基于小波过程神经网络的飞机发动机状态监视

Condition Monitoring of Aeroengine Based on Wavelet Process Neural Networks

针对飞机发动机状态监视问题,提出了小波过程神经网络模型。其隐层和输出层为过程神经元,隐层激活函数采用小波函数。该模型结合了过程神经网络可以处理连续输入信号的特点及小波变换良好的时频局域化性质,有更强的学习能力和更高的预测精度。文中给出了相应的学习算法,并以飞机发动机状态监视中排气温度裕度的预测为例,分别利用3层前向过程神经网络和小波过程神经网络进行预测。结果表明,小波过程神经网络结构更简单,收敛速度更快,优于过程神经网络,因而为飞机发动机状态监视提供了一种有效的方法。

Aiming at the problem of aeroengine condition monitoring, a wavelet process neural network （WPNN） model is proposed. Its hidden layer and output layer are composed of process neuron and the hidden layer function consists of wavelet function. The network has not only the capability to deal with the continuous input signals, but also the time-frequency local property of the wavelet analysis. The learning ability of WPNN is better and the predictive precision is higher. The corresponding learning algorithm is given and the network is compared with three layers feedforward process neural network （PNN） by predicting the exhaust gas temperature （EGT）. The result exhibits good convergence and simple architecture of the network. The prediction capability is superior to PNN. This provides an effective way for the problem of aeroengine condition monitoring.