This study presents a definitive framework for Cocos nucifera (coconut) shell valorization, integrating high-resolution thermogravimetry with advanced machine learning. Physicochemical analysis confirms a high-energy feedstock (45.7% carbon, 71.5% volatiles), with SEM/XEDS and FTIR revealing heterogeneous, lignocellulosic, catalytic-rich structural matrix. TG/DTG analysis identified distinct degradation windows: hemicellulose (135–395 °C), cellulose (270–430 °C), and protracted lignin decomposition (275–675 °C). Kinetic modeling indicates that pyrolysis follows a third-order (F3) continuous degradation mechanism across the studied range, supported by high correlation coefficients (R2 = 0.93–0.96). The mean kinetic and thermodynamic parameters—specifically an activation energy of 165 kJ·mol−1 (calculated across the 10–60 wt% conversion range during hemicellulose and cellulose pyrolysis), a positive activation enthalpy (159 kJ·mol−1), and a Gibbs free energy of activation (155 kJ·mol−1)—suggest that the thermochemical conversion of coconut shell is an endothermic, non-spontaneous process with moderate energy requirements. Furthermore, the integration of kNN machine learning yielded near-perfect predictive metrics (R2 ≈ 1.000) using optimized hyperparameters (k = 85 for TG, k = 100 for DTG, and k = 50 for conversion). These findings suggest that coconut shells can be efficiently valorized as a high-energy feedstock, with data enabling reliable and optimized prediction of thermal degradation to minimize experimental waste.
Otaru et al. (Tue,) studied this question.