Fault diagnosis of hybrid system with an efficient particle filtering estimation approach

Fault diagnosis is one of the central issues in hybrid system study, and the OTPF algorithm has been proposed to handle this problem. However, the performance of OTPF may become weaken in some cases. In this article, a new approach based on particle filtering is proposed to handle these situations. Comparable to OTPF, the method integrates information in modes with similar behavior to obtain better state estimation, and it considers history tracking of the system to make a more wise decision about mode detection at each time step. In addition, the ensemble Kalman filter is introduced to improve the quality of particles in the filtering process. Finally, a numerical simulation is conducted to demonstrate the efficiency of the new approach. The result indicates that the proposed approach can make more accurate estimation of hybrid system with lower computation burden than the OTPF algorithm.