Statistical process monitoring based on nonlocal and multiple neighborhoods preserving embedding model. (May 2018)

• Record Type: Journal Article
• Authors: Tong, Chudong; Lan, Ting; Shi, Xuhua; Chen, Yuwei
• Is Part Of: Journal of process control. Volume 65(2018)
• Page Start: 34
• Page End: 40
• Publication Date: 2018-05
• Subjects: Neighborhood preserving embedding -- Statistical process monitoring -- Dimensionality reduction -- Fault detection; Process control -- Periodicals; Fabrication -- Contrôle -- Périodiques; Process control; Periodicals; Electronic journals; 660.281
• DOI: 10.1016/j.jprocont.2017.10.009
• Languages: English
• ISSNs: 0959-1524

Abstract

Highlights: A novel dimensionality reduction algorithm named NoMNPE is proposed. Time neighborhood, distance neighborhood, and angle neighborhood are embedded. Case studies demonstrate the priority and advantage of the proposed NoMNPE method. Abstract: A novel dimensionality reduction algorithm named nonlocal and multiple neighborhoods preserving embedding (NoMNPE) is proposed for modeling and monitoring industrial processes. The NoMNPE method implements dimensionality reduction by maximizing the variance scattered by nonlocal data points, while simultaneously preserving multiple neighborhoods relationships, which include time neighbors, distance neighbors, and angle neighbors for a given dataset. Therefore, three different manifold characteristics and one additional nonlocal relationship are taken into account in the NoMNPE model. The NoMNPE thus is expected to explore more intrinsic information in contrast to its counterparts, and could achieve enhanced monitoring performance as a result. The comparison studies on two industrial processes have also demonstrated the effectiveness and advantages of the proposed NoMNPE-based process monitoring approach.