ARTIFICIAL NEURAL NETWORK-BASED MODELING OF LIQUID MEMBRANES FOR SEPARATION OF CERIUM (III)

Abstract

The rare earth elements (REEs) have been used in a range of applications from permanent magnets to catalysts, optical devices, electronic materials and many others, with remarkable quantities. The rare earth metal cerium (Ce) is widely used, however, because it is easily accessible in the industry. One of the methods which is promising to replace the conventional methods for separation and recovery of cerium from dilute aqueous solution is Emulsion Liquid Membrane (ELM) due to its high selectivity, low consumption of solvent, extraction-stripping process and high mass transfer efficiency. The process of extraction of Ce (III) ions was predicted and investigated in the present work by ANN modeling technique with ELM as carrier extractant, Span 80 as surfactant, Kerosene as diluent and HNO3 as stripping agent. The extraction time, pH of feed phase, concentration of the feed, and the concentration of the surfactant, stripping phase, stirring speed, phase ratio, treatment ratio and initial cerium concentration were used as input variables in this analysis. Feed forward multi-layer perceptron (ANN) was used to describe the effect of the process parameter to the extraction performance and the Levenberg-Marquardt back propagation algorithm was used for the training of the ANN. Five statistical measures MSE, Pearson correlation coefficient (R), RMSE, MAE and RPE were used to assess the model's predictive ability. The optimized 9–11–1 ANN model achieved overall R = 0.99082 (R² = 0.9817) and MSE = 26.145 (RMSE ≈ 5.11%). The sensitivity analysis shows that the biggest sensitivity is with the pH (28%) and the concentration of D2EHPA (23%). The obtained values obtained from the modelling were in good agreement with the experimental ones and the extraction behavior of Ce (III) was very accurate. This indicates that the modeling based on ANN is well suitable for the modeling of ELM for rare earth elements recovery.