Researches On Fault Diagnosis And Fault-Tolerant Control For Unmanned Aerial Vehicle

At present, unmanned aerial vehicle(UAV) is in a stage of rapid development, both the military and civilian areas have great demand for it. Generally, an UAV system includes mechanical part, electronic control part and ground station, and is very complicated. On the other hand, the meteorological condition of its working environment is quite complex and changeful. So UAVs always have some higher fault probabilities, and it is generally a fatal failure when flying. Considering that the UAV is expensive, as well as its workloads and valuable data, it is of great importance to detect the fault and eliminate its impact quickly.This paper reviews studies on fault diagnose of each control node in UAV, and then focuses on the wing damage fault which is not popular among domestic researches. The behaviors of normal structured fixed wing UAV in damages of three different wings are discussed and the relationship between faults and behaviors shows a relevance of fault and the change of angular velocity in three-axis. Then the theories of system identification, model based fault diagnosis and fault-tolerant control with Pseudo-Inverse method are presented in Chapter 3 briefly. The implementation steps of how to identify the parameters of a system with the recursive least squares identification method, how to diagnose a fault with the design of state observer and residual and how to calculate the pseudo-inverse matrix are given. Using STM32 microcontroller, this paper designed a hardware platform for the fault diagnosis and fault-tolerant control system of small UAV. The operating system, μC/OS-II, is introduced and transplanted in the platform. It can record the flying control data and angular velocity values as a control node in the control line of small UAV. At last, a controlled fault implant component is built to implement on the vertical tail wing, which can make the wing damaged with a remote control command during the flight. The data of normal flight can be used to model three four-input-one-output observers which estimate the angular velocity of three-axis. Then the residual can be designed as the deviation of the estimations and the true values which come from observers and sensors respectively. After that, a fault identification vector which contains three residuals can be constructed to determine the fault quickly according to the judging rules. The fault flight data are processed in Matlab to draw three lines of residuals, which show that the fault is in vertical tail wing under the judging rules. And the forms of pseudo-inverse matrix in each considered fault are presented under priori-knowledge of flight control effect in the end.