Improved estimation and fault detection method for a class of nonlinear hybrid systems using self switched sigma point filter

Abstract

An estimation and fault detection method for a class of nonlinear hybrid systems has been presented in this paper using a self-switched Unscented Kalman Filter. The term ‘estimation of a hybrid system’ implies state estimation as well as mode estimation. In this work a residual based fault detection technique has been adopted. A three tank system has been used to demonstrate the effectiveness of the scheme. Leakage faults of different radiuses have been detected in this paper using the proposed method. If the residuals cross the pre specified threshold value, then the fault is said to be detected. ‘t-test statistical’ method has been used here to specify the threshold value. Performance of the proposed method is compared with the same for an EKF based FDI scheme. Simulation results show that the UKF based scheme can detect a smaller radius of leakage and can also detect the faults faster in compared to the EKF based scheme.