Classification and aging assessment of kraft and thermally upgraded insulating paper via near-infrared spectroscopy

In recent years, researchers propose using near-infrared spectroscopy to achieve non-destructive aging assessment of insulating paper. However, different types of paper samples exhibit spectral differences, which require the consideration of complex and multiple mapping relationships during model training, leading to a conflict between the generalization ability and accuracy of the model. In this paper, a strategy combining paper sample classification with aging assessment has been adopted. Craft paper and thermally upgraded paper samples in different aging states are collected and analyzed. It is observed that the changes in NIR spectra and degree of polymerization (DP) varied among different types of paper samples. The disparities in NIR spectra between different paper types enabled the paper sample classification model based on random forest to achieve high accuracy. In addition, four single type DP prediction model based on partial least squares regression are developed. The validation on the test set demonstrated that our proposed models had lower root mean square error compared to hybrid model (73.8 vs 95.7).