{\xi} + \dfrac{4 \delta}{3 \epsilon} \Pr u_{\xi \xi}, \\
+    - c(T_{\xi} - u_{\xi}) &= q \omega + \delta T_{\xi \xi}, \\
+    - c \lambda_{\xi} &= \omega + \dfrac{\delta}{\text{Le}} \lambda_{\xi \xi}.
+@f}
+Here, $u$ is the pressure, $T$ is the temperature, $\lambda$ is the reaction-progress variable, varying from $0$ in the fresh mixture to $1$ in the burnt products, $c > 0$ is the unknown wave speed, $\Pr$ and $\mathrm{Le}$ are the Prandtl and Lewis numbers, $q$ is the energy of heat release. The model parameters $\epsilon$ and $\delta$ determine the strength of nonlinearity and dissipative effects. The reaction rate $\omega$ is taken as
+@f{align*}{
+\omega = k (1 - \lambda) \exp(-\theta / T) \, \mathrm{H}(T - T_{\mathrm{ign}}),
+@f}
+with activation energy $\theta$, ignition temperature $T_{\mathrm{ign}}$, constant of the reaction rate $k$, and the Heaviside
+step function $\mathrm{H}$.
+
+The boundary conditions at $\xi = -\infty$ are
+@f{align*}{
+    u_{\xi} = 0, \ T = T_l, \ \lambda = 1,
+@f}
+and at $\xi = +\infty$ are
+@f{align*}{
+    u = u_r, \ \lambda = 0.
+@f}
+The right boundary condition for temperature is $T = T_r$ for detonation waves (supersonic regime, i.e. $c > 1$) and $T_{\xi} = 0$ for deflagration waves (subsonic regime, i.e. $c < 1$).
+
+Because of translational invariance, we need to impose another constraint on the system to fix a particular solution. So we choose the following centering condition: $T(0) = T_{\mathrm{ign}}$.
+
+
+## Numerical algorithm
+
+### Newton–Raphson iteration scheme
+The nonlinear boundary value problem is solved numerically on a finite interval $I = [l, r]$ $\left(|l|, |r| \gg 1 \right)$, using a Newton–Raphson iteration scheme, similar to one, described in deal.II tutorial [step-15](https://www.dealii.org/current/doxygen/deal.II/step_15.html). The main difference from step-15 is that we have an additional scalar unknown, the front velocity $c$. So the algorithm has to be modified to take this feature into account.
+
+Rewriting the system in a vector form
+@f{align*}{
+    \mathbf{F}(u, T, \lambda, c) = 
+    \left(\begin{array}{c}
+         \dfrac{4 \delta}{3 \epsilon} \Pr u_{\xi \xi} - u_{\xi} (1 - c + \epsilon u) -\dfrac{\epsilon}{2} T_{\xi}  \\[7pt]
+         \delta T_{\xi \xi} + c(T_{\xi} - u_{\xi}) + q \omega \\[5pt]
+         \dfrac{\delta}{\text{Le}} \lambda_{\xi \xi} + c \lambda_{\xi} + \omega
+    \end{array} \right)
+    = 0,
+@f}
+we define a Newton–Raphson iteration as
+@f{align*}{
+    \mathbf{F'}(\mathbf{x}^k, \mathbf{dx}^k) = - \mathbf{F}(\mathbf{x}^k),
+@f}
+
+@f{align*}{
+    \mathbf{x}^{n+1} = \mathbf{x}^{n} + \alpha^k \mathbf{dx}^k,
+@f}
+where $k$ is the step number, $\mathbf{x}^k = (u^k, T^k, \lambda^k, c^k)^{\top}$ is a vector argument, $\mathbf{dx}^k = (du^k, dT^k, d\lambda^k, dc^k)^{\top}$ is an increment, $\alpha^k$ is some damping parameter for managing the global convergence behavior and $\mathbf{F'}(\mathbf{x}^k, \mathbf{dx}^k)$ is the directional derivative, defined as 
+@f{align*}{
+    \mathbf{F'}(\mathbf{x}, \mathbf{dx}) = \dfrac{\mathrm{d}}{\mathrm{d} \varepsilon} \Big|_{\varepsilon=0} \mathbf{F}(\mathbf{x} &+ \varepsilon \mathbf{dx}).
+@f}
+
+The system to be solved at every iteration step to obtain the increment $\mathbf{dx}^k = (du^k, dT^k, d\lambda^k, dc^k)^{\top}$ can be represented in matrix vector notation as follows
+@f{align*}{
+    \begin{pmatrix}
+    \dfrac{4 \delta}{3 \epsilon} \Pr \partial_{\xi \xi} - (1 - c + \epsilon u)\partial_{\xi} - \epsilon u_{\xi} & -\dfrac{\epsilon}{2} \partial_{\xi} & 0 & u_{\xi} \\[9pt]
+    -c \partial_{\xi} & \delta \partial_{\xi \xi} + c \partial_{\xi} + q \kappa_1 & q \kappa_2 &  T_{\xi} - u_{\xi} \\[9pt]
+    0 & \kappa_1 & \dfrac{\delta}{\text{Le}} \partial_{\xi \xi} + c \partial_{\xi} + \kappa_2 &  \lambda_{\xi}
+    \end{pmatrix} 
+    \begin{pmatrix}
+        du \\[9pt]
+        dT \\[9pt]
+        d\lambda \\[9pt]
+        dc
+    \end{pmatrix} = -\begin{pmatrix}
+        f_1 \\[9pt]
+        f_2 \\[9pt]
+        f_3
+    \end{pmatrix},
+@f}
+where
+@f{align*}{
+    \kappa_1 &= k (1 - \lambda) \exp(-\theta / T) \left[ \dfrac{\theta }{T^2} \, \text{H}(T - T_{\text{ign}}) + \delta(T - T_{\text{ign}}) \right], \\
+    \kappa_2 &= - k \exp(-\theta / T) \, \text{H}(T - T_{\text{ign}}),
+@f}
+in which $\delta(\cdot)$ is a Dirac delta function, and $f_i \, (i=1,2,3)$ are the components of the vector function $\mathbf{F}(u, T, \lambda, c)$. The term $\delta(T - T_{\text{ign}})$ can be rewritten as 
+@f{align*}{
+    \delta(T - T_{\text{ign}}) = \frac{\delta(\xi)}{|T'(0)|}.
+@f}
+
+We choose the initial guess $\mathbf{x}^0$ to include the appropriate boundary values, therefore the update $\mathbf{dx}^k$ uses homogeneous Dirichlet or Neumann boundary conditions.
+
+### Weak formulation
+We multiply both sides of the equation for $\mathbf{dx}^k$ with vector valued test function $\mathbf{v} = (v_1, v_2, v_3)^{\top}$ and integrate over the domain $\Omega$ to obtain a scalar equation
+@f{align*}{
+    J(\mathbf{dx}, \mathbf{v}) = -b(\mathbf{v}).
+@f}
+
+@f{align*}{
+    J(\mathbf{dx}, &\mathbf{v}) = \int \limits_{\Omega} \mathbf{v} \cdot \mathbf{F'}(\mathbf{x}, \mathbf{dx}) \, d\xi = \\
+    = &\dfrac{4 \delta}{3 \epsilon} \Pr (-\partial_{\xi} v_1, \partial_{\xi} du) + (v_1, - (1 - c + \epsilon u)\partial_{\xi} du - \epsilon u_{\xi} du -\dfrac{\epsilon}{2} \partial_{\xi} dT + u_{\xi} dc) + \\ 
+    &+ \delta (-\partial_{\xi} v_2, \partial_{\xi} dT) + (v_2, -c \, \partial_{\xi} du + c \, \partial_{\xi} dT + q \kappa_1 dT + q \kappa_2 d\lambda + T_{\xi} dc - u_{\xi} dc) + \\
+    &+ \dfrac{\delta}{\text{Le}} (-\partial_{\xi} v_3, \partial_{\xi} d\lambda) + (v_3, \kappa_1 dT + c \partial_{\xi} d\lambda + \kappa_2 d\lambda + \lambda_{\xi} dc).
+@f}
+
+@f{align*}{
+    b(\mathbf{v}) = \int \limits_{\Omega} &\mathbf{v} \cdot \mathbf{F}(\mathbf{x}) \, d\xi = \\
+    = &\dfrac{4 \delta}{3 \epsilon} \Pr (-\partial_{\xi} v_1, u_{\xi}) + (v_1, - u_{\xi} (1 - c + \epsilon u) -\dfrac{\epsilon}{2} T_{\xi}) + \\
+    &+ \delta (-\partial_{\xi} v_2, T_{\xi}) + (v_2, c(T_{\xi} - u_{\xi}) + q \omega) + \\
+    &+ \dfrac{\delta}{\text{Le}} (-\partial_{\xi} v_3, \lambda_{\xi}) + (v_3, c \lambda_{\xi} + \omega).
+@f}
+In the above expressions second derivatives disappear due to integration by parts with homogeneous Dirichlet and Neumann boundary conditions. The solution is sought as an expansion
+@f{align*}{
+    \begin{pmatrix}
+        du \\[9pt]
+        dT \\[9pt]
+        d\lambda \\[9pt]
+        dc
+    \end{pmatrix} = 
+    \sum \limits_{i = 1}^{3N} 
+    U_{i}\begin{pmatrix}
+        \phi_i^1 \\[9pt]
+        \phi_i^2 \\[9pt]
+        \phi_i^3 \\[9pt]
+        0
+    \end{pmatrix} +
+    U_{3N + 1}\begin{pmatrix}
+        0 \\[9pt]
+        0 \\[9pt]
+        0 \\[9pt]
+        1
+    \end{pmatrix} \quad \in \quad V_p \times V_p \times V_p \times \mathbb{R}.
+@f}
+where $V_p$ is a finite element space of continuous, piecewise polynomials of degree $p$. The set of vector functions $(\phi_i^1, \phi_i^2, \phi_i^3)^{\top} \in  V_p^3$ form the basis of the corresponding space. We then choose test functions $\mathbf{v}$ to be the same as the basis functions, and obtain the linear system $J U = b$. Elements of the matrix and the right-hand side are computed according to following formulas:
+@f{align*}{
+    J_{ij} = &\dfrac{4 \delta}{3 \epsilon} \Pr (-\partial_{\xi} \phi_i^1, \partial_{\xi} \phi_j^1) + (\phi_i^1, - (1 - c + \epsilon u)\partial_{\xi} \phi_j^1 - \epsilon u_{\xi} \phi_j^1 -\dfrac{\epsilon}{2} \partial_{\xi} \phi_j^2) + \\ 
+    &+ \delta (-\partial_{\xi} \phi_i^2, \partial_{\xi} \phi_j^2) + (\phi_i^2, -c \, \partial_{\xi} \phi_j^1 + c \, \partial_{\xi} \phi_j^2 + q \kappa_1 \phi_j^2 + q \kappa_2 \phi_j^3) + \\
+    &+ \dfrac{\delta}{\text{Le}} (-\partial_{\xi} \phi_i^3, \partial_{\xi} \phi_j^3) + (\phi_i^3, \kappa_1 \phi_j^2 + c \partial_{\xi} \phi_j^3 + \kappa_2 \phi_j^3), 
+@f}
+
+@f{align*}{
+    J_{i, 3N + 1} = (\phi_i^1, u_{\xi}) + (\phi_i^2, T_{\xi} - u_{\xi}) + (\phi_i^3, \lambda_{\xi}),
+@f}
+
+@f{align*}{
+    b_{i} = &\dfrac{4 \delta}{3 \epsilon} \Pr (-\partial_{\xi} \phi_i^1, u_{\xi}) + (\phi_i^1, - u_{\xi} (1 - c + \epsilon u) -\dfrac{\epsilon}{2} T_{\xi}) + \\
+    &+ \delta (-\partial_{\xi} \phi_i^2, T_{\xi}) + (\phi_i^2, c(T_{\xi} - u_{\xi}) + q \omega) + \\
+    &+ \dfrac{\delta}{\text{Le}} (-\partial_{\xi} \phi_i^3, \lambda_{\xi}) + (\phi_i^3, c \lambda_{\xi} + \omega),
+@f}
+for $i, j < 3N + 1$.
