Forecasting the Thermal Degradation Depending on the Kinetics of Dracaena Draco Lignocellulosic Fibers Using an Artificial Neural Network

Abstract

In order to forecast the thermal degradation of Dracaena draco plant fibers (DDFs) using thermogravimetric analysis (TGA) at heating rates ranging from 5 to 30°C/min, this study employed artificial neural networks (ANNs). Hemicellulose, cellulose, and lignin breakdown were represented by the three different degradation stages that were seen. The enhanced ANN27 model successfully captured pyrolysis behavior and degradation patterns, achieving a high prediction accuracy (R2 = 0.99966). The model performed well at lower heating rates (5 and 10°C/min), but because of bias and heteroscedasticity, adjustments are required at higher rates (15–30°C/min). In contrast to the experimental averages of 131.244 kJ/mol, 109.269 kJ/mol, and 131.694 kJ/mol, respectively, kinetic analysis showed that the ANN27-predicted activation energies (Ea) were 133.420 kJ/mol (KAS), 53.692 kJ/mol (FWO), and 133.784 kJ/mol (STR). Without requiring a lot of testing, our ANN method provides insights into DDF thermal behavior and optimizes processing settings by properly forecasting degradation curves.