Integrated Machine Learning Framework for Real-Time Hybrid Engine Optimization and Performance Prediction

Abstract

This paper presents a novel machine learning framework for optimizing hybrid engine performance across multiple competing objectives. We develop a comprehensive computational approach that combines physics-based modeling with advanced machine learning techniques to simultaneously optimize fuel efficiency, power output, and emissions in hybrid powertrains. The study employs a synthetic dataset representing 1,000 operating points with eight input parameters and three target variables. Our comparative analysis of Random Forest, Gradient Boosting, and Neural Network models reveals distinct performance patterns across the three target variables. Notably, all models achieved excellent performance in predicting fuel efficiency (R² > 0.98), with Gradient Boosting demonstrating superior overall performance across all metrics. For engine power prediction, Gradient Boosting again outperformed other models (R² ? 0.95), while Neural Networks showed significantly higher error rates. In emissions prediction, all models demonstrated lower accuracy (R² between 0.7-0.8), with Gradient Boosting maintaining a slight edge. Feature importance analysis identifies the most significant parameters affecting hybrid system performance, enabling us to establish a Pareto-optimal frontier of operating configurations through multi-objective optimization. Our approach visualizes the complex parameter space through interactive 3D representations, facilitating deeper understanding of the trade-offs between efficiency, power, and environmental impact. The proposed framework has potential applications in real-time hybrid engine control systems and can reduce development time in powertrain design, providing a foundation for similar multi-objective optimization problems in automotive engineering.

Introduction

The automotive industry is undergoing a significant transformation, driven by the integration of machine learning (ML) technologies aimed at optimizing hybrid powertrains. These advancements address the complexities of balancing fuel efficiency, emissions, and performance in hybrid engines.

Key Insights:

1. Machine Learning in Hybrid Engine Optimization:
   ML techniques, such as Random Forest, Gradient Boosting, and Neural Networks, are employed to model and predict the intricate interactions within hybrid engine systems. These models process data from various sensors to identify optimal operating conditions, enhancing predictive capabilities and system performance.

2. Challenges in Emission Prediction:
   Predicting emissions remains a complex task due to the non-linear relationships between engine parameters and pollutant outputs. While ML models show promise, achieving high accuracy in emissions prediction necessitates continuous refinement of data inputs and model architectures.

3. Integration with Model Predictive Control (MPC):
   Combining ML with MPC allows for real-time optimization of engine parameters, leading to improved fuel efficiency and reduced emissions. This hybrid approach enables adaptive control strategies that respond dynamically to changing driving conditions.

4. Visualization Tools for Decision Support:
   Advanced visualization techniques, including 3D scatter plots and interactive dashboards, are developed to represent the complex parameter spaces of hybrid engines. These tools assist engineers in understanding trade-offs and making informed decisions during the design and calibration processes.

5. Future Directions:
   Ongoing research focuses on enhancing the interpretability of ML models, integrating real-time data streams, and developing more robust frameworks that can accommodate the evolving demands of hybrid engine systems.

Conclusion

This study presents a comprehensive machine learning framework for optimizing hybrid engine performance across the competing objectives of fuel efficiency, power output, and emissions reduction. Our results demonstrate that machine learning techniques can effectively capture and predict complex relationships in hybrid powertrain systems with varying degrees of accuracy across different performance metrics. The comparative analysis of Random Forest, Gradient Boosting, and Neural Network models reveals that Gradient Boosting consistently delivers superior performance across all target variables, with particularly strong results for fuel efficiency (R² > 0.98) and engine power (R² ? 0.95). The relative difficulty in accurately predicting emissions (R² < 0.80) highlights an important area for future research and suggests the need for more sophisticated modeling approaches or additional input parameters to fully characterize emissions behavior. Our visualization framework provides engineers with powerful tools to understand the multidimensional parameter space and identify optimal operating configurations. The ability to visualize three-dimensional relationships between engine RPM, throttle position, and electric current, and their combined impact on efficiency and power, offers unprecedented insight into hybrid system dynamics. This study establishes a foundation for data-driven optimization in hybrid powertrain development, with potential applications in real-time control systems, calibration processes, and design optimization. Future work should focus on expanding the feature set to improve emissions predictions, validating these approaches with real-world engine data, and developing adaptive models that can account for component aging and environmental variations. The integration of physics-based knowledge with machine learning techniques demonstrated in this work represents a promising direction for accelerating the development and optimization of increasingly complex hybrid propulsion systems.

