电子学报 ›› 2012, Vol. 40 ›› Issue (4): 734-738.DOI: 10.3969/j.issn.0372-2112.2012.04.018

• 学术论文 • 上一篇    下一篇

基于动态权值的多分类器故障诊断系统

黄江涛1,2,3, 王明辉3, 李武劲3, 古博3   

  1. 1. 广西师范学院计算机与信息工程学院,广西南宁 530023;2. 科学计算与智能信息处理 广西高校重点实验室,广西南宁 530023;3. 四川大学计算机学院,四川成都 610065
  • 收稿日期:2010-06-09 修回日期:2011-11-18 出版日期:2012-04-25
    • 基金资助:
    • 国家自然科学基金 (No.61071162)

Multiple Classifier Fault Diagnosis System Based on Dynamic Weight

HUANG Jiang-tao1,2,3, WANG Ming-hui3, LI Wu-jing3, GU Bo3   

  1. 1. College of Computer and Information Engineering,Guangxi Teachers Education University,Nanning,Guangxi 530023,China;2. Key Lab of Scientific Computing and Intelligent Information Processing in Universities of Guangxi,Nanning,Guangxi 530023,China;3. College of Computer Science,Sichuan University,Chengdu,Sichuan 610065,China
  • Received:2010-06-09 Revised:2011-11-18 Online:2012-04-25 Published:2012-04-25

摘要: 为提高动态系统故障诊断的精确性,以及减少系统运行环境对故障诊断带来的影响,本文提出了一种基于动态权值的多分类器故障诊断系统.该方法使用决策支持度来衡量当前诊断任务中各分类器的实时决策可信度,并将其联合分类器性能指标动态地为各分类器赋予融合权值,决策性能好且决策支持度高的分类器决策结果获得较大的融合权值,同时,使不可靠决策结果的融合权值趋近于零.在此基础上,将多分类器系统优化为实时性能较好的分类器组成的子系统进行故障诊断,减少了不可靠决策的干扰,进一步提高了融合决策的精确度.试验表明本文方法具有良好的诊断决策性能,能获得比单个分类器和常用的一些融合算法更高的分类准确度.

关键词: 故障诊断, 多分类器系统, 数据融合, 决策支持度, 动态权值

Abstract: In order to improve the accuracy of fault diagnosis for dynamic system,and reduce the diagnosis influence from operating environment,a new fusion method in multiple classifier system based on dynamic weight is proposed.The new approach dynamic assigns weights to base classifiers according to their classification accuracy and decision support value.Bigger weights are assigned to more reliable decision output,and the weights of unreliable outputs are close to zero.In this sense,a subsystem is used to make final decision instead of system.Experimental results demonstrate that the new fusion method can get good fault diagnosis performance,and it can get higher accuracy than single classifier and some common used fusion methods.

Key words: fault diagnosis, multiple classifier system, data fusion, decision support value, dynamic weight

中图分类号: