Prediction by artificial neural network of insulation performance of eco-treated cork stoppers: Experimental measurement, modeling and optimization

This study aims to predict by artificial neural networks (ANN) the improvement in mass insulation of cork stoppers treated by high temperature thermal (HTT) and/or boiling. Experimental tests have shown that the desorption kinetics are more favorable for smaller molecules D_KCl < D_NaCl. The results validated the developed mathematical model, which accounted for the actual cylindrical shape of the stopper, and quantified the improvement in apparent diffusion coefficients as a function of the maximum temperature of the treatment cycle: D_105°<D_200°<D_350° ≈ D_450° and the protocol type: D_A<D_B<D_C. The results revealed a positive correlation between temperature and the diffusion phenomenon, with a significant influence observed up to 350 °C. Furthermore, to enhance the accuracy of Dapp, the Bat Algorithm optimization method was applied, achieving a precision of the order of 10⁻⁵. An experimental database, composed of 3864 points, was previously optimized to be integrated into an artificial neural network (ANN) model with a specific architecture (5-5-6-1). The model thus developed demonstrated remarkable reliability, displaying a coefficient of correlation R² of 0.9997 and an extremely low root mean square error (RMSE), evaluated at 7.38 × 10⁻¹⁴. These performances underline the robustness and accuracy of the proposed model for the prediction of the studied phenomenon.