| dc.description.abstract | In this study, principal component analysis was used to separate standard and non-standard ethylene vinyl acetate (EVA) based on the infrared spectra, whereas partial least square was used to create sample group’s prediction models. Preprocessing step was performed on the IR spectrum to ensure the best prediction for measurement. The best grouping for water-based EVA was obtained by using the before-manipulation model, which showed the variation value of 96% (PC1 = 84%, PC2 = 12%), whereas for hotmelt EVA, the best grouping was obtained by using the normalization before-manipulation model, showing the variation value of 96% (PC1 = 90%, PC2 = 6%). The best prediction model for both water-based and hotmelt EVA was obtained by using the beforemanipulation model with the r calibration of 0.9801 and 0.9402, r validation of 0.9431 and 0.8414, RMSEC of 0.0694 and 0.0992, RMSEP of 0.1223 and 0.1627, SEC of 0.0702 and 0.0997, SEP of 0.1236 and 0.1634, calibration bias of -3.663 × 10-8 and 2.577 × 10-8, and validation bias of -0.0011 and -0.0067, respectively. Based on these results, this alternative spectroscopic and chemometric-based quality control method can be applied for both water-based and hotmelt materials and can improve the simplicity and practicality of quality control process. | en |
