Main jobs and contributions involved in this dissertation are summarized as follows. 1. Fault detection, fault diagnosis algorithm and fault tolerant control for nonlinear systems. A self-adaptive observer method is adopted for fault detection and fault diagnosis for a class of nonlinear dynamic systems subjected to a linear output structure, this method is proved by the example of the three-tank system. A self-adaptive regulation rules are reached by controller reconfiguration, which enable the system that occurs fault tracks the output of the original system, so that fault tolerant control is realized. 2. Design of the fault diagnosis and the fault-tolerant control for two collaborative subsystems. Faults of two collaborative subsystems include sensor fault and process fault. When faults occur in one of the two subsystems the other healthy subsystem is used to accommodate faults and compensate the influences caused by the fault to the whole system. Theoretical analysis and computer simulations have illustrated the validity of this method. 3. Robustness analysis of the fault diagnosis and the fault-tolerant control for two collaborative subsystems. Robust adaptive observer based fault diagnosis algorithms are applied to estimate the sensor faults and process faults in the subsystems. Two robustness analysis techniques are proposed for the entire system. One is based on the known information of the faulty subsystem and the other only takes advantage of the fault estimation. 4. Fault-tolerant control algorithm for complex systems based on Multi-Agent technology. Classification of complex systems, fault diagnosis of subsystems and fault-tolerant control of whole system are involved in the research base on Multi-Agent technology. Firstly it is researched that how to make a decision of fault-tolerant control under a certain optimized figure and secondly an illustrative example of Robot Soccer Competition is analyzed to identify the structure and implementation of the fault-tolerant control of Multi-Agent systems. 5. Fault diagnosis algorithm and fault tolerant control for bounded dynamic stochastic systems. A novel approach of fault diagnosis and fault-tolerant control in non-Gaussian stochastic systems is proposed by using B-spline model to approximate system output probability density functions. Firstly a control strategy of the healthy system is obtained, then the fault diagnosis algorithms are described using an adaptive observer, finally a controller for fault system is chosen so that the actual probability density function of system output is as close as possible to the initial probability density function.
修改评论