Yoo C K, Lee J M, Lee I B, Vanrolleghem P A
BIOMATH, Ghent University, Coupure Links 653, B-9000 Gent, Belgium.
Water Sci Technol. 2004;50(11):163-71.
This paper proposes a new process monitoring method using dynamic independent component analysis (ICA), ICA is a recently developed technique to extract the hidden factors that underlie sets of measurements, whereas principal component analysis (PCA) is a dimensionality reduction technique in terms of capturing the variance of the data. Its goal is to find a linear representation of non-Gaussian data so that the components are statistically independent. PCA aims at finding PCs that are uncorrelated and are linear combinations of the observed variables, while ICA is designed to separate the ICs that are independent and constitute the observed variables. The dynamic ICA monitoring method is applying ICA to the augmenting matrix with time-lagged variables. The dynamic monitoring method was applied to detect and monitor disturbances in a full-scale biological wastewater treatment (WWTP), which is characterized by a variety of dynamic and non-Gaussian characteristics. The dynamic ICA method showed more powerful monitoring performance on a WWTP application than the dynamic PCA method since it can extract source signals which are independent of time and cross-correlation of variables.
本文提出了一种使用动态独立成分分析（ICA）的新过程监测方法。ICA是最近开发的一种技术，用于提取测量集背后的隐藏因素，而主成分分析（PCA）是一种在捕获数据方差方面的降维技术。其目标是找到非高斯数据的线性表示，以使各成分在统计上相互独立。PCA旨在找到不相关且是观测变量线性组合的主成分，而ICA旨在分离出独立且构成观测变量的独立成分（IC）。动态ICA监测方法是将ICA应用于带有时间滞后变量的增广矩阵。该动态监测方法被应用于检测和监测全尺寸生物废水处理厂（WWTP）中的干扰，该处理厂具有各种动态和非高斯特征。动态ICA方法在WWTP应用中显示出比动态PCA方法更强大的监测性能，因为它可以提取与时间无关且变量间互相关的源信号。
