FlightGear as a Tool for Real Time Fault-injection, Detection and Self-repair

Abstract

The development of fault-detection and self-healing methods at both a hardware and software level in modern aircraft is an attractive prospect. However it is expensive to design and test these techniques using real aircraft. This paper appraises the viability of using FlightGear, an open- source Flight Simulator, as a test-bed for these approaches. The paper characterises the realism of various aspects of a model of the Airbus A380. Interfaces are established to abstract critical control system routines from FlightGear. These functions are replicated in both software and hardware environments. The control data can then be subjected to fault-injection and the control modules modified to enable fault-detection and self-healing. By applying cluster analysis techniques to training sets of data, a fault-detection, diagnosis and self-healing model is designed to address these injected faults. FlightGear is found to provide highly realistic simulation of aircraft systems and instrumentation. Hardware-in-the-loop testing shows promise as an area for future work. The proposed fault-detection model is found to provide 96% accuracy.