Rotorcraft Fault Detection Using Difference Flatness Nan Zhang 1,2 , Juan Luis López Fernandez 1,3 , Andrei Doncescu 2 and Felix Mora-Camino 1,2 1 Ecole Nationale de l’Aviation Civile, Toulouse, France 2 LAAS du CNRS, Toulouse, France 3 ESI, Universidad de Sevilla, Spain Fault detection is essential for the survivability of flying devices. However, flight dynamics of highly maneuvering vehicles are strongly nonlinear. Then the applicability of general fault detection techniques to this case is very questionable. In this communication, after introducing the concept of difference flat systems, a fault detection scheme based on difference flatness is proposed. The proposed scheme is illustrated in the case of a rotorcraft subject to faults characterized by parameter shifts. Nomenclature X k = state of the system at time k Y k = output of the system at time k U k = input of the system at time k x i,k = component i of state X at time k k = time index f, g,η , ξ = generic functions v k = flat output measurement error at time k k υ = state measurement error at time k k ω = state equation discretization error at time k V i = variance of flat output i measurement error Δ i = variance of state component i measurement error W i = variance of state component i modelling error I. Introduction N the last decade a large amount of interest has risen for new fault detection and identification (FDI) approaches for non linear systems. However few results have been obtained through purely non linear approaches. Differential flatness, a property of some nonlinear dynamic systems, introduced by Fliess et al. from the theory of differential geometry, has made possible the development of new tools to control effectively nonlinear systems. Many dynamic non linear systems have been proved to be differentially flat and the differential flatness of conventional and non conventional aircraft dynamics has been proven in different situations. While there are many different approaches to cope with fault detection in the case of linear systems, this is not the case with non linear systems and in this paper we introduce a fault detection technique applicable to difference flat non linear systems. I In the first part of this paper, the main concepts relative to difference flatness applied to discrete dynamical systems are particularly considered. Then a new approach, based on the redundancy between flat outputs and direct state component measurements, is proposed. To take into account measurement errors as well as modelling errors to perform fault detection tests in this non linear context, probabilistic distributions are generated on-line. Finally, the application of the proposed approach is to a rotorcraft subject to faults characterized by parameter shifts is discussed. American Institute of Aeronautics and Astronautics 1 AIAA Guidance, Navigation, and Control Conference 10 - 13 August 2009, Chicago, Illinois AIAA 2009-5631 Copyright © 2009 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.