Aircraft inertial measurement unit fault identification with application to real flight data

Abstract

Using the aircraft kinematic model can reduce the model uncertainties for sensor Fault Detection and Diagnosis (FDD). Previous studies use state augmentation or differerentiation to estimate the faults in the Inertial Measurement Unit (IMU) sensors. However, augmentation increases the computational load and it is difficult to choose the Lipschitz constant especially when the aircraft perform high-amplitude maneuvers. Another problem is that it assumes turbulence is absent. In this paper, the estimation of the IMU faults is addressed using an Optimal Two-Stage Extended Kalman Filter which reduces the compu-tational load and does not require the selection of a Lipschitz constant. The performance of the filter, when coping with nonlinear systems, is improved by proposing a novel Iterated Optimal Two-Stage Extended Kalman Filter. The inuence of the turbulence is reduced by proposing a new set of kinematic model. The performance of using the two different kinematic models with and without the turbulence is compared using the simulated data of the ADMIRE aircraft benchmark model. Furthermore, the IMU fault identification using the two kinematic models is validated by using the real ight test data of a Cessna Citation II aircraft. The results demonstrate that the proposed approaches are effective and can be applied in practice.