1201 lines
59 KiB
C++
1201 lines
59 KiB
C++
#ifndef REDUCEDMODELOPTIMIZER_STRUCTS_HPP
|
|
#define REDUCEDMODELOPTIMIZER_STRUCTS_HPP
|
|
#include "csvfile.hpp"
|
|
#include "drmsimulationmodel.hpp"
|
|
#include "linearsimulationmodel.hpp"
|
|
#include "simulation_structs.hpp"
|
|
#include "unordered_map"
|
|
#include <string>
|
|
|
|
namespace ReducedModelOptimization {
|
|
|
|
enum BaseSimulationScenario {
|
|
Axial,
|
|
Shear,
|
|
Bending,
|
|
Dome,
|
|
Saddle,
|
|
S,
|
|
NumberOfBaseSimulationScenarios
|
|
};
|
|
inline static std::vector<std::string> baseSimulationScenarioNames{"Axial",
|
|
"Shear",
|
|
"Bending",
|
|
"Dome",
|
|
"Saddle",
|
|
"S"};
|
|
|
|
struct Colors
|
|
{
|
|
inline static std::array<double, 3> fullInitial{0.094, 0.094, 0.094};
|
|
// inline static std::array<double, 3> fullDeformed{0.583333, 0.890196, 0.109804};
|
|
inline static std::array<double, 3> fullDeformed{0.094, 0.094, 0.094};
|
|
inline static std::array<double, 3> reducedInitial{0.890196, 0.109804, 0.193138};
|
|
inline static std::array<double, 3> reducedDeformed{0.262, 0.627, 0.910};
|
|
};
|
|
|
|
struct xRange
|
|
{
|
|
std::string label{};
|
|
double min{0};
|
|
double max{0};
|
|
|
|
inline bool operator<(const xRange &other) { return label < other.label; }
|
|
std::string toString() const
|
|
{
|
|
return label + "=[" + std::to_string(min) + "," + std::to_string(max) + "]";
|
|
}
|
|
|
|
void fromString(const std::string &s)
|
|
{
|
|
const std::size_t equalPos = s.find("=");
|
|
label = s.substr(0, equalPos);
|
|
const std::size_t commaPos = s.find(",");
|
|
const size_t minBeginPos = equalPos + 2;
|
|
min = std::stod(s.substr(minBeginPos, commaPos - minBeginPos));
|
|
const size_t maxBeginPos = commaPos + 1;
|
|
const std::size_t closingBracketPos = s.find("]");
|
|
max = std::stod(s.substr(maxBeginPos, closingBracketPos - maxBeginPos));
|
|
}
|
|
|
|
bool operator==(const xRange &xrange) const
|
|
{
|
|
return label == xrange.label && min == xrange.min && max == xrange.max;
|
|
}
|
|
|
|
std::tuple<std::string, double, double> toTuple() const
|
|
{
|
|
return std::make_tuple(label, min, max);
|
|
}
|
|
|
|
void set(const std::tuple<std::string, double, double> &inputTuple)
|
|
{
|
|
if (std::get<1>(inputTuple) > std::get<2>(inputTuple)) {
|
|
std::cerr
|
|
<< "Invalid xRange tuple. Second argument(min) cant be smaller than the third(max)"
|
|
<< std::endl;
|
|
std::terminate();
|
|
// return;
|
|
}
|
|
std::tie(label, min, max) = inputTuple;
|
|
}
|
|
};
|
|
|
|
enum OptimizationParameterIndex { E, A, I2, I3, J, R, Theta, NumberOfOptimizationVariables };
|
|
|
|
inline int getParameterIndex(const std::string &s)
|
|
{
|
|
if ("E" == s) {
|
|
return E;
|
|
} else if ("A" == s) {
|
|
return A;
|
|
} else if ("I2" == s) {
|
|
return I2;
|
|
} else if ("I3" == s) {
|
|
return I3;
|
|
} else if ("J" == s) {
|
|
return J;
|
|
} else if ("R" == s || "HexSize" == s) {
|
|
return R;
|
|
} else if ("Theta" == s || "HexAngle" == s) {
|
|
return Theta;
|
|
} else {
|
|
std::cerr << "Input is not recognized as a valid optimization variable index:" << s
|
|
<< std::endl;
|
|
return -1;
|
|
}
|
|
}
|
|
|
|
struct Settings
|
|
{
|
|
inline static std::string defaultFilename{"OptimizationSettings.json"};
|
|
// std::array<double, NumberOfBaseSimulationScenarios> baseScenarioMaxMagnitudes{0.590241,
|
|
// 0.888372,
|
|
// 0.368304,
|
|
// 0.0127508,
|
|
// 1.18079,
|
|
// 0}; //final
|
|
std::array<double, NumberOfBaseSimulationScenarios> baseScenarioMaxMagnitudes{0.590241 / 3,
|
|
0.588372 / 3,
|
|
0.368304,
|
|
0.1,
|
|
1.18 / 2,
|
|
0}; //final
|
|
// std::array<double, NumberOfBaseSimulationScenarios> baseScenarioMaxMagnitudes{0, 0, 0, 0.1, 0};
|
|
|
|
// std::array<double, NumberOfBaseSimulationScenarios>
|
|
// baseScenarioMaxMagnitudes{20.85302947095844,
|
|
// 1.8073431893126763,
|
|
// 0.2864731720436702,
|
|
// 0.14982243562639147,
|
|
// 0.18514829631059054};//median
|
|
// std::array<double, NumberOfBaseSimulationScenarios>
|
|
// baseScenarioMaxMagnitudes{1.1725844893199244,
|
|
// 0.3464275389927846,
|
|
// 0.09527915004635197,
|
|
// 0.06100757786262501,
|
|
// 0.10631914784812076}; //5_1565 0.03axial
|
|
// std::array<double, NumberOfBaseSimulationScenarios>
|
|
// baseScenarioMaxMagnitudes{/*15*/ 0 /*1.711973658196369*/,
|
|
// 1.878077115238504,
|
|
// 0.8,
|
|
// 0.15851675178327318,
|
|
// 0.8,
|
|
// /*1.711973658196369*/ 0}; //custom
|
|
// std::array<double, NumberOfBaseSimulationScenarios>
|
|
// baseScenarioMaxMagnitudes{0,
|
|
// 14.531854387244818,
|
|
// 0.38321932238436796,
|
|
// 0.21381267870193282,
|
|
// 0.28901381608791094,
|
|
// 1.711973658196369}; //9_14423
|
|
// std::array<double, NumberOfBaseSimulationScenarios>
|
|
// baseScenarioMaxMagnitudes{1.1725844893199244,
|
|
// 0.3464275389927846,
|
|
// 0.09527915004635197,
|
|
// 0.06100757786262501,
|
|
// 0.10631914784812076}; //5_1565 0.03axial
|
|
// std::array<double, NumberOfBaseSimulationScenarios>
|
|
// baseScenarioMagnitudes{9.82679, 0.138652, 0.247242, 0.739443, 0.00675865}; //Hyperparam opt
|
|
// std::array<double, NumberOfBaseSimulationScenarios> baseScenarioMaxMagnitudes{
|
|
// 30, 8, 0.4421382884449713, 0.22758433903942452, 0.3247935583883217};
|
|
|
|
// std::array<double, NumberOfBaseSimulationScenarios>
|
|
// baseScenarioMagnitudes{10 * 6.310485381644259,
|
|
// 10 * 1.7100142258819078,
|
|
// 10 * 0.18857048204421728,
|
|
// 10 * 0.10813697502645818,
|
|
// 10 * 0.11982893539207524}; //6_338
|
|
// std::array<double, NumberOfBaseSimulationScenarios> baseScenarioMagnitudes{7.72224,
|
|
// 7.72224,
|
|
// 0.89468,
|
|
// 0.445912,
|
|
// 0.625905};
|
|
// std::array<double, NumberOfBaseSimulationScenarios> baseScenarioMagnitudes{0.407714,
|
|
// 22.3524,
|
|
// 0.703164,
|
|
// 0.0226138,
|
|
// 0.161316};
|
|
// std::array<double, NumberOfBaseSimulationScenarios>
|
|
// baseScenarioMagnitudes{2, 1, 0.4, 0.2, 0.2}; //8_15444 magnitudes from randomBending0
|
|
|
|
// std::array<double, NumberOfBaseSimulationScenarios>
