A Novel Feature-Extraction-Based Process Monitoring Method for Multimode Processes With Common Features and Its Applications to a Rolling Process

In this article, a novel feature-extraction-based process monitoring method is proposed for multimode processes with common features. Different from the traditional feature extraction methods that consider either common scores or common weightings between different modes, a common-subspace-based method that takes both common scores and weightings into account is developed based on tensor decomposition. In addition, specific features for each mode are extracted by the independent component analysis. Moreover, a moving-window Kullback-Leibler-divergence-based detection statistic is developed to monitor the changes in both common and specific features. The newly proposed methods are applied to a real hot rolling mill (HRM) process, where common setting for different steel slabs and specific configurations for each steel product exist. The practical application performance shows that the proposed methods can accurately capture common features and effectively monitor different fault cases in an HRM process.

Automated summary

A novel feature-extraction-based process monitoring method is proposed for multimode processes with common features, utilizing tensor decomposition to consider both common and specific features. The practical application performance shows that the proposed method can accurately capture common features and effectively monitor different fault cases.