Aeroengine Modeling With Neural Networks By Using Practice Measure Data

Aeroengine modeling is a very important disquisition in aeroengine field. Aeroengine math model is the precondition of engine control. From the point of quick, simple and precise views, this thesis researches on the method of using BP, RBF neural networks to build engine model with surface aeroengine's test-drive data. Further more, Auto-associatinve Neural Network is adopted to filter the noise in the data. The major contents are as below: (1) Firstly, it reviews the development course of Neural Network, introduces the basic structure and mathematic theory of the basic neural network unit-neuron. And it also uses Auto-associative Neural Network to filter the noises of aeroengine's test-drive data. It pointes out both virtue and shortage occurred in the process of filtering data with this methods. It compares the filter effect between practical measure data and academic output of the model. (2) This article discusses the methods of engine modeling and the realization, introduces the theory of principle modeling and distinguishing modeling, builds an aeroengine math model with principle modeling method. (3) Proving the forward neural-network has the ability of approaching nonlinear system at any precise, which is the theoretical base of modeling with BP and RBF. (4) Using BP and RBF neural-network to build aero-engine ground test-drive model separately, comparing and discussing the training speed , precise and generalization ability of these two kinds of models. (5) Summarizing the whole work, Expectation the using of neural-network modeling and intelligent digital filtering technology on the aeroengine.