Research And Application Of Sensor Fault Diagnosis

The methods of sensor faults diagnosis in nonlinear system are summarized and the application of nonlinear filters in fault diagnosis is discussed in this paper. Three schemes based on nonlinear filter technique are presented to detect sensor faults in boiler-turbine system. When the plant model is known, those algorithms of fault diagnosis are implemented by extended kalman filter and unscented kalman filter respectively. The principle and implementation process of algorithms are described, and the effect of identification and detection to multiple sensor faults are analyzed. In addition, the sensitivity, robustness and stability of these algorithms are also discussed and the difference between two kinds of algorithms are given by simulation experiment. Unscented filter can be used to detect incipient faults, isolate sensor faults in MIMO and intense coupling system. Simulation results show it's robust and stable. When the model is unknown, adaptive laguerre filter is designed to identify the plant online. Then the scheme of fault diagnosis is applied to detect sensor faults in boiler-turbine system. Simulation results show the algorithm can properly detect the faults in the system and meet the demand of production. Finally, the overall work is summarized.