|
|
// baseScenarioMagnitudes{1.0600375226325425,
|
|
// 0.6381040280710403,
|
|
// 0.17201755995098306,
|
|
// 0.0706601822149856,
|
|
// 0.13578373479448072}; //8_15444 magnitudes from displacements
|
|
// std::array<double, NumberOfBaseSimulationScenarios>
|
|
// baseScenarioMagnitudes{10 * 1.0600375226325425,
|
|
// 10 * 0.6381040280710403,
|
|
// 10 * 0.17201755995098306,
|
|
// 10 * 0.0706601822149856,
|
|
// 10 * 0.13578373479448072}; //8_15444 magnitudes from displacements
|
|
|
|
// std::array<double, NumberOfBaseSimulationScenarios> baseScenarioMaxMagnitudes;
|
|
std::vector<std::vector<OptimizationParameterIndex>> optimizationStrategy = {
|
|
// {E,A, I2, I3, J, R, Theta}};
|
|
{A, I2, I3, J, R, Theta}};
|
|
std::optional<std::vector<double>>
|
|
optimizationVariablesGroupsWeights; //TODO:should be removed in the future if not a splitting strategy is used for the optimization
|
|
enum NormalizationStrategy { NonNormalized, Epsilon };
|
|
inline static std::vector<std::string> normalizationStrategyStrings{"NonNormalized", "Epsilon"};
|
|
NormalizationStrategy normalizationStrategy{Epsilon};
|
|
std::array<xRange, NumberOfOptimizationVariables> variablesRanges{xRange{"E", 1e-3, 1e3},
|
|
xRange{"A", 1e-3, 1e3},
|
|
xRange{"I2", 1e-3, 1e3},
|
|
xRange{"I3", 1e-3, 1e3},
|
|
xRange{"J", 1e-3, 1e3},
|
|
xRange{"R", 0.05, 0.95},
|
|
xRange{"Theta", -30, 30}};
|
|
struct SettingsPSO
|
|
{
|
|
int numberOfParticles{200};
|
|
#ifdef USE_PSO
|
|
inline static std::string optimizerName{"pso"};
|
|
#else
|
|
inline static std::string optimizerName{"sa"};
|
|
#endif
|
|
} pso;
|
|
|
|
struct SettingsDlibGlobal
|
|
{
|
|
int numberOfFunctionCalls{100000};
|
|
} dlib;
|
|
|
|
double solverAccuracy{1e-2};
|
|
double translationEpsilon{4e-3};
|
|
// double translationEpsilon{0};
|
|
// double angularDistanceEpsilon{vcg::math::ToRad(2.0)};
|
|
double angularDistanceEpsilon{vcg::math::ToRad(0.0)};
|
|
double targetBaseTriangleSize{0.03};
|
|
std::filesystem::path intermediateResultsDirectoryPath;
|
|
struct ObjectiveWeights
|
|
{
|
|
double translational{1.2};
|
|
double rotational{0.8};
|
|
bool operator==(const ObjectiveWeights &other) const;
|
|
};
|
|
std::array<ObjectiveWeights, NumberOfBaseSimulationScenarios> perBaseScenarioObjectiveWeights;
|
|
// std::array<ObjectiveWeights, NumberOfBaseSimulationScenarios> perBaseScenarioObjectiveWeights{
|
|
// {{1.95, 0.05}, {0.87, 1.13}, {0.37, 1.63}, {0.01, 1.99}, {0.94, 1.06}, {1.2, 0.8}}};
|
|
std::array<std::pair<double, double>, NumberOfBaseSimulationScenarios>
|
|
convertObjectiveWeightsToPairs() const
|
|
{
|
|
std::array<std::pair<double, double>, NumberOfBaseSimulationScenarios> objectiveWeightsPairs;
|
|
for (int baseScenario = Axial; baseScenario != NumberOfBaseSimulationScenarios;
|
|
baseScenario++) {
|
|
objectiveWeightsPairs[baseScenario]
|
|
= std::make_pair(perBaseScenarioObjectiveWeights[baseScenario].translational,
|
|
perBaseScenarioObjectiveWeights[baseScenario].rotational);
|
|
}
|
|
return objectiveWeightsPairs;
|
|
}
|
|
|
|
struct JsonKeys
|
|
{
|
|
inline static std::string OptimizationStrategy{"OptimizationStrategy"};
|
|
inline static std::string OptimizationStrategyGroupWeights{
|
|
"OptimizationStrategyGroupWeights"};
|
|
inline static std::string OptimizationVariables{"OptimizationVariables"};
|
|
inline static std::string NumberOfFunctionCalls{"NumberOfFunctionCalls"};
|
|
inline static std::string SolverAccuracy{"SolverAccuracy"};
|
|
inline static std::string ObjectiveWeights{"ObjectiveWeight"};
|
|
};
|
|
|
|
nlohmann::json toJson() const
|
|
{
|
|
nlohmann::json json;
|
|
json[GET_VARIABLE_NAME(optimizationStrategy)] = optimizationStrategy;
|
|
if (optimizationVariablesGroupsWeights.has_value()) {
|
|
json[GET_VARIABLE_NAME(ptimizationStrategyGroupWeights)]
|
|
= optimizationVariablesGroupsWeights.value();
|
|
}
|
|
//write x ranges
|
|
const std::array<std::tuple<std::string, double, double>, NumberOfOptimizationVariables>
|
|
xRangesAsTuples = [=]() {
|
|
std::array<std::tuple<std::string, double, double>, NumberOfOptimizationVariables>
|
|
xRangesAsTuples;
|
|
for (int optimizationParameterIndex = E;
|
|
optimizationParameterIndex != NumberOfOptimizationVariables;
|
|
optimizationParameterIndex++) {
|
|
xRangesAsTuples[optimizationParameterIndex]
|
|
= variablesRanges[optimizationParameterIndex].toTuple();
|
|
}
|
|
return xRangesAsTuples;
|
|
}();
|
|
json[JsonKeys::OptimizationVariables] = xRangesAsTuples;
|
|
// for (size_t xRangeIndex = 0; xRangeIndex < variablesRanges.size(); xRangeIndex++) {
|
|
// const auto &xRange = variablesRanges[xRangeIndex];
|
|
// json[JsonKeys::OptimizationVariables + "_" + std::to_string(xRangeIndex)]
|
|
// = xRange.toString();
|
|
// }
|
|
|
|
json[GET_VARIABLE_NAME(solverAccuracy)] = solverAccuracy;
|
|
//Objective weights
|
|
std::array<std::pair<double, double>, NumberOfBaseSimulationScenarios> objectiveWeightsPairs;
|
|
std::transform(perBaseScenarioObjectiveWeights.begin(),
|
|
perBaseScenarioObjectiveWeights.end(),
|
|
objectiveWeightsPairs.begin(),
|
|
[](const ObjectiveWeights &objectiveWeights) {
|
|
return std::make_pair(objectiveWeights.translational,
|
|
objectiveWeights.rotational);
|
|
});
|
|
json[JsonKeys::ObjectiveWeights] = objectiveWeightsPairs;
|
|
json[GET_VARIABLE_NAME(translationEpsilon)] = translationEpsilon;
|
|
json[GET_VARIABLE_NAME(angularDistanceEpsilon)] = vcg::math::ToDeg(angularDistanceEpsilon);
|
|
json[GET_VARIABLE_NAME(targetBaseTriangleSize)] = targetBaseTriangleSize;
|
|
json[GET_VARIABLE_NAME(baseScenarioMaxMagnitudes)] = baseScenarioMaxMagnitudes;
|
|
|
|
#ifdef USE_ENSMALLEN
|
|
#ifdef USE_PSO
|
|
json[GET_VARIABLE_NAME(pso.numberOfParticles)] = pso.numberOfParticles;
|
|
#endif
|
|
json[GET_VARIABLE_NAME(pso.optimizerName)] = pso.optimizerName;
|
|
#else
|
|
json[GET_VARIABLE_NAME(dlib.numberOfFunctionCalls)] = dlib.numberOfFunctionCalls;
|
|
#endif
|
|
return json;
|
|
}
|
|
void save(const std::filesystem::path &saveToPath)
