Multivariate quality control of batch processes using STATIS

Danilo Marcondes Filho
Statistical Department, Federal University of Rio Grande do Sul. Av. Bento Gonçalves, 9500 – Porto Alegre, RS 91509-900 , Brazil. Tel: +55 (51) 3308-6225 , Fax: +55 (51)3308-7301 . Email: marcondes.danilo@gmail.com

 
Luis Paulo Luna de Oliveira
PPCA, University of Vale do Rio dos Sinos. Av. Unisinos 950, São Leopoldo, RS, Brazil. Tel: +55 (51). Email: lunadeoliveira@gmail.com

This paper presents the description of the STATIS method and its application in a field of Statistical Process Control. STATIS (from French: Structuration des Tableaux A Trois Indices de la Statistique) is an alternative technique for data dimensionality reduction suitable for application in large multivariate datasets. Industrial batch processes are commonly used in the production of a variety of items. Data emerging from such processes present a peculiar structure, and a number of customized multivariate control charts have been proposed for their monitoring. We propose control charts based on STATIS method, an exploratory technique for measuring similarities between data matrices. Data are arranged in such a way that the monitoring along time is prioritized. The methodology easily allows a nonparametric on-line monitoring of complex batch processes in time, where a large number of variables are present. Also, with STATIS as the theoretical basis for the proposed control chart, special situations may be handled, such as when batches present varying durations. In addition, Dual STATIS, an alternative analysis belonging to the STATIS toolbox, could be used for diagnosing disturbances in the process. Besides its presentation, the new approach is illustrated using simulated data. The proposed technique can be of large application in chemical, biochemical, pharmaceutical and food industries, given their usual degree of complexity.

Keywords: Statistical Process Control, Batch processes, Control charts, STATIS method.

2015 QPRC