Comparative Analysis Between Batch Monitoring Strategies. II. MICA vs. OSS

ALVAREZ MEDINA CARLOS RODRIGO; ADRIANA BRANDOLÍN; MABEL SÁNCHEZ

Rio Gallegos – Argentina

Congreso; XII Reunión de Procesamiento de la Información y Control; 2007

Universidad Nacional de la Patagonia Austral

A new statistical method that uses Independent Component Analysis (ICA) has been recently developed for monitoring batch processes. ICA seeks underlying non-Gaussian and mutually independent factors or components from multivariate statistical data. As in-control data of batch processes present this feature, it is reported that the recently developed Multiway Independent Component Analysis technique  (MICA) provides more meaningful information for process monitoring than other strategies, which work in latent variable spaces. Another monitoring approach consists in the detection of the out of control status, in the original variable space, by means of the D statistic and, the identification of suspicious variables by calculating variable contributions to the inflated statistic. A straightforward procedure to decompose the D statistic as a unique sum of each variable contribution was recently developed by Alvarez et al. (2006b). It provides an explanation about the physical meaning of the negative contributions and estimates a bound for them. In this work, detection and identification capabilities of the aforementioned techniques are compared for the monitoring of an emulsion polymerization reactor.