Computer Engineering and Applications ›› 2010, Vol. 46 ›› Issue (8): 245-248.DOI: 10.3778/j.issn.1002-8331.2010.08.071

• 工程与应用 • Previous Articles    

Pattern recognition technology integrating Mahalanobis-Taguchi system and SVM

ZENG Jiang-hui1,ZENG Feng-zhang2,CHEN Song-hui3   

  1. 1.China Aero-Polytechnology Establishment,Beijing 100028,China
    2.School of Management and Economics,Beijing Institute of Technology,Beijing 100081,China
    3.China Aerospace Times Electronics Corporation,Beijing 100070,China
  • Received:2008-09-12 Revised:2008-12-08 Online:2010-03-11 Published:2010-03-11
  • Contact: ZENG Jiang-hui

马田系统与SVM相集成的模式识别技术研究

曾江辉1,曾凤章2,陈嵩辉3   

  1. 1.中国航空综合技术研究所,北京 100028
    2.北京理工大学 管理与经济学院,北京 100081
    3.中国航天时代电子公司,北京 100070
  • 通讯作者: 曾江辉

Abstract: For solving the existing problems of the establishing threshold methods in MTS,applying the theory of hyper sphere support vector machine,the one-class and two-class hyper ellipsoidal support vector machine models are put forward.The two models can be proved theoretically to transform into quadratic programming problems.The function of threshold of MTS based on the hyper ellipsoidal support vector machine is established.At last,a fault diagnosis example is illustrated applying the new method,and high accuracy rating is reached.

Key words: Mahalanobis-Taguchi System(MTS), Support Vector Machine(SVM), Mahalanobis distance, threshold, fault diagnosis

摘要: 为了解决现有马田系统阈值确定方法的不足,基于超球面支持向量机算法,构建了单类及二类超椭球面支持向量机算法,并理论上证明了此算法可转换为二次规划模型。推导了基于超椭球面支持向量机的马田系统阈值确定公式。将所提出的方法应用于故障诊断,得到了较高的判别正确率。

关键词: 马田系统(MTS), 支持向量机(SVM), 马氏距离, 阈值, 故障诊断

CLC Number: