Soft-sensor for Monitoring an Industrial Styrene-Butadiene Rubber Process (SBR)

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

Girona

Congreso; 10º Congreso Interamericano de Computación Aplicada a la Industria de Procesos (CAIP?2011); 2011

This work presents a full methodology to build and evaluate a soft sensor capable of monitoring theproduction of Styrene-Butadiene Rubber in an industrial train of 7 continuously-stirred tankreactors. The aim is to develop a device for on-line estimation of production and quality variablesusing a multivariate statistical technique like partial least squared. Besides pursuing the soft sensordevelopment, this paper attempts to provide a guide for similar developments by suggestingattention to several specific methodological points. Since a wide range validation space is desiredfor this sensor and the actual plant cannot be arbitrarily disturbed, an existing complexfundamental model is used to explore different possible operating conditions. The approach usedto develop the soft sensor includes a distributed sampling of multivariate steady-state conditions tocollect the calibration data set, and the use of a filter for excluding outliers. Then, it is shown thatthe analysis of how the explained variability progress when latent variables are included in themodel allows the detection of poor predictor variables providing the chance for improving themultivariable regression by eliminating interfering contributions.