Sensor Fault Detection for Aircraft Using a Single Kalman Filter and Hidden Markov Models
Résumé
This paper presents a new scheme for sensor fault
detection and isolation. It uses a single Kalman filter and a
Gaussian hidden Markov model for each of the monitored
sensors. This combination is able to simultaneously detect single
and multiple sensor faults, still guaranteeing optimal system
state estimation. This algorithm also can run on systems with
limited computational power. The efficiency of the approach is
evaluated through simulation of an aircraft to detect airspeed
and GPS sensor faults. The results show fast fault detection
and low false-alarm rate.