基于簇特征加权的航空发动机状态监视方法  

Aero-engine condition monitoring method based on cluster features weighting

作　　者：周媛[1,2] 左洪福[1] 王丽娜[2] 

机构地区：[1]南京航空航天大学民航学院,南京210016 [2]南京信息工程大学电子与信息工程学院,南京210044

出　　处：《航空动力学报》2015年第7期1759-1765,共7页Journal of Aerospace Power

基　　金：国家自然科学基金委员会与中国民用航空总局联合项目(60939003);国家自然科学基金(61079013);江苏省自然科学基金(BK2011737)

摘　　要：提出了基于簇特征加权模糊C-均值聚类算法(FWFCM)的航空发动机状态监视模型,该模型主要分为离线学习和在线监视两个部分,离线学习模块计算出模型参数输出到在线监视模块,在线监视模块根据模型参数对实时数据进行分类,实时数据又输入到离线学习模块中参与更新模型参数.结果表明:相比基于数据加权策略的模糊聚类算法(DWFCM)以及经典模糊C-均值聚类算法(FCM),该方法平均离线状态识别率和在线状态识别率分别提高了5.233%和8.358%.实验证明此方法性能好且有很好的鲁棒性和泛化能力,对于不确定性的航空发动机在线状态监视有较好的应用价值.A model for aeroengine condition monitoring based on fuzzy C-means clustering algorithm with cluster features weighting（FWFCM）was proposed.This model is composed of two parts：one is offline-learning module,in which the module parameters are iteratively computed,and the other is online-monitoring module,in which the realtime data can be classified according to the parameters.Then the realtime data are inputed into the offlinemodule and the module parameters are updated.Result shows that,compared with the data weighted fuzzy clustering algorithm（DWFCM）and the classic fuzzy C-means clustering algorithm（FCM）,the average condition recognition accuracy of the proposed method in offline module and in online module is 5.233% and 8.358% higher than the other two algorithms,respectively.It can be shown that the FWFCM works well for aeroengine condition monitoring with good rebustness and generalization,and has practical application value for uncertained aeroengine condition monitoring.

关 键 词：状态监视 模糊聚类 簇特征加权 鲁棒性 泛化能力 

分 类 号：V37[航空宇航科学与技术—航空宇航推进理论与工程] TP391[自动化与计算机技术—计算机应用技术]