Transflection Near-infrared Spectroscopy Combined with Machine Learning for Mechanical Stability Time Evaluation in Concentrated Rubber Latex

Abstract

This study aims to apply near-infrared spectroscopy (NIRS) in transflection mode combined with a machine learning approach to evaluate the mechanical stability time (MST) in Para concentrated rubber latex. Four supervised learning algorithms, including principal component regression (PCR), partial least squares regression (PLSR), support vector regression (SVR) and random forest regression (RFR), were employed to relate the NIR spectra with the MST degree of the latex samples. A comparison of predictive performance among these different algorithms was performed. The RFR model exhibited the best fitting performance with a coefficient of determination for calibration (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>) and root mean square error of calibration (RMSEC) of 0.95 and 37 seconds, respectively. In addition, the RFR-based model outperformed all others with its predictive performance, presenting coefficient of determination for prediction (r<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>) and root mean square error of prediction (RMSEP) of 0.64 and 91 seconds, respectively. Based on these results, this study could imply that the relationship between the NIR spectra and the change in the MST degree of the samples tends to be nonlinear.