Control-based fault detection and isolation for autonomous aircraft

Abstract

This article describes a new method to perform fault detection and isolation for a closed-loop-controlled autonomous aircraft. This vehicle is equipped with standard sensors and actuators, and its dynamics is non-linear. It is assumed that a guidance law and a control loop have been designed to achieve a given mission. The diagnosis method uses the resulting control objectives to generate residuals indicative of the presence of faults. Two classical guidance laws are considered, leading to different control constraints and diagnosis signals. A structural sensitivity analysis shows that all sensor and actuator faults can be detected and all sensor faults isolated, for both laws. The fault diagnosis procedure does not require the costly integration of the model of the system, and the closed-loop scheme makes it robust to model uncertainty. Realistic simulation results with strong model and measurement uncertainty demonstrate the potential of the approach. A theoretical analogy with observer-based fault diagnosis is also derived.