Detection and classification of abnormal process situations using multidimensional wavelet domain hidden Markov trees

被引：28

作者：

Bakhtazad, A

Palazoglu, A

Romagnoli, JA ^{[1
]}

机构：

[1] Univ Sydney, Dept Chem Engn, ORICA Lab Proc Syst Engn, Sydney, NSW 2006, Australia

[2] Univ Calif Davis, Dept Chem Engn & Mat Sci, Davis, CA 95616 USA

来源：

COMPUTERS & CHEMICAL ENGINEERING | 2000年 / 24卷 / 2-7期

关键词：

process trend analysis; wavelet transform; hidden Markov model;

D O I：

10.1016/S0098-1354(00)00370-7

中图分类号：

TP39 [计算机的应用];

学科分类号：

081203 ; 0835 ;

摘要：

This paper addresses the detection of abnormal process situations during plant operation via an effective trending strategy. Wavelet-domain hidden Markov models (HMMs) are exploited as a powerful tool for statistical modeling and processing of wavelet coefficients. We focus on the multivariate problem as many variables contribute to the decision regarding process status. A simulation study illustrates the salient features of the proposed framework. (C) 2000 Elsevier Science Ltd. All rights reserved.

引用

页码：769 / 775

页数：7

共 10 条

[1]

BAKHSI BR, 1998, AICHE J, V44, P1596

[2]

BAKHSI BR, 1994, COMPUT CHEM ENG, V18, P267

[3]

Bakhtazad A., 1999, Intelligent Data Analysis, V3, P267, DOI 10.1016/S1088-467X(99)00023-2

[4] Wavelet-based statistical signal processing using hidden Markov models [J].