Comparative Analysis of Lignocellulose Agricultural Waste and Pre-treatment Conditions with FTIR and Machine Learning Modeling.

Autor: Pancholi, Manali J., Khristi, Anand, M., Athira K., Bagchi, Debjani
Předmět:
Zdroj: BioEnergy Research; Mar2023, Vol. 16 Issue 1, p123-137, 15p
Abstrakt: Resource-efficient production of value-added products from lignocellulosic waste is an important requisite for sustainable development. Since constituent separation of lignocellulosic waste is challenging due to the energetically robust structure of the cellulose-hemicellulose-lignin network, Fourier transform infrared (FTIR) spectroscopy is used for rapid, non-invasive analysis of cellulose and lignin for lignocellulose farm waste procured from about 30 different crops post-harvest. FTIR mode peak heights show linear concentration dependence over a narrow range, and yield 44 ± 5% cellulose and 9.95 ± 2% lignin content in one variety of rice straw, with high standard deviation for inter-species yield. Due to compositional heterogeneity and multivariate nature, FTIR data is analyzed with machine learning models. Species-wide pattern classification for cellulose and lignin in the lignocellulose data is first conducted with linear discriminant analysis, decision tree, and random forest algorithms. For the present data, the best classification accuracy is obtained with random forest algorithm with an accuracy score of 0.75, with least mismatches between the predicted and true values. Convolutional neural network modeling with Bayesian regularization training algorithm using FTIR absorption peak parameters resulted in a good representation of the lignocellulose data with root mean square error ~ 0.11. The structural changes produced in cellulose due to different pre-treatments is analyzed with peak heights of FTIR modes and correlated with efficiency of enzymatic hydrolysis to form glucose. Lignocellulose pre-treatment with deep eutectic solvent improves cellulose accessibility to enzymes, with 38% glucose yield enhancement in comparison to acid and alkali pre-treatments. [ABSTRACT FROM AUTHOR]
Databáze: Complementary Index
Externí odkaz: