Una combinación de KICA y filtrado EWMA-ED para la detección de fallos de pequeña magnitud en procesos químicos.

The current industry requires of the fault diagnostic systems with a detection stage that can ensure a small number of false alarms and a quick detection of faults, regardless of their nature and magnitude. However, a common problem of the approaches based on the Multivariate Statistical Process Monitoring (MSPM) which are currently used for this purpose is the difficulty of them to detect small magnitude faults without incurring a high number of false alarms, and faults not detected. To overcome this drawback, in this paper is proposed a novel approach that combines the advantages of kernel methods, and a novel EWMA with enhanced dynamic to filter the T² Hotelling statistic used as the fault detection mechanism in a complex chemical process. [ABSTRACT FROM AUTHOR]