A new method of developing ANN-isotherm hybrid models for the determination of thermodynamic parameters in the adsorption of ions Ag+, Co2+ and Cu2+ onto zeolites ZSM-5, HY, and 4A

Abstract A new methodology for simultaneous parameter estimation of several adsorption curves through a hybrid ANN-isotherm model is proposed. For that, the ANN model was trained with 36 adsorption isotherm curves with three different adsorbates (Ag+, Co2+, and Cu2+), three different adsorbents (zeolites ZSM-5, HY, and 4A), and four different temperatures. Two ANN-isotherm hybrid models were created. One was based on the Langmuir equation, and the other was based on Sips. A pseudo-outputs method was implemented for isotherm and thermodynamic parameters estimation. The hybrid models reached Pearson's linear correlation coefficient values between 0.98 and 1 for measured and predicted data. The Sips heterogeneity factor indicated that only zeolite 4A performed the adsorption process with a certain heterogeneity degree. In contrast, zeolites ZSM-5 and HY had adsorptions fully described by the Langmuir isotherm. The assessment of adsorption thermodynamic parameters compared the isotherm-based values of equilibrium constant (K) and change in Gibbs' free energy (ΔG0) from the pseudo-outputs ANN structure to Van’t Hoff equations calculated K and ΔG0, showing good agreement with high values of R2 and low values of RMSE. It was also possible to determine that zeolite 4A showed the best adsorption efficiency due to the higher affinity with ions Ag+, Co2+, and Cu2+, followed by zeolite HY. Among the metallic ions, Ag+ and Cu2+ had similar adsorption behaviors onto all adsorbents, with improved adsorption as the temperature increased, while the Co2+ showed the lowest sensitivity to temperature variation.