|
|
{
|
|
assert(std::filesystem::is_directory(saveToPath));
|
|
nlohmann::json json = toJson();
|
|
std::filesystem::path jsonFilePath(
|
|
std::filesystem::path(saveToPath).append(defaultFilename));
|
|
std::ofstream jsonFile(jsonFilePath.string());
|
|
jsonFile << json;
|
|
jsonFile.close();
|
|
}
|
|
|
|
bool load(const std::filesystem::path &jsonFilePath)
|
|
{
|
|
if (!std::filesystem::exists(jsonFilePath)) {
|
|
std::cerr << "Optimization settings could not be loaded because input filepath does "
|
|
"not exist:"
|
|
<< jsonFilePath << std::endl;
|
|
assert(false);
|
|
return false;
|
|
}
|
|
std::ifstream ifs(jsonFilePath);
|
|
nlohmann::json json;
|
|
ifs >> json;
|
|
|
|
if (json.contains(GET_VARIABLE_NAME(optimizationStrategy))) {
|
|
optimizationStrategy
|
|
= std::vector<std::vector<ReducedModelOptimization::OptimizationParameterIndex>>(
|
|
(json.at(GET_VARIABLE_NAME(optimizationStrategy))));
|
|
}
|
|
if (json.contains(GET_VARIABLE_NAME(optimizationStrategyGroupWeights))) {
|
|
optimizationVariablesGroupsWeights = std::vector<double>(
|
|
json[GET_VARIABLE_NAME(optimizationStrategyGroupWeights)]);
|
|
}
|
|
//read x ranges
|
|
if (json.contains(JsonKeys::OptimizationVariables)) {
|
|
const std::array<std::tuple<std::string, double, double>, NumberOfOptimizationVariables>
|
|
xRangesAsTuples = json.at(JsonKeys::OptimizationVariables);
|
|
for (const auto &rangeTuple : xRangesAsTuples) {
|
|
variablesRanges[getParameterIndex(std::get<0>(rangeTuple))].set(rangeTuple);
|
|
}
|
|
} else { //NOTE:legacy compatibility
|
|
size_t xRangeIndex = 0;
|
|
while (true) {
|
|
const std::string jsonXRangeKey = JsonKeys::OptimizationVariables + "_"
|
|
+ std::to_string(xRangeIndex++);
|
|
if (!json.contains(jsonXRangeKey)) {
|
|
break;
|
|
}
|
|
xRange x;
|
|
x.fromString(json.at(jsonXRangeKey));
|
|
variablesRanges[getParameterIndex(x.label)] = x;
|
|
}
|
|
}
|
|
|
|
if (json.contains(GET_VARIABLE_NAME(dlib.numberOfFunctionCalls))) {
|
|
dlib.numberOfFunctionCalls = json.at(GET_VARIABLE_NAME(dlib.numberOfFunctionCalls));
|
|
}
|
|
if (json.contains(GET_VARIABLE_NAME(solverAccuracy))) {
|
|
solverAccuracy = json.at(GET_VARIABLE_NAME(solverAccuracy));
|
|
}
|
|
//Objective weights
|
|
if (json.contains(JsonKeys::ObjectiveWeights)) {
|
|
std::array<std::pair<double, double>, NumberOfBaseSimulationScenarios>
|
|
objectiveWeightsPairs = json.at(JsonKeys::ObjectiveWeights);
|
|
std::transform(objectiveWeightsPairs.begin(),
|
|
objectiveWeightsPairs.end(),
|
|
perBaseScenarioObjectiveWeights.begin(),
|
|
[](const std::pair<double, double> &objectiveWeightsPair) {
|
|
return ObjectiveWeights{objectiveWeightsPair.first,
|
|
objectiveWeightsPair.second};
|
|
});
|
|
}
|
|
if (json.contains(GET_VARIABLE_NAME(translationalNormalizationEpsilon))) {
|
|
translationEpsilon
|
|
= json[GET_VARIABLE_NAME(translationalNormalizationEpsilon)];
|
|
}
|
|
|
|
if (json.contains(GET_VARIABLE_NAME(angularDistanceEpsilon))) {
|
|
angularDistanceEpsilon = vcg::math::ToRad(
|
|
static_cast<double>(json[GET_VARIABLE_NAME(angularDistanceEpsilon)]));
|
|
}
|
|
|
|
if (json.contains(GET_VARIABLE_NAME(targetBaseTriangleSize))) {
|
|
targetBaseTriangleSize = static_cast<double>(
|
|
json[GET_VARIABLE_NAME(targetBaseTriangleSize)]);
|
|
}
|
|
|
|
if (json.contains(GET_VARIABLE_NAME(pso.numberOfParticles))) {
|
|
pso.numberOfParticles = static_cast<int>(json[GET_VARIABLE_NAME(pso.numberOfParticles)]);
|
|
}
|
|
|
|
// perBaseScenarioObjectiveWeights = json.at(JsonKeys::ObjectiveWeights);
|
|
// objectiveWeights.translational = json.at(JsonKeys::ObjectiveWeights);
|
|
// objectiveWeights.rotational = 2 - objectiveWeights.translational;
|
|
return true;
|
|
}
|
|
|
|
std::string toString() const { return toJson().dump(); }
|
|
|
|
void writeHeaderTo(csvFile &os) const
|
|
{
|
|
if (!variablesRanges.empty()) {
|
|
for (const xRange &range : variablesRanges) {
|
|
os << range.label + " max";
|
|
os << range.label + " min";
|
|
}
|
|
}
|
|
os << "Function Calls";
|
|
os << "Solution Accuracy";
|
|
os << "Normalization trans epsilon";
|
|
os << "Normalization rot epsilon(deg)";
|
|
os << JsonKeys::ObjectiveWeights;
|
|
os << "Optimization parameters";
|
|
// os << std::endl;
|
|
}
|
|
|
|
void writeSettingsTo(csvFile &os) const
|
|
{
|
|
if (!variablesRanges.empty()) {
|
|
for (const xRange &range : variablesRanges) {
|
|
os << range.max;
|
|
os << range.min;
|
|
}
|
|
}
|
|
os << dlib.numberOfFunctionCalls;
|
|
os << solverAccuracy;
|
|
os << std::to_string(translationEpsilon);
|
|
os << std::to_string(vcg::math::ToDeg(angularDistanceEpsilon));
|
|
std::string objectiveWeightsString;
|
|
objectiveWeightsString += "{";
|
|
for (int baseScenario = Axial; baseScenario != NumberOfBaseSimulationScenarios;
|
|
baseScenario++) {
|
|
objectiveWeightsString
|
|
+= "{" + std::to_string(perBaseScenarioObjectiveWeights[baseScenario].translational)
|
|
+ "," + std::to_string(perBaseScenarioObjectiveWeights[baseScenario].rotational)
|
|
+ "}";
|
|
}
|
|
objectiveWeightsString += "}";
|
|
os << objectiveWeightsString;
|
|
|
|
//export optimization parameters
|
|
std::vector<std::vector<int>> vv;
|
|
for (const std::vector<OptimizationParameterIndex> &v : optimizationStrategy) {
|
|
std::vector<int> vi;
|
|
vi.reserve(v.size());
|
|
for (const OptimizationParameterIndex ¶meter : v) {
|
|
vi.emplace_back(parameter);
|
|
}
|
|
vv.push_back(vi);
|
|
}
|
|
os << Utilities::toString(vv);
|
|
}
|
|
};
|
|
|
|
inline bool operator==(const Settings &settings1, const Settings &settings2) noexcept
|
|
{
|
|
const bool haveTheSameObjectiveWeights
|
|
= std::mismatch(settings1.perBaseScenarioObjectiveWeights.begin(),
|
|
settings1.perBaseScenarioObjectiveWeights.end(),
|
|
settings2.perBaseScenarioObjectiveWeights.begin())
|
|
.first
|
|
== settings1.perBaseScenarioObjectiveWeights.end();
|
|
return settings1.dlib.numberOfFunctionCalls == settings2.dlib.numberOfFunctionCalls
|
|
&& settings1.variablesRanges == settings2.variablesRanges
|
|
