A Transferable Predictive Maintenance Method for Hybrid Powertrains: Evaluation on the Internal Combustion Engine Subsystem

Hybrid powertrains, which tightly couple internal combustion engines with electric motors and power electronics, create intertwined mechanical, thermal, and electrical failure modes that demand predictive maintenance approaches aware of cross-domain dynamics. In this study, we developed and validated a data-driven framework on the ICE subsystem using high-resolution sensor and operating data; CatBoost achieved the best predictive performance (F۱ = ۰.۷۵۸۳), and refined class labeling improved fault separability by approximately ۹%. While the experimental focus was the combustion unit, the methodology is inherently extensible to modern electrified powertrains, enabling joint monitoring of ICE, e-machine, inverters, and battery (SOC/SOH) and supporting integrated diagnostics, adaptive energy management, and safer control strategies. These findings position the approach as a practical, transferable foundation for predictive maintenance in next-generation hybrid and electrified vehicles.

کلیدواژه ها:

Predictive Maintenance ، Hybrid Powertrain ، Electric Vehicle ، Diagnostics ، Machine Learning ، Fault Prediction ، Statistical Analysis ، Sensor Data Analysis ، Class Separability ، Energy Management