Non-Negative Matrix Factorization for Detection and Diagnosis of Plantwide Oscillations

详细信息    查看全文

• 作者：Arun K. Tangirala ; Jitendra Kanodia ; Sirish L. Shah
• 刊名：Industrial & Engineering Chemistry Research
• 出版年：2007
• 出版时间：January 31, 2007
• 年：2007
• 卷：46
• 期：3
• 页码：801 - 817
• 全文大小：813K
• 年卷期：v.46,no.3(January 31, 2007)
• ISSN：1520-5045

文摘

In this paper, we propose the use of non-negative matrix factorization (NMF) of multivariate spectra forplantwide oscillation detection. One of the key features of NMF is that it provides a parts-based representationthat allows us to retain the causal basis spectral shapes or parts that constitute the spectra of measurements,unlike the popular principal component analysis (PCA)-based methods. The contributions of this paper are asfollows: (i) a novel measure known as the pseudo-singular value (PSV) to assess the order of the basis space(the PSV is also useful in determining the most dominant features of a data set); (ii) a power decompositionplot that contains the total power (defined in this work) and its decomposition by NMF (the power plot is auseful and compact visual tool that provides overall spectral characteristics of the plant and shows thedecomposition of these characteristics into well-localized frequency components); and (iii) a novel measuredefined as the strength factor (SF) to assess the strength of the localized features in the variables (it can bealso used in isolating the root cause). Finally, it is shown that the proposed implementation of NMF is powerfuland sensitive enough to capture small oscillations in the measurements. As a result, it largely eliminates theneed to filter the data. Industrial case studies are presented to illustrate the applications of NMF and todemonstrate the utility and practicality of the proposed measures.