Multivariate statistical process control. Process monitoring methods and applications. (Q714895)

The aim of this book is to present an actual panorama of the statistical monitoring methods applied to industrial processes. The book focuses on data-driven methods, i.e. the methods that consider as input a set of data to extract the relevant information. These methods are an alternative to model-driven or knowledge-driven methods that make use of exact process models of the industrial process or expert knowledge. The book starts with the presentation of well established methods like Independent Component Analysis (ICA), Principal Component Analysis (PCA), Factor Analysis (FA), etc. These methods are the main building blocks of refined techniques presented along the book, the scope of each methods is well described. Methods that combine these building blocks are discussed and evaluated. For example, PCA and ICA are used in tandem in order to separate the Gaussian and non-Gaussian data information that are subsequently analyzed with appropriate methods. Extensions of these conventional process monitoring methods are presented that consider the time evolution or the multiple evolution modes of the monitored systems. Adaptive, recursive, moving-window, multimodel or local model approach are presented to monitor dynamical and nonlinear processes. The presentation of the book makes it ready to use for an audience already aware of the vocabulary and main techniques of statistical analysis. Such a reader will find a quick access to recent methods. Besides the presentation of methods, the books contains a wealth of examples and benchmarks that are very valuable for estimating the quality of the methods as well as for supporting further researches in the area. The reader not acquainted with statistical methods will probably need some more resources to learn the subject. Basic techniques are not described. However, the book contains many pointer to the relevant literature that are useful to fill the gaps.