Adaptive unknown input observer approach for aircraft actuator fault detection and isolation

Abstract

In this paper, an adaptive unknown input observer (UIO) approach is developed to detect and isolate aircraft actuator faults. In a multiple-model scheme, a bank of parallel observers are constructed, each of which is based on a model that describes the system in the presence of a particular actuator fault. The observers are constructed based on a modified form of the standard UIO to generate fault-dependant residual signals, such that when a model matches the system, the residual signal will be zero. Otherwise, the residual will be definitely non-zero and governed uniquely by the faulty signal. For locked actuators and loss of actuator effectiveness, in which the locked position and the reduced effectiveness are additional unknowns, we develop an adaptive scheme to estimate these unknown parameters. To the best of our knowledge, this is the first adaptive UIO presented. We prove that the proposed adaptive algorithms guarantee that both the residual signals and the estimation errors of the unknown parameters converge exponentially when a model matches the plant. By further designing a model-matching index, the fault can be isolated accurately. A condition for the approach is that for an nth order system, there must be n independent measurements available. This requirement limits the applicability of our proposed approach. The condition is certainly satisfied by all state-feedback control systems. However, for some other systems, extra efforts may be needed to increase the number of measurements. The method is applied to a linear model of the F-16 aircraft with controller. The results show that the approach is effective. Copyright © 2006 John Wiley & Sons, Ltd.