Fault Detection and Diagnosis in Industrial Systems

Abstract

An alternative group of methods has emerged which do not require the use of an explicit model. This is the key basic construct for the data-driven paradigm. Model-free and non-parametric methods for fault detection, process optimisation and control design are currently at a particularly exciting stage of development.This new advanced textbook by Chiang, Russell and Braatz primarily tackles the data-driven routes to Fault Detection and Diagnosis. It is an outgrowth of a prior Advances in Industrial Control monograph; Russell, Chiang and Braatz. Data-driven Techniques for Fault Detection and Diagnosis in ChemicalProcesses, 2000, ISBN 1-85233-258-1. The new textbook expands the material of the monograph and gives a fuller presentation of some of the alternative modelbased methods, the analytical methods, and of the knowledge-based techniques. vi Series Editors' ForewordThis allows the reader to compare and contrast the different approaches to the problem of fault detection and diagnosis. Thus the text is suitable for advanced courses for process, chemical and control engineers.

Fault Detection of the Tennessee Eastman Process Using Improved PCA and Neural Classifier

“…PRINCIPAL COMPONENT ANALYSIS PCA is a dimensionality reduction technique CITATION Chi01 \l 1033 [1] . PCA is used to linearly project a matrix of data in a low dimensional space, this space spanned by PCs (i.e.…”

“…eigenvectors corresponding to the large eigenvalues) for the distribution of the training data. PCA determines a set of orthogonal vectors, called loading vector, ordered by the amount of variance explained in the loading vector directions CITATION Chi01 \l 1033 [1] . Let X is the original data set, where each row is a single sample of data set and each column is an observation.…”

“…The PRESS statistic for one group is computed based on various dimensions of the space, i, using all the other groups. This is repeated for each group, and the value i associated with the minimum average PRESS statistic determines the dimension of the space CITATION Chi01 \l 1033 [1] . GA introduced in next section.…”

An approach of classification and parameters estimation, using neural network, for lubricant degradation diagnosis

“…Neural networks can filter noise and disturbance; they can provide a stable diagnostic, failures without traditional types of models, extremely sensitive, and economic efficiency due to insignificant computing and design effort. Another desirable feature of neural networks is that exact patterns are not required to reach the decision stage [1], [2], [4], [5], [7], [8], [11]. In a typical operation, the process model can only be approximate and critical measurements may be capable of internally crunching functional relationships that represent processes, filtering noise, and managing correlations.…”

Data-based fault detection in chemical processes: Managing records with operator intervention and uncertain labels

“…principal component analysis (PCA) (Lu, Yao, Gao, & Wang, 2005), partial least squares (PLS) (Chiang, Braatz, & Russell, 2001). Recently, FD has been considered as a classification problem as well, and Machine Learning provides various tools for classification, which are categorized below (Isermann, 2006):  Geometric classifier.…”