Study On Robust Fault Detection And Isolation For Dynamic Systems

With the rapid development of high-tech field such as aerospace, astronautics, nuclear industry and robotics and all kinds of domestic applications, apparatus and control systems become more and more complicated, and the investment in them are more and more high, thus an increasing demand for those systems to be safer and more reliable. Fault detection and diagnosis technique provides a new way for improving the reliability and safety of those systems; therefore more attention has been paid on its rapid development. Robustness of fault detection system is the key issue for fault detection and diagnosis technique, which is vital for its application and must be worked out thoroughly.This thesis is concerned with fault detection and isolation problem for dynamic systems such as norm-bounded uncertain systems, state-delayed uncertain systems, linear parameter-varying systems with time delays, time-delay systems with Markovian jump parameters and nonlinear systems by using fault detection filter, threshold selection method and linear matrix inequalities.First of all, for uncertain systems and state-delayed uncertain systems, the residual generators are constructed based on full-order filters, and the fault detection and isolation problems for these two classes of systems are reduced to H∞filtering problems by using H∞control theory. The fault detection filter design is formulated as convex optimization problem subject to LMI constraints in terms of linear matrix inequalities, which can be solved readily by using standard numerical software. Threshold logic of a decision function is used to evaluate the generated residual so as to enhance the robustness of fault detection systems.Then, robust fault detection and isolation problems for time-delay linear parameter-varying systems are investigated in this thesis. The residual generators are created by parameter-dependent filters, i.e. the designed filters are time-varying whose matrices parameters are known functions of time-varying parameters. Sufficient conditions for the existence of the above filters are established by means of LMIs, and the filters synthesis problems are simplified by introducing a set of auxiliary slack variables to fulfill decoupling. By using the basis functions and...