Data-driven approaches for enhanced on-board fault diagnosis and emission monitoring to support Euro 7 standard implementation

The European Commission has introduced the Euro 7 standard to reduce pollutant emissions in the transport sector. This regulation emphasizes the role of On-Board Monitoring (OBM) in ensuring low emissions throughout a vehicle’s lifespan, considering system aging and potential faults. The research presented in this dissertation explores data-driven methods for detecting emission-related engine faults, supporting On-Board Diagnostics (OBD), and enabling real-time emission monitoring under various operating conditions, key challenges posed by Euro 7 OBM requirements. To achieve this, common emission-related engine faults were simulated using a validated 0-D model of a Diesel Plug-in Hybrid Electric Vehicle (PHEV). The study assessed their impact on NOx emissions and identified useful on-board signals for OBM-oriented models. Various classifiers were evaluated based on accuracy, training time, and prediction speed, with Tree, Ensemble, and Neural Networks emerging as the best-performing ones. These models were further optimized using Bayesian techniques to enhance classification accuracy. The same methodology was applied to develop OBM-oriented regression models for NOx emission estimation. Using the same dataset, regression models were trained to correct the reference ECU model when non-nominal conditions are considered. These models leverage on-board signals to refine NOx predictions in presence of engine faults, significantly reducing estimation errors. Tests across different driving cycles and fault conditions confirmed high accuracy, good interpolation, and robust generalization. Neural Networks outperformed other models, offering the best balance between accuracy, generalization, and complexity. For real-world application, the models were deployed on a Raspberry Pi and tested in a Hardware-in-the-Loop (HiL) environment, demonstrating feasibility for low-cost on-board integration via CAN-bus communication. Further validation was conducted using real test bench data, confirming the models' effectiveness and potential to help manufacturers comply with Euro 7 regulations. This research was conducted at the Green Mobility Research Lab, a collaboration between the University of Bologna and FEV Italia s.r.l.