&& settings1.solverAccuracy == settings2.solverAccuracy && haveTheSameObjectiveWeights
|
|
&& settings1.translationEpsilon == settings2.translationEpsilon;
|
|
}
|
|
|
|
inline void updateMeshWithOptimalVariables(const std::vector<double> &x, SimulationMesh &m)
|
|
{
|
|
assert(CrossSectionType::name == RectangularBeamDimensions::name);
|
|
const double E = x[0];
|
|
const double A = x[1];
|
|
const double beamWidth = std::sqrt(A);
|
|
const double beamHeight = beamWidth;
|
|
|
|
const double J = x[2];
|
|
const double I2 = x[3];
|
|
const double I3 = x[4];
|
|
for (EdgeIndex ei = 0; ei < m.EN(); ei++) {
|
|
Element &e = m.elements[ei];
|
|
e.setDimensions(CrossSectionType(beamWidth, beamHeight));
|
|
e.setMaterial(ElementMaterial(e.material.poissonsRatio, E));
|
|
e.dimensions.inertia.J = J;
|
|
e.dimensions.inertia.I2 = I2;
|
|
e.dimensions.inertia.I3 = I3;
|
|
}
|
|
|
|
CoordType center_barycentric(1, 0, 0);
|
|
CoordType interfaceEdgeMiddle_barycentric(0, 0.5, 0.5);
|
|
CoordType movableVertex_barycentric((center_barycentric + interfaceEdgeMiddle_barycentric)
|
|
* x[x.size() - 2]);
|
|
|
|
CoordType patternCoord0 = CoordType(0, 0, 0);
|
|
double bottomEdgeHalfSize = 0.03 / std::tan(M_PI / 3);
|
|
CoordType interfaceNodePosition(0, -0.03, 0);
|
|
CoordType patternBottomRight = interfaceNodePosition + CoordType(bottomEdgeHalfSize, 0, 0);
|
|
CoordType patternBottomLeft = interfaceNodePosition - CoordType(bottomEdgeHalfSize, 0, 0);
|
|
vcg::Triangle3<double> baseTriangle(patternCoord0, patternBottomLeft, patternBottomRight);
|
|
CoordType baseTriangleMovableVertexPosition = baseTriangle.cP(0)
|
|
* movableVertex_barycentric[0]
|
|
+ baseTriangle.cP(1)
|
|
* movableVertex_barycentric[1]
|
|
+ baseTriangle.cP(2)
|
|
* movableVertex_barycentric[2];
|
|
VectorType patternPlaneNormal(0, 0, 1);
|
|
baseTriangleMovableVertexPosition = vcg::RotationMatrix(patternPlaneNormal,
|
|
vcg::math::ToRad(x[x.size() - 1]))
|
|
* baseTriangleMovableVertexPosition;
|
|
|
|
const int fanSize = 6;
|
|
for (int rotationCounter = 0; rotationCounter < fanSize; rotationCounter++) {
|
|
m.vert[2 * rotationCounter + 1].P() = vcg::RotationMatrix(patternPlaneNormal,
|
|
vcg::math::ToRad(
|
|
60.0 * rotationCounter))
|
|
* baseTriangleMovableVertexPosition;
|
|
}
|
|
|
|
m.reset();
|
|
}
|
|
|
|
struct Results
|
|
{
|
|
std::string label{"EmptyLabel"};
|
|
double time{-1};
|
|
bool wasSuccessful{true};
|
|
// int numberOfSimulationCrashes{0};
|
|
Settings settings;
|
|
|
|
std::vector<std::pair<std::string, double>> optimalXNameValuePairs;
|
|
std::vector<std::shared_ptr<SimulationJob>> fullPatternSimulationJobs;
|
|
std::vector<std::shared_ptr<SimulationJob>> reducedPatternSimulationJobs;
|
|
double fullPatternYoungsModulus{0};
|
|
|
|
PatternGeometry baseTriangleFullPattern; //non-fanned,non-tiled full pattern
|
|
vcg::Triangle3<double> baseTriangle;
|
|
// std::string notes;
|
|
|
|
//Data gathered for csv exporting
|
|
struct ObjectiveValues
|
|
{
|
|
double totalRaw;
|
|
double total;
|
|
std::vector<double> perSimulationScenario_rawTranslational;
|
|
std::vector<double> perSimulationScenario_rawRotational;
|
|
std::vector<double> perSimulationScenario_translational;
|
|
std::vector<double> perSimulationScenario_rotational;
|
|
std::vector<double> perSimulationScenario_total;
|
|
std::vector<double> perSimulationScenario_total_unweighted;
|
|
|
|
} objectiveValue;
|
|
|
|
std::vector<double> perScenario_fullPatternPotentialEnergy;
|
|
std::vector<double> objectiveValueHistory;
|
|
std::vector<size_t> objectiveValueHistory_iteration;
|
|
|
|
inline static std::string DefaultFileName{"OptimizationResults.json"};
|
|
|
|
struct CSVExportingSettings
|
|
{
|
|
bool exportPE{false};
|
|
bool exportIterationOfMinima{false};
|
|
bool exportRawObjectiveValue{false};
|
|
CSVExportingSettings() {}
|
|
};
|
|
|
|
const CSVExportingSettings exportSettings;
|
|
struct JsonKeys
|
|
{
|
|
inline static std::string optimizationVariables{"OptimizationVariables"};
|
|
inline static std::string baseTriangle{"BaseTriangle"};
|
|
inline static std::string Label{"Label"};
|
|
inline static std::string FullPatternLabel{"FullPatternLabel"};
|
|
inline static std::string Settings{"OptimizationSettings"};
|
|
inline static std::string FullPatternPotentialEnergies{"PE"};
|
|
inline static std::string fullPatternYoungsModulus{"youngsModulus"};
|
|
};
|
|
|
|
void saveObjectiveValuePlot(const std::filesystem::path &outputImageDirPath) const
|
|
{
|
|
std::vector<std::string> scenarioLabels(objectiveValue.perSimulationScenario_total.size());
|
|
const double colorAxial = 1;
|
|
const double colorShear = 3;
|
|
const double colorBending = 5;
|
|
const double colorDome = 0.1;
|
|
const double colorSaddle = 0;
|
|
std::vector<double> colors(objectiveValue.perSimulationScenario_total.size());
|
|
for (int scenarioIndex = 0; scenarioIndex < scenarioLabels.size(); scenarioIndex++) {
|
|
scenarioLabels[scenarioIndex] = reducedPatternSimulationJobs[scenarioIndex]->getLabel();
|
|
if (scenarioLabels[scenarioIndex].rfind("Axial", 0) == 0) {
|
|
colors[scenarioIndex] = colorAxial;
|
|
|
|
} else if (scenarioLabels[scenarioIndex].rfind("Shear", 0) == 0) {
|
|
colors[scenarioIndex] = colorShear;
|
|
|
|
} else if (scenarioLabels[scenarioIndex].rfind("Bending", 0) == 0) {
|
|
colors[scenarioIndex] = colorBending;
|
|
|
|
} else if (scenarioLabels[scenarioIndex].rfind("Dome", 0) == 0) {
|
|
colors[scenarioIndex] = colorDome;
|
|
} else if (scenarioLabels[scenarioIndex].rfind("Saddle", 0) == 0) {
|
|
colors[scenarioIndex] = colorSaddle;
|
|
} else {
|
|
std::cerr << "Label could not be identified" << std::endl;
|
|
}
|
|
}
|
|
std::vector<double> y(objectiveValue.perSimulationScenario_total.size());
|
|
for (int scenarioIndex = 0; scenarioIndex < scenarioLabels.size(); scenarioIndex++) {
|
|
y[scenarioIndex]
|
|
// = optimizationResults.objectiveValue.perSimulationScenario_rawTranslational[scenarioIndex]
|
|
// + optimizationResults.objectiveValue.perSimulationScenario_rawRotational[scenarioIndex];
