基于CNN-BES-ELM的航空发动机气路故障诊断研究  

Research on aero-engine gas path fault diagnosis based on CNN-BES-ELM

作　　者：戴郎杰 蔡开龙 王阿久 黄菲 Dai Langjie;Cai Kailong;Wang Ajiu;Huang Fei(College of Aircraft Engineering,Nanchang Hang Kong University,Nanchang 330063,China;College of Aviation,Nanchang Hang Kong University,Nanchang 330063,China)

机构地区：[1]南昌航空大学飞行器工程学院,南昌330063 [2]南昌航空大学通航学院,南昌330063

出　　处：《电子测量技术》2024年第13期27-34,共8页Electronic Measurement Technology

基　　金：江西省双千计划(jxsq2018106057)项目资助。

摘　　要：针对航空发动机运行过程中出现的气路故障问题,提出了一种基于卷积神经网络(CNN)结合秃鹰搜索算法(BES)优化极限学习机(ELM)的航空发动机气路故障诊断模型。通过CNN对航空发动机气路数据进行学习并提取出隐藏在数据中的故障特征,引入BES算法对ELM的权重和偏置进行寻优,使用优化后的ELM对CNN所提取的抽象特征进行分类,从而达到故障诊断的目的。实验结果表明,基于CNN-BES-ELM的模型其平均准确率达到了97.80%,较CNN-ELM、CNN和ELM等方法分别提高了2.7%、5.4%和7.35%,较常用的深度学习模型如深度置信网络(DBN)和堆叠自编码器(SAE),其准确率分别提高了5.4%和3.4%;并且在随机噪声、高斯噪声和椒盐噪声等噪声环境下仍保有90%以上的准确率,整体表现出良好的诊断性能、泛化能力和抗噪能力,为其在航空发动机气路故障诊断中的实际应用提供了理论依据。features extracted by the CNN,so as to achieve the purpose of fault diagnosis.The experimental results show that the CNN-BES-ELM-based model achieves an average accuracy of 97.80%,which is 2.7%,5.4%and 7.35%higher than that of CNN-ELM,CNN and ELM,respectively,and compared with commonly used deep learning models such as Deep Belief Network(DBN)and Stacked Auto Encoder(SAE),the accuracy is improved by 5.4%and 3.4%;and still retains more than 90%accuracy in noise environments such as random noise,Gaussian noise and pretzel noise,which overall shows good diagnostic performance,generalization ability and noise immunity,and provides a theoretical basis for its practical application in aero-engine airway fault diagnosis.

关 键 词：航空发动机 故障诊断 卷积神经网络 秃鹰搜索算法 极限学习机 

分 类 号：TN707[电子电信—电路与系统]