Electric Vertical Take-Off and Landing (eVTOL) aircraft are poised to transform urban airspace, enabling both commercial deliveries and passenger transportation. Ensuring the safety of this future airspace necessitates highly precise health management systems that can actively predict and prevent potential failures in these vehicles. Such proactive measures are crucial not only for maintaining high safety standards but also for optimizing maintenance and enabling autonomous decision-making in Urban Air Mobility (UAM) systems.
Real-time understanding of an aircraft's health condition is key to moving from fixed maintenance schedules to a condition-based predictive maintenance model. Accurate health prediction relies on both the current health state and an understanding of future usage patterns.
Traditionally, complex system health prediction relied on either model-based or data-driven approaches. Hybrid modeling, combining the strengths of both, is gaining traction. This approach leverages existing knowledge of the system's physics-based principles with the data-driven learning capabilities of machine learning. Particularly suited to the challenges of complex, evolving electric aircraft propulsion systems, Hybrid Physics-Informed Neural Networks (H-PINNs) offer the potential for accurate and adaptable health prediction models. This hybrid approach has been successfully applied to predicting the health of electric powertrains in Unmanned Aerial Vehicles (UAVs), using deep learning to learn the uncertain, degrading parameters in the physics-based model.
In conclusion, this hybrid framework represents a significant advancement in monitoring and predicting the health of complex systems. This has far-reaching implications for improving safety, optimizing maintenance, and enhancing the reliability and efficiency of electric aircraft.
