Disturbance detection and isolation by dynamic principal component analysis

Ku, Wenfu; Storer, Robert H.; Georgakis, Christos

doi:10.1016/0169-7439(95)00076-3

Cited by 1,323 publications

(867 citation statements)

References 20 publications

Supporting

Mentioning

843

Contrasting

Unclassified

Order By: Relevance

“…The dynamic characteristic is achieved by introducing time-lagged variables into the data matrices in a similar manner as in time series analysis. The dynamic methods are especially suitable for continuous processes with long time delays and varying throughputs on process variables (Ku, Storer, & Georgakis, 1995;Wold, Sjöström, & Eriksson, 2001). Chen, McAvoy, and Pivoso (1998) proposed a multivariate statistical controller based on dynamic PCA.…”

Section: Fig 1 Categories Of the Diagnostic Methodsmentioning

confidence: 99%

An online application of dynamic PLS to a dearomatization process

Komulainen

Sourander

Jämsä‐Jounela

2004

Computers & Chemical Engineering

View full text Add to dashboard Cite

Early detection of process disturbances and prediction of malfunctions in process equipment improve the safety of the process, minimize the time and resources needed for maintenance, and increase the uniform quality of the products. The objective of online-monitoring is to trace the state of the process and the condition of process equipment in real-time, and to detect faults as early as possible.In this article the different properties of the online-monitoring methods applied in the process industries are first reviewed. A description of the systematic development of the online-monitoring system for an industrial dearomatization process, specifically for flash point and distillation curve analysers, is then presented. Finally, the results of offline and online tests of the monitoring system using real industrial data from the Fortum Naantali Refinery in Finland, are described and discussed. The developed onlinemonitoring application was successful in real-time process monitoring and it fulfilled the industrial requirements. PACS: 07.05.Mh; 07.05.Tp; 83.85.Ns

show abstract

Section: Fig 1 Categories Of the Diagnostic Methodsmentioning

confidence: 99%

An online application of dynamic PLS to a dearomatization process

Komulainen

Sourander

Jämsä‐Jounela

2004

Computers & Chemical Engineering

View full text Add to dashboard Cite

show abstract

“…The associated cross-correlation matrices contain the information for principal components analysis, PCA [15]. Standard matrix decomposition techniques are the basis for a dimension reduction, which helps to focus on the important aspects of a model for the whole system.…”

Section: Time Series and Multi-variable Datamentioning

confidence: 99%

Model selection: An overview of practices in chemical engineering

Verheijen

2003

Computer Aided Chemical Engineering

View full text Add to dashboard Cite

The problem of choosing out of a large set of models is considered here. Five criteria are mentioned: the level of rigor that a model describes phenomena; the accuracy of the model with respect to data; the adequacy of the model for the purpose of simulation, optimization or design; the flexibility of a model; and the computational complexity. The focus is on the purely statistical criteria. A systematic procedure is presented to eliminate sequentially hierarchical models and equivalent models, based on the F-test and Bartlett's χ 2 -test. The use of Bayesian statistics as a tool is reviewed. And recent developments in optimization-based approaches are listed. An overview of available measures for the distance between models is given. It is shown that one should not accept a single model, as is engineering practice, but rather allow for a set of models when the selection does not give sufficient reason to discriminate between models. Examples from the area of reaction kinetics and adsorption are given.

show abstract

“…Some of these approaches are heuristic and rely on the relative magnitude of the eigenvalues to estimate the number of retained principal components. 24,25 Other approaches rely on cross-validation, 26,27 or on modifications of the likelihood function. 25 As shown through illustrative examples in Section 5, the likelihood function increases by retaining more principal components.…”

Section: Estimating the Pca Model Rankmentioning

confidence: 99%

Bayesian principal component analysis

Nounou

Bakshi

Goel

et al. 2002

Journal of Chemometrics

View full text Add to dashboard Cite

Principal component analysis (PCA) is a dimensionality reduction modeling technique that transforms a set of process variables by rotating their axes of representation. Maximum Likelihood PCA (MLPCA) is an extension that accounts for different noise contributions in each variable. Neither PCA nor its extensions utilize external information about the model or data such as the range or distribution of the underlying measurements. Such prior information can be extracted from measured data and can be used to greatly enhance the model accuracy. This paper develops a Bayesian PCA (BPCA) modeling algorithm that improves the accuracy of estimating the parameters and measurements by incorporating prior knowledge about the data and model.The proposed approach integrates modeling and feature extraction by simultaneously solving parameter estimation and data reconciliation optimization problems.Methods for estimating the prior parameters from available data are discussed. Furthermore, BPCA reduces to PCA or MLPCA when a uniform prior is used. Several examples illustrate the benefits of BPCA versus existing methods even when the measurements violate the assumptions about their distribution.

show abstract

Disturbance detection and isolation by dynamic principal component analysis

Cited by 1,323 publications

References 20 publications

An online application of dynamic PLS to a dearomatization process

An online application of dynamic PLS to a dearomatization process

Model selection: An overview of practices in chemical engineering

Bayesian principal component analysis

Contact Info

Product

Resources

About