A Nonlinear Observer for Sensor Fault Detection in an Airplane

Abstract

Abstract A sensor fault detection algorithm for the aerodynamic states of an aircraft will be discussd in this paper. The algorithm consists of a nonlinear deterministic observer with a subsequent polynomial classifier. In this paper the requirements, the design and the validation particularly of the LUENBERGER-observer and in addition of the classifier will be discussed. The observer computes the aircraft velocity vector in polar coordinates with its components airspeed, angle of attack and angle of sideslip from inertial data which are provided by a laser-strap-down platform of the Inertial Navigation System (INS). The difference of each estimated and measured state is an input signal for a polynomial classifier, which is implemented on a transputer network. For the validation of the observer and the verification of the sensor fault detection, flight - test data from the twin engine aircraft DomierDo128 are used. As discussed in this paper, the resulting algorithm is a good tool for the supervision of the aerodynamic states. In addition the quality of the estimation makes it possible to replace a sensor signal of an aerodynamic state by its estimate in case of a perturbation of the physical sensor.