Chaste Commit::f2ff7ee04e70ac9d06c57344df8d017dbb12b97b
BackwardEulerIvpOdeSolver.cpp
1/*
2
3Copyright (c) 2005-2024, University of Oxford.
4All rights reserved.
5
6University of Oxford means the Chancellor, Masters and Scholars of the
7University of Oxford, having an administrative office at Wellington
8Square, Oxford OX1 2JD, UK.
9
10This file is part of Chaste.
11
12Redistribution and use in source and binary forms, with or without
13modification, are permitted provided that the following conditions are met:
14 * Redistributions of source code must retain the above copyright notice,
15 this list of conditions and the following disclaimer.
16 * Redistributions in binary form must reproduce the above copyright notice,
17 this list of conditions and the following disclaimer in the documentation
18 and/or other materials provided with the distribution.
19 * Neither the name of the University of Oxford nor the names of its
20 contributors may be used to endorse or promote products derived from this
21 software without specific prior written permission.
22
23THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
24AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
27LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE
29GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
32OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33
34*/
35
36
37#include "BackwardEulerIvpOdeSolver.hpp"
38#include <cmath>
39
41 double timeStep,
42 double time,
43 std::vector<double>& rCurrentYValues,
44 std::vector<double>& rCurrentGuess)
45{
46 std::vector<double> dy(mSizeOfOdeSystem); //For JC to optimize
47 pAbstractOdeSystem->EvaluateYDerivatives(time+timeStep, rCurrentGuess, dy);
48 for (unsigned i=0; i<mSizeOfOdeSystem; i++)
49 {
50 mResidual[i] = rCurrentGuess[i] - timeStep * dy[i] - rCurrentYValues[i];
51 }
52}
53
55 double timeStep,
56 double time,
57 std::vector<double>& rCurrentYValues,
58 std::vector<double>& rCurrentGuess)
59{
60 for (unsigned i=0; i<mSizeOfOdeSystem; i++)
61 {
62 for (unsigned j=0; j<mSizeOfOdeSystem; j++)
63 {
64 mJacobian[i][j] = 0.0;
65 }
66 }
67
68 if (pAbstractOdeSystem->GetUseAnalyticJacobian() && !mForceUseOfNumericalJacobian)
69 {
70 // The ODE system has an analytic jacobian, so use that
72 = static_cast<AbstractOdeSystemWithAnalyticJacobian*>(pAbstractOdeSystem);
73 p_ode_system->AnalyticJacobian(rCurrentGuess, mJacobian, time, timeStep);
74 }
75 else
76 {
77 ComputeNumericalJacobian(pAbstractOdeSystem,
78 timeStep,
79 time,
80 rCurrentYValues,
81 rCurrentGuess);
82 }
83}
84
86{
87 double fact;
88 for (unsigned i=0; i<mSizeOfOdeSystem; i++)
89 {
90 for (unsigned ii=i+1; ii<mSizeOfOdeSystem; ii++)
91 {
92 fact = mJacobian[ii][i]/mJacobian[i][i];
93 for (unsigned j=i; j<mSizeOfOdeSystem; j++)
94 {
95 mJacobian[ii][j] -= fact*mJacobian[i][j];
96 }
97 mResidual[ii] -= fact*mResidual[i];
98 }
99 }
100 // This needs to int, since a downloop in unsigned won't terminate properly
101 for (int i=mSizeOfOdeSystem-1; i>=0; i--)
102 {
103 mUpdate[i] = mResidual[i];
104 for (unsigned j=i+1; j<mSizeOfOdeSystem; j++)
105 {
106 mUpdate[i] -= mJacobian[i][j]*mUpdate[j];
107 }
108 mUpdate[i] /= mJacobian[i][i];
109 }
110}
111
113{
114 double norm = 0.0;
115 for (unsigned i=0; i<mSizeOfOdeSystem; i++)
116 {
117 if (fabs(pVector[i]) > norm)
118 {
119 norm = fabs(pVector[i]);
120 }
121 }
122 return norm;
123}
124
126 double timeStep,
127 double time,
128 std::vector<double>& rCurrentYValues,
129 std::vector<double>& rCurrentGuess)
130{
131 std::vector<double> residual(mSizeOfOdeSystem);
132 std::vector<double> residual_perturbed(mSizeOfOdeSystem);
133 std::vector<double> guess_perturbed(mSizeOfOdeSystem);
134
135 double epsilon = mNumericalJacobianEpsilon;
136
137 ComputeResidual(pAbstractOdeSystem, timeStep, time, rCurrentYValues, rCurrentGuess);
