New online kernel method with the Tabu search algorithm for process monitoring

Hajer Lahdhiri, Khaoula Ben Abdellafou, Okba Taouali, Majdi Mansouri, Ouajdi Korbaa

Research output: Contribution to journalArticle

4 Citations (Scopus)

Abstract

Process monitoring is an integral part of chemical process, required higher product quality and safety operation. Therefore, the objective of this paper is to ensure the suitable functioning and to improve the fault detection performance of conventional kernel Principal Components Analysis (KPCA). Thus, an online Reduced Rank KPCA (OnRR-KPCA) with adaptive model has been developed to monitor a dynamic nonlinear process. The developed method is proposed. Firstly, to extract the useful observations, from large amount of training data registered in normal operating conditions, in order to construct the reduced reference model. Secondly, to monitor the process online and update the reference model if a new useful observation is available and satisfies the condition of independencies between variables in feature space. To demonstrate the effectiveness of the OnRR-KPCA with adaptive model over the conventional KPCA and the RR-KPCA, the fault detection performances are illustrated through two examples: one using synthetic data, the second using a simulated Tennessee Eastman Process (TEP) data.

Original languageEnglish
Pages (from-to)2687-2698
Number of pages12
JournalTransactions of the Institute of Measurement and Control
Volume41
Issue number10
DOIs
Publication statusPublished - 1 Jun 2019

    Fingerprint

Keywords

  • Online Reduced Rank-KPCA
  • fault detection
  • nonlinear process monitoring

ASJC Scopus subject areas

  • Instrumentation

Cite this