Circular economy: Kinetic-Triplet, thermodynamic, and gradient descent optimisation algorithm of deep learning models for the thermal degradation of walnut shell

The study offers a thorough assessment of the thermal degradation of pulverised walnut shell, utilizing thermo-kinetic analysis and a gradient descent optimisation algorithm. Thermogravimetric analysis (TGA) and differential thermogravimetry (DTG) measurements were obtained at heating rates of 3, 10, 20, and 40 °C.min −1 , with degradation temperatures ranging from 30 to 850 °C. The thermo-kinetic analysis of the TGA traces yielded activation energies of 89.33, 83.71, 84.12, and 92.26 kJ mol −1 , determined through the FWO, KAS, STK, and FR isoconversional model-free kinetic methods, respectively. The power law [P3] nucleation model was identified as the likely reaction mechanism. The estimated thermodynamic data indicate that the formation of activated complexes from the reactants is endothermic (Activation enthalpy, ΔH = 78.13–86.68 kJ mol −1 ), non-spontaneous (Gibbs free energy of activated complex, ΔG = 184.13–288.05 kJ mol −1 ), and less disordered (Entropy of activation, ΔS = −0.31 to −0.16 kJ mol −1 K −1 ), positioning walnut shell as a more viable bioenergy source compared to various agricultural wastes. The application of the gradient descent optimisation algorithm resulted in an optimized deep neural network (DNN) framework and parameters that produced output signals closely aligning with experimental measurements ( R 2 ∼ 0.999 ). This study identified the degradation temperature as the most sensitive parameter influencing the thermal degradation of walnut shell. The findings could provide valuable insights for energy and environmental experts, as well as policymakers, regarding the potential use of walnut shell as a viable feedstock for pyrolysis and bioenergy production.