The Research On Aeroengine Intelligent Modeling And Fault Diagnosis

Modeling and gas path fault diagnosis based on model are important research areas in aeroengine. In this thesis, from the point of quick, simple, exact and practical views, we have studied the dynamic modeling methods of aeroengine. The models are constructed by BP, RBF neural network and wavelet network, respectively. By use of wavelet analysis and wavelet neural network the diagnosis of the gas path fault is studied also. At last a summary and discuss of the future use of wavelet transform in the aeroengine modeling and fault diagnosis in general are given. The main contents are as follows:(1) First we discuss the modeling principle of aeroengine component level model and its realization methods, introduce the methods of analysis and experiment, and get the idea of using wavelet neural network to construct aroengine model.(2) The BP, RBF neural network's theory are introduced. Then apply them to identification of aeroengine, and contrast both methods in the train speed, precision, generalization ability etc.(3) After study of the wavelet transform and the wavelet neural network theory, we realize the multi-input single-output (MISO) wavelet neural network's program, and parallely connect some MISO wavelet neural networks to get the multi-input multi-output (MIMO) wavelet neural network.(4) Combine the MIMO wavelet neural network we get the aeroengine dynamic model. The simulation results show that MIMO wavelet neural network has the best generalization ability and precision, the model has a worth of practical use.(5) Wavelet transform and MIMO wavelet neural network technique are both used to diagnosis the aeroengine gas path components fault. After study of the diagnosis of single and compound fault, it turns out that wavelet neural network has improved the diagnosis accuracy.(6) Summarize the whole paper and discuss the application of wavelet transform to the aeroengine modeling in the future.