|
|
= objectiveValue.perSimulationScenario_total_unweighted[scenarioIndex];
|
|
}
|
|
std::vector<double> x = matplot::linspace(0, y.size() - 1, y.size());
|
|
std::vector<double> markerSizes(y.size(), 5);
|
|
Utilities::createPlot("scenario index",
|
|
"error",
|
|
x,
|
|
y,
|
|
markerSizes,
|
|
colors,
|
|
std::filesystem::path(outputImageDirPath)
|
|
.append("perScenarioObjectiveValues.svg"));
|
|
}
|
|
|
|
void save(const std::string &saveToPath, const bool shouldExportDebugFiles = false)
|
|
{
|
|
//clear directory
|
|
if (std::filesystem::exists(saveToPath)) {
|
|
for (const auto &entry : std::filesystem::directory_iterator(saveToPath)) {
|
|
std::error_code ec;
|
|
std::filesystem::remove_all(entry.path(), ec);
|
|
}
|
|
}
|
|
std::filesystem::create_directories(saveToPath);
|
|
|
|
//Save optimal X
|
|
nlohmann::json json_optimizationResults;
|
|
json_optimizationResults[JsonKeys::Label] = label;
|
|
if (wasSuccessful) {
|
|
std::string jsonValue_optimizationVariables;
|
|
for (const auto &optimalXNameValuePair : optimalXNameValuePairs) {
|
|
jsonValue_optimizationVariables.append(optimalXNameValuePair.first + ",");
|
|
}
|
|
jsonValue_optimizationVariables.pop_back(); // for deleting the last comma
|
|
json_optimizationResults[JsonKeys::optimizationVariables]
|
|
= jsonValue_optimizationVariables;
|
|
|
|
for (const auto &optimalXNameValuePair : optimalXNameValuePairs) {
|
|
json_optimizationResults[optimalXNameValuePair.first] = optimalXNameValuePair
|
|
.second;
|
|
}
|
|
}
|
|
//base triangle
|
|
json_optimizationResults[JsonKeys::baseTriangle]
|
|
= std::vector<double>{baseTriangle.cP0(0)[0],
|
|
baseTriangle.cP0(0)[1],
|
|
baseTriangle.cP0(0)[2],
|
|
baseTriangle.cP1(0)[0],
|
|
baseTriangle.cP1(0)[1],
|
|
baseTriangle.cP1(0)[2],
|
|
baseTriangle.cP2(0)[0],
|
|
baseTriangle.cP2(0)[1],
|
|
baseTriangle.cP2(0)[2]};
|
|
baseTriangleFullPattern.save(std::filesystem::path(saveToPath).string());
|
|
json_optimizationResults[JsonKeys::FullPatternLabel] = baseTriangleFullPattern.getLabel();
|
|
|
|
//potential energies
|
|
// const int numberOfSimulationJobs = fullPatternSimulationJobs.size();
|
|
// std::vector<double> fullPatternPE(numberOfSimulationJobs);
|
|
// for (int simulationScenarioIndex = 0; simulationScenarioIndex < numberOfSimulationJobs;
|
|
// simulationScenarioIndex++) {
|
|
// fullPatternPE[simulationScenarioIndex]
|
|
// = perScenario_fullPatternPotentialEnergy[simulationScenarioIndex];
|
|
// }
|
|
// json_optimizationResults[JsonKeys::FullPatternPotentialEnergies] = fullPatternPE;
|
|
json_optimizationResults[JsonKeys::fullPatternYoungsModulus] = fullPatternYoungsModulus;
|
|
////Save to json file
|
|
std::filesystem::path jsonFilePath(
|
|
std::filesystem::path(saveToPath).append(DefaultFileName));
|
|
std::ofstream jsonFile_optimizationResults(jsonFilePath.string());
|
|
jsonFile_optimizationResults << json_optimizationResults;
|
|
|
|
/*TODO: Refactor since the meshes are saved for each simulation scenario although they do not change*/
|
|
//Save jobs and meshes for each simulation scenario
|
|
|
|
if (shouldExportDebugFiles) {
|
|
//Save the reduced and full patterns
|
|
const std::filesystem::path simulationJobsPath(
|
|
std::filesystem::path(saveToPath).append("SimulationJobs"));
|
|
const int numberOfSimulationJobs = fullPatternSimulationJobs.size();
|
|
for (int simulationScenarioIndex = 0;
|
|
simulationScenarioIndex < numberOfSimulationJobs;
|
|
simulationScenarioIndex++) {
|
|
const std::shared_ptr<SimulationJob> &pFullPatternSimulationJob
|
|
= fullPatternSimulationJobs[simulationScenarioIndex];
|
|
std::filesystem::path simulationJobFolderPath(
|
|
std::filesystem::path(simulationJobsPath)
|
|
.append(std::to_string(simulationScenarioIndex) + "_"
|
|
+ pFullPatternSimulationJob->label));
|
|
std::filesystem::create_directories(simulationJobFolderPath);
|
|
const auto fullPatternDirectoryPath
|
|
= std::filesystem::path(simulationJobFolderPath).append("Full");
|
|
std::filesystem::create_directory(fullPatternDirectoryPath);
|
|
pFullPatternSimulationJob->save(fullPatternDirectoryPath.string());
|
|
const std::shared_ptr<SimulationJob> &pReducedPatternSimulationJob
|
|
= reducedPatternSimulationJobs[simulationScenarioIndex];
|
|
const auto reducedPatternDirectoryPath
|
|
= std::filesystem::path(simulationJobFolderPath).append("Reduced");
|
|
if (!std::filesystem::exists(reducedPatternDirectoryPath)) {
|
|
std::filesystem::create_directory(reducedPatternDirectoryPath);
|
|
}
|
|
pReducedPatternSimulationJob->save(reducedPatternDirectoryPath.string());
|
|
}
|
|
|
|
// constexpr bool shouldSaveObjectiveValuePlot = shouldExportDebugFiles;
|
|
// if (shouldSaveObjectiveValuePlot) {
|
|
saveObjectiveValuePlot(saveToPath);
|
|
// }
|
|
}
|
|
csvFile csv_resultsLocalFile(std::filesystem::path(saveToPath).append("results.csv"),
|
|
true);
|
|
csv_resultsLocalFile << "Name";
|
|
writeHeaderTo(csv_resultsLocalFile);
|
|
settings.writeHeaderTo(csv_resultsLocalFile);
|
|
csv_resultsLocalFile << endrow;
|
|
csv_resultsLocalFile << baseTriangleFullPattern.getLabel();
|
|
writeResultsTo(csv_resultsLocalFile);
|
|
settings.writeSettingsTo(csv_resultsLocalFile);
|
|
csv_resultsLocalFile << endrow;
|
|
|
|
//save minima info
|
|
// std::filesystem::path csvFilepathMinimaInfo = std::filesystem::path(saveToPath)
|
|
// .append("minimaInfo.csv");
|
|
// csvFile csv_minimaInfo(csvFilepathMinimaInfo, false);
|
|
// writeMinimaInfoTo(csv_minimaInfo);
|
|
|
|
settings.save(saveToPath);
|
|
|
|
#ifdef POLYSCOPE_DEFINED
|
|
std::cout << "Saved optimization results to:" << saveToPath << std::endl;
|
|
#endif
|
|
}
|
|
|
|
bool load(const std::filesystem::path &loadFromPath, const bool &shouldLoadDebugFiles = false)
|
|
{
|
|
assert(std::filesystem::is_directory(loadFromPath));
|
|
std::filesystem::path jsonFilepath(
|
|
std::filesystem::path(loadFromPath).append(DefaultFileName));
|
|
if (!std::filesystem::exists(jsonFilepath)) {
|
|