References

[1] Soori, M., Jough, F.K.G., Dastres, R. and Arezoo, B., 2025. Additive Manufacturing Modification by Artificial Intelligence, Machine Learning, and Deep Learning: A Review. Additive Manufacturing Frontiers, p.200198. [2] Chuenmee, N., Phothi, N., Chamniprasart, K., Khaengkarn, S. and Srisertpol, J., 2025. Machine learning for predicting resistance spot weld quality in automotive manufacturing. Results in Engineering, 25, p.103570. [3] Kazmi, K.H., Chandra, M., Rajak, S., Sharma, S.K., Mandal, A. and Das, A.K., 2025. Implementing machine learning in robotic wire arc additive manufacturing for minimizing surface roughness. International Journal of Computer Integrated Manufacturing, 38(2), pp.255-270. [4] Mishra, A., Jatti, V.S., Sefene, E.M. and Paliwal, S., 2023. Explainable artificial intelligence (XAI) and supervised machine learning-based algorithms for prediction of surface roughness of additively manufactured polylactic acid (PLA) specimens. Applied Mechanics, 4(2), pp.668-698. [5] Batu, T., Lemu, H.G. and Shimels, H., 2023. Application of artificial intelligence for surface roughness prediction of additively manufactured components. Materials, 16(18), p.6266. [6] Sawant, D.A., Jatti, V.S., Mishra, A., Sefene, E.M. and Jatti, A.V., 2023. Surface roughness and surface crack length prediction using supervised machine learning–based approach of electrical discharge machining of deep cryogenically treated NiTi, NiCu, and BeCu alloys. The International Journal of Advanced Manufacturing Technology, 128(11-12), pp.5595-5612. [7] Du, Y., Mukherjee, T. and DebRoy, T., 2021. Physics-informed machine learning and mechanistic modeling of additive manufacturing to reduce defects. Applied Materials Today, 24, p.101123. [8] Yan, W., Lin, S., Kafka, O.L., Lian, Y., Yu, C., Liu, Z., Yan, J., Wolff, S., Wu, H., Ndip-Agbor, E. and Mozaffar, M., 2018. Data-driven multi-scale multi-physics models to derive process–structure–property relationships for additive manufacturing. Computational Mechanics, 61, pp.521-541. [9] Farrag, A., Yang, Y., Cao, N., Won, D. and Jin, Y., 2025. Physics-informed machine learning for metal additive manufacturing. Progress in Additive Manufacturing, 10(1), pp.171-185. [10] Balluchi, A., Benvenuti, L., Di Benedetto, M.D., Pinello, C. and Sangiovanni-Vincentelli, A.L., 2002. Automotive engine control and hybrid systems: Challenges and opportunities. Proceedings of the IEEE, 88(7), pp.888-912. [11] Kawamoto, N., Naiki, K., Kawai, T., Shikida, T. and Tomatsuri, M., 2009. Development of new 1.8-liter engine for hybrid vehicles. [12] Shahpouri, S., Norouzi, A., Hayduk, C., Rezaei, R., Shahbakhti, M. and Koch, C.R., 2021. Hybrid machine learning approaches and a systematic model selection process for predicting soot emissions in compression ignition engines. Energies, 14(23), p.7865. [13] Mohammad, A., Rezaei, R., Hayduk, C., Delebinski, T.O., Shahpouri, S. and Shahbakhti, M., 2021. Hybrid physical and machine learning-oriented modeling approach to predict emissions in a diesel compression ignition engine (No. 2021-01-0496). SAE Technical Paper. [14] Xu, M., Wang, J., Liu, J., Li, M., Geng, J., Wu, Y. and Song, Z., 2020. An improved hybrid modeling method based on extreme learning machine for gas turbine engine. Aerospace Science and Technology, 107, p.106333. [15] Shahpouri, S., Gordon, D., Hayduk, C., Rezaei, R., Koch, C.R. and Shahbakhti, M., 2023. Hybrid emission and combustion modeling of hydrogen fueled engines. International Journal of Hydrogen Energy, 48(62), pp.24037-24053.

Copyright

Copyright © 2025 Manu Yadav, Dr. Devesh Kumar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.