Robust Fault Diagnosis Of Uncertain System

With the development and application of control technique, the safety and reliability of the control system are increasingly important. The fault diagnosis has become one of the important technique to ensure the safety and reliability. Since 1970's more and more attention has been paid to the fault diagnosis of the control system. Nowadays plenty of methods have been proposed and there are some successful applications. Thereinto the model-based method has been widely studied. However because of the model uncertainty, this method is lack of robustness, and its application is limited.In order to overcome aforementioned proplem, this thesis is concern with the robust fault diagnosis, the fault estimation, the application of fault diagnosis techniques and so on.Firstly the development of fault diagnosis techniques is introduced and diagnosis methods have been summarized systematically. Then the new concept called"the composite fault diagnosis method based on observer"is proposed to revise and reinforce the old category. The problems remained for further study and the orientation in this area is also pointed in the thesis.To multi-actuator fault in the stochastic linear system, a new fault diagnosis method based on the unknown inputs Kalman filter and the wavelet fault extraction is proposed in the thesis. The method can effectively diagnose multi-actuator faults that occur synchronously or continuously. It can not only detect the faults timely but also estimate them exactly. The method is robust to unknown inputs.To the stochastic discrete bilinear system with unknown disturbance inputs, the disturbance decoupling technique and the bias-free technique are utilized to realize the robust fault detection and estimation. Compared with the extended state variable method, the proposed fault diagnosis algorithm can reduce the compatations significantly. Moreover it can combine the fault detection with the fault diagnosis, i.e., the amplitude of the fault can be estimate directly.