std::cerr << "Input path does not exist:" << loadFromPath << std::endl;
|
|
return false;
|
|
}
|
|
//Load optimal X
|
|
nlohmann::json json_optimizationResults;
|
|
std::ifstream ifs(std::filesystem::path(loadFromPath).append(DefaultFileName));
|
|
ifs >> json_optimizationResults;
|
|
|
|
// std::cout << json_optimizationResults.dump() << std::endl;
|
|
|
|
label = json_optimizationResults.at(JsonKeys::Label);
|
|
std::string optimizationVariablesString = *json_optimizationResults.find(
|
|
JsonKeys::optimizationVariables);
|
|
std::string optimizationVariablesDelimeter = ",";
|
|
size_t pos = 0;
|
|
std::vector<std::string> optimizationVariablesNames;
|
|
while ((pos = optimizationVariablesString.find(optimizationVariablesDelimeter))
|
|
!= std::string::npos) {
|
|
const std::string variableName = optimizationVariablesString.substr(0, pos);
|
|
optimizationVariablesNames.push_back(variableName);
|
|
optimizationVariablesString.erase(0, pos + optimizationVariablesDelimeter.length());
|
|
}
|
|
optimizationVariablesNames.push_back(optimizationVariablesString); //add last variable name
|
|
|
|
optimalXNameValuePairs.resize(optimizationVariablesNames.size());
|
|
for (int xVariable_index = 0; xVariable_index < optimizationVariablesNames.size();
|
|
xVariable_index++) {
|
|
const std::string xVariable_name = optimizationVariablesNames[xVariable_index];
|
|
const double xVariable_value = *json_optimizationResults.find(xVariable_name);
|
|
optimalXNameValuePairs[xVariable_index] = std::make_pair(xVariable_name,
|
|
xVariable_value);
|
|
}
|
|
|
|
const std::string fullPatternLabel = json_optimizationResults.at(
|
|
JsonKeys::FullPatternLabel);
|
|
if (!baseTriangleFullPattern.load(
|
|
std::filesystem::path(loadFromPath).append(fullPatternLabel + ".ply").string())) {
|
|
!baseTriangleFullPattern.load(std::filesystem::path(loadFromPath)
|
|
.append(loadFromPath.stem().string() + ".ply")
|
|
.string());
|
|
}
|
|
|
|
std::vector<double> baseTriangleVertices = json_optimizationResults.at(
|
|
JsonKeys::baseTriangle);
|
|
baseTriangle.P0(0) = CoordType(baseTriangleVertices[0],
|
|
baseTriangleVertices[1],
|
|
baseTriangleVertices[2]);
|
|
baseTriangle.P1(0) = CoordType(baseTriangleVertices[3],
|
|
baseTriangleVertices[4],
|
|
baseTriangleVertices[5]);
|
|
baseTriangle.P2(0) = CoordType(baseTriangleVertices[6],
|
|
baseTriangleVertices[7],
|
|
baseTriangleVertices[8]);
|
|
if (json_optimizationResults.contains(JsonKeys::fullPatternYoungsModulus)) {
|
|
fullPatternYoungsModulus = json_optimizationResults.at(
|
|
JsonKeys::fullPatternYoungsModulus);
|
|
} else {
|
|
fullPatternYoungsModulus = 1 * 1e9;
|
|
}
|
|
if (shouldLoadDebugFiles) {
|
|
const std::filesystem::path simulationJobsFolderPath(
|
|
std::filesystem::path(loadFromPath).append("SimulationJobs"));
|
|
|
|
const std::vector<std::filesystem::path> scenariosSortedByName = [&]() {
|
|
std::vector<std::filesystem::path> sortedByName;
|
|
for (auto &entry : std::filesystem::directory_iterator(simulationJobsFolderPath))
|
|
sortedByName.push_back(entry.path());
|
|
|
|
std::sort(sortedByName.begin(), sortedByName.end(), &Utilities::compareNat);
|
|
return sortedByName;
|
|
}();
|
|
for (const auto &simulationScenarioPath : scenariosSortedByName) {
|
|
if (!std::filesystem::is_directory(simulationScenarioPath)) {
|
|
continue;
|
|
}
|
|
// Load full job
|
|
const auto fullJobFilepath = Utilities::getFilepathWithExtension(
|
|
std::filesystem::path(simulationScenarioPath).append("Full"), ".json");
|
|
SimulationJob fullJob;
|
|
fullJob.load(fullJobFilepath.string());
|
|
fullJob.pMesh->setBeamMaterial(0.3, fullPatternYoungsModulus);
|
|
fullPatternSimulationJobs.push_back(std::make_shared<SimulationJob>(fullJob));
|
|
// Load reduced job
|
|
const auto reducedJobFilepath = Utilities::getFilepathWithExtension(
|
|
std::filesystem::path(simulationScenarioPath).append("Reduced"), ".json");
|
|
SimulationJob reducedJob;
|
|
reducedJob.load(reducedJobFilepath.string());
|
|
applyOptimizationResults_elements(*this, reducedJob.pMesh);
|
|
reducedPatternSimulationJobs.push_back(
|
|
std::make_shared<SimulationJob>(reducedJob));
|
|
}
|
|
}
|
|
settings.load(std::filesystem::path(loadFromPath).append(Settings::defaultFilename));
|
|
|
|
return true;
|
|
}
|
|
|
|
template<typename MeshType>
|
|
static void applyOptimizationResults_reducedModel_nonFanned(
|
|
const ReducedModelOptimization::Results &reducedPattern_optimizationResults,
|
|
const vcg::Triangle3<double> &patternBaseTriangle,
|
|
MeshType &reducedPattern)
|
|
{
|
|
std::unordered_map<std::string, double>
|
|
optimalXVariables(reducedPattern_optimizationResults.optimalXNameValuePairs.begin(),
|
|
reducedPattern_optimizationResults.optimalXNameValuePairs.end());
|
|
assert(
|
|
(optimalXVariables.contains("R") && optimalXVariables.contains("Theta"))
|
|
|| (optimalXVariables.contains("HexSize") && optimalXVariables.contains("HexAngle")));
|
|
if (optimalXVariables.contains("HexSize")) {
|
|
applyOptimizationResults_reducedModel_nonFanned(optimalXVariables.at("HexSize"),
|
|
optimalXVariables.at("HexAngle"),
|
|
patternBaseTriangle,
|
|
reducedPattern);
|
|
return;
|
|
}
|
|
applyOptimizationResults_reducedModel_nonFanned(optimalXVariables.at("R"),
|
|
optimalXVariables.at("Theta"),
|
|
patternBaseTriangle,
|
|
reducedPattern);
|
|
}
|
|
|
|
template<typename MeshType>
|
|
static void applyOptimizationResults_reducedModel_nonFanned(
|
|
const double &hexSize,
|
|
const double &hexAngle,
|
|
const vcg::Triangle3<double> &patternBaseTriangle,
|
|
MeshType &reducedPattern)
|
|
{
|
|
//Set optimal geometrical params of the reduced pattern
|
|
// CoordType center_barycentric(1, 0, 0);
|
|
// CoordType interfaceEdgeMiddle_barycentric(0, 0.5, 0.5);
|
|
// CoordType movableVertex_barycentric(
|
|
// (center_barycentric * (1 - hexSize) + interfaceEdgeMiddle_barycentric));
|
|
CoordType movableVertex_barycentric(1 - hexSize, hexSize / 2, hexSize / 2);
|
|
|
|
reducedPattern.vert[0].P() = patternBaseTriangle.cP(0) * movableVertex_barycentric[0]