+
+In order for the system to have a unique solution, we need to supplement it with one more equation, corresponding to the constraint $T(0) = T_{\text{ign}}$. The initial approximation to the solution is set so as to satisfy this condition, so we just need the computed increment function $dT$ to be zero at the specified point. Thus, we add a row of zeros with a value of 1 in the position corresponding to $dT(0)$ to the matrix $J$ and set $b_{3N + 1} = 0$.
+
+The resulting sparsity pattern structure has the form shown in the figure below.
+![Sparsity pattern corresponding to basis polynomial degree $p=1$.](doc/pics/sparsity-pattern_p1.png)
+
+The integration of the terms with $\kappa_1$ need special attention because of the Dirac delta function. If $U_n$ and $U_m$ are the degrees of freedom, associated with the vertex $\xi = 0$ (i.e., $\phi_n^2(0) = 1$ and $\phi_m^3(0) = 1$), we get
+@f{align*}{
+(\phi_n^2, q (k (1 - \lambda) \exp(-\theta / T) \delta(T - T_{\text{ign}})) \phi_n^2) = \dfrac{q k (1 - \lambda(0)) \exp(-\theta / T(0))}{|T'(0)|}
+@f}
+and
+@f{align*}{
+(\phi_m^3, (k (1 - \lambda) \exp(-\theta / T) \delta(T - T_{\text{ign}})) \phi_n^2) = \dfrac{k (1 - \lambda(0)) \exp(-\theta / T(0))}{|T'(0)|}.
+@f}
+
+
+### Initial guess
+The initial guess for detonation wave is obtained from the following problem
+@f{align*}{
+    u_{\xi} (- c + 1 + \epsilon u) &= -\dfrac{\epsilon}{2} T_{\xi} , \\
+    - c(T_{\xi} - u_{\xi}) &= q \omega, \\
+    - c \lambda_{\xi} &= \omega,
+@f}
+which is the limiting case of the system at $\delta = 0$. The problem reduces to the nonlinear initial value problem
+@f{align*}{
+    \lambda_{\xi} = -\dfrac{k}{c} (1 - \lambda) \exp \left( \dfrac{-\theta}{T(\lambda)} \right), 
+@f}
+with initial condition $\lambda(0) = 0$; see [1] for details.
+
+For the deflagration case, the initial guess is taken piecewise constant for $u$ and $T$, and
+@f{align*}{
+    \lambda(\xi) = 
+    \begin{cases}
+    -\exp \left(\xi (1 - c) \Big/ \left(\dfrac{4 \delta}{3 \epsilon} \Pr \right) \right) + 1 \quad &\mathrm{for}\ \xi \in [l, 0],  \\
+    0 \quad &\mathrm{for}\ \xi \in (0, r]
+    \end{cases}
+@f}
+for the reaction-progress variable. The value in the interval $(0, 1)$ is chosen as the initial guess for the front velocity $c$.
+
+
+### Boundary conditions
+In the numerical solution, the boundary conditions described in the [beginning](#problem) are imposed at the ends of the interval $I$. In addition, a homogeneous Neumann condition is applied to the function $d\lambda$ at the left boundary.
+
+
+## Program
+
+### Parameters
+The calculation parameters are set in the `ParametersList.prm` file. To reproduce the results obtained below, you can run the program with the parameter files `ParametersListDeflagrationSlow.prm`, `ParametersListDeflagrationFast.prm` and `ParametersListDetonation.prm`.
+
+### Class `TravelingWaveSolver`
+`TravelingWaveSolver` is the main class for computation of the traveling-wave profiles.
+
+The implementation of Newton's method is based on that described in [step-77](https://www.dealii.org/current/doxygen/deal.II/step_77.html) and relies on SUNDIALS' [KINSOL](https://computing.llnl.gov/projects/sundials/kinsol) package. Because of the additional unknown, the front velocity, we expand the Jacobi matrix by one column and one row (`jacobian_matrix_extended`), and add one more element to the solution vector (`current_solution_extended`). After completing the Newton iterations, we split the resulting extended solution vector `current_solution_extended` into two parts: the solution vector `current_solution`, corresponding to $(u, T, \lambda)$, and the front velocity `current_wave_speed`. After that the adaptive mesh refinement is performed using the `current_solution` vector, which is very important for resolving a narrow transition layer with a large solution gradient in the vicinity of zero. The [KellyErrorEstimator](https://www.dealii.org/current/doxygen/deal.II/classKellyErrorEstimator.html) is used as a refinement indicator.
+
+### Function `calculate_profile`
+The full calculation cycle is done in the `calculate_profile` function. First, we construct an initial guess to the solution depending on the selected wave type and store the result as an object of type `SolutionStruct`. This object, along with the problem parameters, is then passed to the constructor of the `TravelingWaveSolver` class to calculate the traveling wave.
+
+Decreasing the dissipation parameter $\delta$ leads to the appearance of large gradients in solutions in the neighborhood of zero. As a consequence, Newton's method becomes more sensitive to the initial data and ceases to converge. To solve this problem, the `calculate_profile` function implements the method of continuation by the $\delta$ parameter (for an example, see [step-57](https://www.dealii.org/current/doxygen/deal.II/step_57.html)). The solution and the refined triangulation are saved after each step of the method using the `get_solution` and `get_triangulation` functions and then passed to the next step.
+
+
+### Error estimation
+Integration of the [governing equations](#problem) over the real line gives the following relations:
+@f{align*}{
+    u_l (1 - c) + \frac{\epsilon}{2} u_l^2 + \frac{\epsilon}{2} T_l &=  u_r (1 - c) + \frac{\epsilon}{2} u_r^2 + \frac{\epsilon}{2} T_r , \\
+    T_l - u_l &= T_r - u_r + q.
+@f}
+These relations let us express any two parameters of $c, T_l, T_r, u_l, u_r$ in terms of the remaining three. Thus, we can write
+@f{align*}{
+    u_l &= (T_l - T_r) + u_r - q, \\
+    c &= 1 + \epsilon \left( u_r - \dfrac{(q - (T_l - T_r))^2 + (T_l - T_r)}{2 (q - (T_l - T_r))} \right).
+@f}
+This means that since we choose the three parameters $T_l, T_r, u_r$ for the detonation case ourselves, the above formulas give us the exact values of $c$ and $u_l$. These can be used to obtain the value of the error in the calculated $c$ and $u_l$.
+
+For the deflagration case, however, we can only choose two parameters, $T_l$ and $u_r$. The remaining three are determined during the solution, so the formulas can only give us an error estimate.
+
+
+### Initial guess
+To get the initial condition for detonation, we have to solve the nonlinear initial value problem for $\lambda$ we mentioned earlier. This is done in the `LimitSolution` class. Numerical integration is performed using the [odeint](https://www.boost.org/doc/libs/1_85_0/libs/numeric/odeint/doc/html/index.html) library of the [Boost](https://www.boost.org) with its interface in `IntegrateSystem.h`.
+
+
+## Results
+
+To visualize the computed profiles, one can use gnuplot typing
+
+`plot for [i=2:4] "solution_filename" using 1:i w p title word("u T lambda", i-1)`
+
+or execute the python script `plot.py` 
+
+`python plot.py "solution_filename"`
+
+### Slow deflagration for $\delta = 0.01$
+The calculated wave speed is $c = 0.0909$.
+![Slow deflagration, $\delta = 0.01$, $c = 0.0909$](doc/pics/slow_deflagration_delta_0.01.png)
+
+
+### Fast deflagration for $\delta = 0.01$
+The calculated wave speed is $c = 0.8252$.
+![Fast deflagration, $\delta = 0.01$, $0.8252$](doc/pics/fast_deflagration_delta_0.01.png)
+
+
+### Detonation for $\delta = 0.01$ and $\delta = 0.001$
+The calculated wave speed in both cases is the same $c = 1.216481$, as expected. Solid lines represent the detonation profile for the ideal case, when $\delta=0$. 
+
+![Detonation, $\delta = 0.01$, $c = 1.216481$.](doc/pics/detonation_delta_0.01.png)
+
+![Detonation, $\delta = 0.001$, $c = 1.216481$](doc/pics/detonation_delta_0.001.png)
+
+
+## Acknowledgments
+
+I would like to thank my friend Oleg Rogozin for introducing me to the deal.II library and the world of finite elements.
+
+
+## References
+
+1. [Goldin A.Y., Magomedov S.M., Faria L.M., Kasimov A.R. Study of a qualitative model for combustion waves: Flames, detonations, and deflagration-to-detonation transition. Computers & Fluids 2024; 273:106213.](https://doi.org/10.1016/j.compfluid.2024.106213)
\ No newline at end of file
diff --git a/TravelingWaves/Solution.cc b/TravelingWaves/Solution.cc
new file mode 100644 (file)
index 0000000..8087337
--- /dev/null
@@ -0,0 +1,90 @@
+#include "Solution.h"
+
+namespace TravelingWave
+{
+
+  using namespace dealii;
+  
+  SolutionStruct::SolutionStruct() {}
+  SolutionStruct::SolutionStruct(const std::vector<double> &ix, const std::vector<double> &iu, 
+                      const std::vector<double> &iT, const std::vector<double> &ilambda, double iwave_speed)
+    : x(ix)
+    , u(iu)
+    , T(iT)
+    , lambda(ilambda)
+    , wave_speed(iwave_speed)
+  {}
+  SolutionStruct::SolutionStruct(const std::vector<double> &ix, const std::vector<double> &iu, 
+                      const std::vector<double> &iT, const std::vector<double> &ilambda)
+    : SolutionStruct(ix, iu, iT, ilambda, 0.)
