基于ELM的航空发动机故障诊断方法

摘要

以航空发动机主燃油泵为具体研究对象,提出了一种基于基于小波包能量比与极限学习机(Extreme Learning Machine,ELM)的故障诊断方法.对于某型真实航空发动机,采用振动传感器感知发动机附件机匣的振动信号,对获取的发动机附件机匣的振动信号采用DB3小波包对其进行3层小波包分解,求出第3层各频带信号的能量作为原始信号的特征,构建特征向量.用求得的特征向量建立基于ELM的故障诊断模型,对航空发动机主燃油泵进行故障诊断技术研究.为表明该方法的有效性,还设计了基于BP神经网络的故障诊断模型,并对所构建的特征向量进行了诊断.试验结果表明,基于ELM故障诊断方法可以有效提高故障诊断的速度及准确率,具有很好的工程应用前景.%In this paper,the main fuel pump of aeroengine as a specific object,a fault diagnosis method based on the extreme learning machine (ELM) is proposed. Firstly,for a certain type of real aeroengine,the vibration sensors to obtain the vibration signal of the engine casing,and DB3 wavelet packet is used to carry out the 3 layer wavelet packet decomposition for the vibration signals obtained, and find the signal energy of each band,formed feature vector in each fault mode. Then a fault diagnosis model based on ELM is established by using the obtained fault feature vectors to research on fault diagnosis technology of main fuel pump.In order to show the effectiveness of the method,based on BP neural network fault diagnosis model to diagnose the obtained fault feature vectors is designed. The results show that the fault diagnosis method based on ELM can effectively improve the speed and accuracy of fault diagnosis,which has a good engineering application prospect.