|
|
+ patternBaseTriangle.cP(1) * movableVertex_barycentric[1]
|
|
+ patternBaseTriangle.cP(2) * movableVertex_barycentric[2];
|
|
if (hexAngle != 0) {
|
|
reducedPattern.vert[0].P() = vcg::RotationMatrix(CoordType{0, 0, 1},
|
|
vcg::math::ToRad(hexAngle))
|
|
* reducedPattern.vert[0].cP();
|
|
}
|
|
// for (int rotationCounter = 0;
|
|
// rotationCounter < ReducedModelOptimizer::fanSize; rotationCounter++) {
|
|
// pReducedPatternSimulationMesh->vert[2 * rotationCounter].P()
|
|
// = vcg::RotationMatrix(ReducedModelOptimizer::patternPlaneNormal,
|
|
// vcg::math::ToRad(60.0 * rotationCounter))
|
|
// * baseTriangleMovableVertexPosition;
|
|
// }
|
|
// reducedPattern.registerForDrawing();
|
|
// polyscope::show();
|
|
// CoordType p0 = reducedPattern.vert[0].P();
|
|
// CoordType p1 = reducedPattern.vert[1].P();
|
|
// int i = 0;
|
|
// i++;
|
|
}
|
|
|
|
static void applyOptimizationResults_elements(
|
|
const ReducedModelOptimization::Results &reducedPattern_optimizationResults,
|
|
const std::shared_ptr<SimulationMesh> &pReducedPattern_simulationMesh)
|
|
{
|
|
assert(CrossSectionType::name == RectangularBeamDimensions::name);
|
|
// Set reduced parameters fron the optimization results
|
|
std::unordered_map<std::string, double>
|
|
optimalXVariables(reducedPattern_optimizationResults.optimalXNameValuePairs.begin(),
|
|
reducedPattern_optimizationResults.optimalXNameValuePairs.end());
|
|
|
|
const std::string ALabel = "A";
|
|
if (optimalXVariables.contains(ALabel)) {
|
|
const double A = optimalXVariables.at(ALabel);
|
|
const double beamWidth = std::sqrt(A);
|
|
const double beamHeight = beamWidth;
|
|
CrossSectionType elementDimensions(beamWidth, beamHeight);
|
|
for (int ei = 0; ei < pReducedPattern_simulationMesh->EN(); ei++) {
|
|
Element &e = pReducedPattern_simulationMesh->elements[ei];
|
|
e.setDimensions(elementDimensions);
|
|
}
|
|
}
|
|
|
|
const double poissonsRatio = 0.3;
|
|
const std::string ymLabel = "E";
|
|
if (optimalXVariables.contains(ymLabel)) {
|
|
const double E = optimalXVariables.at(ymLabel);
|
|
const ElementMaterial elementMaterial(poissonsRatio, E);
|
|
for (int ei = 0; ei < pReducedPattern_simulationMesh->EN(); ei++) {
|
|
Element &e = pReducedPattern_simulationMesh->elements[ei];
|
|
e.setMaterial(elementMaterial);
|
|
}
|
|
}
|
|
|
|
const std::string JLabel = "J";
|
|
if (optimalXVariables.contains(JLabel)) {
|
|
const double J = optimalXVariables.at(JLabel);
|
|
for (int ei = 0; ei < pReducedPattern_simulationMesh->EN(); ei++) {
|
|
Element &e = pReducedPattern_simulationMesh->elements[ei];
|
|
e.dimensions.inertia.J = J;
|
|
}
|
|
}
|
|
|
|
const std::string I2Label = "I2";
|
|
if (optimalXVariables.contains(I2Label)) {
|
|
const double I2 = optimalXVariables.at(I2Label);
|
|
for (int ei = 0; ei < pReducedPattern_simulationMesh->EN(); ei++) {
|
|
Element &e = pReducedPattern_simulationMesh->elements[ei];
|
|
e.dimensions.inertia.I2 = I2;
|
|
}
|
|
}
|
|
|
|
const std::string I3Label = "I3";
|
|
if (optimalXVariables.contains(I3Label)) {
|
|
const double I3 = optimalXVariables.at(I3Label);
|
|
for (int ei = 0; ei < pReducedPattern_simulationMesh->EN(); ei++) {
|
|
Element &e = pReducedPattern_simulationMesh->elements[ei];
|
|
e.dimensions.inertia.I3 = I3;
|
|
}
|
|
}
|
|
pReducedPattern_simulationMesh->reset();
|
|
}
|
|
|
|
#if POLYSCOPE_DEFINED
|
|
void draw(const std::vector<int> &desiredSimulationScenariosIndices = std::vector<int>()) const
|
|
{
|
|
PolyscopeInterface::init();
|
|
DRMSimulationModel drmSimulator;
|
|
LinearSimulationModel linearSimulator;
|
|
assert(fullPatternSimulationJobs.size() == reducedPatternSimulationJobs.size());
|
|
fullPatternSimulationJobs[0]->pMesh->registerForDrawing(Colors::fullInitial);
|
|
reducedPatternSimulationJobs[0]->pMesh->registerForDrawing(Colors::reducedInitial, false);
|
|
|
|
const int numberOfSimulationJobs = fullPatternSimulationJobs.size();
|
|
const std::vector<int> scenariosToDraw = [&]() {
|
|
if (desiredSimulationScenariosIndices.empty()) {
|
|
std::vector<int> v(numberOfSimulationJobs);
|
|
std::iota(v.begin(), v.end(), 0); //draw all
|
|
return v;
|
|
} else {
|
|
return desiredSimulationScenariosIndices;
|
|
}
|
|
}();
|
|
|
|
for (const int &simulationJobIndex : scenariosToDraw) {
|
|
// Drawing of full pattern results
|
|
const std::shared_ptr<SimulationJob> &pFullPatternSimulationJob
|
|
= fullPatternSimulationJobs[simulationJobIndex];
|
|
pFullPatternSimulationJob->registerForDrawing(
|
|
fullPatternSimulationJobs[0]->pMesh->getLabel());
|
|
DRMSimulationModel::Settings drmSettings;
|
|
SimulationResults fullModelResults
|
|
= drmSimulator.executeSimulation(pFullPatternSimulationJob, drmSettings);
|
|
fullModelResults.registerForDrawing(Colors::fullDeformed, true);
|
|
// SimulationResults fullModelLinearResults =
|
|
// linearSimulator.executeSimulation(pFullPatternSimulationJob);
|
|
// fullModelLinearResults.setLabelPrefix("linear");
|
|
// fullModelLinearResults.registerForDrawing(Colors::fullDeformed,false);
|
|
// Drawing of reduced pattern results
|
|
const std::shared_ptr<SimulationJob> &pReducedPatternSimulationJob
|
|
= reducedPatternSimulationJobs[simulationJobIndex];
|
|
// pReducedPatternSimulationJob->pMesh->registerForDrawing();
|
|
// polyscope::show();
|
|
// SimulationResults reducedModelResults = drmSimulator.executeSimulation(
|
|
// pReducedPatternSimulationJob);
|
|
// reducedModelResults.registerForDrawing(Colors::reducedDeformed, false);
|
|
// SimulationResults reducedModelResults
|
|
// = drmSimulator.executeSimulation(pReducedPatternSimulationJob,
|
|
// DRMSimulationModel::Settings());
|
|
SimulationResults reducedModelResults = linearSimulator.executeSimulation(
|
|
pReducedPatternSimulationJob);
|
|
reducedModelResults.setLabelPrefix("linear");
|
|
reducedModelResults.registerForDrawing(Colors::reducedDeformed, true);
|
|
polyscope::options::programName = fullPatternSimulationJobs[0]->pMesh->getLabel();
|
|
// polyscope::view::resetCameraToHomeView();
