Fault Detection And Diagnosis Methods For Networked Control Systems

The fault detection and diagnosis methods for the networked control systems are studied in this dissertation. The main research works are concluded as follows:(1) The detection and diagnosis for the NCS with output delays and data packet dropout are studied. The sensor, controller and actuator nodes are all time-driven. Based on discrete models, the fault detection problems are studied through the methods of fixed-output delay, delay compensation and equally divided sampling period, respectively. By constructing the fault observers, the stability conditions are given based on Lyapunov stability theory. After the fault detection, the faults are isolated with the parallel robust fault detection observer method and the directional residual vector method, and the fault diagnosis is accomplished.(2) The robust fault detection for the NCS with control delays is studied. The NCS with time-driven sensor node, event-driven actuator node and controller node is modelled as an uncertain discrete time system. Considering the NCS without data dropout, the state estimate error equation is given by constructing the fault observer, and a linear matrix inequation gives the stability condition of the observer. Considering the NCS with data dropout, under fixed packet dropout rate, the fault observer is modelled as a discrete-time switching system, the fault detection is studied and the condition for the mean-square stability of the system is given. For the NCS with control delays, packet dropout and disturbances, the robust H∞fault detection is studied, and the robustness and sensitivity are analyzed.(3) The detection and diagnosis for the NCS with long time delays are studied through the filter method and predictor method. Considering a class of networked control systems with time delays, a novel method is proposed to design Kalman filters with delay compensation, and it is used to estimate the state of the system. Through studying the delay characteristics, the discrete-time Kalman filter formula of the networked control systems is derived by using the variational method, and the specific optimal linear recursive filtering algorithm is given. And in the fault detection, a new integrated chi-square test is proposed. And then, the delay sampling value of the networked control systems is predicted by neural network. The error equation of the fault observer is constructed based on a neural network prediction, and the stability condition of the observer is proved. Finally, a RBF fault isolation method is given. (4) The fault detection for the networked control systems using wireless transmission is studied. The type of wireless network and the research contents of the wireless networked control systems are introduced. Considering a class of networked control systems using wireless transmission between the sensor and the controller, and with nonlinear controlled object, using the concept of the dominant subsystem in fuzzy models, the fuzzy observer is designed based on local T-S fuzzy models, and the sufficient condition of globally asymptotic stability is given. At the same time, considering a class of networked control systems built around wireless sensor networks, the fault detection is carried out when the output delays are present. By studying the characteristics of data transmission with wireless sensor networks, the fault observer is constructed based on discrete switching systems. And a linear matrix inquality gives the stability condition of this observer. When the system is normal, if the given inequality condition is satisfied, the observer system is stable. When a fault is present, the observer residue can change rapidly and detect the occurrence of the fault.(5) The fault diagnosis simulation platform for networked control systems is constructed. Firstly, the properties of an ideal fault diagnosis simulation platform for NCS are summarized and a novel framework for it is proposed, and then fault diagnosis simulation platform for NCS is implemented based on Delphi,ADS 1.2 and hardware component. For simulation platform, network delay can be more than one period and each part of NCS has some optional properties. The platform overcomes some disadvantages of existent simulation platforms and is more effective and flexible. And it accomplishes to simulate the operating process of control systems under network, and to test the fault detection and diagnosis methods for networked control systems.