摘要
为了提前感知滚动轴承故障,避免民用航空发动机非计划维护带来的损失,提出了一种故障预警方法。对轴承振动信号进行特征工程,提取其时域和频域特征,引入梯度提升决策树(Gradient boosting decision tree,GBDT)算法,量化了特征重要度;在特征相关性分析的基础上,利用核主成分分析(Kernel principal component analysis,KPCA)方法实现特征融合与主元提取,再次结合GBDT构建了故障预警模型,使用交叉验证法实现了模型泛化能力评估。结果表明:KPCA+GBDT模型的泛化性能显著,模型的F_(1)分数高达0.991,对应的受试者工作特性(Receiver operating characteristic,ROC)曲线下面积的值为0.998,体现出该方法用于支撑航空发动机健康管理与维护决策工作的工程应用价值。
In order to detect the rolling bearing fault in advance and avoid the loss caused by unscheduled maintenance of civil aero-engine,a fault early warning method is proposed in the present work. Feature engineering is applied to bearing vibration signals,and its time and frequency domain features are extracted. The gradient boosting decision tree(GBDT)algorithm is adopted to quantify the feature importance. On the basis of feature correlation analysis,the kernel principal component analysis(KPCA)method is used to achieve feature fusion and principal component extraction. Combining with GBDT algorithm again,the fault early warning model is established. The cross-validation method is used to evaluate the generalization ability of the proposed model. The results show that the generalization performance of the KPCA+GBDT model is significant. The F_(1) score of the model is as high as 0.991,and the area under the receiver operating characteristic(ROC)curve,area under curve is 0.998,which means this model can be used to support the health management and maintenance decision-making work of civil aero-engine.
作者
刘勇
王朝
周平
LIU Yong;WANG Chao;ZHOU Ping(College of Aeronautical Engineering,Civil Aviation University of China,Tianjin 300300,China;Beijing Aeronautical Science and Technology Research Institute of COM AC,Beijing 102211,China;Chengdu Holy Industry&Commerce Corp.Ltd(Group),Chengdu 640041,China)
出处
《推进技术》
EI
CAS
CSCD
北大核心
2022年第2期289-298,共10页
Journal of Propulsion Technology
基金
中国民航大学科研启动基金项目(2020KYQD76)。
关键词
航空发动机
滚动轴承
故障预警
核主成分分析
梯度提升决策树
Aero-engine
Rolling bearing
Fault early warning
Kernel principal component analysis
Gradient boosting decision tree