|
|
polyscope::show();
|
|
// Save a screensh
|
|
const std::string screenshotFilename
|
|
= "/home/iason/Coding/Projects/Approximating shapes with flat "
|
|
"patterns/RodModelOptimizationForPatterns/Results/Images/"
|
|
+ fullPatternSimulationJobs[0]->pMesh->getLabel() + "_"
|
|
+ pFullPatternSimulationJob->getLabel();
|
|
polyscope::screenshot(screenshotFilename, false);
|
|
pFullPatternSimulationJob->unregister(fullPatternSimulationJobs[0]->pMesh->getLabel());
|
|
fullModelResults.unregister();
|
|
// reducedModelResults.unregister();
|
|
reducedModelResults.unregister();
|
|
// fullModelLinearResults.unregister();
|
|
// double error = computeError(reducedModelLinearResults.displacements,
|
|
// fullModelResults.displacements);
|
|
// std::cout << "Error of simulation scenario "
|
|
// << simulationScenarioStrings[simulationScenarioIndex] << " is " << error
|
|
// << std::endl;
|
|
}
|
|
}
|
|
#endif // POLYSCOPE_DEFINED
|
|
void saveMeshFiles() const {
|
|
const int numberOfSimulationJobs = fullPatternSimulationJobs.size();
|
|
assert(numberOfSimulationJobs != 0 &&
|
|
fullPatternSimulationJobs.size() ==
|
|
reducedPatternSimulationJobs.size());
|
|
|
|
fullPatternSimulationJobs[0]->pMesh->save(
|
|
"undeformed " + fullPatternSimulationJobs[0]->pMesh->getLabel() + ".ply");
|
|
reducedPatternSimulationJobs[0]->pMesh->save(
|
|
"undeformed " + reducedPatternSimulationJobs[0]->pMesh->getLabel() + ".ply");
|
|
DRMSimulationModel simulator_drm;
|
|
LinearSimulationModel simulator_linear;
|
|
for (int simulationJobIndex = 0; simulationJobIndex < numberOfSimulationJobs;
|
|
simulationJobIndex++) {
|
|
// Drawing of full pattern results
|
|
const std::shared_ptr<SimulationJob> &pFullPatternSimulationJob
|
|
= fullPatternSimulationJobs[simulationJobIndex];
|
|
DRMSimulationModel::Settings drmSettings;
|
|
SimulationResults fullModelResults
|
|
= simulator_drm.executeSimulation(pFullPatternSimulationJob, drmSettings);
|
|
fullModelResults.saveDeformedModel();
|
|
|
|
// Drawing of reduced pattern results
|
|
const std::shared_ptr<SimulationJob> &pReducedPatternSimulationJob
|
|
= reducedPatternSimulationJobs[simulationJobIndex];
|
|
SimulationResults reducedModelResults = simulator_linear.executeSimulation(
|
|
pReducedPatternSimulationJob);
|
|
reducedModelResults.saveDeformedModel();
|
|
}
|
|
}
|
|
|
|
void writeHeaderTo(csvFile &os) const
|
|
{
|
|
if (exportSettings.exportRawObjectiveValue) {
|
|
os << "Total raw Obj value";
|
|
}
|
|
os << "Total Obj value";
|
|
if (exportSettings.exportIterationOfMinima) {
|
|
os << "Iteration of minima";
|
|
}
|
|
for (int simulationScenarioIndex = 0;
|
|
simulationScenarioIndex < fullPatternSimulationJobs.size();
|
|
simulationScenarioIndex++) {
|
|
const std::string simulationScenarioName
|
|
= fullPatternSimulationJobs[simulationScenarioIndex]->getLabel();
|
|
os << "Obj value Trans " + simulationScenarioName;
|
|
os << "Obj value Rot " + simulationScenarioName;
|
|
os << "Obj value Total " + simulationScenarioName;
|
|
}
|
|
|
|
if (exportSettings.exportPE) {
|
|
for (int simulationScenarioIndex = 0;
|
|
simulationScenarioIndex < fullPatternSimulationJobs.size();
|
|
simulationScenarioIndex++) {
|
|
const std::string simulationScenarioName
|
|
= fullPatternSimulationJobs[simulationScenarioIndex]->getLabel();
|
|
os << "PE " + simulationScenarioName;
|
|
}
|
|
}
|
|
for (const auto &nameValuePair : optimalXNameValuePairs) {
|
|
os << nameValuePair.first;
|
|
}
|
|
os << "Time(s)";
|
|
// os << "#Crashes";
|
|
// os << "notes";
|
|
}
|
|
|
|
void writeHeaderTo(std::vector<csvFile *> &vectorOfPointersToOutputStreams) const
|
|
{
|
|
for (csvFile *outputStream : vectorOfPointersToOutputStreams) {
|
|
writeHeaderTo(*outputStream);
|
|
}
|
|
}
|
|
|
|
void writeResultsTo(csvFile &os) const
|
|
{
|
|
if (exportSettings.exportRawObjectiveValue) {
|
|
os << objectiveValue.totalRaw;
|
|
}
|
|
os << objectiveValue.total;
|
|
if (exportSettings.exportIterationOfMinima && !objectiveValueHistory_iteration.empty()) {
|
|
os << objectiveValueHistory_iteration.back();
|
|
}
|
|
for (int simulationScenarioIndex = 0;
|
|
simulationScenarioIndex < fullPatternSimulationJobs.size();
|
|
simulationScenarioIndex++) {
|
|
os << objectiveValue.perSimulationScenario_translational[simulationScenarioIndex];
|
|
os << objectiveValue.perSimulationScenario_rotational[simulationScenarioIndex];
|
|
os << objectiveValue.perSimulationScenario_total[simulationScenarioIndex];
|
|
}
|
|
if (exportSettings.exportPE) {
|
|
for (int simulationScenarioIndex = 0;
|
|
simulationScenarioIndex < fullPatternSimulationJobs.size();
|
|
simulationScenarioIndex++) {
|
|
os << perScenario_fullPatternPotentialEnergy[simulationScenarioIndex];
|
|
}
|
|
}
|
|
for (const auto &optimalXNameValuePair : optimalXNameValuePairs) {
|
|
os << optimalXNameValuePair.second;
|
|
}
|
|
|
|
os << time;
|
|
// if (numberOfSimulationCrashes == 0) {
|
|
// os << "-";
|
|
// } else {
|
|
// os << numberOfSimulationCrashes;
|
|
// }
|
|
// os << notes;
|
|
}
|
|
|
|
void writeResultsTo(std::vector<csvFile *> &vectorOfPointersToOutputStreams) const
|
|
{
|
|
for (csvFile *&outputStream : vectorOfPointersToOutputStreams) {
|
|
writeResultsTo(*outputStream);
|
|
}
|
|
}
|
|
|
|
void writeMinimaInfoTo(csvFile &outputCsv)
|
|
{
|
|
outputCsv << "Iteration";
|
|
outputCsv << "Objective value";
|
|
for (int objectiveValueIndex = 0; objectiveValueIndex < objectiveValueHistory.size();
|
|
objectiveValueIndex++) {
|
|
outputCsv.endrow();
|
|
outputCsv << objectiveValueHistory_iteration[objectiveValueIndex];
|
|
outputCsv << objectiveValueHistory[objectiveValueIndex];
|
|
}
|
|
}
|
|
};
|
|
enum SimulationModelTag { DRM, Linear };
|
|
|
|
inline bool Settings::ObjectiveWeights::operator==(const ObjectiveWeights &other) const
|
|
{
|
|
return this->translational == other.translational && this->rotational == other.rotational;
|
|
}
|
|
} // namespace ReducedModelOptimization
|
|
#endif // REDUCEDMODELOPTIMIZER_STRUCTS_HPP
|