Model-based fault diagnosis has attracted considerable attention from researchers and developers of flight control systems, thanks to its hardware simplicity and cost-effectiveness. However, the airplane model, which is adopted commonly in fault diagnosis, only exists theoretically and is linearized in approximation. For this reason, uncertainties such as system non-linearity and subjectivity will degrade the fault diagnosis results. In this paper, we propose a novel actuator fault diagnosis scheme for flight control systems based on model identification techniques. With this scheme, system identification can be achieved with a linear model that uses a closed-loop subspace model identification algorithm, and a non-linear model that uses an extended state observer and neural networks. On this basis, the current actuator fault is estimated using an adaptive two-stage Kalman filter. Finally, the non-linear six-degree-of-freedom model of a B747 airplane is simulated in the Matlab/Simulink environment, where the effectiveness of the proposed scheme is verified from fault diagnosis tests.
Modelling molecular interaction pathways using a two-stage identification algorithm