Robust Fault Detection And Diagnosis On Dynamic Systems

Robust fault detection and diagnosis (FDD) on a class of nonlinear systems and Networked Control Systems (NCS) are studied in this dissertation, of which the main contributions are concluded as follows:(1) Robust FDD on nonlinear systems are studied based on adaptive fuzzy observer. A direct fault detection for a class of nonlinear systems is achieved by designing an adaptive fuzzy output observer and corresponding parameter adjustable algorithm. By applying running-points linearization for such nonlinear systems, a robust FDD approach is then proposed based on adaptive fuzzy states observer and RBF neural network fault-classifier. A new weight matrix adjustable algorithm on observer and an improved robust classifying algorithm on classifier are presented. The identification of the fault which is modeled by adaptive fuzzy systems is also addressed.(2) Robust FDD based on parity equations is studied for a class of nonlinear systems described by T-S fuzzy models. Based on the T-S fuzzy nonlinear models, measuring residual and fuzzy parity equations are firstly constructed, then the fault detection and fault parameters identification are discussed. The existence and search method of parity vector are studied. For the nonlinear uncertain systems with exterior disturbances, robust FDD as well as the solution of optimal parity vectors are studied.(3) A comprehensive fault diagnosis approach on nonlinear systems is studied by using uncertain information theory. A comprehensive fault diagnosis approach based on fuzzy theory is proposed by combining observer-based method with knowledge-based method. Furthermore, the fault information obtained by observer-based method with knowledge-based method respectively is sequenced according to performances of FDD. A comprehensive fault diagnosis approach based on rough set theory is proposed and the diagnosis rules are obtained by using the requirements-oriented reduction algorithm.(4) Robust FDD on NCS with short time-delay are studied by states observer-based approach. By transforming the effect of random short time-delay on NCS into the unknown bounded uncertainness, which can be fulfilled by the matrix decomposition, the robust fault detection on NCS with short time-delay is achieved, and h_∞ and H₂/ H_∞ states observers are designed. In addition, based on the given measuring residual and parity equations on NCS with short time-delay, a method of robust fault diagnosis is presented by...