A mixed H-/H LPV FDD observer for nonlinear aircraft

Abstract

This paper describes the development of a robust residual generator-based linear parameter varying (LPV) mixed H /H1 observer for non-linear systems, to facilitate the robust fault detection and diagnosis (FDD) task for flight control systems. The main design goal is to maximize the robustness of the residual signal to uncertainty and disturbances whilst also achieving the specific minimum sensitivity of the residual signal to faults. The specific minimum sensitivity index used is based on the H index concept and is extended to the LPV FDD system problem. This allows the fault signature for multiple sensor and actuator faults to be reconstructed simultaneously, facilitating the robust isolation of faults rather than just their detection. Instead of using constant gains, parameter-varying gains are used in the LPV observer and the free design parameters generated through using a generalized inverse, can be parameterized to improve the conservatism of finding the robust solution. This approach is combined within the standard affine quadratic LPV framework since it is well known that this framework leads to less conservative results, compared with the standard quadratic LPV case. The FDD approach developed is then applied in an example of a non-linear full force and moment aircraft system simulation to detect and isolate multiple actuator faults. The robust residuals generated approximate closely the actual fault signals.