基于自适应独立成分分析的化工过程监测

针对独立成分分析（independent component analysis，ICA）方法应用于过程监测时如何选择独立成分（independent component，IC）的问题，提出了自适应独立成分分析（adaptation independent component analysis，AICA）方法。AICA方法首先利用分离矩阵建立关联矩阵，该关联矩阵表示IC的相似性，同时通过核密度估计选择概率密度最小的IC作为特殊独立成分（particular independent component，PIC），再通过关联矩阵选择与PIC具有相似变异特征的几个普通独立成分（common independent components，CICs），最后将PIC与CICs用于构造监测统计量。将AICA方法应用于田纳西-伊士曼（Tennessee Eastman，TE）仿真过程，测试结果表明所提方法对于过程监测是有效的。

Chemical process monitoring based on adaptive independent component analysis

In an effort to tackle the problem of how to select independent components (IC) in process monitoring when using independent component analysis (ICA), this paper proposes an adaptation independent component analysis (AICA) method. The AICA method first establishes a correlation matrix by using the separation matrix. The correlation matrix represents the similarities in IC. At the same time, the minimum probability density IC is selected as the particular independent component (PIC) by estimating the nuclear density. Then, by means of the correlation matrix, several common independent components (CICs) with similar variation characteristics to PIC are selected. Finally, PIC and CICs are used to construct monitoring statistics. The AICA method was applied in simulations of the Tennessee Eastman (TE) process. The test results show that the proposed method is an effective means of process monitoring.