Artistic representation of the classification of various technical lignins
Source: BAM
Lignin is one of the largest aromatic carbon sources on our planet. It is found in all woody plants as well as in many grasses and protects them from UV radiation, predators and desiccation and gives them stability. Technical lignin is mainly obtained as a by-product in pulp and paper production and represents the starting point from which it can be further processed for potential applications. In this work, technical lignins were measured using attenuated total reflectance infrared spectroscopy (ATR-IR) and then a mathematical model was created to classify them, allowing unknown samples to be quickly and reliably assigned to their parent plant group (softwoods, hardwoods, straw & grasses). The model is based on the application of principal component analysis (PCA) coupled with classification by the k-nearest neighbour method (k-NN). After recording the IR data, the spectra have to be processed mathematically. The effect of different stages of spectrum processing on the predictive power of the model was compared. The mean-centred raw data, a second derivative and a second derivative combined with a unit vector normalization of the data were compared. Euclidean distance was used as the distance measure for k-NN. The quality criteria accuracy (Acc), sensitivity (TPR) and specificity (TNR) were used to evaluate the predictive power. The different data sets were tested for k = 1...20. A 5-fold cross-validation was used for validation. It could not only be shown that classification using ATR-IR is possible, but also that the predictive power of the model increases with advancing data processing. For the optimized model with k = 4, the following values are obtained for the quality criteria: Acc = 98.9%, TPR = 99.2%, TNR = 99.6%.
Identification and Classification of Technical Lignins by means of Principle Component Analysis and k-Nearest Neighbor Algorithm
published in Chemistry-Methods, Band 1, Heft 8, S. 350 - 396, 2021
BAM division Structure Analysis