Research On Information Fusion And Fault-tolerant Method Of Sensors In Flight Control System

Reliability and safety are the key factors in the design procedure of the civil aircraft. The development and application of the Fly-by-Wire technology and the active flight control method increases the maneuvering capability and flight performance. However, the structure and compostion of the flight control system become more complicated, which brings a challenge to the reliability and safety of the system.Fault-tolerant sensor subsystem with high precision and reliability is the essential assurance of the filght control system’s normal operation. Instead of the traditional voting monitoring method based on redundancy, advanced data fusion technology and fault-tolerant method can satisfy higher requirement of the performance index. Under the normal operational condition, the data fusion technology can synthesize multiple information sources including prior knowledge and measurement data to increase the signal accuracy. Under the sensor failiure condition, fault diagnosis method can detect and isolate the faulty sensor timely in order to alleviate the flight safety hazard caused by fault propagation. The signal reconstruction scheme can maintain the flight control task by the calculation of the analytic signal based on the mathematical relations between flight states. Therefore, the research on the data fusion and fault-tolerant method of the flight control sensor is of great significance to guarantee the reliability and safety of the system.The dissertation takes sensors in the flight control system of the civil aircraft as the research objects, establishes the integrated framework of sensor data fusion in the fault-tolerant flight control system, and discusses the fault diagnosis method of the sensor component, the fault-tolerant design of the pitch angular rate sensor based on the signal reconstruction, the weighted data fusion of sensor units based on the variance change detection, the air data subsystem fusion based on the adaptive nonlinear filtering method and the reliability modeling and analysis method of the hybrid redundancy system. The main contents and innovative points are listed as follows:(1)According to the composition and configuration characteristic of flight control sensors, the data fusion scheme is designed. The scheme is composed of three levels: component level, unit level and subsystem level. Each level can utilize complementary and redundant information, to increase the signal accuracy and enhance the fault-tolerant capability of the flight control system.(2)The fault diagnosis method of redundant sensor components is researched. With the application of the Model Group Switching algorithm to compensate the parity vector of the Average Parity Vector method, the MGS-APV method is proposed to implement fault detection and isolation. The model group cover is designed based on operating modes of the component set. The activation and termination logic of the model group is determined to adjust the model set. Therefore, the influence of sensor errors on the decision function is decreased. Under steady and dynamic flight conditions, the simulation results of the typical sensor unit demonstrate that compared to the classic methods, the MGS-APV method can achieve detection and isolation of the small amplitude fault timely, and reduce the computation cost with less number of filters in each calculating cycle.(3)The pitch angular sensor is the essential component for safe flight. In order to improve the fault tolerance and reliability, the fault-tolerant scheme is proposed, consisting of fault diagnosis assisted by analytical signal and fault-tolerant control functions. According to the different requirements of the two functions, signal reconstruction methods are designed respectively. The fault-tolerant control function has high real-time requirement, so the analytic signal is obtained by the tracking differentiator method. A signal reconstruction method based on the fuzzy current statistical model is presented to meet the relative higher accuracy requirement of the analytic signal. On purpose of improving the matching degree between the fuzzy model and the practial motion mode, Genetic Algorithm with Variable Length Chromosome based on Invasive Weed Cloning is put forward to optimize fuzzy rules and parameters simultaneously. An improved moving data window is designed to realize the fault diagnosis function assisted by the analytic signal. The improvement of the convergence and precision of the modified optimization method is verified by the classic test example. The effectiveness of the signal reconstruction, fault-tolerant control and fault diagnosis methods is proved by simulation under the typical flight condition.(4)To deal with the problem that the accuracy of weighted data fusion may decrease when the measurement noise varies, an improved weighted data fusion method is proposed. The improved method is composed of the variance estimation and hypothesis testing parts. First, the variance estimation is implemented with the adaptive moving data window, the length of which is determined by the hypothesis testing result. Then, in the part of hypothesis testing, the test statistic of the hypothesis testing approximates the normal distribution by utilizing the signal segmenting method and the central limit theorem, which simplifies the following calculation and theoretical deduction. Based on the Markov state transition theory and the maximum a posteriori criterion, the change of the measurement noise variance is detected. Finally, the comparison of simulation results verifies that the improved algorithm solves the limitation of the typical method with higher accuracy of data fusion and variance estimation.(5)To solve the problem that air data sensors have low measurement accuracy and high failure rate, the adaptive Central Differential Kalman Filter(ACDKF) method is presented. Under the normal operation condition, the ACDKF method can improve the precision of air data based on the aircraft kinematics equations and the accurate inertial signal. Under the sensor failure condition, the fault detection and isolation is implemented by the adaptive regulation of the filter gain matrix, with the introduction of the hypothesis test for the distribution variation of the innovation sequence and multiple fading factors. Compared to the CDKF method and voting monitoring method, the simulation results of single sensor failure and multiple sensor failures setting, indicate the effectness and superiority of ACDKF.(6)The reliability of hybrid redundant sensor system is modeled and analyzed, with respect to the sequential failure and complex management logic introduced by data fusion and fault-tolerant method. First, the reliability models of component failure and fault diagnosis processes are established and integrated as a semi-Markov process. Second, combining the algebraic model with supplementary variables, a quantitative analysis method is proposed. The method uses the algebraic model to reduce the reliability model to the logical sum of failure modes and the supplementary variable method to solve the reduced semi-Markov process. Third, we use distribution functions to deduce the time-sequential failure probability calculation formulas, which are used to seek the probability solution of various failure modes. Finally, the numerical analysis and the comparison with other typical methods show the generality of the reliability model and the accuracy and convenience of the quantitative analysis method. And the effect of fault-tolerant design on the reliability is analyzed based on the modeling and quantitive probability calculation.