Machine Learning Scripts – Master Thesis Project

This repository contains Python notebooks and data files developed during my Master’s thesis at TU Delft, focused on Machine Learning applications in Numerical Combustion.

The main objective of this work is the development and analysis of Artificial Neural Network (ANN) models as a computationally efficient alternative to conventional lookup tables used in numerical simulations of turbulent combustion, with particular focus on hydrogen combustion.

Repository Contents

Data

• Information_ANN_Architectures_4D_hydrogen.csv
  Dataset containing information on ANN architectures used for 4D hydrogen combustion cases. This file is updated as new architectures and results are generated.

Jupyter Notebooks

• function_ANNs.ipynb
  Functions developed for training and evaluating Artificial Neural Networks, including data preprocessing and model handling.

• Steady_unsteady_flamelets.ipynb
  Generation and analysis of steady and unsteady flamelet-based lookup tables, including updated 4D datasets.

• Plots_6D.ipynb
  Functions and scripts for post-processing and visualisation of results obtained from 6D ANN models.

• tables to csv file.ipynb
  Jupyter Notebook used for converting lookup tables from .txt format to .csv format for easier processing and analysis.

Documentation

• README.md
  Overview and description of the repository contents.

• LICENSE
  GNU General Public License v3.0 (GPL-3.0).

Background and Reference

The theoretical background of the lookup tables and combustion modelling approaches used in this repository are detailed in my Master’s thesis:

Cristopher Morales,
Using Artificial Intelligence for turbulent combustion modelling: Simplifying the conventional lookup tables,
TU Delft, 2021.
Available at:
https://repository.tudelft.nl/islandora/object/uuid:0e4db0e6-1852-4346-ad25-11f3eed9d0bd?collection=education

Citation

If you find this repository or the associated thesis useful for your research, I would greatly appreciate if you cite the thesis in your work.

Notes

• The notebooks are intended for research and educational purposes.
• Some scripts may require adaptation depending on the dataset size, ANN architecture, or combustion configuration.
• Contributions, issues, and suggestions are welcome.