Predictive Modeling for Degree of Substitution of Cellulose Acetate using Infrared Spectroscopy and Machine Learning

Abstract

The objective of this study is to apply FTIR and machine learning models for the quantitative analysis of the degree of substitution of cellulose acetate. The models used for the degree of substitution analysis include PCA (principal component analysis), PLS-DA (partial least squares discriminant analysis), SVM (support vector machine), and KNN (k-nearest neighbor). The critical findings of this study indicated that it is possible to analyze the degree of substitution for cellulose acetate with a degree of substitution of 2.0 or less using IR spectrum data derived from acetylation, estimated through PCA. The decrease in explanatory power for degrees of substitution higher than 2.0 can be attributed to the chemical reaction rate. However, by applying SVM and utilizing the kernel trick to project the data into a high-dimensional feature space and perform non-linear classification, it was possible to create a degree of substitution discrimination model with excellent performance, regardless of the degree of substitution. As a result, the model for analyzing the degree of substitution of polymer monomers based on machine learning and IR spectrum data was proposed. It is believed that this model can efficiently replace existing analytical methods.