138 for (unsigned i=0; i<mSizeOfOdeSystem; i++)
139 {
140 residual[i] = mResidual[i];
141 }
142
143 for (unsigned global_column=0; global_column<mSizeOfOdeSystem; global_column++)
144 {
145 for (unsigned i=0; i<mSizeOfOdeSystem; i++)
146 {
147 guess_perturbed[i] = rCurrentGuess[i];
148 }
149
150 guess_perturbed[global_column] += epsilon;
151
152 ComputeResidual(pAbstractOdeSystem, timeStep, time, rCurrentYValues, guess_perturbed);
153 for (unsigned i=0; i<mSizeOfOdeSystem; i++)
154 {
155 residual_perturbed[i] = mResidual[i];
156 }
157
158 // Compute residual_perturbed - residual
159 double one_over_eps = 1.0/epsilon;
160 for (unsigned i=0; i<mSizeOfOdeSystem; i++)
161 {
162 mJacobian[i][global_column] = one_over_eps*(residual_perturbed[i] - residual[i]);
163 }
164 }
165}
166
168 double timeStep,
169 double time,
170 std::vector<double>& rCurrentYValues,
171 std::vector<double>& rNextYValues)
172{
173 // Check the size of the ODE system matches the solvers expected
174 assert(mSizeOfOdeSystem == pAbstractOdeSystem->GetNumberOfStateVariables());
175
176 unsigned counter = 0;
177// const double eps = 1e-6 * rCurrentGuess[0]; // Our tolerance (should use min(guess) perhaps?)
178 const double eps = 1e-6; // JonW tolerance
179 double norm = 2*eps;
180
181 std::vector<double> current_guess(mSizeOfOdeSystem);
182 current_guess.assign(rCurrentYValues.begin(), rCurrentYValues.end());
183
184 while (norm > eps)
185 {
186 // Calculate Jacobian and mResidual for current guess
187 ComputeResidual(pAbstractOdeSystem, timeStep, time, rCurrentYValues, current_guess);
188 ComputeJacobian(pAbstractOdeSystem, timeStep, time, rCurrentYValues, current_guess);
189// // Update norm (our style)
190// norm = ComputeNorm(mResidual);
191
192 // Solve Newton linear system
194
195 // Update norm (JonW style)
196 norm = ComputeNorm(mUpdate);
197
198 // Update current guess
199 for (unsigned i=0; i<mSizeOfOdeSystem; i++)
200 {
201 current_guess[i] -= mUpdate[i];
202 }
203
204 counter++;
205 assert(counter < 20); // avoid infinite loops
206 }
207 rNextYValues.assign(current_guess.begin(), current_guess.end());
208}
209
211{
212 mSizeOfOdeSystem = sizeOfOdeSystem;
213
214 // default epsilon
217
218 // allocate memory
219 mResidual = new double[mSizeOfOdeSystem];
220 mUpdate = new double[mSizeOfOdeSystem];
221
222 mJacobian = new double*[mSizeOfOdeSystem];
223 for (unsigned i=0; i<mSizeOfOdeSystem; i++)
224 {
225 mJacobian[i] = new double[mSizeOfOdeSystem];
226 }
227}
228
230{
231 // Delete pointers
232 delete[] mResidual;
233 delete[] mUpdate;
234
235 for (unsigned i=0; i<mSizeOfOdeSystem; i++)
236 {
237 delete[] mJacobian[i];
238 }
239 delete[] mJacobian;
240}
241
243{
244 assert(epsilon > 0);
246}
247
252
253
254// Serialization for Boost >= 1.36
#define CHASTE_CLASS_EXPORT(T)
virtual void AnalyticJacobian(const std::vector< double > &rSolutionGuess, double **jacobian, double time, double timeStep)=0
virtual void EvaluateYDerivatives(double time, const std::vector< double > &rY, std::vector< double > &rDY)=0
void ComputeNumericalJacobian(AbstractOdeSystem *pAbstractOdeSystem, double timeStep, double time, std::vector< double > &rCurrentYValues, std::vector< double > &rCurrentGuess)
void CalculateNextYValue(AbstractOdeSystem *pAbstractOdeSystem, double timeStep, double time, std::vector< double > &rCurrentYValues, std::vector< double > &rNextYValues)
void ComputeResidual(AbstractOdeSystem *pAbstractOdeSystem, double timeStep, double time, std::vector< double > &rCurrentYValues, std::vector< double > &rCurrentGuess)
void SetEpsilonForNumericalJacobian(double epsilon)
void ComputeJacobian(AbstractOdeSystem *pAbstractOdeSystem, double timeStep, double time, std::vector< double > &rCurrentYValues, std::vector< double > &rCurrentGuess)
BackwardEulerIvpOdeSolver(unsigned sizeOfOdeSystem)