Multivariate Statistical Monitoring of an Industrial SBR Process. Part II: Fault Detection and Diagnosis

GODOY, J.L.; MINARI, R.J.; VEGA, J.R.; MARCHETTI, J.L.

Rosario

Congreso; RPIC 2009 - XIII Reunión de Trabajo en Procesamiento de la Información y Control; 2009

The production of Styrene-Butadiene Rubber (SBR) in an industrial continuous 7-reactor train operated under steady-state (SS) conditions is investigated. On-line monitoring of quality variables in the SBR process is difficult because of the lack of specific sen-sing devices. In Part I of this work, a soft-sensor for monitoring both process and quality variables was developed. It was based on a partial least squared (PLS) regression, and then it was validated and compared with a previously-developed neural network soft sensor. In this Part II, we present a PLS-based algorithm for the detection, isolation, and characterization of several faults that frequently dis-turb the SS operation of the reaction train. Two sources of errors that produce erroneous SS opera-tions are considered: 1) faults in the mass- or vol-ume-flow sensors; and 2) the presence of unmeas-ured disturbances (e.g., impurities in some reagents). Simulation results show that the proposed algorithm adequately identify the analyzed faults and/or dis-turbances. The results suggest that the proposed methods are powerful tools for monitoring and con-trol the normal SS operation of the SBR process.