+  {}
+
+  void SolutionStruct::reinit(const unsigned int number_of_elements)
+  {
+    wave_speed = 0.;
+    x.clear();
+    u.clear();
+    T.clear();
+    lambda.clear();
+
+    x.resize(number_of_elements);
+    u.resize(number_of_elements);
+    T.resize(number_of_elements);
+    lambda.resize(number_of_elements);
+  }
+
+  void SolutionStruct::save_to_file(std::string filename = "sol") const
+  {
+    const std::string file_for_solution = filename + ".txt";
+    std::ofstream output(file_for_solution);
+
+    output << std::scientific << std::setprecision(16);
+    for (unsigned int i = 0; i < x.size(); ++i)
+    {
+      output << std::fixed << x[i]; 
+      output << std::scientific << " " << u[i] << " " << T[i] << " " << lambda[i] << "\n";
+    }
+    output.close();
+
+    std::ofstream file_for_wave_speed_output("wave_speed-" + file_for_solution);
+    file_for_wave_speed_output << std::scientific << std::setprecision(16);
+    file_for_wave_speed_output << wave_speed << std::endl;
+    file_for_wave_speed_output.close();
+  }
+
+
+  Interpolant::Interpolant(const std::vector<double> &ix_points, const std::vector<double> &iy_points)
+    : interpolant(ix_points, iy_points)
+  {}
+
+  double Interpolant::value(const Point<1> &p, const unsigned int component) const
+  {
+    double x = p[0];
+    double res = interpolant.value(x);
+
+    return res;
+  }
+
+
+  template <typename InterpolantType>
+  SolutionVectorFunction<InterpolantType>::SolutionVectorFunction(InterpolantType iu_interpolant, InterpolantType iT_interpolant, InterpolantType ilambda_interpolant) 
+    : Function<1>(3)
+    , u_interpolant(iu_interpolant)
+    , T_interpolant(iT_interpolant)
+    , lambda_interpolant(ilambda_interpolant)
+  {}
+
+  template <typename InterpolantType>
+  double SolutionVectorFunction<InterpolantType>::value(const Point<1> &p, const unsigned int component) const
+  {
+    double res = 0.;
+    if (component == 0)                        {       res = u_interpolant.value(p); }
+    else if (component == 1)   { res = T_interpolant.value(p); }
+    else if (component == 2)   {       res = lambda_interpolant.value(p);      }
+
+    return res;
+  }
+
+  template class SolutionVectorFunction<Interpolant>;
+
+} // namespace TravelingWave
\ No newline at end of file
diff --git a/TravelingWaves/Solution.h b/TravelingWaves/Solution.h
new file mode 100644 (file)
index 0000000..2de6619
--- /dev/null
@@ -0,0 +1,67 @@
+#ifndef SOLUTION_STRUCT
+#define SOLUTION_STRUCT
+
+#include <deal.II/base/function.h>
+
+#include "LinearInterpolator.h"
+
+#include <vector>
+#include <string>
+#include <fstream>
+#include <iostream>
+#include <iomanip>
+
+
+namespace TravelingWave
+{
+  using namespace dealii;
+
+  // The structure for keeping the solution: arrays of coordinates $\xi$, solution $u$, $T$, $\lambda$, and the wave speed $c$. 
+  struct SolutionStruct
+  {
+    SolutionStruct();
+    SolutionStruct(const std::vector<double> &ix, const std::vector<double> &iu, 
+              const std::vector<double> &iT, const std::vector<double> &ilambda, const double iwave_speed);
+    SolutionStruct(const std::vector<double> &ix, const std::vector<double> &iu, 
+              const std::vector<double> &iT, const std::vector<double> &ilambda);
+
+    void reinit(const unsigned int number_of_elements);
+    
+    void save_to_file(std::string filename) const;
+
+    std::vector<double> x;                             // mesh coordinates (must be an increasing sequence)
+    std::vector<double> u;                             // array of u components
+    std::vector<double> T;                             // array of T components
+    std::vector<double> lambda;                // array of lambda components
+
+    double wave_speed;                                         // speed of the wave
+  };
+
+  // Interpolation class
+  class Interpolant : public Function<1>
+  {
+  public:
+    Interpolant(const std::vector<double> &ix_points, const std::vector<double> &iy_points);
+    virtual double value(const Point<1> &p, const unsigned int component = 0) const override;
+
+  private:
+    LinearInterpolator<double> interpolant;
+  };
+
+  // Vector function $(u(p), T(p), \lambda(p))$
+  template <typename InterpolantType>
+  class SolutionVectorFunction : public Function<1>
+  {
+  public:
+    SolutionVectorFunction(InterpolantType iu_interpolant, InterpolantType iT_interpolant, InterpolantType ilambda_interpolant);
+    virtual double value(const Point<1> &p, const unsigned int component = 0) const override;
+
+  private:
+    InterpolantType u_interpolant;
+    InterpolantType T_interpolant;
+    InterpolantType lambda_interpolant;
+  };
+
+} // namespace TravelingWave
+
+#endif
\ No newline at end of file
diff --git a/TravelingWaves/TravelingWaveSolver.cc b/TravelingWaves/TravelingWaveSolver.cc
new file mode 100644 (file)
index 0000000..09b67ea
--- /dev/null
@@ -0,0 +1,874 @@
+#include "TravelingWaveSolver.h"
+
+namespace TravelingWave
+{
+  using namespace dealii;
+
+  // Constructor of the class that takes parameters of the problem and an initial guess for Newton's iterations.
+  TravelingWaveSolver::TravelingWaveSolver(const Parameters &parameters, const SolutionStruct &initial_guess_input) 
+    : params(parameters)
+    , problem(params.problem)
+    , number_of_quadrature_points((params.fe.quadrature_points_number > 0) ? params.fe.quadrature_points_number : (params.fe.poly_degree + 1))
+    , triangulation_uploaded(false)
+    , fe(FE_Q<1>(params.fe.poly_degree), 1, 
+          FE_Q<1>(params.fe.poly_degree), 1,
+          FE_Q<1>(params.fe.poly_degree), 1)    // 3 fe basis sets, corresponding to du, dT, dlambda
+    , dof_handler(triangulation)
+    , current_wave_speed(0.)
+    , initial_guess(initial_guess_input)
+    , computing_timer(std::cout, TimerOutput::never, TimerOutput::wall_times)
+  {
+    // Table with values of some parameters to be written to the standard output before calculations.
+    TableHandler table;
+    table.add_value("Parameter name", "number of quadrature points");
+    table.add_value("value", number_of_quadrature_points);
+
+    table.add_value("Parameter name", "delta");
+    table.add_value("value", params.problem.delta);
+
+    table.add_value("Parameter name", "epsilon");
+    table.add_value("value", params.problem.epsilon);
+    
+    table.add_value("Parameter name", "params.problem.wave_speed_init");
+    table.add_value("value", params.problem.wave_speed_init);
+
+    table.add_value("Parameter name", "initial_guess.wave_speed");
+    table.add_value("value", initial_guess.wave_speed);
+    
+    table.add_value("Parameter name", "T_left");
+    table.add_value("value", params.problem.T_left);
+    
+    table.set_precision("value", 2);
+    table.set_scientific("value", true);
+    
+    std::cout << "\n";
+    table.write_text(std::cout, TableHandler::TextOutputFormat::org_mode_table);
+    std::cout << "\n";
+  }
+
+  // A function that takes a triangulation and assigns it to the member variable <code>triangulation </code>.
+  void TravelingWaveSolver::set_triangulation(const Triangulation<1> &itriangulation)
+  {
+    triangulation.clear();
+    triangulation.copy_triangulation(itriangulation);
+    triangulation_uploaded = true;
+  }
+
+  // Here we find the indices of the degrees of freedom, associated with the boundary vertices, and the degree of freedom, associated with the vertex with coordinate $\xi = 0$, and corresponding to temperature.
+  void TravelingWaveSolver::find_boundary_and_centering_dof_numbers()
+  {
+    for (const auto &cell : dof_handler.active_cell_iterators())
+    {
+      for (const auto &v_ind : cell->vertex_indices())
+      {
+        if (isapprox(cell->vertex(v_ind)[0], params.mesh.interval_left))
+        {
+          boundary_and_centering_dof_numbers["u_left"]        = cell->vertex_dof_index(v_ind, 0);
+          boundary_and_centering_dof_numbers["T_left"]        = cell->vertex_dof_index(v_ind, 1);
+          boundary_and_centering_dof_numbers["lambda_left"]   = cell->vertex_dof_index(v_ind, 2);
+        }
+        else if (isapprox(cell->vertex(v_ind)[0], params.mesh.interval_right))
+        {
+          boundary_and_centering_dof_numbers["u_right"]       = cell->vertex_dof_index(v_ind, 0);
+          boundary_and_centering_dof_numbers["T_right"]       = cell->vertex_dof_index(v_ind, 1);
+          boundary_and_centering_dof_numbers["lambda_right"]  = cell->vertex_dof_index(v_ind, 2);
+        }
+        else if (isapprox(cell->vertex(v_ind)[0], 0.))
+        {
+          boundary_and_centering_dof_numbers["T_zero"]        = cell->vertex_dof_index(v_ind, 1);
+        }
+      }
+    }
+  }
+
+  // Set solution values, corresponding to Dirichlet boundary conditions and the centering condition $T(0) = T_{\mathrm{ign}}$.
+  void TravelingWaveSolver::set_boundary_and_centering_values()
+  {
+    current_solution[boundary_and_centering_dof_numbers["u_right"]] = problem.u_right;
+
+    current_solution[boundary_and_centering_dof_numbers["T_left"]] = problem.T_left;
+    if (problem.T_r_bc_type == 1)      // 1 for "Dirichlet"
+    {
+      current_solution[boundary_and_centering_dof_numbers["T_right"]] = problem.T_right;
+    } // else is 0 for "Neumann"
+    current_solution[boundary_and_centering_dof_numbers["T_zero"]] = problem.T_ign;
+
+    current_solution[boundary_and_centering_dof_numbers["lambda_right"]] = 0.;
+  }
+
+
+  void TravelingWaveSolver::setup_system(const bool initial_step)
+  {
+    TimerOutput::Scope t(computing_timer, "set up");
+
+    dof_handler.distribute_dofs(fe);
+
+    std::cout << "Number of dofs : " << dof_handler.n_dofs() << std::endl;
+
+    extended_solution_dim = dof_handler.n_dofs() + 1;
+
+    find_boundary_and_centering_dof_numbers();
+
+    // Boundary condition constraints for $du$, $dT$ and $d\lambda$.
+    zero_boundary_constraints.clear();
+
+    // Dirichlet homogeneous boundary condition for $du$ at the right boundary.
+    zero_boundary_constraints.add_line(boundary_and_centering_dof_numbers["u_right"]);
+
+    // Dirichlet homogeneous boundary condition for $dT$ at the left boundary.
+    zero_boundary_constraints.add_line(boundary_and_centering_dof_numbers["T_left"]);
+    // For the temperature at the left boundary there are two possibilities:
+    if (problem.T_r_bc_type == 1)      // 1 for "Dirichlet"
+    {
+      std::cout << "Dirichlet condition for the temperature at the right boundary." << std::endl;
+      zero_boundary_constraints.add_line(boundary_and_centering_dof_numbers["T_right"]);
+    } // else is 0 for "Neumann"
+    else
+    {
+      std::cout << "Neumann condition for the temperature at the right boundary." << std::endl;
+    }
+
+    // Dirichlet homogeneous boundary condition for $d\lambda$ at the right boundary. (At the left boundary we consider the homogeneous Neumann boundary condition for $d\lambda$.)
+    zero_boundary_constraints.add_line(boundary_and_centering_dof_numbers["lambda_right"]);
+    
+    zero_boundary_constraints.close();
+
+    // We create extended dynamic sparsity pattern with an additional row and an additional column.
+    DynamicSparsityPattern dsp(extended_solution_dim);
+    {
+      std::vector<types::global_dof_index> dofs_on_this_cell;
+      dofs_on_this_cell.reserve(dof_handler.get_fe_collection().max_dofs_per_cell());
+
+      for (const auto &cell : dof_handler.active_cell_iterators())
+      {
+        const unsigned int dofs_per_cell = cell->get_fe().n_dofs_per_cell();
+        dofs_on_this_cell.resize(dofs_per_cell);
+        cell->get_dof_indices(dofs_on_this_cell);
+
+        zero_boundary_constraints.add_entries_local_to_global(dofs_on_this_cell,
+                              dsp,
+                              /*keep_constrained_dofs*/ true);
+      }
+
+      // Adding elements to the last column.
