Robust Fault Detection For Networked Control Systems Based On Analytical Redundancy Technology Research

The dissertation focuses on the analytical redundancy based robust fault detection for Networked Control System with stochastic networked induced delays,packet losses and external disturbances. The main results consist of the following parts:(1) For networked control system with networked induced delays,packet dropout and unknown disturbances, different discrete time mathematical models of the NCS are established.(2) Aiming at the robust fault detection and diagnosis for NCS with stochastic delays less than one sampling period and external disturbance, the unknown input observer (UIO) is designed as the residual generator. A sufficient condition for the residual dynamics stability is given. Based on the UIO, the robust residual decoupling from the unknown inputs is constructed. At last, the fault diagnosis is achieved by applying the directional residual method.(3) The robust fault detection for NCS with stochastic long delays is studied. Based on the model of the NCS, the measurement redundancy function and parity equation of the NCS are constructed. Simultaneously, the sufficient condition and solving method of the parity vector are given. For sensor failures, actuator failures and both of them, a set of specific residual generators which is sensitive to sensor or actuator faults are constructed. The conditions the parity vector satisfied are given simultaneously. At last, the fault diagnosis is achieved.(4) The robust fault detection for NCS with long delays,external disturbances and packet losses is studied. With fixed packet dropout rate, two state observers are designed respectively for event one and event two. The problem about the observed based fault detection is formulated as designing the robust fault detection filter. By utilizing the technology of LMI, the given performance index is optimized. Then the fault detection is realized by the output residuals.