Fault detection in non-linear systems based on GP-EKF and GP-UKF algorithms

In this paper, two new fault detection methods are proposed for non-linear systems. The proposed methods are based on combining an extended Kalman filter (EKF) and an unscented Kalman filter (UKF) with Gaussian processes (GPs). One of the major advantages of these algorithms is that they do not need the system's model while they have an accurate and fast operation in fault detection. In order to show the promising performance of the proposed algorithms, they are applied to an aeroplane tracking system with a highly non-linear dynamics. Superiority of the GP-UKF over GP-EKF in fault detection is also shown based on the simulation results.