From Simulation to AI: Neural Networks for Chemical Kinetics in Turbulent Hydrogen Swirling Flames
Project Overview
This project investigates whether a physically constrained neural network can replace expensive chemistry calculations in simulations of hydrogen flames. Using data from a high-fidelity swirling-flow simulation, the network learns to predict how the chemical composition changes while conserving the elements involved. The aim is to make reacting-flow simulations for applications such as gas turbines faster and more practical.
What were the key results of your research project?
- A complete workflow was developed to turn simulation data into neural-network training examples.
- The network reproduced short-term chemical changes accurately, including in an unseen part of the flow domain.
- Physical constraints helped keep the predictions chemically valid.
- Errors accumulated when the network repeatedly predicted further into the future.
- The results show potential for faster chemistry calculations, but further development is needed before integration into a full flow solver.
A presentation of this research will be shared here and on our YouTube site when available.
How do you feel you have benefitted from completing this internship and has it made you consider future career paths?
I have had to learn a whole new skillset to complete this project. I felt out of my depth at many times so I have definitely learnt a lot.
I would feel more confident attempting challenges in this field now, such as taking it up as a career.
Download slides of this presentation
Internships 2026 - Adam Wilson