CLASSIFICATION OF CHEMICALLY-MODIFIED CELLULOSES USING A NEAR-INFRARED SPECTROMETER AND SOFT INDEPENDENT MODELING OF CLASS ANALOGIES

A method for classification of eleven chemically modified celluloses h as been developed with the use of near-infrared (NIR) spectroscopy and soft independent modeling of class analogies (SIMCA). The sample set consisted of 440 different batches from eleven different cellulose der ivatives. A full factorial design in temperature and moisture was made for one sample from each class in order to introduce climate variatio ns in the calibration sample set. Principal components analysis (PCA) models were made for each class, and samples not present in the calibr ation set were classified according to the SIMCA method. Only one type II error (acceptance of an unacceptable sample) was detected in the c lassification of the different celluloses. The number of type I errors (rejection of an acceptable sample) ranged from 0 to 14%. Subgroups, due to different manufacturers, viscosities, particle sizes, and degre es of substitution, were detected and correctly classified. The sample presentation, focus of the instrument, number of reference measuremen ts, depth of penetration, and selection of training set samples are di scussed.