A novel plant-wide process monitoring framework based on distributed Gap-SVDD with adaptive radius

摘要

With the increasing complexity of modern process industries, plant-wide process monitoring has become a challenging issue. In this paper, a novel monitoring framework based on distributed gap support vector data description (Gap-SVDD) with adaptive radius is proposed for plant-wide processes. Firstly, the plant-wide processes are divided into different subblocks by the mixed similarity measure for handling the heavy coupling process variables. Afterwards, gap metric is introduced as a kind of data preprocessing method, which is combined with SVDD to set up the monitoring model. The feature extraction of Gap-SVDD is more accurate than conventional methods. Finally, an adaptive radius is developed for Gap-SVDD. It is calculated by a modified univariate statistical method with a limited window length. The purpose of this strategy is to enhance the performance of Gap-SVDD. The superiority of the proposed framework is demonstrated by the revised Tennessee Eastman (TE) benchmark. Comparisons with other conventional methods are also provided. (C) 2019 Elsevier B.V. All rights reserved.