上海交通大学学报(自然版) ›› 2015, Vol. 49 ›› Issue (06): 842-848.

• 自动化技术、计算机技术 • 上一篇    下一篇

基于多元统计分析的故障检测方法

纪洪泉,何潇,周东华   

  1. (清华大学 自动化系, 清华信息科学与技术国家实验室, 北京 100084)
  • 收稿日期:2015-01-15 出版日期:2015-06-29 发布日期:2015-06-29
  • 基金资助:

    国家自然科学基金资助项目(61490701,61210012,61290324,61473163)

Fault Detection Techniques Based on Multivariate Statistical Analysis

JI Hongquan,HE Xiao,ZHOU Donghua   

  1. (Department of Automation, TNList, Tsinghua University, Beijing 100084, China)
  • Received:2015-01-15 Online:2015-06-29 Published:2015-06-29

摘要:

摘要:  作为数据驱动故障检测方法中的重要分支,基于多元统计分析的故障检测方法主要包括主元分析、偏最小二乘、独立元素分析和费舍尔判别分析.本文回顾了上述几种方法,包括数据模型、故障检测的原理及方法优劣.仿真实验说明了几种方法的特性及其故障检测的效果,并探讨了基于数据故障检测方法中的一些问题.

关键词:  , 多元统计分析, 主元分析, 偏最小二乘, 独立元素分析, 费舍尔判别分析

Abstract:

Abstract: As an important branch of data-driven fault detection methods, multivariate statistical analysis-based fault detection methods mainly include principal component analysis, partial least squares, independent component analysis and fisher discriminant analysis. In this paper, the data model and fault detection mechanism of each method mentioned above were reviewed. Several properties of these methods were revealed intuitively using simulation results, and their fault detection abilities were illustrated. Finally, several problems related to data-driven fault detection methods were discussed.

Key words: multivariate statistical analysis, principal component analysis, partial least squares, independent component analysis, Fisher discriminant analysis

中图分类号: