北京航空航天大学学报 ›› 2014, Vol. 40 ›› Issue (10): 1366-1371.doi: 10.13700/j.bh.1001-5965.2013.0656

• 论文 • 上一篇    下一篇

基于模糊关联分类器的民机超限事件诊断方法

高小霞, 霍纬纲, 冯兴杰   

  1. 中国民航大学 计算机科学与技术学院, 天津 300300
  • 收稿日期:2013-11-18 出版日期:2014-10-20 发布日期:2014-10-29
  • 作者简介:高小霞(1980-),女,河北临城人,实验师,xxgao@cauc.edu.cn.
  • 基金资助:

    国家自然科学基金资助项目(61301245,61201414,61202018);国家自然科学基金委员会与中国民用航空局联合基金资助项目(U1233113,61179063);中央高校基本科研业务费中国民航大学专项资助项目(ZXH2012N001)

Civil aircraft's exceedance event diagnosis method based on fuzzy associative classifier

Gao Xiaoxia, Huo Weigang, Feng Xingjie   

  1. School of Computer Science and Technology, Civil Aviation University of China, Tianjin 300300, China
  • Received:2013-11-18 Online:2014-10-20 Published:2014-10-29

摘要:

现有的民用飞机超限事件智能诊断模型大多属于“黑盒”模型,不利于分析超限事件发生的原因.为此提出了一种基于模糊关联分类器(FAC,Fuzzy Associative Classifier)的民用飞机超限事件诊断方法.该方法抽取发生超限事件时对应的QAR(Quick Access Recorder)参数快照取值,采用模糊C均值(FCM,Fuzzy C-Means)聚类算法对抽取的QAR参数取值模糊预处理,然后基于Apriori算法生成模糊关联分类规则库,并由遗传算法对其进行裁剪,结合模糊分类推理方法形成FAC.采用B737-800实际样本数据进行了验证.实验结果表明,所提出的FAC能有效诊断超限事件,FAC识别超限事件的错误率与最小二乘支持向量机(LS-SVM,Least Squares Support Vector Machine)模型相当,但其解释性方面优于LS-SVM.

关键词: 飞行品质监控, 模糊关联分类器, 超限事件, 遗传算法, 诊断模型

Abstract:

Most of current civil aircraft's exceedance event intelligent diagnosis models are "black box" model, which donot contribute to analyze the occurrence of civil aircraft's exceedance event. In order to overcome these shortcomings, a civil aircraft's exceedance event diagnosis method based on fuzzy associative classifier (FAC) was proposed. First, quick access recorder's (QAR's) parameters value when exceedance event occured was extracted.Fuzzy C-means (FCM) cluster algorithm was adopted to preprocess extracted QAR's parameters value. Then, the library of fuzzy associative classification rule (FACR) was generated by Apriori algorithm.Genetic algorithm was used to prune the library of FACR.Finally,fuzzy classification reasoning method was integrated to build FAC. The FAC was verified with sample data generated by B737-800. Experiment results show that the FAC can diagnose exceedance event effectively, and its classification error rate is equivalent to least squares support vector machine (LS-SVM), but its interpretability is superior to LS-SVM.

Key words: flight operations quality assurance, fuzzy associative classifier, exceedance event, genetic algorithm, diagnosis model

中图分类号: 


版权所有 © 《北京航空航天大学学报》编辑部
通讯地址:北京市海淀区学院路37号 北京航空航天大学学报编辑部 邮编:100191 E-mail:jbuaa@buaa.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发