Artificial intelligence modeling and experimental studies of oily pollutants uptake from water using ZIF-8/carbon fiber nanostructure.

In this study, the experimental and modeling of oily pollutants (crude oil, asphaltene, and maltene) uptake by ZIF-8/carbon fiber nanostructure was investigated. The influence of pollutant type, concentration, ionic strength, and sorption time on uptake was systematically examined using a batch absorption system. Then, the experimental data of uptake was modeled using cascade forward (CFNN), multi-layer perceptron (MLP), radial basis function (RBF), and generalized regression (GRNN) neural networks. ZIF-8/carbon fiber nanostructure distinguished by its high hydrophobicity (WCA of 150°) and a complex meso-micro pore structure, demonstrated remarkable efficiency in oil pollutant uptake. Furthermore, the modeling results unveiled that the CFNN-LM model yielded superior predictions, achieving an impressive accuracy rate, as approximately 98% of the uptake data demonstrated an average absolute percent relative error (AAPRE,%) below 3%. Moreover, sensitivity analysis showed that the concentration of pollutants had the most notable impact on the pollutant uptake. Furthermore, the uptake values exhibited an upward trend with elevated concentrations of the pollutant and extended process time, while showing a decline with an increase in ionic strength. These results affirm the reliability of the proposed CFNN-LM model in accurately estimating uptake amounts during the separation process. In summary, the ZIF-8/carbon fiber nanostructure stands out as a highly promising remedy for eliminating oil pollutants from oil/water mixtures, with the added benefit of accurate uptake predictions facilitated by the CFNN-LM model.