The CFDNNetAdapt is a hybrid CFD-DNN optimization algorithm for CFD-based shape optimization. The algorithm combines CFD with multi-objective multi-parameter optimization performed via MOEA with directly incorporated DNNs. The DNN architecture is searched for automatically and accelerate the mid-to-late MOEA iterations.
MOEA -- D. Hadka, Platypus, A Free and Open Source Python Library for Multiobjective Optimization, 2020. URL: https://github.com/Project-Platypus/Platypus
DNNs -- D. Atabay, Institute for Energy Economy and Application Technology,665 Technische Universität München, pyrenn: A recurrent neural network tool-box for python and matlab, 2018. URL: https://pyrenn.readthedocs.io/en/latest/.
article in preparation
Prepared for python3 (https://www.python.org/downloads/release/python-31010/) and OpenFOAMv8 (https://openfoam.org/version/8/).
used python3 packages -- os, io, math, sys, shutil, numpy, scipy, re, copy, csv, dill, multiprocessing, glob, subprocess, operator, random, datetime
python3 packages used by thirdParty codes -- six, pandas, functools, traceback, mpi4py, unittest, pickle, abc, time, logging, collections, sets
cd ./example/convDifShapeOptim && python3 Allrun.py
cd ./example/convDifShapeOptim && python3 testRun.py
CFDNNetAdapt is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. See http://www.gnu.org/licenses/, for a description of the GNU General Public License terms under which you can copy the files.