On-line instrumentation for latex production

Abstract

<p>The availability of fast, reproducible on-line instrumentation plays a critical role in paving the way for improvements in the productivity and product quality of industrial processes. In the polymer latex industry, one of the primary reasons for the lack of adequate instrumentation is that the traditional on-line sensors for density, viscosity, light scattering and absorption seldom respond to only one property; instead, their response is a function of many latex properties. If, however, these sensors respond differently to latex property changes, a novel approach to instrumentation is to use information from several of them to infer several quality variables simultaneously. This thesis applies this concept to two latex systems. The first is the continuous emulsion polymerization of poly(vinyl acetate). A density meter and the single wavelength UV turbidity of a diluted sample are used to track solids fraction and mean particle size on a pilot scale reactor. To sensor-latex property relationships are derived from a mechanistic understanding of the system. The standard deviation of the solids fraction measurement is 0.8% of value, for the mean diameter, it is 3%. The second process is the batch copolymerization of styrene and methyl methacrylate (MMA). Two spectrometers, covering a wavelength range from 190 nm to 1800 nm were used to predict the weight fractions of water, styrene and MMA monomer, poly(styrene), and poly(MMA). The standard deviations of the composition measurements were between 0.2 and 1.2 weight percent. The sensor-latex property relationships were arrived at by an empirical, multivariate calibration. The important implications of these results are that: (1) the number of sensors that could be considered for on-line analyzers is greatly increased; and (2) this methodology may be successful for previously intractable instrumentation problems. Also, potentially any latex property that can be measured in the laboratory can be included in the calibration, provided the property of interest has a reproducible effect on the response of the sensor.</p>