Abstract: An analytical method based on fiber optic near infrared spectroscopy (NIR) and partial least squares data analysis (PLS) was developed to monitor a reaction in which a methyl ester is transesterified with polyethylene glycol 300 (PEG 300) to give the corresponding PEG 300 ester and methanol. A series of multivariant modeling techniques which predict the transesterification reaction end point by monitoring the PEG 300 content were evaluated. The models were developed by making a series of transesterification reactions, collecting grab samples, analyzing them by using the NIR spectrophotometer and liquid chromatography (HPLC) to construct a training and validation set consisting of 50 data points. The models were evaluated in the laboratory setting and the PLS model was used in a pilot plant to predict the reaction end point. Use of the NIR modeling system in the pilot plant led to reduced reaction times and a concomitant rise in product quality. A purged NEMA enclosure was manufactured and was used to house the NIR instrumentation in the pilot plant setting.!1
展开▼