Hybrid System State Tracking and Fault Detection Using Particle Filters

• Published in: IEEE transactions on control systems technology Vol. 14; no. 6; pp. 1078 - 1087
• Main Authors: Tafazoli, S., Xuehong Sun
• Format: Journal Article
• Language: English
• Published: New York, NY IEEE 01.11.2006; Institute of Electrical and Electronics Engineers; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Applied sciences; Bayesian inference; Computer science; control theory; systems; Computer simulation; Control systems; Control theory. Systems; Exact sciences and technology; Fault detection; Faults; Filtering; Gaussian distribution; hybrid system; Hybrid systems; Inference algorithms; Linear systems; Mathematical models; Modelling and identification; Nonlinear dynamical systems; Nonlinear systems; Particle filters; Particle tracking; sequential Monte Carlo methods; state-space methods; Sun; Tracking; Bayes estimation; Monte Carlo method; fault detection; Inference; State space method; Modeling; Hybrid system; sequential Monte Carlo methods; Filter; Sequential method; particle filters; Bayesian inference; Defect detection; state-space methods
• ISSN: 1063-6536; 1558-0865
• DOI: 10.1109/TCST.2006.883193

When particle filters are used for fault detection, they have the problem of sample impoverishment, which means there are not enough particles that can transition to a rare-occurring faulty mode. The consequence is that the fault cannot be properly detected. This paper proposes a method to overcome this problem. Essentially, we develop an algorithm for tracking the states of hybrid systems where fault detection is modeled as a special case of the state tracking of a hybrid system. Extensive simulations are carried out to analyze the effects of various parameters on the performance of the algorithm. It is shown that our algorithm can detect both known and unknown faults using a very small number of particles