Mg-Fe-LDH for Aquatic Selenium Treatment: Adsorption, RSM Modeling, and Machine Learning Neural Network

Different aqueous selenium species including selenite, selenate, and selenocyanate, can be harmful to both humans and other life forms. Considering this, the present study investigated the application of Mg-Fe-LDH nanomaterial adsorbent for the removal of aqueous selenite, selenate, and selenocyanate species. The research examined the removal efficiency of selenite, selenate, and selenocyanate under a varying set of competitive conditions including Mg-Fe-LDH dosage (0.5–1.5) g/l, Mg-Fe-LDH calcination temperature (0–500 °C), and concentration of selenium species (2.5–7.5 ppm). The respective experimental results showed a higher removal for the selenite and selenate species as compared to the selenocyanate species. Also, the Freundlich isotherm provided a good fit both for selenite and selenocyanate uptake whereas for selenate the Langmuir isotherm provided a better fit. Furthermore, the study found that the removal of selenite and selenocyanate followed the pseudo-first-order and pseudo-second-order models, respectively, while both models fitted well with selenate. This work also employed response surface methodology (RSM) and back-propagation artificial neural network (BPANN) based modeling techniques to predict the removal efficiency of respective selenium species. The modeling results demonstrated that the BPANN method exhibited a higher accuracy in predicting selenium species removal as compared to the RSM technique. To that end, the BPANN-based models yielded R ² values of 0.9894, 0.9709, and 0.9568 for selenite, selenate, and selenocyanate, respectively. In contrast, the RSM methodology showed lower R ² values, i.e., 0.9464 for selenite, 0.9002 for selenate, and 0.8225 for selenocyanate.