Diagnostic Techniques, Based On Causal Network Avionics System Failure

Because of the complicated structure, mass data transmission and variable correlative faults in an avionic system, it is difficult to localize faults and to guarantee the accuracy and efficiency of the avionic system fault diagnosis. Especially in the case of multiple faults, the propagations of correlative faults caused large difficulty for the failure diagnosis procedure. So the multiple-fault diagnosis research is always a hotspot issue. The Parsimonious Covering Theory and the Causal Network Model is first applied to the multiple-fault diagnosis of an avionic system. And the multiple-fault diagnosis technique of an avionic system based on Causal Network is researched.First, the types of multiple faults in avionic system are investigated, the generic fault model and correlative fault model are set up, the cause-effect relationship and influencing factors among the correlative faults are proposed. The Parsimonious Covering Theory is applied to the avionic system failure diagnosis, and the criterion for parsimony is chosen. Combining the causal network and object-oriented method, an object-oriented representation method of causal network knowledge is put forward.Then, founded on the analysis of multiple-fault model of an avionic system, the methods of causal network construction based on structural model, functional model and fault tree are presented respectively, including the constructing regulation and examples. The difficulty of multiple-fault diagnosis modeling in avionic system is successfully solved. By the investigation of the design and application of avionic system, a causal strength calculation method depending on the fuzzy algorithm of multiple attribution judgment is put forward by using fuzzy theory and Analytical Hierarchy Process, to solve the key problem of causal strength calculation in the causal network.Finally, focused on an avionic system, the algorithm of multiple-fault diagnosis based on causal network is obtained and validated according to the reliability data and expert experience, while the fault diagnosis procedure and the reasoning method is given. The expert system based on causal network is also developed. The validities of the methods presented in the thesis are verified by the application of the algorithms in the multiple-fault and correlative fault diagnosis of an avionic system.