+      for (unsigned int i = 0; i < extended_solution_dim; ++i)
+      {
+        dsp.add(i, extended_solution_dim - 1);
+      }
+      // Adding one element to the last row, corresponding to the T(0).
+      dsp.add(extended_solution_dim - 1, boundary_and_centering_dof_numbers["T_zero"]);
+    }
+    
+    // Initialization
+    sparsity_pattern_extended.copy_from(dsp);
+    jacobian_matrix_extended.reinit(sparsity_pattern_extended); 
+    jacobian_matrix_extended_factorization.reset();
+
+    current_solution_extended.reinit(extended_solution_dim);
+
+    if (initial_step)
+    {
+      current_solution.reinit(dof_handler.n_dofs());
+    }
+
+  }
+
+
+  void TravelingWaveSolver::set_initial_guess()
+  {
+    current_wave_speed = initial_guess.wave_speed;
+
+    // The initial condition is a discrete set of coordinates $\xi$ and values of functions $u$, $T$ and $\lambda$. From the three sets we create three continuous functions using interpolation, which then form one continuous vector function of <code> SolutionVectorFunction </code> type.
+    Interpolant u_interpolant(initial_guess.x, initial_guess.u);
+    Interpolant T_interpolant(initial_guess.x, initial_guess.T);
+    Interpolant lambda_interpolant(initial_guess.x, initial_guess.lambda);
+
+    SolutionVectorFunction init_guess_func(u_interpolant, T_interpolant, lambda_interpolant);
+
+    VectorTools::interpolate(dof_handler, init_guess_func, current_solution); 
+
+    set_boundary_and_centering_values();
+
+    for (unsigned int i = 0; i < extended_solution_dim - 1; ++i)
+    {
+      current_solution_extended(i) = current_solution(i);
+    }
+    current_solution_extended(extended_solution_dim - 1) = current_wave_speed;
+  }
+
+  // Heaviside function.
+  double TravelingWaveSolver::Heaviside_func(double x) const
+  {
+    if (x > 0) 
+    {
+      return 1.;
+    }
+    else
+    {
+      return 0.;
+    }
+  }
+
+
+  void TravelingWaveSolver::compute_and_factorize_jacobian(const Vector<double> &evaluation_point_extended)
+  {
+    {
+      TimerOutput::Scope t(computing_timer, "assembling the Jacobian");
+
+      Vector<double> evaluation_point(dof_handler.n_dofs());
+      for (unsigned int i = 0; i < dof_handler.n_dofs(); ++i)
+      {
+        evaluation_point(i) = evaluation_point_extended(i);
+      }
+
+      const double wave_speed = evaluation_point_extended(extended_solution_dim - 1);
+
+      std::cout << "Computing Jacobian matrix ... " << std::endl;
+  
+      const QGauss<1> quadrature_formula(number_of_quadrature_points);
+  
+      jacobian_matrix_extended = 0;
+
+      FEValues<1> fe_values(fe,
+                            quadrature_formula,
+                            update_values | update_gradients | 
+                            update_quadrature_points | update_JxW_values);
+
+      const unsigned int dofs_per_cell = fe.n_dofs_per_cell();
+      const unsigned int n_q_points    = quadrature_formula.size();
+
+      FullMatrix<double> cell_matrix(dofs_per_cell, dofs_per_cell);
+      Vector<double> row_last_element_vector(dofs_per_cell);
+
+      std::vector<types::global_dof_index> local_dof_indices(dofs_per_cell);
+      
+      const FEValuesExtractors::Scalar velocity(0);
+      const FEValuesExtractors::Scalar temperature(1);
+      const FEValuesExtractors::Scalar lambda(2);
+
+      std::vector<double> current_velocity_values(n_q_points);
+      std::vector<double> current_temperature_values(n_q_points);
+      std::vector<double> current_lambda_values(n_q_points);
+
+      std::vector<Tensor<1, 1>> current_velocity_gradients(n_q_points);
+      std::vector<Tensor<1, 1>> current_temperature_gradients(n_q_points);
+      std::vector<Tensor<1, 1>> current_lambda_gradients(n_q_points);
+
+      std::vector<double> phi_u(dofs_per_cell);
+      std::vector<Tensor<1, 1>> grad_phi_u(dofs_per_cell);
+      std::vector<double> phi_T(dofs_per_cell);
+      std::vector<Tensor<1, 1>> grad_phi_T(dofs_per_cell);
+      std::vector<double> phi_lambda(dofs_per_cell);
+      std::vector<Tensor<1, 1>> grad_phi_lambda(dofs_per_cell);
+
+      for (const auto &cell : dof_handler.active_cell_iterators())
+      {
+        cell_matrix = 0;
+        row_last_element_vector = 0;
+
+        fe_values.reinit(cell);
+
+        fe_values[velocity].get_function_values(evaluation_point, current_velocity_values);
+        fe_values[temperature].get_function_values(evaluation_point, current_temperature_values);
+        fe_values[lambda].get_function_values(evaluation_point, current_lambda_values);
+
+        fe_values[velocity].get_function_gradients(evaluation_point, current_velocity_gradients);
+        fe_values[temperature].get_function_gradients(evaluation_point, current_temperature_gradients);
+        fe_values[lambda].get_function_gradients(evaluation_point, current_lambda_gradients);
+
+        auto kappa_1 = [=](double T, double lambda){
+          return problem.k * (1 - lambda) * std::exp(-problem.theta / T) * (
+            problem.theta / (T * T) * Heaviside_func(T - problem.T_ign) /* + Delta_function(T - problem.T_ign) */
+          );
+        };
+
+        auto kappa_2 = [=](double T, double lambda){
+          return -problem.k * std::exp(-problem.theta / T) * Heaviside_func(T - problem.T_ign);
+        };
+
+        for (unsigned int q = 0; q < n_q_points; ++q)
+        {
+          for (unsigned int k = 0; k < dofs_per_cell; ++k)
+          {
+            phi_u[k]            = fe_values[velocity].value(k, q);
+            grad_phi_u[k]       = fe_values[velocity].gradient(k, q);
+            phi_T[k]            = fe_values[temperature].value(k, q);
+            grad_phi_T[k]       = fe_values[temperature].gradient(k, q);
+            phi_lambda[k]       = fe_values[lambda].value(k, q);
+            grad_phi_lambda[k]  = fe_values[lambda].gradient(k, q);
+          }
+
+          const double del_Pr_eps = (problem.Pr * 4 * problem.delta / (3 * problem.epsilon));
+          const double del_Le = (problem.delta / problem.Le);
+
+          for (unsigned int i = 0; i < dofs_per_cell; ++i)
+          {
+            for (unsigned int j = 0; j < dofs_per_cell; ++j)
+            {
+              cell_matrix(i, j) += (
+
+                del_Pr_eps * (-grad_phi_u[i] * grad_phi_u[j])
+                + phi_u[i] * (
+                  - (1 - wave_speed + problem.epsilon * current_velocity_values[q]) * grad_phi_u[j][0]
+                  - problem.epsilon * current_velocity_gradients[q][0] * phi_u[j]
+                  - problem.epsilon / 2. * grad_phi_T[j][0]
+                )
+                
+                + problem.delta * (-grad_phi_T[i] * grad_phi_T[j])
+                + phi_T[i] * (
+                  - wave_speed * grad_phi_u[j][0]
+                  + wave_speed * grad_phi_T[j][0]
+                  + problem.q * kappa_1(current_temperature_values[q], current_lambda_values[q]) * phi_T[j]
+                  + problem.q * kappa_2(current_temperature_values[q], current_lambda_values[q]) * phi_lambda[j]
+                )
+
+                + del_Le * (-grad_phi_lambda[i] * grad_phi_lambda[j])
+                + phi_lambda[i] * (
+                  kappa_1(current_temperature_values[q], current_lambda_values[q]) * phi_T[j]
+                  + wave_speed * grad_phi_lambda[j][0]
+                  + kappa_2(current_temperature_values[q], current_lambda_values[q]) * phi_lambda[j]
+                )
+
+              ) * fe_values.JxW(q);
+
+            }
+
+            row_last_element_vector(i) += (
+              (phi_u[i] * current_velocity_gradients[q][0])
+              + (phi_T[i] * current_temperature_gradients[q][0])
+              - (phi_T[i] * current_velocity_gradients[q][0])
+              + (phi_lambda[i] * current_lambda_gradients[q][0])
+            ) * fe_values.JxW(q);
+          }
+
+        }
+  
+        cell->get_dof_indices(local_dof_indices);
+        
+        for (const unsigned int i : fe_values.dof_indices())
+        {
+          for (const unsigned int j : fe_values.dof_indices())
+          {
+            jacobian_matrix_extended.add(local_dof_indices[i],
+                                          local_dof_indices[j],
+                                          cell_matrix(i, j));
+          }
+          
+          // Adding elements to the last column.
+          jacobian_matrix_extended.add(local_dof_indices[i],
+                                        extended_solution_dim - 1,
+                                        row_last_element_vector(i));
+        }
+
+      }
+
+      // Global dof indices of dofs for $dT$ and $d\lambda$, associated with vertex $\xi = 0$.
+      types::global_dof_index T_zero_point_dof_ind(0), lambda_zero_point_dof_ind(0);   
+      
+      // Approximating the derivative of $T$ at $\xi = 0$ as done in step-14.
+      double T_point_derivative(0.);
+      double T_at_zero_point(0.);
+      double lambda_at_zero_point(0.);
+      {
+        double derivative_evaluation_point = 0.;       // Point at which T = T_ign.
+
+        const QTrapezoid<1> quadrature_formula;
+        FEValues<1>    fe_values(fe,
+                              quadrature_formula,
+                              update_values | update_gradients | update_quadrature_points);
+
+        const FEValuesExtractors::Scalar temperature(1);
+        const FEValuesExtractors::Scalar lambda(2);
+        
+        const unsigned int n_q_points = quadrature_formula.size();
+        std::vector<double> current_temperature_values(n_q_points);
+        std::vector<Tensor<1, 1>> current_temperature_gradients(n_q_points);
+        std::vector<double> current_lambda_values(n_q_points);
+        
+        unsigned int derivative_evaluation_point_hits = 0;
+
+        for (const auto &cell : dof_handler.active_cell_iterators())
+        {
+          for (const auto &vertex : cell->vertex_indices())
+          {
+            if (isapprox(cell->vertex(vertex)[0], derivative_evaluation_point))
+            {
+              T_zero_point_dof_ind = cell->vertex_dof_index(vertex, 1);
+              lambda_zero_point_dof_ind = cell->vertex_dof_index(vertex, 2);
+
+              fe_values.reinit(cell);
+              fe_values[temperature].get_function_values(current_solution, current_temperature_values);
+              fe_values[temperature].get_function_gradients(current_solution, current_temperature_gradients);
+              fe_values[lambda].get_function_values(current_solution, current_lambda_values);
+
+              unsigned int q_point = 0;
+              for (; q_point < n_q_points; ++q_point)
+              {
+                if (isapprox(fe_values.quadrature_point(q_point)[0], derivative_evaluation_point))
+                {
+                  break;
+                }
+              }
+
+              T_at_zero_point = current_temperature_values[q_point];
+              lambda_at_zero_point = current_lambda_values[q_point];
+
+              T_point_derivative += current_temperature_gradients[q_point][0];
+              ++derivative_evaluation_point_hits;
+
+              break;
+            }
+          }
+        }
+        T_point_derivative /= static_cast<double>(derivative_evaluation_point_hits);
+      }
+
+      // Here we add to the matrix the terms that appear after integrating the terms with the Dirac delta function (which we skipped inside the loop).
