Study Of Aircraft Engine Fault Diagnosis Based On RBF Neural Network

The aircraft engine fault diagnosis is very important to the airline, if can understand the engine performance and fast recognize its fault accurately then have the plan to trade sends as well as determines when the engine send to repair, avoids the significant accident, enhance the company benefit. This article first introduces the current aircraft engine fault diagnosis technology and analysis neural network fault diagnosis present situation, then uses Radial Basis Function (RBF) network for the aircraft engine fault diagnosis.The neural network has the good project application prospect, and it is considered as the most potential diagnostic tool. At present the BP network, SOM network, PNN network etc. are used in the engine fault diagnosis. The BP network depends on the entire network "to remember" the fault pattern, the SOM network uses the sample distance to recognize the fault pattern, the PNN first estimate the classes'probability by the known pattern sample, then obtains the Bayes classification also its network training does not need to iterate.And then uses Radial Basis Function (RBF) network for the aircraft engine fault diagnosis. First introduces the Radial Basis Function (RBF), the RBF network study algorithms. According to the aircraft engine fault actual situation, improve and realize nearest neighbor-clustering algorithm, and compare the several kinds of neural networks, finally uses the integration neural network for the fault diagnosis. The result indicated that, the RBF network has the simple structure, its training speed is quick, the diagnosis precision is higher, has the very good potential in the engine fault diagnosis; But the number of its hide layer, the center and the spread of hide layer function, still need to study.