Optimization and prediction of thermodynamic parameters in co-pyrolysis of banana peel and waste plastics using AIC model and ANN modeling

• Optimization and prediction of thermal behavior of unknown mixtures of biomass and plastics. • ANN and AIC model to understand reactivity, thermodynamic parameters, and synergism. • Identification of complex reaction mechanisms in thermal degradation using Criado master plots. • Suggested ANN model showed high accuracy for prediction with R 2 > 0.998. • Co-pyrolysis of biomass and plastics showed positive synergy by reducing activation energy. The optimization and prediction of thermodynamic parameters including synergistic effects, and kinetic analysis in co-pyrolysis of banana peel (BP) and waste polystyrene (PS) plastic at different heating rates using ANN and AIC models has been performed. Thermogravimetric analysis was performed to determine the initial, maximum, and final degradation temperatures. The synergistic effect was studied using additive formula to determine the theoretical thermal behavior and compared with experimental TGA data. Kinetic parameters were determined by using the advanced isoconversional (AIC) model for estimation of activation energy (E α ), Criado master plot for reaction mechanism ( f (α)), and compensation method for frequency factor (A α ). The analysis showed that the average activation energy values were 182.5, 140.6, and 161.8 kJ mol −1 for PS, BP, and PS+BP, respectively. It also clearly shows positive synergy in co-pyrolysis of PS and BP by reducing 11.3 % activation energy compared to that of PS alone. The frequency factor was found to be 1.0 × 10 14 , 1.0 × 10 15 , and 1.0 × 10 23 s −1 for PS, BP, and PS+BP, respectively. The reaction mechanism was identified as R3, D4, and D4+R3 for PS, BP, and PS+BP, respectively. Further, the obtained kinetic parameters were used to determine the thermodynamic parameters such as enthalpy (ΔH), Gibbs energy (ΔG), and Entropy (ΔS). Finally, ANN was designed to address the co-pyrolysis behavior subjected to various heating rates. Subsequently, the trained ANN model (5 × 4×4 × 4) was employed to forecast thermal degradation behavior. Impressively, the model yielded highly accurate results with a correlation coefficient R 2 > 0.998 in each case. The optimized model was further used to predict TGA data and activation energy for unknown mixtures of PS and BP. The suggested ANN model showed a great advantage in optimizing to avoid extensive experiments at various heating rates to achieve the goal.