Artificial Neural Networks-Based Ignition Delay Time Prediction for Natural Gas Blends

ABSTRACT

Energy production from natural gas has considerably increased worldwide. The composition of natural gas is diverse, and each blend has a different reactivity. Understanding the reactivities of these fuels enables the designer to develop fuel-source independent combustors. Also, future engines such as scramjet and homogeneous charge compression ignition (HCCI) will rely heavily on the reactivity of the fuel. Ignition delay time (IDT) is a direct measure of a fuel’s reactivity. In the current study, an artificial neural network (ANN) based model is developed to predict the IDT of different natural gas blends. The model has 13 inputs and three hidden layers and is trained using a back-propagation approach. The developed model is superior compared to a multiple linear regression approach and is validated with shock tube experiments. Furthermore, the model is used to predict the IDT of six different liquified natural gas blends (LNG), and the predicted results match the experimental data accurately. Additionally, the IDTs of four different commercial natural gas blends are predicted using the ANN model, showcasing the application of the tool in a real-world scenario.

KEYWORDS:

• Natural gas
• ignition delay time
• machine learning
• artificial neural networks

Acknowledgements

We gratefully acknowledge the research funding from the project BAS/1/1337-01-01. We also thank Prof. Henry Curran for useful discussions.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/00102202.2023.2239467

Notes

¹ The overall composition does not add up to 100% due to the presence of impurities.

Additional information

Funding

The work was supported by the King Abdullah University of Science and Technology [BAS/1/1337-01-01].