As carbon-free clean combustion has become crucial to help achieve the increasingly stringent climate targets, alternative fuels are gaining more interests in the technology sector of internal combustion engines. Hydrogen and ammonia are considered as promising zero-carbon fuels in addressing the decarbonisation challenges of the global mobility turnaround. The present study introduces the method of developing neural network models to predict the laminar flame speed (LFS) of ammonia and hydrogen mixtures. Representing the LFS in a neural network increases fidelity in predictive combustion simulation without a big penalty in calculation time. The original data-generation process, based on detailed chemistry calculation in a 1-D simulation tool, is also discussed. This illustrates a wide range of engine-like operating conditions and how the quality and quantity of the data sets influence the neural network model fidelity. A trendwise study on LFS varying as hydrogen fractions in the fuel (ranging from 0 vol% to 100 vol%) was carried out using the introduced methodology. Results show increasing laminar flame speeds as hydrogen fraction increases. The developed artificial neural network (ANN) model achieves acceptable quality with a coefficient of determination > 0.99.
Simulation Study on Laminar Flame Speed of Ammonia-Hydrogen Blends Using Artificial Neural Networks
Proceedings
Internationaler Motorenkongress ; 2024 ; baden-baden, Deutschland April 19, 2024 - April 29, 2024
2024-07-01
15 pages
Article/Chapter (Book)
Electronic Resource
German