Model-Based Robust Fault Diagnosis Of Dynamic Systems

With the development of technology and the improvement of productivity, modern control systems are becoming larger and more complicated. The possibility of fault arising in systems is also increasing. In order to improve the reliability and security of the systems, it is urgent to build a monitoring system. The monitoring system is used to monitor the system states, to detect the fault in system in time, to analyze and judge the reason and characteristics of the fault, and to take the measures to avoid catastrophic accident. Fault detection and diagnosis (FDD) algorithms and their applications have been attracted much attention over the past thirty years.Model-based fault diagnosis theory was developed in the early 1970s in 20th century and a large amount of nature theoretical knowledge on model-based fault diagnosis has been accumulated during the past three decades, such as parameter estimation approach, unknown input observer approach, eigenstructure assignment approach, parity space (relation) approach, H_∞, optimization approach, strong tracking filtering approach and differential geometry approach etc.This thesis is based on the Project "Study of Modelling and Identification of a Class of Nonlinear Systems (The National Nature Science Foundation of China)", and fault diagnosis for control system based on model is investigated. The main results are as follows:1. The latest development of fault detection and diagnosis is briefly introduced. The basic fault diagnosis approach based on observers for linear control systems is discussed, including the basic theory of the approach of unknown input observer and fault detection filter.2. Secondly, a bilinear fault detection filter (BFDF) is proposed for a bilinear system with an uncertain output. Sufficient conditions are given for the existence of a BFDF and the design procedure of BFDF is presented. A fault reconstruction approach based on equivalent control for the direction of the residual is studied. By using the equivalent control and not the residual to reconstruct the fault signal, the fault can be detected.3. Thirdly, the fault diagnosis problem for a class of nonlinear systems with uncertainty which depends on states, inputs and unknown parameters is discussed. By using the estimations of both the states and parameters with adaptive observers, we can detect the fault.4. Finally, the problem of observer-based fault detection filter (FDF) for a class of non-uniformly sampled-data multirate systems is studied. A lifting technique and unknown input observer are used to design the FDF. The design of FDF is formulated as a H_∞optimization problem and a solvable condition as well as an optimal solution are derived. All the approaches discussed in this thesis are demonstrated through corresponding simulations.