基于双隐层过程神经网络的飞机发动机故障检测

利用双隐层过程神经网络模型可以直接处理时变信号的特点，提出了一种用双隐层过程神经网络模型对飞机发动机进行故障检测的方法。由过程神经元隐层完成对输入信息过程模式特征的提取和对时间的聚合运算，非时变一般神经元隐层用于提高网络对系统输入输出之间复杂关系的映射能力。分别利用递归神经网络和双隐层过程神经网络对发动机排气温度裕度进行仿真预测。结果表明，双隐层过程神经网络收敛速度快、精度高，优于递归神经网络的预测结果。为飞机发动机状态监测问题提供了一种有效的方法。

Failure detection of aeroengine based on process neural network with double hidden-layers

Process neural network(PNN) with double hidden-layers model was proposed to detect aeroengine failure.The network can deal with the time-varied signals.The hidden layer of process neuron executes time aggregation operation while the hidden layer of generic neuron raises the mapping capability of the network to complex relation between the system input and output.The network was compared with recurrent neural network(RNN) by predicting exhaust gas temperature(EGT).The results exhibit good convergence and high accuracy of the network and the predictive capability is superior to RNN.This provides an effective way for aeroengine failure detection.