The Research Of Nolinear System Fault Diagnosis Based On Adaptive Sliding Mode Observer

For high safety requirements systems, the consequences of faults are often extremely serious. Thus increasing the reliability and maintainability of such systems is necessary, and fault diagnosis technology is a new way to improve system reliability, maintainability and validity, Wherein the fault diagnosis technology based on the system model and the residual is widely applied in the control system. However, the problem of the system-modeling error, the interference from external, the changes of the parameters which caused by-internal device wear, and non-linear are inevitable in the actual environment, therefore robust fault diagnosis method based on the analytical model has become a new research direction of fault diagnosis for nonlinear uncertain systems. For a class of nonlinear uncertain systems and the upper bound of input disturbance unknown, a fault diagnosis method, combined with adaptive technology and sliding mode observer technique, has been presented.Firstly, the thesis briefly describes the basic methods and main content of the fault diagnosis, and summaries the current status of robust fault diagnosis method. Then, the sliding mode variable structure control theory is introduced and the principles and processes of multiple fault diagnosis method are analyzed using several typical fault diagnosis methods. The sliding mode observer design, robust residual generation and fault detection method have been analyzed in detail for a class of Nonlinear Uncertain Systems which satisfies the Lipschitz condition., and the chattering of the sliding mode observer is analyzed via the contrast of the simulation results obtained by setting the same parameter using different values.For the problem of single actuator fault detection in the situation that the upper bound of input disturbance is unknown, combining with adaptive technology and sliding mode observer technique, the adaptive sliding mode observer is constructed. The observer realizes the observer gain self-tunning, which reduces the design requirements. The designed observer can detect faults in a system in the condition of its upper bound unknown, and could restrain the sliding mode observer vibration effectively.Finally, through the method of equivalent transformation, the sensor failure is transformed to the virtual actuator failure, therefore actuator and sensor faults can be processed together. Combined by dedicated observers technology, the fault detection method based on adaptive sliding mode observer is extended to the case of multiple fault diagnosis, and multiple faults can be isolated and unknown fault can be detected. In addition, the reconstructed signals of known faults are computed and the residual signal is obtained that can reflect the type of fault qualitatively...