The Research For Aircraft Fault Diagnosis System Based On Integration Of FTA With BAM Neural Networks

The aircraft is a large complex system, with a large number of componentsassociated with each other. This feature makes fault diagnosis a challenge, seriouslythreatening the flight safety. So it is important and necessary to implement accurateand fast aircraft fault diagnosis. The work done by this thesis is as follows：First, it introduces the development status and trend of the traditional faultdiagnosis technology. Simultaneously, it points out the traditional fault diagnosistechnology has the limitation of self-studying and self-adaption. Fault diagnosisscheme based on fault tree analysis is unsuitable for aircraft, as the numerousassociated components will lead to a very large knowledge base of the fault tree.Therefore, its inference speed will turn out to sharply slow. As a result, it can’tprovide timely diagnosis. In contrast, Fault diagnosis scheme based on the artificialneural network has many advantages, such as: parallel processing, self-learningfunction and high accuracy, but it is very difficult to get the required trainingsamples.Second, it proposes a scheme, which integrates FTA with BAM, to implementfast and accurate fault diagnosis for aircraft. In the scheme, the characteristics thatFTA can analyze and process each failure phenomenon, and remove redundant faultdata, are utilized to get independent and orthogonal fault samples for BAM. And thenBAM can diagnose faults fast and accurately by associative memory.Third, it designs a new aircraft diagnosis system based on this scheme. It couldcome to conclusion that BAM neural network diagnosis could cover the shortage offaulty tree diagnosis by the fault diagnosis cases. The results of diagnosis shows thatthe system meets the functional requirement, simplicity of operator, good stability,strong fault tolerance and flexibility.