推进技术 ›› 2008, Vol. 29 ›› Issue (1): 79-83.

• • 上一篇    下一篇

基于核函数主元分析的航空发动机故障检测方法

胡金海,谢寿生,陈卫,侯胜利,蔡开龙   

  1. 空军工程大学工程学院 陕西西安710038;空军工程大学工程学院 陕西西安710039;空军工程大学工程学院 陕西西安710040;空军工程大学工程学院 陕西西安710041;空军工程大学工程学院 陕西西安710042
  • 发布日期:2021-08-15
  • 基金资助:
    国家自然科学基金资助项目(60672179);军队重点科研基金资助项目(2003KJ01705)

An aeroengine fault detection method based on kernel principal component analysis

  1. Engineering Inst.,Air force Engineering Univ.,Xi’an 710038,China;Engineering Inst.,Air force Engineering Univ.,Xi’an 710039,China;Engineering Inst.,Air force Engineering Univ.,Xi’an 710040,China;Engineering Inst.,Air force Engineering Univ.,Xi’an 710041,China;Engineering Inst.,Air force Engineering Univ.,Xi’an 710042,China
  • Published:2021-08-15

摘要: 航空发动机性能由正常到异常、再由异常发展到完全故障的阶段,其参数变化具有一定非线性特征。为了有效检测这种具有非线性特征的故障,提出一种基于核函数主元分析(KPCA)的非线性故障检测方法。该方法通过核函数完成非线性变换,将变量由非线性的输入空间转换到线性的特征空间,在特征空间中使用线性主元分析(PCA)方法计算主元,构造T2和SPE统计量检测故障的发生。通过对某型涡扇发动机进行实例验证分析,结果表明,KPCA方法一方面克服了综合参数法由于没有确定的警戒值而无法有效地进行故障检测的不足;另一方面KPCA方法在非线性故障检测过程中能够提取重要的非线性特征信息,因而比PCA方法能更早地检测到早期潜在故障,且KPCA方法检测错误率更低。因此,KPCA方法更适合于具有非线性特征的航空发动机故障检测。

关键词: 航空发动机;性能监控;故障检测;核主元分析法;主元分析法

Abstract: The aeroengine parameters possess some nonlinear features when the performance of aeroengine goes from normal to abnormal,and further from abnormal to completely faulty.In order to effectively detect the fault owned nonlinear feature,a novel approach of fault detection of aeroengine based on kernel principal component analysis(KPCA) model is presented.KPCA performs nonlinear transformation by kernel function to map the nonlinear input space into linear feature space and computes principal component by performing principal component analysis(PCA) in feature space,and detects faults by utilizing statistics T2 and SPE.The practical applications in monitoring certain type of turbine-fan engine show that KPCA is superior to PCA in fault detection and is more suitable to fault detection of aerogengine owned nonlinear feature.

Key words: Aeroengine;Performance monitoring;Fault detection;Kernel principal component analysis;Principal component analysis