Machine Learning-Assisted performance prediction and analysis of Ammonia-Diesel Dual-Fuel engines

This study predicted and analyzed the Brake Thermal Efficiency (BTE) and emission characteristics (NH3, NOx, and N2O) of a 12.5L marine low-pressure ammonia-diesel dual-fuel engine under various operating conditions. To predict engine efficiency and emission performance, a comprehensive dataset comprising 461 experimental conditions was utilized to train and evaluate multiple machine learning (ML) models. The comparative analysis revealed that differences in predictive accuracy among the models were primarily governed by the dataset size, inherent biases within the feature space, and the distinct mathematical formulations of each algorithm. Among the evaluated models, XGBoost demonstrated the highest predictive performance (R2>0.91) and was identified as the optimal model. Subsequently, a feature importance analysis was conducted to provide a physical interpretation of the experimental results through the ML framework. The proposed ML approach effectively deciphered the complex combustion phenomena, revealing that air excess ratio (λ) was the primary factor dictating the trade-off between NOx and NH3 emissions. This methodology significantly enhances the understanding of inter-variable correlations and facilitates the identification of the root causes governing engine performance and emissions.

Authors

Institutions

Publication Details

Journal
Fuel
Published
2026-11-16
DOI
https://doi.org/10.1016/j.fuel.2026.142137
Primary Topic
Advanced Combustion Engine Technologies
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Machine Learning-Assisted performance prediction and analysis of Ammonia-Diesel Dual-Fuel engines

Kyusung Kim, Kyunghun Kim, Chansoo Park, Cheolwoong Park et al.
2 links
Fuel
Advanced Combustion Engine Technologies
article

Machine Learning-Assisted performance prediction and analysis of Ammonia-Diesel Dual-Fuel engines

Kyusung Kim, Kyunghun Kim, Chansoo Park, Cheolwoong Park, Hyeongjun Jang, Kyoungdoug Min, Minki Kim
article en

Abstract

This study predicted and analyzed the Brake Thermal Efficiency (BTE) and emission characteristics (NH3, NOx, and N2O) of a 12.5L marine low-pressure ammonia-diesel dual-fuel engine under various operating conditions. To predict engine efficiency and emission performance, a comprehensive dataset comprising 461 experimental conditions was utilized to train and evaluate multiple machine learning (ML) models. The comparative analysis revealed that differences in predictive accuracy among the models were primarily governed by the dataset size, inherent biases within the feature space, and the distinct mathematical formulations of each algorithm. Among the evaluated models, XGBoost demonstrated the highest predictive performance (R2>0.91) and was identified as the optimal model. Subsequently, a feature importance analysis was conducted to provide a physical interpretation of the experimental results through the ML framework. The proposed ML approach effectively deciphered the complex combustion phenomena, revealing that air excess ratio (λ) was the primary factor dictating the trade-off between NOx and NH3 emissions. This methodology significantly enhances the understanding of inter-variable correlations and facilitates the identification of the root causes governing engine performance and emissions.

FuelVol. 431
Chungnam National University (KR), Korea Institute of Machinery & Materials (KR), Jeonbuk National University (KR)
Korea Institute of Machinery and Materials
Affordable and clean energy, Climate action
Advanced Combustion Engine Technologies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon