基于重要点的飞机发动机数据异常子序列检测  

Abnormal sub-sequence detection for aircraft-engine data based on important point

作　　者：杨慧[1] 王光霞[1] 

机构地区：[1]中国民航大学计算机科学与技术学院,天津300300

出　　处：《计算机工程与设计》2016年第9期2543-2547,共5页Computer Engineering and Design

基　　金：国家自然科学基金与中国民航联合基金项目(61179063);国家自然科学基金项目(61301245)

摘　　要：为从飞机快速存储记录器(QAR)数据中发现飞机发动机异常数据并预测飞机发动机故障,针对飞机发动机数据量大、结构复杂导致的异常检测复杂以及各个航班的数据量不等等问题,提出一种基于重要点的自适应分段算法。针对不同长度的数据自动生成合适数量的重要点,对QAR数据分段,对分段后的子序列进行模式表示,采用基于局部密度的时间序列异常模式检测算法对飞机发动机数据进行异常检测。实验结果表明,该方法降低了运算复杂度,能够准确发现异常子序列,方便飞机发动机故障诊断和维修。To find the abnormal data of the aircraft-engine quick access recorder（QAR）data and predict problems of aircraft-engine,considering that the aircraft-engine data are time series data with large data volume and complex structure,which is very complex to detect the anomaly in the raw data directly,and the amount of data for each flight is unequal,an adaptive segmentation algorithm based on important points was proposed.For different length of data,a suitable number of important points were automatically generated to split the QAR data,and the model representation of the sub-sequences was made.The anomaly detection of the aircraft-engine data was carried out using the time series anomaly pattern detection algorithm based on local density.Experimental results show that this method can reduce the complexity of the operation and find the abnormal sub-sequence accurately,so that the fault diagnosis and maintenance of aircraft-engine can be easy.

关 键 词：飞机发动机数据 重要点 自适应分段 异常检测 故障诊断 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]