+      double term_with_delta_func(0.);
+      term_with_delta_func = problem.k * std::exp(-problem.theta / T_at_zero_point) * (1 - lambda_at_zero_point) / std::abs(T_point_derivative);
+      jacobian_matrix_extended.add(T_zero_point_dof_ind, T_zero_point_dof_ind, problem.q * term_with_delta_func);
+      jacobian_matrix_extended.add(lambda_zero_point_dof_ind, T_zero_point_dof_ind, term_with_delta_func);
+
+      // Add 1 to the position <code> T_zero_point_dof_ind </code> of the last row of the matrix.
+      jacobian_matrix_extended.add(extended_solution_dim - 1, T_zero_point_dof_ind, 1.);
+      
+      zero_boundary_constraints.condense(jacobian_matrix_extended);
+    }
+
+    {
+      TimerOutput::Scope t(computing_timer, "factorizing the Jacobian");
+
+      std::cout << "Factorizing Jacobian matrix" << std::endl;
+
+      jacobian_matrix_extended_factorization = std::make_unique<SparseDirectUMFPACK>();
+      jacobian_matrix_extended_factorization->factorize(jacobian_matrix_extended);
+    }
+
+  }
+
+
+  double TravelingWaveSolver::compute_residual(const Vector<double> &evaluation_point_extended, Vector<double> &residual)
+  {
+    TimerOutput::Scope t(computing_timer, "assembling the residual");
+  
+    std::cout << "Computing residual vector ... " << std::endl; 
+    
+    residual = 0;
+
+    Vector<double> evaluation_point(dof_handler.n_dofs());
+    for (unsigned int i = 0; i < dof_handler.n_dofs(); ++i)
+    {
+      evaluation_point(i) = evaluation_point_extended(i);
+    }
+
+    const double wave_speed = evaluation_point_extended(extended_solution_dim - 1);
+
+    const QGauss<1> quadrature_formula(number_of_quadrature_points);
+    FEValues<1> fe_values(fe,
+                quadrature_formula,
+                update_values | update_gradients | 
+                update_quadrature_points | update_JxW_values);
+    const unsigned int dofs_per_cell = fe.n_dofs_per_cell();
+    const unsigned int n_q_points    = quadrature_formula.size();
+   
+    Vector<double> cell_residual(dofs_per_cell); 
+    std::vector<types::global_dof_index> local_dof_indices(dofs_per_cell);
+
+    const FEValuesExtractors::Scalar velocity(0);
+    const FEValuesExtractors::Scalar temperature(1);
+    const FEValuesExtractors::Scalar lambda(2);
+
+    std::vector<double> current_velocity_values(n_q_points);
+    std::vector<Tensor<1, 1>> current_velocity_gradients(n_q_points);
+    std::vector<double> current_temperature_values(n_q_points);
+    std::vector<Tensor<1, 1>> current_temperature_gradients(n_q_points);
+    std::vector<double> current_lambda_values(n_q_points);
+    std::vector<Tensor<1, 1>> current_lambda_gradients(n_q_points);
+
+    std::vector<double> phi_u(dofs_per_cell);
+    std::vector<Tensor<1, 1>> grad_phi_u(dofs_per_cell);
+    std::vector<double> phi_T(dofs_per_cell);
+    std::vector<Tensor<1, 1>> grad_phi_T(dofs_per_cell);
+    std::vector<double> phi_lambda(dofs_per_cell);
+    std::vector<Tensor<1, 1>> grad_phi_lambda(dofs_per_cell);
+
+    for (const auto &cell : dof_handler.active_cell_iterators())
+    {
+      cell_residual = 0;
+
+      fe_values.reinit(cell);
+
+      fe_values[velocity].get_function_values(evaluation_point, current_velocity_values);
+      fe_values[velocity].get_function_gradients(evaluation_point, current_velocity_gradients);
+      fe_values[temperature].get_function_values(evaluation_point, current_temperature_values);
+      fe_values[temperature].get_function_gradients(evaluation_point, current_temperature_gradients);
+      fe_values[lambda].get_function_values(evaluation_point, current_lambda_values);
+      fe_values[lambda].get_function_gradients(evaluation_point, current_lambda_gradients);
+
+      auto omega = [=](double T, double lambda){
+        return problem.k * (1 - lambda) * std::exp(-problem.theta / T) * Heaviside_func(T - problem.T_ign);
+      };
+
+      for (unsigned int q = 0; q < n_q_points; ++q)
+      {
+        for (unsigned int k = 0; k < dofs_per_cell; ++k)
+        {
+          phi_u[k]            = fe_values[velocity].value(k, q);
+          grad_phi_u[k]       = fe_values[velocity].gradient(k, q);
+          phi_T[k]            = fe_values[temperature].value(k, q);
+          grad_phi_T[k]       = fe_values[temperature].gradient(k, q);
+          phi_lambda[k]       = fe_values[lambda].value(k, q);
+          grad_phi_lambda[k]  = fe_values[lambda].gradient(k, q);
+        }
+
+        double del_Pr_eps = (problem.Pr * 4 * problem.delta / (3 * problem.epsilon));
+        double del_Le = (problem.delta / problem.Le);
+
+        for (unsigned int i = 0; i < dofs_per_cell; ++i)
+        {
+          cell_residual(i) += (
+
+            del_Pr_eps * (-grad_phi_u[i] * current_velocity_gradients[q])
+            + phi_u[i] * (
+              - current_velocity_gradients[q][0] * (1 - wave_speed + problem.epsilon * current_velocity_values[q]) 
+              - problem.epsilon / 2. * current_temperature_gradients[q][0]
+            )
+
+            + problem.delta * (-grad_phi_T[i] * current_temperature_gradients[q])
+            + phi_T[i] * (
+              wave_speed * (current_temperature_gradients[q][0] - current_velocity_gradients[q][0])
+              + problem.q * omega(current_temperature_values[q], current_lambda_values[q])
+            )
+
+            + del_Le * (-grad_phi_lambda[i] * current_lambda_gradients[q])
+            + phi_lambda[i] * (
+              wave_speed * current_lambda_gradients[q][0] + omega(current_temperature_values[q], current_lambda_values[q])
+            )
+
+          ) * fe_values.JxW(q);
+        }
+
+      }                                                                                        
+
+      cell->get_dof_indices(local_dof_indices);
+
+      for (const unsigned int i : fe_values.dof_indices())
+      {                                        
+        residual(local_dof_indices[i]) += cell_residual(i);
+      }
+    }
+
+    residual(extended_solution_dim - 1) = 0.;
+
+    zero_boundary_constraints.condense(residual);
+
+    double residual_norm = residual.l2_norm();
+    
+    std::cout << std::defaultfloat;
+    std::cout << "norm of residual = " << residual_norm << std::endl;
+
+    return residual_norm;
+  }
+
+  // Split the solution vector into two parts: one part is the solution $u$, $T$ and $\lambda$, and another part is the wave speed.
+  void TravelingWaveSolver::split_extended_solution_vector()
+  {
+    for (unsigned int i = 0; i < extended_solution_dim - 1; ++i)
+    {
+      current_solution(i) = current_solution_extended(i);
+    }
+
+    current_wave_speed = current_solution_extended(extended_solution_dim - 1);
+  }
+
+
+  void TravelingWaveSolver::solve(const Vector<double> &rhs, Vector<double> &solution_extended, const double /*tolerance*/)
+  {
+    TimerOutput::Scope t(computing_timer, "linear system solve");
+    std::cout << "Solving linear system ... " << std::endl;
+
+    jacobian_matrix_extended_factorization->vmult(solution_extended, rhs);
+    zero_boundary_constraints.distribute(solution_extended);
+
+  }
+
+
+  // Function for adaptive mesh refinement based on <code> KellyErrorEstimator </code>.
+  void TravelingWaveSolver::refine_mesh()
+  {
+    Vector<float> estimated_error_per_cell(triangulation.n_active_cells());
+
+    const FEValuesExtractors::Scalar lambda(2);
+
+    KellyErrorEstimator<1>::estimate(
+      dof_handler,
+      QGauss<0>( 0 /* number_of_quadrature_points */),
+      {},
+      current_solution,
+      estimated_error_per_cell,
+      fe.component_mask(lambda)
+    );
+
+    GridRefinement::refine_and_coarsen_fixed_number(triangulation,
+                            estimated_error_per_cell,
+                            0.1,
+                            0.05);
+
+    triangulation.prepare_coarsening_and_refinement();
+
+    SolutionTransfer<1> solution_transfer(dof_handler);
+    solution_transfer.prepare_for_coarsening_and_refinement(current_solution);
+
+    triangulation.execute_coarsening_and_refinement();
+
+    setup_system(/*initial_step=*/ false);
+
+    Vector<double> tmp(dof_handler.n_dofs());
+    solution_transfer.interpolate(current_solution, tmp);
+    current_solution = std::move(tmp);
+
+    set_boundary_and_centering_values();
+
+    for (unsigned int i = 0; i < extended_solution_dim - 1; ++i)
+    {
+      current_solution_extended(i) = current_solution(i);
+    }
+    current_solution_extended(extended_solution_dim - 1) = current_wave_speed;
+
+  }
+
+
+  double TravelingWaveSolver::run_newton_iterations(const double target_tolerance)
+  {
+    
+    double residual_norm = 0.;
+    {
+      typename SUNDIALS::KINSOL< Vector<double> >::AdditionalData additional_data;
+      additional_data.function_tolerance = target_tolerance;
+  
+      SUNDIALS::KINSOL<Vector<double>> nonlinear_solver(additional_data);
+  
+      nonlinear_solver.reinit_vector = [&](Vector<double> &x) {
+        x.reinit(extended_solution_dim);
+      };
+
+      nonlinear_solver.residual = [&](const Vector<double> &evaluation_point, Vector<double> &residual) {
+        residual_norm = compute_residual(evaluation_point, residual);
+
+        return 0;
+      };
+  
+      nonlinear_solver.setup_jacobian = [&](const Vector<double> &evaluation_point, const Vector<double> & /*current_f*/) {
+        compute_and_factorize_jacobian(evaluation_point);
+
+        return 0;
+      };
+
+      nonlinear_solver.solve_with_jacobian = [&](const Vector<double> &rhs, Vector<double> &solution, const double tolerance) {
+        this->solve(rhs, solution, tolerance);
+
+        return 0;
+      };
+
+      nonlinear_solver.solve(current_solution_extended);
+    }
+
+    return residual_norm;
+
+  }
+
+  // Output the solution ($u$, $T$ and $\lambda$) and the wave speed into two separate files with double precision. The files can be read by gnuplot.
