A neighbour in control technique for batch process monitoring

ALVAREZ RODRIGO; BRANDOLIN ADRIANA; SANCHEZ MABEL

Salvador-Bahia-Brazil

Simposio; 10th International Symposium on Process Systems Engineering (PSE 2009); 2009

Universidade de Sao Pablo

In this work a Nearest in Control Neighbour (NICN) based strategy is presented for monitoring batch processes. The technique detects a fault if the T2 statistic value exceeds the critical one for three successive observations. In this case, the identification of suspicious measurements is determined in terms of the distance between the current observation and its NICN. Furthermore a comparison between the NICN approach and the classical Multiway Principal Component Analysis (MPCA) is provided. Both techniques are applied to supervise the operation of a semibatch reactor for methyl-methacrylate polymerization. The analysis of results reveals that NICN strategy is suitable for fault detection of batch processes and has the advantage of avoiding ambiguous identifications when compared with classical techniques.