Research Of Model Based Fault Diagnosis For Flight Control Systems

As common defense and civilian aerial platform, the basic features of aircraft include three main aspects, such as task management, fault diagnosis and payload operations. With the development of aviation technology, more and more aircrafts perform the complex flight mission. They often are used for high mobility or long-endurance flight, and even explore the high/low temperature, high radiation areas and other harsh environments region. So that flight control systems(FCS) are becoming more and more sophisticated and complex. And at designning step, they are requiremented to have the ability of the fault detection and diagnosis. At U.S. Unmanned Aerial Vehicle(UAV) autonomous control Level, the Level 2 is the capability of real-time fault diagnostic,and level 3 is adapt to failure and flight conditions. It can be seen that the developed country attached great importance to FCS fault diagnostic and regulation functions. This thesis aims the actuator faults and sensor faults of FCS, bases on the observer theory to study the fault detection, diagnosis, localization and regulation method, and builds a FCS fault diagnosis semi-physical simulation platform to test FCS fault diagnosis technology.At first the research background and development of dynamic systems fault diagnosis technology are introduced. Then fault diagnosis classification and the application of the fault diagnosis technology in the field of aviation are investigated, and the common failure modes and features of FCS are analysed.The basic approach of control system fault diagnosis based on observers is introduced, the general structure of fault diagnosis observer are described, and a equivalent definitions of observer status is given. Then based on the definition, the necessary conditions of effective observer state are proposed. Afterwards, simulation examples are used to show the efficiency of the proposed observer operation and fault detection.The fault detection in the sensor subsystem of FCS are researched. From the perspective of the time series correlation, the definition and conditions of the diagnosability of sensor channels are proved, and the algorithm of dynamically adjust the length of the time window based on parity space method are proposed to detect and isolate flight control sensor group faults. Then, the satellite navigation system is used as an example to reasearch the extended Kalman filter method for nonlinear sensor fault detection and location.A robust fault diagnosis approach based on nonlinear observer is presented for actuator fault in affine nonlinear flight control system. By using a differential-geometric approach, affine nonlinear FCS is transformed in state and output space. The original system is reduced to a “observable quotient system”. Then by designning fault detection observer, the residuals are decoupled from fault signals in other channels and uncertainty, and actuator fault in different input channels are detected and isolated. And a simulation example of a certain type of aircraft shows that the algorithms have good performance, and also work well on multi-fault situation.For the discrete model of FCS, a fault diagnosis method based on adaptive observer is proposed, and fault estimation algorithms are given. For nonlinear characteristics which meet Lipschiz conditions, the observer error convergence conditions are also investigated.Finally, a set of UAV fault diagnosis semi-physical simulation platform has been constructed, which used the embedded system to build flight control computer, and built a computer of the aircraft model and a fault diagnosis computer with CPCI bus Industrial Personal Computer(IPC). The platform working principle and software design process have been described.