+  void TravelingWaveSolver::output_with_double_precision(const Vector<double> &solution, const double wave_speed, const std::string filename)
+  {
+    TimerOutput::Scope t(computing_timer, "graphical output txt");
+
+    const std::string file_for_solution = filename + ".txt";
+    std::ofstream output(file_for_solution);
+
+    for (const auto &cell : dof_handler.active_cell_iterators())
+    {
+      for (const auto &v_ind : cell->vertex_indices())
+      {
+        double u = solution(cell->vertex_dof_index(v_ind, 0));
+        double T = solution(cell->vertex_dof_index(v_ind, 1));
+        double lambda = solution(cell->vertex_dof_index(v_ind, 2));
+
+        output << std::scientific << std::setprecision(16);
+        output << cell->vertex(v_ind)[0];
+
+        output << std::scientific << std::setprecision(16);
+        output << std::scientific << " " << u << " " << T << " " << lambda << "\n";
+      }
+      output  << "\n";
+    }
+
+    output.close();
+
+    std::ofstream file_for_wave_speed_output("wave_speed-" + file_for_solution);
+    file_for_wave_speed_output << std::scientific << std::setprecision(16);
+    file_for_wave_speed_output << wave_speed << std::endl;
+    file_for_wave_speed_output.close();
+  }
+
+  // Copy the solution into the <code> SolutionStruct </code> object, that stores the solution in an ordered manner. 
+  void TravelingWaveSolver::get_solution(SolutionStruct &solution) const
+  {
+    // To obtain an ordered solution array, we first create a set consisting of the elements <code> {x, u, T, lambda} </code> in which the sorting is done by coordinate, and then copy the contents of the set into the arrays of the <code> SolutionStruct </code> object.
+    auto comp = [](const std::vector<double> &a, const std::vector<double> &b) {
+      return a[0] < b[0];
+    };
+    std::set<std::vector<double>, decltype(comp)> solution_set(comp);
+    for (const auto &cell : dof_handler.active_cell_iterators())
+    {
+      for (const auto &v_ind : cell->vertex_indices())
+      {
+        double x = cell->vertex(v_ind)[0];
+        double u = current_solution(cell->vertex_dof_index(v_ind, 0));
+        double T = current_solution(cell->vertex_dof_index(v_ind, 1));
+        double lambda = current_solution(cell->vertex_dof_index(v_ind, 2));
+        solution_set.insert({x, u, T, lambda});
+      }
+    }
+
+    solution.x.clear();
+    solution.u.clear();
+    solution.T.clear();
+    solution.lambda.clear();
+
+    solution.x.reserve(solution_set.size());
+    solution.u.reserve(solution_set.size());
+    solution.T.reserve(solution_set.size());
+    solution.lambda.reserve(solution_set.size());
+
+    for (auto it = solution_set.begin(); it != solution_set.end(); ++it)
+    {
+      solution.x.push_back((*it)[0]);
+      solution.u.push_back((*it)[1]);
+      solution.T.push_back((*it)[2]);
+      solution.lambda.push_back((*it)[3]);
+    }
+
+    solution.wave_speed = current_wave_speed;
+
+  }
+
+
+  void TravelingWaveSolver::get_triangulation(Triangulation<1> &otriangulation) const
+  {
+    otriangulation.clear();
+    otriangulation.copy_triangulation(triangulation);
+  }
+
+
+  void TravelingWaveSolver::run(const std::string filename, const bool save_solution_to_file)
+  {
+    const int mesh_refinement_type    = params.mesh.adaptive;
+    const unsigned int n_refinements  = params.mesh.refinements_number;
+    const double tol                  = params.solver.tol;
+
+    if (triangulation_uploaded == false)  // If the triangulation is not loaded from outside, we will create one.
+    {
+      // We create two triangulations: one to the left and one to the right of zero coordinate. After that we merge them to obtain one triangulation, which contains zero point.
+      Triangulation<1> triangulation_left;
+      GridGenerator::subdivided_hyper_cube(
+        triangulation_left,
+        static_cast<unsigned int>(std::abs( 0. - params.mesh.interval_left )),
+        params.mesh.interval_left, 0.
+      );
+
+      Triangulation<1> triangulation_right;
+      GridGenerator::subdivided_hyper_cube(
+        triangulation_right,
+        static_cast<unsigned int>(std::abs( params.mesh.interval_right - 0. )),
+        0., params.mesh.interval_right
+      );
+
+      GridGenerator::merge_triangulations(triangulation_left, triangulation_right, triangulation);
+
+    }
+
+    if (triangulation_uploaded == false)
+    {
+      if (mesh_refinement_type == 1)        // For ADAPTIVE mesh refinement.
+      {
+        triangulation.refine_global(1);     // refine initial mesh globally, before adaptive refinement cycles.
+      }
+      else if (mesh_refinement_type == 0)   // For GLOBAL mesh refinement.
+      {
+        triangulation.refine_global(n_refinements);
+      }
+    }
+
+    setup_system(/*initial step*/ true);
+    set_initial_guess();
+
+    if (save_solution_to_file)
+    {
+      output_with_double_precision(current_solution, current_wave_speed, "solution_initial_data");
+    }
+
+    if (mesh_refinement_type == 1)    // Compute with ADAPTIVE mesh refinement.
+    {
+      double residual_norm = 0.;
+      {
+        Vector<double> tmp_residual(extended_solution_dim);
+        residual_norm = compute_residual(current_solution_extended, tmp_residual);
+      }
+
+      unsigned int refinement_cycle = 0;
+      while ((residual_norm > tol) && (refinement_cycle < n_refinements))
+      {
+        computing_timer.reset();
+        std::cout << "Mesh refinement step " << refinement_cycle << std::endl;
+
+        if (refinement_cycle != 0) { refine_mesh(); }
+        
+        const double target_tolerance = 0.1 * std::pow(0.1, refinement_cycle);    // Decrease tolerance for Newton solver at each refinement step.
+        std::cout << "  Target_tolerance: " << target_tolerance << std::endl;
+  
+        residual_norm = run_newton_iterations(target_tolerance);
+        split_extended_solution_vector();
+
+        {
+          std::cout << std::scientific << std::setprecision(16);
+          std::cout << "current_wave_speed = " << current_wave_speed << std::endl;
+          std::cout << std::defaultfloat;
+        }
+
+        computing_timer.print_summary();
+
+        ++refinement_cycle;
+      }
+      if (save_solution_to_file)
+      {
+        output_with_double_precision(current_solution, current_wave_speed, filename);
+      }
+
+    }
+    else if (mesh_refinement_type == 0)   // Compute with GLOBAL mesh refinement.
+    {
+      run_newton_iterations(tol);
+      split_extended_solution_vector();
+      
+      if (save_solution_to_file)
+      {
+        output_with_double_precision(current_solution, current_wave_speed, filename);
+      }
+      
+      {
+        std::cout << std::scientific << std::setprecision(16);
+        std::cout << "current_wave_speed = " << current_wave_speed << std::endl;
+        std::cout << std::defaultfloat;
+      }
+
+      computing_timer.print_summary();
+
+    }
+
+  }
+
+
+} // namespace TravelingWave
diff --git a/TravelingWaves/TravelingWaveSolver.h b/TravelingWaves/TravelingWaveSolver.h
new file mode 100644 (file)
index 0000000..b031567
--- /dev/null
@@ -0,0 +1,115 @@
+#ifndef WAVE_CONSTRUCTOR
+#define WAVE_CONSTRUCTOR
+
+#include <deal.II/base/timer.h>
+#include <deal.II/base/function.h>
+#include <deal.II/base/table_handler.h>
+#include <deal.II/grid/tria.h>
+#include <deal.II/grid/grid_generator.h>
+#include <deal.II/grid/grid_refinement.h>
+#include <deal.II/grid/grid_out.h>
+#include <deal.II/dofs/dof_handler.h>
+#include <deal.II/dofs/dof_tools.h>
+#include <deal.II/fe/fe_q.h>
+#include <deal.II/fe/fe_system.h>
+#include <deal.II/fe/fe_values.h>
+#include <deal.II/lac/vector.h>
+#include <deal.II/lac/full_matrix.h>
+#include <deal.II/lac/sparse_matrix.h>
+#include <deal.II/lac/sparse_direct.h>
+#include <deal.II/lac/dynamic_sparsity_pattern.h>
+#include <deal.II/numerics/vector_tools.h>
+#include <deal.II/numerics/error_estimator.h>
+#include <deal.II/numerics/solution_transfer.h>
+#include <deal.II/numerics/data_out.h>
+#include <deal.II/sundials/kinsol.h>
+
+#include "Parameters.h"
+#include "Solution.h"
+#include "AuxiliaryFunctions.h"
+
+#include <cmath>
+#include <algorithm>
+#include <string>
+#include <fstream>
+#include <iostream>
+#include <iomanip>
+#include <set>
+
+// Namespace of the program
+namespace TravelingWave
+{
+  using namespace dealii;
+
+  // The main class for construction of the traveling wave solutions.
+  class TravelingWaveSolver
+  {
+  public:
+    TravelingWaveSolver(const Parameters &parameters, const SolutionStruct &initial_guess_input);
+    
+    void set_triangulation(const Triangulation<1> &itriangulation);
+    
+    void run(const std::string filename="solution", const bool save_solution_to_file=true);
+    void get_solution(SolutionStruct &solution) const;
+    void get_triangulation(Triangulation<1> &otriangulation) const;
+
+  private:
+    void setup_system(const bool initial_step);
+    void find_boundary_and_centering_dof_numbers();
+    void set_boundary_and_centering_values();
+
+    void set_initial_guess();
+
+    double Heaviside_func(double x) const;
+    
+    void compute_and_factorize_jacobian(const Vector<double> &evaluation_point_extended);
+    double compute_residual(const Vector<double> &evaluation_point_extended, Vector<double> &residual);
+    void split_extended_solution_vector();
+
+    void solve(const Vector<double> &rhs, Vector<double> &solution, const double /*tolerance*/);
+    void refine_mesh();
+    double run_newton_iterations(const double target_tolerance=1e-5);
+
+    void output_with_double_precision(const Vector<double> &solution, const double wave_speed, const std::string filename="solution");
+
+    // The dimension of the finite element solution increased by one to account for the value corresponding to the wave speed.
+    unsigned int extended_solution_dim;
+    std::map<std::string, unsigned int> boundary_and_centering_dof_numbers;
+
+    // Parameters of the problem, taken from a .prm file.
+    const Parameters &params;
+    const Problem               &problem; // Reference variable, just for convenience.
+
+    unsigned int number_of_quadrature_points;
+
+    Triangulation<1> triangulation;
+    // The flag indicating whether the triangulation was uploaded externally or created within the <code> run </code> member function.
+    bool                                                triangulation_uploaded;
+    FESystem<1>      fe;
+    DoFHandler<1>    dof_handler;
+
+    // Constraints for Dirichlet boundary conditions.
+    AffineConstraints<double> zero_boundary_constraints;       
+
+    SparsityPattern                                                                                    sparsity_pattern_extended;
+    SparseMatrix<double>                                                               jacobian_matrix_extended;
+    std::unique_ptr<SparseDirectUMFPACK>       jacobian_matrix_extended_factorization;
+
+    // Finite element solution of the problem.
+    Vector<double>     current_solution;
+
+    // Value of the wave speed $c$.
+    double                                     current_wave_speed;
+
+    // Solution with an additional term, corresponding to the variable wave_speed.
+    Vector<double>     current_solution_extended;
+
+    // Initial guess for Newton's iterations.
