Development of analytical redundancy for fault detection and diagnosis of air data sensors in civil aircraft

Abstract

Fault detection and diagnosis (FDD) for Primary Air Data Sensors newline(PADS) on civil aircraft, crucial for determining airspeeds and aerodynamic newlineangles, remains a significant challenge, particularly when faced with newlinesimultaneous faults caused by adverse weather conditions, such as icing at newlinehigh altitudes. The sensors, including Pitot-static probes and airflow angle newlinevanes, are inherently exposed to environmental conditions due to their newlinenecessity to interact with the surrounding air mass. This makes them newlinesusceptible to blockages and measurement inaccuracies caused by factors like newlinesensor noise and external disturbances. Traditional FDD approaches in civil newlineaviation rely on redundancy from multiple air data sensor measurements and newlinedynamic model-based air data estimators as backups. However, these methods newlineare vulnerable to common-mode failures, which could lead to critical newlinesituations, such as the loss of aircraft control. This vulnerability is newlineexacerbated by the potential for simultaneous faults in sensors exposed to the newlinesame environmental conditions, coupled with the limitations of existing newlineschemes to robustly account for atmospheric disturbances and uncertainties. newlineThis research proposes an innovative FDD scheme that employs newlineanalytical redundancy through a kinematic model-based estimator, offering a newlinesolution that does not depend on inputs from aircraft controls and redundant newlinePADS measurements. This approach aims to mitigate the risk of complete newlineloss of redundant sensor data and stall conditions by providing a reliable newlinealgorithm for PADS-FDD. The kinematic air data model uses aircraft newlineEquations of Motion (EOM) and leverages data from the Inertial newlineMeasurement Unit (IMU) to offer a more disturbance-resistant estimation newlinemethod compared to traditional dynamic model-based approaches. newline