Premium
Multimode non‐Gaussian process monitoring based on local entropy independent component analysis
Author(s) -
Zhong Na,
Deng Xiaogang
Publication year - 2017
Publication title -
the canadian journal of chemical engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.404
H-Index - 67
eISSN - 1939-019X
pISSN - 0008-4034
DOI - 10.1002/cjce.22651
Subject(s) - independent component analysis , entropy (arrow of time) , multi mode optical fiber , gaussian process , computer science , gaussian , fault detection and isolation , principal component analysis , transfer entropy , data mining , entropy estimation , differential entropy , principle of maximum entropy , algorithm , pattern recognition (psychology) , maximum entropy spectral estimation , artificial intelligence , mathematics , statistics , physics , optical fiber , telecommunications , quantum mechanics , actuator , estimator
Abstract Traditionally, independent component analysis (ICA) as a multivariate statistical process monitoring (MSPM) method has attracted considerable attention due to its excellent ability in analysis of non‐Gaussian datasets. However, it may degrade fault detection performance for multimode operating process because of its assumption of one single steady mode. In order to supervise the non‐Gaussian process with multiple steady modes more effectively, this paper proposes a process monitoring method based on local entropy independent component analysis (LEICA). This method applies local probability density estimation to remove the effects of multimode characteristics. Furthermore, information entropy theory is used to extract the feature information of process data by calculating their local information entropies. Based on these local entropy data, ICA is applied to establish the local entropy component model for fault detection. Lastly, a numerical example and the Tennessee Eastman (TE) process are used to verify the proposed method and the results demonstrate the superiority of LEICA method.