+    SolutionStruct initial_guess;
+
+    TimerOutput computing_timer;
+  };
+
+} // namespace TravelingWave
+
+#endif
\ No newline at end of file
diff --git a/TravelingWaves/calculate_profile.cc b/TravelingWaves/calculate_profile.cc
new file mode 100644 (file)
index 0000000..a6a9e2e
--- /dev/null
@@ -0,0 +1,276 @@
+#include "TravelingWaveSolver.h"
+#include "calculate_profile.h"
+
+namespace TravelingWave
+{
+  using namespace dealii;
+
+  // Computation of the limit case (ideal) solution, corresponding to $\delta = 0$, by solving the ODE. The output is the part of the solution to the left of zero. Here u_0, T_0, lambda_0 are the values of the medium state to the right of zero.
+  void compute_limit_sol_left_part(const Parameters &parameters, 
+                                    const double wave_speed, 
+                                    const double u_0, 
+                                    const double T_0, 
+                                    const double lambda_0, 
+                                    SolutionStruct &LimitSol, 
+                                    const double root_sign)
+  {
+    LimitSolution limit_sol(parameters, lambda_0, u_0, T_0, root_sign);
+    limit_sol.set_wave_speed(wave_speed);
+    
+    {
+      // We take more integration points to better resolve the transition layer.
+      std::vector<double> t_span(static_cast<unsigned int>(std::abs( 0. - parameters.mesh.interval_left )));
+      double finer_mesh_starting_value = -9.1;
+      linspace(parameters.mesh.interval_left, finer_mesh_starting_value, t_span); 
+      std::vector<double> fine_grid(10000);
+      linspace(finer_mesh_starting_value + 1e-4, 0., fine_grid); 
+      t_span.insert(t_span.end(), fine_grid.begin(), fine_grid.end());
+
+      // Reverse the order of the elements (because we need to perform back in time integration).
+      std::reverse(t_span.begin(), t_span.end());
+
+      state_type lambda_val(1);
+      lambda_val[0] = lambda_0;                // initial value
+      IntegrateSystemAtTimePoints(limit_sol.lambda_vec, limit_sol.t_vec, t_span,
+        limit_sol, 
+        lambda_val,
+        -1e-2, Integrator_Type::dopri5);
+    }
+
+    limit_sol.calculate_u_T_omega();
+
+    // Reverse the order of elements
+    std::reverse(limit_sol.t_vec.begin(), limit_sol.t_vec.end());
+    std::reverse(limit_sol.lambda_vec.begin(), limit_sol.lambda_vec.end());
+    std::reverse(limit_sol.u_vec.begin(), limit_sol.u_vec.end());
+    std::reverse(limit_sol.T_vec.begin(), limit_sol.T_vec.end());
+    std::reverse(limit_sol.omega_vec.begin(), limit_sol.omega_vec.end());
+
+    SaveSolutionIntoFile(limit_sol.lambda_vec, limit_sol.t_vec, "solution_lambda_limit.txt");
+    SaveSolutionIntoFile(limit_sol.u_vec, limit_sol.t_vec, "solution_u_limit.txt");
+    SaveSolutionIntoFile(limit_sol.T_vec, limit_sol.t_vec, "solution_T_limit.txt");
+    SaveSolutionIntoFile(limit_sol.omega_vec, limit_sol.t_vec, "solution_omega_limit.txt");
+
+    LimitSol.reinit(limit_sol.t_vec.size());
+    LimitSol.wave_speed = wave_speed;
+    for (unsigned int i=0; i < limit_sol.t_vec.size(); ++i)
+    {
+      LimitSol.x[i] = limit_sol.t_vec[i];
+      LimitSol.u[i] = limit_sol.u_vec[i][0];
+      LimitSol.T[i] = limit_sol.T_vec[i][0];
+      LimitSol.lambda[i] = limit_sol.lambda_vec[i][0];
+    }
+  }
+
+
+  // Construction of an initial guess for detonation wave solution. The ODE is solved for the ideal system with $\delta = 0$.
+  void compute_initial_guess_detonation(const Parameters &params, SolutionStruct &initial_guess, const double root_sign)
+  {
+    const Problem &problem = params.problem;
+    double current_wave_speed(problem.wave_speed_init);
+
+    {  // Here we compute the exact value of the wave speed $c$ for the detonation case. We can do this because we have the Dirichlet boundary conditions $T_l$, $T_r$ and $u_r$. Exact values of $u_l$ and $c$ are obtained using the integral relations.
+      double DeltaT = problem.T_left - problem.T_right;
+      double qDT = problem.q - DeltaT;
+      current_wave_speed = 1. + problem.epsilon * (problem.u_right - (qDT * qDT + DeltaT) / (2 * qDT));
+    }
+
+    double u_0 = problem.u_right;
+    double T_0 = problem.T_right;
+    double lambda_0 = 0.;
+
+    compute_limit_sol_left_part(params, current_wave_speed, u_0, T_0, lambda_0, initial_guess, root_sign);
+
+    initial_guess.wave_speed = current_wave_speed;
+
+    for (int i = initial_guess.x.size() - 1; i > - 1; --i)
+    {
+      if (isapprox(initial_guess.x[i], 0.))
+      {
+        initial_guess.u[i] = problem.u_right;
+        initial_guess.T[i] = problem.T_ign;
+        initial_guess.lambda[i] = 0.;
+        break;
+      }
+    }
+
+    // Adding the points to the right part of the interval (w.r.t. $\xi = 0$).
+    unsigned int number_of_additional_points = 5;
+    for (unsigned int i = 0; i < number_of_additional_points; ++i)
+    {
+      initial_guess.x.push_back(params.mesh.interval_right / (std::pow(2., number_of_additional_points - 1 - i)));
+      initial_guess.u.push_back(problem.u_right);
+      initial_guess.T.push_back(problem.T_right);
+      initial_guess.lambda.push_back(0.);
+    }
+
+  }
+
+
+  // Construction of a piecewise constant initial guess for deflagration wave solution.
+  void compute_initial_guess_deflagration(const Parameters &params, SolutionStruct &initial_guess)
+  {
+    const Problem &problem = params.problem;
+    double current_wave_speed(problem.wave_speed_init);
+
+    double del_Pr_eps = (problem.Pr * 4 * problem.delta / (3 * problem.epsilon));
+    double del_Le = (problem.delta / problem.Le);
+
+    auto u_init_guess_func = [&](double x) {
+      if (x < 0.)
+      {
+        return problem.u_left;
+      }
+      else
+      {
+        return problem.u_right;
+      }
+    };
+
+    auto T_init_guess_func = [&](double x) {
+      if (x < 0.)
+      {
+        return problem.T_left;
+      }
+      else if (isapprox(x, 0.))
+      {
+        return problem.T_ign;
+      }
+      else
+      {
+        return problem.T_right;
+      }
+    };
+
+    auto lambda_init_guess_func = [=](double x) {
+      if (x < 0.)
+      {
+        return -std::exp(x * std::abs(1 - current_wave_speed) / del_Pr_eps) + 1;
+      }
+      else 
+      {
+        return 0.;
+      }
+    };
+
+    unsigned int multiplier_for_number_of_points = 7;
+    unsigned int number_of_points = multiplier_for_number_of_points * static_cast<unsigned int>(std::trunc(std::abs( params.mesh.interval_right - params.mesh.interval_left )));
+    std::vector<double> x_span(number_of_points);
+    linspace(params.mesh.interval_left, params.mesh.interval_right, x_span);
+
+    std::vector<double> u_init_arr(number_of_points);
+    std::vector<double> T_init_arr(number_of_points);
+    std::vector<double> lambda_init_arr(number_of_points);
+
+    for (unsigned int i = 0; i < number_of_points; ++i)
+    {
+      u_init_arr[i] = u_init_guess_func(x_span[i]);
+      T_init_arr[i] = T_init_guess_func(x_span[i]);
+      lambda_init_arr[i] = lambda_init_guess_func(x_span[i]);
+    }
+
+    initial_guess.x = x_span;
+    initial_guess.u = u_init_arr;
+    initial_guess.T = T_init_arr;
+    initial_guess.lambda = lambda_init_arr;
+    initial_guess.wave_speed = current_wave_speed;
+
+  }
+
+
+  // Compute the traveling-wave profile. The continuation method can be switched on by setting the argument <code> continuation_for_delta </code> as <code> true </code>.
+  void calculate_profile(Parameters& parameters,
+                                    const bool continuation_for_delta /* Compute with the continuation. */, 
+                                    const double delta_start /* The starting value of delta for the continuation method. */, 
+                                    const unsigned int number_of_continuation_points)
+  {
+    SolutionStruct sol;
+
+    if (parameters.problem.wave_type == 1)                             // detonation wave
+    {
+      compute_initial_guess_detonation(parameters, sol);
+    }
+    else if (parameters.problem.wave_type == 0)                // deflagration wave
+    {
+      compute_initial_guess_deflagration(parameters, sol);
+    }
+    
+    if (continuation_for_delta == false)
+    {
+      TravelingWaveSolver wave(parameters, sol);
+      std::string filename = "solution_delta-" + Utilities::to_string(parameters.problem.delta) + "_eps-" 
+                                                        + Utilities::to_string(parameters.problem.epsilon);
+      wave.run(filename);
+      wave.get_solution(sol);
+    }
+    else       // Run with continuation_for_delta.
+    {
+      double delta_target = parameters.problem.delta;
+      parameters.problem.delta = delta_start;
+
+      std::vector<double> delta_span(number_of_continuation_points);
+
+      // Generate a sequence of delta values being uniformly distributed in log10 scale.
+      {
+        double delta_start_log10 = std::log10(delta_start);
+        double delta_target_log10 = std::log10(delta_target);
+
+        std::vector<double> delta_log_span(delta_span.size());
+        linspace(delta_start_log10, delta_target_log10, delta_log_span);
+
+        for (unsigned int i = 0; i < delta_span.size(); ++i)
+        {
+          delta_span[i] = std::pow(10, delta_log_span[i]);
+        }
+      }
+
+      Triangulation<1> refined_triangulation;
+      bool first_iter_flag = true;
+
+      for (double delta : delta_span)
+      {
+        parameters.problem.delta = delta;
+        std::string filename = "solution_delta-" + Utilities::to_string(parameters.problem.delta) + "_eps-" 
+                                                    + Utilities::to_string(parameters.problem.epsilon);
+
+        TravelingWaveSolver wave(parameters, sol);
+
+        if (first_iter_flag)
+        {
+          first_iter_flag = false;
+        }
+        else
+        {
+          wave.set_triangulation(refined_triangulation);       
+        }
+
+        wave.run(filename);
+        wave.get_solution(sol);
+        wave.get_triangulation(refined_triangulation);
+      }
+
+    }
+
+    // Error estimation.
