Surface To Air Missiles Intelligent Fault Diagnosis Expert System Design And Implementation

Nowadays, with the development of more kinds of air defense missiles and the enhancement of equipment complexity and automatization, missile fault diagnosis has become a more extruding problem. Especially during the lauching test period, with limited lauching test times and fault diagnosis experience, how to fully utilize current missile fault finding knowledge to locate the exact missile fault from measurable singnals and presented omens, has become a key factor in accelerating missile development.Due to the complexity of the air defense missile system and the absence of diagnosis knowledge, applying only traditional and single method in missile fault diagnosis can hardly achieve satisfactory results. In this paper, a hybrid approach combining two styles of fault diagnosis expert systems (FDES), that is the FDES based on rule inference and the one based on integrated neural networks (INN), is presented which proved to be effective in missile fault diagnosis. During diagnosis, the FDES based on rule inferece is first used to detect up-level missile fauls, and then the FDES based on INN is applied to make further diagnosis according to the missile control system where knowledge for diagnosis is scarce. The main research contents are list as follows:1. Theories and designing methods of the two kinds of FDES as related above are studied and according to the speciality of air defense missile system, the hybrid intelligent FDES solution is presented.2. Air defense missile FDES based on rule inferece is studied and designed in detail, including the construction of knowledge base system, design of the inference algorithm and the building of knowledge base management system.3. Energy spectrum ananlysis using wavelet technique is studied and applied to deal with feature exstraction of missile fault signals, from which a feasible feature vector is created, to be used by the FDES based on INN. Parameters for diagnosis are also selected based on two different criteria: the cluster divergence of sample datas and the diagnosis reliability of parameter candidates.4. Integrated neural networks, which is based on information fusion is investigated and applied to the construction of intelligent FDES.5. Software development and realization is also touched on at the end of this paper.