ANN Modelling of the Adsorption of Herbicides and Pesticides Based on Sorbate-Sorbent Interphase

Herbicides and pesticides (H & P) are commonly used in agricultural practice and is a serious environmental pollutant in contemporary times. Studies have shown that it can be efficiently mitigated from the environment by adsorption. The aim of this study was to utilise Artificial Neural Networks (ANN) to model the adsorption of H & P from aqueous media based on the sorbate-sorbent interphase. The sorbate-sorbent interphase was characterised by the relative molecular mass (g/mol), specific surface area (m(2)/g), effective surface area (mol/m(2)), solubility (mol/l), and preferential adsorption (sorbate mol on sorbent/sorbate mol in solution). The coefficient of determination (R-2) at training, validation and testing were 0.9825, 0.9428 and 0.9793 respectively. The accuracy of the model was substantiated by direct comparison and parity plots. The paired samples correlation showed a strong positive correlation (0.980) and statistical significance (p < 0.05) between the model predictions and experimental results. This study reveals that information on the adsorbent specific surface area, adsorbent effective surface area, adsorbate preferential adsorption, adsorbate solubility and adsorbate relative molecular mass can be used to accurately predict the mass adsorption capacity for any H & P from aqueous media.

Automated summary

This study utilized Artificial Neural Networks to model the adsorption of herbicides and pesticides from aqueous media and found that parameters such as adsorbent specific surface area, effective surface area, adsorbate preferential adsorption, solubility, and relative molecular mass play important roles in predicting the mass adsorption capacity accurately. The model's accuracy was verified through various statistical analyses.