Network Control System Fault Diagnosis And Fault-tolerant Control Research

With the rapid development of the information technology, networked control systems (NCSs) expand continually and become more and more complexity. In practice, security and reliability are very significant, so the study on fault diagnosis and fault tolerant control of NCS has important theoretical value and practical significance. Based on the existing theoretical research results, we study fault diagnosis and fault tolerant control of NCS, and propose the new ideas and methods.In the process of data transmission, time-delay, packet-loss, and error sequence inevitably exist, due to the factors bandwidth and information collision. For NCS with random time-delay, we construct a discrete model with random delays. Two general optimal fault detection filters are designed:Kalman filter and H-infinity filter. In simulations, it shows the differences of the two methods simply and intuitively. Moreover, the former is obtained using an iteration method, and the latter is realized with linear matrix inequalities (LMIs) and augmented matrices to ensure the smallest possible error between the residual and the fault.Considering a single network-induced factor, there are many research results on fault diagnosis and fault tolerant control of NCS. However, taking into account time-delay, packet-loss, error sequence and other factors at the same time, the research results are rare. Therefore, fault detection for networked control systems with limited quality of services (QoS), such as network-induced time delay, data dropout, and error sequence, is studied. We develop an augmented Markov Jump Systems (MJSs) model. The robust fault detection filter is designed by making use of Ricatti inequalities to guarantee the error dynamics system is stochastic stable and the residual is sensitive to the fault and robust to the disturbances.For the common sensor and actuator failure, the fault-tolerant controller design of networked control system is investigated. The common actuator faults are described as a "two-state fault model". Then, the continuous model of NCS with networked-induced delay, packet-loss, error sequence and other factors, is established. We construct Lyapunov-Krasovskii function, with the free weighting matrices to eliminate the cross terms and by utilizing the convexity of the matrix function in our derivation to obtain less conservative results. Finally, a simulation verifies the validity of the conclusion.With the rapid development of control systems, nonlinear system theory has received wide attention. For uncertain nonlinear NCS with random sensor failure, the T-S fuzzv model and the random sensor failure model are established. Then, based on the theories of the stochastic system and the fuzzy system, a new delay-dependent method of fault-tolerant controller design is proposed. In our theoretical derivation, neither any model transformation nor free weighting matrices are employed. The basic ideas are to use appropriate Lyapunov-Krasovskii function and augmented Lyapunov-Krasovskii functional, to construct uncorrelated augmented matrix (UAM). In addition, we deal with cross terms in the UAM through improved Jensen inequality and consider the uncertain parameters of the system as the disturbance. Simulation results show the effectiveness of the method.