Independent Component Analysis with Application to Hot Galvanizing Pickling Waste Liquor Treatment Process

摘要

In this paper, the application of independent component analysis (ICA) to statistical process monitoring is studied. This paper mainly focuses on studying on the fault detection and isolation principle based on the data model of ICA. Contributions of this paper are: (1) for the purpose of fault detection, two monitoring statistics are designated by detailed analysis on the data model of ICA; (2) a semi-supervised Laplacian regularization (SLR) kernel density estimation approach is proposed to determine the normal operation region; (3) Under fault condition, a fault isolation method based on expert knowledge is introduced; (4) Hot Galvanizing Pickling Waste Liquor Treatment Process (HGPWLTP) is taken to evaluate the monitoring performance of the proposed approaches. Encouraging experimental results of the proposed process monitoring approaches are achieved.