Feature Extraction And Fault Diagnosis Of Air Power Failure

Aircraft Electric Power Source System (AEPSS) is an important equipment transforming the engine energy to electric energy for all electro-equipments in airplane. Its right working state is a key to ensure normal and secure flight of airplane. Condition detection and fault diagnosis is a main research in field of aircraft electric power source system. Aircraft electric power source is usually constituted by three grades of generator, but it is difficult to build fault model for its complexity structure. At present, examine and repair on the ground depends on some traditional instruments and routine examine, whose disadvantages are few inspecting items and low efficiency. This off-line testing method cannot exactly get important states and parameters of faults.In this paper, a condition detection and fault diagnosis system simulating "on-line" on the ground of the aircraft electric power source system is designed which based-on multi-sensor information fusion technique. Analyzed three kinds of familiar electric faults of aircraft electric power source, general inspecting items are confirmed. For different data of multi-sensor, different signal processing methods extract fault features and improve SNR. The foundation of fault feature database provides a good base for fault diagnosis and classifying.Evidence reasoning is used in fault diagnosis and classifying of the aircraft electric power source system. Theory of evidence and Dempster's rule are introduced, and the method getting probability evaluate is studied. The results of the simulations show the effectiveness of evidence reasoning method and the condition detection and fault diagnosis system on the ground.