REMOTE IN-LINE MONITORING OF EMULSION POLYMERIZATION OF STYRENE BY SHORT-WAVELENGTH NEAR-INFRARED SPECTROSCOPY .1. PERFORMANCE DURING NORMAL RUNS

Citation
Cc. Wu et al., REMOTE IN-LINE MONITORING OF EMULSION POLYMERIZATION OF STYRENE BY SHORT-WAVELENGTH NEAR-INFRARED SPECTROSCOPY .1. PERFORMANCE DURING NORMAL RUNS, Process control and quality, 8(1), 1996, pp. 1-23
Citations number
34
Categorie Soggetti
Instument & Instrumentation","Engineering, Chemical
Journal title
ISSN journal
09243089
Volume
8
Issue
1
Year of publication
1996
Pages
1 - 23
Database
ISI
SICI code
0924-3089(1996)8:1<1:RIMOEP>2.0.ZU;2-O
Abstract
Near-infrared (NIR) spectroscopy in the wavelength range 700-1100 nm i s used with an in-line fiber-optic diffuse reflectance probe to monito r the emulsion polymerization of styrene. A reactor and spectroscopic monitoring system is constructed that provides computer control of the spectrometer, delivery of the monomer, temperature and stir rate. Bas ed on comparison with the spectra of pure components, the diffuse refl ectance spectra obtained exhibit bands that can be assigned to water, styrene monomer and polystyrene. However, the original zero derivative spectra are distorted and suffer large variable offsets due to the ef fects of light scattering. By application of the second derivative tra nsformation, improved results are obtained and a good correlation is f ound between the NTR spectra and percentage of solid polymer as measur ed by the gravimetric method. Predictions between succeeding runs resu lt in relative standard deviations of 4.5% for linear regression at 88 0 nm, and 3.2% for two-latent-variable principal component regression and partial least squares, As a further check on the value of NIR spec tra for emulsion process monitoring, an attempt was made to correlate spectral features with the concentrations of the three major chemical species in the reactor. Since reference analytical methods were not av ailable for these components, a novel alternative calibration method w as developed. The spectra were correlated to the concentrations of the species as predicted by a computer simulation of the process. The cor relation between model predictions and spectra proved quite good, resu lting in NIR methods for monomer, polymer and water with standard erro rs of prediction of 7-27%, 2-7% and 0.2-0.6%, respectively.