基于ART2神经网络的发动机故障诊断方法

Fault Diagnosis of Engine Based on Modified ART2 Neural Network

发动机的故障诊断是一个动态的故障分类过程,许多故障诊断方法在对动态故障模式进行识别和分类时,存在对未知故障模式无法识别的问题。针对这一问题,引入ART2神经网络,利用db6小波包对发动机气缸盖的振动信号提取的特征向量作为网络的输入,应用实例证明,ART2神经网络不仅能正确识别学习过的故障模式,对突发、未知的故障模式也能很好地识别。

Because of the dynamic characteristic of engine fault diagnosis,many methods of fault diagnosis which can't recognize unknown fault patterns can't be used in dynamic diagnosis.In order to solve the problem,ART2 neural network is used.By means of db6 wavelet packet the characteristic vectors are extracted from vibration signals as the network input.The application results show that ART2 neural network can recognize known and unknown fault patterns.