Air data system fault modeling and detection

Abstract

Aircraft pitot-static probes are essential to airspeed and altitude measurements and safe flight. Measurement integrity is typically achieved via sensor hardware redundancy and a voting system. Hardware redundancy imposes a cost and payload penalty. This paper investigates an analytical alternative to hardware redundancy requiring a mathematical model of faulted and unfaulted pitot-static probes. The most common probe faults–debris, ice, or water blockages–are modeled using physical air data relationships and experimental wind tunnel data. These models are used with a linear model of the NASA GTM aircraft at one flight condition to design robust fault detection filters. Two linear H∞ filters are designed to detect faults, reject disturbances, and provide robustness to model errors. Performance is evaluated using experimentally derived fault models with nonlinear aircraft simulations that incorporate actuator uncertainty.