中国机械工程 ›› 2014, Vol. 25 ›› Issue (11): 1433-1437.

• 机械基础工程 • 上一篇    下一篇

基于主成分分析和线性判别的航空发动机状态监视

周媛1,2;左洪福1   

  1. 1.南京航空航天大学,南京,210016
    2.南京信息工程大学,南京,210044
  • 出版日期:2014-06-10 发布日期:2014-06-23
  • 基金资助:
    国家自然科学基金资助重点项目(60939003)

Condition Monitoring of Aero-engine Based on PCA and LDA

Zhou Yuan1,2;Zuo Hongfu1   

  1. 1.Nanjing University of Aeronautics and Astronautics,Nanjing,210016
    2.Nanjing University of Information Science and Technology,Nanjing,210044
  • Online:2014-06-10 Published:2014-06-23
  • Supported by:
    National Natural Science Foundation of China(No. 60939003)

摘要:

利用航空发动机传感器数据对发动机状态进行监视,采用主成分分析(PCA)方法和线性判别法(LDA)对发动机传感器数据进行二次特征提取,按照最优近邻思想进行分类。将2008年IEEE PHM数据作为实验数据,将基于PCA和LDA的分类结果与基于PCA的分类方法以及深度信念网(DBN)分类方法的结果进行了对比分析,结果表明,基于PCA和LDA方法的识别率综合最优且结构简单,对于工程应用该方法有效可行。

关键词: 航空发动机, 状态监视, 主成分分析, 线性判别, 深度信念网

Abstract:

This paper proposed an approach for aero-engine condition monitoring by sensors data based on PCA and LDA. The failure features of sensors data were re-extracted via PCA and LDA, and the engine conditions were classified by the nearest neighbor algorithm. The experiments were conducted on the 2008 IEEE PHM challenge data, and the proposed approach was compared with a PCA based classification method and a DBN based  classification method. The results show that the proposed method exhibits a higher recognition accuracy with simpler structure which is practicable and efficient to engineering apllications.

Key words: aero-engine, condition monitoring, principal component analysis(PCA), linear discriminant analysis(LDA);deep belief net(DBN)

中图分类号: