On-Line Control of a Semibatch Emulsion Polymerization Reactor Based on Calorimetry z Isabel Saenz de Buruaga, Antonio Echevarha, Philip D. Armitage, JosC C. de la Cal, JosC R. Leiza, and JosC M. Asua zyxwv Grupo de Ingenieria Quimica, Dept. de Quimica Aplicada, Facultad de Ciencias Quimicas, Universidad del Pais Vasco, Apdo. 1072, 20080 San SebastiBn, Spain zyxw In this work, on-line calorimetry was used to estimate the conversion and the copoly- mer composition in emulsionpolymerizationsystems. Real-timefeedback control of the polymer composition during the semibatch emulsion copolymerization of vinyl acetate and butyl acrylate was cam'ed out. Both homogeneouscopolymersand copolymers with previous defined composition profiles were obtained. It was shown that the feedback control was able to avoid monomer accumulation in the reactor, and hence potentially dangerous thermal runaways, without any deleterious effect on the polymer composi- tion, when a sudden inhibition was caused by deliberately adding a solution of hydro- quinone. The use of feedback controlfor the maximization of the production rate under safe conditions in the unseeded emulsion copolymerizationof vinyl acetate and VeoValO (an alkyl vinyl ester from Shell) is also presented. Introduction Emulsion polymerization is currently the predominant process used in industry to produce a great variety of.poly- mers with many uses (paints, adhesives, coatings, binders) that only a few years ago were prepared by other polymerization processes. The importance of the emulsion polymerization process is that, because of its multiphase and compartmental- ized nature, it offers the possibility of preparing polymers with unique properties that cannot be produced by other polymer- ization techniques. In addition, environmental regulations have led to the substitution of solvent-based polymers by wa- terborne latexes, increasing the importance of emulsion poly- merization. As a consequence, more and more products are prepared by emulsion polymerization, and the quality and properties of such products are a matter of great importance. Therefore, considerable effort has been devoted to control- ling the properties of polymers produced by emulsion poly- merization processes. However, developments in the moni- toring and control of emulsion polymerization processes have been even slower than those in other polymerization pro- cesses where homogeneous reactions are carried out. The main difficulty encountered is the lack of robust, on-line sen- sors to measure the properties of the polymers. Correspondence concerning this article should be addressed to J. M. Asua. At first sight, it may seem that because of the relatively low viscosity of polymeric emulsions, the development of on-line sensors for emulsion polymerization is easier than for bulk and solution processes. However, on-line sensors for emul- sion polymerization are more difficult to develop, due to the thermodynamically unstable nature of the latex particles (Chien and Penlidis, 1990). Consequently, most of the control strategies developed in the last two decades have been open- loop, based on mathematical models of the process or on ex- tensive experimental work (Broadhead et al., 1985; Hamielec et al., 1987; Arzamendi and Asua, 1989, 1990, 1991; Arza- mendi et al., 1991; Leiza et al., 1993a; Van Doremaele et al., 1992; Schoonbrood et al., 1993; Canu et al., 1994; Gugliotta et al., 1995b). On the other hand, few closed-loop strategies using on-line measurements have been reported in the litera- ture (Dimitratos, 1989; Dimitratos et al., 1989; Guyot et al., 1981; Kozub and MacGregor, 1992; Soroush and Kravaris, 1992; Leiza et al., 1993; Urretabizkaia et al., 1994). Gas chromatography (GC) has been successfully applied to on-line measurement and control of copolymer composition (Guyot et al., 1981). This measurement technique was com- bined with an extended Kalman filter (Leiza et al., L993b) to estimate monomer concentrations and calculate the required monomer feed rates at low-to-medium latex solids contents z ( < 35%). Urretabizkaia et al. (1994) used a nonlinear adap- AIChE Journal April zyxwvu 1997 Vol. 43, No. 4 1069