Project Details
Projekt Print View

Dynamic data-driven assessment of technical mission risk for unmanned aircraft systems

Subject Area Traffic and Transport Systems, Intelligent and Automated Traffic
Term from 2020 to 2025
Project identifier Deutsche Forschungsgemeinschaft (DFG) - Project number 447676110
 
Final Report Year 2024

Final Report Abstract

In this project, the risk assessment of the operation of complex technical systems is investigated using data-based methods. This requires the consideration of degraded states of various subsystems and their interactions. As a use case, the monitoring of the mission risk of unmanned aerial vehicles (UAVs) is considered. Using the example of a hybrid quad-plane configuration, which combines the advantages of a multicopter with those of a fixed-wing aircraft, the subsystems are analyzed with regard to their criticality and subsequently, the actuators (drive and control systems) and the battery are selected as subsystems to be monitored. Diagnostic and prognostic approaches for monitoring the health of these subsystems are developed using test bench data, simulation models and public data sets. A flight simulation model is used to analyze the overall system. This enables an analysis of the influence of functional impairments of several UAV actuators on the flight stability. To simulate the degradation, a decreasing efficiency of the drive motors over time, an increasing variance of the servo actuator control and a decreasing battery capacity over time are modeled. Several UAV mission sequences are simulated, with randomly generated mission profiles supplemented by path planning taking airspace restrictions into account. The influence of wind and different flight modes, including take-off, landing and transition phases between fixed-wing and multicopter flight, is also modelled. For the risk assessment, a method is being developed that estimates the current system status using hidden semi-Markov models (HSMM). For training, the simulated flight data is divided into segments extending until mission failure and clustered using K-Means or decision tree algorithms. The trained HSMMs and decision trees then enable to analyze the probability of a mission failure.

Publications

 
 

Additional Information

Textvergrößerung und Kontrastanpassung