Machine learning classification of THz spectra of commercial rubber

Abstract

Rubber is a widely used industrial material whose performance and durability are strongly influenced by thermal processing, making reliable quality assessment essential. This study investigated machine learning techniques for classifying commercial rubber samples using terahertz time-domain spectroscopy (THz-TDS) absorption coefficient and refractive index spectra obtained after controlled thermal treatments at 150°C, 175°C, and 200°C. Each spectral representation formed a separate 70-feature dataset. Five supervised classifiers were trained on the multiclass spectral data: logistic regression (LR), k-nearest neighbors (KNN), random forest (RF), support vector machine (SVM), and neural network (NN). Models were evaluated on an independent held-out test set (20% of the data), with five-fold cross-validation performed on the training set (80%) during hyperparameter optimization. Performance was assessed using accuracy, precision, recall, and F1-score, while uncertainty was quantified using nonparametric bootstrap confidence intervals. Absorption coefficient spectra consistently provided stronger discriminatory information than refractive index spectra. For the absorption dataset, SVM achieved the highest performance, with an accuracy of 0.93 (95% CI: 0.82–1.00) and F1-score of 0.92 (95% CI: 0.80–1.00), followed closely by RF and NN. In contrast, refractive index models showed lower predictive performance and greater uncertainty, with RF achieving the highest F1-score of 0.77 (95% CI: 0.60–0.93). These results demonstrate that THz absorption is more sensitive to thermally induced molecular and microstructural changes in rubber. Overall, integrating THz-TDS absorption spectroscopy with machine learning offers a non-destructive and automated framework for rubber quality evaluation, industrial process monitoring, and intelligent quality-control applications.