UAV Analytical Redundancy based fault detection of the Airspeed Sensor via Generalized Likelihood Ratio Test

The purpose of this paper is to evaluate the performance of a sequential Generalized Likelihood Ratio Test (GLRT) for the detection of possible faults occurring on the airspeed velocity sensor of an UAV based on real flight data from the WVU YF-22 research aircraft. In the first part of the paper a novel Analytical Redundancy based method is proposed to provide a real-time estimation of the airspeed velocity to be use later for fault detection purpose. In this approach, assuming the availability of the state and input signals governing the UAV dynamics, the angle of attack equation is rearranged as a quadratic equation in V(k) which is solved to provide estimates of the airspeed. The parameters governing the airspeed equation are identified via LS optimization from a record of flight data. The error signal between the measured airspeed velocity V(k) and its estimation provides the residual signal r(k) that is used for fault detection. In order to apply a sequential GLRT it was necessary to design a residual whitening filter so that rigorous statistical tests can be performed on the Gaussian whitened signal. Under this assumption it is possible to compute the probability of false alarms. In the experimental part, given a probability of false alarm the detectability of hard and soft failures was studied by varying the amplitude and rate of faults artificially injected on the experimental flight data. The experimental results confirm the capabilities of the proposed method to cope with the important problem of the airspeed sensor fault detection.