Real-Time Fault Detection and Classification in Engine Components using Machine Learning

The reliability of automotive engines is critical for vehicle performance and safety. This research presents a machine learning-based approach for real-time fault detection and classification in four key engine components: oxygen sensors, spark plugs, turbochargers, and radiators. We employ XGBoost regression for fault prediction and classification models to categorize component failures. A digital twin model of the engine cooling system was developed to enable real-time visualization, monitoring, and performance prediction of critical engine components using live sensor input data. Developing a Remaining Useful Life (RUL) prediction model for oxygen sensors is identified as a promising direction for future work. The dataset is synthesized with threshold values according to real-world sensor readings, ensuring high accuracy and interpretability using explainable AI techniques. Experimental results demonstrate the effectiveness of the proposed methodology in fault identification and performance prediction, improving predictive maintenance strategies.