Cross-layer Bayesian Network for UAV Health Monitoring

The growing use of Unmanned Aerial Vehicles (UAVs) implies high reliability and safety requirements, particularly for safety- and mission-critical applications. To ensure flawless operation of a UAV, it is essential to recognize and isolate faults at all layers before they cause system failures. This paper presents an integrated Bayesian network-based method for UAV health management, considering the cross-layer dependencies of various sub modules such as avionics, propulsion, sensors and actuators, communication modules, and onboard computers. The approach employs Failure Mode and Effect Analysis (FMEA) in a cross-layer manner, factoring in dependencies across various subsystems to enhance Fault Detection and Isolation (FDI) performance. By converting FMEA-derived faults and failure events into a cohesive Bayesian network, the proposed methodology facilitates efficient identification and quantification of fault probabilities based on evidence gathered through sensor data. The paper includes case studies and numerical examples that illustrate the efficacy of the proposed methodology in analysing UAV health and isolating faults in intricate, interdependent systems.