聚类HMM模型在QAR数据分析中的应用研究
|
摘要
快速存取记录器QAR(Quick Access Recorder)数据是飞机飞行过程中从传感器获取的流数据。面对大量的QAR数据,提出一种基于聚类的HMM模型。针对QAR数据特点,分析发生故障或异常时QAR数据中不同属性的变化特点,提取主要影响属性进行分析。通过对其聚类进行数据离散化,得出数据的状态趋势,即将其分为多个状态趋势。对故障或异常发生的过程进行HMM建模,以状态序列的形式描述故障或异常发生的过程,并以飞机空中颠簸故障为例,建立空中颠簸相关QAR数据的HMM模型,并检验了该模型的有效性。
QAR data is the streaming data obtained from the sensor during the flight. Facing with the huge QAR data,a clustering-based HMM model is proposed. According to the characteristics of QAR data, the changing characteristics of different attributes of QAR data during malfunction or abnormality are analysed and the main impacted attributes are extracted. The state trend of the data is obtained from the data discretization of its clustering,which means,the cluster is divided into multiple state trends and the HMM model is constructed during malfunction or abnormality which the malfunction or abnormality is described in the form of status switch. Also,the HMM model of related QAR data of air jolt is constructed to verify the effectiveness of the proposed model with the example of aircraft malfunction in air jolt.
QAR data is the streaming data obtained from the sensor during the flight. Facing with the huge QAR data,a clustering-based HMM model is proposed. According to the characteristics of QAR data, the changing characteristics of different attributes of QAR data during malfunction or abnormality are analysed and the main impacted attributes are extracted. The state trend of the data is obtained from the data discretization of its clustering,which means,the cluster is divided into multiple state trends and the HMM model is constructed during malfunction or abnormality which the malfunction or abnormality is described in the form of status switch. Also,the HMM model of related QAR data of air jolt is constructed to verify the effectiveness of the proposed model with the example of aircraft malfunction in air jolt.
引文
[1]曹惠玲,周百政.QAR数据在航空发动机监控中的应用研究[J].中国民航大学学报,2010,28(3):15-19.
[2]Jiang H,Li X,Liu C.Large margin hidden Markov models for speech recognition[J].IEEE Transactions on Audio Speech&Language Processing,2005,14(5):1584-1595.
[3]黄晓彬,王春峰,房振明,等.基于隐马尔科夫模型的中国股票信息探测[J].系统工程理论与实践,2012,32(4):713-720.
[4]张璇,周峰.隐马尔科夫模型应用领域、热点及趋势分析---基于Cite Space II[J].现代商贸工业,2015,36(15):63-65.
[5]于天剑,陈特放,陈雅婷,等.HMM在电机轴承上的故障诊断[J].哈尔滨工业大学学报,2016,48(2):184-188.
[6]柳姣姣,禹素萍,吴波,等.基于隐马尔科夫模型的时空序列预测方法[J].微型机与应用,2016,35(1):74-76.
[7]廖俊,于雷,罗寰,等.基于趋势转折点的时间序列分段线性表示[J].计算机工程与应用,2010,46(30):50-53.
[8]章登义,欧阳黜霏,吴文李.针对时间序列多步预测的聚类隐马尔科夫模型[J].电子学报,2014(12):2359-2364.
[9]杨慧,王光霞.基于重要点的飞机发动机异常子序列检测[J].计算机工程与设计,2016,37(9):2543-2547.
[10]钱江波,董逸生.一种基于广度优先搜索邻居的聚类算法[J].东南大学学报(自然科学版),2004,34(1):109-112.
[11]倪巍伟,陆介平,陈耿,等.基于k均值分区的数据流离群点检测算法[J].计算机研究与发展,2006,43(9):1639-1643.
[12]任江涛,何武,印鉴,等.一种时间序列快速分段及符号化方法[J].计算机科学,2005,32(9):166-169.
[2]Jiang H,Li X,Liu C.Large margin hidden Markov models for speech recognition[J].IEEE Transactions on Audio Speech&Language Processing,2005,14(5):1584-1595.
[3]黄晓彬,王春峰,房振明,等.基于隐马尔科夫模型的中国股票信息探测[J].系统工程理论与实践,2012,32(4):713-720.
[4]张璇,周峰.隐马尔科夫模型应用领域、热点及趋势分析---基于Cite Space II[J].现代商贸工业,2015,36(15):63-65.
[5]于天剑,陈特放,陈雅婷,等.HMM在电机轴承上的故障诊断[J].哈尔滨工业大学学报,2016,48(2):184-188.
[6]柳姣姣,禹素萍,吴波,等.基于隐马尔科夫模型的时空序列预测方法[J].微型机与应用,2016,35(1):74-76.
[7]廖俊,于雷,罗寰,等.基于趋势转折点的时间序列分段线性表示[J].计算机工程与应用,2010,46(30):50-53.
[8]章登义,欧阳黜霏,吴文李.针对时间序列多步预测的聚类隐马尔科夫模型[J].电子学报,2014(12):2359-2364.
[9]杨慧,王光霞.基于重要点的飞机发动机异常子序列检测[J].计算机工程与设计,2016,37(9):2543-2547.
[10]钱江波,董逸生.一种基于广度优先搜索邻居的聚类算法[J].东南大学学报(自然科学版),2004,34(1):109-112.
[11]倪巍伟,陆介平,陈耿,等.基于k均值分区的数据流离群点检测算法[J].计算机研究与发展,2006,43(9):1639-1643.
[12]任江涛,何武,印鉴,等.一种时间序列快速分段及符号化方法[J].计算机科学,2005,32(9):166-169.