Efficient Nonlinear Actuator Fault Detection and Isolation System for Unmanned Aerial Vehicles

In this paper, three main limitations of the classical implementation of the multiple-model adaptive-estimation method to isolate faults based on predefined fault hypotheses are highlighted. The first concerns the number of filters that must be designed to span the range of possible fault scenarios, which must be limited due to computational load. The second limitation appears when an actuator is locked at an arbitrary nonzero position that biases the residuals of the Kalman filters, leading to inaccurate fault detection and state estimation. Third, most of the implementations of a multiple-model adaptive-estimation method only work efficiently around predefined operating conditions. This paper presents a nonlinear actuator-fault detection and isolation system, which properly works over the entire operating envelope of an aircraft. Locked-in-place and floating actuator faults can be handled. The robustness of the fault detection and isolation system is enhanced by the usage of auxiliary excitation signals. The fault detection and isolation system is also capable of handling two simultaneous actuator failures with no increase of the computational load. The complete system was demonstrated in simulation with a nonlinear model of a model aircraft in moderate to severe wind conditions.