Classification of Industrial Tannin Extracts Using NIR Spectroscopy and PLS-DA

Vegetable tannins are plant extracts that have traditionally been used to tan leather. This technology was the most important leather making process in western world until the development and implementation of mineral tanning process with basic chrome salts, in late 19th century and early 20th century. These vegetable tanning materials differ greatly in chemical constitution and tanning properties. Noteworthy are the recent applications where the Infrared spectra were employed in the construction of the multivariate regression models developed to classification of the natural products. On the other hand the Near Infrared Spectroscopy (NIR) is one of the most popular techniques due to its easy handling and great applicability in quantitative analysis. In this work Partial Least Squares with Discriminant Analysis (PLS-DA) was used for developed the clean methodology for classification of industrial tannin extracts using NIR spectra. Experimental Part: Six of the most used industrial tannin extracts have been investigated. Ten samples of black wattle (Acacia mearnsii), eight of quebracho (Schinopsis lorentzii), eight of tara (Caesalpinia spinosa), eight of chestnut (Castanea sativa), seven of myrobalan (Terminalia chebula) and five of valonea (Quercus Aegilops) were analysed. The extracts were provided by Tanac (Brazil), Silvateam (Argentina), Exandal (Peru), Tannin Sevnica (Slovenia), Samrat Chemical (India) and Air-tu Kymia (Turkey). For the NIR analysis the samples were dried at 80 °C and analyzed at room temperature. Spectroscopic analysis were performed on a spectrophotometer with near-infrared reflection, Foss Near-IR Systems Infrared Spectrometer, Model 6500 with spinning sample module for solid state in the spectral range of 1200-2500 nm, with a resolution of 4 cm-1 with 32 scans. The analyses were performed in triplicate and for the PLS-DA the average spectra was used. Results and Discussion: The main spectral regions where differences were observed in the spectra of different extracts of tannin, match the range of 1600-1800 nm related to absorption of the first overtone of CH, CH2, CH3, and the region 2200-2400 nm the combination of bands related to CH, CH2, CH3. The algorithm of Kennard–Stone was used for the selection of training and test sets. Before multivariate analysis, the reflectance spectra were smoothed (1nd polynomial and 15 points per window), normalized (standard normal variation – SNV), derived (1st order, 2nd polynomial, 15 points per window) and mean centered using PLS-Toolbox. For all six classes patterned were obtained 100% sensitivity and 100% selectivity using the average spectra of the each sample. Conclusion: The results of this study, this method has proved suitable for the characterization of the six types of industrial extracts, the main advantages to using a non-invasive technique, low cost and with reduced scanning time.