Alberto Ferrer

The course introduces the student in the problem of the analysis, monitoring and fault diagnosis in modern industrial processes (continuous or batch), highly automated, with a massive amount of process variables collected (every minute, second ...), with control loops and some quality variables collected less frequently and with a certain delay between the time they are registered and the one the product is obtained, frequently, after a laboratory analysis. In addition, limitations that traditional multivariate analysis techniques have in these contexts are discussed, introducing multivariate statistical methods based on projection to latent structures as efficient tools for dealing with these types of problems. Potential applications of these latent-based multivariate statistical techniques to analyse mega databases in other fields of Science and Technology are also discussed.