基于小波神经网络的航空发动机建模研究

提出将多个多输入单输出小波神经网络(WNN)组合构造多输入多输出(MIMO)的WNN来逼近MIMO非线性动态系统的快速而简单的实现方法，并采用高效率的初始化方法缩短了训练时间。采用某型航空发动机在飞行包线内均匀分布的工作点参数来训练，建立了全包线适用的动态小波神经网络航空发动机模型，用交叉验证的方法检验表明在全包线内有较高的精度及泛化能力。与反传算法神经网络(BPNN)、径向基函数神经网络(RBFNN)建立的动态模型在精度及泛化能力等方面做比较，结果表明WNN建立的模型训练精度高而且泛化能力强。

Aeroengine Modeling Based on Wavelet Neural Network

A quick and simple method for mapping a multi-input multi-output (MIMO) dynamic system is presented by combination of several multi-input single-output wavelet neural networks (WNN), and an initialization method of high efficiency is utilized to shorten the train time. Then a nonlinear dynamical WNN aeroengine model is established. The model can simulate the aeroegine in the full envelope. Simulation of cross validation shows good generalization ability and high accuracy of the WNN aeroengine model in full envelope. Compared with the model constructed by back forwards neural network (BPNN) and radial based function neural network (RBFNN), the model constructed by WNN has a higher accuracy and better generalization ability than BPNN and RBFNN. It is indicated that the method of constructing aeroengine model by WNN is a very feasible.