Emissions-Based Predictive Maintenance Framework for Hybrid Electric Vehicles Using Laboratory-Simulated Driving Conditions

This study presents a predictive maintenance framework for hybrid electric vehicles (HEVs) based on emissions behaviour under laboratory-simulated driving conditions. Vehicle speed, road gradient, and ambient temperature were selected as the principal input variables affecting emission levels. Using simulated datasets, three machine learning model, specifically Linear Regression, Multilayer Perceptron (MLP), as well as Random Forest, were trained and evaluated. Within that set, the Random Forest model demonstrated the best performance, achieving an R2 score of 0.79, Mean Absolute Error (MAE) of 12.57 g/km, and root mean square error (RMSE) of 15.4 g/km, significantly outperforming both Linear Regression and MLP. A MATLAB-based graphical interface was developed to allow real-time classification of emission severity using defined thresholds (Normal ≤ 150 g/km, Warning ≤ 220 g/km, Critical > 220 g/km) and to provide automatic maintenance recommendations derived from the predicted emissions. Scenario-based validation confirmed the system’s ability to detect emission anomalies, which might function as early indicators of mechanical degradation when interpreted relative to operating conditions. The proposed framework, developed using laboratory-simulated datasets, provides a practical, interpretable, and accurate solution for emissions-based predictive maintenance. Although the results demonstrate feasibility, the framework should be further confirmed with real-world on-road data prior to large-scale use.