City Research Online

Abstract

Fuel flexibility ensures reliable operation and improves the implementation of a dual-fuel strategy alongside the Diesel-only model in marine powertrains. This versatile approach, however, imposes limitations related to the complexity of the injection system, underpinning the necessity of comprehending the relationship between design and performance to facilitate the injector optimization process. The present study introduces a data-driven predictor utilizing a deep neural network to predict spray tip penetration and cone angle in marine injectors. This neural network is trained using experimental data from serial and prototype industrial designs, incorporating operating conditions and a wide range of geometrical parameters as primary input features. Issues, as for example overfitting of the training dataset, were mitigated via regularization, enhancing generalization. The deep neural network accurately predicts spray characteristics across short, medium, and long penetration ranges, achieving 95% accuracy for unseen data. Furthermore, a feature importance analysis indicates that the injection pressure, number of spray holes, outlet spray hole diameter, and sac hole volume are the primary parameters influencing spray behavior. This neural network provides a computationally efficient alternative to conventional approaches, such as time-consuming Computational Fluid Dynamics simulations or test measurements. The model is tailored to support the marine injector design workflow, allowing the fast exploration of design space in the early design phase at operation conditions relevant for fuel flexibility, and contributing to accelerate the injector development process.

Publication Type:	Article
Additional Information:	© 2025 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Publisher Keywords:	Machine-learning, Large diesel injectors, Spray penetration, Spray cone angle
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science T Technology > TJ Mechanical engineering and machinery
Departments:	School of Science & Technology School of Science & Technology > Department of Engineering
SWORD Depositor:	Symplectic Administrator

Export

Downloads