Fault Diagnosis And Fault Accommodation Based On Non-uniformly Sampling

Modern engineering system is becoming more and more large-scale and complicated, but all kinds of fault often happen inevitably which has lead to several serious accidents. In this situation, the technology of fault diagnosis and fault accommodation will be more crucial. On the other hand, with the popularity of the network and sensor technology in the complex systems, the system outputs often present the phenomenon of different interval sampling which is usually named this phenomenon the non-uniform sampling because of some facts such as the delay and loss of communication network, the deficiency of sensor performance, the delay of experimental data analysis, this kind of passive non-uniform sampling apparently increases the difficulties in fault diagnosis and fault-tolerant control. Furthermore, in order to reduce the consumption of computing and communicating resources, the active adjustment to the sampling period is a kind of effective and challenging means online which has been verified in the feedback control and state estimation and so on. However, considering the perspective of fault diagnosis and fault-tolerant control, the design of the sampling period is almost fewBased on above, this thesis will make a new research on fault diagnosis and fault accommodation design, it aims to adopt new methods including passive non-uniform sampling and active non-uniform sampling which is based on event-triggered sampling and self-triggered sampling. The main contents of this thesis are as following: 1. For a class of non-uniformly sampled-data systems with continuoustime-varying fault, an active fault tolerant control design scheme ispresented based on an optimal continuous augmented observer and aniterative algorithm. This approach can effectively treat failure andstate estimation error, the unknown time-varying failure withinter-sample behavior and so on, and guarantee the stability of thefault system. 2. The problem of fault estimation for dynamic system is preliminaryinvestigated from the event-triggered viewpoint. On the basic ofmatrix dimensions and LMI technique, a novel filter error system isfirstly designed, which can realize accurate estimates of the constantand time-varying failure. 3. For the linear time invariant system with the actuator fault,self-triggered sampling is designed to fault accommodation fordynamic systems. By defining an attenuation condition of Lyapunovfunction and using the iterative technology, this proposed schemedevelops the technology of self-trigger sampling with state estimation,which has implemented using as little as possible sampling output toestimate fault accurately and timely, and can ensure the stability ofthe fault system. 4. The feasibility and effectiveness of the proposed methods aredemonstrated by four water benchmark example and a drive trainsystem of the wind turbine.