Kinetic and Thermodynamic Insights into Sapodilla Leaves Pyrolysis via Experimental Analysis and Machine Learning Models

Abstract Measuring the kinetic and thermodynamic parameters precisely is essential to the understanding of biomass pyrolysis and the efficient conversion of thermochemical conversion. Sapodilla leaves were used in this study as a representative feedstock and studied by using experimental and machine learning (ML) methodologies. Thermogravimetric analysis (TGA), with several heating rates, which was examined using the Coats–Redfern method, indicated that the activation energy (Ea) and the pre-exponential factor (A) were very sensitive to model choice and heating rate. Low-temperature decomposition best fitted the diffusion- and reaction-order models, but the high-temperature stages were modeled using nucleation models. Thermodynamic analysis proved that pyrolysis is endothermic and non-spontaneous (ΔG = -104107 kJmol⁻ ¹) and negative entropy values (ΔS = -0.224 to -0.234 kJmol⁻ ¹K⁻ ¹) indicated greater structural organization in the formation of char. To supplement experimental fitting, an ML model was also created to estimate the kinetic triplet (Ea, A, R²) directly using proximate and ultimate analysis. Nonlinear regressors e.g., Random Forest, Support Vector Regression, and XGBoost, were more accurate, and stacked ensembles have the most stable predictions. The ML models consistently gave close Ea ranges (~ 4245 kJmol⁻ ¹), compared to the variability of Coats–Redfern results, and using SHAP to interpret results; volatile matter, carbon content, and O / C and H / C ratios were found to be dominant. This paper shows that mechanistic and ML methods play a complementary role: Coats-Redfern analysis offers stage-specific mechanistic descriptions, whereas ML offers stable predictions based on composition that can be used to screen biomass quickly. The combined approach improves mechanistic insights and predictive ability, facilitating system design and optimization of renewable thermochemical transformation.