Abstract Future (hybrid) electric aircraft have the potential to highly impact the mobility sector in rural or congested areas. With the ability of vertical take off and landing of many aircraft configurations, the reliability and integrity of the motors play a crucial role within the system. Therefore, we propose a real-time-capable fault diagnosis system including the underlying IT-architecture, a data-based fault diagnosis algorithm and a final real-time-assessment on real-time hardware. The main element of the IT-architecture is an Avionics Full Duplex Switched Ethernet (AFDX) network for component connectivity with the ability of high data rates at an acceptable jitter rate. An exemplary data-based fault diagnosis algorithm is developed using feature engineering and deep learning methods to detect faults. The algorithm is developed on the bearing fault dataset provided by the University of Paderborn since bearing faults are responsible for 40%–70% of electric motor failures. The final algorithm is trained on bearings with healthy state and bearings with artificially induced faults, and is tested on bearings with real faults created by accelerated lifetime tests. In this case, the developed algorithm reaches an accuracy of 97.8%. Afterwards, preprocessing steps and the trained algorithm are integrated into a Simulink pipeline and is assessed on a Speedgoat real-time-capable hardware. The results show that the developed architecture and algorithm are well suited for detecting faults in electric aircraft propulsion systems in real-time.
Coors et al. (Thu,) studied this question.