关键词: FONT-SIZE: 10.5pt, mso-ascii-font-family: Times New Roman, mso-bidi-font-family: Times New Roman, mso-font-kerning: 1.0pt, mso-ansi-language: EN-US, mso-fareast-language: ZH-CN, mso-bidi-language: AR-SA, mso-hansi-font-family: Times New Roman">非高斯；动态特性；主元分析；独立因子分析；过程监控

Abstract:

According to the dynamic characteristics of industrial production process and process variables of Gaussian distribution problems, this paper proposes a new monitoring method based on Dynamic Independent Factor Analysis(DIFA).Firstly, the paper chooses suitable time lags for each variable by auto-correlation analysis to construct the augmented matrix; secondly, compresses the augmented variable matrix by Principal Component Analysis(PCA); at last, computes the parameters by the Expectation Maximization Algorithm, establishes a mixed model of adaptive Gaussian, constructs the monitoring indices including GI² and GSPE and presents the non-parametric algorithm for confidence bounds.This new monitoring method was used in an adsorption separation process and compared with the methods based on Modified Dynamic Factor Analysis(MDFA)and Independent Component Analysis(ICA), the validity and superiority were proved.

Key words: FONT-SIZE: 10.5pt, mso-ascii-font-family: Times New Roman, mso-bidi-font-family: Times New Roman, mso-font-kerning: 1.0pt, mso-ansi-language: EN-US, mso-fareast-language: ZH-CN, mso-bidi-language: AR-SA, mso-hansi-font-family: Times New Roman">非高斯；动态特性；主元分析；独立因子分析；过程监控