On Observer-based Robust Fault Detection And Reconstruction For Nonlinear Systems

State monitoring, fault diagnosis and fault-tolerant control have drawn great public interest and been a key problem to be solved in theories and engineering application of control systems. The fault diagnosis technology based on a combination of the analytic model of system and modern control theories and methodologies is widely applied in control systems. Factors such as nonlinearity, modeling errors and unknown input disturbance, etc., however, have greatly limited the application of such technologies. Granted by the Natural Science Foundation of China and the Ministerial-level Preparatory Research Project, a systematic and deep investigation is done in this paper on the robust fault diagnosis and reconstructing technologies of uncertain nonlinear systems, focusing on the integration of parameters estimation capability of state observation based on analytic model and the invariance to external disturbance of the sliding mode variable structure control theory, so as to increase the fault diagnosis robustness, the diagnosis capacity of the actuator faults and sensor faults, the diagnosis accuracy for weak and incipient faults, and to achieve fault-tolerant control in fault diagnosis systems.The major contents and conclusions can be summarized as follows:The features and existing problems in this field are investigated and summarized after an in-depth survey of current theories and technological development. It is proposed that robust fault diagnosis and reconstruction method is effective to improve the performance indices of fault diagnosis system and solve the application problems in observer-based fault diagnosis for mechatronic nonlinear systems.For the contradiction between robustness and sensitivity of diagnosis found in the residual generation fault diagnosis of nonlinear systems, a robust residual generation fault diagnosis based on a sliding mode observer is proposed, in which the sliding observer is designed to generate residues and the invariance of the sliding mode variable structure to unknown input disturbance is used to control the influence of disturbance on residue signals so that sensitivity of the systems to faults can be increased. An extension of this method is made by designing correspondent sliding mode observers for each fault type so as to realize multiple faults detection and isolation. The resolution of the contradiction between diagnosis robustness and sensitivity has laid a good foundation for the fault reconstruction mentioned in subsequent chapters.For the actuator fault diagnosis of the common affine nonlinear systems, a synthetic design is proposed to realize robust fault diagnosis and tolerant control. The sliding mode observer is used to decrease the influence of unknown input disturbance on the control system and realize robust state observation, the adaptive observer is used to separate and reconstruct faults for model uncertain systems, and the design of observers adopted naturally prevents the controlling states of system from being affected by faults, all of which guarantee the realization of a synthetic function of fault reconstruction and tolerant control.The research on sensor fault diagnosis for nonlinear systems is lagging far behind that on actuator fault diagnosis and most of the methodologies adopted cannot satisfy the needs in practice. For this case a synthetic design is made to realize robust fault reconstruction and tolerant control on the basis of the equivalent transformation between two different fault types. The first-order filter transfroms the sensor faults equivalently into virtual actuator faults. The parameter selecting principle for the filter is for the first time approached from the perspective of observability of system. Then, a robust fault tolerant control and reconstructing method is proposed for sensors faults of normal form nonlinear systems, which guarantees a considerably high sensitivity to incipient faults as well as the robustness of system. An enhanced sensor fault reconstructing algorithm is put forward for general nonlinear systems, in which unknown input disturbance is involved in both the forward channel and the feedback channel. A typical feature of this algorithm is its improved engineering practicability.For the contradiction between fault reconstruction accuracy and its robustness arising from the correlations between unknown input disturbance and faults, a precisely fault reconstructing method is proposed by using the disturbance decoupling technology concerning the sliding mode observer and the equivalence theory of sliding mode variable structure. To guarantee the insensitivity of the system to unknown input disturbance and the precise reconstruction of the weakest fault, that is, to realize robust fault reconstruction, a multiple faults reconstructing algorithm is achieved on the basis of diffeomorphism coordinate transformation, where the derivatives need not be calculated and the applicability greatly improved. On the basis of linear coordinate transformation, a precisely reconstructing method for of is constructed, where the faults and unknown input disturbances are completely decoupled and the precise reconstruction of them can be respectively achieved. It is also proposed that the sensitivity to the faults can be improved by eliminating the effects of sliding mode chattering on fault signals and that weak and incipient faults can be robustly reconstructed.Application study is carried out on the proposed robust fault detecting and reconstructing technology used in uncertain mechatronic nonlinear systems with a consideration of both engineering requirements and the system features of mechatronic tracking and steadying servo platform. For the DC servomotor control system on the platform, a fault reconstructing and tolerant control program based on slide mode observer is designed. Experimental results indicate the correctness and efficiency of the theories and technologies proposed in this paper.Research done in this paper has filled in and expanded the theories, technologies and application scope of fault diagnosing and reconstructing technologies for nonlinear systems, which will be of great value both academically and practically.