Fault detection and diagnosis in industrial systems (Q5926461)

This textbook presents the theoretical background and practical techniques for process monitoring. The intended audience is engineering students and practical engineers. The book is appropriate for a first-year graduate or advanced undergraduate course on data-driven approaches to fault detection and diagnosis. Numerous simple examples and a simulator for a large-scale industrial plant are used to illustrate the methods. The process monitoring techniques can be implemented by using commercial software packages. The first chapter gives an overview of process monitoring procedures and methods. In Part II ``Background'', Chapter 2 provides the prerequisites in multivariable statistics, including univariate control and a discussion of data requirements. Then, Chapter 3 discusses pattern classification, including discriminant analysis and feature extraction, which are fundamental to fault diagnosis techniques. Part III on ``Data-driven methods'', containing Chapters 4-7, cover data-driven process monitoring methods: Principal component analysis, partial least squares, Fisher discriminant analysis, cvnonical variate analysis. In Part IV on ``Application'', with Chapters 8-10, the methods are compared through application to a large-scale system, a plant simulator. Here, also some homework problems can be found. The last Part V on ``Analytical and knowledge-based methods'' describes analytical methods, including parameter estimation, state estimation, and parity relations in Chapter 11. Finally, Chapter 12 describes knowledge-based methods, including causal analysis, expert systems, and pattern recognition. Furthermore, various combinations of process monitoring techniques are discussed. The present textbook allows the reader to compare and contrast the different approaches to the problem of fault detection and diagnosis. Hence, the text can be recommended for advanced courses for process, chemical and control engineers.