Non-parametric fault detection methods in non-linear systems

In this study, two fault detection methods are proposed for non-linear systems. In these methods, Gaussian process (GP) is integrated into extended Kalman filter (EKF) and square root cubature Kalman filter (SCKF), which are called GP-EKF and GP-SCKF, respectively. The most important advantage of the proposed methods is that there is no need to know the accurate model of the system. Therefore, these methods are considered as non-parametric approaches of fault detection in non-linear systems. Moreover, by applying these methods, it is possible to detect the fault with high accuracy at early stage. First, GP-EKF and GP-SCKF are proposed for non-linear state estimation, and then GP-SCKF is compared with the GP-EKF and the results of this comparison prove the superiority of GP-SCKF regarding the complexity of computations and accuracy. In addition, simulation results show a good performance of GP-EKF and GP-SCKF in non-linear system's fault detection. To illustrate performance of these algorithms in state estimation and fault detection, they are used in aircraft tracking system. Both of the proposed methods are able to detect the sensors faults at early stage.