+    {
+      unsigned int sol_length = sol.x.size();
+      double u_r = sol.u[sol_length-1];                        // Dirichlet boundary condition
+      double T_r = sol.T[sol_length-1];                        // Dirichlet condition only for detonation case
+      double u_l = sol.u[0];
+      double T_l = sol.T[0];                                                           // Dirichlet boundary condition
+      double wave_speed = sol.wave_speed;
+
+      std::cout << "Error estimates:" << std::endl;
+      double DeltaT = T_l - T_r;
+      double qDT = parameters.problem.q - DeltaT;
+
+      double wave_speed_formula = 1. + parameters.problem.epsilon * (u_r - (qDT * qDT + DeltaT) / (2 * qDT));
+      std::cout << std::setw(18) << std::left << "For wave speed" << " :  " << std::setw(5) << wave_speed - wave_speed_formula << std::endl;
+
+      double u_l_formula = DeltaT + u_r - parameters.problem.q;
+      std::cout << std::setw(18) << std::left << "For u_l" << " :  " << std::setw(5) << u_l - u_l_formula << std::endl;
+    }
+
+  }
+
+} // namespace TravelingWave
\ No newline at end of file
diff --git a/TravelingWaves/calculate_profile.h b/TravelingWaves/calculate_profile.h
new file mode 100644 (file)
index 0000000..ff5761d
--- /dev/null
@@ -0,0 +1,30 @@
+#ifndef INITIAL_GUESS
+#define INITIAL_GUESS
+
+#include "Parameters.h"
+#include "Solution.h"
+#include "LimitSolution.h"
+#include "IntegrateSystem.h"
+#include "AuxiliaryFunctions.h"
+
+namespace TravelingWave
+{
+  void compute_limit_sol_left_part(const Parameters &parameters, 
+                                    const double wave_speed, 
+                                    const double u_0, 
+                                    const double T_0, 
+                                    const double lambda_0, 
+                                    SolutionStruct &LimitSol, 
+                                    const double root_sign = 1.);
+
+       void compute_initial_guess_detonation(const Parameters &params, SolutionStruct &initial_guess, const double root_sign = 1.);
+       void compute_initial_guess_deflagration(const Parameters &params, SolutionStruct &initial_guess);
+
+  void calculate_profile(Parameters& parameters,
+                                                                                                                                       const bool continuation_for_delta=false /* Compute with the continuation. */, 
+                                                                                                                                       const double delta_start=0.01 /* The starting value of delta for the continuation method. */, 
+                                                                                                                                       const unsigned int number_of_continuation_points=10);
+
+} // namespace TravelingWave
+
+#endif
\ No newline at end of file
diff --git a/TravelingWaves/doc/author b/TravelingWaves/doc/author
new file mode 100644 (file)
index 0000000..bf84f88
--- /dev/null
@@ -0,0 +1 @@
+Shamil Magomedov <magomedov.shamil.m@gmail.com>
diff --git a/TravelingWaves/doc/builds-on b/TravelingWaves/doc/builds-on
new file mode 100644 (file)
index 0000000..b5ab7fb
--- /dev/null
@@ -0,0 +1 @@
+step-15 step-77
diff --git a/TravelingWaves/doc/dependencies b/TravelingWaves/doc/dependencies
new file mode 100644 (file)
index 0000000..076e977
--- /dev/null
@@ -0,0 +1,2 @@
+DEAL_II_WITH_SUNDIALS
+DEAL_II_WITH_UMFPACK
\ No newline at end of file
diff --git a/TravelingWaves/doc/entry-name b/TravelingWaves/doc/entry-name
new file mode 100644 (file)
index 0000000..66779e2
--- /dev/null
@@ -0,0 +1 @@
+Traveling-wave solutions of a qualitative model for combustion waves.
diff --git a/TravelingWaves/doc/pics/detonation_delta_0.001.png b/TravelingWaves/doc/pics/detonation_delta_0.001.png
new file mode 100644 (file)
index 0000000..9b1bff4
Binary files /dev/null and b/TravelingWaves/doc/pics/detonation_delta_0.001.png differ
diff --git a/TravelingWaves/doc/pics/detonation_delta_0.01.png b/TravelingWaves/doc/pics/detonation_delta_0.01.png
new file mode 100644 (file)
index 0000000..794a723
Binary files /dev/null and b/TravelingWaves/doc/pics/detonation_delta_0.01.png differ
diff --git a/TravelingWaves/doc/pics/fast_deflagration_delta_0.01.png b/TravelingWaves/doc/pics/fast_deflagration_delta_0.01.png
new file mode 100644 (file)
index 0000000..3f133e2
Binary files /dev/null and b/TravelingWaves/doc/pics/fast_deflagration_delta_0.01.png differ
diff --git a/TravelingWaves/doc/pics/slow_deflagration_delta_0.01.png b/TravelingWaves/doc/pics/slow_deflagration_delta_0.01.png
new file mode 100644 (file)
index 0000000..0f590a3
Binary files /dev/null and b/TravelingWaves/doc/pics/slow_deflagration_delta_0.01.png differ
diff --git a/TravelingWaves/doc/pics/sparsity-pattern_p1.png b/TravelingWaves/doc/pics/sparsity-pattern_p1.png
new file mode 100644 (file)
index 0000000..900561a
Binary files /dev/null and b/TravelingWaves/doc/pics/sparsity-pattern_p1.png differ
diff --git a/TravelingWaves/doc/tooltip b/TravelingWaves/doc/tooltip
new file mode 100644 (file)
index 0000000..68513cd
--- /dev/null
@@ -0,0 +1 @@
+Traveling waves
diff --git a/TravelingWaves/main.cc b/TravelingWaves/main.cc
new file mode 100644 (file)
index 0000000..59431a7
--- /dev/null
@@ -0,0 +1,77 @@
+#include "calculate_profile.h"
+
+int main(int argc, char *argv[])
+{
+
+  try
+  {
+    using namespace TravelingWave;
+  
+    Parameters parameters;
+
+    std::string prm_filename = "ParametersList.prm";
+    if (argc > 1)
+    {
+      // Check if file argv[1] exists.
+      if (file_exists(argv[1])) 
+      {
+        prm_filename = argv[1];
+      }
+      else
+      {
+        std::string errorMessage = "File \"" + std::string(argv[1]) + "\" is not found.";
+        throw std::runtime_error(errorMessage);
+      }
+    }
+    else
+    {
+      // Check if the file "ParametersList.prm" exists in the current or in the parent directory.
+      if (!(file_exists(prm_filename) || file_exists("../" + prm_filename)))
+      {
+        std::string errorMessage = "File \"" + prm_filename + "\" is not found.";
+        throw std::runtime_error(errorMessage);
+      }
+      else
+      {
+        if (!file_exists(prm_filename))
+        {
+          prm_filename = "../" + prm_filename;
+        }
+      }
+    }
+
+    std::cout << "Reading parameters... " << std::flush;
+    ParameterAcceptor::initialize(prm_filename);
+    std::cout << "done" << std::endl;
+    
+    calculate_profile(parameters, /* With continuation_for_delta */ false, 0.1, 3);
+    
+  }
+  catch (std::exception &exc)
+  {
+    std::cerr << std::endl
+              << std::endl
+              << "----------------------------------------------------"
+              << std::endl;
+    std::cerr << "Exception on processing: " << std::endl
+              << exc.what() << std::endl
+              << "Aborting!" << std::endl
+              << "----------------------------------------------------"
+              << std::endl;
+    return 1;
+  }
+  catch (...)
+  {
+    std::cerr << std::endl
+              << std::endl
+              << "----------------------------------------------------"
+              << std::endl;
+    std::cerr << "Unknown exception!" << std::endl
+              << "Aborting!" << std::endl
+              << "----------------------------------------------------"
+              << std::endl;
+    return 1;
+  }
+
+  return 0;
+}
diff --git a/TravelingWaves/plot.py b/TravelingWaves/plot.py
new file mode 100644 (file)
index 0000000..6c80f62
--- /dev/null
@@ -0,0 +1,91 @@
+import numpy as np
+import matplotlib.pyplot as plt
+import os
+import sys
+
+plot_params = {
+  #'backend': 'pdf',
+  # 'lines.marker' : 'x',
+  'scatter.marker' : 'x',
+  'lines.markersize' : 4,
+  'lines.linewidth' : 1,
+  'axes.labelsize': 16,
+  # 'textfontsize': 12,
+  'font.size' : 16,
+  'legend.fontsize': 16,
+  'xtick.labelsize': 14,
+  'ytick.labelsize': 14,
+  'text.usetex': True,
+  'figure.figsize': [9,6],
+  'axes.grid': True
+}
+
+plt.rcParams.update(plot_params)
+
+
+if len(sys.argv) > 1:
+
+  filename = sys.argv[1]
+
+  if os.path.exists(filename):
+    data = np.loadtxt(filename, np.float64)
+    data_unique = np.unique(data, axis=0)
+    data_unique = np.array(sorted(data_unique, key=lambda x : x[0]))
+    x = data_unique[:, 0]
+    u_sol = data_unique[:, 1]
+    T_sol = data_unique[:, 2]
+    lambda_sol = data_unique[:, 3]
+
+    fig, ax = plt.subplots(nrows=1, ncols=1)
+
+    ax.scatter(x, u_sol, label=r"$u$", color='blue')
+    ax.scatter(x, T_sol, label=r"$T$", color='red')
+    ax.scatter(x, lambda_sol, label=r"$\lambda$", color='green')
+
+
+    # Plot of limit solutions for the detonation case. Uncomment, if needed.
+    #===============================================================#
+    '''
+
+    path_to_solution_files = os.path.split(filename)[0]
+    u_limit_path = os.path.join(path_to_solution_files, 'solution_u_limit.txt')
+    T_limit_path = os.path.join(path_to_solution_files, 'solution_T_limit.txt')
+    lambda_limit_path = os.path.join(path_to_solution_files, 'solution_lambda_limit.txt')
+
+    if os.path.exists(u_limit_path):
+      u_limit = np.loadtxt(u_limit_path, np.float64)
+      ax.plot(u_limit[:, 0], u_limit[:, 1], label=r"$u_{\mathrm{lim}}$", color='blue')
+      ax.plot([0, x[-1]], [u_sol[-1], u_sol[-1]], color='blue')
+    else:
+      print("No such file:", u_limit_path)
+
+    if os.path.exists(T_limit_path):
+      T_limit = np.loadtxt(T_limit_path, np.float64)
+      ax.plot(T_limit[:, 0], T_limit[:, 1], label=r"$T_{\mathrm{lim}}$", color='red')
+      ax.plot([0, x[-1]], [T_sol[-1], T_sol[-1]], color='red')
+    else:
+      print("No such file:", T_limit_path)
+
+    if os.path.exists(lambda_limit_path):
+      lambda_limit = np.loadtxt(lambda_limit_path, np.float64)
+      ax.plot(lambda_limit[:, 0], lambda_limit[:, 1], label=r"$\lambda_{\mathrm{lim}}$", color='green')
+      ax.plot([0, x[-1]], [lambda_sol[-1], lambda_sol[-1]], color='green')
+    else:
+      print("No such file:", lambda_limit_path)
+
+
+    '''
+    #===============================================================#
+    
+
+    ax.set_xlabel(r"$\xi$")
+    ax.set_ylabel(r"$u, T, \lambda$")
+    ax.legend()
+
+    # plt.savefig("fast_deflagration_delta_0.01.png", bbox_inches='tight', dpi=500)
+    # plt.savefig('slow_deflagration_delta_0.01.png', bbox_inches='tight', dpi=500)
+    # plt.savefig('detonation_delta_0.01.png', bbox_inches='tight', dpi=500)
+
+    plt.show()
+  else:
+    print("No such file:", filename)

In the beginning the Universe was created. This has made a lot of people very angry and has been widely regarded as a bad move.

Douglas Adams


Typeset in Trocchi and Trocchi Bold Sans Serif.