Development of a Hybrid Digital Twin for Electric Engines in Aircraft Propulsion Systems

Abstract

This study presents the development of a hybrid digital twin (DT) for an aircraft propulsion electric engine, employing physics-based modeling, machine learning (ML), and experimental methodologies. In doing so, four different DT models are first developed: a physics-based digital twin, a purely MLdriven digital twin, and two hybrid digital twins. The hybrid approaches fuse physics-based predictions and sensor data through residual learning and Kalman filtering to improve ML predictive accuracy. The performances of the models are then evaluated in predicting experimental data obtained from a 200 kW hybrid-electric ground-testing platform at a voltage of 800 VDC across various speed and torque settings, emphasizing transient operating conditions. Results demonstrate that the hybrid DT, which integrates both Kalman filtering and ML, achieves the highest predictive accuracy-effectively capturing transient and steady-state behaviors while minimizing noise effects. This research highlights the advantages of integrating domain knowledge with data-driven methodologies to support onboard monitoring and management of electric aero-engines for anomaly detection, predictive maintenance, and performance optimization.