Artificial neural network modeling and response surface methodology as efficient tools for predicting the effective parameters of uranium extraction from chloride media

This study focuses on optimizing solvent extraction parameters to efficiently recover uranium from chloride solutions, a process with significant potential for the nuclear industry. Initially, various extractants were evaluated using a single-variable method, which identified di(2-ethylhexyl) phosphoric acid (D2EHPA) in kerosene as having high potential for extracting uranium from chloride media. Subsequently, other significant process parameters were assessed: chloride ion concentration (0.05–2 mol/L), extractant concentration (0.01–0.2 mol/L), pH of the aqueous phase (0.1–2), and contact time (5–60 min). Response Surface Methodology (RSM) and an Artificial Neural Network (ANN) were employed to model the data and compare the results of the extraction process. The RSM model a second-degree polynomial equation with an R 2 value of 0.9846, fitted the experimental data excellently. The ANN model with two hidden layers predicted the extraction efficiency with an R 2 of 0.9911, indicating a strong fit. Under the optimum conditions, the extraction efficiency was 98.87 % according to the RSM and 99.63 % according to the ANN, both of which are very close to the experimental value of 99.43 %. This research contributes to optimizing uranium extraction processes, enhancing efficiency, and potentially reducing the environmental impact of the nuclear industry.