摘要
滑油中的金属颗粒成分及含量反映了发动机部件磨损程度,利用光谱分析技术监测诊断发动机部件磨损故障。在分析发射光谱原始数据的基础上,提出基于BP神经网络的航空发动机磨损部位识别方法,并通过实例阐述了部位磨损识别的步骤。将待识别样本输入已经训练好的神经网络中,得到低压压气机轴承支座磨损故障模式。待识别样本中含有Fe、Al、Cr、Cu、Mg,与低压压气机轴承支座磨损故障模式存在的元素完全一致。与原始识别方法相比,本文方法得到的故障特征更加明显,所需训练样本更少,识别精度达到96.67%。
Element and content of mental grain in the oil show important information of wearing,therefore it is helpful to use spectra analysis technology to monitor and diagnose wearing fault.On the base of initial data in spectra analysis,the method of judging the wearing parts of aero-engine based on BP neutral net⁃work was put forward.The identification process of wearing parts by this method was discussed by an exam⁃ple.Identification mode was input to the trained neural network,and bearing support wearing was attained.Identification mode contains Fe,Al,Cr,Cu,Mg,which is consistent with the element of bearing support wearing.Compared with the original identification method,the fault features obtained by this method are more obvious,and the identification accuracy reaches 96.67%.
作者
孙涛
李冬
SUN Tao;LI Dong(Naval Aviation Academy Fundamental College,Yantai 264001,China;The 91899 Unit of Army,Huludao 125001,China)
出处
《燃气涡轮试验与研究》
北大核心
2019年第6期50-53,60,共5页
Gas Turbine Experiment and Research
关键词
航空发动机
光谱分析
磨损故障
神经网络
相似矩阵
聚类
aero-engine
spectra analysis
wearing failure
neural network
similar matrix
cluster